Trying to run my code in GPU

Hello, I am trying to run my code in GPU. I just installed parallel toolbox. Here is a part of my code:

w=zeros(1000);
A=gpuArray(w);

for z=1:3000
    
    for i = 2:length(A)-1
        for n=2:width(A)-1
            A(i,n)=(A(i+1,n)+A(i-1,n)+A(i,n+1))+A(i,n-1)/4;
        end
        z
    end
    
end

I don't think my code is running in GPU because the Task Manager shows GPU is not working.

And the code is running much slower than not running it in GPU.

Can someone help me fix it?

Thanks

[Rik]

You probably intended to use height instead of length.

Other than that, you didn't define the array A and you're not using the array you created on the GPU, so why exactly did you expect this to work?

It also looks like your final result might be reproducible with a convolution or another array operation, although I'm not entirely certain which one exactly. The fastest loop is of course replacing the loop with array operations.

[Jingqi Sun]

@Rik Hi, I defined 'A' outside the matrix , sorry I forgot to copy and paste that line. I will updated my question right now. I know built in array operations could operate much faster, but I don't think there have a built in function for what I am trying to do. Thanks.

Answer 1

[Matt J]

Your code isn't gettng the benefit of the GPU because you are not applying any parallelized functions to your gpuArray, A. In this case, conv2() would be appropriate.

w=zeros(1000);
A=gpuArray(w);

k=[0 1 0;
    1 0 1;
    0 1 0]/4;

for z=1:3000
    
 A=conv2(A,k,'valid');
    
end

[Jingqi Sun]

Hi thank you for your reply! I think you are right because iteration time even takes longer when I use gpuArray.