mirrors/dynarmic
1
0
Fork 0
Code Issues Pull requests Projects Releases Packages Wiki Activity

A32: Implement ASIMD VST{1,2,3,4} (single n-element structure from one lane)

Browse source
This commit is contained in:
MerryMage 2020-06-21 16:27:33 +01:00
parent fa145ae401
commit 8f0f1cfd66
3 changed files with 64 additions and 2 deletions
Show stats Download patch file Download diff file Expand all files Collapse all files

2
src/frontend/A32/decoder/asimd.inc
View file

@ -127,5 +127,5 @@ INST(arm_UDF, "UNALLOCATED", "111101000--0--------11-
INST(arm_UDF, "UNALLOCATED", "111101001-00--------11----------") // v8
INST(v8_VLD_all_lanes, "VLD{1-4} (all lanes)", "111101001D10nnnndddd11nnzzTammmm") // v8
INST(arm_UDF, "UNALLOCATED", "111101001-10--------1110---1----") // v8
//INST(v8_VST_single, "VST{1-4} (single)", "111101001D00nnnnddddzzNNaaaammmm") // v8
INST(v8_VST_single, "VST{1-4} (single)", "111101001D00nnnnddddzzNNaaaammmm") // v8
INST(v8_VLD_single, "VLD{1-4} (single)", "111101001D10nnnnddddzzNNaaaammmm") // v8

63
src/frontend/A32/translate/impl/asimd_load_store_structures.cpp
View file

@ -235,6 +235,68 @@ bool ArmTranslatorVisitor::v8_VLD_all_lanes(bool D, Reg n, size_t Vd, size_t nn,
return true;
}
bool ArmTranslatorVisitor::v8_VST_single(bool D, Reg n, size_t Vd, size_t sz, size_t nn, size_t index_align, Reg m) {
const size_t nelem = nn + 1;
ASSERT_MSG(sz != 0b11, "Decode Error");
if (nelem == 1 && Common::Bit(sz, index_align)) {
return UndefinedInstruction();
}
const size_t ebytes = 1 << sz;
const size_t index = Common::Bits(sz + 1, 3, index_align);
const size_t inc = (sz != 0 && Common::Bit(sz, index_align)) ? 2 : 1;
const size_t a = Common::Bits(0, sz ? sz - 1 : 0, index_align);
if (nelem == 1 && inc == 2) {
return UndefinedInstruction();
}
if (nelem == 1 && (a != 0b00 && a != 0b11)) {
return UndefinedInstruction();
}
if (nelem == 2 && Common::Bit<1>(a)) {
return UndefinedInstruction();
}
if (nelem == 3 && a != 0b00) {
return UndefinedInstruction();
}
if (nelem == 4 && a == 0b11) {
return UndefinedInstruction();
}
// TODO: alignment
const ExtReg d = ToExtRegD(Vd, D);
const size_t d_last = RegNumber(d) + inc * (nelem - 1);
if (n == Reg::R15 || d_last + 1 > 32) {
return UnpredictableInstruction();
}
const bool wback = m != Reg::R15;
const bool register_index = m != Reg::R15 && m != Reg::R13;
auto address = ir.GetRegister(n);
for (size_t i = 0; i < nelem; i++) {
const ExtReg ext_reg = d + i * inc;
const auto element = ir.VectorGetElement(ebytes * 8, ir.GetVector(ext_reg), index);
ir.WriteMemory(ebytes * 8, address, element);
address = ir.Add(address, ir.Imm32(static_cast<u32>(ebytes)));
}
if (wback) {
if (register_index) {
ir.SetRegister(n, ir.Add(ir.GetRegister(n), ir.GetRegister(m)));
} else {
ir.SetRegister(n, ir.Add(ir.GetRegister(n), ir.Imm32(static_cast<u32>(nelem * ebytes))));
}
}
return true;
}
bool ArmTranslatorVisitor::v8_VLD_single(bool D, Reg n, size_t Vd, size_t sz, size_t nn, size_t index_align, Reg m) {
const size_t nelem = nn + 1;
@ -297,5 +359,4 @@ bool ArmTranslatorVisitor::v8_VLD_single(bool D, Reg n, size_t Vd, size_t sz, si
return true;
}
} // namespace Dynarmic::A32

1
src/frontend/A32/translate/impl/translate_arm.h
View file

@ -545,6 +545,7 @@ struct ArmTranslatorVisitor final {
bool v8_VST_multiple(bool D, Reg n, size_t Vd, Imm<4> type, size_t sz, size_t align, Reg m);
bool v8_VLD_multiple(bool D, Reg n, size_t Vd, Imm<4> type, size_t sz, size_t align, Reg m);
bool v8_VLD_all_lanes(bool D, Reg n, size_t Vd, size_t nn, size_t sz, bool T, bool a, Reg m);
bool v8_VST_single(bool D, Reg n, size_t Vd, size_t sz, size_t nn, size_t index_align, Reg m);
bool v8_VLD_single(bool D, Reg n, size_t Vd, size_t sz, size_t nn, size_t index_align, Reg m);
};