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

emit_x64_vector: SSSE3 implementation of EmitVectorExtract

Browse source
This commit is contained in:
MerryMage 2020-06-01 15:41:29 +01:00
parent f3845cea9a
commit b47adaee1d
1 changed files with 17 additions and 6 deletions
Show stats Download patch file Download diff file Expand all files Collapse all files

23
src/backend/x64/emit_x64_vector.cpp
View file

@ -1041,19 +1041,30 @@ void EmitX64::EmitVectorEqual128(EmitContext& ctx, IR::Inst* inst) {
void EmitX64::EmitVectorExtract(EmitContext& ctx, IR::Inst* inst) { void EmitX64::EmitVectorExtract(EmitContext& ctx, IR::Inst* inst) {
auto args = ctx.reg_alloc.GetArgumentInfo(inst); auto args = ctx.reg_alloc.GetArgumentInfo(inst);
const Xbyak::Xmm xmm_a = ctx.reg_alloc.UseScratchXmm(args[0]);
const u8 position = args[2].GetImmediateU8(); const u8 position = args[2].GetImmediateU8();
ASSERT(position % 8 == 0); ASSERT(position % 8 == 0);
if (position != 0) { if (position == 0) {
ctx.reg_alloc.DefineValue(inst, args[0]);
return;
}
if (code.DoesCpuSupport(Xbyak::util::Cpu::tSSSE3)) {
const Xbyak::Xmm xmm_a = ctx.reg_alloc.UseXmm(args[0]);
const Xbyak::Xmm xmm_b = ctx.reg_alloc.UseScratchXmm(args[1]); const Xbyak::Xmm xmm_b = ctx.reg_alloc.UseScratchXmm(args[1]);
code.psrldq(xmm_a, position / 8); code.palignr(xmm_b, xmm_a, position / 8);
code.pslldq(xmm_b, (128 - position) / 8); ctx.reg_alloc.DefineValue(inst, xmm_b);
code.por(xmm_a, xmm_b); return;
} }
const Xbyak::Xmm xmm_a = ctx.reg_alloc.UseScratchXmm(args[0]);
const Xbyak::Xmm xmm_b = ctx.reg_alloc.UseScratchXmm(args[1]);
code.psrldq(xmm_a, position / 8);
code.pslldq(xmm_b, (128 - position) / 8);
code.por(xmm_a, xmm_b);
ctx.reg_alloc.DefineValue(inst, xmm_a); ctx.reg_alloc.DefineValue(inst, xmm_a);
} }