Skip to content

HIP: rocWMMA documentation and enabling in workflow builds#12179

Merged
JohannesGaessler merged 3 commits into
ggml-org:masterfrom
hjc4869:ci-rocwmma
Mar 6, 2025
Merged

HIP: rocWMMA documentation and enabling in workflow builds#12179
JohannesGaessler merged 3 commits into
ggml-org:masterfrom
hjc4869:ci-rocwmma

Conversation

@hjc4869
Copy link
Copy Markdown
Contributor

@hjc4869 hjc4869 commented Mar 4, 2025 •
edited
Loading

  1. Enable rocWMMA for Windows and Ubuntu in official GitHub Actions.
  2. Add documentation of GGML_HIP_ROCWMMA_FATTN CMake option.

build.yml changes validated in forked branches

Related: #12032

@github-actions github-actions Bot added documentation Improvements or additions to documentation devops improvements to build systems and github actions labels Mar 4, 2025
@hjc4869 hjc4869 mentioned this pull request Mar 4, 2025
Merged
@JohannesGaessler JohannesGaessler merged commit 3ffbbd5 into ggml-org:master Mar 6, 2025
@JohannesGaessler
Copy link
Copy Markdown
Contributor

JohannesGaessler commented Mar 6, 2025

Sorry, I merged this PR but it seems the CI on master is now failing because of it due to the rocWMMA files not being found. The CI was successful for the PR though; are there different setups?

@hjc4869
Copy link
Copy Markdown
Contributor Author

hjc4869 commented Mar 6, 2025

Let me take a look what's possibly going wrong.

@hjc4869 hjc4869 deleted the ci-rocwmma branch March 6, 2025 14:13
@hjc4869 hjc4869 restored the ci-rocwmma branch March 6, 2025 14:14
@hjc4869
Copy link
Copy Markdown
Contributor Author

hjc4869 commented Mar 6, 2025

windows-latest-cmake-hip job didn't specify AMDGPU_TARGETS, thus it was using gfx906 as default target, and it didn't support WMMA so it skipped compiling related device code.

windows-latest-cmake-hip-release has AMDGPU_TARGETS specified, but the job was skipped in non-master branches, so neither my private build nor PR check has actually validated the WMMA build under Windows.

The issue is likely due to not using absolute paths in -DCMAKE_CXX_FLAGS. There will also be some problems by using $PWD (contains \ backslash on Windows) with Unix Makefiles under Windows. I'm validating a fix on my fork branch and will send a fix after it's ready.

@hjc4869
Copy link
Copy Markdown
Contributor Author

hjc4869 commented Mar 6, 2025

PR sent: #12230

@hjc4869 hjc4869 deleted the ci-rocwmma branch March 6, 2025 18:53
@hjc4869 hjc4869 mentioned this pull request Mar 7, 2025
Merged
arthw pushed a commit to arthw/llama.cpp that referenced this pull request Mar 19, 2025
@hjc4869 @arthw
HIP: rocWMMA documentation and enabling in workflow builds (ggml-org#…
de6fff6 
…12179)

* Enable rocWMMA for Windows CI build

* Enable for Ubuntu

* GGML_HIP_ROCWMMA_FATTN documentation work
Seunghhon pushed a commit to Seunghhon/llama.cpp that referenced this pull request Apr 26, 2026
@hjc4869
HIP: rocWMMA documentation and enabling in workflow builds (ggml-org#…
72c8230 
…12179)

* Enable rocWMMA for Windows CI build

* Enable for Ubuntu

* GGML_HIP_ROCWMMA_FATTN documentation work
ljubomirj pushed a commit to ljubomirj/llama.cpp that referenced this pull request May 6, 2026
@hjc4869
HIP: rocWMMA documentation and enabling in workflow builds (ggml-org#…
f30e030 
…12179)

* Enable rocWMMA for Windows CI build

* Enable for Ubuntu

* GGML_HIP_ROCWMMA_FATTN documentation work
my-other-github-account pushed a commit to my-other-github-account/llama.cpp that referenced this pull request May 15, 2026
@hjc4869
HIP: rocWMMA documentation and enabling in workflow builds (ggml-org#…
f5038d5 
…12179)

* Enable rocWMMA for Windows CI build

* Enable for Ubuntu

* GGML_HIP_ROCWMMA_FATTN documentation work
my-other-github-account pushed a commit to my-other-github-account/llama.cpp that referenced this pull request May 15, 2026
@hjc4869
HIP: rocWMMA documentation and enabling in workflow builds (ggml-org#…
5c881d5 
…12179)

* Enable rocWMMA for Windows CI build

* Enable for Ubuntu

* GGML_HIP_ROCWMMA_FATTN documentation work
phibya pushed a commit to ziee-ai/llama.cpp that referenced this pull request May 29, 2026
@hjc4869
HIP: rocWMMA documentation and enabling in workflow builds (ggml-org#…
9cb0c9c 
…12179)

* Enable rocWMMA for Windows CI build

* Enable for Ubuntu

* GGML_HIP_ROCWMMA_FATTN documentation work
AlexiAlp pushed a commit to minghaop/llama.cpp that referenced this pull request Jun 2, 2026
@hjc4869
HIP: rocWMMA documentation and enabling in workflow builds (ggml-org#…
9100106 
…12179)

* Enable rocWMMA for Windows CI build

* Enable for Ubuntu

* GGML_HIP_ROCWMMA_FATTN documentation work
AlexiAlp pushed a commit to minghaop/llama.cpp that referenced this pull request Jun 2, 2026
@hjc4869
HIP: rocWMMA documentation and enabling in workflow builds (ggml-org#…
2e6dbd2 
…12179)

* Enable rocWMMA for Windows CI build

* Enable for Ubuntu

* GGML_HIP_ROCWMMA_FATTN documentation work
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Reviewers

@slaren slaren slaren approved these changes

@JohannesGaessler JohannesGaessler JohannesGaessler approved these changes

Assignees

No one assigned

Labels

devops improvements to build systems and github actions documentation Improvements or additions to documentation

Projects

None yet

Milestone

No milestone

Development

Successfully merging this pull request may close these issues.

3 participants

@hjc4869 @JohannesGaessler @slaren