Fix FutureNCCL's completed() disagreeing with wait() #48503
Closed
Conversation
This file contains 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 commit is part of a stack that reworks FutureNCCL in order to extract a generic CUDA-aware Future subclass. The stack deliberately breaks up this transition into elementary changes, to make it easier to verify that the behavior is preserved (or to highlight how it gets changed). --- My impression is that one property of the upstream Future class is that once .wait() returns, or once a callback is invoked, then .completed() should return True. This was not the case for FutureNCCL because .wait() would return immediately, and callbacks would be invoked inline, but .completed() could return False if the CUDA async operations hadn't completed yet. That was odd and confusing. Since there are other ways for users to check the status of CUDA operations (if they really need, and typically I don't think it's so common), perhaps it's best to avoid checking the status of CUDA events in .completed(). Differential Revision: [D25180531](https://our.internmc.facebook.com/intern/diff/D25180531/) [ghstack-poisoned]
lw
requested review from
mingzhe09088,
mrshenli,
pietern,
pritamdamania87,
rohan-varma and
zhaojuanmao
as code owners
facebook-github-bot
added
cla signed
oncall: distributed
This was referenced Nov 26, 2020
💊 CI failures summary and remediationsAs of commit 48673db (more details on the Dr. CI page): 💚 💚 Looks good so far! There are no failures yet. 💚 💚 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 on the GitHub issue tracker or post in the (internal) Dr. CI Users group. This comment has been revised 27 times. |
mrshenli
reviewed
Nov 26, 2020
| // Checking the work's corresponding CUDA events' status | ||
| auto ret = cudaEventQuery((*cudaEvents_)[0]); | ||
| return ret != cudaErrorNotReady || ret == cudaSuccess; | ||
| return true; |
There was a problem hiding this comment.
there is no regression here, and I agree that user code won't trigger any problem. But C++ code can still use the second ctor of FutureNCCL to create an empty future without marking it as completed?
There was a problem hiding this comment.
Yes, correct. As I mentioned earlier, the second constructor is "meant" to be private and only invoked by then(). If you want we can make that explicit. However, in #48505 we'll fix this issue for good by having FutureNCCL share the same logic as ivalue::Future (and thus have a real complete-vs-incomplete status).
This commit is part of a stack that reworks FutureNCCL in order to extract a generic CUDA-aware Future subclass. The stack deliberately breaks up this transition into elementary changes, to make it easier to verify that the behavior is preserved (or to highlight how it gets changed). --- My impression is that one property of the upstream Future class is that once .wait() returns, or once a callback is invoked, then .completed() should return True. This was not the case for FutureNCCL because .wait() would return immediately, and callbacks would be invoked inline, but .completed() could return False if the CUDA async operations hadn't completed yet. That was odd and confusing. Since there are other ways for users to check the status of CUDA operations (if they really need, and typically I don't think it's so common), perhaps it's best to avoid checking the status of CUDA events in .completed(). Differential Revision: [D25180531](https://our.internmc.facebook.com/intern/diff/D25180531/) [ghstack-poisoned]
This commit is part of a stack that reworks FutureNCCL in order to extract a generic CUDA-aware Future subclass. The stack deliberately breaks up this transition into elementary changes, to make it easier to verify that the behavior is preserved (or to highlight how it gets changed). --- My impression is that one property of the upstream Future class is that once .wait() returns, or once a callback is invoked, then .completed() should return True. This was not the case for FutureNCCL because .wait() would return immediately, and callbacks would be invoked inline, but .completed() could return False if the CUDA async operations hadn't completed yet. That was odd and confusing. Since there are other ways for users to check the status of CUDA operations (if they really need, and typically I don't think it's so common), perhaps it's best to avoid checking the status of CUDA events in .completed(). Differential Revision: [D25180531](https://our.internmc.facebook.com/intern/diff/D25180531/) [ghstack-poisoned]
This commit is part of a stack that reworks FutureNCCL in order to extract a generic CUDA-aware Future subclass. The stack deliberately breaks up this transition into elementary changes, to make it easier to verify that the behavior is preserved (or to highlight how it gets changed). --- My impression is that one property of the upstream Future class is that once .wait() returns, or once a callback is invoked, then .completed() should return True. This was not the case for FutureNCCL because .wait() would return immediately, and callbacks would be invoked inline, but .completed() could return False if the CUDA async operations hadn't completed yet. That was odd and confusing. Since there are other ways for users to check the status of CUDA operations (if they really need, and typically I don't think it's so common), perhaps it's best to avoid checking the status of CUDA events in .completed(). Differential Revision: [D25180531](https://our.internmc.facebook.com/intern/diff/D25180531/) [ghstack-poisoned]
This was referenced Nov 29, 2020
This commit is part of a stack that reworks FutureNCCL in order to extract a generic CUDA-aware Future subclass. The stack deliberately breaks up this transition into elementary changes, to make it easier to verify that the behavior is preserved (or to highlight how it gets changed). --- My impression is that one property of the upstream Future class is that once .wait() returns, or once a callback is invoked, then .completed() should return True. This was not the case for FutureNCCL because .wait() would return immediately, and callbacks would be invoked inline, but .completed() could return False if the CUDA async operations hadn't completed yet. That was odd and confusing. Since there are other ways for users to check the status of CUDA operations (if they really need, and typically I don't think it's so common), perhaps it's best to avoid checking the status of CUDA events in .completed(). Differential Revision: [D25180531](https://our.internmc.facebook.com/intern/diff/D25180531/) [ghstack-poisoned]
This commit is part of a stack that reworks FutureNCCL in order to extract a generic CUDA-aware Future subclass. The stack deliberately breaks up this transition into elementary changes, to make it easier to verify that the behavior is preserved (or to highlight how it gets changed). --- My impression is that one property of the upstream Future class is that once .wait() returns, or once a callback is invoked, then .completed() should return True. This was not the case for FutureNCCL because .wait() would return immediately, and callbacks would be invoked inline, but .completed() could return False if the CUDA async operations hadn't completed yet. That was odd and confusing. Since there are other ways for users to check the status of CUDA operations (if they really need, and typically I don't think it's so common), perhaps it's best to avoid checking the status of CUDA events in .completed(). Differential Revision: [D25180531](https://our.internmc.facebook.com/intern/diff/D25180531/) [ghstack-poisoned]
This was referenced Dec 3, 2020
This commit is part of a stack that reworks FutureNCCL in order to extract a generic CUDA-aware Future subclass. The stack deliberately breaks up this transition into elementary changes, to make it easier to verify that the behavior is preserved (or to highlight how it gets changed). --- My impression is that one property of the upstream Future class is that once .wait() returns, or once a callback is invoked, then .completed() should return True. This was not the case for FutureNCCL because .wait() would return immediately, and callbacks would be invoked inline, but .completed() could return False if the CUDA async operations hadn't completed yet. That was odd and confusing. Since there are other ways for users to check the status of CUDA operations (if they really need, and typically I don't think it's so common), perhaps it's best to avoid checking the status of CUDA events in .completed(). Differential Revision: [D25180531](https://our.internmc.facebook.com/intern/diff/D25180531/) [ghstack-poisoned]
This commit is part of a stack that reworks FutureNCCL in order to extract a generic CUDA-aware Future subclass. The stack deliberately breaks up this transition into elementary changes, to make it easier to verify that the behavior is preserved (or to highlight how it gets changed). --- My impression is that one property of the upstream Future class is that once .wait() returns, or once a callback is invoked, then .completed() should return True. This was not the case for FutureNCCL because .wait() would return immediately, and callbacks would be invoked inline, but .completed() could return False if the CUDA async operations hadn't completed yet. That was odd and confusing. Since there are other ways for users to check the status of CUDA operations (if they really need, and typically I don't think it's so common), perhaps it's best to avoid checking the status of CUDA events in .completed(). Differential Revision: [D25180531](https://our.internmc.facebook.com/intern/diff/D25180531/) [ghstack-poisoned]
This commit is part of a stack that reworks FutureNCCL in order to extract a generic CUDA-aware Future subclass. The stack deliberately breaks up this transition into elementary changes, to make it easier to verify that the behavior is preserved (or to highlight how it gets changed). --- My impression is that one property of the upstream Future class is that once .wait() returns, or once a callback is invoked, then .completed() should return True. This was not the case for FutureNCCL because .wait() would return immediately, and callbacks would be invoked inline, but .completed() could return False if the CUDA async operations hadn't completed yet. That was odd and confusing. Since there are other ways for users to check the status of CUDA operations (if they really need, and typically I don't think it's so common), perhaps it's best to avoid checking the status of CUDA events in .completed(). Differential Revision: [D25180531](https://our.internmc.facebook.com/intern/diff/D25180531/) [ghstack-poisoned]
|
This pull request has been merged in 003c30b. |
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 commit is part of a stack that reworks FutureNCCL in order to extract a generic CUDA-aware Future subclass. The stack deliberately breaks up this transition into elementary changes, to make it easier to verify that the behavior is preserved (or to highlight how it gets changed).
My impression is that one property of the upstream Future class is that once .wait() returns, or once a callback is invoked, then .completed() should return True. This was not the case for FutureNCCL because .wait() would return immediately, and callbacks would be invoked inline, but .completed() could return False if the CUDA async operations hadn't completed yet.
That was odd and confusing. Since there are other ways for users to check the status of CUDA operations (if they really need, and typically I don't think it's so common), perhaps it's best to avoid checking the status of CUDA events in .completed().
Differential Revision: D25180531