Skip to content

[Draft] Fix compute_type xf32_r issue for override file#1838

Open
tomchengchitang wants to merge 1 commit into
developfrom
override-file-for-xf32
Open

[Draft] Fix compute_type xf32_r issue for override file#1838
tomchengchitang wants to merge 1 commit into
developfrom
override-file-for-xf32

Conversation

@tomchengchitang

Copy link
Copy Markdown
Contributor
  • This is a draft to test, not the final version.

msujon-AMD pushed a commit to msujon-AMD/hipBLASLt that referenced this pull request Apr 1, 2025
@jayhawk-commits

Copy link
Copy Markdown
Contributor

Please resolve merge conflicts or close this PR to complete the task of importing PRs from this repo to the monorepo.

assistant-librarian Bot pushed a commit that referenced this pull request Oct 2, 2025
[hipblastlt] PTS script: added --targetArch and
 auto-detection to run different suites according to the arch.  (#1838)

## Motivation

<!-- Explain the purpose of this PR and the goals it aims to achieve.
-->
Performance CI Job runs "**`--suite all`**" by default. But in current
scenario, different architectures have their own focused default problem
sizes (but we'd still like to use --suite all). It is possible that "No
Solution Found" error will happen in different architectures since the
problem-sizes in the bench-yaml are not supported.

## Technical Details

<!-- Explain the changes along with any relevant GitHub links. -->
Added an argument **`--targetArch`** to **`./hipblaslt-perf`**.
When running the Perf CI Job, rockJenkins can still use `--suite all`,
but with this PR, we can add an additional "**`--targetArch gfxXXX`**"
option in the groovy. So the Perf Job can run the specific problem sizes
according to the GPU of the running nodes.

When --targetArch is not specified (this modification needs to be done
in rocJenkins), this PR provides a simple auto-detection way to detect
the ISA of the current architecture.

But for PTS, the most ideal case is to pass "**`--targetArch gfxXXX`**"
in the performance CI job; the auto-detection functionality currently
works as a workaround when the groovy file is not updated.

## Test Plan

<!-- Explain any relevant testing done to verify this PR. -->
Tested by common CI routine.

## Submission Checklist

- [x] Look over the contributing guidelines at
https://github.com/ROCm/ROCm/blob/develop/CONTRIBUTING.md#pull-requests.
ammallya pushed a commit that referenced this pull request Nov 21, 2025
… different suites according to the arch. (#1838)

## Motivation

<!-- Explain the purpose of this PR and the goals it aims to achieve.
-->
Performance CI Job runs "**`--suite all`**" by default. But in current
scenario, different architectures have their own focused default problem
sizes (but we'd still like to use --suite all). It is possible that "No
Solution Found" error will happen in different architectures since the
problem-sizes in the bench-yaml are not supported.

## Technical Details

<!-- Explain the changes along with any relevant GitHub links. -->
Added an argument **`--targetArch`** to **`./hipblaslt-perf`**. 
When running the Perf CI Job, rockJenkins can still use `--suite all`,
but with this PR, we can add an additional "**`--targetArch gfxXXX`**"
option in the groovy. So the Perf Job can run the specific problem sizes
according to the GPU of the running nodes.

When --targetArch is not specified (this modification needs to be done
in rocJenkins), this PR provides a simple auto-detection way to detect
the ISA of the current architecture.

But for PTS, the most ideal case is to pass "**`--targetArch gfxXXX`**"
in the performance CI job; the auto-detection functionality currently
works as a workaround when the groovy file is not updated.

## Test Plan

<!-- Explain any relevant testing done to verify this PR. -->
Tested by common CI routine. 

## Submission Checklist

- [x] Look over the contributing guidelines at
https://github.com/ROCm/ROCm/blob/develop/CONTRIBUTING.md#pull-requests.
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