a function slows down my profiler: I would like to speed it up

%{
hi, i've create a large code..
it's fast but there is a function that slows everything down for me (my application does several simulations by executing this function several times)
I want to give it a try by asking someone experienced if it can be speeded up
it's function name "Calcola_MinTrade". and in profile it's that's what slows everything down
%}
RP_bin=load('matlab_RP_bin.mat');
MinNtrad=load('matlab_Ntrades.mat');

tic
RP_bin=RP_bin.RP_bin;

Ntradess=MinNtrad.Ntradess;
period=2;
minTrades=16;

 z=find(RP_bin);
  [~,c]=size(Ntradess);
  MinNtrad=zeros(numel(z),c);
   for x=1:c
    for i=period+1:numel(z) 
          id=z(i); 
          a=Ntradess(id,x);   
          a1=a-minTrades+1;
         %%******************
           kk=z(max(1,i-period));
           if ~minTrades
                     b=kk;
           else  
             idx=find(flip(Ntradess(1:id,x))<=a1,1,'first');
             b=id-idx+1;
             b=min(b,kk);
             %%*****************
           end  
           if ~isempty(b) 
                MinNtrad(i,x)= b; 
    
           end                
     end 
   end

   
   %the function return MinNtrad
   toc

Elapsed time is 1.038010 seconds.

Answer 1

[the cyclist]

It won't be a big speedup, but I would expect

idx=find(Ntradess(id:-1:1,x)<=a1,1,'first');

to be consistently faster than

idx=find(flipud(Ntradess(1:id,x))<=a1,1,'first');

(Note that I reversed the indexing in the first dimension.)

[Walter Roberson]

idx = (id+1) - find(Ntrades(1:id, x), 1, 'last')

might possibly be faster.

[shamal]

yes..it's more fast but it's not equal