Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Jump to bottom

[TESTS] Use FileCheck from LLVM package. #1275

Closed
wants to merge 2 commits into from
Closed

[TESTS] Use FileCheck from LLVM package. #1275

wants to merge 2 commits into from

Conversation

dfukalov
Copy link
Contributor

@dfukalov dfukalov commented Mar 3, 2023

Needed to move to 8e5a41e8271f package that now contains the utility.

@dfukalov
[TESTS] Use FileCheck from LLVM package.
43cb3c9 
Needed to move to 8e5a41e8271f package that now contains the utility.
@dfukalov dfukalov requested a review from ptillet as a code owner March 3, 2023 17:59
@ptillet
Copy link
Collaborator

ptillet commented Mar 3, 2023

Seems like the latest LLVM build is giving me (and a few other people) a GLIBC error. I'll look into that

@dfukalov
Copy link
Contributor Author

dfukalov commented Mar 3, 2023 •
edited
Loading

I'm pretty sure it's related to gcc-11 upgrade. There are topics in internet about gcc-11 and "GLIBCXX_3.4.29 not found" issue.
Also, I had something similar on Ubuntu 18, but no issues on Ubuntu 22 (although I didn't check the Ubuntu 20).

Perhaps, your llvm package + environment can actually move to Ubuntu 20 or even 22?

@dfukalov
Copy link
Contributor Author

dfukalov commented Mar 3, 2023

Btw, why not use clang+lld build? For the llvm package build it's at least faster.
I use it locally to build own debug llvm package since gcc+ld gets enormous memory and time to build it.

@ptillet
Copy link
Collaborator

ptillet commented Mar 3, 2023

I'm pretty sure it's related to gcc-11 upgrade

At first sight it seems like it, but I believe the last llvm-17 binaries before that came from this workflow https://github.com/ptillet/triton-llvm-releases/actions/runs/4217355325/jobs/7322723517 that used g++-11 too. Either way I'm investigating. For now I've tried reverting the LLVM_UTILS PR you submitted, to see if it makes a difference.

Btw, why not use clang+lld build? For the llvm package build it's at least faster.

No particular reason. I'd accept a PR that makes the workflow use clang instead of gcc

@ptillet
Copy link
Collaborator

ptillet commented Mar 6, 2023

Re-implemented with minor fixes in #1282 so I wouldn't bother you on a weekend

@ptillet ptillet closed this Mar 6, 2023
ZzEeKkAa pushed a commit to ZzEeKkAa/triton that referenced this pull request Aug 5, 2024
@alexbaden
[TMP]: Disable layout rematerialization cache (triton-lang#1275)
eb51a81 
Temporarily disables the layout rematerialization cache for backward
slice rematerialization. The remat cache is exposing a bug on systems
with the LTS driver which results in an accuracy error with two
HuggingFace models under the PyTorch Inductor benchmarks. 

Issue triton-lang#1255
ZzEeKkAa pushed a commit to ZzEeKkAa/triton that referenced this pull request Aug 5, 2024
@whitneywhtsang
Revert "[TMP]: Disable layout rematerialization cache (triton-lang#1275…
c45830b 
…)"

This reverts commit eb51a81.
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Reviewers

@ptillet ptillet Awaiting requested review from ptillet ptillet is a code owner

Assignees
No one assigned
Labels
None yet
Projects
None yet
Milestone
No milestone
Development

Successfully merging this pull request may close these issues.
2 participants
@dfukalov @ptillet