Reduce wheel size by excluding source files (#1588)

Use opt-in package-data approach to control what goes into wheels:
- Set include-package-data = false to prevent MANIFEST.in auto-inclusion
- Explicitly include only .pxd files (for downstream cimport) and headers

cuda_core wheel size reduced from ~5 MB to ~1.6 MB (68% reduction).

Also fix MANIFEST.in for both packages to include all files needed
for sdist builds (headers, template files).

Co-authored-by: Cursor <cursoragent@cursor.com>

Author: Andy-Jost

cuda_bindings/MANIFEST.in

@@ -1,7 +1,7 @@
 # SPDX-FileCopyrightText: Copyright (c) 2024-2025 NVIDIA CORPORATION & AFFILIATES. All rights reserved.
 # SPDX-License-Identifier: LicenseRef-NVIDIA-SOFTWARE-LICENSE
 
-recursive-include cuda/ *.pyx *.pxd *.pxi
+recursive-include cuda/ *.pyx *.pxd *.pxi *.pyx.in *.pxd.in *.pxi.in *.h
 # at least with setuptools 75.0.0 this folder was added erroneously
 # to the payload, causing file copying to the build environment failed
 exclude cuda/bindings cuda?bindings

cuda_bindings/docs/source/release/13.1.2-notes.rst

@@ -1,4 +1,4 @@
-.. SPDX-FileCopyrightText: Copyright (c) 2025 NVIDIA CORPORATION & AFFILIATES. All rights reserved.
+.. SPDX-FileCopyrightText: Copyright (c) 2025-2026 NVIDIA CORPORATION & AFFILIATES. All rights reserved.
 .. SPDX-License-Identifier: LicenseRef-NVIDIA-SOFTWARE-LICENSE
 
 .. module:: cuda.bindings
@@ -22,6 +22,13 @@ Bugfixes
   (`PR #1336 <https://github.com/NVIDIA/cuda-python/pull/1336>`_)
 * Fixed ABI incompatibility bugs in cuFILE bindings introduced in v13.1.0.
 
+Miscellaneous
+-------------
+
+* Wheel and installed package sizes significantly reduced (e.g., on a typical Linux x86_64
+  build, wheel from ~16.6 MB to ~5.7 MB and installed from ~152 MB to ~23 MB) by excluding
+  Cython source files, generated C++ files, and template files from distribution packages.
+
 Known issues
 ------------
 

cuda_bindings/pyproject.toml

@@ -56,8 +56,13 @@ Documentation = "https://nvidia.github.io/cuda-python/"
 [tool.setuptools.packages.find]
 include = ["cuda*"]
 
-[tool.setuptools.exclude-package-data]
-"*" = ["*.cpp"]
+[tool.setuptools]
+include-package-data = false
+
+[tool.setuptools.package-data]
+"*" = ["*.pxd", "*.pxi"]
+"cuda.bindings._bindings" = ["*.h"]
+"cuda.bindings._lib" = ["*.h"]
 
 [tool.setuptools.dynamic]
 readme = { file = ["DESCRIPTION.rst"], content-type = "text/x-rst" }

cuda_core/MANIFEST.in

@@ -4,3 +4,4 @@
 
 recursive-include cuda/core *.pyx *.pxd *.pxi
 recursive-include cuda/core/_cpp *.cpp *.hpp
+recursive-include cuda/core/_include *.h *.hpp

cuda_core/docs/source/release/0.6.x-notes.rst

@@ -36,4 +36,7 @@ None.
 Fixes and enhancements
 ----------------------
 
-None.
+- Wheel and installed package sizes significantly reduced (e.g., on a typical Linux x86_64
+  build, wheel from ~4.6 MB to ~1.6 MB and installed from ~26 MB to ~4.4 MB) by excluding
+  Cython source files, generated C++ files, and other build artifacts from distribution
+  packages.

cuda_core/pyproject.toml

@@ -72,9 +72,13 @@ issues = "https://github.com/NVIDIA/cuda-python/issues/"
 [tool.setuptools.packages.find]
 include = ["cuda.core*"]
 
+[tool.setuptools]
+include-package-data = false
+
 [tool.setuptools.package-data]
-"cuda.core._include" = ["*.h", "*.hpp", "*.cuh"]
-"cuda.core._cpp" = ["*.cpp", "*.hpp"]
+"*" = ["*.pxd"]
+"cuda.core._include" = ["*.h", "*.hpp"]
+"cuda.core._cpp" = ["*.hpp"]
 
 [tool.setuptools.dynamic]
 readme = { file = ["DESCRIPTION.rst"], content-type = "text/x-rst" }