Idea to speed up crunching

Message boards : Number crunching : Idea to speed up crunching

To post messages, you must log in.

AuthorMessage
TJ

Send message
Joined: 29 Mar 09
Posts: 127
Credit: 4,799,890
RAC: 0
Message 74142 - Posted: 31 Oct 2012, 15:09:03 UTC

As Dr. Baker is posting on the home page that crucnhing is the delaying factor in the process for new medicine, I should sugest to change the Rosetta software so that it can cruch on multiple cores of the processor.
MilkyWay has this for some tasks and that reduces the crunhing time significantly, as I have seen on a i7 processor.
Greetings,
TJ.
ID: 74142 · Rating: 0 · rate: Rate + / Rate - Report as offensive    Reply Quote
Mod.Sense
Volunteer moderator

Send message
Joined: 22 Aug 06
Posts: 4018
Credit: 0
RAC: 0
Message 74144 - Posted: 31 Oct 2012, 22:36:53 UTC

Rosetta achieves the net benefit of multi-core machines when you configure the BOINC Manager to allow multiple CPUs to be used by BOINC, and you run multiple tasks at the same time.

What you are referring to is referred to as multi-threading, and is where a single task can use several CPUs at the same time.

By running multiple BOINC tasks at the same time, you actually assure a higher utilization rate of the available CPUs. Multi-threaded applications typically have specific areas of the program (such as start-up for example) where they cannot take full advantage of the CPUs available.
Rosetta Moderator: Mod.Sense
ID: 74144 · Rating: 0 · rate: Rate + / Rate - Report as offensive    Reply Quote
mikey
Avatar

Send message
Joined: 5 Jan 06
Posts: 1895
Credit: 9,169,305
RAC: 3,078
Message 74150 - Posted: 1 Nov 2012, 12:08:53 UTC - in response to Message 74144.  

Rosetta achieves the net benefit of multi-core machines when you configure the BOINC Manager to allow multiple CPUs to be used by BOINC, and you run multiple tasks at the same time.

What you are referring to is referred to as multi-threading, and is where a single task can use several CPUs at the same time.

By running multiple BOINC tasks at the same time, you actually assure a higher utilization rate of the available CPUs. Multi-threaded applications typically have specific areas of the program (such as start-up for example) where they cannot take full advantage of the CPUs available.


Which usually results in each task being much larger to offset some of that initial loss. Larger individual tasks can then have their own problems, ie crashing, resulting in a large loss of time and credits for the user. The other problem is to design a unit flexible enough to work on a dual core pc and also work on an 8 or 32 core machine, or anywhere in between. AMD came out a few years ago with 3 core cpu machines, so it is not an easy thing to design a work unit to handle all of those as easily as it is to just make it run on one core of the cpu and then also be able to just run several units at once.
ID: 74150 · Rating: 0 · rate: Rate + / Rate - Report as offensive    Reply Quote
Profile Chilean
Avatar

Send message
Joined: 16 Oct 05
Posts: 711
Credit: 26,694,507
RAC: 0
Message 74180 - Posted: 5 Nov 2012, 16:45:51 UTC

The best way to speed things up is to fix the bug that isn't letting a bunch of crunch with 4+ cores machines.
ID: 74180 · Rating: 0 · rate: Rate + / Rate - Report as offensive    Reply Quote

Message boards : Number crunching : Idea to speed up crunching



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