From 9c111584b5384cfd2e87d0fb7a84185ae01b23ab Mon Sep 17 00:00:00 2001
From: TC1995 <mariogplayer90@gmail.com>
Date: Sat, 26 Oct 2024 21:32:47 +0200
Subject: [PATCH 01/26] Big video changes of the day (October 26th, 2024)

8514/A/Mach8/32:
Reworked the acceleration a bit as well as the mode switches from VGA to ATI/8514/A mode and viceversa based on the documentation. Fixes the Linux GUI fonts using the Mach32 driver (possibly Mach8 too) and other anomalies (still needs more work on the acceleration though).

PVGA:
Minor fixes to the banking.

S3-based (Pre-ViRGE):
1. Made the chip class use its own banking so that the Enhanced Mode mapping (equivalent to 64K-based A0000) is taken into account (per bit 3 of CRTC31, which forces the mapping to be 64K A0000, regardless of the GDCREG6 bits). Fixes NT 3.1 347.1's S3 driver.
2. Initial rework on 15bpp/16bpp accelerated mode of the 911/924 chips (currently not 100% bug free, I need help with this chips).

XGA-1/2:
Get rid of the linear endian reverse variable hack used by OS/2 and, instead, use the already declared ones more wisely, fixes OS/2 2.1 reversed fonts and keeps everything else working as it should.

Misc:
Added the BT481 RAMDAC for future card use.
---
 src/include/86box/vid_8514a.h     |   22 +-
 src/include/86box/vid_ati_mach8.h |   21 +-
 src/include/86box/vid_svga.h      |    5 +
 src/include/86box/vid_xga.h       |    5 +-
 src/video/CMakeLists.txt          |    2 +-
 src/video/vid_8514a.c             | 2973 ++++++++++++-----------------
 src/video/vid_ati68860_ramdac.c   |   16 +-
 src/video/vid_ati_mach8.c         | 2061 ++++++++++----------
 src/video/vid_bt481_ramdac.c      |  160 ++
 src/video/vid_paradise.c          |   22 +-
 src/video/vid_s3.c                | 1466 ++++----------
 src/video/vid_svga.c              |   64 +-
 src/video/vid_xga.c               |  412 ++--
 13 files changed, 3029 insertions(+), 4200 deletions(-)
 create mode 100644 src/video/vid_bt481_ramdac.c

diff --git a/src/include/86box/vid_8514a.h b/src/include/86box/vid_8514a.h
index d834a6f42..39af01fd0 100644
--- a/src/include/86box/vid_8514a.h
+++ b/src/include/86box/vid_8514a.h
@@ -52,7 +52,7 @@ typedef struct ibm8514_t {
     int type;
     int local;
     int bpp;
-    int on[2];
+    int on;
     int accel_bpp;
 
     uint32_t vram_size;
@@ -102,6 +102,8 @@ typedef struct ibm8514_t {
         uint16_t frgd_mix;
         uint16_t multifunc_cntl;
         uint16_t multifunc[16];
+        uint16_t clip_right;
+        uint16_t clip_bottom;
         int16_t  clip_left;
         int16_t  clip_top;
         uint8_t  pix_trans[2];
@@ -112,8 +114,6 @@ typedef struct ibm8514_t {
         int      x3;
         int      y1;
         int      y2;
-        int      sys_cnt;
-        int      sys_cnt2;
         int      temp_cnt;
         int16_t  cx;
         int16_t  cx_back;
@@ -127,20 +127,14 @@ typedef struct ibm8514_t {
         int16_t  err;
         uint32_t src;
         uint32_t dest;
-        uint32_t newsrc_blt;
-        uint32_t newdest_blt;
-        uint32_t newdest_in;
-        uint32_t newdest_out;
-        uint8_t *writemono;
-        uint8_t *nibbleset;
         int      x_count;
         int      xx_count;
         int      y_count;
         int      input;
+        int      input2;
         int      output;
+        int      output2;
 
-        uint16_t cur_x_bit12;
-        uint16_t cur_y_bit12;
         int      ssv_len;
         uint8_t  ssv_dir;
         uint8_t  ssv_draw;
@@ -156,7 +150,6 @@ typedef struct ibm8514_t {
     } accel;
 
     uint16_t test;
-    int      vendor_mode[2];
     int      h_blankstart;
     int      h_blank_end_val;
     int      hblankstart;
@@ -227,6 +220,11 @@ typedef struct ibm8514_t {
     int      ext_pitch;
     int      ext_crt_pitch;
     int      extensions;
+    int      linear;
+    int      _4bpp;
+    uint32_t vram_amount;
+    int      vram_512k_8514;
+    PALETTE  _8514pal;
 
     latch8514_t latch;
 } ibm8514_t;
diff --git a/src/include/86box/vid_ati_mach8.h b/src/include/86box/vid_ati_mach8.h
index a25432920..849446def 100644
--- a/src/include/86box/vid_ati_mach8.h
+++ b/src/include/86box/vid_ati_mach8.h
@@ -41,8 +41,6 @@ typedef struct mach_t {
     int ramdac_type;
     int old_mode;
 
-    uint32_t memory;
-
     uint16_t config1;
     uint16_t config2;
 
@@ -73,9 +71,7 @@ typedef struct mach_t {
     uint8_t  bank_r;
     uint16_t shadow_set;
     uint16_t shadow_cntl;
-    int ext_on[2];
-    int extended_mode;
-    int compat_mode;
+    int override_resolution;
 
     struct {
         uint8_t  line_idx;
@@ -84,6 +80,12 @@ typedef struct mach_t {
         uint8_t  patt_len;
         uint8_t  pix_trans[2];
         uint8_t  eeprom_control;
+        uint8_t  alu_bg_fn;
+        uint8_t  alu_fg_fn;
+        uint16_t clip_left;
+        uint16_t clip_right;
+        uint16_t clip_top;
+        uint16_t clip_bottom;
         uint16_t dest_x_end;
         uint16_t dest_x_start;
         uint16_t dest_y_end;
@@ -102,14 +104,14 @@ typedef struct mach_t {
         uint16_t ge_offset_hi;
         uint16_t linedraw_opt;
         uint16_t max_waitstates;
-        uint8_t  patt_data_idx;
-        uint8_t  patt_data[0x18];
         uint16_t scan_to_x;
         uint16_t scratch0;
         uint16_t scratch1;
         uint16_t test;
         uint16_t pattern;
         uint16_t test2;
+        int      patt_data_idx_reg;
+        int      patt_data_idx;
         int      src_y_dir;
         int      cmd_type;
         int      block_write_mono_pattern_enable;
@@ -144,9 +146,8 @@ typedef struct mach_t {
         int      stepx;
         int      stepy;
         int      src_stepx;
-        uint8_t  color_pattern[16];
-        uint8_t  color_pattern_full[32];
-        uint16_t color_pattern_word[8];
+        uint8_t  mono_pattern_normal[16];
+        uint8_t  color_pattern[32];
         int      mono_pattern[8][8];
         uint32_t ge_offset;
         uint32_t crt_offset;
diff --git a/src/include/86box/vid_svga.h b/src/include/86box/vid_svga.h
index 18b7b672a..c91db4fe8 100644
--- a/src/include/86box/vid_svga.h
+++ b/src/include/86box/vid_svga.h
@@ -369,6 +369,7 @@ uint32_t svga_mask_addr(uint32_t addr, svga_t *svga);
 uint32_t svga_mask_changedaddr(uint32_t addr, svga_t *svga);
 
 void svga_doblit(int wx, int wy, svga_t *svga);
+void svga_poll(void *priv);
 
 enum {
     RAMDAC_6BIT = 0,
@@ -396,6 +397,9 @@ extern void    att498_ramdac_out(uint16_t addr, int rs2, uint8_t val, void *priv
 extern uint8_t att498_ramdac_in(uint16_t addr, int rs2, void *priv, svga_t *svga);
 extern float   av9194_getclock(int clock, void *priv);
 
+extern void    bt481_ramdac_out(uint16_t addr, int rs2, uint8_t val, void *priv, svga_t *svga);
+extern uint8_t bt481_ramdac_in(uint16_t addr, int rs2, void *priv, svga_t *svga);
+
 extern void    bt48x_ramdac_out(uint16_t addr, int rs2, int rs3, uint8_t val, void *priv, svga_t *svga);
 extern uint8_t bt48x_ramdac_in(uint16_t addr, int rs2, int rs3, void *priv, svga_t *svga);
 extern void    bt48x_recalctimings(void *priv, svga_t *svga);
@@ -452,6 +456,7 @@ extern const device_t att491_ramdac_device;
 extern const device_t att492_ramdac_device;
 extern const device_t att498_ramdac_device;
 extern const device_t av9194_device;
+extern const device_t bt481_ramdac_device;
 extern const device_t bt484_ramdac_device;
 extern const device_t att20c504_ramdac_device;
 extern const device_t bt485_ramdac_device;
diff --git a/src/include/86box/vid_xga.h b/src/include/86box/vid_xga.h
index 90a3c69aa..b90b53978 100644
--- a/src/include/86box/vid_xga.h
+++ b/src/include/86box/vid_xga.h
@@ -31,13 +31,11 @@ typedef struct xga_hwcursor_t {
 } xga_hwcursor_t;
 
 typedef struct xga_t {
-    mem_mapping_t  membios_mapping;
     mem_mapping_t  memio_mapping;
     mem_mapping_t  linear_mapping;
     mem_mapping_t  video_mapping;
     rom_t          bios_rom;
-    rom_t          membios_rom;
-    rom_t          vga_bios_rom;
+    rom_t          bios_rom2;
     xga_hwcursor_t hwcursor;
     xga_hwcursor_t hwcursor_latch;
     PALETTE        extpal;
@@ -153,6 +151,7 @@ typedef struct xga_t {
     int a5_test;
     int type;
     int bus;
+    int busy;
 
     uint32_t linear_base;
     uint32_t linear_size;
diff --git a/src/video/CMakeLists.txt b/src/video/CMakeLists.txt
index 2f7607ad6..bbd329b7e 100644
--- a/src/video/CMakeLists.txt
+++ b/src/video/CMakeLists.txt
@@ -19,7 +19,7 @@ add_library(vid OBJECT agpgart.c video.c vid_table.c vid_cga.c vid_cga_comp.c
     vid_sigma.c vid_wy700.c vid_ega.c vid_ega_render.c vid_svga.c vid_8514a.c
     vid_svga_render.c vid_ddc.c vid_vga.c vid_ati_eeprom.c vid_ati18800.c
     vid_ati28800.c vid_ati_mach8.c vid_ati_mach64.c vid_ati68875_ramdac.c
-    vid_ati68860_ramdac.c vid_bt48x_ramdac.c vid_chips_69000.c
+    vid_ati68860_ramdac.c vid_bt481_ramdac.c vid_bt48x_ramdac.c vid_chips_69000.c
     vid_av9194.c vid_icd2061.c vid_ics2494.c vid_ics2595.c vid_cl54xx.c
     vid_et3000.c vid_et4000.c vid_sc1148x_ramdac.c vid_sc1502x_ramdac.c
     vid_et4000w32.c vid_stg_ramdac.c vid_ht216.c vid_oak_oti.c vid_paradise.c
diff --git a/src/video/vid_8514a.c b/src/video/vid_8514a.c
index 944a14a88..3e758b952 100644
--- a/src/video/vid_8514a.c
+++ b/src/video/vid_8514a.c
@@ -106,6 +106,9 @@ ibm8514_log(const char *fmt, ...)
         dat = (dev->vram[(addr) & (dev->vram_mask)]); \
     }
 
+#define READ_HIGH(addr, dat) \
+    dat |= (dev->vram[(addr) & (dev->vram_mask)] << 8);
+
 #define MIX(mixmode, dest_dat, src_dat)                                                       \
     {                                                                                         \
         switch ((mixmode) ? (dev->accel.frgd_mix & 0x1f) : (dev->accel.bkgd_mix & 0x1f)) {    \
@@ -191,7 +194,10 @@ ibm8514_log(const char *fmt, ...)
                 dest_dat = MAX(0, (src_dat - dest_dat));                                      \
                 break;                                                                        \
             case 0x1b:                                                                        \
-                dest_dat = MIN(~0, (dest_dat + src_dat));                                     \
+                if (dev->bpp)                                                                 \
+                    dest_dat = MIN(0xffff, (dest_dat + src_dat));                             \
+                else                                                                          \
+                    dest_dat = MIN(0xff, (dest_dat + src_dat));                               \
                 break;                                                                        \
             case 0x1c:                                                                        \
                 dest_dat = MAX(0, (dest_dat - src_dat)) / 2;                                  \
@@ -203,7 +209,10 @@ ibm8514_log(const char *fmt, ...)
                 dest_dat = MAX(0, (src_dat - dest_dat)) / 2;                                  \
                 break;                                                                        \
             case 0x1f:                                                                        \
-                dest_dat = (~0 < (src_dat + dest_dat)) ? ~0 : ((src_dat + dest_dat) >> 1); 	  \
+                if (dev->bpp)                                                                 \
+                    dest_dat = (0xffff < (src_dat + dest_dat)) ? 0xffff : ((src_dat + dest_dat) / 2); \
+                else                                                                          \
+                    dest_dat = (0xff < (src_dat + dest_dat)) ? 0xff : ((src_dat + dest_dat) / 2); \
                 break;                                                                        \
         }                                                                                     \
     }
@@ -259,10 +268,9 @@ ibm8514_accel_out_pixtrans(svga_t *svga, UNUSED(uint16_t port), uint32_t val, in
     int        frgd_mix  = (dev->accel.frgd_mix >> 5) & 3;
     int        bkgd_mix  = (dev->accel.bkgd_mix >> 5) & 3;
     int        cmd       = dev->accel.cmd >> 13;
-    int        and3      = dev->accel.cur_x & 3;
 
     if (dev->accel.cmd & 0x100) {
-        if (len != 1) {
+        if (len == 2) {
             /*Bus size*/
             if (dev->accel.cmd & 0x200) /*16-bit*/
                 pixcnt = 16;
@@ -271,200 +279,49 @@ ibm8514_accel_out_pixtrans(svga_t *svga, UNUSED(uint16_t port), uint32_t val, in
 
             /*Pixel transfer data mode, can't be the same as Foreground/Background CPU data*/
             if (pixcntl == 2) {
-                if ((frgd_mix == 2) || (bkgd_mix == 2)) {
+                if ((frgd_mix == 2) || (bkgd_mix == 2))
                     pixelxfer = val;
-                } else {
-                    if (dev->accel.cmd & 2) {
+                else {
+                    if (dev->accel.cmd & 0x02) {
                         if (pixcnt == 16) {
                             if ((cmd >= 2) && (dev->accel.cmd & 0x1000))
                                 val = (val >> 8) | (val << 8);
                         }
-                        if (and3 == 3) {
-                            if (dev->accel.cmd & 0x1000)
-                                goto regular_nibble;
-                            if (val & 0x02)
-                                nibble |= 0x10;
-                            if (val & 0x04)
-                                nibble |= 0x08;
-                            if (val & 0x08)
-                                nibble |= 0x04;
-                            if (val & 0x10)
-                                nibble |= 0x02;
-                            if (val & 0x200)
-                                nibble |= 0x01;
-                            if (val & 0x400)
-                                nibble |= 0x80;
-                            if (val & 0x800)
-                                nibble |= 0x40;
-                            if (val & 0x1000)
-                                nibble |= 0x20;
-                        } else if (and3 == 2) {
-                            if (dev->accel.cmd & 0x1000)
-                                goto regular_nibble;
-                            if (val & 0x02)
-                                nibble |= 0x20;
-                            if (val & 0x04)
-                                nibble |= 0x10;
-                            if (val & 0x08)
-                                nibble |= 0x08;
-                            if (val & 0x10)
-                                nibble |= 0x04;
-                            if (val & 0x200)
-                                nibble |= 0x02;
-                            if (val & 0x400)
-                                nibble |= 0x01;
-                            if (val & 0x800)
-                                nibble |= 0x80;
-                            if (val & 0x1000)
-                                nibble |= 0x40;
-                        } else if (and3 == 1) {
-                            if (dev->accel.cmd & 0x1000)
-                                goto regular_nibble;
-                            if (val & 0x02)
-                                nibble |= 0x40;
-                            if (val & 0x04)
-                                nibble |= 0x20;
-                            if (val & 0x08)
-                                nibble |= 0x10;
-                            if (val & 0x10)
-                                nibble |= 0x08;
-                            if (val & 0x200)
-                                nibble |= 0x04;
-                            if (val & 0x400)
-                                nibble |= 0x02;
-                            if (val & 0x800)
-                                nibble |= 0x01;
-                            if (val & 0x1000)
-                                nibble |= 0x80;
-                        } else {
-regular_nibble:
-                            if (val & 0x02)
-                                nibble |= 0x80;
-                            if (val & 0x04)
-                                nibble |= 0x40;
-                            if (val & 0x08)
-                                nibble |= 0x20;
-                            if (val & 0x10)
-                                nibble |= 0x10;
-                            if (val & 0x200)
-                                nibble |= 0x08;
-                            if (val & 0x400)
-                                nibble |= 0x04;
-                            if (val & 0x800)
-                                nibble |= 0x02;
-                            if (val & 0x1000)
-                                nibble |= 0x01;
-                        }
-
-                        if ((and3 == 0) || (dev->accel.cmd & 0x1000) || ((dev->accel.cmd & 8) && ibm8514_cpu_src(svga))) {
-                            if ((dev->accel.cmd & 8) && ibm8514_cpu_src(svga)) {
+                        if ((cmd <= 2) || (cmd == 4)) {
+                            if ((dev->accel.cmd & 0x08) && (cmd >= 2))
                                 monoxfer = val;
-                            } else
-                                monoxfer = nibble;
-                            ibm8514_accel_start(pixcnt, 1, monoxfer, pixelxfer, svga, len);
-                            if (dev->accel.nibbleset != NULL) {
-                                free(dev->accel.nibbleset);
-                                dev->accel.nibbleset = NULL;
-                            }
-                            if (dev->accel.writemono != NULL) {
-                                free(dev->accel.writemono);
-                                dev->accel.writemono = NULL;
-                            }
-                            return;
-                        }
-
-                        dev->accel.writemono[dev->accel.x_count] = nibble;
-                        if (val & 0x1c00) {
-                            if (and3 == 1) {
+                            else {
+                                if (val & 0x02)
+                                    nibble |= 0x80;
+                                if (val & 0x04)
+                                    nibble |= 0x40;
+                                if (val & 0x08)
+                                    nibble |= 0x20;
+                                if (val & 0x10)
+                                    nibble |= 0x10;
+                                if (val & 0x200)
+                                    nibble |= 0x08;
+                                if (val & 0x400)
+                                    nibble |= 0x04;
+                                if (val & 0x800)
+                                    nibble |= 0x02;
                                 if (val & 0x1000)
-                                    dev->accel.nibbleset[dev->accel.x_count] = 0x80;
-                                else
-                                    dev->accel.nibbleset[dev->accel.x_count] = 0;
-                            } else if (and3 == 2) {
-                                if (val & 0x1000) {
-                                    if (val & 0x800)
-                                        dev->accel.nibbleset[dev->accel.x_count] = 0xc0;
-                                    else
-                                        dev->accel.nibbleset[dev->accel.x_count] = 0x40;
-                                } else if (val & 0x800) {
-                                    if (val & 0x1000)
-                                        dev->accel.nibbleset[dev->accel.x_count] = 0xc0;
-                                    else
-                                        dev->accel.nibbleset[dev->accel.x_count] = 0x80;
-                                } else
-                                    dev->accel.nibbleset[dev->accel.x_count] = 0;
-                            } else if (and3 == 3) {
-                                if (val & 0x1000) {
-                                    if (val & 0x800) {
-                                        if (val & 0x400)
-                                            dev->accel.nibbleset[dev->accel.x_count] = 0xe0;
-                                        else
-                                            dev->accel.nibbleset[dev->accel.x_count] = 0x60;
-                                    } else if (val & 0x400) {
-                                        if (val & 0x800)
-                                            dev->accel.nibbleset[dev->accel.x_count] = 0xe0;
-                                        else
-                                            dev->accel.nibbleset[dev->accel.x_count] = 0xa0;
-                                    } else
-                                        dev->accel.nibbleset[dev->accel.x_count] = 0x20;
-                                } else if (val & 0x800) {
-                                    if (val & 0x400) {
-                                        if (val & 0x1000)
-                                            dev->accel.nibbleset[dev->accel.x_count] = 0xe0;
-                                        else
-                                            dev->accel.nibbleset[dev->accel.x_count] = 0xc0;
-                                    } else if (val & 0x1000) {
-                                        if (val & 0x400)
-                                            dev->accel.nibbleset[dev->accel.x_count] = 0xe0;
-                                        else
-                                            dev->accel.nibbleset[dev->accel.x_count] = 0x60;
-                                    } else
-                                        dev->accel.nibbleset[dev->accel.x_count] = 0x40;
-                                } else if (val & 0x400) {
-                                    if (val & 0x800) {
-                                        if (val & 0x1000)
-                                            dev->accel.nibbleset[dev->accel.x_count] = 0xe0;
-                                        else
-                                            dev->accel.nibbleset[dev->accel.x_count] = 0xc0;
-                                    } else if (val & 0x1000) {
-                                        if (val & 0x800)
-                                            dev->accel.nibbleset[dev->accel.x_count] = 0xe0;
-                                        else
-                                            dev->accel.nibbleset[dev->accel.x_count] = 0xa0;
-                                    } else
-                                        dev->accel.nibbleset[dev->accel.x_count] = 0x80;
-                                } else
-                                    dev->accel.nibbleset[dev->accel.x_count] = 0;
+                                    nibble |= 0x01;
+
+                                monoxfer = nibble;
                             }
                         } else
-                            dev->accel.nibbleset[dev->accel.x_count] = 0;
-
-                        dev->accel.x_count++;
-                        if (dev->accel.x_count == dev->accel.sys_cnt) {
-                            for (int i = 0; i < dev->accel.x_count; i++) {
-                                dev->accel.writemono[i] &= ~dev->accel.nibbleset[i];
-                                dev->accel.writemono[i] |= dev->accel.nibbleset[i + 1];
-                                ibm8514_accel_start(pixcnt, 1, dev->accel.writemono[i], pixelxfer, svga, len);
-                            }
-
-                            dev->accel.x_count = 0;
-                            if (dev->accel.nibbleset != NULL) {
-                                free(dev->accel.nibbleset);
-                                dev->accel.nibbleset = NULL;
-                            }
-                            if (dev->accel.writemono != NULL) {
-                                free(dev->accel.writemono);
-                                dev->accel.writemono = NULL;
-                            }
-                        }
-                        return;
-                    }
-                    monoxfer = val;
+                            monoxfer = val;
+                    } else
+                        monoxfer = val;
                 }
-            } else {
+            } else
                 pixelxfer = val;
-            }
-            ibm8514_accel_start(pixcnt, 1, monoxfer, pixelxfer, svga, len);
+
+            if (dev->accel.input)
+                ibm8514_accel_start(pixcnt >> 1, 1, monoxfer & 0xff, pixelxfer & 0xff, svga, len);
+            else
+                ibm8514_accel_start(pixcnt, 1, monoxfer & 0xffff, pixelxfer & 0xffff, svga, len);
         }
     }
 }
@@ -480,115 +337,61 @@ ibm8514_accel_out_fifo(svga_t *svga, uint16_t port, uint32_t val, int len)
     switch (port) {
         case 0x82e8:
         case 0xc2e8:
-            if (len == 1)
-                dev->accel.cur_y = (dev->accel.cur_y & 0x700) | val;
-            else
+            if (len == 2)
                 dev->accel.cur_y = val & 0x7ff;
             break;
-        case 0x82e9:
-        case 0xc2e9:
-            if (len == 1)
-                dev->accel.cur_y = (dev->accel.cur_y & 0xff) | ((val & 0x07) << 8);
-            break;
 
         case 0x86e8:
         case 0xc6e8:
-            if (len == 1)
-                dev->accel.cur_x = (dev->accel.cur_x & 0x700) | val;
-            else
+            if (len == 2)
                 dev->accel.cur_x = val & 0x7ff;
             break;
-        case 0x86e9:
-        case 0xc6e9:
-            if (len == 1) {
-                dev->accel.cur_x = (dev->accel.cur_x & 0xff) | ((val & 0x07) << 8);
-            }
-            break;
 
         case 0x8ae8:
         case 0xcae8:
-            if (len == 1)
-                dev->accel.desty_axstp = (dev->accel.desty_axstp & 0x3f00) | val;
-            else {
-                dev->accel.desty       = val & 0x07ff;
+            if (len == 2) {
+                dev->accel.desty       = val & 0x7ff;
                 dev->accel.desty_axstp = val & 0x3fff;
                 if (val & 0x2000)
                     dev->accel.desty_axstp |= ~0x1fff;
             }
             break;
-        case 0x8ae9:
-        case 0xcae9:
-            if (len == 1) {
-                dev->accel.desty_axstp = (dev->accel.desty_axstp & 0xff) | ((val & 0x3f) << 8);
-                if (val & 0x20)
-                    dev->accel.desty_axstp |= ~0x1fff;
-            }
-            break;
 
         case 0x8ee8:
         case 0xcee8:
-            if (len == 1)
-                dev->accel.destx_distp = (dev->accel.destx_distp & 0x3f00) | val;
-            else {
-                dev->accel.destx       = val & 0x07ff;
+            if (len == 2) {
+                dev->accel.destx       = val & 0x7ff;
                 dev->accel.destx_distp = val & 0x3fff;
                 if (val & 0x2000)
                     dev->accel.destx_distp |= ~0x1fff;
             }
             break;
-        case 0x8ee9:
-        case 0xcee9:
-            if (len == 1) {
-                dev->accel.destx_distp = (dev->accel.destx_distp & 0xff) | ((val & 0x3f) << 8);
-                if (val & 0x20)
-                    dev->accel.destx_distp |= ~0x1fff;
-            }
-            break;
 
         case 0x92e8:
-            if (len != 1)
+            if (len == 2)
                 dev->test = val;
             fallthrough;
 
         case 0xd2e8:
-            if (len == 1)
-                dev->accel.err_term = (dev->accel.err_term & 0x3f00) | val;
-            else {
+            if (len == 2) {
                 dev->accel.err_term = val & 0x3fff;
                 if (val & 0x2000)
                     dev->accel.err_term |= ~0x1fff;
             }
             break;
-        case 0x92e9:
-        case 0xd2e9:
-            if (len == 1) {
-                dev->accel.err_term = (dev->accel.err_term & 0xff) | ((val & 0x3f) << 8);
-                if (val & 0x20)
-                    dev->accel.err_term |= ~0x1fff;
-            }
-            break;
 
         case 0x96e8:
         case 0xd6e8:
-            if (len == 1)
-                dev->accel.maj_axis_pcnt = (dev->accel.maj_axis_pcnt & 0x700) | val;
-            else {
+            if (len == 2) {
                 dev->accel.maj_axis_pcnt = val & 0x7ff;
                 dev->accel.maj_axis_pcnt_no_limit = val;
             }
             break;
-        case 0x96e9:
-        case 0xd6e9:
-            if (len == 1)
-                dev->accel.maj_axis_pcnt = (dev->accel.maj_axis_pcnt & 0xff) | ((val & 0x07) << 8);
-            break;
 
         case 0x9ae8:
         case 0xdae8:
             dev->accel.ssv_state = 0;
-            if (len == 1)
-                dev->accel.cmd = (dev->accel.cmd & 0xff00) | val;
-            else {
+            if (len == 2) {
                 dev->data_available  = 0;
                 dev->data_available2 = 0;
                 dev->accel.cmd       = val;
@@ -596,20 +399,7 @@ ibm8514_accel_out_fifo(svga_t *svga, uint16_t port, uint32_t val, int len)
                     if (dev->accel.cmd & 0x100)
                         dev->accel.cmd_back = 0;
                 }
-                ibm8514_log("8514/A CMD=%04x.\n", dev->accel.cmd);
-                ibm8514_accel_start(-1, 0, -1, 0, svga, len);
-            }
-            break;
-        case 0x9ae9:
-        case 0xdae9:
-            if (len == 1) {
-                dev->data_available  = 0;
-                dev->data_available2 = 0;
-                dev->accel.cmd       = (dev->accel.cmd & 0xff) | (val << 8);
-                if (port == 0xdae9) {
-                    if (dev->accel.cmd & 0x100)
-                        dev->accel.cmd_back = 0;
-                }
+                ibm8514_log("8514/A CMD=%04x, back=%d.\n", dev->accel.cmd, dev->accel.cmd_back);
                 ibm8514_accel_start(-1, 0, -1, 0, svga, len);
             }
             break;
@@ -617,35 +407,14 @@ ibm8514_accel_out_fifo(svga_t *svga, uint16_t port, uint32_t val, int len)
         case 0x9ee8:
         case 0xdee8:
             dev->accel.ssv_state = 1;
-            if (len == 1)
-                dev->accel.short_stroke = (dev->accel.short_stroke & 0xff00) | val;
-            else {
+            if (len == 2) {
                 dev->accel.short_stroke = val;
-                dev->accel.cx           = dev->accel.cur_x;
-                dev->accel.cy           = dev->accel.cur_y;
-                if (dev->accel.cur_x >= 0x600)
-                    dev->accel.cx |= ~0x5ff;
-                if (dev->accel.cur_y >= 0x600)
-                    dev->accel.cy |= ~0x5ff;
 
-                if (dev->accel.cmd & 0x1000) {
-                    ibm8514_short_stroke_start(-1, 0, -1, 0, svga, dev->accel.short_stroke & 0xff, len);
-                    ibm8514_short_stroke_start(-1, 0, -1, 0, svga, dev->accel.short_stroke >> 8, len);
-                } else {
-                    ibm8514_short_stroke_start(-1, 0, -1, 0, svga, dev->accel.short_stroke >> 8, len);
-                    ibm8514_short_stroke_start(-1, 0, -1, 0, svga, dev->accel.short_stroke & 0xff, len);
-                }
-            }
-            break;
-        case 0x9ee9:
-        case 0xdee9:
-            dev->accel.ssv_state = 1;
-            if (len == 1) {
-                dev->accel.short_stroke = (dev->accel.short_stroke & 0xff) | (val << 8);
-                dev->accel.cx           = dev->accel.cur_x;
-                dev->accel.cy           = dev->accel.cur_y;
+                dev->accel.cx = dev->accel.cur_x;
                 if (dev->accel.cur_x >= 0x600)
                     dev->accel.cx |= ~0x5ff;
+
+                dev->accel.cy = dev->accel.cur_y;
                 if (dev->accel.cur_y >= 0x600)
                     dev->accel.cy |= ~0x5ff;
 
@@ -662,93 +431,56 @@ ibm8514_accel_out_fifo(svga_t *svga, uint16_t port, uint32_t val, int len)
         case 0xa2e8:
         case 0xe2e8:
             if (port == 0xe2e8) {
-                if (dev->accel.cmd_back) {
-                    if (len == 1)
-                        dev->accel.bkgd_color = (dev->accel.bkgd_color & 0x00ff) | val;
-                    else
+                if (len == 2) {
+                    if (dev->accel.cmd_back)
                         dev->accel.bkgd_color = val;
-                } else {
-                    if (ibm8514_cpu_dest(svga))
-                        break;
-                    ibm8514_accel_out_pixtrans(svga, port, val, len);
+                    else {
+                        if (ibm8514_cpu_dest(svga))
+                            break;
+                        ibm8514_accel_out_pixtrans(svga, port, val, len);
+                    }
                 }
             } else {
-                if (len == 1)
-                    dev->accel.bkgd_color = (dev->accel.bkgd_color & 0x00ff) | val;
-                else
+                if (len == 2)
                     dev->accel.bkgd_color = val;
             }
             break;
-        case 0xa2e9:
-        case 0xe2e9:
-            if (len == 1)
-                dev->accel.bkgd_color = (dev->accel.bkgd_color & 0xff00) | (val << 8);
-            break;
 
         case 0xa6e8:
         case 0xe6e8:
             if (port == 0xe6e8) {
-                if (dev->accel.cmd_back) {
-                    if (len == 1)
-                        dev->accel.frgd_color = (dev->accel.frgd_color & 0x00ff) | val;
-                    else
+                if (len == 2) {
+                    if (dev->accel.cmd_back)
                         dev->accel.frgd_color = val;
-                } else {
-                    if (ibm8514_cpu_dest(svga))
-                        break;
-                    ibm8514_accel_out_pixtrans(svga, port, val, len);
+                    else {
+                        if (ibm8514_cpu_dest(svga))
+                            break;
+                        ibm8514_accel_out_pixtrans(svga, port, val, len);
+                    }
                 }
             } else {
-                if (len == 1)
-                    dev->accel.frgd_color = (dev->accel.frgd_color & 0x00ff) | val;
-                else
+                if (len == 2)
                     dev->accel.frgd_color = val;
             }
             break;
-        case 0xa6e9:
-        case 0xe6e9:
-            if (len == 1)
-                dev->accel.frgd_color = (dev->accel.frgd_color & 0xff00) | (val << 8);
-            break;
 
         case 0xaae8:
         case 0xeae8:
-            if (len == 1)
-                dev->accel.wrt_mask = (dev->accel.wrt_mask & 0x00ff) | val;
-            else
+            if (len == 2)
                 dev->accel.wrt_mask = val;
             break;
-        case 0xaae9:
-        case 0xeae9:
-            if (len == 1)
-                dev->accel.wrt_mask = (dev->accel.wrt_mask & 0xff00) | (val << 8);
-            break;
 
         case 0xaee8:
         case 0xeee8:
-            if (len == 1)
-                dev->accel.rd_mask = (dev->accel.rd_mask & 0x00ff) | val;
-            else
+            if (len == 2)
                 dev->accel.rd_mask = val;
             break;
-        case 0xaee9:
-        case 0xeee9:
-            if (len == 1)
-                dev->accel.rd_mask = (dev->accel.rd_mask & 0xff00) | (val << 8);
-            break;
 
         case 0xb2e8:
         case 0xf2e8:
-            if (len == 1)
-                dev->accel.color_cmp = (dev->accel.color_cmp & 0x00ff) | val;
-            else
+            if (len == 2)
                 dev->accel.color_cmp = val;
             break;
-        case 0xb2e9:
-        case 0xf2e9:
-            if (len == 1)
-                dev->accel.color_cmp = (dev->accel.color_cmp & 0xff00) | (val << 8);
-            break;
 
         case 0xb6e8:
         case 0xf6e8:
@@ -762,45 +494,10 @@ ibm8514_accel_out_fifo(svga_t *svga, uint16_t port, uint32_t val, int len)
 
         case 0xbee8:
         case 0xfee8:
-            if (len == 1)
-                dev->accel.multifunc_cntl = (dev->accel.multifunc_cntl & 0xff00) | val;
-            else {
+            if (len == 2) {
                 dev->accel.multifunc_cntl                             = val;
                 dev->accel.multifunc[dev->accel.multifunc_cntl >> 12] = dev->accel.multifunc_cntl & 0xfff;
-                if ((dev->accel.multifunc_cntl >> 12) == 1)
-                    dev->accel.clip_top = val & 0x7ff;
-
-                if ((dev->accel.multifunc_cntl >> 12) == 2)
-                    dev->accel.clip_left = val & 0x7ff;
-
-                if ((dev->accel.multifunc_cntl >> 12) == 3)
-                    dev->accel.multifunc[3] = val & 0x7ff;
-
-                if ((dev->accel.multifunc_cntl >> 12) == 4)
-                    dev->accel.multifunc[4] = val & 0x7ff;
-
-                ibm8514_log("CLIPBOTTOM=%d, CLIPRIGHT=%d, bpp=%d, pitch=%d.\n", dev->accel.multifunc[3], dev->accel.multifunc[4], dev->accel_bpp, dev->pitch);
-                if (port == 0xfee8)
-                    dev->accel.cmd_back = 1;
-                else
-                    dev->accel.cmd_back = 0;
-            }
-            break;
-        case 0xbee9:
-        case 0xfee9:
-            if (len == 1) {
-                dev->accel.multifunc_cntl                             = (dev->accel.multifunc_cntl & 0xff) | (val << 8);
-                dev->accel.multifunc[dev->accel.multifunc_cntl >> 12] = dev->accel.multifunc_cntl & 0xfff;
-                if ((dev->accel.multifunc_cntl >> 12) == 1)
-                    dev->accel.clip_top = dev->accel.multifunc_cntl & 0x7ff;
-
-                if ((dev->accel.multifunc_cntl >> 12) == 2)
-                    dev->accel.clip_left = dev->accel.multifunc_cntl & 0x7ff;
-
-                if (port == 0xfee9)
-                    dev->accel.cmd_back = 1;
-                else
-                    dev->accel.cmd_back = 0;
+                dev->accel.cmd_back = !!(port == 0xfee8);
             }
             break;
 
@@ -902,10 +599,8 @@ ibm8514_accel_out(uint16_t port, uint32_t val, svga_t *svga, int len)
 
             case 0x6e8:
                 /*In preparation to switch from VGA to 8514/A mode*/
-                if (!dev->on[0] || !dev->on[1]) {
-                    dev->hdisped = val;
-                    dev->hdisp = (dev->hdisped + 1) << 3;
-                }
+                dev->hdisped = val;
+                dev->hdisp = (dev->hdisped + 1) << 3;
                 ibm8514_log("[%04X:%08X]: IBM 8514/A: (0x%04x): hdisp=0x%02x.\n", CS, cpu_state.pc, port, val);
                 svga_recalctimings(svga);
                 break;
@@ -938,11 +633,9 @@ ibm8514_accel_out(uint16_t port, uint32_t val, svga_t *svga, int len)
             case 0x16e8:
             case 0x16e9:
                 /*In preparation to switch from VGA to 8514/A mode*/
-                if (!dev->on[0] || !dev->on[1]) {
-                    WRITE8(port, dev->v_disp, val);
-                    dev->v_disp &= 0x1fff;
-                    dev->vdisp = (dev->v_disp + 1) >> 1;
-                }
+                WRITE8(port, dev->v_disp, val);
+                dev->v_disp &= 0x1fff;
+                dev->vdisp = (dev->v_disp + 1) >> 1;
                 ibm8514_log("IBM 8514/A: V_DISP write 16E8 = %d\n", dev->v_disp);
                 ibm8514_log("IBM 8514/A: (0x%04x): vdisp=0x%02x.\n", port, val);
                 svga_recalctimings(svga);
@@ -1000,12 +693,10 @@ ibm8514_accel_out(uint16_t port, uint32_t val, svga_t *svga, int len)
                 break;
 
             case 0x4ae8:
-            case 0x4ae9:
                 WRITE8(port, dev->accel.advfunc_cntl, val);
-                dev->on[0] = dev->accel.advfunc_cntl & 0x01;
-                vga_on = !dev->on[0];
-                dev->vendor_mode[0] = 0;
-                ibm8514_log("[%04X:%08X]: IBM 8514/A: (0x%04x): ON=%d, shadow crt=%x, hdisp=%d, vdisp=%d.\n", CS, cpu_state.pc, port, dev->on[port & 1], dev->accel.advfunc_cntl & 0x04, dev->hdisp, dev->vdisp);
+                dev->on = dev->accel.advfunc_cntl & 0x01;
+                vga_on = !dev->on;
+                ibm8514_log("[%04X:%08X]: IBM 8514/A: (0x%04x): ON=%d, shadow crt=%x, hdisp=%d, vdisp=%d.\n", CS, cpu_state.pc, port, dev->on, dev->accel.advfunc_cntl & 0x04, dev->hdisp, dev->vdisp);
                 ibm8514_log("IBM mode set %s resolution.\n", (dev->accel.advfunc_cntl & 0x04) ? "2: 1024x768" : "1: 640x480");
                 svga_recalctimings(svga);
                 break;
@@ -1042,29 +733,29 @@ ibm8514_accel_in_fifo(svga_t *svga, uint16_t port, int len)
     switch (port) {
         case 0x82e8:
         case 0xc2e8:
-            if (len != 1)
+            if (len == 2)
                 temp = dev->accel.cur_y;
             break;
 
         case 0x86e8:
         case 0xc6e8:
-            if (len != 1)
+            if (len == 2)
                 temp = dev->accel.cur_x;
             break;
 
         case 0x92e8:
-            if (len != 1)
+            if (len == 2)
                 temp = dev->test;
             break;
 
         case 0x96e8:
-            if (len != 1)
+            if (len == 2)
                 temp = dev->accel.maj_axis_pcnt;
             break;
 
         case 0x9ae8:
         case 0xdae8:
-            if (len != 1) {
+            if (len == 2) {
                 if (dev->force_busy)
                     temp |= 0x200; /*Hardware busy*/
                 dev->force_busy = 0;
@@ -1090,14 +781,19 @@ ibm8514_accel_in_fifo(svga_t *svga, uint16_t port, int len)
         case 0xe2e8:
         case 0xe6e8:
             if (ibm8514_cpu_dest(svga)) {
-                if (len != 1) {
+                if (len == 2) {
                     cmd = (dev->accel.cmd >> 13);
                     READ_PIXTRANS_WORD(dev->accel.cx, 0);
-                    if (dev->accel.input && !dev->accel.odd_in && !dev->accel.sx) {
-                        temp &= ~0xff00;
-                        temp |= (dev->vram[(dev->accel.newdest_in + dev->accel.cur_x) & dev->vram_mask] << 8);
-                    }
-                    ibm8514_accel_out_pixtrans(svga, port, temp, len);
+                    if (dev->accel.input) {
+                        ibm8514_accel_out_pixtrans(svga, port, temp & 0xff, len);
+                        if (dev->accel.odd_in) { /*WORDs on odd destination scan lengths.*/
+                            dev->accel.odd_in = 0;
+                            temp &= ~0xff00;
+                            READ_HIGH(dev->accel.dest + dev->accel.cx, temp);
+                        }
+                        ibm8514_accel_out_pixtrans(svga, port, (temp >> 8) & 0xff, len);
+                    } else
+                        ibm8514_accel_out_pixtrans(svga, port, temp, len);
                 }
             }
             break;
@@ -1113,9 +809,6 @@ ibm8514_accel_in(uint16_t port, svga_t *svga)
 {
     ibm8514_t *dev       = (ibm8514_t *) svga->dev8514;
     uint8_t   temp       = 0;
-    uint16_t  clip_b_ibm = dev->accel.multifunc[3];
-    uint16_t  clip_r_ibm = dev->accel.multifunc[4];
-    int       cmd        = (dev->accel.cmd >> 13);
 
     switch (port) {
         case 0x2e8:
@@ -1150,19 +843,8 @@ ibm8514_accel_in(uint16_t port, svga_t *svga)
             if (dev->vc == dev->v_syncstart)
                 dev->subsys_stat |= 1;
 
-            if (cmd == 6) {
-                if (((dev->accel.dx) >= dev->accel.clip_left) && ((dev->accel.dx) <= clip_r_ibm) && ((dev->accel.dy) >= dev->accel.clip_top) && ((dev->accel.dy) <= clip_b_ibm))
-                    temp |= 2;
-            } else {
-                if (((dev->accel.cx) >= dev->accel.clip_left) && ((dev->accel.dx) <= clip_r_ibm) && ((dev->accel.cy) >= dev->accel.clip_top) && ((dev->accel.cy) <= clip_b_ibm))
-                    temp |= 2;
-            }
-
-            if (!dev->force_busy)
-                temp |= 8;
-
             if (port & 1)
-                temp = 0x80;
+                temp = dev->vram_512k_8514 ? 0x00 : 0x80;
             else {
                 temp |= (dev->subsys_stat | 0x80);
                 temp |= 0x20;
@@ -1232,8 +914,10 @@ ibm8514_accel_start(int count, int cpu_input, uint32_t mix_dat, uint32_t cpu_dat
     uint16_t   old_dest_dat;
     int        frgd_mix;
     int        bkgd_mix;
-    uint16_t   clip_b          = dev->accel.multifunc[3];
-    uint16_t   clip_r          = dev->accel.multifunc[4];
+    int16_t    clip_t          = dev->accel.multifunc[1] & 0x7ff;
+    int16_t    clip_l          = dev->accel.multifunc[2] & 0x7ff;
+    uint16_t   clip_b          = dev->accel.multifunc[3] & 0x7ff;
+    uint16_t   clip_r          = dev->accel.multifunc[4] & 0x7ff;
     int        pixcntl         = (dev->accel.multifunc[0x0a] >> 6) & 3;
     uint16_t   mix_mask        = dev->bpp ? 0x8000 : 0x80;
     uint16_t   compare         = dev->accel.color_cmp;
@@ -1265,11 +949,11 @@ ibm8514_accel_start(int count, int cpu_input, uint32_t mix_dat, uint32_t cpu_dat
     bkgd_mix = (dev->accel.bkgd_mix >> 5) & 3;
 
     if (cpu_input) {
-        if ((dev->accel.cmd & 2) || (pixcntl == 2)) {
+        if ((dev->accel.cmd & 0x02) || (pixcntl == 2)) {
             if ((frgd_mix == 2) || (bkgd_mix == 2))
                 count >>= 3;
             else if (pixcntl == 2) {
-                if (dev->accel.cmd & 2)
+                if (dev->accel.cmd & 0x02)
                     count >>= 1;
                 else
                     count >>= 3;
@@ -1369,9 +1053,12 @@ ibm8514_accel_start(int count, int cpu_input, uint32_t mix_dat, uint32_t cpu_dat
             if (dev->accel.ssv_state == 0)
                 break;
 
-            if (dev->accel.cmd & 8) {
+            if (dev->accel.cmd & 0x08) {
                 while (count-- && dev->accel.ssv_len >= 0) {
-                    if (dev->accel.cx >= dev->accel.clip_left && dev->accel.cx <= clip_r && dev->accel.cy >= dev->accel.clip_top && dev->accel.cy <= clip_b) {
+                    if ((dev->accel.cx >= clip_l) &&
+                        (dev->accel.cx <= clip_r) &&
+                        (dev->accel.cy >= clip_t) &&
+                        (dev->accel.cy <= clip_b)) {
                         switch ((mix_dat & mix_mask) ? frgd_mix : bkgd_mix) {
                             case 0:
                                 src_dat = bkgd_color;
@@ -1389,17 +1076,24 @@ ibm8514_accel_start(int count, int cpu_input, uint32_t mix_dat, uint32_t cpu_dat
                             default:
                                 break;
                         }
+
                         READ((dev->accel.cy * dev->pitch) + dev->accel.cx, dest_dat);
 
-                        if ((compare_mode == 0) || ((compare_mode == 0x10) && (dest_dat >= compare)) || ((compare_mode == 0x18) && (dest_dat < compare)) || ((compare_mode == 0x20) && (dest_dat != compare)) || ((compare_mode == 0x28) && (dest_dat == compare)) || ((compare_mode == 0x30) && (dest_dat <= compare)) || ((compare_mode == 0x38) && (dest_dat > compare))) {
+                        if ((compare_mode == 0) ||
+                            ((compare_mode == 0x10) && (dest_dat >= compare)) ||
+                            ((compare_mode == 0x18) && (dest_dat < compare)) ||
+                            ((compare_mode == 0x20) && (dest_dat != compare)) ||
+                            ((compare_mode == 0x28) && (dest_dat == compare)) ||
+                            ((compare_mode == 0x30) && (dest_dat <= compare)) ||
+                            ((compare_mode == 0x38) && (dest_dat > compare))) {
                             old_dest_dat = dest_dat;
                             MIX(mix_dat & mix_mask, dest_dat, src_dat);
                             dest_dat = (dest_dat & wrt_mask) | (old_dest_dat & ~wrt_mask);
 
                             if (dev->accel.ssv_draw) {
-                                if ((dev->accel.cmd & 4) && dev->accel.ssv_len) {
+                                if ((dev->accel.cmd & 0x04) && dev->accel.ssv_len) {
                                     WRITE((dev->accel.cy * dev->pitch) + dev->accel.cx, dest_dat);
-                                } else if (!(dev->accel.cmd & 4)) {
+                                } else if (!(dev->accel.cmd & 0x04)) {
                                     WRITE((dev->accel.cy * dev->pitch) + dev->accel.cx, dest_dat);
                                 }
                             }
@@ -1454,8 +1148,10 @@ ibm8514_accel_start(int count, int cpu_input, uint32_t mix_dat, uint32_t cpu_dat
                 }
             } else {
                 while (count-- && (dev->accel.ssv_len >= 0)) {
-                    if ((dev->accel.cx >= dev->accel.clip_left) && (dev->accel.cx <= clip_r) &&
-                        (dev->accel.cy >= dev->accel.clip_top) && (dev->accel.cy <= clip_b)) {
+                    if ((dev->accel.cx >= clip_l) &&
+                        (dev->accel.cx <= clip_r) &&
+                        (dev->accel.cy >= clip_t) &&
+                        (dev->accel.cy <= clip_b)) {
                         switch ((mix_dat & mix_mask) ? frgd_mix : bkgd_mix) {
                             case 0:
                                 src_dat = bkgd_color;
@@ -1476,15 +1172,21 @@ ibm8514_accel_start(int count, int cpu_input, uint32_t mix_dat, uint32_t cpu_dat
 
                         READ((dev->accel.cy * dev->pitch) + dev->accel.cx, dest_dat);
 
-                        if ((compare_mode == 0) || ((compare_mode == 0x10) && (dest_dat >= compare)) || ((compare_mode == 0x18) && (dest_dat < compare)) || ((compare_mode == 0x20) && (dest_dat != compare)) || ((compare_mode == 0x28) && (dest_dat == compare)) || ((compare_mode == 0x30) && (dest_dat <= compare)) || ((compare_mode == 0x38) && (dest_dat > compare))) {
+                        if ((compare_mode == 0) ||
+                            ((compare_mode == 0x10) && (dest_dat >= compare)) ||
+                            ((compare_mode == 0x18) && (dest_dat < compare)) ||
+                            ((compare_mode == 0x20) && (dest_dat != compare)) ||
+                            ((compare_mode == 0x28) && (dest_dat == compare)) ||
+                            ((compare_mode == 0x30) && (dest_dat <= compare)) ||
+                            ((compare_mode == 0x38) && (dest_dat > compare))) {
                             old_dest_dat = dest_dat;
                             MIX(mix_dat & mix_mask, dest_dat, src_dat);
                             dest_dat = (dest_dat & wrt_mask) | (old_dest_dat & ~wrt_mask);
 
                             if (dev->accel.ssv_draw) {
-                                if ((dev->accel.cmd & 4) && dev->accel.ssv_len) {
+                                if ((dev->accel.cmd & 0x04) && dev->accel.ssv_len) {
                                     WRITE((dev->accel.cy * dev->pitch) + dev->accel.cx, dest_dat);
-                                } else if (!(dev->accel.cmd & 4)) {
+                                } else if (!(dev->accel.cmd & 0x04)) {
                                     WRITE((dev->accel.cy * dev->pitch) + dev->accel.cx, dest_dat);
                                 }
                             }
@@ -1576,47 +1278,25 @@ ibm8514_accel_start(int count, int cpu_input, uint32_t mix_dat, uint32_t cpu_dat
 
         case 1: /*Draw line*/
             if (!cpu_input) {
-                dev->accel.xx_count = 0;
-                dev->accel.cx       = dev->accel.cur_x;
-                dev->accel.cy       = dev->accel.cur_y;
+                dev->accel.output = 0;
+                dev->accel.x_count = 0;
 
+                dev->accel.cx = dev->accel.cur_x;
                 if (dev->accel.cur_x >= 0x600)
                     dev->accel.cx |= ~0x5ff;
 
+                dev->accel.cy = dev->accel.cur_y;
                 if (dev->accel.cur_y >= 0x600)
                     dev->accel.cy |= ~0x5ff;
 
                 dev->accel.sy = dev->accel.maj_axis_pcnt;
 
-                ibm8514_log("Line Draw 8514/A, frgdmix=%d, bkgdmix=%d, c(%d,%d), pixcntl=%d, sy=%d, polyfill=%x, selfrmix=%02x, selbkmix=%02x, bkgdcol=%02x, frgdcol=%02x, clipt=%d, clipb=%d.\n", frgd_mix, bkgd_mix, dev->accel.cx, dev->accel.cy, pixcntl, dev->accel.sy, dev->accel.multifunc[0x0a] & 6, dev->accel.frgd_mix & 0x1f, dev->accel.bkgd_mix & 0x1f, bkgd_color, frgd_color, dev->accel.clip_top, clip_b);
+                ibm8514_log("Line Draw 8514/A CMD=%04x, frgdmix=%d, bkgdmix=%d, c(%d,%d), pixcntl=%d, sy=%d, polyfill=%x, selfrmix=%02x, selbkmix=%02x, bkgdcol=%02x, frgdcol=%02x, clipt=%d, clipb=%d.\n", dev->accel.cmd, frgd_mix, bkgd_mix, dev->accel.cx, dev->accel.cy, pixcntl, dev->accel.sy, dev->accel.multifunc[0x0a] & 6, dev->accel.frgd_mix & 0x1f, dev->accel.bkgd_mix & 0x1f, bkgd_color, frgd_color, dev->accel.clip_top, clip_b);
                 if (ibm8514_cpu_src(svga)) {
-                    if (dev->accel.cmd & 2) {
-                        if (dev->accel.cmd & 8) {
-                            if (and3 == 1) {
-                                dev->accel.sy += 4;
-                                if (dev->accel.cmd & 0x20)
-                                    dev->accel.cx += 4;
-                                else
-                                    dev->accel.cx -= 4;
-                            } else if (and3 == 2) {
-                                dev->accel.sy += 5;
-                                if (dev->accel.cmd & 0x20)
-                                    dev->accel.cx += 5;
-                                else
-                                    dev->accel.cx -= 5;
-                            } else if (and3 == 3) {
-                                dev->accel.sy += 6;
-                                if (dev->accel.cmd & 0x20)
-                                    dev->accel.cx += 6;
-                                else
-                                    dev->accel.cx -= 6;
-                            } else {
-                                dev->accel.sy += 3;
-                                if (dev->accel.cmd & 0x20)
-                                    dev->accel.cx += 3;
-                                else
-                                    dev->accel.cx -= 3;
-                            }
+                    if (dev->accel.cmd & 0x02) {
+                        if (!(dev->accel.cmd & 0x1000)) {
+                            if (dev->accel.cmd & 0x08)
+                                dev->accel.output = 1;
                         }
                     }
                     dev->data_available  = 0;
@@ -1629,12 +1309,18 @@ ibm8514_accel_start(int count, int cpu_input, uint32_t mix_dat, uint32_t cpu_dat
                 }
             }
 
-            if (dev->accel.cmd & 8) { /*Vector Line*/
-                if (ibm8514_cpu_dest(svga) && cpu_input && (dev->accel.cmd & 2))
+            if (dev->accel.cmd & 0x08) { /*Vector Line*/
+                if (ibm8514_cpu_dest(svga) && cpu_input && (dev->accel.cmd & 0x02))
                     count >>= 1;
-                dev->accel.xx_count++;
+
+                if (dev->accel.cmd & 0x02)
+                    ibm8514_log("Line Draw Vector Single pixtrans=%04x, count=%d.\n", mix_dat, count);
+
                 while (count-- && (dev->accel.sy >= 0)) {
-                    if (dev->accel.cx >= dev->accel.clip_left && dev->accel.cx <= clip_r && dev->accel.cy >= dev->accel.clip_top && dev->accel.cy <= clip_b) {
+                    if ((dev->accel.cx >= clip_l) &&
+                        (dev->accel.cx <= clip_r) &&
+                        (dev->accel.cy >= clip_t) &&
+                        (dev->accel.cy <= clip_b)) {
                         if (ibm8514_cpu_dest(svga) && (pixcntl == 0)) {
                             mix_dat = mix_mask; /* Mix data = forced to foreground register. */
                         } else if (ibm8514_cpu_dest(svga) && (pixcntl == 3)) {
@@ -1649,94 +1335,72 @@ ibm8514_accel_start(int count, int cpu_input, uint32_t mix_dat, uint32_t cpu_dat
                             if (pixcntl == 3)
                                 src_dat = ((src_dat & rd_mask) == rd_mask);
                         } else {
-                            switch ((mix_dat & mix_mask) ? frgd_mix : bkgd_mix) {
-                                case 0:
-                                    src_dat = bkgd_color;
-                                    break;
-                                case 1:
-                                    src_dat = frgd_color;
-                                    break;
-                                case 2:
-                                    src_dat = cpu_dat;
-                                    break;
-                                case 3:
-                                    src_dat = 0;
-                                    break;
+                            if (dev->accel.output) {
+                                switch ((mix_dat & 0x01) ? frgd_mix : bkgd_mix) {
+                                    case 0:
+                                        src_dat = bkgd_color;
+                                        break;
+                                    case 1:
+                                        src_dat = frgd_color;
+                                        break;
+                                    case 2:
+                                        src_dat = cpu_dat;
+                                        break;
+                                    case 3:
+                                        src_dat = 0;
+                                        break;
 
-                                default:
-                                    break;
+                                    default:
+                                        break;
+                                }
+                            } else {
+                                switch ((mix_dat & mix_mask) ? frgd_mix : bkgd_mix) {
+                                    case 0:
+                                        src_dat = bkgd_color;
+                                        break;
+                                    case 1:
+                                        src_dat = frgd_color;
+                                        break;
+                                    case 2:
+                                        src_dat = cpu_dat;
+                                        break;
+                                    case 3:
+                                        src_dat = 0;
+                                        break;
+
+                                    default:
+                                        break;
+                                }
                             }
                         }
 
                         READ((dev->accel.cy * dev->pitch) + dev->accel.cx, dest_dat);
 
-                        if ((compare_mode == 0) || ((compare_mode == 0x10) && (dest_dat >= compare)) || ((compare_mode == 0x18) && (dest_dat < compare)) || ((compare_mode == 0x20) && (dest_dat != compare)) || ((compare_mode == 0x28) && (dest_dat == compare)) || ((compare_mode == 0x30) && (dest_dat <= compare)) || ((compare_mode == 0x38) && (dest_dat > compare))) {
+                        if ((compare_mode == 0) ||
+                            ((compare_mode == 0x10) && (dest_dat >= compare)) ||
+                            ((compare_mode == 0x18) && (dest_dat < compare)) ||
+                            ((compare_mode == 0x20) && (dest_dat != compare)) ||
+                            ((compare_mode == 0x28) && (dest_dat == compare)) ||
+                            ((compare_mode == 0x30) && (dest_dat <= compare)) ||
+                            ((compare_mode == 0x38) && (dest_dat > compare))) {
                             old_dest_dat = dest_dat;
-                            MIX(mix_dat & mix_mask, dest_dat, src_dat);
+                            if (dev->accel.output) {
+                                MIX(mix_dat & 0x01, dest_dat, src_dat);
+                            } else {
+                                MIX(mix_dat & mix_mask, dest_dat, src_dat);
+                            }
                             dest_dat = (dest_dat & wrt_mask) | (old_dest_dat & ~wrt_mask);
-                            if ((dev->accel.cmd & 2) && ibm8514_cpu_src(svga)) {
-                                if (and3 == 1) {
-                                    if (dev->accel.xx_count >= 2) {
-                                        if ((dev->accel.cmd & 4) && dev->accel.sy) {
-                                            WRITE((dev->accel.cy * dev->pitch) + dev->accel.cx, dest_dat);
-                                        } else if (!(dev->accel.cmd & 4)) {
-                                            WRITE((dev->accel.cy * dev->pitch) + dev->accel.cx, dest_dat);
-                                        }
-                                    }
-                                } else if (and3 == 2) {
-                                    if (dev->accel.xx_count == 2) {
-                                        if (count <= 2) {
-                                            if ((dev->accel.cmd & 4) && dev->accel.sy) {
-                                                WRITE((dev->accel.cy * dev->pitch) + dev->accel.cx, dest_dat);
-                                            } else if (!(dev->accel.cmd & 4)) {
-                                                WRITE((dev->accel.cy * dev->pitch) + dev->accel.cx, dest_dat);
-                                            }
-                                        }
-                                    } else if (dev->accel.xx_count >= 3) {
-                                        if ((dev->accel.cmd & 4) && dev->accel.sy) {
-                                            WRITE((dev->accel.cy * dev->pitch) + dev->accel.cx, dest_dat);
-                                        } else if (!(dev->accel.cmd & 4)) {
-                                            WRITE((dev->accel.cy * dev->pitch) + dev->accel.cx, dest_dat);
-                                        }
-                                    }
-                                } else if (and3 == 3) {
-                                    if (dev->accel.xx_count == 2) {
-                                        if (count <= 1) {
-                                            if ((dev->accel.cmd & 4) && dev->accel.sy) {
-                                                WRITE((dev->accel.cy * dev->pitch) + dev->accel.cx, dest_dat);
-                                            } else if (!(dev->accel.cmd & 4)) {
-                                                WRITE((dev->accel.cy * dev->pitch) + dev->accel.cx, dest_dat);
-                                            }
-                                        }
-                                    } else if (dev->accel.xx_count >= 3) {
-                                        if ((dev->accel.cmd & 4) && dev->accel.sy) {
-                                            WRITE((dev->accel.cy * dev->pitch) + dev->accel.cx, dest_dat);
-                                        } else if (!(dev->accel.cmd & 4)) {
-                                            WRITE((dev->accel.cy * dev->pitch) + dev->accel.cx, dest_dat);
-                                        }
-                                    }
-                                } else {
-                                    if (dev->accel.xx_count == 1) {
-                                        if (!count) {
-                                            if ((dev->accel.cmd & 4) && dev->accel.sy) {
-                                                WRITE((dev->accel.cy * dev->pitch) + dev->accel.cx, dest_dat);
-                                            } else if (!(dev->accel.cmd & 4)) {
-                                                WRITE((dev->accel.cy * dev->pitch) + dev->accel.cx, dest_dat);
-                                            }
-                                        }
-                                    } else if (dev->accel.xx_count >= 2) {
-                                        if ((dev->accel.cmd & 4) && dev->accel.sy) {
-                                            WRITE((dev->accel.cy * dev->pitch) + dev->accel.cx, dest_dat);
-                                        } else if (!(dev->accel.cmd & 4)) {
-                                            WRITE((dev->accel.cy * dev->pitch) + dev->accel.cx, dest_dat);
-                                        }
-                                    }
+                            if ((dev->accel.cmd & 0x02) && ibm8514_cpu_src(svga)) {
+                                if ((dev->accel.cmd & 0x04) && dev->accel.sy) {
+                                    WRITE((dev->accel.cy * dev->pitch) + dev->accel.cx, dest_dat);
+                                } else if (!(dev->accel.cmd & 0x04)) {
+                                    WRITE((dev->accel.cy * dev->pitch) + dev->accel.cx, dest_dat);
                                 }
                             } else {
                                 if (ibm8514_cpu_src(svga) || !cpu_input) {
-                                    if ((dev->accel.cmd & 4) && dev->accel.sy) {
+                                    if ((dev->accel.cmd & 0x04) && dev->accel.sy) {
                                         WRITE((dev->accel.cy * dev->pitch) + dev->accel.cx, dest_dat);
-                                    } else if (!(dev->accel.cmd & 4)) {
+                                    } else if (!(dev->accel.cmd & 0x04)) {
                                         WRITE((dev->accel.cy * dev->pitch) + dev->accel.cx, dest_dat);
                                     }
                                 }
@@ -1744,17 +1408,21 @@ ibm8514_accel_start(int count, int cpu_input, uint32_t mix_dat, uint32_t cpu_dat
                         }
                     }
 
-                    mix_dat <<= 1;
-                    mix_dat |= 1;
+                    if (!dev->accel.sy)
+                        break;
+
+                    if (dev->accel.output)
+                        mix_dat >>= 1;
+                    else {
+                        mix_dat <<= 1;
+                        mix_dat |= 1;
+                    }
+
                     if (dev->bpp)
                         cpu_dat >>= 16;
                     else
                         cpu_dat >>= 8;
 
-                    if (dev->accel.sy == 0) {
-                        break;
-                    }
-
                     switch (dev->accel.cmd & 0xe0) {
                         case 0x00:
                             dev->accel.cx++;
@@ -1793,19 +1461,24 @@ ibm8514_accel_start(int count, int cpu_input, uint32_t mix_dat, uint32_t cpu_dat
                 }
                 dev->accel.cur_x = dev->accel.cx;
                 dev->accel.cur_y = dev->accel.cy;
-            } else { /*Bresenham*/
+                dev->accel.x_count = 0;
+                dev->accel.output = 0;
+            } else { /*Bresenham Line*/
                 if (pixcntl == 1) {
                     dev->accel.temp_cnt = 8;
                     while (count-- && (dev->accel.sy >= 0)) {
-                        if (dev->accel.temp_cnt == 0) {
+                        if (!dev->accel.temp_cnt) {
                             dev->accel.temp_cnt = 8;
-                            mix_dat             = old_mix_dat;
+                            mix_dat = old_mix_dat;
                         }
-                        if (dev->accel.cx >= dev->accel.clip_left && dev->accel.cx <= clip_r && dev->accel.cy >= dev->accel.clip_top && dev->accel.cy <= clip_b) {
+                        if ((dev->accel.cx >= clip_l) &&
+                            (dev->accel.cx <= clip_r) &&
+                            (dev->accel.cy >= clip_t) &&
+                            (dev->accel.cy <= clip_b)) {
                             if (ibm8514_cpu_dest(svga)) {
                                 READ((dev->accel.cy * dev->pitch) + dev->accel.cx, src_dat);
                             } else
-                                switch ((mix_dat & 1) ? frgd_mix : bkgd_mix) {
+                                switch ((mix_dat & 0x01) ? frgd_mix : bkgd_mix) {
                                     case 0:
                                         src_dat = bkgd_color;
                                         break;
@@ -1825,13 +1498,19 @@ ibm8514_accel_start(int count, int cpu_input, uint32_t mix_dat, uint32_t cpu_dat
 
                             READ((dev->accel.cy * dev->pitch) + dev->accel.cx, dest_dat);
 
-                            if ((compare_mode == 0) || ((compare_mode == 0x10) && (dest_dat >= compare)) || ((compare_mode == 0x18) && (dest_dat < compare)) || ((compare_mode == 0x20) && (dest_dat != compare)) || ((compare_mode == 0x28) && (dest_dat == compare)) || ((compare_mode == 0x30) && (dest_dat <= compare)) || ((compare_mode == 0x38) && (dest_dat > compare))) {
+                            if ((compare_mode == 0) ||
+                                ((compare_mode == 0x10) && (dest_dat >= compare)) ||
+                                ((compare_mode == 0x18) && (dest_dat < compare)) ||
+                                ((compare_mode == 0x20) && (dest_dat != compare)) ||
+                                ((compare_mode == 0x28) && (dest_dat == compare)) ||
+                                ((compare_mode == 0x30) && (dest_dat <= compare)) ||
+                                ((compare_mode == 0x38) && (dest_dat > compare))) {
                                 old_dest_dat = dest_dat;
-                                MIX(mix_dat & 1, dest_dat, src_dat);
+                                MIX(mix_dat & 0x01, dest_dat, src_dat);
                                 dest_dat = (dest_dat & wrt_mask) | (old_dest_dat & ~wrt_mask);
-                                if ((dev->accel.cmd & 4) && dev->accel.sy) {
+                                if ((dev->accel.cmd & 0x04) && dev->accel.sy) {
                                     WRITE((dev->accel.cy * dev->pitch) + dev->accel.cx, dest_dat);
-                                } else if (!(dev->accel.cmd & 4)) {
+                                } else if (!(dev->accel.cmd & 0x04)) {
                                     WRITE((dev->accel.cy * dev->pitch) + dev->accel.cx, dest_dat);
                                 }
                             }
@@ -1844,9 +1523,8 @@ ibm8514_accel_start(int count, int cpu_input, uint32_t mix_dat, uint32_t cpu_dat
                         else
                             cpu_dat >>= 8;
 
-                        if (dev->accel.sy == 0) {
+                        if (dev->accel.sy == 0)
                             break;
-                        }
 
                         if (dev->accel.cmd & 0x40) {
                             if (dev->accel.cmd & 0x80)
@@ -1882,7 +1560,10 @@ ibm8514_accel_start(int count, int cpu_input, uint32_t mix_dat, uint32_t cpu_dat
                     }
                 } else {
                     while (count-- && (dev->accel.sy >= 0)) {
-                        if ((dev->accel.cx) >= dev->accel.clip_left && (dev->accel.cx) <= clip_r && (dev->accel.cy) >= dev->accel.clip_top && (dev->accel.cy) <= clip_b) {
+                        if ((dev->accel.cx >= clip_l) &&
+                            (dev->accel.cx <= clip_r) &&
+                            (dev->accel.cy >= clip_t) &&
+                            (dev->accel.cy <= clip_b)) {
                             if (ibm8514_cpu_dest(svga) && (pixcntl == 0)) {
                                 mix_dat = mix_mask; /* Mix data = forced to foreground register. */
                             } else if (ibm8514_cpu_dest(svga) && (pixcntl == 3)) {
@@ -1917,13 +1598,19 @@ ibm8514_accel_start(int count, int cpu_input, uint32_t mix_dat, uint32_t cpu_dat
 
                             READ((dev->accel.cy * dev->pitch) + dev->accel.cx, dest_dat);
 
-                            if ((compare_mode == 0) || ((compare_mode == 0x10) && (dest_dat >= compare)) || ((compare_mode == 0x18) && (dest_dat < compare)) || ((compare_mode == 0x20) && (dest_dat != compare)) || ((compare_mode == 0x28) && (dest_dat == compare)) || ((compare_mode == 0x30) && (dest_dat <= compare)) || ((compare_mode == 0x38) && (dest_dat > compare))) {
+                            if ((compare_mode == 0) ||
+                                ((compare_mode == 0x10) && (dest_dat >= compare)) ||
+                                ((compare_mode == 0x18) && (dest_dat < compare)) ||
+                                ((compare_mode == 0x20) && (dest_dat != compare)) ||
+                                ((compare_mode == 0x28) && (dest_dat == compare)) ||
+                                ((compare_mode == 0x30) && (dest_dat <= compare)) ||
+                                ((compare_mode == 0x38) && (dest_dat > compare))) {
                                 old_dest_dat = dest_dat;
                                 MIX(mix_dat & mix_mask, dest_dat, src_dat);
                                 dest_dat = (dest_dat & wrt_mask) | (old_dest_dat & ~wrt_mask);
-                                if ((dev->accel.cmd & 4) && dev->accel.sy) {
+                                if ((dev->accel.cmd & 0x04) && dev->accel.sy) {
                                     WRITE((dev->accel.cy * dev->pitch) + dev->accel.cx, dest_dat);
-                                } else if (!(dev->accel.cmd & 4)) {
+                                } else if (!(dev->accel.cmd & 0x04)) {
                                     WRITE((dev->accel.cy * dev->pitch) + dev->accel.cx, dest_dat);
                                 }
                             }
@@ -1981,79 +1668,60 @@ ibm8514_accel_start(int count, int cpu_input, uint32_t mix_dat, uint32_t cpu_dat
         case 3: /*Rectangle fill (Y direction)*/
         case 4: /*Rectangle fill (Y direction using nibbles)*/
             if (!cpu_input) {
-                dev->accel.x_count     = 0;
-                dev->accel.xx_count    = 0;
-                dev->accel.odd_out     = 0;
-                dev->accel.odd_in      = 0;
-                dev->accel.input       = 0;
-                dev->accel.output      = 0;
-                dev->accel.newdest_out = 0;
-                dev->accel.newdest_in  = 0;
-
-                dev->accel.sx = dev->accel.maj_axis_pcnt & 0x7ff;
-                dev->accel.sy = dev->accel.multifunc[0] & 0x7ff;
+                dev->accel.x_count = 0;
+                dev->accel.output = 0;
+                dev->accel.input = 0;
+                dev->accel.input2 = 0;
+                dev->accel.odd_in = 0;
 
                 dev->accel.cx = dev->accel.cur_x;
                 if (dev->accel.cur_x >= 0x600)
                     dev->accel.cx |= ~0x5ff;
+
                 dev->accel.cy = dev->accel.cur_y;
                 if (dev->accel.cur_y >= 0x600)
                     dev->accel.cy |= ~0x5ff;
 
-                if (((dev->local & 0xff) >= 0x02) && dev->accel.ge_offset && ((dev->accel_bpp == 24) || (dev->accel_bpp == 8)))
+                dev->accel.sx = dev->accel.maj_axis_pcnt & 0x7ff;
+                dev->accel.sy = dev->accel.multifunc[0] & 0x7ff;
+
+                if ((dev->accel_bpp == 24) || (dev->accel_bpp <= 8))
                     dev->accel.dest = (dev->accel.ge_offset << 2) + (dev->accel.cy * dev->pitch);
+                else if (dev->bpp)
+                    dev->accel.dest = (dev->accel.ge_offset << 1) + (dev->accel.cy * dev->pitch);
                 else
-                    dev->accel.dest = dev->accel.cy * dev->pitch;
+                    dev->accel.dest = (dev->accel.ge_offset << 2) + (dev->accel.cy * dev->pitch);
 
                 if (cmd == 4)
-                    dev->accel.cmd |= 2;
+                    dev->accel.cmd |= 0x02;
                 else if (cmd == 3)
-                    dev->accel.cmd &= ~2;
+                    dev->accel.cmd &= ~0x02;
 
                 if (ibm8514_cpu_src(svga)) {
-                    if (dev->accel.cmd & 2) {
+                    if (dev->accel.cmd & 0x02) {
                         if (!(dev->accel.cmd & 0x1000)) {
-                            if (!(dev->accel.cmd & 8)) {
-                                dev->accel.sx += and3;
-                                dev->accel.nibbleset = (uint8_t *) calloc(1, (dev->accel.sx >> 3) + 1);
-                                dev->accel.writemono = (uint8_t *) calloc(1, (dev->accel.sx >> 3) + 1);
-                                dev->accel.sys_cnt   = (dev->accel.sx >> 3) + 1;
+                            if (dev->accel.cmd & 0x08) {
+                                dev->accel.x_count = dev->accel.cx - (and3 + 3);
+                                dev->accel.sx += (and3 + 3);
                             } else {
-                                if (and3 == 1) {
-                                    dev->accel.sx += 4;
+                                dev->accel.x_count = dev->accel.cx;
+                                if (and3) {
                                     if (dev->accel.cmd & 0x20)
-                                        dev->accel.cx += 4;
+                                        dev->accel.x_count -= and3;
                                     else
-                                        dev->accel.cx -= 4;
-                                } else if (and3 == 2) {
-                                    dev->accel.sx += 5;
-                                    if (dev->accel.cmd & 0x20)
-                                        dev->accel.cx += 5;
-                                    else
-                                        dev->accel.cx -= 5;
-                                } else if (and3 == 3) {
-                                    dev->accel.sx += 6;
-                                    if (dev->accel.cmd & 0x20)
-                                        dev->accel.cx += 6;
-                                    else
-                                        dev->accel.cx -= 6;
-                                } else {
-                                    dev->accel.sx += 3;
-                                    if (dev->accel.cmd & 0x20)
-                                        dev->accel.cx += 3;
-                                    else
-                                        dev->accel.cx -= 3;
+                                        dev->accel.x_count += and3;
+
+                                    dev->accel.sx += 8;
                                 }
                             }
                         }
                     } else {
-                        if (!(dev->accel.cmd & 0x40) && (frgd_mix == 2) && (bkgd_mix == 2) && (pixcntl == 0) && (cmd == 2)) {
-                            if (!(dev->accel.sx & 1)) {
-                                dev->accel.output = 1;
-                                if (((dev->local & 0xff) >= 0x02) && dev->accel.ge_offset && ((dev->accel_bpp == 24) || (dev->accel_bpp == 8)))
-                                    dev->accel.newdest_out = (dev->accel.ge_offset << 2) + ((dev->accel.cy + 1) * dev->pitch);
-                                else
-                                    dev->accel.newdest_out = (dev->accel.cy + 1) * dev->pitch;
+                        if (dev->accel.cmd & 0x1000) {
+                            if (dev->accel.cmd & 0x200) {
+                                if (dev->accel.cmd & 0x04) {
+                                    dev->accel.output = 1;
+                                    dev->accel.sx -= 2;
+                                }
                             }
                         }
                     }
@@ -2061,35 +1729,35 @@ ibm8514_accel_start(int count, int cpu_input, uint32_t mix_dat, uint32_t cpu_dat
                     dev->data_available2 = 0;
                     return; /*Wait for data from CPU*/
                 } else if (ibm8514_cpu_dest(svga)) {
-                    if (!(dev->accel.cmd & 2) && (frgd_mix == 2) && (pixcntl == 0) && (cmd == 2)) {
-                        if (!(dev->accel.sx & 1)) {
-                            dev->accel.input      = 1;
-                            if (((dev->local & 0xff) >= 0x02) && dev->accel.ge_offset && ((dev->accel_bpp == 24) || (dev->accel_bpp == 8)))
-                                dev->accel.newdest_in = (dev->accel.ge_offset << 2) + ((dev->accel.cy + 1) * dev->pitch);
-                            else
-                                dev->accel.newdest_in = (dev->accel.cy + 1) * dev->pitch;
-                        }
-                    } else if (dev->accel.cmd & 2) {
-                        if (dev->accel.cmd & 8) {
-                            dev->accel.sx += and3;
-                            dev->accel.nibbleset = (uint8_t *) calloc(1, (dev->accel.sx >> 3) + 1);
-                            dev->accel.writemono = (uint8_t *) calloc(1, (dev->accel.sx >> 3) + 1);
-                            dev->accel.sys_cnt   = (dev->accel.sx >> 3) + 1;
+                    if (!(dev->accel.cmd & 0x02)) {
+                        if (dev->accel.cmd & 0x1000) {
+                            if (dev->accel.cmd & 0x200) {
+                                if (!(dev->accel.sx & 1) && !(dev->accel.cmd & 0x04)) {
+                                    dev->accel.input = 1;
+                                } else if (dev->accel.cmd & 0x04) {
+                                    dev->accel.input2 = 1;
+                                    dev->accel.sx -= 2;
+                                }
+                            }
                         }
                     }
+                    ibm8514_log("INPUT=%d.\n", dev->accel.input);
                     dev->data_available  = 1;
                     dev->data_available2 = 1;
                     return; /*Wait for data from CPU*/
                 }
             }
 
-            if (dev->accel.cmd & 2) {
+            ibm8514_log("Rectangle %d: full=%04x, odd=%d, c(%d,%d), frgdmix=%d, bkgdmix=%d, xcount=%d, and3=%d, len(%d,%d), CURX=%d, Width=%d, pixcntl=%d, mix_dat=%08x, count=%d, cpu_data=%08x, cpu_input=%d.\n", cmd, dev->accel.cmd, dev->accel.input, dev->accel.cx, dev->accel.cy, frgd_mix, bkgd_mix, dev->accel.x_count, and3, dev->accel.sx, dev->accel.sy, dev->accel.cur_x, dev->accel.maj_axis_pcnt, pixcntl, mix_dat, count, cpu_dat, cpu_input);
+
+            if (dev->accel.cmd & 0x08) { /*Vectored Rectangle*/
                 if (cpu_input) {
-rect_fill_pix:
-                    if ((dev->accel.cmd & 8) && ibm8514_cpu_src(svga)) {
-                        dev->accel.xx_count++;
+                    if (ibm8514_cpu_src(svga)) {
                         while (count-- && (dev->accel.sy >= 0)) {
-                            if (dev->accel.cx >= dev->accel.clip_left && dev->accel.cx <= clip_r && dev->accel.cy >= dev->accel.clip_top && dev->accel.cy <= clip_b) {
+                            if ((dev->accel.cx >= clip_l) &&
+                                (dev->accel.cx <= clip_r) &&
+                                (dev->accel.cy >= clip_t) &&
+                                (dev->accel.cy <= clip_b)) {
                                 switch ((mix_dat & mix_mask) ? frgd_mix : bkgd_mix) {
                                     case 0:
                                         src_dat = bkgd_color;
@@ -2110,77 +1778,27 @@ rect_fill_pix:
 
                                 READ(dev->accel.dest + dev->accel.cx, dest_dat);
 
-                                if ((compare_mode == 0) || ((compare_mode == 0x10) && (dest_dat >= compare)) || ((compare_mode == 0x18) && (dest_dat < compare)) || ((compare_mode == 0x20) && (dest_dat != compare)) || ((compare_mode == 0x28) && (dest_dat == compare)) || ((compare_mode == 0x30) && (dest_dat <= compare)) || ((compare_mode == 0x38) && (dest_dat > compare))) {
+                                if ((compare_mode == 0) ||
+                                    ((compare_mode == 0x10) && (dest_dat >= compare)) ||
+                                    ((compare_mode == 0x18) && (dest_dat < compare)) ||
+                                    ((compare_mode == 0x20) && (dest_dat != compare)) ||
+                                    ((compare_mode == 0x28) && (dest_dat == compare)) ||
+                                    ((compare_mode == 0x30) && (dest_dat <= compare)) ||
+                                    ((compare_mode == 0x38) && (dest_dat > compare))) {
                                     old_dest_dat = dest_dat;
                                     MIX(mix_dat & mix_mask, dest_dat, src_dat);
                                     dest_dat = (dest_dat & wrt_mask) | (old_dest_dat & ~wrt_mask);
-                                    if (and3 == 1) {
-                                        if (dev->accel.xx_count >= 2) {
-                                            if ((dev->accel.cmd & 4) && dev->accel.sx) {
-                                                WRITE(dev->accel.dest + dev->accel.cx, dest_dat);
-                                            } else if (!(dev->accel.cmd & 4)) {
-                                                WRITE(dev->accel.dest + dev->accel.cx, dest_dat);
-                                            }
-                                        }
-                                    } else if (and3 == 2) {
-                                        if (dev->accel.xx_count == 2) {
-                                            if (count <= 2) {
-                                                if ((dev->accel.cmd & 4) && dev->accel.sx) {
-                                                    WRITE(dev->accel.dest + dev->accel.cx, dest_dat);
-                                                } else if (!(dev->accel.cmd & 4)) {
-                                                    WRITE(dev->accel.dest + dev->accel.cx, dest_dat);
-                                                }
-                                            }
-                                        } else if (dev->accel.xx_count >= 3) {
-                                            if ((dev->accel.cmd & 4) && dev->accel.sx) {
-                                                WRITE(dev->accel.dest + dev->accel.cx, dest_dat);
-                                            } else if (!(dev->accel.cmd & 4)) {
-                                                WRITE(dev->accel.dest + dev->accel.cx, dest_dat);
-                                            }
-                                        }
-                                    } else if (and3 == 3) {
-                                        if (dev->accel.xx_count == 2) {
-                                            if (count <= 1) {
-                                                if ((dev->accel.cmd & 4) && dev->accel.sx) {
-                                                    WRITE(dev->accel.dest + dev->accel.cx, dest_dat);
-                                                } else if (!(dev->accel.cmd & 4)) {
-                                                    WRITE(dev->accel.dest + dev->accel.cx, dest_dat);
-                                                }
-                                            }
-                                        } else if (dev->accel.xx_count >= 3) {
-                                            if ((dev->accel.cmd & 4) && dev->accel.sx) {
-                                                WRITE(dev->accel.dest + dev->accel.cx, dest_dat);
-                                            } else if (!(dev->accel.cmd & 4)) {
-                                                WRITE(dev->accel.dest + dev->accel.cx, dest_dat);
-                                            }
-                                        }
-                                    } else {
-                                        if (dev->accel.xx_count == 1) {
-                                            if (!count) {
-                                                if ((dev->accel.cmd & 4) && dev->accel.sx) {
-                                                    WRITE(dev->accel.dest + dev->accel.cx, dest_dat);
-                                                } else if (!(dev->accel.cmd & 4)) {
-                                                    WRITE(dev->accel.dest + dev->accel.cx, dest_dat);
-                                                }
-                                            }
-                                        } else if (dev->accel.xx_count >= 2) {
-                                            if ((dev->accel.cmd & 4) && dev->accel.sx) {
-                                                WRITE(dev->accel.dest + dev->accel.cx, dest_dat);
-                                            } else if (!(dev->accel.cmd & 4)) {
-                                                WRITE(dev->accel.dest + dev->accel.cx, dest_dat);
-                                            }
-                                        }
+                                    if ((dev->accel.cmd & 0x02) && (dev->accel.x_count != dev->accel.cx))
+                                        goto skip_vector_rect_write;
+
+                                    if ((dev->accel.cmd & 0x04) && dev->accel.sx) {
+                                        WRITE(dev->accel.dest + dev->accel.cx, dest_dat);
+                                    } else if (!(dev->accel.cmd & 0x04)) {
+                                        WRITE(dev->accel.dest + dev->accel.cx, dest_dat);
                                     }
                                 }
                             }
 
-                            mix_dat <<= 1;
-                            mix_dat |= 1;
-                            if (dev->bpp)
-                                cpu_dat >>= 16;
-                            else
-                                cpu_dat >>= 8;
-
                             switch (dev->accel.cmd & 0xe0) {
                                 case 0x00:
                                     dev->accel.cx++;
@@ -2205,17 +1823,42 @@ rect_fill_pix:
                                     break;
                             }
 
+skip_vector_rect_write:
+                            switch (dev->accel.cmd & 0xe0) {
+                                case 0x00:
+                                    dev->accel.x_count++;
+                                    break;
+                                case 0x20:
+                                    dev->accel.x_count++;
+                                    break;
+                                case 0x60:
+                                    dev->accel.x_count--;
+                                    break;
+                                case 0x80:
+                                    dev->accel.x_count--;
+                                    break;
+                                case 0xa0:
+                                    dev->accel.x_count--;
+                                    break;
+                                case 0xe0:
+                                    dev->accel.x_count++;
+                                    break;
+
+                                default:
+                                    break;
+                            }
+
+                            if (dev->bpp)
+                                cpu_dat >>= 16;
+                            else
+                                cpu_dat >>= 8;
+
+                            mix_dat <<= 1;
+                            mix_dat |= 1;
+
                             dev->accel.sx--;
                             if (dev->accel.sx < 0) {
                                 dev->accel.sx = dev->accel.maj_axis_pcnt & 0x7ff;
-                                if (and3 == 1)
-                                    dev->accel.sx += 4;
-                                else if (and3 == 2)
-                                    dev->accel.sx += 5;
-                                else if (and3 == 3)
-                                    dev->accel.sx += 6;
-                                else
-                                    dev->accel.sx += 3;
 
                                 if (dev->accel.cmd & 0x20)
                                     dev->accel.cx -= (dev->accel.sx + 1);
@@ -2246,34 +1889,62 @@ rect_fill_pix:
                                         break;
                                 }
 
-                                if (((dev->local & 0xff) >= 0x02) && dev->accel.ge_offset && ((dev->accel_bpp == 24) || (dev->accel_bpp == 8)))
+                                if ((dev->accel_bpp == 24) || (dev->accel_bpp <= 8))
                                     dev->accel.dest = (dev->accel.ge_offset << 2) + (dev->accel.cy * dev->pitch);
+                                else if (dev->bpp)
+                                    dev->accel.dest = (dev->accel.ge_offset << 1) + (dev->accel.cy * dev->pitch);
                                 else
-                                    dev->accel.dest = dev->accel.cy * dev->pitch;
+                                    dev->accel.dest = (dev->accel.ge_offset << 2) + (dev->accel.cy * dev->pitch);
 
                                 dev->accel.sy--;
+                                dev->accel.x_count = 0;
                                 return;
                             }
                         }
-                        break;
-                    }
-                    if (count < 8) {
-                        while (count-- && (dev->accel.sy >= 0)) {
-                            if (dev->accel.cx >= dev->accel.clip_left && dev->accel.cx <= clip_r && dev->accel.cy >= dev->accel.clip_top && dev->accel.cy <= clip_b) {
-                                if (ibm8514_cpu_dest(svga) && (pixcntl == 0)) {
-                                    mix_dat = mix_mask; /* Mix data = forced to foreground register. */
-                                } else if (ibm8514_cpu_dest(svga) && (pixcntl == 3)) {
-                                    /* Mix data = current video memory value. */
-                                    READ(dev->accel.dest + dev->accel.cx, mix_dat);
-                                    mix_dat = ((mix_dat & rd_mask) == rd_mask);
-                                    mix_dat = mix_dat ? mix_mask : 0;
-                                }
+                    } else
+                        ibm8514_log("Vectored Rectangle with destination reads (TODO).\n");
+                } else
+                    ibm8514_log("Vectored Rectangle with normal processing (TODO).\n");
+            } else { /*Normal Rectangle*/
+                if (cpu_input) {
+                    while (count-- && (dev->accel.sy >= 0)) {
+                        if ((dev->accel.cx >= clip_l) &&
+                            (dev->accel.cx <= clip_r) &&
+                            (dev->accel.cy >= clip_t) &&
+                            (dev->accel.cy <= clip_b)) {
+                            if (ibm8514_cpu_dest(svga) && (pixcntl == 0)) {
+                                mix_dat = mix_mask; /* Mix data = forced to foreground register. */
+                            } else if (ibm8514_cpu_dest(svga) && (pixcntl == 3)) {
+                                /* Mix data = current video memory value. */
+                                READ(dev->accel.dest + dev->accel.cx, mix_dat);
+                                mix_dat = ((mix_dat & rd_mask) == rd_mask);
+                                mix_dat = mix_dat ? mix_mask : 0;
+                            }
 
-                                if (ibm8514_cpu_dest(svga)) {
-                                    READ(dev->accel.dest + dev->accel.cx, src_dat);
-                                    if (pixcntl == 3)
-                                        src_dat = ((src_dat & rd_mask) == rd_mask);
-                                } else
+                            if (ibm8514_cpu_dest(svga)) {
+                                READ(dev->accel.dest + dev->accel.cx, src_dat);
+                                if (pixcntl == 3)
+                                    src_dat = ((src_dat & rd_mask) == rd_mask);
+                            } else {
+                                if (dev->accel.cmd & 0x02) {
+                                    switch ((mix_dat & 0x01) ? frgd_mix : bkgd_mix) {
+                                        case 0:
+                                            src_dat = bkgd_color;
+                                            break;
+                                        case 1:
+                                            src_dat = frgd_color;
+                                            break;
+                                        case 2:
+                                            src_dat = cpu_dat;
+                                            break;
+                                        case 3:
+                                            src_dat = 0;
+                                            break;
+
+                                        default:
+                                            break;
+                                    }
+                                } else {
                                     switch ((mix_dat & mix_mask) ? frgd_mix : bkgd_mix) {
                                         case 0:
                                             src_dat = bkgd_color;
@@ -2291,418 +1962,114 @@ rect_fill_pix:
                                         default:
                                             break;
                                     }
+                                }
+                            }
 
-                                READ(dev->accel.dest + dev->accel.cx, dest_dat);
+                            READ(dev->accel.dest + dev->accel.cx, dest_dat);
+
+                            if ((compare_mode == 0) ||
+                                ((compare_mode == 0x10) && (dest_dat >= compare)) ||
+                                ((compare_mode == 0x18) && (dest_dat < compare)) ||
+                                ((compare_mode == 0x20) && (dest_dat != compare)) ||
+                                ((compare_mode == 0x28) && (dest_dat == compare)) ||
+                                ((compare_mode == 0x30) && (dest_dat <= compare)) ||
+                                ((compare_mode == 0x38) && (dest_dat > compare))) {
+                                old_dest_dat = dest_dat;
+                                if (dev->accel.cmd & 0x02) {
+                                    MIX(mix_dat & 0x01, dest_dat, src_dat);
+                                    if ((dev->accel.x_count != dev->accel.cx) && !(dev->accel.cmd & 0x1000) && and3)
+                                        goto skip_nibble_rect_write;
 
-                                if ((compare_mode == 0) || ((compare_mode == 0x10) && (dest_dat >= compare)) || ((compare_mode == 0x18) && (dest_dat < compare)) || ((compare_mode == 0x20) && (dest_dat != compare)) || ((compare_mode == 0x28) && (dest_dat == compare)) || ((compare_mode == 0x30) && (dest_dat <= compare)) || ((compare_mode == 0x38) && (dest_dat > compare))) {
-                                    old_dest_dat = dest_dat;
-                                    MIX(mix_dat & mix_mask, dest_dat, src_dat);
                                     dest_dat = (dest_dat & wrt_mask) | (old_dest_dat & ~wrt_mask);
-                                    WRITE(dev->accel.dest + dev->accel.cx, dest_dat);
-                                }
-                            }
-                            mix_dat <<= 1;
-                            mix_dat |= 1;
-                            if (dev->bpp)
-                                cpu_dat >>= 16;
-                            else
-                                cpu_dat >>= 8;
-
-                            if (dev->accel.cmd & 0x20)
-                                dev->accel.cx++;
-                            else
-                                dev->accel.cx--;
-
-                            dev->accel.sx--;
-                            if (dev->accel.sx < 0) {
-                                dev->accel.sx = dev->accel.maj_axis_pcnt & 0x7ff;
-
-                                if (dev->accel.cmd & 2) {
-                                    dev->accel.sx += (dev->accel.cur_x & 3);
-                                }
-
-                                if (dev->accel.cmd & 0x20) {
-                                    dev->accel.cx -= (dev->accel.sx) + 1;
-                                } else
-                                    dev->accel.cx += (dev->accel.sx) + 1;
-
-                                if (dev->accel.cmd & 0x80)
-                                    dev->accel.cy++;
-                                else
-                                    dev->accel.cy--;
-
-                                if (((dev->local & 0xff) >= 0x02) && dev->accel.ge_offset && ((dev->accel_bpp == 24) || (dev->accel_bpp == 8)))
-                                    dev->accel.dest = (dev->accel.ge_offset << 2) + (dev->accel.cy * dev->pitch);
-                                else
-                                    dev->accel.dest = dev->accel.cy * dev->pitch;
-
-                                dev->accel.sy--;
-                                return;
-                            }
-                        }
-                    } else {
-                        while (count-- && (dev->accel.sy >= 0)) {
-                            if (dev->accel.cx >= dev->accel.clip_left && dev->accel.cx <= clip_r && dev->accel.cy >= dev->accel.clip_top && dev->accel.cy <= clip_b) {
-                                if (ibm8514_cpu_dest(svga) && (pixcntl == 0)) {
-                                    mix_dat = 1; /* Mix data = forced to foreground register. */
-                                } else if (ibm8514_cpu_dest(svga) && (pixcntl == 3)) {
-                                    /* Mix data = current video memory value. */
-                                    READ(dev->accel.dest + dev->accel.cx, mix_dat);
-                                    mix_dat = ((mix_dat & rd_mask) == rd_mask);
-                                    mix_dat = mix_dat ? 1 : 0;
-                                }
-
-                                if (ibm8514_cpu_dest(svga)) {
-                                    READ(dev->accel.dest + dev->accel.cx, src_dat);
-                                    if (pixcntl == 3)
-                                        src_dat = ((src_dat & rd_mask) == rd_mask);
-                                } else {
-                                    switch ((mix_dat & 1) ? frgd_mix : bkgd_mix) {
-                                        case 0:
-                                            src_dat = bkgd_color;
-                                            break;
-                                        case 1:
-                                            src_dat = frgd_color;
-                                            break;
-                                        case 2:
-                                            src_dat = cpu_dat;
-                                            break;
-                                        case 3:
-                                            src_dat = 0;
-                                            break;
-
-                                        default:
-                                            break;
-                                    }
-                                }
-
-                                READ(dev->accel.dest + dev->accel.cx, dest_dat);
-
-                                if ((compare_mode == 0) || ((compare_mode == 0x10) && (dest_dat >= compare)) || ((compare_mode == 0x18) && (dest_dat < compare)) || ((compare_mode == 0x20) && (dest_dat != compare)) || ((compare_mode == 0x28) && (dest_dat == compare)) || ((compare_mode == 0x30) && (dest_dat <= compare)) || ((compare_mode == 0x38) && (dest_dat > compare))) {
-                                    old_dest_dat = dest_dat;
-                                    MIX(mix_dat & 1, dest_dat, src_dat);
-                                    dest_dat = (dest_dat & wrt_mask) | (old_dest_dat & ~wrt_mask);
-                                    WRITE(dev->accel.dest + dev->accel.cx, dest_dat);
-                                }
-                            }
-                            mix_dat >>= 1;
-                            if (dev->bpp)
-                                cpu_dat >>= 16;
-                            else
-                                cpu_dat >>= 8;
-
-                            if (dev->accel.cmd & 0x20)
-                                dev->accel.cx++;
-                            else
-                                dev->accel.cx--;
-
-                            dev->accel.sx--;
-                            if (dev->accel.sx < 0) {
-                                dev->accel.sx = dev->accel.maj_axis_pcnt & 0x7ff;
-
-                                if (dev->accel.cmd & 2) {
-                                    if (!(dev->accel.cmd & 0x1000))
-                                        dev->accel.sx += (dev->accel.cur_x & 3);
-                                }
-
-                                if (dev->accel.cmd & 0x20)
-                                    dev->accel.cx -= (dev->accel.sx) + 1;
-                                else
-                                    dev->accel.cx += (dev->accel.sx) + 1;
-
-                                if (dev->accel.cmd & 2) {
-                                    if (dev->accel.cmd & 0x1000) {
-                                        dev->accel.cx = dev->accel.cur_x;
-                                        if (dev->accel.cur_x >= 0x600)
-                                            dev->accel.cx |= ~0x5ff;
-                                    }
-                                }
-
-                                if (dev->accel.cmd & 0x80)
-                                    dev->accel.cy++;
-                                else
-                                    dev->accel.cy--;
-
-                                if (((dev->local & 0xff) >= 0x02) && dev->accel.ge_offset && ((dev->accel_bpp == 24) || (dev->accel_bpp == 8)))
-                                    dev->accel.dest = (dev->accel.ge_offset << 2) + (dev->accel.cy * dev->pitch);
-                                else
-                                    dev->accel.dest = dev->accel.cy * dev->pitch;
-
-                                dev->accel.sy--;
-                                return;
-                            }
-                        }
-                    }
-                } else {
-                    goto rect_fill;
-                }
-            } else {
-                if (cpu_input) {
-                    if (pixcntl == 2) {
-                        goto rect_fill_pix;
-                    } else {
-                        if (dev->accel.input && !dev->accel.output) {
-                            while (count-- && (dev->accel.sy >= 0)) {
-                                if (dev->accel.cx >= dev->accel.clip_left && dev->accel.cx <= clip_r && dev->accel.cy >= dev->accel.clip_top && dev->accel.cy <= clip_b) {
-                                    mix_dat = mix_mask; /* Mix data = forced to foreground register. */
-                                    if (!dev->accel.odd_in && !dev->accel.sx) {
-                                        READ(dev->accel.newdest_in + dev->accel.cur_x, src_dat);
-                                        READ(dev->accel.newdest_in + dev->accel.cur_x, dest_dat);
-                                    } else {
-                                        READ(dev->accel.dest + dev->accel.cx, src_dat);
-                                        READ(dev->accel.dest + dev->accel.cx, dest_dat);
-                                    }
-                                    if ((compare_mode == 0) || ((compare_mode == 0x10) && (dest_dat >= compare)) || ((compare_mode == 0x18) && (dest_dat < compare)) || ((compare_mode == 0x20) && (dest_dat != compare)) || ((compare_mode == 0x28) && (dest_dat == compare)) || ((compare_mode == 0x30) && (dest_dat <= compare)) || ((compare_mode == 0x38) && (dest_dat > compare))) {
-                                        old_dest_dat = dest_dat;
-                                        MIX(mix_dat & mix_mask, dest_dat, src_dat);
-                                        dest_dat = (dest_dat & wrt_mask) | (old_dest_dat & ~wrt_mask);
-                                        if (!dev->accel.odd_in && !dev->accel.sx) {
-                                            WRITE(dev->accel.newdest_in + dev->accel.cur_x, dest_dat);
-                                        } else {
-                                            WRITE(dev->accel.dest + dev->accel.cx, dest_dat);
-                                        }
-                                    }
-                                }
-
-                                mix_dat <<= 1;
-                                mix_dat |= 1;
-
-                                if (dev->accel.cmd & 0x20)
-                                    dev->accel.cx++;
-                                else
-                                    dev->accel.cx--;
-
-                                dev->accel.sx--;
-                                if (dev->accel.odd_in) {
-                                    if (dev->accel.sx < 0) {
-                                        dev->accel.sx     = dev->accel.maj_axis_pcnt & 0x7ff;
-                                        dev->accel.odd_in = 0;
-                                        dev->accel.cx     = dev->accel.cur_x;
-                                        if (dev->accel.cmd & 0x80)
-                                            dev->accel.cy++;
-                                        else
-                                            dev->accel.cy--;
-
-                                        if (((dev->local & 0xff) >= 0x02) && dev->accel.ge_offset && ((dev->accel_bpp == 24) || (dev->accel_bpp == 8))) {
-                                            dev->accel.dest = (dev->accel.ge_offset << 2) + (dev->accel.cy * dev->pitch);
-                                            dev->accel.newdest_in = (dev->accel.ge_offset << 2) + ((dev->accel.cy + 1) * dev->pitch);
-                                        } else {
-                                            dev->accel.dest       = dev->accel.cy * dev->pitch;
-                                            dev->accel.newdest_in = (dev->accel.cy + 1) * dev->pitch;
-                                        }
-                                        dev->accel.sy--;
-                                        return;
+                                    if ((dev->accel.cmd & 0x04) && dev->accel.sx) {
+                                        WRITE(dev->accel.dest + dev->accel.cx, dest_dat);
+                                    } else if (!(dev->accel.cmd & 0x04)) {
+                                        WRITE(dev->accel.dest + dev->accel.cx, dest_dat);
                                     }
                                 } else {
-                                    if (dev->accel.sx < 0) {
-                                        dev->accel.sx = dev->accel.maj_axis_pcnt & 0x7ff;
-                                        dev->accel.sx--;
-                                        dev->accel.cx     = dev->accel.cur_x;
-                                        dev->accel.odd_in = 1;
-                                        if (dev->accel.cmd & 0x20)
-                                            dev->accel.cx++;
-                                        else
-                                            dev->accel.cx--;
-                                        if (dev->accel.cmd & 0x80)
-                                            dev->accel.cy++;
-                                        else
-                                            dev->accel.cy--;
-
-                                        if (((dev->local & 0xff) >= 0x02) && dev->accel.ge_offset && ((dev->accel_bpp == 24) || (dev->accel_bpp == 8))) {
-                                            dev->accel.dest = (dev->accel.ge_offset << 2) + (dev->accel.cy * dev->pitch);
-                                            dev->accel.newdest_in = (dev->accel.ge_offset << 2) + ((dev->accel.cy + 1) * dev->pitch);
-                                        } else {
-                                            dev->accel.dest       = dev->accel.cy * dev->pitch;
-                                            dev->accel.newdest_in = (dev->accel.cy + 1) * dev->pitch;
-                                        }
-                                        dev->accel.sy--;
-                                        return;
-                                    }
-                                }
-                            }
-                        } else if (dev->accel.output && !dev->accel.input) {
-                            while (count-- && (dev->accel.sy >= 0)) {
-                                if (dev->accel.cx >= dev->accel.clip_left && dev->accel.cx <= clip_r && dev->accel.cy >= dev->accel.clip_top && dev->accel.cy <= clip_b) {
-                                    src_dat = cpu_dat;
-                                    if (!dev->accel.odd_out && !dev->accel.sx) {
-                                        READ(dev->accel.newdest_out + dev->accel.cur_x, dest_dat);
-                                    } else {
-                                        READ(dev->accel.dest + dev->accel.cx, dest_dat);
-                                    }
-
-                                    if ((compare_mode == 0) || ((compare_mode == 0x10) && (dest_dat >= compare)) || ((compare_mode == 0x18) && (dest_dat < compare)) || ((compare_mode == 0x20) && (dest_dat != compare)) || ((compare_mode == 0x28) && (dest_dat == compare)) || ((compare_mode == 0x30) && (dest_dat <= compare)) || ((compare_mode == 0x38) && (dest_dat > compare))) {
-                                        old_dest_dat = dest_dat;
-                                        MIX(mix_dat & mix_mask, dest_dat, src_dat);
-                                        dest_dat = (dest_dat & wrt_mask) | (old_dest_dat & ~wrt_mask);
-                                        if (!dev->accel.odd_out && !dev->accel.sx) {
-                                            WRITE(dev->accel.newdest_out + dev->accel.cur_x, dest_dat);
-                                        } else {
-                                            WRITE(dev->accel.dest + dev->accel.cx, dest_dat);
-                                        }
-                                    }
-                                }
-
-                                mix_dat <<= 1;
-                                mix_dat |= 1;
-                                if (dev->bpp)
-                                    cpu_dat >>= 16;
-                                else
-                                    cpu_dat >>= 8;
-
-                                if (dev->accel.cmd & 0x20)
-                                    dev->accel.cx++;
-                                else
-                                    dev->accel.cx--;
-
-                                dev->accel.sx--;
-                                if (dev->accel.odd_out) {
-                                    if (dev->accel.sx < 0) {
-                                        dev->accel.sx      = dev->accel.maj_axis_pcnt & 0x7ff;
-                                        dev->accel.odd_out = 0;
-                                        dev->accel.cx      = dev->accel.cur_x;
-                                        if (dev->accel.cmd & 0x80)
-                                            dev->accel.cy++;
-                                        else
-                                            dev->accel.cy--;
-
-                                        if (((dev->local & 0xff) >= 0x02) && dev->accel.ge_offset && ((dev->accel_bpp == 24) || (dev->accel_bpp == 8))) {
-                                            dev->accel.dest = (dev->accel.ge_offset << 2) + (dev->accel.cy * dev->pitch);
-                                            dev->accel.newdest_out = (dev->accel.ge_offset << 2) + ((dev->accel.cy + 1) * dev->pitch);
-                                        } else {
-                                            dev->accel.dest        = dev->accel.cy * dev->pitch;
-                                            dev->accel.newdest_out = (dev->accel.cy + 1) * dev->pitch;
-                                        }
-                                        dev->accel.sy--;
-                                        return;
-                                    }
-                                } else {
-                                    if (dev->accel.sx < 0) {
-                                        dev->accel.sx      = dev->accel.maj_axis_pcnt & 0x7ff;
-                                        dev->accel.odd_out = 1;
-                                        dev->accel.sx--;
-                                        dev->accel.cx = dev->accel.cur_x;
-                                        if (dev->accel.cmd & 0x20)
-                                            dev->accel.cx++;
-                                        else
-                                            dev->accel.cx--;
-                                        if (dev->accel.cmd & 0x80)
-                                            dev->accel.cy++;
-                                        else
-                                            dev->accel.cy--;
-
-                                        if (((dev->local & 0xff) >= 0x02) && dev->accel.ge_offset && ((dev->accel_bpp == 24) || (dev->accel_bpp == 8))) {
-                                            dev->accel.dest = (dev->accel.ge_offset << 2) + (dev->accel.cy * dev->pitch);
-                                            dev->accel.newdest_out = (dev->accel.ge_offset << 2) + ((dev->accel.cy + 1) * dev->pitch);
-                                        } else {
-                                            dev->accel.dest        = dev->accel.cy * dev->pitch;
-                                            dev->accel.newdest_out = (dev->accel.cy + 1) * dev->pitch;
-                                        }
-                                        dev->accel.sy--;
-                                        return;
-                                    }
-                                }
-                            }
-                        } else {
-                            while (count-- && (dev->accel.sy >= 0)) {
-                                if (dev->accel.cx >= dev->accel.clip_left && dev->accel.cx <= clip_r && dev->accel.cy >= dev->accel.clip_top && dev->accel.cy <= clip_b) {
-                                    if (ibm8514_cpu_dest(svga) && (pixcntl == 0)) {
-                                        mix_dat = mix_mask; /* Mix data = forced to foreground register. */
-                                    } else if (ibm8514_cpu_dest(svga) && (pixcntl == 3)) {
-                                        /* Mix data = current video memory value. */
-                                        READ(dev->accel.dest + dev->accel.cx, mix_dat);
-                                        mix_dat = ((mix_dat & rd_mask) == rd_mask);
-                                        mix_dat = mix_dat ? mix_mask : 0;
-                                    }
-
-                                    if (ibm8514_cpu_dest(svga)) {
-                                        READ(dev->accel.dest + dev->accel.cx, src_dat);
+                                    if (ibm8514_cpu_dest(svga) && (cmd == 2)) {
                                         if (pixcntl == 3) {
-                                            src_dat = ((src_dat & rd_mask) == rd_mask);
-                                        }
-                                    } else
-                                        switch ((mix_dat & mix_mask) ? frgd_mix : bkgd_mix) {
-                                            case 0:
-                                                src_dat = bkgd_color;
-                                                break;
-                                            case 1:
-                                                src_dat = frgd_color;
-                                                break;
-                                            case 2:
-                                                src_dat = cpu_dat;
-                                                break;
-                                            case 3:
-                                                src_dat = 0;
-                                                break;
-
-                                            default:
-                                                break;
-                                        }
-
-                                    READ(dev->accel.dest + dev->accel.cx, dest_dat);
-
-                                    if ((compare_mode == 0) || ((compare_mode == 0x10) && (dest_dat >= compare)) || ((compare_mode == 0x18) && (dest_dat < compare)) || ((compare_mode == 0x20) && (dest_dat != compare)) || ((compare_mode == 0x28) && (dest_dat == compare)) || ((compare_mode == 0x30) && (dest_dat <= compare)) || ((compare_mode == 0x38) && (dest_dat > compare))) {
-                                        old_dest_dat = dest_dat;
-                                        if (ibm8514_cpu_dest(svga)) {
-                                            if (pixcntl == 3) {
-                                                MIX(mix_dat & mix_mask, dest_dat, src_dat);
-                                            }
-                                        } else {
                                             MIX(mix_dat & mix_mask, dest_dat, src_dat);
                                         }
-                                        dest_dat = (dest_dat & wrt_mask) | (old_dest_dat & ~wrt_mask);
+                                    } else {
+                                        MIX(mix_dat & mix_mask, dest_dat, src_dat);
+                                    }
+                                    dest_dat = (dest_dat & wrt_mask) | (old_dest_dat & ~wrt_mask);
+                                    if ((dev->accel.cmd & 0x04) && dev->accel.sx) {
+                                        WRITE(dev->accel.dest + dev->accel.cx, dest_dat);
+                                    } else if (!(dev->accel.cmd & 0x04)) {
                                         WRITE(dev->accel.dest + dev->accel.cx, dest_dat);
                                     }
                                 }
-
-                                mix_dat <<= 1;
-                                mix_dat |= 1;
-                                if (dev->bpp)
-                                    cpu_dat >>= 16;
-                                else
-                                    cpu_dat >>= 8;
-
-                                if (dev->accel.cmd & 0x20)
-                                    dev->accel.cx++;
-                                else
-                                    dev->accel.cx--;
-
-                                dev->accel.sx--;
-                                if (dev->accel.sx < 0) {
-                                    dev->accel.sx = dev->accel.maj_axis_pcnt & 0x7ff;
-
-                                    if (dev->accel.cmd & 0x20) {
-                                        dev->accel.cx -= (dev->accel.sx) + 1;
-                                    } else
-                                        dev->accel.cx += (dev->accel.sx) + 1;
-
-                                    if (dev->accel.cmd & 0x80)
-                                        dev->accel.cy++;
-                                    else
-                                        dev->accel.cy--;
-
-                                    if (((dev->local & 0xff) >= 0x02) && dev->accel.ge_offset && ((dev->accel_bpp == 24) || (dev->accel_bpp == 8)))
-                                        dev->accel.dest = (dev->accel.ge_offset << 2) + (dev->accel.cy * dev->pitch);
-                                    else
-                                        dev->accel.dest = dev->accel.cy * dev->pitch;
-
-                                    dev->accel.sy--;
-                                    return;
-                                }
                             }
                         }
+                        if (dev->accel.cmd & 0x20)
+                            dev->accel.cx++;
+                        else
+                            dev->accel.cx--;
+
+skip_nibble_rect_write:
+                        if (dev->accel.cmd & 0x20)
+                            dev->accel.x_count++;
+                        else
+                            dev->accel.x_count--;
+
+                        if (dev->accel.cmd & 0x02)
+                            mix_dat >>= 1;
+                        else {
+                            mix_dat <<= 1;
+                            mix_dat |= 1;
+                        }
+
+                        if (dev->bpp)
+                            cpu_dat >>= 16;
+                        else
+                            cpu_dat >>= 8;
+
+                        dev->accel.sx--;
+                        if (dev->accel.sx < 0) {
+                            dev->accel.sx = dev->accel.maj_axis_pcnt & 0x7ff;
+                            if (dev->accel.input)
+                                dev->accel.odd_in = 1;
+                            if (dev->accel.output || dev->accel.input2)
+                                dev->accel.sx -= 2;
+
+                            if (dev->accel.cmd & 0x20)
+                                dev->accel.cx -= (dev->accel.sx + 1);
+                            else
+                                dev->accel.cx += (dev->accel.sx + 1);
+
+                            if (dev->accel.cmd & 0x80)
+                                dev->accel.cy++;
+                            else
+                                dev->accel.cy--;
+
+                            if ((dev->accel_bpp == 24) || (dev->accel_bpp <= 8))
+                                dev->accel.dest = (dev->accel.ge_offset << 2) + (dev->accel.cy * dev->pitch);
+                            else if (dev->bpp)
+                                dev->accel.dest = (dev->accel.ge_offset << 1) + (dev->accel.cy * dev->pitch);
+                            else
+                                dev->accel.dest = (dev->accel.ge_offset << 2) + (dev->accel.cy * dev->pitch);
+
+                            dev->accel.sy--;
+                            dev->accel.x_count = 0;
+                            return;
+                        }
                     }
                 } else {
-rect_fill:
                     if (pixcntl == 1) {
                         if (dev->accel.cmd & 0x40) {
-                            count               = dev->accel.maj_axis_pcnt + 1;
+                            count = (dev->accel.maj_axis_pcnt & 0x7ff) + 1;
                             dev->accel.temp_cnt = 8;
                             while (count-- && dev->accel.sy >= 0) {
-                                if (dev->accel.temp_cnt == 0) {
+                                if (!dev->accel.temp_cnt) {
                                     mix_dat >>= 8;
                                     dev->accel.temp_cnt = 8;
                                 }
-                                if (dev->accel.cx >= dev->accel.clip_left && dev->accel.cx <= clip_r && dev->accel.cy >= dev->accel.clip_top && dev->accel.cy <= clip_b) {
+                                if ((dev->accel.cx >= clip_l) &&
+                                    (dev->accel.cx <= clip_r) &&
+                                    (dev->accel.cy >= clip_t) &&
+                                    (dev->accel.cy <= clip_b)) {
                                     switch ((mix_dat & mix_mask) ? frgd_mix : bkgd_mix) {
                                         case 0:
                                             src_dat = bkgd_color;
@@ -2711,8 +2078,6 @@ rect_fill:
                                             src_dat = frgd_color;
                                             break;
                                         case 2:
-                                            src_dat = 0;
-                                            break;
                                         case 3:
                                             src_dat = 0;
                                             break;
@@ -2723,11 +2088,16 @@ rect_fill:
 
                                     READ(dev->accel.dest + dev->accel.cx, dest_dat);
 
-                                    if ((compare_mode == 0) || ((compare_mode == 0x10) && (dest_dat >= compare)) || ((compare_mode == 0x18) && (dest_dat < compare)) || ((compare_mode == 0x20) && (dest_dat != compare)) || ((compare_mode == 0x28) && (dest_dat == compare)) || ((compare_mode == 0x30) && (dest_dat <= compare)) || ((compare_mode == 0x38) && (dest_dat > compare))) {
+                                    if ((compare_mode == 0) ||
+                                        ((compare_mode == 0x10) && (dest_dat >= compare)) ||
+                                        ((compare_mode == 0x18) && (dest_dat < compare)) ||
+                                        ((compare_mode == 0x20) && (dest_dat != compare)) ||
+                                        ((compare_mode == 0x28) && (dest_dat == compare)) ||
+                                        ((compare_mode == 0x30) && (dest_dat <= compare)) ||
+                                        ((compare_mode == 0x38) && (dest_dat > compare))) {
                                         old_dest_dat = dest_dat;
                                         MIX(mix_dat & mix_mask, dest_dat, src_dat);
                                         dest_dat = (dest_dat & wrt_mask) | (old_dest_dat & ~wrt_mask);
-
                                         WRITE(dev->accel.dest + dev->accel.cx, dest_dat);
                                     }
                                 }
@@ -2748,36 +2118,38 @@ rect_fill:
                                     dev->accel.sx = dev->accel.maj_axis_pcnt & 0x7ff;
 
                                     if (dev->accel.cmd & 0x20) {
-                                        dev->accel.cx -= (dev->accel.sx) + 1;
+                                        dev->accel.cx -= (dev->accel.sx + 1);
                                     } else
-                                        dev->accel.cx += (dev->accel.sx) + 1;
+                                        dev->accel.cx += (dev->accel.sx + 1);
 
                                     if (dev->accel.cmd & 0x80)
                                         dev->accel.cy++;
                                     else
                                         dev->accel.cy--;
 
-                                    if (((dev->local & 0xff) >= 0x02) && dev->accel.ge_offset && ((dev->accel_bpp == 24) || (dev->accel_bpp == 8)))
+                                    if ((dev->accel_bpp == 24) || (dev->accel_bpp <= 8))
                                         dev->accel.dest = (dev->accel.ge_offset << 2) + (dev->accel.cy * dev->pitch);
+                                    else if (dev->bpp)
+                                        dev->accel.dest = (dev->accel.ge_offset << 1) + (dev->accel.cy * dev->pitch);
                                     else
-                                        dev->accel.dest = dev->accel.cy * dev->pitch;
+                                        dev->accel.dest = (dev->accel.ge_offset << 2) + (dev->accel.cy * dev->pitch);
 
                                     dev->accel.sy--;
-
-                                    dev->accel.cur_x = dev->accel.cx;
-                                    dev->accel.cur_y = dev->accel.cy;
                                     return;
                                 }
                             }
                         } else {
                             dev->accel.temp_cnt = 8;
-                            while (count-- && dev->accel.sy >= 0) {
+                            while (count-- && (dev->accel.sy >= 0)) {
                                 if (!dev->accel.temp_cnt) {
                                     dev->accel.temp_cnt = 8;
-                                    mix_dat             = old_mix_dat;
+                                    mix_dat = old_mix_dat;
                                 }
-                                if (dev->accel.cx >= dev->accel.clip_left && dev->accel.cx <= clip_r && dev->accel.cy >= dev->accel.clip_top && dev->accel.cy <= clip_b) {
-                                    switch ((mix_dat & 1) ? frgd_mix : bkgd_mix) {
+                                if ((dev->accel.cx >= clip_l) &&
+                                    (dev->accel.cx <= clip_r) &&
+                                    (dev->accel.cy >= clip_t) &&
+                                    (dev->accel.cy <= clip_b)) {
+                                    switch ((mix_dat & 0x01) ? frgd_mix : bkgd_mix) {
                                         case 0:
                                             src_dat = bkgd_color;
                                             break;
@@ -2785,8 +2157,6 @@ rect_fill:
                                             src_dat = frgd_color;
                                             break;
                                         case 2:
-                                            src_dat = 0;
-                                            break;
                                         case 3:
                                             src_dat = 0;
                                             break;
@@ -2797,11 +2167,16 @@ rect_fill:
 
                                     READ(dev->accel.dest + dev->accel.cx, dest_dat);
 
-                                    if ((compare_mode == 0) || ((compare_mode == 0x10) && (dest_dat >= compare)) || ((compare_mode == 0x18) && (dest_dat < compare)) || ((compare_mode == 0x20) && (dest_dat != compare)) || ((compare_mode == 0x28) && (dest_dat == compare)) || ((compare_mode == 0x30) && (dest_dat <= compare)) || ((compare_mode == 0x38) && (dest_dat > compare))) {
+                                    if ((compare_mode == 0) ||
+                                        ((compare_mode == 0x10) && (dest_dat >= compare)) ||
+                                        ((compare_mode == 0x18) && (dest_dat < compare)) ||
+                                        ((compare_mode == 0x20) && (dest_dat != compare)) ||
+                                        ((compare_mode == 0x28) && (dest_dat == compare)) ||
+                                        ((compare_mode == 0x30) && (dest_dat <= compare)) ||
+                                        ((compare_mode == 0x38) && (dest_dat > compare))) {
                                         old_dest_dat = dest_dat;
-                                        MIX(mix_dat & 1, dest_dat, src_dat);
+                                        MIX(mix_dat & 0x01, dest_dat, src_dat);
                                         dest_dat = (dest_dat & wrt_mask) | (old_dest_dat & ~wrt_mask);
-
                                         WRITE(dev->accel.dest + dev->accel.cx, dest_dat);
                                     }
                                 }
@@ -2819,167 +2194,193 @@ rect_fill:
                                     dev->accel.sx = dev->accel.maj_axis_pcnt & 0x7ff;
 
                                     if (dev->accel.cmd & 0x20) {
-                                        dev->accel.cx -= (dev->accel.sx) + 1;
+                                        dev->accel.cx -= (dev->accel.sx + 1);
                                     } else
-                                        dev->accel.cx += (dev->accel.sx) + 1;
+                                        dev->accel.cx += (dev->accel.sx + 1);
 
                                     if (dev->accel.cmd & 0x80)
                                         dev->accel.cy++;
                                     else
                                         dev->accel.cy--;
 
-                                    if (((dev->local & 0xff) >= 0x02) && dev->accel.ge_offset && ((dev->accel_bpp == 24) || (dev->accel_bpp == 8)))
+                                    if ((dev->accel_bpp == 24) || (dev->accel_bpp <= 8))
                                         dev->accel.dest = (dev->accel.ge_offset << 2) + (dev->accel.cy * dev->pitch);
+                                    else if (dev->bpp)
+                                        dev->accel.dest = (dev->accel.ge_offset << 1) + (dev->accel.cy * dev->pitch);
                                     else
-                                        dev->accel.dest = dev->accel.cy * dev->pitch;
+                                        dev->accel.dest = (dev->accel.ge_offset << 2) + (dev->accel.cy * dev->pitch);
 
                                     dev->accel.sy--;
 
                                     if (dev->accel.sy < 0) {
-                                        dev->accel.cur_x = dev->accel.cx;
-                                        dev->accel.cur_y = dev->accel.cy;
+                                        if (cmd != 4) {
+                                            dev->accel.cur_x = dev->accel.cx;
+                                            dev->accel.cur_y = dev->accel.cy;
+                                        }
                                         return;
                                     }
                                 }
                             }
                         }
-                    } else {
-                        if ((dev->accel.multifunc[0x0a] & 6) == 4) {
-                            while (count-- && (dev->accel.sy >= 0)) {
-                                if (dev->accel.cx >= dev->accel.clip_left && dev->accel.cx <= clip_r && dev->accel.cy >= dev->accel.clip_top && dev->accel.cy <= clip_b) {
-                                    READ(dev->accel.dest + dev->accel.cx, mix_dat);
-                                    if ((mix_dat & rd_mask_polygon) == rd_mask_polygon)
-                                        dev->accel.fill_state = !dev->accel.fill_state;
+                    } else if ((dev->accel.multifunc[0x0a] & 0x06) == 0x04) { /*Polygon Draw Type A*/
+                        while (count-- && (dev->accel.sy >= 0)) {
+                            if ((dev->accel.cx >= clip_l) &&
+                                (dev->accel.cx <= clip_r) &&
+                                (dev->accel.cy >= clip_t) &&
+                                (dev->accel.cy <= clip_b)) {
+                                READ(dev->accel.dest + dev->accel.cx, mix_dat);
+                                if ((mix_dat & rd_mask_polygon) == rd_mask_polygon)
+                                    dev->accel.fill_state ^= 1;
 
-                                    READ(dev->accel.dest + dev->accel.cx, dest_dat);
-                                    old_dest_dat = dest_dat;
-                                    if (dev->accel.fill_state) {
-                                        if (!(rd_mask_polygon & 1) && (wrt_mask & 1)) {
-                                            MIX(mix_dat ^ rd_mask_polygon, dest_dat, mix_dat);
-                                            ibm8514_log("Filling c(%d,%d) without bit 0 of rdmask=%02x, wrtmask=%02x, mixdat=%02x, dest=%02x, old=%02x.\n", dev->accel.cx, dev->accel.cy, rd_mask_polygon, wrt_mask, mix_dat, dest_dat, old_dest_dat);
-                                            dest_dat &= ~rd_mask_polygon;
-                                        } else if ((rd_mask_polygon & 1) && (wrt_mask & 1)) {
-                                            ibm8514_log("Filling c(%d,%d) with bit 0 of rdmask=%02x, wrtmask=%02x.\n", dev->accel.cx, dev->accel.cy, rd_mask_polygon, wrt_mask);
-                                            dest_dat &= ~(rd_mask_polygon & wrt_mask);
-                                        }
-                                    } else {
-                                        if (!(rd_mask_polygon & 1) && (wrt_mask & 1))
-                                            dest_dat &= ~rd_mask_polygon;
-                                        else if ((rd_mask_polygon & 1) && (wrt_mask & 1))
-                                            dest_dat &= ~(rd_mask_polygon & wrt_mask);
+                                READ(dev->accel.dest + dev->accel.cx, dest_dat);
+                                old_dest_dat = dest_dat;
+                                if (dev->accel.fill_state) {
+                                    if (!(rd_mask_polygon & 0x01) && (wrt_mask & 0x01)) {
+                                        MIX(mix_dat ^ rd_mask_polygon, dest_dat, mix_dat);
+                                        ibm8514_log("Filling c(%d,%d) without bit 0 of rdmask=%02x, wrtmask=%02x, mixdat=%02x, dest=%02x, old=%02x.\n", dev->accel.cx, dev->accel.cy, rd_mask_polygon, wrt_mask, mix_dat, dest_dat, old_dest_dat);
+                                        dest_dat &= ~rd_mask_polygon;
+                                    } else if ((rd_mask_polygon & 0x01) && (wrt_mask & 0x01)) {
+                                        ibm8514_log("Filling c(%d,%d) with bit 0 of rdmask=%02x, wrtmask=%02x.\n", dev->accel.cx, dev->accel.cy, rd_mask_polygon, wrt_mask);
+                                        dest_dat &= ~(rd_mask_polygon & wrt_mask);
                                     }
+                                } else {
+                                    if (!(rd_mask_polygon & 0x01) && (wrt_mask & 0x01))
+                                        dest_dat &= ~rd_mask_polygon;
+                                    else if ((rd_mask_polygon & 0x01) && (wrt_mask & 0x01))
+                                        dest_dat &= ~(rd_mask_polygon & wrt_mask);
+                                }
 
+                                dest_dat = (dest_dat & wrt_mask) | (old_dest_dat & ~wrt_mask);
+
+                                if ((compare_mode == 0) ||
+                                    ((compare_mode == 0x10) && (dest_dat >= compare)) ||
+                                    ((compare_mode == 0x18) && (dest_dat < compare)) ||
+                                    ((compare_mode == 0x20) && (dest_dat != compare)) ||
+                                    ((compare_mode == 0x28) && (dest_dat == compare)) ||
+                                    ((compare_mode == 0x30) && (dest_dat <= compare)) ||
+                                    ((compare_mode == 0x38) && (dest_dat > compare))) {
+                                    ibm8514_log("Results c(%d,%d):rdmask=%02x, wrtmask=%02x, mix=%02x, destdat=%02x, nowrite=%d.\n", dev->accel.cx, dev->accel.cy, rd_mask_polygon, wrt_mask, mix_dat, dest_dat, dev->accel.cx_back);
+                                    WRITE(dev->accel.dest + dev->accel.cx, dest_dat);
+                                }
+                            }
+
+                            if (dev->accel.cmd & 0x20)
+                                dev->accel.cx++;
+                            else
+                                dev->accel.cx--;
+
+                            dev->accel.sx--;
+                            if (dev->accel.sx < 0) {
+                                dev->accel.fill_state = 0;
+                                dev->accel.sx = dev->accel.maj_axis_pcnt & 0x7ff;
+
+                                if (dev->accel.cmd & 0x20)
+                                    dev->accel.cx -= (dev->accel.sx + 1);
+                                else
+                                    dev->accel.cx += (dev->accel.sx + 1);
+
+                                if (dev->accel.cmd & 0x80)
+                                    dev->accel.cy++;
+                                else
+                                    dev->accel.cy--;
+
+                                if ((dev->accel_bpp == 24) || (dev->accel_bpp <= 8))
+                                    dev->accel.dest = (dev->accel.ge_offset << 2) + (dev->accel.cy * dev->pitch);
+                                else if (dev->bpp)
+                                    dev->accel.dest = (dev->accel.ge_offset << 1) + (dev->accel.cy * dev->pitch);
+                                else
+                                    dev->accel.dest = (dev->accel.ge_offset << 2) + (dev->accel.cy * dev->pitch);
+
+                                dev->accel.sy--;
+
+                                if (dev->accel.sy < 0) {
+                                    ibm8514_log(".\n");
+                                    return;
+                                }
+                            }
+                        }
+                    } else {
+                        ibm8514_log("Polygon Draw Type=%02x, CL=%d, CR=%d.\n", dev->accel.multifunc[0x0a] & 0x06, clip_l, clip_r);
+                        while (count-- && dev->accel.sy >= 0) {
+                            if ((dev->accel.cx >= clip_l) &&
+                                (dev->accel.cx <= clip_r) &&
+                                (dev->accel.cy >= clip_t) &&
+                                (dev->accel.cy <= clip_b)) {
+                                switch ((mix_dat & mix_mask) ? frgd_mix : bkgd_mix) {
+                                    case 0:
+                                        src_dat = bkgd_color;
+                                        if (!bkgd_mix && (dev->accel.cmd & 0x40) && ((dev->accel.frgd_mix & 0x1f) == 7) && ((dev->accel.bkgd_mix & 0x1f) == 3) && !dev->bpp && (bkgd_color == 0x00)) /*For some reason, the September 1992 Mach8/32 drivers for Win3.x don't set the background colors properly.*/
+                                            src_dat = frgd_color;
+                                        break;
+                                    case 1:
+                                        src_dat = frgd_color;
+                                        break;
+                                    case 2:
+                                    case 3:
+                                        src_dat = 0;
+                                        break;
+
+                                    default:
+                                        break;
+                                }
+
+                                READ(dev->accel.dest + dev->accel.cx, dest_dat);
+
+                                if ((compare_mode == 0) ||
+                                    ((compare_mode == 0x10) && (dest_dat >= compare)) ||
+                                    ((compare_mode == 0x18) && (dest_dat < compare)) ||
+                                    ((compare_mode == 0x20) && (dest_dat != compare)) ||
+                                    ((compare_mode == 0x28) && (dest_dat == compare)) ||
+                                    ((compare_mode == 0x30) && (dest_dat <= compare)) ||
+                                    ((compare_mode == 0x38) && (dest_dat > compare))) {
+                                    old_dest_dat = dest_dat;
+                                    MIX(mix_dat & mix_mask, dest_dat, src_dat);
                                     dest_dat = (dest_dat & wrt_mask) | (old_dest_dat & ~wrt_mask);
 
-                                    if ((compare_mode == 0) || ((compare_mode == 0x10) && (dest_dat >= compare)) || ((compare_mode == 0x18) && (dest_dat < compare)) || ((compare_mode == 0x20) && (dest_dat != compare)) || ((compare_mode == 0x28) && (dest_dat == compare)) || ((compare_mode == 0x30) && (dest_dat <= compare)) || ((compare_mode == 0x38) && (dest_dat > compare))) {
-                                        ibm8514_log("Results c(%d,%d):rdmask=%02x, wrtmask=%02x, mix=%02x, destdat=%02x, nowrite=%d.\n", dev->accel.cx, dev->accel.cy, rd_mask_polygon, wrt_mask, mix_dat, dest_dat, dev->accel.cx_back);
+                                    if (dev->accel.cmd & 0x10) {
                                         WRITE(dev->accel.dest + dev->accel.cx, dest_dat);
                                     }
                                 }
+                            } else
+                                ibm8514_log("Outside clipping.\n");
+
+                            mix_dat <<= 1;
+                            mix_dat |= 1;
+
+                            if (dev->accel.cmd & 0x20)
+                                dev->accel.cx++;
+                            else
+                                dev->accel.cx--;
+
+                            dev->accel.sx--;
+                            if (dev->accel.sx < 0) {
+                                dev->accel.fill_state = 0;
+                                dev->accel.sx = dev->accel.maj_axis_pcnt & 0x7ff;
 
                                 if (dev->accel.cmd & 0x20)
-                                    dev->accel.cx++;
+                                    dev->accel.cx -= (dev->accel.sx + 1);
                                 else
-                                    dev->accel.cx--;
+                                    dev->accel.cx += (dev->accel.sx + 1);
 
-                                dev->accel.sx--;
-                                if (dev->accel.sx < 0) {
-                                    dev->accel.fill_state = 0;
-                                    dev->accel.sx = dev->accel.maj_axis_pcnt & 0x7ff;
-
-                                    if (dev->accel.cmd & 0x20)
-                                        dev->accel.cx -= (dev->accel.sx) + 1;
-                                    else
-                                        dev->accel.cx += (dev->accel.sx) + 1;
-
-                                    if (dev->accel.cmd & 0x80)
-                                        dev->accel.cy++;
-                                    else
-                                        dev->accel.cy--;
-
-                                    dev->accel.dest = dev->accel.cy * dev->pitch;
-
-                                    dev->accel.sy--;
-
-                                    if (dev->accel.sy < 0) {
-                                        ibm8514_log(".\n");
-                                        return;
-                                    }
-                                }
-                            }
-                        } else {
-                            ibm8514_log("Rectangle Fill Normal CMD=%04x, CURRENT(%d,%d), sx=%d, FR(%02x), linedraw=%d.\n", dev->accel.cmd, dev->accel.cx, dev->accel.cy, dev->accel.sx, frgd_color, dev->accel.linedraw);
-                            while (count-- && dev->accel.sy >= 0) {
-                                if (dev->accel.cx >= dev->accel.clip_left && dev->accel.cx <= clip_r && dev->accel.cy >= dev->accel.clip_top && dev->accel.cy <= clip_b) {
-                                    switch ((mix_dat & mix_mask) ? frgd_mix : bkgd_mix) {
-                                        case 0:
-                                            src_dat = bkgd_color;
-                                            if (!bkgd_mix && (dev->accel.cmd & 0x40) && ((dev->accel.frgd_mix & 0x1f) == 7) && ((dev->accel.bkgd_mix & 0x1f) == 3) && !dev->bpp && (bkgd_color == 0x00)) /*For some reason, the September 1992 Mach8/32 drivers for Win3.x don't set the background colors properly.*/
-                                                src_dat = frgd_color;
-                                            break;
-                                        case 1:
-                                            src_dat = frgd_color;
-                                            break;
-                                        case 2:
-                                            src_dat = 0;
-                                            break;
-                                        case 3:
-                                            src_dat = 0;
-                                            break;
-
-                                        default:
-                                            break;
-                                    }
-
-
-                                    READ(dev->accel.dest + dev->accel.cx, dest_dat);
-
-                                    if ((compare_mode == 0) || ((compare_mode == 0x10) && (dest_dat >= compare)) || ((compare_mode == 0x18) && (dest_dat < compare)) || ((compare_mode == 0x20) && (dest_dat != compare)) || ((compare_mode == 0x28) && (dest_dat == compare)) || ((compare_mode == 0x30) && (dest_dat <= compare)) || ((compare_mode == 0x38) && (dest_dat > compare))) {
-                                        old_dest_dat = dest_dat;
-                                        MIX(mix_dat & mix_mask, dest_dat, src_dat);
-                                        dest_dat = (dest_dat & wrt_mask) | (old_dest_dat & ~wrt_mask);
-
-                                        if (dev->accel.cmd & 0x10) {
-                                            WRITE(dev->accel.dest + dev->accel.cx, dest_dat);
-                                        }
-                                    }
-                                }
-
-                                mix_dat <<= 1;
-                                mix_dat |= 1;
-
-                                if (dev->accel.cmd & 0x20)
-                                    dev->accel.cx++;
+                                if (dev->accel.cmd & 0x80)
+                                    dev->accel.cy++;
                                 else
-                                    dev->accel.cx--;
+                                    dev->accel.cy--;
 
-                                dev->accel.sx--;
-                                if (dev->accel.sx < 0) {
-                                    dev->accel.fill_state = 0;
-                                    dev->accel.sx = dev->accel.maj_axis_pcnt & 0x7ff;
+                                if ((dev->accel_bpp == 24) || (dev->accel_bpp <= 8))
+                                    dev->accel.dest = (dev->accel.ge_offset << 2) + (dev->accel.cy * dev->pitch);
+                                else if (dev->bpp)
+                                    dev->accel.dest = (dev->accel.ge_offset << 1) + (dev->accel.cy * dev->pitch);
+                                else
+                                    dev->accel.dest = (dev->accel.ge_offset << 2) + (dev->accel.cy * dev->pitch);
 
-                                    if (dev->accel.cmd & 0x20)
-                                        dev->accel.cx -= (dev->accel.sx) + 1;
-                                    else
-                                        dev->accel.cx += (dev->accel.sx) + 1;
-
-                                    if (dev->accel.cmd & 0x80)
-                                        dev->accel.cy++;
-                                    else
-                                        dev->accel.cy--;
-
-                                    if (((dev->local & 0xff) >= 0x02) && dev->accel.ge_offset && ((dev->accel_bpp == 24) || (dev->accel_bpp == 8)))
-                                        dev->accel.dest = (dev->accel.ge_offset << 2) + (dev->accel.cy * dev->pitch);
-                                    else
-                                        dev->accel.dest = dev->accel.cy * dev->pitch;
-
-                                    dev->accel.sy--;
-
-                                    if (dev->accel.sy < 0) {
+                                dev->accel.sy--;
+                                if (dev->accel.sy < 0) {
+                                    if (cmd != 4) {
                                         dev->accel.cur_x = dev->accel.cx;
                                         dev->accel.cur_y = dev->accel.cy;
-                                        return;
                                     }
+                                    return;
                                 }
                             }
                         }
@@ -2990,15 +2391,16 @@ rect_fill:
 
         case 5: /*Draw Polygon Boundary Line*/ {
             if (!cpu_input) {
+                dev->accel.sy = dev->accel.maj_axis_pcnt_no_limit;
+
                 dev->accel.cx = dev->accel.cur_x;
                 if (dev->accel.cur_x >= 0x600)
                     dev->accel.cx |= ~0x5ff;
+
                 dev->accel.cy = dev->accel.cur_y;
                 if (dev->accel.cur_y >= 0x600)
                     dev->accel.cy |= ~0x5ff;
 
-                dev->accel.sy = dev->accel.maj_axis_pcnt_no_limit;
-
                 if (dev->accel.cmd & 0x80)
                     dev->accel.oldcy = dev->accel.cy + 1;
                 else
@@ -3017,14 +2419,15 @@ rect_fill:
                 }
             }
 
-            if (dev->accel.cmd & 8) {
+            if (dev->accel.cmd & 0x08) { /*Vectored Boundary Line*/
                 while (count-- && (dev->accel.sy >= 0)) {
-                    if (dev->accel.cx < 0)
-                        dev->accel.cx = 0;
-                    if (dev->accel.cy < 0)
-                        dev->accel.cy = 0;
+                    if (dev->accel.cx < clip_l)
+                        dev->accel.cx = clip_l;
 
-                    if (dev->accel.cx >= dev->accel.clip_left && dev->accel.cx <= clip_r && dev->accel.cy >= dev->accel.clip_top && dev->accel.cy <= clip_b) {
+                    if ((dev->accel.cx >= clip_l) &&
+                        (dev->accel.cx <= clip_r) &&
+                        (dev->accel.cy >= clip_t) &&
+                        (dev->accel.cy <= clip_b)) {
                         switch ((mix_dat & mix_mask) ? frgd_mix : bkgd_mix) {
                             case 0:
                                 src_dat = bkgd_color;
@@ -3046,16 +2449,22 @@ rect_fill:
 
                         READ((dev->accel.cy * dev->pitch) + dev->accel.cx, dest_dat);
 
-                        if ((compare_mode == 0) || ((compare_mode == 0x10) && (dest_dat >= compare)) || ((compare_mode == 0x18) && (dest_dat < compare)) || ((compare_mode == 0x20) && (dest_dat != compare)) || ((compare_mode == 0x28) && (dest_dat == compare)) || ((compare_mode == 0x30) && (dest_dat <= compare)) || ((compare_mode == 0x38) && (dest_dat > compare))) {
+                        if ((compare_mode == 0) ||
+                            ((compare_mode == 0x10) && (dest_dat >= compare)) ||
+                            ((compare_mode == 0x18) && (dest_dat < compare)) ||
+                            ((compare_mode == 0x20) && (dest_dat != compare)) ||
+                            ((compare_mode == 0x28) && (dest_dat == compare)) ||
+                            ((compare_mode == 0x30) && (dest_dat <= compare)) ||
+                            ((compare_mode == 0x38) && (dest_dat > compare))) {
                             old_dest_dat = dest_dat;
                             MIX(mix_dat & mix_mask, dest_dat, src_dat);
                             dest_dat = (dest_dat & wrt_mask) | (old_dest_dat & ~wrt_mask);
                             if (dev->accel.cmd & 0x10) {
-                                if (dev->accel.sy && (dev->accel.cmd & 4)) {
+                                if (dev->accel.sy && (dev->accel.cmd & 0x04)) {
                                     if (dev->accel.oldcy != dev->accel.cy) {
                                         WRITE((dev->accel.cy * dev->pitch) + (dev->accel.cx), dest_dat);
                                     }
-                                } else if (!(dev->accel.cmd & 4)) {
+                                } else if (!(dev->accel.cmd & 0x04)) {
                                     if (dev->accel.oldcy != dev->accel.cy) {
                                         WRITE((dev->accel.cy * dev->pitch) + (dev->accel.cx), dest_dat);
                                     }
@@ -3116,12 +2525,15 @@ rect_fill:
 
                     dev->accel.sy--;
                 }
-            } else {
+            } else { /*Vectored Bresenham*/
                 while (count-- && (dev->accel.sy >= 0)) {
-                    if (dev->accel.cx < dev->accel.clip_left)
-                        dev->accel.cx = dev->accel.clip_left;
+                    if (dev->accel.cx < clip_l)
+                        dev->accel.cx = clip_l;
 
-                    if (dev->accel.cx >= dev->accel.clip_left && dev->accel.cx <= clip_r && dev->accel.cy >= dev->accel.clip_top && dev->accel.cy <= clip_b) {
+                    if ((dev->accel.cx >= clip_l) &&
+                        (dev->accel.cx <= clip_r) &&
+                        (dev->accel.cy >= clip_t) &&
+                        (dev->accel.cy <= clip_b)) {
                         switch ((mix_dat & mix_mask) ? frgd_mix : bkgd_mix) {
                             case 0:
                                 src_dat = bkgd_color;
@@ -3142,7 +2554,13 @@ rect_fill:
 
                         READ((dev->accel.cy * dev->pitch) + dev->accel.cx, dest_dat);
 
-                        if ((compare_mode == 0) || ((compare_mode == 0x10) && (dest_dat >= compare)) || ((compare_mode == 0x18) && (dest_dat < compare)) || ((compare_mode == 0x20) && (dest_dat != compare)) || ((compare_mode == 0x28) && (dest_dat == compare)) || ((compare_mode == 0x30) && (dest_dat <= compare)) || ((compare_mode == 0x38) && (dest_dat > compare))) {
+                        if ((compare_mode == 0) ||
+                            ((compare_mode == 0x10) && (dest_dat >= compare)) ||
+                            ((compare_mode == 0x18) && (dest_dat < compare)) ||
+                            ((compare_mode == 0x20) && (dest_dat != compare)) ||
+                            ((compare_mode == 0x28) && (dest_dat == compare)) ||
+                            ((compare_mode == 0x30) && (dest_dat <= compare)) ||
+                            ((compare_mode == 0x38) && (dest_dat > compare))) {
                             old_dest_dat = dest_dat;
                             MIX(mix_dat & mix_mask, dest_dat, src_dat);
                             dest_dat = (dest_dat & wrt_mask) | (old_dest_dat & ~wrt_mask);
@@ -3209,38 +2627,50 @@ rect_fill:
             if (!cpu_input) /*!cpu_input is trigger to start operation*/
             {
                 dev->accel.x_count = 0;
-                dev->accel.output  = 0;
-
-                dev->accel.sx = dev->accel.maj_axis_pcnt & 0x7ff;
-                dev->accel.sy = dev->accel.multifunc[0] & 0x7ff;
 
                 dev->accel.dx = dev->accel.destx;
-                dev->accel.dy = dev->accel.desty;
-
                 if (dev->accel.destx >= 0x600)
                     dev->accel.dx |= ~0x5ff;
+
+                dev->accel.dy = dev->accel.desty;
                 if (dev->accel.desty >= 0x600)
                     dev->accel.dy |= ~0x5ff;
 
                 dev->accel.cx = dev->accel.cur_x;
-                dev->accel.cy = dev->accel.cur_y;
-
                 if (dev->accel.cur_x >= 0x600)
                     dev->accel.cx |= ~0x5ff;
+
+                dev->accel.cy = dev->accel.cur_y;
                 if (dev->accel.cur_y >= 0x600)
                     dev->accel.cy |= ~0x5ff;
 
-                dev->accel.src  = dev->accel.cy * dev->pitch;
-                dev->accel.dest = dev->accel.dy * dev->pitch;
+                dev->accel.sx = dev->accel.maj_axis_pcnt & 0x7ff;
+                dev->accel.sy = dev->accel.multifunc[0] & 0x7ff;
+
+                if ((dev->accel_bpp == 24) || (dev->accel_bpp <= 8)) {
+                    dev->accel.src  = (dev->accel.ge_offset << 2) + (dev->accel.cy * dev->pitch);
+                    dev->accel.dest = (dev->accel.ge_offset << 2) + (dev->accel.dy * dev->pitch);
+                } else if (dev->bpp) {
+                    dev->accel.src  = (dev->accel.ge_offset << 1) + (dev->accel.cy * dev->pitch);
+                    dev->accel.dest = (dev->accel.ge_offset << 1) + (dev->accel.dy * dev->pitch);
+                } else {
+                    dev->accel.src  = (dev->accel.ge_offset << 2) + (dev->accel.cy * dev->pitch);
+                    dev->accel.dest = (dev->accel.ge_offset << 2) + (dev->accel.dy * dev->pitch);
+                }
                 dev->accel.fill_state = 0;
 
                 if (ibm8514_cpu_src(svga)) {
-                    if (dev->accel.cmd & 2) {
+                    if (dev->accel.cmd & 0x02) {
                         if (!(dev->accel.cmd & 0x1000)) {
-                            dev->accel.sx += (dev->accel.cur_x & 3);
-                            dev->accel.nibbleset = (uint8_t *) calloc(1, (dev->accel.sx >> 3) + 1);
-                            dev->accel.writemono = (uint8_t *) calloc(1, (dev->accel.sx >> 3) + 1);
-                            dev->accel.sys_cnt   = (dev->accel.sx >> 3) + 1;
+                            dev->accel.x_count = dev->accel.cx;
+                            if (and3) {
+                                if (dev->accel.cmd & 0x20)
+                                    dev->accel.x_count -= and3;
+                                else
+                                    dev->accel.x_count += and3;
+
+                                dev->accel.sx += 8;
+                            }
                         }
                     }
                     dev->data_available  = 0;
@@ -3253,495 +2683,389 @@ rect_fill:
                 }
             }
 
-            if (dev->accel.cmd & 2) {
-                if (cpu_input) {
-bitblt_pix:
-                    if (count < 8) {
-                        while (count-- && (dev->accel.sy >= 0)) {
-                            if (dev->accel.dx >= dev->accel.clip_left && dev->accel.dx <= clip_r && dev->accel.dy >= dev->accel.clip_top && dev->accel.dy <= clip_b) {
-                                if (pixcntl == 3) {
-                                    if (!(dev->accel.cmd & 0x10) && ((frgd_mix != 3) || (bkgd_mix != 3))) {
-                                        READ(dev->accel.src + dev->accel.cx, mix_dat);
-                                        mix_dat = ((mix_dat & rd_mask) == rd_mask);
-                                        mix_dat = mix_dat ? mix_mask : 0;
-                                    } else if (dev->accel.cmd & 0x10) {
-                                        READ(dev->accel.src + dev->accel.cx, mix_dat);
-                                        mix_dat = ((mix_dat & rd_mask) == rd_mask);
-                                        mix_dat = mix_dat ? mix_mask : 0;
-                                    }
-                                }
-                                switch ((mix_dat & mix_mask) ? frgd_mix : bkgd_mix) {
-                                    case 0:
-                                        src_dat = bkgd_color;
-                                        break;
-                                    case 1:
-                                        src_dat = frgd_color;
-                                        break;
-                                    case 2:
-                                        src_dat = cpu_dat;
-                                        break;
-                                    case 3:
-                                        READ(dev->accel.src + dev->accel.cx, src_dat);
-                                        if (pixcntl == 3) {
-                                            if (dev->accel.cmd & 0x10) {
-                                                src_dat = ((src_dat & rd_mask) == rd_mask);
-                                            }
-                                        }
-                                        break;
-
-                                    default:
-                                        break;
-                                }
-
-                                READ(dev->accel.dest + dev->accel.dx, dest_dat);
-
-                                if ((compare_mode == 0) || ((compare_mode == 0x10) && (dest_dat >= compare)) || ((compare_mode == 0x18) && (dest_dat < compare)) || ((compare_mode == 0x20) && (dest_dat != compare)) || ((compare_mode == 0x28) && (dest_dat == compare)) || ((compare_mode == 0x30) && (dest_dat <= compare)) || ((compare_mode == 0x38) && (dest_dat > compare))) {
-                                    old_dest_dat = dest_dat;
-
-                                    MIX(mix_dat & mix_mask, dest_dat, src_dat);
-                                    dest_dat = (dest_dat & wrt_mask) | (old_dest_dat & ~wrt_mask);
-                                    WRITE(dev->accel.dest + dev->accel.dx, dest_dat);
-                                }
-                            }
-
-                            mix_dat <<= 1;
-                            mix_dat |= 1;
-                            if (dev->bpp)
-                                cpu_dat >>= 16;
-                            else
-                                cpu_dat >>= 8;
-
-                            if (dev->accel.cmd & 0x20) {
-                                dev->accel.dx++;
-                                dev->accel.cx++;
-                            } else {
-                                dev->accel.dx--;
-                                dev->accel.cx--;
-                            }
-
-                            dev->accel.sx--;
-                            if (dev->accel.sx < 0) {
-                                dev->accel.sx = dev->accel.maj_axis_pcnt & 0x7ff;
-
-                                if (dev->accel.cmd & 2) {
-                                    dev->accel.sx += (dev->accel.cur_x & 3);
-                                }
-
-                                if (dev->accel.cmd & 0x20) {
-                                    dev->accel.dx -= (dev->accel.sx) + 1;
-                                    dev->accel.cx -= (dev->accel.sx) + 1;
-                                } else {
-                                    dev->accel.dx += (dev->accel.sx) + 1;
-                                    dev->accel.cx += (dev->accel.sx) + 1;
-                                }
-
-                                if (dev->accel.cmd & 0x80) {
-                                    dev->accel.dy++;
-                                    dev->accel.cy++;
-                                } else {
-                                    dev->accel.dy--;
-                                    dev->accel.cy--;
-                                }
-
-                                dev->accel.src  = dev->accel.cy * dev->pitch;
-                                dev->accel.dest = dev->accel.dy * dev->pitch;
-                                dev->accel.sy--;
-                                return;
+            ibm8514_log("BitBLT: full=%04x, odd=%d, c(%d,%d), d(%d,%d), xcount=%d, and3=%d, len(%d,%d), CURX=%d, Width=%d, pixcntl=%d, mix_dat=%08x, count=%d, cpu_data=%08x, cpu_input=%d.\n", dev->accel.cmd, dev->accel.input, dev->accel.cx, dev->accel.cy, dev->accel.dx, dev->accel.dy, dev->accel.x_count, and3, dev->accel.sx, dev->accel.sy, dev->accel.cur_x, dev->accel.maj_axis_pcnt, pixcntl, mix_dat, count, cpu_dat, cpu_input);
+            if (cpu_input) {
+                while (count-- && (dev->accel.sy >= 0)) {
+                    if ((dev->accel.dx >= clip_l) &&
+                        (dev->accel.dx <= clip_r) &&
+                        (dev->accel.dy >= clip_t) &&
+                        (dev->accel.dy <= clip_b)) {
+                        if (pixcntl == 3) {
+                            if (!(dev->accel.cmd & 0x10) && ((frgd_mix != 3) || (bkgd_mix != 3))) {
+                                READ(dev->accel.src + dev->accel.cx, mix_dat);
+                                mix_dat = ((mix_dat & rd_mask) == rd_mask);
+                                if (dev->accel.cmd & 0x02)
+                                    mix_dat = mix_dat ? 0x01 : 0x00;
+                                else
+                                    mix_dat = mix_dat ? mix_mask : 0x00;
+                            } else if (dev->accel.cmd & 0x10) {
+                                READ(dev->accel.src + dev->accel.cx, mix_dat);
+                                mix_dat = ((mix_dat & rd_mask) == rd_mask);
+                                if (dev->accel.cmd & 0x02)
+                                    mix_dat = mix_dat ? 0x01 : 0x00;
+                                else
+                                    mix_dat = mix_dat ? mix_mask : 0x00;
                             }
                         }
-                    } else {
-                        while (count-- && (dev->accel.sy >= 0)) {
-                            if (dev->accel.dx >= dev->accel.clip_left && dev->accel.dx <= clip_r && dev->accel.dy >= dev->accel.clip_top && dev->accel.dy <= clip_b) {
-                                if (pixcntl == 3) {
-                                    if (!(dev->accel.cmd & 0x10) && ((frgd_mix != 3) || (bkgd_mix != 3))) {
-                                        READ(dev->accel.src + dev->accel.cx, mix_dat);
-                                        mix_dat = ((mix_dat & rd_mask) == rd_mask);
-                                        mix_dat = mix_dat ? 1 : 0;
-                                    } else if (dev->accel.cmd & 0x10) {
-                                        READ(dev->accel.src + dev->accel.cx, mix_dat);
-                                        mix_dat = ((mix_dat & rd_mask) == rd_mask);
-                                        mix_dat = mix_dat ? 1 : 0;
+                        if (dev->accel.cmd & 0x02) {
+                            switch ((mix_dat & 0x01) ? frgd_mix : bkgd_mix) {
+                                case 0:
+                                    src_dat = bkgd_color;
+                                    break;
+                                case 1:
+                                    src_dat = frgd_color;
+                                    break;
+                                case 2:
+                                    src_dat = cpu_dat;
+                                    break;
+                                case 3:
+                                    READ(dev->accel.src + dev->accel.cx, src_dat);
+                                    if (pixcntl == 3) {
+                                        if (dev->accel.cmd & 0x10)
+                                            src_dat = ((src_dat & rd_mask) == rd_mask);
                                     }
-                                }
-                                switch ((mix_dat & 1) ? frgd_mix : bkgd_mix) {
-                                    case 0:
-                                        src_dat = bkgd_color;
-                                        break;
-                                    case 1:
-                                        src_dat = frgd_color;
-                                        break;
-                                    case 2:
-                                        src_dat = cpu_dat;
-                                        break;
-                                    case 3:
-                                        READ(dev->accel.src + dev->accel.cx, src_dat);
-                                        if (pixcntl == 3) {
-                                            if (dev->accel.cmd & 0x10) {
-                                                src_dat = ((src_dat & rd_mask) == rd_mask);
-                                            }
-                                        }
-                                        break;
+                                    break;
 
-                                    default:
-                                        break;
-                                }
-
-                                READ(dev->accel.dest + dev->accel.dx, dest_dat);
-
-                                if ((compare_mode == 0) || ((compare_mode == 0x10) && (dest_dat >= compare)) || ((compare_mode == 0x18) && (dest_dat < compare)) || ((compare_mode == 0x20) && (dest_dat != compare)) || ((compare_mode == 0x28) && (dest_dat == compare)) || ((compare_mode == 0x30) && (dest_dat <= compare)) || ((compare_mode == 0x38) && (dest_dat > compare))) {
-                                    old_dest_dat = dest_dat;
-                                    MIX(mix_dat & 1, dest_dat, src_dat);
-                                    dest_dat = (dest_dat & wrt_mask) | (old_dest_dat & ~wrt_mask);
-                                    WRITE(dev->accel.dest + dev->accel.dx, dest_dat);
-                                }
-                            }
-
-                            mix_dat >>= 1;
-                            if (dev->bpp)
-                                cpu_dat >>= 16;
-                            else
-                                cpu_dat >>= 8;
-
-                            if (dev->accel.cmd & 0x20) {
-                                dev->accel.dx++;
-                                dev->accel.cx++;
-                            } else {
-                                dev->accel.dx--;
-                                dev->accel.cx--;
-                            }
-
-                            dev->accel.sx--;
-                            if (dev->accel.sx < 0) {
-                                dev->accel.sx = dev->accel.maj_axis_pcnt & 0x7ff;
-
-                                if (dev->accel.cmd & 2) {
-                                    if (!(dev->accel.cmd & 0x1000))
-                                        dev->accel.sx += (dev->accel.cur_x & 3);
-                                }
-
-                                if (dev->accel.cmd & 0x20) {
-                                    dev->accel.dx -= (dev->accel.sx) + 1;
-                                    dev->accel.cx -= (dev->accel.sx) + 1;
-                                } else {
-                                    dev->accel.dx += (dev->accel.sx) + 1;
-                                    dev->accel.cx += (dev->accel.sx) + 1;
-                                }
-
-                                if (dev->accel.cmd & 2) {
-                                    if (dev->accel.cmd & 0x1000) {
-                                        dev->accel.cx = dev->accel.cur_x;
-                                        if (dev->accel.cur_x >= 0x600)
-                                            dev->accel.cx |= ~0x5ff;
-                                        dev->accel.dx = dev->accel.destx;
-                                        if (dev->accel.destx >= 0x600)
-                                            dev->accel.dx |= ~0x5ff;
-                                    }
-                                }
-
-                                if (dev->accel.cmd & 0x80) {
-                                    dev->accel.dy++;
-                                    dev->accel.cy++;
-                                } else {
-                                    dev->accel.dy--;
-                                    dev->accel.cy--;
-                                }
-
-                                dev->accel.dest = dev->accel.dy * dev->pitch;
-                                dev->accel.src  = dev->accel.cy * dev->pitch;
-                                dev->accel.sy--;
-                                return;
-                            }
-                        }
-                    }
-                } else {
-                    goto bitblt;
-                }
-            } else {
-                if (cpu_input) {
-                    if (pixcntl == 2) {
-                        goto bitblt_pix;
-                    } else {
-                        while (count-- && (dev->accel.sy >= 0)) {
-                            if (dev->accel.dx >= dev->accel.clip_left && dev->accel.dx <= clip_r && dev->accel.dy >= dev->accel.clip_top && dev->accel.dy <= clip_b) {
-                                if (pixcntl == 3) {
-                                    if (!(dev->accel.cmd & 0x10) && ((frgd_mix != 3) || (bkgd_mix != 3))) {
-                                        READ(dev->accel.src + dev->accel.cx, mix_dat);
-                                        mix_dat = ((mix_dat & rd_mask) == rd_mask);
-                                        mix_dat = mix_dat ? mix_mask : 0;
-                                    } else if (dev->accel.cmd & 0x10) {
-                                        READ(dev->accel.src + dev->accel.cx, mix_dat);
-                                        mix_dat = ((mix_dat & rd_mask) == rd_mask);
-                                        mix_dat = mix_dat ? mix_mask : 0;
-                                    }
-                                }
-                                switch ((mix_dat & mix_mask) ? frgd_mix : bkgd_mix) {
-                                    case 0:
-                                        src_dat = bkgd_color;
-                                        break;
-                                    case 1:
-                                        src_dat = frgd_color;
-                                        break;
-                                    case 2:
-                                        src_dat = cpu_dat;
-                                        break;
-                                    case 3:
-                                        READ(dev->accel.src + dev->accel.cx, src_dat);
-                                        if (pixcntl == 3) {
-                                            if (dev->accel.cmd & 0x10) {
-                                                src_dat = ((src_dat & rd_mask) == rd_mask);
-                                            }
-                                        }
-                                        break;
-
-                                    default:
-                                        break;
-                                }
-
-                                READ(dev->accel.dest + dev->accel.dx, dest_dat);
-
-                                if ((compare_mode == 0) || ((compare_mode == 0x10) && (dest_dat >= compare)) || ((compare_mode == 0x18) && (dest_dat < compare)) || ((compare_mode == 0x20) && (dest_dat != compare)) || ((compare_mode == 0x28) && (dest_dat == compare)) || ((compare_mode == 0x30) && (dest_dat <= compare)) || ((compare_mode == 0x38) && (dest_dat > compare))) {
-                                    old_dest_dat = dest_dat;
-                                    MIX(mix_dat & mix_mask, dest_dat, src_dat);
-                                    dest_dat = (dest_dat & wrt_mask) | (old_dest_dat & ~wrt_mask);
-                                    WRITE(dev->accel.dest + dev->accel.dx, dest_dat);
-                                }
-                            }
-
-                            mix_dat <<= 1;
-                            mix_dat |= 1;
-                            if (dev->bpp)
-                                cpu_dat >>= 16;
-                            else
-                                cpu_dat >>= 8;
-
-                            if (dev->accel.cmd & 0x20) {
-                                dev->accel.dx++;
-                                dev->accel.cx++;
-                            } else {
-                                dev->accel.dx--;
-                                dev->accel.cx--;
-                            }
-
-                            dev->accel.sx--;
-                            if (dev->accel.sx < 0) {
-                                dev->accel.sx = dev->accel.maj_axis_pcnt & 0x7ff;
-
-                                if (dev->accel.cmd & 0x20) {
-                                    dev->accel.dx -= (dev->accel.sx) + 1;
-                                    dev->accel.cx -= (dev->accel.sx) + 1;
-                                } else {
-                                    dev->accel.dx += (dev->accel.sx) + 1;
-                                    dev->accel.cx += (dev->accel.sx) + 1;
-                                }
-
-                                if (dev->accel.cmd & 0x80) {
-                                    dev->accel.dy++;
-                                    dev->accel.cy++;
-                                } else {
-                                    dev->accel.dy--;
-                                    dev->accel.cy--;
-                                }
-
-                                dev->accel.dest = dev->accel.dy * dev->pitch;
-                                dev->accel.src  = dev->accel.cy * dev->pitch;
-                                dev->accel.sy--;
-                                return;
-                            }
-                        }
-                    }
-                } else {
-bitblt:
-                    if (pixcntl == 1) {
-                        if (dev->accel.cmd & 0x40) {
-                            count               = dev->accel.maj_axis_pcnt + 1;
-                            dev->accel.temp_cnt = 8;
-                            while (count-- && dev->accel.sy >= 0) {
-                                if (!dev->accel.temp_cnt) {
-                                    mix_dat >>= 8;
-                                    dev->accel.temp_cnt = 8;
-                                }
-                                if (dev->accel.dx >= dev->accel.clip_left && dev->accel.dx <= clip_r && dev->accel.dy >= dev->accel.clip_top && dev->accel.dy <= clip_b) {
-                                    switch ((mix_dat & mix_mask) ? frgd_mix : bkgd_mix) {
-                                        case 0:
-                                            src_dat = bkgd_color;
-                                            break;
-                                        case 1:
-                                            src_dat = frgd_color;
-                                            break;
-                                        case 2:
-                                            src_dat = 0;
-                                            break;
-                                        case 3:
-                                            READ(dev->accel.src + dev->accel.cx, src_dat);
-                                            break;
-
-                                        default:
-                                            break;
-                                    }
-
-                                    READ(dev->accel.dest + dev->accel.dx, dest_dat);
-
-                                    if ((compare_mode == 0) || ((compare_mode == 0x10) && (dest_dat >= compare)) || ((compare_mode == 0x18) && (dest_dat < compare)) || ((compare_mode == 0x20) && (dest_dat != compare)) || ((compare_mode == 0x28) && (dest_dat == compare)) || ((compare_mode == 0x30) && (dest_dat <= compare)) || ((compare_mode == 0x38) && (dest_dat > compare))) {
-                                        old_dest_dat = dest_dat;
-                                        MIX(mix_dat & mix_mask, dest_dat, src_dat);
-                                        dest_dat = (dest_dat & wrt_mask) | (old_dest_dat & ~wrt_mask);
-
-                                        WRITE(dev->accel.dest + dev->accel.dx, dest_dat);
-                                    }
-                                }
-
-                                if (dev->accel.temp_cnt > 0) {
-                                    dev->accel.temp_cnt--;
-                                    mix_dat <<= 1;
-                                    mix_dat |= 1;
-                                }
-
-                                if (dev->accel.cmd & 0x20) {
-                                    dev->accel.dx++;
-                                    dev->accel.cx++;
-                                } else {
-                                    dev->accel.dx--;
-                                    dev->accel.cx--;
-                                }
-
-                                dev->accel.sx--;
-                                if (dev->accel.sx < 0) {
-                                    dev->accel.sx = dev->accel.maj_axis_pcnt & 0x7ff;
-
-                                    if (dev->accel.cmd & 0x20) {
-                                        dev->accel.dx -= (dev->accel.sx) + 1;
-                                        dev->accel.cx -= (dev->accel.sx) + 1;
-                                    } else {
-                                        dev->accel.dx += (dev->accel.sx) + 1;
-                                        dev->accel.cx += (dev->accel.sx) + 1;
-                                    }
-
-                                    if (dev->accel.cmd & 0x80) {
-                                        dev->accel.dy++;
-                                        dev->accel.cy++;
-                                    } else {
-                                        dev->accel.dy--;
-                                        dev->accel.cy--;
-                                    }
-
-                                    dev->accel.dest = dev->accel.dy * dev->pitch;
-                                    dev->accel.src  = dev->accel.cy * dev->pitch;
-                                    dev->accel.sy--;
-                                    return;
-                                }
+                                default:
+                                    break;
                             }
                         } else {
-                            dev->accel.temp_cnt = 8;
-                            while (count-- && dev->accel.sy >= 0) {
-                                if (!dev->accel.temp_cnt) {
-                                    dev->accel.temp_cnt = 8;
-                                    mix_dat             = old_mix_dat;
-                                }
-                                if (dev->accel.dx >= dev->accel.clip_left && dev->accel.dx <= clip_r && dev->accel.dy >= dev->accel.clip_top && dev->accel.dy <= clip_b) {
-                                    switch ((mix_dat & 1) ? frgd_mix : bkgd_mix) {
-                                        case 0:
-                                            src_dat = bkgd_color;
-                                            break;
-                                        case 1:
-                                            src_dat = frgd_color;
-                                            break;
-                                        case 2:
-                                            src_dat = 0;
-                                            break;
-                                        case 3:
-                                            READ(dev->accel.src + dev->accel.cx, src_dat);
-                                            break;
-
-                                        default:
-                                            break;
+                            switch ((mix_dat & mix_mask) ? frgd_mix : bkgd_mix) {
+                                case 0:
+                                    src_dat = bkgd_color;
+                                    break;
+                                case 1:
+                                    src_dat = frgd_color;
+                                    break;
+                                case 2:
+                                    src_dat = cpu_dat;
+                                    break;
+                                case 3:
+                                    READ(dev->accel.src + dev->accel.cx, src_dat);
+                                    if (pixcntl == 3) {
+                                        if (dev->accel.cmd & 0x10)
+                                            src_dat = ((src_dat & rd_mask) == rd_mask);
                                     }
+                                    break;
 
-                                    READ(dev->accel.dest + dev->accel.dx, dest_dat);
+                                default:
+                                    break;
+                            }
+                        }
 
-                                    if ((compare_mode == 0) || ((compare_mode == 0x10) && (dest_dat >= compare)) || ((compare_mode == 0x18) && (dest_dat < compare)) || ((compare_mode == 0x20) && (dest_dat != compare)) || ((compare_mode == 0x28) && (dest_dat == compare)) || ((compare_mode == 0x30) && (dest_dat <= compare)) || ((compare_mode == 0x38) && (dest_dat > compare))) {
-                                        old_dest_dat = dest_dat;
-                                        MIX(mix_dat & 1, dest_dat, src_dat);
-                                        dest_dat = (dest_dat & wrt_mask) | (old_dest_dat & ~wrt_mask);
+                        READ(dev->accel.dest + dev->accel.dx, dest_dat);
 
-                                        WRITE(dev->accel.dest + dev->accel.dx, dest_dat);
-                                    }
+                        if ((compare_mode == 0) ||
+                            ((compare_mode == 0x10) && (dest_dat >= compare)) ||
+                            ((compare_mode == 0x18) && (dest_dat < compare)) ||
+                            ((compare_mode == 0x20) && (dest_dat != compare)) ||
+                            ((compare_mode == 0x28) && (dest_dat == compare)) ||
+                            ((compare_mode == 0x30) && (dest_dat <= compare)) ||
+                            ((compare_mode == 0x38) && (dest_dat > compare))) {
+                            old_dest_dat = dest_dat;
+                            if (dev->accel.cmd & 0x02) {
+                                MIX(mix_dat & 0x01, dest_dat, src_dat);
+                                if ((dev->accel.x_count != dev->accel.cx) && !(dev->accel.cmd & 0x1000) && and3)
+                                    goto skip_nibble_bitblt_write;
+
+                                dest_dat = (dest_dat & wrt_mask) | (old_dest_dat & ~wrt_mask);
+                                WRITE(dev->accel.dest + dev->accel.dx, dest_dat);
+                            } else {
+                                MIX(mix_dat & mix_mask, dest_dat, src_dat);
+                                dest_dat = (dest_dat & wrt_mask) | (old_dest_dat & ~wrt_mask);
+                                WRITE(dev->accel.dest + dev->accel.dx, dest_dat);
+                            }
+                        }
+                    }
+
+                    if (dev->accel.cmd & 0x20) {
+                        dev->accel.dx++;
+                        dev->accel.cx++;
+                    } else {
+                        dev->accel.dx--;
+                        dev->accel.cx--;
+                    }
+
+skip_nibble_bitblt_write:
+                    if (dev->accel.cmd & 0x20)
+                        dev->accel.x_count++;
+                    else
+                        dev->accel.x_count--;
+
+                    if (dev->accel.cmd & 0x02)
+                        mix_dat >>= 1;
+                    else {
+                        mix_dat <<= 1;
+                        mix_dat |= 1;
+                    }
+
+                    if (dev->bpp)
+                        cpu_dat >>= 16;
+                    else
+                        cpu_dat >>= 8;
+
+                    dev->accel.sx--;
+                    if (dev->accel.sx < 0) {
+                        dev->accel.sx = dev->accel.maj_axis_pcnt & 0x7ff;
+
+                        if (dev->accel.cmd & 0x20) {
+                            dev->accel.dx -= (dev->accel.sx + 1);
+                            dev->accel.cx -= (dev->accel.sx + 1);
+                        } else {
+                            dev->accel.dx += (dev->accel.sx + 1);
+                            dev->accel.cx += (dev->accel.sx + 1);
+                        }
+
+                        if (dev->accel.cmd & 0x80) {
+                            dev->accel.dy++;
+                            dev->accel.cy++;
+                        } else {
+                            dev->accel.dy--;
+                            dev->accel.cy--;
+                        }
+
+                        if ((dev->accel_bpp == 24) || (dev->accel_bpp <= 8)) {
+                            dev->accel.src  = (dev->accel.ge_offset << 2) + (dev->accel.cy * dev->pitch);
+                            dev->accel.dest = (dev->accel.ge_offset << 2) + (dev->accel.dy * dev->pitch);
+                        } else if (dev->bpp) {
+                            dev->accel.src  = (dev->accel.ge_offset << 1) + (dev->accel.cy * dev->pitch);
+                            dev->accel.dest = (dev->accel.ge_offset << 1) + (dev->accel.dy * dev->pitch);
+                        } else {
+                            dev->accel.src  = (dev->accel.ge_offset << 2) + (dev->accel.cy * dev->pitch);
+                            dev->accel.dest = (dev->accel.ge_offset << 2) + (dev->accel.dy * dev->pitch);
+                        }
+
+                        dev->accel.sy--;
+                        dev->accel.x_count = 0;
+                        return;
+                    }
+                }
+            } else {
+                if (pixcntl == 1) {
+                    if (dev->accel.cmd & 0x40) {
+                        count = (dev->accel.maj_axis_pcnt & 0x7ff) + 1;
+                        dev->accel.temp_cnt = 8;
+                        while (count-- && (dev->accel.sy >= 0)) {
+                            if (!dev->accel.temp_cnt) {
+                                mix_dat >>= 8;
+                                dev->accel.temp_cnt = 8;
+                            }
+                            if ((dev->accel.dx >= clip_l) &&
+                                (dev->accel.dx <= clip_r) &&
+                                (dev->accel.dy >= clip_t) &&
+                                (dev->accel.dy <= clip_b)) {
+                                switch ((mix_dat & mix_mask) ? frgd_mix : bkgd_mix) {
+                                    case 0:
+                                        src_dat = bkgd_color;
+                                        break;
+                                    case 1:
+                                        src_dat = frgd_color;
+                                        break;
+                                    case 2:
+                                        src_dat = 0;
+                                        break;
+                                    case 3:
+                                        READ(dev->accel.src + dev->accel.cx, src_dat);
+                                        break;
+
+                                    default:
+                                        break;
                                 }
 
+                                READ(dev->accel.dest + dev->accel.dx, dest_dat);
+
+                                if ((compare_mode == 0) ||
+                                    ((compare_mode == 0x10) && (dest_dat >= compare)) ||
+                                    ((compare_mode == 0x18) && (dest_dat < compare)) ||
+                                    ((compare_mode == 0x20) && (dest_dat != compare)) ||
+                                    ((compare_mode == 0x28) && (dest_dat == compare)) ||
+                                    ((compare_mode == 0x30) && (dest_dat <= compare)) ||
+                                    ((compare_mode == 0x38) && (dest_dat > compare))) {
+                                    old_dest_dat = dest_dat;
+                                    MIX(mix_dat & mix_mask, dest_dat, src_dat);
+                                    dest_dat = (dest_dat & wrt_mask) | (old_dest_dat & ~wrt_mask);
+                                    WRITE(dev->accel.dest + dev->accel.dx, dest_dat);
+                                }
+                            }
+
+                            if (dev->accel.temp_cnt > 0) {
                                 dev->accel.temp_cnt--;
-                                mix_dat >>= 1;
+                                mix_dat <<= 1;
+                                mix_dat |= 1;
+                            }
+
+                            if (dev->accel.cmd & 0x20) {
+                                dev->accel.dx++;
+                                dev->accel.cx++;
+                            } else {
+                                dev->accel.dx--;
+                                dev->accel.cx--;
+                            }
+
+                            dev->accel.sx--;
+                            if (dev->accel.sx < 0) {
+                                dev->accel.sx = dev->accel.maj_axis_pcnt & 0x7ff;
 
                                 if (dev->accel.cmd & 0x20) {
-                                    dev->accel.dx++;
-                                    dev->accel.cx++;
+                                    dev->accel.dx -= (dev->accel.sx + 1);
+                                    dev->accel.cx -= (dev->accel.sx + 1);
                                 } else {
-                                    dev->accel.dx--;
-                                    dev->accel.cx--;
+                                    dev->accel.dx += (dev->accel.sx + 1);
+                                    dev->accel.cx += (dev->accel.sx + 1);
                                 }
 
-                                dev->accel.sx--;
-                                if (dev->accel.sx < 0) {
-                                    dev->accel.sx = dev->accel.maj_axis_pcnt & 0x7ff;
-
-                                    if (dev->accel.cmd & 0x20) {
-                                        dev->accel.dx -= (dev->accel.sx) + 1;
-                                        dev->accel.cx -= (dev->accel.sx) + 1;
-                                    } else {
-                                        dev->accel.dx += (dev->accel.sx) + 1;
-                                        dev->accel.cx += (dev->accel.sx) + 1;
-                                    }
-
-                                    if (dev->accel.cmd & 0x80) {
-                                        dev->accel.dy++;
-                                        dev->accel.cy++;
-                                    } else {
-                                        dev->accel.dy--;
-                                        dev->accel.cy--;
-                                    }
-
-                                    dev->accel.dest = dev->accel.dy * dev->pitch;
-                                    dev->accel.src  = dev->accel.cy * dev->pitch;
-                                    dev->accel.sy--;
-
-                                    if (dev->accel.sy < 0) {
-                                        return;
-                                    }
+                                if (dev->accel.cmd & 0x80) {
+                                    dev->accel.dy++;
+                                    dev->accel.cy++;
+                                } else {
+                                    dev->accel.dy--;
+                                    dev->accel.cy--;
                                 }
+
+                                if ((dev->accel_bpp == 24) || (dev->accel_bpp <= 8)) {
+                                    dev->accel.src  = (dev->accel.ge_offset << 2) + (dev->accel.cy * dev->pitch);
+                                    dev->accel.dest = (dev->accel.ge_offset << 2) + (dev->accel.dy * dev->pitch);
+                                } else if (dev->bpp) {
+                                    dev->accel.src  = (dev->accel.ge_offset << 1) + (dev->accel.cy * dev->pitch);
+                                    dev->accel.dest = (dev->accel.ge_offset << 1) + (dev->accel.dy * dev->pitch);
+                                } else {
+                                    dev->accel.src  = (dev->accel.ge_offset << 2) + (dev->accel.cy * dev->pitch);
+                                    dev->accel.dest = (dev->accel.ge_offset << 2) + (dev->accel.dy * dev->pitch);
+                                }
+
+                                dev->accel.sy--;
+                                return;
                             }
                         }
                     } else {
-                        if ((dev->accel_bpp == 24) && ((dev->local & 0xff) >= 0x02) && (dev->accel.cmd == 0xc2b5)) {
-                            int64_t cx;
-                            int64_t dx;
+                        dev->accel.temp_cnt = 8;
+                        while (count-- && (dev->accel.sy >= 0)) {
+                            if (!dev->accel.temp_cnt) {
+                                dev->accel.temp_cnt = 8;
+                                mix_dat = old_mix_dat;
+                            }
+                            if ((dev->accel.dx >= clip_l) &&
+                                (dev->accel.dx <= clip_r) &&
+                                (dev->accel.dy >= clip_t) &&
+                                (dev->accel.dy <= clip_b)) {
+                                switch ((mix_dat & 0x01) ? frgd_mix : bkgd_mix) {
+                                    case 0:
+                                        src_dat = bkgd_color;
+                                        break;
+                                    case 1:
+                                        src_dat = frgd_color;
+                                        break;
+                                    case 2:
+                                        src_dat = 0;
+                                        break;
+                                    case 3:
+                                        READ(dev->accel.src + dev->accel.cx, src_dat);
+                                        break;
 
-                            cx = (int64_t) dev->accel.cx;
-                            dx = (int64_t) dev->accel.dx;
-
-                            while (1) {
-                                if ((dx >= (((int64_t)dev->accel.clip_left) * 3)) && (dx <= (((uint64_t)clip_r) * 3)) &&
-                                    (dev->accel.dy >= (dev->accel.clip_top << 1)) && (dev->accel.dy <= (clip_b << 1))) {
-
-                                    READ(dev->accel.src + (dev->accel.ge_offset << 2) + cx, src_dat);
-                                    READ(dev->accel.dest + (dev->accel.ge_offset << 2) + dx, dest_dat);
-
-                                    dest_dat = (src_dat & wrt_mask) | (dest_dat & ~wrt_mask);
-
-                                    WRITE(dev->accel.dest + (dev->accel.ge_offset << 2) + dx, dest_dat);
+                                    default:
+                                        break;
                                 }
 
-                                cx++;
-                                dx++;
+                                READ(dev->accel.dest + dev->accel.dx, dest_dat);
 
-                                dev->accel.sx--;
-                                if (dev->accel.sx < 0)
-                                    return;
+                                if ((compare_mode == 0) ||
+                                    ((compare_mode == 0x10) && (dest_dat >= compare)) ||
+                                    ((compare_mode == 0x18) && (dest_dat < compare)) ||
+                                    ((compare_mode == 0x20) && (dest_dat != compare)) ||
+                                    ((compare_mode == 0x28) && (dest_dat == compare)) ||
+                                    ((compare_mode == 0x30) && (dest_dat <= compare)) ||
+                                    ((compare_mode == 0x38) && (dest_dat > compare))) {
+                                    old_dest_dat = dest_dat;
+                                    MIX(mix_dat & 0x01, dest_dat, src_dat);
+                                    dest_dat = (dest_dat & wrt_mask) | (old_dest_dat & ~wrt_mask);
+                                    WRITE(dev->accel.dest + dev->accel.dx, dest_dat);
+                                }
                             }
-                            return;
-                        }
 
-                        ibm8514_log("BitBLT 8514/A=%04x, selfrmix=%d, selbkmix=%d, d(%d,%d), c(%d,%d), pixcntl=%d, sy=%d, frgdmix=%02x, bkgdmix=%02x, rdmask=%02x, wrtmask=%02x, linedraw=%d.\n", dev->accel.cmd, frgd_mix, bkgd_mix, dev->accel.dx, dev->accel.dy, dev->accel.cx, dev->accel.cy, pixcntl, dev->accel.sy, dev->accel.frgd_mix & 0x1f, dev->accel.bkgd_mix & 0x1f, dev->accel.rd_mask, wrt_mask, dev->accel.linedraw);
+                            dev->accel.temp_cnt--;
+                            mix_dat >>= 1;
+
+                            if (dev->accel.cmd & 0x20) {
+                                dev->accel.dx++;
+                                dev->accel.cx++;
+                            } else {
+                                dev->accel.dx--;
+                                dev->accel.cx--;
+                            }
+
+                            dev->accel.sx--;
+                            if (dev->accel.sx < 0) {
+                                dev->accel.sx = dev->accel.maj_axis_pcnt & 0x7ff;
+
+                                if (dev->accel.cmd & 0x20) {
+                                    dev->accel.dx -= (dev->accel.sx + 1);
+                                    dev->accel.cx -= (dev->accel.sx + 1);
+                                } else {
+                                    dev->accel.dx += (dev->accel.sx + 1);
+                                    dev->accel.cx += (dev->accel.sx + 1);
+                                }
+
+                                if (dev->accel.cmd & 0x80) {
+                                    dev->accel.dy++;
+                                    dev->accel.cy++;
+                                } else {
+                                    dev->accel.dy--;
+                                    dev->accel.cy--;
+                                }
+
+                                if ((dev->accel_bpp == 24) || (dev->accel_bpp <= 8)) {
+                                    dev->accel.src  = (dev->accel.ge_offset << 2) + (dev->accel.cy * dev->pitch);
+                                    dev->accel.dest = (dev->accel.ge_offset << 2) + (dev->accel.dy * dev->pitch);
+                                } else if (dev->bpp) {
+                                    dev->accel.src  = (dev->accel.ge_offset << 1) + (dev->accel.cy * dev->pitch);
+                                    dev->accel.dest = (dev->accel.ge_offset << 1) + (dev->accel.dy * dev->pitch);
+                                } else {
+                                    dev->accel.src  = (dev->accel.ge_offset << 2) + (dev->accel.cy * dev->pitch);
+                                    dev->accel.dest = (dev->accel.ge_offset << 2) + (dev->accel.dy * dev->pitch);
+                                }
+                                dev->accel.sy--;
+
+                                if (dev->accel.sy < 0) {
+                                    dev->accel.destx = dev->accel.dx;
+                                    dev->accel.desty = dev->accel.dy;
+                                    return;
+                                }
+                            }
+                        }
+                    }
+                } else {
+                    if ((dev->accel_bpp == 24) && (dev->accel.cmd == 0xc2b5)) {
+                        int64_t cx;
+                        int64_t dx;
+
+                        cx = (int64_t) dev->accel.cx;
+                        dx = (int64_t) dev->accel.dx;
+
+                        while (1) {
+                            if ((dx >= (((int64_t)clip_l) * 3)) &&
+                                (dx <= (((uint64_t)clip_r) * 3)) &&
+                                (dev->accel.dy >= (clip_t << 1)) &&
+                                (dev->accel.dy <= (clip_b << 1))) {
+
+                                READ(dev->accel.src + cx, src_dat);
+                                READ(dev->accel.dest + dx, dest_dat);
+                                dest_dat = (src_dat & wrt_mask) | (dest_dat & ~wrt_mask);
+                                WRITE(dev->accel.dest + dx, dest_dat);
+                            }
+
+                            cx++;
+                            dx++;
+
+                            dev->accel.sx--;
+                            if (dev->accel.sx < 0)
+                                return;
+                        }
+                    } else {
                         while (count-- && dev->accel.sy >= 0) {
-                            if ((dev->accel.dx >= dev->accel.clip_left) && (dev->accel.dx <= clip_r) &&
-                                (dev->accel.dy >= dev->accel.clip_top) && (dev->accel.dy <= clip_b)) {
+                            if ((dev->accel.dx >= clip_l) &&
+                                (dev->accel.dx <= clip_r) &&
+                                (dev->accel.dy >= clip_t) &&
+                                (dev->accel.dy <= clip_b)) {
                                 if (pixcntl == 3) {
                                     if (!(dev->accel.cmd & 0x10) && ((frgd_mix != 3) || (bkgd_mix != 3))) {
                                         READ(dev->accel.src + dev->accel.cx, mix_dat);
@@ -3766,9 +3090,8 @@ bitblt:
                                     case 3:
                                         READ(dev->accel.src + dev->accel.cx, src_dat);
                                         if (pixcntl == 3) {
-                                            if ((dev->accel.cmd & 0x10) && !(dev->accel.cmd & 0x40)) {
+                                            if ((dev->accel.cmd & 0x10) && !(dev->accel.cmd & 0x40))
                                                 src_dat = ((src_dat & rd_mask) == rd_mask);
-                                            }
                                         }
                                         break;
 
@@ -3778,12 +3101,18 @@ bitblt:
 
                                 READ(dev->accel.dest + dev->accel.dx, dest_dat);
 
-                                if ((compare_mode == 0) || ((compare_mode == 0x10) && (dest_dat >= compare)) || ((compare_mode == 0x18) && (dest_dat < compare)) || ((compare_mode == 0x20) && (dest_dat != compare)) || ((compare_mode == 0x28) && (dest_dat == compare)) || ((compare_mode == 0x30) && (dest_dat <= compare)) || ((compare_mode == 0x38) && (dest_dat > compare))) {
+                                if ((compare_mode == 0) ||
+                                    ((compare_mode == 0x10) && (dest_dat >= compare)) ||
+                                    ((compare_mode == 0x18) && (dest_dat < compare)) ||
+                                    ((compare_mode == 0x20) && (dest_dat != compare)) ||
+                                    ((compare_mode == 0x28) && (dest_dat == compare)) ||
+                                    ((compare_mode == 0x30) && (dest_dat <= compare)) ||
+                                    ((compare_mode == 0x38) && (dest_dat > compare))) {
                                     old_dest_dat = dest_dat;
                                     MIX(mix_dat & mix_mask, dest_dat, src_dat);
                                     dest_dat = (dest_dat & wrt_mask) | (old_dest_dat & ~wrt_mask);
-                                    if (dev->accel.cmd & 4) {
-                                        if (dev->accel.sx > 0) {
+                                    if (dev->accel.cmd & 0x04) {
+                                        if (dev->accel.sx) {
                                             WRITE(dev->accel.dest + dev->accel.dx, dest_dat);
                                         }
                                     } else {
@@ -3808,15 +3137,13 @@ bitblt:
                                 dev->accel.fill_state = 0;
                                 dev->accel.sx = dev->accel.maj_axis_pcnt & 0x7ff;
 
-                                dev->accel.dx = dev->accel.destx;
-
-                                if (dev->accel.destx >= 0x600)
-                                    dev->accel.dx |= ~0x5ff;
-
-                                dev->accel.cx = dev->accel.cur_x;
-
-                                if (dev->accel.cur_x >= 0x600)
-                                    dev->accel.cx |= ~0x5ff;
+                                if (dev->accel.cmd & 0x20) {
+                                    dev->accel.dx -= (dev->accel.sx + 1);
+                                    dev->accel.cx -= (dev->accel.sx + 1);
+                                } else {
+                                    dev->accel.dx += (dev->accel.sx + 1);
+                                    dev->accel.cx += (dev->accel.sx + 1);
+                                }
 
                                 if (dev->accel.cmd & 0x80) {
                                     dev->accel.dy++;
@@ -3826,12 +3153,23 @@ bitblt:
                                     dev->accel.cy--;
                                 }
 
-                                dev->accel.dest = dev->accel.dy * dev->pitch;
-                                dev->accel.src  = dev->accel.cy * dev->pitch;
+                                if ((dev->accel_bpp == 24) || (dev->accel_bpp <= 8)) {
+                                    dev->accel.src  = (dev->accel.ge_offset << 2) + (dev->accel.cy * dev->pitch);
+                                    dev->accel.dest = (dev->accel.ge_offset << 2) + (dev->accel.dy * dev->pitch);
+                                } else if (dev->bpp) {
+                                    dev->accel.src  = (dev->accel.ge_offset << 1) + (dev->accel.cy * dev->pitch);
+                                    dev->accel.dest = (dev->accel.ge_offset << 1) + (dev->accel.dy * dev->pitch);
+                                } else {
+                                    dev->accel.src  = (dev->accel.ge_offset << 2) + (dev->accel.cy * dev->pitch);
+                                    dev->accel.dest = (dev->accel.ge_offset << 2) + (dev->accel.dy * dev->pitch);
+                                }
                                 dev->accel.sy--;
 
-                                if (dev->accel.sy < 0)
+                                if (dev->accel.sy < 0) {
+                                    dev->accel.destx = dev->accel.dx;
+                                    dev->accel.desty = dev->accel.dy;
                                     return;
+                                }
                             }
                         }
                     }
@@ -3882,16 +3220,16 @@ ibm8514_render_8bpp(svga_t *svga)
 
         for (int x = 0; x <= dev->h_disp; x += 8) {
             dat  = *(uint32_t *) (&dev->vram[dev->ma & dev->vram_mask]);
-            p[0] = dev->pallook[dat & 0xff];
-            p[1] = dev->pallook[(dat >> 8) & 0xff];
-            p[2] = dev->pallook[(dat >> 16) & 0xff];
-            p[3] = dev->pallook[(dat >> 24) & 0xff];
+            p[0] = dev->pallook[dat & dev->dac_mask];
+            p[1] = dev->pallook[(dat >> 8) & dev->dac_mask];
+            p[2] = dev->pallook[(dat >> 16) & dev->dac_mask];
+            p[3] = dev->pallook[(dat >> 24) & dev->dac_mask];
 
             dat  = *(uint32_t *) (&dev->vram[(dev->ma + 4) & dev->vram_mask]);
-            p[4] = dev->pallook[dat & 0xff];
-            p[5] = dev->pallook[(dat >> 8) & 0xff];
-            p[6] = dev->pallook[(dat >> 16) & 0xff];
-            p[7] = dev->pallook[(dat >> 24) & 0xff];
+            p[4] = dev->pallook[dat & dev->dac_mask];
+            p[5] = dev->pallook[(dat >> 8) & dev->dac_mask];
+            p[6] = dev->pallook[(dat >> 16) & dev->dac_mask];
+            p[7] = dev->pallook[(dat >> 24) & dev->dac_mask];
 
             dev->ma += 8;
             p += 8;
@@ -4147,7 +3485,7 @@ ibm8514_poll(void *priv)
     int      wy;
 
     ibm8514_log("IBM 8514/A poll.\n");
-    if (dev->on[0] || dev->on[1]) {
+    if (dev->on) {
         ibm8514_log("ON!\n");
         if (!dev->linepos) {
             if ((dev->displine == ((dev->hwcursor_latch.y < 0) ? 0 : dev->hwcursor_latch.y)) && dev->hwcursor_latch.ena) {
@@ -4302,7 +3640,7 @@ ibm8514_recalctimings(svga_t *svga)
     } else
 #endif
     {
-        if (dev->on[0]) {
+        if (dev->on) {
             dev->h_total = dev->htotal + 1;
             dev->rowcount = !!(dev->disp_cntl & 0x08);
 
@@ -4336,6 +3674,11 @@ ibm8514_recalctimings(svga_t *svga)
             dev->pitch = 1024;
             dev->rowoffset = 0x80;
             svga->map8 = dev->pallook;
+            if (dev->vram_512k_8514) {
+                if (dev->h_disp == 640)
+                    dev->pitch = 640;
+            }
+            dev->accel_bpp = 8;
             svga->render8514 = ibm8514_render_8bpp;
             ibm8514_log("BPP=%d, Pitch = %d, rowoffset = %d, crtc13 = %02x, highres bit = %02x, has_vga? = %d.\n", dev->bpp, dev->pitch, dev->rowoffset, svga->crtc[0x13], dev->accel.advfunc_cntl & 4, !ibm8514_standalone_enabled);
         }
@@ -4382,8 +3725,7 @@ ibm8514_mca_reset(void *priv)
     ibm8514_t       *dev  = (ibm8514_t *) svga->dev8514;
 
     ibm8514_log("MCA reset.\n");
-    dev->on[0] = 0;
-    dev->on[1] = 0;
+    dev->on = 0;
     vga_on = 1;
 #ifdef ATI_8514_ULTRA
     if (dev->extensions)
@@ -4391,6 +3733,8 @@ ibm8514_mca_reset(void *priv)
     else
 #endif
         ibm8514_mca_write(0x102, 0, svga);
+
+    timer_set_callback(&svga->timer, svga_poll);
 }
 
 static void *
@@ -4409,12 +3753,15 @@ ibm8514_init(const device_t *info)
     svga->dev8514    = dev;
     svga->ext8514    = NULL;
 
-    dev->vram_size   = 1024 << 10;
+    dev->vram_amount = device_get_config_int("memory");
+    dev->vram_512k_8514 = dev->vram_amount == 512;
+    dev->vram_size   = dev->vram_amount << 10;
     dev->vram        = calloc(dev->vram_size, 1);
     dev->changedvram = calloc((dev->vram_size >> 12) + 1, 1);
     dev->vram_mask   = dev->vram_size - 1;
     dev->map8        = dev->pallook;
     dev->local       = 0;
+    dev->accel_bpp   = 8;
 
     dev->type     = info->flags;
     dev->bpp      = 0;
@@ -4450,6 +3797,7 @@ ibm8514_init(const device_t *info)
                              0x1000, MEM_MAPPING_EXTERNAL);
                     ati_eeprom_load(&mach->eeprom, "ati8514.nvr", 0);
                     mach->accel.scratch0 = (((bios_addr >> 7) - 0x1000) >> 4);
+                    mach->accel.scratch1 = mach->accel.scratch0 - 0x80;
                 }
                 ati8514_init(svga, svga->ext8514, svga->dev8514);
                 break;
@@ -4519,6 +3867,25 @@ ibm8514_force_redraw(void *priv)
 #ifdef ATI_8514_ULTRA
 // clang-format off
 static const device_config_t ext8514_config[] = {
+    {
+        .name = "memory",
+        .description = "Memory size",
+        .type = CONFIG_SELECTION,
+        .default_int = 1024,
+        .selection = {
+            {
+                .description = "512 KB",
+                .value = 512
+            },
+            {
+                .description = "1 MB",
+                .value = 1024
+            },
+            {
+                .description = ""
+            }
+        }
+    },
     {
         .name = "extensions",
         .description = "Vendor",
@@ -4566,6 +3933,32 @@ static const device_config_t ext8514_config[] = {
         .type = CONFIG_END
     }
 };
+#else
+// clang-format off
+static const device_config_t ext8514_config[] = {
+    {
+        .name = "memory",
+        .description = "Memory size",
+        .type = CONFIG_SELECTION,
+        .default_int = 1024,
+        .selection = {
+            {
+                .description = "512 KB",
+                .value = 512
+            },
+            {
+                .description = "1 MB",
+                .value = 1024
+            },
+            {
+                .description = ""
+            }
+        }
+    },
+    {
+        .type = CONFIG_END
+    }
+};
 #endif
 
 // clang-format off
@@ -4580,11 +3973,7 @@ const device_t gen8514_isa_device = {
     { .available = NULL },
     .speed_changed = ibm8514_speed_changed,
     .force_redraw = ibm8514_force_redraw,
-#ifdef ATI_8514_ULTRA
     .config = ext8514_config
-#else
-    .config = NULL
-#endif
 };
 
 const device_t ibm8514_mca_device = {
@@ -4598,11 +3987,7 @@ const device_t ibm8514_mca_device = {
     { .available = NULL },
     .speed_changed = ibm8514_speed_changed,
     .force_redraw = ibm8514_force_redraw,
-#ifdef ATI_8514_ULTRA
     .config = ext8514_config
-#else
-    .config = NULL
-#endif
 };
 
 
diff --git a/src/video/vid_ati68860_ramdac.c b/src/video/vid_ati68860_ramdac.c
index ac2a0f7cb..cb6de4353 100644
--- a/src/video/vid_ati68860_ramdac.c
+++ b/src/video/vid_ati68860_ramdac.c
@@ -74,16 +74,16 @@ ati68860_ramdac_out(uint16_t addr, uint8_t val, void *priv, svga_t *svga)
 
     switch (addr) {
         case 0:
-            svga_out((dev && (dev->on[0] || dev->on[1])) ? 0x2ec : 0x3c8, val, svga);
+            svga_out((dev && dev->on) ? 0x2ec : 0x3c8, val, svga);
             break;
         case 1:
-            svga_out((dev && (dev->on[0] || dev->on[1])) ? 0x2ed : 0x3c9, val, svga);
+            svga_out((dev && dev->on) ? 0x2ed : 0x3c9, val, svga);
             break;
         case 2:
-            svga_out((dev && (dev->on[0] || dev->on[1])) ? 0x2ea : 0x3c6, val, svga);
+            svga_out((dev && dev->on) ? 0x2ea : 0x3c6, val, svga);
             break;
         case 3:
-            svga_out((dev && (dev->on[0] || dev->on[1])) ? 0x2eb : 0x3c7, val, svga);
+            svga_out((dev && dev->on) ? 0x2eb : 0x3c7, val, svga);
             break;
         default:
             ramdac->regs[addr & 0xf] = val;
@@ -181,16 +181,16 @@ ati68860_ramdac_in(uint16_t addr, void *priv, svga_t *svga)
 
     switch (addr) {
         case 0:
-            temp = svga_in((dev && (dev->on[0] || dev->on[1])) ? 0x2ec : 0x3c8, svga);
+            temp = svga_in((dev && dev->on) ? 0x2ec : 0x3c8, svga);
             break;
         case 1:
-            temp = svga_in((dev && (dev->on[0] || dev->on[1])) ? 0x2ed : 0x3c9, svga);
+            temp = svga_in((dev && dev->on) ? 0x2ed : 0x3c9, svga);
             break;
         case 2:
-            temp = svga_in((dev && (dev->on[0] || dev->on[1])) ? 0x2ea : 0x3c6, svga);
+            temp = svga_in((dev && dev->on) ? 0x2ea : 0x3c6, svga);
             break;
         case 3:
-            temp = svga_in((dev && (dev->on[0] || dev->on[1])) ? 0x2eb : 0x3c7, svga);
+            temp = svga_in((dev && dev->on) ? 0x2eb : 0x3c7, svga);
             break;
         case 4:
         case 8:
diff --git a/src/video/vid_ati_mach8.c b/src/video/vid_ati_mach8.c
index d41f65ed9..7ef0d2475 100644
--- a/src/video/vid_ati_mach8.c
+++ b/src/video/vid_ati_mach8.c
@@ -117,9 +117,8 @@ mach_log(const char *fmt, ...)
                 mach->accel.pix_trans[(n)] = vram_w[(dev->accel.dest + (cx) + (n)) & (dev->vram_mask >> 1)] & 0xff; \
             else \
                 mach->accel.pix_trans[(n)] = vram_w[(dev->accel.dest + (cx) + (n)) & (dev->vram_mask >> 1)] >> 8; \
-        } else { \
+        } else \
             mach->accel.pix_trans[(n)] = dev->vram[(dev->accel.dest + (cx) + (n)) & dev->vram_mask]; \
-        } \
     }
 
 #define READ_PIXTRANS_WORD(cx, n)                                                                    \
@@ -152,6 +151,9 @@ mach_log(const char *fmt, ...)
         else \
             dat = dev->vram[(addr) & (dev->vram_mask)];
 
+#define READ_HIGH(addr, dat) \
+        dat |= (dev->vram[(addr) & (dev->vram_mask)] << 8);
+
 #define MIX(mixmode, dest_dat, src_dat)                                                       \
     {                                                                                         \
         switch ((mixmode) ? (dev->accel.frgd_mix & 0x1f) : (dev->accel.bkgd_mix & 0x1f)) {    \
@@ -237,7 +239,10 @@ mach_log(const char *fmt, ...)
                 dest_dat = MAX(0, (src_dat - dest_dat));                                      \
                 break;                                                                        \
             case 0x1b:                                                                        \
-                dest_dat = MIN(0xff, (dest_dat + src_dat));                                   \
+                if (dev->bpp)                                                                 \
+                    dest_dat = MIN(0xffff, (dest_dat + src_dat));                             \
+                else                                                                          \
+                    dest_dat = MIN(0xff, (dest_dat + src_dat));                               \
                 break;                                                                        \
             case 0x1c:                                                                        \
                 dest_dat = MAX(0, (dest_dat - src_dat)) / 2;                                  \
@@ -249,7 +254,10 @@ mach_log(const char *fmt, ...)
                 dest_dat = MAX(0, (src_dat - dest_dat)) / 2;                                  \
                 break;                                                                        \
             case 0x1f:                                                                        \
-                dest_dat = (0xff < (src_dat + dest_dat)) ? 0xff : ((src_dat + dest_dat) / 2); \
+                if (dev->bpp)                                                                 \
+                    dest_dat = (0xffff < (src_dat + dest_dat)) ? 0xffff : ((src_dat + dest_dat) / 2); \
+                else                                                                          \
+                    dest_dat = (0xff < (src_dat + dest_dat)) ? 0xff : ((src_dat + dest_dat) / 2); \
                 break;                                                                        \
         }                                                                                     \
     }
@@ -299,12 +307,12 @@ mach_accel_start(int cmd_type, int cpu_input, int count, uint32_t mix_dat, uint3
     uint16_t      old_dest_dat;
     uint16_t     *vram_w    = (uint16_t *) dev->vram;
     uint16_t      mix       = 0;
-    int16_t       clip_l    = dev->accel.clip_left & 0x7ff;
-    int16_t       clip_t    = dev->accel.clip_top & 0x7ff;
-    int16_t       clip_r    = dev->accel.multifunc[4] & 0x7ff;
-    int16_t       clip_b    = dev->accel.multifunc[3] & 0x7ff;
     uint32_t      mono_dat0 = 0;
     uint32_t      mono_dat1 = 0;
+    int16_t      clip_t    = dev->accel.multifunc[1] & 0x7ff;
+    int16_t      clip_l    = dev->accel.multifunc[2] & 0x7ff;
+    int16_t      clip_b    = dev->accel.multifunc[3] & 0x7ff;
+    int16_t      clip_r    = dev->accel.multifunc[4] & 0x7ff;
 
     if (!dev->bpp) {
         rd_mask &= 0xff;
@@ -324,17 +332,13 @@ mach_accel_start(int cmd_type, int cpu_input, int count, uint32_t mix_dat, uint3
 
     if (cpu_input) {
         if (dev->bpp) {
-            if ((mach->accel.dp_config & 0x200) && (count == 2)) {
+            if ((mach->accel.dp_config & 0x200) && (count == 2))
                 count >>= 1;
-            }
         }
     }
 
-    if ((dev->accel_bpp == 8) || (dev->accel_bpp == 15) || (dev->accel_bpp == 16) || (dev->accel_bpp == 24)) {
-        if ((cmd_type == 2) && cpu_input) {
-            mach_log("RdMask=%04x, DPCONFIG=%04x, Clipping: l=%d, r=%d, t=%d, b=%d, LineDrawOpt=%04x, BPP=%d, CMDType = %d, offs=%08x, cnt = %d, input = %d, mono_src = %d, frgdsel = %d, d(%d,%d), dstxend = %d, pitch = %d, extcrt = %d, rw = %x, monpattern = %x.\n", rd_mask, mach->accel.dp_config, clip_l, clip_r, clip_t, clip_b, mach->accel.linedraw_opt, dev->accel_bpp, cmd_type, mach->accel.ge_offset, count, cpu_input, mono_src, frgd_sel, dev->accel.cur_x, dev->accel.cur_y, mach->accel.dest_x_end, dev->ext_pitch, dev->ext_crt_pitch, mach->accel.dp_config & 1, mach->accel.mono_pattern_enable);
-        }
-    }
+    if ((dev->accel_bpp == 8) || (dev->accel_bpp == 15) || (dev->accel_bpp == 16) || (dev->accel_bpp == 24))
+        mach_log("RdMask=%04x, DPCONFIG=%04x, Clipping: l=%d, r=%d, t=%d, b=%d, LineDrawOpt=%04x, BPP=%d, CMDType = %d, offs=%08x, cnt = %d, input = %d, mono_src = %d, frgdsel = %d, d(%d,%d), dstxend = %d, pitch = %d, extcrt = %d, rw = %x, monpattern = %x.\n", dev->accel.rd_mask, mach->accel.dp_config, clip_l, clip_r, clip_t, clip_b, mach->accel.linedraw_opt, dev->accel_bpp, cmd_type, mach->accel.ge_offset, count, cpu_input, mono_src, frgd_sel, dev->accel.cur_x, dev->accel.cur_y, mach->accel.dest_x_end, dev->ext_pitch, dev->ext_crt_pitch, mach->accel.dp_config & 1, mach->accel.mono_pattern_enable);
 
     switch (cmd_type) {
         case 1: /*Extended Raw Linedraw from bres_count register (0x96ee)*/
@@ -342,6 +346,7 @@ mach_accel_start(int cmd_type, int cpu_input, int count, uint32_t mix_dat, uint3
                 dev->accel.dx = dev->accel.cur_x;
                 if (dev->accel.cur_x >= 0x600)
                     dev->accel.dx |= ~0x5ff;
+
                 dev->accel.dy = dev->accel.cur_y;
                 if (dev->accel.cur_y >= 0x600)
                     dev->accel.dy |= ~0x5ff;
@@ -349,6 +354,7 @@ mach_accel_start(int cmd_type, int cpu_input, int count, uint32_t mix_dat, uint3
                 dev->accel.cx = dev->accel.destx_distp;
                 if (dev->accel.destx_distp >= 0x600)
                     dev->accel.cx |= ~0x5ff;
+
                 dev->accel.cy = dev->accel.desty_axstp;
                 if (dev->accel.desty_axstp >= 0x600)
                     dev->accel.cy |= ~0x5ff;
@@ -357,8 +363,6 @@ mach_accel_start(int cmd_type, int cpu_input, int count, uint32_t mix_dat, uint3
                 dev->accel.sx         = 0;
                 mach->accel.poly_fill = 0;
 
-                mach->accel.color_pattern_idx = ((dev->accel.cx + (dev->accel.cy << 3)) & mach->accel.patt_len);
-
                 mach->accel.stepx = (mach->accel.linedraw_opt & 0x20) ? 1 : -1;
                 mach->accel.stepy = (mach->accel.linedraw_opt & 0x80) ? 1 : -1;
 
@@ -377,19 +381,9 @@ mach_accel_start(int cmd_type, int cpu_input, int count, uint32_t mix_dat, uint3
                 }
             }
 
-            if (frgd_sel == 5) {
-                for (int x = 0; x <= mach->accel.patt_len; x++) {
-                    mach->accel.color_pattern[x] = mach->accel.patt_data[x & mach->accel.patt_len];
-                }
-
-                /*The destination coordinates should match the pattern index.*/
-                if (mach->accel.color_pattern_idx != mach->accel.patt_idx)
-                    mach->accel.color_pattern_idx = mach->accel.patt_idx;
-            }
-
             if (mono_src == 1) {
-                count               = mach->accel.width;
-                mix_dat             = mach->accel.patt_data[0x10];
+                count = mach->accel.width;
+                mix_dat = mach->accel.mono_pattern_normal[0];
                 dev->accel.temp_cnt = 8;
             }
 
@@ -404,7 +398,7 @@ mach_accel_start(int cmd_type, int cpu_input, int count, uint32_t mix_dat, uint3
                                 dev->accel.temp_cnt = 8;
                                 mix_dat >>= 8;
                             }
-                            mix = (mix_dat & 0x80);
+                            mix = !!(mix_dat & 0x80);
                             dev->accel.temp_cnt--;
                             mix_dat <<= 1;
                             mix_dat |= 1;
@@ -415,10 +409,10 @@ mach_accel_start(int cmd_type, int cpu_input, int count, uint32_t mix_dat, uint3
                                 mix_dat <<= 1;
                             } else {
                                 if (mach->accel.dp_config & 0x200) {
-                                    mix = mix_dat & 1;
+                                    mix = mix_dat & 0x01;
                                     mix_dat >>= 1;
                                 } else {
-                                    mix = mix_dat & 0x80;
+                                    mix = !!(mix_dat & 0x80);
                                     mix_dat <<= 1;
                                     mix_dat |= 1;
                                 }
@@ -437,7 +431,10 @@ mach_accel_start(int cmd_type, int cpu_input, int count, uint32_t mix_dat, uint3
                             break;
                     }
 
-                    if (((dev->accel.dx) >= clip_l) && ((dev->accel.dx) <= clip_r) && ((dev->accel.dy) >= clip_t) && ((dev->accel.dy) <= clip_b)) {
+                    if ((dev->accel.dx >= clip_l) &&
+                        (dev->accel.dx <= clip_r) &&
+                        (dev->accel.dy >= clip_t) &&
+                        (dev->accel.dy <= clip_b)) {
                         switch (mix ? frgd_sel : bkgd_sel) {
                             case 0:
                                 src_dat = dev->accel.bkgd_color;
@@ -453,20 +450,18 @@ mach_accel_start(int cmd_type, int cpu_input, int count, uint32_t mix_dat, uint3
                                     src_dat = cpu_dat;
                                 else {
                                     if (dev->bpp) {
-                                        READ((mach->accel.ge_offset << 1) + ((dev->accel.cy) * (dev->pitch)) + (dev->accel.cx), src_dat);
+                                        READ((mach->accel.ge_offset << 1) + (dev->accel.cy * dev->pitch) + dev->accel.cx, src_dat);
                                     } else {
-                                        READ((mach->accel.ge_offset << 2) + ((dev->accel.cy) * (dev->pitch)) + (dev->accel.cx), src_dat);
+                                        READ((mach->accel.ge_offset << 2) + (dev->accel.cy * dev->pitch) + dev->accel.cx, src_dat);
                                     }
-                                    if (mono_src == 3) {
+                                    if (mono_src == 3)
                                         src_dat = (src_dat & rd_mask) == rd_mask;
-                                    }
                                 }
                                 break;
                             case 5:
-                                if (mix) {
-                                    src_dat = mach->accel.color_pattern[((dev->accel.dx) + ((dev->accel.dy) << 3)) & mach->accel.patt_len];
-                                } else
-                                    src_dat = 0;
+                                src_dat = mach->accel.color_pattern[mach->accel.color_pattern_idx];
+                                if (dev->bpp)
+                                    src_dat |= (mach->accel.color_pattern[mach->accel.color_pattern_idx + 1] << 8);
                                 break;
 
                             default:
@@ -556,6 +551,14 @@ mach_accel_start(int cmd_type, int cpu_input, int count, uint32_t mix_dat, uint3
                     else
                         cpu_dat >>= 8;
 
+                    if (dev->bpp)
+                        mach->accel.color_pattern_idx += 2;
+                    else
+                        mach->accel.color_pattern_idx++;
+
+                    if (mach->accel.color_pattern_idx > mach->accel.patt_len)
+                        mach->accel.color_pattern_idx = 0;
+
                     switch (mach->accel.linedraw_opt & 0xe0) {
                         case 0x00:
                             dev->accel.cx++;
@@ -648,7 +651,10 @@ mach_accel_start(int cmd_type, int cpu_input, int count, uint32_t mix_dat, uint3
                             break;
                     }
 
-                    if (((dev->accel.dx) >= clip_l) && ((dev->accel.dx) <= clip_r) && ((dev->accel.dy) >= clip_t) && ((dev->accel.dy) <= clip_b)) {
+                    if ((dev->accel.dx >= clip_l) &&
+                        (dev->accel.dx <= clip_r) &&
+                        (dev->accel.dy >= clip_t) &&
+                        (dev->accel.dy <= clip_b)) {
                         switch (mix ? frgd_sel : bkgd_sel) {
                             case 0:
                                 src_dat = dev->accel.bkgd_color;
@@ -674,10 +680,9 @@ mach_accel_start(int cmd_type, int cpu_input, int count, uint32_t mix_dat, uint3
                                 }
                                 break;
                             case 5:
-                                if (mix) {
-                                    src_dat = mach->accel.color_pattern[((dev->accel.dx) + ((dev->accel.dy) << 3)) & mach->accel.patt_len];
-                                } else
-                                    src_dat = 0;
+                                src_dat = mach->accel.color_pattern[mach->accel.color_pattern_idx];
+                                if (dev->bpp)
+                                    src_dat |= (mach->accel.color_pattern[mach->accel.color_pattern_idx + 1] << 8);
                                 break;
 
                             default:
@@ -767,6 +772,14 @@ mach_accel_start(int cmd_type, int cpu_input, int count, uint32_t mix_dat, uint3
                     else
                         cpu_dat >>= 8;
 
+                    if (dev->bpp)
+                        mach->accel.color_pattern_idx += 2;
+                    else
+                        mach->accel.color_pattern_idx++;
+
+                    if (mach->accel.color_pattern_idx > mach->accel.patt_len)
+                        mach->accel.color_pattern_idx = 0;
+
                     if (mach->accel.linedraw_opt & 0x40) {
                         dev->accel.dy += mach->accel.stepy;
                         if ((frgd_sel == 3) || (bkgd_sel == 3))
@@ -817,9 +830,8 @@ mach_accel_start(int cmd_type, int cpu_input, int count, uint32_t mix_dat, uint3
                     dev->accel.dy |= ~0x5ff;
 
                 if (mach->accel.dp_config == 0x5211) {
-                    if (mach->accel.dest_x_end == 1024) {
+                    if (mach->accel.dest_x_end == 1024)
                         goto skip_dx;
-                    }
                 }
                 /*Destination Width*/
                 if (mach->accel.dest_x_start != dev->accel.dx)
@@ -849,14 +861,8 @@ skip_dx:
                     mach_log("BitBLT: Dst Indeterminate X, dpconfig = %04x, destxend = %d, destxstart = %d.\n", mach->accel.dp_config, mach->accel.dest_x_end, mach->accel.dest_x_start);
                 }
 
-                dev->accel.sx                 = 0;
-                mach->accel.poly_fill         = 0;
-                mach->accel.color_pattern_idx = ((dev->accel.dx + (dev->accel.dy << 3)) & mach->accel.patt_len);
-                if ((dev->accel_bpp == 24) && (mono_src != 1)) {
-                    if (mach->accel.color_pattern_idx == mach->accel.patt_len)
-                        mach->accel.color_pattern_idx = mach->accel.patt_data_idx;
-                } else if ((dev->accel_bpp == 24) && (frgd_sel == 5) && (mono_src == 1) && (mach->accel.patt_len_reg & 0x4000))
-                    mach->accel.color_pattern_idx = 0;
+                dev->accel.sx = 0;
+                mach->accel.poly_fill = 0;
 
                 /*Height*/
                 mach->accel.dy_start = dev->accel.cur_y;
@@ -878,13 +884,16 @@ skip_dx:
                 }
 
                 dev->accel.sy = 0;
-                if (dev->bpp)
-                    dev->accel.dest = (mach->accel.ge_offset << 1) + (dev->accel.dy * (dev->pitch));
-                else
-                    dev->accel.dest = (mach->accel.ge_offset << 2) + (dev->accel.dy * (dev->pitch));
+                if (mach->accel.dp_config & 0x02)
+                    dev->accel.dest = (dev->accel.dy * dev->pitch);
+                else {
+                    if (dev->bpp)
+                        dev->accel.dest = (mach->accel.ge_offset << 1) + (dev->accel.dy * dev->pitch);
+                    else
+                        dev->accel.dest = (mach->accel.ge_offset << 2) + (dev->accel.dy * dev->pitch);
+                }
 
                 mach->accel.src_stepx = 0;
-
                 /*Source Width*/
                 dev->accel.cx = mach->accel.src_x;
                 if (mach->accel.src_x >= 0x600)
@@ -918,27 +927,28 @@ skip_dx:
                     mach_log("BitBLT: Src Indeterminate X: width = %d, coordinates: %d,%d px, end: %d px, stepx = %d, dpconfig = %04x, oddwidth = %d.\n", mach->accel.src_width, dev->accel.cx, dev->accel.cy, mach->accel.src_x_end, mach->accel.src_stepx, mach->accel.dp_config, mach->accel.src_width & 1);
                 }
                 mach->accel.sx = 0;
-                if (dev->bpp)
-                    dev->accel.src = (mach->accel.ge_offset << 1) + (dev->accel.cy * (dev->pitch));
+                if (mach->accel.patt_data_idx < 16)
+                    mach->accel.color_pattern_idx = mach->accel.patt_idx;
                 else
-                    dev->accel.src = (mach->accel.ge_offset << 2) + (dev->accel.cy * (dev->pitch));
+                    mach->accel.color_pattern_idx = 0;
 
-                if ((dev->accel_bpp == 24) && (frgd_sel == 5))
-                    mach_log("BitBLT=%04x, WH(%d,%d), SRCWidth=%d, c(%d,%d), s(%d,%d).\n", mach->accel.dp_config, mach->accel.width, mach->accel.height, mach->accel.src_width, dev->accel.dx, dev->accel.dy, dev->accel.cx, dev->accel.cy);
-                else if (mach->accel.dp_config & 0x02)
-                    mach_log("BitBLT=%04x, Pitch=%d, C(%d,%d), D(%d,%d), SRCWidth=%d, SRCXStep=%d, WH(%d,%d), clipt=%d, clipb=%d, geoffset=%08x.\n", mach->accel.dp_config, dev->ext_pitch, mach->accel.src_x, mach->accel.src_y, dev->accel.cur_x, dev->accel.cur_y, mach->accel.src_width, mach->accel.src_stepx, mach->accel.width, mach->accel.height, clip_t, clip_b, (mach->accel.ge_offset << 2));
+                if (dev->bpp)
+                    dev->accel.src = (mach->accel.ge_offset << 1) + (dev->accel.cy * dev->pitch);
+                else
+                    dev->accel.src = (mach->accel.ge_offset << 2) + (dev->accel.cy * dev->pitch);
 
                 if (mono_src == 1) {
-                    if ((mach->accel.mono_pattern_enable) && !(mach->accel.patt_len_reg & 0x4000)) {
-                        mono_dat0 = mach->accel.patt_data[0x10];
-                        mono_dat0 |= (mach->accel.patt_data[0x11] << 8);
-                        mono_dat0 |= (mach->accel.patt_data[0x12] << 16);
-                        mono_dat0 |= (mach->accel.patt_data[0x13] << 24);
-                        mono_dat1 = mach->accel.patt_data[0x14];
-                        mono_dat1 |= (mach->accel.patt_data[0x15] << 8);
-                        mono_dat1 |= (mach->accel.patt_data[0x16] << 16);
-                        mono_dat1 |= (mach->accel.patt_data[0x17] << 24);
+                    if (mach->accel.mono_pattern_enable || mach->accel.block_write_mono_pattern_enable) {
+                        mono_dat0 = mach->accel.mono_pattern_normal[0];
+                        mono_dat0 |= (mach->accel.mono_pattern_normal[1] << 8);
+                        mono_dat0 |= (mach->accel.mono_pattern_normal[2] << 16);
+                        mono_dat0 |= (mach->accel.mono_pattern_normal[3] << 24);
+                        mono_dat1 = mach->accel.mono_pattern_normal[4];
+                        mono_dat1 |= (mach->accel.mono_pattern_normal[5] << 8);
+                        mono_dat1 |= (mach->accel.mono_pattern_normal[6] << 16);
+                        mono_dat1 |= (mach->accel.mono_pattern_normal[7] << 24);
 
+                        mach_log("MonoData0=%x, MonoData1=%x.\n", mono_dat0, mono_dat1);
                         for (uint8_t y = 0; y < 8; y++) {
                             for (uint8_t x = 0; x < 8; x++) {
                                 uint32_t temp                      = (y & 4) ? mono_dat1 : mono_dat0;
@@ -962,35 +972,12 @@ skip_dx:
             }
 
             if (mono_src == 1) {
-                if (!mach->accel.mono_pattern_enable && !(mach->accel.patt_len_reg & 0x4000)) {
-                    count               = mach->accel.width;
-                    mix_dat             = mach->accel.patt_data[0x10] ^ ((mach->accel.patt_idx & 1) ? 0xff : 0);
-                    dev->accel.temp_cnt = 8;
-                }
-            }
-
-            if (frgd_sel == 5) {
-                if (dev->bpp) {
-                    for (int x = 0; x <= mach->accel.patt_len; x += 2) {
-                        mach->accel.color_pattern_word[x + (mach->accel.color_pattern_idx & 1)] = (mach->accel.patt_data[x & mach->accel.patt_len] & 0xff);
-                        mach->accel.color_pattern_word[x + (mach->accel.color_pattern_idx & 1)] |= (mach->accel.patt_data[(x + 1) & mach->accel.patt_len] << 8);
-                    }
-                } else {
-                    if ((dev->accel_bpp == 24) && (mach->accel.patt_len < 3)) {
-                        for (int x = 0; x <= mach->accel.patt_len; x++) {
-                            mach->accel.color_pattern[x] = mach->accel.patt_data[x];
-                            mach_log("BITBLT: Color Pattern 24bpp[%d]=%02x, dataidx=%d, pattlen=%d.\n", x, mach->accel.color_pattern[x], mach->accel.patt_data_idx, mach->accel.patt_len);
-                        }
-                    } else {
-                        for (int x = 0; x <= mach->accel.patt_len; x++) {
-                            mach->accel.color_pattern[x] = mach->accel.patt_data[x & mach->accel.patt_len];
-                        }
+                if (!mach->accel.mono_pattern_enable && !mach->accel.block_write_mono_pattern_enable) {
+                    if (((dev->accel_bpp == 24) && (frgd_sel != 5)) || (dev->accel_bpp != 24)) {
+                        mix_dat = mach->accel.mono_pattern_normal[0] ^ ((mach->accel.patt_idx & 0x01) ? 0xff : 0);
+                        dev->accel.temp_cnt = 8;
                     }
                 }
-
-                /*The destination coordinates should match the pattern index.*/
-                if (mach->accel.color_pattern_idx != mach->accel.patt_idx)
-                    mach->accel.color_pattern_idx = mach->accel.patt_idx;
             }
 
             if (mach->accel.dy_end == mach->accel.dy_start) {
@@ -1016,24 +1003,22 @@ skip_dx:
             while (count--) {
                 switch (mono_src) {
                     case 0:
-                        mix = 1;
+                        mix = 0x01;
                         break;
                     case 1:
-                        if (mach->accel.mono_pattern_enable)
+                        if (mach->accel.mono_pattern_enable || mach->accel.block_write_mono_pattern_enable)
                             mix = mach->accel.mono_pattern[dev->accel.dy & 7][dev->accel.dx & 7];
+                        else if ((dev->accel_bpp == 24) && (frgd_sel == 5))
+                            mix = 0x01;
                         else {
-                            if ((dev->accel_bpp == 24) && (frgd_sel == 5) && (mach->accel.patt_len_reg & 0x4000))
-                                mix = 1;
-                            else {
-                                if (!dev->accel.temp_cnt) {
-                                    dev->accel.temp_cnt = 8;
-                                    mix_dat >>= 8;
-                                }
-                                mix = (mix_dat & 0x80);
-                                dev->accel.temp_cnt--;
-                                mix_dat <<= 1;
-                                mix_dat |= 1;
+                            if (!dev->accel.temp_cnt) {
+                                dev->accel.temp_cnt = 8;
+                                mix_dat >>= 8;
                             }
+                            mix = !!(mix_dat & 0x80);
+                            dev->accel.temp_cnt--;
+                            mix_dat <<= 1;
+                            mix_dat |= 1;
                         }
                         break;
                     case 2:
@@ -1042,10 +1027,10 @@ skip_dx:
                             mix_dat <<= 1;
                         } else {
                             if (mach->accel.dp_config & 0x200) {
-                                mix = mix_dat & 1;
+                                mix = mix_dat & 0x01;
                                 mix_dat >>= 1;
                             } else {
-                                mix = mix_dat & 0x80;
+                                mix = !!(mix_dat & 0x80);
                                 mix_dat <<= 1;
                                 mix_dat |= 1;
                             }
@@ -1060,15 +1045,18 @@ skip_dx:
                         break;
                 }
 
-                if (((dev->accel.dx) >= clip_l) && ((dev->accel.dx) <= clip_r) && ((dev->accel.dy) >= clip_t) && ((dev->accel.dy) <= clip_b)) {
-                    if ((mach->accel.dp_config & 0x02) || (mach->accel.linedraw_opt & 0x02)) {
+                if ((dev->accel.dx >= clip_l) &&
+                    (dev->accel.dx <= clip_r) &&
+                    (dev->accel.dy >= clip_t) &&
+                    (dev->accel.dy <= clip_b)) {
+                    if (mach->accel.dp_config & 0x02) {
                         READ(dev->accel.src + dev->accel.cx, poly_src);
                         poly_src = ((poly_src & rd_mask) == rd_mask);
                         if (poly_src)
                             mach->accel.poly_fill ^= 1;
                     }
 
-                    if (mach->accel.poly_fill || !(mach->accel.dp_config & 0x02) || !(mach->accel.linedraw_opt & 0x02)) {
+                    if (mach->accel.poly_fill || !(mach->accel.dp_config & 0x02)) {
                         switch (mix ? frgd_sel : bkgd_sel) {
                             case 0:
                                 src_dat = dev->accel.bkgd_color;
@@ -1089,22 +1077,18 @@ skip_dx:
                                 }
                                 break;
                             case 5:
-                                if (mix) {
-                                    if (dev->bpp)
-                                        src_dat = mach->accel.color_pattern_word[mach->accel.color_pattern_idx];
-                                    else
-                                        src_dat = mach->accel.color_pattern[mach->accel.color_pattern_idx];
-                                } else
-                                    src_dat = 0;
+                                src_dat = mach->accel.color_pattern[mach->accel.color_pattern_idx];
+                                if (dev->bpp)
+                                    src_dat |= (mach->accel.color_pattern[mach->accel.color_pattern_idx + 1] << 8);
                                 break;
 
                             default:
                                 break;
                         }
 
-                        if ((dev->accel_bpp == 24) && (mono_src == 1) && (frgd_sel == 5) && (mach->accel.patt_len_reg & 0x4000)) {
+                        if ((dev->accel_bpp == 24) && (mono_src == 1) && (frgd_sel == 5) && !mach->accel.mono_pattern_enable) {
                             if (dev->accel.sy & 1) {
-                                READ(dev->accel.dest + dev->accel.dx - dev->ext_pitch, dest_dat);
+                                READ(dev->accel.dest + dev->accel.dx - dev->pitch, dest_dat);
                             } else {
                                 READ(dev->accel.dest + dev->accel.dx, dest_dat);
                             }
@@ -1148,14 +1132,20 @@ skip_dx:
                         }
 
                         if (mach->accel.dp_config & 0x10) {
-                            if ((dev->accel_bpp == 24) && (mono_src == 1) && (frgd_sel == 5) && (mach->accel.patt_len_reg & 0x4000)) {
-                                if (dev->accel.sy & 1) {
-                                    WRITE(dev->accel.dest + dev->accel.dx - dev->ext_pitch, dest_dat);
-                                } else {
+                            if (mach->accel.block_write_mono_pattern_enable) {
+                                if (mix) {
                                     WRITE(dev->accel.dest + dev->accel.dx, dest_dat);
                                 }
                             } else {
-                                WRITE(dev->accel.dest + dev->accel.dx, dest_dat);
+                                if ((dev->accel_bpp == 24) && (mono_src == 1) && (frgd_sel == 5) && !mach->accel.mono_pattern_enable) {
+                                    if (dev->accel.sy & 1) {
+                                        WRITE(dev->accel.dest + dev->accel.dx - dev->pitch, dest_dat);
+                                    } else {
+                                        WRITE(dev->accel.dest + dev->accel.dx, dest_dat);
+                                    }
+                                } else {
+                                    WRITE(dev->accel.dest + dev->accel.dx, dest_dat);
+                                }
                             }
                         }
                     }
@@ -1166,54 +1156,45 @@ skip_dx:
                 else
                     cpu_dat >>= 8;
 
+                if (mach->accel.dp_config == 0x2071 || (mach->accel.dp_config == 0x2011))
+                    mach_log("FontBlit: SX=%d, C(%d,%d), SRCWidth=%d, frgdmix=%d, bkgdmix=%d, rdmask=%04x, D(%d,%d), geoffset=%x, addr=%08x,.\n", mach->accel.sx, dev->accel.cx, dev->accel.cy, mach->accel.src_width, dev->accel.frgd_mix & 0x1f, dev->accel.bkgd_mix & 0x1f, rd_mask, dev->accel.dx, dev->accel.dy, dev->accel.ge_offset, (dev->accel.src + dev->accel.cx) & dev->vram_mask);
+
                 if ((mono_src == 3) || (frgd_sel == 3) || (bkgd_sel == 3) || (mach->accel.dp_config & 0x02)) {
                     dev->accel.cx += mach->accel.src_stepx;
                     mach->accel.sx++;
                     if (mach->accel.sx >= mach->accel.src_width) {
                         mach->accel.sx = 0;
-                        if (mach->accel.src_stepx == -1)
+                        if (mach->accel.src_stepx < 0)
                             dev->accel.cx += mach->accel.src_width;
                         else
                             dev->accel.cx -= mach->accel.src_width;
 
                         dev->accel.cy += (mach->accel.src_y_dir ? 1 : -1);
                         if (dev->bpp)
-                            dev->accel.src = (mach->accel.ge_offset << 1) + (dev->accel.cy * (dev->pitch));
+                            dev->accel.src = (mach->accel.ge_offset << 1) + (dev->accel.cy * dev->pitch);
                         else
-                            dev->accel.src = (mach->accel.ge_offset << 2) + (dev->accel.cy * (dev->pitch));
+                            dev->accel.src = (mach->accel.ge_offset << 2) + (dev->accel.cy * dev->pitch);
                     }
                 }
 
-                dev->accel.dx += mach->accel.stepx;
-
-                if ((dev->accel_bpp == 8) || ((dev->accel_bpp == 24) && (mach->accel.patt_len >= 3) && (mono_src != 1)))
-                    mach->accel.color_pattern_idx = (mach->accel.color_pattern_idx + mach->accel.stepx) & mach->accel.patt_len;
-
-                if ((dev->accel_bpp == 24) && (mach->accel.color_pattern_idx == mach->accel.patt_len) && (mach->accel.patt_len >= 3) && (mono_src != 1)) {
-                    mach->accel.color_pattern_idx = mach->accel.patt_data_idx;
-                } else if ((dev->accel_bpp == 24) && (mach->accel.patt_len < 3)) {
-                    if (mach->accel.patt_len == 2) {
-                        mach->accel.color_pattern_idx++;
-                        if (mach->accel.color_pattern_idx == 3)
-                            mach->accel.color_pattern_idx = 0;
-                    } else
-                        mach->accel.color_pattern_idx = (mach->accel.color_pattern_idx + mach->accel.stepx) & mach->accel.patt_len;
-
-                } else if ((dev->accel_bpp == 24) && (mach->accel.patt_len_reg & 0x4000) && (frgd_sel == 5)) {
+                if (dev->bpp)
+                    mach->accel.color_pattern_idx += 2;
+                else
                     mach->accel.color_pattern_idx++;
-                    if (mach->accel.color_pattern_idx == 3)
+
+                if ((mono_src == 1) && !mach->accel.mono_pattern_enable && !mach->accel.block_write_mono_pattern_enable && (frgd_sel == 5) && (dev->accel_bpp == 24)) {
+                    if (mach->accel.color_pattern_idx > 2)
+                        mach->accel.color_pattern_idx = 0;
+                } else {
+                    if (mach->accel.color_pattern_idx > mach->accel.patt_len)
                         mach->accel.color_pattern_idx = 0;
                 }
 
-                if (dev->bpp) {
-                    mach->accel.color_pattern_idx = (mach->accel.color_pattern_idx + mach->accel.stepx) & mach->accel.patt_len;
-                    mach->accel.color_pattern_idx = (mach->accel.color_pattern_idx + mach->accel.stepx) & mach->accel.patt_len;
-                }
-
+                dev->accel.dx += mach->accel.stepx;
                 dev->accel.sx++;
                 if ((dev->accel.sx >= mach->accel.width) || (dev->accel.dx >= 0x600)) {
                     dev->accel.sx         = 0;
-                    if (mach->accel.stepx == -1)
+                    if (mach->accel.stepx < 0)
                         dev->accel.dx += mach->accel.width;
                     else
                         dev->accel.dx -= mach->accel.width;
@@ -1222,27 +1203,19 @@ skip_dx:
                     dev->accel.sy++;
 
                     mach->accel.poly_fill = 0;
-                    if (dev->bpp)
-                        dev->accel.dest = (mach->accel.ge_offset << 1) + (dev->accel.dy * (dev->pitch));
-                    else
-                        dev->accel.dest = (mach->accel.ge_offset << 2) + (dev->accel.dy * (dev->pitch));
-
-                    if ((mono_src == 1) && (dev->accel_bpp == 24) && (frgd_sel == 5))
-                        mach->accel.color_pattern_idx = 0;
-                    else
-                        mach->accel.color_pattern_idx = ((dev->accel.dx + (dev->accel.dy << 3)) & mach->accel.patt_len);
-
-                    if ((dev->accel_bpp == 24) && (mach->accel.color_pattern_idx == mach->accel.patt_len) && (mono_src != 1))
-                        mach->accel.color_pattern_idx = 0;
-                    if ((mono_src == 1) && !mach->accel.mono_pattern_enable && !(mach->accel.patt_len_reg & 0x4000)) {
-                        dev->accel.cur_x = dev->accel.dx;
-                        dev->accel.cur_y = dev->accel.dy;
-                        return;
+                    if (mach->accel.dp_config & 0x02)
+                        dev->accel.dest = (dev->accel.dy * dev->pitch);
+                    else {
+                        if (dev->bpp)
+                            dev->accel.dest = (mach->accel.ge_offset << 1) + (dev->accel.dy * dev->pitch);
+                        else
+                            dev->accel.dest = (mach->accel.ge_offset << 2) + (dev->accel.dy * dev->pitch);
                     }
+
                     if (dev->accel.sy >= mach->accel.height) {
-                        if ((mono_src == 2) || (mono_src == 3) || (frgd_sel == 3) || (bkgd_sel == 3) || (mach->accel.dp_config & 0x02) || (mach->accel.linedraw_opt & 0x02))
+                        if ((mono_src == 2) || (mono_src == 3) || (frgd_sel == 3) || (bkgd_sel == 3) || (mach->accel.dp_config & 0x02))
                             return;
-                        if ((mono_src == 1) && (frgd_sel == 5) && (dev->accel_bpp == 24) && (mach->accel.patt_len_reg & 0x4000))
+                        if ((mono_src == 1) && (frgd_sel == 5) &&  (dev->accel_bpp == 24))
                             return;
                         dev->accel.cur_x = dev->accel.dx;
                         dev->accel.cur_y = dev->accel.dy;
@@ -1290,18 +1263,12 @@ skip_dx:
                 }
             }
 
-            if (frgd_sel == 5) {
-                for (int x = 0; x <= mach->accel.patt_len; x++) {
-                    mach->accel.color_pattern[x] = mach->accel.patt_data[x & mach->accel.patt_len];
-                }
-            }
-
             if (mono_src == 1) {
-                mix_dat             = mach->accel.patt_data[0x10];
+                mix_dat = mach->accel.mono_pattern_normal[0];
                 dev->accel.temp_cnt = 8;
             }
 
-            count             = (dev->accel.dx > dev->accel.dy) ? (dev->accel.dx >> 1) : (dev->accel.dy >> 1);
+            count = (dev->accel.dx > dev->accel.dy) ? (dev->accel.dx >> 1) : (dev->accel.dy >> 1);
             mach->accel.width = count;
 
             if (dev->accel.dx > dev->accel.dy) {
@@ -1317,7 +1284,10 @@ skip_dx:
                         mix_dat <<= 1;
                         mix_dat |= 1;
 
-                        if ((dev->accel.cx >= clip_l) && (dev->accel.cx <= clip_r) && (dev->accel.cy >= clip_t) && (dev->accel.cy <= clip_b)) {
+                        if ((dev->accel.cx >= clip_l) &&
+                            (dev->accel.cx <= clip_r) &&
+                            (dev->accel.cy >= clip_t) &&
+                            (dev->accel.cy <= clip_b)) {
                             mach->accel.clip_overrun = 0;
                             switch (mix ? frgd_sel : bkgd_sel) {
                                 case 0:
@@ -1336,10 +1306,9 @@ skip_dx:
                                         src_dat = 0;
                                     break;
                                 case 5:
-                                    if (mix)
-                                        src_dat = mach->accel.color_pattern[((dev->accel.cx) + ((dev->accel.cy) << 3)) & mach->accel.patt_len];
-                                    else
-                                        src_dat = 0;
+                                    src_dat = mach->accel.color_pattern[mach->accel.color_pattern_idx];
+                                    if (dev->bpp)
+                                        src_dat |= (mach->accel.color_pattern[mach->accel.color_pattern_idx + 1] << 8);
                                     break;
 
                                 default:
@@ -1404,6 +1373,14 @@ skip_dx:
                         else
                             cpu_dat >>= 8;
 
+                        if (dev->bpp)
+                            mach->accel.color_pattern_idx += 2;
+                        else
+                            mach->accel.color_pattern_idx++;
+
+                        if (mach->accel.color_pattern_idx > mach->accel.patt_len)
+                            mach->accel.color_pattern_idx = 0;
+
                         if (mach->accel.err >= 0) {
                             dev->accel.cy += mach->accel.stepy;
                             mach->accel.err -= dev->accel.dx;
@@ -1438,12 +1415,19 @@ skip_dx:
                                 break;
                         }
 
-                        if ((dev->accel.cx >= clip_l) && (dev->accel.cx <= clip_r) && (dev->accel.cy >= clip_t) && (dev->accel.cy <= clip_b)) {
+                        if ((dev->accel.cx >= clip_l) &&
+                            (dev->accel.cx <= clip_r) &&
+                            (dev->accel.cy >= clip_t) &&
+                            (dev->accel.cy <= clip_b)) {
                             mach->accel.clip_overrun = 0;
                             if (mach->accel.linedraw_opt & 0x02) {
-                                READ((mach->accel.ge_offset << 2) + ((dev->accel.cy) * (dev->pitch)) + (dev->accel.cx), poly_src);
+                                if (dev->bpp) {
+                                    READ((mach->accel.ge_offset << 1) + ((dev->accel.cy) * (dev->pitch)) + (dev->accel.cx), poly_src);
+                                } else {
+                                    READ((mach->accel.ge_offset << 2) + ((dev->accel.cy) * (dev->pitch)) + (dev->accel.cx), poly_src);
+                                }
                                 if (poly_src)
-                                    mach->accel.poly_fill = !mach->accel.poly_fill;
+                                    mach->accel.poly_fill ^= 1;
                             }
 
                             switch (mix ? frgd_sel : bkgd_sel) {
@@ -1464,20 +1448,23 @@ skip_dx:
                                     }
                                     break;
                                 case 5:
-                                    if (mix) {
-                                        src_dat = mach->accel.color_pattern[((dev->accel.cx) + ((dev->accel.cy) << 3)) & mach->accel.patt_len];
-                                    } else
-                                        src_dat = 0;
+                                    src_dat = mach->accel.color_pattern[mach->accel.color_pattern_idx];
+                                    if (dev->bpp)
+                                        src_dat |= (mach->accel.color_pattern[mach->accel.color_pattern_idx + 1] << 8);
                                     break;
 
                                 default:
                                     break;
                             }
 
-                            if (dev->bpp) {
-                                READ((mach->accel.ge_offset << 1) + ((dev->accel.cy) * (dev->pitch)) + (dev->accel.cx), dest_dat);
+                            if (mach->accel.linedraw_opt & 0x02) {
+                                if (dev->bpp) {
+                                    READ((mach->accel.ge_offset << 1) + ((dev->accel.cy) * (dev->pitch)) + (dev->accel.cx), dest_dat);
+                                } else {
+                                    READ((mach->accel.ge_offset << 2) + ((dev->accel.cy) * (dev->pitch)) + (dev->accel.cx), dest_dat);
+                                }
                             } else {
-                                READ((mach->accel.ge_offset << 2) + ((dev->accel.cy) * (dev->pitch)) + (dev->accel.cx), dest_dat);
+                                READ(((dev->accel.cy) * (dev->pitch)) + (dev->accel.cx), dest_dat);
                             }
 
                             switch (compare_mode) {
@@ -1520,17 +1507,25 @@ skip_dx:
                             if ((mach->accel.dp_config & 0x10) && (cmd_type == 3)) {
                                 if (mach->accel.linedraw_opt & 0x04) {
                                     if (dev->accel.sx < mach->accel.width) {
+                                        if (mach->accel.linedraw_opt & 0x02) {
+                                            if (dev->bpp) {
+                                                WRITE((mach->accel.ge_offset << 1) + ((dev->accel.cy) * (dev->pitch)) + (dev->accel.cx), dest_dat);
+                                            } else {
+                                                WRITE((mach->accel.ge_offset << 2) + ((dev->accel.cy) * (dev->pitch)) + (dev->accel.cx), dest_dat);
+                                            }
+                                        } else {
+                                            WRITE(((dev->accel.cy) * (dev->pitch)) + (dev->accel.cx), dest_dat);
+                                        }
+                                    }
+                                } else {
+                                    if (mach->accel.linedraw_opt & 0x02) {
                                         if (dev->bpp) {
                                             WRITE((mach->accel.ge_offset << 1) + ((dev->accel.cy) * (dev->pitch)) + (dev->accel.cx), dest_dat);
                                         } else {
                                             WRITE((mach->accel.ge_offset << 2) + ((dev->accel.cy) * (dev->pitch)) + (dev->accel.cx), dest_dat);
                                         }
-                                    }
-                                } else {
-                                    if (dev->bpp) {
-                                        WRITE((mach->accel.ge_offset << 1) + ((dev->accel.cy) * (dev->pitch)) + (dev->accel.cx), dest_dat);
                                     } else {
-                                        WRITE((mach->accel.ge_offset << 2) + ((dev->accel.cy) * (dev->pitch)) + (dev->accel.cx), dest_dat);
+                                        WRITE(((dev->accel.cy) * (dev->pitch)) + (dev->accel.cx), dest_dat);
                                     }
                                 }
                             }
@@ -1545,6 +1540,14 @@ skip_dx:
                         else
                             cpu_dat >>= 8;
 
+                        if (dev->bpp)
+                            mach->accel.color_pattern_idx += 2;
+                        else
+                            mach->accel.color_pattern_idx++;
+
+                        if (mach->accel.color_pattern_idx > mach->accel.patt_len)
+                            mach->accel.color_pattern_idx = 0;
+
                         if (mach->accel.err >= 0) {
                             dev->accel.cy += mach->accel.stepy;
                             mach->accel.err -= dev->accel.dx;
@@ -1568,7 +1571,10 @@ skip_dx:
                         mix_dat <<= 1;
                         mix_dat |= 1;
 
-                        if ((dev->accel.cx >= clip_l) && (dev->accel.cx <= clip_r) && (dev->accel.cy >= clip_t) && (dev->accel.cy <= clip_b)) {
+                        if ((dev->accel.cx >= clip_l) &&
+                            (dev->accel.cx <= clip_r) &&
+                            (dev->accel.cy >= clip_t) &&
+                            (dev->accel.cy <= clip_b)) {
                             mach->accel.clip_overrun = 0;
                             switch (mix ? frgd_sel : bkgd_sel) {
                                 case 0:
@@ -1588,10 +1594,9 @@ skip_dx:
                                     }
                                     break;
                                 case 5:
-                                    if (mix) {
-                                        src_dat = mach->accel.color_pattern[((dev->accel.cx) + ((dev->accel.cy) << 3)) & mach->accel.patt_len];
-                                    } else
-                                        src_dat = 0;
+                                    src_dat = mach->accel.color_pattern[mach->accel.color_pattern_idx];
+                                    if (dev->bpp)
+                                        src_dat |= (mach->accel.color_pattern[mach->accel.color_pattern_idx + 1] << 8);
                                     break;
 
                                 default:
@@ -1656,6 +1661,14 @@ skip_dx:
                         else
                             cpu_dat >>= 8;
 
+                        if (dev->bpp)
+                            mach->accel.color_pattern_idx += 2;
+                        else
+                            mach->accel.color_pattern_idx++;
+
+                        if (mach->accel.color_pattern_idx > mach->accel.patt_len)
+                            mach->accel.color_pattern_idx = 0;
+
                         if (mach->accel.err >= 0) {
                             dev->accel.cx += mach->accel.stepx;
                             mach->accel.err -= dev->accel.dy;
@@ -1690,9 +1703,11 @@ skip_dx:
                                 break;
                         }
 
-                        if ((dev->accel.cx >= clip_l) && (dev->accel.cx <= clip_r) && (dev->accel.cy >= clip_t) && (dev->accel.cy <= clip_b)) {
+                        if ((dev->accel.cx >= clip_l) &&
+                            (dev->accel.cx <= clip_r) &&
+                            (dev->accel.cy >= clip_t) &&
+                            (dev->accel.cy <= clip_b)) {
                             mach->accel.clip_overrun = 0;
-
                             switch (mix ? frgd_sel : bkgd_sel) {
                                 case 0:
                                     src_dat = dev->accel.bkgd_color;
@@ -1711,20 +1726,23 @@ skip_dx:
                                     }
                                     break;
                                 case 5:
-                                    if (mix) {
-                                        src_dat = mach->accel.color_pattern[((dev->accel.cx) + ((dev->accel.cy) << 3)) & mach->accel.patt_len];
-                                    } else
-                                        src_dat = 0;
+                                    src_dat = mach->accel.color_pattern[mach->accel.color_pattern_idx];
+                                    if (dev->bpp)
+                                        src_dat |= (mach->accel.color_pattern[mach->accel.color_pattern_idx + 1] << 8);
                                     break;
 
                                 default:
                                     break;
                             }
 
-                            if (dev->bpp) {
-                                READ((mach->accel.ge_offset << 1) + ((dev->accel.cy) * (dev->pitch)) + (dev->accel.cx), dest_dat);
+                            if (mach->accel.linedraw_opt & 0x02) {
+                                if (dev->bpp) {
+                                    READ((mach->accel.ge_offset << 1) + ((dev->accel.cy) * (dev->pitch)) + (dev->accel.cx), dest_dat);
+                                } else {
+                                    READ((mach->accel.ge_offset << 2) + ((dev->accel.cy) * (dev->pitch)) + (dev->accel.cx), dest_dat);
+                                }
                             } else {
-                                READ((mach->accel.ge_offset << 2) + ((dev->accel.cy) * (dev->pitch)) + (dev->accel.cx), dest_dat);
+                                READ(((dev->accel.cy) * (dev->pitch)) + (dev->accel.cx), dest_dat);
                             }
 
                             switch (compare_mode) {
@@ -1765,17 +1783,25 @@ skip_dx:
                             if ((mach->accel.dp_config & 0x10) && (cmd_type == 3)) {
                                 if (mach->accel.linedraw_opt & 0x04) {
                                     if (dev->accel.sx < mach->accel.width) {
+                                        if (mach->accel.linedraw_opt & 0x02) {
+                                            if (dev->bpp) {
+                                                WRITE((mach->accel.ge_offset << 1) + ((dev->accel.cy) * (dev->pitch)) + (dev->accel.cx), dest_dat);
+                                            } else {
+                                                WRITE((mach->accel.ge_offset << 2) + ((dev->accel.cy) * (dev->pitch)) + (dev->accel.cx), dest_dat);
+                                            }
+                                        } else {
+                                            WRITE((dev->accel.cy * dev->pitch) + dev->accel.cx, dest_dat);
+                                        }
+                                    }
+                                } else {
+                                    if (mach->accel.linedraw_opt & 0x02) {
                                         if (dev->bpp) {
                                             WRITE((mach->accel.ge_offset << 1) + ((dev->accel.cy) * (dev->pitch)) + (dev->accel.cx), dest_dat);
                                         } else {
                                             WRITE((mach->accel.ge_offset << 2) + ((dev->accel.cy) * (dev->pitch)) + (dev->accel.cx), dest_dat);
                                         }
-                                    }
-                                } else {
-                                    if (dev->bpp) {
-                                        WRITE((mach->accel.ge_offset << 1) + ((dev->accel.cy) * (dev->pitch)) + (dev->accel.cx), dest_dat);
                                     } else {
-                                        WRITE((mach->accel.ge_offset << 2) + ((dev->accel.cy) * (dev->pitch)) + (dev->accel.cx), dest_dat);
+                                        WRITE(((dev->accel.cy) * (dev->pitch)) + (dev->accel.cx), dest_dat);
                                     }
                                 }
                             }
@@ -1790,6 +1816,14 @@ skip_dx:
                         else
                             cpu_dat >>= 8;
 
+                        if (dev->bpp)
+                            mach->accel.color_pattern_idx += 2;
+                        else
+                            mach->accel.color_pattern_idx++;
+
+                        if (mach->accel.color_pattern_idx > mach->accel.patt_len)
+                            mach->accel.color_pattern_idx = 0;
+
                         if (mach->accel.err >= 0) {
                             dev->accel.cx += mach->accel.stepx;
                             mach->accel.err -= dev->accel.dy;
@@ -1842,8 +1876,6 @@ skip_dx:
                 }
 
                 dev->accel.sx = 0;
-                if ((dev->accel_bpp == 24) && (mach->accel.patt_len < 0x17))
-                    mach->accel.color_pattern_idx = 0;
 
                 /*Step Y*/
                 mach->accel.dy_start = dev->accel.cur_y;
@@ -1853,13 +1885,12 @@ skip_dx:
                 if (mach->accel.dest_y_end >= 0x600)
                     mach->accel.dy_end |= ~0x5ff;
 
-                if (mach->accel.dy_end > mach->accel.dy_start) {
+                if (mach->accel.dy_end > mach->accel.dy_start)
                     mach->accel.stepy = 1;
-                } else if (mach->accel.dy_end < mach->accel.dy_start) {
+                else if (mach->accel.dy_end < mach->accel.dy_start)
                     mach->accel.stepy = -1;
-                } else {
+                else
                     mach->accel.stepy = 0;
-                }
 
                 if (dev->bpp)
                     dev->accel.dest = (mach->accel.ge_offset << 1) + (dev->accel.dy * (dev->pitch));
@@ -1903,22 +1934,7 @@ skip_dx:
                 else
                     dev->accel.src = (mach->accel.ge_offset << 2) + (dev->accel.cy * (dev->pitch));
 
-                if ((dev->accel_bpp == 24) && (frgd_sel == 5)) {
-                    if (mach->accel.patt_len == 0x17)
-                        mach->accel.color_pattern_idx = 0;
-                    dev->accel.x1 = dev->accel.dx + mach->accel.width;
-                    if (dev->accel.x1 == dev->pitch) {
-                        dev->accel.x2 = mach->accel.width & 1;
-                    } else if ((dev->accel.x1 == mach->accel.width) && (dev->accel.dy & 1) && !dev->accel.y1 && dev->accel.x2) {
-                        if (mach->accel.patt_len == 0x17)
-                            mach->accel.color_pattern_idx = 3;
-                        dev->accel.x3 = 1;
-                    } else
-                        dev->accel.x3 = 0;
-                } else
-                    mach_log("ScanToX=%04x, Pitch=%d, C(%d,%d), SRCWidth=%d, WH(%d,%d), geoffset=%08x.\n", mach->accel.dp_config, dev->ext_pitch, dev->accel.cx, dev->accel.cy, mach->accel.src_width, mach->accel.width, mach->accel.height, (mach->accel.ge_offset << 1));
-
-                dev->accel.y1 = 0;
+                mach_log("ScanToX=%04x, MonoSRC=%d, FrgdSel=%d, BkgdSel=%d, Pitch=%d, C(%d,%d), SRCWidth=%d, WH(%d,%d), colorpattidx=%d, pattlen=%d.\n", mach->accel.dp_config, mono_src, frgd_sel, bkgd_sel, dev->ext_pitch, dev->accel.cx, dev->accel.cy, mach->accel.src_width, mach->accel.width, mach->accel.height, mach->accel.color_pattern_idx, mach->accel.patt_len);
 
                 if ((mono_src == 2) || (bkgd_sel == 2) || (frgd_sel == 2) || mach_pixel_read(mach)) {
                     if (mach_pixel_write(mach)) {
@@ -1935,30 +1951,10 @@ skip_dx:
 
             if (mono_src == 1) {
                 count               = mach->accel.width;
-                mix_dat             = mach->accel.patt_data[0x10];
+                mix_dat             = mach->accel.mono_pattern_normal[0];
                 dev->accel.temp_cnt = 8;
             }
 
-            if (frgd_sel == 5) {
-                if (dev->accel_bpp != 24) {
-                    for (int x = 0; x <= mach->accel.patt_len; x++) {
-                        mach->accel.color_pattern[x] = mach->accel.patt_data[x & mach->accel.patt_len];
-                    }
-                } else {
-                    if (mach->accel.patt_len == 0x17) {
-                        for (int x = 0; x <= mach->accel.patt_len; x++) {
-                            mach->accel.color_pattern_full[x] = mach->accel.patt_data[x];
-                            mach_log("ScanToX: Color Pattern 24bpp[%d]=%02x, dataidx=%d, pattlen=%d.\n", x, mach->accel.color_pattern_full[x], mach->accel.patt_data_idx, mach->accel.patt_len);
-                        }
-                    } else {
-                        for (int x = 0; x <= mach->accel.patt_len; x++) {
-                            mach->accel.color_pattern[x] = mach->accel.patt_data[x];
-                            mach_log("ScanToX: Color Pattern 24bpp[%d]=%02x, dataidx=%d, pattlen=%d.\n", x, mach->accel.color_pattern[x], mach->accel.patt_data_idx, mach->accel.patt_len);
-                        }
-                    }
-                }
-            }
-
             while (count--) {
                 switch (mono_src) {
                     case 0:
@@ -1998,7 +1994,10 @@ skip_dx:
                         break;
                 }
 
-                if (((dev->accel.dx) >= clip_l) && ((dev->accel.dx) <= clip_r) && ((dev->accel.dy) >= clip_t) && ((dev->accel.dy) <= clip_b)) {
+                if ((dev->accel.dx >= clip_l) &&
+                    (dev->accel.dx <= clip_r) &&
+                    (dev->accel.dy >= clip_t) &&
+                    (dev->accel.dy <= clip_b)) {
                     switch (mix ? frgd_sel : bkgd_sel) {
                         case 0:
                             src_dat = dev->accel.bkgd_color;
@@ -2019,16 +2018,9 @@ skip_dx:
                             }
                             break;
                         case 5:
-                            if (mix) {
-                                if (dev->accel_bpp == 24) {
-                                    if (mach->accel.patt_len == 0x17)
-                                        src_dat = mach->accel.color_pattern_full[mach->accel.color_pattern_idx];
-                                    else
-                                        src_dat = mach->accel.color_pattern[mach->accel.color_pattern_idx];
-                                } else
-                                    src_dat = mach->accel.color_pattern[(dev->accel.dx + (dev->accel.dy << 3)) & mach->accel.patt_len];
-                            } else
-                                src_dat = 0;
+                            src_dat = mach->accel.color_pattern[mach->accel.color_pattern_idx];
+                            if (dev->bpp)
+                                src_dat |= (mach->accel.color_pattern[mach->accel.color_pattern_idx + 1] << 8);
                             break;
 
                         default:
@@ -2073,7 +2065,7 @@ skip_dx:
                     }
 
                     if (mach->accel.dp_config & 0x10) {
-                        WRITE(dev->accel.dest + (dev->accel.dx), dest_dat);
+                        WRITE(dev->accel.dest + dev->accel.dx, dest_dat);
                     }
                 }
 
@@ -2097,23 +2089,15 @@ skip_dx:
                         dev->accel.src = (mach->accel.ge_offset << 2) + (dev->accel.cy * (dev->pitch));
                 }
 
-                dev->accel.dx += mach->accel.stepx;
-                if ((dev->accel_bpp == 24) && (mach->accel.patt_len == 0x17)) {
+                if (dev->bpp)
+                    mach->accel.color_pattern_idx += 2;
+                else
                     mach->accel.color_pattern_idx++;
-                    if (dev->accel.x3) {
-                        if (mach->accel.color_pattern_idx == 9)
-                            mach->accel.color_pattern_idx = 3;
-                    } else {
-                        if (mach->accel.color_pattern_idx == 6)
-                            mach->accel.color_pattern_idx = 0;
-                    }
-                } else if ((dev->accel_bpp == 24) && (mach->accel.patt_len < 3)) {
-                    mach->accel.color_pattern_idx++;
-                    if (mach->accel.color_pattern_idx == 3)
-                        mach->accel.color_pattern_idx = 0;
-                } else
-                    mach->accel.color_pattern_idx = (mach->accel.color_pattern_idx + mach->accel.stepx) & mach->accel.patt_len;
 
+                if (mach->accel.color_pattern_idx > mach->accel.patt_len)
+                    mach->accel.color_pattern_idx = 0;
+
+                dev->accel.dx += mach->accel.stepx;
                 dev->accel.sx++;
                 if (dev->accel.sx >= mach->accel.width) {
                     dev->accel.sx = 0;
@@ -2142,21 +2126,26 @@ mach_accel_out_pixtrans(mach_t *mach, ibm8514_t *dev, uint16_t val)
     int frgd_sel;
     int bkgd_sel;
     int mono_src;
+    int swap = 0;
 
     frgd_sel = (mach->accel.dp_config >> 13) & 7;
     bkgd_sel = (mach->accel.dp_config >> 7) & 3;
     mono_src = (mach->accel.dp_config >> 5) & 3;
 
-    if ((mach->accel.dp_config & 4) && (mach->accel.cmd_type != 5)) {
+    if ((mach->accel.dp_config & 0x04) && (mach->accel.cmd_type != 5)) {
+        mach_log("Read Host Monochrome Data.\n");
         val = (val >> 8) | (val << 8);
+        swap = 1;
     }
 
     switch (mach->accel.dp_config & 0x200) {
         case 0x000: /*8-bit size*/
             if (mono_src == 2) {
                 if ((frgd_sel != 2) && (bkgd_sel != 2)) {
-                    if (mach->accel.dp_config & 0x1000)
+                    if ((mach->accel.dp_config & 0x1000) && !swap) {
+                        mach_log("8-bit bus size swap.\n");
                         val = (val >> 8) | (val << 8);
+                    }
                     mach_accel_start(mach->accel.cmd_type, 1, 8, val | (val << 16), 0, mach, dev);
                 } else
                     mach_accel_start(mach->accel.cmd_type, 1, 1, -1, val | (val << 16), mach, dev);
@@ -2166,8 +2155,10 @@ mach_accel_out_pixtrans(mach_t *mach, ibm8514_t *dev, uint16_t val)
         case 0x200: /*16-bit size*/
             if (mono_src == 2) {
                 if ((frgd_sel != 2) && (bkgd_sel != 2)) {
-                    if (mach->accel.dp_config & 0x1000)
+                    if ((mach->accel.dp_config & 0x1000) && !swap) {
+                        mach_log("16-bit bus size swap.\n");
                         val = (val >> 8) | (val << 8);
+                    }
                     mach_accel_start(mach->accel.cmd_type, 1, 16, val | (val << 16), 0, mach, dev);
                 } else
                     mach_accel_start(mach->accel.cmd_type, 1, 2, -1, val | (val << 16), mach, dev);
@@ -2231,15 +2222,15 @@ mach_out(uint16_t addr, uint8_t val, void *priv)
                             mach->bank_r |= ((mach->regs[0xae] & 0x0c) << 2);
                             mach->bank_w |= ((mach->regs[0xae] & 3) << 4);
                         }
-                        if (dev->on[0] || dev->on[1])
+                        if (dev->on)
                             mach_log("Separate B2Bank = %02x, AEbank = %02x.\n", mach->regs[0xb2], mach->regs[0xae]);
                     } else { /* Single bank mode */
                         mach->bank_w = ((mach->regs[0xb2] & 0x1e) >> 1);
-                        if ((dev->local & 0xff) >= 0x02) {
+                        if ((dev->local & 0xff) >= 0x02)
                             mach->bank_w |= ((mach->regs[0xae] & 3) << 4);
-                        }
+
                         mach->bank_r = mach->bank_w;
-                        if (dev->on[0] || dev->on[1])
+                        if (dev->on)
                             mach_log("Single B2Bank = %02x, AEbank = %02x.\n", mach->regs[0xb2], mach->regs[0xae]);
                     }
                     svga->read_bank  = mach->bank_r << 16;
@@ -2356,7 +2347,7 @@ mach_in(uint16_t addr, void *priv)
     if (((addr & 0xFFF0) == 0x3D0 || (addr & 0xFFF0) == 0x3B0) && !(svga->miscout & 1))
         addr ^= 0x60;
 
-    if ((addr >= 0x3c6) && (addr <= 0x3c9) && (dev->on[0] || dev->on[1]))
+    if ((addr >= 0x3c6) && (addr <= 0x3c9) && dev->on)
         addr -= 0xdc;
 
     switch (addr) {
@@ -2457,8 +2448,8 @@ ati8514_recalctimings(svga_t *svga)
     mach_t       *mach = (mach_t *) svga->ext8514;
     ibm8514_t    *dev  = (ibm8514_t *) svga->dev8514;
 
-    mach_log("ON0=%d, ON1=%d, vgahdisp=%d.\n", dev->on[0], dev->on[1], svga->hdisp);
-    if (dev->on[0] || dev->on[1]) {
+    mach_log("ON=%d, vgahdisp=%d.\n", dev->on, svga->hdisp);
+    if (dev->on) {
         mach_log("8514/A ON.\n");
         dev->h_total                    = dev->htotal + 1;
         dev->rowcount                   = !!(dev->disp_cntl & 0x08);
@@ -2466,17 +2457,18 @@ ati8514_recalctimings(svga_t *svga)
         mach->accel.ge_offset           = dev->accel.ge_offset;
 
         mach_log("HDISP=%d, VDISP=%d, shadowset=%x, 8514/A mode=%x, clocksel=%02x.\n", dev->hdisp, dev->vdisp, mach->shadow_set & 0x03, dev->accel.advfunc_cntl & 0x04, mach->accel.clock_sel & 0xfe);
-        if (dev->accel.advfunc_cntl & 0x01) {
+
+        if (dev->hdisp) {
+            dev->h_disp = dev->hdisp;
+            dev->dispend = dev->vdisp;
+        } else {
             if (dev->accel.advfunc_cntl & 0x04) {
-                dev->h_disp = dev->hdisp;
-                dev->dispend = dev->vdisp;
+                dev->h_disp = 1024;
+                dev->dispend = 768;
             } else {
                 dev->h_disp = 640;
                 dev->dispend = 480;
             }
-        } else {
-            dev->h_disp = dev->hdisp;
-            dev->dispend = dev->vdisp;
         }
 
         if (dev->accel.advfunc_cntl & 0x04)
@@ -2500,6 +2492,16 @@ ati8514_recalctimings(svga_t *svga)
         dev->rowoffset = dev->ext_crt_pitch;
         mach_log("cntl=%d, hv(%d,%d), pitch=%d, rowoffset=%d, gextconfig=%03x, shadow=%x interlace=%d.\n", dev->accel.advfunc_cntl & 0x04, dev->h_disp, dev->dispend, dev->pitch, dev->rowoffset, mach->accel.ext_ge_config & 0xcec0, mach->shadow_set & 3, dev->interlace);
         svga->map8 = dev->pallook;
+        if (dev->vram_512k_8514) {
+            if (dev->h_disp == 640) {
+                dev->ext_pitch = 640;
+                dev->pitch = dev->ext_pitch;
+            } else {
+                dev->ext_pitch = 1024;
+                dev->pitch = dev->ext_pitch;
+            }
+        }
+        dev->accel_bpp = 8;
         svga->render8514 = ibm8514_render_8bpp;
     } else {
         if (!(svga->gdcreg[6] & 1) && !(svga->attrregs[0x10] & 1)) { /*Text mode*/
@@ -2567,17 +2569,39 @@ mach_recalctimings(svga_t *svga)
     }
 
     svga->render8514 = ibm8514_render_blank;
-    mach_log("ON[0]=%d, ON[1]=%d, exton[0]=%d, exton[1]=%d, vendormode0=%d, vendormode1=%d.\n", dev->on[0], dev->on[1], mach->ext_on[0], mach->ext_on[1], dev->vendor_mode[0], dev->vendor_mode[1]);
-    if (dev->on[0] || dev->on[1]) {
-        mach_log("8514/A ON.\n");
+    if (dev->on) {
+        mach_log("8514/A ON, extpitch=%d, devma=%x, vgamalatch=%x.\n", dev->ext_pitch, dev->ma, svga->ma_latch);
+        dev->pitch                      = dev->ext_pitch;
+        dev->rowoffset                  = dev->ext_crt_pitch;
         dev->h_total                    = dev->htotal + 1;
         dev->rowcount                   = !!(dev->disp_cntl & 0x08);
         dev->accel.ge_offset            = (mach->accel.ge_offset_lo | (mach->accel.ge_offset_hi << 16));
         mach->accel.ge_offset           = dev->accel.ge_offset;
 
         mach_log("HDISP=%d, VDISP=%d, shadowset=%x, 8514/A mode=%x, clocksel=%02x.\n", dev->hdisp, dev->vdisp, mach->shadow_set & 0x03, dev->accel.advfunc_cntl & 0x04, mach->accel.clock_sel & 0xfe);
-        if (!dev->bpp) {
-            if (dev->accel.advfunc_cntl & 0x01) {
+        if ((dev->local & 0xff) >= 0x02) {
+            if (dev->bpp || (dev->accel_bpp >= 24) || (mach->accel.clock_sel & 0x01)) {
+                dev->h_disp = dev->hdisp;
+                dev->dispend = dev->vdisp;
+            } else {
+                if (dev->interlace) { /*Interlaced displays are only for 800x600 and up.*/
+                    if (dev->accel.advfunc_cntl & 0x04) {
+                        dev->h_disp = dev->hdisp;
+                        dev->dispend = dev->vdisp;
+                    } else {
+                        dev->h_disp = 640;
+                        dev->dispend = 480;
+                    }
+                } else {
+                    dev->h_disp = dev->hdisp;
+                    dev->dispend = dev->vdisp;
+                }
+            }
+        } else {
+            if (mach->accel.clock_sel & 0x01) {
+                dev->h_disp = dev->hdisp;
+                dev->dispend = dev->vdisp;
+            } else {
                 if (dev->accel.advfunc_cntl & 0x04) {
                     dev->h_disp = dev->hdisp;
                     dev->dispend = dev->vdisp;
@@ -2585,13 +2609,7 @@ mach_recalctimings(svga_t *svga)
                     dev->h_disp = 640;
                     dev->dispend = 480;
                 }
-            } else {
-                dev->h_disp = dev->hdisp;
-                dev->dispend = dev->vdisp;
             }
-        } else {
-            dev->h_disp = dev->hdisp;
-            dev->dispend = dev->vdisp;
         }
 
         svga->clock8514 = (cpuclock * (double) (1ULL << 32)) / svga->getclock((mach->accel.clock_sel >> 2) & 0x0f, svga->clock_gen);
@@ -2614,9 +2632,7 @@ mach_recalctimings(svga_t *svga)
             dev->dispend += 2;
 
         if ((dev->local & 0xff) >= 0x02) {
-            mach_log("HDISP=%d.\n", dev->h_disp);
-            dev->pitch = dev->ext_pitch;
-            dev->rowoffset = dev->ext_crt_pitch;
+            mach_log("HDISP=%d, mask=%02x.\n", dev->h_disp, dev->dac_mask);
             if ((mach->accel.ext_ge_config & 0x800) || (!(mach->accel.ext_ge_config & 0x8000) && !(mach->accel.ext_ge_config & 0x800))) {
                 if ((mach->accel.ext_ge_config & 0x30) == 0x20) {
                     if ((mach->accel.ext_ge_config & 0xc0) == 0x40)
@@ -2628,9 +2644,20 @@ mach_recalctimings(svga_t *svga)
                         dev->accel_bpp = 32;
                     else
                         dev->accel_bpp = 24;
-                } else
+                } else if ((mach->accel.ext_ge_config & 0x30) == 0x10)
                     dev->accel_bpp = 8;
-
+                else {
+                    if (dev->vram_512k_8514) {
+                        if (dev->h_disp == 640) {
+                            dev->ext_pitch = 640;
+                            dev->pitch = dev->ext_pitch;
+                        } else {
+                            dev->ext_pitch = 1024;
+                            dev->pitch = dev->ext_pitch;
+                        }
+                    }
+                    dev->accel_bpp = 8;
+                }
                 mach_log("hv(%d,%d), pitch=%d, rowoffset=%d, gextconfig=%03x, bpp=%d, shadow=%x, vgahdisp=%d.\n", dev->h_disp, dev->dispend, dev->pitch, dev->ext_crt_pitch, mach->accel.ext_ge_config & 0xcec0, dev->accel_bpp, mach->shadow_set & 3, svga->hdisp);
                 switch (dev->accel_bpp) {
                     case 8:
@@ -2662,10 +2689,18 @@ mach_recalctimings(svga_t *svga)
                 }
             }
         } else {
-            dev->pitch = dev->ext_pitch;
-            dev->rowoffset = dev->ext_crt_pitch;
             mach_log("cntl=%d, hv(%d,%d), pitch=%d, rowoffset=%d, gextconfig=%03x, shadow=%x interlace=%d.\n", dev->accel.advfunc_cntl & 0x04, dev->h_disp, dev->dispend, dev->pitch, dev->rowoffset, mach->accel.ext_ge_config & 0xcec0, mach->shadow_set & 3, dev->interlace);
             svga->map8 = dev->pallook;
+            if (dev->vram_512k_8514) {
+                if (dev->h_disp == 640) {
+                    dev->ext_pitch = 640;
+                    dev->pitch = dev->ext_pitch;
+                } else {
+                    dev->ext_pitch = 1024;
+                    dev->pitch = dev->ext_pitch;
+                }
+            }
+            dev->accel_bpp = 8;
             svga->render8514 = ibm8514_render_8bpp;
         }
     }
@@ -2744,155 +2779,63 @@ mach_accel_out_fifo(mach_t *mach, svga_t *svga, ibm8514_t *dev, uint16_t port, u
 
     switch (port) {
         case 0x82e8:
-        case 0x82e9:
         case 0x86e8:
         case 0x86e9:
         case 0xc2e8:
-        case 0xc2e9:
         case 0xc6e8:
-        case 0xc6e9:
             ibm8514_accel_out_fifo(svga, port, val, len);
             break;
         case 0xf6ee:
             ibm8514_accel_out_fifo(svga, 0x82e8, val, len);
             break;
-        case 0xf6ef:
-            ibm8514_accel_out_fifo(svga, 0x82e9, val, len);
-            break;
 
         case 0x8ae8:
         case 0xcae8:
             ibm8514_accel_out_fifo(svga, port, val, len);
-            if (len != 1)
-                mach->accel.src_y = val;
-            break;
-        case 0x8ae9:
-        case 0x8ee9:
-        case 0xcae9:
-        case 0xcee9:
-            ibm8514_accel_out_fifo(svga, port, val, len);
+            if (len == 2) {
+                mach_log("SRCY=%d.\n", val & 0x07ff);
+                mach->accel.src_y = val & 0x07ff;
+            }
             break;
 
         case 0x8ee8:
         case 0xcee8:
             ibm8514_accel_out_fifo(svga, port, val, len);
-            if (len != 1)
-                mach->accel.src_x = val;
+            if (len == 2) {
+                mach_log("SRCX=%d.\n", val & 0x07ff);
+                mach->accel.src_x = val & 0x07ff;
+            }
             break;
 
         case 0x92e8:
-        case 0x92e9:
-        case 0x96e9:
         case 0xd2e8:
-        case 0xd2e9:
-        case 0xd6e9:
             ibm8514_accel_out_fifo(svga, port, val, len);
             break;
 
         case 0x96e8:
         case 0xd6e8:
             ibm8514_accel_out_fifo(svga, port, val, len);
-            if (len != 1)
+            if (len == 2)
                 mach->accel.test = val & 0x1fff;
             break;
 
         case 0x9ae8:
         case 0xdae8:
-            dev->accel.ssv_state = 0;
-            if (len == 1)
-                dev->accel.cmd = (dev->accel.cmd & 0xff00) | val;
-            else {
-                dev->data_available  = 0;
-                dev->data_available2 = 0;
-                dev->accel.cmd       = val;
-                mach_log("CMD8514=%04x, len=%d, pixcntl=%02x.\n", val, len, dev->accel.multifunc[0x0a]);
-                mach->accel.cmd_type = -1;
-                if (port == 0xdae8) {
-                    if (dev->accel.cmd & 0x100)
-                        dev->accel.cmd_back = 0;
-                }
-                ibm8514_accel_start(-1, 0, -1, 0, svga, len);
-            }
-            break;
-        case 0x9ae9:
-        case 0xdae9:
-            if (len == 1) {
-                dev->data_available  = 0;
-                dev->data_available2 = 0;
-                dev->accel.cmd       = (dev->accel.cmd & 0xff) | (val << 8);
-                mach->accel.cmd_type = -1;
-                if (port == 0xdae9) {
-                    if (dev->accel.cmd & 0x100)
-                        dev->accel.cmd_back = 0;
-                }
-                ibm8514_accel_start(-1, 0, -1, 0, svga, len);
-            }
+            mach->accel.cmd_type = -1;
+            ibm8514_accel_out_fifo(svga, port, val, len);
             break;
 
         case 0x9ee8:
         case 0xdee8:
-            dev->accel.ssv_state = 1;
-            if (len == 1)
-                dev->accel.short_stroke = (dev->accel.short_stroke & 0xff00) | val;
-            else {
-                dev->accel.short_stroke = val;
-                dev->accel.cx           = dev->accel.cur_x;
-                dev->accel.cy           = dev->accel.cur_y;
-
-                if (dev->accel.cur_x >= 0x600)
-                    dev->accel.cx |= ~0x5ff;
-
-                if (dev->accel.cur_y >= 0x600)
-                    dev->accel.cy |= ~0x5ff;
-
-                if (dev->accel.cmd & 0x1000) {
-                    ibm8514_short_stroke_start(-1, 0, -1, 0, svga, dev->accel.short_stroke & 0xff, len);
-                    ibm8514_short_stroke_start(-1, 0, -1, 0, svga, dev->accel.short_stroke >> 8, len);
-                } else {
-                    ibm8514_short_stroke_start(-1, 0, -1, 0, svga, dev->accel.short_stroke >> 8, len);
-                    ibm8514_short_stroke_start(-1, 0, -1, 0, svga, dev->accel.short_stroke & 0xff, len);
-                }
-            }
-            break;
-        case 0x9ee9:
-        case 0xdee9:
-            if (len == 1) {
-                dev->accel.short_stroke = (dev->accel.short_stroke & 0xff) | (val << 8);
-                dev->accel.cx           = dev->accel.cur_x;
-                dev->accel.cy           = dev->accel.cur_y;
-
-                if (dev->accel.cur_x >= 0x600) {
-                    dev->accel.cx |= ~0x5ff;
-                }
-                if (dev->accel.cur_y >= 0x600) {
-                    dev->accel.cy |= ~0x5ff;
-                }
-
-                if (dev->accel.cmd & 0x1000) {
-                    ibm8514_short_stroke_start(-1, 0, -1, 0, svga, dev->accel.short_stroke & 0xff, len);
-                    ibm8514_short_stroke_start(-1, 0, -1, 0, svga, dev->accel.short_stroke >> 8, len);
-                } else {
-                    ibm8514_short_stroke_start(-1, 0, -1, 0, svga, dev->accel.short_stroke >> 8, len);
-                    ibm8514_short_stroke_start(-1, 0, -1, 0, svga, dev->accel.short_stroke & 0xff, len);
-                }
-            }
+            ibm8514_accel_out_fifo(svga, port, val, len);
             break;
 
         case 0xa2e8:
         case 0xe2e8:
             if (port == 0xe2e8) {
-                if (dev->accel.cmd_back) {
-                    if (len == 1)
-                        dev->accel.bkgd_color = (dev->accel.bkgd_color & 0x00ff) | val;
-                    else
+                if (len == 2) {
+                    if (dev->accel.cmd_back) {
                         dev->accel.bkgd_color = val;
-                } else {
-                    if (len == 1) {
-                        if (mach->accel.cmd_type >= 0) {
-                            if (mach_pixel_read(mach))
-                                break;
-                            mach->accel.pix_trans[1] = val;
-                        }
                     } else {
                         if (mach->accel.cmd_type >= 0) {
                             if (mach_pixel_read(mach))
@@ -2904,80 +2847,25 @@ mach_accel_out_fifo(mach_t *mach, svga_t *svga, ibm8514_t *dev, uint16_t port, u
                             ibm8514_accel_out_pixtrans(svga, port, val, len);
                         }
                     }
-                }
-            } else {
-                if (len == 1)
-                    dev->accel.bkgd_color = (dev->accel.bkgd_color & 0x00ff) | val;
-                else
-                    dev->accel.bkgd_color = val;
-            }
-            break;
-        case 0xa2e9:
-        case 0xe2e9:
-            if (port == 0xe2e9) {
-                if (dev->accel.cmd_back) {
-                    if (len == 1)
-                        dev->accel.bkgd_color = (dev->accel.bkgd_color & 0xff00) | (val << 8);
                 } else {
-                    if (len == 1) {
-                        if (mach->accel.cmd_type >= 0) {
-                            if (mach_pixel_read(mach))
-                                break;
-                            mach->accel.pix_trans[0] = val;
-                            frgd_sel                 = (mach->accel.dp_config >> 13) & 7;
-                            bkgd_sel                 = (mach->accel.dp_config >> 7) & 3;
-                            mono_src                 = (mach->accel.dp_config >> 5) & 3;
-
-                            switch (mach->accel.dp_config & 0x200) {
-                                case 0x000: /*8-bit size*/
-                                    if (mono_src == 2) {
-                                        if ((frgd_sel != 2) && (bkgd_sel != 2)) {
-                                            mach_accel_start(mach->accel.cmd_type, 1, 8, mach->accel.pix_trans[0] | (mach->accel.pix_trans[1] << 8), 0, mach, dev);
-                                        } else
-                                            mach_accel_start(mach->accel.cmd_type, 1, 1, -1, mach->accel.pix_trans[0] | (mach->accel.pix_trans[1] << 8), mach, dev);
-                                    } else
-                                        mach_accel_start(mach->accel.cmd_type, 1, 1, -1, mach->accel.pix_trans[0] | (mach->accel.pix_trans[1] << 8), mach, dev);
-                                    break;
-                                case 0x200: /*16-bit size*/
-                                    if (mono_src == 2) {
-                                        if ((frgd_sel != 2) && (bkgd_sel != 2)) {
-                                            if (mach->accel.dp_config & 0x1000)
-                                                mach_accel_start(mach->accel.cmd_type, 1, 16, mach->accel.pix_trans[1] | (mach->accel.pix_trans[0] << 8), 0, mach, dev);
-                                            else
-                                                mach_accel_start(mach->accel.cmd_type, 1, 16, mach->accel.pix_trans[0] | (mach->accel.pix_trans[1] << 8), 0, mach, dev);
-                                        } else
-                                            mach_accel_start(mach->accel.cmd_type, 1, 2, -1, mach->accel.pix_trans[0] | (mach->accel.pix_trans[1] << 8), mach, dev);
-                                    } else
-                                        mach_accel_start(mach->accel.cmd_type, 1, 2, -1, mach->accel.pix_trans[0] | (mach->accel.pix_trans[1] << 8), mach, dev);
-                                    break;
-
-                                default:
-                                    break;
-                            }
-                        }
+                    if (mach->accel.cmd_type >= 0) {
+                        if (mach_pixel_read(mach))
+                            break;
+                        mach->accel.pix_trans[1] = val;
                     }
                 }
             } else {
-                if (len == 1)
-                    dev->accel.bkgd_color = (dev->accel.bkgd_color & 0xff00) | (val << 8);
+                if (len == 2)
+                    dev->accel.bkgd_color = val;
             }
             break;
 
         case 0xa6e8:
         case 0xe6e8:
             if (port == 0xe6e8) {
-                if (dev->accel.cmd_back) {
-                    if (len == 1)
-                        dev->accel.frgd_color = (dev->accel.frgd_color & 0x00ff) | val;
-                    else
+                if (len == 2) {
+                    if (dev->accel.cmd_back) {
                         dev->accel.frgd_color = val;
-                } else {
-                    if (len == 1) {
-                        if (mach->accel.cmd_type >= 0) {
-                            if (mach_pixel_read(mach))
-                                break;
-                            mach->accel.pix_trans[1] = val;
-                        }
                     } else {
                         if (mach->accel.cmd_type >= 0) {
                             if (mach_pixel_read(mach))
@@ -2989,62 +2877,58 @@ mach_accel_out_fifo(mach_t *mach, svga_t *svga, ibm8514_t *dev, uint16_t port, u
                             ibm8514_accel_out_pixtrans(svga, port, val, len);
                         }
                     }
-                }
-            } else {
-                if (len == 1)
-                    dev->accel.frgd_color = (dev->accel.frgd_color & 0x00ff) | val;
-                else
-                    dev->accel.frgd_color = val;
-            }
-            break;
-        case 0xa6e9:
-        case 0xe6e9:
-            if (port == 0xe6e9) {
-                if (dev->accel.cmd_back) {
-                    if (len == 1)
-                        dev->accel.frgd_color = (dev->accel.frgd_color & 0xff00) | (val << 8);
                 } else {
-                    if (len == 1) {
-                        if (mach->accel.cmd_type >= 0) {
-                            if (mach_pixel_read(mach))
-                                break;
-                            mach->accel.pix_trans[0] = val;
-                            frgd_sel                 = (mach->accel.dp_config >> 13) & 7;
-                            bkgd_sel                 = (mach->accel.dp_config >> 7) & 3;
-                            mono_src                 = (mach->accel.dp_config >> 5) & 3;
-
-                            switch (mach->accel.dp_config & 0x200) {
-                                case 0x000: /*8-bit size*/
-                                    if (mono_src == 2) {
-                                        if ((frgd_sel != 2) && (bkgd_sel != 2)) {
-                                            mach_accel_start(mach->accel.cmd_type, 1, 8, mach->accel.pix_trans[0] | (mach->accel.pix_trans[1] << 8), 0, mach, dev);
-                                        } else
-                                            mach_accel_start(mach->accel.cmd_type, 1, 1, -1, mach->accel.pix_trans[0] | (mach->accel.pix_trans[1] << 8), mach, dev);
-                                    } else
-                                        mach_accel_start(mach->accel.cmd_type, 1, 1, -1, mach->accel.pix_trans[0] | (mach->accel.pix_trans[1] << 8), mach, dev);
-                                    break;
-                                case 0x200: /*16-bit size*/
-                                    if (mono_src == 2) {
-                                        if ((frgd_sel != 2) && (bkgd_sel != 2)) {
-                                            if (mach->accel.dp_config & 0x1000)
-                                                mach_accel_start(mach->accel.cmd_type, 1, 16, mach->accel.pix_trans[1] | (mach->accel.pix_trans[0] << 8), 0, mach, dev);
-                                            else
-                                                mach_accel_start(mach->accel.cmd_type, 1, 16, mach->accel.pix_trans[0] | (mach->accel.pix_trans[1] << 8), 0, mach, dev);
-                                        } else
-                                            mach_accel_start(mach->accel.cmd_type, 1, 2, -1, mach->accel.pix_trans[0] | (mach->accel.pix_trans[1] << 8), mach, dev);
-                                    } else
-                                        mach_accel_start(mach->accel.cmd_type, 1, 2, -1, mach->accel.pix_trans[0] | (mach->accel.pix_trans[1] << 8), mach, dev);
-                                    break;
-
-                                default:
-                                    break;
-                            }
-                        }
+                    if (mach->accel.cmd_type >= 0) {
+                        if (mach_pixel_read(mach))
+                            break;
+                        mach->accel.pix_trans[1] = val;
                     }
                 }
             } else {
-                if (len == 1)
-                    dev->accel.frgd_color = (dev->accel.frgd_color & 0xff00) | (val << 8);
+                if (len == 2)
+                    dev->accel.frgd_color = val;
+            }
+            break;
+
+        case 0xe2e9:
+        case 0xe6e9:
+            mach_log("Write PORT=%04x, 8514/A=%x, val=%04x, len=%d.\n", port, dev->accel.cmd_back, val, len);
+            if (len == 1) {
+                if (mach->accel.cmd_type >= 0) {
+                    if (mach_pixel_read(mach))
+                        break;
+                    mach->accel.pix_trans[0] = val;
+                    frgd_sel = (mach->accel.dp_config >> 13) & 7;
+                    bkgd_sel = (mach->accel.dp_config >> 7) & 3;
+                    mono_src = (mach->accel.dp_config >> 5) & 3;
+
+                    switch (mach->accel.dp_config & 0x200) {
+                        case 0x000: /*8-bit size*/
+                            if (mono_src == 2) {
+                                if ((frgd_sel != 2) && (bkgd_sel != 2)) {
+                                    mach_accel_start(mach->accel.cmd_type, 1, 8, mach->accel.pix_trans[0] | (mach->accel.pix_trans[1] << 8), 0, mach, dev);
+                                } else
+                                    mach_accel_start(mach->accel.cmd_type, 1, 1, -1, mach->accel.pix_trans[0] | (mach->accel.pix_trans[1] << 8), mach, dev);
+                            } else
+                                mach_accel_start(mach->accel.cmd_type, 1, 1, -1, mach->accel.pix_trans[0] | (mach->accel.pix_trans[1] << 8), mach, dev);
+                            break;
+                        case 0x200: /*16-bit size*/
+                            if (mono_src == 2) {
+                                if ((frgd_sel != 2) && (bkgd_sel != 2)) {
+                                    if (mach->accel.dp_config & 0x1000)
+                                        mach_accel_start(mach->accel.cmd_type, 1, 16, mach->accel.pix_trans[1] | (mach->accel.pix_trans[0] << 8), 0, mach, dev);
+                                    else
+                                        mach_accel_start(mach->accel.cmd_type, 1, 16, mach->accel.pix_trans[0] | (mach->accel.pix_trans[1] << 8), 0, mach, dev);
+                                } else
+                                    mach_accel_start(mach->accel.cmd_type, 1, 2, -1, mach->accel.pix_trans[0] | (mach->accel.pix_trans[1] << 8), mach, dev);
+                            } else
+                                mach_accel_start(mach->accel.cmd_type, 1, 2, -1, mach->accel.pix_trans[0] | (mach->accel.pix_trans[1] << 8), mach, dev);
+                            break;
+
+                        default:
+                            break;
+                    }
+                }
             }
             break;
 
@@ -3069,118 +2953,47 @@ mach_accel_out_fifo(mach_t *mach, svga_t *svga, ibm8514_t *dev, uint16_t port, u
 
         case 0xbee8:
         case 0xfee8:
-            if (len == 1)
-                dev->accel.multifunc_cntl = (dev->accel.multifunc_cntl & 0xff00) | val;
-            else {
-                dev->accel.multifunc_cntl                             = val;
-                dev->accel.multifunc[dev->accel.multifunc_cntl >> 12] = dev->accel.multifunc_cntl & 0xfff;
-                if ((dev->accel.multifunc_cntl >> 12) == 1)
-                    dev->accel.clip_top = val & 0x7ff;
-
-                if ((dev->accel.multifunc_cntl >> 12) == 2)
-                    dev->accel.clip_left = val & 0x7ff;
-
-                if ((dev->accel.multifunc_cntl >> 12) == 3)
-                    dev->accel.multifunc[3] = val & 0x7ff;
-
-                if ((dev->accel.multifunc_cntl >> 12) == 4)
-                    dev->accel.multifunc[4] = val & 0x7ff;
-
-                mach_log("CLIPBOTTOM=%d, CLIPRIGHT=%d, bpp=%d, pitch=%d.\n", dev->accel.multifunc[3], dev->accel.multifunc[4], dev->accel_bpp, dev->pitch);
+            ibm8514_accel_out_fifo(svga, port, val, len);
+            if (len == 2) {
                 if ((dev->accel.multifunc_cntl >> 12) == 5) {
                     if ((dev->local & 0xff) < 0x02)
                         dev->ext_crt_pitch = 128;
                 }
-                if (port == 0xfee8)
-                    dev->accel.cmd_back = 1;
-                else
-                    dev->accel.cmd_back = 0;
-            }
-            break;
-        case 0xbee9:
-        case 0xfee9:
-            if (len == 1) {
-                dev->accel.multifunc_cntl                             = (dev->accel.multifunc_cntl & 0xff) | (val << 8);
-                dev->accel.multifunc[dev->accel.multifunc_cntl >> 12] = dev->accel.multifunc_cntl & 0xfff;
-                if ((dev->accel.multifunc_cntl >> 12) == 1)
-                    dev->accel.clip_top = dev->accel.multifunc_cntl & 0x7ff;
-
-                if ((dev->accel.multifunc_cntl >> 12) == 2)
-                    dev->accel.clip_left = dev->accel.multifunc_cntl & 0x7ff;
-
-                if ((dev->accel.multifunc_cntl >> 12) == 5) {
-                    if ((dev->local & 0xff) < 0x02)
-                        dev->ext_crt_pitch = 128;
-                }
-                if (port == 0xfee9)
-                    dev->accel.cmd_back = 1;
-                else
-                    dev->accel.cmd_back = 0;
             }
             break;
 
             /*ATI Mach8/32 specific registers*/
         case 0x82ee:
-            mach->accel.patt_data_idx = val & 0x1f;
-            mach_log("Pattern Data Index = %d.\n", val & 0x1f);
+            mach->accel.patt_data_idx_reg = val & 0x1f;
+            mach->accel.patt_data_idx = mach->accel.patt_data_idx_reg;
+
+            if (mach->accel.patt_data_idx_reg < 0x10)
+                mach->accel.color_pattern_idx = mach->accel.patt_idx;
+            else
+                mach->accel.color_pattern_idx = 0;
+
+            mach_log("Write Port 82ee: Pattern Data Index=%d.\n", val & 0x1f);
             break;
 
         case 0x8eee:
-            if (len == 1)
-                mach->accel.patt_data[mach->accel.patt_data_idx] = val;
-            else {
-                mach->accel.patt_data[mach->accel.patt_data_idx]     = val & 0xff;
-                mach->accel.patt_data[mach->accel.patt_data_idx + 1] = (val >> 8) & 0xff;
-                if (mach->accel.mono_pattern_enable)
-                    mach->accel.patt_data_idx = (mach->accel.patt_data_idx + 2) & 0x17;
-                else {
-                    frgd_sel = (mach->accel.dp_config >> 13) & 7;
-                    mono_src = (mach->accel.dp_config >> 5) & 3;
-                    if ((dev->accel_bpp == 24) && (mach->accel.patt_len == 0x17) && (frgd_sel == 5)) {
-                        mach->accel.patt_data_idx += 2;
-                        dev->accel.y1 = 1;
-                    } else {
-                        if (dev->accel_bpp == 24)
-                            mach->accel.patt_data_idx += 2;
-                        else
-                            mach->accel.patt_data_idx = (mach->accel.patt_data_idx + 2) & mach->accel.patt_len;
-                    }
-                    mach_log("ExtCONFIG = %04x, Pattern Mono = %04x, selidx = %d, dataidx = %d, bit 0 = %02x len = %d.\n", mach->accel.ext_ge_config, val, mach->accel.patt_idx, mach->accel.patt_data_idx, val & 1, mach->accel.patt_len);
-                }
-            }
-            break;
-        case 0x8eef:
-            if (len == 1) {
-                mach->accel.patt_data[mach->accel.patt_data_idx + 1] = val;
-                if (mach->accel.mono_pattern_enable)
-                    mach->accel.patt_data_idx = (mach->accel.patt_data_idx + 2) & 7;
-                else {
-                    frgd_sel = (mach->accel.dp_config >> 13) & 7;
-                    if ((dev->accel_bpp == 24) && (mach->accel.patt_len == 0x17) && (frgd_sel == 5)) {
-                        mach->accel.patt_data_idx += 2;
-                        dev->accel.y1 = 1;
-                    } else
-                        mach->accel.patt_data_idx = (mach->accel.patt_data_idx + 2) & mach->accel.patt_len;
+            if (len == 2) {
+                if (mach->accel.patt_data_idx_reg < 0x10) {
+                    mach->accel.color_pattern[mach->accel.patt_data_idx] = val & 0xff;
+                    mach->accel.color_pattern[mach->accel.patt_data_idx + 1] = (val >> 8) & 0xff;
+                    mach_log("Write Port 8eee: Color Pattern Word Data[%d]=%04x.\n", mach->accel.patt_data_idx, val);
+                } else {
+                    mach->accel.mono_pattern_normal[mach->accel.patt_data_idx - 0x10] = val & 0xff;
+                    mach->accel.mono_pattern_normal[(mach->accel.patt_data_idx + 1) - 0x10] = (val >> 8) & 0xff;
+                    mach_log("Write Port 8eee: Mono Pattern Word Data[%d]=%04x.\n", mach->accel.patt_data_idx - 0x10, val);
                 }
+                mach->accel.patt_data_idx += 2;
             }
             break;
 
         case 0x96ee:
-            if (len == 1)
-                mach->accel.bres_count = (mach->accel.bres_count & 0x700) | val;
-            else {
+            if (len == 2) {
                 mach->accel.bres_count = val & 0x7ff;
-                mach_log("BresenhamDraw = %04x.\n", mach->accel.dp_config);
-                dev->data_available  = 0;
-                dev->data_available2 = 0;
-                mach->accel.cmd_type = 1;
-                mach_accel_start(mach->accel.cmd_type, 0, -1, -1, 0, mach, dev);
-            }
-            break;
-        case 0x96ef:
-            if (len == 1) {
-                mach->accel.bres_count = (mach->accel.bres_count & 0xff) | ((val & 0x07) << 8);
-                mach_log("96EE (2) line draw.\n");
+                mach_log("BresenhamDraw=%04x.\n", mach->accel.dp_config);
                 dev->data_available  = 0;
                 dev->data_available2 = 0;
                 mach->accel.cmd_type = 1;
@@ -3193,15 +3006,13 @@ mach_accel_out_fifo(mach_t *mach, svga_t *svga, ibm8514_t *dev, uint16_t port, u
             break;
 
         case 0xa2ee:
-            mach_log("Line OPT = %04x\n", val);
-            if (len == 1)
-                mach->accel.linedraw_opt = (mach->accel.linedraw_opt & 0xff00) | val;
-            else {
+            mach_log("Line OPT=%04x.\n", val);
+            if (len == 2) {
                 mach->accel.linedraw_opt = val;
-                mach->accel.bbottom = dev->accel.multifunc[3] & 0x7ff;
-                mach->accel.btop = dev->accel.clip_top & 0x7ff;
-                mach->accel.bleft = dev->accel.clip_left & 0x7ff;
-                mach->accel.bright = dev->accel.multifunc[4] & 0x7ff;
+                mach->accel.bbottom = dev->accel.multifunc[3];
+                mach->accel.btop = dev->accel.multifunc[1];
+                mach->accel.bleft = dev->accel.multifunc[2];
+                mach->accel.bright = dev->accel.multifunc[4];
                 if (mach->accel.linedraw_opt & 0x100) {
                     mach->accel.bbottom = 2047;
                     mach->accel.btop = 0;
@@ -3210,88 +3021,50 @@ mach_accel_out_fifo(mach_t *mach, svga_t *svga, ibm8514_t *dev, uint16_t port, u
                 }
             }
             break;
-        case 0xa2ef:
-            if (len == 1)
-                mach->accel.linedraw_opt = (mach->accel.linedraw_opt & 0x00ff) | (val << 8);
-            break;
 
         case 0xa6ee:
-            if (len == 1)
-                mach->accel.dest_x_start = (mach->accel.dest_x_start & 0x700) | val;
-            else
+            if (len == 2)
                 mach->accel.dest_x_start = val & 0x7ff;
             break;
-        case 0xa6ef:
-            if (len == 1)
-                mach->accel.dest_x_start = (mach->accel.dest_x_start & 0x0ff) | ((val & 0x07) << 8);
-            break;
 
         case 0xaaee:
-            if (len == 1)
-                mach->accel.dest_x_end = (mach->accel.dest_x_end & 0x700) | val;
-            else
+            if (len == 2)
                 mach->accel.dest_x_end = val & 0x7ff;
             break;
-        case 0xaaef:
-            if (len == 1)
-                mach->accel.dest_x_end = (mach->accel.dest_x_end & 0x0ff) | ((val & 0x07) << 8);
-            break;
 
         case 0xaeee:
-            if (len == 1)
-                mach->accel.dest_y_end = (mach->accel.dest_y_end & 0x700) | val;
-            else {
+            if (len == 2) {
                 mach->accel.dest_y_end = val & 0x7ff;
                 if ((val + 1) == 0x10000) {
-                    mach_log("Dest_Y_end overflow val = %04x\n", val);
+                    mach_log("Dest_Y_end overflow val=%04x, DPCONFIG=%04x\n", val, mach->accel.dp_config);
                     mach->accel.dest_y_end = 0;
                 }
                 dev->data_available  = 0;
                 dev->data_available2 = 0;
-                mach_log("BitBLT = %04x.\n", mach->accel.dp_config);
+                mach_log("BitBLT=%04x.\n", mach->accel.dp_config);
+                mach_log(".\n");
                 mach->accel.cmd_type = 2; /*Non-conforming BitBLT from dest_y_end register (0xaeee)*/
                 mach_accel_start(mach->accel.cmd_type, 0, -1, -1, 0, mach, dev);
             }
             break;
-        case 0xaeef:
-            if (len == 1) {
-                mach->accel.dest_y_end = (mach->accel.dest_y_end & 0x0ff) | ((val & 0x07) << 8);
-                dev->data_available    = 0;
-                dev->data_available2   = 0;
-                mach->accel.cmd_type   = 2; /*Non-conforming BitBLT from dest_y_end register (0xaeee)*/
-                mach_accel_start(mach->accel.cmd_type, 0, -1, -1, 0, mach, dev);
-            }
-            break;
 
         case 0xb2ee:
-            if (len == 1)
-                mach->accel.src_x_start = (mach->accel.src_x_start & 0x700) | val;
-            else
+            if (len == 2)
                 mach->accel.src_x_start = val & 0x7ff;
             break;
-        case 0xb2ef:
-            if (len == 1)
-                mach->accel.src_x_start = (mach->accel.src_x_start & 0x0ff) | ((val & 0x07) << 8);
-            break;
 
         case 0xb6ee:
-            dev->accel.bkgd_mix = val & 0xff;
+            ibm8514_accel_out_fifo(svga, 0xb6e8, val, len);
             break;
 
         case 0xbaee:
-            dev->accel.frgd_mix = val & 0xff;
+            ibm8514_accel_out_fifo(svga, 0xbae8, val, len);
             break;
 
         case 0xbeee:
-            if (len == 1)
-                mach->accel.src_x_end = (mach->accel.src_x_end & 0x700) | val;
-            else
+            if (len == 2)
                 mach->accel.src_x_end = val & 0x7ff;
             break;
-        case 0xbeef:
-            if (len == 1)
-                mach->accel.src_x_end = (mach->accel.src_x_end & 0x0ff) | ((val & 0x07) << 8);
-            break;
 
         case 0xc2ee:
             mach->accel.src_y_dir = val & 1;
@@ -3303,9 +3076,7 @@ mach_accel_out_fifo(mach_t *mach, svga_t *svga, ibm8514_t *dev, uint16_t port, u
             break;
 
         case 0xcaee:
-            if (len == 1)
-                mach->accel.scan_to_x = (mach->accel.scan_to_x & 0x700) | val;
-            else {
+            if (len == 2) {
                 mach->accel.scan_to_x = (val & 0x7ff);
                 if ((val + 1) == 0x10000) {
                     mach_log("Scan_to_X overflow val = %04x\n", val);
@@ -3314,128 +3085,74 @@ mach_accel_out_fifo(mach_t *mach, svga_t *svga, ibm8514_t *dev, uint16_t port, u
                 dev->data_available  = 0;
                 dev->data_available2 = 0;
                 mach->accel.cmd_type = 5; /*Horizontal Raster Draw from scan_to_x register (0xcaee)*/
-                mach_log("ScanToX = %04x.\n", mach->accel.dp_config);
-                mach_accel_start(mach->accel.cmd_type, 0, -1, -1, 0, mach, dev);
-            }
-            break;
-        case 0xcaef:
-            if (len == 1) {
-                mach->accel.scan_to_x = (mach->accel.scan_to_x & 0x0ff) | ((val & 0x07) << 8);
-                dev->data_available   = 0;
-                dev->data_available2  = 0;
-                mach->accel.cmd_type  = 5; /*Horizontal Raster Draw from scan_to_x register (0xcaee)*/
+                mach_log("ScanToX=%04x.\n", mach->accel.dp_config);
+                mach_log(".\n");
                 mach_accel_start(mach->accel.cmd_type, 0, -1, -1, 0, mach, dev);
             }
             break;
 
         case 0xceee:
             mach_log("CEEE write val = %04x.\n", val);
-            if (len == 1)
-                mach->accel.dp_config = (mach->accel.dp_config & 0xff00) | val;
-            else {
+            if (len == 2) {
                 dev->data_available  = 0;
                 dev->data_available2 = 0;
                 mach->accel.dp_config = val;
             }
             break;
-        case 0xceef:
-            if (len == 1)
-                mach->accel.dp_config = (mach->accel.dp_config & 0x00ff) | (val << 8);
-            break;
 
         case 0xd2ee:
             mach->accel.patt_len = val & 0x1f;
-            mach_log("Pattern Length = %d, val = %04x.\n", val & 0x1f, val);
+            mach_log("Write Port d2ee: Pattern Length=%d, val=%04x.\n", val & 0x1f, val);
             mach->accel.mono_pattern_enable = !!(val & 0x80);
-            if (len != 1)
+            if (len == 2) {
+                mach->accel.block_write_mono_pattern_enable = !!(val & 0x8000);
                 mach->accel.patt_len_reg = val;
-            else
-                mach->accel.patt_len_reg = (mach->accel.patt_len_reg & 0xff00) | val;
-            break;
-        case 0xd2ef:
-            if (len == 1)
-                mach->accel.patt_len_reg = (mach->accel.patt_len_reg & 0x00ff) | (val << 8);
+            }
             break;
 
         case 0xd6ee:
             mach->accel.patt_idx = val & 0x1f;
-            mach_log("Pattern Index = %d, val = %02x.\n", val & 0x1f, val);
+            mach_log("Write Port d6ee: Pattern Index=%d.\n", val & 0x1f);
             break;
 
         case 0xdaee:
             mach_log("DAEE (extclipl) write val = %d\n", val);
-            if (len == 1)
-                dev->accel.clip_left = (dev->accel.clip_left & 0x700) | val;
-            else
-                dev->accel.clip_left = val & 0x7ff;
-            break;
-        case 0xdaef:
-            if (len == 1)
-                dev->accel.clip_left = (dev->accel.clip_left & 0x0ff) | ((val & 0x07) << 8);
+            if (len == 2)
+                dev->accel.multifunc[2] = val & 0x7ff;
             break;
 
         case 0xdeee:
             mach_log("DEEE (extclipt) write val = %d\n", val);
-            if (len == 1)
-                dev->accel.clip_top = (dev->accel.clip_top & 0x700) | val;
-            else
-                dev->accel.clip_top = val & 0x7ff;
-            break;
-        case 0xdeef:
-            if (len == 1)
-                dev->accel.clip_top = (dev->accel.clip_top & 0x0ff) | ((val & 0x07) << 8);
+            if (len == 2)
+                dev->accel.multifunc[1] = val & 0x7ff;
             break;
 
         case 0xe2ee:
             mach_log("E2EE (extclipr) write val = %d\n", val);
-            if (len == 1)
-                dev->accel.multifunc[4] = (dev->accel.multifunc[4] & 0x700) | val;
-            else
+            if (len == 2)
                 dev->accel.multifunc[4] = val & 0x7ff;
             break;
-        case 0xe2ef:
-            if (len == 1)
-                dev->accel.multifunc[4] = (dev->accel.multifunc[4] & 0x0ff) | ((val & 0x07) << 8);
-            break;
 
         case 0xe6ee:
             mach_log("E6EE (extclipb) write val = %d\n", val);
-            if (len == 1)
-                dev->accel.multifunc[3] = (dev->accel.multifunc[3] & 0x700) | val;
-            else
+            if (len == 2)
                 dev->accel.multifunc[3] = val & 0x7ff;
             break;
-        case 0xe6ef:
-            if (len == 1)
-                dev->accel.multifunc[3] = (dev->accel.multifunc[3] & 0x0ff) | ((val & 0x07) << 8);
-            break;
 
         case 0xeeee:
-            if (len == 1)
-                mach->accel.dest_cmp_fn = (mach->accel.dest_cmp_fn & 0xff00) | val;
-            else
+            if (len == 2)
                 mach->accel.dest_cmp_fn = val;
             break;
-        case 0xeeef:
-            if (len == 1)
-                mach->accel.dest_cmp_fn = (mach->accel.dest_cmp_fn & 0x00ff) | (val << 8);
-            break;
 
         case 0xf2ee:
             mach_log("F2EE.\n");
-            if (len == 1)
-                mach->accel.dst_clr_cmp_mask = (mach->accel.dst_clr_cmp_mask & 0xff00) | val;
-            else
+            if (len == 2)
                 mach->accel.dst_clr_cmp_mask = val;
             break;
-        case 0xf2ef:
-            if (len == 1)
-                mach->accel.dst_clr_cmp_mask = (mach->accel.dst_clr_cmp_mask & 0x00ff) | (val << 8);
-            break;
 
         case 0xfeee:
-            mach_log("LineDraw = %04x.\n", mach->accel.dp_config);
-            if (len != 1) {
+            mach_log("LineDraw=%04x.\n", mach->accel.dp_config);
+            if (len == 2) {
                 mach->accel.line_array[mach->accel.line_idx] = val;
                 dev->accel.cur_x                             = mach->accel.line_array[(mach->accel.line_idx == 4) ? 4 : 0];
                 dev->accel.cur_y                             = mach->accel.line_array[(mach->accel.line_idx == 5) ? 5 : 1];
@@ -3459,7 +3176,7 @@ mach_accel_out_fifo(mach_t *mach, svga_t *svga, ibm8514_t *dev, uint16_t port, u
 static void
 mach_accel_out_call(uint16_t port, uint8_t val, mach_t *mach, svga_t *svga, ibm8514_t *dev)
 {
-    if (port != 0x7aee && port != 0x7aef && port != 0x42e8 && port != 0x42e9)
+    if (port != 0x42e8 && port != 0x42e9)
         mach_log("[%04X:%08X]: Port CALL OUT=%04x, val=%02x.\n", CS, cpu_state.pc, port, val);
 
     switch (port) {
@@ -3467,10 +3184,6 @@ mach_accel_out_call(uint16_t port, uint8_t val, mach_t *mach, svga_t *svga, ibm8
         case 0x6e9:
         case 0xae8:
         case 0xee8:
-        case 0x12e8:
-        case 0x12e9:
-        case 0x1ae8:
-        case 0x1ae9:
         case 0x1ee8:
         case 0x1ee9:
         case 0x22e8:
@@ -3481,9 +3194,27 @@ mach_accel_out_call(uint16_t port, uint8_t val, mach_t *mach, svga_t *svga, ibm8
 
         case 0x6e8:
             /*In preparation to switch from VGA to 8514/A mode*/
-            if (!dev->on[0] || !dev->on[1] || (mach->accel.clock_sel & 0x01)) {
-                dev->hdisped = val;
-                dev->hdisp = (dev->hdisped + 1) << 3;
+            if (!(mach->shadow_cntl & 0x08)) {
+                if ((mach->shadow_set & 0x03) || (mach->accel.clock_sel & 0x01)) {
+                    dev->hdisped = val;
+                    dev->hdisp = (dev->hdisped + 1) << 3;
+                }
+            }
+            mach_log("[%04X:%08X]: ATI 8514/A: (0x%04x): hdisp=0x%02x.\n", CS, cpu_state.pc, port, val);
+            svga_recalctimings(svga);
+            break;
+
+        case 0x12e8:
+        case 0x12e9:
+            /*In preparation to switch from VGA to 8514/A mode*/
+            if (!(mach->shadow_cntl & 0x10)) {
+                if ((mach->shadow_set & 0x03) || (mach->accel.clock_sel & 0x01)) {
+                    WRITE8(port, dev->v_total_reg, val);
+                    dev->v_total_reg &= 0x1fff;
+                    dev->v_total = dev->v_total_reg + 1;
+                    if (dev->interlace)
+                        dev->v_total >>= 1;
+                }
             }
             mach_log("[%04X:%08X]: ATI 8514/A: (0x%04x): hdisp=0x%02x.\n", CS, cpu_state.pc, port, val);
             svga_recalctimings(svga);
@@ -3492,29 +3223,46 @@ mach_accel_out_call(uint16_t port, uint8_t val, mach_t *mach, svga_t *svga, ibm8
         case 0x16e8:
         case 0x16e9:
             /*In preparation to switch from VGA to 8514/A mode*/
-            if (!dev->on[0] || !dev->on[1] || (mach->accel.clock_sel & 0x01)) {
-                WRITE8(port, dev->v_disp, val);
-                dev->v_disp &= 0x1fff;
-                dev->vdisp = (dev->v_disp + 1) >> 1;
+            if (!(mach->shadow_cntl & 0x20)) {
+                if ((mach->shadow_set & 0x03) || (mach->accel.clock_sel & 0x01)) {
+                    WRITE8(port, dev->v_disp, val);
+                    dev->v_disp &= 0x1fff;
+                    dev->vdisp = (dev->v_disp + 1) >> 1;
+                }
             }
             mach_log("ATI 8514/A: V_DISP write 16E8 = %d\n", dev->v_disp);
             mach_log("ATI 8514/A: (0x%04x): vdisp=0x%02x.\n", port, val);
             svga_recalctimings(svga);
             break;
 
+        case 0x1ae8:
+        case 0x1ae9:
+            /*In preparation to switch from VGA to 8514/A mode*/
+            if (!(mach->shadow_cntl & 0x10)) {
+                if ((mach->shadow_set & 0x03) || (mach->accel.clock_sel & 0x01)) {
+                    WRITE8(port, dev->v_sync_start, val);
+                    dev->v_sync_start &= 0x1fff;
+                    dev->v_syncstart = dev->v_sync_start + 1;
+                    if (dev->interlace)
+                        dev->v_syncstart >>= 1;
+                }
+            }
+            mach_log("ATI 8514/A: V_SYNCSTART write 1AE8 = %d\n", dev->v_syncstart);
+            mach_log("ATI 8514/A: (0x%04x): vsyncstart=0x%02x.\n", port, val);
+            svga_recalctimings(svga);
+            break;
+
         case 0x4ae8:
-        case 0x4ae9:
             WRITE8(port, dev->accel.advfunc_cntl, val);
-            dev->on[port & 1] = dev->accel.advfunc_cntl & 0x01;
+            dev->on = dev->accel.advfunc_cntl & 0x01;
+            vga_on = !dev->on;
+
             if ((dev->local & 0xff) < 0x02)
                 dev->ext_crt_pitch = 128;
 
-            vga_on = !dev->on[port & 1];
-            mach->ext_on[port & 1] = dev->on[port & 1];
-            mach32_updatemapping(mach, svga);
-            dev->vendor_mode[port & 1] = 0;
-            mach_log("[%04X:%08X]: ATI 8514/A: (0x%04x): ON=%d, shadow crt=%x, hdisp=%d, vdisp=%d.\n", CS, cpu_state.pc, port, dev->on[port & 1], dev->accel.advfunc_cntl & 0x04, dev->hdisp, dev->vdisp);
-            mach_log("IBM mode set %s resolution.\n", (dev->accel.advfunc_cntl & 0x04) ? "2: 1024x768" : "1: 640x480");
+            mach_log("[%04X:%08X]: ATI 8514/A: (0x%04x): ON=%d, shadow crt=%x, hdisp=%d, vdisp=%d.\n", CS, cpu_state.pc, port, val & 0x01, dev->accel.advfunc_cntl & 0x04, dev->hdisp, dev->vdisp);
+            mach_log("Vendor IBM mode set %s resolution.\n", (dev->accel.advfunc_cntl & 0x04) ? "2: 1024x768" : "1: 640x480");
+
             svga_recalctimings(svga);
             break;
 
@@ -3546,6 +3294,7 @@ mach_accel_out_call(uint16_t port, uint8_t val, mach_t *mach, svga_t *svga, ibm8
             mach->cursor_offset_hi = mach->cursor_offset_hi_reg & 0x0f;
             dev->hwcursor.addr = ((mach->cursor_offset_lo | (mach->cursor_offset_hi << 16)) << 2);
             dev->hwcursor.ena = !!(mach->cursor_offset_hi_reg & 0x8000);
+            mach_log("HWCursorEnabled=%x.\n", dev->hwcursor.ena);
             break;
 
         case 0x12ee:
@@ -3593,6 +3342,8 @@ mach_accel_out_call(uint16_t port, uint8_t val, mach_t *mach, svga_t *svga, ibm8
                 else
                     dev->ext_crt_pitch <<= 1;
             }
+            dev->on = 1;
+            vga_on = 0;
             mach_log("ATI 8514/A: (0x%04x) val=0x%02x.\n", port, val);
             svga_recalctimings(svga);
             break;
@@ -3634,18 +3385,20 @@ mach_accel_out_call(uint16_t port, uint8_t val, mach_t *mach, svga_t *svga, ibm8
         case 0x46ef:
             WRITE8(port, mach->shadow_cntl, val);
             mach_log("ATI 8514/A: (0x%04x) val=%02x.\n", port, val);
+            mach32_updatemapping(mach, svga);
             break;
 
         case 0x4aee:
         case 0x4aef:
             WRITE8(port, mach->accel.clock_sel, val);
-            dev->on[port & 1] = mach->accel.clock_sel & 0x01;
-            mach_log("ATI 8514/A: (0x%04x): ON=%d, val=%04x, hdisp=%d, vdisp=%d, val=0x%02x.\n", port, dev->on[port & 1], val, dev->hdisp, dev->vdisp, val & 0xfe);
-            mach_log("ATI mode set %s resolution.\n", (dev->accel.advfunc_cntl & 0x04) ? "2: 1024x768" : "1: 640x480");
-            mach->ext_on[port & 1] = dev->on[port & 1];
-            vga_on = !dev->on[port & 1];
-            dev->vendor_mode[port & 1] = 1;
+            dev->on = mach->accel.clock_sel & 0x01;
+            mach_log("ATI 8514/A: (0x%04x): ON=%d, val=%04x, hdisp=%d, vdisp=%d, val=0x%02x.\n", port, dev->on, val, dev->hdisp, dev->vdisp, val & 0xfe);
+            if (!(port & 1))
+                mach_log("Vendor ATI mode set %s resolution.\n", (dev->accel.advfunc_cntl & 0x04) ? "2: 1024x768" : "1: 640x480");
+
+            vga_on = !dev->on;
             svga_recalctimings(svga);
+            mach32_updatemapping(mach, svga);
             break;
 
         case 0x52ee:
@@ -3670,6 +3423,8 @@ mach_accel_out_call(uint16_t port, uint8_t val, mach_t *mach, svga_t *svga, ibm8
                 mach_log("CRT Shadow Set 1: 640x480.\n");
             else if ((mach->shadow_set & 0x03) == 0x02)
                 mach_log("CRT Shadow Set 2: 1024x768.\n");
+
+            mach32_updatemapping(mach, svga);
             break;
 
         case 0x5eee:
@@ -3693,19 +3448,29 @@ mach_accel_out_call(uint16_t port, uint8_t val, mach_t *mach, svga_t *svga, ibm8
         case 0x6aee:
         case 0x6aef:
             WRITE8(port, mach->accel.max_waitstates, val);
+            mach->override_resolution = !!(mach->accel.max_waitstates & 0x400);
+            if (mach->override_resolution) {
+                dev->on = 1;
+                vga_on = !dev->on;
+                svga_recalctimings(svga);
+                mach32_updatemapping(mach, svga);
+            }
             break;
 
         case 0x6eee:
         case 0x6eef:
             WRITE8(port, mach->accel.ge_offset_lo, val);
             svga_recalctimings(svga);
+            mach_log("ATI 8514/A: (0x%04x) val=0x%02x, geoffset=%04x.\n", port, val, dev->accel.ge_offset);
+            mach32_updatemapping(mach, svga);
             break;
 
         case 0x72ee:
         case 0x72ef:
             WRITE8(port, mach->accel.ge_offset_hi, val);
-            mach_log("ATI 8514/A: (0x%04x) val=0x%02x, geoffset=%04x.\n", port, val, dev->accel.ge_offset);
             svga_recalctimings(svga);
+            mach_log("ATI 8514/A: (0x%04x) val=0x%02x, geoffset=%04x.\n", port, val, dev->accel.ge_offset);
+            mach32_updatemapping(mach, svga);
             break;
 
         case 0x76ee:
@@ -3714,6 +3479,7 @@ mach_accel_out_call(uint16_t port, uint8_t val, mach_t *mach, svga_t *svga, ibm8
             dev->ext_pitch = ((mach->accel.ge_pitch & 0xff) << 3);
             mach_log("ATI 8514/A: (0x%04x) val=0x%02x, extpitch=%d.\n", port, val, dev->ext_pitch);
             svga_recalctimings(svga);
+            mach32_updatemapping(mach, svga);
             break;
 
         case 0x7aee:
@@ -3722,8 +3488,9 @@ mach_accel_out_call(uint16_t port, uint8_t val, mach_t *mach, svga_t *svga, ibm8
             if ((dev->local & 0xff) >= 0x02) {
                 if (mach->accel.crt_pitch & 0xff)
                     dev->ext_crt_pitch = mach->accel.crt_pitch & 0xff;
+
                 switch (mach->accel.ext_ge_config & 0x30) {
-                    case 0:
+                    case 0x00:
                     case 0x10:
                         dev->bpp = 0;
                         break;
@@ -3743,11 +3510,11 @@ mach_accel_out_call(uint16_t port, uint8_t val, mach_t *mach, svga_t *svga, ibm8
                         break;
                 }
                 svga_set_ramdac_type(svga, !!(mach->accel.ext_ge_config & 0x4000));
-                dev->vendor_mode[port & 1] = 1;
-                mach32_updatemapping(mach, svga);
                 mach_log("ATI 8514/A: (0x%04x) val=%02x.\n", port, val);
                 svga_recalctimings(svga);
+                mach32_updatemapping(mach, svga);
             } else {
+                mach_log("ATI 8514/A: (0x%04x) val=%02x.\n", port, val & 0x30);
                 ati_eeprom_write(&mach->eeprom, !!(mach->accel.ext_ge_config & 0x4040), !!(mach->accel.ext_ge_config & 0x2020), !!(mach->accel.ext_ge_config & 0x1010));
             }
             break;
@@ -3796,9 +3563,10 @@ mach_accel_in_fifo(mach_t *mach, svga_t *svga, ibm8514_t *dev, uint16_t port, in
 
         case 0x9ae8:
         case 0xdae8:
-            if (len != 1) {
+            if (len == 2) {
                 if (dev->force_busy)
                     temp |= 0x200; /*Hardware busy*/
+
                 dev->force_busy = 0;
                 if (dev->data_available) {
                     temp |= 0x100; /*Read Data available*/
@@ -3829,10 +3597,11 @@ mach_accel_in_fifo(mach_t *mach, svga_t *svga, ibm8514_t *dev, uint16_t port, in
         case 0xdae9:
             if (len == 1) {
                 if (dev->force_busy2)
-                    temp |= 2; /*Hardware busy*/
+                    temp |= 0x02; /*Hardware busy*/
+
                 dev->force_busy2 = 0;
                 if (dev->data_available2) {
-                    temp |= 1; /*Read Data available*/
+                    temp |= 0x01; /*Read Data available*/
                     if (mach->accel.cmd_type >= 0) {
                         switch (mach->accel.cmd_type) {
                             case 2:
@@ -3877,18 +3646,23 @@ mach_accel_in_fifo(mach_t *mach, svga_t *svga, ibm8514_t *dev, uint16_t port, in
             } else {
                 if (ibm8514_cpu_dest(svga)) {
                     cmd = (dev->accel.cmd >> 13);
-                    if (len != 1) {
+                    if (len == 2) {
                         READ_PIXTRANS_WORD(dev->accel.cx, 0)
-                        if (dev->accel.input && !dev->accel.odd_in && !dev->accel.sx) {
-                            temp &= ~0xff00;
-                            temp |= (dev->vram[(dev->accel.newdest_in + dev->accel.cur_x) & dev->vram_mask] << 8);
-                        }
-                        if (dev->subsys_stat & 1) {
+                        if (dev->subsys_stat & 0x01) {
                             dev->force_busy = 1;
                             dev->data_available = 1;
                         }
+                        if (dev->accel.input) {
+                            ibm8514_accel_out_pixtrans(svga, port, temp & 0xff, len);
+                            if (dev->accel.odd_in) { /*WORDs on odd destination scan lengths.*/
+                                dev->accel.odd_in = 0;
+                                temp &= ~0xff00;
+                                READ_HIGH(dev->accel.dest + dev->accel.cx, temp);
+                            }
+                            ibm8514_accel_out_pixtrans(svga, port, (temp >> 8) & 0xff, len);
+                        } else
+                            ibm8514_accel_out_pixtrans(svga, port, temp, len);
                     }
-                    ibm8514_accel_out_pixtrans(svga, port, temp, len);
                 }
             }
             break;
@@ -3937,47 +3711,9 @@ mach_accel_in_fifo(mach_t *mach, svga_t *svga, ibm8514_t *dev, uint16_t port, in
             }
             break;
 
-        case 0xbee8:
-        case 0xfee8:
-            if (len != 1) {
-                mach_log("Multifunc_cntl = %d.\n", dev->accel.multifunc_cntl >> 12);
-                switch ((dev->accel.multifunc_cntl >> 12) & 0x0f) {
-                    case 0:
-                        temp = dev->accel.multifunc[0];
-                        break;
-                    case 1:
-                        temp = dev->accel.clip_top;
-                        break;
-                    case 2:
-                        temp = dev->accel.clip_left;
-                        break;
-                    case 3:
-                        temp = dev->accel.multifunc[3];
-                        break;
-                    case 4:
-                        temp = dev->accel.multifunc[4];
-                        break;
-                    case 5:
-                        temp = dev->accel.multifunc[5];
-                        break;
-                    case 8:
-                        temp = dev->accel.multifunc[8];
-                        break;
-                    case 9:
-                        temp = dev->accel.multifunc[9];
-                        break;
-                    case 0x0a:
-                        temp = dev->accel.multifunc[0x0a];
-                        break;
-
-                    default:
-                        break;
-                }
-            }
-            break;
-
         case 0x82ee:
-            temp = mach->accel.patt_data_idx;
+            if (len == 2)
+                temp = mach->accel.patt_data_idx;
             break;
 
         case 0x86ee:
@@ -4106,10 +3842,9 @@ mach_accel_in_fifo(mach_t *mach, svga_t *svga, ibm8514_t *dev, uint16_t port, in
             break;
 
         case 0xdaee:
-            if (len != 1) {
-                temp = mach->accel.src_x;
+            if (len == 2) {
                 if ((dev->local & 0xff) >= 0x02)
-                    temp &= 0x7ff;
+                    temp = mach->accel.src_x;
             } else
                 temp = mach->accel.src_x & 0xff;
             break;
@@ -4119,10 +3854,9 @@ mach_accel_in_fifo(mach_t *mach, svga_t *svga, ibm8514_t *dev, uint16_t port, in
             break;
 
         case 0xdeee:
-            if (len != 1) {
-                temp = mach->accel.src_y;
+            if (len == 2) {
                 if ((dev->local & 0xff) >= 0x02)
-                    temp &= 0x7ff;
+                    temp = mach->accel.src_y;
             } else
                 temp = mach->accel.src_y & 0xff;
             break;
@@ -4132,7 +3866,7 @@ mach_accel_in_fifo(mach_t *mach, svga_t *svga, ibm8514_t *dev, uint16_t port, in
             break;
 
         case 0xfaee:
-            if (len != 1) {
+            if (len == 2) {
                 if ((dev->local & 0xff) >= 0x02) {
                     if (mach->pci_bus)
                         temp = 0x0017;
@@ -4163,7 +3897,8 @@ mach_accel_in_fifo(mach_t *mach, svga_t *svga, ibm8514_t *dev, uint16_t port, in
             break;
     }
 
-    mach_log("[%04X:%08X]: Port FIFO IN=%04x, temp=%04x, len=%d.\n", CS, cpu_state.pc, port, temp, len);
+    if (port != 0x9aee && port != 0x62ee)
+        mach_log("[%04X:%08X]: Port FIFO IN=%04x, temp=%04x, len=%d.\n", CS, cpu_state.pc, port, temp, len);
 
     return temp;
 }
@@ -4171,14 +3906,7 @@ mach_accel_in_fifo(mach_t *mach, svga_t *svga, ibm8514_t *dev, uint16_t port, in
 static uint8_t
 mach_accel_in_call(uint16_t port, mach_t *mach, svga_t *svga, ibm8514_t *dev)
 {
-    uint8_t         temp        = 0;
-    uint16_t        clip_b_ibm  = dev->accel.multifunc[3];
-    uint16_t        clip_r_ibm  = dev->accel.multifunc[4];
-    int16_t         clip_l      = dev->accel.clip_left & 0x7ff;
-    int16_t         clip_t      = dev->accel.clip_top & 0x7ff;
-    int16_t         clip_r      = dev->accel.multifunc[4] & 0x7ff;
-    int16_t         clip_b      = dev->accel.multifunc[3] & 0x7ff;
-    int             cmd         = (dev->accel.cmd >> 13);
+    uint8_t temp = 0;
 
     switch (port) {
         case 0x2e8:
@@ -4196,24 +3924,8 @@ mach_accel_in_call(uint16_t port, mach_t *mach, svga_t *svga, ibm8514_t *dev)
             if (dev->vc == dev->v_syncstart)
                 dev->subsys_stat |= 1;
 
-            if (mach->accel.cmd_type >= 0) {
-                if (((dev->accel.dx) >= clip_l) && ((dev->accel.dx) <= clip_r) && ((dev->accel.dy) >= clip_t) && ((dev->accel.dy) <= clip_b))
-                    temp |= 2;
-            } else {
-                if (cmd == 6) {
-                    if (((dev->accel.dx) >= dev->accel.clip_left) && ((dev->accel.dx) <= clip_r_ibm) && ((dev->accel.dy) >= dev->accel.clip_top) && ((dev->accel.dy) <= clip_b_ibm))
-                        temp |= 2;
-                } else {
-                    if (((dev->accel.cx) >= dev->accel.clip_left) && ((dev->accel.dx) <= clip_r_ibm) && ((dev->accel.cy) >= dev->accel.clip_top) && ((dev->accel.cy) <= clip_b_ibm))
-                        temp |= 2;
-                }
-            }
-
-            if (!dev->force_busy)
-                temp |= 8;
-
             if (port & 1)
-                temp = 0x80;
+                temp = dev->vram_512k_8514 ? 0x00 : 0x80;
             else {
                 temp |= (dev->subsys_stat | 0x80);
                 if (mach->accel.ext_ge_config & 0x08)
@@ -4256,7 +3968,7 @@ mach_accel_in_call(uint16_t port, mach_t *mach, svga_t *svga, ibm8514_t *dev)
 
             if (!(port & 1)) {
                 temp &= ~0x0c;
-                switch (mach->memory) {
+                switch (dev->vram_amount) {
                     case 1024:
                         temp |= 0x04;
                         break;
@@ -4268,8 +3980,6 @@ mach_accel_in_call(uint16_t port, mach_t *mach, svga_t *svga, ibm8514_t *dev)
                         break;
 
                     default:
-                        if ((dev->local & 0xff) < 0x02)
-                            temp |= 0x04;
                         break;
                 }
             }
@@ -4351,7 +4061,7 @@ mach_accel_in_call(uint16_t port, mach_t *mach, svga_t *svga, ibm8514_t *dev)
         default:
             break;
     }
-    if (port != 0x62ee && port != 0x62ef && port != 0x42e8 && port != 0x42e9 && port != 0x02e8 && port != 0x02e9)
+    if (port != 0x42e8 && port != 0x42e9 && port != 0x62ee && port != 0x62ef && port != 0x02e8 && port != 0x02e9)
         mach_log("[%04X:%08X]: Port NORMAL IN=%04x, temp=%04x.\n", CS, cpu_state.pc, port, temp);
 
     return temp;
@@ -4568,43 +4278,6 @@ mach_accel_inl(uint16_t port, void *priv)
     return temp;
 }
 
-static uint32_t
-mach32_decode_addr(svga_t *svga, uint32_t addr, int write)
-{
-    int memory_map_mode = (svga->gdcreg[6] >> 2) & 3;
-
-    addr &= 0x1ffff;
-
-    switch (memory_map_mode) {
-        case 0:
-            break;
-        case 1:
-            if (addr >= 0x10000)
-                return 0xffffffff;
-            break;
-        case 2:
-            addr -= 0x10000;
-            if (addr >= 0x8000)
-                return 0xffffffff;
-            break;
-        default:
-        case 3:
-            addr -= 0x18000;
-            if (addr >= 0x8000)
-                return 0xffffffff;
-            break;
-    }
-
-    if (memory_map_mode <= 1) {
-        if (write)
-            addr = (addr & svga->banked_mask) + svga->write_bank;
-        else
-            addr = (addr & svga->banked_mask) + svga->read_bank;
-    }
-
-    return addr;
-}
-
 static __inline void
 mach32_write_common(uint32_t addr, uint8_t val, int linear, mach_t *mach, svga_t *svga)
 {
@@ -4613,37 +4286,51 @@ mach32_write_common(uint32_t addr, uint8_t val, int linear, mach_t *mach, svga_t
     int     reset_wm   = 0;
     latch8514_t vall;
     uint8_t wm = svga->writemask;
-    uint8_t count;
     uint8_t i;
 
     cycles -= svga->monitor->mon_video_timing_write_b;
 
     if (linear) {
+        if (!dev->vram_512k_8514 && ((mach->accel.ext_ge_config & 0x30) == 0x00))
+            addr <<= 1;
+
         addr &= dev->vram_mask;
         dev->changedvram[addr >> 12] = svga->monitor->mon_changeframecount;
-        dev->vram[addr]              = val;
+        if (!dev->vram_512k_8514 && ((mach->accel.ext_ge_config & 0x30) == 0x00)) {
+            switch (addr & 0x06) {
+                case 0x00:
+                case 0x06:
+                    dev->vram[addr] = val & 0x0f;
+                    dev->vram[addr + 1] = (val >> 4) & 0x0f;
+                    break;
+                case 0x02:
+                    dev->vram[addr + 2] = val & 0x0f;
+                    dev->vram[addr + 3] = (val >> 4) & 0x0f;
+                    break;
+                case 0x04:
+                    dev->vram[addr - 2] = val & 0x0f;
+                    dev->vram[addr - 1] = (val >> 4) & 0x0f;
+                    break;
+                default:
+                    break;
+            }
+        } else
+            dev->vram[addr] = val;
+
         return;
-    } else {
-        xga_write_test(addr, val, svga);
-        addr = mach32_decode_addr(svga, addr, 1);
-        if (addr == 0xffffffff)
-            return;
     }
 
     if (!(svga->gdcreg[6] & 1))
         svga->fullchange = 2;
 
-    if (((svga->chain4 && (svga->packed_chain4 || svga->force_old_addr)) || svga->fb_only) && (svga->writemode < 4)) {
+    if (svga->chain4) {
         writemask2 = 1 << (addr & 3);
         addr &= ~3;
-    } else if (svga->chain4 && (svga->writemode < 4)) {
-        writemask2 = 1 << (addr & 3);
-        addr &= ~3;
-        addr = ((addr & 0xfffc) << 2) | ((addr & 0x30000) >> 14) | (addr & ~0x3ffff);
     } else if (svga->chain2_write) {
         writemask2 &= ~0xa;
         if (addr & 1)
             writemask2 <<= 1;
+
         addr &= ~1;
         addr &= dev->vram_mask;
     } else {
@@ -4660,90 +4347,53 @@ mach32_write_common(uint32_t addr, uint8_t val, int linear, mach_t *mach, svga_t
 
     dev->changedvram[addr >> 12] = svga->monitor->mon_changeframecount;
 
-    count = 4;
+    val = ((val >> (svga->gdcreg[3] & 7)) | (val << (8 - (svga->gdcreg[3] & 7))));
+    if ((svga->gdcreg[8] == 0xff) && !(svga->gdcreg[3] & 0x18) && (!svga->gdcreg[1] || svga->set_reset_disabled)) {
+        for (i = 0; i < 4; i++) {
+            if (writemask2 & (1 << i))
+                dev->vram[addr | i] = val;
+        }
+    } else {
+        for (i = 0; i < 4; i++) {
+            if (svga->gdcreg[1] & (1 << i))
+                vall.b[i] = !!(svga->gdcreg[0] & (1 << i)) * 0xff;
+            else
+                vall.b[i] = val;
+        }
 
-    switch (svga->writemode) {
-        case 0:
-            val = ((val >> (svga->gdcreg[3] & 7)) | (val << (8 - (svga->gdcreg[3] & 7))));
-            if ((svga->gdcreg[8] == 0xff) && !(svga->gdcreg[3] & 0x18) && (!svga->gdcreg[1] || svga->set_reset_disabled)) {
-                for (i = 0; i < count; i++) {
-                    if (writemask2 & (1 << i))
-                        dev->vram[addr | i] = val;
-                }
-                return;
-            } else {
-                for (i = 0; i < count; i++) {
-                    if (svga->gdcreg[1] & (1 << i))
-                        vall.b[i] = !!(svga->gdcreg[0] & (1 << i)) * 0xff;
-                    else
-                        vall.b[i] = val;
-                }
-            }
-            break;
-        case 1:
-            for (i = 0; i < count; i++) {
-                if (writemask2 & (1 << i))
-                    dev->vram[addr | i] = dev->latch.b[i];
-            }
-            return;
-        case 2:
-            for (i = 0; i < count; i++)
-                vall.b[i] = !!(val & (1 << i)) * 0xff;
-
-            if (!(svga->gdcreg[3] & 0x18) && (!svga->gdcreg[1] || svga->set_reset_disabled)) {
-                for (i = 0; i < count; i++) {
+        switch (svga->gdcreg[3] & 0x18) {
+            case 0x00: /* Set */
+                for (i = 0; i < 4; i++) {
                     if (writemask2 & (1 << i))
                         dev->vram[addr | i] = (vall.b[i] & svga->gdcreg[8]) | (dev->latch.b[i] & ~svga->gdcreg[8]);
                 }
-                return;
-            }
-            break;
-        case 3:
-            val = ((val >> (svga->gdcreg[3] & 7)) | (val << (8 - (svga->gdcreg[3] & 7))));
-            wm  = svga->gdcreg[8];
-            svga->gdcreg[8] &= val;
+                break;
+            case 0x08: /* AND */
+                for (i = 0; i < 4; i++) {
+                    if (writemask2 & (1 << i))
+                        dev->vram[addr | i] = (vall.b[i] | ~svga->gdcreg[8]) & dev->latch.b[i];
+                }
+                break;
+            case 0x10: /* OR */
+                for (i = 0; i < 4; i++) {
+                    if (writemask2 & (1 << i))
+                        dev->vram[addr | i] = (vall.b[i] & svga->gdcreg[8]) | dev->latch.b[i];
+                }
+                break;
+            case 0x18: /* XOR */
+                for (i = 0; i < 4; i++) {
+                    if (writemask2 & (1 << i))
+                        dev->vram[addr | i] = (vall.b[i] & svga->gdcreg[8]) ^ dev->latch.b[i];
+                }
+                break;
 
-            for (i = 0; i < count; i++)
-                vall.b[i] = !!(svga->gdcreg[0] & (1 << i)) * 0xff;
+            default:
+                break;
+        }
 
-            reset_wm = 1;
-            break;
-        default:
-            break;
+        if (reset_wm)
+            svga->gdcreg[8] = wm;
     }
-
-    switch (svga->gdcreg[3] & 0x18) {
-        case 0x00: /* Set */
-            for (i = 0; i < count; i++) {
-                if (writemask2 & (1 << i))
-                    dev->vram[addr | i] = (vall.b[i] & svga->gdcreg[8]) | (dev->latch.b[i] & ~svga->gdcreg[8]);
-            }
-            break;
-        case 0x08: /* AND */
-            for (i = 0; i < count; i++) {
-                if (writemask2 & (1 << i))
-                    dev->vram[addr | i] = (vall.b[i] | ~svga->gdcreg[8]) & dev->latch.b[i];
-            }
-            break;
-        case 0x10: /* OR */
-            for (i = 0; i < count; i++) {
-                if (writemask2 & (1 << i))
-                    dev->vram[addr | i] = (vall.b[i] & svga->gdcreg[8]) | dev->latch.b[i];
-            }
-            break;
-        case 0x18: /* XOR */
-            for (i = 0; i < count; i++) {
-                if (writemask2 & (1 << i))
-                    dev->vram[addr | i] = (vall.b[i] & svga->gdcreg[8]) ^ dev->latch.b[i];
-            }
-            break;
-
-        default:
-            break;
-    }
-
-    if (reset_wm)
-        svga->gdcreg[8] = wm;
 }
 
 #ifdef ATI_8514_ULTRA
@@ -4780,25 +4430,90 @@ static void
 mach32_write(uint32_t addr, uint8_t val, void *priv)
 {
     mach_t *mach = (mach_t *) priv;
-    mach32_write_common(addr, val, 0, mach, &mach->svga);
+    svga_t *svga = &mach->svga;
+    ibm8514_t *dev = (ibm8514_t *) svga->dev8514;
+
+    xga_write_test(addr, val, svga);
+    addr = (addr & svga->banked_mask) + svga->write_bank;
+
+    if ((((dev->local & 0xff) >= 0x02) && !dev->vram_512k_8514) && ((mach->accel.ext_ge_config & 0x30) == 0x00)) {
+        addr <<= 1;
+        switch (addr & 0x06) {
+            case 0x00:
+            case 0x06:
+                mach32_write_common(addr, val & 0x0f, 0, mach, svga);
+                mach32_write_common(addr + 1, (val >> 4) & 0x0f, 0, mach, svga);
+                break;
+            case 0x02:
+                mach32_write_common(addr + 2, val & 0x0f, 0, mach, svga);
+                mach32_write_common(addr + 3, (val >> 4) & 0x0f, 0, mach, svga);
+                break;
+            case 0x04:
+                mach32_write_common(addr - 2, val & 0x0f, 0, mach, svga);
+                mach32_write_common(addr - 1, (val >> 4) & 0x0f, 0, mach, svga);
+                break;
+            default:
+                break;
+        }
+    } else
+        mach32_write_common(addr, val, 0, mach, svga);
 }
 
 static void
 mach32_writew(uint32_t addr, uint16_t val, void *priv)
 {
     mach_t *mach = (mach_t *) priv;
-    mach32_write_common(addr, val & 0xff, 0, mach, &mach->svga);
-    mach32_write_common(addr + 1, val >> 8, 0, mach, &mach->svga);
+    svga_t *svga = &mach->svga;
+    ibm8514_t *dev = (ibm8514_t *) svga->dev8514;
+
+    xga_write_test(addr, val, svga);
+    addr = (addr & svga->banked_mask) + svga->write_bank;
+
+    if ((((dev->local & 0xff) >= 0x02) && !dev->vram_512k_8514) && ((mach->accel.ext_ge_config & 0x30) == 0x00)) {
+        addr <<= 1;
+        if (addr & 0x04) {
+            mach32_write_common(addr - 2, val & 0x0f, 0, mach, svga);
+            mach32_write_common(addr - 1, (val >> 4) & 0x0f, 0, mach, svga);
+            mach32_write_common(addr + 2, (val >> 8) & 0x0f, 0, mach, svga);
+            mach32_write_common(addr + 3, (val >> 12) & 0x0f, 0, mach, svga);
+        } else {
+            mach32_write_common(addr, val & 0x0f, 0, mach, svga);
+            mach32_write_common(addr + 1, (val >> 4) & 0x0f, 0, mach, svga);
+            mach32_write_common(addr + 4, (val >> 8) & 0x0f, 0, mach, svga);
+            mach32_write_common(addr + 5, (val >> 12) & 0x0f, 0, mach, svga);
+        }
+    } else {
+        mach32_write_common(addr, val & 0xff, 0, mach, svga);
+        mach32_write_common(addr + 1, val >> 8, 0, mach, svga);
+    }
 }
 
 static void
 mach32_writel(uint32_t addr, uint32_t val, void *priv)
 {
     mach_t *mach = (mach_t *) priv;
-    mach32_write_common(addr, val & 0xff, 0, mach, &mach->svga);
-    mach32_write_common(addr + 1, val >> 8, 0, mach, &mach->svga);
-    mach32_write_common(addr + 2, val >> 16, 0, mach, &mach->svga);
-    mach32_write_common(addr + 3, val >> 24, 0, mach, &mach->svga);
+    svga_t *svga = &mach->svga;
+    ibm8514_t *dev = (ibm8514_t *) svga->dev8514;
+
+    xga_write_test(addr, val, svga);
+    addr = (addr & svga->banked_mask) + svga->write_bank;
+
+    if ((((dev->local & 0xff) >= 0x02) && !dev->vram_512k_8514) && ((mach->accel.ext_ge_config & 0x30) == 0x00)) {
+        addr <<= 1;
+        mach32_write_common(addr, val & 0x0f, 0, mach, svga);
+        mach32_write_common(addr + 1, (val >> 4) & 0x0f, 0, mach, svga);
+        mach32_write_common(addr + 4, (val >> 8) & 0x0f, 0, mach, svga);
+        mach32_write_common(addr + 5, (val >> 12) & 0x0f, 0, mach, svga);
+        mach32_write_common(addr + 2, (val >> 16) & 0x0f, 0, mach, svga);
+        mach32_write_common(addr + 3, (val >> 20) & 0x0f, 0, mach, svga);
+        mach32_write_common(addr + 6, (val >> 24) & 0x0f, 0, mach, svga);
+        mach32_write_common(addr + 7, (val >> 28) & 0x0f, 0, mach, svga);
+    } else {
+        mach32_write_common(addr, val & 0xff, 0, mach, svga);
+        mach32_write_common(addr + 1, val >> 8, 0, mach, svga);
+        mach32_write_common(addr + 2, val >> 16, 0, mach, svga);
+        mach32_write_common(addr + 3, val >> 24, 0, mach, svga);
+    }
 }
 
 static __inline void
@@ -4808,10 +4523,25 @@ mach32_writew_linear(uint32_t addr, uint16_t val, mach_t *mach)
     ibm8514_t *dev = (ibm8514_t *) svga->dev8514;
 
     cycles -= svga->monitor->mon_video_timing_write_w;
+    if (!dev->vram_512k_8514 && ((mach->accel.ext_ge_config & 0x30) == 0x00))
+        addr <<= 1;
 
     addr &= dev->vram_mask;
     dev->changedvram[addr >> 12] = svga->monitor->mon_changeframecount;
-    *(uint16_t *) &dev->vram[addr] = val;
+    if (!dev->vram_512k_8514 && ((mach->accel.ext_ge_config & 0x30) == 0x00)) {
+        if (addr & 0x04) {
+            dev->vram[addr - 2] = val & 0x0f;
+            dev->vram[addr - 1] = (val >> 4) & 0x0f;
+            dev->vram[addr + 2] = (val >> 8) & 0x0f;
+            dev->vram[addr + 3] = (val >> 12) & 0x0f;
+        } else {
+            dev->vram[addr] = val & 0x0f;
+            dev->vram[addr + 1] = (val >> 4) & 0x0f;
+            dev->vram[addr + 4] = (val >> 8) & 0x0f;
+            dev->vram[addr + 5] = (val >> 12) & 0x0f;
+        }
+    } else
+        *(uint16_t *) &dev->vram[addr] = val;
 }
 
 static __inline void
@@ -4822,9 +4552,22 @@ mach32_writel_linear(uint32_t addr, uint32_t val, mach_t *mach)
 
     cycles -= svga->monitor->mon_video_timing_write_l;
 
+    if (!dev->vram_512k_8514 && ((mach->accel.ext_ge_config & 0x30) == 0x00))
+        addr <<= 1;
+
     addr &= dev->vram_mask;
     dev->changedvram[addr >> 12] = svga->monitor->mon_changeframecount;
-    *(uint32_t *) &dev->vram[addr] = val;
+    if (!dev->vram_512k_8514 && ((mach->accel.ext_ge_config & 0x30) == 0x00)) {
+        dev->vram[addr] = val & 0x0f;
+        dev->vram[addr + 1] = (val >> 4) & 0x0f;
+        dev->vram[addr + 4] = (val >> 8) & 0x0f;
+        dev->vram[addr + 5] = (val >> 12) & 0x0f;
+        dev->vram[addr + 2] = (val >> 16) & 0x0f;
+        dev->vram[addr + 3] = (val >> 20) & 0x0f;
+        dev->vram[addr + 6] = (val >> 24) & 0x0f;
+        dev->vram[addr + 7] = (val >> 28) & 0x0f;
+    } else
+        *(uint32_t *) &dev->vram[addr] = val;
 }
 
 static __inline uint8_t
@@ -4835,17 +4578,37 @@ mach32_read_common(uint32_t addr, int linear, mach_t *mach, svga_t *svga)
     int              readplane  = svga->readplane;
     uint8_t          count;
     uint8_t          temp;
-    uint8_t          ret;
+    uint8_t          ret = 0x00;
 
     cycles -= svga->monitor->mon_video_timing_read_b;
 
     if (linear) {
-        return dev->vram[addr & dev->vram_mask];
-    } else {
-        (void) xga_read_test(addr, svga);
-        addr = mach32_decode_addr(svga, addr, 0);
-        if (addr == 0xffffffff)
-            return 0xff;
+        if (!dev->vram_512k_8514 && ((mach->accel.ext_ge_config & 0x30) == 0x00))
+            addr <<= 1;
+
+        addr &= dev->vram_mask;
+        if (!dev->vram_512k_8514 && ((mach->accel.ext_ge_config & 0x30) == 0x00)) {
+            switch (addr & 0x06) {
+                case 0x00:
+                case 0x06:
+                    ret = dev->vram[addr] & 0x0f;
+                    ret |= (dev->vram[addr + 1] << 4);
+                    break;
+                case 0x02:
+                    ret = dev->vram[addr + 2] & 0x0f;
+                    ret |= (dev->vram[addr + 3] << 4);
+                    break;
+                case 0x04:
+                    ret = dev->vram[addr - 2] & 0x0f;
+                    ret |= (dev->vram[addr - 1] << 4);
+                    break;
+                default:
+                    break;
+            }
+        } else
+            ret = dev->vram[addr];
+
+        return ret;
     }
 
     count = 2;
@@ -4853,7 +4616,7 @@ mach32_read_common(uint32_t addr, int linear, mach_t *mach, svga_t *svga)
     latch_addr = (addr << count) & svga->decode_mask;
     count      = (1 << count);
 
-    if ((svga->chain4 && (svga->packed_chain4 || svga->force_old_addr)) || svga->fb_only) {
+    if (svga->chain4) {
         addr &= svga->decode_mask;
         if (addr >= dev->vram_size)
             return 0xff;
@@ -4861,9 +4624,6 @@ mach32_read_common(uint32_t addr, int linear, mach_t *mach, svga_t *svga)
         for (uint8_t i = 0; i < count; i++)
             dev->latch.b[i] = dev->vram[latch_addr | i];
         return dev->vram[addr & dev->vram_mask];
-    } else if (svga->chain4 && !svga->force_old_addr) {
-        readplane = addr & 3;
-        addr      = ((addr & 0xfffc) << 2) | ((addr & 0x30000) >> 14) | (addr & ~0x3ffff);
     } else if (svga->chain2_read) {
         readplane = (readplane & 2) | (addr & 1);
         addr &= ~1;
@@ -4896,6 +4656,7 @@ mach32_read_common(uint32_t addr, int linear, mach_t *mach, svga_t *svga)
 
     addr &= dev->vram_mask;
 
+    mach_log("ReadMode=%02x.\n", svga->readmode);
     if (svga->readmode) {
         temp = 0xff;
 
@@ -4959,9 +4720,35 @@ static uint8_t
 mach32_read(uint32_t addr, void *priv)
 {
     mach_t *mach = (mach_t *) priv;
-    uint8_t          ret;
+    svga_t *svga = &mach->svga;
+    ibm8514_t *dev = (ibm8514_t *) svga->dev8514;
+    uint8_t ret = 0x00;
+
+    (void) xga_read_test(addr, svga);
+    addr = (addr & svga->banked_mask) + svga->read_bank;
+
+    if ((((dev->local & 0xff) >= 0x02) && !dev->vram_512k_8514) && ((mach->accel.ext_ge_config & 0x30) == 0x00)) {
+        addr <<= 1;
+        switch (addr & 0x06) {
+            case 0x00:
+            case 0x06:
+                ret = mach32_read_common(addr, 0, mach, svga) & 0x0f;
+                ret |= (mach32_read_common(addr + 1, 0, mach, svga) << 4);
+                break;
+            case 0x02:
+                ret = mach32_read_common(addr + 2, 0, mach, svga) & 0x0f;
+                ret |= (mach32_read_common(addr + 3, 0, mach, svga) << 4);
+                break;
+            case 0x04:
+                ret = mach32_read_common(addr - 2, 0, mach, svga) & 0x0f;
+                ret |= (mach32_read_common(addr - 1, 0, mach, svga) << 4);
+                break;
+            default:
+                break;
+        }
+    } else
+        ret = mach32_read_common(addr, 0, mach, svga);
 
-    ret  = mach32_read_common(addr, 0, mach, &mach->svga);
     return ret;
 }
 
@@ -4969,10 +4756,30 @@ static uint16_t
 mach32_readw(uint32_t addr, void *priv)
 {
     mach_t *mach = (mach_t *) priv;
-    uint16_t         ret;
+    svga_t *svga = &mach->svga;
+    ibm8514_t *dev = (ibm8514_t *) svga->dev8514;
+    uint16_t ret;
 
-    ret  = mach32_read_common(addr, 0, mach, &mach->svga);
-    ret  |= (mach32_read_common(addr + 1, 0, mach, &mach->svga) << 8);
+    (void) xga_read_test(addr, svga);
+    addr = (addr & svga->banked_mask) + svga->read_bank;
+
+    if ((((dev->local & 0xff) >= 0x02) && !dev->vram_512k_8514) && ((mach->accel.ext_ge_config & 0x30) == 0x00)) {
+        addr <<= 1;
+        if (addr & 0x04) {
+            ret = mach32_read_common(addr - 2, 0, mach, svga) & 0x0f;
+            ret |= (mach32_read_common(addr - 1, 0, mach, svga) << 4);
+            ret |= (mach32_read_common(addr + 2, 0, mach, svga) << 8);
+            ret |= (mach32_read_common(addr + 3, 0, mach, svga) << 12);
+        } else {
+            ret = mach32_read_common(addr, 0, mach, svga) & 0x0f;
+            ret |= (mach32_read_common(addr + 1, 0, mach, svga) << 4);
+            ret |= (mach32_read_common(addr + 4, 0, mach, svga) << 8);
+            ret |= (mach32_read_common(addr + 5, 0, mach, svga) << 12);
+        }
+    } else {
+        ret = mach32_read_common(addr, 0, mach, svga);
+        ret |= (mach32_read_common(addr + 1, 0, mach, svga) << 8);
+    }
     return ret;
 }
 
@@ -4980,12 +4787,29 @@ static uint32_t
 mach32_readl(uint32_t addr, void *priv)
 {
     mach_t *mach = (mach_t *) priv;
-    uint32_t         ret;
+    svga_t *svga = &mach->svga;
+    ibm8514_t *dev = (ibm8514_t *) svga->dev8514;
+    uint32_t ret;
 
-    ret  = mach32_read_common(addr, 0, mach, &mach->svga);
-    ret  |= (mach32_read_common(addr + 1, 0, mach, &mach->svga) << 8);
-    ret  |= (mach32_read_common(addr + 2, 0, mach, &mach->svga) << 16);
-    ret  |= (mach32_read_common(addr + 3, 0, mach, &mach->svga) << 24);
+    (void) xga_read_test(addr, svga);
+    addr = (addr & svga->banked_mask) + svga->read_bank;
+
+    if ((((dev->local & 0xff) >= 0x02) && !dev->vram_512k_8514) && ((mach->accel.ext_ge_config & 0x30) == 0x00)) {
+        addr <<= 1;
+        ret = mach32_read_common(addr, 0, mach, svga) & 0x0f;
+        ret |= (mach32_read_common(addr + 1, 0, mach, svga) << 4);
+        ret |= (mach32_read_common(addr + 4, 0, mach, svga) << 8);
+        ret |= (mach32_read_common(addr + 5, 0, mach, svga) << 12);
+        ret |= (mach32_read_common(addr + 2, 0, mach, svga) << 16);
+        ret |= (mach32_read_common(addr + 3, 0, mach, svga) << 20);
+        ret |= (mach32_read_common(addr + 6, 0, mach, svga) << 24);
+        ret |= (mach32_read_common(addr + 7, 0, mach, svga) << 28);
+    } else {
+        ret = mach32_read_common(addr, 0, mach, svga);
+        ret |= (mach32_read_common(addr + 1, 0, mach, svga) << 8);
+        ret |= (mach32_read_common(addr + 2, 0, mach, svga) << 16);
+        ret |= (mach32_read_common(addr + 3, 0, mach, svga) << 24);
+    }
     return ret;
 }
 
@@ -4994,10 +4818,27 @@ mach32_readw_linear(uint32_t addr, mach_t *mach)
 {
     svga_t          *svga       = &mach->svga;
     ibm8514_t       *dev        = (ibm8514_t *) svga->dev8514;
+    uint16_t         ret;
 
     cycles -= svga->monitor->mon_video_timing_read_w;
+    if (!dev->vram_512k_8514 && ((mach->accel.ext_ge_config & 0x30) == 0x00)) {
+        addr <<= 1;
+        addr &= dev->vram_mask;
+        if (addr & 0x04) {
+            ret = dev->vram[addr - 2] & 0x0f;
+            ret |= (dev->vram[addr - 1] << 4);
+            ret |= (dev->vram[addr + 2] << 8);
+            ret |= (dev->vram[addr + 3] << 12);
+        } else {
+            ret = dev->vram[addr] & 0x0f;
+            ret |= (dev->vram[addr + 1] << 4);
+            ret |= (dev->vram[addr + 4] << 8);
+            ret |= (dev->vram[addr + 5] << 12);
+        }
+    } else
+        ret = *(uint16_t *) &dev->vram[addr & dev->vram_mask];
 
-    return *(uint16_t *) &dev->vram[addr & dev->vram_mask];
+    return ret;
 }
 
 static __inline uint32_t
@@ -5005,10 +4846,25 @@ mach32_readl_linear(uint32_t addr, mach_t *mach)
 {
     svga_t          *svga       = &mach->svga;
     ibm8514_t       *dev        = (ibm8514_t *) svga->dev8514;
+    uint32_t         ret;
 
     cycles -= svga->monitor->mon_video_timing_read_l;
 
-    return *(uint32_t *) &dev->vram[addr & dev->vram_mask];
+    if (!dev->vram_512k_8514 && ((mach->accel.ext_ge_config & 0x30) == 0x00)) {
+        addr <<= 1;
+        addr &= dev->vram_mask;
+        ret = dev->vram[addr] & 0x0f;
+        ret |= (dev->vram[addr + 1] << 4);
+        ret |= (dev->vram[addr + 4] << 8);
+        ret |= (dev->vram[addr + 5] << 12);
+        ret |= (dev->vram[addr + 2] << 16);
+        ret |= (dev->vram[addr + 3] << 20);
+        ret |= (dev->vram[addr + 6] << 24);
+        ret |= (dev->vram[addr + 7] << 28);
+    } else
+        ret = *(uint32_t *) &dev->vram[addr & dev->vram_mask];
+
+    return ret;
 }
 
 static void
@@ -5029,10 +4885,10 @@ mach32_ap_writeb(uint32_t addr, uint8_t val, void *priv)
             mach_accel_outb(0x02e8 + (addr & 1) + (port_dword << 8), val, mach);
         }
     } else {
-        mach_log("Linear WORDB Write=%08x, val=%02x.\n", addr, val);
-        if (dev->on[0] || dev->on[1])
+        if (dev->on) {
+            mach_log("Linear WORDB Write=%08x, val=%02x, ON=%x, dpconfig=%04x, apsize=%08x.\n", addr & dev->vram_mask, val, dev->on, mach->accel.dp_config, mach->ap_size << 20);
             mach32_write_common(addr, val, 1, mach, svga);
-        else
+        } else
             svga_write_linear(addr, val, svga);
     }
 }
@@ -5055,8 +4911,8 @@ mach32_ap_writew(uint32_t addr, uint16_t val, void *priv)
             mach_accel_outw(0x02e8 + (port_dword << 8), val, mach);
         }
     } else {
-        mach_log("Linear WORDW Write=%08x, val=%04x.\n", addr, val);
-        if (dev->on[0] || dev->on[1]) {
+        if (dev->on) {
+            mach_log("Linear WORDW Write=%08x, val=%04x, ON=%x, dpconfig=%04x, apsize=%08x.\n", addr & dev->vram_mask, val, dev->on, mach->accel.dp_config, mach->ap_size << 20);
             mach32_writew_linear(addr, val, mach);
         } else
             svga_writew_linear(addr, val, svga);
@@ -5083,8 +4939,8 @@ mach32_ap_writel(uint32_t addr, uint32_t val, void *priv)
             mach_accel_outw(0x02e8 + (port_dword << 8) + 4, val >> 16, mach);
         }
     } else {
-        mach_log("Linear WORDL Write=%08x, val=%08x.\n", addr, val);
-        if (dev->on[0] || dev->on[1]) {
+        if (dev->on) {
+            mach_log("Linear WORDL Write=%08x, val=%08x, ON=%x, dpconfig=%04x, apsize=%08x.\n", addr & dev->vram_mask, val, dev->on, mach->accel.dp_config, mach->ap_size << 20);
             mach32_writel_linear(addr, val, mach);
         } else
             svga_writel_linear(addr, val, svga);
@@ -5107,7 +4963,7 @@ mach32_ap_readb(uint32_t addr, void *priv)
         else
             temp = mach_accel_inb(0x02e8 + (addr & 1) + (port_dword << 8), mach);
     } else {
-        if (dev->on[0] || dev->on[1])
+        if (dev->on)
             temp = mach32_read_common(addr, 1, mach, svga);
         else
             temp = svga_read_linear(addr, svga);
@@ -5134,7 +4990,7 @@ mach32_ap_readw(uint32_t addr, void *priv)
         else
             temp = mach_accel_inw(0x02e8 + (port_dword << 8), mach);
     } else {
-        if (dev->on[0] || dev->on[1]) {
+        if (dev->on) {
             temp = mach32_readw_linear(addr, mach);
         } else
             temp = svga_readw_linear(addr, svga);
@@ -5164,12 +5020,12 @@ mach32_ap_readl(uint32_t addr, void *priv)
             temp |= (mach_accel_inw(0x02e8 + (port_dword << 8) + 4, mach) << 8);
         }
     } else {
-        if (dev->on[0] || dev->on[1]) {
+        if (dev->on) {
             temp = mach32_readl_linear(addr, mach);
         } else
             temp = svga_readl_linear(addr, svga);
 
-        mach_log("Linear WORDL Read=%08x, ret=%08x, ON0=%d, ON1=%d.\n", addr, temp, dev->on[0], dev->on[1]);
+        mach_log("Linear WORDL Read=%08x, ret=%08x, ON%d.\n", addr, temp, dev->on);
     }
 
     return temp;
@@ -5186,7 +5042,7 @@ mach32_updatemapping(mach_t *mach, svga_t *svga)
         return;
     }
 
-    if (mach->regs[0xbd] & 4) {
+    if (mach->regs[0xbd] & 0x04) {
         mem_mapping_set_addr(&svga->mapping, 0xa0000, 0x20000);
         svga->banked_mask = 0xffff;
     } else {
@@ -5218,18 +5074,21 @@ mach32_updatemapping(mach_t *mach, svga_t *svga)
         if (((mach->memory_aperture & 3) == 1) && !mach->pci_bus) {
             /*1 MB aperture*/
             mach->ap_size = 1;
+            mach_log("Linear Enabled APSIZE=1.\n");
             mem_mapping_set_addr(&mach->mmio_linear_mapping, mach->linear_base, mach->ap_size << 20);
         } else {
             /*4 MB aperture*/
             mach->ap_size = 4;
+            mach_log("Linear Enabled APSIZE=4.\n");
             mem_mapping_set_addr(&mach->mmio_linear_mapping, mach->linear_base, mach->ap_size << 20);
         }
     } else {
         mach->ap_size = 4;
+        mach_log("Linear Disabled APSIZE=4.\n");
         mem_mapping_disable(&mach->mmio_linear_mapping);
     }
-    if ((dev->on[0] || dev->on[1]) && (mach->ext_on[0] || mach->ext_on[1]) && (dev->vendor_mode[0] || dev->vendor_mode[1])) {
-        mach_log("ExtON.\n");
+    if (dev->on && ((dev->local & 0xff) >= 0x02)) {
+        mach_log("Mach32 banked mapping.\n");
 #ifdef ATI_8514_ULTRA
         if (svga->ext8514 != NULL) {
             mem_mapping_set_handler(&svga->mapping, ati8514_read, ati8514_readw, ati8514_readl, ati8514_write, ati8514_writew, ati8514_writel);
@@ -5241,7 +5100,7 @@ mach32_updatemapping(mach_t *mach, svga_t *svga)
             mem_mapping_set_p(&svga->mapping, mach);
         }
     } else {
-        mach_log("ExtOFF.\n");
+        mach_log("IBM compatible banked mapping.\n");
         mem_mapping_set_handler(&svga->mapping, svga_read, svga_readw, svga_readl, svga_write, svga_writew, svga_writel);
         mem_mapping_set_p(&svga->mapping, svga);
     }
@@ -5267,6 +5126,7 @@ mach32_hwcursor_draw(svga_t *svga, int displine)
         case 8:
             color0 = dev->pallook[mach->cursor_col_0];
             color1 = dev->pallook[mach->cursor_col_1];
+            mach_log("4/8BPP: Color0=%08x, Color1=%08x.\n", color0, color1);
             break;
         case 15:
             color0 = video_15to32[((mach->ext_cur_col_0_r << 16) | (mach->ext_cur_col_0_g << 8) | mach->cursor_col_0) & 0xffff];
@@ -5288,6 +5148,7 @@ mach32_hwcursor_draw(svga_t *svga, int displine)
 
     for (int x = 0; x < 64; x += 8) {
         dat = dev->vram[dev->hwcursor_latch.addr & dev->vram_mask] | (dev->vram[(dev->hwcursor_latch.addr + 1) & dev->vram_mask] << 8);
+
         for (int xx = 0; xx < 8; xx++) {
             comb = (dat >> (xx << 1)) & 0x03;
 
@@ -5598,13 +5459,11 @@ mach_mca_reset(void *priv)
     svga_t    *svga = &mach->svga;
     ibm8514_t *dev  = (ibm8514_t *) svga->dev8514;
 
-    mem_mapping_disable(&mach->bios_rom.mapping);
-    mem_mapping_disable(&mach->bios_rom2.mapping);
     mach_log("MCA reset.\n");
-    dev->on[0] = 0;
-    dev->on[1] = 0;
+    dev->on = 0;
     vga_on = 1;
     mach_mca_write(0x102, 0, mach);
+    timer_set_callback(&svga->timer, svga_poll);
 }
 
 #ifdef ATI_8514_ULTRA
@@ -5794,8 +5653,9 @@ mach8_init(const device_t *info)
     dev->type        = info->flags;
     dev->local       = info->local & 0xff;
     mach->has_bios   = !(info->local & 0xff00);
-    mach->memory     = device_get_config_int("memory");
     mach->ramdac_type = mach->pci_bus ? device_get_config_int("ramdac") : 1;
+    dev->vram_amount = device_get_config_int("memory");
+    dev->vram_512k_8514 = dev->vram_amount == 512;
 
     if ((dev->local & 0xff) >= 0x02) {
         if (mach->pci_bus) {
@@ -5833,12 +5693,12 @@ mach8_init(const device_t *info)
                  0, MEM_MAPPING_EXTERNAL);
 
     if ((dev->local & 0xff) >= 0x02) {
-        svga_init(info, svga, mach, mach->memory << 10, /*default: 2MB for Mach32*/
+        svga_init(info, svga, mach, dev->vram_amount << 10, /*default: 2MB for Mach32*/
                       mach_recalctimings,
                       mach_in, mach_out,
                       mach32_hwcursor_draw,
                       NULL);
-        dev->vram_size   = mach->memory << 10;
+        dev->vram_size   = dev->vram_amount << 10;
         dev->vram        = calloc(dev->vram_size, 1);
         dev->changedvram = calloc((dev->vram_size >> 12) + 1, 1);
         dev->vram_mask   = dev->vram_size - 1;
@@ -5886,20 +5746,22 @@ mach8_init(const device_t *info)
                       mach_in, mach_out,
                       NULL,
                       NULL);
-        dev->vram_size   = (1024 << 10);
+        dev->vram_size   = (dev->vram_amount << 10);
         dev->vram        = calloc(dev->vram_size, 1);
         dev->changedvram = calloc((dev->vram_size >> 12) + 1, 1);
         dev->vram_mask   = dev->vram_size - 1;
         video_inform(VIDEO_FLAG_TYPE_8514, &timing_gfxultra_isa);
-        mach->config1 = 0x01 | 0x02 | 0x20 | 0x08 | 0x80;
+        mach->config1 = 0x01 | 0x02 | 0x08 | 0x80;
+        if (dev->vram_amount >= 1024)
+            mach->config1 |= 0x20;
+
         mach->config2 = 0x02;
         svga->clock_gen = device_add(&ati18811_0_device);
     }
     dev->bpp            = 0;
     svga->getclock      = ics2494_getclock;
 
-    dev->on[0] = 0;
-    dev->on[1] = 0;
+    dev->on = 0;
     dev->ext_pitch = 1024;
     dev->ext_crt_pitch = 0x80;
     dev->accel_bpp = 8;
@@ -5915,6 +5777,7 @@ mach8_init(const device_t *info)
 
     if ((dev->local & 0xff) >= 0x02) {
         svga->decode_mask     = (4 << 20) - 1;
+        mach->accel.cmd_type  = -1;
         mach->cursor_col_1    = 0xff;
         mach->ext_cur_col_1_r = 0xff;
         mach->ext_cur_col_1_g = 0xff;
@@ -5954,23 +5817,25 @@ ati8514_init(svga_t *svga, void *ext8514, void *dev8514)
     mach_t *mach = (mach_t *) ext8514;
     ibm8514_t *dev = (ibm8514_t *) dev8514;
 
-    dev->on[0] = 0;
-    dev->on[1] = 0;
+    dev->on = 0;
     dev->ext_pitch = 1024;
     dev->ext_crt_pitch = 0x80;
     dev->accel_bpp = 8;
     dev->rowoffset = 0x80;
-    dev->hdisp = 1024;
-    dev->vdisp = 768;
+    dev->hdisp = 0;
+    dev->vdisp = 0;
 
     io_sethandler(0x02ea, 4, ati8514_in, NULL, NULL, ati8514_out, NULL, NULL, svga);
     ati8514_io_set(svga);
     mach->mca_bus = !!(dev->type & DEVICE_MCA);
 
     if (mach->mca_bus)
-        mach->config1 = 0x02 | 0x04 | 0x08 | 0x20 | 0x80;
+        mach->config1 = 0x02 | 0x04;
     else
-        mach->config1 = 0x02 | 0x08 | 0x20 | 0x80;
+        mach->config1 = 0x02 | 0x2000;
+
+    if (dev->vram_amount >= 1024)
+        mach->config1 |= 0x20;
 
     mach->config2 = 0x01 | 0x02;
 }
@@ -6042,6 +5907,32 @@ mach_force_redraw(void *priv)
     svga->fullchange = svga->monitor->mon_changeframecount;
 }
 
+// clang-format off
+static const device_config_t mach8_config[] = {
+    {
+        .name = "memory",
+        .description = "Memory size",
+        .type = CONFIG_SELECTION,
+        .default_int = 1024,
+        .selection = {
+            {
+                .description = "512 KB",
+                .value = 512
+            },
+            {
+                .description = "1 MB",
+                .value = 1024
+            },
+            {
+                .description = ""
+            }
+        }
+    },
+    {
+        .type = CONFIG_END
+    }
+};
+
 // clang-format off
 static const device_config_t mach32_config[] = {
     {
@@ -6140,7 +6031,7 @@ const device_t mach8_vga_isa_device = {
     { .available = mach8_vga_available },
     .speed_changed = mach_speed_changed,
     .force_redraw = mach_force_redraw,
-    .config = NULL
+    .config = mach8_config
 };
 
 const device_t mach32_isa_device = {
diff --git a/src/video/vid_bt481_ramdac.c b/src/video/vid_bt481_ramdac.c
new file mode 100644
index 000000000..6cf5f2b1b
--- /dev/null
+++ b/src/video/vid_bt481_ramdac.c
@@ -0,0 +1,160 @@
+/*
+ * 86Box    A hypervisor and IBM PC system emulator that specializes in
+ *          running old operating systems and software designed for IBM
+ *          PC systems and compatibles from 1981 through fairly recent
+ *          system designs based on the PCI bus.
+ *
+ *          This file is part of the 86Box distribution.
+ *
+ *          Emulation of the Brooktree BT481 true colour RAMDAC
+ *          family.
+ *
+ *
+ *
+ * Authors: TheCollector1995.
+ *
+ *          Copyright 2024 TheCollector1995.
+ */
+#include <stdio.h>
+#include <stdint.h>
+#include <stdlib.h>
+#include <string.h>
+#include <wchar.h>
+#include <86box/86box.h>
+#include <86box/device.h>
+#include <86box/mem.h>
+#include <86box/timer.h>
+#include <86box/video.h>
+#include <86box/vid_svga.h>
+
+typedef struct bt481_ramdac_t {
+    int     state;
+    uint8_t cmd;
+} bt481_ramdac_t;
+
+static void
+bt481_ramdac_command(uint8_t val, void *priv, svga_t *svga)
+{
+    bt481_ramdac_t *ramdac  = (bt481_ramdac_t *) priv;
+    ramdac->cmd             = val;
+    pclog("RAMDAC CMD=%02x.\n", val);
+    switch ((ramdac->cmd >> 4) & 0x0f) {
+        default:
+        case 0x00:
+            svga->bpp = 8;
+            break;
+        case 0x08:
+        case 0x0a:
+            svga->bpp = 15;
+            break;
+        case 0x09:
+        case 0x0c:
+            svga->bpp = 16;
+            break;
+        case 0x0e:
+        case 0x0f:
+            svga->bpp = 24;
+            break;
+    }
+    svga_recalctimings(svga);
+}
+
+void
+bt481_ramdac_out(uint16_t addr, int rs2, uint8_t val, void *priv, svga_t *svga)
+{
+    bt481_ramdac_t *ramdac = (bt481_ramdac_t *) priv;
+    uint8_t             rs = (addr & 0x03) | ((!!rs2) << 2);
+
+    switch (rs) {
+        case 0x00:
+        case 0x01:
+        case 0x03:
+        case 0x04:
+        case 0x05:
+        case 0x07:
+            svga_out(addr, val, svga);
+            ramdac->state = 0;
+            break;
+        case 0x02:
+            pclog("RAMDAC Write State=%x.\n", ramdac->state);
+            switch (ramdac->state) {
+                case 4:
+                    bt481_ramdac_command(val, ramdac, svga);
+                    break;
+                default:
+                    svga_out(addr, val, svga);
+                    break;
+            }
+            break;
+        case 0x06:
+            bt481_ramdac_command(val, ramdac, svga);
+            ramdac->state = 0;
+            break;
+
+        default:
+            break;
+    }
+}
+
+uint8_t
+bt481_ramdac_in(uint16_t addr, int rs2, void *priv, svga_t *svga)
+{
+    bt481_ramdac_t * ramdac = (bt481_ramdac_t *) priv;
+    uint8_t          temp   = 0xff;
+    uint8_t          rs     = (addr & 0x03) | ((!!rs2) << 2);
+
+    switch (rs) {
+        case 0x02:
+        case 0x06:
+            switch (ramdac->state) {
+                case 4:
+                    temp = ramdac->cmd;
+                    break;
+                default:
+                    temp = svga_in(addr, svga);
+                    ramdac->state++;
+                    break;
+            }
+            break;
+
+        default:
+            temp          = svga_in(addr, svga);
+            ramdac->state = 0;
+            break;
+    }
+
+    pclog("RAMDAC IN=%02x, ret=%02x.\n", rs, temp);
+    return temp;
+}
+
+static void *
+bt481_ramdac_init(const device_t *info)
+{
+    bt481_ramdac_t *ramdac = (bt481_ramdac_t *) malloc(sizeof(bt481_ramdac_t));
+    memset(ramdac, 0, sizeof(bt481_ramdac_t));
+
+    return ramdac;
+}
+
+static void
+bt481_ramdac_close(void *priv)
+{
+    bt481_ramdac_t *ramdac = (bt481_ramdac_t *) priv;
+
+    if (ramdac)
+        free(ramdac);
+}
+
+const device_t bt481_ramdac_device = {
+    .name          = "Brooktree Bt481 RAMDAC",
+    .internal_name = "bt481_ramdac",
+    .flags         = 0,
+    .local         = 0,
+    .init          = bt481_ramdac_init,
+    .close         = bt481_ramdac_close,
+    .reset         = NULL,
+    { .available = NULL },
+    .speed_changed = NULL,
+    .force_redraw  = NULL,
+    .config        = NULL
+};
diff --git a/src/video/vid_paradise.c b/src/video/vid_paradise.c
index 73902345d..053ef2651 100644
--- a/src/video/vid_paradise.c
+++ b/src/video/vid_paradise.c
@@ -303,7 +303,7 @@ paradise_remap(paradise_t *paradise)
 void
 paradise_recalctimings(svga_t *svga)
 {
-    const paradise_t *paradise = (paradise_t *) svga->priv;
+    paradise_t *paradise = (paradise_t *) svga->priv;
 
     svga->lowres = !(svga->gdcreg[0x0e] & 0x01);
 
@@ -337,15 +337,11 @@ paradise_recalctimings(svga_t *svga)
                     svga->hdisp >>= 1;
                     if (svga->hdisp == 788)
                         svga->hdisp += 12;
-                    if (svga->hdisp == 800)
-                        svga->ma_latch -= 3;
                 } else if (svga->bpp == 15) {
                     svga->render = svga_render_15bpp_highres;
                     svga->hdisp >>= 1;
                     if (svga->hdisp == 788)
                         svga->hdisp += 12;
-                    if (svga->hdisp == 800)
-                        svga->ma_latch -= 3;
                 } else
                     svga->render = svga_render_8bpp_highres;
 
@@ -366,6 +362,11 @@ paradise_write(uint32_t addr, uint8_t val, void *priv)
     uint32_t    prev_addr;
     uint32_t    prev_addr2;
 
+    if (!(svga->attrregs[0x10] & 0x40)) {
+        svga_write(addr, val, svga);
+        return;
+    }
+
     addr = (addr & 0x7fff) + paradise->write_bank[(addr >> 15) & 3];
 
     /*Could be done in a better way but it works.*/
@@ -399,6 +400,11 @@ paradise_writew(uint32_t addr, uint16_t val, void *priv)
     uint32_t    prev_addr;
     uint32_t    prev_addr2;
 
+    if (!(svga->attrregs[0x10] & 0x40)) {
+        svga_writew(addr, val, svga);
+        return;
+    }
+
     addr = (addr & 0x7fff) + paradise->write_bank[(addr >> 15) & 3];
 
     /*Could be done in a better way but it works.*/
@@ -432,6 +438,9 @@ paradise_read(uint32_t addr, void *priv)
     uint32_t    prev_addr;
     uint32_t    prev_addr2;
 
+    if (!(svga->attrregs[0x10] & 0x40))
+        return svga_read(addr, svga);
+
     addr = (addr & 0x7fff) + paradise->read_bank[(addr >> 15) & 3];
 
     /*Could be done in a better way but it works.*/
@@ -465,6 +474,9 @@ paradise_readw(uint32_t addr, void *priv)
     uint32_t    prev_addr;
     uint32_t    prev_addr2;
 
+    if (!(svga->attrregs[0x10] & 0x40))
+        return svga_readw(addr, svga);
+
     addr = (addr & 0x7fff) + paradise->read_bank[(addr >> 15) & 3];
 
     /*Could be done in a better way but it works.*/
diff --git a/src/video/vid_s3.c b/src/video/vid_s3.c
index 2b14284c7..c91eab589 100644
--- a/src/video/vid_s3.c
+++ b/src/video/vid_s3.c
@@ -253,6 +253,7 @@ typedef struct s3_t {
         uint16_t multifunc_cntl;
         uint16_t multifunc[16];
         uint8_t  pix_trans[4];
+        uint16_t pix_trans_val;
         int      ssv_state;
 
         int16_t  cx, cy;
@@ -278,14 +279,8 @@ typedef struct s3_t {
         uint8_t  frgd_color_actual[2];
         uint8_t  bkgd_color_actual[2];
         uint8_t  wrt_mask_actual[2];
-        uint8_t  rd_mask_actual[2];
-        uint8_t *pix_trans_ptr;
-        int      pix_trans_ptr_cnt;
-        int      pix_trans_x_count;
-        int      pix_trans_x_count2;
         int      color_16bit_check;
-        int      color_16bit_check_rectfill;
-        uint16_t minus, srcminus;
+        int16_t  minus;
 
         /*For non-threaded FIFO*/
         int setup_fifo_slot;
@@ -412,6 +407,7 @@ static uint32_t s3_accel_read_l(uint32_t addr, void *priv);
 static void    s3_out(uint16_t addr, uint8_t val, void *priv);
 static uint8_t s3_in(uint16_t addr, void *priv);
 
+static void     s3_accel_out_fifo(s3_t *s3, uint16_t port, uint8_t val);
 static void     s3_accel_out(uint16_t port, uint8_t val, void *priv);
 static void     s3_accel_out_w(uint16_t port, uint16_t val, void *priv);
 static void     s3_accel_out_l(uint16_t port, uint32_t val, void *priv);
@@ -508,11 +504,10 @@ s3_update_irqs(s3_t *s3)
     if (!s3->pci)
         return;
 
-    if (s3->subsys_cntl & s3->subsys_stat & INT_MASK) {
+    if (s3->subsys_cntl & s3->subsys_stat & INT_MASK)
         pci_set_irq(s3->pci_slot, PCI_INTA, &s3->irq_state);
-    } else {
+    else
         pci_clear_irq(s3->pci_slot, PCI_INTA, &s3->irq_state);
-    }
 }
 
 void        s3_accel_start(int count, int cpu_input, uint32_t mix_dat, uint32_t cpu_dat, void *priv);
@@ -536,21 +531,20 @@ static void s3_visionx68_video_engine_op(uint32_t cpu_dat, s3_t *s3);
     }
 
 #define READ_PIXTRANS_BYTE_IO(n) \
-    s3->accel.pix_trans[n] = svga->vram[dword_remap(svga, (s3->accel.dest + s3->accel.cx + n - s3->accel.minus)) & s3->vram_mask];
+    s3->accel.pix_trans[n] = svga->vram[dword_remap(svga, (s3->accel.dest + s3->accel.cx - s3->accel.minus + n)) & s3->vram_mask];
 
 #define READ_PIXTRANS_BYTE_MM \
     temp = svga->vram[dword_remap(svga, (s3->accel.dest + s3->accel.cx)) & s3->vram_mask];
 
 #define READ_PIXTRANS_WORD                                                                                 \
-    if ((s3->bpp == 0) && !s3->color_16bit) {                                                                \
+    if ((s3->bpp == 0) && !s3->color_16bit) {                   \
         temp = svga->vram[dword_remap(svga, (s3->accel.dest + s3->accel.cx)) & s3->vram_mask];             \
         temp |= (svga->vram[dword_remap(svga, (s3->accel.dest + s3->accel.cx + 1)) & s3->vram_mask] << 8); \
-    } else {                                                                                               \
-        temp = vram_w[dword_remap_w(svga, (s3->accel.dest + s3->accel.cx - s3->accel.minus)) & (s3->vram_mask >> 1)];        \
-    }
+    } else                                                                                               \
+        temp = vram_w[dword_remap_w(svga, (s3->accel.dest + s3->accel.cx - s3->accel.minus)) & (s3->vram_mask >> 1)];
 
 #define READ_PIXTRANS_LONG                                                                                       \
-    if ((s3->bpp == 0) && !s3->color_16bit) {                                                                      \
+    if ((s3->bpp == 0) && !s3->color_16bit) {                         \
         temp = svga->vram[dword_remap(svga, (s3->accel.dest + s3->accel.cx)) & s3->vram_mask];                   \
         temp |= (svga->vram[dword_remap(svga, (s3->accel.dest + s3->accel.cx + 1)) & s3->vram_mask] << 8);       \
         temp |= (svga->vram[dword_remap(svga, (s3->accel.dest + s3->accel.cx + 2)) & s3->vram_mask] << 16);      \
@@ -612,111 +606,62 @@ s3_accel_out_pixtrans_w(s3_t *s3, uint16_t val)
                s3->accel.frgd_mix, s3->accel.bkgd_mix, s3->accel.cur_x, val);
         switch (s3->accel.cmd & 0x600) {
             case 0x000:
-                if (((s3->accel.multifunc[0xa] & 0xc0) == 0x80) || (s3->accel.cmd & 2)) {
+                if (((s3->accel.multifunc[0xa] & 0xc0) == 0x80) || (s3->accel.cmd & 0x02)) {
                     if (((s3->accel.frgd_mix & 0x60) != 0x40) || ((s3->accel.bkgd_mix & 0x60) != 0x40)) {
                         if (s3->accel.cmd & 0x1000)
                             val = (val >> 8) | (val << 8);
                         s3->accel_start(8, 1, val | (val << 16), 0, s3);
+                    } else {
+                        if ((s3->bpp == 0) && s3->color_16bit)
+                            s3->accel_start(2, 1, 0xffffffff, val | (val << 16), s3);
+                        else
+                            s3->accel_start(1, 1, 0xffffffff, val | (val << 16), s3);
+                    }
+                } else {
+                    if ((s3->bpp == 0) && s3->color_16bit) {
+                        if (s3->accel.cur_x & 0x400)
+                            val = (val >> 8) | (val << 8);
+
+                        s3->accel_start(2, 1, 0xffffffff, val | (val << 16), s3);
                     } else
                         s3->accel_start(1, 1, 0xffffffff, val | (val << 16), s3);
-                } else {
-                    if (s3->accel.color_16bit_check_rectfill) {
-                        if (s3->accel.color_16bit_check) {
-                            if (s3->accel.pix_trans_x_count < s3->accel.pix_trans_ptr_cnt) {
-                                s3_log("Word: CPU data CMD=%04x, byte write=%02x, "
-                                       "cnt=%d, check=%d.\n", s3->accel.cmd, val & 0xff,
-                                       s3->accel.pix_trans_x_count, s3->accel.color_16bit_check);
-                                s3->accel.pix_trans_ptr[s3->accel.pix_trans_x_count] = val & 0xff;
-                                s3->accel.pix_trans_ptr[s3->accel.pix_trans_x_count + 1] = val >> 8;
-                                s3->accel.pix_trans_x_count += 2;
-                            }
-                        }
-                        break;
-                    }
-                    s3->accel_start(1, 1, 0xffffffff, val | (val << 16), s3);
                 }
                 break;
             case 0x200:
-                if (((s3->accel.multifunc[0xa] & 0xc0) == 0x80) || (s3->accel.cmd & 2)) {
+                if (((s3->accel.multifunc[0xa] & 0xc0) == 0x80) || (s3->accel.cmd & 0x02)) {
                     if (((s3->accel.frgd_mix & 0x60) != 0x40) || ((s3->accel.bkgd_mix & 0x60) != 0x40)) {
                         if (s3->accel.cmd & 0x1000)
                             val = (val >> 8) | (val << 8);
+
                         s3->accel_start(16, 1, val | (val << 16), 0, s3);
                     } else
                         s3->accel_start(2, 1, 0xffffffff, val | (val << 16), s3);
                 } else {
-                    if (s3->accel.color_16bit_check_rectfill) {
-                        if (s3->accel.color_16bit_check) {
-                            if (s3->accel.pix_trans_x_count < s3->accel.pix_trans_ptr_cnt) {
-                                s3_log("Word: CPU data CMD=%04x, word write=%04x, cnt=%d, check=%d, "
-                                       "totalptrcnt=%d.\n", s3->accel.cmd, val,
-                                       s3->accel.pix_trans_x_count, s3->accel.color_16bit_check,
-                                       s3->accel.pix_trans_ptr_cnt);
-                                s3->accel.pix_trans_ptr[s3->accel.pix_trans_x_count] = val & 0xff;
-                                s3->accel.pix_trans_ptr[s3->accel.pix_trans_x_count + 1] = val >> 8;
-                                s3->accel.pix_trans_x_count += 2;
-                                s3->accel.pix_trans_x_count2 = s3->accel.pix_trans_x_count;
-                            }
-                        } else {
-                            if (s3->accel.pix_trans_x_count < s3->accel.pix_trans_ptr_cnt) {
-                                s3_log("Word: CPU data CMD=%04x, word write=%04x, cnt=%d, check=%d, "
-                                       "totalptrcnt=%d.\n", s3->accel.cmd, val,
-                                       s3->accel.pix_trans_x_count, s3->accel.color_16bit_check,
-                                       s3->accel.pix_trans_ptr_cnt);
-                                s3->accel.pix_trans_ptr[s3->accel.pix_trans_x_count2] = val & 0xff;
-                                s3->accel.pix_trans_ptr[s3->accel.pix_trans_x_count2 + 1] = val >> 8;
-                                s3->accel.pix_trans_x_count += 2;
-                            }
-                            if (s3->accel.pix_trans_x_count2 == s3->accel.pix_trans_ptr_cnt) {
-                                for (int i = 0; i < s3->accel.pix_trans_ptr_cnt; i += 2) {
-                                    s3_log("Transferring write count=%d, bytes=%08x.\n", i,
-                                           s3->accel.pix_trans_ptr[i] |
-                                           (s3->accel.pix_trans_ptr[i + 1] << 8) |
-                                           (s3->accel.pix_trans_ptr[i + 2] << 16) |
-                                           (s3->accel.pix_trans_ptr[i + 3] << 24));
-                                    s3->accel_start(2, 1, 0xffffffff, s3->accel.pix_trans_ptr[i] |
-                                                    (s3->accel.pix_trans_ptr[i + 1] << 8), s3);
-                                }
-
-                                s3->accel.pix_trans_x_count2 = 0;
-                                s3->accel.color_16bit_check_rectfill = 0;
-                                if (s3->accel.pix_trans_ptr != NULL) {
-                                    free(s3->accel.pix_trans_ptr);
-                                    s3->accel.pix_trans_ptr = NULL;
-                                }
-                            }
-                        }
-                        break;
+                    if ((s3->bpp == 0) && s3->color_16bit) {
+                        if (s3->accel.cur_x & 0x400)
+                            val = (val >> 8) | (val << 8);
                     }
                     s3->accel_start(2, 1, 0xffffffff, val | (val << 16), s3);
                 }
                 break;
             case 0x400:
-                if (svga->crtc[0x53] & 0x08) {
-                    if (((s3->accel.multifunc[0xa] & 0xc0) == 0x80) || (s3->accel.cmd & 2)) {
-                        if (((s3->accel.frgd_mix & 0x60) != 0x40) || ((s3->accel.bkgd_mix & 0x60) != 0x40)) {
-                            if (s3->accel.cmd & 0x1000)
-                                val = (val >> 8) | (val << 8);
+                if (((s3->accel.multifunc[0xa] & 0xc0) == 0x80) || (s3->accel.cmd & 0x02)) {
+                    if (((s3->accel.frgd_mix & 0x60) != 0x40) || ((s3->accel.bkgd_mix & 0x60) != 0x40)) {
+                        if (s3->accel.cmd & 0x1000)
+                            val = (val >> 8) | (val << 8);
+
+                        if (svga->crtc[0x53] & 0x08)
                             s3->accel_start(32, 1, val | (val << 16), 0, s3);
-                        } else
-                            s3->accel_start(4, 1, 0xffffffff, val | (val << 16), s3);
-                    } else
-                        s3->accel_start(4, 1, 0xffffffff, val | (val << 16), s3);
-                } else {
-                    if (((s3->accel.multifunc[0xa] & 0xc0) == 0x80) || (s3->accel.cmd & 2)) {
-                        if (((s3->accel.frgd_mix & 0x60) != 0x40) || ((s3->accel.bkgd_mix & 0x60) != 0x40)) {
-                            if (s3->accel.cmd & 0x1000)
-                                val = (val >> 8) | (val << 8);
+                        else
                             s3->accel_start(16, 1, val | (val << 16), 0, s3);
-                        } else
-                            s3->accel_start(4, 1, 0xffffffff, val | (val << 16), s3);
                     } else
                         s3->accel_start(4, 1, 0xffffffff, val | (val << 16), s3);
-                }
+                } else
+                    s3->accel_start(4, 1, 0xffffffff, val | (val << 16), s3);
                 break;
             case 0x600:
                 if (s3->chip == S3_TRIO32 || s3->chip == S3_VISION968 || s3->chip == S3_VISION868 || s3->chip >= S3_TRIO64V) {
-                    if (((s3->accel.multifunc[0xa] & 0xc0) == 0x80) || (s3->accel.cmd & 2)) {
+                    if (((s3->accel.multifunc[0xa] & 0xc0) == 0x80) || (s3->accel.cmd & 0x02)) {
                         if (((s3->accel.frgd_mix & 0x60) != 0x40) || ((s3->accel.bkgd_mix & 0x60) != 0x40)) {
                             if (s3->accel.cmd & 0x1000)
                                 val = (val >> 8) | (val << 8);
@@ -805,7 +750,7 @@ s3_accel_out_pixtrans_l(s3_t *s3, uint32_t val)
 static void
 s3_accel_out_fifo(s3_t *s3, uint16_t port, uint8_t val)
 {
-    const svga_t *svga = &s3->svga;
+    svga_t *svga = &s3->svga;
 
     switch (port) {
         case 0x8148:
@@ -840,6 +785,7 @@ s3_accel_out_fifo(s3_t *s3, uint16_t port, uint8_t val)
             s3->accel.cur_x        = (s3->accel.cur_x & 0xff) | ((val & 0x0f) << 8);
             s3->accel.poly_cx = s3->accel.poly_x = s3->accel.cur_x << 20;
             s3->accel.poly_x                     = s3->accel.poly_cx >> 20;
+            s3_log("[%04X:%08X] OUT PORTB=%04x, val=%04x.\n", CS, cpu_state.pc, port - 1, s3->accel.cur_x);
             break;
         case 0x854a:
         case 0x86ea:
@@ -954,6 +900,7 @@ s3_accel_out_fifo(s3_t *s3, uint16_t port, uint8_t val)
                 if (!(s3->accel.multifunc[0xe] & 0x200))
                     s3->accel.multifunc[0xe] &= ~0x10;
             }
+            s3_log("[%04X:%08X] OUT PORTB=%04x, val=%04x.\n", CS, cpu_state.pc, port - 1, s3->accel.cmd);
             s3->accel_start(-1, 0, 0xffffffff, 0, s3);
             break;
 
@@ -989,6 +936,7 @@ s3_accel_out_fifo(s3_t *s3, uint16_t port, uint8_t val)
 
         case 0xa148:
         case 0xa2e8:
+            s3_log("[%04X:%08X] OUT PORTB=%04x (Background Color), val=%02x.\n", CS, cpu_state.pc, port, val);
             if (s3->bpp == 3) {
                 if ((s3->chip >= S3_86C928) && (s3->chip < S3_VISION964)) {
                     if (s3->accel.multifunc[0xe] & 0x10)
@@ -1006,6 +954,7 @@ s3_accel_out_fifo(s3_t *s3, uint16_t port, uint8_t val)
             break;
         case 0xa149:
         case 0xa2e9:
+            s3_log("[%04X:%08X] OUT PORTB=%04x (Background Color), val=%02x.\n", CS, cpu_state.pc, port, val);
             if (s3->bpp == 3) {
                 if ((s3->chip >= S3_86C928) && (s3->chip < S3_VISION964)) {
                     if (s3->accel.multifunc[0xe] & 0x10)
@@ -1069,6 +1018,7 @@ s3_accel_out_fifo(s3_t *s3, uint16_t port, uint8_t val)
 
         case 0xa548:
         case 0xa6e8:
+            s3_log("[%04X:%08X] OUT PORTB=%04x (Foreground Color), val=%02x.\n", CS, cpu_state.pc, port, val);
             if (s3->bpp == 3) {
                 if ((s3->chip >= S3_86C928) && (s3->chip < S3_VISION964)) {
                     if (s3->accel.multifunc[0xe] & 0x10)
@@ -1086,6 +1036,7 @@ s3_accel_out_fifo(s3_t *s3, uint16_t port, uint8_t val)
             break;
         case 0xa549:
         case 0xa6e9:
+            s3_log("[%04X:%08X] OUT PORTB=%04x (Foreground Color), val=%02x.\n", CS, cpu_state.pc, port, val);
             if (s3->bpp == 3) {
                 if ((s3->chip >= S3_86C928) && (s3->chip < S3_VISION964)) {
                     if (s3->accel.multifunc[0xe] & 0x10)
@@ -1113,7 +1064,6 @@ s3_accel_out_fifo(s3_t *s3, uint16_t port, uint8_t val)
                         s3->accel.multifunc[0xe] ^= 0x10;
                 }
             }
-
             if (s3->accel.color_16bit_check)
                 s3->accel.frgd_color_actual[1] = s3->accel.frgd_color & 0xff;
             else
@@ -1150,6 +1100,7 @@ s3_accel_out_fifo(s3_t *s3, uint16_t port, uint8_t val)
 
         case 0xa948:
         case 0xaae8:
+            s3_log("[%04X:%08X] OUT PORTB=%04x (Write Mask), val=%02x.\n", CS, cpu_state.pc, port, val);
             if (s3->bpp == 3) {
                 if ((s3->chip >= S3_86C928) && (s3->chip < S3_VISION964)) {
                     if (s3->accel.multifunc[0xe] & 0x10)
@@ -1167,6 +1118,7 @@ s3_accel_out_fifo(s3_t *s3, uint16_t port, uint8_t val)
             break;
         case 0xa949:
         case 0xaae9:
+            s3_log("[%04X:%08X] OUT PORTB=%04x (Write Mask), val=%02x.\n", CS, cpu_state.pc, port, val);
             if (s3->bpp == 3) {
                 if ((s3->chip >= S3_86C928) && (s3->chip < S3_VISION964)) {
                     if (s3->accel.multifunc[0xe] & 0x10)
@@ -1194,7 +1146,6 @@ s3_accel_out_fifo(s3_t *s3, uint16_t port, uint8_t val)
                         s3->accel.multifunc[0xe] ^= 0x10;
                 }
             }
-
             if (s3->accel.color_16bit_check)
                 s3->accel.wrt_mask_actual[1] = s3->accel.wrt_mask & 0xff;
             else
@@ -1307,7 +1258,8 @@ s3_accel_out_fifo(s3_t *s3, uint16_t port, uint8_t val)
 
         case 0xb148:
         case 0xb2e8:
-            if ((s3->accel.multifunc[0xe] & 0x100) || (s3->chip >= S3_86C928)) {
+            s3_log("[%04X:%08X] OUT PORTB=%04x, val=%02x, CMD=%04x, C(%d,%d).\n", CS, cpu_state.pc, port, val, s3->accel.cmd, s3->accel.cur_x, s3->accel.cur_y);
+            if (s3->accel.multifunc[0xe] & 0x100) {
                 s3->accel.b2e8_pix = 0;
                 if (s3->bpp == 3) {
                     if ((s3->chip >= S3_86C928) && (s3->chip < S3_VISION964)) {
@@ -1336,48 +1288,8 @@ s3_accel_out_fifo(s3_t *s3, uint16_t port, uint8_t val)
                                     s3->accel_start(8, 1, s3->accel.pix_trans[0], 0, s3);
                                 else
                                     s3->accel_start(1, 1, 0xffffffff, s3->accel.pix_trans[0], s3);
-                            } else {
-                                if (s3->accel.color_16bit_check_rectfill) {
-                                    if (s3->accel.color_16bit_check) {
-                                        if (s3->accel.pix_trans_x_count < s3->accel.pix_trans_ptr_cnt) {
-                                            s3_log("Byte: CPU data CMD=%04x, byte write=%02x, cnt=%d, "
-                                                   "check=%d.\n", s3->accel.cmd, val,
-                                                   s3->accel.pix_trans_x_count, s3->accel.color_16bit_check);
-                                            s3->accel.pix_trans_ptr[s3->accel.pix_trans_x_count] = val;
-                                            s3->accel.pix_trans_x_count++;
-                                            s3->accel.pix_trans_x_count2 = s3->accel.pix_trans_x_count;
-                                        }
-                                    } else {
-                                        if (s3->accel.pix_trans_x_count2 < s3->accel.pix_trans_ptr_cnt) {
-                                            s3_log("Byte: CPU data CMD=%04x, byte write=%02x, cnt=%d, "
-                                                   "check=%d.\n", s3->accel.cmd, val,
-                                                   s3->accel.pix_trans_x_count2, s3->accel.color_16bit_check);
-                                            s3->accel.pix_trans_ptr[s3->accel.pix_trans_x_count2] = val;
-                                            s3->accel.pix_trans_x_count2++;
-                                        }
-                                        s3_log("WriteCNT=%d, TotalCNT=%d.\n", s3->accel.pix_trans_x_count2,
-                                               s3->accel.pix_trans_ptr_cnt);
-                                        if (s3->accel.pix_trans_x_count2 == s3->accel.pix_trans_ptr_cnt) {
-                                            for (int i = 0; i < s3->accel.pix_trans_ptr_cnt; i += 2) {
-                                                s3_log("Transferring write count=%d, bytes=%04x.\n", i,
-                                                       s3->accel.pix_trans_ptr[i] |
-                                                       (s3->accel.pix_trans_ptr[i + 1] << 8));
-                                                s3->accel_start(1, 1, 0xffffffff, s3->accel.pix_trans_ptr[i] |
-                                                                (s3->accel.pix_trans_ptr[i + 1] << 8), s3);
-                                            }
-
-                                            s3->accel.pix_trans_x_count2 = 0;
-                                            s3->accel.color_16bit_check_rectfill = 0;
-                                            if (s3->accel.pix_trans_ptr != NULL) {
-                                                free(s3->accel.pix_trans_ptr);
-                                                s3->accel.pix_trans_ptr = NULL;
-                                            }
-                                        }
-                                    }
-                                    break;
-                                }
+                            } else
                                 s3->accel_start(1, 1, 0xffffffff, s3->accel.pix_trans[0], s3);
-                            }
                             break;
 
                         default:
@@ -1388,7 +1300,8 @@ s3_accel_out_fifo(s3_t *s3, uint16_t port, uint8_t val)
             break;
         case 0xb149:
         case 0xb2e9:
-            if ((s3->accel.multifunc[0xe] & 0x100) || (s3->chip >= S3_86C928)) {
+            s3_log("[%04X:%08X] OUT PORTB=%04x, val=%02x, CMD=%04x, C(%d,%d).\n", CS, cpu_state.pc, port, val, s3->accel.cmd, s3->accel.cur_x, s3->accel.cur_y);
+            if (s3->accel.multifunc[0xe] & 0x100) {
                 s3->accel.b2e8_pix = 0;
                 if (s3->bpp == 3) {
                     if ((s3->chip >= S3_86C928) && (s3->chip < S3_VISION964)) {
@@ -1643,6 +1556,8 @@ s3_accel_out_fifo(s3_t *s3, uint16_t port, uint8_t val)
 
         case 0xe148:
         case 0xe2e8:
+            s3_log("[%04X:%08X] OUT PORTB=%04x, val=%02x, CMD=%04x, C(%d,%d), WRTMASK=%04x.\n", CS, cpu_state.pc, port, val, s3->accel.cmd, s3->accel.cur_x, s3->accel.cur_y, s3->accel.wrt_mask);
+            s3_log(".\n");
             s3->accel.b2e8_pix = 0;
             if (s3_cpu_dest(s3))
                 break;
@@ -1656,46 +1571,17 @@ s3_accel_out_fifo(s3_t *s3, uint16_t port, uint8_t val)
                             else
                                 s3->accel_start(1, 1, 0xffffffff, s3->accel.pix_trans[0], s3);
                         } else {
-                            if (s3->accel.color_16bit_check_rectfill) {
-                                if (s3->accel.color_16bit_check) {
-                                    if (s3->accel.pix_trans_x_count < s3->accel.pix_trans_ptr_cnt) {
-                                        s3_log("Byte: CPU data CMD=%04x, byte write=%02x, cnt=%d, "
-                                               "check=%d.\n", s3->accel.cmd, val,
-                                               s3->accel.pix_trans_x_count, s3->accel.color_16bit_check);
-                                        s3->accel.pix_trans_ptr[s3->accel.pix_trans_x_count] = val;
-                                        s3->accel.pix_trans_x_count++;
-                                        s3->accel.pix_trans_x_count2 = s3->accel.pix_trans_x_count;
-                                    }
+                            if ((s3->bpp == 0) && s3->color_16bit) {
+                                s3->accel.pix_trans[1] = svga->vram[dword_remap(svga, (s3->accel.dest + s3->accel.cx - s3->accel.minus)) & s3->vram_mask];
+                                if (s3->accel.cur_x & 0x400) {
+                                    s3_log("Last Pixel Written=%02x (1024).\n", s3->accel.pix_trans[1]);
+                                    s3->accel_start(2, 1, 0xffffffff, s3->accel.pix_trans[1] | (s3->accel.pix_trans[0] << 8), s3);
                                 } else {
-                                    if (s3->accel.pix_trans_x_count2 < s3->accel.pix_trans_ptr_cnt) {
-                                        s3_log("Byte: CPU data CMD=%04x, byte write=%02x, cnt=%d, "
-                                               "check=%d.\n", s3->accel.cmd, val,
-                                               s3->accel.pix_trans_x_count2, s3->accel.color_16bit_check);
-                                        s3->accel.pix_trans_ptr[s3->accel.pix_trans_x_count2] = val;
-                                        s3->accel.pix_trans_x_count2++;
-                                    }
-                                    s3_log("WriteCNT=%d, TotalCNT=%d.\n", s3->accel.pix_trans_x_count2,
-                                           s3->accel.pix_trans_ptr_cnt);
-                                    if (s3->accel.pix_trans_x_count2 == s3->accel.pix_trans_ptr_cnt) {
-                                        for (int i = 0; i < s3->accel.pix_trans_ptr_cnt; i += 2) {
-                                            s3_log("Transferring write count=%d, bytes=%04x.\n", i,
-                                                   s3->accel.pix_trans_ptr[i] |
-                                                   (s3->accel.pix_trans_ptr[i + 1] << 8));
-                                            s3->accel_start(1, 1, 0xffffffff, s3->accel.pix_trans_ptr[i] |
-                                                            (s3->accel.pix_trans_ptr[i + 1] << 8), s3);
-                                        }
-
-                                        s3->accel.pix_trans_x_count2 = 0;
-                                        s3->accel.color_16bit_check_rectfill = 0;
-                                        if (s3->accel.pix_trans_ptr != NULL) {
-                                            free(s3->accel.pix_trans_ptr);
-                                            s3->accel.pix_trans_ptr = NULL;
-                                        }
-                                    }
+                                    s3_log("Last Pixel Written=%02x (0).\n", s3->accel.pix_trans[1]);
+                                    s3->accel_start(2, 1, 0xffffffff, s3->accel.pix_trans[0] | (s3->accel.pix_trans[1] << 8), s3);
                                 }
-                                break;
-                            }
-                            s3->accel_start(1, 1, 0xffffffff, s3->accel.pix_trans[0], s3);
+                            } else
+                                s3->accel_start(1, 1, 0xffffffff, s3->accel.pix_trans[0], s3);
                         }
                         break;
 
@@ -1706,6 +1592,8 @@ s3_accel_out_fifo(s3_t *s3, uint16_t port, uint8_t val)
             break;
         case 0xe149:
         case 0xe2e9:
+            s3_log("[%04X:%08X] OUT PORTB=%04x, val=%02x, CMD=%04x, C(%d,%d), WRTMASK=%04x.\n", CS, cpu_state.pc, port, val, s3->accel.cmd, s3->accel.cur_x, s3->accel.cur_y, s3->accel.wrt_mask);
+            s3_log(".\n");
             s3->accel.b2e8_pix = 0;
             if (s3_cpu_dest(s3))
                 break;
@@ -1859,6 +1747,8 @@ static void
 s3_accel_out_fifo_w(s3_t *s3, uint16_t port, uint16_t val)
 {
     if ((port != 0x9ee8) && (port != 0x9d48)) {
+        s3_log("[%04X:%08X] OUT PORTW=%04x, val=%04x, CMD=%04x, C(%d,%d), WRTMASK=%04x.\n", CS, cpu_state.pc, port, val, s3->accel.cmd, s3->accel.cur_x, s3->accel.cur_y, s3->accel.wrt_mask);
+        s3_log(".\n");
         if ((port == 0xb2e8) || (port == 0xb148)) {
             if ((s3->accel.multifunc[0xe] & 0x100) || (s3->chip >= S3_86C928)) {
                 s3->accel.b2e8_pix = 0;
@@ -1869,7 +1759,6 @@ s3_accel_out_fifo_w(s3_t *s3, uint16_t port, uint16_t val)
                 s3_accel_out_pixtrans_w(s3, val);
             }
         } else {
-            s3_log("Port WriteW FIFO=%04x, val=%04x, bussize=%03x.\n", port, val, s3->accel.cmd & 0x600);
             s3->accel.b2e8_pix = 0;
             if ((port == 0xe2e8) || (port == 0xe2ea) || (port == 0xe148) || (port == 0xe14a))
                 s3_accel_out_pixtrans_w(s3, val);
@@ -1899,6 +1788,7 @@ static void
 s3_accel_out_fifo_l(s3_t *s3, uint16_t port, uint32_t val)
 {
     if ((port == 0xb2e8) || (port == 0xb148)) {
+        s3_log("[%04X:%08X] OUT PORTL=%04x, val=%08x, CMD=%04x, C(%d,%d).\n", CS, cpu_state.pc, port, val, s3->accel.cmd, s3->accel.cur_x, s3->accel.cur_y);
         if ((s3->accel.multifunc[0xe] & 0x100) || (s3->chip >= S3_86C928)) {
             s3->accel.b2e8_pix = 0;
             s3_accel_out_fifo(s3, port, val);
@@ -1911,7 +1801,6 @@ s3_accel_out_fifo_l(s3_t *s3, uint16_t port, uint32_t val)
         }
     } else {
         s3->accel.b2e8_pix = 0;
-        s3_log("Port WriteL FIFO=%04x, val=%08x, bussize=%03x.\n", port, val, s3->accel.cmd & 0x600);
         if ((port == 0xe2e8) || (port == 0xe148))
             s3_accel_out_pixtrans_l(s3, val);
         else {
@@ -2867,6 +2756,8 @@ s3_out(uint16_t addr, uint8_t val, void *priv)
                     svga->write_bank = svga->read_bank = s3->bank << 16;
                 else
                     svga->write_bank = svga->read_bank = s3->bank << 14;
+
+                svga_recalctimings(svga);
             } else if (svga->seqaddr == 9) {
                 svga->seqregs[svga->seqaddr] = val & 0x80;
                 s3_io_set(s3);
@@ -2979,6 +2870,8 @@ s3_out(uint16_t addr, uint8_t val, void *priv)
                         svga->write_bank = svga->read_bank = s3->bank << 16;
                     else
                         svga->write_bank = svga->read_bank = s3->bank << 14;
+
+                    s3_log("CRTC35 Chain4=%x, Bank=%02x.\n", svga->chain4, s3->bank);
                     break;
 
                 case 0x51:
@@ -3105,21 +2998,9 @@ s3_out(uint16_t addr, uint8_t val, void *priv)
 
                 case 0x43:
                     if (s3->chip < S3_VISION964) {
-                        if (s3->chip <= S3_86C805) {
-                            s3->color_16bit = !!(val & 8);
-                            if (s3->color_16bit) {
-                                s3->width = 1024;
-                            } else {
-                                if (s3->chip <= S3_86C924)
-                                    s3->width = 1024;
-                                else {
-                                    if (s3->accel.advfunc_cntl & 4)
-                                        s3->width = 1024;
-                                    else
-                                        s3->width = 640;
-                                }
-                            }
-                        }
+                        if (s3->chip <= S3_86C805)
+                            svga_recalctimings(svga);
+
                         s3_io_remove_alt(s3);
                         s3->translate = !!(val & 0x10);
                         s3_io_set_alt(s3);
@@ -3332,6 +3213,140 @@ s3_in(uint16_t addr, void *priv)
     return svga_in(addr, svga);
 }
 
+uint32_t
+s3_decode_addr(svga_t *svga, uint32_t addr, int write)
+{
+    int memory_map_mode = (svga->gdcreg[6] >> 2) & 3;
+
+    s3_log("CRTC31 bit 3=%x, map=%x, write=%x, wrtbank=%x, chain4=%x.\n", svga->crtc[0x31] & 0x08, memory_map_mode, write, svga->write_bank, svga->chain4);
+
+    if (svga->crtc[0x31] & 0x08)
+        memory_map_mode = 1;
+
+    addr &= 0x1ffff;
+
+    switch (memory_map_mode) {
+        case 0:
+            break;
+        case 1:
+            if (addr >= 0x10000)
+                return 0xffffffff;
+            break;
+        case 2:
+            addr -= 0x10000;
+            if (addr >= 0x8000)
+                return 0xffffffff;
+            break;
+        default:
+        case 3:
+            addr -= 0x18000;
+            if (addr >= 0x8000)
+                return 0xffffffff;
+            break;
+    }
+
+    if (memory_map_mode <= 1) {
+        if (write)
+            addr += svga->write_bank;
+        else
+            addr += svga->read_bank;
+    }
+
+    return addr;
+}
+
+static void
+s3_write(uint32_t addr, uint8_t val, void *priv)
+{
+    s3_t *s3 = (s3_t *) priv;
+    svga_t *svga = &s3->svga;
+
+    xga_write_test(addr, val, svga);
+    addr = s3_decode_addr(svga, addr, 1);
+    if (addr == 0xffffffff)
+        return;
+
+    svga_write_linear(addr, val, svga);
+}
+
+static void
+s3_writew(uint32_t addr, uint16_t val, void *priv)
+{
+    s3_t *s3 = (s3_t *) priv;
+    svga_t *svga = &s3->svga;
+
+    xga_write_test(addr, val, svga);
+    xga_write_test(addr + 1, val >> 8, svga);
+    addr = s3_decode_addr(svga, addr, 1);
+    if (addr == 0xffffffff)
+        return;
+
+    svga_writew_linear(addr, val, svga);
+}
+
+static void
+s3_writel(uint32_t addr, uint32_t val, void *priv)
+{
+    s3_t *s3 = (s3_t *) priv;
+    svga_t *svga = &s3->svga;
+
+    xga_write_test(addr, val, svga);
+    xga_write_test(addr + 1, val >> 8, svga);
+    xga_write_test(addr + 2, val >> 16, svga);
+    xga_write_test(addr + 3, val >> 24, svga);
+    addr = s3_decode_addr(svga, addr, 1);
+    if (addr == 0xffffffff)
+        return;
+
+    svga_writel_linear(addr, val, svga);
+}
+
+static uint8_t
+s3_read(uint32_t addr, void *priv)
+{
+    s3_t *s3 = (s3_t *) priv;
+    svga_t *svga = &s3->svga;
+
+    (void) xga_read_test(addr, svga);
+    addr = s3_decode_addr(svga, addr, 0);
+    if (addr == 0xffffffff)
+        return 0xff;
+
+    return svga_read_linear(addr, svga);
+}
+
+static uint16_t
+s3_readw(uint32_t addr, void *priv)
+{
+    s3_t *s3 = (s3_t *) priv;
+    svga_t *svga = &s3->svga;
+
+    (void) xga_read_test(addr, svga);
+    (void) xga_read_test(addr + 1, svga);
+    addr = s3_decode_addr(svga, addr, 0);
+    if (addr == 0xffffffff)
+        return 0xffff;
+
+    return svga_readw_linear(addr, svga);
+}
+
+static uint32_t
+s3_readl(uint32_t addr, void *priv)
+{
+    s3_t *s3 = (s3_t *) priv;
+    svga_t *svga = &s3->svga;
+
+    (void) xga_read_test(addr, svga);
+    (void) xga_read_test(addr + 1, svga);
+    (void) xga_read_test(addr + 2, svga);
+    (void) xga_read_test(addr + 3, svga);
+    addr = s3_decode_addr(svga, addr, 0);
+    if (addr == 0xffffffff)
+        return 0xffffffff;
+
+    return svga_readl_linear(addr, svga);
+}
+
 static void
 s3_recalctimings(svga_t *svga)
 {
@@ -3416,33 +3431,48 @@ s3_recalctimings(svga_t *svga)
             break;
     }
 
-    if (s3->chip != S3_86C801)
-        mask |= 0x01;
-    switch (svga->crtc[0x50] & mask) {
-        case 0x00:
-            if (s3->color_16bit)
-                s3->width = 1024;
-            else
-                s3->width = (svga->crtc[0x31] & 2) ? 2048 : 1024;
-            break;
-        case 0x01:
-            s3->width = 1152;
-            break;
-        case 0x40:
-            s3->width = 640;
-            break;
-        case 0x80:
-            s3->width = ((s3->chip > S3_86C805) && (s3->accel.advfunc_cntl & 4)) ? 1600 : 800;
-            break;
-        case 0x81:
-            s3->width = 1600;
-            break;
-        case 0xc0:
-            s3->width = 1280;
-            break;
+    if (s3->chip <= S3_86C805) {
+        s3->color_16bit = !!(svga->crtc[0x43] & 0x08);
+        if (svga->bpp == 24)
+            s3->color_16bit = 0;
 
-        default:
-            break;
+        if (s3->color_16bit)
+            s3->width = 1024;
+        else {
+            if (s3->chip <= S3_86C924)
+                s3->width = 1024;
+        }
+    }
+
+    if (s3->chip >= S3_86C928) {
+        if (s3->chip != S3_86C801)
+            mask |= 0x01;
+
+        switch (svga->crtc[0x50] & mask) {
+            case 0x00:
+                s3->width = (svga->crtc[0x31] & 0x02) ? 2048 : 1024;
+                if (s3->color_16bit)
+                    s3->width = 1024;
+                break;
+            case 0x01:
+                s3->width = 1152;
+                break;
+            case 0x40:
+                s3->width = 640;
+                break;
+            case 0x80:
+                s3->width = ((s3->chip > S3_86C805) && (s3->accel.advfunc_cntl & 0x04)) ? 1600 : 800;
+                break;
+            case 0x81:
+                s3->width = 1600;
+                break;
+            case 0xc0:
+                s3->width = 1280;
+                break;
+
+            default:
+                break;
+        }
     }
 
     if (svga->crtc[0x33] & 0x20) {
@@ -4275,14 +4305,27 @@ s3_recalctimings(svga_t *svga)
                         /*Enhanced 4bpp mode, just like the 8bpp mode per the spec. */
                         svga->render = svga_render_8bpp_highres;
                         svga->rowoffset <<= 1;
+                        svga->vram_display_mask = s3->vram_mask;
+                    } else {
+                        svga->vram_display_mask = (svga->crtc[0x32] & 0x40) ? 0x3ffff : s3->vram_mask;
+                        svga->write_bank = 0;
+                        svga->read_bank = svga->write_bank;
                     }
-                    svga->vram_display_mask = s3->vram_mask;
-                } else
+                } else {
                     svga->vram_display_mask = (svga->crtc[0x32] & 0x40) ? 0x3ffff : s3->vram_mask;
-            } else
+                    svga->write_bank = 0;
+                    svga->read_bank = svga->write_bank;
+                }
+            } else {
                 svga->vram_display_mask = (svga->crtc[0x32] & 0x40) ? 0x3ffff : s3->vram_mask;
-        } else
+                svga->write_bank = 0;
+                svga->read_bank = svga->write_bank;
+            }
+        } else {
             svga->vram_display_mask = (svga->crtc[0x32] & 0x40) ? 0x3ffff : s3->vram_mask;
+            svga->write_bank = 0;
+            svga->read_bank = svga->write_bank;
+        }
     }
 
     if ((s3->chip == S3_TRIO32) || (s3->chip == S3_TRIO64) || (s3->chip == S3_VISION864))
@@ -4421,8 +4464,11 @@ s3_trio64v_recalctimings(svga_t *svga)
                     break;
             }
             svga->vram_display_mask = s3->vram_mask;
-        } else
+        } else {
             svga->vram_display_mask = (svga->crtc[0x32] & 0x40) ? 0x3ffff : s3->vram_mask;
+            svga->write_bank = 0;
+            svga->read_bank = svga->write_bank;
+        }
     } else /*Streams mode*/
     {
         if (s3->streams.buffer_ctrl & 1)
@@ -4724,7 +4770,7 @@ s3_accel_in(uint16_t port, void *priv)
     s3_t   *s3   = (s3_t *) priv;
     svga_t *svga = &s3->svga;
     int     temp;
-    uint8_t temp2;
+    uint8_t temp2 = 0x00;
 
     if (!s3->enable_8514)
         return 0xff;
@@ -5605,7 +5651,7 @@ s3_accel_in(uint16_t port, void *priv)
             if (s3->accel.cmd & 0x100) {
                 switch (s3->accel.cmd & 0x600) {
                     case 0x000:
-                        if (((s3->accel.multifunc[0xa] & 0xc0) == 0x80) || (s3->accel.cmd & 2)) {
+                        if (((s3->accel.multifunc[0xa] & 0xc0) == 0x80) || (s3->accel.cmd & 0x02)) {
                             if (((s3->accel.frgd_mix & 0x60) != 0x40) || ((s3->accel.bkgd_mix & 0x60) != 0x40))
                                 s3->accel_start(8, 1, s3->accel.pix_trans[0], 0, s3);
                             else
@@ -5614,7 +5660,7 @@ s3_accel_in(uint16_t port, void *priv)
                             s3->accel_start(1, 1, 0xffffffff, s3->accel.pix_trans[0], s3);
                         break;
                     case 0x200:
-                        if (((s3->accel.multifunc[0xa] & 0xc0) == 0x80) || (s3->accel.cmd & 2)) {
+                        if (((s3->accel.multifunc[0xa] & 0xc0) == 0x80) || (s3->accel.cmd & 0x02)) {
                             if (((s3->accel.frgd_mix & 0x60) != 0x40) || ((s3->accel.bkgd_mix & 0x60) != 0x40))
                                 s3->accel_start(16, 1, s3->accel.pix_trans[0], 0, s3);
                             else
@@ -5637,7 +5683,7 @@ s3_accel_in(uint16_t port, void *priv)
             if (s3->accel.cmd & 0x100) {
                 switch (s3->accel.cmd & 0x600) {
                     case 0x000:
-                        if (((s3->accel.multifunc[0xa] & 0xc0) == 0x80) || (s3->accel.cmd & 2)) {
+                        if (((s3->accel.multifunc[0xa] & 0xc0) == 0x80) || (s3->accel.cmd & 0x02)) {
                             if (((s3->accel.frgd_mix & 0x60) != 0x40) || ((s3->accel.bkgd_mix & 0x60) != 0x40))
                                 s3->accel_start(8, 1, s3->accel.pix_trans[0] | (s3->accel.pix_trans[1] << 8), 0, s3);
                             else
@@ -5646,7 +5692,7 @@ s3_accel_in(uint16_t port, void *priv)
                             s3->accel_start(1, 1, 0xffffffff, s3->accel.pix_trans[0] | (s3->accel.pix_trans[1] << 8), s3);
                         break;
                     case 0x200:
-                        if (((s3->accel.multifunc[0xa] & 0xc0) == 0x80) || (s3->accel.cmd & 2)) {
+                        if (((s3->accel.multifunc[0xa] & 0xc0) == 0x80) || (s3->accel.cmd & 0x02)) {
                             if (((s3->accel.frgd_mix & 0x60) != 0x40) || ((s3->accel.bkgd_mix & 0x60) != 0x40)) {
                                 if (s3->accel.cmd & 0x1000)
                                     s3->accel_start(16, 1, s3->accel.pix_trans[1] | (s3->accel.pix_trans[0] << 8), 0, s3);
@@ -5760,10 +5806,21 @@ s3_accel_in_w(uint16_t port, void *priv)
                             if (s3->accel.cmd & 0x1000)
                                 temp = (temp >> 8) | (temp << 8);
                             s3->accel_start(8, 1, temp | (temp << 16), 0, s3);
+                        } else {
+                            if ((s3->bpp == 0) && s3->color_16bit)
+                                s3->accel_start(2, 1, 0xffffffff, temp | (temp << 16), s3);
+                            else
+                                s3->accel_start(1, 1, 0xffffffff, temp | (temp << 16), s3);
+                        }
+                    } else {
+                        if ((s3->bpp == 0) && s3->color_16bit) {
+                            if (s3->accel.cur_x & 0x400)
+                                temp = ((temp >> 8) | (temp << 8)) & 0xffff;
+
+                            s3->accel_start(2, 1, 0xffffffff, temp | (temp << 16), s3);
                         } else
                             s3->accel_start(1, 1, 0xffffffff, temp | (temp << 16), s3);
-                    } else
-                        s3->accel_start(1, 1, 0xffffffff, temp | (temp << 16), s3);
+                    }
                     break;
                 case 0x200:
                     if (((s3->accel.multifunc[0xa] & 0xc0) == 0x80) || (s3->accel.cmd & 2)) {
@@ -6479,15 +6536,11 @@ polygon_setup(s3_t *s3)
 }
 
 #define READ(addr, dat)                                                 \
-    if ((s3->bpp == 0) && !s3->color_16bit)                               \
+    if (((s3->bpp == 0) && !s3->color_16bit) || (s3->bpp == 2))         \
         dat = svga->vram[dword_remap(svga, addr) & s3->vram_mask];      \
-    else if ((s3->bpp == 1) || (s3->color_16bit && (svga->bpp < 24))) \
+    else if ((s3->bpp == 1) || s3->color_16bit)   \
         dat = vram_w[dword_remap_w(svga, addr) & (s3->vram_mask >> 1)]; \
-    else if (s3->bpp == 2)  \
-        dat = svga->vram[dword_remap(svga, addr) & s3->vram_mask];      \
-    else if (s3->color_16bit && (svga->bpp == 24)) {                     \
-        dat = vram_w[dword_remap_w(svga, addr) & (s3->vram_mask >> 1)]; \
-    } else                                                                \
+    else                                                               \
         dat = vram_l[dword_remap_l(svga, addr) & (s3->vram_mask >> 2)];
 
 #define MIX_READ                                                                                  \
@@ -6548,7 +6601,7 @@ polygon_setup(s3_t *s3)
     {                                                                                      \
         old_dest_dat = dest_dat;                                                           \
         MIX_READ                                                                           \
-        dest_dat = (dest_dat & wrt_mask) | (old_dest_dat & ~wrt_mask);                     \
+        dest_dat = (dest_dat & wrt_mask) | (old_dest_dat & ~wrt_mask);                 \
     }
 
 #define ROPMIX_READ(D, P, S)                       \
@@ -7333,20 +7386,14 @@ polygon_setup(s3_t *s3)
     }
 
 #define WRITE(addr, dat)                                                                                                   \
-    if ((s3->bpp == 0) && !s3->color_16bit) {                                                                                \
-        svga->vram[dword_remap(svga, addr) & s3->vram_mask]                = dat;                                          \
+    if (((s3->bpp == 0) && !s3->color_16bit) || (s3->bpp == 2)) {                   \
+        svga->vram[dword_remap(svga, addr) & s3->vram_mask] = dat;                                          \
         svga->changedvram[(dword_remap(svga, addr) & s3->vram_mask) >> 12] = svga->monitor->mon_changeframecount;          \
-    } else if ((s3->bpp == 1) || (s3->color_16bit && (svga->bpp < 24))) {                                                   \
+    } else if ((s3->bpp == 1) || s3->color_16bit) {                                                   \
         vram_w[dword_remap_w(svga, addr) & (s3->vram_mask >> 1)]                    = dat;                                 \
         svga->changedvram[(dword_remap_w(svga, addr) & (s3->vram_mask >> 1)) >> 11] = svga->monitor->mon_changeframecount; \
-    } else if (s3->bpp == 2) {                                                  \
-        svga->vram[dword_remap(svga, addr) & s3->vram_mask]                = dat;                                          \
-        svga->changedvram[(dword_remap(svga, addr) & s3->vram_mask) >> 12] = svga->monitor->mon_changeframecount;          \
-    } else if (s3->color_16bit && (svga->bpp == 24)) { \
-        vram_w[dword_remap_w(svga, addr) & (s3->vram_mask >> 1)]           = dat;                                 \
-        svga->changedvram[(dword_remap_w(svga, addr) & (s3->vram_mask >> 1)) >> 11] = svga->monitor->mon_changeframecount; \
     } else {                                                                                                               \
-        vram_l[dword_remap_l(svga, addr) & (s3->vram_mask >> 2)]                    = dat;                                 \
+        vram_l[dword_remap_l(svga, addr) & (s3->vram_mask >> 2)] = dat;                                                \
         svga->changedvram[(dword_remap_l(svga, addr) & (s3->vram_mask >> 2)) >> 10] = svga->monitor->mon_changeframecount; \
     }
 
@@ -7731,745 +7778,6 @@ s3_visionx68_video_engine_op(uint32_t cpu_dat, s3_t *s3)
     }
 }
 
-void
-s3_911_accel_start(int count, int cpu_input, uint32_t mix_dat, uint32_t cpu_dat, void *priv)
-{
-    s3_t *s3 = (s3_t *)priv;
-    svga_t *svga = &s3->svga;
-    uint32_t  src_dat = 0;
-    uint32_t  dest_dat;
-    uint32_t  old_dest_dat;
-    int       frgd_mix;
-    int       bkgd_mix;
-    int       clip_t       = s3->accel.multifunc[1] & 0xfff;
-    int       clip_l       = s3->accel.multifunc[2] & 0xfff;
-    int       clip_b       = s3->accel.multifunc[3] & 0xfff;
-    int       clip_r       = s3->accel.multifunc[4] & 0xfff;
-    int       vram_mask    = (s3->accel.multifunc[0xa] & 0xc0) == 0xc0;
-    uint32_t  mix_mask     = (s3->accel.cmd & 0x200) ? 0x8000 : 0x80;
-    uint16_t *vram_w       = (uint16_t *) svga->vram;
-    uint32_t *vram_l       = (uint32_t *) svga->vram;
-    uint32_t  rd_mask      = s3->accel.rd_mask;
-    uint32_t  wrt_mask     = s3->accel.wrt_mask;
-    uint32_t  frgd_color   = s3->accel.frgd_color;
-    uint32_t  bkgd_color   = s3->accel.bkgd_color;
-    int       cmd          = s3->accel.cmd >> 13;
-
-    if ((s3->accel.cmd & 0x100) && (s3_cpu_src(s3) || (s3_cpu_dest(s3))) && (!cpu_input || (s3_enable_fifo(s3) == 0)))
-        s3->force_busy = 1;
-
-    if (cpu_input && (((s3->accel.multifunc[0xa] & 0xc0) != 0x80) || (!(s3->accel.cmd & 2)))) {
-        if (s3->color_16bit) {
-            if (count > 1)
-                count >>= 1;
-        }
-    }
-
-    if (s3->color_16bit)
-        rd_mask &= 0xffff;
-    else
-        rd_mask &= 0xff;
-
-    /*Bit 4 of the Command register is the draw yes bit, which enables writing to memory/reading from memory when enabled.
-      When this bit is disabled, no writing to memory/reading from memory is allowed. (This bit is almost meaningless on
-      the NOP command)*/
-
-    switch (cmd) {
-        case 0: /*NOP (Short Stroke Vectors)*/
-            if (s3->accel.ssv_state == 0)
-                break;
-
-            frgd_mix = (s3->accel.frgd_mix >> 5) & 3;
-            bkgd_mix = (s3->accel.bkgd_mix >> 5) & 3;
-
-            if (s3->accel.cmd & 8) { /*Radial*/
-                while (count-- && s3->accel.ssv_len >= 0) {
-                    if ((s3->accel.cx & 0xfff) >= clip_l && (s3->accel.cx & 0xfff) <= clip_r && (s3->accel.cy & 0xfff) >= clip_t && (s3->accel.cy & 0xfff) <= clip_b) {
-                        switch ((mix_dat & mix_mask) ? frgd_mix : bkgd_mix) {
-                            case 0:
-                                src_dat = bkgd_color;
-                                break;
-                            case 1:
-                                src_dat = frgd_color;
-                                break;
-                            case 2:
-                                src_dat = cpu_dat;
-                                break;
-                            case 3:
-                                src_dat = 0;
-                                break;
-
-                            default:
-                                break;
-                        }
-
-                        READ((s3->accel.cy * s3->width) + s3->accel.cx, dest_dat);
-
-                        MIX
-
-                        if (s3->accel.ssv_draw) {
-                            WRITE((s3->accel.cy * s3->width) + s3->accel.cx, dest_dat);
-                        }
-                    }
-
-                    mix_dat <<= 1;
-                    mix_dat |= 1;
-                    if (s3->bpp == 0)
-                        cpu_dat >>= 8;
-                    else
-                        cpu_dat >>= 16;
-
-                    if (!s3->accel.ssv_len)
-                        break;
-
-                    switch (s3->accel.ssv_dir & 0xe0) {
-                        case 0x00:
-                            s3->accel.cx++;
-                            break;
-                        case 0x20:
-                            s3->accel.cx++;
-                            s3->accel.cy--;
-                            break;
-                        case 0x40:
-                            s3->accel.cy--;
-                            break;
-                        case 0x60:
-                            s3->accel.cx--;
-                            s3->accel.cy--;
-                            break;
-                        case 0x80:
-                            s3->accel.cx--;
-                            break;
-                        case 0xa0:
-                            s3->accel.cx--;
-                            s3->accel.cy++;
-                            break;
-                        case 0xc0:
-                            s3->accel.cy++;
-                            break;
-                        case 0xe0:
-                            s3->accel.cx++;
-                            s3->accel.cy++;
-                            break;
-
-                        default:
-                            break;
-                    }
-
-                    s3->accel.ssv_len--;
-                    s3->accel.cx &= 0xfff;
-                    s3->accel.cy &= 0xfff;
-                }
-
-                s3->accel.cur_x = s3->accel.cx;
-                s3->accel.cur_y = s3->accel.cy;
-            }
-            break;
-
-        case 1: /*Draw line*/
-            if (!cpu_input) {
-                s3->accel.cx = s3->accel.cur_x & 0xfff;
-                s3->accel.cy = s3->accel.cur_y & 0xfff;
-                s3->accel.sy = s3->accel.maj_axis_pcnt & 0x7ff;
-
-                if (s3->color_16bit && (svga->bpp < 24)) {
-                    if (s3->accel.wrt_mask != 0xffff) {
-                        if (s3->accel.cur_x & 0x400) {
-                            s3->accel.color_16bit_check = 0;
-                            s3->accel.minus = 0x400;
-                        } else {
-                            s3->accel.color_16bit_check = 1;
-                            s3->accel.minus = 0;
-                        }
-                    } else {
-                        if (s3->accel.cur_x & 0x400)
-                            s3->accel.color_16bit_check = 1;
-                        else
-                            s3->accel.color_16bit_check = 0;
-
-                        s3->accel.minus = 0;
-                    }
-                } else {
-                    s3->accel.color_16bit_check = 0;
-                    s3->accel.minus = 0;
-                }
-
-                if (s3_cpu_src(s3))
-                    return; /*Wait for data from CPU*/
-            }
-
-            frgd_mix = (s3->accel.frgd_mix >> 5) & 3;
-            bkgd_mix = (s3->accel.bkgd_mix >> 5) & 3;
-
-            if (s3->accel.cmd & 8) { /*Radial*/
-                if (s3->color_16bit && (svga->bpp < 24)) {
-                    if (s3->accel.color_16bit_check)
-                        return;
-                    if (s3->accel.wrt_mask != 0xffff)
-                        wrt_mask = (s3->accel.wrt_mask_actual[0] | (s3->accel.wrt_mask_actual[1] << 8));
-                }
-
-                while (count-- && s3->accel.sy >= 0) {
-                    if ((s3->accel.cx & 0xfff) >= clip_l && (s3->accel.cx & 0xfff) <= clip_r && (s3->accel.cy & 0xfff) >= clip_t && (s3->accel.cy & 0xfff) <= clip_b) {
-                        switch ((mix_dat & mix_mask) ? frgd_mix : bkgd_mix) {
-                            case 0:
-                                src_dat = bkgd_color;
-                                if (s3->color_16bit && (svga->bpp < 24))
-                                    src_dat = s3->accel.bkgd_color_actual[0] | (s3->accel.bkgd_color_actual[1] << 8);
-                                break;
-                            case 1:
-                                src_dat = frgd_color;
-                                if (s3->color_16bit && (svga->bpp < 24))
-                                    src_dat = s3->accel.frgd_color_actual[0] | (s3->accel.frgd_color_actual[1] << 8);
-                                break;
-                            case 2:
-                                src_dat = cpu_dat;
-                                break;
-                            case 3:
-                                src_dat = 0;
-                                break;
-
-                            default:
-                                break;
-                        }
-
-                        READ((s3->accel.cy * s3->width) + s3->accel.cx - s3->accel.minus, dest_dat);
-
-                        MIX
-
-                        WRITE((s3->accel.cy * s3->width) + s3->accel.cx - s3->accel.minus, dest_dat);
-                    }
-
-                    mix_dat <<= 1;
-                    mix_dat |= 1;
-                    if ((s3->bpp == 0) && !s3->color_16bit)
-                        cpu_dat >>= 8;
-                    else
-                        cpu_dat >>= 16;
-
-                    if (!s3->accel.sy)
-                        break;
-
-                    switch (s3->accel.cmd & 0xe0) {
-                        case 0x00:
-                            s3->accel.cx++;
-                            break;
-                        case 0x20:
-                            s3->accel.cx++;
-                            s3->accel.cy--;
-                            break;
-                        case 0x40:
-                            s3->accel.cy--;
-                            break;
-                        case 0x60:
-                            s3->accel.cx--;
-                            s3->accel.cy--;
-                            break;
-                        case 0x80:
-                            s3->accel.cx--;
-                            break;
-                        case 0xa0:
-                            s3->accel.cx--;
-                            s3->accel.cy++;
-                            break;
-                        case 0xc0:
-                            s3->accel.cy++;
-                            break;
-                        case 0xe0:
-                            s3->accel.cx++;
-                            s3->accel.cy++;
-                            break;
-
-                        default:
-                            break;
-                    }
-                    s3->accel.sy--;
-                    s3->accel.cx &= 0xfff;
-                    s3->accel.cy &= 0xfff;
-                }
-                s3->accel.cur_x = s3->accel.cx;
-                s3->accel.cur_y = s3->accel.cy;
-            } else { /*Bresenham*/
-                if (s3->accel.b2e8_pix && s3_cpu_src(s3) && (count == 16)) { /*Stupid undocumented 0xB2E8 on 911/924*/
-                    count              = s3->accel.maj_axis_pcnt + 1;
-                    s3->accel.temp_cnt = 16;
-                }
-
-                if (s3->color_16bit && (svga->bpp < 24)) {
-                    if (!s3->accel.b2e8_pix) {
-                        if (!s3->accel.color_16bit_check)
-                            wrt_mask = (s3->accel.wrt_mask_actual[0] | (s3->accel.wrt_mask_actual[1] << 8));
-                        else
-                            return;
-                    }
-                }
-
-                s3_log("CMD=%04x, curx=%d, lwrtmask=%04x, actual wrtmask=%04x, frgdmix=%d, "
-                       "bkgdmix=%d, input=%d, cnt=%d.\n", s3->accel.cmd, s3->accel.cur_x,
-                       wrt_mask, s3->accel.wrt_mask, frgd_mix, bkgd_mix, cpu_input, count);
-                while (count-- && s3->accel.sy >= 0) {
-                    if (s3->accel.b2e8_pix && s3_cpu_src(s3) && !s3->accel.temp_cnt) {
-                        mix_dat >>= 16;
-                        s3->accel.temp_cnt = 16;
-                    }
-
-                    if ((s3->accel.cx & 0xfff) >= clip_l && (s3->accel.cx & 0xfff) <= clip_r &&
-                        (s3->accel.cy & 0xfff) >= clip_t && (s3->accel.cy & 0xfff) <= clip_b) {
-                        switch ((mix_dat & mix_mask) ? frgd_mix : bkgd_mix) {
-                            case 0:
-                                src_dat = bkgd_color;
-                                if (s3->color_16bit && (svga->bpp < 24) && !s3->accel.b2e8_pix) {
-                                    if (!s3->accel.color_16bit_check)
-                                        src_dat = s3->accel.bkgd_color_actual[0] |
-                                                  (s3->accel.bkgd_color_actual[1] << 8);
-                                }
-                                break;
-                            case 1:
-                                src_dat = frgd_color;
-                                if (s3->color_16bit && (svga->bpp < 24) && !s3->accel.b2e8_pix) {
-                                    if (!s3->accel.color_16bit_check)
-                                        src_dat = s3->accel.frgd_color_actual[0] |
-                                                  (s3->accel.frgd_color_actual[1] << 8);
-                                }
-                                break;
-                            case 2:
-                                src_dat = cpu_dat;
-                                break;
-                            case 3:
-                                src_dat = 0;
-                                break;
-
-                            default:
-                                break;
-                        }
-
-                        READ((s3->accel.cy * s3->width) + s3->accel.cx - s3->accel.minus, dest_dat);
-
-                        MIX
-
-                        if (s3->accel.cmd & 0x10) {
-                            WRITE((s3->accel.cy * s3->width) + s3->accel.cx - s3->accel.minus, dest_dat);
-                        }
-                    }
-
-                    if (s3->accel.b2e8_pix && s3_cpu_src(s3)) {
-                        if (s3->accel.temp_cnt > 0) {
-                            s3->accel.temp_cnt--;
-                            mix_dat <<= 1;
-                            mix_dat |= 1;
-                        }
-                    } else {
-                        mix_dat <<= 1;
-                        mix_dat |= 1;
-                    }
-
-                    if (s3->color_16bit)
-                        cpu_dat >>= 16;
-                    else
-                        cpu_dat >>= 8;
-
-                    if (!s3->accel.sy)
-                        break;
-
-                    if (s3->accel.cmd & 0x40) {
-                        if (s3->accel.cmd & 0x80)
-                            s3->accel.cy++;
-                        else
-                            s3->accel.cy--;
-
-                        if (s3->accel.err_term >= 0) {
-                            s3->accel.err_term += s3->accel.destx_distp;
-                            if (s3->accel.cmd & 0x20)
-                                s3->accel.cx++;
-                            else
-                                s3->accel.cx--;
-                        } else
-                            s3->accel.err_term += s3->accel.desty_axstp;
-                    } else {
-                        if (s3->accel.cmd & 0x20)
-                            s3->accel.cx++;
-                        else
-                            s3->accel.cx--;
-
-                        if (s3->accel.err_term >= 0) {
-                            s3->accel.err_term += s3->accel.destx_distp;
-                            if (s3->accel.cmd & 0x80)
-                                s3->accel.cy++;
-                            else
-                                s3->accel.cy--;
-                        } else
-                            s3->accel.err_term += s3->accel.desty_axstp;
-                    }
-
-                    s3->accel.sy--;
-                    s3->accel.cx &= 0xfff;
-                    s3->accel.cy &= 0xfff;
-                }
-                s3->accel.cur_x = s3->accel.cx;
-                s3->accel.cur_y = s3->accel.cy;
-            }
-            break;
-
-        case 2:             /*Rectangle fill*/
-            if (!cpu_input) { /*!cpu_input is trigger to start operation*/
-                s3->accel.sx = s3->accel.maj_axis_pcnt & 0xfff;
-                s3->accel.sy = s3->accel.multifunc[0] & 0xfff;
-                s3->accel.cx = s3->accel.cur_x & 0xfff;
-                s3->accel.cy = s3->accel.cur_y & 0xfff;
-                s3->accel.pix_trans_x_count = 0;
-
-                s3->accel.dest = s3->accel.cy * s3->width;
-
-                if (s3->color_16bit && (svga->bpp < 24)) {
-                    if (s3->accel.cur_x & 0x400) {
-                        s3->accel.color_16bit_check = 0;
-                        s3->accel.minus = 0x400;
-                    } else {
-                        s3->accel.color_16bit_check = 1;
-                        s3->accel.minus = 0;
-                    }
-
-                    if (s3->accel.color_16bit_check) {
-                        if (((s3->accel.multifunc[0xa] & 0xc0) == 0x00) && !(s3->accel.cmd & 2))
-                            s3->accel.color_16bit_check_rectfill = !!s3_cpu_src(s3);
-                        else
-                            s3->accel.color_16bit_check_rectfill = 0;
-                    }
-
-                    if (s3->accel.color_16bit_check_rectfill) {
-                        if (s3->accel.color_16bit_check) {
-                            s3->accel.pix_trans_ptr = (uint8_t *) calloc(1, (s3->accel.sx + 1) << 1);
-                            s3->accel.pix_trans_ptr_cnt = (s3->accel.sx + 1) << 1;
-                        }
-                    } else
-                        s3->accel.pix_trans_x_count = 0;
-                } else {
-                    s3->accel.pix_trans_x_count = 0;
-                    s3->accel.color_16bit_check = 0;
-                    s3->accel.color_16bit_check_rectfill = 0;
-                    s3->accel.minus = 0;
-                }
-
-                if (s3_cpu_src(s3)) {
-                    s3->data_available = 0;
-                    return; /*Wait for data from CPU*/
-                } else if (s3_cpu_dest(s3)) {
-                    s3->data_available = 1;
-                    return;
-                }
-            }
-
-            frgd_mix = (s3->accel.frgd_mix >> 5) & 3;
-            bkgd_mix = (s3->accel.bkgd_mix >> 5) & 3;
-
-            if (s3->accel.b2e8_pix && s3_cpu_src(s3) && count == 16) { /*Stupid undocumented 0xB2E8 on 911/924*/
-                count              = s3->accel.maj_axis_pcnt + 1;
-                s3->accel.temp_cnt = 16;
-            }
-
-            if (s3->color_16bit && (svga->bpp < 24)) {
-                if (!s3->accel.b2e8_pix) {
-                    if (!s3->accel.color_16bit_check) {
-                        wrt_mask = (s3->accel.wrt_mask_actual[0] | (s3->accel.wrt_mask_actual[1] << 8));
-                    } else if (s3->accel.color_16bit_check && (s3->accel.cmd == 0x40f3))
-                        return;
-                }
-            }
-
-            while (count-- && s3->accel.sy >= 0) {
-                if (s3->accel.b2e8_pix && s3_cpu_src(s3) && !s3->accel.temp_cnt) {
-                    mix_dat >>= 16;
-                    s3->accel.temp_cnt = 16;
-                }
-
-                if ((s3->accel.cx & 0xfff) >= clip_l && (s3->accel.cx & 0xfff) <= clip_r && (s3->accel.cy & 0xfff) >= clip_t && (s3->accel.cy & 0xfff) <= clip_b) {
-                    if (s3_cpu_dest(s3) && ((s3->accel.multifunc[0xa] & 0xc0) == 0x00)) {
-                        mix_dat = mix_mask; /* Mix data = forced to foreground register. */
-                    } else if (s3_cpu_dest(s3) && vram_mask) {
-                        /* Mix data = current video memory value. */
-                        READ(s3->accel.dest + s3->accel.cx - s3->accel.minus, mix_dat);
-                        mix_dat = ((mix_dat & rd_mask) == rd_mask);
-                        mix_dat = mix_dat ? mix_mask : 0;
-                    }
-
-                    if (s3_cpu_dest(s3)) {
-                        READ(s3->accel.dest + s3->accel.cx - s3->accel.minus, src_dat);
-                        if (vram_mask)
-                            src_dat = ((src_dat & rd_mask) == rd_mask);
-                    } else {
-                        switch ((mix_dat & mix_mask) ? frgd_mix : bkgd_mix) {
-                            case 0:
-                                src_dat = bkgd_color;
-                                if (s3->color_16bit && (svga->bpp < 24) && !s3->accel.b2e8_pix && (s3->accel.cmd != 0x41b3)) {
-                                    if (!s3->accel.color_16bit_check)
-                                        src_dat = s3->accel.bkgd_color_actual[0] | (s3->accel.bkgd_color_actual[1] << 8);
-                                }
-                                break;
-                            case 1:
-                                src_dat = frgd_color;
-                                if (s3->color_16bit && (svga->bpp < 24) && !s3->accel.b2e8_pix && (s3->accel.cmd != 0x41b3)) {
-                                    if (!s3->accel.color_16bit_check)
-                                        src_dat = s3->accel.frgd_color_actual[0] | (s3->accel.frgd_color_actual[1] << 8);
-                                }
-                                break;
-                            case 2:
-                                src_dat = cpu_dat;
-                                break;
-                            case 3:
-                                src_dat = 0;
-                                break;
-
-                            default:
-                                break;
-                        }
-                    }
-
-                    READ(s3->accel.dest + s3->accel.cx - s3->accel.minus, dest_dat);
-
-                    MIX
-
-                    if (s3->accel.cmd & 0x10) {
-                        WRITE(s3->accel.dest + s3->accel.cx - s3->accel.minus, dest_dat);
-                    }
-                }
-
-                if (s3->accel.b2e8_pix && s3_cpu_src(s3)) {
-                    if (s3->accel.temp_cnt > 0) {
-                        s3->accel.temp_cnt--;
-                        mix_dat <<= 1;
-                        mix_dat |= 1;
-                    }
-                } else {
-                    mix_dat <<= 1;
-                    mix_dat |= 1;
-                }
-
-                if (s3->color_16bit)
-                    cpu_dat >>= 16;
-                else
-                    cpu_dat >>= 8;
-
-                if (s3->accel.cmd & 0x20)
-                    s3->accel.cx++;
-                else
-                    s3->accel.cx--;
-
-                s3->accel.cx &= 0xfff;
-                s3->accel.sx--;
-                if (s3->accel.sx < 0) {
-                    if (s3->accel.cmd & 0x20)
-                        s3->accel.cx -= (s3->accel.maj_axis_pcnt & 0xfff) + 1;
-                    else
-                        s3->accel.cx += (s3->accel.maj_axis_pcnt & 0xfff) + 1;
-
-                    s3->accel.sx = s3->accel.maj_axis_pcnt & 0xfff;
-                    if (s3->accel.cmd & 0x80)
-                        s3->accel.cy++;
-                    else
-                        s3->accel.cy--;
-
-                    s3->accel.cy &= 0xfff;
-                    s3->accel.dest = s3->accel.cy * s3->width;
-                    s3->accel.sy--;
-
-                    if (cpu_input) {
-                        if (s3->accel.b2e8_pix) {
-                            s3->accel.cur_x = s3->accel.cx;
-                            s3->accel.cur_y = s3->accel.cy;
-                        }
-                        return;
-                    }
-                    if (s3->accel.sy < 0) {
-                        s3->accel.cur_x = s3->accel.cx;
-                        s3->accel.cur_y = s3->accel.cy;
-                        return;
-                    }
-                }
-            }
-            break;
-
-        case 6:             /*BitBlt*/
-            if (!cpu_input) { /*!cpu_input is trigger to start operation*/
-                s3->accel.sx = s3->accel.maj_axis_pcnt & 0xfff;
-                s3->accel.sy = s3->accel.multifunc[0] & 0xfff;
-
-                s3->accel.dx = s3->accel.destx_distp & 0xfff;
-                s3->accel.dy = s3->accel.desty_axstp & 0xfff;
-
-                s3->accel.cx = s3->accel.cur_x & 0xfff;
-                s3->accel.cy = s3->accel.cur_y & 0xfff;
-
-                s3->accel.src  = s3->accel.cy * s3->width;
-                s3->accel.dest = s3->accel.dy * s3->width;
-
-                if (s3->color_16bit && (svga->bpp < 24)) {
-                    if (s3->accel.destx_distp & 0x400) {
-                        s3->accel.color_16bit_check = 0;
-                        s3->accel.minus = 0x400;
-                    } else {
-                        s3->accel.color_16bit_check = 1;
-                        s3->accel.minus = 0;
-                    }
-                    s3->accel.srcminus = 0x400;
-                } else {
-                    s3->accel.color_16bit_check = 0;
-                    s3->accel.minus = 0;
-                    s3->accel.srcminus = 0;
-                }
-            }
-
-            if ((s3->accel.cmd & 0x100) && !cpu_input)
-                return; /*Wait for data from CPU*/
-
-            frgd_mix = (s3->accel.frgd_mix >> 5) & 3;
-            bkgd_mix = (s3->accel.bkgd_mix >> 5) & 3;
-
-            if (s3->color_16bit && (svga->bpp < 24)) {
-                if (!s3->accel.color_16bit_check)
-                    wrt_mask = (s3->accel.wrt_mask_actual[0] | (s3->accel.wrt_mask_actual[1] << 8));
-                else
-                    return;
-            }
-
-            if (!cpu_input && (frgd_mix == 3) && !vram_mask && ((s3->accel.cmd & 0xa0) == 0xa0) && ((s3->accel.frgd_mix & 0xf) == 7) && ((s3->accel.bkgd_mix & 0xf) == 7)) {
-                while (1) {
-                    if ((s3->accel.dx & 0xfff) >= clip_l && (s3->accel.dx & 0xfff) <= clip_r && (s3->accel.dy & 0xfff) >= clip_t && (s3->accel.dy & 0xfff) <= clip_b) {
-                        READ(s3->accel.src + s3->accel.cx - s3->accel.srcminus, src_dat);
-                        READ(s3->accel.dest + s3->accel.dx - s3->accel.minus, dest_dat);
-
-                        dest_dat = (src_dat & wrt_mask) | (dest_dat & ~wrt_mask);
-
-                        WRITE(s3->accel.dest + s3->accel.dx - s3->accel.minus, dest_dat);
-                    }
-
-                    s3->accel.cx++;
-                    s3->accel.dx++;
-                    s3->accel.sx--;
-                    s3->accel.dx &= 0xfff;
-                    if (s3->accel.sx < 0) {
-                        s3->accel.cx -= (s3->accel.maj_axis_pcnt & 0xfff) + 1;
-                        s3->accel.dx -= (s3->accel.maj_axis_pcnt & 0xfff) + 1;
-                        s3->accel.sx = s3->accel.maj_axis_pcnt & 0xfff;
-
-                        s3->accel.cy++;
-                        s3->accel.dy++;
-
-                        s3->accel.dy &= 0xfff;
-                        s3->accel.src  = s3->accel.cy * s3->width;
-                        s3->accel.dest = s3->accel.dy * s3->width;
-
-                        s3->accel.sy--;
-
-                        if (s3->accel.sy < 0) { /*It's evident that this is a clear undocumented difference compared to later chips, per what NT 3.5+ does to DX/DY.*/
-                            s3->accel.destx_distp = s3->accel.dx;
-                            s3->accel.desty_axstp = s3->accel.dy;
-                            return;
-                        }
-                    }
-                }
-            } else {
-                while (count-- && s3->accel.sy >= 0) {
-                    if ((s3->accel.dx & 0xfff) >= clip_l && (s3->accel.dx & 0xfff) <= clip_r && ((s3->accel.dy & 0xfff) >= clip_t && (s3->accel.dy & 0xfff) <= clip_b)) {
-                        if (vram_mask && (s3->accel.cmd & 0x10)) {
-                            READ(s3->accel.src + s3->accel.cx - s3->accel.srcminus, mix_dat);
-                            mix_dat = ((mix_dat & rd_mask) == rd_mask);
-                            mix_dat = mix_dat ? mix_mask : 0;
-                        }
-                        switch ((mix_dat & mix_mask) ? frgd_mix : bkgd_mix) {
-                            case 0:
-                                src_dat = bkgd_color;
-                                if (s3->color_16bit && (svga->bpp < 24)) {
-                                    if (!s3->accel.color_16bit_check)
-                                        src_dat = s3->accel.bkgd_color_actual[0] | (s3->accel.bkgd_color_actual[1] << 8);
-                                }
-                                break;
-                            case 1:
-                                src_dat = frgd_color;
-                                if (s3->color_16bit && (svga->bpp < 24)) {
-                                    if (!s3->accel.color_16bit_check)
-                                        src_dat = s3->accel.frgd_color_actual[0] | (s3->accel.frgd_color_actual[1] << 8);
-                                }
-                                break;
-                            case 2:
-                                src_dat = cpu_dat;
-                                break;
-                            case 3:
-                                READ(s3->accel.src + s3->accel.cx - s3->accel.srcminus, src_dat);
-                                if (vram_mask && (s3->accel.cmd & 0x10))
-                                    src_dat = ((src_dat & rd_mask) == rd_mask);
-                                break;
-
-                            default:
-                                break;
-                        }
-
-                        READ(s3->accel.dest + s3->accel.dx - s3->accel.minus, dest_dat);
-
-                        MIX
-
-                        if ((!(s3->accel.cmd & 0x10) && vram_mask) || (s3->accel.cmd & 0x10)) {
-                            WRITE(s3->accel.dest + s3->accel.dx - s3->accel.minus, dest_dat);
-                        }
-                    }
-
-                    mix_dat <<= 1;
-                    mix_dat |= 1;
-
-                    if (s3->color_16bit)
-                        cpu_dat >>= 16;
-                    else
-                        cpu_dat >>= 8;
-
-                    if (s3->accel.cmd & 0x20) {
-                        s3->accel.cx++;
-                        s3->accel.dx++;
-                    } else {
-                        s3->accel.cx--;
-                        s3->accel.dx--;
-                    }
-                    s3->accel.dx &= 0xfff;
-                    s3->accel.sx--;
-                    if (s3->accel.sx < 0) {
-                        if (s3->accel.cmd & 0x20) {
-                            s3->accel.cx -= ((s3->accel.maj_axis_pcnt & 0xfff) + 1);
-                            s3->accel.dx -= ((s3->accel.maj_axis_pcnt & 0xfff) + 1);
-                        } else {
-                            s3->accel.cx += ((s3->accel.maj_axis_pcnt & 0xfff) + 1);
-                            s3->accel.dx += ((s3->accel.maj_axis_pcnt & 0xfff) + 1);
-                        }
-                        s3->accel.sx = s3->accel.maj_axis_pcnt & 0xfff;
-
-                        if (s3->accel.cmd & 0x80) {
-                            s3->accel.cy++;
-                            s3->accel.dy++;
-                        } else {
-                            s3->accel.cy--;
-                            s3->accel.dy--;
-                        }
-                        s3->accel.dy  &= 0xfff;
-                        s3->accel.src  = s3->accel.cy * s3->width;
-                        s3->accel.dest = s3->accel.dy * s3->width;
-
-                        s3->accel.sy--;
-
-                        if (cpu_input)
-                            return;
-
-                        if (s3->accel.sy < 0) { /*It's evident that this is a clear undocumented difference compared to later chips, per what NT 3.5+ does to DX/DY.*/
-                            s3->accel.destx_distp = s3->accel.dx;
-                            s3->accel.desty_axstp = s3->accel.dy;
-                            return;
-                        }
-                    }
-                }
-            }
-            break;
-
-        default:
-            break;
-    }
-}
-
 void
 s3_short_stroke_start(int count, int cpu_input, uint32_t mix_dat, uint32_t cpu_dat, s3_t *s3, uint8_t ssv)
 {
@@ -8517,20 +7825,17 @@ s3_accel_start(int count, int cpu_input, uint32_t mix_dat, uint32_t cpu_dat, voi
     uint32_t  srcbase;
     uint32_t  dstbase;
 
-    s3->accel.srcminus = 0;
-    s3->accel.minus = 0;
-
     if ((s3->chip >= S3_TRIO64 || s3->chip == S3_VISION968 || s3->chip == S3_VISION868) && (s3->accel.cmd & (1 << 11)))
-        cmd |= 8;
+        cmd |= 0x08;
 
     // SRC-BASE/DST-BASE
-    if ((s3->accel.multifunc[0xd] >> 4) & 7)
-        srcbase = 0x100000 * ((s3->accel.multifunc[0xd] >> 4) & 3);
+    if (((s3->accel.multifunc[0xd] >> 4) & 7) && (s3->chip >= S3_VISION964))
+        srcbase = 0x100000 * ((s3->accel.multifunc[0xd] >> 4) & 7);
     else
         srcbase = 0x100000 * ((s3->accel.multifunc[0xe] >> 2) & 3);
 
-    if ((s3->accel.multifunc[0xd] >> 0) & 7)
-        dstbase = 0x100000 * ((s3->accel.multifunc[0xd] >> 0) & 3);
+    if (((s3->accel.multifunc[0xd] >> 0) & 7) && (s3->chip >= S3_VISION964))
+        dstbase = 0x100000 * ((s3->accel.multifunc[0xd] >> 0) & 7);
     else
         dstbase = 0x100000 * ((s3->accel.multifunc[0xe] >> 0) & 3);
 
@@ -8545,10 +7850,36 @@ s3_accel_start(int count, int cpu_input, uint32_t mix_dat, uint32_t cpu_dat, voi
     if ((s3->accel.cmd & 0x100) && (s3_cpu_src(s3) || (s3_cpu_dest(s3))) && (!cpu_input || (s3_enable_fifo(s3) == 0)))
         s3->force_busy = 1;
 
+    if ((s3->bpp == 0) && s3->color_16bit && !s3->accel.b2e8_pix) {
+        if (cmd <= 2) {
+            if (s3->accel.cur_x & 0x400) {
+                if (s3->accel.cmd != 0x41b3)
+                    wrt_mask = (wrt_mask << 8) & 0xff00;
+                else
+                    wrt_mask &= 0xff;
+
+                frgd_color = (frgd_color << 8) & 0xff00;
+                bkgd_color = (bkgd_color << 8) & 0xff00;
+            } else {
+                if (clip_r >= 0x400) {
+                    wrt_mask &= 0xff;
+                    frgd_color &= 0xff;
+                    bkgd_color &= 0xff;
+                }
+            }
+        } else if (cmd == 6) {
+            if (s3->accel.destx_distp & 0x400) {
+                wrt_mask = (wrt_mask << 8) & 0xff00;
+                frgd_color = (frgd_color << 8) & 0xff00;
+                bkgd_color = (bkgd_color << 8) & 0xff00;
+            }
+        }
+    }
+
     if (!cpu_input)
         s3->accel.dat_count = 0;
 
-    if (cpu_input && (((s3->accel.multifunc[0xa] & 0xc0) != 0x80) || (!(s3->accel.cmd & 2)))) {
+    if (cpu_input && (((s3->accel.multifunc[0xa] & 0xc0) != 0x80) || (!(s3->accel.cmd & 0x02)))) {
         if ((s3->bpp == 3) && (count == 2)) {
             if (s3->accel.dat_count) {
                 cpu_dat             = ((cpu_dat & 0xffff) << 16) | s3->accel.dat_buf;
@@ -8570,9 +7901,9 @@ s3_accel_start(int count, int cpu_input, uint32_t mix_dat, uint32_t cpu_dat, voi
     else if ((s3->bpp == 1) || s3->color_16bit)
         rd_mask &= 0xffff;
 
-    if (s3->bpp == 0)
+    if ((s3->bpp == 0) && !s3->color_16bit)
         compare &= 0xff;
-    else if (s3->bpp == 1)
+    else if ((s3->bpp == 1) || s3->color_16bit)
         compare &= 0xffff;
 
     switch (s3->accel.cmd & 0x600) {
@@ -8593,19 +7924,21 @@ s3_accel_start(int count, int cpu_input, uint32_t mix_dat, uint32_t cpu_dat, voi
             break;
     }
 
+    frgd_mix = (s3->accel.frgd_mix >> 5) & 3;
+    bkgd_mix = (s3->accel.bkgd_mix >> 5) & 3;
+
     /*Bit 4 of the Command register is the draw yes bit, which enables writing to memory/reading from memory when enabled.
       When this bit is disabled, no writing to memory/reading from memory is allowed. (This bit is almost meaningless on
       the NOP command)*/
-    s3_log("CMD=%d, full=%04x, s3bpp=%x, multifuncE=%03x, sourcedisplay=%x, mmio=%02x, srcbase=%08x, dstbase=%08x.\n", cmd, s3->accel.cmd, s3->bpp, s3->accel.multifunc[0x0e] & 0x230, vram_mask, svga->crtc[0x53] & 0x18, srcbase, dstbase);
+    if (s3->accel.cmd == 0x41b3 || (cmd == 6))
+        s3_log("CMD=%d, full=%04x, s3bpp=%x, multifuncE=%03x, sourcedisplay=%x, mmio=%02x, srcbase=%08x, dstbase=%08x, cpu=%04x, mix=%04x, count=%d, rd_mask=%04x, wrt_mask=%04x, width=%d, s=%d,%d, c=%d,%d, d=%d,%d, 16bitcolor=%x, frgdcolor=%04x, bkgdcolor=%04x, frgdsel=%d, bkgdsel=%d, frgdmix=%02x, svgabpp=%d.\n", cmd, s3->accel.cmd, s3->bpp, s3->accel.multifunc[0x0e], vram_mask, svga->crtc[0x53] & 0x18, srcbase, dstbase, cpu_dat & 0xffff, mix_dat & 0xffff, count, rd_mask, wrt_mask, s3->width, s3->accel.sx, s3->accel.sy, s3->accel.cx, s3->accel.cy, s3->accel.dx, s3->accel.dy, s3->color_16bit, frgd_color, bkgd_color, frgd_mix, bkgd_mix, s3->accel.frgd_mix & 0x0f, svga->bpp);
+
     switch (cmd) {
         case 0: /*NOP (Short Stroke Vectors)*/
             if (s3->accel.ssv_state == 0)
                 break;
 
-            frgd_mix = (s3->accel.frgd_mix >> 5) & 3;
-            bkgd_mix = (s3->accel.bkgd_mix >> 5) & 3;
-
-            if (s3->accel.cmd & 8) /*Radial*/
+            if (s3->accel.cmd & 0x08) /*Radial*/
             {
                 while (count-- && s3->accel.ssv_len >= 0) {
                     if ((s3->accel.cx & 0xfff) >= clip_l && (s3->accel.cx & 0xfff) <= clip_r && (s3->accel.cy & 0xfff) >= clip_t && (s3->accel.cy & 0xfff) <= clip_b) {
@@ -8655,7 +7988,7 @@ s3_accel_start(int count, int cpu_input, uint32_t mix_dat, uint32_t cpu_dat, voi
 
                     mix_dat <<= 1;
                     mix_dat |= 1;
-                    if (s3->bpp == 0 && !s3->color_16bit)
+                    if ((s3->bpp == 0) && !s3->color_16bit)
                         cpu_dat >>= 8;
                     else
                         cpu_dat >>= 16;
@@ -8709,17 +8042,18 @@ s3_accel_start(int count, int cpu_input, uint32_t mix_dat, uint32_t cpu_dat, voi
 
         case 1: /*Draw line*/
             if (!cpu_input) {
+                s3->accel.minus = 0;
                 s3->accel.cx = s3->accel.cur_x & 0xfff;
                 s3->accel.cy = s3->accel.cur_y & 0xfff;
                 s3->accel.sy = s3->accel.maj_axis_pcnt;
+                if ((s3->bpp == 0) && s3->color_16bit && (s3->accel.cur_x & 0x400))
+                    s3->accel.minus = 0x400;
 
                 if (s3_cpu_src(s3))
                     return; /*Wait for data from CPU*/
             }
-            frgd_mix = (s3->accel.frgd_mix >> 5) & 3;
-            bkgd_mix = (s3->accel.bkgd_mix >> 5) & 3;
 
-            if (s3->accel.cmd & 8) { /*Radial*/
+            if (s3->accel.cmd & 0x08) { /*Radial*/
                 while (count-- && s3->accel.sy >= 0) {
                     if ((s3->accel.cx & 0xfff) >= clip_l && (s3->accel.cx & 0xfff) <= clip_r && (s3->accel.cy & 0xfff) >= clip_t && (s3->accel.cy & 0xfff) <= clip_b) {
                         switch ((mix_dat & mix_mask) ? frgd_mix : bkgd_mix) {
@@ -8756,11 +8090,11 @@ s3_accel_start(int count, int cpu_input, uint32_t mix_dat, uint32_t cpu_dat, voi
                             update = 1;
 
                         if (update) {
-                            READ((s3->accel.cy * s3->width) + s3->accel.cx, dest_dat);
+                            READ((s3->accel.cy * s3->width) + s3->accel.cx - s3->accel.minus, dest_dat);
 
                             MIX
 
-                            WRITE((s3->accel.cy * s3->width) + s3->accel.cx, dest_dat);
+                            WRITE((s3->accel.cy * s3->width) + s3->accel.cx - s3->accel.minus, dest_dat);
                         }
                     }
 
@@ -8860,11 +8194,11 @@ s3_accel_start(int count, int cpu_input, uint32_t mix_dat, uint32_t cpu_dat, voi
                             update = 1;
 
                         if (update) {
-                            READ((s3->accel.cy * s3->width) + s3->accel.cx, dest_dat);
+                            READ((s3->accel.cy * s3->width) + s3->accel.cx - s3->accel.minus, dest_dat);
 
                             MIX
 
-                            WRITE((s3->accel.cy * s3->width) + s3->accel.cx, dest_dat);
+                            WRITE((s3->accel.cy * s3->width) + s3->accel.cx - s3->accel.minus, dest_dat);
                         }
                     }
 
@@ -8879,7 +8213,7 @@ s3_accel_start(int count, int cpu_input, uint32_t mix_dat, uint32_t cpu_dat, voi
                         mix_dat |= 1;
                     }
 
-                    if (s3->bpp == 0 && !s3->color_16bit)
+                    if ((s3->bpp == 0) && !s3->color_16bit)
                         cpu_dat >>= 8;
                     else
                         cpu_dat >>= 16;
@@ -8926,13 +8260,21 @@ s3_accel_start(int count, int cpu_input, uint32_t mix_dat, uint32_t cpu_dat, voi
             }
             break;
 
-        case 2:             /*Rectangle fill*/
+        case 2: /*Rectangle fill*/
             if (!cpu_input) /*!cpu_input is trigger to start operation*/
             {
+                s3->accel.minus = 0;
                 s3->accel.sx = s3->accel.maj_axis_pcnt & 0xfff;
                 s3->accel.sy = s3->accel.multifunc[0] & 0xfff;
                 s3->accel.cx = s3->accel.cur_x & 0xfff;
                 s3->accel.cy = s3->accel.cur_y & 0xfff;
+                if ((s3->bpp == 0) && s3->color_16bit && (s3->accel.cur_x & 0x400))
+                    s3->accel.minus = 0x400;
+
+                if (s3->accel.cur_x & 0x400)
+                    s3_log("Rectangle Fill + 1024 FULLCMD=%04x: frgdcolor=%04x, s(%d,%d), c(%d,%d).\n", s3->accel.cmd, frgd_color, s3->accel.sx, s3->accel.sy, s3->accel.cx, s3->accel.cy);
+                else
+                    s3_log("Rectangle Fill + 0 FULLCMD=%04x: frgdcolor=%04x, s(%d,%d), c(%d,%d).\n", s3->accel.cmd, frgd_color, s3->accel.sx, s3->accel.sy, s3->accel.cx, s3->accel.cy);
 
                 s3->accel.dest = dstbase + s3->accel.cy * s3->width;
 
@@ -8945,23 +8287,19 @@ s3_accel_start(int count, int cpu_input, uint32_t mix_dat, uint32_t cpu_dat, voi
                 }
             }
 
-            frgd_mix = (s3->accel.frgd_mix >> 5) & 3;
-            bkgd_mix = (s3->accel.bkgd_mix >> 5) & 3;
-
             if (s3->accel.b2e8_pix && s3_cpu_src(s3) && (count == 16)) { /*Stupid undocumented 0xB2E8 on 911/924*/
                 count              = s3->accel.maj_axis_pcnt + 1;
                 s3->accel.temp_cnt = 16;
-                s3_log("CMD2: B2E8 pixtrans, frgdmix=%d, bkgdmix=%d.\n", frgd_mix, bkgd_mix);
             }
 
-            while (count-- && s3->accel.sy >= 0) {
-                if (s3->accel.b2e8_pix && s3_cpu_src(s3) && s3->accel.temp_cnt == 0) {
+            while (count-- && (s3->accel.sy >= 0)) {
+                if (s3->accel.b2e8_pix && s3_cpu_src(s3) && !s3->accel.temp_cnt) {
                     mix_dat >>= 16;
                     s3->accel.temp_cnt = 16;
                 }
 
-                if ((((s3->accel.cx) >= clip_l && (s3->accel.cx) <= clip_r && (s3->accel.cy) >= clip_t && (s3->accel.cy) <= clip_b) && !(s3->accel.multifunc[0xe] & 0x20)) ||
-                    (((s3->accel.cx) < clip_l && (s3->accel.cx) > clip_r && (s3->accel.cy) < clip_t && (s3->accel.cy) > clip_b) && (s3->accel.multifunc[0xe] & 0x20)) ) {
+                if ((((s3->accel.cx >= clip_l) && (s3->accel.cx <= clip_r) && (s3->accel.cy >= clip_t) && (s3->accel.cy <= clip_b)) && !(s3->accel.multifunc[0xe] & 0x20)) ||
+                    (((s3->accel.cx < clip_l) && (s3->accel.cx > clip_r) && (s3->accel.cy < clip_t) && (s3->accel.cy > clip_b)) && (s3->accel.multifunc[0xe] & 0x20)) ) {
                     if (s3_cpu_dest(s3) && ((s3->accel.multifunc[0xa] & 0xc0) == 0x00)) {
                         mix_dat = mix_mask; /* Mix data = forced to foreground register. */
                     } else if (s3_cpu_dest(s3) && vram_mask) {
@@ -8971,8 +8309,8 @@ s3_accel_start(int count, int cpu_input, uint32_t mix_dat, uint32_t cpu_dat, voi
                         mix_dat = mix_dat ? mix_mask : 0;
                     }
 
-                    if (s3_cpu_dest(s3)) {
-                        READ(s3->accel.dest + s3->accel.cx, src_dat);
+                    if (s3_cpu_dest(s3) || ((s3_cpu_src(s3)) && s3->color_16bit && (s3->bpp == 0) && (s3->accel.cmd == 0x41b3))) {
+                        READ(s3->accel.dest + s3->accel.cx - s3->accel.minus, src_dat);
                         if (vram_mask)
                             src_dat = ((src_dat & rd_mask) == rd_mask);
                     } else {
@@ -9011,12 +8349,15 @@ s3_accel_start(int count, int cpu_input, uint32_t mix_dat, uint32_t cpu_dat, voi
                         update = 1;
 
                     if (update) {
-                        READ(s3->accel.dest + s3->accel.cx, dest_dat);
+                        READ(s3->accel.dest + s3->accel.cx - s3->accel.minus, dest_dat);
 
                         MIX
 
                         if (s3->accel.cmd & 0x10) {
-                            WRITE(s3->accel.dest + s3->accel.cx, dest_dat);
+                            if (s3->accel.cmd == 0x41b3)
+                                s3_log("Full=%04x: Destination=%04x, OldDest=%04x, c=%d,%d.\n", s3->accel.cmd, dest_dat, old_dest_dat, s3->accel.cx, s3->accel.cy);
+
+                            WRITE(s3->accel.dest + s3->accel.cx - s3->accel.minus, dest_dat);
                         }
                     }
                 }
@@ -9032,7 +8373,7 @@ s3_accel_start(int count, int cpu_input, uint32_t mix_dat, uint32_t cpu_dat, voi
                     mix_dat |= 1;
                 }
 
-                if (s3->bpp == 0 && !s3->color_16bit)
+                if ((s3->bpp == 0) && !s3->color_16bit)
                     cpu_dat >>= 8;
                 else
                     cpu_dat >>= 16;
@@ -9045,13 +8386,13 @@ s3_accel_start(int count, int cpu_input, uint32_t mix_dat, uint32_t cpu_dat, voi
                 s3->accel.cx &= 0xfff;
                 s3->accel.sx--;
                 if (s3->accel.sx < 0) {
-                    if (s3->accel.cmd & 0x20)
-                        s3->accel.cx -= ((s3->accel.maj_axis_pcnt & 0xfff) + 1);
-                    else
-                        s3->accel.cx += ((s3->accel.maj_axis_pcnt & 0xfff) + 1);
-
                     s3->accel.sx = s3->accel.maj_axis_pcnt & 0xfff;
 
+                    if (s3->accel.cmd & 0x20)
+                        s3->accel.cx -= (s3->accel.sx + 1);
+                    else
+                        s3->accel.cx += (s3->accel.sx + 1);
+
                     if (s3->accel.cmd & 0x80)
                         s3->accel.cy++;
                     else
@@ -9093,8 +8434,6 @@ s3_accel_start(int count, int cpu_input, uint32_t mix_dat, uint32_t cpu_dat, voi
                 end_y1 = s3->accel.desty_axstp;
                 end_y2 = s3->accel.desty_axstp2;
 
-                frgd_mix = (s3->accel.frgd_mix >> 5) & 3;
-
                 while ((s3->accel.poly_cy < end_y1) && (s3->accel.poly_cy2 < end_y2)) {
                     int y       = s3->accel.poly_cy;
                     int x_count = ABS((s3->accel.poly_cx2 >> 20) - s3->accel.poly_x) + 1;
@@ -9177,36 +8516,36 @@ s3_accel_start(int count, int cpu_input, uint32_t mix_dat, uint32_t cpu_dat, voi
 
         case 6: /*BitBlt*/
             if (!cpu_input) { /*!cpu_input is trigger to start operation*/
+                s3->accel.minus = 0;
                 s3->accel.sx = s3->accel.maj_axis_pcnt & 0xfff;
                 s3->accel.sy = s3->accel.multifunc[0] & 0xfff;
 
                 s3->accel.dx = s3->accel.destx_distp & 0xfff;
                 s3->accel.dy = s3->accel.desty_axstp & 0xfff;
 
+                if ((s3->bpp == 0) && s3->color_16bit && (clip_r > 0x3ff) && (s3->accel.destx_distp & 0x400))
+                    s3->accel.minus = 0x400;
+
                 s3->accel.cx = s3->accel.cur_x & 0xfff;
                 s3->accel.cy = s3->accel.cur_y & 0xfff;
 
                 s3->accel.src  = srcbase + s3->accel.cy * s3->width;
                 s3->accel.dest = dstbase + s3->accel.dy * s3->width;
+                s3_log("BitBLT: D(%d,%d).\n", s3->accel.dx, s3->accel.dy);
             }
 
-            if ((s3->accel.cmd & 0x100) && !cpu_input) {
+            if ((s3->accel.cmd & 0x100) && !cpu_input)
                 return; /*Wait for data from CPU*/
-            }
-
-            frgd_mix = (s3->accel.frgd_mix >> 5) & 3;
-            bkgd_mix = (s3->accel.bkgd_mix >> 5) & 3;
 
             if (!cpu_input && (frgd_mix == 3) && !vram_mask && !(s3->accel.multifunc[0xe] & 0x100) && ((s3->accel.cmd & 0xa0) == 0xa0) && ((s3->accel.frgd_mix & 0xf) == 7) && ((s3->accel.bkgd_mix & 0xf) == 7)) {
                 while (1) {
-                    if ((s3->accel.dx) >= clip_l && (s3->accel.dx) <= clip_r && (s3->accel.dy) >= clip_t && (s3->accel.dy) <= clip_b) {
-                        READ(s3->accel.src + s3->accel.cx, src_dat);
-                        READ(s3->accel.dest + s3->accel.dx, dest_dat);
-
+                    if ((s3->accel.dx >= clip_l) && (s3->accel.dx <= clip_r) && (s3->accel.dy >= clip_t) && (s3->accel.dy <= clip_b)) {
+                        READ(s3->accel.src + s3->accel.cx - s3->accel.minus, src_dat);
+                        READ(s3->accel.dest + s3->accel.dx - s3->accel.minus, dest_dat);
                         dest_dat = (src_dat & wrt_mask) | (dest_dat & ~wrt_mask);
 
                         if (s3->accel.cmd & 0x10) {
-                            WRITE(s3->accel.dest + s3->accel.dx, dest_dat);
+                            WRITE(s3->accel.dest + s3->accel.dx - s3->accel.minus, dest_dat);
                         }
                     }
 
@@ -9222,8 +8561,8 @@ s3_accel_start(int count, int cpu_input, uint32_t mix_dat, uint32_t cpu_dat, voi
                         s3->accel.cy++;
                         s3->accel.dy++;
 
-                        s3->accel.src  = srcbase + s3->accel.cy * s3->width;
-                        s3->accel.dest = dstbase + s3->accel.dy * s3->width;
+                        s3->accel.src  = srcbase + (s3->accel.cy * s3->width);
+                        s3->accel.dest = dstbase + (s3->accel.dy * s3->width);
 
                         s3->accel.sy--;
 
@@ -9235,8 +8574,8 @@ s3_accel_start(int count, int cpu_input, uint32_t mix_dat, uint32_t cpu_dat, voi
                     }
                 }
             } else {
-                while (count-- && s3->accel.sy >= 0) {
-                    if ((s3->accel.dx) >= clip_l && (s3->accel.dx) <= clip_r && ((s3->accel.dy) >= clip_t && (s3->accel.dy) <= clip_b)) {
+                while (count-- && (s3->accel.sy >= 0)) {
+                    if ((s3->accel.dx >= clip_l) && (s3->accel.dx <= clip_r) && (s3->accel.dy >= clip_t) && (s3->accel.dy <= clip_b)) {
                         if (vram_mask && (s3->accel.cmd & 0x10)) {
                             READ(s3->accel.src + s3->accel.cx, mix_dat);
                             mix_dat = ((mix_dat & rd_mask) == rd_mask);
@@ -9253,7 +8592,7 @@ s3_accel_start(int count, int cpu_input, uint32_t mix_dat, uint32_t cpu_dat, voi
                                 src_dat = cpu_dat;
                                 break;
                             case 3:
-                                READ(s3->accel.src + s3->accel.cx, src_dat);
+                                READ(s3->accel.src + s3->accel.cx - s3->accel.minus, src_dat);
                                 if (vram_mask && (s3->accel.cmd & 0x10))
                                     src_dat = ((src_dat & rd_mask) == rd_mask);
                                 break;
@@ -9278,12 +8617,12 @@ s3_accel_start(int count, int cpu_input, uint32_t mix_dat, uint32_t cpu_dat, voi
                             update = 1;
 
                         if (update) {
-                            READ(s3->accel.dest + s3->accel.dx, dest_dat);
+                            READ(s3->accel.dest + s3->accel.dx - s3->accel.minus, dest_dat);
 
                             MIX
 
                             if ((!(s3->accel.cmd & 0x10) && vram_mask) || (s3->accel.cmd & 0x10)) {
-                                WRITE(s3->accel.dest + s3->accel.dx, dest_dat);
+                                WRITE(s3->accel.dest + s3->accel.dx - s3->accel.minus, dest_dat);
                             }
                         }
                     }
@@ -9291,7 +8630,7 @@ s3_accel_start(int count, int cpu_input, uint32_t mix_dat, uint32_t cpu_dat, voi
                     mix_dat <<= 1;
                     mix_dat |= 1;
 
-                    if (s3->bpp == 0 && !s3->color_16bit)
+                    if ((s3->bpp == 0) && !s3->color_16bit)
                         cpu_dat >>= 8;
                     else
                         cpu_dat >>= 16;
@@ -9366,9 +8705,6 @@ s3_accel_start(int count, int cpu_input, uint32_t mix_dat, uint32_t cpu_dat, voi
                 return; /*Wait for data from CPU*/
             }
 
-            frgd_mix = (s3->accel.frgd_mix >> 5) & 3;
-            bkgd_mix = (s3->accel.bkgd_mix >> 5) & 3;
-
             while (count-- && s3->accel.sy >= 0) {
                 if ((s3->accel.dx) >= clip_l && (s3->accel.dx) <= clip_r && (s3->accel.dy) >= clip_t && (s3->accel.dy) <= clip_b) {
                     if (vram_mask) {
@@ -9424,7 +8760,7 @@ s3_accel_start(int count, int cpu_input, uint32_t mix_dat, uint32_t cpu_dat, voi
 
                 mix_dat <<= 1;
                 mix_dat |= 1;
-                if (s3->bpp == 0 && !s3->color_16bit)
+                if ((s3->bpp == 0) && !s3->color_16bit)
                     cpu_dat >>= 8;
                 else
                     cpu_dat >>= 16;
@@ -9616,9 +8952,6 @@ s3_accel_start(int count, int cpu_input, uint32_t mix_dat, uint32_t cpu_dat, voi
                 end_y1 = s3->accel.desty_axstp;
                 end_y2 = s3->accel.desty_axstp2;
 
-                frgd_mix = (s3->accel.frgd_mix >> 5) & 3;
-                bkgd_mix = (s3->accel.bkgd_mix >> 5) & 3;
-
                 while ((s3->accel.poly_cy < end_y1) && (s3->accel.poly_cy2 < end_y2)) {
                     int y       = s3->accel.poly_cy;
                     int x_count = ABS((s3->accel.poly_cx2 >> 20) - s3->accel.poly_x) + 1;
@@ -9742,9 +9075,6 @@ s3_accel_start(int count, int cpu_input, uint32_t mix_dat, uint32_t cpu_dat, voi
             if ((s3->accel.cmd & 0x100) && !cpu_input)
                 return; /*Wait for data from CPU*/
 
-            frgd_mix = (s3->accel.frgd_mix >> 5) & 3;
-            bkgd_mix = (s3->accel.bkgd_mix >> 5) & 3;
-
             while (count-- && s3->accel.sy >= 0) {
                 if ((s3->accel.dx & 0xfff) >= clip_l && (s3->accel.dx & 0xfff) <= clip_r && (s3->accel.dy & 0xfff) >= clip_t && (s3->accel.dy & 0xfff) <= clip_b) {
                     switch ((mix_dat & mix_mask) ? frgd_mix : bkgd_mix) {
@@ -10462,6 +9792,7 @@ s3_init(const device_t *info)
     s3->pci = !!(info->flags & DEVICE_PCI);
     s3->vlb = !!(info->flags & DEVICE_VLB);
 
+
     mem_mapping_add(&s3->linear_mapping, 0, 0,
                     svga_read_linear, svga_readw_linear, svga_readl_linear,
                     svga_write_linear, svga_writew_linear, svga_writel_linear,
@@ -10505,6 +9836,8 @@ s3_init(const device_t *info)
                       NULL);
         }
     }
+    mem_mapping_set_handler(&svga->mapping, s3_read, s3_readw, s3_readl, s3_write, s3_writew, s3_writel);
+    mem_mapping_set_p(&svga->mapping, s3);
 
     svga->hwcursor.cur_ysize = 64;
 
@@ -10603,10 +9936,7 @@ s3_init(const device_t *info)
 
     svga->force_old_addr = 1;
 
-    if (s3->chip <= S3_86C924)
-        s3->accel_start = s3_911_accel_start;
-    else
-        s3->accel_start = s3_accel_start;
+    s3->accel_start = s3_accel_start;
 
     switch (s3->card_type) {
         case S3_ORCHID_86C911:
@@ -10617,7 +9947,6 @@ s3_init(const device_t *info)
             s3->id_ext        = stepping;
             s3->id_ext_pci    = 0;
             s3->packed_mmio   = 0;
-            s3->width         = 1024;
 
             svga->ramdac    = device_add(&sc11483_ramdac_device);
             if (s3->card_type == S3_ORCHID_86C911) {
@@ -10637,7 +9966,6 @@ s3_init(const device_t *info)
             s3->id_ext        = stepping;
             s3->id_ext_pci    = 0;
             s3->packed_mmio   = 0;
-            s3->width         = 1024;
 
             svga->ramdac    = device_add(&att490_ramdac_device);
             svga->clock_gen = device_add(&ics2494an_305_device);
diff --git a/src/video/vid_svga.c b/src/video/vid_svga.c
index cecdcf840..32776f1eb 100644
--- a/src/video/vid_svga.c
+++ b/src/video/vid_svga.c
@@ -149,13 +149,13 @@ svga_out(uint16_t addr, uint8_t val, void *priv)
                 case 2:
                     index                 = dev->dac_addr & 0xff;
                     dev->dac_b = val;
-                    svga->vgapal[index].r = dev->dac_r;
-                    svga->vgapal[index].g = dev->dac_g;
-                    svga->vgapal[index].b = dev->dac_b;
+                    dev->_8514pal[index].r = dev->dac_r;
+                    dev->_8514pal[index].g = dev->dac_g;
+                    dev->_8514pal[index].b = dev->dac_b;
                     if (svga->ramdac_type == RAMDAC_8BIT)
-                        dev->pallook[index] = makecol32(svga->vgapal[index].r, svga->vgapal[index].g, svga->vgapal[index].b);
+                        dev->pallook[index] = makecol32(dev->_8514pal[index].r, dev->_8514pal[index].g, dev->_8514pal[index].b);
                     else
-                        dev->pallook[index] = makecol32(video_6to8[svga->vgapal[index].r & 0x3f], video_6to8[svga->vgapal[index].g & 0x3f], video_6to8[svga->vgapal[index].b & 0x3f]);
+                        dev->pallook[index] = makecol32(video_6to8[dev->_8514pal[index].r & 0x3f], video_6to8[dev->_8514pal[index].g & 0x3f], video_6to8[dev->_8514pal[index].b & 0x3f]);
                     dev->dac_pos  = 0;
                     dev->dac_addr = (dev->dac_addr + 1) & 0xff;
                     break;
@@ -221,11 +221,8 @@ svga_out(uint16_t addr, uint8_t val, void *priv)
         case 0x3c3:
             if (xga_active && xga)
                 xga->on = (val & 0x01) ? 0 : 1;
-            if (ibm8514_active && dev) {
-                dev->on[0] = (val & 0x01) ? 0 : 1;
-                if (dev->local & 0xff)
-                    dev->on[1] = dev->on[0];
-            }
+            if (ibm8514_active && dev)
+                dev->on = (val & 0x01) ? 0 : 1;
 
             svga_log("3C3: VGA ON = %d.\n", val & 0x01);
             vga_on = val & 0x01;
@@ -399,24 +396,24 @@ svga_in(uint16_t addr, void *priv)
                 case 0:
                     dev->dac_pos++;
                     if (svga->ramdac_type == RAMDAC_8BIT)
-                        ret = svga->vgapal[index].r;
+                        ret = dev->_8514pal[index].r;
                     else
-                        ret = svga->vgapal[index].r & 0x3f;
+                        ret = dev->_8514pal[index].r & 0x3f;
                     break;
                 case 1:
                     dev->dac_pos++;
                     if (svga->ramdac_type == RAMDAC_8BIT)
-                        ret = svga->vgapal[index].g;
+                        ret = dev->_8514pal[index].g;
                     else
-                        ret = svga->vgapal[index].g & 0x3f;
+                        ret = dev->_8514pal[index].g & 0x3f;
                     break;
                 case 2:
                     dev->dac_pos  = 0;
                     dev->dac_addr = (dev->dac_addr + 1) & 0xff;
                     if (svga->ramdac_type == RAMDAC_8BIT)
-                        ret = svga->vgapal[index].b;
+                        ret = dev->_8514pal[index].b;
                     else
-                        ret = svga->vgapal[index].b & 0x3f;
+                        ret = dev->_8514pal[index].b & 0x3f;
                     break;
 
                 default:
@@ -551,11 +548,11 @@ svga_set_ramdac_type(svga_t *svga, int type)
         for (int c = 0; c < 256; c++) {
             if (ibm8514_active && dev) {
                 if (svga->ramdac_type == RAMDAC_8BIT)
-                    dev->pallook[c] = makecol32(svga->vgapal[c].r, svga->vgapal[c].g, svga->vgapal[c].b);
+                    dev->pallook[c] = makecol32(dev->_8514pal[c].r, dev->_8514pal[c].g, dev->_8514pal[c].b);
                 else
-                    dev->pallook[c] = makecol32((svga->vgapal[c].r & 0x3f) * 4,
-                                                 (svga->vgapal[c].g & 0x3f) * 4,
-                                                 (svga->vgapal[c].b & 0x3f) * 4);
+                    dev->pallook[c] = makecol32((dev->_8514pal[c].r & 0x3f) * 4,
+                                                 (dev->_8514pal[c].g & 0x3f) * 4,
+                                                 (dev->_8514pal[c].b & 0x3f) * 4);
             }
             if (xga_active && xga) {
                 if (svga->ramdac_type == RAMDAC_8BIT)
@@ -829,7 +826,7 @@ svga_recalctimings(svga_t *svga)
 
 #ifdef TBD
     if (ibm8514_active && (svga->dev8514 != NULL)) {
-        if (dev->on[0] || dev->on[1]) {
+        if (dev->on) {
             uint32_t dot8514 = dev->h_blankstart;
             uint32_t adj_dot8514 = dev->h_blankstart;
             uint32_t eff_mask8514 = 0x0000003f;
@@ -867,7 +864,7 @@ svga_recalctimings(svga_t *svga)
 
     crtcconst = svga->clock * svga->char_width;
     if (ibm8514_active && (svga->dev8514 != NULL)) {
-        if (dev->on[0] || dev->on[1])
+        if (dev->on)
             crtcconst8514 = svga->clock8514;
     }
 
@@ -912,7 +909,7 @@ svga_recalctimings(svga_t *svga)
     _dispontime = svga->hdisp_time;
 
     if (ibm8514_active && (svga->dev8514 != NULL)) {
-        if (dev->on[0] || dev->on[1]) {
+        if (dev->on) {
             disptime8514 = dev->h_total ? dev->h_total : TIMER_USEC;
             _dispontime8514 = dev->hdisped;
         }
@@ -935,7 +932,7 @@ svga_recalctimings(svga_t *svga)
         svga->dispofftime = TIMER_USEC;
 
     if (ibm8514_active && (svga->dev8514 != NULL)) {
-        if (dev->on[0] || dev->on[1]) {
+        if (dev->on) {
             _dispofftime8514 = disptime8514 - _dispontime8514;
             _dispontime8514 *= crtcconst8514;
             _dispofftime8514 *= crtcconst8514;
@@ -1464,9 +1461,10 @@ svga_write_common(uint32_t addr, uint8_t val, uint8_t linear, void *priv)
     if (!linear) {
         xga_write_test(addr, val, svga);
         addr = svga_decode_addr(svga, addr, 1);
-
-        if (addr == 0xffffffff)
+        if (addr == 0xffffffff) {
+            svga_log("WriteCommon Over.\n");
             return;
+        }
     }
 
     if (!(svga->gdcreg[6] & 1))
@@ -1498,8 +1496,10 @@ svga_write_common(uint32_t addr, uint8_t val, uint8_t linear, void *priv)
     if (svga->translate_address)
         addr = svga->translate_address(addr, priv);
 
-    if (addr >= svga->vram_max)
+    if (addr >= svga->vram_max) {
+        svga_log("WriteBankedOver=%08x, val=%02x.\n", addr & svga->vram_mask, val);
         return;
+    }
 
     addr &= svga->vram_mask;
 
@@ -1651,7 +1651,6 @@ svga_read_common(uint32_t addr, uint8_t linear, void *priv)
     if (!linear) {
         (void) xga_read_test(addr, svga);
         addr = svga_decode_addr(svga, addr, 0);
-
         if (addr == 0xffffffff)
             return 0xff;
     }
@@ -1873,6 +1872,8 @@ svga_writew_common(uint32_t addr, uint16_t val, uint8_t linear, void *priv)
     cycles -= svga->monitor->mon_video_timing_write_w;
 
     if (!linear) {
+        xga_write_test(addr, val & 0xff, svga);
+        xga_write_test(addr + 1, val >> 8, svga);
         addr = svga_decode_addr(svga, addr, 1);
 
         if (addr == 0xffffffff)
@@ -1929,6 +1930,10 @@ svga_writel_common(uint32_t addr, uint32_t val, uint8_t linear, void *priv)
     cycles -= svga->monitor->mon_video_timing_write_l;
 
     if (!linear) {
+        xga_write_test(addr, val & 0xff, svga);
+        xga_write_test(addr + 1, (val >> 8) & 0xff, svga);
+        xga_write_test(addr + 2, (val >> 16) & 0xff, svga);
+        xga_write_test(addr + 3, (val >> 24) & 0xff, svga);
         addr = svga_decode_addr(svga, addr, 1);
 
         if (addr == 0xffffffff)
@@ -1961,6 +1966,7 @@ svga_writel_common(uint32_t addr, uint32_t val, uint8_t linear, void *priv)
     }
     if (addr >= svga->vram_max)
         return;
+
     addr &= svga->vram_mask;
 
     svga->changedvram[addr >> 12]   = svga->monitor->mon_changeframecount;
@@ -2006,7 +2012,6 @@ svga_readw_common(uint32_t addr, uint8_t linear, void *priv)
 
     if (!linear) {
         addr = svga_decode_addr(svga, addr, 0);
-
         if (addr == 0xffffffff)
             return 0xffff;
     }
@@ -2053,7 +2058,6 @@ svga_readl_common(uint32_t addr, uint8_t linear, void *priv)
 
     if (!linear) {
         addr = svga_decode_addr(svga, addr, 0);
-
         if (addr == 0xffffffff)
             return 0xffffffff;
     }
diff --git a/src/video/vid_xga.c b/src/video/vid_xga.c
index e8d6ef715..e71ed42a7 100644
--- a/src/video/vid_xga.c
+++ b/src/video/vid_xga.c
@@ -54,6 +54,13 @@ static void xga_render_4bpp(svga_t *svga);
 static void xga_render_8bpp(svga_t *svga);
 static void xga_render_16bpp(svga_t *svga);
 
+static void xga_write(uint32_t addr, uint8_t val, void *priv);
+static void xga_writew(uint32_t addr, uint16_t val, void *priv);
+static void xga_writel(uint32_t addr, uint32_t val, void *priv);
+static uint8_t xga_read(uint32_t addr, void *priv);
+static uint16_t xga_readw(uint32_t addr, void *priv);
+static uint32_t xga_readl(uint32_t addr, void *priv);
+
 int xga_active = 0;
 
 #ifdef ENABLE_XGA_LOG
@@ -184,7 +191,7 @@ xga_updatemapping(svga_t *svga)
             xga_log("XGA: Extended Graphics mode.\n");
             switch (xga->aperture_cntl) {
                 case 0:
-                    xga_log("XGA: No 64KB aperture: 1MB=%x, 4MB=%x.\n", xga->base_addr_1mb, xga->linear_base);
+                    xga_log("XGA: No 64KB aperture: 1MB=%x, 4MB=%x, SVGA Mapping Base=%x.\n", xga->base_addr_1mb, xga->linear_base, svga->mapping.base);
                     if (xga->base_addr_1mb) {
                         mem_mapping_set_addr(&xga->linear_mapping, xga->base_addr_1mb, 0x100000);
                         mem_mapping_enable(&xga->linear_mapping);
@@ -194,18 +201,39 @@ xga_updatemapping(svga_t *svga)
                     } else
                         mem_mapping_disable(&xga->linear_mapping);
 
-                    mem_mapping_disable(&xga->video_mapping);
+                    mem_mapping_set_handler(&svga->mapping, svga_read, svga_readw, svga_readl, svga_write, svga_writew, svga_writel);
+                    switch (svga->gdcreg[6] & 0xc) {
+                        case 0x0: /*128k at A0000*/
+                            mem_mapping_set_addr(&svga->mapping, 0xa0000, 0x20000);
+                            svga->banked_mask = 0xffff;
+                            break;
+                        case 0x4: /*64k at A0000*/
+                            mem_mapping_set_addr(&svga->mapping, 0xa0000, 0x10000);
+                            svga->banked_mask = 0xffff;
+                            break;
+                        case 0x8: /*32k at B0000*/
+                            mem_mapping_set_addr(&svga->mapping, 0xb0000, 0x08000);
+                            svga->banked_mask = 0x7fff;
+                            break;
+                        case 0xC: /*32k at B8000*/
+                            mem_mapping_set_addr(&svga->mapping, 0xb8000, 0x08000);
+                            svga->banked_mask = 0x7fff;
+                            break;
+
+                        default:
+                            break;
+                    }
                     break;
                 case 1:
                     xga_log("XGA: 64KB aperture at A0000.\n");
-                    mem_mapping_set_addr(&xga->video_mapping, 0xa0000, 0x10000);
-                    mem_mapping_enable(&xga->video_mapping);
+                    mem_mapping_set_handler(&svga->mapping, xga_read, xga_readw, xga_readl, xga_write, xga_writew, xga_writel);
+                    mem_mapping_set_addr(&svga->mapping, 0xa0000, 0x10000);
                     xga->banked_mask = 0xffff;
                     break;
                 case 2:
                     xga_log("XGA: 64KB aperture at B0000.\n");
-                    mem_mapping_set_addr(&xga->video_mapping, 0xb0000, 0x10000);
-                    mem_mapping_enable(&xga->video_mapping);
+                    mem_mapping_set_handler(&svga->mapping, xga_read, xga_readw, xga_readl, xga_write, xga_writew, xga_writel);
+                    mem_mapping_set_addr(&svga->mapping, 0xb0000, 0x10000);
                     xga->banked_mask = 0xffff;
                     break;
                 default:
@@ -244,7 +272,7 @@ xga_recalctimings(svga_t *svga)
 
         xga->ma_latch = xga->disp_start_addr;
 
-        xga_log("XGA ClkSel1 = %d, ClkSel2 = %02x.\n", (xga->clk_sel_1 >> 2) & 3, xga->clk_sel_2 & 0x80);
+        xga_log("XGA ClkSel1 = %d, ClkSel2 = %02x, dispcntl2=%02x.\n", (xga->clk_sel_1 >> 2) & 3, xga->clk_sel_2 & 0x80, xga->disp_cntl_2 & 0xc0);
         switch ((xga->clk_sel_1 >> 2) & 3) {
             case 0:
                 xga_log("HDISP VGA0 = %d, XGA = %d.\n", svga->hdisp, xga->h_disp);
@@ -798,19 +826,10 @@ xga_ext_inb(uint16_t addr, void *priv)
 #define READW(addr, dat) \
     dat = *(uint16_t *) &xga->vram[(addr) & (xga->vram_mask)];
 
-#define READW_INV(addr, dat)                        \
-    dat = xga->vram[(addr + 1) & (xga->vram_mask)]; \
-    dat |= (xga->vram[(addr) & (xga->vram_mask)] << 8);
-
 #define WRITEW(addr, dat)                                        \
     *(uint16_t *) &xga->vram[((addr)) & (xga->vram_mask)] = dat; \
     xga->changedvram[(((addr)) & (xga->vram_mask)) >> 12] = svga->monitor->mon_changeframecount;
 
-#define WRITEW_INV(addr, dat)                                           \
-    xga->vram[((addr + 1)) & (xga->vram_mask)]            = dat & 0xff; \
-    xga->vram[((addr)) & (xga->vram_mask)]                = dat >> 8;   \
-    xga->changedvram[(((addr)) & (xga->vram_mask)) >> 12] = svga->monitor->mon_changeframecount;
-
 #define ROP(mix, d, s)                                                                 \
     {                                                                                  \
         switch ((mix) ? (xga->accel.frgd_mix & 0x1f) : (xga->accel.bkgd_mix & 0x1f)) { \
@@ -884,7 +903,7 @@ xga_ext_inb(uint16_t addr, void *priv)
     }
 
 static uint32_t
-xga_accel_read_pattern_map_pixel(svga_t *svga, int x, int y, int map, uint32_t base, int width)
+xga_accel_read_pattern_map_pixel(svga_t *svga, int x, int y, uint32_t base, int width)
 {
     const xga_t *xga  = (xga_t *) svga->xga;
     uint32_t     addr = base;
@@ -903,27 +922,19 @@ xga_accel_read_pattern_map_pixel(svga_t *svga, int x, int y, int map, uint32_t b
     } else
         byte = mem_readb_phys(addr);
 
-    if (xga->linear_endian_reverse)
-        bits = 7 - (x & 7);
-    else {
-        if (xga->accel.px_map_format[xga->accel.dst_map] & 8) {
-            if ((xga->accel.px_map_format[xga->accel.src_map] & 8) && (xga->accel.px_map_format[map] & 8))
-                bits = (x & 7);
-            else
-                bits = 7 - (x & 7);
-        } else {
-            if ((xga->accel.px_map_format[map] & 8) && !(xga->access_mode & 8))
-                bits = (x & 7);
-            else
-                bits = 7 - (x & 7);
-        }
+    bits = 7 - (x & 7);
+
+    if (!(xga->accel.px_map_format[xga->accel.src_map] & 0x08) && !(xga->accel.px_map_format[xga->accel.dst_map] & 0x08)) {
+        if (((xga->accel.px_map_format[xga->accel.src_map] & 0x07) >= 0x02) && ((xga->accel.px_map_format[xga->accel.dst_map] & 0x07) >= 0x02))
+            bits ^= 7;
     }
+
     px = (byte >> bits) & 1;
     return px;
 }
 
 static uint32_t
-xga_accel_read_map_pixel(svga_t *svga, int x, int y, int map, uint32_t base, int width, UNUSED(int usesrc))
+xga_accel_read_map_pixel(svga_t *svga, int x, int y, int map, uint32_t base, int width)
 {
     xga_t   *xga  = (xga_t *) svga->xga;
     uint32_t addr = base;
@@ -935,7 +946,7 @@ xga_accel_read_map_pixel(svga_t *svga, int x, int y, int map, uint32_t base, int
     if ((addr < xga->linear_base) || (addr > (xga->linear_base + 0xfffff)))
         skip = 1;
 
-    switch (xga->accel.px_map_format[map] & 7) {
+    switch (xga->accel.px_map_format[map] & 0x07) {
         case 0: /*1-bit*/
             addr += (y * (width >> 3));
             addr += (x >> 3);
@@ -944,14 +955,11 @@ xga_accel_read_map_pixel(svga_t *svga, int x, int y, int map, uint32_t base, int
             } else
                 byte = mem_readb_phys(addr);
 
-            if (xga->linear_endian_reverse)
+            if (xga->accel.px_map_format[map] & 0x08)
+                bits = (x & 7);
+            else
                 bits = 7 - (x & 7);
-            else {
-                if ((xga->accel.px_map_format[map] & 8) && !(xga->access_mode & 8))
-                    bits = (x & 7);
-                else
-                    bits = 7 - (x & 7);
-            }
+
             px = (byte >> bits) & 1;
             return px;
         case 2: /*4-bit*/
@@ -975,17 +983,19 @@ xga_accel_read_map_pixel(svga_t *svga, int x, int y, int map, uint32_t base, int
         case 4: /*16-bit*/
             addr += (y * (width << 1));
             addr += (x << 1);
-            if (xga->linear_endian_reverse) {
-                byte = mem_readw_phys(addr);
-                if ((xga->access_mode & 7) == 4)
-                    byte = ((byte & 0xff00) >> 8) | ((byte & 0x00ff) << 8);
-                else if (xga->access_mode & 8)
-                    byte = ((byte & 0xff00) >> 8) | ((byte & 0x00ff) << 8);
-            } else {
+            if (xga->access_mode & 0x08) {
                 if (!skip) {
                     READW(addr, byte);
                 } else
-                    byte = mem_readb_phys(addr) | (mem_readb_phys(addr + 1) << 8);
+                    byte = mem_readw_phys(addr);
+            } else {
+                if (!skip) {
+                    READW(addr, byte);
+                } else {
+                    byte = mem_readw_phys(addr);
+                    if ((xga->access_mode & 0x07) == 0x04)
+                        byte = ((byte & 0xff00) >> 8) | ((byte & 0x00ff) << 8);
+                }
             }
             return byte;
 
@@ -1007,7 +1017,7 @@ xga_accel_write_map_pixel(svga_t *svga, int x, int y, int map, uint32_t base, ui
     if ((addr < xga->linear_base) || (addr > (xga->linear_base + 0xfffff)))
         skip = 1;
 
-    switch (xga->accel.px_map_format[map] & 7) {
+    switch (xga->accel.px_map_format[map] & 0x07) {
         case 0: /*1-bit*/
             addr += (y * (width >> 3));
             addr += (x >> 3);
@@ -1016,15 +1026,11 @@ xga_accel_write_map_pixel(svga_t *svga, int x, int y, int map, uint32_t base, ui
             } else
                 byte = mem_readb_phys(addr);
 
-            if (xga->linear_endian_reverse) {
+            if ((xga->accel.px_map_format[map] & 0x08) || (xga->accel.px_map_format[xga->accel.src_map] & 0x08))
+                mask = 1 << (x & 7);
+            else
                 mask = 1 << (7 - (x & 7));
-            } else {
-                if ((xga->accel.px_map_format[map] & 8) && !(xga->access_mode & 8)) {
-                    mask = 1 << (x & 7);
-                } else {
-                    mask = 1 << (7 - (x & 7));
-                }
-            }
+
             byte = (byte & ~mask) | ((pixel ? 0xff : 0) & mask);
             if (pixel & 1) {
                 if (!skip) {
@@ -1044,18 +1050,14 @@ xga_accel_write_map_pixel(svga_t *svga, int x, int y, int map, uint32_t base, ui
             addr += (x >> 1);
             if (!skip) {
                 READ(addr, byte);
-            } else {
+            } else
                 byte = mem_readb_phys(addr);
-            }
 
-            if (xga->linear_endian_reverse)
+            if (xga->accel.px_map_format[map] & 0x08)
+                mask = 0x0f << ((x & 1) << 2);
+            else
                 mask = 0x0f << ((1 - (x & 1)) << 2);
-            else {
-                if ((xga->accel.px_map_format[map] & 8) && !(xga->access_mode & 8))
-                    mask = 0x0f << ((x & 1) << 2);
-                else
-                    mask = 0x0f << ((1 - (x & 1)) << 2);
-            }
+
             byte = (byte & ~mask) | (pixel & mask);
             if (!skip) {
                 WRITE(addr, byte);
@@ -1073,14 +1075,16 @@ xga_accel_write_map_pixel(svga_t *svga, int x, int y, int map, uint32_t base, ui
         case 4: /*16-bit*/
             addr += (y * width << 1);
             addr += (x << 1);
-            if (xga->linear_endian_reverse) {
-                if ((xga->access_mode & 7) == 4)
-                    pixel = ((pixel & 0xff00) >> 8) | ((pixel & 0x00ff) << 8);
-                else if (xga->access_mode & 8)
-                    pixel = ((pixel & 0xff00) >> 8) | ((pixel & 0x00ff) << 8);
+            if (xga->access_mode & 0x08) {
+                if (!skip) {
+                    WRITEW(addr, pixel);
+                }
             } else {
                 if (!skip) {
                     WRITEW(addr, pixel);
+                } else {
+                    if ((xga->access_mode & 0x07) == 0x04)
+                        pixel = ((pixel & 0xff00) >> 8) | ((pixel & 0x00ff) << 8);
                 }
             }
             mem_writew_phys(addr, pixel);
@@ -1159,8 +1163,8 @@ xga_short_stroke(svga_t *svga, uint8_t ssv)
         while (y >= 0) {
             if (xga->accel.command & 0xc0) {
                 if ((dx >= xga->accel.mask_map_origin_x_off) && (dx <= ((xga->accel.px_map_width[0] & 0xfff) + xga->accel.mask_map_origin_x_off)) && (dy >= xga->accel.mask_map_origin_y_off) && (dy <= ((xga->accel.px_map_height[0] & 0xfff) + xga->accel.mask_map_origin_y_off))) {
-                    src_dat  = (((xga->accel.command >> 28) & 3) == 2) ? xga_accel_read_map_pixel(svga, xga->accel.src_map_x & 0xfff, xga->accel.src_map_y & 0xfff, xga->accel.src_map, srcbase, xga->accel.px_map_width[xga->accel.src_map] + 1, 1) : xga->accel.frgd_color;
-                    dest_dat = xga_accel_read_map_pixel(svga, dx, dy, xga->accel.dst_map, dstbase, xga->accel.px_map_width[xga->accel.dst_map] + 1, 0);
+                    src_dat  = (((xga->accel.command >> 28) & 3) == 2) ? xga_accel_read_map_pixel(svga, xga->accel.src_map_x & 0xfff, xga->accel.src_map_y & 0xfff, xga->accel.src_map, srcbase, xga->accel.px_map_width[xga->accel.src_map] + 1) : xga->accel.frgd_color;
+                    dest_dat = xga_accel_read_map_pixel(svga, dx, dy, xga->accel.dst_map, dstbase, xga->accel.px_map_width[xga->accel.dst_map] + 1);
 
                     if ((xga->accel.cc_cond == 4) || ((xga->accel.cc_cond == 1) && (dest_dat > color_cmp)) || ((xga->accel.cc_cond == 2) && (dest_dat == color_cmp)) || ((xga->accel.cc_cond == 3) && (dest_dat < color_cmp)) || ((xga->accel.cc_cond == 5) && (dest_dat >= color_cmp)) || ((xga->accel.cc_cond == 6) && (dest_dat != color_cmp)) || ((xga->accel.cc_cond == 7) && (dest_dat <= color_cmp))) {
                         old_dest_dat = dest_dat;
@@ -1179,8 +1183,8 @@ xga_short_stroke(svga_t *svga, uint8_t ssv)
                     }
                 }
             } else {
-                src_dat  = (((xga->accel.command >> 28) & 3) == 2) ? xga_accel_read_map_pixel(svga, xga->accel.src_map_x & 0xfff, xga->accel.src_map_y & 0xfff, xga->accel.src_map, srcbase, xga->accel.px_map_width[xga->accel.src_map] + 1, 1) : xga->accel.frgd_color;
-                dest_dat = xga_accel_read_map_pixel(svga, dx, dy, xga->accel.dst_map, dstbase, xga->accel.px_map_width[xga->accel.dst_map] + 1, 0);
+                src_dat  = (((xga->accel.command >> 28) & 3) == 2) ? xga_accel_read_map_pixel(svga, xga->accel.src_map_x & 0xfff, xga->accel.src_map_y & 0xfff, xga->accel.src_map, srcbase, xga->accel.px_map_width[xga->accel.src_map] + 1) : xga->accel.frgd_color;
+                dest_dat = xga_accel_read_map_pixel(svga, dx, dy, xga->accel.dst_map, dstbase, xga->accel.px_map_width[xga->accel.dst_map] + 1);
 
                 if ((xga->accel.cc_cond == 4) || ((xga->accel.cc_cond == 1) && (dest_dat > color_cmp)) || ((xga->accel.cc_cond == 2) && (dest_dat == color_cmp)) || ((xga->accel.cc_cond == 3) && (dest_dat < color_cmp)) || ((xga->accel.cc_cond == 5) && (dest_dat >= color_cmp)) || ((xga->accel.cc_cond == 6) && (dest_dat != color_cmp)) || ((xga->accel.cc_cond == 7) && (dest_dat <= color_cmp))) {
                     old_dest_dat = dest_dat;
@@ -1270,8 +1274,8 @@ xga_line_draw_write(svga_t *svga)
                 if (xga->accel.command & 0xc0) {
                     if ((dx >= xga->accel.mask_map_origin_x_off) && (dx <= ((xga->accel.px_map_width[0] & 0xfff) + xga->accel.mask_map_origin_x_off)) && (dy >= xga->accel.mask_map_origin_y_off) && (dy <= ((xga->accel.px_map_height[0] & 0xfff) + xga->accel.mask_map_origin_y_off))) {
                         if (draw_pixel) {
-                            src_dat  = (((xga->accel.command >> 28) & 3) == 2) ? xga_accel_read_map_pixel(svga, cx, cy, xga->accel.src_map, srcbase, xga->accel.px_map_width[xga->accel.src_map] + 1, 1) : xga->accel.frgd_color;
-                            dest_dat = xga_accel_read_map_pixel(svga, dx, dy, xga->accel.dst_map, dstbase, xga->accel.px_map_width[xga->accel.dst_map] + 1, 0);
+                            src_dat  = (((xga->accel.command >> 28) & 3) == 2) ? xga_accel_read_map_pixel(svga, cx, cy, xga->accel.src_map, srcbase, xga->accel.px_map_width[xga->accel.src_map] + 1) : xga->accel.frgd_color;
+                            dest_dat = xga_accel_read_map_pixel(svga, dx, dy, xga->accel.dst_map, dstbase, xga->accel.px_map_width[xga->accel.dst_map] + 1);
 
                             if ((xga->accel.cc_cond == 4) || ((xga->accel.cc_cond == 1) && (dest_dat > color_cmp)) || ((xga->accel.cc_cond == 2) && (dest_dat == color_cmp)) || ((xga->accel.cc_cond == 3) && (dest_dat < color_cmp)) || ((xga->accel.cc_cond == 5) && (dest_dat >= color_cmp)) || ((xga->accel.cc_cond == 6) && (dest_dat != color_cmp)) || ((xga->accel.cc_cond == 7) && (dest_dat <= color_cmp))) {
                                 ROP(1, dest_dat, src_dat);
@@ -1281,8 +1285,8 @@ xga_line_draw_write(svga_t *svga)
                     }
                 } else {
                     if (draw_pixel) {
-                        src_dat  = (((xga->accel.command >> 28) & 3) == 2) ? xga_accel_read_map_pixel(svga, cx, cy, xga->accel.src_map, srcbase, xga->accel.px_map_width[xga->accel.src_map] + 1, 1) : xga->accel.frgd_color;
-                        dest_dat = xga_accel_read_map_pixel(svga, dx, dy, xga->accel.dst_map, dstbase, xga->accel.px_map_width[xga->accel.dst_map] + 1, 0);
+                        src_dat  = (((xga->accel.command >> 28) & 3) == 2) ? xga_accel_read_map_pixel(svga, cx, cy, xga->accel.src_map, srcbase, xga->accel.px_map_width[xga->accel.src_map] + 1) : xga->accel.frgd_color;
+                        dest_dat = xga_accel_read_map_pixel(svga, dx, dy, xga->accel.dst_map, dstbase, xga->accel.px_map_width[xga->accel.dst_map] + 1);
 
                         if ((xga->accel.cc_cond == 4) || ((xga->accel.cc_cond == 1) && (dest_dat > color_cmp)) || ((xga->accel.cc_cond == 2) && (dest_dat == color_cmp)) || ((xga->accel.cc_cond == 3) && (dest_dat < color_cmp)) || ((xga->accel.cc_cond == 5) && (dest_dat >= color_cmp)) || ((xga->accel.cc_cond == 6) && (dest_dat != color_cmp)) || ((xga->accel.cc_cond == 7) && (dest_dat <= color_cmp))) {
                             ROP(1, dest_dat, src_dat);
@@ -1330,8 +1334,8 @@ xga_line_draw_write(svga_t *svga)
             while (y >= 0) {
                 if (xga->accel.command & 0xc0) {
                     if ((dx >= xga->accel.mask_map_origin_x_off) && (dx <= ((xga->accel.px_map_width[0] & 0xfff) + xga->accel.mask_map_origin_x_off)) && (dy >= xga->accel.mask_map_origin_y_off) && (dy <= ((xga->accel.px_map_height[0] & 0xfff) + xga->accel.mask_map_origin_y_off))) {
-                        src_dat  = (((xga->accel.command >> 28) & 3) == 2) ? xga_accel_read_map_pixel(svga, xga->accel.src_map_x & 0xfff, xga->accel.src_map_y & 0xfff, xga->accel.src_map, srcbase, xga->accel.px_map_width[xga->accel.src_map] + 1, 1) : xga->accel.frgd_color;
-                        dest_dat = xga_accel_read_map_pixel(svga, dx, dy, xga->accel.dst_map, dstbase, xga->accel.px_map_width[xga->accel.dst_map] + 1, 0);
+                        src_dat  = (((xga->accel.command >> 28) & 3) == 2) ? xga_accel_read_map_pixel(svga, xga->accel.src_map_x & 0xfff, xga->accel.src_map_y & 0xfff, xga->accel.src_map, srcbase, xga->accel.px_map_width[xga->accel.src_map] + 1) : xga->accel.frgd_color;
+                        dest_dat = xga_accel_read_map_pixel(svga, dx, dy, xga->accel.dst_map, dstbase, xga->accel.px_map_width[xga->accel.dst_map] + 1);
 
                         if ((xga->accel.cc_cond == 4) || ((xga->accel.cc_cond == 1) && (dest_dat > color_cmp)) || ((xga->accel.cc_cond == 2) && (dest_dat == color_cmp)) || ((xga->accel.cc_cond == 3) && (dest_dat < color_cmp)) || ((xga->accel.cc_cond == 5) && (dest_dat >= color_cmp)) || ((xga->accel.cc_cond == 6) && (dest_dat != color_cmp)) || ((xga->accel.cc_cond == 7) && (dest_dat <= color_cmp))) {
                             old_dest_dat = dest_dat;
@@ -1346,8 +1350,8 @@ xga_line_draw_write(svga_t *svga)
                         }
                     }
                 } else {
-                    src_dat  = (((xga->accel.command >> 28) & 3) == 2) ? xga_accel_read_map_pixel(svga, xga->accel.src_map_x & 0xfff, xga->accel.src_map_y & 0xfff, xga->accel.src_map, srcbase, xga->accel.px_map_width[xga->accel.src_map] + 1, 1) : xga->accel.frgd_color;
-                    dest_dat = xga_accel_read_map_pixel(svga, dx, dy, xga->accel.dst_map, dstbase, xga->accel.px_map_width[xga->accel.dst_map] + 1, 0);
+                    src_dat  = (((xga->accel.command >> 28) & 3) == 2) ? xga_accel_read_map_pixel(svga, xga->accel.src_map_x & 0xfff, xga->accel.src_map_y & 0xfff, xga->accel.src_map, srcbase, xga->accel.px_map_width[xga->accel.src_map] + 1) : xga->accel.frgd_color;
+                    dest_dat = xga_accel_read_map_pixel(svga, dx, dy, xga->accel.dst_map, dstbase, xga->accel.px_map_width[xga->accel.dst_map] + 1);
 
                     if ((xga->accel.cc_cond == 4) || ((xga->accel.cc_cond == 1) && (dest_dat > color_cmp)) || ((xga->accel.cc_cond == 2) && (dest_dat == color_cmp)) || ((xga->accel.cc_cond == 3) && (dest_dat < color_cmp)) || ((xga->accel.cc_cond == 5) && (dest_dat >= color_cmp)) || ((xga->accel.cc_cond == 6) && (dest_dat != color_cmp)) || ((xga->accel.cc_cond == 7) && (dest_dat <= color_cmp))) {
                         old_dest_dat = dest_dat;
@@ -1465,7 +1469,7 @@ xga_bitblt(svga_t *svga)
         else {
             if ((dstwidth == (xga->h_disp - 1)) && (srcwidth == 1)) {
                 if ((xga->accel.dst_map == 1) && (xga->accel.src_map == 2)) {
-                    if ((xga->accel.px_map_format[xga->accel.dst_map] >= 0x0b) && (xga->accel.px_map_format[xga->accel.src_map] >= 0x0b))
+                    if ((xga->accel.px_map_format[xga->accel.dst_map] >= 0x0a) && (xga->accel.px_map_format[xga->accel.src_map] >= 0x0a))
                         xga->accel.pattern = 1;
                 }
             }
@@ -1486,8 +1490,8 @@ xga_bitblt(svga_t *svga)
         while (xga->accel.y >= 0) {
             if (xga->accel.command & 0xc0) {
                 if ((dx >= xga->accel.mask_map_origin_x_off) && (dx <= ((xga->accel.px_map_width[0] & 0xfff) + xga->accel.mask_map_origin_x_off)) && (dy >= xga->accel.mask_map_origin_y_off) && (dy <= ((xga->accel.px_map_height[0] & 0xfff) + xga->accel.mask_map_origin_y_off))) {
-                    src_dat  = (((xga->accel.command >> 28) & 3) == 2) ? xga_accel_read_map_pixel(svga, xga->accel.sx, xga->accel.sy, xga->accel.src_map, srcbase, srcwidth + 1, 1) : frgdcol;
-                    dest_dat = xga_accel_read_map_pixel(svga, dx, dy, xga->accel.dst_map, dstbase, dstwidth + 1, 0);
+                    src_dat  = (((xga->accel.command >> 28) & 3) == 2) ? xga_accel_read_map_pixel(svga, xga->accel.sx, xga->accel.sy, xga->accel.src_map, srcbase, srcwidth + 1) : frgdcol;
+                    dest_dat = xga_accel_read_map_pixel(svga, dx, dy, xga->accel.dst_map, dstbase, dstwidth + 1);
                     if ((xga->accel.cc_cond == 4) || ((xga->accel.cc_cond == 1) && (dest_dat > color_cmp)) || ((xga->accel.cc_cond == 2) && (dest_dat == color_cmp)) || ((xga->accel.cc_cond == 3) && (dest_dat < color_cmp)) || ((xga->accel.cc_cond == 5) && (dest_dat >= color_cmp)) || ((xga->accel.cc_cond == 6) && (dest_dat != color_cmp)) || ((xga->accel.cc_cond == 7) && (dest_dat <= color_cmp))) {
                         old_dest_dat = dest_dat;
                         ROP(1, dest_dat, src_dat);
@@ -1497,8 +1501,8 @@ xga_bitblt(svga_t *svga)
                 }
             } else {
                 if ((dx >= 0) && (dx <= dstwidth) && (dy >= 0) && (dy <= dstheight)) {
-                    src_dat  = (((xga->accel.command >> 28) & 3) == 2) ? xga_accel_read_map_pixel(svga, xga->accel.sx, xga->accel.sy, xga->accel.src_map, srcbase, srcwidth + 1, 1) : frgdcol;
-                    dest_dat = xga_accel_read_map_pixel(svga, dx, dy, xga->accel.dst_map, dstbase, dstwidth + 1, 0);
+                    src_dat  = (((xga->accel.command >> 28) & 3) == 2) ? xga_accel_read_map_pixel(svga, xga->accel.sx, xga->accel.sy, xga->accel.src_map, srcbase, srcwidth + 1) : frgdcol;
+                    dest_dat = xga_accel_read_map_pixel(svga, dx, dy, xga->accel.dst_map, dstbase, dstwidth + 1);
                     if ((xga->accel.cc_cond == 4) || ((xga->accel.cc_cond == 1) && (dest_dat > color_cmp)) || ((xga->accel.cc_cond == 2) && (dest_dat == color_cmp)) || ((xga->accel.cc_cond == 3) && (dest_dat < color_cmp)) || ((xga->accel.cc_cond == 5) && (dest_dat >= color_cmp)) || ((xga->accel.cc_cond == 6) && (dest_dat != color_cmp)) || ((xga->accel.cc_cond == 7) && (dest_dat <= color_cmp))) {
                         old_dest_dat = dest_dat;
                         ROP(1, dest_dat, src_dat);
@@ -1549,12 +1553,12 @@ xga_bitblt(svga_t *svga)
             if (dstwidth == (xga->h_disp - 1)) {
                 if (srcwidth == (xga->h_disp - 1)) {
                     if ((xga->accel.src_map == 1) && (xga->accel.dst_map == 1) && (xga->accel.pat_src == 2)) {
-                        if ((xga->accel.px_map_format[xga->accel.dst_map] >= 0x0b) && (xga->accel.px <= 7) && (xga->accel.py <= 3))
+                        if ((xga->accel.px_map_format[xga->accel.dst_map] >= 0x0a) && (xga->accel.px <= 7) && (xga->accel.py <= 3))
                             xga->accel.pattern = 1;
                     }
                 } else {
                     if (!xga->accel.src_map && (xga->accel.dst_map == 1) && (xga->accel.pat_src == 2)) {
-                        if ((xga->accel.px_map_format[xga->accel.dst_map] >= 0x0b) && (xga->accel.px <= 7) && (xga->accel.py <= 3)) {
+                        if ((xga->accel.px_map_format[xga->accel.dst_map] >= 0x0a) && (xga->accel.px <= 7) && (xga->accel.py <= 3)) {
                             if ((patwidth >= 7) && ((xga->accel.command & 0xc0) == 0x40))
                                 xga->accel.pattern = 0;
                             else
@@ -1585,15 +1589,15 @@ xga_bitblt(svga_t *svga)
 
         if ((((xga->accel.command >> 24) & 0x0f) == 0x0a) && ((xga->accel.bkgd_mix & 0x1f) == 5)) {
             while (xga->accel.y >= 0) {
-                mix = xga_accel_read_pattern_map_pixel(svga, xga->accel.px, xga->accel.py, xga->accel.pat_src, patbase, patwidth + 1);
+                mix = xga_accel_read_pattern_map_pixel(svga, xga->accel.px, xga->accel.py, patbase, patwidth + 1);
                 if (mix)
                     xga->accel.filling = !xga->accel.filling;
 
                 if (xga->accel.command & 0xc0) {
                     if ((dx >= xga->accel.mask_map_origin_x_off) && (dx <= ((xga->accel.px_map_width[0] & 0xfff) + xga->accel.mask_map_origin_x_off)) && (dy >= xga->accel.mask_map_origin_y_off) && (dy <= ((xga->accel.px_map_height[0] & 0xfff) + xga->accel.mask_map_origin_y_off))) {
-                        src_dat  = (((xga->accel.command >> 28) & 3) == 2) ? xga_accel_read_map_pixel(svga, xga->accel.sx, xga->accel.sy, xga->accel.src_map, srcbase, srcwidth + 1, 1) : frgdcol;
+                        src_dat  = (((xga->accel.command >> 28) & 3) == 2) ? xga_accel_read_map_pixel(svga, xga->accel.sx, xga->accel.sy, xga->accel.src_map, srcbase, srcwidth + 1) : frgdcol;
                         if (xga->accel.filling) {
-                            dest_dat = xga_accel_read_map_pixel(svga, dx, dy, xga->accel.dst_map, dstbase, 1024, 0);
+                            dest_dat = xga_accel_read_map_pixel(svga, dx, dy, xga->accel.dst_map, dstbase, 1024);
                             if ((xga->accel.cc_cond == 4) || ((xga->accel.cc_cond == 1) && (dest_dat > color_cmp)) || ((xga->accel.cc_cond == 2) && (dest_dat == color_cmp)) || ((xga->accel.cc_cond == 3) && (dest_dat < color_cmp)) || ((xga->accel.cc_cond == 5) && (dest_dat >= color_cmp)) || ((xga->accel.cc_cond == 6) && (dest_dat != color_cmp)) || ((xga->accel.cc_cond == 7) && (dest_dat <= color_cmp))) {
                                 old_dest_dat = dest_dat;
                                 ROP(1, dest_dat, src_dat);
@@ -1605,9 +1609,9 @@ xga_bitblt(svga_t *svga)
                     }
                 } else {
                     if ((dx >= 0) && (dx <= dstwidth) && (dy >= 0) && (dy <= dstheight)) {
-                        src_dat  = (((xga->accel.command >> 28) & 3) == 2) ? xga_accel_read_map_pixel(svga, xga->accel.sx, xga->accel.sy, xga->accel.src_map, srcbase, srcwidth + 1, 1) : frgdcol;
+                        src_dat  = (((xga->accel.command >> 28) & 3) == 2) ? xga_accel_read_map_pixel(svga, xga->accel.sx, xga->accel.sy, xga->accel.src_map, srcbase, srcwidth + 1) : frgdcol;
                         if (xga->accel.filling) {
-                            dest_dat = xga_accel_read_map_pixel(svga, dx, dy, xga->accel.dst_map, dstbase, dstwidth + 1, 0);
+                            dest_dat = xga_accel_read_map_pixel(svga, dx, dy, xga->accel.dst_map, dstbase, dstwidth + 1);
                             if ((xga->accel.cc_cond == 4) || ((xga->accel.cc_cond == 1) && (dest_dat > color_cmp)) || ((xga->accel.cc_cond == 2) && (dest_dat == color_cmp)) || ((xga->accel.cc_cond == 3) && (dest_dat < color_cmp)) || ((xga->accel.cc_cond == 5) && (dest_dat >= color_cmp)) || ((xga->accel.cc_cond == 6) && (dest_dat != color_cmp)) || ((xga->accel.cc_cond == 7) && (dest_dat <= color_cmp))) {
                                 old_dest_dat = dest_dat;
                                 ROP(1, dest_dat, src_dat);
@@ -1647,16 +1651,16 @@ xga_bitblt(svga_t *svga)
             }
         } else {
             while (xga->accel.y >= 0) {
-                mix = xga_accel_read_pattern_map_pixel(svga, xga->accel.px, xga->accel.py, xga->accel.pat_src, patbase, patwidth + 1);
+                mix = xga_accel_read_pattern_map_pixel(svga, xga->accel.px, xga->accel.py, patbase, patwidth + 1);
 
                 if (xga->accel.command & 0xc0) {
                     if ((dx >= xga->accel.mask_map_origin_x_off) && (dx <= ((xga->accel.px_map_width[0] & 0xfff) + xga->accel.mask_map_origin_x_off)) && (dy >= xga->accel.mask_map_origin_y_off) && (dy <= ((xga->accel.px_map_height[0] & 0xfff) + xga->accel.mask_map_origin_y_off))) {
                         if (mix)
-                            src_dat = (((xga->accel.command >> 28) & 3) == 2) ? xga_accel_read_map_pixel(svga, xga->accel.sx, xga->accel.sy, xga->accel.src_map, srcbase, srcwidth + 1, 1) : frgdcol;
+                            src_dat = (((xga->accel.command >> 28) & 3) == 2) ? xga_accel_read_map_pixel(svga, xga->accel.sx, xga->accel.sy, xga->accel.src_map, srcbase, srcwidth + 1) : frgdcol;
                         else
-                            src_dat = (((xga->accel.command >> 30) & 3) == 2) ? xga_accel_read_map_pixel(svga, xga->accel.sx, xga->accel.sy, xga->accel.src_map, srcbase, srcwidth + 1, 1) : bkgdcol;
+                            src_dat = (((xga->accel.command >> 30) & 3) == 2) ? xga_accel_read_map_pixel(svga, xga->accel.sx, xga->accel.sy, xga->accel.src_map, srcbase, srcwidth + 1) : bkgdcol;
 
-                        dest_dat = xga_accel_read_map_pixel(svga, dx, dy, xga->accel.dst_map, dstbase, dstwidth + 1, 0);
+                        dest_dat = xga_accel_read_map_pixel(svga, dx, dy, xga->accel.dst_map, dstbase, dstwidth + 1);
                         if ((xga->accel.cc_cond == 4) || ((xga->accel.cc_cond == 1) && (dest_dat > color_cmp)) || ((xga->accel.cc_cond == 2) && (dest_dat == color_cmp)) || ((xga->accel.cc_cond == 3) && (dest_dat < color_cmp)) || ((xga->accel.cc_cond == 5) && (dest_dat >= color_cmp)) || ((xga->accel.cc_cond == 6) && (dest_dat != color_cmp)) || ((xga->accel.cc_cond == 7) && (dest_dat <= color_cmp))) {
                             old_dest_dat = dest_dat;
                             ROP(mix, dest_dat, src_dat);
@@ -1667,11 +1671,11 @@ xga_bitblt(svga_t *svga)
                 } else {
                     if ((dx >= 0) && (dx <= dstwidth) && (dy >= 0) && (dy <= dstheight)) {
                         if (mix)
-                            src_dat = (((xga->accel.command >> 28) & 3) == 2) ? xga_accel_read_map_pixel(svga, xga->accel.sx, xga->accel.sy, xga->accel.src_map, srcbase, srcwidth + 1, 1) : frgdcol;
+                            src_dat = (((xga->accel.command >> 28) & 3) == 2) ? xga_accel_read_map_pixel(svga, xga->accel.sx, xga->accel.sy, xga->accel.src_map, srcbase, srcwidth + 1) : frgdcol;
                         else
-                            src_dat = (((xga->accel.command >> 30) & 3) == 2) ? xga_accel_read_map_pixel(svga, xga->accel.sx, xga->accel.sy, xga->accel.src_map, srcbase, srcwidth + 1, 1) : bkgdcol;
+                            src_dat = (((xga->accel.command >> 30) & 3) == 2) ? xga_accel_read_map_pixel(svga, xga->accel.sx, xga->accel.sy, xga->accel.src_map, srcbase, srcwidth + 1) : bkgdcol;
 
-                        dest_dat = xga_accel_read_map_pixel(svga, dx, dy, xga->accel.dst_map, dstbase, dstwidth + 1, 0);
+                        dest_dat = xga_accel_read_map_pixel(svga, dx, dy, xga->accel.dst_map, dstbase, dstwidth + 1);
                         if ((xga->accel.cc_cond == 4) || ((xga->accel.cc_cond == 1) && (dest_dat > color_cmp)) || ((xga->accel.cc_cond == 2) && (dest_dat == color_cmp)) || ((xga->accel.cc_cond == 3) && (dest_dat < color_cmp)) || ((xga->accel.cc_cond == 5) && (dest_dat >= color_cmp)) || ((xga->accel.cc_cond == 6) && (dest_dat != color_cmp)) || ((xga->accel.cc_cond == 7) && (dest_dat <= color_cmp))) {
                             old_dest_dat = dest_dat;
                             ROP(mix, dest_dat, src_dat);
@@ -1800,6 +1804,10 @@ xga_mem_write(uint32_t addr, uint32_t val, xga_t *xga, svga_t *svga, int len)
 
             case 0x1c:
                 xga->accel.px_map_format[xga->accel.px_map_idx] = val;
+                if (val & 0x08)
+                    xga_log("Big Endian Pixel Format=%d, AccessMode=%x.\n", xga->accel.px_map_idx, xga->access_mode & 0x08);
+                else
+                    xga_log("Little Endian Pixel Format=%d, AccessMode=%x.\n", xga->accel.px_map_idx, xga->access_mode & 0x08);
                 break;
 
             case 0x20:
@@ -2155,6 +2163,7 @@ exec_command:
                     xga->accel.pat_src   = ((xga->accel.command >> 12) & 0x0f);
                     xga->accel.dst_map   = ((xga->accel.command >> 16) & 0x0f);
                     xga->accel.src_map   = ((xga->accel.command >> 20) & 0x0f);
+                    xga_log("PATMAP=%x, DSTMAP=%x, SRCMAP=%x.\n", xga->accel.px_map_format[xga->accel.pat_src], xga->accel.px_map_format[xga->accel.dst_map], xga->accel.px_map_format[xga->accel.src_map]);
 
 #ifdef ENABLE_XGA_LOG
                     if (xga->accel.pat_src)
@@ -2531,29 +2540,21 @@ xga_render_4bpp(svga_t *svga)
 
         xga->lastline_draw = xga->displine;
 
-        for (int x = 0; x <= xga->h_disp; x += 16) {
+        for (int x = 0; x <= xga->h_disp; x += 8) {
             dat  = *(uint32_t *) (&xga->vram[xga->ma & xga->vram_mask]);
             p[0] = xga->pallook[dat & 0x0f];
-            p[1] = xga->pallook[(dat >> 4) & 0x0f];
-            p[2] = xga->pallook[(dat >> 8) & 0x0f];
-            p[3] = xga->pallook[(dat >> 12) & 0x0f];
-            p[4] = xga->pallook[(dat >> 16) & 0x0f];
-            p[5] = xga->pallook[(dat >> 20) & 0x0f];
-            p[6] = xga->pallook[(dat >> 24) & 0x0f];
-            p[7] = xga->pallook[(dat >> 28) & 0x0f];
+            p[1] = xga->pallook[(dat >> 8) & 0x0f];
+            p[2] = xga->pallook[(dat >> 16) & 0x0f];
+            p[3] = xga->pallook[(dat >> 24) & 0x0f];
 
-            dat  = *(uint32_t *) (&xga->vram[(xga->ma + 4) & xga->vram_mask]);
-            p[8] = xga->pallook[dat & 0x0f];
-            p[9] = xga->pallook[(dat >> 4) & 0x0f];
-            p[10] = xga->pallook[(dat >> 8) & 0x0f];
-            p[11] = xga->pallook[(dat >> 12) & 0x0f];
-            p[12] = xga->pallook[(dat >> 16) & 0x0f];
-            p[13] = xga->pallook[(dat >> 20) & 0x0f];
-            p[14] = xga->pallook[(dat >> 24) & 0x0f];
-            p[15] = xga->pallook[(dat >> 28) & 0x0f];
+            dat  = *(uint32_t *) (&xga->vram[(xga->ma + 2) & xga->vram_mask]);
+            p[4] = xga->pallook[dat & 0x0f];
+            p[5] = xga->pallook[(dat >> 8) & 0x0f];
+            p[6] = xga->pallook[(dat >> 16) & 0x0f];
+            p[7] = xga->pallook[(dat >> 24) & 0x0f];
 
             xga->ma += 8;
-            p += 16;
+            p += 8;
         }
         xga->ma &= xga->vram_mask;
     }
@@ -2643,7 +2644,8 @@ xga_write_test(uint32_t addr, uint8_t val, void *priv)
     xga_t  *xga        = (xga_t *) svga->xga;
 
     if (xga_active && xga) {
-        if (((xga->op_mode & 7) >= 1) && (xga->aperture_cntl >= 1)) {
+        if (((xga->op_mode & 7) >= 1) && xga->aperture_cntl) {
+            xga_log("WriteAddr=%05x.\n", addr);
             if (val == 0xa5) { /*Memory size test of XGA*/
                 xga->test    = val;
                 if (addr == 0xa0001)
@@ -2653,28 +2655,39 @@ xga_write_test(uint32_t addr, uint8_t val, void *priv)
 
                 xga->on = 0;
                 vga_on = 1;
-                xga->disp_cntl_2 = 0;
                 xga_log("XGA test1 addr=%05x, test=%02x.\n", addr, xga->a5_test);
             } else if (val == 0x5a) {
                 xga->test = val;
                 xga->on = 0;
                 vga_on = 1;
-                xga->disp_cntl_2 = 0;
                 xga_log("XGA test2 addr = %05x.\n", addr);
             } else if ((addr == 0xa0000) || (addr == 0xa0010)) {
                 addr += xga->write_bank;
                 xga->vram[addr & xga->vram_mask] = val;
                 xga_log("XGA Linear endian reverse write, val = %02x, addr = %05x, banked mask = %04x, a5test=%d.\n", val, addr, svga->banked_mask, xga->a5_test);
-                if (!xga->a5_test)
-                    xga->linear_endian_reverse = 1;
             }
-        } else {
+        } else if (xga->aperture_cntl) {
             xga->on = 0;
             vga_on = 1;
         }
     }
 }
 
+static void
+xga_write_banked(uint32_t addr, uint8_t val, void *priv)
+{
+    svga_t *svga = (svga_t *) priv;
+    xga_t  *xga  = (xga_t *) svga->xga;
+
+    if (xga->access_mode & 0x08) {
+        if ((xga->access_mode & 0x07) == 0x04)
+            addr ^= 1;
+    }
+
+    xga->changedvram[(addr & xga->vram_mask) >> 12] = svga->monitor->mon_changeframecount;
+    xga->vram[addr & xga->vram_mask]                = val;
+}
+
 static void
 xga_write(uint32_t addr, uint8_t val, void *priv)
 {
@@ -2689,33 +2702,45 @@ xga_write(uint32_t addr, uint8_t val, void *priv)
 
     cycles -= svga->monitor->mon_video_timing_write_b;
 
-    if (xga->access_mode & 8) {
-        if ((xga->access_mode & 7) == 4)
-            addr ^= 1;
-    }
-
-    xga->changedvram[(addr & xga->vram_mask) >> 12] = svga->monitor->mon_changeframecount;
-    xga->vram[addr & xga->vram_mask]                = val;
+    xga_write_banked(addr, val, svga);
 }
 
 static void
 xga_writew(uint32_t addr, uint16_t val, void *priv)
 {
-    svga_t      *svga = (svga_t *) priv;
+    svga_t *svga = (svga_t *) priv;
+    xga_t  *xga  = (xga_t *) svga->xga;
 
-    xga_write(addr, val & 0xff, svga);
-    xga_write(addr + 1, val >> 8, svga);
+    addr &= xga->banked_mask;
+    addr += xga->write_bank;
+
+    if (addr >= xga->vram_size)
+        return;
+
+    cycles -= svga->monitor->mon_video_timing_write_w;
+
+    xga_write_banked(addr, val & 0xff, svga);
+    xga_write_banked(addr + 1, val >> 8, svga);
 }
 
 static void
 xga_writel(uint32_t addr, uint32_t val, void *priv)
 {
-    svga_t      *svga = (svga_t *) priv;
+    svga_t *svga = (svga_t *) priv;
+    xga_t  *xga  = (xga_t *) svga->xga;
 
-    xga_write(addr, val & 0xff, svga);
-    xga_write(addr + 1, (val >> 8) & 0xff, svga);
-    xga_write(addr + 2, (val >> 16) & 0xff, svga);
-    xga_write(addr + 3, (val >> 24) & 0xff, svga);
+    addr &= xga->banked_mask;
+    addr += xga->write_bank;
+
+    if (addr >= xga->vram_size)
+        return;
+
+    cycles -= svga->monitor->mon_video_timing_write_l;
+
+    xga_write_banked(addr, val & 0xff, svga);
+    xga_write_banked(addr + 1, val >> 8, svga);
+    xga_write_banked(addr + 2, val >> 16, svga);
+    xga_write_banked(addr + 3, val >> 24, svga);
 }
 
 uint8_t
@@ -2726,7 +2751,7 @@ xga_read_test(uint32_t addr, void *priv)
     uint8_t ret = 0x00;
 
     if (xga_active && xga) {
-        if (((xga->op_mode & 7) >= 1) && (xga->aperture_cntl >= 1)) {
+        if (((xga->op_mode & 7) >= 1) && xga->aperture_cntl) {
             if (xga->test == 0xa5) { /*Memory size test of XGA*/
                 if (addr == 0xa0001) {
                     ret = xga->test;
@@ -2753,7 +2778,7 @@ xga_read_test(uint32_t addr, void *priv)
                 addr += xga->read_bank;
                 return xga->vram[addr & xga->vram_mask];
             }
-        } else {
+        } else if (xga->aperture_cntl) {
             xga->on = 0;
             vga_on = 1;
         }
@@ -2762,12 +2787,29 @@ xga_read_test(uint32_t addr, void *priv)
 }
 
 static uint8_t
-xga_read(uint32_t addr, void *priv)
+xga_read_banked(uint32_t addr, void *priv)
 {
     svga_t      *svga = (svga_t *) priv;
     xga_t *xga  = (xga_t *) svga->xga;
     uint8_t      ret  = 0xff;
 
+    if (xga->access_mode & 0x08) {
+        if ((xga->access_mode & 0x07) == 0x04)
+            addr ^= 1;
+    }
+
+    ret = xga->vram[addr & xga->vram_mask];
+
+    return ret;
+}
+
+static uint8_t
+xga_read(uint32_t addr, void *priv)
+{
+    svga_t      *svga = (svga_t *) priv;
+    xga_t *xga  = (xga_t *) svga->xga;
+    uint8_t     ret  = 0xff;
+
     addr &= xga->banked_mask;
     addr += xga->read_bank;
 
@@ -2778,13 +2820,7 @@ xga_read(uint32_t addr, void *priv)
 
     cycles -= svga->monitor->mon_video_timing_read_b;
 
-    if (xga->access_mode & 8) {
-        if ((xga->access_mode & 7) == 4)
-            addr ^= 1;
-    }
-
-    ret = xga->vram[addr & xga->vram_mask];
-
+    ret = xga_read_banked(addr, svga);
     return ret;
 }
 
@@ -2792,11 +2828,21 @@ static uint16_t
 xga_readw(uint32_t addr, void *priv)
 {
     svga_t      *svga = (svga_t *) priv;
+    xga_t *xga  = (xga_t *) svga->xga;
     uint16_t     ret  = 0xffff;
 
-    ret = xga_read(addr, svga);
-    ret |= (xga_read(addr + 1, svga) << 8);
+    addr &= xga->banked_mask;
+    addr += xga->read_bank;
 
+    if (addr >= xga->vram_size) {
+        xga_log("Over Read ADDR=%x.\n", addr);
+        return ret;
+    }
+
+    cycles -= svga->monitor->mon_video_timing_read_w;
+
+    ret = xga_read_banked(addr, svga);
+    ret |= (xga_read_banked(addr + 1, svga) << 8);
     return ret;
 }
 
@@ -2804,13 +2850,23 @@ static uint32_t
 xga_readl(uint32_t addr, void *priv)
 {
     svga_t      *svga = (svga_t *) priv;
+    xga_t *xga  = (xga_t *) svga->xga;
     uint32_t     ret  = 0xffffffff;
 
-    ret = xga_read(addr, svga);
-    ret |= (xga_read(addr + 1, svga) << 8);
-    ret |= (xga_read(addr + 2, svga) << 16);
-    ret |= (xga_read(addr + 3, svga) << 24);
+    addr &= xga->banked_mask;
+    addr += xga->read_bank;
 
+    if (addr >= xga->vram_size) {
+        xga_log("Over Read ADDR=%x.\n", addr);
+        return ret;
+    }
+
+    cycles -= svga->monitor->mon_video_timing_read_l;
+
+    ret = xga_read_banked(addr, svga);
+    ret |= (xga_read_banked(addr + 1, svga) << 8);
+    ret |= (xga_read_banked(addr + 2, svga) << 16);
+    ret |= (xga_read_banked(addr + 3, svga) << 24);
     return ret;
 }
 
@@ -2820,6 +2876,7 @@ xga_write_linear(uint32_t addr, uint8_t val, void *priv)
     svga_t *svga = (svga_t *) priv;
     xga_t  *xga  = (xga_t *) svga->xga;
 
+    xga_log("WrtieLL XGA=%d.\n", xga->on);
     if (!xga->on) {
         svga_write_linear(addr, val, svga);
         return;
@@ -2834,12 +2891,9 @@ xga_write_linear(uint32_t addr, uint8_t val, void *priv)
 
     cycles -= svga->monitor->mon_video_timing_write_b;
 
-    if (xga->linear_endian_reverse) {
-        if ((xga->access_mode & 7) == 4) {
-            if ((xga->accel.px_map_format[xga->accel.dst_map] & 7) == 4)
-                addr ^= 1;
-        } else if (xga->access_mode & 8) {
-            if ((xga->accel.px_map_format[xga->accel.dst_map] & 7) == 4)
+    if (!(xga->access_mode & 0x08)) {
+        if ((xga->access_mode & 0x07) == 0x04) {
+            if ((xga->accel.px_map_format[xga->accel.dst_map] & 0x07) == 0x04)
                 addr ^= 1;
         }
     }
@@ -2899,12 +2953,9 @@ xga_read_linear(uint32_t addr, void *priv)
 
     cycles -= svga->monitor->mon_video_timing_read_b;
 
-    if (xga->linear_endian_reverse) {
-        if ((xga->access_mode & 7) == 4) {
-            if ((xga->accel.px_map_format[xga->accel.dst_map] & 7) == 4)
-                addr ^= 1;
-        } else if (xga->access_mode & 8) {
-            if ((xga->accel.px_map_format[xga->accel.dst_map] & 7) == 4)
+    if (!(xga->access_mode & 0x08)) {
+        if ((xga->access_mode & 0x07) == 0x04) {
+            if ((xga->accel.px_map_format[xga->accel.dst_map] & 0x07) == 0x04)
                 addr ^= 1;
         }
     }
@@ -3145,11 +3196,12 @@ xga_mca_write(int port, uint8_t val, void *priv)
     mem_mapping_disable(&xga->memio_mapping);
     xga->on                    = 0;
     vga_on                     = 1;
-    xga->linear_endian_reverse = 0;
     xga->a5_test               = 0;
 
     /* Save the MCA register value. */
     xga->pos_regs[port & 7] = val;
+    if (!(xga->pos_regs[4] & 1) && (mem_size >= 15360)) /*MCA 4MB addressing on systems with more than 16MB of memory*/
+        xga->pos_regs[4] |= 1;
 
     if (xga->pos_regs[2] & 1) {
         xga->instance      = (xga->pos_regs[2] & 0x0e) >> 1;
@@ -3184,15 +3236,10 @@ static void
 xga_mca_reset(void *priv)
 {
     svga_t *svga = (svga_t *) priv;
-    xga_t  *xga  = (xga_t *) svga->xga;
 
-    mem_mapping_disable(&xga->memio_mapping);
-    xga->on = 0;
-    vga_on  = 1;
     xga_log("MCA Reset.\n");
+    mem_mapping_set_handler(&svga->mapping, svga_read, svga_readw, svga_readl, svga_write, svga_writew, svga_writel);
     xga_mca_write(0x102, 0, svga);
-    xga->linear_endian_reverse = 0;
-    xga->a5_test               = 0;
 }
 
 static void
@@ -3207,8 +3254,8 @@ xga_reset(void *priv)
 
     xga->on                    = 0;
     vga_on                     = 1;
-    xga->linear_endian_reverse = 0;
     xga->a5_test               = 0;
+    mem_mapping_set_handler(&svga->mapping, svga_read, svga_readw, svga_readl, svga_write, svga_writew, svga_writel);
 }
 
 static uint8_t
@@ -3345,8 +3392,11 @@ xga_pos_out(uint16_t addr, uint8_t val, void *priv)
                 break;
             case 0x0104:
                 xga_log("104Write=%02x.\n", val);
-                if ((xga->pos_idx & 3) == 0)
+                if ((xga->pos_idx & 3) == 0) {
                     xga->pos_regs[4] = val;
+                    if (!(xga->pos_regs[4] & 0x01) && (mem_size >= 15360)) /*4MB addressing on systems with more than 15MB of memory*/
+                        xga->pos_regs[4] |= 0x01;
+                }
                 break;
             case 0x0105:
                 xga_log("105Write=%02x.\n", val);
@@ -3405,7 +3455,6 @@ xga_init(const device_t *info)
     xga->on                    = 0;
     xga->hwcursor.cur_xsize    = 64;
     xga->hwcursor.cur_ysize    = 64;
-    xga->linear_endian_reverse = 0;
     xga->a5_test               = 0;
 
     if (info->flags & DEVICE_MCA) {
@@ -3431,7 +3480,7 @@ xga_init(const device_t *info)
             xga->rom_addr    = 0xc0000 + (((xga->pos_regs[2] & 0xf0) >> 4) * 0x2000);
             xga->instance    = (xga->pos_regs[2] & 0x0e) >> 1;
             xga->pos_regs[2] |= 0x01;
-            if (mem_size >= 16384) {
+            if (mem_size >= 15360) {
                 xga->pos_regs[4] |= 0x01;
                 xga->pos_regs[5] = 0;
             } else {
@@ -3450,9 +3499,6 @@ xga_init(const device_t *info)
         }
     }
 
-    mem_mapping_add(&xga->video_mapping, 0, 0, xga_read, xga_readw, xga_readl,
-                    xga_write, xga_writew, xga_writel,
-                    NULL, MEM_MAPPING_EXTERNAL, svga);
     mem_mapping_add(&xga->linear_mapping, 0, 0, xga_read_linear, xga_readw_linear, xga_readl_linear,
                     xga_write_linear, xga_writew_linear, xga_writel_linear,
                     NULL, MEM_MAPPING_EXTERNAL, svga);

From f280a0ee6e3cd5909f3e031a5138af7e1918e0c0 Mon Sep 17 00:00:00 2001
From: OBattler <oubattler@gmail.com>
Date: Sat, 26 Oct 2024 22:58:38 +0200
Subject: [PATCH 02/26] 3M Microtouch: #include <time.h> to fix the warnings.

---
 src/device/mouse_microtouch_touchscreen.c | 1 +
 1 file changed, 1 insertion(+)

diff --git a/src/device/mouse_microtouch_touchscreen.c b/src/device/mouse_microtouch_touchscreen.c
index 1100937ef..442b25b91 100644
--- a/src/device/mouse_microtouch_touchscreen.c
+++ b/src/device/mouse_microtouch_touchscreen.c
@@ -28,6 +28,7 @@
 #include <stdio.h>
 #include <string.h>
 #include <stdbool.h>
+#include <time.h>
 #include <86box/86box.h>
 #include <86box/device.h>
 #include <86box/timer.h>

From 9109a9d66e50272a8b43b3b258a8963be3c6f325 Mon Sep 17 00:00:00 2001
From: altiereslima <altieres.lima@gmail.com>
Date: Sat, 26 Oct 2024 20:55:26 -0300
Subject: [PATCH 03/26] Update pt-BR.po

---
 src/qt/languages/pt-BR.po | 8 +++++++-
 1 file changed, 7 insertions(+), 1 deletion(-)

diff --git a/src/qt/languages/pt-BR.po b/src/qt/languages/pt-BR.po
index 0b26d5ce5..a9fa7e19b 100644
--- a/src/qt/languages/pt-BR.po
+++ b/src/qt/languages/pt-BR.po
@@ -625,7 +625,7 @@ msgid "Cassette"
 msgstr "Cassete"
 
 msgid "Vision Systems LBA Enhancer"
-msgstr "Aprimorador de Sistema Vision para LBA"
+msgstr "Aprimorador LBA Vision Sistems"
 
 msgid "Hard disks:"
 msgstr "Discos rígidos:"
@@ -981,6 +981,12 @@ msgstr "Não reiniciar"
 msgid "CD-ROM images"
 msgstr "Imagens de CD-ROM"
 
+msgid "Host CD/DVD Drive (%1)"
+msgstr "Unidade de CD/DVD do anfitrião (%1)"
+
+msgid "CD-ROM images"
+msgstr "Imagens de CD-ROM"
+
 msgid "%1 Device Configuration"
 msgstr "Configuração do dispositivo %1"
 

From f32365395764cde34914022bdbe1d61a6fe4c365 Mon Sep 17 00:00:00 2001
From: altiereslima <altieres.lima@gmail.com>
Date: Sun, 27 Oct 2024 00:00:28 -0300
Subject: [PATCH 04/26] ops

---
 src/qt/languages/pt-BR.po | 3 ---
 1 file changed, 3 deletions(-)

diff --git a/src/qt/languages/pt-BR.po b/src/qt/languages/pt-BR.po
index a9fa7e19b..1062d44a9 100644
--- a/src/qt/languages/pt-BR.po
+++ b/src/qt/languages/pt-BR.po
@@ -984,9 +984,6 @@ msgstr "Imagens de CD-ROM"
 msgid "Host CD/DVD Drive (%1)"
 msgstr "Unidade de CD/DVD do anfitrião (%1)"
 
-msgid "CD-ROM images"
-msgstr "Imagens de CD-ROM"
-
 msgid "%1 Device Configuration"
 msgstr "Configuração do dispositivo %1"
 

From 6e212fbd957054b4d12b7c1895b91dae0a8130dd Mon Sep 17 00:00:00 2001
From: altiereslima <altieres.lima@gmail.com>
Date: Sun, 27 Oct 2024 00:07:52 -0300
Subject: [PATCH 05/26] Update pt-BR.po

---
 src/qt/languages/pt-BR.po | 10 +++++-----
 1 file changed, 5 insertions(+), 5 deletions(-)

diff --git a/src/qt/languages/pt-BR.po b/src/qt/languages/pt-BR.po
index 1062d44a9..0cadad77c 100644
--- a/src/qt/languages/pt-BR.po
+++ b/src/qt/languages/pt-BR.po
@@ -13,7 +13,7 @@ msgid "&Keyboard requires capture"
 msgstr "&Teclado requer captura"
 
 msgid "&Right CTRL is left ALT"
-msgstr "CTRL &direito é o ALT esquerdo"
+msgstr "&CTRL direito é o ALT esquerdo"
 
 msgid "&Hard Reset..."
 msgstr "&Reinicialização completa..."
@@ -163,7 +163,7 @@ msgid "Apply fullscreen stretch mode when maximized"
 msgstr "Aplicar modo de ampliação em tela cheia quando maximizado"
 
 msgid "E&GA/(S)VGA settings"
-msgstr "Configurações E&GA/(S)VGA"
+msgstr "Configurações de E&GA/(S)VGA"
 
 msgid "&Inverted VGA monitor"
 msgstr "Monitor VGA &invertido"
@@ -172,7 +172,7 @@ msgid "VGA screen &type"
 msgstr "&Tipo de tela VGA"
 
 msgid "RGB &Color"
-msgstr "&Cores RGB"
+msgstr "&Cor RGB"
 
 msgid "&RGB Grayscale"
 msgstr "Tons de cinza &RGB"
@@ -337,7 +337,7 @@ msgid "Settings"
 msgstr "Configurações"
 
 msgid "Specify Main Window Dimensions"
-msgstr "Especifique as dimensões da janela principal"
+msgstr "Especificar dimensões da janela principal"
 
 msgid "OK"
 msgstr "OK"
@@ -1282,4 +1282,4 @@ msgid "WinBox is no longer supported"
 msgstr "O WinBox não é mais suportado"
 
 msgid "Development of the WinBox manager stopped in 2022 due to a lack of maintainers. As we direct our efforts towards making 86Box even better, we have made the decision to no longer support WinBox as a manager.\n\nNo further updates will be provided through WinBox, and you may encounter incorrect behavior should you continue using it with newer versions of 86Box. Any bug reports related to WinBox behavior will be closed as invalid.\n\nGo to 86box.net for a list of other managers you can use."
-msgstr "O desenvolvimento do gerenciador WinBox foi interrompido em 2022 devido à falta de mantenedores. À medida que direcionamos nossos esforços para tornar o 86Box ainda melhor, tomamos a decisão de não oferecer mais suporte ao WinBox como gerenciador.\n\nNão serão mais fornecidas atualizações através do WinBox, e você poderá encontrar comportamentos incorretos caso continue a usá-lo com versões mais recentes do 86Box. Quaisquer relatórios de bugs relacionados ao comportamento do WinBox serão fechados como inválidos.\n\nAcesse 86box.net para obter uma lista de outros gerenciadores que você pode usar."
+msgstr "O desenvolvimento do gerenciador WinBox foi interrompido em 2022 devido à falta de mantenedores. À medida que direcionamos nossos esforços para tornar o 86Box ainda melhor, tomamos a decisão de não oferecer mais suporte ao WinBox como gerenciador.\n\nNão serão mais fornecidas atualizações através do WinBox, e você poderá encontrar comportamentos incorretos caso continue a usá-lo com versões mais recentes do 86Box. Quaisquer relatórios de erros relacionados ao comportamento do WinBox serão fechados como inválidos.\n\nAcesse 86box.net para obter uma lista de outros gerenciadores que você pode usar."

From 31beadd103e4dde96d2b6977271fd70cd420a521 Mon Sep 17 00:00:00 2001
From: OBattler <oubattler@gmail.com>
Date: Sun, 27 Oct 2024 16:55:05 +0100
Subject: [PATCH 06/26] ISA PnP: Return bit 0 of status register set between
 Wake[0] and Set CSN, fixes Creative ISA PnP sound card drivers installer for
 Windows 3.1 freezing.

---
 src/device/isapnp.c | 9 +++++++--
 1 file changed, 7 insertions(+), 2 deletions(-)

diff --git a/src/device/isapnp.c b/src/device/isapnp.c
index da9e623c1..0139eeb23 100644
--- a/src/device/isapnp.c
+++ b/src/device/isapnp.c
@@ -110,6 +110,7 @@ typedef struct _isapnp_card_ {
 } isapnp_card_t;
 
 typedef struct {
+    uint8_t  in_isolation;
     uint8_t  reg;
     uint8_t  key_pos : 5;
     uint16_t read_data_addr;
@@ -313,6 +314,8 @@ isapnp_read_common(isapnp_t *dev, isapnp_card_t *card, isapnp_device_t *ld, uint
 
         case 0x05: /* Status */
             ret = 0x00;
+            if (dev->in_isolation)
+                ret = 0x01;
             CHECK_CURRENT_CARD();
 
             isapnp_log("ISAPnP: Query status for CSN %02X\n", card->csn);
@@ -485,15 +488,16 @@ isapnp_write_common(isapnp_t *dev, isapnp_card_t *card, isapnp_device_t *ld, uin
         case 0x03: /* Wake[CSN] */
             isapnp_log("ISAPnP: Wake[%02X]\n", val);
             card = dev->first_card;
+            if (val == 0)
+                dev->in_isolation |= 1;
             while (card) {
                 if (card->csn == val) {
                     card->rom_pos     = 0;
                     card->id_checksum = isapnp_init_key[0];
                     if (card->state == PNP_STATE_SLEEP)
                         card->state = (val == 0) ? PNP_STATE_ISOLATION : PNP_STATE_CONFIG;
-                } else {
+                } else
                     card->state = PNP_STATE_SLEEP;
-                }
 
                 card = card->next;
             }
@@ -505,6 +509,7 @@ isapnp_write_common(isapnp_t *dev, isapnp_card_t *card, isapnp_device_t *ld, uin
                 isapnp_set_csn(dev->isolated_card, val);
                 dev->isolated_card->state = PNP_STATE_CONFIG;
                 dev->isolated_card        = NULL;
+                dev->in_isolation         = 0;
             } else {
                 isapnp_log("ISAPnP: Set CSN %02X but no card is isolated\n", val);
             }

From aeaaa09b16fb47c9d805de26bd3ac368a6c9c5ae Mon Sep 17 00:00:00 2001
From: RichardG867 <richardg867@gmail.com>
Date: Sun, 27 Oct 2024 13:17:55 -0300
Subject: [PATCH 07/26] CD-ROM: Report image backend I/O errors as media CRC
 errors to the drive

---
 src/cdrom/cdrom.c               | 73 ++++++++++++++++++++-------------
 src/cdrom/cdrom_image_backend.c | 10 ++---
 src/cdrom/cdrom_image_viso.c    |  7 +++-
 src/cdrom/cdrom_mitsumi.c       |  2 +-
 src/include/86box/scsi_device.h |  3 ++
 src/qt/dummy_cdrom_ioctl.c      |  2 +-
 src/qt/win_cdrom_ioctl.c        |  4 +-
 src/scsi/scsi_cdrom.c           | 13 ++++++
 8 files changed, 75 insertions(+), 39 deletions(-)

diff --git a/src/cdrom/cdrom.c b/src/cdrom/cdrom.c
index 41cc1db98..4dc03885e 100644
--- a/src/cdrom/cdrom.c
+++ b/src/cdrom/cdrom.c
@@ -1520,13 +1520,13 @@ track_type_is_valid(UNUSED(uint8_t id), int type, int flags, int audio, int mode
     return 1;
 }
 
-static void
+static int
 read_sector_to_buffer(cdrom_t *dev, uint8_t *rbuf, uint32_t msf, uint32_t lba, int mode2, int len)
 {
     uint8_t *bb = rbuf;
     const int offset = (!!(mode2 & 0x03)) ? 24 : 16;
 
-    dev->ops->read_sector(dev, CD_READ_DATA, rbuf + offset, lba);
+    int ret = dev->ops->read_sector(dev, CD_READ_DATA, rbuf + offset, lba);
 
     /* Sync bytes */
     bb[0] = 0;
@@ -1546,25 +1546,30 @@ read_sector_to_buffer(cdrom_t *dev, uint8_t *rbuf, uint32_t msf, uint32_t lba, i
         memset(bb, 0, 280);
     else if (!mode2)
         memset(bb, 0, 288);
+
+    return ret;
 }
 
-static void
+static int
 read_audio(cdrom_t *dev, uint32_t lba, uint8_t *b)
 {
-    dev->ops->read_sector(dev, CD_READ_RAW, raw_buffer, lba);
+    int ret = dev->ops->read_sector(dev, CD_READ_RAW, raw_buffer, lba);
 
     memcpy(b, raw_buffer, 2352);
 
     cdrom_sector_size = 2352;
+
+    return ret;
 }
 
-static void
+static int
 read_mode1(cdrom_t *dev, int cdrom_sector_flags, uint32_t lba, uint32_t msf, int mode2, uint8_t *b)
 {
+    int ret;
     if ((dev->cd_status == CD_STATUS_DATA_ONLY) || (dev->ops->sector_size(dev, lba) == 2048))
-        read_sector_to_buffer(dev, raw_buffer, msf, lba, mode2, 2048);
+        ret = read_sector_to_buffer(dev, raw_buffer, msf, lba, mode2, 2048);
     else
-        dev->ops->read_sector(dev, CD_READ_RAW, raw_buffer, lba);
+        ret = dev->ops->read_sector(dev, CD_READ_RAW, raw_buffer, lba);
 
     cdrom_sector_size = 0;
 
@@ -1610,15 +1615,18 @@ read_mode1(cdrom_t *dev, int cdrom_sector_flags, uint32_t lba, uint32_t msf, int
         cdrom_sector_size += 288;
         b += 288;
     }
+
+    return ret;
 }
 
-static void
+static int
 read_mode2_non_xa(cdrom_t *dev, int cdrom_sector_flags, uint32_t lba, uint32_t msf, int mode2, uint8_t *b)
 {
+    int ret;
     if ((dev->cd_status == CD_STATUS_DATA_ONLY) || (dev->ops->sector_size(dev, lba) == 2336))
-        read_sector_to_buffer(dev, raw_buffer, msf, lba, mode2, 2336);
+        ret = read_sector_to_buffer(dev, raw_buffer, msf, lba, mode2, 2336);
     else
-        dev->ops->read_sector(dev, CD_READ_RAW, raw_buffer, lba);
+        ret = dev->ops->read_sector(dev, CD_READ_RAW, raw_buffer, lba);
 
     cdrom_sector_size = 0;
 
@@ -1654,15 +1662,18 @@ read_mode2_non_xa(cdrom_t *dev, int cdrom_sector_flags, uint32_t lba, uint32_t m
         cdrom_sector_size += 2336;
         b += 2336;
     }
+
+    return ret;
 }
 
-static void
+static int
 read_mode2_xa_form1(cdrom_t *dev, int cdrom_sector_flags, uint32_t lba, uint32_t msf, int mode2, uint8_t *b)
 {
+    int ret;
     if ((dev->cd_status == CD_STATUS_DATA_ONLY) || (dev->ops->sector_size(dev, lba) == 2048))
-        read_sector_to_buffer(dev, raw_buffer, msf, lba, mode2, 2048);
+        ret = read_sector_to_buffer(dev, raw_buffer, msf, lba, mode2, 2048);
     else
-        dev->ops->read_sector(dev, CD_READ_RAW, raw_buffer, lba);
+        ret = dev->ops->read_sector(dev, CD_READ_RAW, raw_buffer, lba);
 
     cdrom_sector_size = 0;
 
@@ -1705,15 +1716,18 @@ read_mode2_xa_form1(cdrom_t *dev, int cdrom_sector_flags, uint32_t lba, uint32_t
         cdrom_sector_size += 280;
         b += 280;
     }
+
+    return ret;
 }
 
-static void
+static int
 read_mode2_xa_form2(cdrom_t *dev, int cdrom_sector_flags, uint32_t lba, uint32_t msf, int mode2, uint8_t *b)
 {
+    int ret;
     if ((dev->cd_status == CD_STATUS_DATA_ONLY) || (dev->ops->sector_size(dev, lba) == 2324))
-        read_sector_to_buffer(dev, raw_buffer, msf, lba, mode2, 2324);
+        ret = read_sector_to_buffer(dev, raw_buffer, msf, lba, mode2, 2324);
     else
-        dev->ops->read_sector(dev, CD_READ_RAW, raw_buffer, lba);
+        ret = dev->ops->read_sector(dev, CD_READ_RAW, raw_buffer, lba);
 
     cdrom_sector_size = 0;
 
@@ -1748,6 +1762,8 @@ read_mode2_xa_form2(cdrom_t *dev, int cdrom_sector_flags, uint32_t lba, uint32_t
         cdrom_sector_size += 2328;
         b += 2328;
     }
+
+    return ret;
 }
 
 int
@@ -1763,6 +1779,7 @@ cdrom_readsector_raw(cdrom_t *dev, uint8_t *buffer, int sector, int ismsf, int c
     int      m;
     int      s;
     int      f;
+    int      ret = 0;
 
     if (dev->cd_status == CD_STATUS_EMPTY)
         return 0;
@@ -1827,21 +1844,21 @@ cdrom_readsector_raw(cdrom_t *dev, uint8_t *buffer, int sector, int ismsf, int c
             return 0;
         }
 
-        read_audio(dev, lba, temp_b);
+        ret = read_audio(dev, lba, temp_b);
     } else if (cdrom_sector_type == 2) {
         if (audio || mode2) {
             cdrom_log("CD-ROM %i: [Mode 1] Attempting to read a sector of another type\n", dev->id);
             return 0;
         }
 
-        read_mode1(dev, cdrom_sector_flags, lba, msf, mode2, temp_b);
+        ret = read_mode1(dev, cdrom_sector_flags, lba, msf, mode2, temp_b);
     } else if (cdrom_sector_type == 3) {
         if (audio || !mode2 || (mode2 & 0x03)) {
             cdrom_log("CD-ROM %i: [Mode 2 Formless] Attempting to read a sector of another type\n", dev->id);
             return 0;
         }
 
-        read_mode2_non_xa(dev, cdrom_sector_flags, lba, msf, mode2, temp_b);
+        ret = read_mode2_non_xa(dev, cdrom_sector_flags, lba, msf, mode2, temp_b);
     } else if (cdrom_sector_type == 4) {
         if (audio || !mode2 || ((mode2 & 0x03) != 1)) {
             cdrom_log("CD-ROM %i: [XA Mode 2 Form 1] Attempting to read a sector of another type\n", dev->id);
@@ -1855,7 +1872,7 @@ cdrom_readsector_raw(cdrom_t *dev, uint8_t *buffer, int sector, int ismsf, int c
             return 0;
         }
 
-        read_mode2_xa_form2(dev, cdrom_sector_flags, lba, msf, mode2, temp_b);
+        ret = read_mode2_xa_form2(dev, cdrom_sector_flags, lba, msf, mode2, temp_b);
     } else if (cdrom_sector_type == 8) {
         if (audio) {
             cdrom_log("CD-ROM %i: [Any Data] Attempting to read a data sector from an audio track\n", dev->id);
@@ -1863,9 +1880,9 @@ cdrom_readsector_raw(cdrom_t *dev, uint8_t *buffer, int sector, int ismsf, int c
         }
 
         if (mode2 && ((mode2 & 0x03) == 1))
-            read_mode2_xa_form1(dev, cdrom_sector_flags, lba, msf, mode2, temp_b);
+            ret = read_mode2_xa_form1(dev, cdrom_sector_flags, lba, msf, mode2, temp_b);
         else if (!mode2)
-            read_mode1(dev, cdrom_sector_flags, lba, msf, mode2, temp_b);
+            ret = read_mode1(dev, cdrom_sector_flags, lba, msf, mode2, temp_b);
         else {
             cdrom_log("CD-ROM %i: [Any Data] Attempting to read a data sector whose cooked size is not 2048 bytes\n", dev->id);
             return 0;
@@ -1873,16 +1890,16 @@ cdrom_readsector_raw(cdrom_t *dev, uint8_t *buffer, int sector, int ismsf, int c
     } else {
         if (mode2) {
             if ((mode2 & 0x03) == 0x01)
-                read_mode2_xa_form1(dev, cdrom_sector_flags, lba, msf, mode2, temp_b);
+                ret = read_mode2_xa_form1(dev, cdrom_sector_flags, lba, msf, mode2, temp_b);
             else if ((mode2 & 0x03) == 0x02)
-                read_mode2_xa_form2(dev, cdrom_sector_flags, lba, msf, mode2, temp_b);
+                ret = read_mode2_xa_form2(dev, cdrom_sector_flags, lba, msf, mode2, temp_b);
             else
-                read_mode2_non_xa(dev, cdrom_sector_flags, lba, msf, mode2, temp_b);
+                ret = read_mode2_non_xa(dev, cdrom_sector_flags, lba, msf, mode2, temp_b);
         } else {
             if (audio)
-                read_audio(dev, lba, temp_b);
+                ret = read_audio(dev, lba, temp_b);
             else
-                read_mode1(dev, cdrom_sector_flags, lba, msf, mode2, temp_b);
+                ret = read_mode1(dev, cdrom_sector_flags, lba, msf, mode2, temp_b);
         }
     }
 
@@ -1914,7 +1931,7 @@ cdrom_readsector_raw(cdrom_t *dev, uint8_t *buffer, int sector, int ismsf, int c
 
     *len = cdrom_sector_size;
 
-    return 1;
+    return ret;
 }
 
 /* Peform a master init on the entire module. */
diff --git a/src/cdrom/cdrom_image_backend.c b/src/cdrom/cdrom_image_backend.c
index 7dcde1cdb..a1223bd17 100644
--- a/src/cdrom/cdrom_image_backend.c
+++ b/src/cdrom/cdrom_image_backend.c
@@ -181,13 +181,13 @@ bin_read(void *priv, uint8_t *buffer, uint64_t seek, size_t count)
     if (fseeko64(tf->fp, seek, SEEK_SET) == -1) {
         cdrom_image_backend_log("CDROM: binary_read failed during seek!\n");
 
-        return 0;
+        return -1;
     }
 
     if (fread(buffer, count, 1, tf->fp) != 1) {
         cdrom_image_backend_log("CDROM: binary_read failed during read!\n");
 
-        return 0;
+        return -1;
     }
 
     if (UNLIKELY(tf->motorola)) {
@@ -502,8 +502,8 @@ cdi_read_sector(cd_img_t *cdi, uint8_t *buffer, int raw, uint32_t sector)
     if (raw && !track_is_raw) {
         memset(buffer, 0x00, 2448);
         const int ret = trk->file->read(trk->file, buffer + offset, seek, length);
-        if (!ret)
-            return 0;
+        if (ret <= 0)
+            return ret;
         /* Construct the rest of the raw sector. */
         memset(buffer + 1, 0xff, 10);
         buffer += 12;
@@ -534,7 +534,7 @@ cdi_read_sectors(cd_img_t *cdi, uint8_t *buffer, int raw, uint32_t sector, uint3
 
     for (uint32_t i = 0; i < num; i++) {
         success = cdi_read_sector(cdi, &buf[i * sector_size], raw, sector + i);
-        if (!success)
+        if (success <= 0)
             break;
         /* Based on the DOSBox patch, but check all 8 bytes and makes sure it's not an
            audio track. */
diff --git a/src/cdrom/cdrom_image_viso.c b/src/cdrom/cdrom_image_viso.c
index 90bd83cf7..edca09d91 100644
--- a/src/cdrom/cdrom_image_viso.c
+++ b/src/cdrom/cdrom_image_viso.c
@@ -716,8 +716,11 @@ viso_read(void *priv, uint8_t *buffer, uint64_t seek, size_t count)
                 }
 
                 /* Read data. */
-                if (entry->file && (fseeko64(entry->file, seek - entry->data_offset, SEEK_SET) != -1))
-                    read = fread(buffer, 1, sector_remain, entry->file);
+                if (!entry->file || (fseeko64(entry->file, seek - entry->data_offset, SEEK_SET) == -1))
+                    return -1;
+                read = fread(buffer, 1, sector_remain, entry->file);
+                if (sector_remain && !read)
+                    return -1;
             }
 
             /* Fill remainder with 00 bytes if needed. */
diff --git a/src/cdrom/cdrom_mitsumi.c b/src/cdrom/cdrom_mitsumi.c
index 01a9cf047..0cebfe7f6 100644
--- a/src/cdrom/cdrom_mitsumi.c
+++ b/src/cdrom/cdrom_mitsumi.c
@@ -197,7 +197,7 @@ mitsumi_cdrom_read_sector(mcd_t *dev, int first)
     }
     cdrom_stop(&cdrom);
     ret = cdrom_readsector_raw(&cdrom, dev->buf, cdrom.seek_pos, 0, 2, 0x10, (int *) &dev->readcount, 0);
-    if (!ret)
+    if (ret <= 0)
         return 0;
     if (dev->mode & 0x40) {
         dev->buf[12] = CD_BCD((dev->readmsf >> 16) & 0xff);
diff --git a/src/include/86box/scsi_device.h b/src/include/86box/scsi_device.h
index 64d3bc853..abcd415f4 100644
--- a/src/include/86box/scsi_device.h
+++ b/src/include/86box/scsi_device.h
@@ -189,6 +189,7 @@
 /* SCSI Sense Keys */
 #define SENSE_NONE            0
 #define SENSE_NOT_READY       2
+#define SENSE_MEDIUM_ERROR    3
 #define SENSE_ILLEGAL_REQUEST 5
 #define SENSE_UNIT_ATTENTION  6
 
@@ -196,6 +197,7 @@
 #define ASC_NONE                               0x00
 #define ASC_AUDIO_PLAY_OPERATION               0x00
 #define ASC_NOT_READY                          0x04
+#define ASC_UNRECOVERED_READ_ERROR             0x11
 #define ASC_ILLEGAL_OPCODE                     0x20
 #define ASC_LBA_OUT_OF_RANGE                   0x21
 #define ASC_INV_FIELD_IN_CMD_PACKET            0x24
@@ -212,6 +214,7 @@
 #define ASCQ_NONE                              0x00
 #define ASCQ_UNIT_IN_PROCESS_OF_BECOMING_READY 0x01
 #define ASCQ_INITIALIZING_COMMAND_REQUIRED     0x02
+#define ASCQ_CIRC_UNRECOVERED_ERROR            0x06
 #define ASCQ_CAPACITY_DATA_CHANGED             0x09
 #define ASCQ_AUDIO_PLAY_OPERATION_IN_PROGRESS  0x11
 #define ASCQ_AUDIO_PLAY_OPERATION_PAUSED       0x12
diff --git a/src/qt/dummy_cdrom_ioctl.c b/src/qt/dummy_cdrom_ioctl.c
index 5c533b216..fed1444d1 100644
--- a/src/qt/dummy_cdrom_ioctl.c
+++ b/src/qt/dummy_cdrom_ioctl.c
@@ -207,7 +207,7 @@ plat_cdrom_read_sector(uint8_t *buffer, int raw, uint32_t sector)
     plat_cdrom_close();
     dummy_cdrom_ioctl_log("ReadSector status=%d, sector=%d, size=%" PRId64 ".\n", status, sector, (long long) size);
 
-    return 0;
+    return -1;
 }
 
 void
diff --git a/src/qt/win_cdrom_ioctl.c b/src/qt/win_cdrom_ioctl.c
index ff422bcd2..995876c1d 100644
--- a/src/qt/win_cdrom_ioctl.c
+++ b/src/qt/win_cdrom_ioctl.c
@@ -342,7 +342,7 @@ plat_cdrom_get_sector_size(UNUSED(uint32_t sector))
 int
 plat_cdrom_read_sector(uint8_t *buffer, int raw, uint32_t sector)
 {
-    BOOL status;
+    int  status;
     long size   = 0;
     int  buflen = raw ? RAW_SECTOR_SIZE : COOKED_SECTOR_SIZE;
 
@@ -370,7 +370,7 @@ plat_cdrom_read_sector(uint8_t *buffer, int raw, uint32_t sector)
     plat_cdrom_close();
     win_cdrom_ioctl_log("ReadSector status=%d, sector=%d, size=%" PRId64 ".\n", status, sector, (long long) size);
 
-    return (size == buflen) && (status > 0);
+    return (status > 0) ? (size == buflen) : -1;
 }
 
 void
diff --git a/src/scsi/scsi_cdrom.c b/src/scsi/scsi_cdrom.c
index d014945a7..c527465fc 100644
--- a/src/scsi/scsi_cdrom.c
+++ b/src/scsi/scsi_cdrom.c
@@ -1244,6 +1244,15 @@ scsi_cdrom_not_ready(scsi_cdrom_t *dev)
     scsi_cdrom_cmd_error(dev);
 }
 
+static void
+scsi_cdrom_circ_error(scsi_cdrom_t *dev)
+{
+    scsi_cdrom_sense_key = SENSE_MEDIUM_ERROR;
+    scsi_cdrom_asc       = ASC_UNRECOVERED_READ_ERROR;
+    scsi_cdrom_ascq      = ASCQ_CIRC_UNRECOVERED_ERROR;
+    scsi_cdrom_cmd_error(dev);
+}
+
 static void
 scsi_cdrom_invalid_lun(scsi_cdrom_t *dev)
 {
@@ -1367,6 +1376,10 @@ scsi_cdrom_read_data(scsi_cdrom_t *dev, int msf, int type, int flags, int32_t *l
         if (!ret) {
             scsi_cdrom_illegal_mode(dev);
             return 0;
+        } else if (ret < 0) {
+            pclog("we have a circ\n");
+            scsi_cdrom_circ_error(dev);
+            return -1;
         }
     }
 

From b0e8647753c3a981c3cd87de2805354827e2765b Mon Sep 17 00:00:00 2001
From: richardg867 <richardg867@gmail.com>
Date: Sun, 27 Oct 2024 14:10:28 -0300
Subject: [PATCH 08/26] scsi_cdrom: Remove debug logging

---
 src/scsi/scsi_cdrom.c | 1 -
 1 file changed, 1 deletion(-)

diff --git a/src/scsi/scsi_cdrom.c b/src/scsi/scsi_cdrom.c
index c527465fc..358901847 100644
--- a/src/scsi/scsi_cdrom.c
+++ b/src/scsi/scsi_cdrom.c
@@ -1377,7 +1377,6 @@ scsi_cdrom_read_data(scsi_cdrom_t *dev, int msf, int type, int flags, int32_t *l
             scsi_cdrom_illegal_mode(dev);
             return 0;
         } else if (ret < 0) {
-            pclog("we have a circ\n");
             scsi_cdrom_circ_error(dev);
             return -1;
         }

From 8a7e0ea5477db180d9013cd8c494056a39f313e3 Mon Sep 17 00:00:00 2001
From: TC1995 <mariogplayer90@gmail.com>
Date: Sun, 27 Oct 2024 23:36:11 +0100
Subject: [PATCH 09/26] Small but important video changes of the day (October
 27th, 2024)

8514/A compatibles:
1. The mode switch (from VGA to 8514/A/ATI and vice-versa) has been fixed again (for the Nth time).
2. Removed a pattern hack used on DPCONFIG = 0x5211 on bitblt, now patterns work properly using the ATI Mach8 3.0 win3.1x drivers.
3. Clipping regs are more accurate for acceleration.

XGA 1-2:
A picky OS is a picky OS, fixes to the Win95 fonts (which uses the pattern sources) applied.
---
 src/include/86box/vid_8514a.h |   3 +
 src/video/vid_8514a.c         |  26 ++-
 src/video/vid_ati_mach8.c     | 300 +++++++++++++++++++++-------------
 src/video/vid_xga.c           |   4 +-
 4 files changed, 217 insertions(+), 116 deletions(-)

diff --git a/src/include/86box/vid_8514a.h b/src/include/86box/vid_8514a.h
index 39af01fd0..1c64309cd 100644
--- a/src/include/86box/vid_8514a.h
+++ b/src/include/86box/vid_8514a.h
@@ -198,13 +198,16 @@ typedef struct ibm8514_t {
     int     hsync_width;
     int     htotal;
     int     hdisp;
+    int     hdisp2;
     int     hdisped;
     int     sc;
     int     vsyncstart;
     int     vsyncwidth;
     int     vtotal;
     int     v_disp;
+    int     v_disp2;
     int     vdisp;
+    int     vdisp2;
     int     disp_cntl;
     int     interlace;
     uint8_t subsys_cntl;
diff --git a/src/video/vid_8514a.c b/src/video/vid_8514a.c
index 3e758b952..6d7aa6b94 100644
--- a/src/video/vid_8514a.c
+++ b/src/video/vid_8514a.c
@@ -498,6 +498,24 @@ ibm8514_accel_out_fifo(svga_t *svga, uint16_t port, uint32_t val, int len)
                 dev->accel.multifunc_cntl                             = val;
                 dev->accel.multifunc[dev->accel.multifunc_cntl >> 12] = dev->accel.multifunc_cntl & 0xfff;
                 dev->accel.cmd_back = !!(port == 0xfee8);
+
+                if ((dev->accel.multifunc_cntl >> 12) == 1) {
+                    dev->accel.clip_top = dev->accel.multifunc[1] & 0x3ff;
+                    if (dev->accel.multifunc[1] & 0x400)
+                        dev->accel.clip_top |= ~0x3ff;
+                }
+
+                if ((dev->accel.multifunc_cntl >> 12) == 2) {
+                    dev->accel.clip_left = dev->accel.multifunc[2] & 0x3ff;
+                    if (dev->accel.multifunc[2] & 0x400)
+                        dev->accel.clip_left |= ~0x3ff;
+                }
+                if ((dev->accel.multifunc_cntl >> 12) == 3)
+                    dev->accel.clip_bottom = dev->accel.multifunc[3] & 0x7ff;
+
+                if ((dev->accel.multifunc_cntl >> 12) == 4)
+                    dev->accel.clip_right = dev->accel.multifunc[4] & 0x7ff;
+
             }
             break;
 
@@ -914,10 +932,10 @@ ibm8514_accel_start(int count, int cpu_input, uint32_t mix_dat, uint32_t cpu_dat
     uint16_t   old_dest_dat;
     int        frgd_mix;
     int        bkgd_mix;
-    int16_t    clip_t          = dev->accel.multifunc[1] & 0x7ff;
-    int16_t    clip_l          = dev->accel.multifunc[2] & 0x7ff;
-    uint16_t   clip_b          = dev->accel.multifunc[3] & 0x7ff;
-    uint16_t   clip_r          = dev->accel.multifunc[4] & 0x7ff;
+    int16_t    clip_t          = dev->accel.clip_top;
+    int16_t    clip_l          = dev->accel.clip_left;
+    uint16_t   clip_b          = dev->accel.clip_bottom;
+    uint16_t   clip_r          = dev->accel.clip_right;
     int        pixcntl         = (dev->accel.multifunc[0x0a] >> 6) & 3;
     uint16_t   mix_mask        = dev->bpp ? 0x8000 : 0x80;
     uint16_t   compare         = dev->accel.color_cmp;
diff --git a/src/video/vid_ati_mach8.c b/src/video/vid_ati_mach8.c
index 7ef0d2475..f39dda6d7 100644
--- a/src/video/vid_ati_mach8.c
+++ b/src/video/vid_ati_mach8.c
@@ -309,10 +309,10 @@ mach_accel_start(int cmd_type, int cpu_input, int count, uint32_t mix_dat, uint3
     uint16_t      mix       = 0;
     uint32_t      mono_dat0 = 0;
     uint32_t      mono_dat1 = 0;
-    int16_t      clip_t    = dev->accel.multifunc[1] & 0x7ff;
-    int16_t      clip_l    = dev->accel.multifunc[2] & 0x7ff;
-    int16_t      clip_b    = dev->accel.multifunc[3] & 0x7ff;
-    int16_t      clip_r    = dev->accel.multifunc[4] & 0x7ff;
+    int16_t      clip_t    = dev->accel.clip_top;
+    int16_t      clip_l    = dev->accel.clip_left;
+    int16_t      clip_b    = dev->accel.clip_bottom;
+    int16_t      clip_r    = dev->accel.clip_right;
 
     if (!dev->bpp) {
         rd_mask &= 0xff;
@@ -829,15 +829,7 @@ mach_accel_start(int cmd_type, int cpu_input, int count, uint32_t mix_dat, uint3
                 if (dev->accel.cur_y >= 0x600)
                     dev->accel.dy |= ~0x5ff;
 
-                if (mach->accel.dp_config == 0x5211) {
-                    if (mach->accel.dest_x_end == 1024)
-                        goto skip_dx;
-                }
                 /*Destination Width*/
-                if (mach->accel.dest_x_start != dev->accel.dx)
-                    mach->accel.dest_x_start = dev->accel.dx;
-
-skip_dx:
                 mach->accel.dx_start = mach->accel.dest_x_start;
                 if (mach->accel.dest_x_start >= 0x600)
                     mach->accel.dx_start |= ~0x5ff;
@@ -868,6 +860,7 @@ skip_dx:
                 mach->accel.dy_start = dev->accel.cur_y;
                 if (dev->accel.cur_y >= 0x600)
                     mach->accel.dy_start |= ~0x5ff;
+
                 mach->accel.dy_end = mach->accel.dest_y_end;
                 if (mach->accel.dest_y_end >= 0x600)
                     mach->accel.dy_end |= ~0x5ff;
@@ -995,8 +988,10 @@ skip_dx:
             if (cpu_input) {
                 if (mach->accel.dp_config == 0x3251) {
                     mach_log("DPCONFIG 3251: monosrc=%d, frgdsel=%d, bkgdsel=%d, pitch=%d.\n", mono_src, frgd_sel, bkgd_sel, dev->pitch);
-                    if (dev->accel.sy == mach->accel.height)
+                    if (dev->accel.sy == mach->accel.height) {
+                        mach_log("No Blit on DPCONFIG=3251.\n");
                         return;
+                    }
                 }
             }
 
@@ -2569,6 +2564,7 @@ mach_recalctimings(svga_t *svga)
     }
 
     svga->render8514 = ibm8514_render_blank;
+    mach_log("ON?=%d.\n", dev->on);
     if (dev->on) {
         mach_log("8514/A ON, extpitch=%d, devma=%x, vgamalatch=%x.\n", dev->ext_pitch, dev->ma, svga->ma_latch);
         dev->pitch                      = dev->ext_pitch;
@@ -2578,7 +2574,7 @@ mach_recalctimings(svga_t *svga)
         dev->accel.ge_offset            = (mach->accel.ge_offset_lo | (mach->accel.ge_offset_hi << 16));
         mach->accel.ge_offset           = dev->accel.ge_offset;
 
-        mach_log("HDISP=%d, VDISP=%d, shadowset=%x, 8514/A mode=%x, clocksel=%02x.\n", dev->hdisp, dev->vdisp, mach->shadow_set & 0x03, dev->accel.advfunc_cntl & 0x04, mach->accel.clock_sel & 0xfe);
+        mach_log("HDISP=%d, VDISP=%d, shadowset=%x, 8514/A mode=%x, clocksel=%02x, interlace=%x.\n", dev->hdisp, dev->vdisp, mach->shadow_set & 0x03, dev->accel.advfunc_cntl & 0x04, mach->accel.clock_sel & 0xfe, dev->interlace);
         if ((dev->local & 0xff) >= 0x02) {
             if (dev->bpp || (dev->accel_bpp >= 24) || (mach->accel.clock_sel & 0x01)) {
                 dev->h_disp = dev->hdisp;
@@ -2593,8 +2589,13 @@ mach_recalctimings(svga_t *svga)
                         dev->dispend = 480;
                     }
                 } else {
-                    dev->h_disp = dev->hdisp;
-                    dev->dispend = dev->vdisp;
+                    if (((mach->shadow_set & 0x03) == 0x00) && ((dev->hdisp2 == 640) || (dev->hdisp2 == 1280)) && (dev->hdisp != 800)) {
+                        dev->h_disp = dev->hdisp2;
+                        dev->dispend = dev->vdisp2;
+                    } else {
+                        dev->h_disp = dev->hdisp;
+                        dev->dispend = dev->vdisp;
+                    }
                 }
             }
         } else {
@@ -2619,18 +2620,6 @@ mach_recalctimings(svga_t *svga)
         if (dev->interlace)
             dev->dispend >>= 1;
 
-        if (dev->dispend == 478)
-            dev->dispend += 2;
-
-        if (dev->dispend == 598)
-            dev->dispend += 2;
-
-        if (dev->dispend == 766)
-            dev->dispend += 2;
-
-        if (dev->dispend == 1022)
-            dev->dispend += 2;
-
         if ((dev->local & 0xff) >= 0x02) {
             mach_log("HDISP=%d, mask=%02x.\n", dev->h_disp, dev->dac_mask);
             if ((mach->accel.ext_ge_config & 0x800) || (!(mach->accel.ext_ge_config & 0x8000) && !(mach->accel.ext_ge_config & 0x800))) {
@@ -3116,27 +3105,43 @@ mach_accel_out_fifo(mach_t *mach, svga_t *svga, ibm8514_t *dev, uint16_t port, u
             break;
 
         case 0xdaee:
-            mach_log("DAEE (extclipl) write val = %d\n", val);
-            if (len == 2)
+            mach_log("DAEE (extclipl) write val = %d\n", val & 0x7ff);
+            if (len == 2) {
                 dev->accel.multifunc[2] = val & 0x7ff;
+                dev->accel.clip_left = dev->accel.multifunc[2];
+                if (val & 0x800)
+                    dev->accel.clip_left |= ~0x7ff;
+            }
             break;
 
         case 0xdeee:
-            mach_log("DEEE (extclipt) write val = %d\n", val);
-            if (len == 2)
+            mach_log("DEEE (extclipt) write val = %d\n", val & 0x7ff);
+            if (len == 2) {
                 dev->accel.multifunc[1] = val & 0x7ff;
+                dev->accel.clip_top = dev->accel.multifunc[1];
+                if (val & 0x800)
+                    dev->accel.clip_top |= ~0x7ff;
+            }
             break;
 
         case 0xe2ee:
-            mach_log("E2EE (extclipr) write val = %d\n", val);
-            if (len == 2)
+            mach_log("E2EE (extclipr) write val = %d\n", val & 0x7ff);
+            if (len == 2) {
                 dev->accel.multifunc[4] = val & 0x7ff;
+                dev->accel.clip_right = dev->accel.multifunc[4];
+                if (val & 0x800)
+                    dev->accel.clip_right |= ~0x7ff;
+            }
             break;
 
         case 0xe6ee:
-            mach_log("E6EE (extclipb) write val = %d\n", val);
-            if (len == 2)
+            mach_log("E6EE (extclipb) write val = %d\n", val & 0x7ff);
+            if (len == 2) {
                 dev->accel.multifunc[3] = val & 0x7ff;
+                dev->accel.clip_bottom = dev->accel.multifunc[3];
+                if (val & 0x800)
+                    dev->accel.clip_bottom |= ~0x7ff;
+            }
             break;
 
         case 0xeeee:
@@ -3186,7 +3191,6 @@ mach_accel_out_call(uint16_t port, uint8_t val, mach_t *mach, svga_t *svga, ibm8
         case 0xee8:
         case 0x1ee8:
         case 0x1ee9:
-        case 0x22e8:
         case 0x42e8:
         case 0x42e9:
             ibm8514_accel_out(port, val, svga, 2);
@@ -3197,10 +3201,11 @@ mach_accel_out_call(uint16_t port, uint8_t val, mach_t *mach, svga_t *svga, ibm8
             if (!(mach->shadow_cntl & 0x08)) {
                 if ((mach->shadow_set & 0x03) || (mach->accel.clock_sel & 0x01)) {
                     dev->hdisped = val;
-                    dev->hdisp = (dev->hdisped + 1) << 3;
-                }
+                    dev->hdisp = (val + 1) << 3;
+                } else if (((mach->shadow_set & 0x03) == 0x00) && !(mach->accel.clock_sel & 0x01))
+                    dev->hdisp2 = (val + 1) << 3;
             }
-            mach_log("[%04X:%08X]: ATI 8514/A: (0x%04x): hdisp=0x%02x.\n", CS, cpu_state.pc, port, val);
+            mach_log("[%04X:%08X]: ATI 8514/A: (0x%04x): hdisp=0x%02x, shadowcntl=%02x, shadowset=%02x.\n", CS, cpu_state.pc, port, val, mach->shadow_cntl & 0x08, mach->shadow_set & 0x03);
             svga_recalctimings(svga);
             break;
 
@@ -3208,7 +3213,8 @@ mach_accel_out_call(uint16_t port, uint8_t val, mach_t *mach, svga_t *svga, ibm8
         case 0x12e9:
             /*In preparation to switch from VGA to 8514/A mode*/
             if (!(mach->shadow_cntl & 0x10)) {
-                if ((mach->shadow_set & 0x03) || (mach->accel.clock_sel & 0x01)) {
+                if ((mach->shadow_set & 0x03) || (mach->accel.clock_sel & 0x01) ||
+                    (((mach->shadow_set & 0x03) == 0x00) && !(mach->accel.clock_sel & 0x01))) {
                     WRITE8(port, dev->v_total_reg, val);
                     dev->v_total_reg &= 0x1fff;
                     dev->v_total = dev->v_total_reg + 1;
@@ -3228,9 +3234,17 @@ mach_accel_out_call(uint16_t port, uint8_t val, mach_t *mach, svga_t *svga, ibm8
                     WRITE8(port, dev->v_disp, val);
                     dev->v_disp &= 0x1fff;
                     dev->vdisp = (dev->v_disp + 1) >> 1;
+                    if ((dev->vdisp == 478) || (dev->vdisp == 598) || (dev->vdisp == 766) || (dev->vdisp == 1022))
+                        dev->vdisp += 2;
+                } else if (((mach->shadow_set & 0x03) == 0x00) && !(mach->accel.clock_sel & 0x01)) {
+                    WRITE8(port, dev->v_disp2, val);
+                    dev->v_disp2 &= 0x1fff;
+                    dev->vdisp2 = (dev->v_disp2 + 1) >> 1;
+                    if ((dev->vdisp2 == 478) || (dev->vdisp2 == 598) || (dev->vdisp2 == 766) || (dev->vdisp2 == 1022))
+                        dev->vdisp2 += 2;
                 }
             }
-            mach_log("ATI 8514/A: V_DISP write 16E8 = %d\n", dev->v_disp);
+            mach_log("ATI 8514/A: V_DISP write 16E8=%d, vdisp2=%d.\n", dev->v_disp, dev->v_disp2);
             mach_log("ATI 8514/A: (0x%04x): vdisp=0x%02x.\n", port, val);
             svga_recalctimings(svga);
             break;
@@ -3239,7 +3253,8 @@ mach_accel_out_call(uint16_t port, uint8_t val, mach_t *mach, svga_t *svga, ibm8
         case 0x1ae9:
             /*In preparation to switch from VGA to 8514/A mode*/
             if (!(mach->shadow_cntl & 0x10)) {
-                if ((mach->shadow_set & 0x03) || (mach->accel.clock_sel & 0x01)) {
+                if ((mach->shadow_set & 0x03) || (mach->accel.clock_sel & 0x01) ||
+                    (((mach->shadow_set & 0x03) == 0x00) && !(mach->accel.clock_sel & 0x01))) {
                     WRITE8(port, dev->v_sync_start, val);
                     dev->v_sync_start &= 0x1fff;
                     dev->v_syncstart = dev->v_sync_start + 1;
@@ -3252,15 +3267,29 @@ mach_accel_out_call(uint16_t port, uint8_t val, mach_t *mach, svga_t *svga, ibm8
             svga_recalctimings(svga);
             break;
 
+        case 0x22e8:
+            if (!(mach->shadow_cntl & 0x03)) {
+                if ((mach->shadow_set & 0x03) || (mach->accel.clock_sel & 0x01) ||
+                    (((mach->shadow_set & 0x03) == 0x00) && !(mach->accel.clock_sel & 0x01))) {
+                    dev->disp_cntl = val;
+                    dev->interlace = !!(dev->disp_cntl & 0x10);
+                }
+            }
+            mach_log("ATI 8514/A: DISP_CNTL write %04x=%02x, interlace=%d.\n", port, dev->disp_cntl, dev->interlace);
+            svga_recalctimings(svga);
+            break;
+
         case 0x4ae8:
-            WRITE8(port, dev->accel.advfunc_cntl, val);
-            dev->on = dev->accel.advfunc_cntl & 0x01;
-            vga_on = !dev->on;
+            dev->accel.advfunc_cntl = val;
+            mach_log("[%04X:%08X]: ATI 8514/A: (0x%04x): ON=%d, shadow crt=%x, hdisp=%d, vdisp=%d.\n", CS, cpu_state.pc, port, dev->accel.advfunc_cntl & 0x01, dev->accel.advfunc_cntl & 0x04, dev->hdisp, dev->vdisp);
 
             if ((dev->local & 0xff) < 0x02)
                 dev->ext_crt_pitch = 128;
-
-            mach_log("[%04X:%08X]: ATI 8514/A: (0x%04x): ON=%d, shadow crt=%x, hdisp=%d, vdisp=%d.\n", CS, cpu_state.pc, port, val & 0x01, dev->accel.advfunc_cntl & 0x04, dev->hdisp, dev->vdisp);
+            break;
+        case 0x4ae9:
+            dev->on = dev->accel.advfunc_cntl & 0x01;
+            vga_on = !dev->on;
+            mach_log("[%04X:%08X]: ATI 8514/A: (0x%04x): ON=%d, shadow crt=%x, hdisp=%d, vdisp=%d.\n", CS, cpu_state.pc, port, dev->accel.advfunc_cntl & 0x01, dev->accel.advfunc_cntl & 0x04, dev->hdisp, dev->vdisp);
             mach_log("Vendor IBM mode set %s resolution.\n", (dev->accel.advfunc_cntl & 0x04) ? "2: 1024x768" : "1: 640x480");
 
             svga_recalctimings(svga);
@@ -3342,8 +3371,6 @@ mach_accel_out_call(uint16_t port, uint8_t val, mach_t *mach, svga_t *svga, ibm8
                 else
                     dev->ext_crt_pitch <<= 1;
             }
-            dev->on = 1;
-            vga_on = 0;
             mach_log("ATI 8514/A: (0x%04x) val=0x%02x.\n", port, val);
             svga_recalctimings(svga);
             break;
@@ -3391,7 +3418,9 @@ mach_accel_out_call(uint16_t port, uint8_t val, mach_t *mach, svga_t *svga, ibm8
         case 0x4aee:
         case 0x4aef:
             WRITE8(port, mach->accel.clock_sel, val);
-            dev->on = mach->accel.clock_sel & 0x01;
+            if (port & 1)
+                dev->on = mach->accel.clock_sel & 0x01;
+
             mach_log("ATI 8514/A: (0x%04x): ON=%d, val=%04x, hdisp=%d, vdisp=%d, val=0x%02x.\n", port, dev->on, val, dev->hdisp, dev->vdisp, val & 0xfe);
             if (!(port & 1))
                 mach_log("Vendor ATI mode set %s resolution.\n", (dev->accel.advfunc_cntl & 0x04) ? "2: 1024x768" : "1: 640x480");
@@ -3449,6 +3478,7 @@ mach_accel_out_call(uint16_t port, uint8_t val, mach_t *mach, svga_t *svga, ibm8
         case 0x6aef:
             WRITE8(port, mach->accel.max_waitstates, val);
             mach->override_resolution = !!(mach->accel.max_waitstates & 0x400);
+            mach_log("Override=%d.\n", mach->override_resolution);
             if (mach->override_resolution) {
                 dev->on = 1;
                 vga_on = !dev->on;
@@ -3845,12 +3875,16 @@ mach_accel_in_fifo(mach_t *mach, svga_t *svga, ibm8514_t *dev, uint16_t port, in
             if (len == 2) {
                 if ((dev->local & 0xff) >= 0x02)
                     temp = mach->accel.src_x;
-            } else
-                temp = mach->accel.src_x & 0xff;
+            } else {
+                if ((dev->local & 0xff) >= 0x02)
+                    temp = mach->accel.src_x & 0xff;
+            }
             break;
         case 0xdaef:
-            if (len == 1)
-                temp = mach->accel.src_x >> 8;
+            if (len == 1) {
+                if ((dev->local & 0xff) >= 0x02)
+                    temp = mach->accel.src_x >> 8;
+            }
             break;
 
         case 0xdeee:
@@ -4340,60 +4374,97 @@ mach32_write_common(uint32_t addr, uint8_t val, int linear, mach_t *mach, svga_t
     }
     addr &= svga->decode_mask;
 
-    if (addr >= dev->vram_size)
+    if (addr >= dev->vram_size) {
+        mach_log("WriteOver! %x.\n", addr);
         return;
+    }
 
     addr &= dev->vram_mask;
 
     dev->changedvram[addr >> 12] = svga->monitor->mon_changeframecount;
 
-    val = ((val >> (svga->gdcreg[3] & 7)) | (val << (8 - (svga->gdcreg[3] & 7))));
-    if ((svga->gdcreg[8] == 0xff) && !(svga->gdcreg[3] & 0x18) && (!svga->gdcreg[1] || svga->set_reset_disabled)) {
-        for (i = 0; i < 4; i++) {
-            if (writemask2 & (1 << i))
-                dev->vram[addr | i] = val;
-        }
-    } else {
-        for (i = 0; i < 4; i++) {
-            if (svga->gdcreg[1] & (1 << i))
-                vall.b[i] = !!(svga->gdcreg[0] & (1 << i)) * 0xff;
-            else
-                vall.b[i] = val;
-        }
+    switch (svga->writemode) {
+        case 0:
+            val = ((val >> (svga->gdcreg[3] & 7)) | (val << (8 - (svga->gdcreg[3] & 7))));
+            if ((svga->gdcreg[8] == 0xff) && !(svga->gdcreg[3] & 0x18) && (!svga->gdcreg[1] || svga->set_reset_disabled)) {
+                for (i = 0; i < 4; i++) {
+                    if (writemask2 & (1 << i))
+                        dev->vram[addr | i] = val;
+                }
+                return;
+            } else {
+                for (i = 0; i < 4; i++) {
+                    if (svga->gdcreg[1] & (1 << i))
+                        vall.b[i] = !!(svga->gdcreg[0] & (1 << i)) * 0xff;
+                    else
+                        vall.b[i] = val;
+                }
+            }
+            break;
+        case 1:
+            for (i = 0; i < 4; i++) {
+                if (writemask2 & (1 << i))
+                    dev->vram[addr | i] = dev->latch.b[i];
+            }
+            return;
+        case 2:
+            for (i = 0; i < 4; i++)
+                vall.b[i] = !!(val & (1 << i)) * 0xff;
 
-        switch (svga->gdcreg[3] & 0x18) {
-            case 0x00: /* Set */
+            if (!(svga->gdcreg[3] & 0x18) && (!svga->gdcreg[1] || svga->set_reset_disabled)) {
                 for (i = 0; i < 4; i++) {
                     if (writemask2 & (1 << i))
                         dev->vram[addr | i] = (vall.b[i] & svga->gdcreg[8]) | (dev->latch.b[i] & ~svga->gdcreg[8]);
                 }
-                break;
-            case 0x08: /* AND */
-                for (i = 0; i < 4; i++) {
-                    if (writemask2 & (1 << i))
-                        dev->vram[addr | i] = (vall.b[i] | ~svga->gdcreg[8]) & dev->latch.b[i];
-                }
-                break;
-            case 0x10: /* OR */
-                for (i = 0; i < 4; i++) {
-                    if (writemask2 & (1 << i))
-                        dev->vram[addr | i] = (vall.b[i] & svga->gdcreg[8]) | dev->latch.b[i];
-                }
-                break;
-            case 0x18: /* XOR */
-                for (i = 0; i < 4; i++) {
-                    if (writemask2 & (1 << i))
-                        dev->vram[addr | i] = (vall.b[i] & svga->gdcreg[8]) ^ dev->latch.b[i];
-                }
-                break;
+                return;
+            }
+            break;
+        case 3:
+            val = ((val >> (svga->gdcreg[3] & 7)) | (val << (8 - (svga->gdcreg[3] & 7))));
+            wm  = svga->gdcreg[8];
+            svga->gdcreg[8] &= val;
 
-            default:
-                break;
-        }
+            for (i = 0; i < 4; i++)
+                vall.b[i] = !!(svga->gdcreg[0] & (1 << i)) * 0xff;
 
-        if (reset_wm)
-            svga->gdcreg[8] = wm;
+            reset_wm = 1;
+            break;
+        default:
+            break;
     }
+
+    switch (svga->gdcreg[3] & 0x18) {
+        case 0x00: /* Set */
+            for (i = 0; i < 4; i++) {
+                if (writemask2 & (1 << i))
+                    dev->vram[addr | i] = (vall.b[i] & svga->gdcreg[8]) | (dev->latch.b[i] & ~svga->gdcreg[8]);
+            }
+            break;
+        case 0x08: /* AND */
+            for (i = 0; i < 4; i++) {
+                if (writemask2 & (1 << i))
+                    dev->vram[addr | i] = (vall.b[i] | ~svga->gdcreg[8]) & dev->latch.b[i];
+            }
+            break;
+        case 0x10: /* OR */
+            for (i = 0; i < 4; i++) {
+                if (writemask2 & (1 << i))
+                    dev->vram[addr | i] = (vall.b[i] & svga->gdcreg[8]) | dev->latch.b[i];
+            }
+            break;
+        case 0x18: /* XOR */
+            for (i = 0; i < 4; i++) {
+                if (writemask2 & (1 << i))
+                    dev->vram[addr | i] = (vall.b[i] & svga->gdcreg[8]) ^ dev->latch.b[i];
+            }
+            break;
+
+        default:
+            break;
+    }
+
+    if (reset_wm)
+        svga->gdcreg[8] = wm;
 }
 
 #ifdef ATI_8514_ULTRA
@@ -4618,8 +4689,10 @@ mach32_read_common(uint32_t addr, int linear, mach_t *mach, svga_t *svga)
 
     if (svga->chain4) {
         addr &= svga->decode_mask;
-        if (addr >= dev->vram_size)
+        if (addr >= dev->vram_size) {
+            mach_log("ReadOver! (chain4) %x.\n", addr);
             return 0xff;
+        }
         latch_addr = (addr & dev->vram_mask) & ~3;
         for (uint8_t i = 0; i < count; i++)
             dev->latch.b[i] = dev->vram[latch_addr | i];
@@ -4630,8 +4703,10 @@ mach32_read_common(uint32_t addr, int linear, mach_t *mach, svga_t *svga)
         addr &= dev->vram_mask;
     } else {
         addr &= svga->decode_mask;
-        if (addr >= dev->vram_size)
+        if (addr >= dev->vram_size) {
+            mach_log("ReadOver! (normal) %x.\n", addr);
             return 0xff;
+        }
         latch_addr = (addr & dev->vram_mask) & ~3;
         for (uint8_t i = 0; i < count; i++)
             dev->latch.b[i] = dev->vram[latch_addr | i];
@@ -4642,6 +4717,7 @@ mach32_read_common(uint32_t addr, int linear, mach_t *mach, svga_t *svga)
 
     /* standard VGA latched access */
     if (latch_addr >= dev->vram_size) {
+        mach_log("Over VRAM Latch addr=%x.\n", latch_addr);
         for (uint8_t i = 0; i < count; i++)
             dev->latch.b[i] = 0xff;
     } else {
@@ -4651,8 +4727,10 @@ mach32_read_common(uint32_t addr, int linear, mach_t *mach, svga_t *svga)
             dev->latch.b[i] = dev->vram[latch_addr | i];
     }
 
-    if (addr >= dev->vram_size)
+    if (addr >= dev->vram_size) {
+        mach_log("ReadOver! (chain2) %x.\n", addr);
         return 0xff;
+    }
 
     addr &= dev->vram_mask;
 
@@ -4885,10 +4963,10 @@ mach32_ap_writeb(uint32_t addr, uint8_t val, void *priv)
             mach_accel_outb(0x02e8 + (addr & 1) + (port_dword << 8), val, mach);
         }
     } else {
-        if (dev->on) {
-            mach_log("Linear WORDB Write=%08x, val=%02x, ON=%x, dpconfig=%04x, apsize=%08x.\n", addr & dev->vram_mask, val, dev->on, mach->accel.dp_config, mach->ap_size << 20);
+        mach_log("Linear WORDB Write=%08x, val=%02x, ON=%x, dpconfig=%04x, apsize=%08x.\n", addr & dev->vram_mask, val, dev->on, mach->accel.dp_config, mach->ap_size << 20);
+        if (dev->on)
             mach32_write_common(addr, val, 1, mach, svga);
-        } else
+        else
             svga_write_linear(addr, val, svga);
     }
 }
@@ -4911,10 +4989,10 @@ mach32_ap_writew(uint32_t addr, uint16_t val, void *priv)
             mach_accel_outw(0x02e8 + (port_dword << 8), val, mach);
         }
     } else {
-        if (dev->on) {
-            mach_log("Linear WORDW Write=%08x, val=%04x, ON=%x, dpconfig=%04x, apsize=%08x.\n", addr & dev->vram_mask, val, dev->on, mach->accel.dp_config, mach->ap_size << 20);
+        mach_log("Linear WORDW Write=%08x, val=%04x, ON=%x, dpconfig=%04x, apsize=%08x.\n", addr & dev->vram_mask, val, dev->on, mach->accel.dp_config, mach->ap_size << 20);
+        if (dev->on)
             mach32_writew_linear(addr, val, mach);
-        } else
+        else
             svga_writew_linear(addr, val, svga);
     }
 }
@@ -4939,10 +5017,10 @@ mach32_ap_writel(uint32_t addr, uint32_t val, void *priv)
             mach_accel_outw(0x02e8 + (port_dword << 8) + 4, val >> 16, mach);
         }
     } else {
-        if (dev->on) {
-            mach_log("Linear WORDL Write=%08x, val=%08x, ON=%x, dpconfig=%04x, apsize=%08x.\n", addr & dev->vram_mask, val, dev->on, mach->accel.dp_config, mach->ap_size << 20);
+        mach_log("Linear WORDL Write=%08x, val=%08x, ON=%x, dpconfig=%04x, apsize=%08x.\n", addr & dev->vram_mask, val, dev->on, mach->accel.dp_config, mach->ap_size << 20);
+        if (dev->on)
             mach32_writel_linear(addr, val, mach);
-        } else
+        else
             svga_writel_linear(addr, val, svga);
     }
 }
@@ -4990,9 +5068,9 @@ mach32_ap_readw(uint32_t addr, void *priv)
         else
             temp = mach_accel_inw(0x02e8 + (port_dword << 8), mach);
     } else {
-        if (dev->on) {
+        if (dev->on)
             temp = mach32_readw_linear(addr, mach);
-        } else
+        else
             temp = svga_readw_linear(addr, svga);
 
         mach_log("Linear WORDW Read=%08x, ret=%04x.\n", addr, temp);
@@ -5020,9 +5098,9 @@ mach32_ap_readl(uint32_t addr, void *priv)
             temp |= (mach_accel_inw(0x02e8 + (port_dword << 8) + 4, mach) << 8);
         }
     } else {
-        if (dev->on) {
+        if (dev->on)
             temp = mach32_readl_linear(addr, mach);
-        } else
+        else
             temp = svga_readl_linear(addr, svga);
 
         mach_log("Linear WORDL Read=%08x, ret=%08x, ON%d.\n", addr, temp, dev->on);
diff --git a/src/video/vid_xga.c b/src/video/vid_xga.c
index e71ed42a7..6687238da 100644
--- a/src/video/vid_xga.c
+++ b/src/video/vid_xga.c
@@ -925,8 +925,10 @@ xga_accel_read_pattern_map_pixel(svga_t *svga, int x, int y, uint32_t base, int
     bits = 7 - (x & 7);
 
     if (!(xga->accel.px_map_format[xga->accel.src_map] & 0x08) && !(xga->accel.px_map_format[xga->accel.dst_map] & 0x08)) {
-        if (((xga->accel.px_map_format[xga->accel.src_map] & 0x07) >= 0x02) && ((xga->accel.px_map_format[xga->accel.dst_map] & 0x07) >= 0x02))
+        if (((xga->accel.px_map_format[xga->accel.src_map] & 0x07) >= 0x02) && ((xga->accel.px_map_format[xga->accel.dst_map] & 0x07) >= 0x02) && (xga->accel.pat_src <= 2)) {
+            xga_log("Reverse, access mode=%02x, dstmap=%x, srcmap=%x, pat=%x.\n", xga->access_mode & 0x08, xga->accel.px_map_format[xga->accel.dst_map], xga->accel.px_map_format[xga->accel.src_map], xga->accel.pat_src);
             bits ^= 7;
+        }
     }
 
     px = (byte >> bits) & 1;

From 9685cdfa925e0d9bd03ed87da128b24f439cc643 Mon Sep 17 00:00:00 2001
From: TC1995 <mariogplayer90@gmail.com>
Date: Mon, 28 Oct 2024 02:23:43 +0100
Subject: [PATCH 10/26] XGA changes of the early day (October 28th, 2024)

1. Pattern/font fixes for Windows 3.1's third party XGA drivers as well as OS/2's win3.x' box (seamless and full).
2. The 4MB aperture is still used internally even if the 1MB one is in use, fixes OS/2's XGA driver on boot when less than 16MB of memory is used.
---
 src/video/vid_xga.c | 43 ++++++++++++++++++++++++-------------------
 1 file changed, 24 insertions(+), 19 deletions(-)

diff --git a/src/video/vid_xga.c b/src/video/vid_xga.c
index 6687238da..f126ba9bd 100644
--- a/src/video/vid_xga.c
+++ b/src/video/vid_xga.c
@@ -924,11 +924,10 @@ xga_accel_read_pattern_map_pixel(svga_t *svga, int x, int y, uint32_t base, int
 
     bits = 7 - (x & 7);
 
+    xga_log("0. AccessMode=%02x, SRCMAP=%02x, DSTMAP=%02x, PAT=%02x.\n", xga->access_mode & 0x0f, (xga->accel.px_map_format[xga->accel.src_map] & 0x0f), (xga->accel.px_map_format[xga->accel.dst_map] & 0x0f), (xga->accel.px_map_format[xga->accel.pat_src] & 0x08));
     if (!(xga->accel.px_map_format[xga->accel.src_map] & 0x08) && !(xga->accel.px_map_format[xga->accel.dst_map] & 0x08)) {
-        if (((xga->accel.px_map_format[xga->accel.src_map] & 0x07) >= 0x02) && ((xga->accel.px_map_format[xga->accel.dst_map] & 0x07) >= 0x02) && (xga->accel.pat_src <= 2)) {
-            xga_log("Reverse, access mode=%02x, dstmap=%x, srcmap=%x, pat=%x.\n", xga->access_mode & 0x08, xga->accel.px_map_format[xga->accel.dst_map], xga->accel.px_map_format[xga->accel.src_map], xga->accel.pat_src);
+        if (((xga->accel.px_map_format[xga->accel.src_map] & 0x07) >= 0x02) && ((xga->accel.px_map_format[xga->accel.dst_map] & 0x07) >= 0x02) && (xga->accel.pat_src <= 2))
             bits ^= 7;
-        }
     }
 
     px = (byte >> bits) & 1;
@@ -957,7 +956,8 @@ xga_accel_read_map_pixel(svga_t *svga, int x, int y, int map, uint32_t base, int
             } else
                 byte = mem_readb_phys(addr);
 
-            if (xga->accel.px_map_format[map] & 0x08)
+            xga_log("1. AccessMode=%02x, SRCMAP=%02x, DSTMAP=%02x, PAT=%02x.\n", xga->access_mode & 0x0f, (xga->accel.px_map_format[xga->accel.src_map] & 0x0f), (xga->accel.px_map_format[xga->accel.dst_map] & 0x0f), xga->accel.pat_src);
+            if ((xga->accel.px_map_format[xga->accel.src_map] & 0x08) && !(xga->access_mode & 0x08))
                 bits = (x & 7);
             else
                 bits = 7 - (x & 7);
@@ -1028,10 +1028,15 @@ xga_accel_write_map_pixel(svga_t *svga, int x, int y, int map, uint32_t base, ui
             } else
                 byte = mem_readb_phys(addr);
 
-            if ((xga->accel.px_map_format[map] & 0x08) || (xga->accel.px_map_format[xga->accel.src_map] & 0x08))
-                mask = 1 << (x & 7);
-            else
+            if (xga->access_mode & 0x08)
                 mask = 1 << (7 - (x & 7));
+            else {
+                if ((xga->accel.px_map_format[map] & 0x08) || (xga->accel.px_map_format[xga->accel.src_map] & 0x08)) {
+                    xga_log("2. AccessMode=%02x, SRCMAP=%02x, DSTMAP=%02x, PAT=%02x.\n", xga->access_mode & 0x0f, (xga->accel.px_map_format[xga->accel.src_map] & 0x0f), (xga->accel.px_map_format[map] & 0x0f), xga->accel.pat_src);
+                    mask = 1 << (x & 7);
+                } else
+                    mask = 1 << (7 - (x & 7));
+            }
 
             byte = (byte & ~mask) | ((pixel ? 0xff : 0) & mask);
             if (pixel & 1) {
@@ -1115,11 +1120,11 @@ xga_short_stroke(svga_t *svga, uint8_t ssv)
     int      dirx = 0;
     int      diry = 0;
 
-    dx = xga->accel.dst_map_x & 0x1fff;
+    dx = xga->accel.dst_map_x;
     if (xga->accel.dst_map_x >= 0x1800)
         dx |= ~0x17ff;
 
-    dy = xga->accel.dst_map_y & 0x1fff;
+    dy = xga->accel.dst_map_y;
     if (xga->accel.dst_map_y >= 0x1800)
         dy |= ~0x17ff;
 
@@ -1243,11 +1248,11 @@ xga_line_draw_write(svga_t *svga)
     cx = xga->accel.src_map_x & 0xfff;
     cy = xga->accel.src_map_y & 0xfff;
 
-    dx = xga->accel.dst_map_x & 0x1fff;
+    dx = xga->accel.dst_map_x;
     if (xga->accel.dst_map_x >= 0x1800)
         dx |= ~0x17ff;
 
-    dy = xga->accel.dst_map_y & 0x1fff;
+    dy = xga->accel.dst_map_y;
     if (xga->accel.dst_map_y >= 0x1800)
         dy |= ~0x17ff;
 
@@ -1443,12 +1448,12 @@ xga_bitblt(svga_t *svga)
     xga->accel.sy = xga->accel.src_map_y & 0xfff;
     xga->accel.px = xga->accel.pat_map_x & 0xfff;
     xga->accel.py = xga->accel.pat_map_y & 0xfff;
-    dx = xga->accel.dst_map_x & 0x1fff;
-    dy = xga->accel.dst_map_y & 0x1fff;
+    dx = xga->accel.dst_map_x;
+    dy = xga->accel.dst_map_y;
     if (xga->accel.dst_map_x >= 0x1800)
         dx |= ~0x17ff;
     if (xga->accel.dst_map_y >= 0x1800)
-        dy -= 0x1800;
+        dy |= ~0x17ff;
 
     xga_log("D(%d,%d), SWH(%d,%d), BLT(%d,%d), dstwidth=%d.\n", dx, dy, xga->accel.x, xga->accel.y, srcwidth, srcheight, dstwidth);
 
@@ -1524,7 +1529,7 @@ xga_bitblt(svga_t *svga)
             if (xga->accel.x < 0) {
                 xga->accel.x = xga->accel.blt_width & 0xfff;
 
-                dx = xga->accel.dst_map_x & 0x1fff;
+                dx = xga->accel.dst_map_x;
                 if (xga->accel.dst_map_x >= 0x1800)
                     dx |= ~0x17ff;
                 xga->accel.sx = xga->accel.src_map_x & 0xfff;
@@ -1634,7 +1639,7 @@ xga_bitblt(svga_t *svga)
                     xga->accel.y--;
                     xga->accel.x = xga->accel.blt_width & 0xfff;
 
-                    dx = xga->accel.dst_map_x & 0x1fff;
+                    dx = xga->accel.dst_map_x;
                     if (xga->accel.dst_map_x >= 0x1800)
                         dx |= ~0x17ff;
 
@@ -1699,7 +1704,7 @@ xga_bitblt(svga_t *svga)
                     xga->accel.y--;
                     xga->accel.x = xga->accel.blt_width & 0xfff;
 
-                    dx = xga->accel.dst_map_x & 0x1fff;
+                    dx = xga->accel.dst_map_x;
                     if (xga->accel.dst_map_x >= 0x1800)
                         dx |= ~0x17ff;
 
@@ -3202,7 +3207,7 @@ xga_mca_write(int port, uint8_t val, void *priv)
 
     /* Save the MCA register value. */
     xga->pos_regs[port & 7] = val;
-    if (!(xga->pos_regs[4] & 1) && (mem_size >= 15360)) /*MCA 4MB addressing on systems with more than 16MB of memory*/
+    if (!(xga->pos_regs[4] & 1)) /*MCA 4MB addressing on systems with more than 16MB of memory*/
         xga->pos_regs[4] |= 1;
 
     if (xga->pos_regs[2] & 1) {
@@ -3396,7 +3401,7 @@ xga_pos_out(uint16_t addr, uint8_t val, void *priv)
                 xga_log("104Write=%02x.\n", val);
                 if ((xga->pos_idx & 3) == 0) {
                     xga->pos_regs[4] = val;
-                    if (!(xga->pos_regs[4] & 0x01) && (mem_size >= 15360)) /*4MB addressing on systems with more than 15MB of memory*/
+                    if (!(xga->pos_regs[4] & 0x01)) /*4MB addressing on systems with more than 15MB of memory*/
                         xga->pos_regs[4] |= 0x01;
                 }
                 break;

From b0fd34749c6322fca3e52c2168e62e915fbc710d Mon Sep 17 00:00:00 2001
From: Nelson Kerber Hennemann Filho
 <87081197+nelsonhef@users.noreply.github.com>
Date: Mon, 28 Oct 2024 08:33:13 -0300
Subject: [PATCH 11/26] Update pt-BR.po

Fix some typos
---
 src/qt/languages/pt-BR.po | 4 ++--
 1 file changed, 2 insertions(+), 2 deletions(-)

diff --git a/src/qt/languages/pt-BR.po b/src/qt/languages/pt-BR.po
index 0cadad77c..81a9a5a65 100644
--- a/src/qt/languages/pt-BR.po
+++ b/src/qt/languages/pt-BR.po
@@ -79,7 +79,7 @@ msgid "Use target framerate:"
 msgstr "Usar taxa de quadros alvo:"
 
 msgid "VSync"
-msgstr "VSync (sincronização virtual)"
+msgstr "VSync (sincronização vertical)"
 
 msgid "Shaders"
 msgstr "Shaders"
@@ -625,7 +625,7 @@ msgid "Cassette"
 msgstr "Cassete"
 
 msgid "Vision Systems LBA Enhancer"
-msgstr "Aprimorador LBA Vision Sistems"
+msgstr "Aprimorador LBA Vision Systems"
 
 msgid "Hard disks:"
 msgstr "Discos rígidos:"

From f0e036811523bb588e96b80152b282d32e53219c Mon Sep 17 00:00:00 2001
From: altiereslima <altieres.lima@gmail.com>
Date: Mon, 28 Oct 2024 15:30:55 -0300
Subject: [PATCH 12/26] Update pt-BR.po

---
 src/qt/languages/pt-BR.po | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/src/qt/languages/pt-BR.po b/src/qt/languages/pt-BR.po
index 81a9a5a65..9e6bb4016 100644
--- a/src/qt/languages/pt-BR.po
+++ b/src/qt/languages/pt-BR.po
@@ -625,7 +625,7 @@ msgid "Cassette"
 msgstr "Cassete"
 
 msgid "Vision Systems LBA Enhancer"
-msgstr "Aprimorador LBA Vision Systems"
+msgstr "Vision Systems LBA Enhancer"
 
 msgid "Hard disks:"
 msgstr "Discos rígidos:"

From 4f472445f286e783b5f42aa9165de232960cda08 Mon Sep 17 00:00:00 2001
From: TC1995 <mariogplayer90@gmail.com>
Date: Mon, 28 Oct 2024 22:14:58 +0100
Subject: [PATCH 13/26] Video and emulator GUI changes of the day (October
 28th, 2024)

IBM 8514/A (MCA/ISA).
Added a configure button to the QT frontend for the amount of memory used by said add-on chips.

XGA:
Forgot a change about the 4mb aperture.
---
 src/include/86box/vid_8514a_device.h | 25 +++++++++++++++++++++++++
 src/qt/qt_settingsdisplay.cpp        | 20 ++++++++++++++++++++
 src/qt/qt_settingsdisplay.hpp        |  2 ++
 src/qt/qt_settingsdisplay.ui         |  7 +++++++
 src/video/vid_8514a.c                |  1 +
 src/video/vid_xga.c                  | 10 ++++------
 6 files changed, 59 insertions(+), 6 deletions(-)
 create mode 100644 src/include/86box/vid_8514a_device.h

diff --git a/src/include/86box/vid_8514a_device.h b/src/include/86box/vid_8514a_device.h
new file mode 100644
index 000000000..b4102b17a
--- /dev/null
+++ b/src/include/86box/vid_8514a_device.h
@@ -0,0 +1,25 @@
+/*
+ * 86Box    A hypervisor and IBM PC system emulator that specializes in
+ *          running old operating systems and software designed for IBM
+ *          PC systems and compatibles from 1981 through fairly recent
+ *          system designs based on the PCI bus.
+ *
+ *          This file is part of the 86Box distribution.
+ *
+ *          Emulation of the 8514/A card from IBM for the MCA bus and
+ *          generic ISA bus clones without vendor extensions.
+ *
+ *
+ *
+ * Authors: TheCollector1995
+ *
+ *          Copyright 2024 TheCollector1995.
+ */
+#ifndef VIDEO_8514A_DEVICE_H
+#define VIDEO_8514A_DEVICE_H
+
+#ifdef EMU_DEVICE_H
+extern const device_t ibm8514_mca_device;
+extern const device_t gen8514_isa_device;
+#endif
+#endif /*VIDEO_XGA_DEVICE_H*/
diff --git a/src/qt/qt_settingsdisplay.cpp b/src/qt/qt_settingsdisplay.cpp
index db8c30e5d..50489ea61 100644
--- a/src/qt/qt_settingsdisplay.cpp
+++ b/src/qt/qt_settingsdisplay.cpp
@@ -24,6 +24,7 @@ extern "C" {
 #include <86box/device.h>
 #include <86box/machine.h>
 #include <86box/video.h>
+#include <86box/vid_8514a_device.h>
 #include <86box/vid_xga_device.h>
 }
 
@@ -53,6 +54,7 @@ SettingsDisplay::save()
     // TODO
     for (uint8_t i = 1; i < GFXCARD_MAX; i ++)
         gfxcard[i]                 = ui->comboBoxVideoSecondary->currentData().toInt();
+
     voodoo_enabled             = ui->checkBoxVoodoo->isChecked() ? 1 : 0;
     ibm8514_standalone_enabled = ui->checkBox8514->isChecked() ? 1 : 0;
     xga_standalone_enabled     = ui->checkBoxXga->isChecked() ? 1 : 0;
@@ -127,6 +129,16 @@ SettingsDisplay::on_pushButtonConfigureVoodoo_clicked()
     DeviceConfig::ConfigureDevice(&voodoo_device, 0, qobject_cast<Settings *>(Settings::settings));
 }
 
+void
+SettingsDisplay::on_pushButtonConfigure8514_clicked()
+{
+    if (machine_has_bus(machineId, MACHINE_BUS_MCA) > 0) {
+        DeviceConfig::ConfigureDevice(&ibm8514_mca_device, 0, qobject_cast<Settings *>(Settings::settings));
+    } else {
+        DeviceConfig::ConfigureDevice(&gen8514_isa_device, 0, qobject_cast<Settings *>(Settings::settings));
+    }
+}
+
 void
 SettingsDisplay::on_pushButtonConfigureXga_clicked()
 {
@@ -166,6 +178,8 @@ SettingsDisplay::on_comboBoxVideo_currentIndexChanged(int index)
     ui->checkBox8514->setEnabled(machineSupports8514);
     ui->checkBox8514->setChecked(ibm8514_standalone_enabled && machineSupports8514);
 
+    ui->pushButtonConfigure8514->setEnabled(ui->checkBox8514->isEnabled() && ui->checkBox8514->isChecked());
+
     ui->checkBoxXga->setEnabled(machineSupportsXga);
     ui->checkBoxXga->setChecked(xga_standalone_enabled && machineSupportsXga);
 
@@ -238,6 +252,12 @@ SettingsDisplay::on_checkBoxVoodoo_stateChanged(int state)
     ui->pushButtonConfigureVoodoo->setEnabled(state == Qt::Checked);
 }
 
+void
+SettingsDisplay::on_checkBox8514_stateChanged(int state)
+{
+    ui->pushButtonConfigure8514->setEnabled(state == Qt::Checked);
+}
+
 void
 SettingsDisplay::on_checkBoxXga_stateChanged(int state)
 {
diff --git a/src/qt/qt_settingsdisplay.hpp b/src/qt/qt_settingsdisplay.hpp
index 6297f7eae..3c28c52e0 100644
--- a/src/qt/qt_settingsdisplay.hpp
+++ b/src/qt/qt_settingsdisplay.hpp
@@ -29,9 +29,11 @@ private slots:
 
 private slots:
     void on_checkBoxVoodoo_stateChanged(int state);
+    void on_checkBox8514_stateChanged(int state);
     void on_checkBoxXga_stateChanged(int state);
     void on_comboBoxVideo_currentIndexChanged(int index);
     void on_pushButtonConfigureVoodoo_clicked();
+    void on_pushButtonConfigure8514_clicked();
     void on_pushButtonConfigureXga_clicked();
     void on_pushButtonConfigure_clicked();
 
diff --git a/src/qt/qt_settingsdisplay.ui b/src/qt/qt_settingsdisplay.ui
index a9b7e6e2c..f6bbeeae0 100644
--- a/src/qt/qt_settingsdisplay.ui
+++ b/src/qt/qt_settingsdisplay.ui
@@ -106,6 +106,13 @@
      </property>
     </widget>
    </item>
+   <item row="5" column="2">
+    <widget class="QPushButton" name="pushButtonConfigure8514">
+     <property name="text">
+      <string>Configure</string>
+     </property>
+    </widget>
+   </item>
    <item row="6" column="2">
     <widget class="QPushButton" name="pushButtonConfigureXga">
      <property name="text">
diff --git a/src/video/vid_8514a.c b/src/video/vid_8514a.c
index 6d7aa6b94..accedd131 100644
--- a/src/video/vid_8514a.c
+++ b/src/video/vid_8514a.c
@@ -35,6 +35,7 @@
 #include <86box/thread.h>
 #include <86box/video.h>
 #include <86box/vid_8514a.h>
+#include <86box/vid_8514a_device.h>
 #include <86box/vid_xga.h>
 #include <86box/vid_svga.h>
 #include <86box/vid_svga_render.h>
diff --git a/src/video/vid_xga.c b/src/video/vid_xga.c
index f126ba9bd..bd18beae2 100644
--- a/src/video/vid_xga.c
+++ b/src/video/vid_xga.c
@@ -3487,15 +3487,13 @@ xga_init(const device_t *info)
             xga->rom_addr    = 0xc0000 + (((xga->pos_regs[2] & 0xf0) >> 4) * 0x2000);
             xga->instance    = (xga->pos_regs[2] & 0x0e) >> 1;
             xga->pos_regs[2] |= 0x01;
-            if (mem_size >= 15360) {
-                xga->pos_regs[4] |= 0x01;
+            xga->pos_regs[4] |= 0x01;
+            if (mem_size >= 15360)
                 xga->pos_regs[5] = 0;
-            } else {
+            else {
                 xga->pos_regs[5] = ((mem_size * 64) >> 0x10) + 1;
-                if (xga->pos_regs[5] == 0x10) {
+                if (xga->pos_regs[5] == 0x10)
                     xga->pos_regs[5] = 0x00;
-                    xga->pos_regs[4] |= 0x01;
-                }
             }
             xga->base_addr_1mb = (xga->pos_regs[5] & 0x0f) << 20;
             xga->linear_base = ((xga->pos_regs[4] & 0xfe) * 0x1000000) + (xga->instance << 22);

From 451ece0074272093800d92c48addf8545a704d67 Mon Sep 17 00:00:00 2001
From: RichardG867 <richardg867@gmail.com>
Date: Mon, 28 Oct 2024 19:57:26 -0300
Subject: [PATCH 14/26] scsi_cdrom: Fix reporting of read errors on ATAPI
 drives

---
 src/scsi/scsi_cdrom.c | 4 ++--
 1 file changed, 2 insertions(+), 2 deletions(-)

diff --git a/src/scsi/scsi_cdrom.c b/src/scsi/scsi_cdrom.c
index 358901847..103ce3161 100644
--- a/src/scsi/scsi_cdrom.c
+++ b/src/scsi/scsi_cdrom.c
@@ -1416,7 +1416,7 @@ scsi_cdrom_read_blocks(scsi_cdrom_t *dev, int32_t *len, int first_batch, int ven
     scsi_cdrom_log("Read %i bytes of blocks...\n", *len);
 
     if (ret == -1)
-        return 0;
+        return ret;
     else if (!ret || (!first_batch && (dev->old_len != *len))) {
         if (!first_batch && (dev->old_len != *len))
             scsi_cdrom_illegal_mode(dev);
@@ -2219,7 +2219,7 @@ begin:
 
             if (ret <= 0) {
                 scsi_cdrom_set_phase(dev, SCSI_PHASE_STATUS);
-                dev->packet_status = PHASE_COMPLETE;
+                dev->packet_status = (ret < 0) ? PHASE_ERROR : PHASE_COMPLETE;
                 dev->callback      = 20.0 * CDROM_TIME;
                 scsi_cdrom_set_callback(dev);
                 scsi_cdrom_buf_free(dev);

From 36e42af7b25151b188f9e47a5657d35c740f0bd2 Mon Sep 17 00:00:00 2001
From: TC1995 <mariogplayer90@gmail.com>
Date: Tue, 29 Oct 2024 01:16:21 +0100
Subject: [PATCH 15/26] S3 bank update (October 29th, 2024)

1. Simplified 4bpp rendering mode.
2. Implement CRTC31 bit 0 functionality which enables/disables the vendor banks. Fixes Diamond Stealth 64 964 VRAM (VLB/PCI) detection on NT 3.1 (and possibly other cards) while keeping NT 3.1 347.1's full screen Command Prompt test normal without garbage using the 911/924 chips.
---
 src/video/vid_s3.c | 46 ++++++++++++++++------------------------------
 1 file changed, 16 insertions(+), 30 deletions(-)

diff --git a/src/video/vid_s3.c b/src/video/vid_s3.c
index c91eab589..295c65bd4 100644
--- a/src/video/vid_s3.c
+++ b/src/video/vid_s3.c
@@ -3218,7 +3218,7 @@ s3_decode_addr(svga_t *svga, uint32_t addr, int write)
 {
     int memory_map_mode = (svga->gdcreg[6] >> 2) & 3;
 
-    s3_log("CRTC31 bit 3=%x, map=%x, write=%x, wrtbank=%x, chain4=%x.\n", svga->crtc[0x31] & 0x08, memory_map_mode, write, svga->write_bank, svga->chain4);
+    s3_log("CRTC31 bit 3=%x, map=%x, write=%x, wrtbank=%x, chain4=%x, vrammask=%08x.\n", svga->crtc[0x31] & 0x08, memory_map_mode, write, svga->write_bank, svga->chain4, svga->vram_display_mask);
 
     if (svga->crtc[0x31] & 0x08)
         memory_map_mode = 1;
@@ -4295,36 +4295,20 @@ s3_recalctimings(svga_t *svga)
         }
         svga->vram_display_mask = s3->vram_mask;
     } else {
-        if (!svga->scrblank && (svga->crtc[0x17] & 0x80) && svga->attr_palette_enable) {
-            if ((svga->gdcreg[6] & 1) || (svga->attrregs[0x10] & 1)) {
-                if (svga->crtc[0x31] & 0x08) {
-                    if (svga->bpp == 8) {
-                        if (!(svga->crtc[0x5e] & 0x04))
-                            svga->vblankstart = svga->dispend; /*Applies only to Enhanced modes*/
+        if (svga->crtc[0x31] & 0x08) {
+            if (!(svga->crtc[0x5e] & 0x04))
+                svga->vblankstart = svga->dispend; /*Applies only to Enhanced modes*/
 
-                        /*Enhanced 4bpp mode, just like the 8bpp mode per the spec. */
-                        svga->render = svga_render_8bpp_highres;
-                        svga->rowoffset <<= 1;
-                        svga->vram_display_mask = s3->vram_mask;
-                    } else {
-                        svga->vram_display_mask = (svga->crtc[0x32] & 0x40) ? 0x3ffff : s3->vram_mask;
-                        svga->write_bank = 0;
-                        svga->read_bank = svga->write_bank;
-                    }
-                } else {
-                    svga->vram_display_mask = (svga->crtc[0x32] & 0x40) ? 0x3ffff : s3->vram_mask;
-                    svga->write_bank = 0;
-                    svga->read_bank = svga->write_bank;
-                }
-            } else {
-                svga->vram_display_mask = (svga->crtc[0x32] & 0x40) ? 0x3ffff : s3->vram_mask;
-                svga->write_bank = 0;
-                svga->read_bank = svga->write_bank;
-            }
+            /*Enhanced 4bpp mode, just like the 8bpp mode per the spec. */
+            svga->render = svga_render_8bpp_highres;
+            svga->rowoffset <<= 1;
+            svga->vram_display_mask = s3->vram_mask;
         } else {
             svga->vram_display_mask = (svga->crtc[0x32] & 0x40) ? 0x3ffff : s3->vram_mask;
-            svga->write_bank = 0;
-            svga->read_bank = svga->write_bank;
+            if (!(svga->crtc[0x31] & 0x01)) { /*Bank Enable bit*/
+                svga->write_bank = 0;
+                svga->read_bank = 0;
+            }
         }
     }
 
@@ -4466,8 +4450,10 @@ s3_trio64v_recalctimings(svga_t *svga)
             svga->vram_display_mask = s3->vram_mask;
         } else {
             svga->vram_display_mask = (svga->crtc[0x32] & 0x40) ? 0x3ffff : s3->vram_mask;
-            svga->write_bank = 0;
-            svga->read_bank = svga->write_bank;
+            if (!(svga->crtc[0x31] & 0x01)) { /*Bank Enable bit*/
+                svga->write_bank = 0;
+                svga->read_bank = 0;
+            }
         }
     } else /*Streams mode*/
     {

From 416125390bf8af07500dc5a98eb4a11ff6864cab Mon Sep 17 00:00:00 2001
From: RichardG867 <richardg867@gmail.com>
Date: Tue, 29 Oct 2024 11:41:59 -0300
Subject: [PATCH 16/26] hdd_image: Add framework for reporting I/O errors, and
 stop removing drives with locked images

---
 src/disk/hdd_image.c    | 59 ++++++++++++++++++++++++++++-------------
 src/include/86box/hdd.h |  8 +++---
 2 files changed, 44 insertions(+), 23 deletions(-)

diff --git a/src/disk/hdd_image.c b/src/disk/hdd_image.c
index 7c2618969..8b0949c9b 100644
--- a/src/disk/hdd_image.c
+++ b/src/disk/hdd_image.c
@@ -184,6 +184,9 @@ hdd_image_calc_chs(uint32_t *c, uint32_t *h, uint32_t *s, uint32_t size)
 static int
 prepare_new_hard_disk(uint8_t id, uint64_t full_size)
 {
+    if (!hdd_images[id].file)
+        return -1;
+
     uint64_t target_size = (full_size + hdd_images[id].base) - ftello64(hdd_images[id].file);
 
 #ifndef __unix__
@@ -390,7 +393,8 @@ retry_vhd:
         } else {
             /* Failed for another reason */
             hdd_image_log("Failed for another reason\n");
-            return 0;
+            hdd_images[id].type = HDD_IMAGE_RAW;
+            return 1;
         }
     } else {
         if (image_is_hdi(fn)) {
@@ -500,7 +504,7 @@ retry_vhd:
     return ret;
 }
 
-void
+int
 hdd_image_seek(uint8_t id, uint32_t sector)
 {
     off64_t addr = sector;
@@ -508,12 +512,16 @@ hdd_image_seek(uint8_t id, uint32_t sector)
 
     hdd_images[id].pos = sector;
     if (hdd_images[id].type != HDD_IMAGE_VHD) {
-        if (fseeko64(hdd_images[id].file, addr + hdd_images[id].base, SEEK_SET) == -1)
-            fatal("hdd_image_seek(): Error seeking\n");
+        if (!hdd_images[id].file || (fseeko64(hdd_images[id].file, addr + hdd_images[id].base, SEEK_SET) == -1)) {
+            hdd_image_log("hdd_image_seek(): Error seeking\n");
+            return -1;
+        }
     }
+
+    return 0;
 }
 
-void
+int
 hdd_image_read(uint8_t id, uint32_t sector, uint32_t count, uint8_t *buffer)
 {
     int    non_transferred_sectors;
@@ -523,14 +531,18 @@ hdd_image_read(uint8_t id, uint32_t sector, uint32_t count, uint8_t *buffer)
         non_transferred_sectors = mvhd_read_sectors(hdd_images[id].vhd, sector, count, buffer);
         hdd_images[id].pos      = sector + count - non_transferred_sectors - 1;
     } else {
-        if (fseeko64(hdd_images[id].file, ((uint64_t) (sector) << 9LL) + hdd_images[id].base, SEEK_SET) == -1) {
-            fatal("Hard disk image %i: Read error during seek\n", id);
-            return;
+        if (!hdd_images[id].file || (fseeko64(hdd_images[id].file, ((uint64_t) (sector) << 9LL) + hdd_images[id].base, SEEK_SET) == -1)) {
+            hdd_image_log("Hard disk image %i: Read error during seek\n", id);
+            return -1;
         }
 
         num_read           = fread(buffer, 512, count, hdd_images[id].file);
+        if (count && !num_read)
+            return -1;
         hdd_images[id].pos = sector + num_read;
     }
+
+    return 0;
 }
 
 uint32_t
@@ -554,14 +566,15 @@ hdd_image_read_ex(uint8_t id, uint32_t sector, uint32_t count, uint8_t *buffer)
     if ((sectors - sector) < transfer_sectors)
         transfer_sectors = sectors - sector;
 
-    hdd_image_read(id, sector, transfer_sectors, buffer);
+    if (hdd_image_read(id, sector, transfer_sectors, buffer) < 0)
+        return -1;
 
     if (count != transfer_sectors)
         return 1;
     return 0;
 }
 
-void
+int
 hdd_image_write(uint8_t id, uint32_t sector, uint32_t count, uint8_t *buffer)
 {
     int    non_transferred_sectors;
@@ -571,15 +584,19 @@ hdd_image_write(uint8_t id, uint32_t sector, uint32_t count, uint8_t *buffer)
         non_transferred_sectors = mvhd_write_sectors(hdd_images[id].vhd, sector, count, buffer);
         hdd_images[id].pos      = sector + count - non_transferred_sectors - 1;
     } else {
-        if (fseeko64(hdd_images[id].file, ((uint64_t) (sector) << 9LL) + hdd_images[id].base, SEEK_SET) == -1) {
-            fatal("Hard disk image %i: Write error during seek\n", id);
-            return;
+        if (!hdd_images[id].file || (fseeko64(hdd_images[id].file, ((uint64_t) (sector) << 9LL) + hdd_images[id].base, SEEK_SET) == -1)) {
+            hdd_image_log("Hard disk image %i: Write error during seek\n", id);
+            return -1;
         }
 
         num_write          = fwrite(buffer, 512, count, hdd_images[id].file);
+        if (count && !num_write)
+            return -1;
         fflush(hdd_images[id].file);
         hdd_images[id].pos = sector + num_write;
     }
+
+    return 0;
 }
 
 int
@@ -591,14 +608,15 @@ hdd_image_write_ex(uint8_t id, uint32_t sector, uint32_t count, uint8_t *buffer)
     if ((sectors - sector) < transfer_sectors)
         transfer_sectors = sectors - sector;
 
-    hdd_image_write(id, sector, transfer_sectors, buffer);
+    if (hdd_image_write(id, sector, transfer_sectors, buffer) < 0)
+        return -1;
 
     if (count != transfer_sectors)
         return 1;
     return 0;
 }
 
-void
+int
 hdd_image_zero(uint8_t id, uint32_t sector, uint32_t count)
 {
     if (hdd_images[id].type == HDD_IMAGE_VHD) {
@@ -607,9 +625,9 @@ hdd_image_zero(uint8_t id, uint32_t sector, uint32_t count)
     } else {
         memset(empty_sector, 0, 512);
 
-        if (fseeko64(hdd_images[id].file, ((uint64_t) (sector) << 9LL) + hdd_images[id].base, SEEK_SET) == -1) {
-            fatal("Hard disk image %i: Zero error during seek\n", id);
-            return;
+        if (!hdd_images[id].file || (fseeko64(hdd_images[id].file, ((uint64_t) (sector) << 9LL) + hdd_images[id].base, SEEK_SET) == -1)) {
+            hdd_image_log("Hard disk image %i: Zero error during seek\n", id);
+            return -1;
         }
 
         for (uint32_t i = 0; i < count; i++) {
@@ -617,11 +635,14 @@ hdd_image_zero(uint8_t id, uint32_t sector, uint32_t count)
                 break;
 
             hdd_images[id].pos = sector + i;
-            fwrite(empty_sector, 512, 1, hdd_images[id].file);
+            if (!fwrite(empty_sector, 512, 1, hdd_images[id].file))
+                return -1;
         }
 
         fflush(hdd_images[id].file);
     }
+
+    return 0;
 }
 
 int
diff --git a/src/include/86box/hdd.h b/src/include/86box/hdd.h
index 89a6cf1ff..25c32355e 100644
--- a/src/include/86box/hdd.h
+++ b/src/include/86box/hdd.h
@@ -197,12 +197,12 @@ extern int   hdd_is_valid(int c);
 
 extern void     hdd_image_init(void);
 extern int      hdd_image_load(int id);
-extern void     hdd_image_seek(uint8_t id, uint32_t sector);
-extern void     hdd_image_read(uint8_t id, uint32_t sector, uint32_t count, uint8_t *buffer);
+extern int      hdd_image_seek(uint8_t id, uint32_t sector);
+extern int      hdd_image_read(uint8_t id, uint32_t sector, uint32_t count, uint8_t *buffer);
 extern int      hdd_image_read_ex(uint8_t id, uint32_t sector, uint32_t count, uint8_t *buffer);
-extern void     hdd_image_write(uint8_t id, uint32_t sector, uint32_t count, uint8_t *buffer);
+extern int      hdd_image_write(uint8_t id, uint32_t sector, uint32_t count, uint8_t *buffer);
 extern int      hdd_image_write_ex(uint8_t id, uint32_t sector, uint32_t count, uint8_t *buffer);
-extern void     hdd_image_zero(uint8_t id, uint32_t sector, uint32_t count);
+extern int      hdd_image_zero(uint8_t id, uint32_t sector, uint32_t count);
 extern int      hdd_image_zero_ex(uint8_t id, uint32_t sector, uint32_t count);
 extern uint32_t hdd_image_get_last_sector(uint8_t id);
 extern uint32_t hdd_image_get_pos(uint8_t id);

From d5b0ba9511eb0d03429178fc3537bcd113f3f9bb Mon Sep 17 00:00:00 2001
From: RichardG867 <richardg867@gmail.com>
Date: Tue, 29 Oct 2024 11:47:47 -0300
Subject: [PATCH 17/26] hdc_ide: Report image read/write errors

---
 src/disk/hdc_ide.c | 46 ++++++++++++++++++++++++++++++++++------------
 1 file changed, 34 insertions(+), 12 deletions(-)

diff --git a/src/disk/hdc_ide.c b/src/disk/hdc_ide.c
index 1ec054de9..660093a8e 100644
--- a/src/disk/hdc_ide.c
+++ b/src/disk/hdc_ide.c
@@ -2214,8 +2214,10 @@ ide_callback(void *priv)
                 if (ide->do_initial_read) {
                     ide->do_initial_read = 0;
                     ide->sector_pos      = 0;
-                    hdd_image_read(ide->hdd_num, ide_get_sector(ide),
-                                   ide->tf->secount ? ide->tf->secount : 256, ide->sector_buffer);
+                    ret = hdd_image_read(ide->hdd_num, ide_get_sector(ide),
+                                         ide->tf->secount ? ide->tf->secount : 256, ide->sector_buffer);
+                } else {
+                    ret = 0;
                 }
 
                 memcpy(ide->buffer, &ide->sector_buffer[ide->sector_pos * 512], 512);
@@ -2224,6 +2226,10 @@ ide_callback(void *priv)
 
                 ide->tf->pos = 0;
                 ide->tf->atastat = DRQ_STAT | DRDY_STAT | DSC_STAT;
+                if (ret < 0) {
+                    ide_log("IDE %i: Read aborted (image read error)\n", ide->channel);
+                    err = UNC_ERR;
+                }
 
                 ide_irq_raise(ide);
 
@@ -2245,11 +2251,13 @@ ide_callback(void *priv)
                     ide->sector_pos = ide->tf->secount;
                 else
                     ide->sector_pos = 256;
-                hdd_image_read(ide->hdd_num, ide_get_sector(ide), ide->sector_pos, ide->sector_buffer);
 
                 ide->tf->pos = 0;
 
-                if (!ide_boards[ide->board]->force_ata3 && bm->dma) {
+                if (hdd_image_read(ide->hdd_num, ide_get_sector(ide), ide->sector_pos, ide->sector_buffer) < 0) {
+                    ide_log("IDE %i: DMA read aborted (image read error)\n", ide->channel);
+                    err = UNC_ERR;
+                } else if (!ide_boards[ide->board]->force_ata3 && bm->dma) {
                     /* We should not abort - we should simply wait for the host to start DMA. */
                     ret = bm->dma(ide->sector_buffer, ide->sector_pos * 512, 0, bm->priv);
                     if (ret == 2) {
@@ -2292,8 +2300,10 @@ ide_callback(void *priv)
                 if (ide->do_initial_read) {
                     ide->do_initial_read = 0;
                     ide->sector_pos      = 0;
-                    hdd_image_read(ide->hdd_num, ide_get_sector(ide),
-                                   ide->tf->secount ? ide->tf->secount : 256, ide->sector_buffer);
+                    ret = hdd_image_read(ide->hdd_num, ide_get_sector(ide),
+                                         ide->tf->secount ? ide->tf->secount : 256, ide->sector_buffer);
+                } else {
+                    ret = 0;
                 }
 
                 memcpy(ide->buffer, &ide->sector_buffer[ide->sector_pos * 512], 512);
@@ -2302,6 +2312,10 @@ ide_callback(void *priv)
                 ide->tf->pos     = 0;
 
                 ide->tf->atastat = DRQ_STAT | DRDY_STAT | DSC_STAT;
+                if (ret < 0) {
+                    ide_log("IDE %i: Read aborted (image read error)\n", ide->channel);
+                    err = UNC_ERR;
+                }
                 if (!ide->blockcount)
                     ide_irq_raise(ide);
                 ide->blockcount++;
@@ -2320,7 +2334,7 @@ ide_callback(void *priv)
             else if (!ide->tf->lba && (ide->cfg_spt == 0))
                 err = IDNF_ERR;
             else {
-                hdd_image_write(ide->hdd_num, ide_get_sector(ide), 1, (uint8_t *) ide->buffer);
+                ret = hdd_image_write(ide->hdd_num, ide_get_sector(ide), 1, (uint8_t *) ide->buffer);
                 ide_irq_raise(ide);
                 ide->tf->secount--;
                 if (ide->tf->secount) {
@@ -2332,6 +2346,8 @@ ide_callback(void *priv)
                     ide->tf->atastat = DRDY_STAT | DSC_STAT;
                     ui_sb_update_icon(SB_HDD | hdd[ide->hdd_num].bus, 0);
                 }
+                if (ret < 0)
+                    err = UNC_ERR;
             }
             ide_log("Write: %02X, %i, %08X, %" PRIi64 "\n", err, ide->hdd_num, ide->lba_addr, sector);
             break;
@@ -2360,12 +2376,14 @@ ide_callback(void *priv)
                         return;
                     } else if (ret == 1) {
                         /* DMA successful */
-                        ide_log("IDE %i: DMA write successful\n", ide->channel);
+                        ret = hdd_image_write(ide->hdd_num, ide_get_sector(ide),
+                                              ide->sector_pos, ide->sector_buffer);
 
-                        hdd_image_write(ide->hdd_num, ide_get_sector(ide),
-                                        ide->sector_pos, ide->sector_buffer);
+                        ide_log("IDE %i: DMA write %ssuccessful\n", ide->channel, (ret < 0) ? "un" : "");
 
                         ide->tf->atastat = DRDY_STAT | DSC_STAT;
+                        if (ret < 0)
+                            err = UNC_ERR;
 
                         ide_irq_raise(ide);
                         ui_sb_update_icon(SB_HDD | hdd[ide->hdd_num].bus, 0);
@@ -2393,7 +2411,7 @@ ide_callback(void *priv)
             else if (!ide->tf->lba && (ide->cfg_spt == 0))
                 err = IDNF_ERR;
             else {
-                hdd_image_write(ide->hdd_num, ide_get_sector(ide), 1, (uint8_t *) ide->buffer);
+                ret = hdd_image_write(ide->hdd_num, ide_get_sector(ide), 1, (uint8_t *) ide->buffer);
                 ide->blockcount++;
                 if (ide->blockcount >= ide->blocksize || ide->tf->secount == 1) {
                     ide->blockcount = 0;
@@ -2408,6 +2426,8 @@ ide_callback(void *priv)
                     ide->tf->atastat = DRDY_STAT | DSC_STAT;
                     ui_sb_update_icon(SB_HDD | hdd[ide->hdd_num].bus, 0);
                 }
+                if (ret < 0)
+                    err = UNC_ERR;
             }
             break;
 
@@ -2431,9 +2451,11 @@ ide_callback(void *priv)
             else if (!ide->tf->lba && (ide->cfg_spt == 0))
                 err = IDNF_ERR;
             else {
-                hdd_image_zero(ide->hdd_num, ide_get_sector_format(ide), ide->tf->secount);
+                ret = hdd_image_zero(ide->hdd_num, ide_get_sector_format(ide), ide->tf->secount);
 
                 ide->tf->atastat = DRDY_STAT | DSC_STAT;
+                if (ret < 0)
+                    err = UNC_ERR;
                 ide_irq_raise(ide);
 
                 ui_sb_update_icon(SB_HDD | hdd[ide->hdd_num].bus, 1);

From 94e5fded317e8e4a4386627cb7726d0eef698e82 Mon Sep 17 00:00:00 2001
From: RichardG867 <richardg867@gmail.com>
Date: Tue, 29 Oct 2024 12:12:01 -0300
Subject: [PATCH 18/26] hdd_image: Populate last sector for failed images

---
 src/disk/hdd_image.c | 3 ++-
 1 file changed, 2 insertions(+), 1 deletion(-)

diff --git a/src/disk/hdd_image.c b/src/disk/hdd_image.c
index 8b0949c9b..2b463a23c 100644
--- a/src/disk/hdd_image.c
+++ b/src/disk/hdd_image.c
@@ -393,7 +393,8 @@ retry_vhd:
         } else {
             /* Failed for another reason */
             hdd_image_log("Failed for another reason\n");
-            hdd_images[id].type = HDD_IMAGE_RAW;
+            hdd_images[id].type        = HDD_IMAGE_RAW;
+            hdd_images[id].last_sector = (uint32_t) (((uint64_t) hdd[id].spt) * ((uint64_t) hdd[id].hpc) * ((uint64_t) hdd[id].tracks)) - 1;
             return 1;
         }
     } else {

From 7b786a98fa9c0dfc1f573d615898cf8504ebbe3c Mon Sep 17 00:00:00 2001
From: RichardG867 <richardg867@gmail.com>
Date: Tue, 29 Oct 2024 12:13:04 -0300
Subject: [PATCH 19/26] scsi_disk: Report image read/write errors

---
 src/include/86box/scsi_device.h |  1 +
 src/scsi/scsi_disk.c            | 38 +++++++++++++++++++++++++++------
 2 files changed, 33 insertions(+), 6 deletions(-)

diff --git a/src/include/86box/scsi_device.h b/src/include/86box/scsi_device.h
index abcd415f4..e69d3cc93 100644
--- a/src/include/86box/scsi_device.h
+++ b/src/include/86box/scsi_device.h
@@ -197,6 +197,7 @@
 #define ASC_NONE                               0x00
 #define ASC_AUDIO_PLAY_OPERATION               0x00
 #define ASC_NOT_READY                          0x04
+#define ASC_WRITE_ERROR                        0x0c
 #define ASC_UNRECOVERED_READ_ERROR             0x11
 #define ASC_ILLEGAL_OPCODE                     0x20
 #define ASC_LBA_OUT_OF_RANGE                   0x21
diff --git a/src/scsi/scsi_disk.c b/src/scsi/scsi_disk.c
index 792dad31b..10fdf04a6 100644
--- a/src/scsi/scsi_disk.c
+++ b/src/scsi/scsi_disk.c
@@ -647,6 +647,24 @@ scsi_disk_bus_master_error(scsi_common_t *sc)
     scsi_disk_cmd_error(dev);
 }
 
+static void
+scsi_disk_write_error(scsi_disk_t *dev)
+{
+    scsi_disk_sense_key = SENSE_MEDIUM_ERROR;
+    scsi_disk_asc       = ASC_WRITE_ERROR;
+    scsi_disk_ascq      = 0;
+    scsi_disk_cmd_error(dev);
+}
+
+static void
+scsi_disk_read_error(scsi_disk_t *dev)
+{
+    scsi_disk_sense_key = SENSE_MEDIUM_ERROR;
+    scsi_disk_asc       = ASC_UNRECOVERED_READ_ERROR;
+    scsi_disk_ascq      = 0;
+    scsi_disk_cmd_error(dev);
+}
+
 static void
 scsi_disk_invalid_lun(scsi_disk_t *dev)
 {
@@ -727,10 +745,17 @@ scsi_disk_blocks(scsi_disk_t *dev, int32_t *len, UNUSED(int first_batch), int ou
     *len = dev->requested_blocks << 9;
 
     for (int i = 0; i < dev->requested_blocks; i++) {
-        if (out)
-            hdd_image_write(dev->id, dev->sector_pos + i, 1, dev->temp_buffer + (i << 9));
-        else
-            hdd_image_read(dev->id, dev->sector_pos + i, 1, dev->temp_buffer + (i << 9));
+        if (out) {
+            if (hdd_image_write(dev->id, dev->sector_pos + i, 1, dev->temp_buffer + (i << 9)) < 0) {
+                scsi_disk_write_error(dev);
+                return -1;
+            }
+        } else {
+            if (hdd_image_read(dev->id, dev->sector_pos + i, 1, dev->temp_buffer + (i << 9)) < 0) {
+                scsi_disk_read_error(dev);
+                return -1;
+            }
+        }
     }
 
     scsi_disk_log("%s %i bytes of blocks...\n", out ? "Written" : "Read", *len);
@@ -1040,7 +1065,7 @@ scsi_disk_command(scsi_common_t *sc, uint8_t *cdb)
             ret = scsi_disk_blocks(dev, &alloc_length, 1, 0);
             if (ret <= 0) {
                 scsi_disk_set_phase(dev, SCSI_PHASE_STATUS);
-                dev->packet_status = PHASE_COMPLETE;
+                dev->packet_status = (ret < 0) ? PHASE_ERROR : PHASE_COMPLETE;
                 dev->callback      = 20.0 * SCSI_TIME;
                 scsi_disk_set_callback(dev);
                 scsi_disk_buf_free(dev);
@@ -1501,7 +1526,8 @@ scsi_disk_phase_data_out(scsi_common_t *sc)
                     dev->temp_buffer[6] = (s >> 8) & 0xff;
                     dev->temp_buffer[7] = s & 0xff;
                 }
-                hdd_image_write(dev->id, i, 1, dev->temp_buffer);
+                if (hdd_image_write(dev->id, i, 1, dev->temp_buffer) < 0)
+                    scsi_disk_write_error(dev);
             }
             break;
         case GPCMD_MODE_SELECT_6:

From df945636863142f2eb7d4f08e14e82c76d13f1d3 Mon Sep 17 00:00:00 2001
From: RichardG867 <richardg867@gmail.com>
Date: Tue, 29 Oct 2024 12:39:35 -0300
Subject: [PATCH 20/26] hdc_esdi_at: Report image read/write errors

---
 src/disk/hdc_esdi_at.c | 24 ++++++++++++++++++++----
 1 file changed, 20 insertions(+), 4 deletions(-)

diff --git a/src/disk/hdc_esdi_at.c b/src/disk/hdc_esdi_at.c
index e16d4d729..c08ec636a 100644
--- a/src/disk/hdc_esdi_at.c
+++ b/src/disk/hdc_esdi_at.c
@@ -606,12 +606,16 @@ esdi_callback(void *priv)
             } else {
                 if (get_sector(esdi, &addr)) {
                     esdi->error  = ERR_ID_NOT_FOUND;
+read_error:
                     esdi->status = STAT_READY | STAT_DSC | STAT_ERR;
                     irq_raise(esdi);
                     break;
                 }
 
-                hdd_image_read(drive->hdd_num, addr, 1, (uint8_t *) esdi->buffer);
+                if (hdd_image_read(drive->hdd_num, addr, 1, (uint8_t *) esdi->buffer) < 0) {
+                    esdi->error = ERR_BAD_BLOCK;
+                    goto read_error;
+                }
                 esdi->pos    = 0;
                 esdi->status = STAT_DRQ | STAT_READY | STAT_DSC;
                 irq_raise(esdi);
@@ -628,13 +632,17 @@ esdi_callback(void *priv)
             } else {
                 if (get_sector(esdi, &addr)) {
                     esdi->error  = ERR_ID_NOT_FOUND;
+write_error:
                     esdi->status = STAT_READY | STAT_DSC | STAT_ERR;
                     irq_raise(esdi);
                     ui_sb_update_icon(SB_HDD | HDD_BUS_ESDI, 0);
                     break;
                 }
 
-                hdd_image_write(drive->hdd_num, addr, 1, (uint8_t *) esdi->buffer);
+                if (hdd_image_write(drive->hdd_num, addr, 1, (uint8_t *) esdi->buffer) < 0) {
+                    esdi->error = ERR_BAD_BLOCK;
+                    goto write_error;
+                }
                 irq_raise(esdi);
                 esdi->secount = (esdi->secount - 1) & 0xff;
                 if (esdi->secount) {
@@ -659,13 +667,17 @@ esdi_callback(void *priv)
             } else {
                 if (get_sector(esdi, &addr)) {
                     esdi->error  = ERR_ID_NOT_FOUND;
+verify_error:
                     esdi->status = STAT_READY | STAT_DSC | STAT_ERR;
                     irq_raise(esdi);
                     ui_sb_update_icon(SB_HDD | HDD_BUS_ESDI, 0);
                     break;
                 }
 
-                hdd_image_read(drive->hdd_num, addr, 1, (uint8_t *) esdi->buffer);
+                if (hdd_image_read(drive->hdd_num, addr, 1, (uint8_t *) esdi->buffer) < 0) {
+                    esdi->error = ERR_BAD_BLOCK;
+                    goto verify_error;
+                }
                 ui_sb_update_icon(SB_HDD | HDD_BUS_ESDI, 1);
                 next_sector(esdi);
                 esdi->secount = (esdi->secount - 1) & 0xff;
@@ -692,12 +704,16 @@ esdi_callback(void *priv)
             } else {
                 if (get_sector_format(esdi, &addr)) {
                     esdi->error  = ERR_ID_NOT_FOUND;
+format_error:
                     esdi->status = STAT_READY | STAT_DSC | STAT_ERR;
                     irq_raise(esdi);
                     break;
                 }
 
-                hdd_image_zero(drive->hdd_num, addr, esdi->secount);
+                if (hdd_image_zero(drive->hdd_num, addr, esdi->secount) < 0) {
+                    esdi->error = ERR_BAD_BLOCK;
+                    goto format_error;
+                }
                 esdi->status = STAT_READY | STAT_DSC;
                 irq_raise(esdi);
             }

From ed0c571a8e1798a531b5847b9d2c0592dd51f22a Mon Sep 17 00:00:00 2001
From: RichardG867 <richardg867@gmail.com>
Date: Tue, 29 Oct 2024 12:42:59 -0300
Subject: [PATCH 21/26] hdc_esdi_mca: Report image read/write errors

---
 src/disk/hdc_esdi_mca.c | 31 +++++++++++++++++++++++++++++--
 1 file changed, 29 insertions(+), 2 deletions(-)

diff --git a/src/disk/hdc_esdi_mca.c b/src/disk/hdc_esdi_mca.c
index 1061cbc78..c63c80721 100644
--- a/src/disk/hdc_esdi_mca.c
+++ b/src/disk/hdc_esdi_mca.c
@@ -327,6 +327,27 @@ rba_out_of_range(esdi_t *dev)
     ui_sb_update_icon(SB_HDD | HDD_BUS_ESDI, 0);
 }
 
+static void
+defective_block(esdi_t *dev)
+{
+    dev->status_len     = 9;
+    dev->status_data[0] = dev->command | STATUS_LEN(9) | dev->cmd_dev;
+    dev->status_data[1] = 0x0e01; /*Command block error, invalid parameter*/
+    dev->status_data[2] = 0x0009; /*Defective block*/
+    dev->status_data[3] = 0;
+    dev->status_data[4] = 0;
+    dev->status_data[5] = 0;
+    dev->status_data[6] = 0;
+    dev->status_data[7] = 0;
+    dev->status_data[8] = 0;
+
+    dev->status          = STATUS_IRQ | STATUS_STATUS_OUT_FULL;
+    dev->irq_status      = dev->cmd_dev | IRQ_CMD_COMPLETE_FAILURE;
+    dev->irq_in_progress = 1;
+    set_irq(dev);
+    ui_sb_update_icon(SB_HDD | HDD_BUS_ESDI, 0);
+}
+
 static void
 complete_command_status(esdi_t *dev)
 {
@@ -423,7 +444,10 @@ esdi_callback(void *priv)
                         if (!dev->data_pos) {
                             if (dev->rba >= drive->sectors)
                                 fatal("Read past end of drive\n");
-                            hdd_image_read(drive->hdd_num, dev->rba, 1, (uint8_t *) dev->data);
+                            if (hdd_image_read(drive->hdd_num, dev->rba, 1, (uint8_t *) dev->data) < 0) {
+                                defective_block(dev);
+                                return;
+                            }
                             cmd_time += hdd_timing_read(&hdd[drive->hdd_num], dev->rba, 1);
                             cmd_time += esdi_mca_get_xfer_time(dev, 1);
                         }
@@ -512,7 +536,10 @@ esdi_callback(void *priv)
 
                         if (dev->rba >= drive->sectors)
                             fatal("Write past end of drive\n");
-                        hdd_image_write(drive->hdd_num, dev->rba, 1, (uint8_t *) dev->data);
+                        if (hdd_image_write(drive->hdd_num, dev->rba, 1, (uint8_t *) dev->data) < 0) {
+                            defective_block(dev);
+                            return;
+                        }
                         cmd_time += hdd_timing_write(&hdd[drive->hdd_num], dev->rba, 1);
                         cmd_time += esdi_mca_get_xfer_time(dev, 1);
                         dev->rba++;

From 5734fd9a556ff8c400ebe5b8c1da99e4e64e6bad Mon Sep 17 00:00:00 2001
From: RichardG867 <richardg867@gmail.com>
Date: Tue, 29 Oct 2024 12:49:21 -0300
Subject: [PATCH 22/26] hdc_st506: Report image read/write errors

---
 src/disk/hdc_st506_at.c | 12 ++++++++++--
 src/disk/hdc_st506_xt.c | 32 ++++++++++++++++++++++++--------
 2 files changed, 34 insertions(+), 10 deletions(-)

diff --git a/src/disk/hdc_st506_at.c b/src/disk/hdc_st506_at.c
index d5bbf24d0..9e5b82336 100644
--- a/src/disk/hdc_st506_at.c
+++ b/src/disk/hdc_st506_at.c
@@ -585,12 +585,16 @@ do_callback(void *priv)
             do_seek(mfm);
             if (get_sector(mfm, &addr)) {
                 mfm->error  = ERR_ID_NOT_FOUND;
+read_error:
                 mfm->status = STAT_READY | STAT_DSC | STAT_ERR;
                 irq_raise(mfm);
                 break;
             }
 
-            hdd_image_read(drive->hdd_num, addr, 1, (uint8_t *) mfm->buffer);
+            if (hdd_image_read(drive->hdd_num, addr, 1, (uint8_t *) mfm->buffer) < 0) {
+                mfm->error = ERR_BAD_BLOCK;
+                goto read_error;
+            }
 
             mfm->pos    = 0;
             mfm->status = STAT_DRQ | STAT_READY | STAT_DSC;
@@ -604,12 +608,16 @@ do_callback(void *priv)
             do_seek(mfm);
             if (get_sector(mfm, &addr)) {
                 mfm->error  = ERR_ID_NOT_FOUND;
+write_error:
                 mfm->status = STAT_READY | STAT_DSC | STAT_ERR;
                 irq_raise(mfm);
                 break;
             }
 
-            hdd_image_write(drive->hdd_num, addr, 1, (uint8_t *) mfm->buffer);
+            if (hdd_image_write(drive->hdd_num, addr, 1, (uint8_t *) mfm->buffer) < 0) {
+                mfm->error = ERR_BAD_BLOCK;
+                goto write_error;
+            }
             irq_raise(mfm);
             mfm->secount = (mfm->secount - 1) & 0xff;
 
diff --git a/src/disk/hdc_st506_xt.c b/src/disk/hdc_st506_xt.c
index 47e6f045a..b78466c3b 100644
--- a/src/disk/hdc_st506_xt.c
+++ b/src/disk/hdc_st506_xt.c
@@ -715,6 +715,7 @@ st506_callback(void *priv)
                                  dev->head, dev->sector, dev->count);
 
                     if (!get_sector(dev, drive, &addr)) {
+read_error_start:
                         st506_error(dev, dev->error);
                         st506_complete(dev);
                         return;
@@ -722,8 +723,11 @@ st506_callback(void *priv)
                     ui_sb_update_icon(SB_HDD | HDD_BUS_MFM, 1);
 
                     /* Read data from the image. */
-                    hdd_image_read(drive->hdd_num, addr, 1,
-                                   (uint8_t *) dev->buff);
+                    if (hdd_image_read(drive->hdd_num, addr, 1,
+                                       (uint8_t *) dev->buff) < 0) {
+                        dev->error = ERR_UNC_ERR;
+                        goto read_error_start;
+                    }
 
                     /* Set up the data transfer. */
                     dev->buff_pos = 0;
@@ -765,6 +769,7 @@ st506_callback(void *priv)
                     next_sector(dev, drive);
 
                     if (!get_sector(dev, drive, &addr)) {
+read_error_sent:
                         ui_sb_update_icon(SB_HDD | HDD_BUS_MFM, 0);
                         st506_error(dev, dev->error);
                         st506_complete(dev);
@@ -772,8 +777,11 @@ st506_callback(void *priv)
                     }
 
                     /* Read data from the image. */
-                    hdd_image_read(drive->hdd_num, addr, 1,
-                                   (uint8_t *) dev->buff);
+                    if (hdd_image_read(drive->hdd_num, addr, 1,
+                                       (uint8_t *) dev->buff) < 0) {
+                        dev->error = ERR_UNC_ERR;
+                        goto read_error_sent;
+                    }
 
                     /* Set up the data transfer. */
                     dev->buff_pos = 0;
@@ -856,6 +864,7 @@ st506_callback(void *priv)
 
                 case STATE_RECEIVED_DATA:
                     if (!get_sector(dev, drive, &addr)) {
+write_error:
                         ui_sb_update_icon(SB_HDD | HDD_BUS_MFM, 0);
                         st506_error(dev, dev->error);
                         st506_complete(dev);
@@ -863,8 +872,11 @@ st506_callback(void *priv)
                     }
 
                     /* Write data to image. */
-                    hdd_image_write(drive->hdd_num, addr, 1,
-                                    (uint8_t *) dev->buff);
+                    if (hdd_image_write(drive->hdd_num, addr, 1,
+                                        (uint8_t *) dev->buff) < 0) {
+                        dev->error = ERR_UNC_ERR;
+                        goto write_error;
+                    }
 
                     if (--dev->count == 0) {
                         ui_sb_update_icon(SB_HDD | HDD_BUS_MFM, 0);
@@ -1156,8 +1168,12 @@ st506_callback(void *priv)
                 ui_sb_update_icon(SB_HDD | HDD_BUS_MFM, 1);
 
                 /* Write data to image. */
-                hdd_image_write(drive->hdd_num, addr, 1,
-                                (uint8_t *) dev->buff);
+                if (hdd_image_write(drive->hdd_num, addr, 1,
+                                    (uint8_t *) dev->buff) < 0) {
+                    st506_error(dev, ERR_UNC_ERR);
+                    st506_complete(dev);
+                    return;
+                }
 
                 if (--dev->count == 0) {
                     ui_sb_update_icon(SB_HDD | HDD_BUS_MFM, 0);

From e95b0d1e012fec04f7883eed76b237783eacaf49 Mon Sep 17 00:00:00 2001
From: RichardG867 <richardg867@gmail.com>
Date: Tue, 29 Oct 2024 12:53:54 -0300
Subject: [PATCH 23/26] hdc_xta: Report image read/write errors

---
 src/disk/hdc_xta.c | 21 +++++++++++++++------
 1 file changed, 15 insertions(+), 6 deletions(-)

diff --git a/src/disk/hdc_xta.c b/src/disk/hdc_xta.c
index ede21caf2..a4ef45c43 100644
--- a/src/disk/hdc_xta.c
+++ b/src/disk/hdc_xta.c
@@ -408,8 +408,9 @@ do_fmt:
                         break;
 
                     /* Write the block to the image. */
-                    hdd_image_write(drive->hdd_num, addr, 1,
-                                    (uint8_t *) dev->sector_buf);
+                    if (hdd_image_write(drive->hdd_num, addr, 1,
+                                        (uint8_t *) dev->sector_buf) < 0)
+                        dev->sense = ERR_BADTRK;
                 }
             }
 
@@ -522,6 +523,7 @@ hdc_callback(void *priv)
 do_send:
                     /* Get address of sector to load. */
                     if (get_sector(dev, drive, &addr)) {
+read_error:
                         /* De-activate the status icon. */
                         ui_sb_update_icon(SB_HDD | HDD_BUS_XTA, 0);
                         dev->comp |= COMP_ERR;
@@ -530,8 +532,11 @@ do_send:
                     }
 
                     /* Read the block from the image. */
-                    hdd_image_read(drive->hdd_num, addr, 1,
-                                   (uint8_t *) dev->sector_buf);
+                    if (hdd_image_read(drive->hdd_num, addr, 1,
+                                       (uint8_t *) dev->sector_buf) < 0) {
+                        dev->sense = ERR_BADTRK;
+                        goto read_error;
+                    }
 
                     /* Ready to transfer the data out. */
                     dev->state   = STATE_SDATA;
@@ -673,6 +678,7 @@ do_recv:
 
                     /* Get address of sector to write. */
                     if (get_sector(dev, drive, &addr)) {
+write_error:
                         /* De-activate the status icon. */
                         ui_sb_update_icon(SB_HDD | HDD_BUS_XTA, 0);
 
@@ -682,8 +688,11 @@ do_recv:
                     }
 
                     /* Write the block to the image. */
-                    hdd_image_write(drive->hdd_num, addr, 1,
-                                    (uint8_t *) dev->sector_buf);
+                    if (hdd_image_write(drive->hdd_num, addr, 1,
+                                        (uint8_t *) dev->sector_buf) < 0) {
+                        dev->sense = ERR_BADTRK;
+                        goto write_error;
+                    }
 
                     dev->buf_idx = 0;
                     if (--dev->count == 0) {

From b155c50e96cf76af7857f5834ef1d7891fc0041d Mon Sep 17 00:00:00 2001
From: RichardG867 <richardg867@gmail.com>
Date: Tue, 29 Oct 2024 13:57:21 -0300
Subject: [PATCH 24/26] zip/mo: Report image read/write errors

---
 src/disk/mo.c  | 45 +++++++++++++++++++++++++++++++++++++--------
 src/disk/zip.c | 45 +++++++++++++++++++++++++++++++++++++--------
 2 files changed, 74 insertions(+), 16 deletions(-)

diff --git a/src/disk/mo.c b/src/disk/mo.c
index ed1ab2472..5f757c8af 100644
--- a/src/disk/mo.c
+++ b/src/disk/mo.c
@@ -874,6 +874,24 @@ mo_write_protected(mo_t *dev)
     mo_cmd_error(dev);
 }
 
+static void
+mo_write_error(mo_t *dev)
+{
+    mo_sense_key = SENSE_MEDIUM_ERROR;
+    mo_asc       = ASC_WRITE_ERROR;
+    mo_ascq      = 0;
+    mo_cmd_error(dev);
+}
+
+static void
+mo_read_error(mo_t *dev)
+{
+    mo_sense_key = SENSE_MEDIUM_ERROR;
+    mo_asc       = ASC_UNRECOVERED_READ_ERROR;
+    mo_ascq      = 0;
+    mo_cmd_error(dev);
+}
+
 static void
 mo_invalid_lun(mo_t *dev)
 {
@@ -928,7 +946,7 @@ mo_blocks(mo_t *dev, int32_t *len, UNUSED(int first_batch), int out)
 
     if (!dev->sector_len) {
         mo_command_complete(dev);
-        return -1;
+        return 0;
     }
 
     mo_log("%sing %i blocks starting from %i...\n", out ? "Writ" : "Read", dev->requested_blocks, dev->sector_pos);
@@ -942,20 +960,31 @@ mo_blocks(mo_t *dev, int32_t *len, UNUSED(int first_batch), int out)
     *len = dev->requested_blocks * dev->drv->sector_size;
 
     for (int i = 0; i < dev->requested_blocks; i++) {
-        if (fseek(dev->drv->fp, dev->drv->base + (dev->sector_pos * dev->drv->sector_size) + (i * dev->drv->sector_size), SEEK_SET) == 1)
-            break;
+        if (fseek(dev->drv->fp, dev->drv->base + (dev->sector_pos * dev->drv->sector_size) + (i * dev->drv->sector_size), SEEK_SET) == -1) {
+            if (out)
+                mo_write_error(dev);
+            else
+                mo_read_error(dev);
+            return -1;
+        }
 
         if (feof(dev->drv->fp))
             break;
 
         if (out) {
-            if (fwrite(dev->buffer + (i * dev->drv->sector_size), 1, dev->drv->sector_size, dev->drv->fp) != dev->drv->sector_size)
-                fatal("mo_blocks(): Error writing data\n");
+            if (fwrite(dev->buffer + (i * dev->drv->sector_size), 1, dev->drv->sector_size, dev->drv->fp) != dev->drv->sector_size) {
+                mo_log("mo_blocks(): Error writing data\n");
+                mo_write_error(dev);
+                return -1;
+            }
 
             fflush(dev->drv->fp);
         } else {
-            if (fread(dev->buffer + (i * dev->drv->sector_size), 1, dev->drv->sector_size, dev->drv->fp) != dev->drv->sector_size)
-                fatal("mo_blocks(): Error reading data\n");
+            if (fread(dev->buffer + (i * dev->drv->sector_size), 1, dev->drv->sector_size, dev->drv->fp) != dev->drv->sector_size) {
+                mo_log("mo_blocks(): Error reading data\n");
+                mo_read_error(dev);
+                return -1;
+            }
         }
     }
 
@@ -1433,7 +1462,7 @@ mo_command(scsi_common_t *sc, uint8_t *cdb)
             ret = mo_blocks(dev, &alloc_length, 1, 0);
             if (ret <= 0) {
                 mo_set_phase(dev, SCSI_PHASE_STATUS);
-                dev->packet_status = PHASE_COMPLETE;
+                dev->packet_status = (ret < 0) ? PHASE_ERROR : PHASE_COMPLETE;
                 dev->callback      = 20.0 * MO_TIME;
                 mo_set_callback(dev);
                 mo_buf_free(dev);
diff --git a/src/disk/zip.c b/src/disk/zip.c
index a948cab2f..4a5a9b968 100644
--- a/src/disk/zip.c
+++ b/src/disk/zip.c
@@ -1048,6 +1048,24 @@ zip_write_protected(zip_t *dev)
     zip_cmd_error(dev);
 }
 
+static void
+zip_write_error(zip_t *dev)
+{
+    zip_sense_key = SENSE_MEDIUM_ERROR;
+    zip_asc       = ASC_WRITE_ERROR;
+    zip_ascq      = 0;
+    zip_cmd_error(dev);
+}
+
+static void
+zip_read_error(zip_t *dev)
+{
+    zip_sense_key = SENSE_MEDIUM_ERROR;
+    zip_asc       = ASC_UNRECOVERED_READ_ERROR;
+    zip_ascq      = 0;
+    zip_cmd_error(dev);
+}
+
 static void
 zip_invalid_lun(zip_t *dev)
 {
@@ -1111,7 +1129,7 @@ zip_blocks(zip_t *dev, int32_t *len, UNUSED(int first_batch), int out)
 
     if (!dev->sector_len) {
         zip_command_complete(dev);
-        return -1;
+        return 0;
     }
 
     zip_log("%sing %i blocks starting from %i...\n", out ? "Writ" : "Read", dev->requested_blocks, dev->sector_pos);
@@ -1125,20 +1143,31 @@ zip_blocks(zip_t *dev, int32_t *len, UNUSED(int first_batch), int out)
     *len = dev->requested_blocks << 9;
 
     for (int i = 0; i < dev->requested_blocks; i++) {
-        if (fseek(dev->drv->fp, dev->drv->base + (dev->sector_pos << 9) + (i << 9), SEEK_SET) == 1)
-            break;
+        if (fseek(dev->drv->fp, dev->drv->base + (dev->sector_pos << 9) + (i << 9), SEEK_SET) == -1) {
+            if (out)
+                zip_write_error(dev);
+            else
+                zip_read_error(dev);
+            return -1;
+        }
 
         if (feof(dev->drv->fp))
             break;
 
         if (out) {
-            if (fwrite(dev->buffer + (i << 9), 1, 512, dev->drv->fp) != 512)
-                fatal("zip_blocks(): Error writing data\n");
+            if (fwrite(dev->buffer + (i << 9), 1, 512, dev->drv->fp) != 512) {
+                zip_log("zip_blocks(): Error writing data\n");
+                zip_write_error(dev);
+                return -1;
+            }
 
             fflush(dev->drv->fp);
         } else {
-            if (fread(dev->buffer + (i << 9), 1, 512, dev->drv->fp) != 512)
-                fatal("zip_blocks(): Error reading data\n");
+            if (fread(dev->buffer + (i << 9), 1, 512, dev->drv->fp) != 512) {
+                zip_log("zip_blocks(): Error reading data\n");
+                zip_read_error(dev);
+                return -1;
+            }
         }
     }
 
@@ -1551,7 +1580,7 @@ zip_command(scsi_common_t *sc, uint8_t *cdb)
             ret = zip_blocks(dev, &alloc_length, 1, 0);
             if (ret <= 0) {
                 zip_set_phase(dev, SCSI_PHASE_STATUS);
-                dev->packet_status = PHASE_COMPLETE;
+                dev->packet_status = (ret < 0) ? PHASE_ERROR : PHASE_COMPLETE;
                 dev->callback      = 20.0 * ZIP_TIME;
                 zip_set_callback(dev);
                 zip_buf_free(dev);

From cc9c48711a036334d2a7effcac86d68d062bc917 Mon Sep 17 00:00:00 2001
From: RichardG867 <richardg867@gmail.com>
Date: Tue, 29 Oct 2024 14:15:08 -0300
Subject: [PATCH 25/26] hdd_image: Improve I/O error detection

---
 src/disk/hdd_image.c | 10 +++++-----
 1 file changed, 5 insertions(+), 5 deletions(-)

diff --git a/src/disk/hdd_image.c b/src/disk/hdd_image.c
index 2b463a23c..4607e2b14 100644
--- a/src/disk/hdd_image.c
+++ b/src/disk/hdd_image.c
@@ -538,9 +538,9 @@ hdd_image_read(uint8_t id, uint32_t sector, uint32_t count, uint8_t *buffer)
         }
 
         num_read           = fread(buffer, 512, count, hdd_images[id].file);
-        if (count && !num_read)
-            return -1;
         hdd_images[id].pos = sector + num_read;
+        if (num_read < count)
+            return -1;
     }
 
     return 0;
@@ -591,10 +591,10 @@ hdd_image_write(uint8_t id, uint32_t sector, uint32_t count, uint8_t *buffer)
         }
 
         num_write          = fwrite(buffer, 512, count, hdd_images[id].file);
-        if (count && !num_write)
-            return -1;
-        fflush(hdd_images[id].file);
         hdd_images[id].pos = sector + num_write;
+        fflush(hdd_images[id].file);
+        if (num_write < count)
+            return -1;
     }
 
     return 0;

From 2184a982f83c54dcbb1600fb7c82131e2d270331 Mon Sep 17 00:00:00 2001
From: RichardG867 <richardg867@gmail.com>
Date: Tue, 29 Oct 2024 15:14:27 -0300
Subject: [PATCH 26/26] MiniVHD: Report read/write errors

---
 src/disk/hdd_image.c          | 17 ++++++--
 src/disk/minivhd/internal.h   |  3 +-
 src/disk/minivhd/minivhd_io.c | 80 +++++++++++++++++++++++------------
 3 files changed, 69 insertions(+), 31 deletions(-)

diff --git a/src/disk/hdd_image.c b/src/disk/hdd_image.c
index 4607e2b14..402140408 100644
--- a/src/disk/hdd_image.c
+++ b/src/disk/hdd_image.c
@@ -529,8 +529,11 @@ hdd_image_read(uint8_t id, uint32_t sector, uint32_t count, uint8_t *buffer)
     size_t num_read;
 
     if (hdd_images[id].type == HDD_IMAGE_VHD) {
-        non_transferred_sectors = mvhd_read_sectors(hdd_images[id].vhd, sector, count, buffer);
-        hdd_images[id].pos      = sector + count - non_transferred_sectors - 1;
+        hdd_images[id].vhd->error = 0;
+        non_transferred_sectors   = mvhd_read_sectors(hdd_images[id].vhd, sector, count, buffer);
+        hdd_images[id].pos        = sector + count - non_transferred_sectors - 1;
+        if (hdd_images[id].vhd->error)
+            return -1;
     } else {
         if (!hdd_images[id].file || (fseeko64(hdd_images[id].file, ((uint64_t) (sector) << 9LL) + hdd_images[id].base, SEEK_SET) == -1)) {
             hdd_image_log("Hard disk image %i: Read error during seek\n", id);
@@ -582,8 +585,11 @@ hdd_image_write(uint8_t id, uint32_t sector, uint32_t count, uint8_t *buffer)
     size_t num_write;
 
     if (hdd_images[id].type == HDD_IMAGE_VHD) {
-        non_transferred_sectors = mvhd_write_sectors(hdd_images[id].vhd, sector, count, buffer);
-        hdd_images[id].pos      = sector + count - non_transferred_sectors - 1;
+        hdd_images[id].vhd->error = 0;
+        non_transferred_sectors   = mvhd_write_sectors(hdd_images[id].vhd, sector, count, buffer);
+        hdd_images[id].pos        = sector + count - non_transferred_sectors - 1;
+        if (hdd_images[id].vhd->error)
+            return -1;
     } else {
         if (!hdd_images[id].file || (fseeko64(hdd_images[id].file, ((uint64_t) (sector) << 9LL) + hdd_images[id].base, SEEK_SET) == -1)) {
             hdd_image_log("Hard disk image %i: Write error during seek\n", id);
@@ -621,8 +627,11 @@ int
 hdd_image_zero(uint8_t id, uint32_t sector, uint32_t count)
 {
     if (hdd_images[id].type == HDD_IMAGE_VHD) {
+        hdd_images[id].vhd->error   = 0;
         int non_transferred_sectors = mvhd_format_sectors(hdd_images[id].vhd, sector, count);
         hdd_images[id].pos          = sector + count - non_transferred_sectors - 1;
+        if (hdd_images[id].vhd->error)
+            return -1;
     } else {
         memset(empty_sector, 0, 512);
 
diff --git a/src/disk/minivhd/internal.h b/src/disk/minivhd/internal.h
index d3f930110..4caef7c47 100644
--- a/src/disk/minivhd/internal.h
+++ b/src/disk/minivhd/internal.h
@@ -115,6 +115,7 @@ typedef struct MVHDSparseHeader {
 struct MVHDMeta {
     FILE*            f;
     bool             readonly;
+    bool             error;
     char             filename[MVHD_MAX_PATH_BYTES];
     struct MVHDMeta* parent;
     MVHDFooter       footer;
@@ -271,7 +272,7 @@ struct MVHDMeta* mvhd_create_fixed_raw(const char* path, FILE* raw_img, uint64_t
  * \param [in] f File to write sectors to
  * \param [in] sector_count The number of sectors to write
  */
-void mvhd_write_empty_sectors(FILE* f, int sector_count);
+bool mvhd_write_empty_sectors(FILE* f, int sector_count);
 
 /**
  * \brief Read a fixed VHD image
diff --git a/src/disk/minivhd/minivhd_io.c b/src/disk/minivhd/minivhd_io.c
index 1a349dbc4..cdf6694a8 100644
--- a/src/disk/minivhd/minivhd_io.c
+++ b/src/disk/minivhd/minivhd_io.c
@@ -82,15 +82,18 @@ check_sectors(uint32_t offset, int num_sectors, uint32_t total_sectors, int *tra
     }
 }
 
-void
+bool
 mvhd_write_empty_sectors(FILE *f, int sector_count)
 {
     uint8_t zero_bytes[MVHD_SECTOR_SIZE] = {0};
 
-    for (int i = 0; i < sector_count; i++)
-        fwrite(zero_bytes, sizeof zero_bytes, 1, f);
+    for (int i = 0; i < sector_count; i++) {
+        if (!fwrite(zero_bytes, sizeof zero_bytes, 1, f))
+            return 0;
+    }
 
     fflush(f);
+    return 1;
 }
 
 /**
@@ -107,7 +110,8 @@ read_sect_bitmap(MVHDMeta *vhdm, int blk)
 {
     if (vhdm->block_offset[blk] != MVHD_SPARSE_BLK) {
         mvhd_fseeko64(vhdm->f, (uint64_t)vhdm->block_offset[blk] * MVHD_SECTOR_SIZE, SEEK_SET);
-        (void) !fread(vhdm->bitmap.curr_bitmap, vhdm->bitmap.sector_count * MVHD_SECTOR_SIZE, 1, vhdm->f);
+        if (!fread(vhdm->bitmap.curr_bitmap, vhdm->bitmap.sector_count * MVHD_SECTOR_SIZE, 1, vhdm->f))
+            vhdm->error = 1;
     } else
         memset(vhdm->bitmap.curr_bitmap, 0, vhdm->bitmap.sector_count * MVHD_SECTOR_SIZE);
 
@@ -124,8 +128,10 @@ write_curr_sect_bitmap(MVHDMeta* vhdm)
 {
     if (vhdm->bitmap.curr_block >= 0) {
         int64_t abs_offset = (int64_t)vhdm->block_offset[vhdm->bitmap.curr_block] * MVHD_SECTOR_SIZE;
-        mvhd_fseeko64(vhdm->f, abs_offset, SEEK_SET);
-        fwrite(vhdm->bitmap.curr_bitmap, MVHD_SECTOR_SIZE, vhdm->bitmap.sector_count, vhdm->f);
+        if (mvhd_fseeko64(vhdm->f, abs_offset, SEEK_SET) == -1)
+            vhdm->error = 1;
+        if (!fwrite(vhdm->bitmap.curr_bitmap, MVHD_SECTOR_SIZE, vhdm->bitmap.sector_count, vhdm->f))
+            vhdm->error = 1;
     }
 }
 
@@ -141,8 +147,10 @@ write_bat_entry(MVHDMeta *vhdm, int blk)
     uint64_t table_offset = vhdm->sparse.bat_offset + ((uint64_t)blk * sizeof *vhdm->block_offset);
     uint32_t offset = mvhd_to_be32(vhdm->block_offset[blk]);
 
-    mvhd_fseeko64(vhdm->f, table_offset, SEEK_SET);
-    fwrite(&offset, sizeof offset, 1, vhdm->f);
+    if (mvhd_fseeko64(vhdm->f, table_offset, SEEK_SET) == -1)
+        vhdm->error = 1;
+    if (!fwrite(&offset, sizeof offset, 1, vhdm->f))
+        vhdm->error = 1;
     fflush(vhdm->f);
 }
 
@@ -173,7 +181,8 @@ create_block(MVHDMeta *vhdm, int blk)
     if (!mvhd_is_conectix_str(footer)) {
         /* Oh dear. We use the header instead, since something has gone wrong at the footer */
         mvhd_fseeko64(vhdm->f, 0, SEEK_SET);
-        (void) !fread(footer, sizeof footer, 1, vhdm->f);
+        if (!fread(footer, sizeof footer, 1, vhdm->f))
+            vhdm->error = 1;
         mvhd_fseeko64(vhdm->f, 0, SEEK_END);
     }
 
@@ -182,20 +191,25 @@ create_block(MVHDMeta *vhdm, int blk)
         /* Yikes! We're supposed to be on a sector boundary. Add some padding */
         int64_t padding_amount = ((int64_t) MVHD_SECTOR_SIZE) - (abs_offset % MVHD_SECTOR_SIZE);
         uint8_t zero_byte = 0;
-        for (int i = 0; i < padding_amount; i++)
-            fwrite(&zero_byte, sizeof zero_byte, 1, vhdm->f);
+        for (int i = 0; i < padding_amount; i++) {
+            if (!fwrite(&zero_byte, sizeof zero_byte, 1, vhdm->f))
+                vhdm->error = 1;
+        }
         abs_offset += padding_amount;
     }
 
     uint32_t sect_offset = (uint32_t)(abs_offset / MVHD_SECTOR_SIZE);
     int blk_size_sectors = vhdm->sparse.block_sz / MVHD_SECTOR_SIZE;
-    mvhd_write_empty_sectors(vhdm->f, vhdm->bitmap.sector_count + blk_size_sectors);
+    if (!mvhd_write_empty_sectors(vhdm->f, vhdm->bitmap.sector_count + blk_size_sectors))
+        vhdm->error = 1;
 
     /* Add a bit of padding. That's what Windows appears to do, although it's not strictly necessary... */
-    mvhd_write_empty_sectors(vhdm->f, 5);
+    if (!mvhd_write_empty_sectors(vhdm->f, 5))
+        vhdm->error = 1;
 
     /* And we finish with the footer */
-    fwrite(footer, sizeof footer, 1, vhdm->f);
+    if (!fwrite(footer, sizeof footer, 1, vhdm->f))
+        vhdm->error = 1;
 
     /* We no longer have a sparse block. Update that BAT! */
     vhdm->block_offset[blk] = sect_offset;
@@ -214,8 +228,10 @@ mvhd_fixed_read(MVHDMeta *vhdm, uint32_t offset, int num_sectors, void *out_buff
     check_sectors(offset, num_sectors, total_sectors, &transfer_sectors, &truncated_sectors);
 
     addr = ((int64_t) offset) * MVHD_SECTOR_SIZE;
-    mvhd_fseeko64(vhdm->f, addr, SEEK_SET);
-    (void) !fread(out_buff, transfer_sectors * MVHD_SECTOR_SIZE, 1, vhdm->f);
+    if (mvhd_fseeko64(vhdm->f, addr, SEEK_SET) == -1)
+        vhdm->error = 1;
+    if (!fread(out_buff, transfer_sectors * MVHD_SECTOR_SIZE, 1, vhdm->f))
+        vhdm->error = 1;
 
     return truncated_sectors;
 }
@@ -245,17 +261,20 @@ mvhd_sparse_read(MVHDMeta *vhdm, uint32_t offset, int num_sectors, void *out_buf
             prev_blk = blk;
             if (vhdm->bitmap.curr_block != blk) {
                 read_sect_bitmap(vhdm, blk);
-                mvhd_fseeko64(vhdm->f, (uint64_t)sib * MVHD_SECTOR_SIZE, SEEK_CUR);
+                if (mvhd_fseeko64(vhdm->f, (uint64_t)sib * MVHD_SECTOR_SIZE, SEEK_CUR) == -1)
+                    vhdm->error = 1;
             } else {
                 addr = (((int64_t) vhdm->block_offset[blk]) + vhdm->bitmap.sector_count + sib) *
                        MVHD_SECTOR_SIZE;
-                mvhd_fseeko64(vhdm->f, addr, SEEK_SET);
+                if (mvhd_fseeko64(vhdm->f, addr, SEEK_SET) == -1)
+                    vhdm->error = 1;
             }
         }
 
-        if (VHD_TESTBIT(vhdm->bitmap.curr_bitmap, sib))
-            (void) !fread(buff, MVHD_SECTOR_SIZE, 1, vhdm->f);
-        else {
+        if (VHD_TESTBIT(vhdm->bitmap.curr_bitmap, sib)) {
+            if (!fread(buff, MVHD_SECTOR_SIZE, 1, vhdm->f))
+                vhdm->error = 1;
+        } else {
             memset(buff, 0, MVHD_SECTOR_SIZE);
             mvhd_fseeko64(vhdm->f, MVHD_SECTOR_SIZE, SEEK_CUR);
         }
@@ -301,6 +320,10 @@ mvhd_diff_read(MVHDMeta *vhdm, uint32_t offset, int num_sectors, void *out_buff)
             mvhd_sparse_read(curr_vhdm, s, 1, buff);
         else
             mvhd_fixed_read(curr_vhdm, s, 1, buff);
+        if (curr_vhdm->error) {
+            curr_vhdm->error = 0;
+            vhdm->error = 1;
+        }
 
         curr_vhdm = vhdm;
         buff += MVHD_SECTOR_SIZE;
@@ -320,8 +343,10 @@ mvhd_fixed_write(MVHDMeta *vhdm, uint32_t offset, int num_sectors, void *in_buff
     check_sectors(offset, num_sectors, total_sectors, &transfer_sectors, &truncated_sectors);
 
     addr = (int64_t)offset * MVHD_SECTOR_SIZE;
-    mvhd_fseeko64(vhdm->f, addr, SEEK_SET);
-    fwrite(in_buff, transfer_sectors * MVHD_SECTOR_SIZE, 1, vhdm->f);
+    if (mvhd_fseeko64(vhdm->f, addr, SEEK_SET) == -1)
+        vhdm->error = 1;
+    if (!fwrite(in_buff, transfer_sectors * MVHD_SECTOR_SIZE, 1, vhdm->f))
+        vhdm->error = 1;
     fflush(vhdm->f);
 
     return truncated_sectors;
@@ -364,16 +389,19 @@ mvhd_sparse_diff_write(MVHDMeta *vhdm, uint32_t offset, int num_sectors, void *i
             if (blk != prev_blk) {
                 if (vhdm->bitmap.curr_block != blk) {
                     read_sect_bitmap(vhdm, blk);
-                    mvhd_fseeko64(vhdm->f, (uint64_t)sib * MVHD_SECTOR_SIZE, SEEK_CUR);
+                    if (mvhd_fseeko64(vhdm->f, (uint64_t)sib * MVHD_SECTOR_SIZE, SEEK_CUR) == -1)
+                        vhdm->error = 1;
                 } else {
                     addr = (((int64_t) vhdm->block_offset[blk]) + vhdm->bitmap.sector_count + sib) *
                            MVHD_SECTOR_SIZE;
-                    mvhd_fseeko64(vhdm->f, addr, SEEK_SET);
+                    if (mvhd_fseeko64(vhdm->f, addr, SEEK_SET) == -1)
+                        vhdm->error = 1;
                 }
                 prev_blk = blk;
             }
 
-            fwrite(buff, MVHD_SECTOR_SIZE, 1, vhdm->f);
+            if (!fwrite(buff, MVHD_SECTOR_SIZE, 1, vhdm->f))
+                vhdm->error = 1;
             VHD_SETBIT(vhdm->bitmap.curr_bitmap, sib);
             buff += MVHD_SECTOR_SIZE;
         }