CODEWITHSUNDEEP
arraysmatlab
Golovach Grach
Golovach Grach

Reputation: 168

What I need to replace to improve perfomance of the alghorithm?

I am new in Matlab but I am trying. I have the following code:

for t = 1:size(data,2)
  b = data(t)/avevalue;
  if b >= 1
     cat1 = [repmat((avevalue),floor(b),1)',mod(data(t),15)];
  else
     cat1 = data(t);
  end
  modified = [modified,cat1];
end

The answer for 

data=[16 18 16 25 17 7 15];
avevalue=15;

is

    15     1    15     3    15     1    15    10    15     2     7    15     0

But when my array is more than 10000 elements it working very, impossibly slow (for 100000 nearly 3 minutes, for example). How can I increase its speed?

Upvotes: 1

Views: 62

Answers (1)

ibezito
ibezito

Reputation: 5822

There are two main reasons for the slowness:

  1. The fact that you are using a loop.
  2. The output array is growing on each iteration.

you can improve runtime by trying the following approach:

%auxilliary array
divSumArray = ceil((data+1)/avevalue);

%defines output array
newArr = ones(1,sum(divSumArray))*avevalue;

%calculates modulo
moduloDataIndices = cumsum(divSumArray);

%assigning modulo in proper location
newArr(moduloDataIndices) = mod(data,avevalue);

the final result

15     1    15     3    15     1    15    10    15     2     7    15     0

Time measurement

I measured runtime for the following input:

n = 30000;
data = randi([0 99],n,1);
avevalue=15;

original algo:

Elapsed time is 11.783951 seconds.

optimized algo: 

Elapsed time is 0.007728 seconds.

Upvotes: 2

Related Questions