Strange benchmarks. about crystalline HOT 3 CLOSED

@yxhuvud thanks for the analysis. I did some poking around and found some unneeded code that was eating cycles. Can you take a look and verify better performance?

from crystalline.

Yes, the performance is much better now.

crystalline - push 0.030000 0.000000 0.030000 ( 0.022083)
crystalline - pop 0.350000 0.000000 0.350000 ( 0.237548)
crystalline - mixed 0.050000 0.000000 0.050000 ( 0.043344)

Which makes a lot more sense. Still worse than the values I get of my priority queue, but that makes sense as that particular heap type I use is optimized to the needs of simple priority queues:

mine - insert 0.010000 0.000000 0.010000 ( 0.006840)
mine - delete 0.010000 0.000000 0.010000 ( 0.019307)
mine - mixed 0.010000 0.000000 0.010000 ( 0.005895)

from crystalline.

Cool. I'm closing this, as the strange behavior seems to be under control.

@yxhuvud, feel free to submit a PR with improvements to the priority-queue/heap implementation. Also, it might be cool to experiment with making the priority queue generic over an abstract heap (e.g. using the built-in PQ by default, but could use a pairing-heap instead).

from crystalline.