Why better hardware performance doesn’t speed up Matlab programs?

I implemented my optimization and neural network problems with Matlab toolbox My program took a lot of time to be executed. So I used Amazon EC2 high performance servers (quadruple) and installed Matlab on Windows server 2008 and ran my program on it. But the speed of computations wasn’t so different with my HP laptop!!

My HP Laptop Configuration: 4 GB of memory Intel Core 2 due 2.6 GHz 64-bit platform

Amazon High-Memory Quadruple Extra Large Instance Configuration: 68.4 GB of memory 26 EC2 Compute Units (8 virtual cores with 3.25 EC2 Compute Units each) 1690 GB of instance storage 64-bit platform I/O Performance: High API name: m2.4xlarge

Answer 1

[Walter Roberson]

Which MATLAB version are you using? The newer the MATLAB, the more it can automatically parallelize; the last few releases have made significant improvements in the automatic use of cores.

The worst-case scenario is probably R2009a or earlier and working with for loops with on small portions of arrays. Releases after that improved for loops progressively, and even that release is not bad working with large (5000+ element) vectorized operations.

NN and optimization are not operations that would typically benefit from JIT speedups. They might, however, benefit from explicit parallelization with the Parallel Programming Toolbox.