lkml.org 
[lkml]   [2007]   [Jun]   [8]   [last100]   RSS Feed
Views: [wrap][no wrap]   [headers]  [forward] 
 
Messages in this thread
/
From
Subject[PATCH 0/3] Add optimized SHA-1 implementations for x86 and x86_64
Date
The following 3-part series adds assembly implementations of the SHA-1
transform for x86 and x86_64. For x86_64 the optimized code is always
selected; on x86 it is selected if the kernel is compiled for i486 or above
(since the code needs BSWAP). These changes primarily improve the
performance of the CryptoAPI SHA-1 module and of /dev/urandom. I've
included some performance data from my test boxes below.

This version incorporates feedback from Herbert Xu. Andrew, I'm sending
this to you because of the (admittedly tiny) intersection with arm and s390
in part 1.

-

tcrypt performance tests:

=== Pentium IV in 32-bit mode, average of 5 trials ===
Test# Bytes/ Bytes/ Cyc/B Cyc/B Change
block update (C) (asm)
0 16 16 229 114 50%
1 64 16 142 76 46%
2 64 64 79 35 56%
3 256 16 59 34 42%
4 256 64 44 24 45%
5 256 256 43 17 60%
6 1024 16 51 36 29%
7 1024 256 30 13 57%
8 1024 1024 28 12 57%
9 2048 16 66 30 55%
10 2048 256 31 12 61%
11 2048 1024 27 13 52%
12 2048 2048 26 13 50%
13 4096 16 49 30 39%
14 4096 256 28 12 57%
15 4096 1024 28 11 61%
16 4096 4096 26 13 50%
17 8192 16 49 29 41%
18 8192 256 27 11 59%
19 8192 1024 26 11 58%
20 8192 4096 25 10 60%
21 8192 8192 25 10 60%

=== Intel Core 2 in 64-bit mode, average of 5 trials ===
Test# Bytes/ Bytes/ Cyc/B Cyc/B Change
block update (C) (asm)
0 16 16 112 81 28%
1 64 16 55 39 29%
2 64 64 42 27 36%
3 256 16 35 25 29%
4 256 64 24 14 42%
5 256 256 22 12 45%
6 1024 16 31 22 29%
7 1024 256 17 9 47%
8 1024 1024 16 9 44%
9 2048 16 30 22 27%
10 2048 256 16 8 50%
11 2048 1024 16 8 50%
12 2048 2048 16 8 50%
13 4096 16 29 21 28%
14 4096 256 16 8 50%
15 4096 1024 15 8 47%
16 4096 4096 15 7 53%
17 8192 16 29 22 24%
18 8192 256 16 8 50%
19 8192 1024 15 7 53%
20 8192 4096 15 7 53%
21 8192 8192 15 7 53%

I've also done informal tests on other boxes, and the performance
improvement has been in the same ballpark.

On the aforementioned Pentium IV, /dev/urandom throughput goes from 3.7 MB/s
to 5.6 MB/s with the patches; on the Core 2, it increases from 5.5 MB/s to
8.1 MB/s.

Signed-off-by: Benjamin Gilbert <bgilbert@cs.cmu.edu>
-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at http://vger.kernel.org/majordomo-info.html
Please read the FAQ at http://www.tux.org/lkml/

\
 
 \ /
  Last update: 2007-06-09 00:31    [from the cache]
©2003-2011 Jasper Spaans