Michel Daviot
Michel Daviot
Yep, this is the alpha beta optimization of negamax.
On the branch based on 1.2.14 it was working below 100ms  In our tests (and in production), we allow 200ms for either BruteForce or LAFF to return a result...
To elaborate a bit on our use case : we want to know if a list of boxes can reasonnably (in a way which a human would find) fit in...
@skjolber what do you mean exactly with v1 ? In the branch 1.X ?
I compared the performance between 1.2.16 and 1.2.19 and it is very similar
Cool ! I guess this was released in 2.0.11 ? I'll try again with my benchmarks in the next days
Well, 2.0.11 sounds promising. FastBruteForce can now pack efficiently this problem: container [300x350x300] with 32 boxes: 15x182x182 8 and 50x66x231 24 (1.X does not find a solution) However I found...
I ran my benchmarks to compare 1.2.19 and 2.0.11, and on real data I'm losing up to 90% 😦 
This was done with 2 implementations running concurrently to solve each problem - BruteForce 1.X or FastBruteForce 2.X - LAFF 1.X or PlainPackager 2.X - this implementation runs over all...