Re: [Gretl-users] mysterious error discrepancy between two systems
On Wed, 14 Oct 2015, Sven Schreiber wrote:
Hi,
I'm puzzled by this error message that I'm getting from 'gretlcli' on a
high-performance computing server:
"
gretl_matrix_dot_op: Numerical result out of range
*** error in function otherfcstats, line 15
> denom = sqrt(mean(mfc.^2)) + sqrt(mean(my.^2))
"
The thing is, when I run exactly the same script(s), with the same seed to
the RNG, on my local gretl copy (run in the GUI, so gretlw32.exe), up to the
error message I get the same results on both systems, but on my local system
(Windows laptop) I don't get the error message and the script finishes
gracefully.
In both cases it's the released version 1.10.2. Of course, some background
libraries will be different, maybe that's the cause? What could I tell the
system admin to get rid of the error on the server?
I think this is probably the same issue that's discussed in
http://lists.wfu.edu/pipermail/gretl-users/2015-September/011139.html
I suspect that your code ends up squaring a very small number and
hence producing a "subnormal" floating-point value (underflow). As
noted in the post above, different C libraries can respond to this
differently: some will set the variable errno to ERANGE in this case
and some will not.
So I don't think there's anything the sysadmin can do to fix the
problem. However, we could "relax" the check in the relevant
gretl_matrix function(s), so that we don't necessarily flag an error
when the C library raises ERANGE.
Allin