The compensated summation worked well in eliminating errors in the computation of t, x, and v in ah18 (as confirmed by stepsize_precision.jl). However, this is not yet added to the versions of these routines which compute the Jacobian, so I need to add this.
The compensated summation worked well in eliminating errors in the computation of
t
,x
, andv
inah18
(as confirmed bystepsize_precision.jl
). However, this is not yet added to the versions of these routines which compute the Jacobian, so I need to add this.