[rc5] Linux Client Optimizations...
bshoe at plinet.net
Fri Jun 20 13:50:17 EDT 1997
I'm not sure if you know who I am, but you did once respond to a question I
had asked on the mailing list about why DESCHALL is much faster than the
rc5 clients for comparable platforms. Since then, I've read more about how
the algorythems work and it makes since.
However, I have been working on optimizing the Linux Client(based on the
-original- genx source) to see if I can make it faster. I have "borrowed"
ideas from other rc5 clients, mostly TimC's win32 clients, and tried to
implement them into this version.
Since then I have developed a client that from my tests is about 5% to 10%
faster on the Intel/Linux platform(PPro 200) than the original genx client.
However, the speed still does not compare to the optimization you've
included into the "rc5-linux-i486-bovine". I was wondering if it would be
possible if you could share with me some of what you've done to your client
to make it faster.
Currently, what I have built into the version I have developed are the
keyspace buffering: i noticed you guys had plans to put this in all
version, so i just jumped right in and did it too ;-)
multithreading: based on the POSIX thread library, however, since
LinuxThreads is not 100% POSIX compatible it is not portable
to any other platforms.(I only have access to Linux machines,
I still need to work on streamlining these two functions, but currently,
then seem to behaving well.
Basically, its like this, I can do alot in C, but I'm not well versed in
asm, so I wouldn't be able to think of any new ways to increased the speeds
in this area.
At any rate, just thought I'd let you know what one of us "non-official"
developers is up to.
- Ben Schumacher <bshoe at plinet.net>
To unsubscribe, send email to majordomo at llamas.net with 'unsubscribe rc5' in the body.
More information about the rc5