The Powder Toy

Tiny Divider by mark2222

mark2222
mark2222
91 / 2
1st Apr 2019
4th Apr 2019
TPT's first zero-latency 26-bit subframe division circuit. It even computes the remainder! This circuit is based on the radix-2 SRT division algorithm, using LDTC to compress each iteration into a single row.
subframe mark2222 electronic calculator computer electronics divider 60hz divide division

Comments

  • R33sesK1ng
    R33sesK1ng     4th Apr 2019
    @mark2222, This would require stacking, though, correct?
  • Powderman773
    Powderman773     4th Apr 2019
    *Puts virus on the dividers* I'm in. *Numbers go out* I'm broke.
  • mark2222
    mark2222     4th Apr 2019
    @R33sesK1ng You could either not put LDTC directly adjacent (e.g. use two rows of alternating readers), make your memory diagonally-shaped (so all your LDTCs would point towards the top-right, which is good enough for e.g. a single-row PHOT ROM), or use a CONV-LDTC-CONV stack like in this divider (separate the LDTC from the FILT by a single row of alternating INSLs and CRMCs, and use CONV to "select" a single column just before each LDTC updates).
  • R33sesK1ng
    R33sesK1ng     4th Apr 2019
    My concern, though, is how you've been able to use it so successfully. It seems like when LDTC particles are directly adjacent to one another they detect spectra ctypes diagonal from them as well as those on top of them. Idk what to do at this juncture.
  • R33sesK1ng
    R33sesK1ng     4th Apr 2019
    @mark2222, I've been working with some LDTC. I've found it quite easy to compact memory storage b/c of how it can detect FILT spectra.
  • mark2222
    mark2222     4th Apr 2019
    @NoVIcE We generally prefer wide over tall in subframe, because we want data available as soon as possible in subframe order.
  • NoVIcE
    NoVIcE     4th Apr 2019
    If you see a subframe thingery that is a quarter of the screen in length, you know its something ridiculous.
  • mark2222
    mark2222     4th Apr 2019
    Here's the same code, with comments: https://pastebin.com/Df6KU4vr .
  • mark2222
    mark2222     4th Apr 2019
    Update: I've fixed the overflow problem. Now the divider should return the correct value for any divisor under 26 bits (yes, that's one less bit than the original version). I've also uploaded the simulation code here: https://pastebin.com/jLy08z8s , which closely follows the actual FILT computations that the divider performs.
  • Nunnhem
    Nunnhem     3rd Apr 2019
    couldnt figure it out, but still cool +1