News for hackers. Stuff that matters.
Re:Ask HN: What is the most exciting development in your fie... by solidangle on Sunday February 05, 2017 @12:35PM (#13573413)
> This means that the behavior is somewhat unpredictable, sometimes you can have a formula with 60,000 free variables solved in under half a second, sometimes you can have a formula with 10 that takes an hour.

It sadly also depends a lot on the solver used and the way the problem was encoded in SMT. For a class in college I once tried to solve Fillomino puzzles using SMT. I programmed two solutions, one used a SAT encoding of Warshall's algorithm and another constructed spanning trees. One some puzzles the first solver required multiple hours whereas the second only needed a few seconds, while on other puzzles it was the complete opposite. My second encoding needed on hours for a puzzle which I could solve by hand in literally a few seconds. SAT and SMT solvers are extremely cool, but way incredibly unpredictable.

Re:Ask HN: What is the most exciting development in your fie... by munin on Sunday February 05, 2017 @01:46PM (#13573806)
Absolutely! For another example, Z3 changes what heuristics it has and which it prefers to use from version to version. What happens when you keep your compiler the same but use a newer Z3? Researchers that make these tools will flatly tell you not to do that.

It's frustrating because this stuff really works. Making it work probably doesn't have to be hard, but researchers that know both about verification and usability basically don't exist. I blame the CS community's disdain for HCI as a field.