Parellel computing toolbox speedup
显示 更早的评论
Hello, I've got some computationally intensive problems to solve so i downloaded the parallel computing toolbox but when testing, parfor loops take longer than for loops even in large loops (for loop takes 55secs while parfor takes 60secs). I would think these loops are big enough to overcome the overhead costs of palatalization. I'm running an iMac with an i7 quad core and 32 gb ram. I'm not using a cluster, i'm trying to utilize all of my cores. Is there some sort of configuration I need to change? Might it be intel's hyperthreading causing a problem? Any input would be appriciated
5 个评论
Sean de Wolski
2012-2-24
How many workers are you opening? Are your variables sliced? How much data is being transferred back and forth relative to computation time? Posting a small snippet of code might help us diagnose this as well.
Jonathan Sullivan
2012-2-24
perhaps this for loop and be vectorized. The possible time savings through vectorization can be quite significant, depending on the code. If you post your code, maybe we can take a crack at it.
Walter Roberson
2012-2-24
_Potentially_ related: http://www.mathworks.com/matlabcentral/answers/30073-extremely-slow-script-execution-with-new-laptop
Sarah Wait Zaranek
2012-3-26
Did you make sure to open up a matlab pool?
Jan
2012-3-27
4 very good comments. Actually good enough to be votable answers.
回答（1 个）
Daniel Shub
2012-3-27
0 个投票
Not all problems can be sped up with parallel processing on a single computer. MATLAB automatically utilizes all cores for a number of its standard functions. If your processing is not processor limited, or if MATLAB is already using all your processing power, the PCT will not help.
类别
在 帮助中心 和 File Exchange 中查找有关 Parallel for-Loops (parfor) 的更多信息
Community Treasure Hunt
Find the treasures in MATLAB Central and discover how the community can help you!
Start Hunting!
Translated by 
