Merge branch 'main' into copilot/fix-756

leofang · web-flow · commit e907c78b0461 · 2025-08-22T16:59:42.000-04:00
diff --git a/cuda_bindings/docs/source/install.md b/cuda_bindings/docs/source/install.md
@@ -10,7 +10,7 @@
 * Optionally, NVRTC, nvJitLink, NVVM, and cuFile from CUDA Toolkit 13.x
 
 ```{note}
-The optional CUDA Toolkit components can be installed via PyPI, Conda, OS-specific package managers, or local installers (as described in the CUDA Toolkit [Windows](https://docs.nvidia.com/cuda/cuda-installation-guide-microsoft-windows/index.html) and [Linux](https://docs.nvidia.com/cuda/cuda-installation-guide-linux/index.html) Installation Guides).
+The optional CUDA Toolkit components are now installed via the `cuda-toolkit` metapackage from PyPI for improved dependency resolution. Components can also be installed via Conda, OS-specific package managers, or local installers (as described in the CUDA Toolkit [Windows](https://docs.nvidia.com/cuda/cuda-installation-guide-microsoft-windows/index.html) and [Linux](https://docs.nvidia.com/cuda/cuda-installation-guide-linux/index.html) Installation Guides).
 ```
 
 Starting from v12.8.0, `cuda-python` becomes a meta package which currently depends only on `cuda-bindings`; in the future more sub-packages will be added to `cuda-python`. In the instructions below, we still use `cuda-python` as example to serve existing users, but everything is applicable to `cuda-bindings` as well.
@@ -27,12 +27,14 @@ Install all optional dependencies with:
 pip install -U cuda-python[all]
 ```
 
-Where the optional dependencies are:
+Where the optional dependencies include:
 
-* nvidia-cuda-nvrtc (Provides NVRTC shared library)
-* nvidia-nvjitlink (Provides nvJitLink shared library)
-* nvidia-cuda-nvcc (Provides NVVM shared library)
-* nvidia-cufile (Provides cuFile shared library)
+* `nvidia-cuda-nvrtc` (NVRTC runtime compilation library)  
+* `nvidia-nvjitlink` (nvJitLink library)
+* `nvidia-nvvm` (NVVM library)
+* `nvidia-cufile` (cuFile library, Linux only)
+
+These are now installed through the `cuda-toolkit` metapackage for improved dependency resolution.
 
 
 ## Installing from Conda
@@ -41,6 +43,15 @@ Where the optional dependencies are:
 $ conda install -c conda-forge cuda-python
 ```
 
+```{note}
+When using conda, the `cuda-version` metapackage can be used to control the versions of CUDA Toolkit components that are installed to the conda environment.
+```
+
+For example:
+```console
+$ conda install -c conda-forge cuda-python cuda-version=13
+```
+
 
 ## Installing from Source
 
diff --git a/cuda_bindings/docs/source/release/13.X.Y-notes.rst b/cuda_bindings/docs/source/release/13.X.Y-notes.rst
@@ -12,10 +12,12 @@ Released on TBD
 Highlights
 ----------
 
+* Migrated wheel dependencies from individual NVIDIA packages to the ``cuda-toolkit`` metapackage for improved dependency resolution and version constraints.
 * Automatic CUDA library path detection based on ``CUDA_HOME``, eliminating the need to manually set ``LIBRARY_PATH`` environment variables for installation.
+* The ``[all]`` optional dependencies now use ``cuda-toolkit`` with appropriate extras instead of individual packages. The NVCC compiler is no longer automatically installed with ``pip install cuda-python[all]`` as it was previously included only to access the NVVM library, which now has its own dedicated wheel. Users who need the NVCC compiler should explicitly install it with ``pip install cuda-toolkit[nvcc]==X.Y`` with the appropriate version for their needs.
 
 
 Known issues
 ------------
 
-* Updating from older versions (v12.6.2.post1 and below) via ``pip install -U cuda-python`` might not work. Please do a clean re-installation by uninstalling ``pip uninstall -y cuda-python`` followed by installing ``pip install cuda-python``.
+* Updating from older versions (v12.6.2.post1 and below) via ``pip install -U cuda-python`` might not work. Please do a clean re-installation by uninstalling ``pip uninstall -y cuda-python`` followed by installing ``pip install cuda-python``.
diff --git a/cuda_bindings/pyproject.toml b/cuda_bindings/pyproject.toml
@@ -33,11 +33,8 @@ dependencies = [
 
 [project.optional-dependencies]
 all = [
-    "nvidia-cuda-nvcc~=13.0",
-    "nvidia-cuda-nvrtc~=13.0",
-    "nvidia-nvjitlink~=13.0",
-    "nvidia-nvvm~=13.0",
-    "nvidia-cufile; sys_platform == 'linux'",
+    "cuda-toolkit[nvrtc,nvjitlink,nvvm]==13.*",
+    "cuda-toolkit[cufile]==13.*; sys_platform == 'linux'",
 ]
 
 test = [
diff --git a/cuda_core/cuda/core/experimental/_linker.py b/cuda_core/cuda/core/experimental/_linker.py
@@ -205,15 +205,15 @@ def _init_nvjitlink(self):
             self.formatted_options.append("-time")
         if self.verbose is not None:
             self.formatted_options.append("-verbose")
-        if self.link_time_optimization is not None:
+        if self.link_time_optimization is not None and self.link_time_optimization:
             self.formatted_options.append("-lto")
         if self.ptx is not None:
             self.formatted_options.append("-ptx")
         if self.optimization_level is not None:
             self.formatted_options.append(f"-O{self.optimization_level}")
-        if self.debug is not None:
+        if self.debug is not None and self.debug:
             self.formatted_options.append("-g")
-        if self.lineinfo is not None:
+        if self.lineinfo is not None and self.lineinfo:
             self.formatted_options.append("-lineinfo")
         if self.ftz is not None:
             self.formatted_options.append(f"-ftz={'true' if self.ftz else 'false'}")
diff --git a/cuda_core/cuda/core/experimental/_program.py b/cuda_core/cuda/core/experimental/_program.py
@@ -244,8 +244,8 @@ def __post_init__(self):
             self._formatted_options.append("--device-debug")
         if self.lineinfo is not None and self.lineinfo:
             self._formatted_options.append("--generate-line-info")
-        if self.device_code_optimize is not None:
-            self._formatted_options.append(f"--dopt={'on' if self.device_code_optimize else 'off'}")
+        if self.device_code_optimize is not None and self.device_code_optimize:
+            self._formatted_options.append("--dopt=on")
         if self.ptxas_options is not None:
             opt_name = "--ptxas-options"
             if isinstance(self.ptxas_options, str):
@@ -351,7 +351,7 @@ def _as_bytes(self):
 
     def __repr__(self):
         # __TODO__ improve this
-        return self._formatted_options
+        return str(self._formatted_options)
 
 
 ProgramHandleT = Union["cuda.bindings.nvrtc.nvrtcProgram", LinkerHandleT]
diff --git a/cuda_core/docs/source/install.md b/cuda_core/docs/source/install.md
@@ -17,11 +17,11 @@ dependencies are as follows:
 
 ## Installing from PyPI
 
-`cuda.core` works with `cuda.bindings` (part of `cuda-python`) 11 or 12. For example with CUDA 12:
+`cuda.core` works with `cuda.bindings` (part of `cuda-python`) 11 or 12. Test dependencies now use the ``cuda-toolkit`` metapackage for improved dependency resolution. For example with CUDA 12:
 ```console
 $ pip install cuda-core[cu12]
 ```
-and likewise use `[cu11]` for CUDA 11.
+and likewise use `[cu11]` for CUDA 11, or `[cu13]` for CUDA 13.
 
 Note that using `cuda.core` with NVRTC installed from PyPI via `pip install` requires
 `cuda.bindings` 12.8.0+ or 11.8.6+. Likewise, with nvJitLink it requires 12.8.0+.
diff --git a/cuda_core/docs/source/release/0.X.Y-notes.rst b/cuda_core/docs/source/release/0.X.Y-notes.rst
@@ -26,6 +26,7 @@ New features
 ------------
 
 - Added :attr:`Device.arch` property that returns the compute capability as a string (e.g., '75' for CC 7.5), providing a convenient alternative to manually concatenating the compute capability tuple.
+- CUDA 13.x testing support through new ``test-cu13`` dependency group.
 
 
 New examples
diff --git a/cuda_core/pyproject.toml b/cuda_core/pyproject.toml
@@ -50,9 +50,9 @@ cu11 = ["cuda-bindings[all]==11.8.*"]
 cu12 = ["cuda-bindings[all]==12.*"]
 cu13 = ["cuda-bindings[all]==13.*"]
 test = ["cython>=3.0", "setuptools", "pytest>=6.2.4"]
-test-cu11 = ["cuda-core[test]", "cupy-cuda11x", "nvidia-cuda-runtime-cu11"]  # runtime headers needed by CuPy
-test-cu12 = ["cuda-core[test]", "cupy-cuda12x", "nvidia-cuda-runtime-cu12"]  # runtime headers needed by CuPy
-# TODO add test-cu13 once CuPy is ready
+test-cu11 = ["cuda-core[test]", "cupy-cuda11x", "cuda-toolkit[cudart]==11.*"]  # runtime headers needed by CuPy
+test-cu12 = ["cuda-core[test]", "cupy-cuda12x", "cuda-toolkit[cudart]==12.*"]  # runtime headers needed by CuPy
+test-cu13 = ["cuda-core[test]", "cupy-cuda13x", "cuda-toolkit[cudart]==13.*"]  # runtime headers needed by CuPy
 
 [project.urls]
 homepage = "https://nvidia.github.io/cuda-python/"
diff --git a/cuda_pathfinder/pyproject.toml b/cuda_pathfinder/pyproject.toml
@@ -15,35 +15,12 @@ test = [
     "pytest>=6.2.4",
 ]
 nvidia_wheels_cu12 = [
-    "nvidia-cublas-cu12",
-    "nvidia-cuda-nvcc-cu12",
-    "nvidia-cuda-nvrtc-cu12",
-    "nvidia-cuda-runtime-cu12",
-    "nvidia-cufft-cu12",
-    "nvidia-cufile-cu12; sys_platform != 'win32'",
-    "nvidia-curand-cu12",
-    "nvidia-cusolver-cu12",
-    "nvidia-cusparse-cu12",
-    "nvidia-npp-cu12",
-    "nvidia-nvfatbin-cu12",
-    "nvidia-nvjitlink-cu12",
-    "nvidia-nvjpeg-cu12",
+    "cuda-toolkit[nvcc,cublas,nvrtc,cudart,cufft,curand,cusolver,cusparse,npp,nvfatbin,nvjitlink,nvjpeg]==12.*",
+    "cuda-toolkit[cufile]==12.*; sys_platform != 'win32'",
 ]
 nvidia_wheels_cu13 = [
-    "nvidia-cublas",
-    "nvidia-cuda-nvcc",
-    "nvidia-cuda-nvrtc",
-    "nvidia-cuda-runtime",
-    "nvidia-cufft",
-    "nvidia-cufile; sys_platform != 'win32'",
-    "nvidia-curand",
-    "nvidia-cusolver",
-    "nvidia-cusparse",
-    "nvidia-npp",
-    "nvidia-nvfatbin",
-    "nvidia-nvjitlink",
-    "nvidia-nvjpeg",
-    "nvidia-nvvm",
+    "cuda-toolkit[nvcc,cublas,nvrtc,cudart,cufft,curand,cusolver,cusparse,npp,nvfatbin,nvjitlink,nvjpeg,nvvm]==13.*",
+    "cuda-toolkit[cufile]==13.*; sys_platform != 'win32'",
 ]
 
 [project.urls]
