Skip to content

Add performance results section to README#15

Merged
sandlbn merged 2 commits intomainfrom
attn
Apr 16, 2026
Merged

Add performance results section to README#15
sandlbn merged 2 commits intomainfrom
attn

Conversation

@sandlbn
Copy link
Copy Markdown
Contributor

@sandlbn sandlbn commented Apr 9, 2026

Showcase FlashAttention and KernelBench L2 results in README #4

@sandlbn sandlbn requested a review from danielfleischer April 9, 2026 14:22
Copy link
Copy Markdown
Member

@danielfleischer danielfleischer left a comment

Choose a reason for hiding this comment

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

Thanks, some remarks about the plots and being specific with the HW, instead of BMG.

Comment thread plots/attention_panel.png Outdated
Copy link
Copy Markdown
Member

Choose a reason for hiding this comment

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

Please add a legend to the right figure as well as it's not clear whether the left one holds as well.

Comment thread plots/l2_roofline_gemm.png Outdated
Copy link
Copy Markdown
Member

Choose a reason for hiding this comment

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

You can focus on the interesting region, expand it a little so we see more; no need to start with x=0.000

Copy link
Copy Markdown
Contributor Author

Choose a reason for hiding this comment

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

It should be better now.

Comment thread plots/l2_roofline_matmul.png Outdated
Copy link
Copy Markdown
Member

Choose a reason for hiding this comment

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

Same: focus on the interesting region, no need to start with x,y = 0,0

Copy link
Copy Markdown
Contributor Author

Choose a reason for hiding this comment

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

Did it better now.

Comment thread README.md Outdated

---

## Performance Results on Intel BMG (FP16)
Copy link
Copy Markdown
Member

Choose a reason for hiding this comment

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

Replace BMG with the specific card we used.

Copy link
Copy Markdown
Contributor Author

Choose a reason for hiding this comment

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

Done

@sandlbn
Copy link
Copy Markdown
Contributor Author

sandlbn commented Apr 15, 2026

Thanks, some remarks about the plots and being specific with the HW, instead of BMG.

Done it’s B70 now

Copy link
Copy Markdown
Member

@danielfleischer danielfleischer left a comment

Choose a reason for hiding this comment

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

Nice improvements.

@sandlbn sandlbn merged commit 07da513 into main Apr 16, 2026
2 checks passed
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