[PyTorch] Use real if constexpr behind macro in hot template #51368
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
This seems to noticably reduce build times, at least for RegisterCPU.cpp. It makes sense that a compiler builtin would be faster than simulating the same builtin with templates. Identified with templight. Differential Revision: [D26154964](https://our.internmc.facebook.com/intern/diff/D26154964/) [ghstack-poisoned]
💊 CI failures summary and remediationsAs of commit f309519 (more details on the Dr. CI page):
Extra GitHub checks: 1 failed
This comment was automatically generated by Dr. CI (expand for details).Follow this link to opt-out of these comments for your Pull Requests.Please report bugs/suggestions to the (internal) Dr. CI Users group. |
This was referenced Jan 29, 2021
This seems to noticably reduce build times, at least for RegisterCPU.cpp. It makes sense that a compiler builtin would be faster than simulating the same builtin with templates. Identified with templight. Differential Revision: [D26154964](https://our.internmc.facebook.com/intern/diff/D26154964/) [ghstack-poisoned]
swolchok
added a commit
that referenced
this pull request
Jan 29, 2021
Pull Request resolved: #51368 This seems to noticably reduce build times, at least for RegisterCPU.cpp. It makes sense that a compiler builtin would be faster than simulating the same builtin with templates. Identified with templight. ghstack-source-id: 120697490 Differential Revision: [D26154964](https://our.internmc.facebook.com/intern/diff/D26154964/)
This seems to noticably reduce build times, at least for RegisterCPU.cpp. It makes sense that a compiler builtin would be faster than simulating the same builtin with templates. Identified with templight. Differential Revision: [D26154964](https://our.internmc.facebook.com/intern/diff/D26154964/) [ghstack-poisoned]
swolchok
added a commit
that referenced
this pull request
Feb 1, 2021
Pull Request resolved: #51368 This seems to noticably reduce build times, at least for RegisterCPU.cpp. It makes sense that a compiler builtin would be faster than simulating the same builtin with templates. Identified with templight. ghstack-source-id: 120761381 Differential Revision: [D26154964](https://our.internmc.facebook.com/intern/diff/D26154964/)
This seems to noticably reduce build times, at least for RegisterCPU.cpp. It makes sense that a compiler builtin would be faster than simulating the same builtin with templates. Identified with templight. Differential Revision: [D26154964](https://our.internmc.facebook.com/intern/diff/D26154964/) [ghstack-poisoned]
swolchok
added a commit
that referenced
this pull request
Feb 8, 2021
Pull Request resolved: #51368 This seems to noticably reduce build times, at least for RegisterCPU.cpp. It makes sense that a compiler builtin would be faster than simulating the same builtin with templates. Identified with templight. ghstack-source-id: 121256867 Differential Revision: [D26154964](https://our.internmc.facebook.com/intern/diff/D26154964/)
This seems to noticably reduce build times, at least for RegisterCPU.cpp. It makes sense that a compiler builtin would be faster than simulating the same builtin with templates. Identified with templight. Differential Revision: [D26154964](https://our.internmc.facebook.com/intern/diff/D26154964/) [ghstack-poisoned]
swolchok
added a commit
that referenced
this pull request
Feb 9, 2021
Pull Request resolved: #51368 This seems to noticably reduce build times, at least for RegisterCPU.cpp. It makes sense that a compiler builtin would be faster than simulating the same builtin with templates. Identified with templight. ghstack-source-id: 121300917 Differential Revision: [D26154964](https://our.internmc.facebook.com/intern/diff/D26154964/)
This seems to noticably reduce build times, at least for RegisterCPU.cpp. It makes sense that a compiler builtin would be faster than simulating the same builtin with templates. Identified with templight. Differential Revision: [D26154964](https://our.internmc.facebook.com/intern/diff/D26154964/) [ghstack-poisoned]
swolchok
added a commit
that referenced
this pull request
Feb 9, 2021
Pull Request resolved: #51368 This seems to noticably reduce build times, at least for RegisterCPU.cpp. It makes sense that a compiler builtin would be faster than simulating the same builtin with templates. Identified with templight. ghstack-source-id: 121363469 Differential Revision: [D26154964](https://our.internmc.facebook.com/intern/diff/D26154964/)
…hind macro in hot template" This seems to noticably reduce build times, at least for RegisterCPU.cpp. It makes sense that a compiler builtin would be faster than simulating the same builtin with templates. Identified with templight. Differential Revision: [D26154964](https://our.internmc.facebook.com/intern/diff/D26154964/) [ghstack-poisoned]
swolchok
added a commit
that referenced
this pull request
Feb 9, 2021
Pull Request resolved: #51368 This seems to noticably reduce build times, at least for RegisterCPU.cpp. It makes sense that a compiler builtin would be faster than simulating the same builtin with templates. Identified with templight. ghstack-source-id: 121378959 Differential Revision: [D26154964](https://our.internmc.facebook.com/intern/diff/D26154964/)
|
Do we have any data on the perf benefit from this? I can believe the lambdas in the polyfill API add overhead (or inline disruption risk, etc.) over the native feature, but given that we'll want to use this kind of surface-level #ifdef sparingly for readability's sake, it'd be good to establish a ballpark quantitative threshold for when it's justified. |
bhosmer
approved these changes
Feb 14, 2021
|
@bhosmer note that IIUC, Ubuntu 16.04 LTS is the blocker for raising our minimum C++ version to C++17. It hits end of life in April, so we should be able to move to C++17 and use |
Yup! @smessmer actually has the move to C++17 on his H1 roadmap, so you're right, it'll clean up pretty quick. |
|
This pull request has been merged in 6e1a5b1. |
xsacha
pushed a commit
to xsacha/pytorch
that referenced
this pull request
Mar 31, 2021
…#51368) Summary: Pull Request resolved: pytorch#51368 This seems to noticably reduce build times, at least for RegisterCPU.cpp. It makes sense that a compiler builtin would be faster than simulating the same builtin with templates. Identified with templight. ghstack-source-id: 121378959 Test Plan: Confirmed this speeds up RegisterCPU.cpp optimized build by simply running builds under `time(1)`: previous diff: [50.53, 50.41, 50.57, 50.67, 50.94] mean: 50.6 std: 0.179 this diff: [45.71, 45.89, 46.21, 48.51, 45.84] mean: 46.4 std: 1.05 Reviewed By: bhosmer Differential Revision: D26154964 fbshipit-source-id: 62ee2f5a872007db032dfebf7ad4d1b6e1ce63d1
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
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.
Stack from ghstack:
This seems to noticably reduce build times, at least for
RegisterCPU.cpp. It makes sense that a compiler builtin would be
faster than simulating the same builtin with templates.
Identified with templight.
Differential Revision: D26154964