Pablo Zimmermann

Making it draft while I'm still investigating. Feel free to give me input anyways.

I made a few strict unit tests in ScopeTreeTimerData to see if we had an issue with the method, but tests are still passing, so I will set this PR...

In newer versions, you can use scale the Capture Window using (Ctrl + and Ctrl -). @bartlomiejwolk @PrototypeNM1

> Do we need to update LLVM hash in the same PR? It's actually not needed since not updating the llvm-hash will mean an old (in the cache) LLVM built...

Updated the llvm-hash (after https://github.com/triton-lang/triton/pull/4791). Everything still green. Waiting for your input @ThomasRaoux :)

This should be fixed after https://github.com/openxla/xla/commit/763244822bf01b1e33f5352b0eea5cb13259af58. We were using the non-sorted algorithm which is very slow for fp16 (See https://github.com/openxla/xla/issues/22233#issuecomment-2884486719 if you want more information).

The build starts to fail with "the > error" after this commit: https://github.com/llvm/llvm-project/commit/ce80c80dca45c7b4636a3e143973e2c6cbdb2884. It's quite a simple commit, maybe you have a better idea why using a non-deterministic seed could...

> BTW, looking at the cmake file of llvm it seems like it will override the pre-processor and the right way to set this cmake option: > > ``` >...

Created https://github.com/triton-lang/triton/pull/4512 to set LLVM_ABI_BREAKING_CHECKS=FORCE_OFF as it was suggested here. It's the same we are doing inside Google and I think that being able to update llvm is a priority....