[Numpy-discussion] masked ufuncs in C: on github

[David Cournapeau]

Eric Firing wrote:
> That would incur the overhead of an extra function call for each 
> element; I suspect it would slow it down a lot. My motivation is to make 
> masked array overhead negligible, at least for medium to large arrays.
>   

You can use inline in that case - starting with numpy 1.3.0, inline can
be used in C code in a portable way (it is a macro which points to
compiler specific inline if the C99 inline is not available so it works
even on broken compilers).

cheers,

David