[Beignet] [PATCH] fix built-in function "fast_normalize"

[Xing, Homer]

I think the same as you. You should assume that if x is NaN, then the result should be a vector full of NaNs.

Homer Hsing

-----Original Message-----
From: Lu, Guanqun 
Sent: Monday, October 28, 2013 10:41 AM
To: Xing, Homer; beignet at lists.freedesktop.org
Subject: RE: [Beignet] [PATCH] fix built-in function "fast_normalize"

hi homer,

how about if x is NaN, according to the opencl-1.2 spec.

  "normalize ( v ) returns a vector full of NaNs if any element is a NaN."

so i would assume that if x is NaN, the result should be NaN. HOw do you think?


--guanqun



> -----Original Message-----
> From: beignet-bounces+guanqun.lu=intel.com at lists.freedesktop.org
> [mailto:beignet-bounces+guanqun.lu=intel.com at lists.freedesktop.org] On 
> Behalf Of Homer Hsing
> Sent: Monday, October 28, 2013 9:03 AM
> To: beignet at lists.freedesktop.org
> Subject: [Beignet] [PATCH] fix built-in function "fast_normalize"
> 
> if the parameter is zero, then return zero if the parameter is 
> positive, then return 1.
> for other cases, return -1.
> 
> Signed-off-by: Homer Hsing <homer.xing at intel.com>
> ---
>  backend/src/ocl_stdlib.tmpl.h | 2 +-
>  1 file changed, 1 insertion(+), 1 deletion(-)
> 
> diff --git a/backend/src/ocl_stdlib.tmpl.h 
> b/backend/src/ocl_stdlib.tmpl.h index a3bf0a0..38d1ab0 100644
> --- a/backend/src/ocl_stdlib.tmpl.h
> +++ b/backend/src/ocl_stdlib.tmpl.h
> @@ -1779,7 +1779,7 @@ INLINE_OVERLOADABLE float fast_distance(float x, 
> float y) { return length(x-y);  INLINE_OVERLOADABLE float 
> fast_distance(float2 x, float2 y) { return length(x-y); }  
> INLINE_OVERLOADABLE float fast_distance(float3 x, float3 y) { return 
> length(x-y); }  INLINE_OVERLOADABLE float fast_distance(float4 x, 
> float4 y) { return length(x-y); } -INLINE_OVERLOADABLE float 
> fast_normalize(float x) { return 1.f; }
> +INLINE_OVERLOADABLE float fast_normalize(float x) { return x > 0 ? 
> +1.f : (x <
> 0 ? -1.f : 0.f); }
>  INLINE_OVERLOADABLE float2 fast_normalize(float2 x) { return x * 
> rsqrt(dot(x, x)); }  INLINE_OVERLOADABLE float3 fast_normalize(float3 
> x) { return x * rsqrt(dot(x, x)); }  INLINE_OVERLOADABLE float4 
> fast_normalize(float4 x) { return x * rsqrt(dot(x, x)); }
> --
> 1.8.3.2
> 
> _______________________________________________
> Beignet mailing list
> Beignet at lists.freedesktop.org
> http://lists.freedesktop.org/mailman/listinfo/beignet