Multithreaded speedups for CPU models. #517
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
On the CPU, we were not achieving full thread utilization. The problem seems to be that
using the 'guided' schedule in openmp assigns a large range of ids to a single thread,
and its pretty common to have ids sorted by frequency. This caused 1 thread to process
all the high activity users/items - and left the others starved for work as they finished
early.
Fix by switching to a dynamic openmp schedule. For the ALS model on cpu, this change is 3.8x
faster on training movielens-20m, and 2.2x faster training the Github Stars dataset -
while being neutral on lastfm. For cosine the model, this change is 2.2x training movielens20m.
* 'schedule=dynamic' times
* ALS model:
* GitHub 60.96 s/it
* Movielens: 1.05 it/s
* LastFM 2.19 s/it
* Cosine model:
* Movielens: 49603.48 it/s
* 'schedule=guided' openmp times (before this change)
* ALS
* Github: 133 s/it
* Lastfm: 2.19 s/it
* Movielens: 3.62 s/it
* Cosine with change
* Movielens 22804.29 it/s
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Outdated suggestions cannot be applied.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.
On the CPU, we were not achieving full thread utilization. The problem seems to be that
using the 'guided' schedule in openmp assigns a large range of ids to a single thread,
and its pretty common to have ids sorted by frequency. This caused 1 thread to process
all the high activity users/items - and left the others starved for work as they finished
early.
Fix by switching to a dynamic openmp schedule. For the ALS model on cpu, this change is 3.8x
faster on training movielens-20m, and 2.2x faster training the Github Stars dataset -
while being neutral on lastfm. For the cosine model, this change is 2.2x training movielens20m.