Rockbox mail archiveSubject: Re: Playback Engine Rework - The basic rundown
Re: Playback Engine Rework - The basic rundown
From: Michael Sevakis <jethead71_at_sbcglobal.net>
Date: Mon, 25 Apr 2011 21:44:11 -0400
----- Original Message -----
> From: Bryan Childs
> To: Michael Sevakis ; Rockbox development
> Sent: Monday, April 25, 2011 3:25 PM
> Subject: Re: Playback Engine Rework - The basic rundown
> With a list of pros that long, I find it hard to say no to this.
> Is there a list of cons that go with it too?
It's a sale? :-)
Bigger binsize I guess though not necessarily bigger mem usage (last time I
looked at certain targets) and more is allocated for during playback only
rather than statically; number of track items statically allocated follows
More code of course but nothing added for the heck of it. On the plus side
for that, apparent causes (code executed) and effects (results of that code)
are more directly coupled together. Then again, maybe I see it that way just
because it's my code and not another's; let the reader decide I guess. :-)
A bit more data copying due to isolating the areas and keeping playing track
info coherent but it's not done very frequently.
There are some warty bits necessitated by the surrounding interface it must
We're still in flux!
Have to update the "How to write codecs" wiki. Fun!
Might have shiny new bugs (but the old ones were getting soooo boring).
There is a tiny glitch introduced with cuesheets where skipping back might
briefly show the cue two before the current instead of the previous one that
didn't seem to be present in the current implementation with what I used to
test (hadn't bothered to look into concealing the seek imprecision again
just yet, if it even should be concealed - maybe a better way exists). It
doesn't last long since the codec progresses to the cue point pretty
quickly. It does look a bit odd to me but the cuesheets still work well.
As mentioned, somehow I or someone that knows has to deal with those pesky
codecs (NSF and one line in RAAC I believe is all there is in SVN) depending
on being able to poke information directly into the WPS. I propose just
tempering what can be done until some sort of more proper interface is
provided because allowing "metadata push" could still be useful.
There was another minor glitch I mentioned in IRC. Not sure if I fixed it
but I definitely seem to have forgetten what it was. :-\
Received on 2011-04-26