2008/4/23, Albert Santoni <[EMAIL PROTECTED]>:
>
> Hi guys,
>
> For those of you who haven't been on IRC recently, I just wanted to give
> a brief update on our quest to figure out this memory corruption
> business:
>
> 1) Adam, Pawel, and I figured out how EngineBufferScaleDummy and the
> EngineScaling classes in general are supposed to work. Our dummy
> implementation doesn't do any scaling, it just copies memory into the
> right places and scales the play position accordingly. I'm fairly
> confident that the code we have in there is bug-free.
>
> 2) EngineBufferScaleLinear looks OK to me. With this in mind, I'm going
> to switch "vinyl emulation" mode over to using our own linear
> interpolation code (via EngineBufferScaleLinear).
>
> 3) The only problems I've encountered while using the linear
> interpolation code and the dummy scaling code is a deadlock, and it's
> only happened when I'm using vinyl control. To be clear, I haven't had
> any random asserts or crashes with the dummy/lerping code. I'm
> optimistic that we may have found the (or a?) source of the memory
> corruption (EngineBufferScaleST).
>
> So, all of this is going to need testing soon, whenever these changes
> hit trunk. If other people find the linear interpolation code to be
> stable for them as well, then some of us should try to take a hard look
> at EngineBufferScaleST and figure out what's wrong there.
>
> In the worst case, if we can't figure out what the problem is, I might
> take a look at Chris Cannam's Rubberband library. Rubberband does the
> same sort of stretching as Soundtouch, but is much newer. From what
> we've learnt from EngineBufferScaleDummy, we should be able to throw
> together some code that uses Rubberband without too much trouble
> (*crosses fingers*).
>
> Anyways, we're making progress!
> (I'm a little disappointed by the vinyl control deadlock, but I only
> have myself to blame, hah) :)
>
Great work so far, you three!
I'll be sure to keep testing the stability of trunk.
--
Wesley Stessens <[EMAIL PROTECTED]>
Human Knowledge Belongs To The World - Antitrust (2001)
http://wesley.debianbox.be
-------------------------------------------------------------------------
This SF.net email is sponsored by the 2008 JavaOne(SM) Conference
Don't miss this year's exciting event. There's still time to save $100.
Use priority code J8TL2D2.
http://ad.doubleclick.net/clk;198757673;13503038;p?http://java.sun.com/javaone
_______________________________________________
Mixxx-devel mailing list
[email protected]
https://lists.sourceforge.net/lists/listinfo/mixxx-devel
