- fft for long vectors
- The two-input form of conv2
- Integer conversion and arithmetic
Matlab with dual core uses only 50% cpu (R2010b, Intel Core 2 duo T9300)
3 vues (au cours des 30 derniers jours)
Afficher commentaires plus anciens
I'm doing a finite difference simulation with matlab which takes around 15 minutes at this moment. When looking at the resource monitor of Windows 7 (32 bit), I see matlab never uses more than 50% of the cpu.
I'm using Matlab R2010b. The computer is a HP with a Intel core 2 duo T9300, 2.5GHz 4GB RAM, of which 3GB is usable (due to 32bit Windows 7)
The question is, how can I make the computer use 100% of its computing power so I can run my simulations faster?
In Windows I already turned all power saving options off, but this didn't help.
0 commentaires
Réponse acceptée
Sarah Wait Zaranek
le 14 Mar 2011
It depends which version of MATLAB you are using. MATLAB started including multithreaded functions in R2007a. I believe multithreading was always supported on all platforms.
By R2009a, multithreading became a startup option (-singleCompThread ). Before then, it could toggled in the preferences. As for a list of the functions that are multithreaded, look here:
This list hasn't been updated for the last 2 releases so you can add the following functions:
For your particular case, you may not be using many of these functions and/or large enough matrices to see much of a difference. You may want to explore using the Parallel Computing Toolbox.
3 commentaires
Cris Luengo
le 14 Mar 2011
You might want to look into "vectorizing" your code, substituting the loops over x and y by matrix operations. Matrix operations are multi-threaded. For example, "for x=1:10, a(x)=b(x)*c(x); end" uses only one processor core, but "a(1:10)=b(1:10).*c(1:10);" uses up to 10 cores.
Plus de réponses (3)
the cyclist
le 14 Mar 2011
From poking around on Google, I gather that there is a multithreading option in Preferences, but that not all MATLAB functions will multithread.
Not sure on this, and I cannot check because I don't think these options exist on a Mac, which is what I use.
Jason Ross
le 14 Mar 2011
Keep in mind also that "computing power" is not measured by CPU alone. If you look in Resource Monitor, examine the disk, memory and network connection metrics, as well.
For example, if you are accessing data from a network it's entirely possible that you could be waiting on that operation to complete and the CPU has no data to process.
Richard Crozier
le 17 Mar 2011
Have a look at the multicore package on the file exchange to speed up computation using multiple cores.
0 commentaires
Voir également
Community Treasure Hunt
Find the treasures in MATLAB Central and discover how the community can help you!
Start Hunting!