WoDet 3
WoDet 3: Third Workshop on Determinism and Correctness in Parallel Programming
March 4, 2012
Co-located with ASPLOS 12, London, England, UK 
Unintentional non-determinism is the bane of multithreaded software development. Defective software might execute correctly hundreds of times before a subtle synchronization bug appears, and when it does, developers often cannot readily reproduce it while debugging. Nondeterminism also complicates testing as good coverage requires both a wide range of program inputs and a large number of possible interleavings for each input. These problems have taken on renewed urgency as multicore systems have driven parallel programming to become mainstream.
Determinism is emerging as an important research area, ranging from techniques for existing code (including deterministic execution models, parallelizing compilers, and deterministic replay for debugging) to new programming models (including deterministic general purpose languages and run-time systems). Deterministic multiprocessing yields deep open questions in programming languages, compilers, operating systems, runtime systems and architecture.
While there is a growing consensus that determinism would greatly help with the programmability challenges of multicore systems, there is still little consensus on many important questions. What are the performance and programmability trade-offs for enforcing deterministic semantics with different approaches? Should deterministic semantics be strictly enforced or guaranteed only for programs that are "well-behaved" in certain ways? How can we support truly non-deterministic algorithms, where non-determinism is intentionally used for improved parallel performance? How can each layer of the system stack contribute to these goals? What are other safety guarantees useful in making parallel programming easier and less error prone (e.g., race-freedom, atomicity, etc..)?
The Third Workshop on Determinism and Correctness in Parallel Programming is an across-the-stack forum to discuss the role of a wide range of correctness properties in parallel and concurrent programming. While determinism is an important theme, the scope of the workshop includes other correctness properties for parallel programs and systems. The workshop will be a full day event with a few invited talks, a moderated debate, and technical sessions for short peer-reviewed papers discussing ideas, positions, or preliminary research results.
In addition to answers to the questions above, topics of interest include:
- Language extensions for disciplined parallel programming models (deterministic, data race-free, etc.)
- Architecture, operating system, runtime system and compiler support for parallel program correctness
- Concurrency debugging techniques
- New properties of parallel programs
- Limit studies and empirical studies of the cost of safety properties
- Studies of the applicability of correctness properties in parallel programs and algorithms
- Concurrency bug avoidance techniques
- Real-world experience with safe parallel programming models, systems, or tools
Program
Accepted papers
The Case For Merging Execution- and Language-level Determinism with MELD, J. Devietti, D. Grossman, L. CezeAutomatic Empirical Failure Avoidance for Concurrent Software, B. Lucia, L. Ceze
KUDA: GPU Accelerated Split Race Checker, C. Bekar, T. Elmas, S. Okur, S. Tasiran
On Justifying and Verifying Relaxed Detection of Conflicts in Concurrent Programs, O. Subasi, T. Elmas, S. Tasiran
Tasks with Effects: A Model for Disciplined Concurrent Programming, S. Heumann, V. Adve
A Generalized Reduction Construct for Deterministic OpenMP, A. Aviram, B. Ford
All papers in one ZIP file
Organizer
Emery Berger, University of Massachusetts Amherst
Program Committee
Vikram Adve, University of IllinoisEmery Berger, University of Massachusetts Amherst
Luis Ceze, University of Washington
Jason Flinn, University of Michigan
Bryan Ford, Yale University
Suresh Jagannathan, Purdue University
Shan Lu, University of Wisconsin
Madan Musuvathi, Microsoft Research (Redmond)
Simon Peyton Jones, Microsoft Research (Cambridge)
Koushik Sen, University of California, Berkeley
Martin Vechev, ETH Zurich
Eran Yahav, Technion
Junfeng Yang, Columbia University