[Mesa-dev] [PATCH 05/13] nir: Avoid double promition.

[Emil Velikov]

On 14 July 2015 at 00:22, Matt Turner <mattst88 at gmail.com> wrote:
> ---
>  src/glsl/nir/nir_opcodes.py | 4 ++--
>  1 file changed, 2 insertions(+), 2 deletions(-)
>
> diff --git a/src/glsl/nir/nir_opcodes.py b/src/glsl/nir/nir_opcodes.py
> index 56e96d9..df5b7e2 100644
> --- a/src/glsl/nir/nir_opcodes.py
> +++ b/src/glsl/nir/nir_opcodes.py
> @@ -474,10 +474,10 @@ else
>  """)
>
>  opcode("ldexp", 0, tfloat, [0, 0], [tfloat, tint], "", """
> -dst = ldexp(src0, src1);
> +dst = ldexpf(src0, src1);
>  /* flush denormals to zero. */
>  if (!isnormal(dst))
> -   dst = copysign(0.0f, src0);
> +   dst = copysignf(0.0f, src0);
Unless I missed something src/glsl/ir_constant_expression.cpp could
use a similar fix.
If you're adding ldexpf/copysignf to c99_math.h can you move the
copysign/isnormal wrappers as well please.

Thanks
Emil