[FFmpeg-devel] [PATCH 2/3] x86: asm for sign/zero_extend()

[Michael Niedermayer]

On Sun, Feb 21, 2010 at 12:35:58AM +0000, Mans Rullgard wrote:
> ---
>  libavcodec/x86/mathops.h |   18 ++++++++++++++++++
>  1 files changed, 18 insertions(+), 0 deletions(-)
> 
> diff --git a/libavcodec/x86/mathops.h b/libavcodec/x86/mathops.h
> index 010cfb7..0c17f35 100644
> --- a/libavcodec/x86/mathops.h
> +++ b/libavcodec/x86/mathops.h
> @@ -97,4 +97,22 @@ static inline uint32_t NEG_USR32(uint32_t a, int8_t s){
>      return a;
>  }
>  
> +#define sign_extend sign_extend
> +static inline int sign_extend(int val, unsigned bits)
> +{
> +    __asm__ ("shll %1, %0 \n\t"
> +             "sarl %1, %0 \n\t"
> +             : "+&r" (val) : "ic" ((uint8_t)-bits));
> +    return val;
> +}
> +

> +#define zero_extend zero_extend
> +static inline unsigned zero_extend(unsigned val, unsigned bits)
> +{
> +    __asm__ ("shll %1, %0 \n\t"
> +             "shrl %1, %0 \n\t"
> +             : "+&r" (val) : "ic" ((uint8_t)-bits));
> +    return val;

if bits is a constant (which i guess it is quite often)
then this is quite inefficient.
val & 0x00007FFF
for example is more efficient in that case.
also its not certain 2 shifts are faster than =-1, >>, &
on all x86 cpus

[...]
-- 
Michael     GnuPG fingerprint: 9FF2128B147EF6730BADF133611EC787040B0FAB

The real ebay dictionary, page 1
"Used only once"    - "Some unspecified defect prevented a second use"
"In good condition" - "Can be repaird by experienced expert"
"As is" - "You wouldnt want it even if you were payed for it, if you knew ..."
-------------- next part --------------
A non-text attachment was scrubbed...
Name: not available
Type: application/pgp-signature
Size: 189 bytes
Desc: Digital signature
URL: <http://lists.mplayerhq.hu/pipermail/ffmpeg-devel/attachments/20100221/66a9d9cc/attachment.pgp>