void Interpreter::ps_abs(UGeckoInstruction inst)
{
const auto& b = rPS(inst.FB);
rPS(inst.FD).SetBoth(b.PS0AsU64() & ~(UINT64_C(1) << 63), b.PS1AsU64() & ~(UINT64_C(1) << 63));
if (inst.Rc)
Helper_UpdateCR1();
}
void Interpreter::ps_merge11(UGeckoInstruction inst)
{
const auto& a = rPS(inst.FA);
const auto& b = rPS(inst.FB);
rPS(inst.FD).SetBoth(a.PS1AsDouble(), b.PS1AsDouble());
if (inst.Rc)
Helper_UpdateCR1();
}
// These "binary instructions" do not alter FPSCR.
void Interpreter::ps_sel(UGeckoInstruction inst)
{
const auto& a = rPS(inst.FA);
const auto& b = rPS(inst.FB);
const auto& c = rPS(inst.FC);
rPS(inst.FD).SetBoth(a.PS0AsDouble() >= -0.0 ? c.PS0AsDouble() : b.PS0AsDouble(),
a.PS1AsDouble() >= -0.0 ? c.PS1AsDouble() : b.PS1AsDouble());
if (inst.Rc)
Helper_UpdateCR1();
}
void JitArm::faddx(UGeckoInstruction inst)
{
INSTRUCTION_START
JITDISABLE(FloatingPoint)
ARMReg vD = fpr.R0(inst.FD);
ARMReg vA = fpr.R0(inst.FA);
ARMReg vB = fpr.R0(inst.FB);
VADD(vD, vA, vB);
if (inst.Rc) Helper_UpdateCR1(vD);
}
void JitArm::fmulx(UGeckoInstruction inst)
{
INSTRUCTION_START
JITDISABLE(FloatingPoint)
ARMReg vD0 = fpr.R0(inst.FD);
ARMReg vA = fpr.R0(inst.FA);
ARMReg vC = fpr.R0(inst.FC);
VMUL(vD0, vA, vC);
if (inst.Rc) Helper_UpdateCR1(vD0);
}
void JitArm::fmrx(UGeckoInstruction inst)
{
INSTRUCTION_START
JITDISABLE(FloatingPoint)
Default(inst); return;
ARMReg vD = fpr.R0(inst.FD);
ARMReg vB = fpr.R0(inst.FB);
VMOV(vD, vB);
if (inst.Rc) Helper_UpdateCR1(vD);
}
void Interpreter::ps_add(UGeckoInstruction inst)
{
const auto& a = rPS(inst.FA);
const auto& b = rPS(inst.FB);
const double ps0 = ForceSingle(NI_add(a.PS0AsDouble(), b.PS0AsDouble()).value);
const double ps1 = ForceSingle(NI_add(a.PS1AsDouble(), b.PS1AsDouble()).value);
rPS(inst.FD).SetBoth(ps0, ps1);
PowerPC::UpdateFPRF(ps0);
if (inst.Rc)
Helper_UpdateCR1();
}
void JitArm::fsubsx(UGeckoInstruction inst)
{
INSTRUCTION_START
JITDISABLE(FloatingPoint)
ARMReg vA = fpr.R0(inst.FA);
ARMReg vB = fpr.R0(inst.FB);
ARMReg vD0 = fpr.R0(inst.FD);
ARMReg vD1 = fpr.R1(inst.FD);
VSUB(vD0, vA, vB);
VMOV(vD1, vD0);
if (inst.Rc) Helper_UpdateCR1(vD0);
}
void Interpreter::ps_muls1(UGeckoInstruction inst)
{
const auto& a = rPS(inst.FA);
const auto& c = rPS(inst.FC);
const double c1 = Force25Bit(c.PS1AsDouble());
const double ps0 = ForceSingle(NI_mul(a.PS0AsDouble(), c1).value);
const double ps1 = ForceSingle(NI_mul(a.PS1AsDouble(), c1).value);
rPS(inst.FD).SetBoth(ps0, ps1);
PowerPC::UpdateFPRF(ps0);
if (inst.Rc)
Helper_UpdateCR1();
}
void Interpreter::ps_nmadd(UGeckoInstruction inst)
{
const auto& a = rPS(inst.FA);
const auto& b = rPS(inst.FB);
const auto& c = rPS(inst.FC);
const double c0 = Force25Bit(c.PS0AsDouble());
const double c1 = Force25Bit(c.PS1AsDouble());
const double tmp0 = ForceSingle(NI_madd(a.PS0AsDouble(), c0, b.PS0AsDouble()).value);
const double tmp1 = ForceSingle(NI_madd(a.PS1AsDouble(), c1, b.PS1AsDouble()).value);
const double ps0 = std::isnan(tmp0) ? tmp0 : -tmp0;
const double ps1 = std::isnan(tmp1) ? tmp1 : -tmp1;
rPS(inst.FD).SetBoth(ps0, ps1);
PowerPC::UpdateFPRF(ps0);
if (inst.Rc)
Helper_UpdateCR1();
}
void JitArm::fctiwx(UGeckoInstruction inst)
{
INSTRUCTION_START
JITDISABLE(bJITFloatingPointOff)
u32 b = inst.FB;
u32 d = inst.FD;
ARMReg vB = fpr.R0(b);
ARMReg vD = fpr.R0(d);
ARMReg V0 = fpr.GetReg();
ARMReg V1 = fpr.GetReg();
ARMReg V2 = fpr.GetReg();
ARMReg rA = gpr.GetReg();
ARMReg fpscrReg = gpr.GetReg();
FixupBranch DoneMax, DoneMin;
LDR(fpscrReg, R9, PPCSTATE_OFF(fpscr));
MOVI2R(rA, (u32)minmaxFloat);
// Check if greater than max float
{
VLDR(V0, rA, 8); // Load Max
VCMPE(vB, V0);
VMRS(_PC); // Loads in to APSR
FixupBranch noException = B_CC(CC_LE);
VMOV(vD, V0); // Set to max
SetFPException(fpscrReg, FPSCR_VXCVI);
DoneMax = B();
SetJumpTarget(noException);
}
// Check if less than min float
{
VLDR(V0, rA, 0);
VCMPE(vB, V0);
VMRS(_PC);
FixupBranch noException = B_CC(CC_GE);
VMOV(vD, V0);
SetFPException(fpscrReg, FPSCR_VXCVI);
DoneMin = B();
SetJumpTarget(noException);
}
// Within ranges, convert to integer
// Set rounding mode first
// PPC <-> ARM rounding modes
// 0, 1, 2, 3 <-> 0, 3, 1, 2
ARMReg rB = gpr.GetReg();
VMRS(rA);
// Bits 22-23
BIC(rA, rA, Operand2(3, 5));
LDR(rB, R9, PPCSTATE_OFF(fpscr));
AND(rB, rB, 0x3); // Get the FPSCR rounding bits
CMP(rB, 1);
SetCC(CC_EQ); // zero
ORR(rA, rA, Operand2(3, 5));
SetCC(CC_NEQ);
CMP(rB, 2); // +inf
SetCC(CC_EQ);
ORR(rA, rA, Operand2(1, 5));
SetCC(CC_NEQ);
CMP(rB, 3); // -inf
SetCC(CC_EQ);
ORR(rA, rA, Operand2(2, 5));
SetCC();
VMSR(rA);
ORR(rA, rA, Operand2(3, 5));
VCVT(vD, vB, TO_INT | IS_SIGNED);
VMSR(rA);
gpr.Unlock(rB);
VCMPE(vD, vB);
VMRS(_PC);
SetCC(CC_EQ);
BIC(fpscrReg, fpscrReg, FRFIMask);
FixupBranch DoneEqual = B();
SetCC();
SetFPException(fpscrReg, FPSCR_XX);
ORR(fpscrReg, fpscrReg, FIMask);
VABS(V1, vB);
VABS(V2, vD);
VCMPE(V2, V1);
VMRS(_PC);
SetCC(CC_GT);
ORR(fpscrReg, fpscrReg, FRMask);
SetCC();
SetJumpTarget(DoneEqual);
SetJumpTarget(DoneMax);
SetJumpTarget(DoneMin);
MOVI2R(rA, (u32)&doublenum);
VLDR(V0, rA, 0);
NEONXEmitter nemit(this);
nemit.VORR(vD, vD, V0);
if (inst.Rc) Helper_UpdateCR1(fpscrReg, rA);
STR(fpscrReg, R9, PPCSTATE_OFF(fpscr));
gpr.Unlock(rA);
gpr.Unlock(fpscrReg);
fpr.Unlock(V0);
fpr.Unlock(V1);
fpr.Unlock(V2);
}
void JitArm::fctiwzx(UGeckoInstruction inst)
{
INSTRUCTION_START
JITDISABLE(bJITFloatingPointOff)
u32 b = inst.FB;
u32 d = inst.FD;
ARMReg vB = fpr.R0(b);
ARMReg vD = fpr.R0(d);
ARMReg V0 = fpr.GetReg();
ARMReg V1 = fpr.GetReg();
ARMReg V2 = fpr.GetReg();
ARMReg rA = gpr.GetReg();
ARMReg fpscrReg = gpr.GetReg();
FixupBranch DoneMax, DoneMin;
LDR(fpscrReg, R9, PPCSTATE_OFF(fpscr));
MOVI2R(rA, (u32)minmaxFloat);
// Check if greater than max float
{
VLDR(V0, rA, 8); // Load Max
VCMPE(vB, V0);
VMRS(_PC); // Loads in to APSR
FixupBranch noException = B_CC(CC_LE);
VMOV(vD, V0); // Set to max
SetFPException(fpscrReg, FPSCR_VXCVI);
DoneMax = B();
SetJumpTarget(noException);
}
// Check if less than min float
{
VLDR(V0, rA, 0);
VCMPE(vB, V0);
VMRS(_PC);
FixupBranch noException = B_CC(CC_GE);
VMOV(vD, V0);
SetFPException(fpscrReg, FPSCR_VXCVI);
DoneMin = B();
SetJumpTarget(noException);
}
// Within ranges, convert to integer
VCVT(vD, vB, TO_INT | IS_SIGNED | ROUND_TO_ZERO);
VCMPE(vD, vB);
VMRS(_PC);
SetCC(CC_EQ);
BIC(fpscrReg, fpscrReg, FRFIMask);
FixupBranch DoneEqual = B();
SetCC();
SetFPException(fpscrReg, FPSCR_XX);
ORR(fpscrReg, fpscrReg, FIMask);
VABS(V1, vB);
VABS(V2, vD);
VCMPE(V2, V1);
VMRS(_PC);
SetCC(CC_GT);
ORR(fpscrReg, fpscrReg, FRMask);
SetCC();
SetJumpTarget(DoneEqual);
SetJumpTarget(DoneMax);
SetJumpTarget(DoneMin);
MOVI2R(rA, (u32)&doublenum);
VLDR(V0, rA, 0);
NEONXEmitter nemit(this);
nemit.VORR(vD, vD, V0);
if (inst.Rc) Helper_UpdateCR1(fpscrReg, rA);
STR(fpscrReg, R9, PPCSTATE_OFF(fpscr));
gpr.Unlock(rA);
gpr.Unlock(fpscrReg);
fpr.Unlock(V0);
fpr.Unlock(V1);
fpr.Unlock(V2);
}
