mirrors/dynarmic
1
0
Fork 0
Code Issues Pull requests Projects Releases Packages Wiki Activity

backend/x64/emit_*: Apply const where applicable

Browse source
This commit is contained in:
Lioncash 2019-05-25 17:35:05 -04:00 committed by MerryMage
parent dd315e89eb
commit cba9351b82
5 changed files with 36 additions and 36 deletions
Show stats Download patch file Download diff file Expand all files Collapse all files

14
src/backend/x64/emit_x64_data_processing.cpp
View file

@ -38,7 +38,7 @@ void EmitX64::EmitPack2x64To1x128(EmitContext& ctx, IR::Inst* inst) {
code.movq(result, lo); code.movq(result, lo);
code.pinsrq(result, hi, 1); code.pinsrq(result, hi, 1);
} else { } else {
Xbyak::Xmm tmp = ctx.reg_alloc.ScratchXmm(); const Xbyak::Xmm tmp = ctx.reg_alloc.ScratchXmm();
code.movq(result, lo); code.movq(result, lo);
code.movq(tmp, hi); code.movq(tmp, hi);
code.punpcklqdq(result, tmp); code.punpcklqdq(result, tmp);
@ -53,7 +53,7 @@ void EmitX64::EmitLeastSignificantWord(EmitContext& ctx, IR::Inst* inst) {
} }
void EmitX64::EmitMostSignificantWord(EmitContext& ctx, IR::Inst* inst) { void EmitX64::EmitMostSignificantWord(EmitContext& ctx, IR::Inst* inst) {
auto carry_inst = inst->GetAssociatedPseudoOperation(IR::Opcode::GetCarryFromOp); const auto carry_inst = inst->GetAssociatedPseudoOperation(IR::Opcode::GetCarryFromOp);
auto args = ctx.reg_alloc.GetArgumentInfo(inst); auto args = ctx.reg_alloc.GetArgumentInfo(inst);
const Xbyak::Reg64 result = ctx.reg_alloc.UseScratchGpr(args[0]); const Xbyak::Reg64 result = ctx.reg_alloc.UseScratchGpr(args[0]);
@ -220,7 +220,7 @@ void EmitX64::EmitExtractRegister64(Dynarmic::BackendX64::EmitContext& ctx, IR::
} }
void EmitX64::EmitLogicalShiftLeft32(EmitContext& ctx, IR::Inst* inst) { void EmitX64::EmitLogicalShiftLeft32(EmitContext& ctx, IR::Inst* inst) {
auto carry_inst = inst->GetAssociatedPseudoOperation(IR::Opcode::GetCarryFromOp); const auto carry_inst = inst->GetAssociatedPseudoOperation(IR::Opcode::GetCarryFromOp);
auto args = ctx.reg_alloc.GetArgumentInfo(inst); auto args = ctx.reg_alloc.GetArgumentInfo(inst);
auto& operand_arg = args[0]; auto& operand_arg = args[0];
@ -353,7 +353,7 @@ void EmitX64::EmitLogicalShiftLeft64(EmitContext& ctx, IR::Inst* inst) {
} }
void EmitX64::EmitLogicalShiftRight32(EmitContext& ctx, IR::Inst* inst) { void EmitX64::EmitLogicalShiftRight32(EmitContext& ctx, IR::Inst* inst) {
auto carry_inst = inst->GetAssociatedPseudoOperation(IR::Opcode::GetCarryFromOp); const auto carry_inst = inst->GetAssociatedPseudoOperation(IR::Opcode::GetCarryFromOp);
auto args = ctx.reg_alloc.GetArgumentInfo(inst); auto args = ctx.reg_alloc.GetArgumentInfo(inst);
auto& operand_arg = args[0]; auto& operand_arg = args[0];
@ -485,7 +485,7 @@ void EmitX64::EmitLogicalShiftRight64(EmitContext& ctx, IR::Inst* inst) {
} }
void EmitX64::EmitArithmeticShiftRight32(EmitContext& ctx, IR::Inst* inst) { void EmitX64::EmitArithmeticShiftRight32(EmitContext& ctx, IR::Inst* inst) {
auto carry_inst = inst->GetAssociatedPseudoOperation(IR::Opcode::GetCarryFromOp); const auto carry_inst = inst->GetAssociatedPseudoOperation(IR::Opcode::GetCarryFromOp);
auto args = ctx.reg_alloc.GetArgumentInfo(inst); auto args = ctx.reg_alloc.GetArgumentInfo(inst);
auto& operand_arg = args[0]; auto& operand_arg = args[0];
@ -604,7 +604,7 @@ void EmitX64::EmitArithmeticShiftRight64(EmitContext& ctx, IR::Inst* inst) {
} }
void EmitX64::EmitRotateRight32(EmitContext& ctx, IR::Inst* inst) { void EmitX64::EmitRotateRight32(EmitContext& ctx, IR::Inst* inst) {
auto carry_inst = inst->GetAssociatedPseudoOperation(IR::Opcode::GetCarryFromOp); const auto carry_inst = inst->GetAssociatedPseudoOperation(IR::Opcode::GetCarryFromOp);
auto args = ctx.reg_alloc.GetArgumentInfo(inst); auto args = ctx.reg_alloc.GetArgumentInfo(inst);
auto& operand_arg = args[0]; auto& operand_arg = args[0];
@ -706,7 +706,7 @@ void EmitX64::EmitRotateRight64(EmitContext& ctx, IR::Inst* inst) {
} }
void EmitX64::EmitRotateRightExtended(EmitContext& ctx, IR::Inst* inst) { void EmitX64::EmitRotateRightExtended(EmitContext& ctx, IR::Inst* inst) {
auto carry_inst = inst->GetAssociatedPseudoOperation(IR::Opcode::GetCarryFromOp); const auto carry_inst = inst->GetAssociatedPseudoOperation(IR::Opcode::GetCarryFromOp);
auto args = ctx.reg_alloc.GetArgumentInfo(inst); auto args = ctx.reg_alloc.GetArgumentInfo(inst);
const Xbyak::Reg32 result = ctx.reg_alloc.UseScratchGpr(args[0]).cvt32(); const Xbyak::Reg32 result = ctx.reg_alloc.UseScratchGpr(args[0]).cvt32();

18
src/backend/x64/emit_x64_floating_point.cpp
View file

@ -1013,7 +1013,7 @@ void EmitX64::EmitFPSub64(EmitContext& ctx, IR::Inst* inst) {
static Xbyak::Reg64 SetFpscrNzcvFromFlags(BlockOfCode& code, EmitContext& ctx) { static Xbyak::Reg64 SetFpscrNzcvFromFlags(BlockOfCode& code, EmitContext& ctx) {
ctx.reg_alloc.ScratchGpr({HostLoc::RCX}); // shifting requires use of cl ctx.reg_alloc.ScratchGpr({HostLoc::RCX}); // shifting requires use of cl
Xbyak::Reg64 nzcv = ctx.reg_alloc.ScratchGpr(); const Xbyak::Reg64 nzcv = ctx.reg_alloc.ScratchGpr();
// x64 flags ARM flags // x64 flags ARM flags
// ZF PF CF NZCV // ZF PF CF NZCV
@ -1040,9 +1040,9 @@ static Xbyak::Reg64 SetFpscrNzcvFromFlags(BlockOfCode& code, EmitContext& ctx) {
void EmitX64::EmitFPCompare32(EmitContext& ctx, IR::Inst* inst) { void EmitX64::EmitFPCompare32(EmitContext& ctx, IR::Inst* inst) {
auto args = ctx.reg_alloc.GetArgumentInfo(inst); auto args = ctx.reg_alloc.GetArgumentInfo(inst);
Xbyak::Xmm reg_a = ctx.reg_alloc.UseXmm(args[0]); const Xbyak::Xmm reg_a = ctx.reg_alloc.UseXmm(args[0]);
Xbyak::Xmm reg_b = ctx.reg_alloc.UseXmm(args[1]); const Xbyak::Xmm reg_b = ctx.reg_alloc.UseXmm(args[1]);
bool exc_on_qnan = args[2].GetImmediateU1(); const bool exc_on_qnan = args[2].GetImmediateU1();
if (exc_on_qnan) { if (exc_on_qnan) {
code.comiss(reg_a, reg_b); code.comiss(reg_a, reg_b);
@ -1050,15 +1050,15 @@ void EmitX64::EmitFPCompare32(EmitContext& ctx, IR::Inst* inst) {
code.ucomiss(reg_a, reg_b); code.ucomiss(reg_a, reg_b);
} }
Xbyak::Reg64 nzcv = SetFpscrNzcvFromFlags(code, ctx); const Xbyak::Reg64 nzcv = SetFpscrNzcvFromFlags(code, ctx);
ctx.reg_alloc.DefineValue(inst, nzcv); ctx.reg_alloc.DefineValue(inst, nzcv);
} }
void EmitX64::EmitFPCompare64(EmitContext& ctx, IR::Inst* inst) { void EmitX64::EmitFPCompare64(EmitContext& ctx, IR::Inst* inst) {
auto args = ctx.reg_alloc.GetArgumentInfo(inst); auto args = ctx.reg_alloc.GetArgumentInfo(inst);
Xbyak::Xmm reg_a = ctx.reg_alloc.UseXmm(args[0]); const Xbyak::Xmm reg_a = ctx.reg_alloc.UseXmm(args[0]);
Xbyak::Xmm reg_b = ctx.reg_alloc.UseXmm(args[1]); const Xbyak::Xmm reg_b = ctx.reg_alloc.UseXmm(args[1]);
bool exc_on_qnan = args[2].GetImmediateU1(); const bool exc_on_qnan = args[2].GetImmediateU1();
if (exc_on_qnan) { if (exc_on_qnan) {
code.comisd(reg_a, reg_b); code.comisd(reg_a, reg_b);
@ -1066,7 +1066,7 @@ void EmitX64::EmitFPCompare64(EmitContext& ctx, IR::Inst* inst) {
code.ucomisd(reg_a, reg_b); code.ucomisd(reg_a, reg_b);
} }
Xbyak::Reg64 nzcv = SetFpscrNzcvFromFlags(code, ctx); const Xbyak::Reg64 nzcv = SetFpscrNzcvFromFlags(code, ctx);
ctx.reg_alloc.DefineValue(inst, nzcv); ctx.reg_alloc.DefineValue(inst, nzcv);
} }

18
src/backend/x64/emit_x64_packed.cpp
View file

@ -18,7 +18,7 @@ using namespace Xbyak::util;
void EmitX64::EmitPackedAddU8(EmitContext& ctx, IR::Inst* inst) { void EmitX64::EmitPackedAddU8(EmitContext& ctx, IR::Inst* inst) {
auto args = ctx.reg_alloc.GetArgumentInfo(inst); auto args = ctx.reg_alloc.GetArgumentInfo(inst);
auto ge_inst = inst->GetAssociatedPseudoOperation(IR::Opcode::GetGEFromOp); const auto ge_inst = inst->GetAssociatedPseudoOperation(IR::Opcode::GetGEFromOp);
const Xbyak::Xmm xmm_a = ctx.reg_alloc.UseScratchXmm(args[0]); const Xbyak::Xmm xmm_a = ctx.reg_alloc.UseScratchXmm(args[0]);
const Xbyak::Xmm xmm_b = ctx.reg_alloc.UseXmm(args[1]); const Xbyak::Xmm xmm_b = ctx.reg_alloc.UseXmm(args[1]);
@ -45,7 +45,7 @@ void EmitX64::EmitPackedAddU8(EmitContext& ctx, IR::Inst* inst) {
void EmitX64::EmitPackedAddS8(EmitContext& ctx, IR::Inst* inst) { void EmitX64::EmitPackedAddS8(EmitContext& ctx, IR::Inst* inst) {
auto args = ctx.reg_alloc.GetArgumentInfo(inst); auto args = ctx.reg_alloc.GetArgumentInfo(inst);
auto ge_inst = inst->GetAssociatedPseudoOperation(IR::Opcode::GetGEFromOp); const auto ge_inst = inst->GetAssociatedPseudoOperation(IR::Opcode::GetGEFromOp);
const Xbyak::Xmm xmm_a = ctx.reg_alloc.UseScratchXmm(args[0]); const Xbyak::Xmm xmm_a = ctx.reg_alloc.UseScratchXmm(args[0]);
const Xbyak::Xmm xmm_b = ctx.reg_alloc.UseXmm(args[1]); const Xbyak::Xmm xmm_b = ctx.reg_alloc.UseXmm(args[1]);
@ -72,7 +72,7 @@ void EmitX64::EmitPackedAddS8(EmitContext& ctx, IR::Inst* inst) {
void EmitX64::EmitPackedAddU16(EmitContext& ctx, IR::Inst* inst) { void EmitX64::EmitPackedAddU16(EmitContext& ctx, IR::Inst* inst) {
auto args = ctx.reg_alloc.GetArgumentInfo(inst); auto args = ctx.reg_alloc.GetArgumentInfo(inst);
auto ge_inst = inst->GetAssociatedPseudoOperation(IR::Opcode::GetGEFromOp); const auto ge_inst = inst->GetAssociatedPseudoOperation(IR::Opcode::GetGEFromOp);
const Xbyak::Xmm xmm_a = ctx.reg_alloc.UseScratchXmm(args[0]); const Xbyak::Xmm xmm_a = ctx.reg_alloc.UseScratchXmm(args[0]);
const Xbyak::Xmm xmm_b = ctx.reg_alloc.UseXmm(args[1]); const Xbyak::Xmm xmm_b = ctx.reg_alloc.UseXmm(args[1]);
@ -114,7 +114,7 @@ void EmitX64::EmitPackedAddU16(EmitContext& ctx, IR::Inst* inst) {
void EmitX64::EmitPackedAddS16(EmitContext& ctx, IR::Inst* inst) { void EmitX64::EmitPackedAddS16(EmitContext& ctx, IR::Inst* inst) {
auto args = ctx.reg_alloc.GetArgumentInfo(inst); auto args = ctx.reg_alloc.GetArgumentInfo(inst);
auto ge_inst = inst->GetAssociatedPseudoOperation(IR::Opcode::GetGEFromOp); const auto ge_inst = inst->GetAssociatedPseudoOperation(IR::Opcode::GetGEFromOp);
const Xbyak::Xmm xmm_a = ctx.reg_alloc.UseScratchXmm(args[0]); const Xbyak::Xmm xmm_a = ctx.reg_alloc.UseScratchXmm(args[0]);
const Xbyak::Xmm xmm_b = ctx.reg_alloc.UseXmm(args[1]); const Xbyak::Xmm xmm_b = ctx.reg_alloc.UseXmm(args[1]);
@ -141,7 +141,7 @@ void EmitX64::EmitPackedAddS16(EmitContext& ctx, IR::Inst* inst) {
void EmitX64::EmitPackedSubU8(EmitContext& ctx, IR::Inst* inst) { void EmitX64::EmitPackedSubU8(EmitContext& ctx, IR::Inst* inst) {
auto args = ctx.reg_alloc.GetArgumentInfo(inst); auto args = ctx.reg_alloc.GetArgumentInfo(inst);
auto ge_inst = inst->GetAssociatedPseudoOperation(IR::Opcode::GetGEFromOp); const auto ge_inst = inst->GetAssociatedPseudoOperation(IR::Opcode::GetGEFromOp);
const Xbyak::Xmm xmm_a = ctx.reg_alloc.UseScratchXmm(args[0]); const Xbyak::Xmm xmm_a = ctx.reg_alloc.UseScratchXmm(args[0]);
const Xbyak::Xmm xmm_b = ctx.reg_alloc.UseXmm(args[1]); const Xbyak::Xmm xmm_b = ctx.reg_alloc.UseXmm(args[1]);
@ -164,7 +164,7 @@ void EmitX64::EmitPackedSubU8(EmitContext& ctx, IR::Inst* inst) {
void EmitX64::EmitPackedSubS8(EmitContext& ctx, IR::Inst* inst) { void EmitX64::EmitPackedSubS8(EmitContext& ctx, IR::Inst* inst) {
auto args = ctx.reg_alloc.GetArgumentInfo(inst); auto args = ctx.reg_alloc.GetArgumentInfo(inst);
auto ge_inst = inst->GetAssociatedPseudoOperation(IR::Opcode::GetGEFromOp); const auto ge_inst = inst->GetAssociatedPseudoOperation(IR::Opcode::GetGEFromOp);
const Xbyak::Xmm xmm_a = ctx.reg_alloc.UseScratchXmm(args[0]); const Xbyak::Xmm xmm_a = ctx.reg_alloc.UseScratchXmm(args[0]);
const Xbyak::Xmm xmm_b = ctx.reg_alloc.UseXmm(args[1]); const Xbyak::Xmm xmm_b = ctx.reg_alloc.UseXmm(args[1]);
@ -191,7 +191,7 @@ void EmitX64::EmitPackedSubS8(EmitContext& ctx, IR::Inst* inst) {
void EmitX64::EmitPackedSubU16(EmitContext& ctx, IR::Inst* inst) { void EmitX64::EmitPackedSubU16(EmitContext& ctx, IR::Inst* inst) {
auto args = ctx.reg_alloc.GetArgumentInfo(inst); auto args = ctx.reg_alloc.GetArgumentInfo(inst);
auto ge_inst = inst->GetAssociatedPseudoOperation(IR::Opcode::GetGEFromOp); const auto ge_inst = inst->GetAssociatedPseudoOperation(IR::Opcode::GetGEFromOp);
if (!ge_inst) { if (!ge_inst) {
const Xbyak::Xmm xmm_a = ctx.reg_alloc.UseScratchXmm(args[0]); const Xbyak::Xmm xmm_a = ctx.reg_alloc.UseScratchXmm(args[0]);
@ -242,7 +242,7 @@ void EmitX64::EmitPackedSubU16(EmitContext& ctx, IR::Inst* inst) {
void EmitX64::EmitPackedSubS16(EmitContext& ctx, IR::Inst* inst) { void EmitX64::EmitPackedSubS16(EmitContext& ctx, IR::Inst* inst) {
auto args = ctx.reg_alloc.GetArgumentInfo(inst); auto args = ctx.reg_alloc.GetArgumentInfo(inst);
auto ge_inst = inst->GetAssociatedPseudoOperation(IR::Opcode::GetGEFromOp); const auto ge_inst = inst->GetAssociatedPseudoOperation(IR::Opcode::GetGEFromOp);
const Xbyak::Xmm xmm_a = ctx.reg_alloc.UseScratchXmm(args[0]); const Xbyak::Xmm xmm_a = ctx.reg_alloc.UseScratchXmm(args[0]);
const Xbyak::Xmm xmm_b = ctx.reg_alloc.UseXmm(args[1]); const Xbyak::Xmm xmm_b = ctx.reg_alloc.UseXmm(args[1]);
@ -512,7 +512,7 @@ void EmitX64::EmitPackedHalvingSubS16(EmitContext& ctx, IR::Inst* inst) {
void EmitPackedSubAdd(BlockOfCode& code, EmitContext& ctx, IR::Inst* inst, bool hi_is_sum, bool is_signed, bool is_halving) { void EmitPackedSubAdd(BlockOfCode& code, EmitContext& ctx, IR::Inst* inst, bool hi_is_sum, bool is_signed, bool is_halving) {
auto args = ctx.reg_alloc.GetArgumentInfo(inst); auto args = ctx.reg_alloc.GetArgumentInfo(inst);
auto ge_inst = inst->GetAssociatedPseudoOperation(IR::Opcode::GetGEFromOp); const auto ge_inst = inst->GetAssociatedPseudoOperation(IR::Opcode::GetGEFromOp);
const Xbyak::Reg32 reg_a_hi = ctx.reg_alloc.UseScratchGpr(args[0]).cvt32(); const Xbyak::Reg32 reg_a_hi = ctx.reg_alloc.UseScratchGpr(args[0]).cvt32();
const Xbyak::Reg32 reg_b_hi = ctx.reg_alloc.UseScratchGpr(args[1]).cvt32(); const Xbyak::Reg32 reg_b_hi = ctx.reg_alloc.UseScratchGpr(args[1]).cvt32();

12
src/backend/x64/emit_x64_saturation.cpp
View file

@ -30,7 +30,7 @@ enum class Op {
template<Op op, size_t size> template<Op op, size_t size>
void EmitSignedSaturatedOp(BlockOfCode& code, EmitContext& ctx, IR::Inst* inst) { void EmitSignedSaturatedOp(BlockOfCode& code, EmitContext& ctx, IR::Inst* inst) {
auto overflow_inst = inst->GetAssociatedPseudoOperation(IR::Opcode::GetOverflowFromOp); const auto overflow_inst = inst->GetAssociatedPseudoOperation(IR::Opcode::GetOverflowFromOp);
auto args = ctx.reg_alloc.GetArgumentInfo(inst); auto args = ctx.reg_alloc.GetArgumentInfo(inst);
@ -79,7 +79,7 @@ void EmitSignedSaturatedOp(BlockOfCode& code, EmitContext& ctx, IR::Inst* inst)
template<Op op, size_t size> template<Op op, size_t size>
void EmitUnsignedSaturatedOp(BlockOfCode& code, EmitContext& ctx, IR::Inst* inst) { void EmitUnsignedSaturatedOp(BlockOfCode& code, EmitContext& ctx, IR::Inst* inst) {
auto overflow_inst = inst->GetAssociatedPseudoOperation(IR::Opcode::GetOverflowFromOp); const auto overflow_inst = inst->GetAssociatedPseudoOperation(IR::Opcode::GetOverflowFromOp);
auto args = ctx.reg_alloc.GetArgumentInfo(inst); auto args = ctx.reg_alloc.GetArgumentInfo(inst);
@ -135,7 +135,7 @@ void EmitX64::EmitSignedSaturatedAdd64(EmitContext& ctx, IR::Inst* inst) {
} }
void EmitX64::EmitSignedSaturatedDoublingMultiplyReturnHigh16(EmitContext& ctx, IR::Inst* inst) { void EmitX64::EmitSignedSaturatedDoublingMultiplyReturnHigh16(EmitContext& ctx, IR::Inst* inst) {
auto overflow_inst = inst->GetAssociatedPseudoOperation(IR::Opcode::GetOverflowFromOp); const auto overflow_inst = inst->GetAssociatedPseudoOperation(IR::Opcode::GetOverflowFromOp);
auto args = ctx.reg_alloc.GetArgumentInfo(inst); auto args = ctx.reg_alloc.GetArgumentInfo(inst);
@ -165,7 +165,7 @@ void EmitX64::EmitSignedSaturatedDoublingMultiplyReturnHigh16(EmitContext& ctx,
} }
void EmitX64::EmitSignedSaturatedDoublingMultiplyReturnHigh32(EmitContext& ctx, IR::Inst* inst) { void EmitX64::EmitSignedSaturatedDoublingMultiplyReturnHigh32(EmitContext& ctx, IR::Inst* inst) {
auto overflow_inst = inst->GetAssociatedPseudoOperation(IR::Opcode::GetOverflowFromOp); const auto overflow_inst = inst->GetAssociatedPseudoOperation(IR::Opcode::GetOverflowFromOp);
auto args = ctx.reg_alloc.GetArgumentInfo(inst); auto args = ctx.reg_alloc.GetArgumentInfo(inst);
@ -211,7 +211,7 @@ void EmitX64::EmitSignedSaturatedSub64(EmitContext& ctx, IR::Inst* inst) {
} }
void EmitX64::EmitSignedSaturation(EmitContext& ctx, IR::Inst* inst) { void EmitX64::EmitSignedSaturation(EmitContext& ctx, IR::Inst* inst) {
auto overflow_inst = inst->GetAssociatedPseudoOperation(IR::Opcode::GetOverflowFromOp); const auto overflow_inst = inst->GetAssociatedPseudoOperation(IR::Opcode::GetOverflowFromOp);
auto args = ctx.reg_alloc.GetArgumentInfo(inst); auto args = ctx.reg_alloc.GetArgumentInfo(inst);
const size_t N = args[1].GetImmediateU8(); const size_t N = args[1].GetImmediateU8();
@ -292,7 +292,7 @@ void EmitX64::EmitUnsignedSaturatedSub64(EmitContext& ctx, IR::Inst* inst) {
} }
void EmitX64::EmitUnsignedSaturation(EmitContext& ctx, IR::Inst* inst) { void EmitX64::EmitUnsignedSaturation(EmitContext& ctx, IR::Inst* inst) {
auto overflow_inst = inst->GetAssociatedPseudoOperation(IR::Opcode::GetOverflowFromOp); const auto overflow_inst = inst->GetAssociatedPseudoOperation(IR::Opcode::GetOverflowFromOp);
auto args = ctx.reg_alloc.GetArgumentInfo(inst); auto args = ctx.reg_alloc.GetArgumentInfo(inst);
const size_t N = args[1].GetImmediateU8(); const size_t N = args[1].GetImmediateU8();

10
src/backend/x64/emit_x64_vector.cpp
View file

@ -467,7 +467,7 @@ void EmitX64::EmitVectorArithmeticShiftRight8(EmitContext& ctx, IR::Inst* inst)
void EmitX64::EmitVectorArithmeticShiftRight16(EmitContext& ctx, IR::Inst* inst) { void EmitX64::EmitVectorArithmeticShiftRight16(EmitContext& ctx, IR::Inst* inst) {
auto args = ctx.reg_alloc.GetArgumentInfo(inst); auto args = ctx.reg_alloc.GetArgumentInfo(inst);
Xbyak::Xmm result = ctx.reg_alloc.UseScratchXmm(args[0]); const Xbyak::Xmm result = ctx.reg_alloc.UseScratchXmm(args[0]);
const u8 shift_amount = args[1].GetImmediateU8(); const u8 shift_amount = args[1].GetImmediateU8();
code.psraw(result, shift_amount); code.psraw(result, shift_amount);
@ -478,7 +478,7 @@ void EmitX64::EmitVectorArithmeticShiftRight16(EmitContext& ctx, IR::Inst* inst)
void EmitX64::EmitVectorArithmeticShiftRight32(EmitContext& ctx, IR::Inst* inst) { void EmitX64::EmitVectorArithmeticShiftRight32(EmitContext& ctx, IR::Inst* inst) {
auto args = ctx.reg_alloc.GetArgumentInfo(inst); auto args = ctx.reg_alloc.GetArgumentInfo(inst);
Xbyak::Xmm result = ctx.reg_alloc.UseScratchXmm(args[0]); const Xbyak::Xmm result = ctx.reg_alloc.UseScratchXmm(args[0]);
const u8 shift_amount = args[1].GetImmediateU8(); const u8 shift_amount = args[1].GetImmediateU8();
code.psrad(result, shift_amount); code.psrad(result, shift_amount);
@ -1001,9 +1001,9 @@ void EmitX64::EmitVectorEqual64(EmitContext& ctx, IR::Inst* inst) {
auto args = ctx.reg_alloc.GetArgumentInfo(inst); auto args = ctx.reg_alloc.GetArgumentInfo(inst);
Xbyak::Xmm xmm_a = ctx.reg_alloc.UseScratchXmm(args[0]); const Xbyak::Xmm xmm_a = ctx.reg_alloc.UseScratchXmm(args[0]);
Xbyak::Xmm xmm_b = ctx.reg_alloc.UseXmm(args[1]); const Xbyak::Xmm xmm_b = ctx.reg_alloc.UseXmm(args[1]);
Xbyak::Xmm tmp = ctx.reg_alloc.ScratchXmm(); const Xbyak::Xmm tmp = ctx.reg_alloc.ScratchXmm();
code.pcmpeqd(xmm_a, xmm_b); code.pcmpeqd(xmm_a, xmm_b);
code.pshufd(tmp, xmm_a, 0b10110001); code.pshufd(tmp, xmm_a, 0b10110001);