Skip to content
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.

Sign up for GitHub

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

Jump to bottom

[GPU] Use shared memory for data tiled multi_mma ops #18625

Merged
merged 1 commit into from
Oct 1, 2024
Merged

[GPU] Use shared memory for data tiled multi_mma ops #18625

merged 1 commit into from
Oct 1, 2024

Conversation

Max191
Copy link
Contributor

@Max191 Max191 commented Sep 27, 2024 •
edited
Loading

This enables shared memory usage for data tiled multi_mma ops. The only change is that promote_operands is now set on data tiled multi_mma configs.

@Max191 Max191 mentioned this pull request Sep 27, 2024
Merged
qedawkins
qedawkins reviewed Sep 27, 2024
View reviewed changes
Copy link
Contributor

@qedawkins qedawkins left a comment

Choose a reason for hiding this comment

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

Sorry, I had forgot I hadn't landed this yet: #18576

(That change was made specifically for this reason, so you can rebase on top of it now).

@Max191
Copy link
Contributor Author

Max191 commented Sep 27, 2024

Sorry, I had forgot I hadn't landed this yet: #18576

(That change was made specifically for this reason, so you can rebase on top of it now).

Nice, I'll wait on this until my other kernel config PR lands then. The e2e tests work without shared memory too, so it's just an optimzation right now.

@Max191 Max191 marked this pull request as draft September 27, 2024 19:40
@Max191 Max191 force-pushed the data-tiled-mma-use-shared-mem branch from 046d3c4 to 63ba6c2 Compare September 30, 2024 15:03
@Max191
Copy link
Contributor Author

Max191 commented Sep 30, 2024 •
edited
Loading

This is based on #18623 now. Changes for this PR are in the last 2 commits.

EDIT: rebased now.

@Max191 Max191 marked this pull request as ready for review September 30, 2024 15:05
@Max191
[GPU] Use shared memory for data tiled multi_mma ops
25d2c05 
Signed-off-by: Max Dawkins <max.dawkins@gmail.com>
@Max191 Max191 force-pushed the data-tiled-mma-use-shared-mem branch from 63ba6c2 to 25d2c05 Compare October 1, 2024 17:12
@Max191 Max191 enabled auto-merge (squash) October 1, 2024 17:19
@Max191 Max191 merged commit 839f7f6 into iree-org:main Oct 1, 2024
35 checks passed
Max191 added a commit that referenced this pull request Oct 7, 2024
@Max191
[GPU][DT] Add e2e matmul tests for GPU data tiling (#18627)
23b63cd 
This adds some matmul e2e tests for GPU data tiling on `gfx94*` chips.

This depends on:
#18626
#18623
#18624
#18625

Signed-off-by: Max Dawkins <max.dawkins@gmail.com>
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Reviewers

@qedawkins qedawkins qedawkins left review comments

@bjacob bjacob bjacob approved these changes

@MaheshRavishankar MaheshRavishankar Awaiting requested review from MaheshRavishankar MaheshRavishankar is a code owner

@kuhar kuhar Awaiting requested review from kuhar kuhar is a code owner

@Groverkss Groverkss Awaiting requested review from Groverkss Groverkss is a code owner

@antiagainst antiagainst Awaiting requested review from antiagainst antiagainst is a code owner

@hanhanW hanhanW Awaiting requested review from hanhanW

Assignees
No one assigned
Labels
None yet
Projects
None yet
Milestone
No milestone
Development

Successfully merging this pull request may close these issues.
3 participants
@Max191 @bjacob @qedawkins