suyu/dynarmic
26
1
Fork 1
Code Issues Pull requests Projects Releases Packages Wiki Activity

emit_x64_vector: Add SSSE3 implementation of VUZP{1,2}.8B

Browse source
This commit is contained in:
zmt00 2023-12-14 18:06:21 -08:00 committed by merry
parent 521bf64ef2
commit 8ef0f2b54f
1 changed files with 30 additions and 14 deletions
Show stats Download patch file Download diff file Expand all files Collapse all files

18
src/dynarmic/backend/x64/emit_x64_vector.cpp
View file

@ -1122,8 +1122,15 @@ void EmitX64::EmitVectorDeinterleaveEven64(EmitContext& ctx, IR::Inst* inst) {
void EmitX64::EmitVectorDeinterleaveEvenLower8(EmitContext& ctx, IR::Inst* inst) { void EmitX64::EmitVectorDeinterleaveEvenLower8(EmitContext& ctx, IR::Inst* inst) {
auto args = ctx.reg_alloc.GetArgumentInfo(inst); auto args = ctx.reg_alloc.GetArgumentInfo(inst);
const Xbyak::Xmm lhs = ctx.reg_alloc.UseScratchXmm(args[0]); const Xbyak::Xmm lhs = ctx.reg_alloc.UseScratchXmm(args[0]);
const Xbyak::Xmm rhs = ctx.reg_alloc.UseScratchXmm(args[1]);
if (code.HasHostFeature(HostFeature::SSSE3)) {
const Xbyak::Xmm rhs = ctx.reg_alloc.UseXmm(args[1]);
code.punpcklbw(lhs, rhs);
code.pshufb(lhs, code.MConst(xword, 0x0D'09'05'01'0C'08'04'00, 0x8080808080808080));
} else {
const Xbyak::Xmm tmp = ctx.reg_alloc.ScratchXmm(); const Xbyak::Xmm tmp = ctx.reg_alloc.ScratchXmm();
const Xbyak::Xmm rhs = ctx.reg_alloc.UseScratchXmm(args[1]);
code.movdqa(tmp, code.MConst(xword, 0x00FF00FF00FF00FF, 0x00FF00FF00FF00FF)); code.movdqa(tmp, code.MConst(xword, 0x00FF00FF00FF00FF, 0x00FF00FF00FF00FF));
code.pand(lhs, tmp); code.pand(lhs, tmp);
@ -1131,6 +1138,7 @@ void EmitX64::EmitVectorDeinterleaveEvenLower8(EmitContext& ctx, IR::Inst* inst)
code.packuswb(lhs, rhs); code.packuswb(lhs, rhs);
code.pshufd(lhs, lhs, 0b11011000); code.pshufd(lhs, lhs, 0b11011000);
code.movq(lhs, lhs); code.movq(lhs, lhs);
}
ctx.reg_alloc.DefineValue(inst, lhs); ctx.reg_alloc.DefineValue(inst, lhs);
} }
@ -1224,6 +1232,13 @@ void EmitX64::EmitVectorDeinterleaveOdd64(EmitContext& ctx, IR::Inst* inst) {
void EmitX64::EmitVectorDeinterleaveOddLower8(EmitContext& ctx, IR::Inst* inst) { void EmitX64::EmitVectorDeinterleaveOddLower8(EmitContext& ctx, IR::Inst* inst) {
auto args = ctx.reg_alloc.GetArgumentInfo(inst); auto args = ctx.reg_alloc.GetArgumentInfo(inst);
const Xbyak::Xmm lhs = ctx.reg_alloc.UseScratchXmm(args[0]); const Xbyak::Xmm lhs = ctx.reg_alloc.UseScratchXmm(args[0]);
if (code.HasHostFeature(HostFeature::SSSE3)) {
const Xbyak::Xmm rhs = ctx.reg_alloc.UseXmm(args[1]);
code.punpcklbw(lhs, rhs);
code.pshufb(lhs, code.MConst(xword, 0x0F'0B'07'03'0E'0A'06'02, 0x8080808080808080));
} else {
const Xbyak::Xmm rhs = ctx.reg_alloc.UseScratchXmm(args[1]); const Xbyak::Xmm rhs = ctx.reg_alloc.UseScratchXmm(args[1]);
code.psraw(lhs, 8); code.psraw(lhs, 8);
@ -1231,6 +1246,7 @@ void EmitX64::EmitVectorDeinterleaveOddLower8(EmitContext& ctx, IR::Inst* inst)
code.packsswb(lhs, rhs); code.packsswb(lhs, rhs);
code.pshufd(lhs, lhs, 0b11011000); code.pshufd(lhs, lhs, 0b11011000);
code.movq(lhs, lhs); code.movq(lhs, lhs);
}
ctx.reg_alloc.DefineValue(inst, lhs); ctx.reg_alloc.DefineValue(inst, lhs);
} }