The 6th Syntax-Guided Synthesis Competition (SyGuS-Comp) will take place as a satellite event of CAV and SYNT 2019.
UPDATE: SyGuS-Comp 2019 has concluded (Results). Congratulations to all winners.
|13/14 July 2019||Solvers presentation (at SYNT’19 workshop)|
Call for Participation
This is a call for participation in the 6th Syntax-Guided Synthesis Competition to be organized as a satellite event of SYNT and CAV 2019.
The classical formulation of the program-synthesis problem is to find a program that meets a correctness specification given as a logical formula. Recent work on program synthesis and program optimization illustrates many potential benefits of allowing the user to supplement the logical specification with a syntactic template that constrains the space of allowed implementation. The motivation is twofold. First, narrowing the space of implementations makes the synthesis problem more tractable. Second, providing a specific syntax can potentially lead to better optimizations.
The input to the syntax-guided synthesis problem (SyGuS) consists of a background theory, a semantic correctness specification for the desired program given by a logical formula, and a syntactic set of candidate implementations given by a grammar. The computational problem then is to find an implementation from the set of candidate expressions that satisfies the specification in the given theory. The formulation of the problem builds on SMT-LIB.
There has been a lot of recent interest in both using SyGuS solvers for various synthesis applications and developing different solving algorithms. The SyGuS competition (SyGuS-Comp’19) will allow solvers to compete on a large collection of benchmarks and advance the state-of-the-art for program-synthesis tools. We invite authors to submit their SyGuS solvers to this year’s SyGuS Competition.
For questions regarding the competition please contact the organizers at email@example.com.
This year’s competition will have 5 tracks:
- General SyGuS track (General),
- Invariant synthesis track (Inv),
- Conditional Linear Integer Arithmetic track (CLIA),
- Programming By Examples [Theory of Strings] track (PBE-Strings), and
- Programming By Examples [Theory of Bit Vectors] track (PBE-BV).
Check the specification language for more details on the formulation of the problems for these tracks.
NOTE: This year’s SyGuS competition will use a lightly-modified version of the SyGuS input format that was used in previous competitions. The new format is more compliant with SMT-LIB version 2.6, includes minor changes to the concrete syntax for commands, and eliminates several deprecated features of the previous format.
A comprehensive description of the new input format and its differences with respect to the previous format is available in the reference document “The SyGuS Language Standard Version 2.0”.
We will evaluate the solvers on a subset of public benchmarks and some secret benchmarks. The domains of benchmarks include bit-vector manipulation, including bit-vector algorithms, concurrency, robotics, string transformations, invariant generation, program repair and cryptographic circuits. We are still finalizing the set of benchmarks, and would appreciate your contribution to the benchmarks as well.
Solvers will be evaluated on the StarExec platform, which provided 200 dual quad-core machines with 256GB memory each. The solvers would be run with a TIMEOUT value. The SyGuS-correctness checker, as well as the solvers from last year’s competition are available on the SyGuS community at StarExec. Candidate participants are invited to register on StarExec, where they can easily compare their solvers to the previous ones against the public benchmarks.
The scoring system is per track and as follows.
A solver that solved benchmarks in the track, out of which benchmarks among the fastest1, and benchmarks with an expression size among the smallest2 will receive a score: . The solver with highest score will be announced the winner of the track.
Tool Submission and Description
We expect tool developers to test their solvers on the public benchmarks, and submit the solver binaries on StarExec by the solver submission deadline. Each solver submission should also be accompanied by a brief (1 – 2 page in IEEE format) description of the key ideas of the solver.
Licensing of Tools and Benchmarks:
All benchmarks will be made public after the competition. We encourage the tool developers to make their solvers open-source, but participants are welcomed to submit binaries of proprietary tools as well.
Four solvers participated in SyGuS-Comp’19:
- CVC4 (in all 5 tracks)
Andrew Reynolds (University of Iowa), Haniel Barbosa (University of Iowa), Andres Noetzli (Stanford University), Clark Barrett (Stanford University) and Cesare Tinelli (University of Iowa)
- DryadSynth (in Inv and CLIA tracks)
KangJing Huang (Purdue University), Xiaokang Qiu (Purdue University), Peiyuan Shen (Purdue University) and Yanjun Wang (Purdue University)
- LoopInvGen (in Inv track)
Saswat Padhi (UCLA), Rahul Sharma (Microsoft Research, India) and Todd Millstein (UCLA)
- OASIS (in Inv track)
Sahil Bhatia (Microsoft Research, India), Saswat Padhi (UCLA), Nagarajan Natarajan (Microsoft Research, India), Prateek Jain (Microsoft Research, India) and Rahul Sharma (Microsoft Research, India)
The winners for various tracks are as follows.
- PBE-BV Track: CVC4
- PBE-Strings Track: CVC4
- Inv Track: CVC4
- CLIA Track: DryadSynth
- General Track: CVC4
Slides for final presentation of results are here.