bwl1289/feat/merge-google-pthreadpool #1
Merged
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
PiperOrigin-RevId: 706596019
PiperOrigin-RevId: 706995443
PiperOrigin-RevId: 707095720
PiperOrigin-RevId: 707224063
PiperOrigin-RevId: 707559093
PiperOrigin-RevId: 707609059
2nd try, fixed subtle difference in `pthreadpool_decrement_fetch_acquire_release_size_t` this time around. PiperOrigin-RevId: 708224758
Unless we're on Android, where sleeping/waking is slow. PiperOrigin-RevId: 708363228
…n defined. PiperOrigin-RevId: 713777960
PiperOrigin-RevId: 714131650
…tegies to `pthreadpool`. The `pthreadpool_parallelize_Xd_tile_Yd_dynamic` strategy differs from the existing non-dynamic `pthreadpool_parallelize_Xd_tile_Yd` in that the `count_*` argument of `function` corresponding to the innermost dimension can be any integer multiple of the corresponding `tile_*` argument, or the remainder of that range. E.g. `pthreadpool_parallelize_3d_tile_2d_dynamic(&threadpool, function, context, range_i, range_j, range_k, tile_j, tile_k)`, results in calls to ``` function(context, offset_i, offset_j, offset_k, count_j, count_k) ``` where `offset_j` and `offset_k` are integer multiples of `tile_j` and `tile_k`, respectively, and `count_k` is an integer multiple of `tile_k` or `range_k - offset_k`. If `range_k <= tile_k`, then `count_j` may be an integer multiple of `tile_j` or `range_j - offset_j`. The number of elements processed in each call is chosen as `max((count - offset) / (chunk_factor * num_threads), 1)` where `count` is the total number of elements and `offset` is the current number of elements already processed. This produces large element chunks initially, when there are lots of elements to process, and successively smaller chunks as the end of the computation nears. The `chunk_factor` used above is the ratio of the fastest to the slowest core in the parallel computation plus one. On heterogeneous systems, this is set to `5`, otherwise to `2`. PiperOrigin-RevId: 715757992
…trategies. PiperOrigin-RevId: 716200764
…gies to match the tile traversal order of the other strategies. Also use the same logic for picking tiles within the range (owner picks from `range_start`, stealing threads from `range_end`) to reduce the number of atomic operations per tile. PiperOrigin-RevId: 723497649
PiperOrigin-RevId: 723539628
PiperOrigin-RevId: 723596220
…ut first check. PiperOrigin-RevId: 724250471
… == 1`. PiperOrigin-RevId: 724342903
Also update yanked version of rules_cc: https://github.com/google/pthreadpool/actions/runs/13320936827/job/37205321364 PiperOrigin-RevId: 726720282
Apple toolchains have been moved to apple_support from Bazel 7. See bazelbuild/bazel#16619. Those values won't match anymore. Also all those config_settings should move to platforms: https://bazel.build/extending/platforms PiperOrigin-RevId: 732118106
PiperOrigin-RevId: 736869281
…d `pthreadpool_parallelize_4d_tile_2d_dynamic_with_uarch`. PiperOrigin-RevId: 736903003
PiperOrigin-RevId: 738347700
…placing the implementation of pthreadpool functions. PiperOrigin-RevId: 740067423
…k symbols PiperOrigin-RevId: 746173951
PiperOrigin-RevId: 750312221
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
7 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.
No description provided.