Drtivá kritika
srovnávací studie HW výkonnosti kandidát? SHA-3
Poznámka: Uvedený link ješt? neobsahuje tuto kritiku, která byla zaslána do emailové conference hash-forum@nist.gov teprve p?ed pár hodinami. Nicmén? tam bude jist? dnes nebo zítra za?azena. Pro zájemce uvádím výtah zemailu.
Když jsem psal ?lánek, který vyjde také dnes ve?er v Crypto-Worldu, o této kritice jsem nev?d?l, ale krásn? tam zapadá.
Vý?atek z emailu:
Thomas Pornins paper Comparative performance review of the SHA-3
second-round candidates (dated 21 June 2010) reports slow-sounding performance figures for the sphlib implementation of CubeHash on four CPUs: Intel Q6600, PowerPC 750, Broadcom BCM3302, and ARM920T. The point of this OFFICIAL COMMENT is that the
documented by Pornin is actually a mistake in the sphlib implementation. I have now posted software that avoids this mistake and thats much faster than the sphlib implementation on a very wide range of CPUs....
The slowness of the sphlib implementation ... is of no relevance to a hash-function-speed competition.
...The actual speeds of (other) SHA-3 candidates on these CPUs will often be much faster than the sphlib speeds ...
.... The eBASH benchmarks (utrecht, 20100802) show
CubeHash16/32 running at 13.24 cycles/byte. The sphlib implementation runs between three and five times
slower...