Skip to content

XPU: Enable new grf_mode settings#1016

Merged
Tcc0403 merged 4 commits intolinkedin:mainfrom
Egor-Krivov:egor/grf_mode
Jan 15, 2026
Merged

XPU: Enable new grf_mode settings#1016
Tcc0403 merged 4 commits intolinkedin:mainfrom
Egor-Krivov:egor/grf_mode

Conversation

@Egor-Krivov
Copy link
Contributor

@Egor-Krivov Egor-Krivov commented Jan 14, 2026

Summary

After this change in triton-xpu grf_mode api changes. This change will be active in triton-xpu>=3.6

Testing Done

I ran Liger-Kernels on new and old triton version on PVC (GPU max 1100). I run tests and relevant benchmarks.

  • Hardware Type: XPU, Intel PVC (GPU max 1100)
  • run make test to ensure correctness
  • run make checkstyle to ensure code style
  • run make test-convergence to ensure convergence

I will run all tests and update the PR description

@Egor-Krivov
Copy link
Contributor Author

When I run tests, some of them fail:

================================================================================= short test summary info =================================================================================
FAILED test/transformers/test_fused_add_rms_norm.py::test_correctness[True-BaseAddRMSNorm-0.0-none-dtype1-0.2-0.02-2-128-512] - RuntimeError: PassManager::run failed
FAILED test/transformers/test_fused_add_rms_norm.py::test_correctness[True-BaseAddRMSNorm-0.0-none-dtype1-0.2-0.02-5-123-123] - RuntimeError: PassManager::run failed
FAILED test/transformers/test_fused_add_rms_norm.py::test_correctness[False-BaseAddRMSNorm-0.0-none-dtype1-0.2-0.02-2-128-512] - RuntimeError: PassManager::run failed
FAILED test/transformers/test_fused_add_rms_norm.py::test_correctness[False-BaseAddRMSNorm-0.0-none-dtype1-0.2-0.02-5-123-123] - RuntimeError: PassManager::run failed
FAILED test/transformers/test_geglu.py::test_correctness[dtype1-0.01-0.01-2-2048-2048-4096] - AssertionError: assert False is True
FAILED test/transformers/test_geglu.py::test_correctness[dtype1-0.01-0.01-9-41-341-4231] - AssertionError: assert False is True
FAILED test/transformers/test_rms_norm.py::test_correctness[True-True-BaseRMSNorm-0.0-none-dtype1-0.2-0.02-2-128-512] - RuntimeError: PassManager::run failed
FAILED test/transformers/test_rms_norm.py::test_correctness[True-True-BaseRMSNorm-0.0-none-dtype1-0.2-0.02-5-123-123] - RuntimeError: PassManager::run failed
FAILED test/transformers/test_rms_norm.py::test_correctness[True-False-BaseRMSNorm-0.0-none-dtype1-0.2-0.02-2-128-512] - RuntimeError: PassManager::run failed
FAILED test/transformers/test_rms_norm.py::test_correctness[True-False-BaseRMSNorm-0.0-none-dtype1-0.2-0.02-5-123-123] - RuntimeError: PassManager::run failed
FAILED test/transformers/test_rms_norm.py::test_correctness[False-True-BaseRMSNorm-0.0-none-dtype1-0.2-0.02-2-128-512] - RuntimeError: PassManager::run failed
FAILED test/transformers/test_rms_norm.py::test_correctness[False-True-BaseRMSNorm-0.0-none-dtype1-0.2-0.02-5-123-123] - RuntimeError: PassManager::run failed
FAILED test/transformers/test_rms_norm.py::test_correctness[False-False-BaseRMSNorm-0.0-none-dtype1-0.2-0.02-2-128-512] - RuntimeError: PassManager::run failed
FAILED test/transformers/test_rms_norm.py::test_correctness[False-False-BaseRMSNorm-0.0-none-dtype1-0.2-0.02-5-123-123] - RuntimeError: PassManager::run failed
========================================== 14 failed, 2360 passed, 269 skipped, 8 xfailed, 2 xpassed, 78 warnings, 37 rerun in 831.33s (0:13:51) ====

I don't think it is caused by this PR though

@Egor-Krivov
Copy link
Contributor Author

@Tcc0403 Ready to merge

Copy link
Collaborator

@Tcc0403 Tcc0403 left a comment

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

lgtm, thanks!

@Tcc0403 Tcc0403 merged commit 71ed8ac into linkedin:main Jan 15, 2026
3 of 7 checks passed
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

2 participants