@@ -42,57 +42,59 @@ export TF_NEED_ROCM=1
4242export ROCM_PATH=" /opt/rocm"
4343
4444EXCLUDED_TESTS=(
45- # //xla/service/gpu/tests:gpu_kernel_tiling_test_gpu_amd_any
46- GpuKernelTilingTest.ColumnReductionWithLayoutChangeTiled
47- GpuKernelTilingTest.ReductionInputTooLarge
48- # //xla/pjrt/c:pjrt_c_api_gpu_test_gpu_amd_any
49- PjrtCAPIGpuExtensionTest.TritonCompile
50- # //xla/backends/gpu/codegen/triton:fusion_emitter_device_test_gpu_amd_any
51- TritonEmitterTest.CheckRocmWarpSize
52- TritonEmitterTest.ConvertF16ToF8E5M2Exhaustive
53- TritonEmitterTest.FP8ToFP8EndToEnd
54- TritonEmitterTest.FusionWithOutputContainingMoreThanInt32MaxElementsExecutesCorrectly
55- BasicDotAlgorithmEmitterTestSuite/BasicDotAlgorithmEmitterTest.BasicAlgorithmIsEmittedCorrectly/ALG_DOT_F64_F64_F64
56- # //xla/backends/gpu/codegen/triton:fusion_emitter_device_legacy_test_gpu_amd_any
57- TritonGemmTest.BroadcastOfVectorConstantIsFused
58- TritonGemmTest.FailIfTooMuchShmem
59- TritonGemmTest.SplitAndTransposeLhsExecutesCorrectly
60- # //xla/backends/gpu/codegen/triton:fusion_emitter_int4_device_test_gpu_amd_any
61- TritonTest.NonstandardLayoutWithManyNonContractingDims
62- TritonTest.NonstandardLayoutWithManyNonContractingDimsReversedLayout
63- # //xla/hlo/builder/lib:self_adjoint_eig_test_gpu_amd_any marked as flaky but randomly red after 3 attempts
64- RandomEighTestInstantiation/RandomEighTest.Random/*
65- # temp excludes for 0.7.1
66- CompareTest.SplitK
67- TritonEmitterTest.RocmWarpSizeIsSetCorrectly
68- MultiOutputFusionTest.MultiOutputReduceFusionMajorWithExtraOutput
69- TestRadixSort/CubSortKeysTest.SortKeys/*
70- GpuIrEmitterUnnestedTest.CanNotEmitTritonCustomCallOnPreAmpereGpu
71- CommandBufferConversionPassTest.ConvertWhileThunkWithAsyncPair
72- CommandBufferConversionPassTest.ConvertWhileThunk
73- TritonFusionNumericsVerifierTest.CompilationSucceedsEvenIfKernelWillSpillRegisters
74- TritonFusionNumericsVerifierTest.VerifyThatDisablingTritonIsFast
75- TritonFusionNumericsVerifierTestSuite/TritonFusionNumericsVerifierTest.VerifyNestedGemmNumerics/1
76- TritonGemmTest.FailForTooComplexTiling
77- TritonAndBlasSupportForDifferentTensorSizes/TritonAndBlasSupportForDifferentTensorSizes.IsDotAlgorithmSupportedByTriton/dot_tf32_tf32_f32
78- TritonAndBlasSupportForDifferentTensorSizes/TritonAndBlasSupportForDifferentTensorSizes.IsDotAlgorithmSupportedByTriton/dot_f32_f32_f32
79- TritonAndBlasSupportForDifferentTensorSizes/TritonAndBlasSupportForDifferentTensorSizes.IsDotAlgorithmSupportedByTriton/dot_tf32_tf32_f32_x3
80- TestRadixSort/CubSortPairsTest.SortPairs/*
81- GpuKernelTilingTest.ReductionInputTooLarge
82- DeterminismTest.Conv
83- TopKTests/TopKKernelTest*
84- DotTestTestSuite/DotTest.IsTritonSupportedExecutesCorrectlyForDot/f8e5m2_dot
85- DotTestTestSuite/DotTest.IsTritonSupportedExecutesCorrectlyForDot/f32_dot
86- TritonNormalizationTest.CanFuseAndEmitDiamondWithBF16Converts
87- ElementwiseTestSuiteF16/UnaryElementwiseTest.ElementwiseUnaryOpExecutesCorrectly/f16_cosine
88- ElementwiseTestSuiteF16/BinaryElementwiseTest.ElementwiseBinaryOpExecutesCorrectly/f16_atan2
89- ElementwiseTestSuiteF16/BinaryElementwiseTest.ElementwiseFusionExecutesCorrectly/f16_atan2
90- TritonTest.FuseSubchannelDequantizationWithTranspose
91- BasicDotAlgorithmEmitterTestSuite/BasicDotAlgorithmEmitterTest.BasicAlgorithmIsEmittedCorrectly/ALG_DOT_F16_F16_F16
92- CommandBufferTests/CommandBufferTest.IndexConditional/*
93- CommandBufferTests/CommandBufferTest.WhileLoop/*
94- CommandBufferTests/CommandBufferTest.TrueFalseConditional/*
95- BufferComparatorTest.VeryLargeArray_Device_U8_Aligned
45+ # //xla/service/gpu/tests:gpu_kernel_tiling_test_gpu_amd_any
46+ GpuKernelTilingTest.ColumnReductionWithLayoutChangeTiled
47+ GpuKernelTilingTest.ReductionInputTooLarge
48+ # //xla/pjrt/c:pjrt_c_api_gpu_test_gpu_amd_any
49+ PjrtCAPIGpuExtensionTest.TritonCompile
50+ # //xla/backends/gpu/codegen/triton:fusion_emitter_device_test_gpu_amd_any
51+ TritonEmitterTest.CheckRocmWarpSize
52+ TritonEmitterTest.ConvertF16ToF8E5M2Exhaustive
53+ TritonEmitterTest.FP8ToFP8EndToEnd
54+ TritonEmitterTest.FusionWithOutputContainingMoreThanInt32MaxElementsExecutesCorrectly
55+ BasicDotAlgorithmEmitterTestSuite/BasicDotAlgorithmEmitterTest.BasicAlgorithmIsEmittedCorrectly/ALG_DOT_F64_F64_F64
56+ # //xla/backends/gpu/codegen/triton:fusion_emitter_device_legacy_test_gpu_amd_any
57+ TritonGemmTest.BroadcastOfVectorConstantIsFused
58+ TritonGemmTest.FailIfTooMuchShmem
59+ TritonGemmTest.SplitAndTransposeLhsExecutesCorrectly
60+ # //xla/backends/gpu/codegen/triton:fusion_emitter_int4_device_test_gpu_amd_any
61+ TritonTest.NonstandardLayoutWithManyNonContractingDims
62+ TritonTest.NonstandardLayoutWithManyNonContractingDimsReversedLayout
63+ # //xla/hlo/builder/lib:self_adjoint_eig_test_gpu_amd_any marked as flaky but randomly red after 3 attempts
64+ RandomEighTestInstantiation/RandomEighTest.Random/*
65+ # temp excludes for 0.7.1
66+ CompareTest.SplitK
67+ TritonEmitterTest.RocmWarpSizeIsSetCorrectly
68+ MultiOutputFusionTest.MultiOutputReduceFusionMajorWithExtraOutput
69+ TestRadixSort/CubSortKeysTest.SortKeys/*
70+ GpuIrEmitterUnnestedTest.CanNotEmitTritonCustomCallOnPreAmpereGpu
71+ CommandBufferConversionPassTest.ConvertWhileThunkWithAsyncPair
72+ CommandBufferConversionPassTest.ConvertWhileThunk
73+ TritonFusionNumericsVerifierTest.CompilationSucceedsEvenIfKernelWillSpillRegisters
74+ TritonFusionNumericsVerifierTest.VerifyThatDisablingTritonIsFast
75+ TritonFusionNumericsVerifierTestSuite/TritonFusionNumericsVerifierTest.VerifyNestedGemmNumerics/1
76+ TritonGemmTest.FailForTooComplexTiling
77+ TritonAndBlasSupportForDifferentTensorSizes/TritonAndBlasSupportForDifferentTensorSizes.IsDotAlgorithmSupportedByTriton/dot_tf32_tf32_f32
78+ TritonAndBlasSupportForDifferentTensorSizes/TritonAndBlasSupportForDifferentTensorSizes.IsDotAlgorithmSupportedByTriton/dot_f32_f32_f32
79+ TritonAndBlasSupportForDifferentTensorSizes/TritonAndBlasSupportForDifferentTensorSizes.IsDotAlgorithmSupportedByTriton/dot_tf32_tf32_f32_x3
80+ TestRadixSort/CubSortPairsTest.SortPairs/*
81+ GpuKernelTilingTest.ReductionInputTooLarge
82+ DeterminismTest.Conv
83+ TopKTests/TopKKernelTest*
84+ DotTestTestSuite/DotTest.IsTritonSupportedExecutesCorrectlyForDot/f8e5m2_dot
85+ DotTestTestSuite/DotTest.IsTritonSupportedExecutesCorrectlyForDot/f32_dot
86+ TritonNormalizationTest.CanFuseAndEmitDiamondWithBF16Converts
87+ ElementwiseTestSuiteF16/UnaryElementwiseTest.ElementwiseUnaryOpExecutesCorrectly/f16_cosine
88+ ElementwiseTestSuiteF16/BinaryElementwiseTest.ElementwiseBinaryOpExecutesCorrectly/f16_atan2
89+ ElementwiseTestSuiteF16/BinaryElementwiseTest.ElementwiseFusionExecutesCorrectly/f16_atan2
90+ TritonTest.FuseSubchannelDequantizationWithTranspose
91+ BasicDotAlgorithmEmitterTestSuite/BasicDotAlgorithmEmitterTest.BasicAlgorithmIsEmittedCorrectly/ALG_DOT_F16_F16_F16
92+ CommandBufferTests/CommandBufferTest.IndexConditional/*
93+ CommandBufferTests/CommandBufferTest.WhileLoop/*
94+ CommandBufferTests/CommandBufferTest.TrueFalseConditional/*
95+ BufferComparatorTest.VeryLargeArray_Device_U8_Aligned
96+ # //xla/tests:collective_ops_e2e_test_amdgpu_any
97+ CollectiveOpsTestE2EShardedUnsharded/*
9698)
9799
98100BAZEL_DISK_CACHE_SIZE=100G
@@ -107,13 +109,10 @@ TAG_FILTERS=$($SCRIPT_DIR/rocm_tag_filters.sh),-multigpu,-multi_gpu_h100,require
107109
108110SANITIZER_ARGS=()
109111if [[ $1 == " asan" ]]; then
110- SANITIZER_ARGS+=(" --test_env=ASAN_OPTIONS=suppressions=build_tools/rocm/asan_ignore_list.txt:use_sigaltstack=0" )
111- SANITIZER_ARGS+=(" --test_env=LSAN_OPTIONS=suppressions=build_tools/rocm/lsan_ignore_list.txt:use_sigaltstack=0" )
112112 SANITIZER_ARGS+=(" --config=asan" )
113113 TAG_FILTERS=$TAG_FILTERS ,-noasan
114114 shift
115115elif [[ $1 == " tsan" ]]; then
116- SANITIZER_ARGS+=(" --test_env=TSAN_OPTIONS=suppressions=build_tools/rocm/tsan_ignore_list.txt::history_size=7:ignore_noninstrumented_modules=1" )
117116 SANITIZER_ARGS+=(" --config=tsan" )
118117 TAG_FILTERS=$TAG_FILTERS ,-notsan
119118 shift
0 commit comments