Technical Solutions

Solution:

For MATLAB 7.3 (R2006b) and previous releases:

MATLAB is a single threaded application. However, multiple processors can be utilized by the BLAS (Basic Linear Algebra Subroutines) if the BLAS version being used supports multi-threading. In this case, significant performance improvement may be observed when running linear algebra intensive applications. This can be achieved on MATLAB 7.3 (R2006b) for both the Intel and AMD family of multiple core processors.

In order to enable BLAS multi-threading, the environment variables for the BLAS need to be set before running MATLAB. The method for doing this is different for each operating system.

On Windows XP, the environment variables can be set using the following steps:

1. Right click on “My Computer” and select “Properties”.
2. Click on the “Advanced” tab.
3. Click on the “Environment Variables” button.
4. Create or edit the following "System variables":

  Variable	 Value
 ----------	-------
BLAS_VERSION	mkl.dll  (for Intel® BLAS)
OMP_NUM_THREADS	   2 

On Linux machines, the environment variables can be set using the “setenv” or “export” commands:

setenv BLAS_VERSION libmkl.so:libguide.so
setenv OMP_NUM_THREADS 2

or

export BLAS_VERSION=libmkl.so:libguide.so
export OMP_NUM_THREADS=2

Note: On the Linux OS, MATLAB has to be run with the "-nojvm" option for multi-threading to work.

After setting the appropriate environmental variables, start MATLAB and ensure that the BLAS version and the number of threads have been set correctly. To do this, execute the following code at the MATLAB Command Prompt:
getenv BLAS_VERSION


ans =

mkl.dll

getenv OMP_NUM_THREADS


ans =
2

As of MATLAB 7.4 (R2007a), MATLAB can be multi-threaded for certain applications. Multi-threading in MATLAB 7.4 (R2007a) can be set via MATLAB's preferences. For more information, see the Related Solution below, or enter the following at the MATLAB command prompt:

web([docroot,'/techdoc/matlab_env/f1-94809.html#bq35sqs'])