# Is there a method of significant preallocation beyond zeros?

1 view (last 30 days)
Kelly Catlin on 24 Mar 2018
Commented: John D'Errico on 24 Mar 2018
I have been calling a function with 83000 iterations (as below):
>>[t1data,t2data,Tdata,Ddata,lamdata] = example_run(2,0.1,25,1.01)
Yet I have let it run for the better part of a day without the program terminating. Is there a way (beyond preallocation with zeros, as I have already done) to decrease the run time significantly?
function [t1data,t2data,Tdata,Ddata,lamdata] = example_run(E0,sigma,F,tau)
%E0=2;sigma=0.1;F=25;tau=1.01; standard reference
%preallocation
Tdata = zeros(1, 100000);
t1data = zeros(1, 100000);
t2data = zeros(1, 100000);
kdata = zeros(1, 100000);
Ddata = zeros(1, 100000);
lamdata = zeros(1, 100000);
[Tc,Ts]=findTcTs(E0,sigma,F,tau);
%T=Ts+0.5;
hold on;
T = Ts;
T0 = Tc;
for k = 1:83001
Tdata = [Tdata,T];
t1data = [t1data,t1];
t2data = [t2data,t2];
kdata = [kdata,k];
Ddata = [Ddata,D];
lamdata = [lamdata, lambda];
T0 = t2;
T = T + 0.1;
plot(T,real(D), 'o')
%iterate by increments to use valid initial guess for t2 for each increment in findt1t2...
end
hold off;

Walter Roberson on 24 Mar 2018
Tdata = [Tdata,T];
Don't do that. You already used
Tdata = zeros(1, 100000);
so each iteration you are expanding Tdata by putting a single new element after all of the zeros. That is not preallocation! Preallocation would be if you had used
Tdata = zeros(1, 100000);
with
Tdata(k) = T;
(It is not obvious why your 100000 does not match your 83001)
John D'Errico on 24 Mar 2018
Actually, what was done was worse than not preallocating at all.