Uh oh!
There was an error while loading. Please reload this page.
[JIT] On importing a CEE_INITOBJ for a simd, construct a VecCon(0) - #81982
Conversation
ghost
commented
Feb 10, 2023
Tagging subscribers to this area: @JulieLeeMSFT, @jakobbotsch, @kunalspathak Issue DetailsWhen handle CEE_INITOBJ in the importer, if we know that the type is a simd type, then constructor a
|
TIHan
commented
Feb 14, 2023
@dotnet/jit-contrib this is ready. cc @tannergooding@kunalspathak A few improvements |
SingleAccretion
left a comment
There was a problem hiding this comment.
I think it is great to make the zero-init form for vectors canonical with CNS_VEC, but we should make this change everywhere and fully disallow ASG(simd, CNS_INT 0). This will allow for some code simplification (e. g. in VN, or the recent local morph fix).
Uh oh!
There was an error while loading. Please reload this page.
TIHan
commented
Feb 14, 2023
I agree, we will make another PR to handle it. @tannergooding knows of the several places to do this so I'll let him handle it. |
Uh oh!
There was an error while loading. Please reload this page.
Co-authored-by: SingleAccretion <62474226+SingleAccretion@users.noreply.github.com>
tannergooding
left a comment
There was a problem hiding this comment.
LGTM. I think we should have an issue tracking the other cleanup so we can get to the desired outcome of no more ASG(SIMD, 0)
TIHan
commented
Feb 22, 2023
@dotnet/jit-contrib this is ready PTAL @jakobbotsch |
When we handle
CEE_INITOBJin the importer, if we know that the type is a simd type, then construct aVecCon(0)for the zero-initialization instead of anIntCon(0).