[Soc-2013-dev] Weekly Report #10 VSE
Alexandr Kuznetsov
kuzsasha at gmail.com
Sat Aug 24 06:20:54 CEST 2013
==Week 10==
* Finished scopes (histogram, waveform, vectorscope)
* Created memory pathways
** Between char and float
** Passing original char to gpu and making conversion there is *much*
faster.
** Added cl_mem to RAM conversion. However, it introduced a lot of
crashes. I need more time to figure out what is wrong.
* Code reorganization and clean up.
* Just started implementing cache.
==Next Week Todo==
* Will commit the code over the weekend
* Cache for input strips and final images
* Add proxies
* Create duplicate scene to get correct values at different time.
== Questions ==
I think adding threading for each seqDeviceCPU. The system works well
when it is in playback mode, rendering multiple frames on separate CPU
cores. However, for a single frame, this approach is limited by the
number of non-dependent buckets. To do better, I propose to have
multiple threads doing the same seqCommand (in a way it was implemented
before). GPU doesn't has this problem, plus, its power should be
sufficient for the most cases.
More information about the Soc-2013-dev
mailing list