How do I reduce the run time of this code?

t=cputime;
pdvect=10.^[-10:0.1:-6];
pgvect=10.^[-10:0.1:-6];

%Variables
psingws=zeros(length(pdvect),length(pgvect));
psingnows=zeros(length(pdvect),length(pgvect));
pmorews=zeros(length(pdvect),length(pgvect));
pmorenows=zeros(length(pdvect),length(pgvect));
pnothing=zeros(length(pdvect),length(pgvect));
psingaws=zeros(length(pdvect),length(pgvect));
psingnoaws=zeros(length(pdvect),length(pgvect));
pmoreaws=zeros(length(pdvect),length(pgvect));
pmorenoaws=zeros(length(pdvect),length(pgvect));
pnothingaws=zeros(length(pdvect),length(pgvect));
pdmat=zeros(length(pdvect),length(pgvect));
pgmat=zeros(length(pdvect),length(pgvect));

%Calculating Ratio
for i1=1:length(pdvect);
    pd=pdvect(i1);
    i1/length(pdvect)
    for j1=1:length(pgvect);
        pg=pgvect(j1);
        [psingws(i1,j1),psingnows(i1,j1),pmorews(i1,j1),pmorenows(i1,j1),pnothingws(i1,j1)]=wsp_v4([pd pd pd pd pd pd pd pd pd pd pd pd pd pd],[pg pg pg pg pg pg pg pg pg pg pg pg pg pg]);
        [psingaws(i1,j1),psingnoaws(i1,j1),pmoreaws(i1,j1),pmorenoaws(i1,j1),pnothingaws(i1,j1)]=aws_v4([pd pd pd pd pd],[pg pg pg pg pg]);
        pdmat(i1,j1)=pd;
        pgmat(i1,j1)=pg;
    end
end

%Ratio of all mutations giving a WS(single and multiple mutation events)[wsp vs. aws]
ratio1=(psingws+pmorews)./(psingaws+pmoreaws);
logratio1=log2(ratio1);

%Ratio of only single mutations giving a WS [wsp vs. aws]
ratio2=psingws./psingaws;
logratio2=log2(ratio2);

%Lower limit
close all
surf(pdvect,pgvect,(psingws)./(psingaws+pmoreaws))
set(gca,'xScale','log','yScale','log');

%Upper limit
figure
surf(pdvect,pgvect,(psingws+pmorews)./(psingaws))
set(gca,'xScale','log','yScale','log');

%Surface of ratio of all mutations giving a WS(single and multiple mutation
%events)[wsp vs. aws]
figure
surf(pdvect,pgvect,logratio1)
set(gca,'xScale','log','yScale','log');

%Surface of ratio of only single mutations giving a WS [wsp vs. aws]
figure
surf(pdvect,pgvect,logratio2)
set(gca,'xScale','log','yScale','log');

e=cputime-t;

[Walter Roberson]

Could you summarize what it does?

Answer 1

[Sean de Wolski]

Have you used the profiler to identify the bottlenecks?

http://www.mathworks.com/help/releases/R2013b/matlab/matlab_prog/profiling-for-improving-performance.html

One small thing is that you calculate the following many times when you could calculate them once and store them:

length(pdvect)
length(pgvect) 
[pd pd pd pd pd pd pd pd pd pd pd pd pd pd]
%etc.

Basically, anything that does not change inside of the for-loop should not be calculated inside of it.

What does this line do?

    i1/length(pdvect)