Message boards :
Number crunching :
question and or idea.....
Message board moderation
Author | Message |
---|---|
Peter M. Ferrie Send message Joined: 28 Mar 03 Posts: 86 Credit: 9,967,062 RAC: 0 |
is it theoretically possible to have a multi core pc, example: (a 2 core or more (atom up to current i7 cpus)) all work on the same workunit together at the same time, to dramatically reduce time it takes for a single core to compute a workunit? what would need to be done make boinc(seti) able to do this? thoughts? comments? |
Mike Send message Joined: 17 Feb 01 Posts: 34257 Credit: 79,922,639 RAC: 80 |
No, its not possible atm. It would require new science apps. Joe, Raistmer and or Jason can explain this much better. With each crime and every kindness we birth our future. |
Aurora Borealis Send message Joined: 14 Jan 01 Posts: 3075 Credit: 5,631,463 RAC: 0 |
No, its not possible atm. My understanding is that the SETI analysis is not conducive to paralleling. I'll leave it to the Lunatic Pros to chime in. |
Richard Haselgrove Send message Joined: 4 Jul 99 Posts: 14650 Credit: 200,643,578 RAC: 874 |
No, its not possible atm. What do you think we do on GPUs? They are parallel devices, par excellence. SETI probably isn't appropriate for the old-style symmetric multi-processor (SMP) style of parallel programming, but it's been suggested that OpenCL could use multiple CPUs in the same way that it uses multiple compute units within a GPU. For traditional (x86-style) CPUs, there probably isn't any point of incurring the extra overhead of the middleware - we have efficient enough CPU applications to perform the work required in a timely fashion on single cores. If OpenCL drivers/runtime support are available, there might be some benefit of from running multi-core applications on Atom and similar devices, which might struggle to meet deadlines in single-core mode. Recent versions of BOINC already have the capability to schedule OpenCL on CPUs, though I'm not aware of anyone testing that facility yet. |
Mike Send message Joined: 17 Feb 01 Posts: 34257 Credit: 79,922,639 RAC: 80 |
No, its not possible atm. In case of programming technique its not the case. If so i could use all my 8 cores plus 3 instances on my GPU to proccess the same task. That was the original question and this is not possible atm. It was discuessed a few years ago at Lunatics that this isn`t an easy task. With each crime and every kindness we birth our future. |
arkayn Send message Joined: 14 May 99 Posts: 4438 Credit: 55,006,323 RAC: 0 |
No, its not possible atm. John over at Collatz does have a OpenCL CPU app available. |
Wiggo Send message Joined: 24 Jan 00 Posts: 34744 Credit: 261,360,520 RAC: 489 |
No, its not possible atm. But how efficient is it compared to just straight out running it without a 3rd party app involved? Cheers. |
Jord Send message Joined: 9 Jun 99 Posts: 15184 Credit: 4,362,181 RAC: 3 |
I'm testing one v6.04 Mini, running on 3 CPU cores of my i5-2500K. It's kind of cool looking: I see these run 'cooler' on my CPU than Multibeam or Astropulse do. Those tend to be around 60 Celsius, these OpenCL only 56 Celsius. |
Wiggo Send message Joined: 24 Jan 00 Posts: 34744 Credit: 261,360,520 RAC: 489 |
Cooler maybe, but maybe not so efficient as the old Collatz app running on a single core? Cheers. |
Jord Send message Joined: 9 Jun 99 Posts: 15184 Credit: 4,362,181 RAC: 3 |
I got three tasks, two single core CPU Mini Collatz tasks and one for OpenCL_CPU. The single core ones sit waiting at 3h 17m 56s initial estimate. While its initial run time was 2 hours 5 minutes on the OpenCL task, best estimates say it's going to take about 2 hours. But I'll have to run one of those single core tasks after it to see what it results into. Will do that tonight. Have to be places in a bit and won't be back home until after 6pm. Ooh... also see it continues to run even with BOINC now set to Suspend - time of day. |
Jord Send message Joined: 9 Jun 99 Posts: 15184 Credit: 4,362,181 RAC: 3 |
Damn, in the end I didn't see what actual elapsed time it had. The task's values are diluted, as run time and CPU time show as 20,902.81 seconds. However, the stderr.txt speaks of CPU time: 34246.5 seconds and Elapsed time: 7545.7 seconds. The latter may be the closest, the last time I checked it the elapsed time was an hour and 47 minutes with 17 minutes to go. But then things sped up, and around the moment supreme I got a phone call. |
©2024 University of California
SETI@home and Astropulse are funded by grants from the National Science Foundation, NASA, and donations from SETI@home volunteers. AstroPulse is funded in part by the NSF through grant AST-0307956.