Message5885

Author malte
Recipients andrew.coles, erez, florian, jendrik, malte, silvia
Date 2016-12-14.18:07:36
Content
Great! For this issue, I can think of two more TODO items:

- Rerun the BJOLP comparison with identical OSI/CPLEX versions. (Ideally we
should use the most recent versions for both, but IIRC, this doesn't work
because the newest CPLEX version has 32-bit issues. But perhaps we can still use
our recommended OSI version for both and couple it with an older CPLEX version
that still supports both architectures. Florian can hopefully advise on this.)

- Run experiments for satisficing configurations. If we do this later when we're
happy with the changes affecting optimal planners, we'll have to run fewer
experiments overall. If we do it earlier, we might be quicker to notice other
parts of the code we need to change before moving to 64-bit compiles. Both
options are fine for me.

Right now I can only think of one obvious change we will want to make for the
satisficing search configurations: replace the generating operator pointer in
lazy open lists with an operator ID, similar to the change for SearchNodeInfo.
We might then also want to change the interface so that it works with these IDs
and we don't have to convert back and forth between operator pointers and IDs
when we actually use IDs on both ends.
History
Date User Action Args
2016-12-14 18:07:36maltesetmessageid: <1481735256.99.0.701949141845.issue213@unibas.ch>
2016-12-14 18:07:36maltesetrecipients: + malte, erez, andrew.coles, silvia, jendrik, florian
2016-12-14 18:07:36maltelinkissue213 messages
2016-12-14 18:07:36maltecreate