[rc5] 2.004 Bug (?) Report

Sean McPherson sean at ntr.net
Fri Aug 8 12:56:53 EDT 1997


> 
> Actually, I've done a benchmark with all the available optimisations,
> and despite misidentifying my AMD 486-160 as a 6x86, the client has it
> right... the 6x86 code is fastest on my machine. I imagine the code for
> different processors works better with some cacheing architectures or
> some such, and thus the "wrong" code is sometimes best.
> 
> Steff

Well, I tested all of them, both with the -benchmark option and with
actual key blocks, and when I use the 'true' Pentium code by giving a
specific type under -config, It's about 20k keys/sec faster! 246k vs 267k.
Overall, this Dual P200 classic is running a total of 540k keys/sec, so
I'm fairly well pleased :) 

--
Sean McPherson
sean at ntr.net
Systems Administration
NTR.Net Corporation
--
REALITY.SYS corrupted. Reboot universe? [Y n]
----
To unsubscribe, send email to majordomo at llamas.net with 'unsubscribe rc5' in the body.



More information about the rc5 mailing list