Re: AES optimized for x86-64

2006-03-16 Thread Herbert Xu 
On Wed, Mar 15, 2006 at 10:00:09PM +0100, Dag Arne Osvik wrote:

 Using the best average encryption time per block from tcrypt's speed 
 tests gave these numbers:
 
 ECB on Intel288.5
 ECB on AMD  286.1
 
 CBC on Intel510.4
 CBC on AMD  334.2
 
 Sorry for the confusion..  All of these numbers are for code currently 
 in the kernel.  However, it seems like only ECB on Intel used the 
 assembly version.  With the asm module running also on the AMD machine, 
 time for ECB on AMD dropped to 187.5 cycles per block, though CBC still 
 runs at 334.2 per block.

Is the 187.5 figure with your code or the existing x86-64 asm module?
If it's the existing one, what are the figures above then?

Cheers,
-- 
Visit Openswan at http://www.openswan.org/
Email: Herbert Xu ~{PmVHI~} [EMAIL PROTECTED]
Home Page: http://gondor.apana.org.au/~herbert/
PGP Key: http://gondor.apana.org.au/~herbert/pubkey.txt
-
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

Re: AES optimized for x86-64

2006-03-15 Thread Dag Arne Osvik 

Herbert Xu wrote:

On Wed, Mar 15, 2006 at 03:32:33PM +0100, Dag Arne Osvik wrote:


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 Intel288.5
ECB on AMD  286.1

CBC on Intel510.4
CBC on AMD  334.2



Sorry I'm confused.  Which one is the existing x86-64 assembly
implementation?


Sorry for the confusion..  All of these numbers are for code currently 
in the kernel.  However, it seems like only ECB on Intel used the 
assembly version.  With the asm module running also on the AMD machine, 
time for ECB on AMD dropped to 187.5 cycles per block, though CBC still 
runs at 334.2 per block.


My code isn't yet ready to be tested by tcrypt, and the number I gave 
for my code (233 cycles/block on Intel's Nocona) is from a simple 
feedback loop, essentially doing CBC encryption of the zero string.



  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