Run One task instance on all cores

Questions and Answers : Wish list : Run One task instance on all cores

To post messages, you must log in.

AuthorMessage
Profile robl9x0

Send message
Joined: 9 Dec 18
Posts: 9
Credit: 6,527,023
RAC: 0
Message 97260 - Posted: 6 Jun 2020, 12:03:58 UTC

Is it possible to run a singular task on all of your CPU's cores and threads? I want to try this mostly for fun. Also does anyone know how the Rosetta computation works, I would love to know.
ID: 97260 · Rating: 0 · rate: Rate + / Rate - Report as offensive    Reply Quote
Brian Nixon

Send message
Joined: 12 Apr 20
Posts: 293
Credit: 8,432,366
RAC: 0
Message 97264 - Posted: 6 Jun 2020, 21:52:23 UTC - in response to Message 97260.  

You might find some information about how Rosetta works at RosettaCommons.

I don’t know of any way to change tasks’ threading or scheduling behaviour. Rosetta’s (and/or BOINC’s) load distribution scheme may well be fundamentally based on running multiple instances of single-threaded applications.
ID: 97264 · Rating: 0 · rate: Rate + / Rate - Report as offensive    Reply Quote
Profile Grant (SSSF)

Send message
Joined: 28 Mar 20
Posts: 1520
Credit: 15,429,740
RAC: 23,842
Message 97266 - Posted: 7 Jun 2020, 0:09:37 UTC - in response to Message 97260.  

Is it possible to run a singular task on all of your CPU's cores and threads?
No, and there would be no performance benefit in doing so compared to running multiple different Tasks on each core/thread.
However that is how applications that run on GPUs work, and make them so fast compared to CPUs for work that is able to be processed that way.

eg- In Seti at home a high end CPU with a heavily optimised AVX application would take 45min to process a particular Work Unit. On a high end GPU with a very highly optimised application it would take 30 seconds to process the same task. That's roughly 90 times faster.
Grant
Darwin NT
ID: 97266 · Rating: 0 · rate: Rate + / Rate - Report as offensive    Reply Quote

Questions and Answers : Wish list : Run One task instance on all cores



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