From Don Smith:

Accomplishments:

By adding several optimizations, I got proverD.pl (focussing + Dyckhoff)
to run two to three orders of magnitude faster on one example.  And it
now runs the example suite (goD) in 0.65 seconds (versus several minutes
before).

The most significant optimization was to AVOID ADDING DUPLICATE
ASSUMPTIONS to the list of synchronous assumptions.  To accomplish this,
I carry around an additional, standardized copy of the synchronous
assumptions, and I check whether a new assumption is already in there.
This check fires often.

As a second optimization, I reordered some calls and added a few cuts.

As a final optimization, I replaced calls to "call_with_depth_limit/3"
with an explicit depth parameter, which I decrement only when switching
from syncrhonous to asynchronous mode, or from left to right, etc.This
turned out to be faster.

By the way, with minimal effort I got the code running under gprolog
(GNU Prolog, which compiles to native code).  Some examples are faster
compiled, some aren't.


From Gosia and Darek,

We are writing to let you know about what we have done so far for the project.
Our goal was to develop a tableaux prover for first-order intuitionistic logic.
As a starting point, we implemented (in Prolog) a labeled tableaux prover for
propositional intuitionistic logic, based on Dyckhoff's calculus. To this end,
we augmented Dyckoff's rules  for implication on the left with labels, derived
from standard rules.
Next, we extended the prover to the first-order case, by adding standard rules
handling quantifiers, and a special case of the rule for implication on the
left, where the predecessor of the implication is a quantified formula.
Now we are testing and optimizing the prover, and comparing its performance
with
Don's provers. We are also going to look at the labeled sequent calculus.