How to avoid for loops using compact array notation?

I am quite new to Matlab and I don't know how to use compact notation, so I use for loops which slows down the execution time. Any suggestions on how to rewrite the following code in a compact way (array notation) to optimize for execution time?

temperatureData and volumes are 2 matrixes with millions of rows and a few tens of columns. The following code adds the volumes that correspond to certain standard ranges of temperatures (ranges specified in partsLimits).

 for o=1:numberBuckets
  for m=1:numberRows
    for n=1:numberColumns 
      if (temperatureData(m,n)<=partsLimits(1,o))&&(temperatureData(m,n)>partsLimits(2,o))
                 standarVolumes(m,o)=standarVolumes(m,o)+volumes(m,n);

      end
    end
  end
end

On the other hand I am not able to use parfor instead of for loops because the standarVolumes variable can not be classified. Any workaround? I hope that I can speed up calculations using the proper notation, but if not possible and I need to keep the for loops I would like to use parfor and parallelize.

Thanks

Answer 1

[Image Analyst]

This should do it faster:

for bucket = 1 : numberBuckets
  % Get the min and max for this bucket from partsLimits.
  maxValue = partsLimits(1,bucket);
  minValue = partsLimits(2,bucket);
  % For this bucket, see where the temperatureData
  % is within the range [minValue, maxValue]
  logicalMatrix = (temperatureData <= maxValue) && ...
      (temperatureData > minValue);
  % For this bucket, add in volumes that meet the criteria.
  standarVolumes(:,bucket)=standarVolumes(:,bucket)+ ...
               sum(volumes(logicalMatrix));    
end

[Joe]

Thanks for that! this was helpful. Also, this helped solving the parfor issue , now I can use parfor without problems.

Answer 2

[Doug Hull]

I find it very unlikely that this will be the bottleneck in your code. Have your run this through the profiler to confirm that of all the code being executed that that this is the part that is best to spend your time speeding up?

It is unclear if these are function calls or matrix indexing operations. That would help to speed this up.

[Joe]

Hi, thanks for your answer. I will definitely run it through the profiler, that's a great idea. It is important to optimize this code as it belongs to a function that the only thing that does is pretty much it, and I need to run it thousands of times...