2019 XCSP3 Competition
Rules and available instances
Detailed results
Preliminary list of descriptions
Resources allocated to solvers and other details
-
The cluster we used is provided
by CRIL and is
composed of nodes with two quad-cores Intel(R) Xeon(R) X5550 @ 2.67GHz with
32 GiB RAM. Hyperthreading was disabled.
-
Sequential solvers were run on one processor (4 cores) and were
allocated 15500 MiB of memory. Two instances of the same solver
were run simultaneously on one host.
- Parallel solvers were run on two processors (8 cores) and were
allocated 31000 MiB of memory.
- For Java programs, the -Xmx parameter was set to 12000 for
sequential solvers, and 24000 for parallel solvers. This parameter
was specified only when the author specified it on the command
line. These parameters allow to stay safely within the memory
limit.
-
In order to facilitate comparison, solvers which were run with the
same time limit and on a same set of instances were run in the same
event (group of tracks). Three events have been defined:
- Main tracks: CSP and COP, sequential and parallel solvers
with a 40 minutes time limit
- Fast COP tracks: COP, sequential and parallel solvers with a 4
minutes time limit
- Mini-solvers tracks: CSP and COP, sequential solvers with a
40 minutes time limit on a restricted set of instances
-
The time limit can be understood either as a CPU limit, or as a
wall-clock (WC) limit. Sequential solvers are best compared with a
CPU time limit. If it is assumed that CPU cores come for free
(which is quite a strong assumption), parallel solvers are best
compared with a WC time limit. Of course sequential and parallel
solvers which were run with the same limits can be compared, and
this can be quite informative.
In order to reach these two limits with a single run, solvers were
given a time limit augmented by 5% to give them a chance to reach
both limits. This was done only in tracks with both sequential and
parallel solvers. The results are post processed to enforce the
initial limit of the track and obtain the final ranking. The
limits given to solvers are summarized in the following table:
| Solver type | CPU limit | WC limit | Memory limit |
|---|
| sequential | track limit*1.05 | track limit*1.05 | 15500 MiB |
| parallel | track limit*nbCores*1.05 | track limit*1.05 | 31000 MiB |