A64: Implement FMUL (by element)'s scalar double/single-precision variant

This commit is contained in:
Lioncash 2018-07-22 17:39:16 -04:00 committed by MerryMage
parent 0ce11b7b15
commit c704acafe4
4 changed files with 33 additions and 3 deletions

View file

@ -152,6 +152,7 @@ add_library(dynarmic
frontend/A64/translate/impl/simd_scalar_shift_by_immediate.cpp frontend/A64/translate/impl/simd_scalar_shift_by_immediate.cpp
frontend/A64/translate/impl/simd_scalar_three_same.cpp frontend/A64/translate/impl/simd_scalar_three_same.cpp
frontend/A64/translate/impl/simd_scalar_two_register_misc.cpp frontend/A64/translate/impl/simd_scalar_two_register_misc.cpp
frontend/A64/translate/impl/simd_scalar_x_indexed_element.cpp
frontend/A64/translate/impl/simd_sha.cpp frontend/A64/translate/impl/simd_sha.cpp
frontend/A64/translate/impl/simd_sha512.cpp frontend/A64/translate/impl/simd_sha512.cpp
frontend/A64/translate/impl/simd_shift_by_immediate.cpp frontend/A64/translate/impl/simd_shift_by_immediate.cpp

View file

@ -504,7 +504,7 @@ INST(FCVTZU_fix_1, "FCVTZU (vector, fixed-point)", "01111
//INST(FMLS_elt_1, "FMLS (by element)", "0101111100LMmmmm0101H0nnnnnddddd") //INST(FMLS_elt_1, "FMLS (by element)", "0101111100LMmmmm0101H0nnnnnddddd")
//INST(FMLS_elt_2, "FMLS (by element)", "010111111zLMmmmm0101H0nnnnnddddd") //INST(FMLS_elt_2, "FMLS (by element)", "010111111zLMmmmm0101H0nnnnnddddd")
//INST(FMUL_elt_1, "FMUL (by element)", "0101111100LMmmmm1001H0nnnnnddddd") //INST(FMUL_elt_1, "FMUL (by element)", "0101111100LMmmmm1001H0nnnnnddddd")
//INST(FMUL_elt_2, "FMUL (by element)", "010111111zLMmmmm1001H0nnnnnddddd") INST(FMUL_elt_2, "FMUL (by element)", "010111111zLMmmmm1001H0nnnnnddddd")
//INST(SQRDMLAH_elt_1, "SQRDMLAH (by element)", "01111111zzLMmmmm1101H0nnnnnddddd") //INST(SQRDMLAH_elt_1, "SQRDMLAH (by element)", "01111111zzLMmmmm1101H0nnnnnddddd")
//INST(SQRDMLSH_elt_1, "SQRDMLSH (by element)", "01111111zzLMmmmm1111H0nnnnnddddd") //INST(SQRDMLSH_elt_1, "SQRDMLSH (by element)", "01111111zzLMmmmm1111H0nnnnnddddd")
//INST(FMULX_elt_1, "FMULX (by element)", "0111111100LMmmmm1001H0nnnnnddddd") //INST(FMULX_elt_1, "FMULX (by element)", "0111111100LMmmmm1001H0nnnnnddddd")

View file

@ -666,8 +666,8 @@ struct TranslatorVisitor final {
bool FMLA_elt_2(bool sz, bool L, bool M, Vec Vm, bool H, Vec Vn, Vec Vd); bool FMLA_elt_2(bool sz, bool L, bool M, Vec Vm, bool H, Vec Vn, Vec Vd);
bool FMLS_elt_1(bool L, bool M, Vec Vm, bool H, Vec Vn, Vec Vd); bool FMLS_elt_1(bool L, bool M, Vec Vm, bool H, Vec Vn, Vec Vd);
bool FMLS_elt_2(bool sz, bool L, bool M, Vec Vm, bool H, Vec Vn, Vec Vd); bool FMLS_elt_2(bool sz, bool L, bool M, Vec Vm, bool H, Vec Vn, Vec Vd);
bool FMUL_elt_1(bool L, bool M, Vec Vm, bool H, Vec Vn, Vec Vd); bool FMUL_elt_1(Imm<1> L, Imm<1> M, Imm<4> Vmlo, Imm<1> H, Vec Vn, Vec Vd);
bool FMUL_elt_2(bool sz, bool L, bool M, Vec Vm, bool H, Vec Vn, Vec Vd); bool FMUL_elt_2(bool sz, Imm<1> L, Imm<1> M, Imm<4> Vmlo, Imm<1> H, Vec Vn, Vec Vd);
bool SQRDMLAH_elt_1(Imm<2> size, bool L, bool M, Vec Vm, bool H, Vec Vn, Vec Vd); bool SQRDMLAH_elt_1(Imm<2> size, bool L, bool M, Vec Vm, bool H, Vec Vn, Vec Vd);
bool SQRDMLSH_elt_1(Imm<2> size, bool L, bool M, Vec Vm, bool H, Vec Vn, Vec Vd); bool SQRDMLSH_elt_1(Imm<2> size, bool L, bool M, Vec Vm, bool H, Vec Vn, Vec Vd);
bool FMULX_elt_1(bool L, bool M, Vec Vm, bool H, Vec Vn, Vec Vd); bool FMULX_elt_1(bool L, bool M, Vec Vm, bool H, Vec Vn, Vec Vd);

View file

@ -0,0 +1,29 @@
/* This file is part of the dynarmic project.
* Copyright (c) 2018 MerryMage
* This software may be used and distributed according to the terms of the GNU
* General Public License version 2 or any later version.
*/
#include "frontend/A64/translate/impl/impl.h"
namespace Dynarmic::A64 {
bool TranslatorVisitor::FMUL_elt_2(bool sz, Imm<1> L, Imm<1> M, Imm<4> Vmlo, Imm<1> H, Vec Vn, Vec Vd) {
if (sz && L == 1) {
return UnallocatedEncoding();
}
const size_t idxdsize = H == 1 ? 128 : 64;
const size_t index = sz ? H.ZeroExtend() : concatenate(H, L).ZeroExtend();
const Vec Vm = concatenate(M, Vmlo).ZeroExtend<Vec>();
const size_t esize = sz ? 64 : 32;
const IR::U32U64 operand = V_scalar(esize, Vn);
const IR::U32U64 element = ir.VectorGetElement(esize, V(idxdsize, Vm), index);
const IR::U32U64 result = ir.FPMul(operand, element, true);
V_scalar(esize, Vd, result);
return true;
}
} // namespace Dynarmic::A64