[x264-devel] Re: [PATCH] Altivec optimizations for quant4x4, quant4x4dc, quant8x8, sub8x8_dct8, sub16x16_dct8, pixel_sa8d_8x8, pixel_sa8d_16x16, *idct8*

[David Wolstencroft]

On the second

> #define ALTIVEC_STORE_SUM_CLIP_ALIGN8_A

and

> vec_s16_t sum16 = vec_adds(idct_sh6, (vec_s16_t)dst16);\

should be

#define ALTIVEC_STORE_SUM_CLIP_ALIGN8_B(dest, idctv) {\

and

vec_s16_t sum16 = vec_adds(idct_sh6, (vec_s16_t)dst16l);\



Sorry - tired :(



On Sep 19, 2006, at 2:27 AM, David Wolstencroft wrote:

> Sooo, ok, since I'm wired (don't know why I can't sleep....)
>
>
> if (dst is 16 byte aligned)
> 	
> #define ALTIVEC_STORE_SUM_CLIP_ALIGN8_A(dest, idctv) {\
>     vec_u8_t dstv = vec_ld(0, dest);\
>     vec_s16_t idct_sh6 = vec_sra(idctv, sixv);\
>     vec_u16_t dst16h = vec_mergeh(zero_u8v, dstv);\
>     vec_u16_t dst16l = vec_mergel(zero_u8v, dstv);\
>     vec_s16_t sum16 = vec_adds(idct_sh6, (vec_s16_t)dst16h);\
>     vec_u8_t sum8 = vec_packsu(sum16, dst16l);\       <- I swear  
> pengvado made a mistake here, if that's possible
>     vec_st(sum8, 0, dest);\
>
> else (8 byte aligned but not 16 byte aligned)
>
> #define ALTIVEC_STORE_SUM_CLIP_ALIGN8_A(dest, idctv) {\
>     vec_u8_t dstv = vec_ld(0, dest);\
>     vec_s16_t idct_sh6 = vec_sra(idctv, sixv);\
>     vec_u16_t dst16h = vec_mergeh(zero_u8v, dstv);\
>     vec_u16_t dst16l = vec_mergel(zero_u8v, dstv);\
>     vec_s16_t sum16 = vec_adds(idct_sh6, (vec_s16_t)dst16);\
>     vec_u8_t sum8 = vec_packsu(dst16h, sum16);\
>     vec_st(sum8, 0, dest);\
>
>
>
> On Sep 18, 2006, at 12:24 PM, Loren Merritt wrote:
>
>> #define ALTIVEC_STORE_SUM_CLIP_ALIGN8_A(dest, idctv) {\
>>     vec_u8_t dstv = vec_ld(0, dest);\
>>     vec_s16_t idct_sh6 = vec_sra(idctv, sixv);\
>>     vec_u16_t dst16h = vec_mergeh(zero_u8v, dstv);\
>>     vec_u16_t dst16l = vec_mergel(zero_u8v, dstv);\
>>     vec_s16_t sum16 = vec_adds(idct_sh6, (vec_s16_t)dst16h);\
>>     vec_u8_t sum8 = vec_packsu(dst16l, sum16);\
>>     vec_st(sum8, 0, dest);\
>

-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://mailman.videolan.org/pipermail/x264-devel/attachments/20060919/4ae295a7/attachment.htm