[cuda] introduce trace tracker callback in cache allocator #112238
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
🔗 Helpful Links🧪 See artifacts and rendered test results at hud.pytorch.org/pr/112238
Note: Links to docs will display an error until the docs builds have been completed. ✅ You can merge normally! (4 Unrelated Failures)As of commit f289e51 with merge base 9d09d29 ( UNSTABLE - The following jobs failed but were likely due to flakiness present on trunk and has been marked as unstable:
This comment was automatically generated by Dr. CI and updates every 15 minutes. |
|
This pull request was exported from Phabricator. Differential Revision: D50726971 |
minsii
force-pushed
the
export-D50726971
branch
2 times, most recently
from
4e2a581 to
3e1dcd0
Compare
|
This pull request was exported from Phabricator. Differential Revision: D50726971 |
minsii
force-pushed
the
export-D50726971
branch
from
3e1dcd0 to
631637b
Compare
|
This pull request was exported from Phabricator. Differential Revision: D50726971 |
zdevito
reviewed
Oct 27, 2023
minsii
force-pushed
the
export-D50726971
branch
from
631637b to
c8855f0
Compare
|
This pull request was exported from Phabricator. Differential Revision: D50726971 |
minsii
force-pushed
the
export-D50726971
branch
from
c8855f0 to
4826432
Compare
|
This pull request was exported from Phabricator. Differential Revision: D50726971 |
|
Addressed comments and added a unit test |
|
@pytorchbot label "topic: not user facing" |
minsii
force-pushed
the
export-D50726971
branch
from
4826432 to
c5e8a61
Compare
|
This pull request was exported from Phabricator. Differential Revision: D50726971 |
minsii
force-pushed
the
export-D50726971
branch
from
c5e8a61 to
74ea813
Compare
|
This pull request was exported from Phabricator. Differential Revision: D50726971 |
|
This pull request was exported from Phabricator. Differential Revision: D50726971 |
minsii
force-pushed
the
export-D50726971
branch
from
74ea813 to
0cc4515
Compare
minsii
force-pushed
the
export-D50726971
branch
2 times, most recently
from
0cc4515 to
213be0b
Compare
|
This pull request was exported from Phabricator. Differential Revision: D50726971 |
minsii
force-pushed
the
export-D50726971
branch
from
213be0b to
b4e6ada
Compare
|
This pull request was exported from Phabricator. Differential Revision: D50726971 |
minsii
force-pushed
the
export-D50726971
branch
from
b4e6ada to
d3a4d68
Compare
|
This pull request was exported from Phabricator. Differential Revision: D50726971 |
zdevito
approved these changes
Nov 1, 2023
…pytorch#112238) Summary: This patch prototypes a trace tracker callback mechanism based on existing TraceEntry records. - It allows external of cache allocator to "attach" trace tracker callbacks. - When a TraceEntry is recorded, it triggers all attached callbacks. Callbacks can selectively behave based on the trace action. - **RISK**: The attached callback would be called within an allocator call stack (e.g., free during an allocate call). Potential deadlock may occur if other locks are called within the callback and has interdependency w/ the device allocator lock. It is the callback developer's responsibility to avoid any potential deadlock. - **ADVICE**: The callback mechanism is designed **only for Pytorch internal use**. We should not expose it to Python layer due to Python GIL that would cause a deadlock. See example in D50726970 that attaches NCCL register/deregister hooks via the trace tracker callback, so that all CUDA segments allocated by the allocator can be registered to NCCL communicators before any NCCL communication happens. This enables fast zero copy algorithms in NCCL. Reviewed By: zdevito Differential Revision: D50726971
minsii
force-pushed
the
export-D50726971
branch
from
d3a4d68 to
f289e51
Compare
|
This pull request was exported from Phabricator. Differential Revision: D50726971 |
minsii
changed the title
|
@pytorchbot merge (Initiating merge automatically since Phabricator Diff has merged) |
Merge startedYour change will be merged once all checks pass (ETA 0-4 Hours). Learn more about merging in the wiki. Questions? Feedback? Please reach out to the PyTorch DevX Team |
xuhancn
pushed a commit
to xuhancn/pytorch
that referenced
this pull request
Nov 7, 2023
…12238) Summary: This patch prototypes a trace tracker callback mechanism based on existing TraceEntry records. - It allows external of cache allocator to "attach" trace tracker callbacks. - When a TraceEntry is recorded, it triggers all attached callbacks. Callbacks can selectively behave based on the trace action. - **RISK**: The attached callback would be called within an allocator call stack (e.g., free during an allocate call). Potential deadlock may occur if other locks are called within the callback and has interdependency w/ the device allocator lock. It is the callback developer's responsibility to avoid any potential deadlock. - **ADVICE**: The callback mechanism is designed **only for Pytorch internal use**. We should not expose it to Python layer due to Python GIL that would cause a deadlock. See example in D50726970 that attaches NCCL register/deregister hooks via the trace tracker callback, so that all CUDA segments allocated by the allocator can be registered to NCCL communicators before any NCCL communication happens. This enables fast zero copy algorithms in NCCL. Differential Revision: D50726971 Pull Request resolved: pytorch#112238 Approved by: https://github.com/zdevito
Skylion007
pushed a commit
to Skylion007/pytorch
that referenced
this pull request
Nov 14, 2023
…12238) Summary: This patch prototypes a trace tracker callback mechanism based on existing TraceEntry records. - It allows external of cache allocator to "attach" trace tracker callbacks. - When a TraceEntry is recorded, it triggers all attached callbacks. Callbacks can selectively behave based on the trace action. - **RISK**: The attached callback would be called within an allocator call stack (e.g., free during an allocate call). Potential deadlock may occur if other locks are called within the callback and has interdependency w/ the device allocator lock. It is the callback developer's responsibility to avoid any potential deadlock. - **ADVICE**: The callback mechanism is designed **only for Pytorch internal use**. We should not expose it to Python layer due to Python GIL that would cause a deadlock. See example in D50726970 that attaches NCCL register/deregister hooks via the trace tracker callback, so that all CUDA segments allocated by the allocator can be registered to NCCL communicators before any NCCL communication happens. This enables fast zero copy algorithms in NCCL. Differential Revision: D50726971 Pull Request resolved: pytorch#112238 Approved by: https://github.com/zdevito
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.
Summary:
This patch prototypes a trace tracker callback mechanism based on existing TraceEntry records.
See example in D50726970 that attaches NCCL register/deregister hooks via the trace tracker callback, so that all CUDA segments allocated by the allocator can be registered to NCCL communicators before any NCCL communication happens. This enables fast zero copy algorithms in NCCL.
Differential Revision: D50726971