How can I make my code run faster?

How can I make this code run faster?

clear all
clc

data = load('SALS.m');

alpha=data(:,1);
mean=data(:,2);

NETC='NET_ALPHA2/%i/net_%i'; % Network Library Path
NETS='Model_784EL_new/net_%i'; % Destination Path


fid=fopen('missing.txt','w');
for i=1:length(alpha)
    NET1= sprintf(NETC,alpha(i),mean(i));
    if exist(NET1, 'file') == 0

    fprintf(fid,'%d\t%d\n',alpha(i),mean(i));
    mean(i)=mean(i)+1;
    NET1= sprintf(NETC,alpha(i),mean(i));
        if exist(NET1, 'file') == 0
            if alpha(i)==0
                alpha(i)=1;
            end
            fprintf(fid,'%d\t%d\n',alpha(i),mean(i));
            mean(i)=mean(i)-2;
            NET1= sprintf(NETC,alpha(i),mean(i));
            if exist(NET1, 'file') == 0
                fprintf(fid,'%d\t%d\n',alpha(i),mean(i));
                mean(i)=mean(i)-1;
                NET1= sprintf(NETC,alpha(i),mean(i));
                if exist(NET1, 'file') == 0
                fprintf(fid,'%d\t%d\n',alpha(i),mean(i));
                mean(i)=mean(i)+4;
                NET1= sprintf(NETC,alpha(i),mean(i));
                end
            end
        end
%     continue

%     else
    end
    NET2= sprintf(NETS,i); 
    fileID = fopen(NET1,'r');
    fileID2 = fopen(NET2,'w');
    B= fscanf(fileID,'%d %d %d\t',[1, 3]);
    fprintf(fileID2,'%d %d %d\n',B);
    formatSpec = '%d %d %d %f %f %f %f %f %f\n';
    A = fscanf(fileID,formatSpec,[9,inf]);
    A = A';
    [m,n] =size(A);
    for row = 1:m
        fprintf(fileID2,formatSpec,A (row,:));
    end
    fclose(fileID);
    fclose(fileID2);
% end
end

[Image Analyst]

Looks like you forgot to attach 'SALS.m'.

[Mina Pakzadmanesh]

Yes you are right. I attached it now.

[Image Analyst]

I'd open as 'wt' and 'rt' since they are known to be text files, not binary files. But that probably won't make it run faster. Since you're doing serial input/output I don't know of anyway to make it faster. It looks like it should talke just milliseconds. How long does it take? Also you can use the new isfile() and isfolder() instead of exist(). Also to be more robust you should check the file handles and take appropriate action if it fails and they equal -1.

[darova]

Can you put outside the for loop this part of the code?

[Mina Pakzadmanesh]

it takes around 3600 seconds.

when I use isfile instead of exist, shall I make any change in fscan ? because I got error.

[Mina Pakzadmanesh]

I tried to put it out put it didnt work

Answer 1

[Steven Lord]

My suspicion is that if you were to profile this code the exist calls would consume the most time. If I understand what you're trying to do correctly, I think calling dir once to obtain the list of files and then checking the list in memory may be more efficient.

But I'd start off profiling the code to determine if my suspicion is correct about where the bottleneck are in the code.