emit_x64_vector_floating_point: Reduce codesize of EmitTwoOpVectorOperation

This commit is contained in:
MerryMage 2018-07-31 20:25:03 +01:00
parent bb93353f94
commit 7731dcdca9

View file

@ -241,17 +241,21 @@ void EmitTwoOpVectorOperation(BlockOfCode& code, EmitContext& ctx, IR::Inst* ins
if (!ctx.AccurateNaN() || ctx.FPSCR_DN()) {
auto args = ctx.reg_alloc.GetArgumentInfo(inst);
const Xbyak::Xmm xmm_a = ctx.reg_alloc.UseScratchXmm(args[0]);
Xbyak::Xmm result;
if constexpr (std::is_member_function_pointer_v<Function>) {
(code.*fn)(xmm_a);
result = ctx.reg_alloc.UseScratchXmm(args[0]);
(code.*fn)(result);
} else {
fn(xmm_a);
const Xbyak::Xmm xmm_a = ctx.reg_alloc.UseXmm(args[0]);
result = ctx.reg_alloc.ScratchXmm();
fn(result, xmm_a);
}
ForceToDefaultNaN<fsize>(code, ctx, xmm_a);
ForceToDefaultNaN<fsize>(code, ctx, result);
ctx.reg_alloc.DefineValue(inst, xmm_a);
ctx.reg_alloc.DefineValue(inst, result);
return;
}
@ -261,15 +265,19 @@ void EmitTwoOpVectorOperation(BlockOfCode& code, EmitContext& ctx, IR::Inst* ins
const Xbyak::Xmm xmm_a = ctx.reg_alloc.UseXmm(args[0]);
const Xbyak::Xmm nan_mask = ctx.reg_alloc.ScratchXmm();
code.movaps(nan_mask, xmm_a);
code.movaps(result, xmm_a);
FCODE(cmpunordp)(nan_mask, nan_mask);
if constexpr (std::is_member_function_pointer_v<Function>) {
code.movaps(result, xmm_a);
(code.*fn)(result);
} else {
fn(result);
fn(result, xmm_a);
}
if (code.DoesCpuSupport(Xbyak::util::Cpu::tAVX)) {
FCODE(vcmpunordp)(nan_mask, result, result);
} else {
code.movaps(nan_mask, result);
FCODE(cmpunordp)(nan_mask, nan_mask);
}
FCODE(cmpunordp)(nan_mask, result);
HandleNaNs<fsize, 1>(code, ctx, {result, xmm_a}, nan_mask, nan_handler);
@ -907,11 +915,11 @@ void EmitFPVectorRoundInt(BlockOfCode& code, EmitContext& ctx, IR::Inst* inst) {
return 0;
}();
EmitTwoOpVectorOperation<fsize, DefaultIndexer>(code, ctx, inst, [&](const Xbyak::Xmm& result){
EmitTwoOpVectorOperation<fsize, DefaultIndexer>(code, ctx, inst, [&](const Xbyak::Xmm& result, const Xbyak::Xmm& xmm_a){
if constexpr (fsize == 32) {
code.roundps(result, result, round_imm);
code.roundps(result, xmm_a, round_imm);
} else {
code.roundpd(result, result, round_imm);
code.roundpd(result, xmm_a, round_imm);
}
});