Lightweight at-most-once logging for API usage #20745
Closed
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
pytorchbot
added
caffe2
oncall: jit
Contributor
facebook-github-bot
left a comment
facebook-github-bot
left a comment
There was a problem hiding this comment.
@dzhulgakov has imported this pull request. If you are a Facebook employee, you can view this diff on Phabricator.
Contributor
|
Is there a way to ensure that we continue not to trigger these trigger points via static initialization? Seems like something subtle that's easy to break. |
ezyang
reviewed
May 21, 2019
c10/util/Logging.cpp
Outdated
ezyang
reviewed
May 21, 2019
torch/csrc/Module.cpp
Outdated
Collaborator
Author
There was a problem hiding this comment.
it's called from python, so it should be ok :) I'm putting it only in construction of classes, so the impact should be minimal
ezyang
approved these changes
May 21, 2019
Contributor
ezyang
left a comment
ezyang
left a comment
There was a problem hiding this comment.
Implementation seems sound
Collaborator
Author
|
Added a simple python test to ensure that we don't add accidentally some static initialization logging (it's a pretty hacky test but it works - also verified that it fails if such logging is actually added) |
Contributor
facebook-github-bot
left a comment
facebook-github-bot
left a comment
There was a problem hiding this comment.
@dzhulgakov has imported this pull request. If you are a Facebook employee, you can view this diff on Phabricator.
dzhulgakov
force-pushed
the
api-metrics
branch
2 times, most recently
from
4702df8 to
d155af7
Compare
Contributor
facebook-github-bot
left a comment
facebook-github-bot
left a comment
There was a problem hiding this comment.
@dzhulgakov has imported this pull request. If you are a Facebook employee, you can view this diff on Phabricator.
dzhulgakov
force-pushed
the
api-metrics
branch
2 times, most recently
from
e63fc5c to
9e81649
Compare
Contributor
facebook-github-bot
left a comment
facebook-github-bot
left a comment
There was a problem hiding this comment.
@dzhulgakov has imported this pull request. If you are a Facebook employee, you can view this diff on Phabricator.
Contributor
facebook-github-bot
left a comment
facebook-github-bot
left a comment
There was a problem hiding this comment.
@dzhulgakov has imported this pull request. If you are a Facebook employee, you can view this diff on Phabricator.
zdevito
pushed a commit
to zdevito/ATen
that referenced
this pull request
May 24, 2019
Summary: Resubmit #20698 which got messed up. Idea is that when PyTorch is used in a custom build environment (e.g. Facebook), it's useful to track usage of various APIs centrally. This PR introduces a simple very lightweight mechanism to do so - only first invocation of a trigger point would be logged. This is significantly more lightweight than #18235 and thus we can allow to put logging in e.g. TensorImpl. Also adds an initial list of trigger points. Trigger points are added in such a way that no static initialization triggers them, i.e. just linking with libtorch.so will not cause any logging. Further suggestions of what to log are welcomed. Pull Request resolved: pytorch/pytorch#20745 Differential Revision: D15429196 Pulled By: dzhulgakov fbshipit-source-id: a5e41a709a65b7ebccc6b95f93854e583cf20aca
Contributor
|
@dzhulgakov merged this pull request in c25e337. |
Contributor
|
ASAN started failing after this PR merged but I am not sure if this diff could have actually caused it: |
ezyang
added a commit
that referenced
this pull request
May 24, 2019
This reverts commit c25e337. gh-metadata: pytorch pytorch 20906 gh/ezyang/139/head
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
5 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.
Resubmit #20698 which got messed up.
Idea is that when PyTorch is used in a custom build environment (e.g. Facebook), it's useful to track usage of various APIs centrally. This PR introduces a simple very lightweight mechanism to do so - only first invocation of a trigger point would be logged. This is significantly more lightweight than #18235 and thus we can allow to put logging in e.g. TensorImpl.
Also adds an initial list of trigger points. Trigger points are added in such a way that no static initialization triggers them, i.e. just linking with libtorch.so will not cause any logging. Further suggestions of what to log are welcomed.
Test plan:
Used PYTORCH_API_USAGE_STDERR=1 env with various scenarios, verified that logging is indeed triggered.
Given the only-once nature of logging, I'm not sure adding unittest would be that beneficial as it might be impact by how multiple unittests are linked together in one binary.
Added a tiny benchmark to verify overhead. Since we're talking about static variable only, it seems to be 0.3ns with warm cache (note it's 1000 invocations reported below). Didn't test with cold cache as it's pretty hard to capture such a small overhead: