[R-sig-dyn-mod] deSolve with compiled code and varying size of parms

[Tim Keitt]

On Wed, Mar 29, 2017 at 3:23 PM, Rampal Etienne <rampaletienne at gmail.com>
wrote:

> to set an evaluation step of 2.5, then RStudio crashes. With other step
> sizes (0.1, 0.5, 1, 1.5, 2) everything worked fine.


You're right.

I believe it is a "bug" in ODEINT (traced it in lldb). Unlike more
integrated solvers, that framework puts more onus on the user to not supply
invalid parameters, or perhaps it would be better to say that the user has
to avoid certain corner cases.

Thanks for finding that corner case. I have added guards to the odeintr
code that adjusts the step size with a warning. I will also file and issue
with ODEINT and see what they think.

You can pull the github version with:
devtools::install_github("thk686/odeintr")

For your case, you might experiment with the openmp bits. It
auto-parallelizes for loops quite nicely.

THK

http://www.keittlab.org/

	[[alternative HTML version deleted]]