emit_x64_vector: Emit VPMULLQ in EmitVectorMultiply64 on AVX-512{DQ, VL} capable CPUs
Shortens code-gen down to a single instruction in the 64-bit path.
This commit is contained in:
parent
9054d1c20b
commit
6de5ed96e5
1 changed files with 11 additions and 0 deletions
|
@ -1459,6 +1459,17 @@ void EmitX64::EmitVectorMultiply32(EmitContext& ctx, IR::Inst* inst) {
|
|||
}
|
||||
|
||||
void EmitX64::EmitVectorMultiply64(EmitContext& ctx, IR::Inst* inst) {
|
||||
if (code.DoesCpuSupport(Xbyak::util::Cpu::tAVX512DQ) && code.DoesCpuSupport(Xbyak::util::Cpu::tAVX512VL)) {
|
||||
auto args = ctx.reg_alloc.GetArgumentInfo(inst);
|
||||
const Xbyak::Xmm a = ctx.reg_alloc.UseScratchXmm(args[0]);
|
||||
const Xbyak::Xmm b = ctx.reg_alloc.UseXmm(args[1]);
|
||||
|
||||
code.vpmullq(a, a, b);
|
||||
|
||||
ctx.reg_alloc.DefineValue(inst, a);
|
||||
return;
|
||||
}
|
||||
|
||||
if (code.DoesCpuSupport(Xbyak::util::Cpu::tSSE41)) {
|
||||
auto args = ctx.reg_alloc.GetArgumentInfo(inst);
|
||||
Xbyak::Xmm a = ctx.reg_alloc.UseScratchXmm(args[0]);
|
||||
|
|
Loading…
Reference in a new issue