How do I preallocate memory in a For loop code?

Question

Fun Dan 2015-5-4

0
链接

此问题的直接链接

https://ww2.mathworks.cn/matlabcentral/answers/214919-how-do-i-preallocate-memory-in-a-for-loop-code

编辑： Stephen23 2015-5-4

How do I preallocate memory for the for loop below? And does preallocating memory makes a simulation run faster?

#1   for Mx = x:50:500;
#2   for My = y:50:400;
#3   %
#4   J1 = [sqrt((Mx-100)^2+(My-20)^2)];
#5   J2 = [sqrt((Mx-20)^2+(My-150)^2)];
#6   J3 = [sqrt((Mx-50)^2+(My-450)^2)];
#7   J4 = [sqrt((Mx-150)^2+(My-450)^2)];
#8   J5 = [sqrt((Mx-7000)^2+(My-200)^2)];
#14   end
#15   end

Thanks

2 个评论
显示无隐藏无

Stephen23 2015-5-4

编辑：Stephen23 2015-5-4

Note that those brackets [] are unnecessary (as the Editor highlighting and message also tells you). There is nothing being concatenated tougher, so they just slow the code down.

Fun Dan 2015-5-4

Thanks Stephen. I took the [] out. I was getting low memory error so I thought if I preallocate a memory that would solve the issue. Also, I notice my simulation keeps running longer than it should, could that be the issue?

请先登录，再进行评论。

请先登录，再回答此问题。

Answer 1

Stephen23 2015-5-4

0
链接

此回答的直接链接

https://ww2.mathworks.cn/matlabcentral/answers/214919-how-do-i-preallocate-memory-in-a-for-loop-code#answer_177681

编辑：Stephen23 2015-5-4

在 MATLAB Online 中打开

Here is a completely vectorized code version using bsxfun:

x = 0;
y = 0;
Vx = x:50:500;
Vy = y:50:400;
Cx = [100,20,50,150,7000];
Cy = [20,150,450,450,200];
Tx = bsxfun(@minus,reshape(Vx,1,[],1),Cx(:)).^2;
Ty = bsxfun(@minus,reshape(Vy,1,1,[]),Cy(:)).^2;
out = sqrt(bsxfun(@plus,Tx,Ty));

where the final variable out is arranged with the constants Cx and Cy along the first dimension (rows), then Vx values along the second dimension (columns), and Vy along the third dimension (pages):

>> size(out)
ans =
   5    11     9

And to answer your question "does preallocating memory makes a simulation run faster?", then answer is yes! In most cases array preallocation will give much faster code than without array preallocation. MATLAB clearly describes and recommends this in their documentation:

http://www.mathworks.com/help/matlab/matlab_prog/techniques-for-improving-performance.html

But you don't have to ask here: try it yourself! Make a small experiment with the same operation in a loop and the same one with preallocation... and you will discover the difference yourself.