Skip to content

ci: bump stable-channel workflows to the ROCm 10.0 image - #528

Open
zichguan-amd wants to merge 3 commits into
amd-stagingfrom
zichguan/rocm-10.0-workflows
Open

zichguan-amd wants to merge 3 commits into
amd-stagingfrom
zichguan/rocm-10.0-workflows

Conversation

@zichguan-amd

Copy link
Copy Markdown
Collaborator

Summary

Stacked on #527 (the Dockerfile bump that builds the :10.0 image). Points the stable-channel CI at the new image and drops skips that only existed because of 7.14 gaps.

  • configure_ci.py stable_release: image :7.14:10.0, label ROCm 7.14ROCm 10.0.

  • 7 native build workflows (build_hip_basic, build_libraries, build_applications, build_tools, build_systems, build_programming_guide, build_hip_documentation): image tag :7.14:10.0.

  • Refreshed stale 7.14 mentions in comments/help text (build-rocm-examples-reusable.yml, build_docker.yml, generate_skip_tests.py).

  • Removed the two 7.14-only stable-channel build skips in skip_manifest.py:

    • HIP-Basic/cooperative_groups_prefix_sum (needed hip/cooperative_groups/hip_scan.h)
    • Libraries/rocProfiler-SDK/thread_trace (needed rocprofiler_thread_trace_shader_data_t)

    Both headers ship in ROCm 10.0 (/opt/rocm/include/...), and both examples were compiled against the locally-built 10.0 image (gfx1100) to confirm before removing the skips.

Notes

Test plan

  • Native stable-channel build workflows pull :10.0 and build green.
  • cooperative_groups_prefix_sum and rocProfiler-SDK/thread_trace now build+test on the stable channel (no longer skipped).
  • Nightly matrix stable_release job uses the 10.0 preinstalled image.

🤖 Generated with Claude Code

@zichguan-amd
zichguan-amd requested a review from a team as a code owner September 18, 2026 01:35
@github-advanced-security

Copy link
Copy Markdown

You are seeing this message because GitHub Code Scanning has recently been set up for this repository, or this pull request contains the workflow file for the Code Scanning tool.

What Enabling Code Scanning Means:

  • The 'Security' tab will display more code scanning analysis results (e.g., for the default branch).
  • Depending on your configuration and choice of analysis tool, future pull requests will be annotated with code scanning analysis results.
  • You will be able to see the analysis results for the pull request's branch on this overview once the scans have completed and the checks have passed.

For more information about GitHub Code Scanning, check out the documentation.

@zichguan-amd
zichguan-amd added this pull request to stack #529 September 18, 2026 14:22
Base automatically changed from zichguan/rocm-10.0-preinstalled-image to amd-staging September 18, 2026 14:26
zichguan-amd and others added 3 commits September 18, 2026 13:58
## Summary
- Bumps the `ubuntu-24.04-rocm` preinstalled image from ROCm **7.14** to
**10.0**.
- The old stable host `repo.amd.com/rocm/packages-multi-arch/ubuntu2404`
only serves up to 7.14 (verified live: `amdrocm-core-sdk10.0` count = 0
there). ROCm 10.0 stable moved to
`stable.repo.amd.com/rocm/core/packages/ubuntu2404`, so the stable repo
URL and its gpg key (`stable.repo.amd.com/rocm/gpg/packages.gpg`) are
repointed there.
- All five installed stable packages (`amdrocm-core-sdk`,
`-hiptensor{,-dev}`, `-rocalution{,-dev}`) exist at `10.0`; names
auto-resolve via `${ROCM_VERSION}`.
- Prerelease repo is left untouched — `amdrocm-decode-test` is still
only published there (now `10.0.0~pre4`); only the stale comment version
is updated.

This is intentionally **Dockerfile-only** so `build_docker.yml` builds
and pushes the `:10.0` image first. A follow-up will repoint
`configure_ci.py`'s `stable_release` entry
(`ghcr.io/rocm/rocm-examples-ubuntu-24.04-rocm:7.14` → `:10.0`, label
`ROCm 7.14` → `10.0`) once the image exists.

The pushed image tag auto-derives from `ARG ROCM_VERSION`
(`build_docker.yml` L91), so no workflow change is needed to tag
`:10.0`.

## Test plan
- [ ] `build_docker.yml` (triggered by this PR touching
`Dockerfiles/**`) builds the `ubuntu-24.04-rocm` image successfully
against the new stable repo.
- [ ] Confirm `apt install amdrocm-core-sdk10.0` (+
hiptensor/rocalution) resolves from `stable.repo.amd.com` and
`amdrocm-decode-test10.0` from the prerelease repo.
- [ ] Confirm the built image populates `/opt/rocm` (incl.
`/opt/rocm/core/.info/version`) so the downstream preinstalled path
stays valid.

🤖 Generated with [Claude Code](https://claude.com/claude-code)

Co-authored-by: Claude Opus 4 <noreply@anthropic.com>
Follow-up to the preinstalled Dockerfile bump: point the stable-channel
CI at the new ghcr :10.0 image tag now that the 10.0 image is built.

- configure_ci.py stable_release: image :7.14 -> :10.0, label ROCm 10.0
- native build workflows (hip_basic, libraries, applications, tools,
  systems, programming_guide, hip_documentation): image tag :7.14 -> :10.0
- refresh stale "7.14" references in comments/help text

The skip_manifest.py entries that reference the 7.14 image (missing
hip_scan.h / rocprofiler_thread_trace_shader_data_t) are intentionally
left untouched; whether they still apply on 10.0 needs verification
against the built image and is tracked as a separate follow-up.

Co-Authored-By: Claude Opus 4 <noreply@anthropic.com>
hip_scan.h and rocprofiler_thread_trace_shader_data_t were absent from
the pinned 7.14 stable image, so cooperative_groups_prefix_sum and
rocProfiler-SDK/thread_trace were build-skipped on the stable channel.
The stable image is now ROCm 10.0, which ships both headers; verified by
compiling both examples against the 10.0 image. Remove the skips so they
build and test on the stable channel again.

Co-Authored-By: Claude Opus 4 <noreply@anthropic.com>
@zichguan-amd
zichguan-amd force-pushed the zichguan/rocm-10.0-workflows branch from 0e8d6cd to 04b3218 Compare September 18, 2026 17:58
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

2 participants