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

emit_x64_vector: AVX512+GNFI implementation of EmitVectorLogicalVShift8

Browse source
This commit is contained in:
Wunkolo 2024-02-04 11:02:15 -08:00 committed by merry
parent f5df599e9d
commit 18717d216c
1 changed files with 41 additions and 0 deletions
Show stats Download patch file Download diff file Expand all files Collapse all files

41
src/dynarmic/backend/x64/emit_x64_vector.cpp
View file

@ -1824,6 +1824,47 @@ void EmitX64::EmitVectorLogicalShiftRight64(EmitContext& ctx, IR::Inst* inst) {
}
void EmitX64::EmitVectorLogicalVShift8(EmitContext& ctx, IR::Inst* inst) {
if (code.HasHostFeature(HostFeature::AVX512_Ortho | HostFeature::AVX512BW | HostFeature::GFNI)) {
auto args = ctx.reg_alloc.GetArgumentInfo(inst);
const Xbyak::Xmm result = ctx.reg_alloc.UseScratchXmm(args[0]);
const Xbyak::Xmm left_shift = ctx.reg_alloc.UseScratchXmm(args[1]);
const Xbyak::Xmm tmp = ctx.reg_alloc.ScratchXmm();
const Xbyak::Opmask negative_mask = k1;
code.pxor(tmp, tmp);
code.vpcmpb(negative_mask, left_shift, tmp, CmpInt::LessThan);
// Reverse bits of negative-shifts
code.vmovaps(xmm0, code.BConst<64>(xword, 0x8040201008040201));
code.vgf2p8affineqb(result | negative_mask, result, xmm0, 0);
// Turn all negative shifts into left-shifts
code.pabsb(left_shift, left_shift);
const Xbyak::Opmask valid_index = k2;
code.vptestnmb(valid_index, left_shift, code.BConst<8>(xword, 0xF8));
// gf2p8mulb's "x8 + x4 + x3 + x + 1"-polynomial-reduction only applies
// when the multiplication overflows. Masking away any bits that would have
// overflowed turns the polynomial-multiplication into regular modulo-multiplication
code.movdqa(tmp, code.Const(xword, 0x01'03'07'0f'1f'3f'7f'ff, 0));
code.vpshufb(tmp | valid_index | T_z, tmp, left_shift);
code.pand(result, tmp);
// n << 0 == n * 1 | n << 1 == n * 2 | n << 2 == n * 4 | etc
code.pxor(tmp, tmp);
code.movsd(tmp, xmm0);
code.pshufb(tmp, left_shift);
code.gf2p8mulb(result, tmp);
// Un-reverse bits of negative-shifts
code.vgf2p8affineqb(result | negative_mask, result, xmm0, 0);
ctx.reg_alloc.DefineValue(inst, result);
return;
}
EmitTwoArgumentFallback(code, ctx, inst, [](VectorArray<u8>& result, const VectorArray<u8>& a, const VectorArray<u8>& b) {
std::transform(a.begin(), a.end(), b.begin(), result.begin(), VShift<u8>);
});