On Fri, 2011-05-27 at 12:20 +0200, Kassen wrote: > Dear list, > > > Right now in the redacted branch and perhaps in fluxa, if these were > merged yet, there is a bug with (sample ). I changed that node to > playback in reverse when the frequency is negative. That's a good > feature, but now it has two issues. Once the node is set to play in > reverse it will keep doing so even when it's recycled. This leads to > interesting "non determinism" but clearly that's wrong.
I quite like the sound of this. > Worse is that if a sample is still loading while this happens Fluxa > will crash. I can't remember the mechanism for not playing a sample until it's loaded, but it shouldn't be any different. > I can fix it, I already had to deal with a related matter for (scrub ) > and hope to do so this weekend. This is more of a warning that right > now this is unsafe. Use (scrub ) with a (ramp ) for your > sample-reversing needs. > > > Other changes I plan to make; > > > (xfade ) when only fed floats should return a float, not a node, this > should speed things up and relieve some strain on the "node-pool" > > > (kas-filter ) is by far the most expensive node, cpu-wise. This is > partially a result of extra care in sound quality (with input > interpolation) but part of the blame is likely in calculating a > cosine, which could be done with a table. I can have a look at some optimisation - I don't think your stuff is in 0.18 yet, I'll try and get some time to look at all this next week - it would be good to have it in puredyne (so I can use it all in scheme bricks workshops :) ... cheers, dave
