Message boards :
LHC@home Science :
GPU Computing
Message board moderation
Previous · 1 · 2
| Author | Message |
|---|---|
 Michael H.W. WeberSend message Joined: 18 Sep 04 Posts: 30 Credit: 5,100,929 RAC: 0  
|
Well, I think maybe ARM might even be interested in funding a student for a few months to port code to MALI GPU to demonstrate their GPUs capabilities in the context of a significant project (SixTRack/LHC/CERN: outreach!) if this is not an intersting enough topic for some Google Summer of code student... Michael. |
|
Send message Joined: 25 Jan 17 Posts: 27 Credit: 3,258,853 RAC: 0  |
SixTrack already runs on arm. For GPUs, this is more involved, so whenever I (or somebody else) gets some time to finish re-writing most of the code. Can't really give an ETA. |
|
Send message Joined: 12 Jul 11 Posts: 857 Credit: 1,619,050 RAC: 0 |
Well we have tried GPUs and results are not brilliant to say the least. For the moment I think it is better to let more efficient apps use the GPU machines. Nonetheless, parallelisation of SixTrack is almost a must although we achieve effective parallelism by just running many WUs at the same time. Excuses, excuses, but we must first get out ARM, AVX, support as well as more physics capability and manage the switch to gfortran. Not forgetting we have several server side and workload management issues to resolve. Eric. |
|
Michael H.W. Weber Send message Joined: 18 Sep 04 Posts: 30 Credit: 5,100,929 RAC: 0
|
SixTrack already runs on arm. I just saw on the application page that, indeed, on May 4th a SixTrack ARM client for both Linux- and Android-on-ARM was released in beta status. If you had announced this on both the LHC@home main page plus in the BOINC clients info panel, a lot more people might already have known and helped testing it. Michael. [edit]: Well, it seems testing isn't an option as at least currently no work for that ARM app appears available. |
|
Send message Joined: 12 Jul 11 Posts: 857 Credit: 1,619,050 RAC: 0 |
You are correct, but we are having problems with the sixtracktest infrastructure. More news soonest. Eric. |
|
Michael H.W. Weber Send message Joined: 18 Sep 04 Posts: 30 Credit: 5,100,929 RAC: 0
|
Well we have tried GPUs and results are not brilliant to say the least. So, then what GPUs have been tested and what was the speed increase factor vs. electricity consumption? I know very well from our own RNA World DC project that there is plenty of code which simply does not benefit from porting to GPUs at all. Michael. |
|
Send message Joined: 12 Jul 11 Posts: 857 Credit: 1,619,050 RAC: 0 |
I shall ask my colleague Riccardo for details (currently on a business trip). Eric. |
|
Michael H.W. Weber Send message Joined: 18 Sep 04 Posts: 30 Credit: 5,100,929 RAC: 0
|
Any news? Michael. |
©2024 CERN