Speeding up the computation

Hi, I have run the following code, which takes ages to complete. Even when I use high performance computers, after several days, the computation has still not completed. Is there any way, I can speed up the computation? Maybe by writing it another way? The code is here:

max_wavelet_level = 8;
nn = 5;

for i = 1:length(Data)
    i
    patient = tensor{1,i};
    for k = 1:size(patient,1)
        for j = 1:size(patient,3)
            WDEC{1,i}(k,:,:,j) = single(modwt(tensor{1,i}(k,:,j),max_wavelet_level,'db4'));
            for l =1:max_wavelet_level+1
                [imf,res] = emd(squeeze(WDEC{1,i}(k,l,:,j)),'Display',0);
                pad_size = max(0,nn-size(imf,2));
                pad = zeros(size(imf,1),pad_size);
                padded_imf = cat(2,imf,pad);
                EMD{1,i}(k,l,:,:,j) = single(padded_imf(:,1:nn));
            end
        end
    end
end

I am running discrete wavelet transform on my signal. Subsequently, I run the EMD algorithm on the subsignals.

[Mohammad Sami]

try profiling to see which part of your code is taking the longest.

Answer 1

[Gaurav Garg]

Hi,

You can use parfor loop to run your for loop on multiple workers.

However, parfor can only be used when there aren’t any dependencies between different wokers/threads which seems to be true in your case. Moreover, parfor loops cannot be nested (parfor cannot be placed inside another parfor).

So, you can consider running parfor on either the first loop (parfor i = 1:length(Data)), or any other loop which should run fine too.