mirror of
https://git.suyu.dev/suyu/dynarmic.git
synced 2026-01-05 22:18:16 +01:00
VFP: Implement VMUL
This commit is contained in:
parent
97b5fa173f
commit
12e7f2c359
8 changed files with 93 additions and 73 deletions
|
|
@ -69,7 +69,7 @@ boost::optional<const VFP2Matcher<V>&> DecodeVFP2(u32 instruction) {
|
|||
// VNMLA
|
||||
// VNMLS
|
||||
// VNMUL
|
||||
// VMUL
|
||||
INST(&V::vfp2_VMUL, "VMUL", "cccc11100D10nnnndddd101zN0M0mmmm"),
|
||||
INST(&V::vfp2_VADD, "VADD", "cccc11100D11nnnndddd101zN0M0mmmm"),
|
||||
INST(&V::vfp2_VSUB, "VSUB", "cccc11100D11nnnndddd101zN1M0mmmm"),
|
||||
// VDIV
|
||||
|
|
|
|||
|
|
@ -560,6 +560,10 @@ public:
|
|||
std::string arm_SRS() { return "ice"; }
|
||||
|
||||
// Floating point arithmetic instructions
|
||||
std::string vfp2_VMUL(Cond cond, bool D, size_t Vn, size_t Vd, bool sz, bool N, bool M, size_t Vm) {
|
||||
return Common::StringFromFormat("vmul%s.%s %s, %s, %s", CondToString(cond), sz ? "f64" : "f32", FPRegStr(sz, Vd, D).c_str(), FPRegStr(sz, Vn, N).c_str(), FPRegStr(sz, Vm, M).c_str());
|
||||
}
|
||||
|
||||
std::string vfp2_VADD(Cond cond, bool D, size_t Vn, size_t Vd, bool sz, bool N, bool M, size_t Vm) {
|
||||
return Common::StringFromFormat("vadd%s.%s %s, %s, %s", CondToString(cond), sz ? "f64" : "f32", FPRegStr(sz, Vd, D).c_str(), FPRegStr(sz, Vn, N).c_str(), FPRegStr(sz, Vm, M).c_str());
|
||||
}
|
||||
|
|
|
|||
|
|
@ -292,6 +292,16 @@ IR::Value IREmitter::FPAdd64(const IR::Value& a, const IR::Value& b, bool fpscr_
|
|||
return Inst(IR::Opcode::FPAdd64, {a, b});
|
||||
}
|
||||
|
||||
IR::Value IREmitter::FPMul32(const IR::Value& a, const IR::Value& b, bool fpscr_controlled) {
|
||||
ASSERT(fpscr_controlled);
|
||||
return Inst(IR::Opcode::FPMul32, {a, b});
|
||||
}
|
||||
|
||||
IR::Value IREmitter::FPMul64(const IR::Value& a, const IR::Value& b, bool fpscr_controlled) {
|
||||
ASSERT(fpscr_controlled);
|
||||
return Inst(IR::Opcode::FPMul64, {a, b});
|
||||
}
|
||||
|
||||
IR::Value IREmitter::FPSub32(const IR::Value& a, const IR::Value& b, bool fpscr_controlled) {
|
||||
ASSERT(fpscr_controlled);
|
||||
return Inst(IR::Opcode::FPSub32, {a, b});
|
||||
|
|
|
|||
|
|
@ -96,6 +96,8 @@ public:
|
|||
IR::Value FPAbs64(const IR::Value& a);
|
||||
IR::Value FPAdd32(const IR::Value& a, const IR::Value& b, bool fpscr_controlled);
|
||||
IR::Value FPAdd64(const IR::Value& a, const IR::Value& b, bool fpscr_controlled);
|
||||
IR::Value FPMul32(const IR::Value& a, const IR::Value& b, bool fpscr_controlled);
|
||||
IR::Value FPMul64(const IR::Value& a, const IR::Value& b, bool fpscr_controlled);
|
||||
IR::Value FPSub32(const IR::Value& a, const IR::Value& b, bool fpscr_controlled);
|
||||
IR::Value FPSub64(const IR::Value& a, const IR::Value& b, bool fpscr_controlled);
|
||||
|
||||
|
|
|
|||
|
|
@ -64,6 +64,8 @@ OPCODE(FPAbs32, T::F32, T::F32
|
|||
OPCODE(FPAbs64, T::F64, T::F64 )
|
||||
OPCODE(FPAdd32, T::F32, T::F32, T::F32 )
|
||||
OPCODE(FPAdd64, T::F64, T::F64, T::F64 )
|
||||
OPCODE(FPMul32, T::F32, T::F32, T::F32 )
|
||||
OPCODE(FPMul64, T::F64, T::F64, T::F64 )
|
||||
OPCODE(FPSub32, T::F32, T::F32, T::F32 )
|
||||
OPCODE(FPSub64, T::F64, T::F64, T::F64 )
|
||||
|
||||
|
|
|
|||
|
|
@ -321,6 +321,7 @@ struct ArmTranslatorVisitor final {
|
|||
// Floating-point three-register data processing instructions
|
||||
bool vfp2_VADD(Cond cond, bool D, size_t Vn, size_t Vd, bool sz, bool N, bool M, size_t Vm);
|
||||
bool vfp2_VSUB(Cond cond, bool D, size_t Vn, size_t Vd, bool sz, bool N, bool M, size_t Vm);
|
||||
bool vfp2_VMUL(Cond cond, bool D, size_t Vn, size_t Vd, bool sz, bool N, bool M, size_t Vm);
|
||||
|
||||
// Floating-point misc instructions
|
||||
bool vfp2_VABS(Cond cond, bool D, size_t Vd, bool sz, bool M, size_t Vm);
|
||||
|
|
|
|||
|
|
@ -55,6 +55,25 @@ bool ArmTranslatorVisitor::vfp2_VSUB(Cond cond, bool D, size_t Vn, size_t Vd, bo
|
|||
return true;
|
||||
}
|
||||
|
||||
bool ArmTranslatorVisitor::vfp2_VMUL(Cond cond, bool D, size_t Vn, size_t Vd, bool sz, bool N, bool M, size_t Vm) {
|
||||
if (ir.current_location.FPSCR_Len() != 1 || ir.current_location.FPSCR_Stride() != 1)
|
||||
return InterpretThisInstruction(); // TODO: Vectorised floating point instructions
|
||||
|
||||
ExtReg d = ToExtReg(sz, Vd, D);
|
||||
ExtReg n = ToExtReg(sz, Vn, N);
|
||||
ExtReg m = ToExtReg(sz, Vm, M);
|
||||
// VMUL.{F32,F64} <{S,D}d>, <{S,D}n>, <{S,D}m>
|
||||
if (ConditionPassed(cond)) {
|
||||
auto a = ir.GetExtendedRegister(n);
|
||||
auto b = ir.GetExtendedRegister(m);
|
||||
auto result = sz
|
||||
? ir.FPMul64(a, b, true)
|
||||
: ir.FPMul32(a, b, true);
|
||||
ir.SetExtendedRegister(d, result);
|
||||
}
|
||||
return true;
|
||||
}
|
||||
|
||||
bool ArmTranslatorVisitor::vfp2_VABS(Cond cond, bool D, size_t Vd, bool sz, bool M, size_t Vm) {
|
||||
if (ir.current_location.FPSCR_Len() != 1 || ir.current_location.FPSCR_Stride() != 1)
|
||||
return InterpretThisInstruction(); // TODO: Vectorised floating point instructions
|
||||
|
|
|
|||
Loading…
Add table
Add a link
Reference in a new issue