nixpkgs/doc/languages-frameworks/cuda.section.md
Connor Baker d5e5246e76 cudaPackages: split outputs
This change which involves creating multiple outputs for CUDA
redistributable packages.

We use a script to find out, ahead of time, the outputs each redist
package provides. From that, we are able to create multiple outputs for
supported redist packages, allowing users to specify exactly which
components they require.

Beyond the script which finds outputs ahead of time, there is some custom
code involved in making this happen. For example, the way Nixpkgs
typically handles multiple outputs involves making `dev` the default
output when available, and adding `out` to `dev`'s
`propagatedBuildInputs`.

Instead, we make each output independent of the others. If a user wants
only to include the headers found in a redist package, they can do so by
choosing the `dev` output. If they want to include dynamic libraries,
they can do so by specifying the `lib` output, or `static` for static
libraries.

To avoid breakages, we continue to provide the `out` output, which
becomes the union of all other outputs, effectively making the split
outputs opt-in.
2023-08-31 03:31:55 +00:00

7.3 KiB

CUDA

CUDA-only packages are stored in the cudaPackages packages set. This set includes the cudatoolkit, portions of the toolkit in separate derivations, cudnn, cutensor and nccl.

A package set is available for each CUDA version, so for example cudaPackages_11_6. Within each set is a matching version of the above listed packages. Additionally, other versions of the packages that are packaged and compatible are available as well. For example, there can be a cudaPackages.cudnn_8_3 package.

To use one or more CUDA packages in an expression, give the expression a cudaPackages parameter, and in case CUDA is optional

{ config
, cudaSupport ? config.cudaSupport
, cudaPackages ? { }
, ...
}:

When using callPackage, you can choose to pass in a different variant, e.g. when a different version of the toolkit suffices

mypkg = callPackage { cudaPackages = cudaPackages_11_5; }

If another version of say cudnn or cutensor is needed, you can override the package set to make it the default. This guarantees you get a consistent package set.

mypkg = let
  cudaPackages = cudaPackages_11_5.overrideScope (final: prev: {
    cudnn = prev.cudnn_8_3;
  }});
in callPackage { inherit cudaPackages; };

The CUDA NVCC compiler requires flags to determine which hardware you want to target for in terms of SASS (real hardware) or PTX (JIT kernels).

Nixpkgs tries to target support real architecture defaults based on the CUDA toolkit version with PTX support for future hardware. Experienced users may optimize this configuration for a variety of reasons such as reducing binary size and compile time, supporting legacy hardware, or optimizing for specific hardware.

You may provide capabilities to add support or reduce binary size through config using cudaCapabilities = [ "6.0" "7.0" ]; and cudaForwardCompat = true; if you want PTX support for future hardware.

Please consult GPUs supported for your specific card(s).

Library maintainers should consult NVCC Docs and release notes for their software package.

Adding a new CUDA release

WARNING

This section of the docs is still very much in progress. Feedback is welcome in GitHub Issues tagging @NixOS/cuda-maintainers or on Matrix.

The CUDA Toolkit is a suite of CUDA libraries and software meant to provide a development environment for CUDA-accelerated applications. Until the release of CUDA 11.4, NVIDIA had only made the CUDA Toolkit available as a multi-gigabyte runfile installer, which we provide through the cudaPackages.cudatoolkit attribute. From CUDA 11.4 and onwards, NVIDIA has also provided CUDA redistributables (“CUDA-redist”): individually packaged CUDA Toolkit components meant to facilitate redistribution and inclusion in downstream projects. These packages are available in the cudaPackages package set.

All new projects should use the CUDA redistributables available in cudaPackages in place of cudaPackages.cudatoolkit, as they are much easier to maintain and update.

Updating CUDA redistributables

  1. Go to NVIDIA's index of CUDA redistributables: https://developer.download.nvidia.com/compute/cuda/redist/

  2. Copy the redistrib_*.json corresponding to the release to pkgs/development/compilers/cudatoolkit/redist/manifests.

  3. Generate the redistrib_features_*.json file by running:

    nix run github:ConnorBaker/cuda-redist-find-features -- <path to manifest>
    

    That command will generate the redistrib_features_*.json file in the same directory as the manifest.

  4. Include the path to the new manifest in pkgs/development/compilers/cudatoolkit/redist/extension.nix.

Updating the CUDA Toolkit runfile installer

WARNING

While the CUDA Toolkit runfile installer is still available in Nixpkgs as the cudaPackages.cudatoolkit attribute, its use is not recommended and should it be considered deprecated. Please migrate to the CUDA redistributables provided by the cudaPackages package set.

To ensure packages relying on the CUDA Toolkit runfile installer continue to build, it will continue to be updated until a migration path is available.

  1. Go to NVIDIA's CUDA Toolkit runfile installer download page: https://developer.nvidia.com/cuda-downloads

  2. Select the appropriate OS, architecture, distribution, and version, and installer type.

    • For example: Linux, x86_64, Ubuntu, 22.04, runfile (local)
    • NOTE: Typically, we use the Ubuntu runfile. It is unclear if the runfile for other distributions will work.
  3. Take the link provided by the installer instructions on the webpage after selecting the installer type and get its hash by running:

    nix store prefetch-file --hash-type sha256 <link>
    
  4. Update pkgs/development/compilers/cudatoolkit/versions.toml to include the release.

Updating the CUDA package set

  1. Include a new cudaPackages_<major>_<minor> package set in pkgs/top-level/all-packages.nix.

    • NOTE: Changing the default CUDA package set should occur in a separate PR, allowing time for additional testing.
  2. Successfully build the closure of the new package set, updating pkgs/development/compilers/cudatoolkit/redist/overrides.nix as needed. Below are some common failures:

Unable to ... During ... Reason Solution Note
Find headers configurePhase or buildPhase Missing dependency on a dev output Add the missing dependency The dev output typically contain the headers
Find libraries configurePhase Missing dependency on a dev output Add the missing dependency The dev output typically contain CMake configuration files
Find libraries buildPhase or patchelf Missing dependency on a lib or static output Add the missing dependency The lib or static output typically contain the libraries

In the scenario you are unable to run the resulting binary: this is arguably the most complicated as it could be any combination of the previous reasons. This type of failure typically occurs when a library attempts to load or open a library it depends on that it does not declare in its DT_NEEDED section. As a first step, ensure that dependencies are patched with cudaPackages.autoAddOpenGLRunpath. Failing that, try running the application with nixGL or a similar wrapper tool. If that works, it likely means that the application is attempting to load a library that is not in the RPATH or RUNPATH of the binary.