Merge pull request #2030 from lioncash/interp-mov

Interpreter: Handle 256-bit VMov
This commit is contained in:
Ryan Houdek authored and GitHub committed 2022-09-28 13:05:16 -07:00
commit 412793c21d
1 file changed
+4 -1
@@ -45,7 +45,10 @@ DEF_OP(VMov) {
auto Op = IROp->C<IR::IROp_VMov>();
const uint8_t OpSize = IROp->Size;
const auto Src = *GetSrc<__uint128_t*>(Data->SSAData, Op->Source);
LOGMAN_THROW_AA_FMT(OpSize <= Core::CPUState::XMM_AVX_REG_SIZE,
"Moves larger than 256-bit aren't supported");
const auto Src = *GetSrc<InterpVector256*>(Data->SSAData, Op->Source);
memcpy(GDP, &Src, OpSize);
}