X-Git-Url: https://notaz.gp2x.de/cgi-bin/gitweb.cgi?a=blobdiff_plain;ds=sidebyside;f=Cyclone%2FOpArith.cpp;fp=Cyclone%2FOpArith.cpp;h=d762049a55344f5892a71bfda33e85ed99a3c661;hb=d9d77995ec88700f438b3638df179a014bf4f6b3;hp=c8144f4113bb3a5a211ed9d69a7406591378f41a;hpb=7441f043a0ba83bd919db25bf3b8e35ae635012a;p=cyclone68000.git

diff --git a/Cyclone/OpArith.cpp b/Cyclone/OpArith.cpp
index c8144f4..d762049 100644
--- a/Cyclone/OpArith.cpp
+++ b/Cyclone/OpArith.cpp
@@ -1,22 +1,25 @@
 
 // This file is part of the Cyclone 68000 Emulator
 
-// Copyright (c) 2011 FinalDave (emudave (at) gmail.com)
+// Copyright (c) 2004,2011 FinalDave (emudave (at) gmail.com)
+// Copyright (c) 2005-2011 Gražvydas "notaz" Ignotas (notasas (at) gmail.com)
 
 // This code is licensed under the GNU General Public License version 2.0 and the MAME License.
 // You can choose the license that has the most advantages for you.
 
 // SVN repository can be found at http://code.google.com/p/cyclone68000/
 
+
 #include "app.h"
 
 // --------------------- Opcodes 0x0000+ ---------------------
-// Emit an Ori/And/Sub/Add/Eor/Cmp Immediate opcode, 0000ttt0 00aaaaaa
+// Emit an Ori/And/Sub/Add/Eor/Cmp Immediate opcode, 0000ttt0 ssaaaaaa
 int OpArith(int op)
 {
   int type=0,size=0;
   int sea=0,tea=0;
   int use=0;
+  const char *shiftstr="";
 
   // Get source and target EA
   type=(op>>9)&7; if (type==4 || type>=7) return 1;
@@ -26,27 +29,28 @@ int OpArith(int op)
 
   // See if we can do this opcode:
   if (EaCanRead(tea,size)==0) return 1;
-  if (type!=6 && EaCanWrite(tea)==0) return 1;
+  if (EaCanWrite(tea)==0 || EaAn(tea)) return 1;
 
-  use=OpBase(op);
+  use=OpBase(op,size);
   if (op!=use) { OpUse(op,use); return 0; } // Use existing handler
 
-  OpStart(op); Cycles=4;
+  OpStart(op, sea, tea); Cycles=4;
 
-  EaCalc(10,0x0000, sea,size);
-  EaRead(10,    10, sea,size,1);
+  // imm must be read first
+  EaCalcReadNoSE(-1,10,sea,size,0);
+  EaCalcReadNoSE((type!=6)?11:-1,0,tea,size,0x003f);
 
-  EaCalc(11,0x003f, tea,size);
-  EaRead(11,     0, tea,size,1);
+  if (size<2) shiftstr=(char *)(size?",asl #16":",asl #24");
+  if (size<2) ot("  mov r10,r10,asl #%i\n",size?16:24);
 
   ot(";@ Do arithmetic:\n");
 
-  if (type==0) ot("  orr r1,r0,r10\n");
-  if (type==1) ot("  and r1,r0,r10\n");
-  if (type==2) ot("  subs r1,r0,r10 ;@ Defines NZCV\n");
-  if (type==3) ot("  adds r1,r0,r10 ;@ Defines NZCV\n");
-  if (type==5) ot("  eor r1,r0,r10\n");
-  if (type==6) ot("  cmp r0,r10 ;@ Defines NZCV\n");
+  if (type==0) ot("  orr r1,r10,r0%s\n",shiftstr);
+  if (type==1) ot("  and r1,r10,r0%s\n",shiftstr);
+  if (type==2||type==6)
+               ot("  rsbs r1,r10,r0%s ;@ Defines NZCV\n",shiftstr);
+  if (type==3) ot("  adds r1,r10,r0%s ;@ Defines NZCV\n",shiftstr);
+  if (type==5) ot("  eor r1,r10,r0%s\n",shiftstr);
 
   if (type<2 || type==5) ot("  adds r1,r1,#0 ;@ Defines NZ, clears CV\n"); // 0,1,5
 
@@ -59,7 +63,7 @@ int OpArith(int op)
 
   if (type!=6)
   {
-    EaWrite(11, 1, tea,size,1);
+    EaWrite(11, 1, tea,size,0x003f,1);
   }
 
   // Correct cycles:
@@ -70,11 +74,11 @@ int OpArith(int op)
   else
   {
     if (size>=2) Cycles+=4;
-    if (tea>=0x10) Cycles+=4;
-    if (Amatch && type==1 && size>=2 && tea<0x10) Cycles-=2;
+    if (tea>=8)  Cycles+=4;
+    if (type==1 && size>=2 && tea<8) Cycles-=2;
   }
 
-  OpEnd();
+  OpEnd(sea,tea);
 
   return 0;
 }
@@ -95,20 +99,22 @@ int OpAddq(int op)
 
   // See if we can do this opcode:
   if (EaCanRead (ea,size)==0) return 1;
-  if (EaCanWrite(ea     )==0) return 1;
+  if (EaCanWrite(ea)     ==0) return 1;
+  if (size == 0 && EaAn(ea) ) return 1;
 
-  use=op; if (ea<0x38) use&=~7;
-  if ((ea&0x38)==0x08) { size=2; use&=~0xc0; } // Every addq #n,An is 32-bit
+  use=OpBase(op,size,1);
 
   if (num!=8) use|=0x0e00; // If num is not 8, use same handler
   if (op!=use) { OpUse(op,use); return 0; } // Use existing handler
 
-  OpStart(op);
+  OpStart(op,ea);
   Cycles=ea<8?4:8;
-  if (size>=2 && ea!=8) Cycles+=4;
+  if(type==0&&size==1) Cycles=ea<0x10?4:8;
+  if(size>=2) Cycles=ea<0x10?8:12;
+
+  if (size>0 && (ea&0x38)==0x08) size=2; // addq.w #n,An is also 32-bit
 
-  EaCalc(10,0x003f, ea,size);
-  EaRead(10,     0, ea,size,1);
+  EaCalcReadNoSE(11,0,ea,size,0x003f);
 
   shift=32-(8<<size);
 
@@ -116,15 +122,19 @@ int OpAddq(int op)
   {
     int lsr=9-shift;
 
-    if (lsr>=0) ot("  mov r2,r8,lsr #%d ;@ Get quick value\n", lsr);
-    else        ot("  mov r2,r8,lsl #%d ;@ Get quick value\n",-lsr);
+    ot("  and r2,r8,#0x0e00 ;@ Get quick value\n");
+
+    if (lsr>=0) sprintf(count,"r2,lsr #%d",  lsr);
+    else        sprintf(count,"r2,lsl #%d", -lsr);
 
-    ot("  and r2,r2,#0x%.4x\n",7<<shift);
     ot("\n");
-    strcpy(count,"r2");
+  }
+  else
+  {
+    sprintf(count,"#0x%.4x",8<<shift);
   }
 
-  if (num==8) sprintf(count,"#0x%.4x",8<<shift);
+  if (size<2)  ot("  mov r0,r0,asl #%d\n\n",size?16:24);
 
   if (type==0) ot("  adds r1,r0,%s\n",count);
   if (type==1) ot("  subs r1,r0,%s\n",count);
@@ -132,9 +142,9 @@ int OpAddq(int op)
   if ((ea&0x38)!=0x08) OpGetFlags(type,1);
   ot("\n");
 
-  EaWrite(10,     1, ea,size,1);
+  EaWrite(11,     1, ea,size,0x003f,1);
 
-  OpEnd();
+  OpEnd(ea);
 
   return 0;
 }
@@ -145,39 +155,43 @@ int OpArithReg(int op)
 {
   int use=0;
   int type=0,size=0,dir=0,rea=0,ea=0;
+  const char *asl="";
+  const char *strop=0;
 
   type=(op>>12)&5;
   rea =(op>> 9)&7;
-  dir =(op>> 8)&1;
+  dir =(op>> 8)&1; // er,re
   size=(op>> 6)&3; if (size>=3) return 1;
   ea  = op&0x3f;
 
   if (dir && ea<0x10) return 1; // addx/subx opcode
 
   // See if we can do this opcode:
-  if (dir==0 && EaCanWrite(rea)==0) return 1;
-  if (dir    && EaCanWrite( ea)==0) return 1;
+  if (dir==0 && EaCanRead (ea,size)==0) return 1;
+  if (dir    && EaCanWrite(ea)==0)      return 1;
+  if ((size==0||!(type&1))&&EaAn(ea))   return 1;
 
-  use=OpBase(op);
+  use=OpBase(op,size);
   use&=~0x0e00; // Use same opcode for Dn
   if (op!=use) { OpUse(op,use); return 0; } // Use existing handler
 
-  OpStart(op); Cycles=4;
+  OpStart(op,ea); Cycles=4;
 
-  ot(";@ Get r10=EA r11=EA value\n");
-  EaCalc(10,0x003f, ea,size);
-  EaRead(10,    11, ea,size,1);
-  ot(";@ Get r0=Register r1=Register value\n");
-  EaCalc( 0,0x0e00,rea,size);
-  EaRead( 0,     1,rea,size,1);
+  EaCalcReadNoSE(dir?11:-1,0,ea,size,0x003f);
+
+  EaCalcReadNoSE(dir?-1:11,1,rea,size,0x0e00);
 
   ot(";@ Do arithmetic:\n");
-  if (type==0) ot("  orr  ");
-  if (type==1) ot("  subs ");
-  if (type==4) ot("  and  ");
-  if (type==5) ot("  adds ");
-  if (dir) ot("r1,r11,r1\n");
-  else     ot("r1,r1,r11\n");
+  if (type==0) strop = "orr";
+  if (type==1) strop = (char *) (dir ? "subs" : "rsbs");
+  if (type==4) strop = "and";
+  if (type==5) strop = "adds";
+
+  if (size==0) asl=",asl #24";
+  if (size==1) asl=",asl #16";
+
+  if (size<2) ot("  mov r0,r0%s\n",asl);
+  ot("  %s r1,r0,r1%s\n",strop,asl);
 
   if ((type&1)==0) ot("  adds r1,r1,#0 ;@ Defines NZ, clears CV\n");
 
@@ -185,13 +199,23 @@ int OpArithReg(int op)
   ot("\n");
 
   ot(";@ Save result:\n");
-  if (dir) EaWrite(10, 1, ea,size,1);
-  else     EaWrite( 0, 1,rea,size,1);
-
-  if (size==1 && ea>=0x10) Cycles+=4;
-  if (size>=2) { if (ea<0x10) Cycles+=4; else Cycles+=2; }
+  if (size<2) ot("  mov r1,r1,asr #%d\n",size?16:24);
+  if (dir) EaWrite(11, 1, ea,size,0x003f,0,0);
+  else     EaWrite(11, 1,rea,size,0x0e00,0,0);
+
+  if(rea==ea) {
+    if(ea<8) Cycles=(size>=2)?8:4; else Cycles+=(size>=2)?26:14;
+  } else if(dir) {
+    Cycles+=4;
+    if(size>=2) Cycles+=4;
+  } else {
+    if(size>=2) {
+      Cycles+=2;
+      if(ea<0x10||ea==0x3c) Cycles+=2;
+    }
+  }
 
-  OpEnd();
+  OpEnd(ea);
 
   return 0;
 }
@@ -209,41 +233,55 @@ int OpMul(int op)
   ea  = op&0x3f;
 
   // See if we can do this opcode:
-  if (EaCanRead(ea,1)==0) return 1;
+  if (EaCanRead(ea,1)==0||EaAn(ea)) return 1;
 
-  use=OpBase(op);
+  use=OpBase(op,1);
   use&=~0x0e00; // Use same for all registers
   if (op!=use) { OpUse(op,use); return 0; } // Use existing handler
 
-  OpStart(op); Cycles=type?70:133;
+  OpStart(op,ea);
+  if(type) Cycles=54;
+  else     Cycles=sign?158:140;
 
-  EaCalc(10,0x003f, ea, 1);
-  EaRead(10,    10, ea, 1);
+  EaCalcReadNoSE(-1,0,ea,1,0x003f);
 
-  EaCalc (0,0x0e00,rea, 2);
-  EaRead (0,     2,rea, 2);
+  EaCalc(11,0x0e00,rea, 2);
+  EaRead(11,     2,rea, 2,0x0e00);
 
-  if (type==0)
+  ot("  movs r1,r0,asl #16\n");
+
+  if (type==0) // div
   {
-    ot("  cmp r10,#0\n");
-    ot("  moveq r10,#1 ;@ Divide by zero\n");
+    // the manual says C is always cleared, but neither Musashi nor FAME do that
+    //ot("  bic r10,r10,#0x20000000 ;@ always clear C\n");
+    ot("  beq divzero%.4x ;@ division by zero\n",op);
     ot("\n");
     
     if (sign)
     {
-      ot("  mov r11,#0 ;@ r11 = 1 if the result is negative\n");
-      ot("  eorlt r11,r11,#1\n");
-      ot("  rsblt r10,r10,#0 ;@ Make r10 positive\n");
+      ot("  mov r12,#0 ;@ r12 = 1 or 2 if the result is negative\n");
+      ot("  tst r2,r2\n");
+      ot("  orrmi r12,r12,#2\n");
+      ot("  rsbmi r2,r2,#0 ;@ Make r2 positive\n");
       ot("\n");
-      ot("  cmp r2,#0\n");
-      ot("  eorlt r11,r11,#1\n");
-      ot("  rsblt r2,r2,#0 ;@ Make r2 positive\n");
+      ot("  movs r0,r1,asr #16\n");
+      ot("  orrmi r12,r12,#1\n");
+      ot("  rsbmi r0,r0,#0 ;@ Make r0 positive\n");
       ot("\n");
+      ot(";@ detect the nasty 0x80000000 / -1 situation\n");
+      ot("  mov r3,r2,asr #31\n");
+      ot("  eors r3,r3,r1,asr #16\n");
+      ot("  beq wrendofop%.4x\n",op);
+    }
+    else
+    {
+      ot("  mov r0,r1,lsr #16 ;@ use only 16 bits of divisor\n");
     }
 
-    ot(";@ Divide r2 by r10\n");
+    ot("\n");
+    ot(";@ Divide r2 by r0\n");
     ot("  mov r3,#0\n");
-    ot("  mov r1,r10\n");
+    ot("  mov r1,r0\n");
     ot("\n");
     ot(";@ Shift up divisor till it's just less than numerator\n");
     ot("Shift%.4x%s\n",op,ms?"":":");
@@ -256,62 +294,85 @@ int OpMul(int op)
     ot("  cmp r2,r1\n");
     ot("  adc r3,r3,r3 ;@ Double r3 and add 1 if carry set\n");
     ot("  subcs r2,r2,r1\n");
-    ot("  teq r1,r10\n");
+    ot("  teq r1,r0\n");
     ot("  movne r1,r1,lsr #1\n");
     ot("  bne Divide%.4x\n",op);
     ot("\n");
+    ot(";@r3==quotient,r2==remainder\n");
 
     if (sign)
     {
-      ot("  tst r11,r11\n");
-      ot("  rsbne r3,r3,#0 ;@ Negate if result is negative\n");
-    }
+      // sign correction
+      ot("  and r1,r12,#1\n");
+      ot("  teq r1,r12,lsr #1\n");
+      ot("  rsbne r3,r3,#0 ;@ negate if quotient is negative\n");
+      ot("  tst r12,#2\n");
+      ot("  rsbne r2,r2,#0 ;@ negate the remainder if divident was negative\n");
+      ot("\n");
 
-    ot("  mov r11,r2 ;@ Remainder\n");
+      // signed overflow check
+      ot("  mov r1,r3,asl #16\n");
+      ot("  cmp r3,r1,asr #16 ;@ signed overflow?\n");
+      ot("  orrne r10,r10,#0x10000000 ;@ set overflow flag\n");
+      ot("  bne endofop%.4x ;@ overflow!\n",op);
+      ot("\n");
+      ot("wrendofop%.4x%s\n",op,ms?"":":");
+    }
+    else
+    {
+      // overflow check
+      ot("  movs r1,r3,lsr #16 ;@ check for overflow condition\n");
+      ot("  orrne r10,r10,#0x10000000 ;@ set overflow flag\n");
+      ot("  bne endofop%.4x ;@ overflow!\n",op);
+      ot("\n");
+    }
 
-    ot("  adds r1,r3,#0 ;@ Defines NZ, clears CV\n");
+    ot("  mov r1,r3,lsl #16 ;@ Clip to 16-bits\n");
+    ot("  adds r1,r1,#0 ;@ Defines NZ, clears CV\n");
     OpGetFlags(0,0);
 
-    ot("  mov r1,r1,lsl #16 ;@ Clip to 16-bits\n");
     ot("  mov r1,r1,lsr #16\n");
-    ot("  orr r1,r1,r11,lsl #16 ;@ Insert remainder\n");
+    ot("  orr r1,r1,r2,lsl #16 ;@ Insert remainder\n");
   }
 
   if (type==1)
   {
-    char *shift="asr";
-
     ot(";@ Get 16-bit signs right:\n");
-    if (sign==0) { ot("  mov r10,r10,lsl #16\n"); shift="lsr"; }
+    ot("  mov r0,r1,%s #16\n",sign?"asr":"lsr");
     ot("  mov r2,r2,lsl #16\n");
-
-    if (sign==0) ot("  mov r10,r10,lsr #16\n");
-    ot("  mov r2,r2,%s #16\n",shift);
+    ot("  mov r2,r2,%s #16\n",sign?"asr":"lsr");
     ot("\n");
 
-    ot("  mul r1,r2,r10\n");
+    ot("  mul r1,r2,r0\n");
     ot("  adds r1,r1,#0 ;@ Defines NZ, clears CV\n");
     OpGetFlags(0,0);
-
-    if (Amatch && ea==0x3c) Cycles-=4;
   }
   ot("\n");
 
-  EaWrite(0,     1,rea, 2);
+  EaWrite(11,    1,rea, 2,0x0e00,1);
 
+  if (type==0) ot("endofop%.4x%s\n",op,ms?"":":");
+  OpEnd(ea);
 
-  OpEnd();
+  if (type==0) // div
+  {
+    ot("divzero%.4x%s\n",op,ms?"":":");
+    ot("  mov r0,#5 ;@ Divide by zero\n");
+    ot("  bl Exception\n");
+    Cycles+=38;
+    OpEnd(ea);
+    ot("\n");
+  }
 
   return 0;
 }
 
 // Get X Bit into carry - trashes r2
-static int GetXBit(int subtract)
+int GetXBit(int subtract)
 {
   ot(";@ Get X bit:\n");
-  ot("  ldrb r2,[r7,#0x45]\n");
-  if (subtract) ot("  mvn r2,r2,lsl #28 ;@ Invert it\n");
-  else          ot("  mov r2,r2,lsl #28\n");
+  ot("  ldr r2,[r7,#0x4c]\n");
+  if (subtract) ot("  mvn r2,r2 ;@ Invert it\n");
   ot("  msr cpsr_flg,r2 ;@ Get into Carry\n");
   ot("\n");
   return 0;
@@ -322,59 +383,158 @@ static int GetXBit(int subtract)
 int OpAbcd(int op)
 {
   int use=0;
-  int type=0,sea=0,addr=0,dea=0;
+  int type=0,sea=0,mem=0,dea=0;
   
-  type=(op>>14)&1;
+  type=(op>>14)&1; // sbcd/abcd
   dea =(op>> 9)&7;
-  addr=(op>> 3)&1;
+  mem =(op>> 3)&1;
   sea = op     &7;
 
-  if (addr) { sea|=0x20; dea|=0x20; }
+  if (mem) { sea|=0x20; dea|=0x20; }
 
-  use=op&~0x0e07; // Use same opcode for all registers
+  use=op&~0x0e07; // Use same opcode for all registers..
+  if (sea==0x27) use|=0x0007; // ___x.b -(a7)
+  if (dea==0x27) use|=0x0e00; // ___x.b -(a7)
   if (op!=use) { OpUse(op,use); return 0; } // Use existing handler
 
-  OpStart(op); Cycles=6;
+  OpStart(op,sea,dea); Cycles=6;
 
-  EaCalc( 0,0x0007, sea,0);
-  EaRead( 0,    10, sea,0,1);
-  EaCalc(11,0x0e00, dea,0);
-  EaRead(11,     1, dea,0,1);
+  if (mem)
+  {
+    ot(";@ Get src/dest EA vals\n");
+    EaCalc (0,0x000f, sea,0,1);
+    EaRead (0,     6, sea,0,0x000f,1);
+    EaCalcReadNoSE(11,0,dea,0,0x0e00);
+  }
+  else
+  {
+    ot(";@ Get src/dest reg vals\n");
+    EaCalcReadNoSE(-1,6,sea,0,0x0007);
+    EaCalcReadNoSE(11,0,dea,0,0x0e00);
+    ot("  mov r6,r6,asl #24\n");
+  }
+  ot("  mov r1,r0,asl #24\n\n");
 
-  ot("  ldrb r2,[r7,#0x45] ;@ Get X bit\n");
-  ot("  tst r2,#2\n");
-  ot("  addne r10,r10,#0x01000000 ;@ Add carry bit\n");
+  ot("  bic r10,r10,#0xb1000000 ;@ clear all flags except old Z\n");
 
   if (type)
   {
-    ot(";@ Add units into r2:\n");
-    ot("  and r2,r1, #0x0f000000\n");
-    ot("  and r0,r10,#0x0f000000\n");
-    ot("  add r2,r2,r0\n");
-    ot("  cmp r2,#0x0a000000\n");
-    ot("  addpl r1,r1,#0x06000000 ;@ Decimal adjust units\n");
-    ot("  add r1,r1,r10 ;@ Add BCD\n");
-    ot("  mov r0,r1,lsr #24\n");
-    ot("  cmp r0,#0xa0\n");
-    ot("  addpl r1,r1,#0x60000000 ;@ Decimal adjust tens\n");
-    OpGetFlags(0,1);
+    ot("  ldr r0,[r7,#0x4c] ;@ Get X bit\n");
+    ot("  mov r3,#0x00f00000\n");
+    ot("  and r2,r3,r1,lsr #4\n");
+    ot("  tst r0,#0x20000000\n");
+    ot("  and r0,r3,r6,lsr #4\n");
+    ot("  add r0,r0,r2\n");
+    ot("  addne r0,r0,#0x00100000\n");
+//    ot("  tst r0,#0x00800000\n");
+//    ot("  orreq r10,r10,#0x01000000 ;@ Undefined V behavior\n");
+    ot("  cmp r0,#0x00900000\n");
+    ot("  addhi r0,r0,#0x00600000 ;@ Decimal adjust units\n");
+
+    ot("  mov r2,r1,lsr #28\n");
+    ot("  add r0,r0,r2,lsl #24\n");
+    ot("  mov r2,r6,lsr #28\n");
+    ot("  add r0,r0,r2,lsl #24\n");
+    ot("  cmp r0,#0x09900000\n");
+    ot("  orrhi r10,r10,#0x20000000 ;@ C\n");
+    ot("  subhi r0,r0,#0x0a000000\n");
+//    ot("  and r3,r10,r0,lsr #3 ;@ Undefined V behavior part II\n");
+//    ot("  orr r10,r10,r3,lsl #4 ;@ V\n");
+    ot("  movs r0,r0,lsl #4\n");
+    ot("  orrmi r10,r10,#0x90000000 ;@ Undefined N+V behavior\n"); // this is what Musashi really does
+    ot("  bicne r10,r10,#0x40000000 ;@ Z flag\n");
   }
   else
   {
-    ot(";@ Sub units into r2:\n");
-    ot("  and r2,r1, #0x0f000000\n");
-    ot("  and r0,r10,#0x0f000000\n");
-    ot("  subs r2,r2,r0\n");
-    ot("  submi r1,r1,#0x06000000 ;@ Decimal adjust units\n");
-    ot("  subs r1,r1,r10 ;@ Subtract BCD\n");
-    ot("  submis r1,r1,#0x60000000 ;@ Decimal adjust tens\n");
-    OpGetFlags(1,1);
+    ot("  ldr r0,[r7,#0x4c] ;@ Get X bit\n");
+    ot("  mov r3,#0x00f00000\n");
+    ot("  and r2,r3,r6,lsr #4\n");
+    ot("  tst r0,#0x20000000\n");
+    ot("  and r0,r3,r1,lsr #4\n");
+    ot("  sub r0,r0,r2\n");
+    ot("  subne r0,r0,#0x00100000\n");
+//    ot("  tst r0,#0x00800000\n");
+//    ot("  orreq r10,r10,#0x01000000 ;@ Undefined V behavior\n");
+    ot("  cmp r0,#0x00900000\n");
+    ot("  subhi r0,r0,#0x00600000 ;@ Decimal adjust units\n");
+
+    ot("  mov r2,r1,lsr #28\n");
+    ot("  add r0,r0,r2,lsl #24\n");
+    ot("  mov r2,r6,lsr #28\n");
+    ot("  sub r0,r0,r2,lsl #24\n");
+    ot("  cmp r0,#0x09900000\n");
+    ot("  orrhi r10,r10,#0xa0000000 ;@ N and C\n");
+    ot("  addhi r0,r0,#0x0a000000\n");
+//    ot("  and r3,r10,r0,lsr #3 ;@ Undefined V behavior part II\n");
+//    ot("  orr r10,r10,r3,lsl #4 ;@ V\n");
+    ot("  movs r0,r0,lsl #4\n");
+//    ot("  orrmi r10,r10,#0x80000000 ;@ Undefined N behavior\n");
+    ot("  bicne r10,r10,#0x40000000 ;@ Z flag\n");
   }
+
+  ot("  str r10,[r7,#0x4c] ;@ Save X bit\n");
+  ot("\n");
+
+  EaWrite(11,     0, dea,0,0x0e00,1);
+
+  ot("  ldr r6,[r7,#0x54]\n");
+  OpEnd(sea,dea);
+
+  return 0;
+}
+
+// 01001000 00eeeeee - nbcd <ea>
+int OpNbcd(int op)
+{
+  int use=0;
+  int ea=0;
+  
+  ea=op&0x3f;
+
+  if(EaCanWrite(ea)==0||EaAn(ea)) return 1;
+
+  use=OpBase(op,0);
+  if(op!=use) { OpUse(op,use); return 0; } // Use existing handler
+
+  OpStart(op,ea); Cycles=6;
+  if(ea >= 8)  Cycles+=2;
+
+  EaCalcReadNoSE(6,0,ea,0,0x003f);
+
+  // this is rewrite of Musashi's code
+  ot("  ldr r2,[r7,#0x4c]\n");
+  ot("  bic r10,r10,#0xb0000000 ;@ clear all flags, except Z\n");
+  ot("  mov r0,r0,asl #24\n");
+  ot("  and r2,r2,#0x20000000\n");
+  ot("  add r2,r0,r2,lsr #5 ;@ add X\n");
+  ot("  rsb r11,r2,#0x9a000000 ;@ do arithmetic\n");
+
+  ot("  cmp r11,#0x9a000000\n");
+  ot("  beq finish%.4x\n",op);
   ot("\n");
 
-  EaWrite(11,     1, dea,0,1);
+  ot("  mvn r3,r11,lsr #31 ;@ Undefined V behavior\n",op);
+  ot("  and r2,r11,#0x0f000000\n");
+  ot("  cmp r2,#0x0a000000\n");
+  ot("  andeq r11,r11,#0xf0000000\n");
+  ot("  addeq r11,r11,#0x10000000\n");
+  ot("  and r3,r3,r11,lsr #31 ;@ Undefined V behavior part II\n",op);
+  ot("  movs r1,r11,asr #24\n");
+  ot("  bicne r10,r10,#0x40000000 ;@ Z\n");
+  ot("  orr r10,r10,r3,lsl #28 ;@ save V\n",op);
+  ot("  orr r10,r10,#0x20000000 ;@ C\n");
+  ot("\n");
+
+  EaWrite(6, 1, ea,0,0x3f,0,0);
+
+  ot("finish%.4x%s\n",op,ms?"":":");
+  ot("  tst r11,r11\n");
+  ot("  orrmi r10,r10,#0x80000000 ;@ N\n");
+  ot("  str r10,[r7,#0x4c] ;@ Save X\n");
+  ot("\n");
 
-  OpEnd();
+  ot("  ldr r6,[r7,#0x54]\n");
+  OpEnd(ea);
 
   return 0;
 }
@@ -385,6 +545,7 @@ int OpAritha(int op)
 {
   int use=0;
   int type=0,size=0,sea=0,dea=0;
+  const char *asr="";
 
   // Suba/Cmpa/Adda/(invalid):
   type=(op>>13)&3; if (type>=3) return 1;
@@ -396,83 +557,118 @@ int OpAritha(int op)
   // See if we can do this opcode:
   if (EaCanRead(sea,size)==0) return 1;
 
-  use=OpBase(op);
+  use=OpBase(op,size);
   use&=~0x0e00; // Use same opcode for An
   if (op!=use) { OpUse(op,use); return 0; } // Use existing handler
 
-  OpStart(op); Cycles=4;
-  EaCalc ( 0,0x003f, sea,size);
-  EaRead ( 0,    10, sea,size);
+  OpStart(op,sea); Cycles=(size==2)?6:8;
+  if(size==2&&(sea<0x10||sea==0x3c)) Cycles+=2;
+  if(type==1) Cycles=6;
+
+  // EA calculation order defines how situations like  suba.w (A0)+, A0 get handled.
+  // different emus act differently in this situation, I couldn't fugure which is right behaviour.
+  //if (type == 1)
+  {
+    EaCalcReadNoSE(-1,0,sea,size,0x003f);
+    EaCalcReadNoSE(type!=1?11:-1,1,dea,2,0x0e00);
+  }
+#if 0
+  else
+  {
+    EaCalcReadNoSE(type!=1?11:-1,1,dea,2,0x0e00);
+    EaCalcReadNoSE(-1,0,sea,size,0x003f);
+  }
+#endif
 
-  EaCalc ( 0,0x0e00, dea,2);
-  EaRead ( 0,     1, dea,2);
+  if (size<2) ot("  mov r0,r0,asl #%d\n\n",size?16:24);
+  if (size<2) asr=(char *)(size?",asr #16":",asr #24");
 
-  if (type==0) ot("  sub r1,r1,r10\n");
-  if (type==1) ot("  cmp r1,r10 ;@ Defines NZCV\n");
+  if (type==0) ot("  sub r1,r1,r0%s\n",asr);
+  if (type==1) ot("  cmp r1,r0%s ;@ Defines NZCV\n",asr);
   if (type==1) OpGetFlags(1,0); // Get Cmp flags
-  if (type==2) ot("  add r1,r1,r10\n");
+  if (type==2) ot("  add r1,r1,r0%s\n",asr);
   ot("\n");
-  
-  EaWrite( 0,     1, dea,2);
 
-  if (Amatch && sea==0x3c) Cycles-=size<2?4:8; // Correct?
-  if (size>=2) { if (sea<0x10) Cycles+=4; else Cycles+=2; }
+  if (type!=1) EaWrite(11, 1, dea,2,0x0e00);
 
-  OpEnd();
+  OpEnd(sea);
 
   return 0;
 }
 
 // --------------------- Opcodes 0x9100+ ---------------------
-// Emit a Subx/Addx opcode, 1t01ddd1 zz000sss addx.z Ds,Dd
+// Emit a Subx/Addx opcode, 1t01ddd1 zz00rsss addx.z Ds,Dd
 int OpAddx(int op)
 {
   int use=0;
-  int type=0,size=0,dea=0,sea=0;
+  int type=0,size=0,dea=0,sea=0,mem=0;
+  const char *asl="";
 
-  type=(op>>12)&5;
+  type=(op>>14)&1;
   dea =(op>> 9)&7;
   size=(op>> 6)&3; if (size>=3) return 1;
-  sea = op&0x3f;
+  sea = op&7;
+  mem =(op>> 3)&1;
 
   // See if we can do this opcode:
   if (EaCanRead(sea,size)==0) return 1;
   if (EaCanWrite(dea)==0) return 1;
 
-  use=OpBase(op);
-  use&=~0x0e00; // Use same opcode for Dn
+  if (mem) { sea+=0x20; dea+=0x20; }
+
+  use=op&~0x0e07; // Use same opcode for Dn
+  if (size==0&&sea==0x27) use|=0x0007; // ___x.b -(a7)
+  if (size==0&&dea==0x27) use|=0x0e00; // ___x.b -(a7)
   if (op!=use) { OpUse(op,use); return 0; } // Use existing handler
 
-  OpStart(op); Cycles=8;
+  OpStart(op,sea,dea); Cycles=4;
+  if(size>=2)   Cycles+=4;
+  if(sea>=0x10) Cycles+=2;
 
-  ot(";@ Get r10=EA r11=EA value\n");
-  EaCalc( 0,0x003f,sea,size);
-  EaRead( 0,    11,sea,size,1);
-  ot(";@ Get r0=Register r1=Register value\n");
-  EaCalc( 0,0x0e00,dea,size);
-  EaRead( 0,     1,dea,size,1);
+  if (mem)
+  {
+    ot(";@ Get src/dest EA vals\n");
+    EaCalc (0,0x000f, sea,size,1);
+    EaRead (0,     6, sea,size,0x000f,1);
+    EaCalcReadNoSE(11,0,dea,size,0x0e00);
+  }
+  else
+  {
+    ot(";@ Get src/dest reg vals\n");
+    EaCalcReadNoSE(-1,6,sea,size,0x0007);
+    EaCalcReadNoSE(11,0,dea,size,0x0e00);
+    if (size<2) ot("  mov r6,r6,asl #%d\n\n",size?16:24);
+  }
+
+  if (size<2) asl=(char *)(size?",asl #16":",asl #24");
 
   ot(";@ Do arithmetic:\n");
-  GetXBit(type==1);
+  GetXBit(type==0);
 
-  if (type==5 && size<2)
+  if (type==1 && size<2)
   {
     ot(";@ Make sure the carry bit will tip the balance:\n");
-    if (size==0) ot("  ldr r2,=0x00ffffff\n");
-    else         ot("  ldr r2,=0x0000ffff\n");
-    ot("  orr r11,r11,r2\n");
+    ot("  mvn r2,#0\n");
+    ot("  orr r6,r6,r2,lsr #%i\n",(size==0)?8:16);
     ot("\n");
   }
 
-  if (type==1) ot("  sbcs r1,r1,r11\n");
-  if (type==5) ot("  adcs r1,r1,r11\n");
-  OpGetFlags(type==1,1); // subtract
+  if (type==0) ot("  rscs r1,r6,r0%s\n",asl);
+  if (type==1) ot("  adcs r1,r6,r0%s\n",asl);
+  ot("  orr r3,r10,#0xb0000000 ;@ for old Z\n");
+  OpGetFlags(type==0,1,0); // subtract
+  if (size<2) {
+    ot("  movs r2,r1,lsr #%i\n", size?16:24);
+    ot("  orreq r10,r10,#0x40000000 ;@ add potentially missed Z\n");
+  }
+  ot("  andeq r10,r10,r3 ;@ fix Z\n");
   ot("\n");
 
   ot(";@ Save result:\n");
-  EaWrite( 0, 1, dea,size,1);
+  EaWrite(11, 1, dea,size,0x0e00,1);
 
-  OpEnd();
+  ot("  ldr r6,[r7,#0x54]\n");
+  OpEnd(sea,dea);
 
   return 0;
 }
@@ -483,6 +679,7 @@ int OpCmpEor(int op)
 {
   int rea=0,eor=0;
   int size=0,ea=0,use=0;
+  const char *asl="";
 
   // Get EA and register EA
   rea=(op>>9)&7;
@@ -490,41 +687,144 @@ int OpCmpEor(int op)
   size=(op>>6)&3; if (size>=3) return 1;
   ea=op&0x3f;
 
+  if (eor && (ea>>3) == 1) return 1; // not a valid mode for eor
+
   // See if we can do this opcode:
   if (EaCanRead(ea,size)==0) return 1;
   if (eor && EaCanWrite(ea)==0) return 1;
+  if (EaAn(ea)&&(eor||size==0)) return 1;
 
-  use=OpBase(op);
+  use=OpBase(op,size);
   use&=~0x0e00; // Use 1 handler for register d0-7
   if (op!=use) { OpUse(op,use); return 0; } // Use existing handler
 
-  OpStart(op); Cycles=eor?8:4;
+  OpStart(op,ea); Cycles=4;
+  if(eor) {
+    if(ea>8)     Cycles+=4;
+    if(size>=2)  Cycles+=4;
+  } else {
+    if(size>=2)  Cycles+=2;
+  }
 
-  ot(";@ Get EA into r10 and value into r0:\n");
-  EaCalc (10,0x003f,  ea,size);
-  EaRead (10,     0,  ea,size,1);
+  ot(";@ Get EA into r11 and value into r0:\n");
+  EaCalcReadNoSE(eor?11:-1,0,ea,size,0x003f);
 
   ot(";@ Get register operand into r1:\n");
-  EaCalc (1 ,0x0e00, rea,size);
-  EaRead (1,      1, rea,size,1);
+  EaCalcReadNoSE(-1,1,rea,size,0x0e00);
+
+  if (size<2) ot("  mov r0,r0,asl #%d\n\n",size?16:24);
+  if (size<2) asl=(char *)(size?",asl #16":",asl #24");
 
   ot(";@ Do arithmetic:\n");
-  if (eor==0) ot("  cmp r1,r0\n");
+  if (eor==0) ot("  rsbs r1,r0,r1%s\n",asl);
   if (eor)
   {
-    ot("  eor r1,r0,r1\n");
+    ot("  eor r1,r0,r1%s\n",asl);
     ot("  adds r1,r1,#0 ;@ Defines NZ, clears CV\n");
   }
 
   OpGetFlags(eor==0,0); // Cmp like subtract
   ot("\n");
 
-  if (size>=2) Cycles+=4; // Correct?
-  if (ea==0x3c) Cycles-=4;
+  if (eor) EaWrite(11, 1,ea,size,0x003f,1);
+
+  OpEnd(ea);
+  return 0;
+}
+
+// Emit a Cmpm opcode, 1011ddd1 xx001sss (rrr=Adst, xx=size extension, sss=Asrc)
+int OpCmpm(int op)
+{
+  int size=0,sea=0,dea=0,use=0;
+  const char *asl="";
+
+  // get size, get EAs
+  size=(op>>6)&3; if (size>=3) return 1;
+  sea=(op&7)|0x18;
+  dea=(op>>9)&0x3f;
+
+  use=op&~0x0e07; // Use 1 handler for all registers..
+  if (size==0&&sea==0x1f) use|=0x0007; // ..except (a7)+
+  if (size==0&&dea==0x1f) use|=0x0e00;
+  if (op!=use) { OpUse(op,use); return 0; } // Use existing handler
+
+  OpStart(op,sea); Cycles=4;
+
+  ot(";@ Get src operand into r11:\n");
+  EaCalc (0,0x0007, sea,size,1);
+  EaRead (0,    11, sea,size,0x0007,1);
+
+  ot(";@ Get dst operand into r0:\n");
+  EaCalcReadNoSE(-1,0,dea,size,0x0e00);
+
+  if (size<2) asl=(char *)(size?",asl #16":",asl #24");
+
+  ot("  rsbs r0,r11,r0%s\n",asl);
+  OpGetFlags(1,0); // Cmp like subtract
+  ot("\n");
+
+  OpEnd(sea);
+  return 0;
+}
+
+
+// Emit a Chk opcode, 0100ddd1 x0eeeeee (rrr=Dn, x=size extension, eeeeee=ea)
+int OpChk(int op)
+{
+  int rea=0;
+  int size=0,ea=0,use=0;
+
+  // Get EA and register EA
+  rea=(op>>9)&7;
+  if((op>>7)&1)
+       size=1; // word operation
+  else size=2; // long
+  ea=op&0x3f;
+
+  if (EaAn(ea)) return 1; // not a valid mode
+  if (size!=1)  return 1; // 000 variant only supports word
+
+  // See if we can do this opcode:
+  if (EaCanRead(ea,size)==0) return 1;
+
+  use=OpBase(op,size);
+  use&=~0x0e00; // Use 1 handler for register d0-7
+  if (op!=use) { OpUse(op,use); return 0; } // Use existing handler
+
+  OpStart(op,ea); Cycles=10;
+
+  ot(";@ Get value into r0:\n");
+  EaCalcReadNoSE(-1,0,ea,size,0x003f);
+
+  ot(";@ Get register operand into r1:\n");
+  EaCalcReadNoSE(-1,1,rea,size,0x0e00);
+
+  if (size<2) ot("  mov r0,r0,asl #%d\n",size?16:24);
+  if (size<2) ot("  mov r1,r1,asl #%d\n\n",size?16:24);
+
+  ot(";@ get flags, including undocumented ones\n");
+  ot("  and r3,r10,#0x80000000\n");
+  ot("  adds r1,r1,#0 ;@ Defines NZ, clears CV\n");
+  OpGetFlags(0,0);
+
+  ot(";@ is reg negative?\n");
+  ot("  bmi chktrap%.4x\n",op);
+
+  ot(";@ Do arithmetic:\n");
+  ot("  bic r10,r10,#0x80000000 ;@ N\n");
+  ot("  cmp r1,r0\n");
+  ot("  bgt chktrap%.4x\n",op);
+
+  ot(";@ old N remains\n");
+  ot("  orr r10,r10,r3\n");
+  OpEnd(ea);
 
-  if (eor) EaWrite(10, 1,ea,size,1);
+  ot("chktrap%.4x%s ;@ CHK exception:\n",op,ms?"":":");
+  ot("  mov r0,#6\n");
+  ot("  bl Exception\n");
+  Cycles+=40;
+  OpEnd(ea);
 
-  OpEnd();
   return 0;
 }