IR: Implement FPVectorAdd
This commit is contained in:
parent
5f77ab28ee
commit
98c8e7d1af
4 changed files with 22 additions and 0 deletions
|
@ -25,6 +25,14 @@ static void EmitVectorOperation(BlockOfCode& code, EmitContext& ctx, IR::Inst* i
|
||||||
ctx.reg_alloc.DefineValue(inst, xmm_a);
|
ctx.reg_alloc.DefineValue(inst, xmm_a);
|
||||||
}
|
}
|
||||||
|
|
||||||
|
void EmitX64::EmitFPVectorAdd32(EmitContext& ctx, IR::Inst* inst) {
|
||||||
|
EmitVectorOperation(code, ctx, inst, &Xbyak::CodeGenerator::addps);
|
||||||
|
}
|
||||||
|
|
||||||
|
void EmitX64::EmitFPVectorAdd64(EmitContext& ctx, IR::Inst* inst) {
|
||||||
|
EmitVectorOperation(code, ctx, inst, &Xbyak::CodeGenerator::addpd);
|
||||||
|
}
|
||||||
|
|
||||||
void EmitX64::EmitFPVectorSub32(EmitContext& ctx, IR::Inst* inst) {
|
void EmitX64::EmitFPVectorSub32(EmitContext& ctx, IR::Inst* inst) {
|
||||||
EmitVectorOperation(code, ctx, inst, &Xbyak::CodeGenerator::subps);
|
EmitVectorOperation(code, ctx, inst, &Xbyak::CodeGenerator::subps);
|
||||||
}
|
}
|
||||||
|
|
|
@ -1154,6 +1154,17 @@ U64 IREmitter::FPU32ToDouble(const U32& a, bool round_to_nearest, bool fpscr_con
|
||||||
return Inst<U64>(Opcode::FPU32ToDouble, a, Imm1(round_to_nearest));
|
return Inst<U64>(Opcode::FPU32ToDouble, a, Imm1(round_to_nearest));
|
||||||
}
|
}
|
||||||
|
|
||||||
|
U128 IREmitter::FPVectorAdd(size_t esize, const U128& a, const U128& b) {
|
||||||
|
switch (esize) {
|
||||||
|
case 32:
|
||||||
|
return Inst<U128>(Opcode::FPVectorAdd32, a, b);
|
||||||
|
case 64:
|
||||||
|
return Inst<U128>(Opcode::FPVectorAdd64, a, b);
|
||||||
|
}
|
||||||
|
UNREACHABLE();
|
||||||
|
return {};
|
||||||
|
}
|
||||||
|
|
||||||
U128 IREmitter::FPVectorSub(size_t esize, const U128& a, const U128& b) {
|
U128 IREmitter::FPVectorSub(size_t esize, const U128& a, const U128& b) {
|
||||||
switch (esize) {
|
switch (esize) {
|
||||||
case 32:
|
case 32:
|
||||||
|
|
|
@ -248,6 +248,7 @@ public:
|
||||||
U64 FPS32ToDouble(const U32& a, bool round_to_nearest, bool fpscr_controlled);
|
U64 FPS32ToDouble(const U32& a, bool round_to_nearest, bool fpscr_controlled);
|
||||||
U64 FPU32ToDouble(const U32& a, bool round_to_nearest, bool fpscr_controlled);
|
U64 FPU32ToDouble(const U32& a, bool round_to_nearest, bool fpscr_controlled);
|
||||||
|
|
||||||
|
U128 FPVectorAdd(size_t esize, const U128& a, const U128& b);
|
||||||
U128 FPVectorSub(size_t esize, const U128& a, const U128& b);
|
U128 FPVectorSub(size_t esize, const U128& a, const U128& b);
|
||||||
|
|
||||||
void Breakpoint();
|
void Breakpoint();
|
||||||
|
|
|
@ -294,6 +294,8 @@ OPCODE(FPU32ToDouble, T::U64, T::U32, T::U1
|
||||||
OPCODE(FPS32ToDouble, T::U64, T::U32, T::U1 )
|
OPCODE(FPS32ToDouble, T::U64, T::U32, T::U1 )
|
||||||
|
|
||||||
// Floating-point vector instructions
|
// Floating-point vector instructions
|
||||||
|
OPCODE(FPVectorAdd32, T::U128, T::U128, T::U128 )
|
||||||
|
OPCODE(FPVectorAdd64, T::U128, T::U128, T::U128 )
|
||||||
OPCODE(FPVectorSub32, T::U128, T::U128, T::U128 )
|
OPCODE(FPVectorSub32, T::U128, T::U128, T::U128 )
|
||||||
OPCODE(FPVectorSub64, T::U128, T::U128, T::U128 )
|
OPCODE(FPVectorSub64, T::U128, T::U128, T::U128 )
|
||||||
|
|
||||||
|
|
Loading…
Reference in a new issue