Some minor cleanup post the addition of TYP_SIMD64 and ZMM support - P1 #83044
Conversation
dotnet-issue-labeler
bot
added
the
area-CodeGen-coreclr
ghost
assigned 
|
Tagging subscribers to this area: @JulieLeeMSFT, @jakobbotsch, @kunalspathak Issue DetailsThe initial support was added in #80960 and a few unrelated cleanup opportunities were seen. This is the first cleanup and covers the simpler things which can be easily relegated to TARGET_XARCH only. There will be a separate PR cleaning up the SIMD handle logic to reduce code duplication and then another which ensures that
|
tannergooding
force-pushed
the
post-zmm-cleanup
branch
from
466a63d to
fd06fc5
Compare
|
Minor TP impact is because we need to call This is called for each We only need to report that info "once" per compilation, so we should probably look at refactoring the |
| @@ -10709,7 +10709,6 @@ void CodeGen::genZeroInitFrameUsingBlockInit(int untrLclHi, int untrLclLo, regNu | |||
| assert(compiler->compGeneratingProlog); | |||
| assert(genUseBlockInit); | |||
| assert(untrLclHi > untrLclLo); | |||
| assert(compiler->getSIMDSupportLevel() >= SIMD_SSE2_Supported); | |||
There was a problem hiding this comment.
Replacing this assert is not needed? Is the idea that if AVX2 is not supported, then the default is SSE2 now?
There was a problem hiding this comment.
The default has been SSE2 for many years (always for x64 and since around .NET Core 2.1 for x86). We require it and consider it part of the baseline ISA, so there is no need to check for its existence outside of supporting the DOTNET_EnableSSE2=0 switch for HWIntrinsic importation.
The initial support was added in #80960 and a few unrelated cleanup opportunities were seen.
This is the first cleanup and covers the simpler things which can be easily relegated to TARGET_XARCH only.
There will be a separate PR cleaning up the SIMD handle logic to reduce code duplication and then another which ensures that
simd32_tandsimd64_tare xarch only as well.