Optimize 3-bit packing #1029
Merged
Optimize 3-bit packing #1029
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
🔗 Helpful Links🧪 See artifacts and rendered test results at hud.pytorch.org/pr/pytorch/ao/1029
Note: Links to docs will display an error until the docs builds have been completed. ✅ No FailuresAs of commit af0ea95 with merge base dec0313 ( This comment was automatically generated by Dr. CI and updates every 15 minutes. |
facebook-github-bot
added
the
CLA Signed
|
This pull request was exported from Phabricator. Differential Revision: D64010666 |
Summary: Optimizes 3-bit packing as outlined here: T199311618 Before change: ---------------------------------------------------------------------------------- benchmark_pack_uint_values<3>/128/8 47.0 ns 46.4 ns 15106555 benchmark_pack_uint_values<3>/128/64 6.94 ns 6.90 ns 101226284 benchmark_pack_uint_values<3>/128/128 3.27 ns 3.24 ns 215022716 benchmark_unpack_uint_values<3>/128/8 22.0 ns 21.9 ns 32585572 benchmark_unpack_uint_values<3>/128/64 6.02 ns 5.98 ns 116910230 benchmark_unpack_uint_values<3>/128/128 2.74 ns 2.73 ns 257088291 After change: ---------------------------------------------------------------------------------- benchmark_pack_uint_values<3>/128/8 19.5 ns 19.5 ns 36050883 benchmark_pack_uint_values<3>/128/64 3.90 ns 3.87 ns 181151919 benchmark_pack_uint_values<3>/128/128 1.57 ns 1.57 ns 447247194 benchmark_unpack_uint_values<3>/128/8 20.5 ns 20.4 ns 34490914 benchmark_unpack_uint_values<3>/128/64 3.19 ns 3.11 ns 228019714 benchmark_unpack_uint_values<3>/128/128 1.71 ns 1.70 ns 408587338 Unpacking perf for 128 values is 1.60x faster (2.74/1.71). Reviewed By: digantdesai Differential Revision: D64010666
metascroy
force-pushed
the
export-D64010666
branch
from
5387317
to
af0ea95
Compare
|
This pull request was exported from Phabricator. Differential Revision: D64010666 |
kirklandsign
approved these changes
Oct 7, 2024
jainapurva
pushed a commit
that referenced
this pull request
Oct 9, 2024
Differential Revision: D64010666 Pull Request resolved: #1029
jainapurva
pushed a commit
that referenced
this pull request
Oct 15, 2024
Differential Revision: D64010666 Pull Request resolved: #1029
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:
Optimizes 3-bit packing as outlined here: T199311618
Before change:
benchmark_pack_uint_values<3>/128/8 47.0 ns 46.4 ns 15106555
benchmark_pack_uint_values<3>/128/64 6.94 ns 6.90 ns 101226284
benchmark_pack_uint_values<3>/128/128 3.27 ns 3.24 ns 215022716
benchmark_unpack_uint_values<3>/128/8 22.0 ns 21.9 ns 32585572
benchmark_unpack_uint_values<3>/128/64 6.02 ns 5.98 ns 116910230
benchmark_unpack_uint_values<3>/128/128 2.74 ns 2.73 ns 257088291
After change:
benchmark_pack_uint_values<3>/128/8 19.5 ns 19.5 ns 36050883
benchmark_pack_uint_values<3>/128/64 3.90 ns 3.87 ns 181151919
benchmark_pack_uint_values<3>/128/128 1.57 ns 1.57 ns 447247194
benchmark_unpack_uint_values<3>/128/8 20.5 ns 20.4 ns 34490914
benchmark_unpack_uint_values<3>/128/64 3.19 ns 3.11 ns 228019714
benchmark_unpack_uint_values<3>/128/128 1.71 ns 1.70 ns 408587338
Unpacking perf for 128 values is 1.60x faster (2.74/1.71).
Differential Revision: D64010666