On 23/07/14 21:43, Anton Khirnov wrote:
> 
> On Mon, 14 Jul 2014 12:55:36 +0200, Luca Barbato <[email protected]> wrote:
>> Square prediction units are the most common.
>> About 0.7% faster on 4k videos.
>> ---
>>  libavcodec/hevc.c             | 64 
>> +++++++++++++++++++++++++++++++++----------
>>  libavcodec/hevcdsp.c          |  6 ++++
>>  libavcodec/hevcdsp.h          |  2 ++
>>  libavcodec/hevcdsp_template.c | 16 +++++++++++
>>  4 files changed, 73 insertions(+), 15 deletions(-)
>>
>> diff --git a/libavcodec/hevc.c b/libavcodec/hevc.c
>> index 1634185..7d15ed3 100644
>> --- a/libavcodec/hevc.c
>> +++ b/libavcodec/hevc.c
>> @@ -27,6 +27,7 @@
>>  #include "libavutil/common.h"
>>  #include "libavutil/display.h"
>>  #include "libavutil/internal.h"
>> +#include "libavutil/intmath.h"
>>  #include "libavutil/md5.h"
>>  #include "libavutil/opt.h"
>>  #include "libavutil/pixdesc.h"
>> @@ -1652,9 +1653,27 @@ static void hevc_await_progress(HEVCContext *s, 
>> HEVCFrame *ref,
>>      ff_thread_await_progress(&ref->tf, y, 0);
>>  }
>>  
>> -static void hls_prediction_unit(HEVCContext *s, int x0, int y0,
>> -                                int nPbW, int nPbH,
>> -                                int log2_cb_size, int partIdx)
>> +av_always_inline
>> +static void put_unweighted_pred(HEVCContext *s,
>> +                                uint8_t *dst, ptrdiff_t dststride,
>> +                                int16_t *src, ptrdiff_t srcstride,
>> +                                int width, int height, int square)
>> +{
>> +    if (!square) {
>> +        s->hevcdsp.put_unweighted_pred(dst, dststride,
>> +                                       src, srcstride, width, height);
>> +    } else {
>> +        const int square = av_log2_16bit(width) - 2;
> 
> Variable shadowing is evil, rename this to something like log_size

ops, you are right :)

> Otherwise looks ok

It depends on the rest of the patchset you have or it got merged while I
wasn't watching?

lu


_______________________________________________
libav-devel mailing list
[email protected]
https://lists.libav.org/mailman/listinfo/libav-devel

Reply via email to