dcpomatic/src/lib/writer.h, branch attic/gpu DCP-o-matic DCP tools https://git.carlh.net/cgit/dcpomatic/atom?h=attic%2Fgpu 2015-12-04T21:14:06Z Use libdcp's compress_j2k; move Data into libdcp. 2015-12-04T21:14:06Z Carl Hetherington cth@carlh.net 2015-11-30T00:05:34Z urn:sha1:6c19b21e12f73ce63edd406ea617ff25bcc9bfea Fix reel that subtitles are put in when they are exactly on a reel boundary. 2015-11-27T20:28:26Z Carl Hetherington cth@carlh.net 2015-11-27T20:28:26Z urn:sha1:84cab4dbc3beba84ba53cec1054ede7c250a4c9c Move quite a lot of stuff out of Writer into a new class 2015-10-22T09:30:45Z Carl Hetherington cth@carlh.net 2015-10-21T09:04:37Z urn:sha1:7a65b08f1b09beacc0e0a8060641f677683d400c ReelWriter, which hopefully cleans things up a bit. Remove ref_write mechanism and instead maintain state for each 2015-10-22T09:30:45Z Carl Hetherington cth@carlh.net 2015-10-21T08:23:14Z urn:sha1:e25bf0c33f4085f6caa3d0d19a083399a422146a reel being written so that we don't need to keep track of frames that are being referenced. Various multi-reel fixes. 2015-10-22T09:30:45Z Carl Hetherington cth@carlh.net 2015-10-20T08:55:24Z urn:sha1:2c5398adaea8bc56fe196144a9a233981657fba0 * _video_frames_enqueued -> _position * some int -> Frame replacements * Call Writer::ref_write when there is nothing to encode / write for one frame. * Assume Encoder::encode is called once per output video frame. * Possibly correct some off-by-1 overlap bugs. * Don't reset _last_written on FAKE write as I believe there is no need. Rename split-by-video content slightly; fix referencing to multi-reel DCPs. 2015-10-12T14:05:09Z Carl Hetherington cth@carlh.net 2015-10-12T14:05:09Z urn:sha1:f9068dcbfbb09082e29e2a779ef1a7a2f6ee849e Basics of reel split. 2015-10-12T10:05:26Z Carl Hetherington cth@carlh.net 2015-09-16T14:38:53Z urn:sha1:dbfbdcba0f5f08a932ba199039f2ca1530e482ac Back-end for very basic and hacky VF support for a DCP imported as content. 2015-09-15T23:52:50Z Carl Hetherington cth@carlh.net 2015-09-15T23:52:50Z urn:sha1:dd59755d86cdb1892bb0ba4d69af520cee4aa964 Optimise checking of existing image data. 2015-09-14T09:21:12Z Carl Hetherington cth@carlh.net 2015-09-03T15:26:52Z urn:sha1:265a02fceb9309338f8ce028ba93bb62cda53746 When checking existing image data go from the end backwards until a good hash is found rather than from the start forwards until a bad hash is found. This is a lot quicker for long files but may miss a corrupted frame somewhere (however unlikely this is). Don't start thread in constructor. (Writer) 2015-09-01T20:05:25Z Carl Hetherington cth@carlh.net 2015-09-01T20:05:25Z urn:sha1:cefba53fe130807e96e608ad3bbb87786b654c6e