Execution time issue r2016

hello there

I was using matlab on my laptop which has integrated intel graphic card , it takes long time to excuite SSIM

then I bought a brand new labtop with high specs and it still taking the same amount of time!

old labtop specs : core i5 2nd u

hdd hard drive

4 GB ram

New labtop specs : core i5 9th h

ssd m.2 hard drive

8 GB ram

Nvidia gtx 1650 GPU

[Rik]

If it is a single thread computation there will be little difference (as there aren't to my knowledge extreem difference in IPC between those two chips, although I might be wrong about that). If the IPC is similar only a drastically higher clock speed will you get a drastically different perfomance. How did you measure the perfomance? What code were you using to compare the speed?

[Mustafa Khaleel]

didn't use a code

just calculated the total execution time .

[Rik]

You should use timeit, or at least tic and toc. And what time scale are we talking about?

[Raymond Norris]

I agree with Rik. One other consideration, if possible, is to upgrade MATLAB.

[Rik]

Although generally execution speed will improve from release to release, this is not always the case and improvement will vary. (example below tested with my RegGrow FEX submission)

%the main calculation of RegGrow happens with convn() and mean()
%so this mainly benchmarks 2D convolutions and means of a vector
sz=[100 100];IM=1:prod(sz);IM=reshape(IM,sz);
testtime=zeros(1,20);for iteration=1:numel(testtime),tic,RegGrow(IM,'waitbar',false);testtime(iteration)=toc;end
clc,fprintf('median time of %d passes: %.3f seconds (version: %s)\n',numel(testtime),median(testtime),version);

% Results on the same computer:
% median time of 20 passes: 0.018 seconds (version: 9.8.0.1451342 (R2020a) Update 5)
% median time of 20 passes: 0.022 seconds (version: 9.4.0.813654 (R2018a))
% median time of 20 passes: 0.024 seconds (version: 8.5.0.197613 (R2015a))
% median time of 20 passes: 0.030 seconds (version: 7.12.0.635 (R2011a))
% median time of 20 passes: 0.117 seconds (version: 6.5.0.180913a (R13))

You need a benchmark like this for your specific application to really quantify the improvement. From time to time someone posts an example of a function or set of functions that are a lot slower, sometimes orders of magnitude.