[FFmpeg-devel] [PATCH v1] fftools/ffmpeg_opt: avoid to display the hwaccels name twice

Marton Balint cus at passwd.hu
Sat Aug 17 19:35:55 EEST 2019



On Sat, 17 Aug 2019, Limin Wang wrote:

> On Sat, Aug 17, 2019 at 05:21:33PM +0200, Marton Balint wrote:
>> 
>> 
>> On Sat, 17 Aug 2019, Limin Wang wrote:
>> 
>> >
>> >
>> >ping, please help to merge the code if it's OK.
>> >
>> >On Thu, Aug 15, 2019 at 09:23:45AM +0800, lance.lmwang at gmail.com wrote:
>> >>From: Limin Wang <lance.lmwang at gmail.com>
>> >>
>> >>videotoolbox and qsv have been defined by hw_type_names[] in hwcontext.c
>> >>
>> >>Fixes ticket #7464
>> >>
>> >>Signed-off-by: Limin Wang <lance.lmwang at gmail.com>
>> >>---
>> >> fftools/ffmpeg_opt.c | 6 ------
>> >> 1 file changed, 6 deletions(-)
>> >>
>> >>diff --git a/fftools/ffmpeg_opt.c b/fftools/ffmpeg_opt.c
>> >>index f5ca18a..8baa898 100644
>> >>--- a/fftools/ffmpeg_opt.c
>> >>+++ b/fftools/ffmpeg_opt.c
>> >>@@ -66,12 +66,6 @@
>> >> }
>> >>
>> >> const HWAccel hwaccels[] = {
>> >>-#if CONFIG_VIDEOTOOLBOX
>> >>-    { "videotoolbox", videotoolbox_init, HWACCEL_VIDEOTOOLBOX, AV_PIX_FMT_VIDEOTOOLBOX },
>> >>-#endif
>> >>-#if CONFIG_LIBMFX
>> >>-    { "qsv",   qsv_init,   HWACCEL_QSV,   AV_PIX_FMT_QSV },
>> >>-#endif
>> 
>> And what about the init functions which are called? Or other cases
>> in the code which iterates over hwaccels? Have you tried if
>> accelerated processing works after your patch?
>
> Yes, I have tested OK with both videotoolbox decode and encode,
> ffmpeg_opt.c:842 will invoke av_hwdevice_find_type_by_name(hwaccel); 
> to get them. For cuvid isn't defined in the hwcontext.c, so keep it
> anyway.

The loop in ffmpeg.c:2845 won't find the hwaccel if you remove them from 
this list.

Regards,
Marton


More information about the ffmpeg-devel mailing list