Dag Arne Osvik wrote:
Herbert Xu wrote:

On Wed, Mar 15, 2006 at 08:57:48AM +0100, Dag Arne Osvik wrote:

I've attached an implementation of AES optimized for 64-bit Pentium 4E. The round function achieves 3 ?ops per cycle, so even with function call overhead its performance is quite good, with best average time for 10000 consecutive encryptions at <233 cycles per block. I'll make it ready for inclusion in the Linux kernel later. In the meantime, comments are welcome.



Cool.  How does this compare with the existing x86-64 AES implementation?


Using the best average encryption time per block from tcrypt's speed tests gave these numbers:

ECB on Intel    288.5
ECB on AMD    286.1

CBC on Intel    510.4
CBC on AMD    334.2

Sorry.  Those AMD numbers are for the C version, not asm.

Btw, those 510 cycles for CBC are rather surprising. Maybe ECB is using the assembler version, while CBC uses C?


  Dag Arne
-
To unsubscribe from this list: send the line "unsubscribe linux-crypto" in
the body of a message to [EMAIL PROTECTED]
More majordomo info at  http://vger.kernel.org/majordomo-info.html

Reply via email to