Speed up a for loop in my programme?

1 次查看(过去 30 天)
Hi, one of the part of my programme contains this piece of code:
size2=2500;
gran=3;
A=ones(size2,size2);
for k=1:gran:(size2-gran)
for j=1:gran:(size2-gran)
X=rand*2*pi-pi;
for h=1:gran
for l=1:gran
A(k+l-1,j+h-1) = A(k+l-1,j+h-1) *exp(+1i*X); %phase in the square gran x gran
end
end
end
end
My pc runs this code in 0.60 seconds but I would like to know if it is possible to speed up this process.
I have already looked similar questions like:
but I don't know if I can apply to my problem
  2 个评论
Yongjian Feng
Yongjian Feng 2021-7-29
Can this be written as matrix operations? If so, then use the corresponding matlab matrix operation. Then Matlab might be able to optimize the process for you.
Alberto Paniate
Alberto Paniate 2021-7-29
I don't know, I have to create a matrix X of random numbers each square of sides gran x gran. In these squares the randoms numbers are equal

请先登录,再进行评论。

回答(1 个)

Eike Blechschmidt
Eike Blechschmidt 2021-7-30
If I understood you right this could do the trick and is about 2.5x faster on my machine.
tic();
X=rand(size2*size2,1)*2*pi-pi;
[row,col] = ind2sub([size2, size2],1:size2*size2);
blockRow = ceil(row/gran);
blockCol = ceil(col/gran);
idx = sub2ind([size2/gran, size2/gran], blockRow, blockCol);
A = ones(size2, size2);
A(1:size2*size2) = exp(+1i*X(idx));
toc();
Important to not is that size2 needs to be a multiple of gran to for this solution to work.

类别

Help CenterFile Exchange 中查找有关 Systems of Nonlinear Equations 的更多信息

Community Treasure Hunt

Find the treasures in MATLAB Central and discover how the community can help you!

Start Hunting!

Translated by