Saving large matrices WITHOUT hdf5

I have a large matrix currently stored in .mat v7.3 format that I need to read into another program. This program can read normal .mat files without problem, but can't handle HDF5 files. This is a problem, since Matlab refuses to save the variable in anything but a -v7.3 file.

How can I force Matlab to save this variable in a non-HDF5 file?

Answer 1

[Chad Greene]

A = rand(4,3,2,5);

dlmwrite('mydata.txt',A)

Answer 2

[Walter Roberson]

What formats are supported by the other program?

You can fwrite() the data as binary. Note that if you fwrite() anything with 2 or more dimensions, the memory is written "down" columns, corresponding to varying the earlier dimensions more rapidly than the later dimensions.

Answer 3

[Jason Corso]

I suppose those are options, but they'd require writing all new I/O code for the other program, which is currently set up quite nicely to read (non-HDF5) .mat files. So really, what I want is to force Matlab to save the matrix to a v7 or earlier .mat file.

[Walter Roberson]

If the matrix is 2 Gb or larger (I think it is, 4 Gb max for sure) then there is no way to store it in a v7 .mat file. v7 .mat files include size fields that are limited to 32 bits.