DSP Robotics Support

Nope.

Definately not an analog system

Think I'm bleeding out of my ears

Two filters in series at high resonance. Didn't think an already clipped sound at the output could pierce the ears that bad holy smokes...

Ok. I was planning to give extra warning through graphics from the beginning. But a soft clipper or somekind of extra limiter in filter series-mode might be in place maybe

Check out these filters, maybe of use for your project?

Check out these filters, maybe of use for your project?

You're amazing! Top notch!

Were they already up on the forum somewhere? I've completely stopped searching on the forum so I would'nt know if they were...

And I was already making measurable gains by sorting my buses as I write this. Thanks to Tepeix test he did (thank you!), i'm starting to suspect an unproportional overhead cost when merging buses, even if there aren't any contained values inside thats subject to addition.
Two whole voices gained at 23-24 played. Haven't increased max voices but maybe even more gained.

Can't wait to check performance with your optimized filters also...

The anticipation was palpable...

...for me

Tested the standalone and with recent changes and MVs ASMs in place I gained another solid 1 voice. So pushing 24 voices on my system with one of the plugins osc modules I know to previously act up at 23-24... only now causes trouble at 25 (very stable limit at next "4 pair" of voices? which seems good?).

But I also gained even more performance on some osc modules (with some settings)... and can push 32 voices no problem at all.
And I've noticed that the same osc module can have different performance depending on settings so that gives me a good lead to find out what differs, and maybe fix.

Otherwise it's hard to compare max voices between my osc modules since they differ so much. And the amount of modulations makes some difference but not that much since all routing is constant. Which was the plan all along.

The performance is more stable than before and looks like <1% per voice (in 4% steps) with an constant overhead cost of around 2-3% in total.
So 24 played voices eats around 19-23% CPU on my system. Which right now is a light i7 laptop with a (short term?) turbo of around 4GHz (or right above) with asio4all @ 512 buffer size. Not very optimal but good for testing purposes...

Haven't payed much attention to it in FS3.09 export since its exports (64bit ones) seems to performs better than the old 32bit exports.
But a little strange to see only one cpu core working... FS maybe doesn't multithread in that way? what do I know.
I'm not even remotely knowledgeable at that stuff.

The Chebychev ASM was a no go and crashes in 3.09 exports. Found the same asm among the synthmaker greatest hits files and there it was commented and made by some Dozius? so it seems to be an old ASM. But doesn't matter, the DSP works just fine.
Going to take a listen to the ZDF filters you provided MV. Must have missed those also if they were posted on the forum somewhere

Here is a 4th order Chebychev lowpass filter. Supposedly similar to Dozius' but leaner and compatible with FS 64bit.
Have fun!

Here is a 4th order Chebychev lowpass filter. Supposedly similar to Dozius' but leaner and compatible with FS 64bit.
Have fun!

Superb!
Throwing that in there immediately!
Edit: I see a green and Ruby... Gonna have to stick to the old DSP on this one I think

I have some work cut out for me regarding the UI so I'm going attend to that for a while. Maybe leave you guys and the forum alone from my cascade of ignorance a few days


MV do you have your "Initials" logo in higher resolution PNG with transparent background? Preferrably without the dropshadow? As I said i'm adjusting the plugins about screen accordingly for next export. There will be a forwarded
link. But only if you like me to add one. Let me know...

Last reflection for now...
If incorporating optimized ASM's for only two simultaneous filters makes a dent of 1 voice? How much cpu does my 5 toolbox envelope-DSP's actually eat?

Still....
I'm really impressed by FS's handling of streams, I don't know if there is somekind of priority going on when
streams are inactive (0) for an amount of time. But, I removed really stupid unnecessary multipliers a few days ago. Thinking that this would make a large impact on performace, the amount of multipliers was 690!!!
Hardly any impact at all. I know floating point operations are blazingly fast. That's what CPUs do. But still...

R&R wrote:I see a green and Ruby... Gonna have to stick to the old DSP on this one I think

This is for the resonance only. The cutoff is all stream, so you can modulate that or control it with an envelope. I have updated the 4th order Cheby Lowpass fsm above with an all ASM variant so you can modulate both cutoff and resonance. The CPU penalty is not severe.

I have updated the 4th order Cheby Lowpass fsm above with an all ASM variant so you can modulate both cutoff and resonance. The CPU penalty is not severe.

Top!!! Splendid!

Wasn't able to throw in the Cheby LP ASM seems it doesn't play nice, atleast with a simple feedback loop...
Might just be the way the ZDF*s acts or an effect of the ripple adjustments. I'm a moron so I wouldn't know either way Haven't had time to test the ZDF yet, so I might find they act the same on their own when running them in a loop...

But doesn't matter since the Cheby isn't that important. The SVF and 303:ish seems to work really well... only had to set a lower limit on the 303 reso I think. I may have adjusted that in the DSP previously I don't remember.

Anyway MV...
You are now a permanent part of the about dialog in the plugin I kept it "low profile" as before so no spelled out names, and no direct linking... only a forward from a subdomain. So I can remove or redirect to any webpage if needed.