crypto/internal/fips140/aes: optimize amd64 #151
Closed
+186
−30
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
Implement overflow-aware optimization in ctrBlocks8Asm: make a fast branch
in case when there is no overflow. One branch per 8 blocks is faster than
7 increments in general purpose registers and transfers from them to XMM.
Added AES-192 and AES-256 modes to the AES-CTR benchmark.
Added a correctness test in ctr_aes_test.go for the overflow optimization.
This improves performance, especially in AES-128 mode.
goos: windows
goarch: amd64
pkg: crypto/cipher
cpu: AMD Ryzen 7 5800H with Radeon Graphics
│ B/s │ B/s vs base
AESCTR/128/50-16 1.377Gi ± 0% 1.384Gi ± 0% +0.51% (p=0.028 n=20)
AESCTR/128/1K-16 6.164Gi ± 0% 6.892Gi ± 1% +11.81% (p=0.000 n=20)
AESCTR/128/8K-16 7.372Gi ± 0% 8.768Gi ± 1% +18.95% (p=0.000 n=20)
AESCTR/192/50-16 1.289Gi ± 0% 1.279Gi ± 0% -0.75% (p=0.001 n=20)
AESCTR/192/1K-16 5.734Gi ± 0% 6.011Gi ± 0% +4.83% (p=0.000 n=20)
AESCTR/192/8K-16 6.889Gi ± 1% 7.437Gi ± 0% +7.96% (p=0.000 n=20)
AESCTR/256/50-16 1.170Gi ± 0% 1.163Gi ± 0% -0.54% (p=0.005 n=20)
AESCTR/256/1K-16 5.235Gi ± 0% 5.391Gi ± 0% +2.98% (p=0.000 n=20)
AESCTR/256/8K-16 6.361Gi ± 0% 6.676Gi ± 0% +4.94% (p=0.000 n=20)
geomean 3.681Gi 3.882Gi +5.46%
The slight slowdown on 50-byte workloads is unrelated to this change,
because such workloads never use ctrBlocks8Asm.
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Outdated suggestions cannot be applied.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.
Implement overflow-aware optimization in ctrBlocks8Asm: make a fast branch
in case when there is no overflow. One branch per 8 blocks is faster than
7 increments in general purpose registers and transfers from them to XMM.
Added AES-192 and AES-256 modes to the AES-CTR benchmark.
Added a correctness test in ctr_test.go for the overflow optimization.
This improves performance, especially in AES-128 mode.
goos: windows
goarch: amd64
pkg: crypto/cipher
cpu: AMD Ryzen 7 5800H with Radeon Graphics
│ B/s │ B/s vs base
AESCTR/128/50-16 1.377Gi ± 0% 1.384Gi ± 0% +0.51% (p=0.028 n=20)
AESCTR/128/1K-16 6.164Gi ± 0% 6.892Gi ± 1% +11.81% (p=0.000 n=20)
AESCTR/128/8K-16 7.372Gi ± 0% 8.768Gi ± 1% +18.95% (p=0.000 n=20)
AESCTR/192/50-16 1.289Gi ± 0% 1.279Gi ± 0% -0.75% (p=0.001 n=20)
AESCTR/192/1K-16 5.734Gi ± 0% 6.011Gi ± 0% +4.83% (p=0.000 n=20)
AESCTR/192/8K-16 6.889Gi ± 1% 7.437Gi ± 0% +7.96% (p=0.000 n=20)
AESCTR/256/50-16 1.170Gi ± 0% 1.163Gi ± 0% -0.54% (p=0.005 n=20)
AESCTR/256/1K-16 5.235Gi ± 0% 5.391Gi ± 0% +2.98% (p=0.000 n=20)
AESCTR/256/8K-16 6.361Gi ± 0% 6.676Gi ± 0% +4.94% (p=0.000 n=20)
geomean 3.681Gi 3.882Gi +5.46%
The slight slowdown on 50-byte workloads is unrelated to this change,
because such workloads never use ctrBlocks8Asm.
Updates golang#76061
🔄 This is a mirror of upstream PR golang#76059