simd_permute: Implement TRN{1,2} in terms of VectorTranspose
This commit is contained in:
parent
7d1e103ff5
commit
a8b481ab63
1 changed files with 2 additions and 40 deletions
|
@ -13,8 +13,7 @@ enum class Transposition {
|
||||||
TRN2,
|
TRN2,
|
||||||
};
|
};
|
||||||
|
|
||||||
bool VectorTranspose(TranslatorVisitor& v, bool Q, Imm<2> size, Vec Vm, Vec Vn, Vec Vd,
|
bool VectorTranspose(TranslatorVisitor& v, bool Q, Imm<2> size, Vec Vm, Vec Vn, Vec Vd, Transposition type) {
|
||||||
Transposition type) {
|
|
||||||
if (!Q && size == 0b11) {
|
if (!Q && size == 0b11) {
|
||||||
return v.ReservedValue();
|
return v.ReservedValue();
|
||||||
}
|
}
|
||||||
|
@ -24,44 +23,7 @@ bool VectorTranspose(TranslatorVisitor& v, bool Q, Imm<2> size, Vec Vm, Vec Vn,
|
||||||
|
|
||||||
const IR::U128 m = v.V(datasize, Vm);
|
const IR::U128 m = v.V(datasize, Vm);
|
||||||
const IR::U128 n = v.V(datasize, Vn);
|
const IR::U128 n = v.V(datasize, Vn);
|
||||||
|
const IR::U128 result = v.ir.VectorTranspose(esize, n, m, type == Transposition::TRN2);
|
||||||
const IR::U128 result = [&] {
|
|
||||||
switch (esize) {
|
|
||||||
case 8:
|
|
||||||
case 16:
|
|
||||||
case 32: {
|
|
||||||
// Create a mask of elements we care about (e.g. for 8-bit: 0x00FF00FF00FF00FF for TRN1
|
|
||||||
// and 0xFF00FF00FF00FF00 for TRN2)
|
|
||||||
const u64 mask_element = [&] {
|
|
||||||
const size_t shift = type == Transposition::TRN1 ? 0 : esize;
|
|
||||||
return Common::Ones<u64>(esize) << shift;
|
|
||||||
}();
|
|
||||||
const size_t doubled_esize = esize * 2;
|
|
||||||
const u64 mask_value = Common::Replicate<u64>(mask_element, doubled_esize);
|
|
||||||
|
|
||||||
const IR::U128 mask = v.ir.VectorBroadcast(64, v.I(64, mask_value));
|
|
||||||
const IR::U128 anded_m = v.ir.VectorAnd(m, mask);
|
|
||||||
const IR::U128 anded_n = v.ir.VectorAnd(n, mask);
|
|
||||||
|
|
||||||
if (type == Transposition::TRN1) {
|
|
||||||
return v.ir.VectorOr(v.ir.VectorLogicalShiftLeft(doubled_esize, anded_m, esize), anded_n);
|
|
||||||
}
|
|
||||||
|
|
||||||
return v.ir.VectorOr(v.ir.VectorLogicalShiftRight(doubled_esize, anded_n, esize), anded_m);
|
|
||||||
}
|
|
||||||
case 64: {
|
|
||||||
default:
|
|
||||||
const auto [src, src_idx, dst, dst_idx] = [type, m, n] {
|
|
||||||
if (type == Transposition::TRN1) {
|
|
||||||
return std::make_tuple(m, 0, n, 1);
|
|
||||||
}
|
|
||||||
return std::make_tuple(n, 1, m, 0);
|
|
||||||
}();
|
|
||||||
|
|
||||||
return v.ir.VectorSetElement(esize, dst, dst_idx, v.ir.VectorGetElement(esize, src, src_idx));
|
|
||||||
}
|
|
||||||
}
|
|
||||||
}();
|
|
||||||
|
|
||||||
v.V(datasize, Vd, result);
|
v.V(datasize, Vd, result);
|
||||||
return true;
|
return true;
|
||||||
|
|
Loading…
Reference in a new issue