Steve Harris wrote: >On Sun, Oct 27, 2002 at 12:23:28 +0200, Tim Goetze wrote: >> that is great to hear! i can almost see the box glowing from >> the heat when the valves are turned on. the iir 'convolver', >> btw, suffers greatly when competing for the cache with other >> plugins now (8-9% @ 64/44.1). think i'll try the floats again. >> though the k6 really is cache-starved compared to modern cpus. > >Well ,the more CPU you have spar, the more tempting it is to un extra >stages and the more cache starved it gets.
live and learn. i'm looking into a few other places right now to see where more cycles can be spared. the alsa out node looks like a good candidate (i'm driving only one of the ten output channels but zeroing the remaining nine at every pass). should turn on -O2 instead of -g for the host, too. :) >> >used for all the stages, I suspect a chebychev would be better for the >> >output, it would also alow us to model different amps with the same code. forgot to ask ... you mean emulate a 'preamp', filter and apply a 'power amp' [and filter again] and a 'cabinet' stage i guess? >> >We'd need to calculate the added harmonics from an amp though, maybe by >> >putting a sin through and FFTing it. >> >> i've uploaded yet another recording, line-sine-16.flac, which >> contains the line out (left) and the sine that went in. seems >> our total transfer function until that point needs more work. btw, the sine was produced by the virus, not the computer. thus it is 'sample-accurate', the output at one point in time directly relates to the sine at the same point. >I'm sure it does. I've been playing with myb ass and an ecasound chain: >ecasound -i /dev/dsp -o /dev/dsp -el:valve,1,1 -el:dcRemove -el:amp,4 >-el:valveRect,0.5,0.5 -el:sinusWavewrapper,2 -el:valve,1,1.5 -el:dcRemove >-el:amp,-10 > >It needs some filters and more work, but it sounds pretty good to me. just tried it, indeed it does. what i like especially is that it sounds quite 'juicy' like good distortion is supposed to. the attack phase is slightly better than with my previous setup, though there's still some auto-wah. still with the mixer pre-eq though, it reads '-15 at 80 Hz' so my guess is the band it attenuates is too low for this purpose. haven't measured its roll-off yet. >I'm having a problem than when the input is 0, the output is infinite. I'm >not sure if its a result of the optimisations I did, or it it would do >that anyway. should be the optimization, i'm getting clean zero out for zero in with your last public release. i still remember the green skies a M$VC optimized POV binary produced, since that day i'm a bit skeptical about compiler optimizations. tim
