Accelerating MATLAB Performance

489Memory-Related Techniques

% Bad idea: allocates 8 MB double array, then converts to 1MB int8 array

data = int8(zeros(1000,1000)); % 1M elements

⇨ Elapsed time is 0.000113 seconds.

% Better: directly allocate the array as a 1MB int8 array - faster

data = zeros(1000,1000,'int8');

⇨ Elapsed time is 0.000072 seconds.

For example, int8(ones(1000,1)) rst allocates 8000 bytes for 1000 double-precision val-

ues, then allocates a separate 1000-bytes block for the converted data.* This is not only

wasteful in CPU cycles and memory, but could also cause severe slowdowns when the

data is large, possibly even thrashing and out-of-memory errors. In ...