Rosetta optimization (SSE2...) ?

Message boards : Number crunching : Rosetta optimization (SSE2...) ?

To post messages, you must log in.

AuthorMessage
Zilli Samuel

Send message
Joined: 2 Mar 06
Posts: 3
Credit: 30,077
RAC: 0
Message 11608 - Posted: 3 Mar 2006, 16:19:27 UTC

Why don't you put optimized versions (SSE2 for Win and Linux for example) in a web page, so users can download a faster application if they own a fast CPU?

This is already done for Seti project, and I think it can be very useful...
ID: 11608 · Rating: -1 · rate: Rate + / Rate - Report as offensive    Reply Quote
Scribe
Avatar

Send message
Joined: 2 Nov 05
Posts: 284
Credit: 157,359
RAC: 0
Message 11609 - Posted: 3 Mar 2006, 16:26:03 UTC

Rosie does not have any optimised versions and as far as I know has no intention of releasing the source code to allow others to do them......
ID: 11609 · Rating: 0 · rate: Rate + / Rate - Report as offensive    Reply Quote
Zilli Samuel

Send message
Joined: 2 Mar 06
Posts: 3
Credit: 30,077
RAC: 0
Message 11611 - Posted: 3 Mar 2006, 16:44:29 UTC - in response to Message 11609.  

Rosie does not have any optimised versions and as far as I know has no intention of releasing the source code to allow others to do them......


Yes, I know there isn't any optimized version.
The reason for my post is that I hope rosetta's programmers read this request so they could compile an optimized version and put a link to download and install it...
ID: 11611 · Rating: 0 · rate: Rate + / Rate - Report as offensive    Reply Quote
BennyRop

Send message
Joined: 17 Dec 05
Posts: 555
Credit: 140,800
RAC: 0
Message 11623 - Posted: 3 Mar 2006, 21:03:33 UTC

Some projects don't use much code that can take advantage of SSE or SSE2. At Distributed Folding, Howard pointed out that 90% of the time the DF client spent traversing pointers. The move to 64 bit clients ended up with a client that was half the speed of the 32 bit client, not twice the speed, which ruled out the argument that 64 bit clients were always better.

It would be interesting to see what kind of improvement in speed a SSE2 optimized client would have on Rosetta - but it may be too tiny an improvement to justify releasing it.

Perhaps David Kim or Dr David Baker could give a response on the usefulness of SSE for their algorithm?


ID: 11623 · Rating: 0 · rate: Rate + / Rate - Report as offensive    Reply Quote
Lee Carre

Send message
Joined: 6 Oct 05
Posts: 96
Credit: 79,331
RAC: 0
Message 11985 - Posted: 13 Mar 2006, 18:03:02 UTC

i use BoincView which can tell me the FLOPS (also called other things like FIOps)

rosetta wasn't bad before the recently released apps, it was just above par with other apps for sheer processing speed

but with v4.82 well, the speed of some units is pretty damn fast, so i think the code is pretty efficient already

currently rosetta is the fastest app i've seen (i crunch for most but not all projects)
Want to search the BOINC Wiki, BOINCstats, or various BOINC forums from within firefox? Try the BOINC related Firefox Search Plugins
ID: 11985 · Rating: 1 · rate: Rate + / Rate - Report as offensive    Reply Quote

Message boards : Number crunching : Rosetta optimization (SSE2...) ?



©2024 University of Washington
https://www.bakerlab.org