Add helper ops to support cache conflict misses #2571
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 pull request was exported from Phabricator. Differential Revision: D55926421 |
❌ Deploy Preview for pytorch-fbgemm-docs failed.
|
sryap
added a commit
to sryap/FBGEMM
that referenced
this pull request
May 8, 2024
Summary: This diff adds helper operators for the cache conflict miss support enablement in SSD TBE. Changes include: - Extend `get_unique_indices_cuda` to compute and return inverse linear indices (the tensor that contains the original positions of lienar indices before sorting) - Extend `lru_cache_find_uncached_cuda` to compute and return the inverse cache sets (the tensor that contains the original positions of cache sets of unique indices before sorting) - Update SSD backend to support cache conflict misses instead of failing. The rows that experience conflict misses will be stored in a scratch pad for TBE kernels to consume. They will be evicted to SSD once the backward+optimizer step of TBE is completed. - Add `ssd_generate_row_addrs` for generating row addresses of data that is fetched from SSD (data can be in either a scratch pad or LXU cache). Differential Revision: D55926421
|
This pull request was exported from Phabricator. Differential Revision: D55926421 |
sryap
added a commit
to sryap/FBGEMM
that referenced
this pull request
May 9, 2024
Summary: This diff adds helper operators for the cache conflict miss support enablement in SSD TBE. Changes include: - Extend `get_unique_indices_cuda` to compute and return inverse linear indices (the tensor that contains the original positions of lienar indices before sorting) - Extend `lru_cache_find_uncached_cuda` to compute and return the inverse cache sets (the tensor that contains the original positions of cache sets of unique indices before sorting) - Update SSD backend to support cache conflict misses instead of failing. The rows that experience conflict misses will be stored in a scratch pad for TBE kernels to consume. They will be evicted to SSD once the backward+optimizer step of TBE is completed. - Add `ssd_generate_row_addrs` for generating row addresses of data that is fetched from SSD (data can be in either a scratch pad or LXU cache). Reviewed By: q10 Differential Revision: D55926421
|
This pull request was exported from Phabricator. Differential Revision: D55926421 |
sryap
added a commit
to sryap/FBGEMM
that referenced
this pull request
May 9, 2024
Summary: This diff adds helper operators for the cache conflict miss support enablement in SSD TBE. Changes include: - Extend `get_unique_indices_cuda` to compute and return inverse linear indices (the tensor that contains the original positions of lienar indices before sorting) - Extend `lru_cache_find_uncached_cuda` to compute and return the inverse cache sets (the tensor that contains the original positions of cache sets of unique indices before sorting) - Update SSD backend to support cache conflict misses instead of failing. The rows that experience conflict misses will be stored in a scratch pad for TBE kernels to consume. They will be evicted to SSD once the backward+optimizer step of TBE is completed. - Add `ssd_generate_row_addrs` for generating row addresses of data that is fetched from SSD (data can be in either a scratch pad or LXU cache). Reviewed By: q10 Differential Revision: D55926421
|
This pull request was exported from Phabricator. Differential Revision: D55926421 |
Summary: This diff adds helper operators for the cache conflict miss support enablement in SSD TBE. Changes include: - Extend `get_unique_indices_cuda` to compute and return inverse linear indices (the tensor that contains the original positions of lienar indices before sorting) - Extend `lru_cache_find_uncached_cuda` to compute and return the inverse cache sets (the tensor that contains the original positions of cache sets of unique indices before sorting) - Update SSD backend to support cache conflict misses instead of failing. The rows that experience conflict misses will be stored in a scratch pad for TBE kernels to consume. They will be evicted to SSD once the backward+optimizer step of TBE is completed. - Add `ssd_generate_row_addrs` for generating row addresses of data that is fetched from SSD (data can be in either a scratch pad or LXU cache). Reviewed By: q10 Differential Revision: D55926421
|
This pull request was exported from Phabricator. Differential Revision: D55926421 |
|
This pull request has been merged in 56d21a0. |
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:
This diff adds helper operators for the cache conflict miss support
enablement in SSD TBE. Changes include:
get_unique_indices_cudato compute and return inverselinear indices (the tensor that contains the original positions of
lienar indices before sorting)
lru_cache_find_uncached_cudato compute and return theinverse cache sets (the tensor that contains the original positions
of cache sets of unique indices before sorting)
failing. The rows that experience conflict misses will be stored in
a scratch pad for TBE kernels to consume. They will be evicted to
SSD once the backward+optimizer step of TBE is completed.
ssd_generate_row_addrsfor generating row addresses of datathat is fetched from SSD (data can be in either a scratch pad or LXU
cache).
Differential Revision: D55926421