BLD: fix issue with POWL for Intel compilers on Windows.

Thanks to Intel (contact: Yolanda Chen) for this patch.

The Intel “powl” on Windows only supports 80 bits longdouble to keep
compatibility with Microsoft VC.

Microsoft VC only support 64 bits longdouble, from “C:\Program Files
(x86)\Microsoft Visual Studio 11.0\VC\include\math.h”, powl is redefined as:

    #define powl(x,y)       ((long double)pow((double)(x), (double)(y)))

However in npy_math.c (generated from npy_math.c.src), this definition is
undefined even though HAVE_POWL is set for the Intel C compilers.

At line 384 of npy_math.c.src, we have::

    #ifdef @kind@@c@
    #undef @kind@@c@
    #endif
    #ifndef HAVE_@KIND@@C@
    @type@ npy_@kind@@c@(@type@ x, @type@ y)
    {
        return (@type@) npy_@kind@((double)x, (double) y);
    }
    #endif

The expanded function generated in “npy_math.c” will look like::

    #ifdef powl
    #undef powl
    #endif
    #ifndef HAVE_POWL
    npy_longdouble npy_powl(npy_longdouble x, npy_longdouble y)
    {
        return (npy_longdouble) npy_pow((double)x, (double) y);
    }
    #endif

0 files changed, 0 insertions, 0 deletions