Efficient System-Enforced Deterministic Parallelism

arXiv: Operating Systems, 2010

The difficulty of developing reliable parallel software is generating interest in deterministic environments, where a given program and input can yield only one possible result. Languages or type systems can enforce determinism in new code, and runtime systems can impose synthetic schedules on legacy parallel code. To parallelize existing serial code, however, we would like a programming model that is naturally deterministic without language restrictions or artificial scheduling. We propose deterministic consistency, a parallel programming model as easy to understand as the “parallel assignment” construct in sequential languages such as Perl and JavaScript, where concurrent threads always read their inputs before writing shared outputs. DC supports common data- and task-parallel synchronization abstractions such as fork/join and barriers, as well as non-hierarchical structures such as producer/consumer pipelines and futures. A preliminary prototype suggests that softwareonly impleme...

Sigplan Notices, 2012

Providing deterministic execution significantly simplifies the debugging, testing, replication, and deployment of multithreaded programs. Recent work has developed deterministic multiprocessor architectures as well as compiler and runtime systems that enforce determinism in current hardware. Such work has incidentally imposed strong memory-ordering properties. Historically, memory ordering has been relaxed in favor of higher performance in shared memory multiprocessors and, interestingly, determinism

Today's parallel architectures provide an execution model that does not guarantee deterministic results for parallel pro-grams. This nondeterminism causes significant challenges for parallel programmers by making it difficult to write par-allel applications with repeatable results. As a consequence, parallel applications are significantly harder to debug, test, and maintain than their sequential predecessors. Deterministic multithreading, which provides an always-on deterministic execution model for parallel applications, has emerged as a potential solution to this problem. In pre-vious work, we introduced the Kendo deterministic multi-threading framework. Kendo provides deterministic multi-threading by enforcing a deterministic ordering of synchro-nization operations, which results in deterministic execution for programs that are correctly synchronized. Our prior re-sults showed that Kendo incurs only minimal overheads on programs running with a small number of threads. However...

Parallel & Distributed Processing …, 2010

Multicore shared-memory architectures are becoming prevalent but bring many programming challenges. Among the biggest is non-determinism: the output of the program does not depend merely on the input, but also on scheduling choices taken by the operating system.

2012

When lock-based parallel programs execute on conventional multicore hardware, faulty software can cause hard-to-debug race conditions in critical sections that violate the contract between locks and their protected shared variables. This paper proposes new hardware support for enforcing isolation of critical section execution. It can detect and tolerate races, allowing programs to execute race-free. Our hardware scheme targets the existing large code base of lockedbased parallel programs written in type unsafe languages such as C and C++. Our approach works directly on unmodified executables. An evaluation of 13 programs from the SPLASH2 and PARSEC suites shows that the cost of the additional hardware and the impact on the overall execution time is minimal for these applications. Our mechanism is complementary to hardware transactional memory in that it uses similar structures but focuses on enhancing the reliability of existing lock-based programs.

2nd Workshop on Determinism and Correctness in Parallel Programming (WoDet), 2011

The difficulty of developing reliable parallel software is generating interest in deterministic environments, where a given program and input can yield only one possible result. Languages or type systems can enforce determinism in new code, and runtime systems can impose synthetic schedules on legacy parallel code. To parallelize existing serial code, however, we would like a programming model that is naturally deterministic without language restrictions or artificial scheduling. We propose "deterministic consistency", a parallel programming model as easy to understand as the "parallel assignment" construct in sequential languages such as Perl and JavaScript, where concurrent threads always read their inputs before writing shared outputs. DC supports common data- and task-parallel synchronization abstractions such as fork/join and barriers, as well as non-hierarchical structures such as producer/consumer pipelines and futures. A preliminary prototype suggests that software-only implementations of DC can run applications written for popular parallel environments such as OpenMP with low (<10%) overhead for some applications.

Operating Systems Design and Implementation, 2010

Current multiprocessor systems execute parallel and concurrent software nondeterministically: even when given precisely the same input, two executions of the same program may produce different output. This severely complicates debugging, testing, and automatic replication for fault-tolerance. Previous efforts to address this issue have focused primarily on record and replay, but making execution actually deterministic would address the problem at the root. Our goals in this work are twofold: (1) to provide fully deterministic execution of arbitrary, unmodified, multithreaded programs as an OS service; and (2) to make all sources of intentional nondeterminism, such as network I/O, be explicit and controllable. To this end we propose a new OS abstraction, the Deterministic Process Group (DPG). All communication between threads and processes internal to a DPG happens deterministically, including implicit communication via sharedmemory accesses, as well as communication via OS channels such as pipes, signals, and the filesystem. To deal with fundamentally nondeterministic external events, our abstraction includes the shim layer, a programmable interface that interposes on all interaction between a DPG and the external world, making determinism useful even for reactive applications. We implemented the DPG abstraction as an extension to Linux and demonstrate its benefits with three use cases: plain deterministic execution; replicated execution; and record and replay by logging just external input. We evaluated our implementation on both parallel and reactive workloads, including Apache, Chromium, and PARSEC.

IEEE International Symposium on High-Performance Comp Architecture, 2012

Recent proposals for determinism-enforcement architectures are able to honor the dependences between threads through a commit step that often becomes a performance bottleneck. As they commit code blocks (or chunks) in a round-robin order, if one chunk gets squashed due to a conflict, its successors also observe a stall. We call this effect transitive squash delay.

Conference on Object-Oriented Programming Systems, Languages, and Applications, 2000

The shift from single to multiple core architectures means that, in order to increase application performance, programmers must write concurrent, multithreaded programs. Unfortunately, multithreaded applications are susceptible to numerous errors, including dead- locks, race conditions, atomicity violations, and order violations. These errors are notoriously difficult for programmers to debug. We present Grace, a runtime system for multithreaded programs written in

Nondeterminism is a key contributor to the difficulty of parallel programming. Many research projects have shown how to provide deterministic parallelism, but with unfortunate trade-offs. Deterministic execution enforces determinism for arbitrary programs but with significant runtime cost, while deterministic languages enforce determinism statically (without runtime overhead) but only for fork-join programs expressible in their static type systems. MELD unifies these approaches. We explain the requirements for soundly integrating a deterministic language into a deterministic execution system, and describe a simple qualifier-based type checker that ensures isolation for code written in a deterministic language. We also extend MELD to incorporate nondeterministic operations without compromising the determinism of the rest of the program. Our experiments with benchmarks from the SPLASH2 and PAR-SEC suites show that a small number of annotations can accelerate the performance of deterministic versions of these programs by 2-6x.