The Baistow solver results change with each run...
emsr opened this issue · comments
This is due to some randomization.
- it would be nice to specify a seed for reproducability
- Results point to a need for a round of root polishing.
I shouldn't create a new urng and distribution every time the randomization ttriggers. Make it a member of the solver.
git commit -m"Make Bairstow take a seed. Make random, things part of the state. Try some refinement tricks for Bairstow. Meh."