feat(cccl): integrate missing CCCL directories — python/, ci/, .agent/, docs/, test/

Sparse-checkout from NVIDIA/cccl main branch to complete cccl_upstream: Added: - python/cuda_cccl/ (226 files) — Python bindings for device-level algorithms Critical for muh toolchain: cuda.compute.reduce_into, scan, radix_sort, etc. Includes 204 .py files with full test coverage for all 27 algorithms - ci/ (163 files) — Build/test infrastructure build_cub.sh, test_cub.sh, build_and_test_targets.sh, matrix.yaml Directly maps to our [INFRA-CI] and [INFRA-BUILD] items - .agent/skills/ (7 files) — NVIDIA's own agent skills for CCCL cccl-style/SKILL.md, cccl-test/SKILL.md, sass-diff/SKILL.md - docs/ (491 files) — Official CCCL documentation CI references, CMake guides, Python compute docs, libcudacxx PTX docs - test/ (12 files) — Top-level integration tests (cuda_smoke, stdpar) - Root configs: .clang-format, .clang-tidy, CONTRIBUTING.md, pyproject.toml - CLAUDE.md symlink → AGENTS.md (NVIDIA's standard) cccl_upstream now mirrors full NVIDIA/cccl structure: Before: 42M (cub + thrust + libcudacxx + cudax + c + examples + benchmarks) After: 53M (+python +ci +docs +.agent +test +configs) This completes the CCCL base needed for: - [muh-bench] items: ci/util/build_and_test_targets.sh for targeted builds - [CCCL-verify] items: python/cuda_cccl/tests/ as reference implementations - [CCCL-test] items: ci/test_cub.sh, ci/test_thrust.sh - Agent workflow: .agent/skills/ for consistent style and test patterns
2026-08-07 02:34:33 +00:00
parent 3f97dca7ad
commit 2a7ca101d7
908 changed files with 121615 additions and 0 deletions
--- a/cccl_upstream/python/cuda_cccl/pyproject.toml
+++ b/cccl_upstream/python/cuda_cccl/pyproject.toml
@@ -0,0 +1,204 @@
+# Copyright (c) 2025-2026, NVIDIA CORPORATION & AFFILIATES. ALL RIGHTS RESERVED.
+#
+# SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
+
+[build-system]
+# cmake>=3.30 is needed for FindCUDAToolkit's CUDA::nvfatbin /
+# CUDA::nvfatbin_static imported targets, which the v2 (HostJIT) backend links
+# against. Listed here so isolated builds (pip's default) pick it up.
+requires = ["scikit-build-core>=0.10", "setuptools_scm", "cython>=3.1.0", "cmake>=3.30"]
+build-backend = "scikit_build_core.build"
+
+[project]
+name = "cuda-cccl"
+description = "CUDA Core Compute Libraries for Python"
+authors = [{ name = "NVIDIA Corporation" }]
+classifiers = [
+  "Development Status :: 4 - Beta",
+  "Intended Audience :: Developers",
+  "Topic :: Software Development :: Libraries :: Python Modules",
+  "Programming Language :: Python :: 3 :: Only",
+  "Programming Language :: Python :: 3.10",
+  "Programming Language :: Python :: 3.11",
+  "Programming Language :: Python :: 3.12",
+  "Programming Language :: Python :: 3.13",
+  "Programming Language :: Python :: 3.14",
+  "Environment :: GPU :: NVIDIA CUDA",
+  "License :: OSI Approved :: Apache Software License",
+  "Operating System :: POSIX :: Linux",
+  "Operating System :: Microsoft :: Windows",
+]
+requires-python = ">=3.10"
+
+dependencies = [
+  "numpy",
+  "cuda-pathfinder>=1.2.3",
+  "cuda-core",
+  "typing_extensions",
+]
+
+dynamic = ["version"]
+readme = { file = "README.md", content-type = "text/markdown" }
+
+[project.optional-dependencies]
+minimal-cu12 = [
+  "cuda-bindings>=12.9.1,<13.0.0",
+  "cuda-toolkit[nvrtc,nvjitlink,cudart,nvcc]==12.*",
+]
+minimal-cu13 = [
+  "cuda-bindings>=13.0.0,<14.0.0",
+  "cuda-toolkit[nvrtc,nvjitlink,cudart,nvcc,nvvm]==13.*",
+]
+# sysctk variants: like cu12/cu13 but without cuda-toolkit — the user is
+# responsible for providing a compatible CUDA toolkit
+minimal-sysctk12 = [
+  "cuda-bindings>=12.9.1,<13.0.0",
+]
+minimal-sysctk13 = [
+  "cuda-bindings>=13.0.0,<14.0.0",
+]
+cu12 = [
+  "cuda-cccl[minimal-cu12]",
+  # numba-cuda currently expects numba.cuda.types.NPDatetime, removed in numba 0.66.
+  "numba>=0.60.0,<0.66",
+  "numba-cuda[cu12]>=0.23.0,!=0.27.*,!=0.28.*,!=0.29.*,!=0.30.0",
+]
+cu13 = [
+  "cuda-cccl[minimal-cu13]",
+  # numba-cuda currently expects numba.cuda.types.NPDatetime, removed in numba 0.66.
+  "numba>=0.60.0,<0.66",
+  "numba-cuda[cu13]>=0.23.0,!=0.27.*,!=0.28.*,!=0.29.*,!=0.30.0"
+]
+sysctk12 = [
+  "cuda-cccl[minimal-sysctk12]",
+  # numba-cuda currently expects numba.cuda.types.NPDatetime, removed in numba 0.66.
+  "numba>=0.60.0,<0.66",
+  # plain numba-cuda (no [cu12]) so it relies on the system-provided CUDA toolkit
+  # instead of pulling the toolkit from pip -- the whole point of the sysctk extra.
+  "numba-cuda>=0.23.0,!=0.27.*,!=0.28.*,!=0.29.*,!=0.30.0"
+]
+sysctk13 = [
+  "cuda-cccl[minimal-sysctk13]",
+  # numba-cuda currently expects numba.cuda.types.NPDatetime, removed in numba 0.66.
+  "numba>=0.60.0,<0.66",
+  # plain numba-cuda (no [cu13]) so it relies on the system-provided CUDA toolkit
+  # instead of pulling the toolkit from pip -- the whole point of the sysctk extra.
+  "numba-cuda>=0.23.0,!=0.27.*,!=0.28.*,!=0.29.*,!=0.30.0",
+]
+# The test extras do not include CuPy. The cuda.compute examples (run via
+# tests/test_examples.py) require it; install it separately (e.g.
+# `pip install cupy-cuda13x`).
+test-cu12 = [
+  # an undocumented way to inherit the dependencies of the cu12 extra.
+  # GH pypa #11296
+  "cuda-cccl[cu12]",
+  "pytest",
+  "pytest-xdist",
+]
+test-cu13 = ["cuda-cccl[cu13]", "pytest", "pytest-xdist"]
+test-sysctk12 = [
+  "cuda-cccl[sysctk12]",
+  "pytest",
+  "pytest-xdist",
+]
+test-sysctk13 = [
+  "cuda-cccl[sysctk13]",
+  "pytest",
+  "pytest-xdist",
+]
+bench-cu12 = [
+  "cuda-cccl[cu12]",
+  "cuda-bench[cu12]",
+  "cupy-cuda12x",
+  "pytest-benchmark",
+]
+bench-cu13 = [
+  "cuda-cccl[cu13]",
+  "cuda-bench[cu13]",
+  "cupy-cuda13x",
+  "pytest-benchmark",
+]
+bench-sysctk12 = [
+  "cuda-cccl[sysctk12]",
+  "cuda-bench[cu12]",
+  "cupy-cuda12x",
+  "pytest-benchmark",
+]
+bench-sysctk13 = [
+  "cuda-cccl[sysctk13]",
+  "cuda-bench[cu13]",
+  "cupy-cuda13x",
+  "pytest-benchmark",
+]
+
+[project.urls]
+Homepage = "https://github.com/NVIDIA/cccl"
+Repository = "https://github.com/NVIDIA/cccl"
+Documentation = "https://nvidia.github.io/cccl"
+Issues = "https://github.com/NVIDIA/cccl/issues"
+
+[tool.scikit-build]
+minimum-version = "build-system.requires"
+build-dir = "build/{wheel_tag}"
+
+[tool.scikit-build.cmake]
+version = ">=3.30"
+args = []
+build-type = "Release"
+source-dir = "."
+
+[tool.scikit-build.ninja]
+version = ">=1.11"
+make-fallback = true
+
+[tool.scikit-build.metadata.version]
+provider = "scikit_build_core.metadata.setuptools_scm"
+
+[tool.setuptools_scm]
+root = "../.."
+git_describe_command = ["git", "describe", "--tags", "--match", "v[0-9]*"]
+
+
+[tool.scikit-build.wheel.packages]
+"cuda" = "cuda"
+
+[tool.mypy]
+cache_dir = "../../.cache/mypy"
+python_version = "3.10"
+
+[[tool.mypy.overrides]]
+module = [
+  # numpy>=2.3 stubs use PEP 695 syntax that mypy can't parse under
+  # python_version=3.10; skip following its imports to keep the 3.10 target.
+  "numpy",
+  "numpy.*",
+  "numba.*",
+  "llvmlite.*",
+  "cuda.cccl.*",
+  "cuda.bindings.*",
+  "cuda.core.*",
+  "cuda.pathfinder.*",
+]
+ignore_missing_imports = true
+follow_imports = "skip"
+
+[tool.ruff]
+extend = "../../pyproject.toml"
+
+[tool.ruff.lint.isort]
+known-first-party = [
+  "cuda.cccl",
+  "cuda.cccl.headers",
+  "cuda.compute",
+]
+
+[tool.pytest.ini_options]
+# Make shared helpers in tests/_utils importable when pytest runs outside tests/.
+pythonpath = ["tests"]
+markers = [
+  "no_verify_sass: skip SASS verification check",
+  "large: tests requiring large device memory allocations",
+  "no_numba: tests that should not import numba or numba.cuda",
+  "free_threading: tests requiring free-threaded CPython with the GIL disabled",
+  "serialization: serialize/deserialize tests (skipped on the v2 HostJIT backend)",
+]