suyu/dynarmic
28
1
Fork 1
Code Issues Pull requests Projects Releases Packages Wiki Activity

A64: Implement SM3TT1B

Browse source
This commit is contained in:
Lioncash 2018-04-10 12:16:10 -04:00 committed by MerryMage
parent 769373b3ed
commit 7fbccabd81
2 changed files with 42 additions and 20 deletions
Show stats Download patch file Download diff file Expand all files Collapse all files

2
src/frontend/A64/decoder/a64.inc
View file

@ -853,7 +853,7 @@ INST(USHLL, "USHLL, USHLL2", "0Q101
// Data Processing - FP and SIMD - Cryptographic three register
INST(SM3TT1A, "SM3TT1A", "11001110010mmmmm10ii00nnnnnddddd")
//INST(SM3TT1B, "SM3TT1B", "11001110010mmmmm10ii01nnnnnddddd")
INST(SM3TT1B, "SM3TT1B", "11001110010mmmmm10ii01nnnnnddddd")
//INST(SM3TT2A, "SM3TT2A", "11001110010mmmmm10ii10nnnnnddddd")
//INST(SM3TT2B, "SM3TT2B", "11001110010mmmmm10ii11nnnnnddddd")

60
src/frontend/A64/translate/impl/simd_crypto_three_register.cpp
View file

@ -8,30 +8,52 @@
namespace Dynarmic::A64 {
bool TranslatorVisitor::SM3TT1A(Vec Vm, Imm<2> imm2, Vec Vn, Vec Vd) {
const IR::U128 d = ir.GetQ(Vd);
const IR::U128 m = ir.GetQ(Vm);
const IR::U128 n = ir.GetQ(Vn);
enum class SM3TT1Variant {
A,
B,
};
static void SM3TT1(TranslatorVisitor& v, Vec Vm, Imm<2> imm2, Vec Vn, Vec Vd, SM3TT1Variant behavior) {
const IR::U128 d = v.ir.GetQ(Vd);
const IR::U128 m = v.ir.GetQ(Vm);
const IR::U128 n = v.ir.GetQ(Vn);
const u32 index = imm2.ZeroExtend();
const IR::U32 top_d = ir.VectorGetElement(32, d, 3);
const IR::U32 before_top_d = ir.VectorGetElement(32, d, 2);
const IR::U32 after_low_d = ir.VectorGetElement(32, d, 1);
const IR::U32 low_d = ir.VectorGetElement(32, d, 0);
const IR::U32 top_n = ir.VectorGetElement(32, n, 3);
const IR::U32 top_d = v.ir.VectorGetElement(32, d, 3);
const IR::U32 before_top_d = v.ir.VectorGetElement(32, d, 2);
const IR::U32 after_low_d = v.ir.VectorGetElement(32, d, 1);
const IR::U32 low_d = v.ir.VectorGetElement(32, d, 0);
const IR::U32 top_n = v.ir.VectorGetElement(32, n, 3);
const IR::U32 wj_prime = ir.VectorGetElement(32, m, index);
const IR::U32 ss2 = ir.Eor(top_n, ir.RotateRight(top_d, ir.Imm8(20)));
const IR::U32 tt1 = ir.Eor(after_low_d, ir.Eor(top_d, before_top_d));
const IR::U32 final_tt1 = ir.Add(tt1, ir.Add(low_d, ir.Add(ss2, wj_prime)));
const IR::U32 wj_prime = v.ir.VectorGetElement(32, m, index);
const IR::U32 ss2 = v.ir.Eor(top_n, v.ir.RotateRight(top_d, v.ir.Imm8(20)));
const IR::U32 tt1 = [&] {
if (behavior == SM3TT1Variant::A) {
return v.ir.Eor(after_low_d, v.ir.Eor(top_d, before_top_d));
}
const IR::U32 tmp1 = v.ir.And(top_d, after_low_d);
const IR::U32 tmp2 = v.ir.And(top_d, before_top_d);
const IR::U32 tmp3 = v.ir.And(after_low_d, before_top_d);
return v.ir.Or(v.ir.Or(tmp1, tmp2), tmp3);
}();
const IR::U32 final_tt1 = v.ir.Add(tt1, v.ir.Add(low_d, v.ir.Add(ss2, wj_prime)));
const IR::U128 zero_vector = ir.ZeroVector();
const IR::U128 tmp1 = ir.VectorSetElement(32, zero_vector, 0, after_low_d);
const IR::U128 tmp2 = ir.VectorSetElement(32, tmp1, 1, ir.RotateRight(before_top_d, ir.Imm8(23)));
const IR::U128 tmp3 = ir.VectorSetElement(32, tmp2, 2, top_d);
const IR::U128 result = ir.VectorSetElement(32, tmp3, 3, final_tt1);
const IR::U128 zero_vector = v.ir.ZeroVector();
const IR::U128 tmp1 = v.ir.VectorSetElement(32, zero_vector, 0, after_low_d);
const IR::U128 tmp2 = v.ir.VectorSetElement(32, tmp1, 1, v.ir.RotateRight(before_top_d, v.ir.Imm8(23)));
const IR::U128 tmp3 = v.ir.VectorSetElement(32, tmp2, 2, top_d);
const IR::U128 result = v.ir.VectorSetElement(32, tmp3, 3, final_tt1);
ir.SetQ(Vd, result);
v.ir.SetQ(Vd, result);
}
bool TranslatorVisitor::SM3TT1A(Vec Vm, Imm<2> imm2, Vec Vn, Vec Vd) {
SM3TT1(*this, Vm, imm2, Vn, Vd, SM3TT1Variant::A);
return true;
}
bool TranslatorVisitor::SM3TT1B(Vec Vm, Imm<2> imm2, Vec Vn, Vec Vd) {
SM3TT1(*this, Vm, imm2, Vn, Vd, SM3TT1Variant::B);
return true;
}