EaCalc(10,0x0e00,rea, 2);\r
EaRead(10, 2,rea, 2,0x0e00);\r
\r
- ot(" movs r0,r0,asl #16\n");\r
+ ot(" movs r1,r0,asl #16\n");\r
\r
if (type==0) // div\r
{\r
if (sign)\r
{\r
ot(" mov r11,#0 ;@ r11 = 1 or 2 if the result is negative\n");\r
- ot(" orrmi r11,r11,#1\n");\r
- ot(" mov r0,r0,asr #16\n");\r
- ot(" rsbmi r0,r0,#0 ;@ Make r0 positive\n");\r
- ot("\n");\r
ot(" tst r2,r2\n");\r
ot(" orrmi r11,r11,#2\n");\r
ot(" rsbmi r2,r2,#0 ;@ Make r2 positive\n");\r
ot("\n");\r
+ ot(" movs r0,r1,asr #16\n");\r
+ ot(" orrmi r11,r11,#1\n");\r
+ ot(" rsbmi r0,r0,#0 ;@ Make r0 positive\n");\r
+ ot("\n");\r
+ ot(";@ detect the nasty 0x80000000 / -1 situation\n");\r
+ ot(" mov r3,r2,asr #31\n");\r
+ ot(" eors r3,r3,r1,asr #16\n");\r
+ ot(" beq wrendofop%.4x\n",op);\r
}\r
else\r
{\r
- ot(" mov r0,r0,lsr #16 ;@ use only 16 bits of divisor\n");\r
+ ot(" mov r0,r1,lsr #16 ;@ use only 16 bits of divisor\n");\r
}\r
\r
+ ot("\n");\r
ot(";@ Divide r2 by r0\n");\r
ot(" mov r3,#0\n");\r
ot(" mov r1,r0\n");\r
ot(" cmp r3,r1,asr #16 ;@ signed overflow?\n");\r
ot(" orrne r9,r9,#0x10000000 ;@ set overflow flag\n");\r
ot(" bne endofop%.4x ;@ overflow!\n",op);\r
+ ot("\n");\r
+ ot("wrendofop%.4x%s\n",op,ms?"":":");\r
}\r
else\r
{\r
ot(" movs r1,r3,lsr #16 ;@ check for overflow condition\n");\r
ot(" orrne r9,r9,#0x10000000 ;@ set overflow flag\n");\r
ot(" bne endofop%.4x ;@ overflow!\n",op);\r
+ ot("\n");\r
}\r
\r
ot(" mov r1,r3,lsl #16 ;@ Clip to 16-bits\n");\r
if (type==1)\r
{\r
ot(";@ Get 16-bit signs right:\n");\r
- ot(" mov r0,r0,%s #16\n",sign?"asr":"lsr");\r
+ ot(" mov r0,r1,%s #16\n",sign?"asr":"lsr");\r
ot(" mov r2,r2,lsl #16\n");\r
ot(" mov r2,r2,%s #16\n",sign?"asr":"lsr");\r
ot("\n");\r
ot(" mov r0,r0,asl #24\n");\r
ot(" and r2,r2,#0x20000000\n");\r
ot(" add r2,r0,r2,lsr #5 ;@ add X\n");\r
- ot(" rsbs r1,r2,#0x9a000000 ;@ do arithmetic\n");\r
+ ot(" rsb r11,r2,#0x9a000000 ;@ do arithmetic\n");\r
\r
- ot(" orrmi r9,r9,#0x80000000 ;@ N\n");\r
- ot(" cmp r1,#0x9a000000\n");\r
+ ot(" cmp r11,#0x9a000000\n");\r
ot(" beq finish%.4x\n",op);\r
ot("\n");\r
\r
- ot(" mvn r3,r9,lsr #3 ;@ Undefined V behavior\n",op);\r
- ot(" and r2,r1,#0x0f000000\n");\r
+ ot(" mvn r3,r11,lsr #31 ;@ Undefined V behavior\n",op);\r
+ ot(" and r2,r11,#0x0f000000\n");\r
ot(" cmp r2,#0x0a000000\n");\r
- ot(" andeq r1,r1,#0xf0000000\n");\r
- ot(" addeq r1,r1,#0x10000000\n");\r
- ot(" and r3,r3,r1,lsr #3 ;@ Undefined V behavior part II\n",op);\r
- ot(" tst r1,r1\n");\r
- ot(" orr r9,r9,r3 ;@ save V\n",op);\r
+ ot(" andeq r11,r11,#0xf0000000\n");\r
+ ot(" addeq r11,r11,#0x10000000\n");\r
+ ot(" and r3,r3,r11,lsr #31 ;@ Undefined V behavior part II\n",op);\r
+ ot(" movs r1,r11,asr #24\n");\r
ot(" bicne r9,r9,#0x40000000 ;@ Z\n");\r
+ ot(" orr r9,r9,r3,lsl #28 ;@ save V\n",op);\r
ot(" orr r9,r9,#0x20000000 ;@ C\n");\r
ot("\n");\r
\r
- EaWrite(10, 1, ea,0,0x3f,1);\r
+ EaWrite(10, 1, ea,0,0x3f,0,0);\r
\r
ot("finish%.4x%s\n",op,ms?"":":");\r
+ ot(" tst r11,r11\n");\r
+ ot(" orrmi r9,r9,#0x80000000 ;@ N\n");\r
ot(" str r9,[r7,#0x4c] ;@ Save X\n");\r
+ ot("\n");\r
\r
OpEnd(ea);\r
\r
if(size==2&&(sea<0x10||sea==0x3c)) Cycles+=2;\r
if(type==1) Cycles=6;\r
\r
- // to handle suba.w (A0)+, A0 properly, must calc reg EA first\r
- EaCalcReadNoSE(type!=1?10:-1,11,dea,2,0x0e00);\r
- EaCalcReadNoSE(-1,0,sea,size,0x003f);\r
+ // EA calculation order defines how situations like suba.w (A0)+, A0 get handled.\r
+ // different emus act differently in this situation, I couldn't fugure which is right behaviour.\r
+ // This is Musashi's behaviour.\r
+ if (type == 1)\r
+ {\r
+ EaCalcReadNoSE(-1,0,sea,size,0x003f);\r
+ EaCalcReadNoSE(type!=1?10:-1,11,dea,2,0x0e00);\r
+ }\r
+ else\r
+ {\r
+ EaCalcReadNoSE(type!=1?10:-1,11,dea,2,0x0e00);\r
+ EaCalcReadNoSE(-1,0,sea,size,0x003f);\r
+ }\r
\r
if (size<2) ot(" mov r0,r0,asl #%d\n\n",size?16:24);\r
if (size<2) asr=(char *)(size?",asr #16":",asr #24");\r
ot(" ldr pc,[r7,#0x64] ;@ Call checkpc()\n");\r
ot(" mov r4,r0\n");\r
#else\r
- if (reg != 4)\r
- ot(" mov r4,r%i\n", reg);\r
+ ot(" bic r4,r%d,#1\n",reg); // we do not emulate address errors\r
#endif\r
ot("\n");\r
}\r
case 6: // trapv\r
OpStart(op,0x10); Cycles=4;\r
ot(" tst r9,#0x10000000\n");\r
- ot(" subne r5,r5,#%i\n",30);\r
+ ot(" subne r5,r5,#%i\n",34);\r
ot(" movne r0,#0x1c ;@ TRAPV exception\n");\r
ot(" blne Exception\n");\r
OpEnd(0x10);\r
ot(" addeq r4,r4,#2 ;@ Skip branch offset\n");\r
ot(" subeq r5,r5,#4 ;@ additional cycles\n");\r
ot(" addne r4,r4,r0 ;@ r4 = New PC\n");\r
+ ot(" bic r4,r4,#1\n"); // we do not emulate address errors\r
ot("\n");\r
#endif\r
Cycles=12-2;\r
if (offset==0) size=1;\r
if (offset==-1) size=2;\r
\r
+ if (size==2) size=0; // 000 model does not support long displacement\r
if (size) use=op; // 16-bit or 32-bit\r
else use=(op&0xff00)+1; // Use same opcode for all 8-bit branches\r
\r
Cycles=18; // always 18\r
}\r
\r
+ ot(" add r0,r4,r11%s ;@ r4 = New PC\n",asr_r11);\r
+\r
#if USE_CHECKPC_CALLBACK && USE_CHECKPC_OFFSETBITS_8\r
if (offset!=0 && offset!=-1) checkpc=1;\r
#endif\r
#endif\r
if (checkpc)\r
{\r
- ot(" add r0,r4,r11%s ;@ r4 = New PC\n",asr_r11);\r
CheckPc(0);\r
}\r
else\r
{\r
- ot(" add r4,r4,r11%s ;@ r4 = New PC\n",asr_r11);\r
+ ot(" bic r4,r0,#1\n"); // we do not emulate address errors\r
ot("\n");\r
}\r
\r