Using Dynare with a Dual Quad
Posted: Mon Nov 02, 2009 8:38 pmDynare with the latest Matlab 2009b has been running very slowly on my Dual Quad Dell. I finally contacted Matlab and received the following help on their website. The best solution for me is to give the command
maxNumCompThreads(4);
before running dynare.
Summary
Linear algebra operations involving sparse matrices or small full matrices may see a decrease in performance on some 8-core machines.
Description
Linear algebra operations that use the BLAS library may experience performance decrease on 8-core Intel Xeon machines. This occurs only for operations on small full matrices or sparse matrices.
Workaround
If you experience slow performance on any machine with 8 or more cores, experiment with decreasing the maximum number of computational threads. The maxNumCompThreads command sets the maximum number of computational threads. For example, to set the maximum number of computational threads to 4, do this:
maxNumCompThreads(4);
Exists in Release (Version)
Exists in R2009b(7.9), R2009a(7.
, R2008b(7.7), R2008a(7.6), R2007b(7.5), R2007a(7.4)
maxNumCompThreads(4);
before running dynare.
Summary
Linear algebra operations involving sparse matrices or small full matrices may see a decrease in performance on some 8-core machines.
Description
Linear algebra operations that use the BLAS library may experience performance decrease on 8-core Intel Xeon machines. This occurs only for operations on small full matrices or sparse matrices.
Workaround
If you experience slow performance on any machine with 8 or more cores, experiment with decreasing the maximum number of computational threads. The maxNumCompThreads command sets the maximum number of computational threads. For example, to set the maximum number of computational threads to 4, do this:
maxNumCompThreads(4);
Exists in Release (Version)
Exists in R2009b(7.9), R2009a(7.
, R2008b(7.7), R2008a(7.6), R2007b(7.5), R2007a(7.4)