[Beignet] [PATCH 1/2] support built-in function "rotate"

[Xing, Homer]

Thanks. Have sent version 2. You will soon receive them.

-----Original Message-----
From: He Junyan [mailto:junyan.he at inbox.com] 
Sent: Thursday, June 27, 2013 11:12 AM
To: Xing, Homer
Cc: beignet at lists.freedesktop.org
Subject: Re: [Beignet] [PATCH 1/2] support built-in function "rotate"

On 06/26/2013 03:51 PM, Homer Hsing wrote:
> Signed-off-by: Homer Hsing <homer.xing at intel.com>
> ---
>   backend/src/ocl_stdlib.h | 33 +++++++++++++++++++++++++++++++++
>   1 file changed, 33 insertions(+)
>
> diff --git a/backend/src/ocl_stdlib.h b/backend/src/ocl_stdlib.h index 
> 3b191ab..b114b46 100644
> --- a/backend/src/ocl_stdlib.h
> +++ b/backend/src/ocl_stdlib.h
> @@ -4355,6 +4355,39 @@ DEC(16)
>   #undef DEC4
>   #undef DEC8
>   #undef DEC16
> +
> +INLINE_OVERLOADABLE uchar __rotate_left(uchar x, uchar y) { return (x 
> +<< y) | (x >> (8 - y)); } INLINE_OVERLOADABLE char __rotate_left(char 
> +x, char y) { return __rotate_left((uchar)x, (uchar)y); } 
> +INLINE_OVERLOADABLE ushort __rotate_left(ushort x, ushort y) { return 
> +(x << y) | (x >> (16 - y)); } INLINE_OVERLOADABLE short 
> +__rotate_left(short x, short y) { return __rotate_left((ushort)x, 
> +(ushort)y); } INLINE_OVERLOADABLE uint __rotate_left(uint x, uint y) 
> +{ return (x << y) | (x >> (32 - y)); } INLINE_OVERLOADABLE int 
> +__rotate_left(int x, int y) { return __rotate_left((uint)x, (uint)y); 
> +} #define DEF(type, n, m) INLINE_OVERLOADABLE type rotate(type x, 
> +type y) { return __rotate_left(x, (type)(y < 0 ? n + y : y & m)); }
^^^^^^^^^^^ if y < -n, seems some problem

> +DEF(char, 8, 7)
> +DEF(uchar, 8, 7)
> +DEF(short, 16, 15)
> +DEF(ushort, 16, 15)
> +DEF(int, 32, 31)
> +DEF(uint, 32, 31)
> +#undef DEF
> +#define DEC2(type) INLINE_OVERLOADABLE type##2 rotate(type##2 a, 
> +type##2 b) { return (rotate(a.s0, b.s0), rotate(a.s1, b.s1)); } 
> +#define DEC3(type) INLINE_OVERLOADABLE type##3 rotate(type##3 a, 
> +type##3 b) { return (rotate(a.s0, b.s0), rotate(a.s1, b.s1), 
> +rotate(a.s2, b.s2)); } #define DEC4(type) INLINE_OVERLOADABLE type##4 
> +rotate(type##4 a, type##4 b) { return (rotate(a.s0, b.s0), 
> +rotate(a.s1, b.s1), rotate(a.s2, b.s2), rotate(a.s3, b.s3)); } 
> +#define DEC8(type) INLINE_OVERLOADABLE type##8 rotate(type##8 a, 
> +type##8 b) { return (rotate(a.s0, b.s0), rotate(a.s1, b.s1), 
> +rotate(a.s2, b.s2), rotate(a.s3, b.s3), rotate(a.s4, b.s4), 
> +rotate(a.s5, b.s5), rotate(a.s6, b.s6), rotate(a.s7, b.s7)); } 
> +#define DEC16(type) INLINE_OVERLOADABLE type##16 rotate(type##16 a, 
> +type##16 b) { return (rotate(a.s0, b.s0), rotate(a.s1, b.s1), 
> +rotate(a.s2, b.s2), rotate(a.s3, b.s3), rotate(a.s4, b.s4), 
> +rotate(a.s5, b.s5), rotate(a.s6, b.s6), rotate(a.s7, b.s7), 
> +rotate(a.s8, b.s8), rotate(a.s9, b.s9), rotate(a.sa, b.sa), 
> +rotate(a.sb, b.sb), rotate(a.sc, b.sc), rotate(a.sd, b.sd), 
> +rotate(a.se, b.se), rotate(a.sf, b.sf)); } #define DEF(n) 
> +DEC##n(char); DEC##n(uchar); DEC##n(short); DEC##n(ushort); 
> +DEC##n(int); DEC##n(uint)
> +DEF(2)
> +DEF(3)
> +DEF(4)
> +DEF(8)
> +DEF(16)
> +#undef DEF
> +#undef DEC2
> +#undef DEC3
> +#undef DEC4
> +#undef DEC8
> +#undef DEC16
> +
>   /////////////////////////////////////////////////////////////////////////////
>   // Work Items functions (see 6.11.1 of OCL 1.1 spec)
>   
> //////////////////////////////////////////////////////////////////////
> ///////