BSNESv115+: apply patch to fix $4203 mul/div behavior

see bsnes-emu/bsnes#256 or https://forums.nesdev.org/viewtopic.php?t=24087 for context
TASEmulators · Dec 11, 2022 · 837592c · 837592c
1 parent 74b37ab
commit 837592c
Show file tree

Hide file tree

Showing 4 changed files with 20 additions and 6 deletions.
diff --git a/Assets/dll/bsnes.wbx.zst b/Assets/dll/bsnes.wbx.zst
diff --git a/waterbox/bsnescore/bsnes/sfc/cpu/cpu.hpp b/waterbox/bsnescore/bsnes/sfc/cpu/cpu.hpp
@@ -165,7 +165,9 @@ struct CPU : Processor::WDC65816, Thread, PPUcounter {
 
   struct ALU {
     uint mpyctr = 0;
+    uint mpylast = 0;
     uint divctr = 0;
+    uint divlast = 0;
     uint shift = 0;
   } alu;
 

diff --git a/waterbox/bsnescore/bsnes/sfc/cpu/io.cpp b/waterbox/bsnescore/bsnes/sfc/cpu/io.cpp
@@ -159,8 +159,10 @@ auto CPU::writeCPU(uint addr, uint8 data) -> void {
     io.rddiv = io.wrmpyb << 8 | io.wrmpya;
 
     if(!configuration.hacks.cpu.fastMath) {
-      alu.mpyctr = 8;  //perform multiplication over the next eight cycles
-      alu.shift = io.wrmpyb;
+      if (!alu.mpylast) {
+        alu.mpyctr = 8;  //perform multiplication over the next eight cycles
+        alu.shift = io.wrmpyb;
+      }
     } else {
       io.rdmpy = io.wrmpya * io.wrmpyb;
     }
@@ -178,12 +180,14 @@ auto CPU::writeCPU(uint addr, uint8 data) -> void {
     io.rdmpy = io.wrdiva;
     if(alu.mpyctr || alu.divctr) return;
 
-    io.wrdivb = data;
-
     if(!configuration.hacks.cpu.fastMath) {
-      alu.divctr = 16;  //perform division over the next sixteen cycles
-      alu.shift = io.wrdivb << 16;
+      if (!alu.divlast) {
+        io.wrdivb = data;
+        alu.divctr = 16;  //perform division over the next sixteen cycles
+        alu.shift = io.wrdivb << 16;
+      }
     } else {
+      io.wrdivb = data;
       if(io.wrdivb) {
         io.rddiv = io.wrdiva / io.wrdivb;
         io.rdmpy = io.wrdiva % io.wrdivb;

diff --git a/waterbox/bsnescore/bsnes/sfc/cpu/timing.cpp b/waterbox/bsnescore/bsnes/sfc/cpu/timing.cpp
@@ -143,20 +143,28 @@ auto CPU::scanline() -> void {
 auto CPU::aluEdge() -> void {
   if(alu.mpyctr) {
     alu.mpyctr--;
+    if (!alu.mpyctr)
+      alu.mpylast = 1;
     if(io.rddiv & 1) io.rdmpy += alu.shift;
     io.rddiv >>= 1;
     alu.shift <<= 1;
   }
+  else
+    alu.mpylast = 0;
 
   if(alu.divctr) {
     alu.divctr--;
+    if (!alu.divctr)
+      alu.divlast = 1;
     io.rddiv <<= 1;
     alu.shift >>= 1;
     if(io.rdmpy >= alu.shift) {
       io.rdmpy -= alu.shift;
       io.rddiv |= 1;
     }
   }
+  else
+    alu.divlast = 0;
 }
 
 auto CPU::dmaEdge() -> void {