Accelerating MATLAB Performance

623I/O Techniques

For MAT-le I/O in Java, consider using the open-source JMATIO library.

1766

Instead of using Java I/O, we might consider using low-level MEX C-code.* MATLAB’s

m-code I/O is closely modeled after the standard C I/O, helping make the transition from

m-code to C-code relatively painless. MATLAB includes a dedicated doc page on using C/

C++ I/O in MEX,

1767

with a couple of corresponding working examples.

†

Here is a simple

example (a multi-threaded version is also available

1768

):

% The original m-code

data = rand(1000); % 1000x1000 double values = ~8MB

fid = fopen('data.bin','w');

len = fwrite(fid,data,'double');

fclose(fid); ...