-
Notifications
You must be signed in to change notification settings - Fork 154
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
[Packaging] Publish AdaptiveCpp package for Ubuntu #1344
base: develop
Are you sure you want to change the base?
[Packaging] Publish AdaptiveCpp package for Ubuntu #1344
Conversation
722a629
to
7541653
Compare
I tried adding a job for a Here it is, in case it helps -
|
@DhruvDh Thanks. I think the space problem can be solved by thinning out the default runner image a little, as we already do here: AdaptiveCpp/.github/workflows/linux.yml Line 145 in b9b9d32
However, I don't think that we need a separate ROCm package. We can just have one that supports everything :) Then what it can offload to only depends on whether users have CUDA/ROCm/OpenCL etc installed. I think the lack of ROCm support in the current version of this PR is just due to it being WIP. (and probably we shouldn't target ROCm 6.0, as it has some issues) |
c4b188c
to
cd4bbe4
Compare
.github/workflows/package-ubuntu.yml
Outdated
- name: install CUDA | ||
run : | | ||
mkdir -p /opt/AdaptiveCpp/cuda | ||
wget -q -O cuda.sh http://developer.download.nvidia.com/compute/cuda/11.0.2/local_installers/cuda_11.0.2_450.51.05_linux.run |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
If we download CUDA from NVIDIA anyway, we might want to pick a newer one. CUDA 11.0 is already pretty old.
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
libcudart12.0
is not available through aptitude, this creates a problem when the Debian helper shlibdeps
is resolving CUDA dependency information (path to dynamic library and version) during the installation phase. So to circumvent that I also downgraded CUDA Toolkit to 11.0.
.github/workflows/package-ubuntu.yml
Outdated
run: | | ||
cd .. | ||
cp -R AdaptiveCpp acpp-24.02.0 | ||
tar -cf acpp_24.02.0.orig.tar.gz acpp-24.02.0/ |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Does it actually need to be versioned here? Ultimately we'd like to have new packages for every push anyway, and it would be easier to maintain if we didn't have to always update version numbers.
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
yes, the tar file needs to carry this name-version scheme. Part of the initialization process is to create an upstream tarball <name>_<version>-<package_revision>.tar.xz
Do I understand correctly that the packages are now ready for testing? If so, I call on everybody reading here to give them a try and see if they work :) |
The packages at this stage in this branch are ready. Would be great to hear some feedback, especially if the configuration even works. |
44af9ab
to
b64bff8
Compare
b64bff8
to
5b34860
Compare
5b34860
to
2cf1d45
Compare
This PR adds a Github workflow for publishing AdaptiveCpp packages on every
git-push
.debian
directory is handled manually: ideally I would want to automate thisany