Closed jkotas closed 2 months ago
@VSadov Are you interested in looking into this?
@VSadov Are you interested in looking into this?
yes, NativeAOT approach is a lot simpler. I’d like try porting it.
assume the indirection isn't arm64 specific? But in theory it might help with cache locality?
assume the indirection isn't arm64 specific? But in theory it might help with cache locality?
Right
Context: https://github.com/dotnet/runtime/issues/99552#issuecomment-1993382384
The allocation helpers in CoreCLR have an unnecessary extra indirection. The alloc_context can be thread static directly. The extra indirection via Thread is unnecessary. The allocation helpers in native AOT do not have this indirection. It would be nice to port the improvement to regular CoreCLR and unify how the allocation helpers work.