Closed ben1979 closed 6 years ago
Thank you for reporting!
You're welcome, Nice job!
Benoît Leloutre
Le 19 juin 2013 à 21:25, "Jonas Hummelstrand" notifications@github.com<mailto:notifications@github.com> a écrit :
Thank you for reporting!
— Reply to this email directly or view it on GitHubhttps://github.com/CasparCG/Server/issues/147#issuecomment-19707928.
I saw something new with this issue. each time we use a "new" mixer fill with decklink (we did mixer 1-1 fill 0 0 0.5 0.5 and mixer 1-1 fill 0 0 1 1) we generate 1 frame delay more between the audio and the video. Maybe it can help you in your investigation.
Thanks!
/ Jonas Hummelstrand +46-733-970000 http://about.me/hummelstrand
any news or update?
Not completely the same, but seems related:
When searching on Google for 'Decklink memory leak', we find lots of issues with Wirecast, Resolme, ... and in their release notes somewhere back in time they resolved memory leaks. They do not mention Decklink memory leaks, but they all use Decklink input cards and have resolved it :-)
Don't know if this is enough info, can add lots of logs, but this is the general gist of things.
2.2
when we use mixer fill command with a decklink producer the ram used memory is not completely restored. In fact, we see a average of 13MB leak.
First, start the server: the process = 225MB play 1-3 decklink device 4 = 363MB 202 PLAY OK stop 1-3 =230MB 202 STOP OK play 1-3 decklink device 4 = 363MB 202 PLAY OK stop 1-3 =230MB (We consider that's negligeable) 202 STOP OK play 1-3 decklink device 4 =363MB 202 PLAY OK mixer 1-3 fill 0 0 0.5 0.5 =390MB 202 MIXER OK mixer 1-3 clear =377MB (and not 363 anymore) 202 MIXER OK mixer 1-3 fill 0 0 0.5 0.5 =402MB 202 MIXER OK mixer 1-3 fill 0 0 1 1 =390MB (and not 363 anymore) 202 MIXER OK mixer 1-3 clear =390MB 202 MIXER OK mixer 1-3 fill 0 0 0.2 0.2 =413MB 202 MIXER OK mixer 1-3 fill 0 0 1 1 =402MB 202 MIXER OK mixer 1-3 fill 0 0 0.2 0.2 =427MB 202 MIXER OK mixer 1-3 fill 0 0 1 1 =414MB 202 MIXER OK mixer 1-3 fill 0 0 0.2 0.2 =439MB 202 MIXER OK mixer 1-3 fill 0 0 1 1 =427MB 202 MIXER OK mixer 1-3 fill 0 0 0.2 0.2 =451MB 202 MIXER OK mixer 1-3 fill 0 0 1 1 =439MB 202 MIXER OK
So with 6 "mixer fill" operations the memory used increases by more or less 70MB. When we do the same operations on a clip, in place of a input SDI, the memory is stable.
What do you think about it?
The same behavior with 2.0.3.f845d1f STABLE