python3Packages.pycuda: add maintainer and clean up #465047
There are no files selected for viewing
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
| Original file line number | Diff line number | Diff line change |
|---|---|---|
| @@ -1,81 +1,233 @@ | ||
| { | ||
| _cuda, | ||
| boost, | ||
| buildPythonPackage, | ||
| cudaPackages, | ||
| fetchFromGitHub, | ||
| lib, | ||
| mako, | ||
| numpy, | ||
| platformdirs, | ||
| python, | ||
| pytools, | ||
| setuptools, | ||
| wheel, | ||
| writableTmpDirAsHomeHook, | ||
| writeShellApplication, | ||
| stdenvNoCC, | ||
| }: | ||
| let | ||
| inherit (_cuda.lib) dropDots; | ||
| inherit (cudaPackages) | ||
| backendStdenv | ||
| cuda_cudart | ||
| cuda_cccl | ||
| cuda_nvcc | ||
| cuda_profiler_api | ||
| libcurand | ||
| ; | ||
| inherit (lib) | ||
| getBin | ||
| getFirstOutput | ||
| getInclude | ||
| getLib | ||
| licenses | ||
| maintainers | ||
| teams | ||
| ; | ||
| in | ||
| buildPythonPackage { | ||
| __structuredAttrs = true; | ||
|
|
||
| pname = "pycuda"; | ||
| version = "2025.1.2"; | ||
|
|
||
| pyproject = true; | ||
|
|
||
| stdenv = backendStdenv; | ||
|
|
||
| src = fetchFromGitHub { | ||
| owner = "inducer"; | ||
| repo = "pycuda"; | ||
| tag = "v2025.1.2"; | ||
| hash = "sha256-JMGVNjiKCAno29df8Zk3njvpgvz9JE8mb0HeJMVTnCQ="; | ||
| # Use the vendored compyte source rather than tracking it as a separate dependency. | ||
| # As an added bonus, this should unbreak the update script added by buildPythonPackage. | ||
| fetchSubmodules = true; | ||
| }; | ||
|
|
||
| build-system = [ | ||
| setuptools | ||
| wheel | ||
|
Contributor
There was a problem hiding this comment. You should be able to get rid of |
||
| ]; | ||
|
|
||
| nativeBuildInputs = [ | ||
| cuda_nvcc | ||
| ]; | ||
|
|
||
| prePatch = '' | ||
| nixLog "removing vendored boost source" | ||
| rm -rf "$PWD/bpl-subset" | ||
|
|
||
| nixLog "patching $PWD/pycuda/compiler.py::compile to fix CUDA include paths" | ||
| substituteInPlace "$PWD/pycuda/compiler.py" \ | ||
| --replace-fail \ | ||
| 'include_dirs = [*include_dirs, _find_pycuda_include_path()]' \ | ||
| 'include_dirs = [ | ||
| *include_dirs, | ||
| "${getInclude cuda_nvcc}/include", | ||
| "${getInclude cuda_cudart}/include", | ||
| "${getInclude cuda_cccl}/include", | ||
| "${getInclude cuda_profiler_api}/include", | ||
| "${getInclude libcurand}/include", | ||
| _find_pycuda_include_path(), | ||
| ]' | ||
|
|
||
| nixLog "patching $PWD/pycuda/compiler.py::compile to fix NVCC path" | ||
| substituteInPlace "$PWD/pycuda/compiler.py" \ | ||
| --replace-fail \ | ||
| 'nvcc="nvcc"' \ | ||
| 'nvcc="${getBin cuda_nvcc}/bin/nvcc"' | ||
|
|
||
| nixLog "patching $PWD/pycuda/compiler.py::DynamicModule.__init__ to fix CUDA runtime library path" | ||
| substituteInPlace "$PWD/pycuda/compiler.py" \ | ||
| --replace-fail \ | ||
| 'cuda_libdir=None,' \ | ||
| 'cuda_libdir="${getLib cuda_cudart}/lib",' | ||
| ''; | ||
|
|
||
| dependencies = [ | ||
| boost | ||
| mako | ||
| numpy | ||
| platformdirs | ||
| pytools | ||
| ]; | ||
|
|
||
| buildInputs = [ | ||
| cuda_cccl | ||
| cuda_cudart | ||
| cuda_nvcc | ||
| cuda_profiler_api | ||
| libcurand | ||
| ]; | ||
|
|
||
| configureScript = "./configure.py"; | ||
|
|
||
| # configure.py doesn't support the installation directory arguments _multioutConfig sets. | ||
| # The other argument provided by configurePhase, like --prefix, --enable-shared, and --disable-static are ignored. | ||
| setOutputFlags = false; | ||
|
|
||
| configureFlags = [ | ||
| # The expected boost python library name is something like boost_python-py313, but our library name doesn't have a | ||
| # hyphen. The pythonVersion is already a major-minor version, so we just need to remove the dot. | ||
| "--no-use-shipped-boost" | ||
| "--boost-python-libname=boost_python${dropDots python.pythonVersion}" | ||
| # Provide paths to our CUDA libraries. | ||
| "--cudadrv-lib-dir=${getFirstOutput [ "stubs" "lib" ] cuda_cudart}/lib/stubs" | ||
| ]; | ||
|
|
||
| # Requires access to libcuda.so.1 which is provided by the driver | ||
| doCheck = false; | ||
|
|
||
| # From setup.py | ||
| pythonImportsCheck = [ | ||
| "pycuda" | ||
| # "pycuda.gl" # Requires the CUDA driver | ||
| "pycuda.sparse" | ||
| "pycuda.compyte" | ||
| ]; | ||
|
|
||
| # TODO: Split into testers and tests. | ||
| # NOTE: Tests take 23m to run on a 4090 and require 18GB of VRAM. | ||
| passthru = { | ||
| testers.tester = writeShellApplication { | ||
| derivationArgs = { | ||
| __structuredAttrs = true; | ||
| strictDeps = true; | ||
| }; | ||
| name = "pycuda-tester"; | ||
| runtimeInputs = [ | ||
| (python.withPackages (ps: [ | ||
| ps.pycuda | ||
| ps.pytest | ||
| ])) | ||
| ]; | ||
| text = '' | ||
| echo "Copying pycuda test sources to $PWD/pycuda_test_sources" | ||
| mkdir -p "$PWD/pycuda_test_sources" | ||
| cp -r "${python.pkgs.pycuda.src}"/test "$PWD/pycuda_test_sources" | ||
| chmod -R u+w "$PWD/pycuda_test_sources" | ||
|
|
||
| pushd "$PWD/pycuda_test_sources" | ||
| pytest "$@" | ||
| popd | ||
| ''; | ||
| }; | ||
| tests = | ||
| let | ||
| makeTest = | ||
| name: testArgs: | ||
| stdenvNoCC.mkDerivation { | ||
| __structuredAttrs = true; | ||
| strictDeps = true; | ||
|
|
||
| inherit name testArgs; | ||
|
|
||
| dontUnpack = true; | ||
|
|
||
| nativeBuildInputs = [ | ||
| # cuda_nvcc | ||
| python.pkgs.pycuda.passthru.testers.tester | ||
| writableTmpDirAsHomeHook | ||
| ]; | ||
|
|
||
| dontConfigure = true; | ||
|
|
||
| buildPhase = '' | ||
| nixLog "using testArgs: ''${testArgs[*]@Q}" | ||
| pycuda-tester "''${testArgs[@]}" || { | ||
| nixErrorLog "pycuda-tester finished with non-zero exit code: $?" | ||
| exit 1 | ||
| } | ||
| ''; | ||
|
|
||
| postInstall = '' | ||
| touch $out | ||
| ''; | ||
|
|
||
| requiredSystemFeatures = [ "cuda" ]; | ||
| }; | ||
| in | ||
| { | ||
| # Fast: ~10s on a 4090 | ||
| driver = makeTest "pycuda-driver-tests" [ "test/test_driver.py" ]; | ||
| # Fast: ~16s on a 4090 | ||
| cumath = makeTest "pycuda-cumath-tests" [ "test/test_cumath.py" ]; | ||
| # Fast: ~3m on a 4090 | ||
| gpuarray-fast = makeTest "pycuda-gpuarray-fast-tests" [ | ||
| "test/test_gpuarray.py" | ||
| "-k" | ||
| "not test_curand_wrappers" | ||
| ]; | ||
| # EXTREMELY SLOW: ~20m on a 4090 | ||
| gpuarray-slow = makeTest "pycuda-gpuarray-slow-tests" [ | ||
| "test/test_gpuarray.py" | ||
| "-k" | ||
| "test_curand_wrappers" | ||
| ]; | ||
| }; | ||
| }; | ||
|
|
||
| meta = { | ||
| description = "CUDA integration for Python"; | ||
| homepage = "https://github.com/inducer/pycuda/"; | ||
| license = licenses.mit; | ||
| platforms = [ | ||
| "aarch64-linux" | ||
| "x86_64-linux" | ||
| ]; | ||
| maintainers = with maintainers; [ connorbaker ]; | ||
| teams = [ teams.cuda ]; | ||
| }; | ||
| } | ||
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Oops, something went wrong.
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Why not using
wrapProgramhere?