Skip to content

Commit 046cefa

Browse files
leofangleofang
authored
CI: retry arm64 GH200 nightly-standard row + add GB300 (#2296)
* ci/test-matrix.yml: re-enable arm64 gh200 nightly-standard row Reverts the disable in 4c70cfa now that the runner team has fixed the pool-side hang on stream-ordered memory allocator calls. * Temporarily add push trigger to ci-nightly.yml for testing Remove before merging. * CI: deselect test_get_bar_size_in_kb on gh200 runners The cufile BAR-size query returns CUDA_ERROR_NOT_SUPPORTED on Grace+Hopper (unified memory, no discrete PCIe BAR). Tracked in #2299; remove this deselect once the test skipif is fixed upstream. Uses PYTEST_ADDOPTS so no changes to run-tests are needed. * also add GB300 * cuda.core tests: build saxpy RDC fixture with -arch=all (covers sm_103 / GB300) -arch=all-major only emits X.0 cubins (sm_100, sm_90, ...), so nvJitLink returns ERROR_INVALID_INPUT when the RDC saxpy fixture is linked on GB300 (sm_103): Blackwell no longer serves an sm_10x cubin from the sm_100 image. -arch=all additionally emits sm_103 (and other minor archs) so the linker finds a usable input. The fixture is built in-job by build_test_binaries.py, so there is no committed artifact to regenerate. * CI: install curand/cublas + cupy for numba-cuda-mlir nightly; extend deselect gate to 0.4.1 test_fortran_contiguous calls cp.random, which dlopens libcurand at runtime; the mlir nightly install didn't provide it, unlike the numba-cuda branch (which installs cuda-toolkit[curand,cublas] + cupy). Mirror that here. Bump the version-gated mlir deselect from <=0.4.0 to <=0.4.1: 0.4.1 still fails the cuobjdump-invoking nvjitlink test (wheel envs lack cuobjdump) and the #135 pytest-contamination tests, so keep them deselected. cu12 rows remain red due to the separate #2320 LTO-IR regression (not in the deselect list). * nit: align columns * Revert "Temporarily add push trigger to ci-nightly.yml for testing" This reverts commit 8d51cf7. * rename variable
1 parent ffbdc7b commit 046cefa

5 files changed

Lines changed: 49 additions & 39 deletions

File tree

‎.github/workflows/test-wheel-linux.yml‎

Lines changed: 4 additions & 1 deletion
Original file line numberDiff line numberDiff line change
@@ -352,6 +352,9 @@ jobs:
352352
env:
353353
CUDA_VER: ${{ matrix.CUDA_VER }}
354354
LOCAL_CTK: ${{ matrix.LOCAL_CTK }}
355+
# #2299: BAR-size query returns CUDA_ERROR_NOT_SUPPORTED on G+H;
356+
# skip the test on gh200 runners until upstream cufile guards it.
357+
PYTEST_ADDOPTS: ${{ matrix.GPU == 'gh200' && '--deselect tests/test_cufile.py::test_get_bar_size_in_kb' || '' }}
355358
run: run-tests bindings
356359

357360
- name: Run cuda.bindings benchmarks (smoke test)
@@ -462,7 +465,7 @@ jobs:
462465
# tests get run automatically. If they still fail on the newer
463466
# version we hear about it loudly (rather than silently masking).
464467
DESELECTS=()
465-
if python -c "from packaging.version import Version; import sys; sys.exit(0 if Version('${NUMBA_CUDA_MLIR_VER}') <= Version('0.4.0') else 1)"; then
468+
if python -c "from packaging.version import Version; import sys; sys.exit(0 if Version('${NUMBA_CUDA_MLIR_VER}') <= Version('0.4.1') else 1)"; then
466469
# NVIDIA/numba-cuda-mlir#135: serial-pytest contamination of
467470
# numba_cuda_mlir.cuda.cudadrv from an xfailed test in
468471
# test_nrt_comprehensive.py contaminates any later test that

‎.github/workflows/test-wheel-windows.yml‎

Lines changed: 1 addition & 1 deletion
Original file line numberDiff line numberDiff line change
@@ -443,7 +443,7 @@ jobs:
443443
# the full union rather than trying to enumerate what happened
444444
# to fail on the most recent nightly.
445445
DESELECTS=()
446-
if python -c "from packaging.version import Version; import sys; sys.exit(0 if Version('${NUMBA_CUDA_MLIR_VER}') <= Version('0.4.0') else 1)"; then
446+
if python -c "from packaging.version import Version; import sys; sys.exit(0 if Version('${NUMBA_CUDA_MLIR_VER}') <= Version('0.4.1') else 1)"; then
447447
DESELECTS+=(
448448
--deselect 'tests/numba_cuda_tests/cudadrv/test_cuda_array_slicing.py::CudaArraySetting::test_no_sync_default_stream'
449449
--deselect 'tests/numba_cuda_tests/cudadrv/test_cuda_array_slicing.py::CudaArraySetting::test_no_sync_supplied_stream'

‎ci/test-matrix.yml‎

Lines changed: 29 additions & 30 deletions
Original file line numberDiff line numberDiff line change
@@ -93,7 +93,7 @@ linux:
9393
- { ARCH: 'arm64', PY_VER: '3.12', CUDA_VER: '13.0.2', LOCAL_CTK: '0', GPU: 'l4', GPU_COUNT: '1', DRIVER: 'latest', ENV: { MODE: 'nightly-pytorch', TORCH_VER: '2.9.1', TORCH_CUDA: 'cu130' } }
9494
# nightly-numba-cuda
9595
- { ARCH: 'amd64', PY_VER: '3.12', CUDA_VER: '12.9.1', LOCAL_CTK: '0', GPU: 'l4', GPU_COUNT: '1', DRIVER: 'latest', ENV: { MODE: 'nightly-numba-cuda' } }
96-
- { ARCH: 'amd64', PY_VER: '3.12', CUDA_VER: '13.3.0', LOCAL_CTK: '0', GPU: 'l4', GPU_COUNT: '1', DRIVER: '580.65.06', ENV: { MODE: 'nightly-numba-cuda' } }
96+
- { ARCH: 'amd64', PY_VER: '3.12', CUDA_VER: '13.3.0', LOCAL_CTK: '0', GPU: 'l4', GPU_COUNT: '1', DRIVER: '580.65.06', ENV: { MODE: 'nightly-numba-cuda' } }
9797
- { ARCH: 'arm64', PY_VER: '3.12', CUDA_VER: '12.9.1', LOCAL_CTK: '0', GPU: 'l4', GPU_COUNT: '1', DRIVER: 'latest', ENV: { MODE: 'nightly-numba-cuda' } }
9898
- { ARCH: 'arm64', PY_VER: '3.12', CUDA_VER: '13.3.0', LOCAL_CTK: '0', GPU: 'l4', GPU_COUNT: '1', DRIVER: 'latest', ENV: { MODE: 'nightly-numba-cuda' } }
9999
# nightly-numba-cuda-mlir (MLIR backend, linux-64 only)
@@ -102,45 +102,44 @@ linux:
102102
# nightly-cuda-core (released cuda-core from PyPI against main pathfinder/bindings)
103103
- { ARCH: 'amd64', PY_VER: '3.12', CUDA_VER: '13.3.0', LOCAL_CTK: '0', GPU: 'a100', GPU_COUNT: '1', DRIVER: 'latest', ENV: { MODE: 'nightly-cuda-core' } }
104104
# nightly-standard (arm64 nightly-only runners — per runner team request)
105-
# TODO: gh200 row disabled — currently hangs on stream-ordered memory
106-
# allocator (cudaMallocAsync); runner pool needs fixing first.
107-
# - { ARCH: 'arm64', PY_VER: '3.14', CUDA_VER: '13.3.0', LOCAL_CTK: '1', GPU: 'gh200', GPU_COUNT: '1', DRIVER: 'latest', ENV: { MODE: 'nightly-standard' } }
105+
- { ARCH: 'arm64', PY_VER: '3.13', CUDA_VER: '13.3.0', LOCAL_CTK: '1', GPU: 'gh200', GPU_COUNT: '1', DRIVER: 'latest', ENV: { MODE: 'nightly-standard' } }
106+
- { ARCH: 'arm64', PY_VER: '3.14', CUDA_VER: '13.3.0', LOCAL_CTK: '1', GPU: 'gb300', GPU_COUNT: '1', DRIVER: 'latest', ENV: { MODE: 'nightly-standard' } }
108107
- { ARCH: 'arm64', PY_VER: '3.14', CUDA_VER: '13.3.0', LOCAL_CTK: '1', GPU: 'l4', GPU_COUNT: '2', DRIVER: 'latest', ENV: { MODE: 'nightly-standard' } }
109108
- { ARCH: 'arm64', PY_VER: '3.14t', CUDA_VER: '13.3.0', LOCAL_CTK: '1', GPU: 'l4', GPU_COUNT: '2', DRIVER: 'latest', ENV: { MODE: 'nightly-standard' } }
110109

111110
windows:
112111
pull-request:
113112
# win-64
114-
- { ARCH: 'amd64', PY_VER: '3.10', CUDA_VER: '12.9.1', LOCAL_CTK: '0', GPU: 'rtx2080', GPU_COUNT: '1', DRIVER: 'latest', DRIVER_MODE: 'WDDM' }
115-
- { ARCH: 'amd64', PY_VER: '3.10', CUDA_VER: '13.0.2', LOCAL_CTK: '1', GPU: 'rtxpro6000', GPU_COUNT: '1', DRIVER: 'latest', DRIVER_MODE: 'TCC' }
116-
- { ARCH: 'amd64', PY_VER: '3.10', CUDA_VER: '13.3.0', LOCAL_CTK: '1', GPU: 'rtxpro6000', GPU_COUNT: '1', DRIVER: 'latest', DRIVER_MODE: 'TCC' }
117-
- { ARCH: 'amd64', PY_VER: '3.11', CUDA_VER: '12.9.1', LOCAL_CTK: '1', GPU: 'v100', GPU_COUNT: '1', DRIVER: 'latest', DRIVER_MODE: 'MCDM' }
118-
- { ARCH: 'amd64', PY_VER: '3.11', CUDA_VER: '13.0.2', LOCAL_CTK: '0', GPU: 'rtx4090', GPU_COUNT: '1', DRIVER: 'latest', DRIVER_MODE: 'WDDM' }
119-
- { ARCH: 'amd64', PY_VER: '3.11', CUDA_VER: '13.3.0', LOCAL_CTK: '0', GPU: 'rtx4090', GPU_COUNT: '1', DRIVER: 'latest', DRIVER_MODE: 'WDDM' }
120-
- { ARCH: 'amd64', PY_VER: '3.12', CUDA_VER: '12.9.1', LOCAL_CTK: '0', GPU: 'l4', GPU_COUNT: '1', DRIVER: 'latest', DRIVER_MODE: 'MCDM' }
121-
- { ARCH: 'amd64', PY_VER: '3.12', CUDA_VER: '13.0.2', LOCAL_CTK: '1', GPU: 'a100', GPU_COUNT: '1', DRIVER: 'latest', DRIVER_MODE: 'TCC' }
122-
- { ARCH: 'amd64', PY_VER: '3.12', CUDA_VER: '13.3.0', LOCAL_CTK: '1', GPU: 'a100', GPU_COUNT: '1', DRIVER: 'latest', DRIVER_MODE: 'TCC' }
123-
- { ARCH: 'amd64', PY_VER: '3.13', CUDA_VER: '12.9.1', LOCAL_CTK: '1', GPU: 'l4', GPU_COUNT: '1', DRIVER: 'latest', DRIVER_MODE: 'TCC' }
124-
- { ARCH: 'amd64', PY_VER: '3.13', CUDA_VER: '13.0.2', LOCAL_CTK: '0', GPU: 'rtxpro6000', GPU_COUNT: '1', DRIVER: 'latest', DRIVER_MODE: 'MCDM' }
125-
- { ARCH: 'amd64', PY_VER: '3.13', CUDA_VER: '13.3.0', LOCAL_CTK: '0', GPU: 'rtxpro6000', GPU_COUNT: '1', DRIVER: 'latest', DRIVER_MODE: 'MCDM', ENV: { CUDA_PYTHON_CUDA_PER_THREAD_DEFAULT_STREAM: '1' } }
126-
- { ARCH: 'amd64', PY_VER: '3.14', CUDA_VER: '12.9.1', LOCAL_CTK: '0', GPU: 'v100', GPU_COUNT: '1', DRIVER: 'latest', DRIVER_MODE: 'TCC' }
127-
- { ARCH: 'amd64', PY_VER: '3.14', CUDA_VER: '13.0.2', LOCAL_CTK: '1', GPU: 'l4', GPU_COUNT: '1', DRIVER: 'latest', DRIVER_MODE: 'MCDM' }
128-
- { ARCH: 'amd64', PY_VER: '3.14', CUDA_VER: '13.3.0', LOCAL_CTK: '1', GPU: 'l4', GPU_COUNT: '1', DRIVER: 'latest', DRIVER_MODE: 'MCDM' }
129-
- { ARCH: 'amd64', PY_VER: '3.14t', CUDA_VER: '12.9.1', LOCAL_CTK: '1', GPU: 'l4', GPU_COUNT: '1', DRIVER: 'latest', DRIVER_MODE: 'TCC' }
130-
- { ARCH: 'amd64', PY_VER: '3.14t', CUDA_VER: '13.0.2', LOCAL_CTK: '0', GPU: 'a100', GPU_COUNT: '1', DRIVER: 'latest', DRIVER_MODE: 'MCDM' }
131-
- { ARCH: 'amd64', PY_VER: '3.14t', CUDA_VER: '13.3.0', LOCAL_CTK: '0', GPU: 'a100', GPU_COUNT: '1', DRIVER: 'latest', DRIVER_MODE: 'MCDM' }
113+
- { ARCH: 'amd64', PY_VER: '3.10', CUDA_VER: '12.9.1', LOCAL_CTK: '0', GPU: 'rtx2080', GPU_COUNT: '1', DRIVER: 'latest', DRIVER_MODE: 'WDDM' }
114+
- { ARCH: 'amd64', PY_VER: '3.10', CUDA_VER: '13.0.2', LOCAL_CTK: '1', GPU: 'rtxpro6000', GPU_COUNT: '1', DRIVER: 'latest', DRIVER_MODE: 'TCC' }
115+
- { ARCH: 'amd64', PY_VER: '3.10', CUDA_VER: '13.3.0', LOCAL_CTK: '1', GPU: 'rtxpro6000', GPU_COUNT: '1', DRIVER: 'latest', DRIVER_MODE: 'TCC' }
116+
- { ARCH: 'amd64', PY_VER: '3.11', CUDA_VER: '12.9.1', LOCAL_CTK: '1', GPU: 'v100', GPU_COUNT: '1', DRIVER: 'latest', DRIVER_MODE: 'MCDM' }
117+
- { ARCH: 'amd64', PY_VER: '3.11', CUDA_VER: '13.0.2', LOCAL_CTK: '0', GPU: 'rtx4090', GPU_COUNT: '1', DRIVER: 'latest', DRIVER_MODE: 'WDDM' }
118+
- { ARCH: 'amd64', PY_VER: '3.11', CUDA_VER: '13.3.0', LOCAL_CTK: '0', GPU: 'rtx4090', GPU_COUNT: '1', DRIVER: 'latest', DRIVER_MODE: 'WDDM' }
119+
- { ARCH: 'amd64', PY_VER: '3.12', CUDA_VER: '12.9.1', LOCAL_CTK: '0', GPU: 'l4', GPU_COUNT: '1', DRIVER: 'latest', DRIVER_MODE: 'MCDM' }
120+
- { ARCH: 'amd64', PY_VER: '3.12', CUDA_VER: '13.0.2', LOCAL_CTK: '1', GPU: 'a100', GPU_COUNT: '1', DRIVER: 'latest', DRIVER_MODE: 'TCC' }
121+
- { ARCH: 'amd64', PY_VER: '3.12', CUDA_VER: '13.3.0', LOCAL_CTK: '1', GPU: 'a100', GPU_COUNT: '1', DRIVER: 'latest', DRIVER_MODE: 'TCC' }
122+
- { ARCH: 'amd64', PY_VER: '3.13', CUDA_VER: '12.9.1', LOCAL_CTK: '1', GPU: 'l4', GPU_COUNT: '1', DRIVER: 'latest', DRIVER_MODE: 'TCC' }
123+
- { ARCH: 'amd64', PY_VER: '3.13', CUDA_VER: '13.0.2', LOCAL_CTK: '0', GPU: 'rtxpro6000', GPU_COUNT: '1', DRIVER: 'latest', DRIVER_MODE: 'MCDM' }
124+
- { ARCH: 'amd64', PY_VER: '3.13', CUDA_VER: '13.3.0', LOCAL_CTK: '0', GPU: 'rtxpro6000', GPU_COUNT: '1', DRIVER: 'latest', DRIVER_MODE: 'MCDM', ENV: { CUDA_PYTHON_CUDA_PER_THREAD_DEFAULT_STREAM: '1' } }
125+
- { ARCH: 'amd64', PY_VER: '3.14', CUDA_VER: '12.9.1', LOCAL_CTK: '0', GPU: 'v100', GPU_COUNT: '1', DRIVER: 'latest', DRIVER_MODE: 'TCC' }
126+
- { ARCH: 'amd64', PY_VER: '3.14', CUDA_VER: '13.0.2', LOCAL_CTK: '1', GPU: 'l4', GPU_COUNT: '1', DRIVER: 'latest', DRIVER_MODE: 'MCDM' }
127+
- { ARCH: 'amd64', PY_VER: '3.14', CUDA_VER: '13.3.0', LOCAL_CTK: '1', GPU: 'l4', GPU_COUNT: '1', DRIVER: 'latest', DRIVER_MODE: 'MCDM' }
128+
- { ARCH: 'amd64', PY_VER: '3.14t', CUDA_VER: '12.9.1', LOCAL_CTK: '1', GPU: 'l4', GPU_COUNT: '1', DRIVER: 'latest', DRIVER_MODE: 'TCC' }
129+
- { ARCH: 'amd64', PY_VER: '3.14t', CUDA_VER: '13.0.2', LOCAL_CTK: '0', GPU: 'a100', GPU_COUNT: '1', DRIVER: 'latest', DRIVER_MODE: 'MCDM' }
130+
- { ARCH: 'amd64', PY_VER: '3.14t', CUDA_VER: '13.3.0', LOCAL_CTK: '0', GPU: 'a100', GPU_COUNT: '1', DRIVER: 'latest', DRIVER_MODE: 'MCDM' }
132131
# special runners
133-
- { ARCH: 'amd64', PY_VER: '3.14', CUDA_VER: '13.3.0', LOCAL_CTK: '1', GPU: 't4', GPU_COUNT: '2', DRIVER: 'latest', DRIVER_MODE: 'TCC' }
134-
- { ARCH: 'amd64', PY_VER: '3.14', CUDA_VER: '13.3.0', LOCAL_CTK: '0', GPU: 'h100', GPU_COUNT: '2', DRIVER: 'latest', DRIVER_MODE: 'MCDM' }
132+
- { ARCH: 'amd64', PY_VER: '3.14', CUDA_VER: '13.3.0', LOCAL_CTK: '1', GPU: 't4', GPU_COUNT: '2', DRIVER: 'latest', DRIVER_MODE: 'TCC' }
133+
- { ARCH: 'amd64', PY_VER: '3.14', CUDA_VER: '13.3.0', LOCAL_CTK: '0', GPU: 'h100', GPU_COUNT: '2', DRIVER: 'latest', DRIVER_MODE: 'MCDM' }
135134
nightly:
136135
# nightly-pytorch
137-
- { ARCH: 'amd64', PY_VER: '3.12', CUDA_VER: '12.6.3', LOCAL_CTK: '0', GPU: 'l4', GPU_COUNT: '1', DRIVER: 'latest', DRIVER_MODE: 'TCC', ENV: { MODE: 'nightly-pytorch', TORCH_VER: '2.12.1', TORCH_CUDA: 'cu126' } }
138-
- { ARCH: 'amd64', PY_VER: '3.12', CUDA_VER: '13.0.2', LOCAL_CTK: '0', GPU: 'l4', GPU_COUNT: '1', DRIVER: 'latest', DRIVER_MODE: 'TCC', ENV: { MODE: 'nightly-pytorch', TORCH_VER: '2.12.1', TORCH_CUDA: 'cu130' } }
139-
- { ARCH: 'amd64', PY_VER: '3.12', CUDA_VER: '12.6.3', LOCAL_CTK: '0', GPU: 'l4', GPU_COUNT: '1', DRIVER: 'latest', DRIVER_MODE: 'TCC', ENV: { MODE: 'nightly-pytorch', TORCH_VER: '2.9.1', TORCH_CUDA: 'cu126' } }
140-
- { ARCH: 'amd64', PY_VER: '3.12', CUDA_VER: '13.0.2', LOCAL_CTK: '0', GPU: 'l4', GPU_COUNT: '1', DRIVER: 'latest', DRIVER_MODE: 'TCC', ENV: { MODE: 'nightly-pytorch', TORCH_VER: '2.9.1', TORCH_CUDA: 'cu130' } }
136+
- { ARCH: 'amd64', PY_VER: '3.12', CUDA_VER: '12.6.3', LOCAL_CTK: '0', GPU: 'l4', GPU_COUNT: '1', DRIVER: 'latest', DRIVER_MODE: 'TCC', ENV: { MODE: 'nightly-pytorch', TORCH_VER: '2.12.1', TORCH_CUDA: 'cu126' } }
137+
- { ARCH: 'amd64', PY_VER: '3.12', CUDA_VER: '13.0.2', LOCAL_CTK: '0', GPU: 'l4', GPU_COUNT: '1', DRIVER: 'latest', DRIVER_MODE: 'TCC', ENV: { MODE: 'nightly-pytorch', TORCH_VER: '2.12.1', TORCH_CUDA: 'cu130' } }
138+
- { ARCH: 'amd64', PY_VER: '3.12', CUDA_VER: '12.6.3', LOCAL_CTK: '0', GPU: 'l4', GPU_COUNT: '1', DRIVER: 'latest', DRIVER_MODE: 'TCC', ENV: { MODE: 'nightly-pytorch', TORCH_VER: '2.9.1', TORCH_CUDA: 'cu126' } }
139+
- { ARCH: 'amd64', PY_VER: '3.12', CUDA_VER: '13.0.2', LOCAL_CTK: '0', GPU: 'l4', GPU_COUNT: '1', DRIVER: 'latest', DRIVER_MODE: 'TCC', ENV: { MODE: 'nightly-pytorch', TORCH_VER: '2.9.1', TORCH_CUDA: 'cu130' } }
141140
# nightly-numba-cuda
142-
- { ARCH: 'amd64', PY_VER: '3.12', CUDA_VER: '12.9.1', LOCAL_CTK: '0', GPU: 'l4', GPU_COUNT: '1', DRIVER: 'latest', DRIVER_MODE: 'TCC', ENV: { MODE: 'nightly-numba-cuda' } }
143-
- { ARCH: 'amd64', PY_VER: '3.12', CUDA_VER: '13.3.0', LOCAL_CTK: '0', GPU: 'l4', GPU_COUNT: '1', DRIVER: '596.36', DRIVER_MODE: 'TCC', ENV: { MODE: 'nightly-numba-cuda' } }
141+
- { ARCH: 'amd64', PY_VER: '3.12', CUDA_VER: '12.9.1', LOCAL_CTK: '0', GPU: 'l4', GPU_COUNT: '1', DRIVER: 'latest', DRIVER_MODE: 'TCC', ENV: { MODE: 'nightly-numba-cuda' } }
142+
- { ARCH: 'amd64', PY_VER: '3.12', CUDA_VER: '13.3.0', LOCAL_CTK: '0', GPU: 'l4', GPU_COUNT: '1', DRIVER: '596.36', DRIVER_MODE: 'TCC', ENV: { MODE: 'nightly-numba-cuda' } }
144143
# nightly-numba-cuda-mlir (MLIR backend, win-64)
145144
- { ARCH: 'amd64', PY_VER: '3.12', CUDA_VER: '12.9.1', LOCAL_CTK: '0', GPU: 'rtxpro6000', GPU_COUNT: '1', DRIVER: 'latest', DRIVER_MODE: 'MCDM', ENV: { MODE: 'nightly-numba-cuda-mlir' } }
146145
- { ARCH: 'amd64', PY_VER: '3.12', CUDA_VER: '13.3.0', LOCAL_CTK: '0', GPU: 'rtxpro6000', GPU_COUNT: '1', DRIVER: 'latest', DRIVER_MODE: 'MCDM', ENV: { MODE: 'nightly-numba-cuda-mlir' } }

‎ci/tools/run-tests‎

Lines changed: 14 additions & 6 deletions
Original file line numberDiff line numberDiff line change
@@ -59,7 +59,7 @@ elif [[ "${test_module}" == "bindings" ]]; then
5959
elif [[ "${test_module}" == "core" || "${test_module}" == nightly-* ]]; then
6060
# Shared setup for core and nightly modes.
6161
TEST_CUDA_MAJOR="$(cut -d '.' -f 1 <<< ${CUDA_VER})"
62-
CUDA_VER_MINOR="$(cut -d '.' -f 1-2 <<< "${CUDA_VER}")"
62+
TEST_CUDA_MAJOR_MINOR="$(cut -d '.' -f 1-2 <<< "${CUDA_VER}")"
6363

6464
FREE_THREADING=""
6565
if python -c 'import sys; assert not sys._is_gil_enabled()' 2> /dev/null; then
@@ -101,7 +101,7 @@ elif [[ "${test_module}" == "core" || "${test_module}" == nightly-* ]]; then
101101
echo "Installing core wheel"
102102
# Constrain cuda-toolkit to the requested CTK version to avoid
103103
# pip pulling in a newer nvidia-cuda-runtime that conflicts with it.
104-
pip install "${CORE_WHL[@]}" --group "test-cu${TEST_CUDA_MAJOR}${FREE_THREADING}" "cuda-toolkit==${CUDA_VER_MINOR}.*"
104+
pip install "${CORE_WHL[@]}" --group "test-cu${TEST_CUDA_MAJOR}${FREE_THREADING}" "cuda-toolkit==${TEST_CUDA_MAJOR_MINOR}.*"
105105
echo "Installed packages before core tests:"
106106
pip list
107107
echo "Running core tests"
@@ -145,7 +145,7 @@ elif [[ "${test_module}" == "core" || "${test_module}" == nightly-* ]]; then
145145
# Use cuda-toolkit[cudart] only — torch brings its own nvcc/nvrtc/etc.
146146
# This avoids version conflicts between our nvidia-* pins and torch's.
147147
echo "Installing pathfinder + bindings + core + test deps + PyTorch ${TORCH_VER} (${TORCH_CUDA})"
148-
PIP_ARGS+=("cuda-toolkit[cudart]==${CUDA_VER_MINOR}.*")
148+
PIP_ARGS+=("cuda-toolkit[cudart]==${TEST_CUDA_MAJOR_MINOR}.*")
149149
if [[ "${TORCH_VER}" == "latest" ]]; then
150150
PIP_ARGS+=(torch)
151151
else
@@ -156,16 +156,24 @@ elif [[ "${test_module}" == "core" || "${test_module}" == nightly-* ]]; then
156156
echo "Installing pathfinder + bindings + core + test deps + numba-cuda"
157157
# numba-cuda's test-cuXX group deps (can't use --group for a wheel install):
158158
PIP_ARGS+=(
159-
"cuda-toolkit[curand,cublas]==${CUDA_VER_MINOR}.*"
159+
"cuda-toolkit[curand,cublas]==${TEST_CUDA_MAJOR_MINOR}.*"
160160
"numba-cuda[cu${TEST_CUDA_MAJOR}]"
161161
"cupy-cuda${TEST_CUDA_MAJOR}x"
162162
psutil cffi pytest-xdist pytest-benchmark filecheck ml_dtypes statistics
163163
)
164164
elif [[ "${test_module}" == "nightly-numba-cuda-mlir" ]]; then
165-
echo "Installing pathfinder + bindings + core + numba-cuda-mlir"
165+
echo "Installing pathfinder + bindings + core + numba-cuda-mlir + cupy"
166166
# numpy<2.5: numba-cuda-mlir 0.4.0 registers np.row_stack, which was
167167
# removed in NumPy 2.5. See NVIDIA/numba-cuda-mlir#154.
168-
PIP_ARGS+=("numba-cuda-mlir[cu${TEST_CUDA_MAJOR}]" "numpy<2.5")
168+
# curand/cublas + cupy: some numba-cuda-mlir tests (e.g.
169+
# test_fortran_contiguous) call cp.random, which dlopens libcurand at
170+
# runtime. Mirror the numba-cuda branch, which installs the same libs.
171+
PIP_ARGS+=(
172+
"numba-cuda-mlir[cu${TEST_CUDA_MAJOR}]"
173+
"numpy<2.5"
174+
"cuda-toolkit[curand,cublas]==${TEST_CUDA_MAJOR_MINOR}.*"
175+
"cupy-cuda${TEST_CUDA_MAJOR}x"
176+
)
169177
fi
170178

171179
pip install "${PIP_ARGS[@]}"

‎cuda_core/tests/test_binaries/build_test_binaries.py‎

Lines changed: 1 addition & 1 deletion
Original file line numberDiff line numberDiff line change
@@ -37,7 +37,7 @@ def main() -> None:
3737
"nvcc",
3838
"-dc",
3939
*nvcc_extra_flags,
40-
"-arch=all-major",
40+
"-arch=all",
4141
"-o",
4242
str(object_path),
4343
str(source_path),

0 commit comments

Comments
 (0)