netlib/blas: adapting to upstream change of CBLAS_INDEX

[gonum]

netlib/blas: adapting to upstream change of CBLAS_INDEX

[kortschak]

I think that's reasonable so long as it doesn't break things. Has OpenBLAS made that change?

[vladimir-ch]

It seems that OpenBLAS has been using #define CBLAS_INDEX size_t for more than 10 years so it's the reference catching up to what other BLAS implementations do.

[kortschak]

Great, then that seems like the thing to do.

[vladimir-ch]

Then the question arises what type should the I?amax routines return. Most naturally it should be uint which means changing our Level1 interfaces. However, if that is done, thenI?amaxcould not return -1 as an invalid index in case of a negative increment or zero size. So either we look away and keep returning int or we panic instead of returning -1. Any other idea?

[kortschak]

I think we keep out signature the same. The internal representation of the index/size type isn't really our concern beyond making sure that it is big enough.

What does OpanBLAS do for I?amax?