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

emit_x64_vector: Vectorize non-SSE4.1 fallback path for VectorMultiply32()

Browse source
This commit is contained in:
Lioncash 2018-04-25 22:44:21 -04:00 • committed by MerryMage
parent d70ee7c0d1
commit 0d20423ad5
1 changed files with 15 additions and 3 deletions
Show stats Download patch file Download diff file Expand all files Collapse all files

18
src/backend_x64/emit_x64_vector.cpp
View file

@ -1117,9 +1117,21 @@ void EmitX64::EmitVectorMultiply32(EmitContext& ctx, IR::Inst* inst) {
return; return;
} }
EmitTwoArgumentFallback(code, ctx, inst, [](std::array<u32, 4>& result, const std::array<u32, 4>& a, const std::array<u32, 4>& b){ auto args = ctx.reg_alloc.GetArgumentInfo(inst);
std::transform(a.begin(), a.end(), b.begin(), result.begin(), std::multiplies<>()); const Xbyak::Xmm a = ctx.reg_alloc.UseScratchXmm(args[0]);
}); const Xbyak::Xmm b = ctx.reg_alloc.UseScratchXmm(args[1]);
const Xbyak::Xmm tmp = ctx.reg_alloc.ScratchXmm();
code.movdqa(tmp, a);
code.psrlq(a, 32);
code.pmuludq(tmp, b);
code.psrlq(b, 32);
code.pmuludq(a, b);
code.pshufd(tmp, tmp, 0b00001000);
code.pshufd(b, a, 0b00001000);
code.punpckldq(tmp, b);
ctx.reg_alloc.DefineValue(inst, tmp);
} }
void EmitX64::EmitVectorMultiply64(EmitContext& ctx, IR::Inst* inst) { void EmitX64::EmitVectorMultiply64(EmitContext& ctx, IR::Inst* inst) {