GPU crunching

Author	Message
mimo Send message Joined: 2 Jan 07 Posts: 4 Credit: 22,491 RAC: 0	Message 18945 - Posted: 31 Jan 2008, 15:34:07 UTC can be LHC ported to GPUs ? i think it is perfectly suitable for paralell computing aka folding@home .... ID: 18945 · Reply Quote

PovAddict Send message Joined: 14 Jul 05 Posts: 275 Credit: 49,291 RAC: 0	Message 18946 - Posted: 31 Jan 2008, 18:24:40 UTC - in response to Message 18945. can be LHC ported to GPUs ? i think it is perfectly suitable for paralell computing aka folding@home .... Do you know what kind of calculations LHC does, or what makes you think it's "perfectly suitable"? ID: 18946 · Reply Quote

mimo Send message Joined: 2 Jan 07 Posts: 4 Credit: 22,491 RAC: 0	Message 18947 - Posted: 31 Jan 2008, 20:15:32 UTC i think this : it is particle physics simulator - so there are calculations of particle interaction - aka particle generators and collision detection in graphics engines - so this can be accelerated on GPU ... this is my view .... ID: 18947 · Reply Quote

Alex Send message Joined: 2 Sep 04 Posts: 378 Credit: 10,765 RAC: 0	Message 18948 - Posted: 1 Feb 2008, 3:22:27 UTC It's actually a good idea to consider GPU computing these days. In 2004, GPU computing would not have been a good fit for this, but today's GPU's now have much better math performance. A quick google search of 'Nvidia Floating Point' shows results that they now have 128 bit double precision floating point math, which could come in handy. I don't see them implementing this for Sixtrack because they have yet to get a fortran compiler that would work for Mac computers, let along GPU's. I'm not the LHC Alex. Just a number cruncher like everyone else here. ID: 18948 · Reply Quote

Keck_Komputers Send message Joined: 1 Sep 04 Posts: 275 Credit: 2,652,452 RAC: 0	Message 18950 - Posted: 1 Feb 2008, 7:59:05 UTC Unfortuantely this project is very poorly suited for running on GPUs. The calculations require very accurate and consistant results. GPUs generally trade this accuracy and consistancy for speed. BOINC WIKI BOINCing since 2002/12/8 ID: 18950 · Reply Quote

mimo Send message Joined: 2 Jan 07 Posts: 4 Credit: 22,491 RAC: 0	Message 18952 - Posted: 1 Feb 2008, 8:39:19 UTC yeah if double precision needed, there is problem ... CUDA is for single precision IEEE compliant ... ID: 18952 · Reply Quote

Alex Send message Joined: 2 Sep 04 Posts: 378 Credit: 10,765 RAC: 0	Message 18953 - Posted: 3 Feb 2008, 3:04:52 UTC - in response to Message 18952. yeah if double precision needed, there is problem ... CUDA is for single precision IEEE compliant ... Damn.. I blame google for giving me the wrong answer. I'm not the LHC Alex. Just a number cruncher like everyone else here. ID: 18953 · Reply Quote

LHC@home