kissat

It's the same overall project, but that result uses kissat[1] solver to complete the solve, this update is that we made our own kissat-style solver specifically for SHA-256. (As of the yesterday's post, I only said "we are working on our own version of the kissat solver based on these properties"[2]). Now we have that solver. It's exciting because we have a large body of Lean-proven theorems, many novel, and we think we can apply many of them directly in this sat solver. Of course, we could be going down a path that doesn't lead to a collision, as we did when we tried to extend the reduced-round records using a mini neural network.[3] We'll just have to see if we can extend the results at all. [1] https://github.com/arminbiere/kissat [2] https://stateofutopia.com/papers/2/we-broke-92-percent-of-sh... [3] https://news.ycombinator.com/it

https://github.com/domschrei/mallob I've seen talks by AWS where they claim that distributed SAT solving is very effective. Do you need it to be parallel or just fast? My impression is that at the single machine level, it's hard to beat kissat and the CPU or GPU parallelism is not that useful. https://github.com/arminbiere/kissat

That is true, but there are many interesting heuristics in modern SAT solvers, such as kissat: https://github.com/arminbiere/kissat (winner for a few years now). LBD ("glues" in glucose) is one of them. Kissat is actually quite readable. More readable is CaDiCaL, also extremely fast and effective: https://github.com/arminbiere/cadical I personally also develop a SAT solver called CryptoMiniSat, but the above two are easier to read :)

For obscure reasons I know that this is the Finnish word for 'cats', and I wondered if there was any connection there. Apparently the answer is yes https://github.com/arminbiere/kissat or at least the developers have already noticed this coincidence for themselves.

It supports the DIMACS input format [1], which you can parse using the kissat_parse_dimacs function [2]. DIMACS is a pretty terrible format, but at least you don't have to build the clauses up yourself using kissat_add. [1] http://www.cs.utexas.edu/users/moore/acl2/manuals/current/ma... [2] https://github.com/arminbiere/kissat/blob/master/src/parse.h

If anyone else is wondering how the hell to use this thing... the main API is here: https://github.com/arminbiere/kissat/blob/master/src/kissat.h In particular all you need to add clauses is this single API entry point: void kissat_add (kissat * solver, int lit); The solver consists of the conjunction of disjunctions (I hope I have that correct — maybe the other way around? I can never remember). Variables are assumed to be any non-zero integer: kissat_add(s, [N != 0]); You create a clause by repeatedly calling kissat_add; the clause is terminated by a 0. For instance if {shirt=1, tie=2}, then "!shirt or tie" is: kissat_add(s, -1); kissat_add(s, 2); kissat_add(s, 0); Once the solver is run, if there's a solution, you get the solution using `kissat_val`: kissat_val(s, 1); kissat_val(s, 2); I hope.