Use unique cache locations in backward for pipeline prefetching #2151
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
✅ Deploy Preview for pytorch-fbgemm-docs canceled.
|
|
This pull request was exported from Phabricator. Differential Revision: D51339208 |
sryap
pushed a commit
to sryap/FBGEMM
that referenced
this pull request
Nov 22, 2023
…rch#2151) Summary: When pipeline prefetching is enabled (`prefetch_pipeline=True`) for `EmbeddingLocation.MANAGED_CACHING`, TBE has to update `lxu_cache_locations` to ensure cache consistency. Prior to this diff, TBE performs the full cache lookup when updating `lxu_cache_locations` (i.e., looking up all indices although they are duplicated). The time complexity of such the lookup grows with the number of indices. Thus, the lookup can be expensive when the number of indices is large. This diff addresses this problem by looking up only the unique indices (which is normally much smaller than the full indices). The number of unique indices tends to stay more or less the same even when the total number of indices grows. Thus, looking up only unique indices can reduce cost of cache lookup effectively. The unique `lxu_cache_locations` are fed directly to TBE backward to consume. Thus, there is no decompression cost. Reviewed By: jianyuh Differential Revision: D51339208
sryap
force-pushed
the
export-D51339208
branch
from
9ba3e7a
to
d5a6b53
Compare
|
This pull request was exported from Phabricator. Differential Revision: D51339208 |
1 similar comment
|
This pull request was exported from Phabricator. Differential Revision: D51339208 |
sryap
pushed a commit
to sryap/FBGEMM
that referenced
this pull request
Nov 22, 2023
…rch#2151) Summary: When pipeline prefetching is enabled (`prefetch_pipeline=True`) for `EmbeddingLocation.MANAGED_CACHING`, TBE has to update `lxu_cache_locations` to ensure cache consistency. Prior to this diff, TBE performs the full cache lookup when updating `lxu_cache_locations` (i.e., looking up all indices although they are duplicated). The time complexity of such the lookup grows with the number of indices. Thus, the lookup can be expensive when the number of indices is large. This diff addresses this problem by looking up only the unique indices (which is normally much smaller than the full indices). The number of unique indices tends to stay more or less the same even when the total number of indices grows. Thus, looking up only unique indices can reduce cost of cache lookup effectively. The unique `lxu_cache_locations` are fed directly to TBE backward to consume. Thus, there is no decompression cost. Reviewed By: jianyuh Differential Revision: D51339208
sryap
force-pushed
the
export-D51339208
branch
from
d5a6b53
to
1b49fbd
Compare
|
This pull request was exported from Phabricator. Differential Revision: D51339208 |
2 similar comments
|
This pull request was exported from Phabricator. Differential Revision: D51339208 |
|
This pull request was exported from Phabricator. Differential Revision: D51339208 |
sryap
added a commit
to sryap/FBGEMM
that referenced
this pull request
Nov 22, 2023
…rch#2151) Summary: Pull Request resolved: pytorch#2151 When pipeline prefetching is enabled (`prefetch_pipeline=True`) for `EmbeddingLocation.MANAGED_CACHING`, TBE has to update `lxu_cache_locations` to ensure cache consistency. Prior to this diff, TBE performs the full cache lookup when updating `lxu_cache_locations` (i.e., looking up all indices although they are duplicated). The time complexity of such the lookup grows with the number of indices. Thus, the lookup can be expensive when the number of indices is large. This diff addresses this problem by looking up only the unique indices (which is normally much smaller than the full indices). The number of unique indices tends to stay more or less the same even when the total number of indices grows. Thus, looking up only unique indices can reduce cost of cache lookup effectively. The unique `lxu_cache_locations` are fed directly to TBE backward to consume. Thus, there is no decompression cost. Reviewed By: jianyuh Differential Revision: D51339208 fbshipit-source-id: a89b54a529a3628eb37e983b7f210d3a29ea315c
sryap
force-pushed
the
export-D51339208
branch
from
1b49fbd
to
5363d8c
Compare
added 2 commits
November 27, 2023 16:36
Summary: This diff adds/updates LXU cache APIs for pipeline prefetching: - Update `lxu_cache_lookup` to allow for unique linear cache indices lookup and external output tensor to be passed to the op - Update `lxu_cache_locations_update` to support unique cache locations update - Add Python binding for `get_unique_indices` Reviewed By: levythu Differential Revision: D51532548
…rch#2151) Summary: When pipeline prefetching is enabled (`prefetch_pipeline=True`) for `EmbeddingLocation.MANAGED_CACHING`, TBE has to update `lxu_cache_locations` to ensure cache consistency. Prior to this diff, TBE performs the full cache lookup when updating `lxu_cache_locations` (i.e., looking up all indices although they are duplicated). The time complexity of such the lookup grows with the number of indices. Thus, the lookup can be expensive when the number of indices is large. This diff addresses this problem by looking up only the unique indices (which is normally much smaller than the full indices). The number of unique indices tends to stay more or less the same even when the total number of indices grows. Thus, looking up only unique indices can reduce cost of cache lookup effectively. The unique `lxu_cache_locations` are fed directly to TBE backward to consume. Thus, there is no decompression cost. Reviewed By: ehsanardestani, jianyuh Differential Revision: D51339208
sryap
force-pushed
the
export-D51339208
branch
from
5363d8c
to
600b56a
Compare
|
This pull request was exported from Phabricator. Differential Revision: D51339208 |
1 similar comment
|
This pull request was exported from Phabricator. Differential Revision: D51339208 |
sryap
pushed a commit
to sryap/FBGEMM
that referenced
this pull request
Nov 28, 2023
…rch#2151) Summary: When pipeline prefetching is enabled (`prefetch_pipeline=True`) for `EmbeddingLocation.MANAGED_CACHING`, TBE has to update `lxu_cache_locations` to ensure cache consistency. Prior to this diff, TBE performs the full cache lookup when updating `lxu_cache_locations` (i.e., looking up all indices although they are duplicated). The time complexity of such the lookup grows with the number of indices. Thus, the lookup can be expensive when the number of indices is large. This diff addresses this problem by looking up only the unique indices (which is normally much smaller than the full indices). The number of unique indices tends to stay more or less the same even when the total number of indices grows. Thus, looking up only unique indices can reduce cost of cache lookup effectively. The unique `lxu_cache_locations` are fed directly to TBE backward to consume. Thus, there is no decompression cost. Reviewed By: ehsanardestani, jianyuh Differential Revision: D51339208
|
This pull request has been merged in 035ed1f. |
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.
Summary:
When pipeline prefetching is enabled (
prefetch_pipeline=True) forEmbeddingLocation.MANAGED_CACHING, TBE has to updatelxu_cache_locationsto ensure cache consistency. Prior to thisdiff, TBE performs the full cache lookup when updating
lxu_cache_locations(i.e., looking up all indices although they areduplicated). The time complexity of such the lookup grows with
the number of indices. Thus, the lookup can be expensive when the
number of indices is large. This diff addresses this problem by
looking up only the unique indices (which is normally much smaller
than the full indices). The number of unique indices tends to stay
more or less the same even when the total number of indices grows.
Thus, looking up only unique indices can reduce cost of cache lookup
effectively. The unique
lxu_cache_locationsare fed directly to TBEbackward to consume. Thus, there is no decompression cost.
Differential Revision: D51339208