[Pytorch] Improve conversion to bfloat16 on aarch64/NEON#166958
Closed
Nicoshev wants to merge 1 commit intopytorch:mainfrom
Closed
[Pytorch] Improve conversion to bfloat16 on aarch64/NEON#166958Nicoshev wants to merge 1 commit intopytorch:mainfrom
Nicoshev wants to merge 1 commit intopytorch:mainfrom
Conversation
🔗 Helpful Links🧪 See artifacts and rendered test results at hud.pytorch.org/pr/166958
Note: Links to docs will display an error until the docs builds have been completed. ✅ You can merge normally! (1 Unrelated Failure)As of commit 444d752 with merge base d273422 ( UNSTABLE - The following job is marked as unstable, possibly due to flakiness on trunk:
This comment was automatically generated by Dr. CI and updates every 15 minutes. |
pytorch-bot
bot
added
the
module: cpu
Nicoshev
added
module: arm
mcfi
approved these changes
Nov 4, 2025
Nicoshev
force-pushed
the
export-D86207189
branch
from
d51ef2d to
27e728b
Compare
Nicoshev
added a commit
to Nicoshev/pytorch
that referenced
this pull request
Nov 4, 2025
) Summary: Conversion by casting to bfloat16_t is broken in clang-[17, 20], fixed in clang-21. We are adding a workaround vectorized code, which improves conversion speed from smaller int data types. We've observed the following performance improvements, when compiling with clang-19 and targeting armv9a+sve2: before: uint8->bfloat16_t ===> 319.433us int8->bfloat16_t ===> 320.216us int16->bfloat16_t ===> 326.899us int32->bfloat16_t ===> 327.925us after: uint8->bfloat16_t ===> 185.189us -----> 72% higher throughput int8->bfloat16_t ===> 169.790us -----> 89% higher throughput int16->bfloat16_t ===> 180.744us -----> 81% higher throughput int32->bfloat16_t ===> 185.129us -----> 77% higher throughput 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: D86207189
Nicoshev
force-pushed
the
export-D86207189
branch
from
27e728b to
587be9a
Compare
Nicoshev
added a commit
to Nicoshev/pytorch
that referenced
this pull request
Nov 4, 2025
) Summary: Conversion by casting to bfloat16_t is broken in clang-[17, 20], fixed in clang-21. We are adding a workaround vectorized code, which improves conversion speed from smaller int data types. We've observed the following performance improvements, when compiling with clang-19 and targeting armv9a+sve2: before: uint8->bfloat16_t ===> 319.433us int8->bfloat16_t ===> 320.216us int16->bfloat16_t ===> 326.899us int32->bfloat16_t ===> 327.925us after: uint8->bfloat16_t ===> 185.189us -----> 72% higher throughput int8->bfloat16_t ===> 169.790us -----> 89% higher throughput int16->bfloat16_t ===> 180.744us -----> 81% higher throughput int32->bfloat16_t ===> 185.129us -----> 77% higher throughput 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: D86207189
Nicoshev
force-pushed
the
export-D86207189
branch
from
587be9a to
d9b2824
Compare
Nicoshev
added a commit
to Nicoshev/pytorch
that referenced
this pull request
Nov 4, 2025
) Summary: Conversion by casting to bfloat16_t is broken in clang-[17, 20], fixed in clang-21. We are adding a workaround vectorized code, which improves conversion speed from smaller int data types. We've observed the following performance improvements, when compiling with clang-19 and targeting armv9a+sve2: before: uint8->bfloat16_t ===> 319.433us int8->bfloat16_t ===> 320.216us int16->bfloat16_t ===> 326.899us int32->bfloat16_t ===> 327.925us after: uint8->bfloat16_t ===> 185.189us -----> 72% higher throughput int8->bfloat16_t ===> 169.790us -----> 89% higher throughput int16->bfloat16_t ===> 180.744us -----> 81% higher throughput int32->bfloat16_t ===> 185.129us -----> 77% higher throughput 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: D86207189
Nicoshev
force-pushed
the
export-D86207189
branch
from
d9b2824 to
01d38f2
Compare
Nicoshev
added a commit
to Nicoshev/pytorch
that referenced
this pull request
Nov 4, 2025
) Summary: Conversion by casting to bfloat16_t is broken in clang-[17, 20], fixed in clang-21. We are adding a workaround vectorized code, which improves conversion speed from smaller int data types. We've observed the following performance improvements, when compiling with clang-19 and targeting armv9a+sve2: before: uint8->bfloat16_t ===> 319.433us int8->bfloat16_t ===> 320.216us int16->bfloat16_t ===> 326.899us int32->bfloat16_t ===> 327.925us after: uint8->bfloat16_t ===> 185.189us -----> 72% higher throughput int8->bfloat16_t ===> 169.790us -----> 89% higher throughput int16->bfloat16_t ===> 180.744us -----> 81% higher throughput int32->bfloat16_t ===> 185.129us -----> 77% higher throughput 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: D86207189
Nicoshev
force-pushed
the
export-D86207189
branch
2 times, most recently
from
893bb57 to
f563627
Compare
Nicoshev
added a commit
to Nicoshev/pytorch
that referenced
this pull request
Nov 4, 2025
) Summary: Conversion by casting to bfloat16_t is broken in clang-[17, 20], fixed in clang-21. We are adding a workaround vectorized code, which improves conversion speed from smaller int data types. We've observed the following performance improvements, when compiling with clang-19 and targeting armv9a+sve2: before: uint8->bfloat16_t ===> 319.433us int8->bfloat16_t ===> 320.216us int16->bfloat16_t ===> 326.899us int32->bfloat16_t ===> 327.925us after: uint8->bfloat16_t ===> 185.189us -----> 72% higher throughput int8->bfloat16_t ===> 169.790us -----> 89% higher throughput int16->bfloat16_t ===> 180.744us -----> 81% higher throughput int32->bfloat16_t ===> 185.129us -----> 77% higher throughput 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: D86207189
Nicoshev
force-pushed
the
export-D86207189
branch
from
f563627 to
59c4cf4
Compare
Nicoshev
added a commit
to Nicoshev/pytorch
that referenced
this pull request
Nov 4, 2025
) Summary: Conversion by casting to bfloat16_t is broken in clang-[17, 20], fixed in clang-21. We are adding a workaround vectorized code, which improves conversion speed from smaller int data types. We've observed the following performance improvements, when compiling with clang-19 and targeting armv9a+sve2: before: uint8->bfloat16_t ===> 319.433us int8->bfloat16_t ===> 320.216us int16->bfloat16_t ===> 326.899us int32->bfloat16_t ===> 327.925us after: uint8->bfloat16_t ===> 185.189us -----> 72% higher throughput int8->bfloat16_t ===> 169.790us -----> 89% higher throughput int16->bfloat16_t ===> 180.744us -----> 81% higher throughput int32->bfloat16_t ===> 185.129us -----> 77% higher throughput 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: D86207189
Nicoshev
added a commit
to Nicoshev/pytorch
that referenced
this pull request
Nov 4, 2025
) Summary: Conversion by casting to bfloat16_t is broken in clang-[17, 20], fixed in clang-21. We are adding a workaround vectorized code, which improves conversion speed from smaller int data types. We've observed the following performance improvements, when compiling with clang-19 and targeting armv9a+sve2: before: uint8->bfloat16_t ===> 319.433us int8->bfloat16_t ===> 320.216us int16->bfloat16_t ===> 326.899us int32->bfloat16_t ===> 327.925us after: uint8->bfloat16_t ===> 185.189us -----> 72% higher throughput int8->bfloat16_t ===> 169.790us -----> 89% higher throughput int16->bfloat16_t ===> 180.744us -----> 81% higher throughput int32->bfloat16_t ===> 185.129us -----> 77% higher throughput 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: D86207189
Nicoshev
force-pushed
the
export-D86207189
branch
2 times, most recently
from
a2ceafe to
c98755c
Compare
Nicoshev
added a commit
to Nicoshev/pytorch
that referenced
this pull request
Nov 4, 2025
) Summary: Conversion by casting to bfloat16_t is broken in clang-[17, 20], fixed in clang-21. We are adding a workaround vectorized code, which improves conversion speed from smaller int data types. We've observed the following performance improvements, when compiling with clang-19 and targeting armv9a+sve2: before: uint8->bfloat16_t ===> 319.433us int8->bfloat16_t ===> 320.216us int16->bfloat16_t ===> 326.899us int32->bfloat16_t ===> 327.925us after: uint8->bfloat16_t ===> 185.189us -----> 72% higher throughput int8->bfloat16_t ===> 169.790us -----> 89% higher throughput int16->bfloat16_t ===> 180.744us -----> 81% higher throughput int32->bfloat16_t ===> 185.129us -----> 77% higher throughput 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 Reviewed By: mcfi Differential Revision: D86207189
Nicoshev
force-pushed
the
export-D86207189
branch
from
c98755c to
a97e4bb
Compare
Nicoshev
added a commit
to Nicoshev/pytorch
that referenced
this pull request
Nov 4, 2025
) Summary: Conversion by casting to bfloat16_t is broken in clang-[17, 20], fixed in clang-21. We are adding a workaround vectorized code, which improves conversion speed from smaller int data types. We've observed the following performance improvements, when compiling with clang-19 and targeting armv9a+sve2: before: uint8->bfloat16_t ===> 319.433us int8->bfloat16_t ===> 320.216us int16->bfloat16_t ===> 326.899us int32->bfloat16_t ===> 327.925us after: uint8->bfloat16_t ===> 185.189us -----> 72% higher throughput int8->bfloat16_t ===> 169.790us -----> 89% higher throughput int16->bfloat16_t ===> 180.744us -----> 81% higher throughput int32->bfloat16_t ===> 185.129us -----> 77% higher throughput 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 Reviewed By: mcfi Differential Revision: D86207189
Nicoshev
added a commit
to Nicoshev/pytorch
that referenced
this pull request
Nov 5, 2025
) Summary: Conversion by casting to bfloat16_t is broken in clang-[17, 20], fixed in clang-21. We are adding a workaround vectorized code, which improves conversion speed from smaller int data types. We've observed the following performance improvements, when compiling with clang-19 and targeting armv9a+sve2: before: uint8->bfloat16_t ===> 319.433us int8->bfloat16_t ===> 320.216us int16->bfloat16_t ===> 326.899us int32->bfloat16_t ===> 327.925us after: uint8->bfloat16_t ===> 185.189us -----> 72% higher throughput int8->bfloat16_t ===> 169.790us -----> 89% higher throughput int16->bfloat16_t ===> 180.744us -----> 81% higher throughput int32->bfloat16_t ===> 185.129us -----> 77% higher throughput 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 Reviewed By: mcfi Differential Revision: D86207189
Nicoshev
force-pushed
the
export-D86207189
branch
2 times, most recently
from
8639998 to
c734960
Compare
Nicoshev
added a commit
to Nicoshev/pytorch
that referenced
this pull request
Nov 5, 2025
) Summary: Conversion by casting to bfloat16_t is broken in clang-[17, 20], fixed in clang-21. We are adding a workaround vectorized code, which improves conversion speed from smaller int data types. We've observed the following performance improvements, when compiling with clang-19 and targeting armv9a+sve2: before: uint8->bfloat16_t ===> 319.433us int8->bfloat16_t ===> 320.216us int16->bfloat16_t ===> 326.899us int32->bfloat16_t ===> 327.925us after: uint8->bfloat16_t ===> 185.189us -----> 72% higher throughput int8->bfloat16_t ===> 169.790us -----> 89% higher throughput int16->bfloat16_t ===> 180.744us -----> 81% higher throughput int32->bfloat16_t ===> 185.129us -----> 77% higher throughput 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 Reviewed By: mcfi Differential Revision: D86207189
malfet
requested changes
Nov 5, 2025
Contributor
malfet
left a comment
malfet
left a comment
There was a problem hiding this comment.
Please add performance benchmark
Nicoshev
force-pushed
the
export-D86207189
branch
from
c734960 to
15c74aa
Compare
pytorch-bot bot
pushed a commit
that referenced
this pull request
Nov 11, 2025
Summary: Conversion by casting to bfloat16_t is broken in clang-[17, 20], fixed in clang-21. We are adding a workaround vectorized code, which improves conversion speed from smaller int data types. We've observed the following performance improvements, when compiling with clang-19 and targeting armv9a+sve2: before: uint8->bfloat16_t ===> 319.433us int8->bfloat16_t ===> 320.216us int16->bfloat16_t ===> 326.899us int32->bfloat16_t ===> 327.925us after: uint8->bfloat16_t ===> 185.189us -----> 72% higher throughput int8->bfloat16_t ===> 169.790us -----> 89% higher throughput int16->bfloat16_t ===> 180.744us -----> 81% higher throughput int32->bfloat16_t ===> 185.129us -----> 77% higher throughput 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 Reviewed By: mcfi Differential Revision: D86207189
Nicoshev
force-pushed
the
export-D86207189
branch
from
15c74aa to
af6bf74
Compare
Nicoshev
added a commit
to Nicoshev/pytorch
that referenced
this pull request
Nov 11, 2025
) Summary: Conversion by casting to bfloat16_t is broken in clang-[17, 20], fixed in clang-21. We are adding a workaround vectorized code, which improves conversion speed from smaller int data types. We've observed the following performance improvements, when compiling with clang-19 and targeting armv9a+sve2: before: uint8->bfloat16_t ===> 319.433us int8->bfloat16_t ===> 320.216us int16->bfloat16_t ===> 326.899us int32->bfloat16_t ===> 327.925us after: uint8->bfloat16_t ===> 185.189us -----> 72% higher throughput int8->bfloat16_t ===> 169.790us -----> 89% higher throughput int16->bfloat16_t ===> 180.744us -----> 81% higher throughput int32->bfloat16_t ===> 185.129us -----> 77% higher throughput 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 Reviewed By: mcfi Differential Revision: D86207189
Nicoshev
force-pushed
the
export-D86207189
branch
from
af6bf74 to
f61269a
Compare
Nicoshev
added a commit
to Nicoshev/pytorch
that referenced
this pull request
Nov 11, 2025
) Summary: Conversion by casting to bfloat16_t is broken in clang-[17, 20], fixed in clang-21. We are adding a workaround vectorized code, which improves conversion speed from smaller int data types. We've observed the following performance improvements, when compiling with clang-19 and targeting armv9a+sve2: before: uint8->bfloat16_t ===> 319.433us int8->bfloat16_t ===> 320.216us int16->bfloat16_t ===> 326.899us int32->bfloat16_t ===> 327.925us after: uint8->bfloat16_t ===> 185.189us -----> 72% higher throughput int8->bfloat16_t ===> 169.790us -----> 89% higher throughput int16->bfloat16_t ===> 180.744us -----> 81% higher throughput int32->bfloat16_t ===> 185.129us -----> 77% higher throughput 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 Reviewed By: mcfi Differential Revision: D86207189
Nicoshev
force-pushed
the
export-D86207189
branch
from
f61269a to
1af3ed3
Compare
pytorch-bot bot
pushed a commit
that referenced
this pull request
Nov 12, 2025
Summary: Conversion by casting to bfloat16_t is broken in clang-[17, 20], fixed in clang-21. We are adding a workaround vectorized code, which improves conversion speed from smaller int data types. We've observed the following performance improvements, when compiling with clang-19 and targeting armv9a+sve2: before: uint8->bfloat16_t ===> 319.433us int8->bfloat16_t ===> 320.216us int16->bfloat16_t ===> 326.899us int32->bfloat16_t ===> 327.925us after: uint8->bfloat16_t ===> 185.189us -----> 72% higher throughput int8->bfloat16_t ===> 169.790us -----> 89% higher throughput int16->bfloat16_t ===> 180.744us -----> 81% higher throughput int32->bfloat16_t ===> 185.129us -----> 77% higher throughput 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 Reviewed By: mcfi Differential Revision: D86207189
Nicoshev
added a commit
to Nicoshev/pytorch
that referenced
this pull request
Nov 13, 2025
) Summary: Conversion by casting to bfloat16_t is broken in clang-[17, 20], fixed in clang-21. We are adding a workaround vectorized code, which improves conversion speed from smaller int data types. We've observed the following performance improvements, when compiling with clang-19 and targeting armv9a+sve2: before: uint8->bfloat16_t ===> 319.433us int8->bfloat16_t ===> 320.216us int16->bfloat16_t ===> 326.899us int32->bfloat16_t ===> 327.925us after: uint8->bfloat16_t ===> 185.189us -----> 72% higher throughput int8->bfloat16_t ===> 169.790us -----> 89% higher throughput int16->bfloat16_t ===> 180.744us -----> 81% higher throughput int32->bfloat16_t ===> 185.129us -----> 77% higher throughput 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 Reviewed By: mcfi Differential Revision: D86207189
Nicoshev
force-pushed
the
export-D86207189
branch
from
1af3ed3 to
29fd6ef
Compare
) Summary: Conversion by casting to bfloat16_t is broken in clang-[17, 20], fixed in clang-21. We are adding a workaround vectorized code, which improves conversion speed from smaller int data types. We've observed the following performance improvements, when compiling with clang-19 and targeting armv9a+sve2: before: uint8->bfloat16_t ===> 319.433us int8->bfloat16_t ===> 320.216us int16->bfloat16_t ===> 326.899us int32->bfloat16_t ===> 327.925us after: uint8->bfloat16_t ===> 185.189us -----> 72% higher throughput int8->bfloat16_t ===> 169.790us -----> 89% higher throughput int16->bfloat16_t ===> 180.744us -----> 81% higher throughput int32->bfloat16_t ===> 185.129us -----> 77% higher throughput 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 Reviewed By: mcfi Differential Revision: D86207189
Nicoshev
force-pushed
the
export-D86207189
branch
from
29fd6ef to
444d752
Compare
Contributor
|
@pytorchbot merge (Initiating merge automatically since Phabricator Diff has merged) |
|
This PR has pending changes requested. Please address the comments and update the PR before merging. |
Contributor
Author
|
@pytorchbot merge -f "Benchmarks added in a separate PR. CI failure was on an x86 build compiled by gcc, it does not execute added path. |
|
❌ 🤖 pytorchbot command failed: |
Contributor
Author
|
@pytorchbot merge -f "Benchs added in PR167099. CI fails on an x86 build compiled by gcc, it doesn't execute new path" |
Collaborator
Merge startedYour change will be merged immediately since you used the force (-f) flag, bypassing any CI checks (ETA: 1-5 minutes). Please use Learn more about merging in the wiki. Questions? Feedback? Please reach out to the PyTorch DevX Team |
Silv3S
pushed a commit
to Silv3S/pytorch
that referenced
this pull request
Nov 18, 2025
) Summary: Autovectorization of casting to bfloat16_t is broken in clang-[17, 20], fixed in clang-21. We are adding a workaround vectorized code, which improves conversion speed from smaller int data types. We've observed the following performance improvements, when compiling with clang-19 and targeting armv9a+sve2: before: uint8->bfloat16_t ===> 319.433us int8->bfloat16_t ===> 320.216us int16->bfloat16_t ===> 326.899us int32->bfloat16_t ===> 327.925us after: uint8->bfloat16_t ===> 185.189us -----> 72% higher throughput int8->bfloat16_t ===> 169.790us -----> 89% higher throughput int16->bfloat16_t ===> 180.744us -----> 81% higher throughput int32->bfloat16_t ===> 185.129us -----> 77% higher throughput 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: D86207189 Pull Request resolved: pytorch#166958 Approved by: https://github.com/mcfi
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:
Autovectorization of casting to bfloat16_t is broken in clang-[17, 20], fixed in clang-21.
We are adding a workaround vectorized code, which improves conversion speed from smaller int data types.
We've observed the following performance improvements, when compiling with clang-19 and targeting armv9a+sve2:
before:
uint8->bfloat16_t ===> 319.433us
int8->bfloat16_t ===> 320.216us
int16->bfloat16_t ===> 326.899us
int32->bfloat16_t ===> 327.925us
after:
uint8->bfloat16_t ===> 185.189us -----> 72% higher throughput
int8->bfloat16_t ===> 169.790us -----> 89% higher throughput
int16->bfloat16_t ===> 180.744us -----> 81% higher throughput
int32->bfloat16_t ===> 185.129us -----> 77% higher throughput
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: D86207189
cc @jgong5 @mingfeima @XiaobingSuper @sanchitintel @ashokei @jingxu10 @jerryzh168 @aditew01 @snadampal @milpuz01 @nikhil-arm @fadara01