<Wanda[cis]>
alright! I have finally implemented the new bitstream reversing algorithm
<Wanda[cis]>
and it managed to find a bunch of SiliconBlue bits, exactly as expected
<Wanda[cis]>
now I need to actually make the SiliconBlue binding emit more reasonable and varied netlists so it can actually find all bits and not just luck out on some unambiguous enough
<Wanda[cis]>
so the plan is to first finish the SiliconBlue binding to see how this thing performs on small devices, and then aim it at the actually interesting target which is UltraScale.
<Wanda[cis]>
UltraScale. UltraPlus. who can tell the difference anyway.
<cr1901>
One's ultrabig the other is ultrasmol
<jn>
the cowards didn't name it UltraMinus
<Wanda[cis]>
it's called UltraLite
<Wanda[cis]>
hrm.
<Wanda[cis]>
testing found rather annoying flaw in the algorithm.
<Wanda[cis]>
probably fixable, but really annoying.
<Wanda[cis]>
so when an interconnect mux has an "enable" bit (set for all used mux configurations), like ice40 does, and the mux solely drives other muxes within the same bittile, my correlation algorithm cannot decide whether the bit should be assigned to the correct mux, or to whatever mux input it is driving
<Wanda[cis]>
this needs to be solved by considering ... mux triplet combinations I think? ugh
<mupuf_>
Wanda[cis]: This is the annoying moment when reality comes crashing down on your best-laid plans :D
<Wanda[cis]>
meh, just needs a bit more violence
<mupuf_>
for sure!
<Wanda[cis]>
I already figured out the solution fwiw, just... there's prjunnamed stuff prioritized in front of it