You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
Need to verify performance, as the manual inliner has been disabled and the entire optimization pipeline is brand new (no issue yet)
I have seen performance regressions in CUDA applications on LLVM 16 vs 18, need to investigate and file a formal issue
It's frankly not obvious if these are worth investigating. The transition to opaque pointers shook looks a bunch of problems in LLVM that were not investigated prior to the release. We may just find that some of these go away with LLVM 18 or subsequent versions.
The text was updated successfully, but these errors were encountered:
Since we've got a couple of them, listing all the (new) known issues in LLVM 17:
It's frankly not obvious if these are worth investigating. The transition to opaque pointers shook looks a bunch of problems in LLVM that were not investigated prior to the release. We may just find that some of these go away with LLVM 18 or subsequent versions.
The text was updated successfully, but these errors were encountered: