Accelerating MATLAB Performance

603I/O Techniques

end

fclose(fid);

⇨ Elapsed time is 8.685103 seconds.

% Using bulk I/O - 170x faster (non-vectorized)

fid = fopen('test.dat');

data = fread(fid,inf,'double');

for idx = 1 : 1e6

data(idx) = sqrt(data(idx));

end

fclose(fid);

⇨ Elapsed time is 0.050718 seconds.

When reading data from an external hardware interface (e.g., an instrument or hardware

device connected via USB or serial port), rather than a le, it makes sense to use buffering.

1699

Using data buffers help to process data in chunks, rather than each data element separately

as it arrives. In essence, data buffering mimics the I/O consolidation recommendation abov ...