Domagoj Babic, Zvonimir Rakamaric. 2013 IFIP Joint International Conference on Formal Techniques for Distributed Systems (33rd FORTE/15th FMOODS), Florence, Italy.
[pdf] [bib]

Abstract: We introduce an automata-based formal model suitable for specifying, modeling, analyzing, and verifying asynchronous task-based and message-passing programs. Our model consists of visibly pushdown automata communicating over unbounded reliable point-to-point first-in-first-out queues. Such a combination unifies two branches of research, one focused on task-based models, and the other on models of message-passing programs. Our model generalizes previously proposed models that have decidable reachability in several ways. Unlike task-based models of asynchronous programs, our model allows sending and receiving of messages even when stacks are not empty, without imposing restrictions on the number of context-switches or communication topology. Our model also generalizes the well-known communicating finite-state machines with recognizable channel property allowing (1) individual components to be visibly pushdown automata, which are more suitable for modeling (possibly recursive) programs, (2) the set of words (i.e., languages) of messages on queues to form a visibly pushdown language, which permits modeling of remote procedure calls and simple forms of counting, and (3) the relations formed by tuples of such languages to be synchronized, which permits modeling of complex interactions among processes. In spite of these generalizations, we prove that the composite configuration and control-state reachability are still decidable for our model.

Bibtex:

@inproceedings{forte2013-br,
  author = {Domagoj Babi\'c and Zvonimir Rakamari\'c},
  title = {Asynchronously Communicating Visibly Pushdown Systems},
  booktitle = {Proceedings of the 2013 IFIP Joint International
    Conference on Formal Techniques for Distributed Systems
    (33rd FORTE/15th FMOODS)},
  year = {2013},
  note = {to appear},
}

Wei-Fan Chiang, Ganesh Gopalakrishnan, Guodong Li, Zvonimir Rakamaric. 5th NASA Formal Methods Symposium (NFM 2013), NASA Ames Research Center, Moffett Field, CA, USA.
[pdf] [bib]

Abstract: GPU based computing has made significant strides in recent years. Unfortunately, GPU program optimizations can introduce subtle concurrency errors, and so incisive formal bug-hunting methods are essential. This paper presents a new formal bug-hunting method for GPU programs that combine barriers and atomics. We present an algorithm called Conflict-directed Delay-bounded scheduling algorithm (CD) that exploits the occurrence of conflicts among atomic synchronization commands to trigger the generation of alternate schedules; these alternate schedules are executed in a delay-bounded manner. We formally describe CD, and present two correctness checking methods, one based on final state comparison, and the other on user assertions. We evaluate our implementation on realistic GPU benchmarks, with encouraging results.

Bibtex:

@inproceedings{nfm2013-cglr,
  author = {Wei-Fan Chiang and Ganesh Gopalakrishnan and Guodong Li
    and Zvonimir Rakamari\'c},
  title = {Formal Analysis of GPU Programs with Atomics via
    Conflict-Directed Delay-Bounding},
  booktitle = {Proceedings of the 5th NASA Formal Methods
    Symposium (NFM 2013)},
  year = {2013},
  note = {to appear},
}

Diego Caminha B. de Oliveira, Zvonimir Rakamaric, Ganesh Gopalakrishnan, Alan Humphrey, Qingyu Meng, Martin Berzins. 5th International Workshop on Software Engineering for Computational Science and Engineering (SE-CSE 2013), San Francisco, CA, USA.
[pdf] [bib]

Abstract: While formal correctness checking methods have been deployed at scale in a number of important practical domains, we believe that such an experiment has yet to occur in the domain of high performance computing at the scale of a million CPU cores. This paper presents preliminary results from the Uintah Runtime Verification (URV) project that has been launched with this objective. Uintah is an asynchronous task-graph based problem-solving environment that has shown promising results on problems as diverse as fluid-structure interaction and turbulent combustion at well over 200K cores to date. Uintah has been tested on leading platforms such as Kraken, Keenland, and Titan consisting of multicore CPUs and GPUs, incorporates several innovative design features, and is following a roadmap for development well into the million core regime. The main results from the URV project to date are crystallized in two observations: (1) A diverse array of well-known ideas from light-weight formal methods and testing/observing HPC systems at scale have an excellent chance of succeeding. The real challenges are in finding out exactly which combinations of ideas to deploy, and where. (2) Large-scale problem solving environments for HPC must be designed such that they can be “crashed early” (at smaller scales of deployment) and “crashed often” (have effective ways of input generation and schedule perturbation that cause vulnerabilities to be attacked with higher probability). Furthermore, following each crash, one must “explain well” (given the extremely obscure ways in which an error finally manifests itself, we must develop ways to record information leading up to the crash in informative ways, to minimize off-site debugging burden). Our plans to achieve these goals and to measure our success are described. We also highlight some of the broadly applicable concepts and approaches.

Bibtex:

@inproceedings{secse2013-orghmb,
  author = {Diego Caminha B. de Oliveira and Zvonimir Rakamari\'c
    and Ganesh Gopalakrishnan and Alan Humphrey and Qingyu Meng
    and Martin Berzins},
  title = {Practical Formal Correctness Checking of Million-core
    Problem Solving Environments for HPC},
  booktitle = {Proceedings of the 5th International Workshop on
    Software Engineering for Computational Science and Engineering
    (SE-CSE 2013)},
  year = {2013},
  note = {to appear},
}

Wei-Fan Chiang, Ganesh Gopalakrishnan, Zvonimir Rakamaric, Dong H. Ahn, Gregory L. Lee. 4th Workshop on Determinism and Correctness in Parallel Programming (WoDet 2013), Houston, TX, USA.
[pdf] [bib]

Abstract: The ability to reproduce simulation results (external determinism) goes a long way towards enhancing the trustworthiness of high performance computing simulations. The ability to replay schedules (internal determinism) greatly facilitates reproducing bugs, and helps reduce wasted programmer productivity. In this paper, we consider these issues in the context of software libraries and APIs used in today’s mainstream high performance computing (HPC) systems as well as expected to be used in upcoming high-end systems. After cataloging the main sources of external and internal nondeterminism, we summarize two thrusts in our current research: (1) mechanisms to control internal nondeterminism by active schedule control, and (2) techniques that may help assess the extent of result nondeterminism in floating point calculations.

Bibtex:

@inproceedings{wodet2013-cgral,
  author = {Wei-Fan Chiang and Ganesh Gopalakrishnan and
    Zvonimir Rakamari\'c and Dong H. Ahn and Gregory L. Lee},
  title = {Determinism and Reproducibility in Large-Scale
    HPC Systems},
  booktitle = {Informal Proceedings of the 4th Workshop on
    Determinism and Correctness in Parallel Programming
    (WoDet 2013)},
  year = {2013},
}

Dimitra Giannakopoulou, Zvonimir Rakamaric, Vishwanath Raman. 19th International Static Analysis Symposium (SAS 2012), Deauville, France.
[pdf] [bib]

Abstract: Given a white-box component C with specified unsafe states, we address the problem of automatically generating an interface that captures safe orderings of invocations of C’s public methods. Method calls in the generated interface are guarded by constraints on their parameters. Unlike previous work, these constraints are generated automatically through an iterative refinement process. Our technique, named PSYCO (Predicate-based SYmbolic COmpositional reasoning), employs a novel combination of the L* automata learning algorithm with symbolic execution. The generated interfaces are three-valued, capturing whether a sequence of method invocations is safe, unsafe, or its effect on the component state is unresolved by the symbolic execution engine. We have implemented PSYCO as a new prototype tool in the JPF open-source software model checking platform, and we have successfully applied it to several examples.

Bibtex:

@inproceedings{sas2012-grr,
  author = {Dimitra Giannakopoulou and Zvonimir Rakamari\'c and
    Vishwanath Raman},
  title = {Symbolic Learning of Component Interfaces},
  booktitle = {Proceedings of the 19th International Static Analysis
    Symposium (SAS 2012)},
  series = {Lecture Notes in Computer Science},
  volume = {7460},
  publisher = {Springer},
  editor = {Antoine Min\'e and David Schmidt},
  year = {2012},
  pages = {248--264},
}

Domagoj Babic, Byron Cook, Alan J. Hu, Zvonimir Rakamaric. Formal Aspects of Computing (FAC), 2012. (Invited.)
[pdf] [bib]

Abstract: We describe a simple and efficient algorithm for proving the termination of a class of loops with nonlinear assignments to variables. The method is based on divergence testing for each variable in the cone-of-influence of the loop’s condition. The analysis allows us to automatically prove the termination of loops that cannot be handled using previous techniques. We also describe a method for integrating our nonlinear termination proving technique into a larger termination proving framework that depends on linear reasoning.

Bibtex:

@article{fac2012-bchr,
  author = {Domagoj Babi\'c and Byron Cook and Alan J. Hu and
    Zvonimir Rakamari\'c},
  title = {Proving Termination of Nonlinear Command Sequences},
  journal = {Formal Aspects of Computing},
  year = {2012},
  pages = {1--15},
  publisher = {Springer-Verlag},
  issn = {0934-5043}
}

Domagoj Babic, Zvonimir Rakamaric. EECS Department, University of California, Berkeley Tech Report UCB/EECS-2011-108, October, 2011.
[pdf] [bib]

Abstract: We introduce an automata-based formal model suitable for specifying, modeling, analyzing, and verifying asynchronous task-based and message-passing programs. Our model consists of visibly pushdown automata communicating over unbounded reliable point-to-point first-in-first-out queues. Such a combination unifies two branches of research, one focused on task-based models, and the other on models of message-passing programs. Our model generalizes previously proposed models that have decidable reachability in several ways. Unlike task-based models of asynchronous programs, our model allows sending and receiving of messages even when stacks are not empty, without imposing restrictions on the number of context-switches or communication topology. Our model is strictly more general than the well-known communicating finite-state machines and allows: (1) individual components to be visibly pushdown automata, which are more suitable for modeling (possibly recursive) programs, (2) the set of words (i.e., languages) of messages on queues to form a visibly pushdown language, which permits modeling of remote procedure calls and simple forms of counting, and (3) the relations formed by tuples of such languages to be synchronized, which permits modeling of complex interactions among processes. In spite of these generalizations, we prove that the composite configuration and control-state reachability are still decidable for our model.

Bibtex:

@techreport{ucb-eecs-2011-108,
  author = {Domagoj Babi\'c and Zvonimir Rakamari\'c},
  title = {Asynchronously Communicating Visibly Pushdown Systems},
  institution = {EECS Department, University of California,
    Berkeley},
  year = {2011},
  month = {October},
  number = {UCB/EECS-2011-108},
}

I successfully defended and wrapped up my PhD! Yaaaay!

Big thanks goes to my advisor Alan Hu, committee members Mark Greenstreet and Ken McMillan, university examiners Gail Murphy and Karthik Pattabiraman, as well as my external examiner Ranjit Jhala. The final version of my PhD thesis is available online and can be found here.

Soon after I wrapped up my PhD, I joined Carnegie Mellon University in Silicon Valley as a postdoctoral fellow. Here I am working on a NASA-funded project and collaborating with Dimitra Giannakopoulou and Vishwanath Raman. The goal of the project is improving coverage of testing and checking of NASA’s air traffic control system using compositional techniques. I am sure that lots of challenging and exciting work is ahead of me!

Zvonimir Rakamaric. Ph.D. Thesis, Department of Computer Science, The University of British Columbia, March, 2011.
[pdf] [bib]

Abstract: Software is large, complex, and error-prone. According to the US National Institute of Standards and Technology, software bugs cost the US economy an estimated $60 billion each year. The trend in hardware design of switching to multi-core architectures makes software development even more complex. Cutting software development costs and ensuring higher reliability of software is of global interest and a grand challenge. This is especially true of the system software that is the foundation beneath all general-purpose application programs.
The verification of system software poses particular challenges: system software is typically written in a low-level programming language with dynamic memory allocation and pointer manipulation, and system software is also highly concurrent, with shared-memory communication being the main concurrent programming paradigm. Available verification tools usually perform poorly when dealing with the aforementioned challenges. This thesis addresses these problems by enabling precise and scalable verification of low-level, shared-memory, concurrent programs. The main contributions are about the interrelated concepts of memory, modularity, and concurrency.
First, because programs use huge amounts of memory, the memory is usually modeled very imprecisely in order to scale to big programs. This imprecise modeling renders most tools almost useless in the memory-intensive parts of code. This thesis describes a scalable, yet precise, memory model that offers on-demand precision only when necessary.
Second, modularity is the key to scalability, but it often comes with a price — a user must manually provide module specifications, making the verification process more tedious. This thesis proposes a light-weight technique for automatically inferring an important family of specifications to make the verification process more automatic.
Third, the number of program behaviors explodes in the presence of concurrency, thereby greatly increasing the complexity of the verification task. This explosion is especially true of shared-memory concurrency. The thesis presents a static context-bounded analysis that combines a number of techniques to successfully solve this problem.
We have implemented the above contributions in the verification tools developed as a part of this thesis. We have applied the tools on real-life system software, and we are already finding critical, previously undiscovered bugs.

Bibtex:

@phdthesis{rakamaric-phd-11,
  author = {Zvonimir Rakamari\'c},
  title = {Modular Verification of Shared-Memory Concurrent
    System Software},
  school = {The University of British Columbia},
  year = {2011},
  month = {March}
}

Michael Emmi, Shaz Qadeer, Zvonimir Rakamaric. 38th ACM SIGPLAN-SIGACT Symposium on Principles of Programming Languages (POPL 2011), Austin, TX, USA.
[pdf] [bib]

Abstract: We provide a new characterization of scheduling nondeterminism by allowing deterministic schedulers to delay their next-scheduled task. In limiting the delays an otherwise-deterministic scheduler is allowed, we discover concurrency bugs efficiently—by exploring few schedules—and robustly—i.e., independent of the number of tasks, context switches, or buffered events. Our characterization elegantly applies to any systematic exploration (e.g., testing, model checking) of concurrent programs with dynamic task-creation. Additionally, we show that certain delaying schedulers admit efficient reductions from concurrent to sequential program analysis.

Bibtex:

@inproceedings{popl2011-eqr,
  author = {Michael Emmi and Shaz Qadeer and Zvonimir Rakamari\'c},
  title = {Delay-Bounded Scheduling},
  booktitle = {Proceedings of the 38th ACM SIGPLAN-SIGACT
    Symposium on Principles of Programming Languages (POPL 2011)},
  publisher = {ACM},
  editor = {Thomas Ball and Mooly Sagiv},
  year = {2011},
  pages = {411--422},
}