Integrate LLVM at 3f0cc068cef26e820b3acbd21b3577817e4bf4ca

[nithinsubbiah]

Carries two revert for 1:N matchAndRewrite overload

1. [mlir][Transforms] Add 1:N matchAndRewrite overload llvm/llvm-project#116470
2. [mlir][Func] Delete DecomposeCallGraphTypes.cpp llvm/llvm-project#117424

[bjacob]

CI failure - I just triggered a rerun to see if it's just transient.

	test_sdxl_rocm_benchmark failed on attempt 1! Retrying!
	Traceback (most recent call last):
	  File "/home/esaimana/actions-runner-2/_work/iree/iree/venv/lib/python3.11/site-packages/pytest_check/plugin.py", line 41, in pytest_runtest_makereport
	    raise AssertionError(report.longrepr)
	AssertionError: FAILURE: check 338.0 <= 330.0: SDXL e2e benchmark time should not regress
	benchmark_sdxl_rocm.py:523 in test_sdxl_rocm_benchmark() -> check.less_equal(
	
	FAILURE: check 58.3 <= 53.0: SDXL punet f8 benchmark time should not regress
	FAILURE: check 15.8 <= 15.5: SDXL prompt encoder benchmark time should not regress

If this persists, we have to decide whether to accept the regression. @ScottTodd

[ScottTodd]

Note that the PkgCI / Regression Test / test_regression_suite :: amdgpu_rocm_mi300_gfx942 job has been unstable a bit lately. I've seen multiple similar flakes... not sure if anyone has investigated or tried to stabilize

• Just main: https://github.com/iree-org/iree/actions/workflows/pkgci.yml?query=branch%3Amain
• All failure statuses: https://github.com/iree-org/iree/actions/workflows/pkgci.yml?query=is%3Afailure

[bjacob]

Indeed, from your link, I found this recent instance on main branch with similar timeouts by similar margins:
https://github.com/iree-org/iree/actions/runs/12150291953/job/33883074393

[bjacob]

Run #2 just completed here, failed again with similar timeout error:
https://github.com/iree-org/iree/actions/runs/12185625006/job/33996379858?pr=19387

Since this is similar to what's already failing today on main, this doesn't need to block this integrate longer. Approving!