[Mesa-dev] [PATCH 2/2] nir/opcodes: Use fpclassify() instead of isnormal() for ldexp

[Ian Romanick]

On 01/26/2015 02:22 PM, Jason Ekstrand wrote:
> Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=88806
> ---
>  src/glsl/nir/nir_opcodes.py | 2 +-
>  1 file changed, 1 insertion(+), 1 deletion(-)
> 
> diff --git a/src/glsl/nir/nir_opcodes.py b/src/glsl/nir/nir_opcodes.py
> index f54a017..3f21539 100644
> --- a/src/glsl/nir/nir_opcodes.py
> +++ b/src/glsl/nir/nir_opcodes.py
> @@ -480,7 +480,7 @@ else
>  opcode("ldexp", 0, tunsigned, [0, 0], [tfloat, tint], "", """
>  dst = ldexp(src0, src1);
>  /* flush denormals to zero. */
> -if (!isnormal(dst))
> +if (fpclassify(dst) != FP_NORMAL)

GCC documentation claims that isnormal is faster than fpclassify(x) ==
FP_NORMAL.  Do we care?

If we care, we should provide isnormal() in macros.h.

If not,

Reviewed-by: Ian Romanick <ian.d.romanick at intel.com>

>     dst = copysign(0.0f, src0);
>  """)
>  
>