If you have enough computing power it´s really worth to render animations with 32

If you have enough computing power it´s really worth to render animations with 32 bit precision.
Together with long palettes and a high fps rate the image quality is just amazing.
Even on a damn 16x16 8 bit matrix. The Teensy 3.6 seriously rocks.
Thanks for this experience @PaulStoffregen .
https://www.youtube.com/watch?v=mubH-w_gwdA&feature=youtu.be

think a 3.2 could handle this?

@Randal_B Not really. My implementation runs @300+ fps on a T3.6 and @17 fps on a T3.2… The code would need some optimisation or the precision (and by that the quality) has to be less.

@Stefan_Petrick oh wow thats a giant leap. Ive got a few 3.6 around, i may need to upgrade my rig to!

@Stefan_Petrick that is gorgeous!

Any chance to see this circles pulsing to some beat :slight_smile: ?

@Nikolay_Hristov When I come back to soundreactive experiments I´ll try this.

would you mind sharing your code? id love to be able to add to my vest

Sure, @Randal_B , is this snippet enough to reproduce it? https://gist.github.com/StefanPetrick/dc666c1b4851d5fb8139b73719b70149
Please show us a video from the result.

the work that i have started for crgb16 would also make crgb32 trivial :slight_smile:

Any ideas on how to get this to work with less precision on a 3.2? it runs well for about a minute and then drops to about 2FPS.

@Chris_Faye The ounly solution I could imagine is to fill all the RAM with a giant precalculated lookup table for the sine data and grab the data from there instead of letting the 3.6 FPU do all the calculations again and again. But the square roots are still expensive.

@Chris_Faye @Stefan_Petrick Have you guys tried using https://github.com/Anon-Penguin/SpeedTrig its only Trig funstions so no sqrt but its butloads quicker than built in fp ops