speed up for loop execution

Hi,

I am trying to run a 3d for loop, which is taking 3 hrs to run. I tried some of the best practices for loops such as pre-allocating the array, indexing and parfor but it does not improve the processing time. Please can you suggest to improve the execution time.

for i1= 1:length(RGI_glacier_temp_C_avg) % length of 216

for j1 = 1:length(elev_bins) % array size 200

for k1 = 1:length(gl_max_elev) % array size 18,000

RGI_downscld_no_thre(i1,j1,k1) = temp2m_all_grids_avg(i1,:) + lr_era * (elev_bins(:,j1) - gl_max_elev(k1,:));

RGI_downscld_no_thre_C(i1,j1,k1) = temp2m_all_grids_C_avg(i1,:) + lr_era * (elev_bins(:,j1) - gl_max_elev(k1,:));

end

end

end

Answer 1

[Bruno Luong]

Try this:

T = reshape(temp2m_all_grids_avg, [], 1, 1);
Tc = reshape(temp2m_all_grids_C_avg, [], 1, 1);
hbins = reshape(elev_bins,1,[],1);
maxh = reshape(gl_max_elev,1,1,[]);
Th = lr_era * (hbins - maxh);
RGI_downscld_no_thre = T + Th;
RGI_downscld_no_thre_C = Tc  + Th;

These outputs require more than 10 Gb on RAM. You might compute the array in single if you run out of memory or it starts to swap on HD.

[Lavanya Ashokkumar]

Thank you very much. It worked.