Commit graph

38 commits

Author SHA1 Message Date
Daniël de Kok b65dfb342c
Merge pull request #113107 from danieldk/remove-ancient-cudas
cudatoolkit-{6,6_5,7,7_5,8}: remove
2021-02-16 08:47:52 +01:00
Daniël de Kok 3c815cf0e6 cudatoolkit-{6,6_5,7,7_5,8}: remove
Remove ancient CUDA toolkits (and corresponding CuDNN versions):

- Not supported by upstream anymore.
- We do not use them in nixpkgs.
- We do not test or actively maintain them.
- Anything but ancient GPUs is supported by newer toolkits.
2021-02-14 11:58:00 +01:00
Daniël de Kok 8dad19aebe cudnn_cudatookit_11: add $ORIGIN/ to library RPATHs
Starting with version 8, CuDNN consists of multiple libraries.
Libraries (except libcudnn.so) are dlopen()ed on demand.

Since the CuDNN library path itself was not added to the rpath of the
libraries, use of CuDNN often fails with the following error:

Could not load library libcudnn_cnn_train.so.8.
Error: libcudnn_ops_train.so.8: cannot open shared object file: No such file or directory
Please make sure libcudnn_cnn_train.so.8 is in your library path!

This change fixes this by adding $ORIGIN/ to the rpath of all CuDNN
libraries.
2021-02-14 11:35:58 +01:00
Daniël de Kok 84786ebecd cudnn_cudatoolkit_11: remove libcudnn_cnn_infer.so exception
With the update to CuDNN 8.1, we do not get the "maximum file size
exceeded" error anymore when patchelf'ing libcudnn_cnn_infer.so. This
change removes the exception for that library.
2021-02-14 11:15:02 +01:00
Daniël de Kok f1db117b9f cudnn_cudatoolkit_11: cudnn_cudatoolkit_11_0 -> cudnn_cudatoolkit_11_2 2021-02-10 13:29:38 +01:00
Daniël de Kok cbbacce70a cudnn_cudatoolkit_11_2: init at 8.1.0 2021-02-10 13:27:55 +01:00
Daniël de Kok 961f531242 cudnn_cudatoolkit_11_1: init at 8.1.0 2021-02-10 13:25:50 +01:00
Daniël de Kok 63cdb44b48 cudnn_cudatoolkit_11_0: 8.0.2 -> 8.1.0 2021-02-10 13:23:51 +01:00
Ben Siraphob 66e44425c6 pkgs/development/libraries: stdenv.lib -> lib 2021-01-21 19:11:02 -08:00
Daniël de Kok 1bdeea7fc8 cudnn_cudatoolkit_10: 10.1 -> 10.2
cudatoolkit_10 is currently CUDA 10.2, whereas cudnn_cudatoolkit_10
was using CuDNN with CUDA 10.1. This change rectifies that
discrepancy.
2020-11-19 10:19:54 +01:00
Matthew Bauer ed950bb104 cudnn: skip libcudnn_cnn_infer.so, instead of removing
This is needed to build the CUDNN samples, so we should make sure it
winds up in $out/lib. The library is too big to support patchelf, so
we need to skip it. Most programs will still get the opengl path
because it also links to other libraries.
2020-08-28 15:16:15 -05:00
Matthew Bauer 164f8024e9 cudatoolkit: init v11.0.3 2020-08-19 13:34:52 -05:00
Michael Reilly 972db874b3 cudnn_cudatoolkit_10_2: init at 7.6.5 2020-03-05 15:14:10 -05:00
Andreas Rammhold 1e72fa5158
Merge pull request #68954 from dguibert/dg/cudatoolkit_10_1
cudatoolkit: add 10.1.243
2019-11-06 08:58:30 +01:00
Joachim Fasting bad07dfac5
tree-wide: replace uses of splitString "." with lib.versions
Quoting from the splitString docstring:

   NOTE: this function is not performant and should never be used.

This replaces trivial uses of splitString for splitting version
strings with the (potentially builtin) splitVersion.
2019-09-26 17:42:49 +02:00
David Guibert 92690a8937 cuddn: add 10.1 (version 7.6.3) 2019-09-17 15:19:41 +02:00
volth 08f68313a4 treewide: remove redundant rec 2019-08-28 11:07:32 +00:00
Ambroz Bizjak 28a0918916 Add driver library path to some packages to find CUDA libraries.
This is to avoid relying on LD_LIBRARY_PATH for finding the CUDA driver libraries.
2019-05-22 18:34:09 +02:00
sveitser e29e315c5f
cudnn_cudatoolkit_10: 7.3.1 -> 7.4.2 2019-03-28 21:49:13 +08:00
Patrick Hilhorst 4cfeb9da53
cudatoolkit, cudnn: many updates
cudatoolkit_9_1: 9.1.85.1 -> 9.1.85.3
cudatoolkit_9_2: init at 9.2.148.1
cudatoolkit_9: 9.1.85.1 -> 9.2.148.1
cudatoolkit_10, cudatoolkit_10_0: init at 10.0.130

cudnn_cudatoolkit_9_0: 7.0.5 -> 7.3.0
cudnn_cudatoolkit_9_1: init at 7.1.3
cudnn_cudatoolkit_9_2: init at 7.2.1
cudnn_cudatoolkit_9: 7.0.5 (9.1) -> 7.3.1 (9.2)
cudnn_cudatoolkit_10, cudnn_cudatoolkit_10_0: init at 7.3.1
2018-10-30 19:09:20 +01:00
Luke Adams eba8679bc5 cudnn*_cudatoolkit*: use public NVIDIA mirror (#31551)
* cudnn*_cudatoolkit*: use NVIDIA mirror instead of requireFile

* cudnn_cudatoolkit7: update hash since public mirror is different
2018-07-18 03:21:57 +00:00
Jean-Philippe Bernardy 397c4b7748 cudatoolkit: add 9.0
This version is a dependency of tensorflow binaries
2018-05-31 22:21:54 +02:00
Jean-Philippe Bernardy 9e730ee254 renaming cudatoolkit version numbers 2018-05-31 22:21:53 +02:00
Anders Kaseorg 6f073154da Give working advice for running nix-prefetch-url
nix-prefetch-url file://… requires a full path.  Suggest an invocation
that will provide one.

Signed-off-by: Anders Kaseorg <andersk@mit.edu>
2018-03-04 17:28:36 +00:00
Nikolay Amiantov efe836160f cudatoolkit: 9.0.176 -> 9.1.85.1 2018-02-25 22:24:09 +03:00
Travis Whitaker 777c9bfdae cuDNN 7.0.3 -> 7.0.5 2017-12-07 15:13:07 -08:00
Nikolay Amiantov 77046bed8a cudnn7: 7.0 -> 7.0.3 2017-10-28 17:39:59 +03:00
Nikolay Amiantov 6309e5fa01 cudnn: refactor to use shared derivation
Also return 8.0-6.0 version for TensorFlow.
2017-10-19 12:47:30 +03:00
Nikolay Amiantov ce9059a120 cudnn_cudatoolkit9: init at 7.0
Make it the default.
2017-10-19 12:47:30 +03:00
Jean-Philippe Bernardy 1ba197f225 make my maintainer handle match my github username 2017-09-12 15:17:18 +02:00
Jean-Philippe Bernardy d51bd36646 cudnn: add 6.0 2017-09-07 09:28:03 +02:00
Silvan Mosberger f5fa5fa4d6 pkgs: refactor needless quoting of homepage meta attribute (#27809)
* pkgs: refactor needless quoting of homepage meta attribute

A lot of packages are needlessly quoting the homepage meta attribute
(about 1400, 22%), this commit refactors all of those instances.

* pkgs: Fixing some links that were wrongfully unquoted in the previous
commit

* Fixed some instances
2017-08-01 22:03:30 +02:00
Matthew Daiter 82835f7eb3 cudnn51_cudatoolkit80: hash change 2017-02-08 18:40:32 +01:00
Matthew Daiter 3897b344f3 cudnn51_cudatoolkit80: Init at 5.1-8.0 2016-11-21 14:51:39 +01:00
Matthew Daiter 541b3fc7d2 cudnn5_cudatoolkit80: adding to main repo 2016-11-07 12:07:29 +01:00
artuuge 465d6a2113 Theano-cuda: init at 0.8.2 2016-06-20 01:06:07 +02:00
Ian-Woo Kim ff8985e7ab cudnn: add cudatoolkit to propagatedBuildInputs 2016-06-02 19:33:03 +02:00
Ian-Woo Kim e47070ad64 cudnn: init at 4.0 2016-05-25 18:26:57 +00:00