Accelerating MATLAB Performance

248 Accelerating MATLAB

®

Performance

As another example:

% Loop variant

data = rand(10,20,30); % 10x20x30

for idx = 1 : size(data,3)

a(idx) = sum(sum(data(:,:,idx)));

end

% Vectorized variant

idx = 1 : size(data,3)

a(idx) = sum(sum(data(:,:,idx)));

% Simpler variant

a = reshape(sum(sum(data)), 1, []); % reshape 1x1x30  1x30

Sometimes the transformation is not so simple, since there are dependencies between

the loop iterators:

% Loop variant

for i = 1 : num_total_days

for j = 1 : num_days_later-1

if (i+j) <= num_total_days+1

for k = 1 : num_stocks

if data(i,j+1,k) >0

data2(i,j,k) = (data(i,j+1,k) / data(i,1,k)) - 1;

end

end

end

end

end

% Vectorized variant, in 2 phases:

% Phase 1 - divide all elements by the ...