8000 [Codegen][AMDGPU] Add pingpong to default gfx942 tuning by qedawkins · Pull Request #20678 · iree-org/iree · GitHub
[go: up one dir, main page]
More Web Proxy on the site http://driver.im/
Skip to content

[Codegen][AMDGPU] Add pingpong to default gfx942 tuning #20678

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Merged
merged 1 commit into from
May 5, 2025

Conversation

qedawkins
Copy link
Contributor
@qedawkins qedawkins commented Apr 29, 2025

This adds 4 pingpong cases

  1. 256x256 MMT f16
  2. [1x]256x256 MMT f16 with an extra unit M dimension
  3. [1x]128x256 MMT f16 with an extra unit M dimension
  4. [1x]256x256 MMT f8 with an extra unit M dimension

The extra unit M dimensions are to support blocked dynamic dims that
come from BlockDynamicDimensions pass based on alignment info.

The strategies look for tile size divisibility and a minimum problem
size when matching.

@qedawkins qedawkins force-pushed the pingpong_tuning_default branch from 6fc9787 to ca240cc Compare April 30, 2025 19:55
This adds 4 pingpong cases
1. 256x256 MMT f16
2. [1x]256x256 MMT f16 with an extra unit M dimension
3. [1x]128x256 MMT f16 with an extra unit M dimension
4. [1x]256x256 MMT f8 with an extra unit M dimension

The extra unit M dimensions are to support blocked dynamic dims that
come from BlockDynamicDimensions pass based on alignment info.

The strategies look for tile size divisibility and a minimum problem
size when matching.
@qedawkins qedawkins force-pushed the pingpong_tuning_default branch from ca240cc to 87e6f23 Compare May 1, 2025 01:09
@qedawkins qedawkins marked this pull request as ready for review May 1, 2025 01:10
@qedawkins qedawkins requested a review from bjacob May 1, 2025 18:58
@qedawkins qedawkins merged commit 6c40355 into iree-org:main May 5, 2025
42 checks passed
@qedawkins qedawkins deleted the pingpong_tuning_default branch May 5, 2025 16:21
KyleHerndon pushed a commit to KyleHerndon/iree that referenced this pull request May 7, 2025
This adds 4 pingpong cases
1. 256x256 MMT f16
2. [1x]256x256 MMT f16 with an extra unit M dimension
3. [1x]128x256 MMT f16 with an extra unit M dimension
4. [1x]256x256 MMT f8 with an extra unit M dimension

The extra unit M dimensions are to support blocked dynamic dims that
come from BlockDynamicDimensions pass based on alignment info.

The strategies look for tile size divisibility and a minimum problem
size when matching.
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
0