How to make my code faster when working with for loop code & large array indexing?

I have a demands vector that gives output as below. It means I have 4992 demands at first time instant and then 5025,etc.

% demands= poissrnd(5000,1,4)
demands =
    4992        5025        5009        5108

And I want to initialize these many number of demands at each time instant. For e.g. at d=1, I create 4992 vectors & at d=2, I create 5025 vectors, etc. Each vector is initialized with a certain values of source, destination, req_slots and a threshold value. This is shown below.

for d=1:numel(demands)   %at each element of demands shown above
 for dd=1:demands(d)   % When d=1, dd=1:4992. And when d=2,dd=1:5025
   [source(dd),destin(dd),required_slots(dd),osnr_th(dd)]=initialize_demand(t,adj) %to initialize each dd numbered demand 
 end    %the initialize_demand fn is not shown here
end

This code works perfectly fine. But it takes a lot of time even with this 5000 value ranges. I may have to increase it to 1 million in later stages. So, is there a smart way of doing this? Like using just one loop or even some other data structure to initialize quickly?

[Matt J]

There's nothing slow about the code that you've shown us. Whatever is slow, it is inside initialize_demand().

[Jaya]

Well, then here is the initialize_demand code. With the same question of faster code requirement, any comments on this please?

 function [source,dest,req_slots,osnr_thr]= initialize_demand(nodes,adj)
%based on the response, pasted below are the details of t and adj. In actual code,
% these are there in the beginning of the code. Just pasted here for reference.
 % j=[ 1 1 2 2 2 3 3 4 5];
 %  t=[ 2 6 3 5 6 5 4 5 6]; %these two are used as node numbers to make a graph.
 % G = addedge(G,j,t) ;   
% adj= full(adjacency(G)); %adj is the adjacency matrix. Is a 6*6 matrix of 1's and 0's
        
src= 1:max(nodes);  snk= 1:max(nodes); x= true;
        
        source= datasample(src,1,2,'Replace',false);
        while x
            dest= datasample(snk,1,2,'Replace',false)
            if ((dest==source)|| (adj(source,dest)==1))
                dest= datasample(snk,1,2,'Replace',false);  %to make have atleast two hops b/w src & dest
            else
                x=false;
             end
            %             x=false;
        end
        
        datarate= unifrnd(12.5,160)   %in gbps
           osnr_thr=12
           req_slots= round(datarate/ (4*12.5))
        end

[Matt J]

Also, the dependence of 't' and 'adj' on the loop variables is not shown.

[Jaya]

I thought maybe they are not important to show. But here they are. Please any comments now?

j=[ 1 1 2 2 2 3 3 4 5];
t=[ 2 6 3 5 6 5 4 5 6]; %these two are used as node numbers to make a graph.
G = addedge(G,j,t) ;   
adj= full(adjacency(G)); %adj is the adjacency matrix. Is a 6*6 matrix of 1's and 0's

Answer 1

[Cris LaPierre]

A couple thoughts.

The biggest speedup will happen by suppressing your outputs. Put a semi-colon after every line. Doing that, your code ran in 17 seconds.

Another thought. If everything is working pefectiy fine, then you can get rid of the outer for loop. Based on the indexing used, your final loop will overwrite most if not all elements in your output variables using the final value in demands (d=numel(demands)). Doing this reduce run time to 3 seconds.

[Jaya]

Thanks a lot! I did put semi-colons later but with a different goal. And I too observed the speed up. I will work on the removal of outer loop idea too.