parrallel computation,4-core machine

I have got an Intel-4core machine recently ,R2011a installed on it.but,I still don't believe that I am using all of my machine's computation ability.actually I couldn't find any multi-threading option in preferences and when I run BENCH command I get a fast and reasonable response but when I run my program it gets sort of same time as my laptop.

Answer 1

[Jason Ross]

Things to know about performance of code:

1. You need to know how many cores you really have. If you have a machine with "hyper-threading", you can have two actual cores, but they appear to the OS as four. Therefore, it's possible to have something like a Core2Duo in your laptop and a 2-core processor in your desktop, and have the same number of cores. Look up the spec sheet for your processor to find out.
2. Parallel execution can be implicit or explicit. Some MATLAB functions are multithreaded, and will automatically take advantage. Others are not. Other functions (in combination with the Parallel Computing Toolbox) will take advantage of multiple cores, but you must change your code to use parallel computing constructs (parfor, spmd, etc)
3. The most valuable thing you can do to search for speedup is to use the profiler to see what's taking up the most time in your code. Otherwise it's just guesswork and pre-optimization. It's possible that you are waiting on something like disk I/O to read a file and parse it, or there is some other bit of code that's written inefficiently, or you aren't preallocating, you could vectorize, etc.

[ahmad]

Thanks for your answer and time,I did some changes to my program up to the level of my knowledge in parallel computing.but by intuition I say my desktop should be faster because it has 6 cores and my supervisor assembled it just for the purpose of work.I don't have any I/O in my code,but i have a 3-d matrix which by increasing its indices the amount of time it takes is way beyond my expectation.I am solving a partial differential eq.

[Jason Ross]

If you've identified the slow section of code, I'd suggest you start a new question that asks "is there a way to increase the performance of this operation" and post the part that's slow. There have been many such questions asked where meaningful results came out -- but people have to be able to see the code.

[Thomas]

Accepted by Binish..

Answer 2

[Thomas]

I do not think bench command is not parallelized.

To make use of all the cores you need the Parallel computing toolbox and you need to parallelize your code (If it can be parallelized..) to actually see a speed up..

see

http://www.mathworks.com/products/parallel-computing/description4.html

[ahmad]

I think that even without parallelizing my code I should get a faster response from my machine just because it has a 12GB memory and 6 cores of 3.2GHz processors.since I am using MATLAB R2011a, I don't find any option to see whether my MATLAB sees all of my cores when it is running the code or not?