Worked a bit on #CommonSubexpressionElimination #CSE today. Saving one addition in the perturbation inner loop makes #BurningShip take 6% less CPU time in that phase. Doing some more for the derivatives saves 17% overall. Other formulas show higher savings, but it's slow doing these by hand (KF has around 100 formulas, today I optimized 5 of them...).

https://code.mathr.co.uk/kalles-fraktaler-2/blobdiff/997b1df0f9d05b8dc0a5da598153fcf6e0ee3f39..45fb14996da61c118e96270e9f7edccb6b769435:/formula/formula_1_2.xml