Preallocating zeros in cell array. and Which one is faster? 3D matrix, cell or structure data?

Hello all, I have to rewrite my Matlab code to make it faster. in some cases I need to used cell structure to keep my data: for example:

wallP = cell(1,wallDensity+1);

for b = wallDensity+1:-1:1

    for u = 1:nWalls

        wallP{b}(u,:) = center + direction(u,:).*(wall+(((1-wall)/wallDensity)*(b-1)));

    end

end

1. I did preallocate my cell matrix, but my question is that it's not like zeros matrix the result in just some empty matrix however I know that after this loop I have a matrix of (nWalls,3) in each cell can I make a zero matrix to enhance the speed? do you recommend it?

(preallocating)

wallP = cell(1,wallDensity+1)

wallP =

    []    []    []

(the final result)

wallP =

    [16x3 double]    [16x3 double]    [16x3 double]

is there any way to make a zeros of (16,3) in each cell before the loop?

2. My second question is that it's possible for me to make a 3D matrix or create a structure to keep these data, for example: wallP(u).b instead of wallP{b}(u,:) or any other way that I don't know??!!!! Thanks for your help

Answer 1

[yonatan gerufi]

Hi Masha, first of all, regarding to code execution time see " profiler ".

it's a matlab feature that analyze the running time of your code.

you can switch part of your code and track the time difference.

regarding the second question (please edit the first one).

read this: cell vs. matrix

Answer 2

[Sean de Wolski]

If each matrix (cell element) is the same size, then you're best off making a 3d matrix

16x3xp

where p is the number cells.

You can then index into it, for example the second one:

x(:,:,2)