[Beignet] [PATCH] libocl: using mad() to implement dot()

[Zhigang Gong]

LGTM, will push latter, thanks.

On Thu, Jan 29, 2015 at 10:36:19AM +0800, Ruiling Song wrote:
> Signed-off-by: Ruiling Song <ruiling.song at intel.com>
> ---
>  backend/src/libocl/src/ocl_geometric.cl |    6 +++---
>  1 file changed, 3 insertions(+), 3 deletions(-)
> 
> diff --git a/backend/src/libocl/src/ocl_geometric.cl b/backend/src/libocl/src/ocl_geometric.cl
> index e469ff9..cd09968 100644
> --- a/backend/src/libocl/src/ocl_geometric.cl
> +++ b/backend/src/libocl/src/ocl_geometric.cl
> @@ -27,13 +27,13 @@ OVERLOADABLE float dot(float p0, float p1) {
>    return p0 * p1;
>  }
>  OVERLOADABLE float dot(float2 p0, float2 p1) {
> -  return p0.x * p1.x + p0.y * p1.y;
> +  return mad(p0.x, p1.x, p0.y * p1.y);
>  }
>  OVERLOADABLE float dot(float3 p0, float3 p1) {
> -  return p0.x * p1.x + p0.y * p1.y + p0.z * p1.z;
> +  return mad(p0.x, p1.x, mad(p0.y, p1.y, p0.z*p1.z));
>  }
>  OVERLOADABLE float dot(float4 p0, float4 p1) {
> -  return p0.x * p1.x + p0.y * p1.y + p0.z * p1.z + p0.w * p1.w;
> +  return mad(p0.x, p1.x, mad(p0.y, p1.y, mad(p0.z, p1.z, p0.w * p1.w)));
>  }
>  OVERLOADABLE float length(float x) { return __gen_ocl_fabs(x); }
>  
> -- 
> 1.7.10.4
> 
> _______________________________________________
> Beignet mailing list
> Beignet at lists.freedesktop.org
> http://lists.freedesktop.org/mailman/listinfo/beignet