GNOME Bugzilla – Bug 687284
audioconvert: prefer output formats with the same depth or at least a higher depth
Last modified: 2012-12-12 16:32:46 UTC
Created attachment 227741 [details] [review] Use sample width to make an educated guess on the output format Audioconvert should attempt to match the outgoing sample width with the ingoing. At the moment it can do things like convert S24BE to S16LE even though S24LE is supported.
Is this still needed after commit a66ff00908432bb984269433aadd27ba5ec86841 Author: Mark Nauwelaerts <mark.nauwelaerts@collabora.co.uk> Date: Thu Oct 18 22:13:09 2012 +0200 audioconvert: enhance transforming caps ... so as to preserve input format precision, and preferably not convert at all. ?
(In reply to comment #1) > Is this still needed after > > commit a66ff00908432bb984269433aadd27ba5ec86841 > Author: Mark Nauwelaerts <mark.nauwelaerts@collabora.co.uk> > Date: Thu Oct 18 22:13:09 2012 +0200 > > audioconvert: enhance transforming caps > > ... so as to preserve input format precision, > and preferably not convert at all. > > ? In my test case S8 gets converted into S32LE without the patch. So unless my test is broken it seems like it is needed.
Created attachment 227751 [details] [review] Enhance current code to prefer an exact match on sample depth if possible
commit bc4389806d17c8ea27d952b5b539ee9235fa3027 Author: Sebastian Dröge <sebastian.droege@collabora.co.uk> Date: Thu Nov 1 14:31:29 2012 +0100 audioconvert: Also ignore the SIGNED flag when matching an output format commit c286f8ffa2fe78aa4c43d7e589271c53b6578a36 Author: Rasmus Rohde <rohde@duff.dk> Date: Wed Oct 31 20:01:05 2012 +0100 audioconvert: Prefer output formats with the same depth or at least a higher depth Enhance current code to prefer an exact match on sample depth if possible. Also ignore GST_AUDIO_FORMAT_FLAG_UNPACK when checking equality on the flags.
Created attachment 227833 [details] [review] Try to match audio format even better on signed/unsigned formats
Comment on attachment 227833 [details] [review] Try to match audio format even better on signed/unsigned formats I don't think this really makes sense as for audioconvert it doesn't make a performance difference if signedness needs to be converted or not with the current code.