This repository was archived by the owner on Oct 31, 2026. It is now read-only.
Repository navigation
Conversation
Forked from Ubuntu-24.04-cuda-12.8-minimal so the recipes stay aligned. Changes are the base image, the description label, and two additions: ARG GPU_APP_COLLECTION_BRANCH lets the image be built against a feature branch. A CUDA 13 image cannot be built from dev until gpu-app-collection supports CUDA 13, and that support is what the image is needed to test -- this breaks the cycle. The 13.1.1 branch added the same ARG for the same reason. The clone honours that ARG via --branch/--single-branch. Deliberately kept from the 12.8 recipe, where the 13.1.1 branch lost both: libzstd-dev, and the parallel `make -j32 -C src/ ci` over a serial test-build.sh. Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to subscribe to this conversation on GitHub.
Already have an account?
Sign in.
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Adds the
Ubuntu-24.04-cuda-13.2-minimalimage. Forked fromUbuntu-24.04-cuda-12.8-minimalrather than from the13.1.1branch, so thetwo live recipes stay aligned.
Why
gpu-app-collectionCI currently has no CUDA 13 job, so nothing exercises thetoolkit on any branch. accel-sim/gpu-app-collection#91 adds one, but with no
13.x tag published it has to start from the stock
nvidia/cudaimage andapt-get installits dependencies at job time -- inconsistent with how everyother CUDA version is added here, and exposed to apt drift. This image is what
lets that job look like the 12.8 one.
The cycle, and the ARG that breaks it
The image build ends in
make -C src/ ci. On CUDA 13 that fails againsttoday's
dev, and fixing it is exactly what #91 does -- so a CUDA 13 imagecannot be built from
devuntil #91 lands, and #91 cannot use the image untilit is built.
ARG GPU_APP_COLLECTION_BRANCH=devbreaks that: the image can be validatedagainst the feature branch first, then rebuilt from
devonce it merges.The
13.1.1branch added the same ARG, presumably after hitting the same wall.Diff from the 12.8 recipe
12.8.0-cudnn-devel->13.2.1-cudnn-develARG GPU_APP_COLLECTION_BRANCH, honoured by--branch/--single-branchTwo things are deliberately kept from 12.8 that the
13.1.1branch dropped,both of which look like accidents of forking rather than decisions:
libzstd-dev, and the parallelmake -j32 -C src/ ciin place of a serialtest-build.sh.Why 13.2.1
It is the toolkit accel-sim/gpu-app-collection#91 was developed and verified
against, so the image and that work agree. The existing
13.1.1branch hasnever been published as a tag; if you would rather consolidate on 13.1.1, the
same diff applies with the base image changed.
Not verified here
I have no container runtime on this machine, so this Dockerfile has not been
built. What is verified is that every step it runs works on CUDA 13.2 outside a
container:
make -C src/ cibuilds all five targets on the #91 branch -- 250apps, including
cuda_samples, which the 12.8 job skips. The packages itinstalls are a superset of what that build needed.
🤖 Generated with Claude Code