diff --git a/doc/build.md b/doc/build.md
index a816e5d98..0424b702b 100644
--- a/doc/build.md
+++ b/doc/build.md
@@ -30,7 +30,7 @@ In order to compile 86Box from this repository, please follow this step-by-step
 
 7. If the compilation succeeded (which it almost always should), you will find `86Box.exe` in the `src` directory.
 
-8. In order to test your fresh build, replace the `86Box.exe` in your current 86Box environment with your freshly built one. If you do not have a pre-existing 86Box environment, download the latest successful build from http://ci.86box.net, and the latest ROM set from https://github.com/86Box/roms.
+8. In order to test your fresh build, replace the `86Box.exe` in your current 86Box environment with your freshly built one. If you do not have a pre-existing 86Box environment, download the latest successful build from https://ci.86box.net, and the latest ROM set from https://github.com/86Box/roms.
 
 9. Enjoy using and testing the emulator! :)
 
diff --git a/src/chipset/ali1217.c b/src/chipset/ali1217.c
index 878ed81df..ab05ac0a0 100644
--- a/src/chipset/ali1217.c
+++ b/src/chipset/ali1217.c
@@ -33,6 +33,7 @@
 #include <86box/port_92.h>
 #include <86box/chipset.h>
 
+
 #ifdef ENABLE_ALI1217_LOG
 int ali1217_do_log = ENABLE_ALI1217_LOG;
 static void
@@ -57,18 +58,12 @@ typedef struct
     int cfg_locked;
 } ali1217_t;
 
-static void ali1217_shadow_recalc(ali1217_t *dev)
+static void ali1217_shadow_recalc(int reg_15, ali1217_t *dev)
 {
     for (uint8_t i = 0; i < 4; i++)
-    {
-        mem_set_mem_state_both(0xc0000 + (i << 15), 0x8000, ((dev->regs[0x14] & (1 << (i * 2))) ? MEM_READ_INTERNAL : MEM_READ_EXTANY) | ((dev->regs[0x14] & (1 << ((i * 2) + 1))) ? MEM_WRITE_INTERNAL : MEM_WRITE_EXTANY));
-        mem_set_mem_state_both(0xe0000 + (i << 15), 0x8000, ((dev->regs[0x15] & (1 << (i * 2))) ? MEM_READ_INTERNAL : MEM_READ_EXTANY) | ((dev->regs[0x15] & (1 << ((i * 2) + 1))) ? MEM_WRITE_INTERNAL : MEM_WRITE_EXTANY));
-    }
+    mem_set_mem_state_both((reg_15 ? 0xe0000 : 0xc0000) + (i << 15), 0x8000, ((dev->regs[0x14 + reg_15] & (1 << (i * 2))) ? MEM_READ_INTERNAL : MEM_READ_EXTANY) | ((dev->regs[0x14 + reg_15] & (1 << ((i * 2) + 1))) ? MEM_WRITE_INTERNAL : MEM_WRITE_EXTANY));
 
-    shadowbios = !!(dev->regs[0x15] & 5);
-    shadowbios_write = !!(dev->regs[0x15] & 0x0a);
-
-    flushmmucache();
+    flushmmucache_nopc();
 }
 
 static void
@@ -84,8 +79,7 @@ ali1217_write(uint16_t addr, uint8_t val, void *priv)
     case 0x23:
         if (dev->index != 0x13)
             ali1217_log("ALi M1217: dev->regs[%02x] = %02x\n", dev->index, val);
-
-        if (dev->index == 0x13)
+        else
             dev->cfg_locked = !(val == 0xc5);
 
         if (!dev->cfg_locked)
@@ -93,7 +87,7 @@ ali1217_write(uint16_t addr, uint8_t val, void *priv)
             dev->regs[dev->index] = val;
 
             if ((dev->index == 0x14) || (dev->index == 0x15))
-                ali1217_shadow_recalc(dev);
+                ali1217_shadow_recalc(dev->index & 1, dev);
         }
         break;
     }
@@ -104,7 +98,7 @@ ali1217_read(uint16_t addr, void *priv)
 {
     ali1217_t *dev = (ali1217_t *)priv;
 
-    return !(addr == 0x22) ? dev->regs[dev->index] : dev->index;
+    return (addr == 0x23) ? dev->regs[dev->index] : 0xff;
 }
 
 static void
@@ -134,7 +128,6 @@ ali1217_init(const device_t *info)
 
     */
     io_sethandler(0x0022, 0x0002, ali1217_read, NULL, NULL, ali1217_write, NULL, NULL, dev);
-    ali1217_shadow_recalc(dev);
 
     return dev;
 }
diff --git a/src/chipset/umc_hb4.c b/src/chipset/umc_hb4.c
index 44f2d4c0e..f9371ae5f 100644
--- a/src/chipset/umc_hb4.c
+++ b/src/chipset/umc_hb4.c
@@ -141,6 +141,8 @@ void hb4_shadow(int cur_addr, hb4_t *dev)
         mem_set_mem_state_both(0xc8000 + ((i - 2) << 14), 0x4000, (dev->pci_conf[0x54] & (1 << i)) ? (CAN_READ | CAN_WRITE) : DISABLE);
 
     mem_set_mem_state_both(0xe0000, 0x20000, CAN_READ | CAN_WRITE);
+	
+	flushmmucache_nopc();
 }
 
 static void
diff --git a/src/cpu/cpu_table.c b/src/cpu/cpu_table.c
index 56685cbf8..407021c0f 100644
--- a/src/cpu/cpu_table.c
+++ b/src/cpu/cpu_table.c
@@ -970,7 +970,7 @@ const cpu_family_t cpu_families[] = {
     }, {
 	.package = CPU_PKG_EBGA368,
 	.manufacturer = "VIA",
-	.name = "Eden(Model 7)",
+	.name = "Eden Model 7",
 	.internal_name = "c3_eden",
 	.cpus = (const CPU[]) {
 		{"100",     CPU_EDEN, fpus_internal, 100000000, 1.5, 2050,   0x673, 0x673, 0, CPU_SUPPORTS_DYNAREC | CPU_FIXED_MULTIPLIER,  9,  9,  4,  4, 12}, /* out of spec */
diff --git a/src/include/86box/snd_sb_dsp.h b/src/include/86box/snd_sb_dsp.h
index 263897711..060b3835a 100644
--- a/src/include/86box/snd_sb_dsp.h
+++ b/src/include/86box/snd_sb_dsp.h
@@ -117,6 +117,6 @@ void sb_dsp_poll(sb_dsp_t *dsp, int16_t *l, int16_t *r);
 void sb_dsp_set_stereo(sb_dsp_t *dsp, int stereo);
 
 void sb_dsp_update(sb_dsp_t *dsp);
-void sb_update_irq(sb_dsp_t *dsp);
+void sb_update_mask(sb_dsp_t *dsp, int irqm8, int irqm16, int irqm401);
 
 #endif /* SOUND_SND_SB_DSP_H */
\ No newline at end of file
diff --git a/src/mem/mem.c b/src/mem/mem.c
index a3ff49148..98ac19f51 100644
--- a/src/mem/mem.c
+++ b/src/mem/mem.c
@@ -793,6 +793,7 @@ writemembl(uint32_t addr, uint8_t val)
 {
     uint64_t addr64 = (uint64_t) addr;
     mem_mapping_t *map;
+
     mem_logical_addr = addr;
 
     if (page_lookup[addr>>12] && page_lookup[addr>>12]->write_b) {
@@ -842,6 +843,7 @@ void
 writemembl_no_mmut(uint32_t addr, uint64_t addr64, uint8_t val)
 {
     mem_mapping_t *map;
+
     mem_logical_addr = addr;
 
     if (page_lookup[addr >> 12] && page_lookup[addr >> 12]->write_b) {
@@ -869,8 +871,6 @@ readmemwl(uint32_t addr)
     uint64_t addr64[2];
     mem_mapping_t *map;
     int i;
-    uint16_t ret = 0x0000;
-    uint32_t prev_page = 0xffffffff;
 
     addr64[0] = (uint64_t) addr;
     addr64[1] = (uint64_t) (addr + 1);
@@ -883,25 +883,15 @@ readmemwl(uint32_t addr)
 	if ((addr & 0xfff) > 0xffe) {
 		if (cr0 >> 31) {
 			for (i = 0; i < 2; i++) {
-				/* If we are on the same page, there is no need to translate again, as we can just
-				   reuse the previous result. */
-				if ((i > 0) && (((addr + i) & ~0xfff) == prev_page))
-					addr64[i] = (addr64[i - 1] & ~0xfffLL) | ((uint64_t) ((addr + i) & 0xfff));
-				else
-					addr64[i] = mmutranslate_read(addr + i);
-
-				prev_page = ((addr + i) & ~0xfff);
+				addr64[i] = mmutranslate_read(addr + i);
 
 				if (addr64[i] > 0xffffffffULL)
 					return 0xffff;
 			}
 		}
-		/* No need to waste precious CPU host cycles on mmutranslate's that were already done, just pass
-		   their result as a parameter to be used if needed. */
-		for (i = 0; i < 2; i++)
-			ret |= (readmembl_no_mmut(addr + i, addr64[i]) << (i << 3));
 
-		return ret;
+		return readmembl_no_mmut(addr, addr64[0]) |
+		       (((uint16_t) readmembl_no_mmut(addr + 1, addr64[1])) << 8);
 	} else if (readlookup2[addr >> 12] != (uintptr_t) LOOKUP_INV)
 		return *(uint16_t *)(readlookup2[addr >> 12] + addr);
     }
@@ -935,7 +925,6 @@ writememwl(uint32_t addr, uint16_t val)
     uint64_t addr64[2];
     mem_mapping_t *map;
     int i;
-    uint32_t prev_page = 0xffffffff;
 
     addr64[0] = (uint64_t) addr;
     addr64[1] = (uint64_t) (addr + 1);
@@ -951,26 +940,18 @@ writememwl(uint32_t addr, uint16_t val)
 				/* Do not translate a page that has a valid lookup, as that is by definition valid
 				   and the whole purpose of the lookup is to avoid repeat identical translations. */
 				if (!page_lookup[(addr + i) >> 12] || !page_lookup[(addr + i) >> 12]->write_b) {
-					/* If we are on the same page, there is no need to translate again, as we can just
-					   reuse the previous result. */
-					if ((i > 0) && (((addr + i) & ~0xfff) == prev_page))
-						addr64[i] = (addr64[i - 1] & ~0xfffLL) | ((uint64_t) ((addr + i) & 0xfff));
-					else
-						addr64[i] = mmutranslate_write(addr + i);
-
-					prev_page = ((addr + i) & ~0xfff);
+					addr64[i] = mmutranslate_write(addr + i);
 
 					if (addr64[i] > 0xffffffffULL)
 						return;
-				} else
-					prev_page = 0xffffffff;
+				}
 			}
 		}
+
 		/* No need to waste precious CPU host cycles on mmutranslate's that were already done, just pass
 		   their result as a parameter to be used if needed. */
-		for (i = 0; i < 2; i++)
-			writemembl_no_mmut(addr + i, addr64[i], val >> (i << 3));
-
+		writemembl_no_mmut(addr, addr64[0], val);
+		writemembl_no_mmut(addr + 1, addr64[1], val >> 8);
 		return;
 	} else if (writelookup2[addr >> 12] != (uintptr_t) LOOKUP_INV) {
 		*(uint16_t *)(writelookup2[addr >> 12] + addr) = val;
@@ -1013,7 +994,6 @@ readmemwl_no_mmut(uint32_t addr, uint64_t *addr64)
 {
     mem_mapping_t *map;
     int i;
-    uint16_t ret = 0x0000;
 
     mem_logical_addr = addr;
 
@@ -1021,14 +1001,15 @@ readmemwl_no_mmut(uint32_t addr, uint64_t *addr64)
 	if (!cpu_cyrix_alignment || (addr & 7) == 7)
 		cycles -= timing_misaligned;
 	if ((addr & 0xfff) > 0xffe) {
-		for (i = 0; i < 2; i++) {
-			if ((cr0 >> 31) && (addr64[i] > 0xffffffffULL))
-				return 0xffff;
-
-			ret |= (readmembl_no_mmut(addr + i, addr64[i]) << (i << 3));
+		if (cr0 >> 31) {
+			for (i = 0; i < 2; i++) {
+				if (addr64[i] > 0xffffffffULL)
+					return 0xffff;
+			}
 		}
 
-		return ret;
+		return readmembl_no_mmut(addr, addr64[0]) |
+		       (((uint16_t) readmembl_no_mmut(addr + 1, addr64[1])) << 8);
 	} else if (readlookup2[addr >> 12] != (uintptr_t) LOOKUP_INV)
 		return *(uint16_t *)(readlookup2[addr >> 12] + addr);
     }
@@ -1068,13 +1049,15 @@ writememwl_no_mmut(uint32_t addr, uint64_t *addr64, uint16_t val)
 	if (!cpu_cyrix_alignment || (addr & 7) == 7)
 		cycles -= timing_misaligned;
 	if ((addr & 0xfff) > 0xffe) {
-		for (i = 0; i < 2; i++) {
-			if ((cr0 >> 31) && (addr64[i] > 0xffffffffULL))
-				return;
-
-			writemembl_no_mmut(addr + i, addr64[i], val >> (i << 3));
+		if (cr0 >> 31) {
+			for (i = 0; i < 2; i++) {
+				if (addr64[i] > 0xffffffffULL)
+					return;
+			}
 		}
 
+		writemembl_no_mmut(addr, addr64[0], val);
+		writemembl_no_mmut(addr + 1, addr64[1], val >> 8);
 		return;
 	} else if (writelookup2[addr >> 12] != (uintptr_t) LOOKUP_INV) {
 		*(uint16_t *)(writelookup2[addr >> 12] + addr) = val;
@@ -1116,9 +1099,7 @@ readmemll(uint32_t addr)
 {
     uint64_t addr64[4];
     mem_mapping_t *map;
-    int i;
-    uint32_t ret = 0x00000000;
-    uint32_t prev_page = 0xffffffff;
+    int i, wrap_i = 1;
 
     for (i = 0; i < 4; i++)
 	addr64[i] = (uint64_t) (addr + i);
@@ -1130,26 +1111,31 @@ readmemll(uint32_t addr)
 		cycles -= timing_misaligned;
 	if ((addr & 0xfff) > 0xffc) {
 		if (cr0 >> 31) {
-			for (i = 0; i < 4; i++) {
-				/* If we are on the same page, there is no need to translate again, as we can just
-				   reuse the previous result. */
-				if ((i > 0) && (((addr + i) & ~0xfff) == prev_page))
-					addr64[i] = (addr64[i - 1] & ~0xfffLL) | ((uint64_t) ((addr + i) & 0xfff));
-				else
-					addr64[i] = mmutranslate_read(addr + i);
+			wrap_i = 4 - (addr & 0x3);
 
-				prev_page = ((addr + i) & ~0xfff);
+			addr64[0] = mmutranslate_read(addr);
 
-				if (addr64[i] > 0xffffffffULL)
-					return 0xffff;
+			if (addr64[0] > 0xffffffffULL)
+				return 0xffffffff;
+
+			for (i = 1; i < wrap_i; i++)
+				addr64[i] = (addr64[0] & ~0xfffLL) | ((uint64_t) ((addr + i) & 0xfff));
+
+			addr64[wrap_i] = mmutranslate_read(addr + wrap_i);
+
+			if (addr64[wrap_i] > 0xffffffffULL)
+				return 0xffffffff;
+
+			if (wrap_i != 3) {
+				for (i = (wrap_i) + 1; i <= 3; i++)
+					addr64[i] = (addr64[wrap_i] & ~0xfffLL) | ((uint64_t) ((addr + i) & 0xfff));
 			}
 		}
+
 		/* No need to waste precious CPU host cycles on mmutranslate's that were already done, just pass
 		   their result as a parameter to be used if needed. */
-		for (i = 0; i < 4; i += 2)
-			ret |= (readmemwl_no_mmut(addr + i, &(addr64[i])) << (i << 3));
-
-		return ret;
+		return readmemwl_no_mmut(addr, addr64) |
+		       (((uint32_t) readmemwl_no_mmut(addr + 2, &(addr64[2]))) << 16);
 	} else if (readlookup2[addr >> 12] != (uintptr_t) LOOKUP_INV)
 		return *(uint32_t *)(readlookup2[addr >> 12] + addr);
     }
@@ -1187,8 +1173,7 @@ writememll(uint32_t addr, uint32_t val)
 {
     uint64_t addr64[4];
     mem_mapping_t *map;
-    int i;
-    uint32_t prev_page = 0xffffffff;
+    int i, wrap_i = 1;
 
     for (i = 0; i < 4; i++)
 	addr64[i] = (uint64_t) (addr + i);
@@ -1200,30 +1185,37 @@ writememll(uint32_t addr, uint32_t val)
 		cycles -= timing_misaligned;
 	if ((addr & 0xfff) > 0xffc) {
 		if (cr0 >> 31) {
-			for (i = 0; i < 4; i++) {
-				/* Do not translate a page that has a valid lookup, as that is by definition valid
-				   and the whole purpose of the lookup is to avoid repeat identical translations. */
-				if (!page_lookup[(addr + i) >> 12] || !page_lookup[(addr + i) >> 12]->write_b) {
-					/* If we are on the same page, there is no need to translate again, as we can just
-					   reuse the previous result. */
-					if ((i > 0) && (((addr + i) & ~0xfff) == prev_page))
-						addr64[i] = (addr64[i - 1] & ~0xfffLL) | ((uint64_t) ((addr + i) & 0xfff));
-					else
-						addr64[i] = mmutranslate_write(addr + i);
+			wrap_i = 4 - (addr & 0x3);
 
-					prev_page = ((addr + i) & ~0xfff);
+			if (!page_lookup[(addr + 0) >> 12] || !page_lookup[(addr + 0) >> 12]->write_b)
+				addr64[0] = mmutranslate_write(addr);
+			else
+				addr64[0] = 0xffffffff;
 
-					if (addr64[i] > 0xffffffffULL)
-						return;
-				} else
-					prev_page = 0xffffffff;
+			if (addr64[0] > 0xffffffffULL)
+				return;
+
+			for (i = 1; i < wrap_i; i++)
+				addr64[i] = (addr64[0] & ~0xfffLL) | ((uint64_t) ((addr + i) & 0xfff));
+
+			if (!page_lookup[(addr + wrap_i) >> 12] || !page_lookup[(addr + wrap_i) >> 12]->write_b)
+				addr64[wrap_i] = mmutranslate_write(addr + wrap_i);
+			else
+				addr64[wrap_i] = 0xffffffff;
+
+			if (addr64[wrap_i] > 0xffffffffULL)
+				return;
+
+			if (wrap_i != 3) {
+				for (i = (wrap_i) + 1; i <= 3; i++)
+					addr64[i] = (addr64[wrap_i] & ~0xfffLL) | ((uint64_t) ((addr + i) & 0xfff));
 			}
 		}
+
 		/* No need to waste precious CPU host cycles on mmutranslate's that were already done, just pass
 		   their result as a parameter to be used if needed. */
-		for (i = 0; i < 4; i += 2)
-			writememwl_no_mmut(addr + i, &(addr64[i]), val >> (i << 3));
-
+		writememwl_no_mmut(addr, &(addr64[0]), val);
+		writememwl_no_mmut(addr + 2, &(addr64[2]), val >> 16);
 		return;
 	} else if (writelookup2[addr >> 12] != (uintptr_t) LOOKUP_INV) {
 		*(uint32_t *)(writelookup2[addr >> 12] + addr) = val;
@@ -1272,7 +1264,6 @@ readmemll_no_mmut(uint32_t addr, uint64_t *addr64)
 {
     mem_mapping_t *map;
     int i;
-    uint32_t ret = 0x00000000;
 
     mem_logical_addr = addr;
 
@@ -1280,14 +1271,15 @@ readmemll_no_mmut(uint32_t addr, uint64_t *addr64)
 	if (!cpu_cyrix_alignment || (addr & 7) > 4)
 		cycles -= timing_misaligned;
 	if ((addr & 0xfff) > 0xffc) {
-		for (i = 0; i < 4; i += 2) {
-			if ((cr0 >> 31) && (addr64[i] > 0xffffffffULL))
-				return 0xffffffff;
-
-			ret |= (readmemwl_no_mmut(addr + i, &(addr64[i])) << (i << 3));
+		if (cr0 >> 31) {
+			for (i = 0; i < 4; i += 2) {
+				if (addr64[i] > 0xffffffffULL)
+					return 0xffffffff;
+			}
 		}
 
-		return ret;
+		return readmemwl_no_mmut(addr, addr64) |
+		       (((uint32_t) readmemwl_no_mmut(addr + 2, &(addr64[2]))) << 16);
 	} else if (readlookup2[addr >> 12] != (uintptr_t) LOOKUP_INV)
 		return *(uint32_t *)(readlookup2[addr >> 12] + addr);
     }
@@ -1332,13 +1324,15 @@ writememll_no_mmut(uint32_t addr, uint64_t *addr64, uint32_t val)
 	if (!cpu_cyrix_alignment || (addr & 7) > 4)
 		cycles -= timing_misaligned;
 	if ((addr & 0xfff) > 0xffc) {
-		for (i = 0; i < 4; i += 2) {
-			if ((cr0 >> 31) && (addr64[i] > 0xffffffffULL))
-				return;
-
-			writememwl_no_mmut(addr + i, &(addr64[i]), val >> (i << 3));
+		if (cr0 >> 31) {
+			for (i = 0; i < 4; i += 2) {
+				if (addr64[i] > 0xffffffffULL)
+					return;
+			}
 		}
 
+		writememwl_no_mmut(addr, &(addr64[0]), val);
+		writememwl_no_mmut(addr + 2, &(addr64[2]), val >> 16);
 		return;
 	} else if (writelookup2[addr >> 12] != (uintptr_t) LOOKUP_INV) {
 		*(uint32_t *)(writelookup2[addr >> 12] + addr) = val;
@@ -1385,9 +1379,7 @@ readmemql(uint32_t addr)
 {
     uint64_t addr64[8];
     mem_mapping_t *map;
-    int i;
-    uint64_t ret = 0x0000000000000000ULL;
-    uint32_t prev_page = 0xffffffff;
+    int i, wrap_i = 1;
 
     for (i = 0; i < 8; i++)
 	addr64[i] = (uint64_t) (addr + i);
@@ -1398,26 +1390,30 @@ readmemql(uint32_t addr)
 	cycles -= timing_misaligned;
 	if ((addr & 0xfff) > 0xff8) {
 		if (cr0 >> 31) {
-			for (i = 0; i < 8; i++) {
-				/* If we are on the same page, there is no need to translate again, as we can just
-				   reuse the previous result. */
-				if ((i > 0) && (((addr + i) & ~0xfff) == prev_page))
-					addr64[i] = (addr64[i - 1] & ~0xfffLL) | ((uint64_t) ((addr + i) & 0xfff));
-				else
-					addr64[i] = mmutranslate_read(addr + i);
+			wrap_i = 8 - (addr & 0x7);
 
-				prev_page = ((addr + i) & ~0xfff);
+			addr64[0] = mmutranslate_read(addr);
 
-				if (addr64[i] > 0xffffffffULL)
-					return 0xffff;
+			if (addr64[0] > 0xffffffffULL)
+				return 0xffffffffffffffffULL;
+
+			for (i = 1; i < wrap_i; i++)
+				addr64[i] = (addr64[0] & ~0xfffLL) | ((uint64_t) ((addr + i) & 0xfff));
+
+			addr64[wrap_i] = mmutranslate_read(addr + wrap_i);
+
+			if (addr64[wrap_i] > 0xffffffffULL)
+				return 0xffffffffffffffffULL;
+
+			if (wrap_i != 7) {
+				for (i = (wrap_i) + 1; i <= 7; i++)
+					addr64[i] = (addr64[wrap_i] & ~0xfffLL) | ((uint64_t) ((addr + i) & 0xfff));
 			}
 		}
 		/* No need to waste precious CPU host cycles on mmutranslate's that were already done, just pass
 		   their result as a parameter to be used if needed. */
-		for (i = 0; i < 8; i += 4)
-			ret |= (readmemll_no_mmut(addr + i, &(addr64[i])) << (i << 3));
-
-		return ret;
+		return readmemll_no_mmut(addr, addr64) |
+		       (((uint64_t) readmemll_no_mmut(addr + 4, &(addr64[4]))) << 32);
 	} else if (readlookup2[addr >> 12] != (uintptr_t) LOOKUP_INV)
 		return *(uint64_t *)(readlookup2[addr >> 12] + addr);
     }
@@ -1444,8 +1440,7 @@ writememql(uint32_t addr, uint64_t val)
 {
     uint64_t addr64[8];
     mem_mapping_t *map;
-    int i;
-    uint32_t prev_page = 0xffffffff;
+    int i, wrap_i;
 
     for (i = 0; i < 8; i++)
 	addr64[i] = (uint64_t) (addr + i);
@@ -1456,30 +1451,37 @@ writememql(uint32_t addr, uint64_t val)
 	cycles -= timing_misaligned;
 	if ((addr & 0xfff) > 0xff8) {
 		if (cr0 >> 31) {
-			for (i = 0; i < 8; i++) {
-				/* Do not translate a page that has a valid lookup, as that is by definition valid
-				   and the whole purpose of the lookup is to avoid repeat identical translations. */
-				if (!page_lookup[(addr + i) >> 12] || !page_lookup[(addr + i) >> 12]->write_b) {
-					/* If we are on the same page, there is no need to translate again, as we can just
-					   reuse the previous result. */
-					if ((i > 0) && (((addr + i) & ~0xfff) == prev_page))
-						addr64[i] = (addr64[i - 1] & ~0xfffLL) | ((uint64_t) ((addr + i) & 0xfff));
-					else
-						addr64[i] = mmutranslate_write(addr + i);
+			wrap_i = 8 - (addr & 0x7);
 
-					prev_page = ((addr + i) & ~0xfff);
+			if (!page_lookup[(addr + 0) >> 12] || !page_lookup[(addr + 0) >> 12]->write_b)
+				addr64[0] = mmutranslate_write(addr);
+			else
+				addr64[0] = 0xffffffff;
 
-					if (addr64[i] > 0xffffffffULL)
-						return;
-				} else
-					prev_page = 0xffffffff;
+			if (addr64[0] > 0xffffffffULL)
+				return;
+
+			for (i = 1; i < wrap_i; i++)
+				addr64[i] = (addr64[wrap_i] & ~0xfffLL) | ((uint64_t) ((addr + i) & 0xfff));
+
+			if (!page_lookup[(addr + wrap_i) >> 12] || !page_lookup[(addr + wrap_i) >> 12]->write_b)
+				addr64[wrap_i] = mmutranslate_write(addr + wrap_i);
+			else
+				addr64[wrap_i] = 0xffffffff;
+
+			if (addr64[wrap_i] > 0xffffffffULL)
+				return;
+
+			if (wrap_i != 7) {
+				for (i = (wrap_i) + 1; i <= 7; i++)
+					addr64[i] = (addr64[wrap_i] & ~0xfffLL) | ((uint64_t) ((addr + i) & 0xfff));
 			}
 		}
+
 		/* No need to waste precious CPU host cycles on mmutranslate's that were already done, just pass
 		   their result as a parameter to be used if needed. */
-		for (i = 0; i < 8; i += 4)
-			writememll_no_mmut(addr + i, &(addr64[i]), val >> (i << 3));
-
+		writememll_no_mmut(addr, &(addr64[0]), val);
+		writememll_no_mmut(addr + 4, &(addr64[4]), val >> 32);
 		return;
 	} else if (writelookup2[addr >> 12] != (uintptr_t) LOOKUP_INV) {
 		*(uint64_t *)(writelookup2[addr >> 12] + addr) = val;
@@ -1533,36 +1535,29 @@ writememql(uint32_t addr, uint64_t val)
 void
 do_mmutranslate(uint32_t addr, uint64_t *addr64, int num, int write)
 {
-    int i, cond = 1;
-    uint32_t prev_page = 0xffffffff;
+    int i, wrap_i;
+    int cond = 1;
 
-    for (i = 0; i < num; i++) {
-	addr64[i] = (uint64_t) (addr + i);
+    if (cr0 >> 31) {
+	wrap_i = (num - (addr & (num - 1))) & (num - 1);
 
-    	if (cr0 >> 31) {
-		/* Do not translate a page that has a valid lookup, as that is by definition valid
-		   and the whole purpose of the lookup is to avoid repeat identical translations. */
-		if (write)
-		    cond = (!page_lookup[(addr + i) >> 12] || !page_lookup[(addr + i) >> 12]->write_b);
+	for (i = 0; i < num; i++) {
+		if (write && ((i == 0) || (wrap_i && (i == wrap_i))))
+		    cond = (!page_lookup[addr >> 12] || !page_lookup[addr >> 12]->write_b);
 
 		if (cond) {
 			/* If we are on the same page, there is no need to translate again, as we can just
 			   reuse the previous result. */
-			if ((i > 0) && (((addr + i) & ~0xfff) == prev_page))
-				addr64[i] = (addr64[i - 1] & ~0xfffLL) | ((uint64_t) ((addr + i) & 0xfff));
-			else
-				addr64[i] = mmutranslatereal(addr + i, write);
+			if (((i == 0) || (wrap_i && (i == wrap_i)))) {
+				addr64[i] = mmutranslatereal(addr, write);
 
-			prev_page = ((addr + i) & ~0xfff);
+				if (addr64[i] > 0xffffffffULL)
+					return;
+			} else
+				addr64[i] = (addr64[i - 1] & ~0xfffLL) | ((uint64_t) (addr & 0xfff));
+		}
 
-			if (addr64[i] == 0xffffffffffffffffULL)
-				return;
-			if (addr64[i] > 0xffffffffULL)
-				return;
-			if (cpu_state.abrt)
-				return;
-		} else
-			prev_page = 0xffffffff;
+		addr++;
 	}
     }
 }
diff --git a/src/sound/snd_sb.c b/src/sound/snd_sb.c
index c68c062ab..2dc147285 100644
--- a/src/sound/snd_sb.c
+++ b/src/sound/snd_sb.c
@@ -858,10 +858,7 @@ sb_ct1745_mixer_write(uint16_t addr, uint8_t val, void *p)
 
 		case 0x83:
 			/* Interrupt mask. */
-			sb->dsp.sb_irqm8 = !(val & 0x01);
-			sb->dsp.sb_irqm16 = !(val & 0x02);
-			sb->dsp.sb_irqm401 = !(val & 0x04);
-			sb_update_irq(&sb->dsp);
+			sb_update_mask(&sb->dsp, !(val & 0x01), !(val & 0x02), !(val & 0x04));
 			break;
 
 		case 0x84:
diff --git a/src/sound/snd_sb_dsp.c b/src/sound/snd_sb_dsp.c
index 4c0bcf6ab..6203def85 100644
--- a/src/sound/snd_sb_dsp.c
+++ b/src/sound/snd_sb_dsp.c
@@ -178,17 +178,21 @@ recalc_sb16_filter(int c, int playback_freq)
 
 
 void
-sb_update_irq(sb_dsp_t *dsp)
+sb_update_mask(sb_dsp_t *dsp, int irqm8, int irqm16, int irqm401)
 {
-    int irq_pending;
+    int clear = 0;
 
-    irq_pending = (dsp->sb_irq8 && !dsp->sb_irqm8) ||
-		  (dsp->sb_irq16 && !dsp->sb_irqm16) ||
-		  (dsp->sb_irq401 && !dsp->sb_irqm401);
+    if (!dsp->sb_irqm8 && irqm8)
+	clear |= 1;
+    dsp->sb_irqm8 = irqm8;
+    if (!dsp->sb_irqm16 && irqm16)
+	clear |= 1;
+    dsp->sb_irqm16 = irqm16;
+    if (!dsp->sb_irqm401 && irqm401)
+	clear |= 1;
+    dsp->sb_irqm401 = irqm401;
 
-    if (irq_pending)
-	picint(1 << dsp->sb_irqnum);
-    else
+    if (clear)
 	picintc(1 << dsp->sb_irqnum);
 }
 
@@ -196,20 +200,28 @@ sb_update_irq(sb_dsp_t *dsp)
 void
 sb_update_status(sb_dsp_t *dsp, int bit, int set)
 {
+    int masked = 0;
+
     switch (bit) {
 	case 0:
 	default:
 		dsp->sb_irq8  = set;
+		masked = dsp->sb_irqm8;
 		break;
 	case 1:
 		dsp->sb_irq16  = set;
+		masked = dsp->sb_irqm16;
 		break;
 	case 2:
 		dsp->sb_irq401 = set;
+		masked = dsp->sb_irqm401;
 		break;
     }
 
-    sb_update_irq(dsp);
+    if (set && !masked)
+	picint(1 << dsp->sb_irqnum);
+    else if (!set)
+	picintc(1 << dsp->sb_irqnum);
 }
 
 
@@ -258,8 +270,8 @@ sb_dsp_set_mpu(sb_dsp_t *dsp, mpu_t *mpu)
 void
 sb_dsp_reset(sb_dsp_t *dsp)
 {
-	midi_clear_buffer();
-	
+    midi_clear_buffer();
+
     timer_disable(&dsp->output_timer);
     timer_disable(&dsp->input_timer);
 
@@ -271,9 +283,8 @@ sb_dsp_reset(sb_dsp_t *dsp)
     dsp->sb_irq8 = 0;
     dsp->sb_irq16 = 0;
     dsp->sb_irq401 = 0;
-    sb_update_irq(dsp);
     dsp->sb_16_pause = 0;
-	dsp->sb_read_wp = dsp->sb_read_rp = 0;
+    dsp->sb_read_wp = dsp->sb_read_rp = 0;
     dsp->sb_data_stat = -1;
     dsp->sb_speaker = 0;
     dsp->sb_pausetime = -1LL;