[llvm] [IPO] Added attributor for identifying invariant loads (PR #141800)

via llvm-commits llvm-commits at lists.llvm.org
Tue Jun 10 06:56:06 PDT 2025


zGoldthorpe wrote:

> I think this is good from what I can tell.
> 
> Although one ask - either here in or in a followup PR - can we make sure that this still works with `ptr addrspace(7)`s created by `make.buffer.rsrc.p7.p1`? (Or that stuff like `invariant.start` doesn't break the analysis?)

I feel like this extension makes more sense in a follow-up PR, since it would probably be nicer to have a more uniform solution to address such intrinsics.

https://github.com/llvm/llvm-project/pull/141800


More information about the llvm-commits mailing list