mirrors/dynarmic
1
0
Fork 0
Code Issues Pull requests Projects Releases Packages Wiki Activity

ir_emitter: Default fpcr_controlled arguments to true

Browse source
This commit is contained in:
MerryMage 2020-06-19 22:46:19 +01:00
parent 794440cf8d
commit d3664b03fe
12 changed files with 62 additions and 68 deletions
Show stats Download patch file Download diff file Expand all files Collapse all files

36
src/frontend/A32/translate/impl/vfp.cpp
View file

@ -97,7 +97,7 @@ bool ArmTranslatorVisitor::vfp_VADD(Cond cond, bool D, size_t Vn, size_t Vd, boo
return EmitVfpVectorOperation(sz, d, n, m, [this](ExtReg d, ExtReg n, ExtReg m) { return EmitVfpVectorOperation(sz, d, n, m, [this](ExtReg d, ExtReg n, ExtReg m) {
const auto reg_n = ir.GetExtendedRegister(n); const auto reg_n = ir.GetExtendedRegister(n);
const auto reg_m = ir.GetExtendedRegister(m); const auto reg_m = ir.GetExtendedRegister(m);
const auto result = ir.FPAdd(reg_n, reg_m, true); const auto result = ir.FPAdd(reg_n, reg_m);
ir.SetExtendedRegister(d, result); ir.SetExtendedRegister(d, result);
}); });
} }
@ -116,7 +116,7 @@ bool ArmTranslatorVisitor::vfp_VSUB(Cond cond, bool D, size_t Vn, size_t Vd, boo
return EmitVfpVectorOperation(sz, d, n, m, [this](ExtReg d, ExtReg n, ExtReg m) { return EmitVfpVectorOperation(sz, d, n, m, [this](ExtReg d, ExtReg n, ExtReg m) {
const auto reg_n = ir.GetExtendedRegister(n); const auto reg_n = ir.GetExtendedRegister(n);
const auto reg_m = ir.GetExtendedRegister(m); const auto reg_m = ir.GetExtendedRegister(m);
const auto result = ir.FPSub(reg_n, reg_m, true); const auto result = ir.FPSub(reg_n, reg_m);
ir.SetExtendedRegister(d, result); ir.SetExtendedRegister(d, result);
}); });
} }
@ -135,7 +135,7 @@ bool ArmTranslatorVisitor::vfp_VMUL(Cond cond, bool D, size_t Vn, size_t Vd, boo
return EmitVfpVectorOperation(sz, d, n, m, [this](ExtReg d, ExtReg n, ExtReg m) { return EmitVfpVectorOperation(sz, d, n, m, [this](ExtReg d, ExtReg n, ExtReg m) {
const auto reg_n = ir.GetExtendedRegister(n); const auto reg_n = ir.GetExtendedRegister(n);
const auto reg_m = ir.GetExtendedRegister(m); const auto reg_m = ir.GetExtendedRegister(m);
const auto result = ir.FPMul(reg_n, reg_m, true); const auto result = ir.FPMul(reg_n, reg_m);
ir.SetExtendedRegister(d, result); ir.SetExtendedRegister(d, result);
}); });
} }
@ -155,7 +155,7 @@ bool ArmTranslatorVisitor::vfp_VMLA(Cond cond, bool D, size_t Vn, size_t Vd, boo
const auto reg_n = ir.GetExtendedRegister(n); const auto reg_n = ir.GetExtendedRegister(n);
const auto reg_m = ir.GetExtendedRegister(m); const auto reg_m = ir.GetExtendedRegister(m);
const auto reg_d = ir.GetExtendedRegister(d); const auto reg_d = ir.GetExtendedRegister(d);
const auto result = ir.FPAdd(reg_d, ir.FPMul(reg_n, reg_m, true), true); const auto result = ir.FPAdd(reg_d, ir.FPMul(reg_n, reg_m));
ir.SetExtendedRegister(d, result); ir.SetExtendedRegister(d, result);
}); });
} }
@ -175,7 +175,7 @@ bool ArmTranslatorVisitor::vfp_VMLS(Cond cond, bool D, size_t Vn, size_t Vd, boo
const auto reg_n = ir.GetExtendedRegister(n); const auto reg_n = ir.GetExtendedRegister(n);
const auto reg_m = ir.GetExtendedRegister(m); const auto reg_m = ir.GetExtendedRegister(m);
const auto reg_d = ir.GetExtendedRegister(d); const auto reg_d = ir.GetExtendedRegister(d);
const auto result = ir.FPAdd(reg_d, ir.FPNeg(ir.FPMul(reg_n, reg_m, true)), true); const auto result = ir.FPAdd(reg_d, ir.FPNeg(ir.FPMul(reg_n, reg_m)));
ir.SetExtendedRegister(d, result); ir.SetExtendedRegister(d, result);
}); });
} }
@ -194,7 +194,7 @@ bool ArmTranslatorVisitor::vfp_VNMUL(Cond cond, bool D, size_t Vn, size_t Vd, bo
return EmitVfpVectorOperation(sz, d, n, m, [this](ExtReg d, ExtReg n, ExtReg m) { return EmitVfpVectorOperation(sz, d, n, m, [this](ExtReg d, ExtReg n, ExtReg m) {
const auto reg_n = ir.GetExtendedRegister(n); const auto reg_n = ir.GetExtendedRegister(n);
const auto reg_m = ir.GetExtendedRegister(m); const auto reg_m = ir.GetExtendedRegister(m);
const auto result = ir.FPNeg(ir.FPMul(reg_n, reg_m, true)); const auto result = ir.FPNeg(ir.FPMul(reg_n, reg_m));
ir.SetExtendedRegister(d, result); ir.SetExtendedRegister(d, result);
}); });
} }
@ -214,7 +214,7 @@ bool ArmTranslatorVisitor::vfp_VNMLA(Cond cond, bool D, size_t Vn, size_t Vd, bo
const auto reg_n = ir.GetExtendedRegister(n); const auto reg_n = ir.GetExtendedRegister(n);
const auto reg_m = ir.GetExtendedRegister(m); const auto reg_m = ir.GetExtendedRegister(m);
const auto reg_d = ir.GetExtendedRegister(d); const auto reg_d = ir.GetExtendedRegister(d);
const auto result = ir.FPAdd(ir.FPNeg(reg_d), ir.FPNeg(ir.FPMul(reg_n, reg_m, true)), true); const auto result = ir.FPAdd(ir.FPNeg(reg_d), ir.FPNeg(ir.FPMul(reg_n, reg_m)));
ir.SetExtendedRegister(d, result); ir.SetExtendedRegister(d, result);
}); });
} }
@ -234,7 +234,7 @@ bool ArmTranslatorVisitor::vfp_VNMLS(Cond cond, bool D, size_t Vn, size_t Vd, bo
const auto reg_n = ir.GetExtendedRegister(n); const auto reg_n = ir.GetExtendedRegister(n);
const auto reg_m = ir.GetExtendedRegister(m); const auto reg_m = ir.GetExtendedRegister(m);
const auto reg_d = ir.GetExtendedRegister(d); const auto reg_d = ir.GetExtendedRegister(d);
const auto result = ir.FPAdd(ir.FPNeg(reg_d), ir.FPMul(reg_n, reg_m, true), true); const auto result = ir.FPAdd(ir.FPNeg(reg_d), ir.FPMul(reg_n, reg_m));
ir.SetExtendedRegister(d, result); ir.SetExtendedRegister(d, result);
}); });
} }
@ -253,7 +253,7 @@ bool ArmTranslatorVisitor::vfp_VDIV(Cond cond, bool D, size_t Vn, size_t Vd, boo
return EmitVfpVectorOperation(sz, d, n, m, [this](ExtReg d, ExtReg n, ExtReg m) { return EmitVfpVectorOperation(sz, d, n, m, [this](ExtReg d, ExtReg n, ExtReg m) {
const auto reg_n = ir.GetExtendedRegister(n); const auto reg_n = ir.GetExtendedRegister(n);
const auto reg_m = ir.GetExtendedRegister(m); const auto reg_m = ir.GetExtendedRegister(m);
const auto result = ir.FPDiv(reg_n, reg_m, true); const auto result = ir.FPDiv(reg_n, reg_m);
ir.SetExtendedRegister(d, result); ir.SetExtendedRegister(d, result);
}); });
} }
@ -273,7 +273,7 @@ bool ArmTranslatorVisitor::vfp_VFNMS(Cond cond, bool D, size_t Vn, size_t Vd, bo
const auto reg_n = ir.GetExtendedRegister(n); const auto reg_n = ir.GetExtendedRegister(n);
const auto reg_m = ir.GetExtendedRegister(m); const auto reg_m = ir.GetExtendedRegister(m);
const auto reg_d = ir.GetExtendedRegister(d); const auto reg_d = ir.GetExtendedRegister(d);
const auto result = ir.FPMulAdd(ir.FPNeg(reg_d), reg_n, reg_m, true); const auto result = ir.FPMulAdd(ir.FPNeg(reg_d), reg_n, reg_m);
ir.SetExtendedRegister(d, result); ir.SetExtendedRegister(d, result);
}); });
} }
@ -293,7 +293,7 @@ bool ArmTranslatorVisitor::vfp_VFNMA(Cond cond, bool D, size_t Vn, size_t Vd, bo
const auto reg_n = ir.GetExtendedRegister(n); const auto reg_n = ir.GetExtendedRegister(n);
const auto reg_m = ir.GetExtendedRegister(m); const auto reg_m = ir.GetExtendedRegister(m);
const auto reg_d = ir.GetExtendedRegister(d); const auto reg_d = ir.GetExtendedRegister(d);
const auto result = ir.FPMulAdd(ir.FPNeg(reg_d), ir.FPNeg(reg_n), reg_m, true); const auto result = ir.FPMulAdd(ir.FPNeg(reg_d), ir.FPNeg(reg_n), reg_m);
ir.SetExtendedRegister(d, result); ir.SetExtendedRegister(d, result);
}); });
} }
@ -313,7 +313,7 @@ bool ArmTranslatorVisitor::vfp_VFMA(Cond cond, bool D, size_t Vn, size_t Vd, boo
const auto reg_n = ir.GetExtendedRegister(n); const auto reg_n = ir.GetExtendedRegister(n);
const auto reg_m = ir.GetExtendedRegister(m); const auto reg_m = ir.GetExtendedRegister(m);
const auto reg_d = ir.GetExtendedRegister(d); const auto reg_d = ir.GetExtendedRegister(d);
const auto result = ir.FPMulAdd(reg_d, reg_n, reg_m, true); const auto result = ir.FPMulAdd(reg_d, reg_n, reg_m);
ir.SetExtendedRegister(d, result); ir.SetExtendedRegister(d, result);
}); });
} }
@ -333,7 +333,7 @@ bool ArmTranslatorVisitor::vfp_VFMS(Cond cond, bool D, size_t Vn, size_t Vd, boo
const auto reg_n = ir.GetExtendedRegister(n); const auto reg_n = ir.GetExtendedRegister(n);
const auto reg_m = ir.GetExtendedRegister(m); const auto reg_m = ir.GetExtendedRegister(m);
const auto reg_d = ir.GetExtendedRegister(d); const auto reg_d = ir.GetExtendedRegister(d);
const auto result = ir.FPMulAdd(reg_d, ir.FPNeg(reg_n), reg_m, true); const auto result = ir.FPMulAdd(reg_d, ir.FPNeg(reg_n), reg_m);
ir.SetExtendedRegister(d, result); ir.SetExtendedRegister(d, result);
}); });
} }
@ -365,7 +365,7 @@ bool ArmTranslatorVisitor::vfp_VMAXNM(bool D, size_t Vn, size_t Vd, bool sz, boo
return EmitVfpVectorOperation(sz, d, n, m, [this](ExtReg d, ExtReg n, ExtReg m) { return EmitVfpVectorOperation(sz, d, n, m, [this](ExtReg d, ExtReg n, ExtReg m) {
const auto reg_n = ir.GetExtendedRegister(n); const auto reg_n = ir.GetExtendedRegister(n);
const auto reg_m = ir.GetExtendedRegister(m); const auto reg_m = ir.GetExtendedRegister(m);
const auto result = ir.FPMaxNumeric(reg_n, reg_m, true); const auto result = ir.FPMaxNumeric(reg_n, reg_m);
ir.SetExtendedRegister(d, result); ir.SetExtendedRegister(d, result);
}); });
} }
@ -380,7 +380,7 @@ bool ArmTranslatorVisitor::vfp_VMINNM(bool D, size_t Vn, size_t Vd, bool sz, boo
return EmitVfpVectorOperation(sz, d, n, m, [this](ExtReg d, ExtReg n, ExtReg m) { return EmitVfpVectorOperation(sz, d, n, m, [this](ExtReg d, ExtReg n, ExtReg m) {
const auto reg_n = ir.GetExtendedRegister(n); const auto reg_n = ir.GetExtendedRegister(n);
const auto reg_m = ir.GetExtendedRegister(m); const auto reg_m = ir.GetExtendedRegister(m);
const auto result = ir.FPMinNumeric(reg_n, reg_m, true); const auto result = ir.FPMinNumeric(reg_n, reg_m);
ir.SetExtendedRegister(d, result); ir.SetExtendedRegister(d, result);
}); });
} }
@ -695,7 +695,7 @@ bool ArmTranslatorVisitor::vfp_VCMP(Cond cond, bool D, size_t Vd, bool sz, bool
const auto exc_on_qnan = E; const auto exc_on_qnan = E;
const auto reg_d = ir.GetExtendedRegister(d); const auto reg_d = ir.GetExtendedRegister(d);
const auto reg_m = ir.GetExtendedRegister(m); const auto reg_m = ir.GetExtendedRegister(m);
const auto nzcv = ir.FPCompare(reg_d, reg_m, exc_on_qnan, true); const auto nzcv = ir.FPCompare(reg_d, reg_m, exc_on_qnan);
ir.SetFpscrNZCV(nzcv); ir.SetFpscrNZCV(nzcv);
return true; return true;
@ -713,10 +713,10 @@ bool ArmTranslatorVisitor::vfp_VCMP_zero(Cond cond, bool D, size_t Vd, bool sz,
const auto reg_d = ir.GetExtendedRegister(d); const auto reg_d = ir.GetExtendedRegister(d);
if (sz) { if (sz) {
const auto nzcv = ir.FPCompare(reg_d, ir.Imm64(0), exc_on_qnan, true); const auto nzcv = ir.FPCompare(reg_d, ir.Imm64(0), exc_on_qnan);
ir.SetFpscrNZCV(nzcv); ir.SetFpscrNZCV(nzcv);
} else { } else {
const auto nzcv = ir.FPCompare(reg_d, ir.Imm32(0), exc_on_qnan, true); const auto nzcv = ir.FPCompare(reg_d, ir.Imm32(0), exc_on_qnan);
ir.SetFpscrNZCV(nzcv); ir.SetFpscrNZCV(nzcv);
} }

2
src/frontend/A64/translate/impl/floating_point_compare.cpp
View file

@ -21,7 +21,7 @@ bool FPCompare(TranslatorVisitor& v, Imm<2> type, Vec Vm, Vec Vn, bool exc_on_qn
operand2 = v.V_scalar(*datasize, Vm); operand2 = v.V_scalar(*datasize, Vm);
} }
const auto nzcv = v.ir.FPCompare(operand1, operand2, exc_on_qnan, true); const auto nzcv = v.ir.FPCompare(operand1, operand2, exc_on_qnan);
v.ir.SetNZCV(nzcv); v.ir.SetNZCV(nzcv);
return true; return true;
} }

2
src/frontend/A64/translate/impl/floating_point_conditional_compare.cpp
View file

@ -17,7 +17,7 @@ bool FPCompare(TranslatorVisitor& v, Imm<2> type, Vec Vm, Cond cond, Vec Vn, Imm
const IR::U32U64 operand1 = v.V_scalar(*datasize, Vn); const IR::U32U64 operand1 = v.V_scalar(*datasize, Vn);
const IR::U32U64 operand2 = v.V_scalar(*datasize, Vm); const IR::U32U64 operand2 = v.V_scalar(*datasize, Vm);
const IR::NZCV then_flags = v.ir.FPCompare(operand1, operand2, exc_on_qnan, true); const IR::NZCV then_flags = v.ir.FPCompare(operand1, operand2, exc_on_qnan);
const IR::NZCV else_flags = v.ir.NZCVFromPackedFlags(v.ir.Imm32(flags)); const IR::NZCV else_flags = v.ir.NZCVFromPackedFlags(v.ir.Imm32(flags));
v.ir.SetNZCV(v.ir.ConditionalSelect(cond, then_flags, else_flags)); v.ir.SetNZCV(v.ir.ConditionalSelect(cond, then_flags, else_flags));
return true; return true;

8
src/frontend/A64/translate/impl/floating_point_data_processing_three_register.cpp
View file

@ -16,7 +16,7 @@ bool TranslatorVisitor::FMADD_float(Imm<2> type, Vec Vm, Vec Va, Vec Vn, Vec Vd)
const IR::U16U32U64 operanda = V_scalar(*datasize, Va); const IR::U16U32U64 operanda = V_scalar(*datasize, Va);
const IR::U16U32U64 operand1 = V_scalar(*datasize, Vn); const IR::U16U32U64 operand1 = V_scalar(*datasize, Vn);
const IR::U16U32U64 operand2 = V_scalar(*datasize, Vm); const IR::U16U32U64 operand2 = V_scalar(*datasize, Vm);
const IR::U16U32U64 result = ir.FPMulAdd(operanda, operand1, operand2, true); const IR::U16U32U64 result = ir.FPMulAdd(operanda, operand1, operand2);
V_scalar(*datasize, Vd, result); V_scalar(*datasize, Vd, result);
return true; return true;
} }
@ -30,7 +30,7 @@ bool TranslatorVisitor::FMSUB_float(Imm<2> type, Vec Vm, Vec Va, Vec Vn, Vec Vd)
const IR::U16U32U64 operanda = V_scalar(*datasize, Va); const IR::U16U32U64 operanda = V_scalar(*datasize, Va);
const IR::U16U32U64 operand1 = V_scalar(*datasize, Vn); const IR::U16U32U64 operand1 = V_scalar(*datasize, Vn);
const IR::U16U32U64 operand2 = V_scalar(*datasize, Vm); const IR::U16U32U64 operand2 = V_scalar(*datasize, Vm);
const IR::U16U32U64 result = ir.FPMulAdd(operanda, ir.FPNeg(operand1), operand2, true); const IR::U16U32U64 result = ir.FPMulAdd(operanda, ir.FPNeg(operand1), operand2);
V_scalar(*datasize, Vd, result); V_scalar(*datasize, Vd, result);
return true; return true;
} }
@ -44,7 +44,7 @@ bool TranslatorVisitor::FNMADD_float(Imm<2> type, Vec Vm, Vec Va, Vec Vn, Vec Vd
const IR::U16U32U64 operanda = V_scalar(*datasize, Va); const IR::U16U32U64 operanda = V_scalar(*datasize, Va);
const IR::U16U32U64 operand1 = V_scalar(*datasize, Vn); const IR::U16U32U64 operand1 = V_scalar(*datasize, Vn);
const IR::U16U32U64 operand2 = V_scalar(*datasize, Vm); const IR::U16U32U64 operand2 = V_scalar(*datasize, Vm);
const IR::U16U32U64 result = ir.FPMulAdd(ir.FPNeg(operanda), ir.FPNeg(operand1), operand2, true); const IR::U16U32U64 result = ir.FPMulAdd(ir.FPNeg(operanda), ir.FPNeg(operand1), operand2);
V_scalar(*datasize, Vd, result); V_scalar(*datasize, Vd, result);
return true; return true;
} }
@ -58,7 +58,7 @@ bool TranslatorVisitor::FNMSUB_float(Imm<2> type, Vec Vm, Vec Va, Vec Vn, Vec Vd
const IR::U16U32U64 operanda = V_scalar(*datasize, Va); const IR::U16U32U64 operanda = V_scalar(*datasize, Va);
const IR::U16U32U64 operand1 = V_scalar(*datasize, Vn); const IR::U16U32U64 operand1 = V_scalar(*datasize, Vn);
const IR::U16U32U64 operand2 = V_scalar(*datasize, Vm); const IR::U16U32U64 operand2 = V_scalar(*datasize, Vm);
const IR::U16U32U64 result = ir.FPMulAdd(ir.FPNeg(operanda), operand1, operand2, true); const IR::U16U32U64 result = ir.FPMulAdd(ir.FPNeg(operanda), operand1, operand2);
V_scalar(*datasize, Vd, result); V_scalar(*datasize, Vd, result);
return true; return true;
} }

18
src/frontend/A64/translate/impl/floating_point_data_processing_two_register.cpp
View file

@ -16,7 +16,7 @@ bool TranslatorVisitor::FMUL_float(Imm<2> type, Vec Vm, Vec Vn, Vec Vd) {
const IR::U32U64 operand1 = V_scalar(*datasize, Vn); const IR::U32U64 operand1 = V_scalar(*datasize, Vn);
const IR::U32U64 operand2 = V_scalar(*datasize, Vm); const IR::U32U64 operand2 = V_scalar(*datasize, Vm);
const IR::U32U64 result = ir.FPMul(operand1, operand2, true); const IR::U32U64 result = ir.FPMul(operand1, operand2);
V_scalar(*datasize, Vd, result); V_scalar(*datasize, Vd, result);
return true; return true;
@ -31,7 +31,7 @@ bool TranslatorVisitor::FDIV_float(Imm<2> type, Vec Vm, Vec Vn, Vec Vd) {
const IR::U32U64 operand1 = V_scalar(*datasize, Vn); const IR::U32U64 operand1 = V_scalar(*datasize, Vn);
const IR::U32U64 operand2 = V_scalar(*datasize, Vm); const IR::U32U64 operand2 = V_scalar(*datasize, Vm);
const IR::U32U64 result = ir.FPDiv(operand1, operand2, true); const IR::U32U64 result = ir.FPDiv(operand1, operand2);
V_scalar(*datasize, Vd, result); V_scalar(*datasize, Vd, result);
return true; return true;
@ -46,7 +46,7 @@ bool TranslatorVisitor::FADD_float(Imm<2> type, Vec Vm, Vec Vn, Vec Vd) {
const IR::U32U64 operand1 = V_scalar(*datasize, Vn); const IR::U32U64 operand1 = V_scalar(*datasize, Vn);
const IR::U32U64 operand2 = V_scalar(*datasize, Vm); const IR::U32U64 operand2 = V_scalar(*datasize, Vm);
const IR::U32U64 result = ir.FPAdd(operand1, operand2, true); const IR::U32U64 result = ir.FPAdd(operand1, operand2);
V_scalar(*datasize, Vd, result); V_scalar(*datasize, Vd, result);
return true; return true;
@ -61,7 +61,7 @@ bool TranslatorVisitor::FSUB_float(Imm<2> type, Vec Vm, Vec Vn, Vec Vd) {
const IR::U32U64 operand1 = V_scalar(*datasize, Vn); const IR::U32U64 operand1 = V_scalar(*datasize, Vn);
const IR::U32U64 operand2 = V_scalar(*datasize, Vm); const IR::U32U64 operand2 = V_scalar(*datasize, Vm);
const IR::U32U64 result = ir.FPSub(operand1, operand2, true); const IR::U32U64 result = ir.FPSub(operand1, operand2);
V_scalar(*datasize, Vd, result); V_scalar(*datasize, Vd, result);
return true; return true;
@ -76,7 +76,7 @@ bool TranslatorVisitor::FMAX_float(Imm<2> type, Vec Vm, Vec Vn, Vec Vd) {
const IR::U32U64 operand1 = V_scalar(*datasize, Vn); const IR::U32U64 operand1 = V_scalar(*datasize, Vn);
const IR::U32U64 operand2 = V_scalar(*datasize, Vm); const IR::U32U64 operand2 = V_scalar(*datasize, Vm);
const IR::U32U64 result = ir.FPMax(operand1, operand2, true); const IR::U32U64 result = ir.FPMax(operand1, operand2);
V_scalar(*datasize, Vd, result); V_scalar(*datasize, Vd, result);
return true; return true;
@ -91,7 +91,7 @@ bool TranslatorVisitor::FMIN_float(Imm<2> type, Vec Vm, Vec Vn, Vec Vd) {
const IR::U32U64 operand1 = V_scalar(*datasize, Vn); const IR::U32U64 operand1 = V_scalar(*datasize, Vn);
const IR::U32U64 operand2 = V_scalar(*datasize, Vm); const IR::U32U64 operand2 = V_scalar(*datasize, Vm);
const IR::U32U64 result = ir.FPMin(operand1, operand2, true); const IR::U32U64 result = ir.FPMin(operand1, operand2);
V_scalar(*datasize, Vd, result); V_scalar(*datasize, Vd, result);
return true; return true;
@ -106,7 +106,7 @@ bool TranslatorVisitor::FMAXNM_float(Imm<2> type, Vec Vm, Vec Vn, Vec Vd) {
const IR::U32U64 operand1 = V_scalar(*datasize, Vn); const IR::U32U64 operand1 = V_scalar(*datasize, Vn);
const IR::U32U64 operand2 = V_scalar(*datasize, Vm); const IR::U32U64 operand2 = V_scalar(*datasize, Vm);
const IR::U32U64 result = ir.FPMaxNumeric(operand1, operand2, true); const IR::U32U64 result = ir.FPMaxNumeric(operand1, operand2);
V_scalar(*datasize, Vd, result); V_scalar(*datasize, Vd, result);
return true; return true;
@ -121,7 +121,7 @@ bool TranslatorVisitor::FMINNM_float(Imm<2> type, Vec Vm, Vec Vn, Vec Vd) {
const IR::U32U64 operand1 = V_scalar(*datasize, Vn); const IR::U32U64 operand1 = V_scalar(*datasize, Vn);
const IR::U32U64 operand2 = V_scalar(*datasize, Vm); const IR::U32U64 operand2 = V_scalar(*datasize, Vm);
const IR::U32U64 result = ir.FPMinNumeric(operand1, operand2, true); const IR::U32U64 result = ir.FPMinNumeric(operand1, operand2);
V_scalar(*datasize, Vd, result); V_scalar(*datasize, Vd, result);
return true; return true;
@ -136,7 +136,7 @@ bool TranslatorVisitor::FNMUL_float(Imm<2> type, Vec Vm, Vec Vn, Vec Vd) {
const IR::U32U64 operand1 = V_scalar(*datasize, Vn); const IR::U32U64 operand1 = V_scalar(*datasize, Vn);
const IR::U32U64 operand2 = V_scalar(*datasize, Vm); const IR::U32U64 operand2 = V_scalar(*datasize, Vm);
const IR::U32U64 result = ir.FPNeg(ir.FPMul(operand1, operand2, true)); const IR::U32U64 result = ir.FPNeg(ir.FPMul(operand1, operand2));
V_scalar(*datasize, Vd, result); V_scalar(*datasize, Vd, result);
return true; return true;

8
src/frontend/A64/translate/impl/simd_across_lanes.cpp
View file

@ -70,13 +70,13 @@ bool FPMinMax(TranslatorVisitor& v, bool Q, bool sz, Vec Vn, Vec Vd, MinMaxOpera
const auto op = [&](const IR::U32U64& lhs, const IR::U32U64& rhs) { const auto op = [&](const IR::U32U64& lhs, const IR::U32U64& rhs) {
switch (operation) { switch (operation) {
case MinMaxOperation::Max: case MinMaxOperation::Max:
return v.ir.FPMax(lhs, rhs, true); return v.ir.FPMax(lhs, rhs);
case MinMaxOperation::MaxNumeric: case MinMaxOperation::MaxNumeric:
return v.ir.FPMaxNumeric(lhs, rhs, true); return v.ir.FPMaxNumeric(lhs, rhs);
case MinMaxOperation::Min: case MinMaxOperation::Min:
return v.ir.FPMin(lhs, rhs, true); return v.ir.FPMin(lhs, rhs);
case MinMaxOperation::MinNumeric: case MinMaxOperation::MinNumeric:
return v.ir.FPMinNumeric(lhs, rhs, true); return v.ir.FPMinNumeric(lhs, rhs);
default: default:
UNREACHABLE(); UNREACHABLE();
} }

10
src/frontend/A64/translate/impl/simd_scalar_pairwise.cpp
View file

@ -23,13 +23,13 @@ bool FPPairwiseMinMax(TranslatorVisitor& v, bool sz, Vec Vn, Vec Vd, MinMaxOpera
const IR::U32U64 result = [&] { const IR::U32U64 result = [&] {
switch (operation) { switch (operation) {
case MinMaxOperation::Max: case MinMaxOperation::Max:
return v.ir.FPMax(element1, element2, true); return v.ir.FPMax(element1, element2);
case MinMaxOperation::MaxNumeric: case MinMaxOperation::MaxNumeric:
return v.ir.FPMaxNumeric(element1, element2, true); return v.ir.FPMaxNumeric(element1, element2);
case MinMaxOperation::Min: case MinMaxOperation::Min:
return v.ir.FPMin(element1, element2, true); return v.ir.FPMin(element1, element2);
case MinMaxOperation::MinNumeric: case MinMaxOperation::MinNumeric:
return v.ir.FPMinNumeric(element1, element2, true); return v.ir.FPMinNumeric(element1, element2);
default: default:
UNREACHABLE(); UNREACHABLE();
} }
@ -57,7 +57,7 @@ bool TranslatorVisitor::FADDP_pair_2(bool size, Vec Vn, Vec Vd) {
const IR::U32U64 operand1 = ir.VectorGetElement(esize, V(128, Vn), 0); const IR::U32U64 operand1 = ir.VectorGetElement(esize, V(128, Vn), 0);
const IR::U32U64 operand2 = ir.VectorGetElement(esize, V(128, Vn), 1); const IR::U32U64 operand2 = ir.VectorGetElement(esize, V(128, Vn), 1);
const IR::U128 result = ir.ZeroExtendToQuad(ir.FPAdd(operand1, operand2, true)); const IR::U128 result = ir.ZeroExtendToQuad(ir.FPAdd(operand1, operand2));
V(128, Vd, result); V(128, Vd, result);
return true; return true;
} }

2
src/frontend/A64/translate/impl/simd_scalar_three_same.cpp
View file

@ -275,7 +275,7 @@ bool TranslatorVisitor::FABD_2(bool sz, Vec Vm, Vec Vn, Vec Vd) {
const IR::U32U64 operand1 = V_scalar(esize, Vn); const IR::U32U64 operand1 = V_scalar(esize, Vn);
const IR::U32U64 operand2 = V_scalar(esize, Vm); const IR::U32U64 operand2 = V_scalar(esize, Vm);
const IR::U32U64 result = ir.FPAbs(ir.FPSub(operand1, operand2, true)); const IR::U32U64 result = ir.FPAbs(ir.FPSub(operand1, operand2));
V_scalar(esize, Vd, result); V_scalar(esize, Vd, result);
return true; return true;

6
src/frontend/A64/translate/impl/simd_scalar_x_indexed_element.cpp
View file

@ -39,7 +39,7 @@ bool MultiplyByElement(TranslatorVisitor& v, bool sz, Imm<1> L, Imm<1> M, Imm<4>
IR::U32U64 operand1 = v.V_scalar(esize, Vn); IR::U32U64 operand1 = v.V_scalar(esize, Vn);
if (extra_behavior == ExtraBehavior::None) { if (extra_behavior == ExtraBehavior::None) {
return v.ir.FPMul(operand1, element, true); return v.ir.FPMul(operand1, element);
} }
if (extra_behavior == ExtraBehavior::MultiplyExtended) { if (extra_behavior == ExtraBehavior::MultiplyExtended) {
@ -51,7 +51,7 @@ bool MultiplyByElement(TranslatorVisitor& v, bool sz, Imm<1> L, Imm<1> M, Imm<4>
} }
const IR::U32U64 operand2 = v.V_scalar(esize, Vd); const IR::U32U64 operand2 = v.V_scalar(esize, Vd);
return v.ir.FPMulAdd(operand2, operand1, element, true); return v.ir.FPMulAdd(operand2, operand1, element);
}(); }();
v.V_scalar(esize, Vd, result); v.V_scalar(esize, Vd, result);
@ -85,7 +85,7 @@ bool MultiplyByElementHalfPrecision(TranslatorVisitor& v, Imm<1> L, Imm<1> M, Im
} }
const IR::U16 operand2 = v.V_scalar(esize, Vd); const IR::U16 operand2 = v.V_scalar(esize, Vd);
return v.ir.FPMulAdd(operand2, operand1, element, true); return v.ir.FPMulAdd(operand2, operand1, element);
}(); }();
v.V_scalar(esize, Vd, result); v.V_scalar(esize, Vd, result);

12
src/frontend/A64/translate/impl/simd_three_same_extra.cpp
View file

@ -115,10 +115,8 @@ bool TranslatorVisitor::FCMLA_vec(bool Q, Imm<2> size, Vec Vm, Imm<2> rot, Vec V
const IR::U32U64 operand3_elem1 = ir.VectorGetElement(esize, operand3, first); const IR::U32U64 operand3_elem1 = ir.VectorGetElement(esize, operand3, first);
const IR::U32U64 operand3_elem2 = ir.VectorGetElement(esize, operand3, second); const IR::U32U64 operand3_elem2 = ir.VectorGetElement(esize, operand3, second);
result = ir.VectorSetElement(esize, result, first, result = ir.VectorSetElement(esize, result, first, ir.FPMulAdd(operand3_elem1, element2, element1));
ir.FPMulAdd(operand3_elem1, element2, element1, true)); result = ir.VectorSetElement(esize, result, second, ir.FPMulAdd(operand3_elem2, element4, element3));
result = ir.VectorSetElement(esize, result, second,
ir.FPMulAdd(operand3_elem2, element4, element3, true));
} }
ir.SetQ(Vd, result); ir.SetQ(Vd, result);
@ -166,10 +164,8 @@ bool TranslatorVisitor::FCADD_vec(bool Q, Imm<2> size, Vec Vm, Imm<1> rot, Vec V
const IR::U32U64 operand1_elem1 = ir.VectorGetElement(esize, operand1, first); const IR::U32U64 operand1_elem1 = ir.VectorGetElement(esize, operand1, first);
const IR::U32U64 operand1_elem3 = ir.VectorGetElement(esize, operand1, second); const IR::U32U64 operand1_elem3 = ir.VectorGetElement(esize, operand1, second);
result = ir.VectorSetElement(esize, result, first, result = ir.VectorSetElement(esize, result, first, ir.FPAdd(operand1_elem1, element1));
ir.FPAdd(operand1_elem1, element1, true)); result = ir.VectorSetElement(esize, result, second, ir.FPAdd(operand1_elem3, element3));
result = ir.VectorSetElement(esize, result, second,
ir.FPAdd(operand1_elem3, element3, true));
} }
ir.SetQ(Vd, result); ir.SetQ(Vd, result);

6
src/frontend/A64/translate/impl/simd_vector_x_indexed_element.cpp
View file

@ -296,10 +296,8 @@ bool TranslatorVisitor::FCMLA_elt(bool Q, Imm<2> size, Imm<1> L, Imm<1> M, Imm<4
const IR::U32U64 operand3_elem1 = ir.VectorGetElement(esize, operand3, first); const IR::U32U64 operand3_elem1 = ir.VectorGetElement(esize, operand3, first);
const IR::U32U64 operand3_elem2 = ir.VectorGetElement(esize, operand3, second); const IR::U32U64 operand3_elem2 = ir.VectorGetElement(esize, operand3, second);
result = ir.VectorSetElement(esize, result, first, result = ir.VectorSetElement(esize, result, first, ir.FPMulAdd(operand3_elem1, element2, element1));
ir.FPMulAdd(operand3_elem1, element2, element1, true)); result = ir.VectorSetElement(esize, result, second, ir.FPMulAdd(operand3_elem2, element4, element3));
result = ir.VectorSetElement(esize, result, second,
ir.FPMulAdd(operand3_elem2, element4, element3, true));
} }
ir.SetQ(Vd, result); ir.SetQ(Vd, result);

20
src/frontend/ir/ir_emitter.h
View file

@ -310,15 +310,15 @@ public:
U128 ZeroVector(); U128 ZeroVector();
U16U32U64 FPAbs(const U16U32U64& a); U16U32U64 FPAbs(const U16U32U64& a);
U32U64 FPAdd(const U32U64& a, const U32U64& b, bool fpcr_controlled); U32U64 FPAdd(const U32U64& a, const U32U64& b, bool fpcr_controlled = true);
NZCV FPCompare(const U32U64& a, const U32U64& b, bool exc_on_qnan, bool fpcr_controlled); NZCV FPCompare(const U32U64& a, const U32U64& b, bool exc_on_qnan, bool fpcr_controlled = true);
U32U64 FPDiv(const U32U64& a, const U32U64& b, bool fpcr_controlled); U32U64 FPDiv(const U32U64& a, const U32U64& b, bool fpcr_controlled = true);
U32U64 FPMax(const U32U64& a, const U32U64& b, bool fpcr_controlled); U32U64 FPMax(const U32U64& a, const U32U64& b, bool fpcr_controlled = true);
U32U64 FPMaxNumeric(const U32U64& a, const U32U64& b, bool fpcr_controlled); U32U64 FPMaxNumeric(const U32U64& a, const U32U64& b, bool fpcr_controlled = true);
U32U64 FPMin(const U32U64& a, const U32U64& b, bool fpcr_controlled); U32U64 FPMin(const U32U64& a, const U32U64& b, bool fpcr_controlled = true);
U32U64 FPMinNumeric(const U32U64& a, const U32U64& b, bool fpcr_controlled); U32U64 FPMinNumeric(const U32U64& a, const U32U64& b, bool fpcr_controlled = true);
U32U64 FPMul(const U32U64& a, const U32U64& b, bool fpcr_controlled); U32U64 FPMul(const U32U64& a, const U32U64& b, bool fpcr_controlled = true);
U16U32U64 FPMulAdd(const U16U32U64& addend, const U16U32U64& op1, const U16U32U64& op2, bool fpcr_controlled); U16U32U64 FPMulAdd(const U16U32U64& addend, const U16U32U64& op1, const U16U32U64& op2, bool fpcr_controlled = true);
U32U64 FPMulX(const U32U64& a, const U32U64& b); U32U64 FPMulX(const U32U64& a, const U32U64& b);
U16U32U64 FPNeg(const U16U32U64& a); U16U32U64 FPNeg(const U16U32U64& a);
U16U32U64 FPRecipEstimate(const U16U32U64& a); U16U32U64 FPRecipEstimate(const U16U32U64& a);
@ -328,7 +328,7 @@ public:
U16U32U64 FPRSqrtEstimate(const U16U32U64& a); U16U32U64 FPRSqrtEstimate(const U16U32U64& a);
U16U32U64 FPRSqrtStepFused(const U16U32U64& a, const U16U32U64& b); U16U32U64 FPRSqrtStepFused(const U16U32U64& a, const U16U32U64& b);
U32U64 FPSqrt(const U32U64& a); U32U64 FPSqrt(const U32U64& a);
U32U64 FPSub(const U32U64& a, const U32U64& b, bool fpcr_controlled); U32U64 FPSub(const U32U64& a, const U32U64& b, bool fpcr_controlled = true);
U16 FPDoubleToHalf(const U64& a, FP::RoundingMode rounding); U16 FPDoubleToHalf(const U64& a, FP::RoundingMode rounding);
U32 FPDoubleToSingle(const U64& a, FP::RoundingMode rounding); U32 FPDoubleToSingle(const U64& a, FP::RoundingMode rounding);
U64 FPHalfToDouble(const U16& a, FP::RoundingMode rounding); U64 FPHalfToDouble(const U16& a, FP::RoundingMode rounding);