Increasing TeX's main memory

Question

I have a plot with a lot of measurement datapoints. I exported this plot from Matlab2tikz.

When I tried to compile, it said that the memory was exceeded. I found this question and it tells me that I can export different tikzpictures to different files. Unfortunately, this didn't help me since I have only 1 tikzpicture causing the problem.

So I actually tried to increase the memory of TeX. I went to /usr/share/texlive/texmf/web2c then in texmf.cnf I changed main_memory = 3000000 to main_memory = 5000000. Then I ran texhash and tried to compile my TeX file again.

My problem is that after I increased the memory limit I got again the same error: ! TeX capacity exceeded, sorry [main memory size=3000000]. However I just changed it to 5000000. I presume that I did something wrong when increasing the memory limit. Did I?

-Edit-

Eventually I just used lualatex. This was due to the fact that I think there is something wrong with my texlive install: it is installed twice on my system for some reason. I think, however, that Heiko Oberdiek's answer is correct and hence marked it as the correct answer.

Answer 1

From texmf.cnf:

% Memory. Must be less than 8,000,000 total.
%
% main_memory is relevant only to initex, extra_mem_* only to non-ini.
% Thus, have to redump the .fmt file after changing main_memory; to add
% to existing fmt files, increase extra_mem_*.  (To get an idea of how
% much, try \tracingstats=2 in your TeX source file;
% web2c/tests/memtest.tex might also be interesting.)

Thus you have to regenerate the format files (fmtutil) to see an effect of changing main_memory.

Answer 2

Increasing the memory may help you to compile your plots but you will end up with very big files that takes lots of time to load by pdf viewers and make scrolling jerky. In addition, as you increase your plots complexity you might hit a performance cap.

Here is another solution that gives very small files while keeping the 'crispiness' of the full tikz approach.

1. Plot each elements that are of differing colour on different graphics. Ensure that the axes limits are the same, remove the axis (axis off in matlab) and export them as high resolution png using the print command

   print -dpng -r800 *name_of_file.png*
   

2. Convert the .png to bitmap with either gimp or imagemagick. With the latter :

   convert *name_of_file.png* *name_of_file.bmp*
   

You might want to directly export in bmp format from Matlab. I have not tested it though and have no license at hand right now.

1. Use potrace to retrieve an .svg vector trace of your .bmp

   potrace *name_of_file.bmp* -s
   

2. Then recombine the .svg with Inkscape, applying wanted colours that have been lost when using potrace.

3. Either export your final result as pdf or as tikz code using Inkscape inkscape2tikz extension. If the former, use \addplot graphic (see pgfplots manual about that command).

I've use that technique on scatter plots with 10 000 points that overlap quite a lot. The resulting files are very small but almost perfectly slick while .png look edgy when one fancy to zoom a lot.