Boost::uBLAS::mapped_matrix max size too small. typedef size_t ulong

[Axon]

Hi,

I'm using Boost::uBLAS::mapped_matrix for working with sparse matrices. At the moment, the maximum dimension of mapped_matrix is 65535x65535 since std::size_t compiles as unsigned int.

I'm looking for a way of increasing this limit. The best way I can think of would be to ensure that size_t uses a bigger (64-bit) type rather than the current UInt32.

Can I force std::size_t to be defined as unsigned long?
I'm using VC 2005 (VC8) on Windows.

Thanks for your support.

Cheers,
Greg

[Banfa]

You don't need 2, the maximum dimension is not 65535, this is the maximum for an unsigned 16 bit integer but on VC++ version 8 std:size which is unsigned int is an unsigned 32 bit value and therefore has a maximum of 4294967295.

[Axon]

You don't need 2, the maximum dimension is not 65535

Please re-read the original post:

the maximum dimension of mapped_matrix is 65535x65535

So I need a way of forcing the use of __uint64 so that my max matrix dimension (limited by the size_t type) is bigger than 65535x65535.

Thanks anyways... Anyone?

[Banfa]

It rather depends on whether you have the library as binary or source code.

If you have the source code of the classes then you may be able to recompile them with an alternately sized size_t, or another variable type used in it's place.

If you only have the binary then the classes have already been compiled with a 32 bit size_t, altering the size of size_t externally to the class is not going to alter its size used internally to the class.