
Hi Andrei,
Cantera will use multithreaded linear algebra routines if they are available when it is compiled. This is not a new feature, and has been around for many years.
The main place where this provides a performance benefit is in the reactor network solver, and perhaps to a more limited extent the 1D solver (the parallel efficiency of the banded solver used here is not great in some LAPACK implementations). You can control how many threads are used for most of these by setting the OMP_NUM_THREADS environment variable, or variables that are particular to the LAPACK implementation you’re using.
Regards,
Ray