[cpu] Modify inductor opt flag --- ftree-loop-vectorize#121782
Closed
Valentine233 wants to merge 6 commits intomainfrom
Closed
[cpu] Modify inductor opt flag --- ftree-loop-vectorize#121782Valentine233 wants to merge 6 commits intomainfrom
Valentine233 wants to merge 6 commits intomainfrom
Conversation
🔗 Helpful Links🧪 See artifacts and rendered test results at hud.pytorch.org/pr/121782
Note: Links to docs will display an error until the docs builds have been completed. ❌ 1 New Failure, 1 Unrelated FailureAs of commit ae50a74 with merge base e7141d1 ( NEW FAILURE - The following job has failed:
BROKEN TRUNK - The following job failed but were present on the merge base:👉 Rebase onto the `viable/strict` branch to avoid these failures
This comment was automatically generated by Dr. CI and updates every 15 minutes. |
jgong5
reviewed
Mar 13, 2024
Collaborator
jgong5
left a comment
There was a problem hiding this comment.
Please share performance numbers to make sure there is no regression.
Comment on lines
+1250
to
+1263
| if not config.cpp.enable_tree_loop_vec_opt_flag: | ||
| base_flags += " -fno-tree-loop-vectorize" |
Collaborator
There was a problem hiding this comment.
Please add the issue links as comment to explain why we have to disable this by default.
Collaborator
Author
There was a problem hiding this comment.
Thanks and added!
Valentine233
commented
Mar 18, 2024
| resnet50,float32,dynamic,default,2.28794694 | ||
| timm_efficientnet,float32,static,cpp,2.72195686 | ||
| mobilenet_v3_large,float32,static,cpp,3.02274304 | ||
| mobilenet_v3_large,float32,static,cpp,2.9000000 |
Collaborator
Author
There was a problem hiding this comment.
To pass CI, modify the speedup. According to the validation, this model doesn't have a perf regression.
3612edd to
ae50a74
Compare
Collaborator
|
@Valentine233 please help to check if whether we can enable the vectorization for the regression models. |
Collaborator
Author
Updated in PR description. |
Collaborator
Author
|
We may wait for all the regressions fixed, if they could be solved in short term. |
Contributor
|
Looks like this PR hasn't been updated in a while so we're going to go ahead and mark this as |
pytorchmergebot
pushed a commit
that referenced
this pull request
Nov 7, 2024
Reopen #121782, as more optimizations have landed. Fixes #115261, #113017. For CPU inductor path, remove -ftree-loop-vectorize from optimization flags to fix functional issues. ### Validation on 3 benchmark suites #### FP32  Outlier models (speedup<0.8, single socket): None. #### BF16  Outlier models (speedup<0.8, single socket multi threads): - functorch_dp_cifar10 0.58 - opacus_cifar10 0.57 Pull Request resolved: #136827 Approved by: https://github.com/jansel, https://github.com/jgong5
pytorchmergebot
pushed a commit
that referenced
this pull request
Nov 12, 2024
Reopen #121782, as more optimizations have landed. Fixes #115261, #113017. For CPU inductor path, remove -ftree-loop-vectorize from optimization flags to fix functional issues. ### Validation on 3 benchmark suites #### FP32  Outlier models (speedup<0.8, single socket): None. #### BF16  Outlier models (speedup<0.8, single socket multi threads): - functorch_dp_cifar10 0.58 - opacus_cifar10 0.57 Pull Request resolved: #136827 Approved by: https://github.com/jansel, https://github.com/jgong5
pobin6
pushed a commit
to pobin6/pytorch
that referenced
this pull request
Dec 5, 2024
Reopen pytorch#121782, as more optimizations have landed. Fixes pytorch#115261, pytorch#113017. For CPU inductor path, remove -ftree-loop-vectorize from optimization flags to fix functional issues. ### Validation on 3 benchmark suites #### FP32  Outlier models (speedup<0.8, single socket): None. #### BF16  Outlier models (speedup<0.8, single socket multi threads): - functorch_dp_cifar10 0.58 - opacus_cifar10 0.57 Pull Request resolved: pytorch#136827 Approved by: https://github.com/jansel, https://github.com/jgong5
pobin6
pushed a commit
to pobin6/pytorch
that referenced
this pull request
Dec 5, 2024
Reopen pytorch#121782, as more optimizations have landed. Fixes pytorch#115261, pytorch#113017. For CPU inductor path, remove -ftree-loop-vectorize from optimization flags to fix functional issues. ### Validation on 3 benchmark suites #### FP32  Outlier models (speedup<0.8, single socket): None. #### BF16  Outlier models (speedup<0.8, single socket multi threads): - functorch_dp_cifar10 0.58 - opacus_cifar10 0.57 Pull Request resolved: pytorch#136827 Approved by: https://github.com/jansel, https://github.com/jgong5
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
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.
Fixes #115261, #113017.
For CPU inductor path, remove
-ftree-loop-vectorizefrom optimization flags to fix functional issues.Validation on 3 benchmark suites
FP32
Outlier models (speedup<0.8, single socket):
atomic_add(scatter_add) @CaoEindex_expr(batch_norm)Expected to be fixed by [inductor][cpp] complete vectorization for int32/int64 #122961
BF16
Outlier models (speedup<0.8, single socket):
atomic_add(scatter_add) @CaoEindex_expr(batch_norm)Expected to be fixed by [inductor][cpp] complete vectorization for int32/int64 #122961
cc @voznesenskym @penguinwu @EikanWang @jgong5 @Guobing-Chen @XiaobingSuper @zhuhaozhe @blzheng @wenzhe-nrv @jiayisunx @peterbell10 @ipiszy @yf225 @chenyang78 @kadeng @muchulee8 @aakhundov @ColinPeppler @amjames @desertfire @chauhang