[rc5] RC5-64 v2.6401 client speed vs RC5-56 v2.004

Tim Charron tcharron at interlog.com
Thu Oct 23 07:36:50 EDT 1997

Joey Sum <jpsum1 at pop.uky.edu> wrote:
> 1. The speed of the RC5-64 v2.6401 client (262 kk/s) is slower than the RC5-56
> v2.004 (280 kk/s) on my Cyrix 6x86-166+.  In both cases, I use the Cyrix
> specific code.

The benchmark timing is more accurate, so it is possible that the 
reported "-benchmark" speed differs while the actual performance does 
not.  Run a few normal blocks and see how long it takes relative 
to rc5v2 2.004 for a more accurate test.  As well, nice settings 
are slightly different in the win32 client...

> 2. The nice settings seem to be different in the RC5-64 v2.6401 client.  I ran
> the RC5-56 v2.004 client on nice 2 (normal) and never noticed any slowdown or
> lag.  Running the RC5-64 v2.6401 client on nice 2 shows a bit of lag and
> slower performance when trying to run something else.  (Nice 1 is fine).

As caspar pointed out, nice 0 is actually supernice in the win32 
client.  In 2.004, nice 0 and nice 1 were the same.  In this version, 
nice 1 remains the same, but nice 0 is actually an even lower 
priority.  This was to address a couple of concerns that there are 
(poorly written) apps that did not run when rc5v2 2.004 was running 
at nice level 0 (video capture software and java garbage collection 
are 2 examples that come to mind).  Nice level 2 has not changed -- 
it will compete for cpu time with all other apps.

-- Tim
Tim Charron
tcharron at interlog.com
tcharron at ctfinance.com

To unsubscribe, send email to majordomo at llamas.net with 'unsubscribe rc5' in the body.

More information about the rc5 mailing list