This is an archive of the discontinued LLVM Phabricator instance.

I noticed buildbot has been restarted, there's a bunch of tests failing at the moment, the HWASAN are unrelated to this, but this patch has exposed issues in the following tests that are failing:

MLIR :: Integration/Dialect/SparseTensor/CPU/concatenate_dim_1.mlir
MLIR :: Integration/Dialect/SparseTensor/CPU/sparse_sorted_coo.mlir
MLIR :: Integration/Dialect/SparseTensor/CPU/sparse_spmm.mlir

The problem common to all of them is they don't use emulation since they're not using substitutions. The top one is now using mlir-cpu-runner since a recent change in D146917, but %lli should have been replaced with an equivalent substitution to support emulation. @awarzynski is looking into this. The bottom two are more straightforward, they simply weren't using a substitution to begin with. I landed 707b6e94b88a that should hopefully fix them.

In D148232#4312482, @c-rhodes wrote:
I noticed buildbot has been restarted, there's a bunch of tests failing at the moment, the HWASAN are unrelated to this, but this patch has exposed issues in the following tests that are failing:
MLIR :: Integration/Dialect/SparseTensor/CPU/concatenate_dim_1.mlir
MLIR :: Integration/Dialect/SparseTensor/CPU/sparse_sorted_coo.mlir
MLIR :: Integration/Dialect/SparseTensor/CPU/sparse_spmm.mlir
The problem common to all of them is they don't use emulation since they're not using substitutions. The top one is now using mlir-cpu-runner since a recent change in D146917, but %lli should have been replaced with an equivalent substitution to support emulation. @awarzynski is looking into this. The bottom two are more straightforward, they simply weren't using a substitution to begin with. I landed 707b6e94b88a that should hopefully fix them.

Thanks @c-rhodes ! I've updated concatenate_dim_1.mlir here: https://github.com/llvm/llvm-project/commit/20bf8c403c5a8db6107358e534fb3ab5dd0ae9a3. https://reviews.llvm.org/D148005 will bring a proper fix (just need to resurrect it first)

Revision Contents

Path

Size

buildbot/

osuosl/

master/

config/

builders.py

9 lines

Diff 516707

buildbot/osuosl/master/config/builders.py

Show First 20 Lines • Show All 444 Lines • ▼ Show 20 Lines	'factory' : ClangBuilder.getClangCMakeBuildFactory(
extra_cmake_args=[		extra_cmake_args=[
"-DCMAKE_C_FLAGS='-mcpu=cortex-a57'",		"-DCMAKE_C_FLAGS='-mcpu=cortex-a57'",
"-DCMAKE_CXX_FLAGS='-mcpu=cortex-a57'",		"-DCMAKE_CXX_FLAGS='-mcpu=cortex-a57'",
"-DCOMPILER_RT_BUILD_SANITIZERS=OFF",		"-DCOMPILER_RT_BUILD_SANITIZERS=OFF",
"-DLLVM_ENABLE_LLD=True",		"-DLLVM_ENABLE_LLD=True",
# lld tests cause us to hit thread limits		# lld tests cause us to hit thread limits
"-DLLVM_ENABLE_THREADS=OFF"])},		"-DLLVM_ENABLE_THREADS=OFF"])},

# AArch64 check-all + flang + compiler-rt + test-suite 2-stage		# AArch64 check-all + flang + compiler-rt + test-suite + SVE/SME
		# mlir-integration-tests 2-stage
{'name' : "clang-aarch64-full-2stage",		{'name' : "clang-aarch64-full-2stage",
'tags' : ["clang"],		'tags' : ["clang"],
'workernames' : ["linaro-clang-aarch64-full-2stage"],		'workernames' : ["linaro-clang-aarch64-full-2stage"],
'builddir': "clang-aarch64-full-2stage",		'builddir': "clang-aarch64-full-2stage",
'factory' : ClangBuilder.getClangCMakeBuildFactory(		'factory' : ClangBuilder.getClangCMakeBuildFactory(
clean=False,		clean=False,
checkout_flang=True,		checkout_flang=True,
checkout_lld=False,		checkout_lld=False,
enable_runtimes="auto",		enable_runtimes="auto",
useTwoStage=True,		useTwoStage=True,
testStage1=False,		testStage1=False,
runTestSuite=True,		runTestSuite=True,
env={		env={
'NO_STOP_MESSAGE':'1', # For Fortran test-suite		'NO_STOP_MESSAGE':'1', # For Fortran test-suite
},		},
testsuite_flags=[		testsuite_flags=[
'--cppflags', '-mcpu=cortex-a57',		'--cppflags', '-mcpu=cortex-a57',
'--threads=32', '--build-threads=32'],		'--threads=32', '--build-threads=32'],
extra_cmake_args=[		extra_cmake_args=[
"-DCMAKE_C_FLAGS='-mcpu=cortex-a57'",		"-DCMAKE_C_FLAGS='-mcpu=cortex-a57'",
"-DCMAKE_CXX_FLAGS='-mcpu=cortex-a57'",		"-DCMAKE_CXX_FLAGS='-mcpu=cortex-a57'",
"-DLLVM_LIT_ARGS='-v'"])},		"-DLLVM_LIT_ARGS='-v'",
		"-DMLIR_INCLUDE_INTEGRATION_TESTS=True",
		"-DMLIR_RUN_ARM_SVE_TESTS=True",
		"-DMLIR_RUN_ARM_SME_TESTS=True",
		"-DARM_EMULATOR_EXECUTABLE=qemu-aarch64"])},


# AArch64 Clang+LLVM+RT+LLD check-all + flang + test-suite +		# AArch64 Clang+LLVM+RT+LLD check-all + flang + test-suite +
# mlir-integration-tests w/SVE-Vector-Length-Agnostic Note that in this and		# mlir-integration-tests w/SVE-Vector-Length-Agnostic Note that in this and
# other clang-aarch64-sve-* builders we set -mllvm		# other clang-aarch64-sve-* builders we set -mllvm
# -treat-scalable-fixed-error-as-warning=false to make compiler fail on		# -treat-scalable-fixed-error-as-warning=false to make compiler fail on
# non-critical SVE codegen issues. This helps us notice and fix SVE		# non-critical SVE codegen issues. This helps us notice and fix SVE
# problems sooner rather than later.		# problems sooner rather than later.
▲ Show 20 Lines • Show All 2,283 Lines • Show Last 20 Lines

This is an archive of the discontinued LLVM Phabricator instance.

[mlir][aarch64] Enable MLIR integration tests for SVE/SME under emulationClosedPublic

Details

Diff Detail

Event Timeline

Revision Contents

Diff 516707

buildbot/osuosl/master/config/builders.py

[mlir][aarch64] Enable MLIR integration tests for SVE/SME under emulation
ClosedPublic