CODEWITHSUNDEEP
matlab
kits
kits

Reputation: 607

MATLAB: short function taking 20+ minutes to run

I'm using a Xeon quad-core processor with 64GB of RAM. The program running the function only has 89 data points. It's now been over 20 minutes and MATLAB is still "busy" computing the program. Does my code below reveal any reason why it's taking so long to compute?

function last15MinsOfDay=last15MinsOfDay(time,price)
% last15MinsOfDay takes the average of prices between 3:45 and 4:00.
timeStr=cellstr(datestr(time));
timeDbl=datevec(timeStr);
times=and(timeDbl(:,4)==14,timeDbl(:,5)>=45)+and(timeDbl(:,4)==15,timeDbl(:,5)==0);
priceIdx=find(times);
z=find(fwdshift(1,priceIdx)~=priceIdx+1);
z=[1; z];
mu=zeros(length(z),1);
for i = 1:length(z);
    while i < length(z)
        mu(i)=mean(price(priceIdx(z(i):z(i+1))));
    end
end
last15MinsOfDay=mu;

Upvotes: 0

Views: 69

Answers (2)

user2849019
user2849019

Reputation:

To supplement esm's answer, you have two loops, each using the same variable i, so this will cause an error in your while loop, as the variable i is redefined by the for loop over each iteration (after you applied esm's correction). To see an example of what I mean, try to following code and look at the output:

z = 1:10;
for i = 1:(length(z));
    while i < length(z)
        disp(['while loop:' num2str(i)]);
        i = i+1;
    end
    disp(['for loop:' num2str(i)]);
end

Furthermore, because of the dual loops, your wasting a lot of processing time by rewriting the same data to the variable m(i)

I have a feeling this is what you were intending to do:

for i = 1:length(z)-1;
        mu(i)=mean(price(priceIdx(z(i):z(i+1))));
end

I just removed the nested while loop, as it doesn't really serve a purpose and eats up processing time and then I changed length(z) to length(z) - 1 to avoid an index out of bounds error.

And finally, your script will only use one of your processor cores, so having a qaud core processor here isn't giving you any speed advantages. To use all 4 cores, try using a parfor instead of for and you can theoretically quadruple your processing speed. http://nl.mathworks.com/help/distcomp/parfor.html

Upvotes: 3

missimer
missimer

Reputation: 4079

I am no expert in matlab but this part looks funny:

for i = 1:length(z);
    while i < length(z)
        mu(i)=mean(price(priceIdx(z(i):z(i+1))));
    end
end

Specifically I do not see i being incremented in the inner loop so the inner loop will run indefinitely.

Upvotes: 4

Related Questions