Skip to content

[RVV] add rvv f32 kernel for ppmm#9962

Open
velonica0 wants to merge 4 commits into
google:masterfrom
velonica0:rvv-ppmm
Open

[RVV] add rvv f32 kernel for ppmm#9962
velonica0 wants to merge 4 commits into
google:masterfrom
velonica0:rvv-ppmm

Conversation

@velonica0
Copy link
Copy Markdown
Contributor

@velonica0 velonica0 commented Apr 14, 2026

Part of #9954

Add RVV kernel for f32-ppmm with MR={1,2,4} and NR={m1,m2,m4} variants. Tested on SpacemiT K1 CPU and K3 CPU, both VLEN=256.

@velonica0 velonica0 mentioned this pull request Apr 14, 2026
Open
dsharlet
dsharlet approved these changes Apr 14, 2026
View reviewed changes
Comment thread cmake/gen/rvv_microkernels.cmake Outdated
src/x32-transposec/gen/x32-transposec-8x8-rvv.c
src/x32-transposec/gen/x32-transposec-16x8-rvv.c
src/x32-transposec/gen/x32-transposec-32x8-rvv.c)
src/x32-transposec/gen/x32-transposec-32x8-rvv.c
Copy link
Copy Markdown
Collaborator

@dsharlet dsharlet Apr 14, 2026

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

In all 3 PRs, there will be merge conflicts because these are all added in the same place. Keeping these lists alphabetized will avoid this.

Copy link
Copy Markdown
Contributor Author

@velonica0 velonica0 Apr 15, 2026

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Thank you for your review, I have made the changes.

@velonica0
Copy link
Copy Markdown
Contributor Author

velonica0 commented Apr 20, 2026

Hi, @dsharlet
My apologies. Fixed a CI error caused by a missing ). It should be fine now for 3 PRs(#9963 #9964).
Sorry for that.

@velonica0
Copy link
Copy Markdown
Contributor Author

velonica0 commented Apr 23, 2026 •
edited
Loading

Hi, @dsharlet
Can these be merged?

@velonica0
Copy link
Copy Markdown
Contributor Author

velonica0 commented Apr 29, 2026

Hi, @dsharlet My apologies. Fixed a CI error caused by a missing ). It should be fine now for 3 PRs(#9963 #9964). Sorry for that.

Hi, @dsharlet
Can these be merged? I see CI is pass!

@velonica0
Copy link
Copy Markdown
Contributor Author

velonica0 commented May 4, 2026

ping @dsharlet
Thanks!

@velonica0
Copy link
Copy Markdown
Contributor Author

velonica0 commented May 14, 2026

Hi, @dsharlet @fbarchard
I see that the PR has been approved. Can it be merged now? Thanks!

@dsharlet
Copy link
Copy Markdown
Collaborator

dsharlet commented May 15, 2026

@velonica0 can you please merge the master branch to re-trigger the tools to run? Same for any other outstanding branches that should be merged.

velonica0 added 3 commits May 18, 2026 11:36
@velonica0
[RVV] add rvv f32 kernel for ppmm
b606c1e 
Add RVV kernel for f32-ppmm with MR={1,2,4} and NR={m1,m2,m4} variants.
Tested on SpacemiT K1 CPU and K3 CPU, both VLEN=256.
@velonica0
Alphabetize f32-ppmm entries in cmake/bzl lists
025108a 
Move f32-ppmm entries from end of PROD/NON_PROD lists to alphabetical
position (between f32-maxpool and f32-qs8-vcvt) to avoid merge conflicts
with other RVV PRs.
@velonica0
fix cmake bug
7dbe384
@velonica0
Copy link
Copy Markdown
Contributor Author

velonica0 commented May 18, 2026

@velonica0 can you please merge the master branch to re-trigger the tools to run? Same for any other outstanding branches that should be merged.

Hi, @dsharlet
Both this PR and PR #9964 have been rebased. Ready for your review again. Thanks a lot for your help!

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

Reviewers

@dsharlet dsharlet dsharlet approved these changes

Assignees

No one assigned

Labels

None yet

Projects

None yet

Milestone

No milestone

Development

Successfully merging this pull request may close these issues.

2 participants

@velonica0 @dsharlet