<div dir="auto"><div><br><br><div class="gmail_quote"><div dir="ltr" class="gmail_attr">ср, 24 апр. 2024 г., 19:43 Andrea paz <<a href="mailto:gamberucci.andrea@gmail.com">gamberucci.andrea@gmail.com</a>>:<br></div><blockquote class="gmail_quote" style="margin:0 0 0 .8ex;border-left:1px #ccc solid;padding-left:1ex">> I assume attached tar archive then only about TEST1 ?<br>
<br>
?<br>
<br>
In test 1 I used a VP8/mov source without audio [4-3_3-2.mov] combined with an<br>
mp3/wav audio and in addition the render result [igor_test3.webm]:<br>
VP9/webm with audio<br>
Vorbis (webm.webm preset). [I didn't notice that the wav audio was<br>
mp3, maybe I need to<br>
use PCM audio?]<br>
In test 2 I used Big Buck Bunny_sunflower in h264/mp4 version with mp3 audio.<br>
<br>
In both tests I simply loaded (in test 1, the project and in test<br>
2, the file) with the “replace current project” option and then did a little<br>
playback/scrolling.<br>
<br>
The messages shown are everything that appeared between starting and<br>
closing CinGG<br>
(I just removed the normal start/end messages that CinGG has).<br>
<br>
The problems with vaapi are probably due to the combination of AMD,<br>
mesa, amdgpu and<br>
rocm, since every time they are updated the behavior always changes and<br>
new problems always arise...<br>
However in CinGG/ffmpeg7 it seems that vaapi is working (test 2<br>
[bbb_sunflower_1080p_30fps_normal.mp4]: no line “HW device init failed, using SW<br>
decode"). In test 1 you have errors with vaapi with VP8 [4-3_3-2.mov],<br>
which I believe is<br>
not supported. But there are no errors with VP9 [igor_test3.webm],<br>
which certainly<br>
supports vaapi.<br></blockquote></div></div><div dir="auto"><br></div><div dir="auto">Ah, sorry I was not sure if error messages show up serially right after loading file or they can show up asynchronous ...</div><div dir="auto"><br></div><div dir="auto"><br></div><div dir="auto">Sorry for taking your time. I know vaapi is fragile, nice to see at least sw fallback in our code still works as intended, even if it makes checking vaapi runtime status harder ...</div><div dir="auto"><br></div><div dir="auto"><br></div><div dir="auto"><div class="gmail_quote"><blockquote class="gmail_quote" style="margin:0 0 0 .8ex;border-left:1px #ccc solid;padding-left:1ex">
<br>
> I wonder if plain system ffmpeg still works for hw decoding?<br>
<br>
In Arch I still have ffmpeg 6.1.1. I have tried transcoding from mp4 to mp4, via<br>
decoding and encoding in vaapi, but I don't understand if decoding in<br>
vaapi works,<br>
while encoding works (<a href="https://trac.ffmpeg.org/wiki/Hardware/VAAPI" rel="noreferrer noreferrer" target="_blank">https://trac.ffmpeg.org/wiki/Hardware/VAAPI</a>):<br>
<br>
$ ffmpeg -hwaccel vaapi -hwaccel_device /dev/dri/renderD128<br>
-hwaccel_output_format vaapi<br>
-i mixkit-worried-and-sad-woman-outdoors-8739.mp4 -c:v h264_vaapi<br>
-b:v 2M -maxrate 2M<br>
output.mp4<br>
<br>
[...]<br>
Input #0, mov,mp4,m4a,3gp,3g2,mj2, from<br>
'mixkit-worried-and-sad-woman-outdoors-8739.mp4':<br>
Metadata:<br>
major_brand : M4V<br>
minor_version : 1<br>
compatible_brands: isomavc1mp42<br>
creation_time : 2020-06-21T18:53:40.000000Z<br>
Duration: 00:00:15.00, start: 0.000000, bitrate: 24801 kb/s<br>
Stream #0:0[0x1](und): Video: h264 (Main) (avc1 / 0x31637661),<br>
yuv420p(tv, bt709,<br>
progressive), 1920x1080 [SAR 1:1 DAR 16:9], 24798 kb/s, 24 fps, 24<br>
tbr, 24k tbn (default)<br>
Metadata:<br>
creation_time : 2020-06-21T18:53:40.000000Z<br>
handler_name : ETI ISO Video Media Handler<br>
vendor_id : [0][0][0][0]<br>
encoder : Elemental H.264<br>
Stream mapping:<br>
Stream #0:0 -> #0:0 (h264 (native) -> h264 (h264_vaapi))<br>
Press [q] to stop, [?] for help<br>
[h264_vaapi @ 0x571d363508c0] Driver does not support some wanted<br>
packed headers (wanted<br>
0xd, found 0x1).<br>
Output #0, mp4, to 'output.mp4':<br>
Metadata:<br>
major_brand : M4V<br>
minor_version : 1<br>
compatible_brands: isomavc1mp42<br>
encoder : Lavf60.16.100<br>
Stream #0:0(und): Video: h264 (High) (avc1 / 0x31637661), vaapi(tv,<br>
bt709, progressive),<br>
1920x1080 [SAR 1:1 DAR 16:9], q=2-31, 2000 kb/s, 24 fps, 12288 tbn (default)<br>
Metadata:<br>
creation_time : 2020-06-21T18:53:40.000000Z<br>
handler_name : ETI ISO Video Media Handler<br>
vendor_id : [0][0][0][0]<br>
encoder : Lavc60.31.102 h264_vaapi<br>
[out#0/mp4 @ 0x571d3632dc80] video:3649kB audio:0kB subtitle:0kB other<br>
streams:0kB global<br>
headers:0kB muxing overhead: 0.064330%<br>
frame= 360 fps=196 q=-0.0 Lsize= 3652kB time=00:00:14.95<br>
bitrate=1999.9kbits/s<br>
speed=8.14x<br></blockquote></div></div><div dir="auto"><br></div><div dir="auto">may be mpv can be used to check decoding? It usually linked to system's ffmpeg. Not sure if tools like radeontop still works with new amdgpu supported AMD hardware.</div><div dir="auto"><br></div><div dir="auto">Anyway, thanks for testing, I hope it not ate too much time from your day.</div><div dir="auto"><div class="gmail_quote"><blockquote class="gmail_quote" style="margin:0 0 0 .8ex;border-left:1px #ccc solid;padding-left:1ex">
</blockquote></div></div></div>