These results are related to domains that use action costs. However,
the encodings currently don’t support this feature. For this reason,
these results were all incorrect and needed to be removed.
These results were affected by a bug in the planner (out-of-range index
access). This was fixed in the meantime, and these measurements will be
repeated now.
This removes all results obtained with our planner where memouts were
encountered. These were mainly due to the fact that we didn’t configure
the planner to respect the memory limit in the first place. With this
configuration now in place, we can redo all the affected measurements in
order to get actual, useful results.
Madagascar automatically manages its memory consumption in order not to
have memouts. The memouts that were recorded by the benchmark runner are
the result of a difference in comupting memory consumption between the
benchmark runner and Madagascar. For this reason, these results are not
representative for Madagascar, and they will be redone with a softer
memory limit (with an extra margin of 256 MB).
Due to minor bugs in the planner, three configurations had errors. This
removes the affected results in order to rerun them with the fixed
planner. As the bugfixes address these configurations only, the other
results are unspoiled by this change.
Some instances have hyphens within variable names. Unfortunately,
support for hyphens in variable names was missing in plasp 3.1.0, which
resulted in incorrect ASP output. Additionally, clingo just printed an
info message in the affected cases, which is why these incorrect
benchmark runs weren’t easy to spot immediately.
This removes all the affected results in order to run them again with
plasp 3.1.1. Note that the new release doesn’t change anything else, and
thus, the other results aren’t invalidated by the plasp update.
patrick