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

feat: add support for gfx1100 devices #2101

Merged
merged 1 commit into from
Jun 12, 2023
Merged

feat: add support for gfx1100 devices #2101

merged 1 commit into from
Jun 12, 2023

Conversation

evshiron
Copy link

Thread ref: ROCm/ROCm#1880

@jayfurmanek jayfurmanek self-requested a review May 15, 2023 21:26
cromefire
cromefire reviewed May 15, 2023
View reviewed changes
tensorflow/compiler/xla/stream_executor/device_description.h
@@ -188,7 +189,7 @@ class RocmComputeCapability {
return {"gfx908", "gfx90a"};
Copy link

@cromefire cromefire May 15, 2023 •
edited
Loading

Choose a reason for hiding this comment

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

BF16 should also work for RDNA 3 according to AMD's documentation, so it should be fine to be enabled. It did work for me to add gfx1100 there, but I'm not sure how to properly test that it's really fine (it didn't make any difference performance wise in my testing).

image

@jayfurmanek
Copy link

jayfurmanek commented May 15, 2023 •
edited
Loading

Hi, Thanks for the PR!
I'll review. There are likely a few more places we'll need to specify the new arch.
It also may be a bit before we can merge as there are some needed miopen updates.

@bondhugula bondhugula mentioned this pull request Jun 4, 2023
Open
jayfurmanek
jayfurmanek approved these changes Jun 12, 2023
View reviewed changes
Copy link

@jayfurmanek jayfurmanek left a comment

Choose a reason for hiding this comment

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

Thanks so much @evshiron! I'll follow up with another PR for a missing piece in XLA and miopen.

@jayfurmanek jayfurmanek merged commit 946fd78 into ROCm:develop-upstream Jun 12, 2023
@evshiron
Copy link
Author

evshiron commented Aug 4, 2023 •
edited
Loading

@jayfurmanek

Greetings. Sorry to bother you, but the wheels from CI doesn't work as indicated here:

And I would like to know when tensorflow-rocm will ship with Navi 31 support?

@evshiron evshiron mentioned this pull request Aug 13, 2023
Open
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Reviewers

@cromefire cromefire cromefire left review comments

@jayfurmanek jayfurmanek jayfurmanek 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.
3 participants
@evshiron @jayfurmanek @cromefire