Code Module - what have I missed

jorge · 15 Jul 2005

whyterabbyt wrote: I see stage(0) and stage(1) sections marked out; I assume stage(0) is an initialisation stage, and stage(1) is runtime, yeah? Are there any other stages?

Yeah, stage(0) runs only once in the beginning to initialise and stage(1) runs at sampling rate, once per sample. There's stage(2) and stage(3), so far they have only been used to implement 1 sample delays or what in DSP terms is called Z^-1. If the 1 sample delay was done all in stage(1) you'd get different behaviour depending on the order you'd insert the components. If you'd insert them left to right, you'd get a cascading effect where your input sample would flow through all of them in one go. The correct way is to do all the calculations in stage(1), use stage(2) to move all the internal states to the outputs and use stage(3) to move all the new input values to the internal states ( or the other way, cant remember now). Stage(1) runs for the whole synth before stage(2) rund for the whole synth and then stage(3) also for the whole synth. In most situations you dont have to worry about stage(2) and stage(3).

whyterabbyt wrote: It also looks like you can handle arrays. Are there any other datatypes which havent been listed in the main docs?

I cant think of any other ones.

whyterabbyt wrote: I also see a loop(n) control, which I hadnt been aware of. is the index fixed, or can it be a variable?

n is fixed, but you can easily implement your own incrementing index variable. Just make it zero before the loop and add 1 to it at the end of the loop.

whyterabbyt wrote: Are there any more instructions.

Maybe, I think log10 was implemented recently. I think we just need to update the docs at some stage soon.

whyterabbyt wrote: One last thing; the comparison available dont seem to include equals and not equals... will these be made possible?

You're right. I think it was implemented at some stage, I vaguely remember using it but I've just tried again and it doesnt seem to work, I'll try to fix it soon.

Shifrin · 15 Jul 2005

I believe there is a hop(N) function too - which only executes the code it contains every N samples IIRC :-

hop(4) { a = b + c; }

So every 4 samples a gets set to b + c!

Shif.

jorge · 15 Jul 2005

whyterabbyt wrote: One last thing; the comparison available dont seem to include equals and not equals... will these be made possible?

Maybe for now you can use:

float a=0; float b=0; float result=0; float true=1; result=(a>=b)&(a<=b)&true; // same as result=(a==b)&true;

I just noticed that we also need "or" to be able to do the following:

result=(a<b)|(a>b)&true; // same as result=(a!=b)&true;

Shifrin · 15 Jul 2005

Yeah that's what I did to check equality!

Shif.

note:-

You can do it in the assembler component afaik using the cmpps command with a type '0' for 'equal to' and a type '4' for 'not equal to'.

whyterabbyt · 15 Jul 2005

jorge quoth Yeah, stage(0) runs only once in the beginning to initialise and stage(1) runs at sampling rate, once per sample. There's stage(2) and stage(3).... In most situations you dont have to worry about stage(2) and stage(3).

Right, ta. BTW when you say 'at the beginning' for stage(0), is that once-only 'ever', or when any input values change.

n is fixed, but you can easily implement your own incrementing index variable. Just make it zero before the loop and add 1 to it at the end of the loop.

But you cant do

loop(myIterations){

}

right?

Are there any more instructions. Maybe, I think log10 was implemented recently. I think we just need to update the docs at some stage soon.

That'd help me a lot

You're right. I think it was implemented at some stage, I vaguely remember using it but I've just tried again and it doesnt seem to work, I'll try to fix it soon.

Ta. I know they can be done with chained comparisons, but I wondered if that was more inefficient...

jorge · 15 Jul 2005

whyterabbyt wrote: when you say 'at the beginning' for stage(0), is that once-only 'ever', or when any input values change.

It runs only once when a new process starts. In a mono section that means when you change the schematic or when you switch directSound on for example. On a poly section that means every 4th note you play.

whyterabbyt · 15 Jul 2005

Many thanks...

Nu Audio Science · 15 Jul 2005

I'm losing it totally with this SSE stuff I thought a new process would be started with every fifth note not every fourth note played. I think i will diss this SSE optomising stuff until i can understand it better

EDIT * Meant to ask Jorge why is it the fourth note played and not the fifth that starts a new process, Is it pre emptive or something ?

jorge · 15 Jul 2005

Nu Audio Science wrote: I thought a new process would be started with every fifth note not every fourth note played.

Each SSE instruction processes 4 values in one go, not 5. The process starts on the 1st note you play, the 5th note, the 9th note etc. If it was every fifth note it would be the 1st, the 6th, the 11th etc

I did simplify my description a bit, I can go more in detail if someone needs it, just thought it would confuse things. For example if you play 4 note, stop one and play your fifth note, it will take the place of the free channel rather than starting a new process. So there's some voice management going on but roughly you can think of it as a process starting with every 4th note.

Nu Audio Science · 15 Jul 2005

I think its the way i talk then because i would say that as every fifth note starts a new process, As in if four are playing no new process but the fifth starts a new process.

Guess i should have paid more attention at school

whyterabbyt · 15 Jul 2005

NAS; think of it as being each 4th note subsequent to the 1st note... Not on each 4th note....

1+4 = 5 5+4 = 9

whyterabbyt · 15 Jul 2005

BTW I forgot to ask; is there any way to use comments in ones code?

skOre · 15 Jul 2005

use the double /:

//comment

whyterabbyt · 15 Jul 2005

Hmm, I thought I'd tried that. Oh well, ta.