A scalable deadlock detection algorithm for UPC collective operations
نویسندگان
چکیده
Unified Parallel C (UPC) is a language used to write parallel programs for shared and distributed memory parallel computers. Deadlock detection in UPC programs requires detecting deadlocks that involve either locks, collective operations, or both. In this paper, a distributed deadlock detection algorithm for UPC programs that uses run-time analysis is presented. The algorithm detects deadlocks in collective operations using a distributed technique with O(1) run-time complexity. The correctness and optimality of the algorithm is proven. For completeness, the algorithm is extended to detect deadlocks involving both locks and collective operations by identifying insolvable dependency chains and cycles in a shared wait-for-graph (WFG). The algorithm is implemented in the run-time error detection tool UPC-CHECK and tested with over 150 functionality test cases. The scalability of this deadlock detection algorithm for UPC collective operations is experimentally verified using up to 8192 threads. Keywords-deadlock, collective, verification, Partitioned Global Address Space (PGAS), Unified Parallel C(UPC).
منابع مشابه
UPC Collective Optimization
Efficient collective operations are an important feature for new parallel programming languages. The size of massively parallel machines is increasing fast and the performance of these operations are crucial to the scalability of many applications. The Unified Parallel C (UPC) language is one of the new parallel programming languages based on the Partitioned Global Address Space (PGAS) model. T...
متن کاملA Scalable Non-blocking Multicast Scheme for Distributed DAG Scheduling
This paper presents an application-level non-blocking multicast scheme for dynamic DAG scheduling on large-scale distributedmemory systems. The multicast scheme takes into account both network topology and space requirement of routing tables to achieve scalability. Specifically, we prove that the scheme is deadlock-free and takes at most logN steps to complete. The routing table chooses appropr...
متن کاملMultidestination Message Passing in Wormhole k-ary n-cube Networks with Base Routing Conformed Paths
This paper proposes multidestination message passing on wormhole k-ary n-cube networks using a new base-routing-conformed-path (BRCP) model. This model allows both unicast (single-destination) and multidestination messages to co-exist in a given network without leading to deadlock. The model is illustrated with several common routing schemes (deterministic, as well as adaptive), and the associa...
متن کاملDesign of Scalable PGAS Collectives for NUMA and Manycore Systems
The increasing number of cores per processor is turning multicore-based systems in pervasive. This involves dealing with multiple levels of memory in NUMA systems, accessible via complex interconnects in order to dispatch the increasing amount of data required. The key for efficient and scalable provision of data is the use of collective communication operations that minimize the impact of bott...
متن کاملUPC-SPIN: A Framework for the Model Checking of UPC Programs
This paper presents a method supported by a software framework for the model checking of Unified Parallel C (UPC) programs. The proposed framework includes a front-end compiler that generates finite models of UPC programs in the modeling language of the SPIN model checker. The model generation is based on a set of sound abstraction rules that transform the UPC synchronization primitives to sema...
متن کامل