etc/yuzu
1
0
Fork 0
Code Issues Pull requests Projects Releases Packages Wiki Activity

Merge pull request #361 from lioncash/moreqops

Browse source 
dyncom/armemu: Implement QADD8/QSUB8.
This commit is contained in:
bunnei 2014-12-29 14:53:04 -05:00
commit 2d2aa2c0be
4 changed files with 142 additions and 65 deletions
Show stats Download patch file Download diff file Expand all files Collapse all files

74
src/core/arm/dyncom/arm_dyncom_interpreter.cpp
View file

@ -2419,8 +2419,7 @@ ARM_INST_PTR INTERPRETER_TRANSLATE(pld)(unsigned int inst, int index)
return inst_base;
}
ARM_INST_PTR INTERPRETER_TRANSLATE(qadd)(unsigned int inst, int index) { UNIMPLEMENTED_INSTRUCTION("QADD"); }
ARM_INST_PTR INTERPRETER_TRANSLATE(qadd8)(unsigned int inst, int index) { UNIMPLEMENTED_INSTRUCTION("QADD8"); }
ARM_INST_PTR INTERPRETER_TRANSLATE(qadd16)(unsigned int inst, int index)
ARM_INST_PTR INTERPRETER_TRANSLATE(qadd8)(unsigned int inst, int index)
{
arm_inst* const inst_base = (arm_inst*)AllocBuffer(sizeof(arm_inst) + sizeof(generic_arm_inst));
generic_arm_inst* const inst_cream = (generic_arm_inst*)inst_base->component;
@ -2438,21 +2437,28 @@ ARM_INST_PTR INTERPRETER_TRANSLATE(qadd16)(unsigned int inst, int index)
return inst_base;
}
ARM_INST_PTR INTERPRETER_TRANSLATE(qadd16)(unsigned int inst, int index)
{
return INTERPRETER_TRANSLATE(qadd8)(inst, index);
}
ARM_INST_PTR INTERPRETER_TRANSLATE(qaddsubx)(unsigned int inst, int index)
{
return INTERPRETER_TRANSLATE(qadd16)(inst, index);
return INTERPRETER_TRANSLATE(qadd8)(inst, index);
}
ARM_INST_PTR INTERPRETER_TRANSLATE(qdadd)(unsigned int inst, int index) { UNIMPLEMENTED_INSTRUCTION("QDADD"); }
ARM_INST_PTR INTERPRETER_TRANSLATE(qdsub)(unsigned int inst, int index) { UNIMPLEMENTED_INSTRUCTION("QDSUB"); }
ARM_INST_PTR INTERPRETER_TRANSLATE(qsub)(unsigned int inst, int index) { UNIMPLEMENTED_INSTRUCTION("QSUB"); }
ARM_INST_PTR INTERPRETER_TRANSLATE(qsub8)(unsigned int inst, int index) { UNIMPLEMENTED_INSTRUCTION("QSUB8"); }
ARM_INST_PTR INTERPRETER_TRANSLATE(qsub8)(unsigned int inst, int index)
{
return INTERPRETER_TRANSLATE(qadd8)(inst, index);
}
ARM_INST_PTR INTERPRETER_TRANSLATE(qsub16)(unsigned int inst, int index)
{
return INTERPRETER_TRANSLATE(qadd16)(inst, index);
return INTERPRETER_TRANSLATE(qadd8)(inst, index);
}
ARM_INST_PTR INTERPRETER_TRANSLATE(qsubaddx)(unsigned int inst, int index)
{
return INTERPRETER_TRANSLATE(qadd16)(inst, index);
return INTERPRETER_TRANSLATE(qadd8)(inst, index);
}
ARM_INST_PTR INTERPRETER_TRANSLATE(rev)(unsigned int inst, int index)
{
@ -5777,55 +5783,60 @@ unsigned InterpreterMainLoop(ARMul_State* state)
GOTO_NEXT_INST;
}
QADD_INST:
QADD8_INST:
QADD8_INST:
QADD16_INST:
QADDSUBX_INST:
QSUB8_INST:
QSUB16_INST:
QSUBADDX_INST:
{
INC_ICOUNTER;
if (inst_base->cond == 0xE || CondPassed(cpu, inst_base->cond)) {
generic_arm_inst* const inst_cream = (generic_arm_inst*)inst_base->component;
const s16 rm_lo = (RM & 0xFFFF);
const s16 rm_hi = ((RM >> 16) & 0xFFFF);
const s16 rn_lo = (RN & 0xFFFF);
const s16 rn_hi = ((RN >> 16) & 0xFFFF);
const u16 rm_lo = (RM & 0xFFFF);
const u16 rm_hi = ((RM >> 16) & 0xFFFF);
const u16 rn_lo = (RN & 0xFFFF);
const u16 rn_hi = ((RN >> 16) & 0xFFFF);
const u8 op2 = inst_cream->op2;
s32 lo_result = 0;
s32 hi_result = 0;
u16 lo_result = 0;
u16 hi_result = 0;
// QADD16
if (op2 == 0x00) {
lo_result = (rn_lo + rm_lo);
hi_result = (rn_hi + rm_hi);
lo_result = ARMul_SignedSaturatedAdd16(rn_lo, rm_lo);
hi_result = ARMul_SignedSaturatedAdd16(rn_hi, rm_hi);
}
// QASX
else if (op2 == 0x01) {
lo_result = (rn_lo - rm_hi);
hi_result = (rn_hi + rm_lo);
lo_result = ARMul_SignedSaturatedSub16(rn_lo, rm_hi);
hi_result = ARMul_SignedSaturatedAdd16(rn_hi, rm_lo);
}
// QSAX
else if (op2 == 0x02) {
lo_result = (rn_lo + rm_hi);
hi_result = (rn_hi - rm_lo);
lo_result = ARMul_SignedSaturatedAdd16(rn_lo, rm_hi);
hi_result = ARMul_SignedSaturatedSub16(rn_hi, rm_lo);
}
// QSUB16
else if (op2 == 0x03) {
lo_result = (rn_lo - rm_lo);
hi_result = (rn_hi - rm_hi);
lo_result = ARMul_SignedSaturatedSub16(rn_lo, rm_lo);
hi_result = ARMul_SignedSaturatedSub16(rn_hi, rm_hi);
}
// QADD8
else if (op2 == 0x04) {
lo_result = ARMul_SignedSaturatedAdd8(rn_lo & 0xFF, rm_lo & 0xFF) |
ARMul_SignedSaturatedAdd8(rn_lo >> 8, rm_lo >> 8) << 8;
hi_result = ARMul_SignedSaturatedAdd8(rn_hi & 0xFF, rm_hi & 0xFF) |
ARMul_SignedSaturatedAdd8(rn_hi >> 8, rm_hi >> 8) << 8;
}
// QSUB8
else if (op2 == 0x07) {
lo_result = ARMul_SignedSaturatedSub8(rn_lo & 0xFF, rm_lo & 0xFF) |
ARMul_SignedSaturatedSub8(rn_lo >> 8, rm_lo >> 8) << 8;
hi_result = ARMul_SignedSaturatedSub8(rn_hi & 0xFF, rm_hi & 0xFF) |
ARMul_SignedSaturatedSub8(rn_hi >> 8, rm_hi >> 8) << 8;
}
if (lo_result > 0x7FFF)
lo_result = 0x7FFF;
else if (lo_result < -0x8000)
lo_result = -0x8000;
if (hi_result > 0x7FFF)
hi_result = 0x7FFF;
else if (hi_result < -0x8000)
hi_result = -0x8000;
RD = (lo_result & 0xFFFF) | ((hi_result & 0xFFFF) << 16);
}
@ -5839,7 +5850,6 @@ unsigned InterpreterMainLoop(ARMul_State* state)
QDADD_INST:
QDSUB_INST:
QSUB_INST:
QSUB8_INST:
REV_INST:
{
INC_ICOUNTER;

68
src/core/arm/interpreter/armemu.cpp
View file

@ -5948,56 +5948,58 @@ L_stm_s_takeabort:
printf("Unhandled v6 insn: %08x", instr);
}
break;
case 0x62: // QADD16, QASX, QSAX, and QSUB16
if ((instr & 0xFF0) == 0xf10 || (instr & 0xFF0) == 0xf30 ||
(instr & 0xFF0) == 0xf50 || (instr & 0xFF0) == 0xf70)
case 0x62: // QADD16, QASX, QSAX, QSUB16, QADD8, and QSUB8
{
const u8 op2 = BITS(5, 7);
const u8 rd_idx = BITS(12, 15);
const u8 rn_idx = BITS(16, 19);
const u8 rm_idx = BITS(0, 3);
const s16 rm_lo = (state->Reg[rm_idx] & 0xFFFF);
const s16 rm_hi = ((state->Reg[rm_idx] >> 0x10) & 0xFFFF);
const s16 rn_lo = (state->Reg[rn_idx] & 0xFFFF);
const s16 rn_hi = ((state->Reg[rn_idx] >> 0x10) & 0xFFFF);
const u16 rm_lo = (state->Reg[rm_idx] & 0xFFFF);
const u16 rm_hi = ((state->Reg[rm_idx] >> 0x10) & 0xFFFF);
const u16 rn_lo = (state->Reg[rn_idx] & 0xFFFF);
const u16 rn_hi = ((state->Reg[rn_idx] >> 0x10) & 0xFFFF);
s32 lo_result;
s32 hi_result;
u16 lo_result = 0;
u16 hi_result = 0;
// QADD16
if ((instr & 0xFF0) == 0xf10) {
lo_result = (rn_lo + rm_lo);
hi_result = (rn_hi + rm_hi);
if (op2 == 0x00) {
lo_result = ARMul_SignedSaturatedAdd16(rn_lo, rm_lo);
hi_result = ARMul_SignedSaturatedAdd16(rn_hi, rm_hi);
}
// QASX
else if ((instr & 0xFF0) == 0xf30) {
lo_result = (rn_lo - rm_hi);
hi_result = (rn_hi + rm_lo);
else if (op2 == 0x01) {
lo_result = ARMul_SignedSaturatedSub16(rn_lo, rm_hi);
hi_result = ARMul_SignedSaturatedAdd16(rn_hi, rm_lo);
}
// QSAX
else if ((instr & 0xFF0) == 0xf50) {
lo_result = (rn_lo + rm_hi);
hi_result = (rn_hi - rm_lo);
else if (op2 == 0x02) {
lo_result = ARMul_SignedSaturatedAdd16(rn_lo, rm_hi);
hi_result = ARMul_SignedSaturatedSub16(rn_hi, rm_lo);
}
// QSUB16
else {
lo_result = (rn_lo - rm_lo);
hi_result = (rn_hi - rm_hi);
else if (op2 == 0x03) {
lo_result = ARMul_SignedSaturatedSub16(rn_lo, rm_lo);
hi_result = ARMul_SignedSaturatedSub16(rn_hi, rm_hi);
}
// QADD8
else if (op2 == 0x04) {
lo_result = ARMul_SignedSaturatedAdd8(rn_lo & 0xFF, rm_lo & 0xFF) |
ARMul_SignedSaturatedAdd8(rn_lo >> 8, rm_lo >> 8) << 8;
hi_result = ARMul_SignedSaturatedAdd8(rn_hi & 0xFF, rm_hi & 0xFF) |
ARMul_SignedSaturatedAdd8(rn_hi >> 8, rm_hi >> 8) << 8;
}
// QSUB8
else if (op2 == 0x07) {
lo_result = ARMul_SignedSaturatedSub8(rn_lo & 0xFF, rm_lo & 0xFF) |
ARMul_SignedSaturatedSub8(rn_lo >> 8, rm_lo >> 8) << 8;
hi_result = ARMul_SignedSaturatedSub8(rn_hi & 0xFF, rm_hi & 0xFF) |
ARMul_SignedSaturatedSub8(rn_hi >> 8, rm_hi >> 8) << 8;
}
if (lo_result > 0x7FFF)
lo_result = 0x7FFF;
else if (lo_result < -0x8000)
lo_result = -0x8000;
if (hi_result > 0x7FFF)
hi_result = 0x7FFF;
else if (hi_result < -0x8000)
hi_result = -0x8000;
state->Reg[rd_idx] = (lo_result & 0xFFFF) | ((hi_result & 0xFFFF) << 16);
return 1;
} else {
printf("Unhandled v6 insn: %08x", BITS(20, 27));
}
break;
case 0x63:

60
src/core/arm/interpreter/armsupp.cpp
View file

@ -478,6 +478,66 @@ ARMul_SubOverflow (ARMul_State * state, ARMword a, ARMword b, ARMword result)
ASSIGNV (SubOverflow (a, b, result));
}
/* 8-bit signed saturated addition */
u8 ARMul_SignedSaturatedAdd8(u8 left, u8 right)
{
u8 result = left + right;
if (((result ^ left) & 0x80) && ((left ^ right) & 0x80) == 0) {
if (left & 0x80)
result = 0x80;
else
result = 0x7F;
}
return result;
}
/* 8-bit signed saturated subtraction */
u8 ARMul_SignedSaturatedSub8(u8 left, u8 right)
{
u8 result = left - right;
if (((result ^ left) & 0x80) && ((left ^ right) & 0x80) != 0) {
if (left & 0x80)
result = 0x80;
else
result = 0x7F;
}
return result;
}
/* 16-bit signed saturated addition */
u16 ARMul_SignedSaturatedAdd16(u16 left, u16 right)
{
u16 result = left + right;
if (((result ^ left) & 0x8000) && ((left ^ right) & 0x8000) == 0) {
if (left & 0x8000)
result = 0x8000;
else
result = 0x7FFF;
}
return result;
}
/* 16-bit signed saturated subtraction */
u16 ARMul_SignedSaturatedSub16(u16 left, u16 right)
{
u16 result = left - right;
if (((result ^ left) & 0x8000) && ((left ^ right) & 0x8000) != 0) {
if (left & 0x8000)
result = 0x8000;
else
result = 0x7FFF;
}
return result;
}
/* 8-bit unsigned saturated addition */
u8 ARMul_UnsignedSaturatedAdd8(u8 left, u8 right)
{

5
src/core/arm/skyeye_common/armdefs.h
View file

@ -790,6 +790,11 @@ extern void ARMul_FixSPSR(ARMul_State*, ARMword, ARMword);
extern void ARMul_ConsolePrint(ARMul_State*, const char*, ...);
extern void ARMul_SelectProcessor(ARMul_State*, unsigned);
extern u8 ARMul_SignedSaturatedAdd8(u8, u8);
extern u8 ARMul_SignedSaturatedSub8(u8, u8);
extern u16 ARMul_SignedSaturatedAdd16(u16, u16);
extern u16 ARMul_SignedSaturatedSub16(u16, u16);
extern u8 ARMul_UnsignedSaturatedAdd8(u8, u8);
extern u16 ARMul_UnsignedSaturatedAdd16(u16, u16);
extern u8 ARMul_UnsignedSaturatedSub8(u8, u8);