[CPU] Improve INT8 SDPA template #3230
Merged
+27
−27
Conversation
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
🔗 Helpful Links🧪 See artifacts and rendered test results at hud.pytorch.org/pr/pytorch/ao/3230
Note: Links to docs will display an error until the docs builds have been completed. ✅ No FailuresAs of commit 64eae18 with merge base f3fc5e7 ( This comment was automatically generated by Dr. CI and updates every 15 minutes. |
meta-cla
bot
added
the
CLA Signed
Xia-Weiwen
added
the
topic: not user facing
| auto tmp2 = tmp1.round(); | ||
| auto tmp3 = tmp2 + vec_beta1; | ||
| auto tmp1 = at::vec::fmadd(tmp0, vec_sum_scale, vec_beta1); | ||
| auto tmp3 = tmp1.round(); |
There was a problem hiding this comment.
Could we also apply the optimization to the below masked vectorization part?
There was a problem hiding this comment.
Updated. Thanks.
| auto tmp6 = tmp5.round(); | ||
| auto tmp7 = tmp6 + vec_beta2; | ||
| auto tmp5 = at::vec::fmadd(tmp4, vec_alpha, vec_beta2); | ||
| auto tmp7 = tmp5.round(); |
There was a problem hiding this comment.
Updated. Thanks.
| auto tmp6 = tmp5.round(); | ||
| auto tmp7 = tmp6 + vec_beta2; | ||
| auto tmp5 = at::vec::fmadd(tmp4, vec_alpha, vec_beta2); | ||
| auto tmp7 = tmp5.round(); |
There was a problem hiding this comment.
Updated. Thanks.
|
CC @mingfeima for review. Thanks. |
mingfeima
approved these changes
Oct 27, 2025
Valentine233
approved these changes
Oct 27, 2025
|
LG, can you include any benchmarking numbers in the summary? |
jerryzh168
approved these changes
Oct 28, 2025
Thanks. I have added numbers. |
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
4 participants
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
It brings about 1% E2E improvement when running int8 VIT on 4 cores.
Benchmark results of int8 ViT
(Tested on an Intel(R) Xeon(R) 6972P machine with 4 cores/instance, totally 15 instances on one socket)