[Pytorch] Use exp_u20 for aarch64's erf#166594
Closed
Nicoshev wants to merge 1 commit intopytorch:mainfrom
Closed
Conversation
pytorch-bot
bot
added
the
module: cpu
🔗 Helpful Links🧪 See artifacts and rendered test results at hud.pytorch.org/pr/166594
Note: Links to docs will display an error until the docs builds have been completed. ❗ 1 Active SEVsThere are 1 currently active SEVs. If your PR is affected, please view them below: ✅ No FailuresAs of commit 2c1cbce with merge base b4403bf ( This comment was automatically generated by Dr. CI and updates every 15 minutes. |
Nicoshev
added
ciflow/trunk
Contributor
Author
|
@pytorchbot label "topic: perf improvements" "release notes: cpu (aarch64)" |
|
Didn't find following labels among repository labels: topic: perf improvements |
pytorch-bot
bot
added
the
release notes: cpu (aarch64)
mcfi
approved these changes
Oct 30, 2025
Nicoshev
added a commit
to Nicoshev/pytorch
that referenced
this pull request
Oct 30, 2025
Summary: After a precision study, we concluded it is ok to use ACL's exp function on f32's erf() We've moved ACL's exp implementation to an inline function, and call it from exp_u20() and erf(). We can keep erf inline this way. Benchmarks show about 108% higher throughput on clang-19: Before: f32 erf: 2539.179us After: f32 erf: 1221.083us Test Plan: Correctness: buck2 test mode/opt //caffe2/test:test_ops buck2 test mode/opt //caffe2/test:torch Performance: buck2 run mode/opt //caffe2/benchmarks/operator_benchmark/fb:operator_benchmark_test Differential Revision: D85730452
Nicoshev
force-pushed
the
export-D85730452
branch
2 times, most recently
from
9235e06 to
36fb135
Compare
Nicoshev
added a commit
to Nicoshev/pytorch
that referenced
this pull request
Oct 30, 2025
Summary: After a precision study, we concluded it is ok to use ACL's exp function on f32's erf() We've moved ACL's exp implementation to an inline function, and call it from exp_u20() and erf(). We can keep erf inline this way. Benchmarks show about 108% higher throughput on clang-19: Before: f32 erf: 2539.179us After: f32 erf: 1221.083us Test Plan: Correctness: buck2 test mode/opt //caffe2/test:test_ops buck2 test mode/opt //caffe2/test:torch Performance: buck2 run mode/opt //caffe2/benchmarks/operator_benchmark/fb:operator_benchmark_test Differential Revision: D85730452
Nicoshev
added a commit
to Nicoshev/pytorch
that referenced
this pull request
Oct 30, 2025
Summary: After a precision study, we concluded it is ok to use ACL's exp function on f32's erf() We've moved ACL's exp implementation to an inline function, and call it from exp_u20() and erf(). We can keep erf inline this way. Benchmarks show about 108% higher throughput on clang-19: Before: f32 erf: 2539.179us After: f32 erf: 1221.083us Test Plan: Correctness: buck2 test mode/opt //caffe2/test:test_ops buck2 test mode/opt //caffe2/test:torch Performance: buck2 run mode/opt //caffe2/benchmarks/operator_benchmark/fb:operator_benchmark_test Differential Revision: D85730452
Nicoshev
force-pushed
the
export-D85730452
branch
2 times, most recently
from
00b4e0b to
f0cfce4
Compare
Nicoshev
added a commit
to Nicoshev/pytorch
that referenced
this pull request
Oct 30, 2025
Summary: After a precision study, we concluded it is ok to use ACL's exp function on f32's erf() We've moved ACL's exp implementation to an inline function, and call it from exp_u20() and erf(). We can keep erf inline this way. Benchmarks show about 108% higher throughput on clang-19, when processing a tensor of 1M elements: Before: f32 erf: 2539.179us After: f32 erf: 1329.063us Test Plan: Correctness: buck2 test mode/opt //caffe2/test:test_ops buck2 test mode/opt //caffe2/test:torch Performance: buck2 run mode/opt //caffe2/benchmarks/operator_benchmark/fb:operator_benchmark_test Differential Revision: D85730452
Summary: After a precision study, we concluded it is ok to use ACL's exp function on f32's erf() We've moved ACL's exp implementation to an inline function, and call it from exp_u20() and erf(). We can keep erf inline this way. Benchmarks show about 91% higher throughput on clang-19, when processing a tensor of 1M elements: Before: f32 erf: 2539.179us After: f32 erf: 1329.063us Test Plan: Correctness: buck2 test mode/opt //caffe2/test:test_ops buck2 test mode/opt //caffe2/test:torch Performance: buck2 run mode/opt //caffe2/benchmarks/operator_benchmark/fb:operator_benchmark_test Differential Revision: D85730452
Nicoshev
force-pushed
the
export-D85730452
branch
from
f0cfce4 to
2c1cbce
Compare
fadara01
approved these changes
Oct 30, 2025
Collaborator
fadara01
left a comment
fadara01
left a comment
There was a problem hiding this comment.
LGTM - You'll see an even higher speedup once we enable SVE128 vectorizer (SVE128 exp_u20() is faster than Neon's).
Also, this implementation doesn't come from Arm Compute Library (ACL), it's from Arm Optimized Routines (AOR)
Contributor
|
@pytorchbot merge (Initiating merge automatically since Phabricator Diff has merged) |
Collaborator
Merge startedYour change will be merged once all checks pass (ETA 0-4 Hours). Learn more about merging in the wiki. Questions? Feedback? Please reach out to the PyTorch DevX Team |
Nicoshev
added
the
module: arm
BoyuanFeng
pushed a commit
that referenced
this pull request
Oct 31, 2025
Summary: After a precision study, we concluded it is ok to use ACL's exp function on f32's erf() We can keep erf inline this way. Benchmarks show about 91% higher throughput when processing a tensor of 1M elements, compiling with clang-19: Before: f32 erf: 2539.179us After: f32 erf: 1329.063us Test Plan: Correctness: buck2 test mode/opt //caffe2/test:test_ops buck2 test mode/opt //caffe2/test:torch Performance: buck2 run mode/opt //caffe2/benchmarks/operator_benchmark/fb:operator_benchmark_test Differential Revision: D85730452 Pull Request resolved: #166594 Approved by: https://github.com/mcfi, https://github.com/fadara01
etaf
pushed a commit
to etaf/pytorch-inductor-xpu
that referenced
this pull request
Nov 4, 2025
Summary: After a precision study, we concluded it is ok to use ACL's exp function on f32's erf() We can keep erf inline this way. Benchmarks show about 91% higher throughput when processing a tensor of 1M elements, compiling with clang-19: Before: f32 erf: 2539.179us After: f32 erf: 1329.063us Test Plan: Correctness: buck2 test mode/opt //caffe2/test:test_ops buck2 test mode/opt //caffe2/test:torch Performance: buck2 run mode/opt //caffe2/benchmarks/operator_benchmark/fb:operator_benchmark_test Differential Revision: D85730452 Pull Request resolved: pytorch#166594 Approved by: https://github.com/mcfi, https://github.com/fadara01
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
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
5 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.
Summary:
After a precision study, we concluded it is ok to use ACL's exp function on f32's erf()
We can keep erf inline this way.
Benchmarks show about 91% higher throughput when processing a tensor of 1M elements, compiling with clang-19:
Before:
f32 erf: 2539.179us
After:
f32 erf: 1329.063us
Test Plan:
Correctness:
buck2 test mode/opt //caffe2/test:test_ops
buck2 test mode/opt //caffe2/test:torch
Performance:
buck2 run mode/opt //caffe2/benchmarks/operator_benchmark/fb:operator_benchmark_test
Differential Revision: D85730452
cc @jgong5 @mingfeima @XiaobingSuper @sanchitintel @ashokei @jingxu10 @jerryzh168 @aditew01 @snadampal @milpuz01 @nikhil-arm @fadara01