suyu/dynarmic
28
1
Fork 1
Code Issues Pull requests Projects Releases Packages Wiki Activity

emit_x64_vector: Add SSE4.1 implementations for VectorZeroExtend

Browse source
This commit is contained in:
MerryMage 2018-02-10 16:24:43 +00:00
parent cbc9f361b0
commit 69de50a878
1 changed files with 32 additions and 25 deletions
Show stats Download patch file Download diff file Expand all files Collapse all files

53
src/backend_x64/emit_x64_vector.cpp
View file

@ -703,45 +703,52 @@ void EmitX64::EmitVectorSub64(EmitContext& ctx, IR::Inst* inst) {
EmitVectorOperation(code, ctx, inst, &Xbyak::CodeGenerator::psubq); EmitVectorOperation(code, ctx, inst, &Xbyak::CodeGenerator::psubq);
} }
static void EmitVectorZeroExtend(BlockOfCode& code, EmitContext& ctx, IR::Inst* inst, int size) { void EmitX64::EmitVectorZeroExtend8(EmitContext& ctx, IR::Inst* inst) {
auto args = ctx.reg_alloc.GetArgumentInfo(inst); auto args = ctx.reg_alloc.GetArgumentInfo(inst);
const Xbyak::Xmm a = ctx.reg_alloc.UseScratchXmm(args[0]); const Xbyak::Xmm a = ctx.reg_alloc.UseScratchXmm(args[0]);
if (code.DoesCpuSupport(Xbyak::util::Cpu::tSSE41)) {
code.pmovzxbw(a, a);
} else {
const Xbyak::Xmm zeros = ctx.reg_alloc.ScratchXmm(); const Xbyak::Xmm zeros = ctx.reg_alloc.ScratchXmm();
code.pxor(zeros, zeros); code.pxor(zeros, zeros);
switch (size) {
case 8:
code.punpcklbw(a, zeros); code.punpcklbw(a, zeros);
break;
case 16:
code.punpcklwd(a, zeros);
break;
case 32:
code.punpckldq(a, zeros);
break;
case 64:
code.punpcklqdq(a, zeros);
break;
} }
ctx.reg_alloc.DefineValue(inst, a); ctx.reg_alloc.DefineValue(inst, a);
} }
void EmitX64::EmitVectorZeroExtend8(EmitContext& ctx, IR::Inst* inst) {
EmitVectorZeroExtend(code, ctx, inst, 8);
}
void EmitX64::EmitVectorZeroExtend16(EmitContext& ctx, IR::Inst* inst) { void EmitX64::EmitVectorZeroExtend16(EmitContext& ctx, IR::Inst* inst) {
EmitVectorZeroExtend(code, ctx, inst, 16); auto args = ctx.reg_alloc.GetArgumentInfo(inst);
const Xbyak::Xmm a = ctx.reg_alloc.UseScratchXmm(args[0]);
if (code.DoesCpuSupport(Xbyak::util::Cpu::tSSE41)) {
code.pmovzxwd(a, a);
} else {
const Xbyak::Xmm zeros = ctx.reg_alloc.ScratchXmm();
code.pxor(zeros, zeros);
code.punpcklwd(a, zeros);
}
ctx.reg_alloc.DefineValue(inst, a);
} }
void EmitX64::EmitVectorZeroExtend32(EmitContext& ctx, IR::Inst* inst) { void EmitX64::EmitVectorZeroExtend32(EmitContext& ctx, IR::Inst* inst) {
EmitVectorZeroExtend(code, ctx, inst, 32); auto args = ctx.reg_alloc.GetArgumentInfo(inst);
const Xbyak::Xmm a = ctx.reg_alloc.UseScratchXmm(args[0]);
if (code.DoesCpuSupport(Xbyak::util::Cpu::tSSE41)) {
code.pmovzxdq(a, a);
} else {
const Xbyak::Xmm zeros = ctx.reg_alloc.ScratchXmm();
code.pxor(zeros, zeros);
code.punpckldq(a, zeros);
}
ctx.reg_alloc.DefineValue(inst, a);
} }
void EmitX64::EmitVectorZeroExtend64(EmitContext& ctx, IR::Inst* inst) { void EmitX64::EmitVectorZeroExtend64(EmitContext& ctx, IR::Inst* inst) {
EmitVectorZeroExtend(code, ctx, inst, 64); auto args = ctx.reg_alloc.GetArgumentInfo(inst);
const Xbyak::Xmm a = ctx.reg_alloc.UseScratchXmm(args[0]);
const Xbyak::Xmm zeros = ctx.reg_alloc.ScratchXmm();
code.pxor(zeros, zeros);
code.punpcklqdq(a, zeros);
ctx.reg_alloc.DefineValue(inst, a);
} }
void EmitX64::EmitVectorZeroUpper(EmitContext& ctx, IR::Inst* inst) { void EmitX64::EmitVectorZeroUpper(EmitContext& ctx, IR::Inst* inst) {