how to add "large arrays" (64 bits) to codegen generated function?

Hi all, is there some flag to tell codegen function to use 64 bits indexes? (Something similar to the -largeArrayDims for the mex compilation) Thanks!

[Sergio Zlotnik]

Ok, I found this thread:

https://www.mathworks.com/matlabcentral/answers/166142-how-can-i-re-mex-matlab-coder-generated-c-h-and-o-files-using-largedimarrays-flag

That answer my question. Did this changed since 2014? Thanks in advance!

Answer 1

[Fangjun Jiang]

I would assume that, if you specify the data type of your index variable as uint64, it will generate the code correctly using 64 bits indexing.

Answer 2

[Jan]

codegen creates C code, as far as I understand. The C code itself is not limited to 32 or 64 bit. 32 bit compilers can work with 64 bit variables also. But the addressing depends on the compiler. But as long as the created code contains e.g. size_t for the addressing, this is adjusted during the compilation.

During the mex compilation it matters, which libraries are used. Hm. I stop poking around in the dark. I think a kind of -largeArrayDims would not make sense for codegen.

[EDITED] See http://www.mathworks.com/help/coder/ug/array-size-restrictions-for-code-generation.html: My assumption was wrong. The code supports 32 bit addressing in arrays only.

[Sergio Zlotnik]

Thanks Jan for answer.

Let me describe my problem: if I run this (stupid) function it works well:

function alloc_large_matrix()
o = ones(2147483647,3);

If I try

codegen alloc_large_matrix.m
??? Maximum variable size allowed by the program is exceeded.

Error in ==> alloc_large_matrix Line: 2 Column: 5

So I guessed that something similar to largeArrayDims was missing...

[Jan]

See [EDITED]