From the original SynthMaker forum (2004–2014), with its members’ names and dates.
rl ·
I failed adapting the formant filter code from musicdsp.org. According to missing precision of floats the filter runs unstable. Doubles would cure this. I know this is in contradiction to SSE optimization, but sometimes float is not good enough.
Shifrin ·
rl wrote: I failed adapting the formant filter code from musicdsp.org. According to missing precision of floats the filter runs unstable. Doubles would cure this. I know this is in contradiction to SSE optimization, but sometimes float is not good enough.
I think jorge said we might get this (doubles in the code component). You can use doubles with SSE, you get two channels instead of the four you get with floats 🙂
Shif.
infuzion ·
Maybe then just open a few double-float assembly commands?
Tom7777 ·
Probably sounding stupid, but remember im not a great coder (yet) but what exactly are doubles? and in what ways can they be used?
Shifrin ·
'Float' numbers are represented using 32bits, these are the floats we use in SM, sometimes we need more accuracy though. Doubles are more precise 64bit floating point numbers. Probably not an exact description but hey 😉
More accurate description :-
http://en.wikipedia.org/wiki/Double_precision
Shif.
Tom7777 ·
Thanks Shifrin
rl ·
Shifrin wrote: I think jorge said we might get this (doubles in the code component). You can use doubles with SSE, you get two channels instead of the four you get with floats 🙂 Shif.
thanks. I've tried some opcodes (e.g. addps, addsd) - but no luck
Wrote something here? To have a post removed or your name hidden, write to [email protected].
SynthMaker