-
Notifications
You must be signed in to change notification settings - Fork 74
New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
compile time issues with nvcc and GCC 10.3 #1297
Comments
CUDA 11.3 does only supports up to GCC 9.x: https://docs.nvidia.com/cuda/cuda-installation-guide-linux/index.html |
I checked CUDA 11.1 include file
|
@alpaka-group/alpaka-maintainers I disabled for now the requirement to pass the CI tests for:
This avoids that other developments get blocked due to a compiler bug. |
fix alpaka-group#1297 Disable tests due to a GCC 10 bug where we have currently no work around.
fix alpaka-group#1297 Disable tests due to a GCC 10 bug where we have currently no work around.
Unluckily it is not enough to mark the tests as not required (you need still admin rights to merge the PR), therefore the combination nvcc+GCC10.3 is currently disabled. #1302 :-( |
fix alpaka-group#1297 Disable tests due to a GCC 10 bug where we have currently no work around.
fix alpaka-group#1297 Disable tests due to a GCC 10 bug where we have currently no work around.
fix #1297 Disable tests due to a GCC 10 bug where we have currently no work around.
fix alpaka-group#1297 Disable tests due to a GCC 10 bug where we have currently no work around.
fix #1297 Disable tests due to a GCC 10 bug where we have currently no work around.
Nvidia's install guide is not updated properly and who knows when it will be. GCC 10 support was mentioned in a few earlier CUDA 11 release notes https://developer.nvidia.com/cuda-toolkit-archive such as
I've had no problems with latest CUDA and GCC 10 until GCC 10.3.0. |
Yes GCC 10 works fine, except for the mentioned issue with 10.3. |
I reopened this issue, the issue is still not solved, only handled by a workaround. |
The same error occurs with GCC 11.0.1, libstdc++ 10.3 and CUDA 11.2, which is weird because nvcc causes GCC 11 to not use libstdc++ 11, but version 10. If I compile a non-CUDA file, the right libstdc++ is selected. |
https://gitlab.kitware.com/cmake/cmake/-/issues/19458 sounds a little bit like your issue. Do you have the issue with alpaka or with a tiny reporducer and command line |
@psychocoderHPC I'm not an Alpaka user - that Gitlab issue didn't solve my problems with the Nvidia stack, for ex. trying to build https://github.com/NVIDIA/nccl-tests Relevant gcc bug reports. Don't know how quickly gcc issues get fixed, so for users on gcc 10.3.0 the Nvidia CUDA stack is essentially unusable for now for some indefinite amount of time. https://gcc.gnu.org/bugzilla/show_bug.cgi?id=100102 |
@edrozenberg Thanks for the clarification 👍 and the links to the corresponding gcc bug reports. |
The GCC project has committed a patch: https://gcc.gnu.org/git/gitweb.cgi?p=gcc.git;h=5357ab75dedef403b0eebf9277d61d1cbeb5898f The patch removes ~ 40 lines from The Nvidia stack works again for me after I built and installed patched gcc 10.3.0 packages. |
Almost a year later there is still no sight of GCC 10.4 :-( |
gcc 10.4 has been released. It includes the above-mentioned bugfix. |
👍 |
The CI is showing compile time issues with nvcc and GCC 10.3
see:
A similar issue was reported:
GCC bug reports:
#1297 (comment)
I am not sure but the error could be fixed with https://gcc.gnu.org/bugzilla/show_bug.cgi?id=100101
ATTENTION Currently two tests in the CI are disabled: #1302
The text was updated successfully, but these errors were encountered: