cudatoolkit: reintroduce version 11.7.0 to master by dguibert · Pull Request #185557 · NixOS/nixpkgs

dguibert · 2022-08-07T14:11:58Z

Description of changes

This patch fixes #179912 which has been reverted on master due to https://gist.github.com/GrahamcOfBorg/45ac7f5bc9e02a74cb1e4264f365417f

Things done

pkgs/development/compilers/cudatoolkit/redist/overrides.nix

pkgs/test/cuda/cuda-samples/extension.nix

samuela

Thanks for resubmitting this @dguibert and apologies that this has been more involved than expected!

I think these changes look good, but I'd actually like us to go a bit further and de-couple TensorRT and cudatoolkit evaluation! It's totally fine for an upgrade to one of the packages to end up marking other derivations as broken, but breaking evaluation is a huge problem. Having to constantly keep the two in sync is an unnecessary burden on both packages and really came back to bite us in #179912.

Could we refactor pkgs/development/libraries/science/math/tensorrt/extension.nix a bit such that upgrading cudatoolkit can be done without breaking the evaluation of tensorrt?

pkgs/development/libraries/science/math/cudnn/extension.nix

aidalgol · 2022-09-25T20:39:42Z

Could we refactor pkgs/development/libraries/science/math/tensorrt/extension.nix a bit such that upgrading cudatoolkit can be done without breaking the evaluation of tensorrt?

@samuela Sorry, I only just saw this PR. Since I introduced TensorRT to nixpkgs, I feel I should take this on. Could you elaborate a bit on what exactly you're proposing?

samuela · 2022-09-26T00:51:50Z

Hi @aidalgol, I think you nailed it in #192958. IIRC the issue was that TensorRT's evaluation would break if cudaVersion was upgraded without editing TensorRT's derivation. But it looks like #192958 should fix that!

samuela · 2022-09-26T05:21:27Z

@dguibert If you rebase off of #192958, then I think this should be good to go. Note that adding 11.7 to the supported versions of cudnn and TensorRT is still necessary in order to prevent those packages from being marked meta.broken but we shouldn't have any evaluation issues from here on out (thanks @aidalgol!!)

This reverts commit 90c5057.

dguibert · 2022-10-04T15:24:53Z

Hi @samuela and @SuperSandro2000,
I took time to rebase this PR and apply your suggestions ;-)

samuela

Nice! I think this looks good overall. Running nixpkgs-review now...

pkgs/development/compilers/cudatoolkit/redist/overrides.nix

samuela · 2022-10-05T04:17:43Z

EDIT: this run is a little wonky since NIXPKGS_ALLOW_BROKEN was not set, running a new nixpkgs-review now.

Result of nixpkgs-review pr 185557 run on x86_64-linux 1

4 packages marked as broken and skipped:

cudaPackages.nvidia_driver
python310Packages.cupy
python39Packages.cupy
truecrack-cuda

9 packages failed to build:

cudaPackages.tensorrt (cudaPackages.tensorrt_8_4_0)
ethminer (ethminer-cuda)
mathematica-cuda
python310Packages.numbaWithCuda
python310Packages.pyrealsense2WithCuda
python310Packages.tensorrt
python39Packages.numbaWithCuda
python39Packages.pyrealsense2WithCuda
python39Packages.tensorrt

59 packages built:

colmapWithCuda
cudaPackages.cuda_cccl
cudaPackages.cuda_cudart
cudaPackages.cuda_cuobjdump
cudaPackages.cuda_cupti
cudaPackages.cuda_cuxxfilt
cudaPackages.cuda_demo_suite
cudaPackages.cuda_documentation
cudaPackages.cuda_gdb
cudaPackages.cuda_memcheck
cudaPackages.cuda_nsight
cudaPackages.cuda_nvcc
cudaPackages.cuda_nvdisasm
cudaPackages.cuda_nvml_dev
cudaPackages.cuda_nvprof
cudaPackages.cuda_nvprune
cudaPackages.cuda_nvrtc
cudaPackages.cuda_nvtx
cudaPackages.cuda_nvvp
cudaPackages.cuda_sanitizer_api
cudatoolkit (cudaPackages.cudatoolkit ,cudatoolkit_11)
cudaPackages.cudnn (cudaPackages.cudnn_8_4_0)
cudaPackages.cudnn_8_3_2
cudaPackages.cutensor
cudaPackages.fabricmanager
cudaPackages.libcublas
cudaPackages.libcufft
cudaPackages.libcufile
cudaPackages.libcurand
cudaPackages.libcusolver
cudaPackages.libcusparse
cudaPackages.libnpp
cudaPackages.libnvidia_nscq
cudaPackages.libnvjpeg
cudaPackages.nccl
cudaPackages.nsight_compute
cudaPackages.nsight_systems
cudaPackages.nvidia_fs
forge
gpu-burn
gromacsCudaMpi
gwe
katagoWithCuda
librealsenseWithCuda
magma
nvtop
nvtop-nvidia
python310Packages.TheanoWithCuda
python310Packages.pycuda
python310Packages.pynvml
python310Packages.tensorflowWithCuda
python310Packages.torchWithCuda
python39Packages.TheanoWithCuda
python39Packages.pycuda
python39Packages.pynvml
python39Packages.tensorflowWithCuda
python39Packages.torchWithCuda
xgboostWithCuda
xpraWithNvenc

samuela · 2022-10-05T05:31:45Z

Result of nixpkgs-review pr 185557 run on x86_64-linux 1

4 packages marked as broken and skipped:

cudaPackages.nvidia_driver
python310Packages.cupy
python39Packages.cupy
truecrack-cuda

9 packages failed to build:

cudaPackages.tensorrt (cudaPackages.tensorrt_8_4_0)
ethminer (ethminer-cuda)
mathematica-cuda
python310Packages.numbaWithCuda
python310Packages.pyrealsense2WithCuda
python310Packages.tensorrt
python39Packages.numbaWithCuda
python39Packages.pyrealsense2WithCuda
python39Packages.tensorrt

59 packages built:

colmapWithCuda
cudaPackages.cuda_cccl
cudaPackages.cuda_cudart
cudaPackages.cuda_cuobjdump
cudaPackages.cuda_cupti
cudaPackages.cuda_cuxxfilt
cudaPackages.cuda_demo_suite
cudaPackages.cuda_documentation
cudaPackages.cuda_gdb
cudaPackages.cuda_memcheck
cudaPackages.cuda_nsight
cudaPackages.cuda_nvcc
cudaPackages.cuda_nvdisasm
cudaPackages.cuda_nvml_dev
cudaPackages.cuda_nvprof
cudaPackages.cuda_nvprune
cudaPackages.cuda_nvrtc
cudaPackages.cuda_nvtx
cudaPackages.cuda_nvvp
cudaPackages.cuda_sanitizer_api
cudatoolkit (cudaPackages.cudatoolkit ,cudatoolkit_11)
cudaPackages.cudnn (cudaPackages.cudnn_8_4_0)
cudaPackages.cudnn_8_3_2
cudaPackages.cutensor
cudaPackages.fabricmanager
cudaPackages.libcublas
cudaPackages.libcufft
cudaPackages.libcufile
cudaPackages.libcurand
cudaPackages.libcusolver
cudaPackages.libcusparse
cudaPackages.libnpp
cudaPackages.libnvidia_nscq
cudaPackages.libnvjpeg
cudaPackages.nccl
cudaPackages.nsight_compute
cudaPackages.nsight_systems
cudaPackages.nvidia_fs
forge
gpu-burn
gromacsCudaMpi
gwe
katagoWithCuda
librealsenseWithCuda
magma
nvtop
nvtop-nvidia
python310Packages.TheanoWithCuda
python310Packages.pycuda
python310Packages.pynvml
python310Packages.tensorflowWithCuda
python310Packages.torchWithCuda
python39Packages.TheanoWithCuda
python39Packages.pycuda
python39Packages.pynvml
python39Packages.tensorflowWithCuda
python39Packages.torchWithCuda
xgboostWithCuda
xpraWithNvenc

samuela · 2022-10-05T05:44:50Z

Here are the failure logs:

error: builder for '/nix/store/ybsmhdrmjy8xarlvf53kzlj4bbfw7b20-Mathematica_13.1.0_BNDL_LINUX.sh.drv' failed with exit code 1;
       last 8 log lines:
       >
       > ***
       > This nix expression requires that Mathematica_13.1.0_BNDL_LINUX.sh is
       > already part of the store. Find the file on your Mathematica CD
       > and add it to the nix store with nix-store --add-fixed sha256 <FILE>.
       >
       > ***
       >
       For full logs, run 'nix log /nix/store/ybsmhdrmjy8xarlvf53kzlj4bbfw7b20-Mathematica_13.1.0_BNDL_LINUX.sh.drv'.
error: 1 dependencies of derivation '/nix/store/nas8dqxwnya0h3bc5ddnygbzhi1ydjr6-mathematica-cuda-13.1.0.drv' failed to build
error: builder for '/nix/store/yg5akr0m7llg7d4ngjwd9i96jw6sz1qw-TensorRT-8.4.0.6.Linux.x86_64-gnu.cuda-11.6.cudnn8.3.tar.gz.drv' failed with exit code 1;
       last 10 log lines:
       > download the 8.4.0.6 Linux x86_64 TAR package for CUDA 11.7 from
       > https://developer.nvidia.com/tensorrt.
       >
       > Once you have downloaded the file, add it to the store with the following
       > command, and try building this derivation again.
       >
       > $ nix-store --add-fixed sha256 TensorRT-8.4.0.6.Linux.x86_64-gnu.cuda-11.6.cudnn8.3.tar.gz
       >
       > ***
       >
       For full logs, run 'nix log /nix/store/yg5akr0m7llg7d4ngjwd9i96jw6sz1qw-TensorRT-8.4.0.6.Linux.x86_64-gnu.cuda-11.6.cudnn8.3.tar.gz.drv'.
error: 1 dependencies of derivation '/nix/store/9lqc02hrkn62fv9nsd6iw81kdyy55qmj-cudatoolkit-11.7-tensorrt-8.4.0.6.drv' failed to build
error: 2 dependencies of derivation '/nix/store/1qlmd5zrk7l2bm77r0p36qg76mnygcr2-python3.10-tensorrt-8.4.0.6.drv' failed to build
error: 2 dependencies of derivation '/nix/store/7lim89bfbd19k8rbgjcpkkkdh4cysh1r-python3.9-tensorrt-8.4.0.6.drv' failed to build
error: builder for '/nix/store/052hqk4l07mmx9vzyh0zx6ppmgr99iw0-python3.10-numba-0.56.2.drv' failed with exit code 1;
       last 10 log lines:
       > unpacking source archive /nix/store/s69sx2va20j7casl68dsf0b2xjjv0ic2-numba-0.56.2.tar.gz
       > source root is numba-0.56.2
       > setting SOURCE_DATE_EPOCH to timestamp 1662090503 of file numba-0.56.2/numba/_version.py
       > patching sources
       > applying patch /nix/store/hmp6p98pg3gkjcr9nvb5a825dhcch41n-cuda_path.patch
       > patching file numba/cuda/cuda_paths.py
       > Hunk #2 FAILED at 32.
       > Hunk #3 succeeded at 61 with fuzz 2 (offset -10 lines).
       > Hunk #4 succeeded at 77 (offset -10 lines).
       > 1 out of 4 hunks FAILED -- saving rejects to file numba/cuda/cuda_paths.py.rej
       For full logs, run 'nix log /nix/store/052hqk4l07mmx9vzyh0zx6ppmgr99iw0-python3.10-numba-0.56.2.drv'.
error: builder for '/nix/store/h9q30xk65f89k4lnw68gdm3zffkka41g-python3.9-numba-0.56.2.drv' failed with exit code 1;
       last 10 log lines:
       > unpacking source archive /nix/store/s69sx2va20j7casl68dsf0b2xjjv0ic2-numba-0.56.2.tar.gz
       > source root is numba-0.56.2
       > setting SOURCE_DATE_EPOCH to timestamp 1662090503 of file numba-0.56.2/numba/_version.py
       > patching sources
       > applying patch /nix/store/hmp6p98pg3gkjcr9nvb5a825dhcch41n-cuda_path.patch
       > patching file numba/cuda/cuda_paths.py
       > Hunk #2 FAILED at 32.
       > Hunk #3 succeeded at 61 with fuzz 2 (offset -10 lines).
       > Hunk #4 succeeded at 77 (offset -10 lines).
       > 1 out of 4 hunks FAILED -- saving rejects to file numba/cuda/cuda_paths.py.rej
       For full logs, run 'nix log /nix/store/h9q30xk65f89k4lnw68gdm3zffkka41g-python3.9-numba-0.56.2.drv'.
error: builder for '/nix/store/ylafyjflmppbnxq5zjypqhxpdyg39ciz-ethminer-0.19.0.drv' failed with exit code 2;
       last 10 log lines:
       > /nix/store/5i10bszy2380gn0z1a6d4gf4pyhg51ps-cli11-2.2.0/include/CLI/App.hpp:594:35: note:   no known conversion for argument 2 from 'unsigned int' to 'CLI::callback_t' {aka 'std::function<bool(const std::vector<std::__cxx11::basic_string<char> >&)>'}
       >   594 |                        callback_t option_callback,
       >       |                        ~~~~~~~~~~~^~~~~~~~~~~~~~~
       > /nix/store/5i10bszy2380gn0z1a6d4gf4pyhg51ps-cli11-2.2.0/include/CLI/App.hpp:701:13: note: candidate: 'CLI::Option* CLI::App::add_option(std::string)'
       >   701 |     Option *add_option(std::string option_name) {
       >       |             ^~~~~~~~~~
       > /nix/store/5i10bszy2380gn0z1a6d4gf4pyhg51ps-cli11-2.2.0/include/CLI/App.hpp:701:13: note:   candidate expects 1 argument, 4 provided
       > make[2]: *** [ethminer/CMakeFiles/ethminer.dir/build.make:76: ethminer/CMakeFiles/ethminer.dir/main.cpp.o] Error 1
       > make[1]: *** [CMakeFiles/Makefile2:516: ethminer/CMakeFiles/ethminer.dir/all] Error 2
       > make: *** [Makefile:156: all] Error 2
       For full logs, run 'nix log /nix/store/ylafyjflmppbnxq5zjypqhxpdyg39ciz-ethminer-0.19.0.drv'.
error: builder for '/nix/store/bam31rgddbrr4k4awcv30f2y8yvskiz6-librealsense-2.45.0.drv' failed with exit code 1;
       last 10 log lines:
       >    84 |             self.deleter = [](void* ptr){ delete[] ptr; };
       >       |                                                    ^~~
       > /build/source/wrappers/python/pyrs_internal.cpp: In lambda function:
       > /build/source/wrappers/python/pyrs_internal.cpp:100:52: warning: deleting 'void*' is undefined [-Wdelete-incomplete]
       >   100 |             self.deleter = [](void* ptr){ delete[] ptr; };
       >       |                                                    ^~~
       > [182/219] Building CXX object wrappers/python/CMakeFiles/pyrealsense2.dir/pyrs_sensor.cpp.o
       > [183/219] Building CXX object wrappers/python/CMakeFiles/pyrealsense2.dir/c_files.cpp.o
       > [184/219] Building CXX object wrappers/python/CMakeFiles/pybackend2.dir/pybackend.cpp.o
       > ninja: build stopped: subcommand failed.
       For full logs, run 'nix log /nix/store/bam31rgddbrr4k4awcv30f2y8yvskiz6-librealsense-2.45.0.drv'.
error: builder for '/nix/store/i5gfqsph1rnff5q674y7wdpf5zc1ql9w-librealsense-2.45.0.drv' failed with exit code 1;
       last 10 log lines:
       > /build/source/wrappers/python/pyrs_internal.cpp:100:52: warning: deleting 'void*' is undefined [-Wdelete-incomplete]
       >   100 |             self.deleter = [](void* ptr){ delete[] ptr; };
       >       |                                                    ^~~
       > [174/219] Building CXX object wrappers/python/CMakeFiles/pyrealsense2.dir/pyrs_frame.cpp.o
       > [175/219] Building CXX object CMakeFiles/realsense2.dir/src/rs.cpp.o
       > [176/219] Building CXX object wrappers/python/CMakeFiles/pyrealsense2.dir/pyrs_sensor.cpp.o
       > [177/219] Building CXX object wrappers/python/CMakeFiles/pyrealsense2.dir/c_files.cpp.o
       > [178/219] Building CXX object wrappers/python/CMakeFiles/pybackend2.dir/pybackend.cpp.o
       > [179/219] Building C object CMakeFiles/realsense2.dir/third-party/sqlite/sqlite3.c.o
       > ninja: build stopped: subcommand failed.
       For full logs, run 'nix log /nix/store/i5gfqsph1rnff5q674y7wdpf5zc1ql9w-librealsense-2.45.0.drv'.
error: 9 dependencies of derivation '/nix/store/pa79sf239g5x61lvv6zbv951gyyq8xc0-env.drv' failed to build
error: 1 dependencies of derivation '/nix/store/394lrlh4bpkaxnafknlfwadp3qardczh-review-shell.drv' failed to build

samuela · 2022-10-05T05:55:38Z

TensorRT and mathematica require special downloads.

ethminer, numbaWithCuda, and pyrealsense2WithCuda are all failing on master. We should mark them as broken.

So this change LGTM modulo the formatting changes mentioned above.

dguibert · 2022-10-05T06:11:41Z

I've amending the formatting changes.
Thx for reviewing this.

(it's time to package version 11.8 released few days ago 😉 )

samuela · 2022-10-05T19:03:28Z

Thanks so much for your PR @dguibert, and thank you for your patience getting it merged! Can't believe that 11.8 is already here haha

dguibert · 2022-10-06T06:02:59Z

11.8 update has been proposed in #194705

dguibert requested a review from samuela August 7, 2022 14:12

ofborg bot added 8.has: package (new) This PR adds a new package 10.rebuild-darwin: 11-100 This PR causes between 11 and 100 packages to rebuild on Darwin. 10.rebuild-linux: 11-100 This PR causes between 11 and 100 packages to rebuild on Linux. labels Aug 7, 2022

SuperSandro2000 reviewed Aug 7, 2022

View reviewed changes

pkgs/development/compilers/cudatoolkit/redist/overrides.nix Outdated Show resolved Hide resolved

SuperSandro2000 reviewed Aug 7, 2022

View reviewed changes

pkgs/test/cuda/cuda-samples/extension.nix Outdated Show resolved Hide resolved

samuela requested changes Aug 10, 2022

View reviewed changes

pkgs/development/libraries/science/math/cudnn/extension.nix Outdated Show resolved Hide resolved

pkgs/development/libraries/science/math/cudnn/extension.nix Outdated Show resolved Hide resolved

aidalgol mentioned this pull request Sep 25, 2022

Prevent eval breakage in CUDA extensions #192958

Merged

13 tasks

SuperSandro2000 added the 2.status: merge conflict This PR has merge conflicts with the target branch label Sep 26, 2022

dguibert added 3 commits October 4, 2022 08:26

Revert "Revert "cudatoolkit_11_7: init at 11.7.0""

df956cf

This reverts commit 90c5057.

tensorrt: fix for cudaVerison 11.7

feccefa

cuda-samples: throw an error for tag 11.7

4f43356

dguibert force-pushed the dg/cudatoolkit_11_7_0 branch from 1a9b818 to 3cb66cf Compare October 4, 2022 15:21

ofborg bot added 8.has: clean-up This PR removes packages or removes other cruft and removed 2.status: merge conflict This PR has merge conflicts with the target branch labels Oct 4, 2022

samuela reviewed Oct 4, 2022

View reviewed changes

pkgs/development/compilers/cudatoolkit/redist/overrides.nix Outdated Show resolved Hide resolved

pkgs/development/compilers/cudatoolkit/redist/overrides.nix Outdated Show resolved Hide resolved

This comment was marked as duplicate.

Sign in to view

cudatoolkit: use if instead of versionOlder+versionAtLeast

a0e9973

dguibert force-pushed the dg/cudatoolkit_11_7_0 branch from 3cb66cf to a0e9973 Compare October 5, 2022 06:06

samuela approved these changes Oct 5, 2022

View reviewed changes

samuela merged commit 6a55613 into NixOS:master Oct 5, 2022

Uh oh!

Conversation

dguibert commented Aug 7, 2022

Description of changes

Things done

Uh oh!

Uh oh!

Uh oh!

samuela left a comment

Choose a reason for hiding this comment

Uh oh!

Uh oh!

Uh oh!

aidalgol commented Sep 25, 2022

Uh oh!

samuela commented Sep 26, 2022

Uh oh!

samuela commented Sep 26, 2022

Uh oh!

dguibert commented Oct 4, 2022

Uh oh!

samuela left a comment

Choose a reason for hiding this comment

Uh oh!

Uh oh!

Uh oh!

samuela commented Oct 5, 2022 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

samuela commented Oct 5, 2022

Uh oh!

This comment was marked as duplicate.

samuela commented Oct 5, 2022

Uh oh!

samuela commented Oct 5, 2022

Uh oh!

dguibert commented Oct 5, 2022

Uh oh!

samuela commented Oct 5, 2022

Uh oh!

dguibert commented Oct 6, 2022

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

4 participants

samuela commented Oct 5, 2022 •

edited

Loading