CUDA , MATLAB AND GPUMAT ? ( Matrix Multiplication)

Accelerated Computing CUDA CUDA Programming and Performance
1

Hi,

I am new to cuda programming, and i have executed matrix multiplication program using visual studio 2008. Now i want to execute matrix multiplication using matlab as my front end.So i downloaded GPUmat and tried out this piece of code

a=rand(500,500);
ad=GPUsingle(a);
b=ad*ad;

Now here i wanted to know if GPUmat executes the code using multiple threads(i.e automatically creates around 500 threads,one for each element multiplication) or do we have to define the kernel and stuff.

Sorry if the question looks a bit stupid.Any help appreciated.

Thank you,
Vaishnav