Commit Graph

12 Commits

Author SHA1 Message Date
Adam Joseph
80472e3754 treewide: add __attrsFailEvaluation and __recurseIntoDerivationForReleaseJobs 2023-12-15 05:13:46 -08:00
Connor Baker
aaf735eac9 cudaPackages.saxpy: only available from CUDA 11.4 and on 2023-12-07 16:45:56 +00:00
Connor Baker
8e800cedaf cudaPackages: move derivations to cuda-modules & support aarch64
cudaPackages.cuda_compat: ignore missing libs provided at runtime

cudaPackages.gpus: Jetson should never build by default

cudaPackages.flags: don't build Jetson capabilities by default

cudaPackages: re-introduce filter for pre-existing CUDA redist packages in overrides

cudaPackages: only recurseIntoAttrs for the latest of each major version

cudaPackages.nvccCompatabilities: use GCC 10 through CUDA 11.5 to avoid a GLIBC incompatability

cudaPackages.cutensor: acquire libcublas through cudatoolkit prior to 11.4

cudaPackages.cuda_compat: mark as broken on aarch64-linux if not targeting Jetson

cudaPackages.cutensor_1_4: fix build

cudaPackages: adjust use of autoPatchelfIgnoreMissingDeps

cudaPackages.cuda_nvprof: remove unecessary override to add addOpenGLRunpath

cudaPackages: use getExe' to avoid patchelf warning about missing meta.mainProgram

cudaPackages: fix evaluation with Nix 2.3

cudaPackages: fix platform detection for Jetson/non-Jetson aarch64-linux

python3Packages.tensorrt: mark as broken if required packages are missing

Note: evaluating the name of the derivation will fail if tensorrt is not present,
which is why we wrap the value in `lib.optionalString`.

cudaPackages.flags.getNixSystem: add guard based on jetsonTargets

cudaPackages.cudnn: use explicit path to patchelf

cudaPackages.tensorrt: use explicit path to patchelf
2023-12-07 16:45:54 +00:00
Someone Serge
0dc161b2f8
cudaPackages_12.cutensor: init and fix 2023-12-04 20:24:32 +00:00
Connor Baker
23144ffa59
Merge pull request #240015 from jmillerpdt/init-nccl-tests
nccl-tests: init at 2.13.6
2023-08-21 10:33:19 -04:00
Artturin
b93da3f4b7 treewide: overrideScope' -> overrideScope
`lib.makeScope` `overrideScope'` has been renamed to `overrideScope`

`fd --type f | xargs sd --string-mode "overrideScope'" "overrideScope"`
2023-08-14 18:46:47 +03:00
Jason Miller
879d6402ac nccl-tests: init at 2.13.6 2023-06-26 23:07:20 +00:00
Artturin
4e3dcf364e treewide: makeSetupHook deps -> propagatedBuildInputs 2023-02-07 21:02:00 +02:00
Thomas Watson
2cec04c8a2 cudaPackages.cutensor: 1.3.1.3 -> 1.5.0.3
Update cutensor to the latest version compatible with its only dependent
in nixpkgs, cupy. Future work might wish to make multiple versions
available. Also update the URL format for all known versions.
2022-10-21 21:45:19 -05:00
Aidan Gauland
c8fba8254a
tensorrt: support multiple CUDA versions
Refactor derivation to pick the version that supports the current CUDA version.
Based on the implementation of the same concept in the cudnn derivation.
2022-07-02 14:12:41 +12:00
Aidan Gauland
d70b4df686
tensorrt: init at 8.4.0.6
Add derivation for TensorRT 8, a high-performance deep learning interface SDK
from NVIDIA, which is at this point non-redistributable.  The current version
aldo requires CUDA 11, so this is left out of the cudaPackages_10* scopes.
2022-06-24 13:02:02 +12:00
Frederik Rietdijk
1d63f89caa cudaPackages: overhaul of how we package cuda packages
There are many different versions of the `cudatoolkit` and related
cuda packages, and it can be tricky to ensure they remain compatible.

- `cudaPackages` is now a package set with `cudatoolkit`, `cudnn`, `cutensor`, `nccl`, as well as `cudatoolkit` split into smaller packages ("redist");
- expressions should now use `cudaPackages` as parameter instead of the individual cuda packages;
- `makeScope` is now used, so it is possible to use `.overrideScope'` to set e.g. a different `cudnn` version;
- `release-cuda.nix` is introduced to easily evaluate cuda packages using hydra.
2022-04-09 08:50:22 +02:00