On 26/12/14 5:31 AM, Christophe Gisquet wrote:
> Is m7 overwritten anywhere in those slice functions?
It's not. I made a couple changes to these two functions compared
to the inline asm version (all the setup before the loops), but
didn't bother checking the simd much while porting it so i didn't
On 26/12/14 3:09 PM, Michael Niedermayer wrote:
> On Fri, Dec 26, 2014 at 02:15:32PM -0300, James Almer wrote:
>> Signed-off-by: James Almer
>> ---
>> libavfilter/vf_fspp.c | 12 +-
>> libavfilter/vf_fspp.h |6 +-
>> libavfilter/x86/Makefile
On Fri, Dec 26, 2014 at 02:15:32PM -0300, James Almer wrote:
> Signed-off-by: James Almer
> ---
> libavfilter/vf_fspp.c | 12 +-
> libavfilter/vf_fspp.h |6 +-
> libavfilter/x86/Makefile |3 +-
> libavfilter/x86/vf_fsp
Signed-off-by: James Almer
---
libavfilter/vf_fspp.c | 12 +-
libavfilter/vf_fspp.h |6 +-
libavfilter/x86/Makefile |3 +-
libavfilter/x86/vf_fspp.asm | 727 +
libavfilter/x86/{vf_fspp.c
On Fri, Dec 26, 2014 at 01:58:11AM -0300, James Almer wrote:
> Signed-off-by: James Almer
> ---
> Checksum of output is the same as far as i could test, but as mentioned in
> the
> other thread it's still different than the output of the C versions.
>
> libavfilter/vf_fspp.c
Hi,
Le 26 déc. 2014 06:00, "James Almer" a écrit :
>
> Signed-off-by: James Almer
> ---
> Checksum of output is the same as far as i could test, but as mentioned
in the
> other thread it's still different than the output of the C versions.
[...]
> +.loop_height:
> +movq m3, [ditherq]
>
Signed-off-by: James Almer
---
Checksum of output is the same as far as i could test, but as mentioned in the
other thread it's still different than the output of the C versions.
libavfilter/vf_fspp.c | 12 +-
libavfilter/vf_fspp.h |6 +-
li
