Tue Nov 20 16:20:02 2001
Marc Mutz wrote:
>-----BEGIN PGP SIGNED MESSAGE-----
>On Monday 19 November 2001 18:43, Lee McKenna wrote:
>>For the record book :) I ran the following tests using gpg 1.0.6,
>>1024bit key, on a PIII/800, 256MB RAM, SCSI/IDE RAID, stock FreeBSD
>>4.3 release, on a file size 3363088179 (3.3GB). CAST5 is the
>>fastest, 3DES slowest. All others are relatively close.
>If you use gnupg from rpm's (or equivalent on FreeBSD), you could try
>recompiling gpg with --march=pentiumpro. This should speed things up a
>bit (maybe even twice; by base64 encoder doubled it's speed by using
>this option). Alternatively, you could use icc, if that runs under
>FreeBSD. That should give you another 15-20%.
I installed the FreeBSD "package", so I am not sure of the compile time
flags that were used. I will grab the sources and try a recompile.
When I use gpg on these large files the CPU is pegged at 98%, and disk
i/o is just under 5MB/s. I'm pretty much convinved at this point that
encrypt/decrypt speed is bound by the CPU and the only thing I can do is
get a faster processor.
What is icc, and how could that help my cause?
Thanks for your help, it is much appreciated.