Speed up for loops for arrays

Hello, Maybe someone could help me to speedup this type of for loop? I did try parfor, I did try to think of way using GPU, but it does not work with indexing.

clear;
a=7811200001;
d=1083355908;
c=zeros(1083355908,1)
parfor i = 1:d
 c(i,1)=a+i; 
end

Answer 1

[madhan ravi]

a=7811200001;
d=1083355908;
c=a+(1:d);
c=(a+1):(a+d); %edited after jan's comment

[Stephen Cobeldick]

This line

c=zeros(1083355908,1);

is completely superfluous and only slows the code down. Get rid of it.

[madhan ravi]

Thank you @Stephen I thought it would enhance speed no?

[Stephen Cobeldick]

"I thought it would enhance speed no?"

No. You simply created one large array in memory, and then on the next line created another large array in memory, that replaced the first array. It had nothing to do with preallocation, if that is what you were thinking, because you were not using indexing anywhere. Preallocation is where there is only one array (with its final size) in memory and its values are accessed using indexing: thus the array does not change size so it does not need to be moved in memory (not really relevant to your code) and its elements are simply accessed in situ using indexing (which is what you were not doing). Read the documentation and check how the examples use indexing:

https://www.mathworks.com/help/matlab/matlab_prog/preallocating-arrays.html

[madhan ravi]

Everything is clear now thank you

[Bruno Luong]

It might even slows down your code since it reduces a RAM for few subsequent instructions.

[madhan ravi]

Useful info @Bruno

[Stephen Cobeldick]

"It might even slows down your code ..."

Creating such a large array (which is entirely unused) will definitely slow down the code.

@madhan ravi: in any case, don't just believe me or Bruno because we wrote something: try it yourself! Use a loop and tic & toc, or use timeit, and find out yourself. Experimentation is always worthwhile.

[madhan ravi]

@Stephen: sure!

[Jan]

Another hint: In a+(1:d) you create the double vector 1:d at first and then add a to each element. It is more efficient to omit the addition and create vector directly:

   (a+1):(a+d)

[madhan ravi]

Selbverständlich , danke schön @Jan