cudaPackages already builts with structuredAttrs, but the cmakeFlags+=
pattern incorrectly appends the additional flags to the first array
argument with a space - which is now part of that argument itself since
#318614, which added support for structuredAttrs to cmake.
On NixOS, the the return status is that of the last command executed within the function or script.
When we're doing tests immediately before the return, the value you end up returning might not be what you want.
The setupCudaHook always checks for the existence of `$prefix/nix-support/include-in-cudatoolkit-root`, so we need to be sure
it always exists. It isn't populated when `strictDeps` is set, but it must exist.
This PR refactor CUDA setup hooks, and in particular
autoAddOpenGLRunpath and autoAddCudaCompatRunpathHook, that were using a
lot of code in common (in fact, I introduced the latter by copy pasting
most of the bash script of the former). This is not satisfying for
maintenance, as a recent patch showed, because we need to duplicate
changes to both hooks.
This commit abstract the common part in a single shell script that
applies a generic patch action to every elf file in the output. For
autoAddOpenGLRunpath the action is just addOpenGLRunpath (now
addDriverRunpath), and is few line function for
autoAddCudaCompatRunpathHook.
Doing so, we also takes the occasion to use the newer addDriverRunpath
instead of the previous addOpenGLRunpath, and rename the CUDA hook to
reflect that as well.
Co-Authored-By: Connor Baker <connor.baker@tweag.io>
Some nvidia devices, such as the Jetson family, support the Nvidia compatibility package (nvidia_compat) which allows to run executables built against a higher CUDA major version on a system with an older CUDA driver. On such platforms, the consensus among CUDA maintainers is that there is no downside in always enabling it by default.
This commit links to the relevant cuda_compat shared libraries by patching the CUDA core packages' runpaths when cuda_compat is available, in the same way as we do for OpenGL drivers currently.
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
