use prefetch to load next mem into cache #21206
Merged
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
test=develop
|
I test it on pyramid_dnn training, see PaddlePaddle/benchmark#151 (comment) for details.
|
luotao1
reviewed
Nov 22, 2019
test=develop
|
Please update the PR description for more details. |
seiriosPlus
pushed a commit
to seiriosPlus/Paddle
that referenced
this pull request
Dec 9, 2019
* use prefetch to load next mem into cache test=develop * remove hard code memcpy om pyramid_hash_ff test=develop
seiriosPlus
pushed a commit
to seiriosPlus/Paddle
that referenced
this pull request
Dec 9, 2019
* use prefetch to load next mem into cache test=develop * remove hard code memcpy om pyramid_hash_ff test=develop
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Outdated suggestions cannot be applied.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.
memcpy usage in hash_embedding_ff of Pyramid DNN is a typical random memory access case, which makes cache miss frequently. Before each memcpy, it uses a hash func to calculate next position for mem read, this position is not continuous, so each time L1 cache need refresh and access cache line to read next mem block, it costs time, and make memcpy perf bad.
The good thing is the hashed position is predictable, we can use XXH32() to predict next position in last memcpy, by using prefetch(), we can let system preload next mem into cache during current memcpy, it saves time and reduce cache miss rate.
test=develop