MATLAB Answers

0

Optimizing nested for loop

Asked by mashtine on 31 Aug 2014
Latest activity Edited by dpb
on 1 Sep 2014
Accepted Answer by dpb
I have the following for loop and it take quite a while to complete when I am sure it can be faster. It looks up an input value from the assigned dataset and extracts parameters from the struct based on the bin it falls into. Once these parameters have been selected for, the output is based on a calculation.
nrows = length(inpdata);
ncols = 2:2:12;
for j = ncols;
for i = 1:nrows;
if j == 2
height = 'height10m';
elseif j == 4
height = 'height20m';
elseif j == 6
height = 'height40m';
elseif j == 8
height = 'height80m';
elseif j == 10
height = 'height120m';
elseif j == 12
height = 'height200m';
end
if inpdata(i,j) > 0
param = strcat('params',num2str(length(bin(bin < inpdata(i,j)))));
gusts(i,j) = inpdata(i,j) + (inpdata2.(height).(param)(1,3)*(inpdata(i,j)^(1 + inpdata2.(height).(param)(1,1))));
elseif inpdata(i,2) == 0
param = 'params1';
gusts(i,j) = inpdata(i,j) + (inpdata2.(height).(param)(1,3)*(inpdata(i,j)^(1 + inpdata2.(height).(param)(1,1))));
end
end
end
I am not too sure how to optimize this for more efficiency but does anyone have any suggestions? The profiler shows that it is the strcat and num2str functions that are the main culprits but I am not sure how to optimize them as I need them in the loop.

  0 Comments

Sign in to comment.

1 Answer

dpb
Answer by dpb
on 31 Aug 2014
 Accepted Answer

If that's the significant fraction, you can try
param=sprintf('params%d',length(bin(bin < inpdata)));
instead.

  2 Comments

mashtine on 31 Aug 2014
that certainly did the trick for the most part! Down from roughly 10 mins to 79 seconds. These minor changes can really alter the memory usage. Thank you dpb
dpb
on 31 Aug 2014
That much is amazing, indeed...just out of curiosity, how does
param=num2str(length(bin(bin < inpdata)),'params%d');
stack up? That is, giving the explicit format but still using num2str?

Sign in to comment.