Closed mezpusz closed 2 months ago
TPTP performance (-i 2000
) seems OK:
Sort by UNS 7341 ['problemsSTD_master8465_lpo_i2000.pkl'] 7344 ['problemsSTD_marton8471_lpo_i2000.pkl'] 7510 ['problemsSTD_master8465_i2000.pkl'] 7514 ['problemsSTD_marton8471_i2000.pkl']
(Of course, 4 problems are typically less than the variance of the used lrs
, but is seems good enough.)
And here the dpc
version of this:
Sort by UNS 7355 ['problemsSTD_master8465_lpo_dpc_i2000.pkl'] 7360 ['problemsSTD_marton8471_lpo_dpc_i2000.pkl'] 7509 ['problemsSTD_master8465_dpc_i2000.pkl'] 7511 ['problemsSTD_marton8471_dpc_i2000.pkl']
This PR fixes some annoying things in
OrderingComparator
and subclasses:KBOComparator
uses info fromKBO::State
to duplicate less effort and code.OrderingComparator
is created but not called byOrdering
anymore, this allows more flexibility.OrderingComparator
has now a default implementation.LPOComparator
has been simplified and now it is created lazily on the fly, as eager creation of the entire tree was a bottleneck in some cases. This is not justified forKBOComparator
yet, so I did not bother doing it now.