Underflow issue with Beast v1.7.2... not with v1.7.1

[julien]

Hi there...

Just wanted to mention about an underflow "issue" I have while using
Beast v1.7.2 (with Beagle CPU, precision double, scaling always) ...

90% lines on the screen are "Underflow calculating like likelihood...
attempting a rescaling"... Yet my analyses run until the end and
finally reach a very decent result...

The most surprising is that when using Beast v1.7.1 (same file & same
command line) the proportion of rescaling attempts is very very
small...

That issue, I guess, is related to the complexity of the model I use
(Yule, 5 genes and independant clocks, 11 partitions for substitution
models + 4 characters under discrete diffusion model + one character
under continuous diffusion)...

But I was wondering about that difference between v1.7.1 and v1.7.2...
as, at a first glance, I don't see anything in the last fixes that
could explain such a difference of behavior..

Regards

[Andrew Rambaut]

This is a reporting issue only - the line that logs the underflow condition somehow moved outside the conditional block so the message is logged whether a rescaling happens or not. To hide the messages pipe the stderr to /dev/null

It is also possible to turn off logging of certain messages at runtime but I will have to remind myself of how to do that.


Andrew

> --
> You received this message because you are subscribed to the Google Groups "beast-users" group.
> To post to this group, send email to beast...@googlegroups.com.
> To unsubscribe from this group, send email to beast-users...@googlegroups.com.
> For more options, visit this group at http://groups.google.com/group/beast-users?hl=en.
>