[AOTI] Support MKLDNN conv ops in cpp wrapper #134475

desertfire · 2024-08-26T16:25:05Z

Stack from ghstack (oldest at bottom):

Summary: Partially fix #123040. In the ABI-compatible mode, MKLDNN fallback ops do not have C shim implementations and thus need to go through the custom ops launch path. Other MLKDNN ops will be fixed in following PRs.

cc @voznesenskym @penguinwu @EikanWang @jgong5 @Guobing-Chen @XiaobingSuper @zhuhaozhe @blzheng @wenzhe-nrv @jiayisunx @ipiszy @yf225 @chenyang78 @kadeng @muchulee8 @ColinPeppler @amjames @chauhang

Summary: Partially fix #123040. In the ABI-compatible mode, MKLDNN fallback ops do not have C shim implementations and thus need to go through the custom ops launch path. Other MLKDNN ops will be fixed in following PRs. [ghstack-poisoned]

pytorch-bot · 2024-08-26T16:25:08Z

🔗 Helpful Links

🧪 See artifacts and rendered test results at hud.pytorch.org/pr/134475

📄 Preview Python docs built from this PR
📄 Preview C++ docs built from this PR
❓ Need help or want to give feedback on the CI? Visit the bot commands wiki or our office hours

Note: Links to docs will display an error until the docs builds have been completed.

✅ No Failures

As of commit b2bb636 with merge base c7328df ():
💚 Looks good so far! There are no failures yet. 💚

This comment was automatically generated by Dr. CI and updates every 15 minutes.

Summary: Partially fix #123040. In the ABI-compatible mode, MKLDNN fallback ops do not have C shim implementations and thus need to go through the custom ops launch path. Other MLKDNN ops will be fixed in following PRs. ghstack-source-id: b61bd38 Pull Request resolved: #134475

Summary: Partially fix #123040. In the ABI-compatible mode, MKLDNN fallback ops do not have C shim implementations and thus need to go through the custom ops launch path. Other MLKDNN ops will be fixed in following PRs. cc voznesenskym penguinwu EikanWang jgong5 Guobing-Chen XiaobingSuper zhuhaozhe blzheng wenzhe-nrv jiayisunx ipiszy yf225 chenyang78 kadeng muchulee8 ColinPeppler amjames chauhang [ghstack-poisoned]

Summary: Partially fix #123040. In the ABI-compatible mode, MKLDNN fallback ops do not have C shim implementations and thus need to go through the custom ops launch path. Other MLKDNN ops will be fixed in following PRs. ghstack-source-id: 24578c1 Pull Request resolved: #134475

Summary: Partially fix #123040. In the ABI-compatible mode, MKLDNN fallback ops do not have C shim implementations and thus need to go through the custom ops launch path. Other MLKDNN ops will be fixed in following PRs. cc voznesenskym penguinwu EikanWang jgong5 Guobing-Chen XiaobingSuper zhuhaozhe blzheng wenzhe-nrv jiayisunx ipiszy yf225 chenyang78 kadeng muchulee8 ColinPeppler amjames chauhang [ghstack-poisoned]

leslie-fang-intel

Thanks for the PR. The changes of Mkldnn_ir part looks good me. @chunyuan-w will take a look for the changes of cpp_wrapper.

Summary: Partially fix #123040. In the ABI-compatible mode, MKLDNN fallback ops do not have C shim implementations and thus need to go through the custom ops launch path. Other MLKDNN ops will be fixed in following PRs. cc voznesenskym penguinwu EikanWang jgong5 Guobing-Chen XiaobingSuper zhuhaozhe blzheng wenzhe-nrv jiayisunx ipiszy yf225 chenyang78 kadeng muchulee8 ColinPeppler amjames chauhang [ghstack-poisoned]

Summary: Partially fix #123040. In the ABI-compatible mode, MKLDNN fallback ops do not have C shim implementations and thus need to go through the custom ops launch path. Other MLKDNN ops will be fixed in following PRs. ghstack-source-id: 0bf4b6d Pull Request resolved: #134475

chunyuan-w · 2024-08-29T12:41:04Z

Just to confirm, do we need to add C shim implementations for MKLDNN OPs as well in the future?

desertfire · 2024-08-29T12:45:48Z

Just to confirm, do we need to add C shim implementations for MKLDNN OPs as well in the future?

You can do that in the future to further improve the performance. My current plan is to clear the remaining ABI-compatible test issues and make ABI-compatible default afterwards.

chunyuan-w · 2024-08-29T12:49:24Z

This case might work as well with the fix in this PR:

pytorch/test/inductor/test_aot_inductor.py

Lines 3462 to 3464 in d01a7a9

    
           # TODO: test_conv_freezing_abi_compatible_cpu fails, 
        
           #   AssertionError: None, i.e. optional output is not supported 
        
           "test_conv_freezing": fail_with_and_without_stack_allocation(is_skip=True),

Summary: Similar to #134475, support linear and qlinear in the ABI-compatible mode for cpp-wrapper Inductor. [ghstack-poisoned]

Summary: Similar to #134475, support linear and qlinear in the ABI-compatible mode for cpp-wrapper Inductor. cc voznesenskym penguinwu EikanWang jgong5 Guobing-Chen XiaobingSuper zhuhaozhe blzheng wenzhe-nrv jiayisunx ipiszy yf225 chenyang78 kadeng muchulee8 ColinPeppler amjames chauhang [ghstack-poisoned]

Summary: Similar to #134475, support qlinear in the ABI-compatible mode for cpp-wrapper Inductor. ghstack-source-id: e41145c Pull Request resolved: #134783

Summary: Similar to #134475, support qconv in the ABI-compatible mode for cpp-wrapper Inductor. [ghstack-poisoned]

Summary: Similar to #134475, support qconv in the ABI-compatible mode for cpp-wrapper Inductor. ghstack-source-id: eada10b Pull Request resolved: #134795

Summary: Similar to #134475, support mkl linear in the ABI-compatible mode for cpp-wrapper Inductor. [ghstack-poisoned]

[ghstack-poisoned]

Summary: Similar to #134475, support mkl linear in the ABI-compatible mode for cpp-wrapper Inductor. ghstack-source-id: 06a726c Pull Request resolved: #134974

Summary: Similar to #134475, support qlinear in the ABI-compatible mode for cpp-wrapper Inductor. Pull Request resolved: #134783 Approved by: https://github.com/leslie-fang-intel, https://github.com/chunyuan-w, https://github.com/angelayi ghstack dependencies: #134475

Summary: Similar to #134475, support qconv in the ABI-compatible mode for cpp-wrapper Inductor. Pull Request resolved: #134795 Approved by: https://github.com/leslie-fang-intel, https://github.com/chunyuan-w, https://github.com/angelayi ghstack dependencies: #134475, #134783

Summary: Similar to #134475, support mkl linear in the ABI-compatible mode for cpp-wrapper Inductor. ghstack-source-id: 8feb940 Pull Request resolved: #134974

Summary: Similar to pytorch/pytorch#134475, support qconv in the ABI-compatible mode for cpp-wrapper Inductor. ghstack-source-id: 6121d59 Pull Request resolved: pytorch/pytorch#134795

Summary: Partially fix pytorch#123040. In the ABI-compatible mode, MKLDNN fallback ops do not have C shim implementations and thus need to go through the custom ops launch path. Other MLKDNN ops will be fixed in following PRs. Pull Request resolved: pytorch#134475 Approved by: https://github.com/leslie-fang-intel, https://github.com/chunyuan-w, https://github.com/angelayi

Summary: Similar to pytorch#134475, support qlinear in the ABI-compatible mode for cpp-wrapper Inductor. Pull Request resolved: pytorch#134783 Approved by: https://github.com/leslie-fang-intel, https://github.com/chunyuan-w, https://github.com/angelayi ghstack dependencies: pytorch#134475

Summary: Similar to pytorch#134475, support qconv in the ABI-compatible mode for cpp-wrapper Inductor. Pull Request resolved: pytorch#134795 Approved by: https://github.com/leslie-fang-intel, https://github.com/chunyuan-w, https://github.com/angelayi ghstack dependencies: pytorch#134475, pytorch#134783

Summary: Similar to #134475, support mkl linear in the ABI-compatible mode for cpp-wrapper Inductor. [ghstack-poisoned]

Summary: Similar to #134475, support mkl linear in the ABI-compatible mode for cpp-wrapper Inductor. ghstack-source-id: 1e8479b Pull Request resolved: #134974

Summary: Similar to #134475, support mkl linear in the ABI-compatible mode for cpp-wrapper Inductor. [ghstack-poisoned]

Summary: Similar to #134475, support mkl linear in the ABI-compatible mode for cpp-wrapper Inductor. ghstack-source-id: 277ea5e Pull Request resolved: #134974

Summary: Similar to #134475, support mkl linear in the ABI-compatible mode for cpp-wrapper Inductor. Differential Revision: [D63322202](https://our.internmc.facebook.com/intern/diff/D63322202) Pull Request resolved: #134974 Approved by: https://github.com/chenyang78, https://github.com/leslie-fang-intel Co-authored-by: leslie-fang-intel <leslie.fang@intel.com>

Summary: Similar to pytorch#134475, support mkl linear in the ABI-compatible mode for cpp-wrapper Inductor. Differential Revision: [D63322202](https://our.internmc.facebook.com/intern/diff/D63322202) Pull Request resolved: pytorch#134974 Approved by: https://github.com/chenyang78, https://github.com/leslie-fang-intel Co-authored-by: leslie-fang-intel <leslie.fang@intel.com>

Summary: Similar to pytorch/pytorch#134475, support mkl linear in the ABI-compatible mode for cpp-wrapper Inductor. ghstack-source-id: 277ea5e Pull Request resolved: pytorch/pytorch#134974

[AOTI] Support MKLDNN conv ops in cpp wrapper

9bb0fb7

Summary: Partially fix #123040. In the ABI-compatible mode, MKLDNN fallback ops do not have C shim implementations and thus need to go through the custom ops launch path. Other MLKDNN ops will be fixed in following PRs. [ghstack-poisoned]

pytorch-bot bot added ciflow/inductor module: inductor labels Aug 26, 2024

desertfire requested review from angelayi, chunyuan-w and leslie-fang-intel August 27, 2024 14:37

leslie-fang-intel approved these changes Aug 29, 2024

View reviewed changes

chunyuan-w approved these changes Aug 29, 2024

View reviewed changes

desertfire added a commit that referenced this pull request Aug 29, 2024

[AOTI] Support MKLDNN linear and qlinear ops

0c7a202

Summary: Similar to #134475, support linear and qlinear in the ABI-compatible mode for cpp-wrapper Inductor. [ghstack-poisoned]

desertfire mentioned this pull request Aug 29, 2024

[AOTI] Support MKLDNN qlinear ops in cpp wrapper #134783

Closed

desertfire added a commit that referenced this pull request Aug 29, 2024

[AOTI] Support MKLDNN qlinear ops

85bf73e

Summary: Similar to #134475, support qlinear in the ABI-compatible mode for cpp-wrapper Inductor. ghstack-source-id: e41145c Pull Request resolved: #134783

desertfire added a commit that referenced this pull request Aug 29, 2024

[AOTI] Support MKLDNN qconv ops

fc89a89

Summary: Similar to #134475, support qconv in the ABI-compatible mode for cpp-wrapper Inductor. [ghstack-poisoned]

desertfire mentioned this pull request Aug 29, 2024

[AOTI] Support MKLDNN qconv ops in cpp wrapper #134795

Closed

desertfire added a commit that referenced this pull request Aug 29, 2024

[AOTI] Support MKLDNN qconv ops

8be35ea

Summary: Similar to #134475, support qconv in the ABI-compatible mode for cpp-wrapper Inductor. ghstack-source-id: eada10b Pull Request resolved: #134795

desertfire added the topic: not user facing topic category label Aug 29, 2024

desertfire mentioned this pull request Aug 30, 2024

[AOTI] Add C shim for aten.mkldnn_rnn_layer in cpp wrapper #134857

Closed

desertfire added a commit that referenced this pull request Sep 3, 2024

[AOTI] Support MKL linear ops

4c01dd6

Summary: Similar to #134475, support mkl linear in the ABI-compatible mode for cpp-wrapper Inductor. [ghstack-poisoned]

Update

b2bb636

[ghstack-poisoned]

desertfire added a commit that referenced this pull request Sep 5, 2024

[AOTI] Support MKL linear ops

b6e30e3

Summary: Similar to #134475, support mkl linear in the ABI-compatible mode for cpp-wrapper Inductor. ghstack-source-id: 06a726c Pull Request resolved: #134974

pytorchmergebot added the Merged label Sep 6, 2024

pytorchmergebot closed this in 0b96dfb Sep 6, 2024

desertfire added a commit that referenced this pull request Sep 9, 2024

[AOTI] Support MKL linear ops

6809c8d

Summary: Similar to #134475, support mkl linear in the ABI-compatible mode for cpp-wrapper Inductor. ghstack-source-id: 8feb940 Pull Request resolved: #134974

desertfire added a commit that referenced this pull request Sep 23, 2024

Update base for Update on "[AOTI] Support MKL linear ops in cpp wrapper"

5453e41

Summary: Similar to #134475, support mkl linear in the ABI-compatible mode for cpp-wrapper Inductor. [ghstack-poisoned]

desertfire added a commit that referenced this pull request Sep 23, 2024

Update on "[AOTI] Support MKL linear ops in cpp wrapper"

7792ea8

Summary: Similar to #134475, support mkl linear in the ABI-compatible mode for cpp-wrapper Inductor. [ghstack-poisoned]

desertfire added a commit that referenced this pull request Sep 23, 2024

Update base for Update on "[AOTI] Support MKL linear ops in cpp wrapper"

a092309

Summary: Similar to #134475, support mkl linear in the ABI-compatible mode for cpp-wrapper Inductor. [ghstack-poisoned]

desertfire added a commit that referenced this pull request Sep 23, 2024

Update on "[AOTI] Support MKL linear ops in cpp wrapper"

83a3160

Summary: Similar to #134475, support mkl linear in the ABI-compatible mode for cpp-wrapper Inductor. [ghstack-poisoned]

desertfire added a commit that referenced this pull request Sep 23, 2024

[AOTI] Support MKL linear ops

d93165a

Summary: Similar to #134475, support mkl linear in the ABI-compatible mode for cpp-wrapper Inductor. ghstack-source-id: 1e8479b Pull Request resolved: #134974

desertfire added a commit that referenced this pull request Sep 23, 2024

Update base for Update on "[AOTI] Support MKL linear ops in cpp wrapper"

665e60b

Summary: Similar to #134475, support mkl linear in the ABI-compatible mode for cpp-wrapper Inductor. [ghstack-poisoned]

desertfire added a commit that referenced this pull request Sep 23, 2024

Update on "[AOTI] Support MKL linear ops in cpp wrapper"

28bbd8f

Summary: Similar to #134475, support mkl linear in the ABI-compatible mode for cpp-wrapper Inductor. [ghstack-poisoned]

leslie-fang-intel pushed a commit that referenced this pull request Sep 24, 2024

[AOTI] Support MKL linear ops

88f5063

Summary: Similar to #134475, support mkl linear in the ABI-compatible mode for cpp-wrapper Inductor. ghstack-source-id: 277ea5e Pull Request resolved: #134974

github-actions bot deleted the gh/desertfire/451/head branch October 6, 2024 02:10

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Uh oh!

[AOTI] Support MKLDNN conv ops in cpp wrapper #134475

[AOTI] Support MKLDNN conv ops in cpp wrapper #134475

Uh oh!

desertfire commented Aug 26, 2024 •

edited by pytorch-bot bot

Loading

Uh oh!

pytorch-bot bot commented Aug 26, 2024 •

edited

Loading

Uh oh!

leslie-fang-intel left a comment

Uh oh!

chunyuan-w commented Aug 29, 2024

Uh oh!

desertfire commented Aug 29, 2024

Uh oh!

chunyuan-w commented Aug 29, 2024

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

5 participants

[AOTI] Support MKLDNN conv ops in cpp wrapper #134475

[AOTI] Support MKLDNN conv ops in cpp wrapper #134475

Uh oh!

Conversation

desertfire commented Aug 26, 2024 • edited by pytorch-bot bot Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

pytorch-bot bot commented Aug 26, 2024 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

🔗 Helpful Links

🧪 See artifacts and rendered test results at hud.pytorch.org/pr/134475

✅ No Failures

Uh oh!

leslie-fang-intel left a comment

Choose a reason for hiding this comment

Uh oh!

chunyuan-w commented Aug 29, 2024

Uh oh!

desertfire commented Aug 29, 2024

Uh oh!

chunyuan-w commented Aug 29, 2024

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

5 participants

desertfire commented Aug 26, 2024 •

edited by pytorch-bot bot

Loading

pytorch-bot bot commented Aug 26, 2024 •

edited

Loading