I'm trying to understand the colorutils fill_gradient_RGB function.  I think datatypes saccum87 is int16_t

Archive FastLED Archive
,
1

I’m trying to understand the colorutils fill_gradient_RGB function.

I think datatypes saccum87 is int16_t (since the uint8_t are multiplied by 128) and accum88 is unit16_t? What is the purpose of introducing these datatypes?

What does this line do?
int16_t divisor = p2 ? p2 : 1;

=
Question behind this question is that I notice some clear steps in my full brightness/saturated hues. Which isn’t really strange, because I use an RYBWheel function over the HSVRaw function (see testRYBwheel): https://gist.github.com/kasperkamperman/c720acf399efbae7e7b1

I think I lose a bit too much resolution. I’m now thinking where it could be smart to solve this. Maybe in this RYB function (where I could keep using the 16bit fractions of the fill gradient function)?

Another option would be using 64 “samples” from the RYB,HSV function and using 4 steps RGB in between with NBLEND. (I tried the 16palette, however that is to much difference).

I probably can also solve it by interpolating between 2 led buffers, however that is rather expensive (because I need to do it on all my 128 leds). However it can give smooth results. Would it be beneficial to even do that in 16bit?

I work now with the Teensy (ARM) so I’m not that limited.

Any thoughts and suggestions are welcome.

2

I’m not an expert on C, but I think

int16_t divisor = p2 ? p2 : 1;

Is a division by 0 protection. That code is a ternary operator that says "if p2 is true, use p2. If p2 is false, use “1”. The only way p2 would be false is if it is 0.

3

For now I’ve implemented interpolation with 3 buffers. It doesn’t seem so expensive as I had imagined (500 us). Please don’t confuse nblend with blend, that costed me a half night of sleep.

Here a working example. A random hue is picked in every new frame. We blend in 255 steps towards that value (we can increase the stepsize of course).

4

You know, I should write that down somewhere in the docs, and my apologies for the confusion: the “n_____” naming convention here means that the function is destructive; it modifies it’s argument (or it’s ‘this’ object) in place.

The convention comes from Common Lisp, where a function like “string-downcase” would return a new copy of it’s argument, but in lower case, and the corresponding function “nstring-downcase” would actual modify the string you passed into it, altering it in place to be all lower case.

It’s probably not a useful convention if no one else knows about it though. Added to my documentation To-Do list.

5

Well, I have to blame myself mostly not understanding pointers well. Luckily you get more skilled making mistakes :slight_smile: Still the best way to learn.

6

An old man told me once: “Experience is the sum of all the mistakes you made yourself.”