Graham Rowlands

Hi Chris, thanks for the replies. This `u_modified` function seems like a nice trick to avoid dealing with discontinuities, but it doesn't seem to change (i.e. improve) the performance. I...

[statprof_no_callbacks.zip](https://github.com/SciML/DifferentialEquations.jl/files/8783241/statprof_no_callbacks.zip) [statprof_with_callbacks.zip](https://github.com/SciML/DifferentialEquations.jl/files/8783242/statprof_with_callbacks.zip) Not knowing a better way of attaching the flamegraphs I've included the output directories here. Happy to package in some other manner. In this call the "with_callbacks" version...

Oops, yes I did flip them. I think that the `PresetTimeCallback` returns a `DiscreteCallback` which doesn't take `interp_points` as an argument. I'd presume they know exactly when to evaluate `u`...

In DiffEqBase `callbacks.jl` I see: ```julia #Base Case: Just one @inline function apply_discrete_callback!(integrator,callback::DiscreteCallback) saved_in_cb = false if callback.condition(integrator.u,integrator.t,integrator) # handle saveat _, savedexactly = savevalues!(integrator) saved_in_cb = true @inbounds if...

Thanks for the clarification! Doc update would indeed be helpful, though you could still consider typing the signal as an `Array{Complex}` to prevent amateurs such as myself getting confused.