Changing the velocity root-finding, line 13 of algorithm 1 in the tandem paper.
Instead of searching for $V$ S.T
$\sigman F(V, \psi{k}^{e}) + \eta V - T = 0$
searching for $\xi$ S.T
$V = 10^{\xi}$
It was tested against bp1 bp3, and my own SSEs simulations and it shows ~40% improvement in run time when running with green functions.
bp1:
new root-finding run time: 554.237
main run time: 888.816
bp3:
new root-finding run time: 4099.94
main run time: 6373.47
Is more testing needed before merging into main? If so, what would you like to test?
yohaimagen
commented
2 days ago
Changing the velocity root-finding, line 13 of algorithm 1 in the tandem paper. Instead of searching for $V$ S.T $\sigman F(V, \psi{k}^{e}) + \eta V - T = 0$ searching for $\xi$ S.T $V = 10^{\xi}$
It was tested against bp1 bp3, and my own SSEs simulations and it shows ~40% improvement in run time when running with green functions.
bp1: new root-finding run time: 554.237 main run time: 888.816
bp3: new root-finding run time: 4099.94 main run time: 6373.47
Is more testing needed before merging into main? If so, what would you like to test?