Re: breaking isis when fitting

On Sep 18, 2007, at 4:56 PM, rgibson_at_email.domain.hidden
> Sometimes ISIS spins for a long time while fitting data, e.g., using 
> the subplex
> method.  I'm sure it's all doing fine, but I want to break it and stop 
> the fit
> process.  Unfortunately, CTRL-C doesn't work.  (I assume it's down in 
> some C
> library that is not interested in CTRL-C?)

Really?  ^C works for me.  Sometimes, for a particularly nasty piece of 
model code, it takes a while for the ^C to register (I believe it's 
registering in the ISIS layer, after it comes back from the Fortran or 
C-code that is the model).  If the ^C is not getting you out of the 
fit, the problem is probably in the model function, not the ISIS layer. 
  (Although perhaps John knows of some odd way you might have built this 
that prevents you from using ^C to break out of a git.)

On a side note, I will say that the strength of (and problem with) 
subplex is that it has a way of getting you into funny bits of 
parameter space that you might otherwise not stumble across.  Thus, for 
some model codes, where the authors haven't put in safety features to 
keep one from crashing it, I simply can't use subplex.  The eqpair 
Comptonization model is a good example.  Subplex will crash that sucker 
99/100 times.  Perhaps for whatever model you are using, subplex has 
gotten you into areas where the model is hitting very long (or infinit) 
model evaluation loops, and it never gets you back into the ISIS layer 
for the ^C to register.

-Mike


----
You received this message because you are
subscribed to the isis-users list.
To unsubscribe, send a message to
isis-users-request_at_email.domain.hiddenwith the first line of the message as:
unsubscribe