template or tutorial for custom c++ scriptnode?

[Morphoice]

@HISEnberg not fast enough. 90% is trial and error and me not understanding a thing I do lol

[d.healey]

@Morphoice said in template or tutorial for custom c++ scriptnode?:

not fast enough

You're plenty fast!

[Morphoice]

@HISEnberg @d-healey actually tried implementing the μ-law compression thing from what I could learn around the web, but it's failing to compile and I cant get a debug version setup on mac as @griffinboy's tutorial only covers PC so far ;) So now I'm stuck lol. But yeah, the problem should be minor and some day I'll find out ;)

                    float val = samples[i];
                      
                    // mu-law compress signal
                    float max = 255;
                    float valmu  = sgn(val) * log(1+max*abs(val)) / log(1+max);
                    
                    // Quantize to 8 bits
                    float valmu8 = valmu - fmodf(valmu,1/max);
                    
                    // mu-law expand signal
                    float valex = sgn(valmu8) * (1/max) * (pow(max+1,abs(valmu8))-1);
                    
                    // Apply Demo Gain
                    samples[i] = valex * smoothGain.advance();

I basically know too litte about c++ to know if the syntax in these formulas is even valid but I tried ;)

[d.healey]

@Morphoice said in template or tutorial for custom c++ scriptnode?:

but I tried

Some of us try for years, don't give up after 2 hours :)

[Morphoice]

@d-healey thanks to your encouraging words it just came to me that actually I just needed to implement the sgn (sign) function as C++ doesn't seem to have one by default

    template <typename T> int sgn(T val) {
        return (T(0) < val) - (val < T(0));
    }

(yeah that one I copied off the internet lol)

and then it suddenly worked. I can't believe it! I have just made a μ-law bitcrusher without even so much as knowing what I do lol. God this thing sounds fly. Massively less artifacts on quieter sounds and sizzle on loud ones!

[Christoph Hart]

@Morphoice use hmath::sign() :)

[Morphoice]

@Christoph-Hart that would have been to easy ;))
In all honesty, I seriously need to learn C++, the things one can do...

[ustk]

@Morphoice everything should work in your code once you:

• use the real math functions as Chris suggested
• respect the variable types. SNEX is very picky about this which is a good thing, but the errors messages are sometimes not helpful (you get the cast issue at the right line for expressions, but not in variable definitions). In your case, when a var type is float, EVERYTHING needs to be float in the expression. (1.0f)

Another suggestion in SNEX, is that you should try to compile for each single line you add to your code, this is especially true at beginning. Because debugging multiple lines you added will be a nightmare as the console writes everything on the same line (remember the Hise Hang )
This reminds me to have a look at least to fix this issue...
So write a new line -> compile, and if it fails you know where you are

[Morphoice]

@ustk aye! I did this in C++ not SNEX though. But I agree it looks more correct this way, although it didn't throw a tantrum without. C++ seems to be a bit forgiving there, which is not at all what I heard about it lol

float val = samples[i];
int max = 255;
float valmu  = sgn(val) * log( 1.0f + max * abs(val) ) / log( 1.0f + max);
float valmu8 = valmu - fmodf(valmu, 1.0f / max);
float valex = sgn(valmu8) * (1.0f / max) * ( pow(max + 1.0f, abs(valmu8)) - 1.0f );                   

[Morphoice]

@Christoph-Hart is there a way to get the custom c++ scriptnode to only process data if there is any sound? as this is intended for one-shots but constantly does the encoding/decoding and a couple of those nodes produce quite a fair amount of cpu

[ustk]

@Morphoice I can think of two ways:

• either build a detector that is waiting for a certain value to pass a threshold, then if it does, do your stuff
• tell the node a signal is coming with a modulated parameter, which will move the detection part to the graph

[Morphoice]

@ustk are all nodes generally on? I've read about a hasTail parameter that stops processing when the note is done etc. probably in my case it needs to work only a few ms after a note was fired

[ustk]

@Morphoice I reckon hasTail is for the case where DAWs cut the plugin processing between the clips to save CPU on big projects. hastTail allows your plugin to still read the buffers until a minimum threshold has been reached before effectively cutting the sound (i.e. for reverbs, etc...)

If your node is optimised, you shouldn't need to bypass it. If the heavy processing is done then you can just do nothing in the process callback to save CPU using just one branching (if statement)

I don't know how your system is built, but if your sound is triggered with MIDI you can use midi events in C++ node with handleHiseEvent, but if it's just reacting to audio then...

[Morphoice]

@ustk don't VST3 plugins in general only waste cpu when they detect the presence of an audio signal?

[Christoph Hart]

@Morphoice might be true but you can‘t rely on every host to implement this, also there are other plugin formats which you need to factor in.

[sletz]

In the meantime, written in Faust: https://faustlibraries.grame.fr/libs/basics/#bamulaw_bitcrusher
and the code here https://github.com/grame-cncm/faustlibraries/blob/a222717339e5f528787080e5e5438ffef8da22a9/basics.lib#L2429 and https://github.com/grame-cncm/faustlibraries/blob/a222717339e5f528787080e5e5438ffef8da22a9/basics.lib#L2371

[griffinboy]

@Morphoice

Yes there are a few ways. The best is probably to use the built in flags at the top of the c++ node. But you can do custom detection too, which I sometimes use if I need specific behaviour.

You'll probably want to optimise the nodes in the first place, I see you've got all the random example scripts (smooth gain ect) that aren't super necessary for your use case. I can give you a more simple template if you want one.

I intend to go back and cover this c++ node stuff better and in more detail but I'm short on time currently. I will be releasing a proper template in the future, with a lightweight library to go along with it, containing efficient approximations of math functions (rather useful for your bitcrusher script), automatic lookuptable creation, and hopefully some tools for SIMD.

And yes - you are progressing really quickly! I get the feeling that you are going to do really well. I don't think I'll have a plugin released until at least two years time Too much perfectionism

[Morphoice]

@griffinboy thanks man! a simpler template would be awesome, I also figured It would make sense to do the calculations only if samples[i] isnt zero, that should take most of the load in my case.

[griffinboy]

@Morphoice

Yep that makes sense, since you don't need to process the zero crossings for your effect.

C++ Node Templates – Google Drive

Google Drive (drive.google.com)

Here's a more up to date version of the template. You'll have to wait until I've got time, for me to release the official version of this, which will have many more improvements

[griffinboy]

@Morphoice Did a small fix there was an error in the template

C++ Node Templates – Google Drive

Google Drive (drive.google.com)