A64: Implement FDIV (vector)
This commit is contained in:
parent
fd075d8d68
commit
49cc6d7fad
6 changed files with 40 additions and 1 deletions
|
@ -187,6 +187,14 @@ void EmitX64::EmitFPVectorAdd64(EmitContext& ctx, IR::Inst* inst) {
|
|||
EmitVectorOperation64(code, ctx, inst, &Xbyak::CodeGenerator::addpd);
|
||||
}
|
||||
|
||||
void EmitX64::EmitFPVectorDiv32(EmitContext& ctx, IR::Inst* inst) {
|
||||
EmitVectorOperation32(code, ctx, inst, &Xbyak::CodeGenerator::divps);
|
||||
}
|
||||
|
||||
void EmitX64::EmitFPVectorDiv64(EmitContext& ctx, IR::Inst* inst) {
|
||||
EmitVectorOperation64(code, ctx, inst, &Xbyak::CodeGenerator::divpd);
|
||||
}
|
||||
|
||||
void EmitX64::EmitFPVectorSub32(EmitContext& ctx, IR::Inst* inst) {
|
||||
EmitVectorOperation32(code, ctx, inst, &Xbyak::CodeGenerator::subps);
|
||||
}
|
||||
|
|
|
@ -770,7 +770,7 @@ INST(MLS_vec, "MLS (vector)", "0Q101
|
|||
//INST(FCMGE_reg_4, "FCMGE (register)", "0Q1011100z1mmmmm111001nnnnnddddd")
|
||||
//INST(FACGE_4, "FACGE", "0Q1011100z1mmmmm111011nnnnnddddd")
|
||||
//INST(FMAXP_vec_2, "FMAXP (vector)", "0Q1011100z1mmmmm111101nnnnnddddd")
|
||||
//INST(FDIV_2, "FDIV (vector)", "0Q1011100z1mmmmm111111nnnnnddddd")
|
||||
INST(FDIV_2, "FDIV (vector)", "0Q1011100z1mmmmm111111nnnnnddddd")
|
||||
INST(EOR_asimd, "EOR (vector)", "0Q101110001mmmmm000111nnnnnddddd")
|
||||
INST(BSL, "BSL", "0Q101110011mmmmm000111nnnnnddddd")
|
||||
//INST(FMINNMP_vec_2, "FMINNMP (vector)", "0Q1011101z1mmmmm110001nnnnnddddd")
|
||||
|
|
|
@ -348,6 +348,23 @@ bool TranslatorVisitor::EOR_asimd(bool Q, Vec Vm, Vec Vn, Vec Vd) {
|
|||
return true;
|
||||
}
|
||||
|
||||
bool TranslatorVisitor::FDIV_2(bool Q, bool sz, Vec Vm, Vec Vn, Vec Vd) {
|
||||
if (sz && !Q) {
|
||||
return ReservedValue();
|
||||
}
|
||||
const size_t esize = sz ? 64 : 32;
|
||||
const size_t datasize = Q ? 128 : 64;
|
||||
|
||||
const IR::U128 operand1 = V(datasize, Vn);
|
||||
const IR::U128 operand2 = V(datasize, Vm);
|
||||
IR::U128 result = ir.FPVectorDiv(esize, operand1, operand2);
|
||||
if (datasize == 64) {
|
||||
result = ir.VectorZeroUpper(result);
|
||||
}
|
||||
V(datasize, Vd, result);
|
||||
return true;
|
||||
}
|
||||
|
||||
bool TranslatorVisitor::BIF(bool Q, Vec Vm, Vec Vn, Vec Vd) {
|
||||
const size_t datasize = Q ? 128 : 64;
|
||||
|
||||
|
|
|
@ -1331,6 +1331,17 @@ U128 IREmitter::FPVectorAdd(size_t esize, const U128& a, const U128& b) {
|
|||
return {};
|
||||
}
|
||||
|
||||
U128 IREmitter::FPVectorDiv(size_t esize, const U128& a, const U128& b) {
|
||||
switch (esize) {
|
||||
case 32:
|
||||
return Inst<U128>(Opcode::FPVectorDiv32, a, b);
|
||||
case 64:
|
||||
return Inst<U128>(Opcode::FPVectorDiv64, a, b);
|
||||
}
|
||||
UNREACHABLE();
|
||||
return {};
|
||||
}
|
||||
|
||||
U128 IREmitter::FPVectorSub(size_t esize, const U128& a, const U128& b) {
|
||||
switch (esize) {
|
||||
case 32:
|
||||
|
|
|
@ -268,6 +268,7 @@ public:
|
|||
U64 FPU32ToDouble(const U32& a, bool round_to_nearest, bool fpscr_controlled);
|
||||
|
||||
U128 FPVectorAdd(size_t esize, const U128& a, const U128& b);
|
||||
U128 FPVectorDiv(size_t esize, const U128& a, const U128& b);
|
||||
U128 FPVectorSub(size_t esize, const U128& a, const U128& b);
|
||||
|
||||
void Breakpoint();
|
||||
|
|
|
@ -343,6 +343,8 @@ OPCODE(FPS32ToDouble, T::U64, T::U32, T::U1
|
|||
// Floating-point vector instructions
|
||||
OPCODE(FPVectorAdd32, T::U128, T::U128, T::U128 )
|
||||
OPCODE(FPVectorAdd64, T::U128, T::U128, T::U128 )
|
||||
OPCODE(FPVectorDiv32, T::U128, T::U128, T::U128 )
|
||||
OPCODE(FPVectorDiv64, T::U128, T::U128, T::U128 )
|
||||
OPCODE(FPVectorSub32, T::U128, T::U128, T::U128 )
|
||||
OPCODE(FPVectorSub64, T::U128, T::U128, T::U128 )
|
||||
|
||||
|
|
Loading…
Reference in a new issue