NixOS 20.0.9's default NVIDIA video driver version is 455.38, which supports CUDA 11.1.1.
**Things to note:**
- [ ] Users with nvtop installed may need to upgrade their nvtop installation due to caching behavior
This has the benefit of being able to override all the inputs to the
build where you were previously only able to override the entire package
set (if at all).
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.
This fixes cudatoolkit building on non-sandboxed system.
The cudatoolkit tests run each of the CUDA binaries as a small smoke
test just to make sure they all can at least somewhat run.
However, the `cuda-gdb` binary doesn't run correctly on non-sandboxed
systems because it picks up versions of Python from /usr/lib.
This PR disables the smoke tests for now. They should be re-enabled
when we figure out how to make them work for cuda-gdb on non-sandboxed
systems.
This PR is for #57939.
This fixes the following error when starting nvvp or nsight:
```
(java:23876): Gtk-WARNING **: 19:12:17.777: Unable to locate theme engine in module_path: "adwaita",
A fatal error has been detected by the Java Runtime Environment:
SIGSEGV (0xb) at pc=0x00007f957f7b67fe, pid=23876, tid=140281059333888
JRE version: Java(TM) SE Runtime Environment (8.0_77-b03) (build 1.8.0_77-b03)
Java VM: Java HotSpot(TM) 64-Bit Server VM (25.77-b03 mixed mode linux-amd64 compressed oops)
Problematic frame:
C [libpixbufloader-bmp.so+0x27fe] gdk_pixbuf__bmp_image_load_increment+0xeee
```
Since GDK_PIXBUF_MODULE_FILE should match the version we actually link
to, we override the environment variable using the one provided by the
gdk_pixbuf setup-hook.
* 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
* Add a "lib" output containing (for the moment) only libOpenCL. This
reduces the closure size of opensubdiv and blender by about 2 GiB.
* Add a "doc" output (about ~200 MiB).
* Remove the "sdk" output since it was worse than useless: all of the
"sdk" output was also included in "out", so it actually increased
disk space usage.
* Run patchelf on ELF binaries only.
- Set a cmake flag to allow cmake to find CUDA automatically.
- Pass -D_FORCE_INLINES to work around
/nix/store/8sl4jfs3nq0pkq4gg655s3axrxdx7z29-glibc-2.24-dev/include/string.h: In function 'void* __mempcpy_inline(void*, const void*, size_t)':
/nix/store/8sl4jfs3nq0pkq4gg655s3axrxdx7z29-glibc-2.24-dev/include/string.h:650:42: error: 'memcpy' was not declared in this scope
https://github.com/BVLC/caffe/issues/4046
This fixes OpenSubdiv and Blender.
This seems to have been confusing people, using both xlibs and xorg, etc.
- Avoided renaming local (and different) xlibs binding in gcc*.
- Fixed cases where both xorg and xlibs were used.
Hopefully everything still works as before.
This effectively reverts 86c283824f
("If cuda headers are presented to nix [...]") and all the following
workarounds that was added due to that commit.
As far as I can tell[1] this hack isn't needed anymore. And moving
includes to $out/usr_include causes pain for cudatoolkit users, so
better get rid of it.
In patches that did more than the $out/usr_include workaround, I only
changed the line back to $out/include instead of re-generating the
patches and fully removing the changed line.
[1]: I build tested blender and caffe, and temporarily added
recurseIntoAttrs to rPackages and haskellPackages so that nox-review
could get proper coverage. However, many of the packages do not build
even before this patch. I also built CUDA samples with cudatoolkit7
that ran fine.
