gbd-tools 4.8.2

GBD Tools: Maintenance and Distribution of Benchmark Instances and their Attributes

GBD Benchmark Database

GBD Benchmark Database (GBD) is about bridging the gap between research on SAT algorithms and data science.

GBD has three interfaces

• Command-line interface gbd
• Web interface gbd serve
• Python interface gbd_core.api.GBD

GBD solves several problems

• benchmark instance identification and identification of equivalence classes of benchmark instances (gbdhash, isohash, instance family, ...)
• distribution of benchmark instances and benchmark meta-data
• a simple query language to provide access to filtered sets of benchmark instances and instance features
• initialization and maintenance of instance feature databases (meta.db, base.db, gate.db, ...)
• transformation algorithms for benchmark instances such as instance sanitization or transformation of cnf to k-isp instances
• keeping track of contexts (cnf, sanitized-cnf, k-isp, ...) and relations of instances between contexts

Programming Language

• Python 3
• SQLite

Installation

• pip install gbd-tools

Configuration

• fetch a database, e.g., https://benchmark-database.de/getdatabase/meta.db
• export GBD_DB=[path/to/database1]:[path/to/database2:..] (and put it in your .bashrc)
• test command-line interface with commands gbd info and gbd --help

GBD Python Interface

from gbd_core.api import GBD
with GBD(['path/to/database1', 'path/to/database2', ..] as gbd:
    df = gbd.query("family = hardware-bmc", resolve=['verified-result', 'runtime-kissat'])

GBD Server

This runs under https://benchmark-database.de/. The command is available in gbd-tools: gbd serve --help

GBD Command-Line Interface

gbd get

We assume https://benchmark-database.de/getdatabase/meta.db is in your gbd path GBD_DB.

Get list of benchmark instances in database:

gbd get

Get list of benchmark instances including some meta-data

gbd get -r result family

Filter for specific benchmark instances with gbd-query

gbd get "family = hardware-bmc" -r filename

gbd init

We assume you installed the python extension module gdbc.

All initialization routines can run in parallel and per-process ressource limits can be set. See gbd init --help for more info.

gbd init local

To initialize a database with local paths to your own benchmarks:

gbd -d my.db init local [path/to/benchmarks]

After that in my.db, the features local and filename exist and are associated with their corresponding gbd-hash:

gbd -d my.db get -r local filename

gbd init isohash

To identify isomorphic instances (approximately by the hash of the sorted degree-sequence of their graph representation):

gbd -d my.db init isohash

After that in my.db, instances can be grouped by their isohash:

gbd -d my.db get -r local filename -g isohash

gbd init base, gbd init gate

gbd -d my.db:mybase.db init --target mybase base

gbd -d my.db:mygate.db init --target mygate gate