Why do I receive an "Out of Memory" error when I use the WAVREAD or WAVWRITE function in MATLAB 7.1 (R14SP3)?

The largest contiguous block of memory on my system is larger than 250 MB. When I try to read in a 250 MB WAV-file using WAVREAD, I receive an "Out of Memory" error.

Answer 1

[MathWorks Support Team]

The ability to read or write WAV-files larger than the fourth largest block of contiguous memory is not available in MATLAB.

Two possible workarounds to memory limitations are:

1. To consider windowing, if it fits the application, by specifying a small subset of samples from a.WAV file at a given time for time-localized processing using the command

wavread(filename, [starting_sample, ending_sample])

2. Read data from the .WAV file in the native file format, instead of MATLAB's default double-precision by executing

wavread(filename, 'native')

This will help minimize memory usage further if the native file format encoded is of a precision less than 64-bit word size.

[Walter Roberson]

The fourth largest block of contiguous memory?

So if I initialize MATLAB and all of the "large block" memory is a single contiguous heap, then the fourth largest block of contiguous memory would be one of the "small block" cache that MATLAB keeps internally. I do not happen to recall exactly the size of the small blocks -- 512 bytes? 4096 bytes? MATLAB never chains those blocks together: if an array grows past the blocksize, MATLAB allocates memory from the heap, looking for best-fit, potentially leading to memory fragmentation.

So, this article seems to be saying that unless I have happened to fragment the heap, that the largest WAV file I would be able to read it would be that small block size such as 512 bytes (minus any overhead.)

I don't think that was the intended meaning. It would make more sense to say something like contiguous memory was required equivalent to 8 bytes per sample per channel plus some temporary storage equal to (whatever). Because if I happen to have a large enough contiguous memory, it should have no problem splitting.