CUDA is viable if each CUDA core runs its own copy of E@H

Message boards : Number crunching : CUDA is viable if each CUDA core runs its own copy of E@H

To post messages, you must log in.

AuthorMessage
HireMe.geek.nz

Send message
Joined: 23 Sep 07
Posts: 11
Credit: 203,838
RAC: 85
Message 1967 - Posted: 5 Mar 2011, 6:41:08 UTC

CUDA is viable if each CUDA core runs its own copy of E@H.

IF you have an MMX + SSE application with a wrapper-dispatcher that can sense CUDA and dispatch a secondary series of (internal) CUDA work units ... then CUDA might be viable.

I only see the CUDA doing the plugboard switching and indexing.
ID: 1967 · Rating: 0 · rate: Rate + / Rate - Report as offensive    Reply Quote
Dragokatzov

Send message
Joined: 17 Mar 11
Posts: 1
Credit: 13,637
RAC: 7
Message 1972 - Posted: 17 Mar 2011, 20:34:36 UTC - in response to Message 1967.  

I agree. even thought i just joined E@H today, I believe it to be a good idea to look at CUDA. CAL would be good too, but from my understanding, there's a real lack of development tools for it. Embrace GPGPU!
ID: 1972 · Rating: 0 · rate: Rate + / Rate - Report as offensive    Reply Quote
Profile TJM
Project administrator
Project developer
Project scientist
Avatar

Send message
Joined: 25 Aug 07
Posts: 843
Credit: 71,330,737
RAC: 382,122
Message 1981 - Posted: 29 Mar 2011, 15:28:54 UTC

The scoring and indexing runs fast on the GPU, it has been tested already long time ago.
Unfortunately the scoring functions need the input data from main loop (hillclimb()), which cannot be pararelized easily.
M4 Project homepage
M4 Project wiki
ID: 1981 · Rating: 0 · rate: Rate + / Rate - Report as offensive    Reply Quote

Message boards : Number crunching : CUDA is viable if each CUDA core runs its own copy of E@H




Copyright © 2017 TJM