[Gretl-users] Re: Example code of 'eigensym' and way faster 'eiggen2'
Am 12.01.2020 um 22:44 schrieb Allin Cottrell:
On Sun, 12 Jan 2020, Sven Schreiber wrote:
> The problem does not go away, ratios stay the same.
I did some testing on Windows 10 today, and I saw the same: eigen() a
lot faster than eigensym(), with the update to openblas 0.3.7 not making
an appreciable difference.
I also found: (a) disabling the symmetry test in eigensym made
essentially no difference; (b) restricting the number of OMP threads to
the number of real cores didn't change the ordering;
...
though nothing like as much as on Windows. I also saw that both
dsyev
and dgeev appear to do multi-threading.
Thanks for the thorough testing. Perhaps it would be interesting to also
restrict the number of OMP threads to 1 (instead of ncores)?
(I don't know how to do that on Windows, except by wrapping the whole
thing in a dummy mpi block with np=1 and omp-threads=1.)
Pending a proper understanding of what's going on. maybe we
should make
eigensym() divert to eigen() on Windows for order less than 90 or so.
Certainly an option.
thanks
sven