Performance of SHA-3 hash functions on Intel processors (Core 2 platform vs Itanium)

Performance of SHA-3 hash functions on Intel processors (Core 2 platform vs Itanium)

Hi,

I am involved in NIST SHA-3 competition for the next cryptographic hash standard as a designer of two functions (Edon-R and Blue Midnight Wish). http://csrc.nist.gov/groups/ST/hash/sha-3/index.html

An extensive benchmarking of various hash functions can be done by a package supercop that can be found at: http://bench.cr.yp.to/results-hash.html

An interesting comparison can be done between performances of the functions on Core 2 and on Itanium platforms. Unfortunately tested Itanium platforms are relatively old (that is the reason of this post).

Example:
Platform Intel Core 2 Duo E4600 (6fd), cobra, supercop-20100120

Cycles/byte for long messages
quartilemedianquartilehash
2.442.502.50edonr512
3.453.463.46md4
3.843.843.84bmw512
4.644.724.73edonr256
5.465.465.47md5
6.286.286.32tiger
6.416.426.42cubehash832
6.476.476.47skein512
7.137.147.19shabal512
7.357.417.49bmw256
7.527.537.57sha1
7.847.887.95sarmal256
9.379.399.51blake64
9.359.439.57cheetah256
9.569.579.59sarmal512
9.829.8610.08blake32...

Platform:
997MHz, HP Itanium II, nmi0020, supercop-20090205

Cycles/byte for long messages
quartilemedianquartilehash
1.221.221.23edonr512
2.782.792.80bmw512
3.423.423.43edonr256
4.804.814.82keccakr1024c576
5.585.585.58md4
7.297.307.30skein512
7.847.857.86bmw256
8.548.548.55blake64
9.329.339.33md5
9.319.339.34sha512
11.1511.1611.16sha1
12.5312.5512.55shabal512
20.6820.6920.69blake32
21.0421.0821.09sha256...

Some dramatic speed-ups and dramatic slow-downs can be detected for various functions. So, I am wondering what would be the performance of the hash functions on the new Itanium - Tukwila? Off coarse performance on new upcomming Intel processors is also of interest. Are there some users in this forum (or maybe Intel engineers) that can run the supercop package on a Tukwila platform and to report the results?

Regards,
Prof. Danilo Gligoroski

2 posts / 0 new
Last post
For more complete information about compiler optimizations, see our Optimization Notice.

Update:

Recent measurement on the new i5 processor shows signifficant improvement over Core 2 performances.

Platform:
2400MHz, Intel Core i5 M 520 (20652), thinkaes,
supercop-20100120
Cycles/byte for long messages
quartilemedianquartilehash
2.102.102.11edonr512
3.333.333.34bmw512
3.883.893.89edonr256
4.134.134.13cubehash832
4.13?4.98?4.98?md5
5.075.085.08skein512
5.095.105.13tiger
5.465.465.46shabal512
5.935.935.93blake32
6.476.476.51sarmal256
6.506.506.50bmw256
7.507.517.51simd256
7.787.797.79blake64
7.967.967.97sarmal512
8.19?8.20?9.91?cheetah256
8.238.238.23cubehash1632
8.268.268.26cubehash816
8.488.548.56lux512
8.618.638.63simd512
9.249.269.28lux256
10.0910.0910.11keccakc512
10.4710.4810.48sha512
Regards,
Prof. Danilo Gligoroski

Leave a Comment

Please sign in to add a comment. Not a member? Join today