asimd_three_regs: Reimplement asimd_VMLAL in terms of WideInstruction

This commit is contained in:
Merry 2020-06-27 13:02:58 +01:00
parent 7997404ee7
commit 4f967387c0

View file

@ -220,27 +220,27 @@ bool AbsoluteDifferenceLong(ArmTranslatorVisitor& v, bool U, bool D, size_t sz,
template <typename Callable> template <typename Callable>
bool WideInstruction(ArmTranslatorVisitor& v, bool U, bool D, size_t sz, size_t Vn, size_t Vd, bool N, bool M, size_t Vm, WidenBehaviour widen_behaviour, Callable fn) { bool WideInstruction(ArmTranslatorVisitor& v, bool U, bool D, size_t sz, size_t Vn, size_t Vd, bool N, bool M, size_t Vm, WidenBehaviour widen_behaviour, Callable fn) {
const size_t esize = 8U << sz; const size_t esize = 8U << sz;
const bool widen_second = widen_behaviour == WidenBehaviour::Both; const bool widen_first = widen_behaviour == WidenBehaviour::Both;
if (sz == 0b11) { if (sz == 0b11) {
// Decode error // Decode error
return v.UndefinedInstruction(); return v.UndefinedInstruction();
} }
if (Common::Bit<0>(Vd) || (!widen_second && Common::Bit<0>(Vn))) { if (Common::Bit<0>(Vd) || (!widen_first && Common::Bit<0>(Vn))) {
return v.UndefinedInstruction(); return v.UndefinedInstruction();
} }
const auto d = ToVector(true, Vd, D); const auto d = ToVector(true, Vd, D);
const auto m = ToVector(false, Vm, M); const auto m = ToVector(false, Vm, M);
const auto n = ToVector(!widen_second, Vn, N); const auto n = ToVector(!widen_first, Vn, N);
const auto reg_d = v.ir.GetVector(d);
const auto reg_m = v.ir.GetVector(m); const auto reg_m = v.ir.GetVector(m);
const auto reg_n = v.ir.GetVector(n); const auto reg_n = v.ir.GetVector(n);
const auto operand_n = widen_second ? (U ? v.ir.VectorZeroExtend(esize, reg_n) : v.ir.VectorSignExtend(esize, reg_n)) const auto wide_n = U ? v.ir.VectorZeroExtend(esize, reg_n) : v.ir.VectorSignExtend(esize, reg_n);
: reg_n; const auto wide_m = U ? v.ir.VectorZeroExtend(esize, reg_m) : v.ir.VectorSignExtend(esize, reg_m);
const auto operand_m = U ? v.ir.VectorZeroExtend(esize, reg_m) : v.ir.VectorSignExtend(esize, reg_m); const auto result = fn(esize * 2, reg_d, widen_first ? wide_n : reg_n, wide_m);
const auto result = fn(esize * 2, operand_n, operand_m);
v.ir.SetVector(d, result); v.ir.SetVector(d, result);
return true; return true;
@ -786,13 +786,13 @@ bool ArmTranslatorVisitor::asimd_VRSQRTS(bool D, bool sz, size_t Vn, size_t Vd,
// ASIMD Three registers of different length // ASIMD Three registers of different length
bool ArmTranslatorVisitor::asimd_VADDL(bool U, bool D, size_t sz, size_t Vn, size_t Vd, bool op, bool N, bool M, size_t Vm) { bool ArmTranslatorVisitor::asimd_VADDL(bool U, bool D, size_t sz, size_t Vn, size_t Vd, bool op, bool N, bool M, size_t Vm) {
return WideInstruction(*this, U, D, sz, Vn, Vd, N, M, Vm, op ? WidenBehaviour::Second : WidenBehaviour::Both, [this](size_t esize, const auto& reg_n, const auto& reg_m) { return WideInstruction(*this, U, D, sz, Vn, Vd, N, M, Vm, op ? WidenBehaviour::Second : WidenBehaviour::Both, [this](size_t esize, const auto&, const auto& reg_n, const auto& reg_m) {
return ir.VectorAdd(esize, reg_n, reg_m); return ir.VectorAdd(esize, reg_n, reg_m);
}); });
} }
bool ArmTranslatorVisitor::asimd_VSUBL(bool U, bool D, size_t sz, size_t Vn, size_t Vd, bool op, bool N, bool M, size_t Vm) { bool ArmTranslatorVisitor::asimd_VSUBL(bool U, bool D, size_t sz, size_t Vn, size_t Vd, bool op, bool N, bool M, size_t Vm) {
return WideInstruction(*this, U, D, sz, Vn, Vd, N, M, Vm, op ? WidenBehaviour::Second : WidenBehaviour::Both, [this](size_t esize, const auto& reg_n, const auto& reg_m) { return WideInstruction(*this, U, D, sz, Vn, Vd, N, M, Vm, op ? WidenBehaviour::Second : WidenBehaviour::Both, [this](size_t esize, const auto&, const auto& reg_n, const auto& reg_m) {
return ir.VectorSub(esize, reg_n, reg_m); return ir.VectorSub(esize, reg_n, reg_m);
}); });
} }
@ -806,33 +806,11 @@ bool ArmTranslatorVisitor::asimd_VABDL(bool U, bool D, size_t sz, size_t Vn, siz
} }
bool ArmTranslatorVisitor::asimd_VMLAL(bool U, bool D, size_t sz, size_t Vn, size_t Vd, bool op, bool N, bool M, size_t Vm) { bool ArmTranslatorVisitor::asimd_VMLAL(bool U, bool D, size_t sz, size_t Vn, size_t Vd, bool op, bool N, bool M, size_t Vm) {
if (sz == 0b11) { return WideInstruction(*this, U, D, sz, Vn, Vd, N, M, Vm, WidenBehaviour::Both, [this, op](size_t esize, const auto& reg_d, const auto& reg_n, const auto& reg_m) {
return UndefinedInstruction(); const auto multiply = ir.VectorMultiply(esize, reg_n, reg_m);
} return op ? ir.VectorSub(esize, reg_d, multiply)
: ir.VectorAdd(esize, reg_d, multiply);
if (Common::Bit<0>(Vd)) { });
return UndefinedInstruction();
}
const size_t esize = 8U << sz;
const auto d = ToVector(true, Vd, D);
const auto m = ToVector(false, Vm, M);
const auto n = ToVector(false, Vn, N);
const auto extend_reg = [&](const auto& reg) {
return U ? ir.VectorZeroExtend(esize, reg)
: ir.VectorSignExtend(esize, reg);
};
const auto reg_d = ir.GetVector(d);
const auto reg_n = ir.GetVector(n);
const auto reg_m = ir.GetVector(m);
const auto multiply = ir.VectorMultiply(2 * esize, extend_reg(reg_n), extend_reg(reg_m));
const auto result = op ? ir.VectorSub(2 * esize, reg_d, multiply)
: ir.VectorAdd(2 * esize, reg_d, multiply);
ir.SetVector(d, result);
return true;
} }
bool ArmTranslatorVisitor::asimd_VMULL(bool U, bool D, size_t sz, size_t Vn, size_t Vd, bool P, bool N, bool M, size_t Vm) { bool ArmTranslatorVisitor::asimd_VMULL(bool U, bool D, size_t sz, size_t Vn, size_t Vd, bool P, bool N, bool M, size_t Vm) {