[openssl-users] openssl-1.1.0 sha1 performance

[c^]

Afternoon,

I have been running some speed tests of openssl 1.0.1, 1.0.2 and 1.1.0
versions against various compiler optimisations. Special interest was given
to the more commonly used primitives, rsa's, aes's etc.

I noticed that SHA1's have some significant performance improvements.
However the multiplier by which it improved by diminishes as you approach
8k/16k block sizes.

Any ideas why this tails off?

I noticed no other 'statistically significant' change in other primitives,
although freely admit i have not exhaustively checked.

openssl-1.0.2g-native-speed.txt:Doing sha1 for 3s on 16 size blocks:
9225205 sha1's in 3.00s
openssl-1.0.2g-native-speed.txt:Doing sha1 for 3s on 64 size blocks:
7275849 sha1's in 3.00s
openssl-1.0.2g-native-speed.txt:Doing sha1 for 3s on 256 size blocks:
4821329 sha1's in 3.00s
openssl-1.0.2g-native-speed.txt:Doing sha1 for 3s on 1024 size blocks:
2059373 sha1's in 3.00s
openssl-1.0.2g-native-speed.txt:Doing sha1 for 3s on 8192 size blocks:
327032 sha1's in 3.00s
openssl-1.1.0-pre4-native-speed.txt:Doing sha1 for 3s on 16 size blocks:
23362218 sha1's in 3.00s
openssl-1.1.0-pre4-native-speed.txt:Doing sha1 for 3s on 64 size blocks:
14131714 sha1's in 2.99s
openssl-1.1.0-pre4-native-speed.txt:Doing sha1 for 3s on 256 size blocks:
7166139 sha1's in 3.00s
openssl-1.1.0-pre4-native-speed.txt:Doing sha1 for 3s on 1024 size blocks:
2413233 sha1's in 3.00s
openssl-1.1.0-pre4-native-speed.txt:Doing sha1 for 3s on 8192 size blocks:
335803 sha1's in 3.00s
openssl-1.1.0-pre4-native-speed.txt:Doing sha1 for 3s on 16384 size blocks:
169210 sha1's in 3.00s

I assume the positive change was part of:
  *) Extensive assembler packs updates, most notably:

- x86[_64]:     AES-NI, PCLMULQDQ, RDRAND support;
- x86[_64]:     SSSE3 support (SHA1, vector-permutation AES);
- x86_64:       bit-sliced AES implementation;
- ARM:          NEON support, contemporary platforms optimizations;
- s390x:        z196 support;
- *:            GHASH and GF(2^m) multiplication implementations;

     [Andy Polyakov]

Has anyone else completed any similar tests?

Thank you,
CraigT
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mta.openssl.org/pipermail/openssl-users/attachments/20160411/876d75eb/attachment.html>