sfizz: Virtual Playing Orchestra bugs
Yesterday I compiled the sfizz lv2 from the develop branch and tried using Virtual Playing Orchestra. There are bugs; I’m not sure if they are due to VPO using non-standard sfZ opcodes, or if sfizz is missing features that VPO depends on.
Any of the VPO sfz files with PERF in the file name causes sfizz to crash immediately upon playing a single note.
The pizzicato articulations aren’t played correctly, they make a very brief sound and then go silent.
There are possibly other issues, these are just the ones I noticed while doing a cursory test.
About this issue
- Original URL
- State: closed
- Created 5 years ago
- Comments: 19 (18 by maintainers)
Memory corruption could well be one reason for part of these problems, especially if it happens with freewheeling. For now its appearance feels extremely random for me, and I haven’t yet managed to point out precisely what could be happening. I will investigate the session and try to reproduce as much as I can, but feel free to start a new issue for these too.
Thanks for looking and stress testing all of these! 🙂
Thanks, this is indeed great news 😃
Yes, the crossfade calculation are not dynamic right now and they’re only checked when the note start. After reading around, this is indeed not what’s expected so I’ll look into it!
I pinpointed alot of possible issues and solved them (hopefully). On my machines your session plays well now, if you’re willing to give a try to the latest HEAD !
Not very good, especially for such a session. This is a good stress test 😃 The strangest thing is that my desktop (which is a much better machine) can reproduce the errors and they seem to originate from the background file loader, so at least I have one place where I can see the issues.
I will try to get to the bottom of this if work allows in the weekend !
@paulfd I made a “stress test” of sorts, using an old project that I made a while back previously using LinuxSampler. I found that with dozens of instances of sfizz, I get all kinds of snap crackle pop in the output, on both my Mac and my Linux machine. I don’t think that they are xruns, since DSP load remains fairly reasonable. I even get them when I export my Ardour session to an audio file, which shouldn’t ever xrun since it doesn’t have to be realtime.
If it helps, here’s my Ardour session: https://www.dropbox.com/s/njkppe16evuus3h/Pumpernickel and Periwinkle.zip?dl=0
The requirements are sfizz, dragonfly reverb, virtual playing orchestra extracted at
/opt/, and the Little Lizard Piano similarly extracted at/opt/(and in actuality, if you can’t find the little lizard piano, the bugs still reproduce without it).I’m guessing that there is some kind of memory corruption problem, since I experienced similar bugs while developing Dragonfly Reverb. In fact, I initially thought that some old bugs in Dragonfly were back with a vengeance, but I was able to reproduce the bug even after deleting my reverb busses.
I just gave sfizz+vpo a more thorough test. From what I can tell, SFizz now performs the VPO instruments as well as LinuxSampler does, which is great news.
There is one minor bug that I also found in LinuxSampler: The instruments with cross-faded dynamics don’t work. I think the problem is that VPO has those instruments set up such that the mod wheel (CC1) controls both volume and cross-fading multiple samples. See this page for more details about the cross-fading feature: http://virtualplaying.com/dynamic-cross-faded-brass/
Edit: To clarify, when I tried to test cross-fading, I adjusted the mod-wheel controller while playing a sustained note. The volume adjusted accordingly, but the cross fade didn’t happen.