diff options
| author | Andy Polyakov <appro@openssl.org> | 2017-07-10 15:21:00 +0200 |
|---|---|---|
| committer | Andy Polyakov <appro@openssl.org> | 2017-07-24 23:29:46 +0200 |
| commit | 1843787173da9b07029d0863e236107b1dd4fdd7 (patch) | |
| tree | dd48012941253be6e112b85915f69bbf58b2a50d | |
| parent | evp/e_aes_cbc_hmac_sha256.c: give SHAEXT right priority. (diff) | |
| download | openssl-1843787173da9b07029d0863e236107b1dd4fdd7.tar.xz openssl-1843787173da9b07029d0863e236107b1dd4fdd7.zip | |
aes/asm/aesni-sha*-x86_64.pl: add SHAEXT performance results.
Reviewed-by: Kurt Roeckx <kurt@roeckx.be>
(Merged from https://github.com/openssl/openssl/pull/3898)
| -rw-r--r-- | crypto/aes/asm/aesni-sha1-x86_64.pl | 7 | ||||
| -rw-r--r-- | crypto/aes/asm/aesni-sha256-x86_64.pl | 15 |
2 files changed, 15 insertions, 7 deletions
diff --git a/crypto/aes/asm/aesni-sha1-x86_64.pl b/crypto/aes/asm/aesni-sha1-x86_64.pl index de4c410f37..b01a4c55c8 100644 --- a/crypto/aes/asm/aesni-sha1-x86_64.pl +++ b/crypto/aes/asm/aesni-sha1-x86_64.pl @@ -34,6 +34,8 @@ # Haswell 4.43[+3.6(4.2)] 8.00(8.58) 4.55(5.21) +75%(+65%) # Skylake 2.63[+3.5(4.1)] 6.17(6.69) 4.23(4.44) +46%(+51%) # Bulldozer 5.77[+6.0] 11.72 6.37 +84% +# Ryzen(**) 2.71[+1.93] 4.64 2.74 +69% +# Goldmont(**) 3.82[+1.70] 5.52 4.20 +31% # # AES-192-CBC # Westmere 4.51 9.81 6.80 +44% @@ -47,13 +49,16 @@ # Sandy Bridge 7.05 12.06(13.15) 7.12(7.72) +69%(+70%) # Ivy Bridge 7.05 11.65 7.12 +64% # Haswell 6.19 9.76(10.34) 6.21(6.25) +57%(+65%) -# Skylake 3.62 7.16(7.68) 4.56(4.76) +57%(+61$) +# Skylake 3.62 7.16(7.68) 4.56(4.76) +57%(+61%) # Bulldozer 8.00 13.95 8.25 +69% +# Ryzen(**) 3.71 5.64 3.72 +52% +# Goldmont(**) 5.35 7.05 5.76 +22% # # (*) There are two code paths: SSSE3 and AVX. See sha1-568.pl for # background information. Above numbers in parentheses are SSSE3 # results collected on AVX-capable CPU, i.e. apply on OSes that # don't support AVX. +# (**) SHAEXT results. # # Needless to mention that it makes no sense to implement "stitched" # *decrypt* subroutine. Because *both* AESNI-CBC decrypt and SHA1 diff --git a/crypto/aes/asm/aesni-sha256-x86_64.pl b/crypto/aes/asm/aesni-sha256-x86_64.pl index 74ec844952..ef46023710 100644 --- a/crypto/aes/asm/aesni-sha256-x86_64.pl +++ b/crypto/aes/asm/aesni-sha256-x86_64.pl @@ -28,18 +28,21 @@ # for standalone AESNI-CBC encrypt, standalone SHA256, and stitched # subroutine: # -# AES-128/-192/-256+SHA256 this(**)gain -# Sandy Bridge 5.05/6.05/7.05+11.6 13.0 +28%/36%/43% -# Ivy Bridge 5.05/6.05/7.05+10.3 11.6 +32%/41%/50% -# Haswell 4.43/5.29/6.19+7.80 8.79 +39%/49%/59% -# Skylake 2.62/3.14/3.62+7.70 8.10 +27%/34%/40% -# Bulldozer 5.77/6.89/8.00+13.7 13.7 +42%/50%/58% +# AES-128/-192/-256+SHA256 this(**) gain +# Sandy Bridge 5.05/6.05/7.05+11.6 13.0 +28%/36%/43% +# Ivy Bridge 5.05/6.05/7.05+10.3 11.6 +32%/41%/50% +# Haswell 4.43/5.29/6.19+7.80 8.79 +39%/49%/59% +# Skylake 2.62/3.14/3.62+7.70 8.10 +27%/34%/40% +# Bulldozer 5.77/6.89/8.00+13.7 13.7 +42%/50%/58% +# Ryzen(***) 2.71/-/3.71+2.05 2.74/-/3.73 +74%/-/54% +# Goldmont(***) 3.82/-/5.35+4.16 4.73/-/5.94 +69%/-/60% # # (*) there are XOP, AVX1 and AVX2 code paths, meaning that # Westmere is omitted from loop, this is because gain was not # estimated high enough to justify the effort; # (**) these are EVP-free results, results obtained with 'speed # -evp aes-256-cbc-hmac-sha256' will vary by percent or two; +# (***) these are SHAEXT results; $flavour = shift; $output = shift; |