Parallel Java: A Unified API for Shared Memory and Cluster Parallel Programming in 100% Java

Kaminsky, Alan

doi:10.1109/ipdps.2007.370421

Cited by 46 publications

(40 citation statements)

References 9 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Several benchmarks are from the Java Grande Forum (JGF) benchmark suite [29] and the Parallel Java (PJ) Library [20]. For our benchmarks, we use the same focus variable and parallel region annotations as in [4].…”

Section: Methodsmentioning

confidence: 99%

NDetermin: Inferring Nondeterministic Sequential Specifications for Parallelism Correctness

Burnim¹,

Elmas²,

Necula³

et al. 2011

View full text Add to dashboard Cite

A key reason for the great difficulty of writing, testing, and verifying parallel programs is the need to reason simultaneously about not only the sequential correctness of each part of a program in isolation, but also about all possible nondeterministic interleavings of the program's parallel threads. Thus, there has been much interest in techniques for separately testing or verifying the correctness of a program's use of parallelism, allowing the program's functional correctness to be tested or verified in a sequential way.Nondeterministic Sequential (NDSeq) specifications have been proposed as a means for achieving this decomposition in testing, debugging, and verifying a program's parallelism correctness and its sequential functional correctness. An NDSeq specification for a parallel program is a sequential version of the program, with no parallel threads but with some limited, controlled nondeterminism. A program's use of parallelism is correct if, for every execution of the parallel program, there exists an execution of the NDSeq specification producing the same result. Functional correctness can then be checked on this NDSeq specification, without any interleaving of parallel threads.While NDSeq specifications have been used successfully to check parallelism correctness, manually writing NDSeq specifications for programs can be a challenging and time-consuming process. Thus, in this work, we propose a technique for automatically inferring a likely NDSeq specification for a parallel program. Given a few representative executions of a parallel program, our technique combines dynamic data flow analysis and Minimum-Cost Boolean Satisfiability (MinCostSAT) solving to infer a likely NDSeq specification for the program's parallelism. We have implemented our technique for Java in a prototype tool called NDETERMIN. For a number of benchmarks, our tool infers equivalent or stronger NDSeq specifications than those previously written manually.

show abstract

Section: Methodsmentioning

confidence: 99%

NDetermin: Inferring Nondeterministic Sequential Specifications for Parallelism Correctness

Burnim¹,

Elmas²,

Necula³

et al. 2011

View full text Add to dashboard Cite

show abstract

“…These tools include Apache Commons 1. 30 , a plain Java array is used. JLargeArrays, as its name suggests, supports only arrays.…”

Section: Third-party Primitive Collectionsmentioning

confidence: 99%

“…Overlapping collections include lists, hash sets, and hash maps, with Object support in hash maps. 30 , regular Java arrays are employed 6 Uses sun.misc.Unsafe…”

Section: Third-party Primitive Collectionsmentioning

confidence: 99%

UPC: Large-Scale Memory Efficient Java Primitive Collections

Hylock¹

2016

JSW

View full text Add to dashboard Cite

Abstract:The Java programming language is versatile and robust, however, for large-scale applications with memory intensive processes, much is still to be desired. The public Java Development Kit (JDK) supports neither forcible object destruction nor large collection sizes (limited to 231-1) as found in choice computational languages such as Fortran and C/C++. Within Java's Hotspot, however, the non-public sun.misc.Unsafe class allows such features through basic off-heap functionality. The submitted UPC collections are fully integrated with Unsafe to provide large (i.e., 263-1), destructible arrays, lists, hash sets, hash maps, and matrices, consistent with advanced computational needs. Additionally, a customized version of OpenJDK 1.9 with bulk-operation Unsafe support and companion collections are provided. These tools are compared to Java and extent third-party primitive collections. Testing indicates UPC performs in a consistent to superior manner compared to the state-of-the-art, with greater improvements realized by way of the modified virtual machine.

show abstract

“…zJava [19] combines compile time analysis with run time support to extract parallelism among methods. Parallel Java [20] provides the same capabilities as OpenMP/MPI in a pure Java API. HPJava [21] uses an array distribution such as High Performance Fortran (HPF).…”

Section: Related Workmentioning

confidence: 99%

Parallel Java Code Generation for Layer-unified Coarse Grain Task Parallel Processing

Yoshida

Ochi

Yamanouchi

2014

IPSJ Online Transactions

View full text Add to dashboard Cite

Multicore processors are widely used for various types of computers. In order to achieve high-performance on such multicore systems, it is necessary to extract coarse grain task parallelism from a target program in addition to loop parallelism. Regarding the development of parallel programs, Java or a Java-extension language represents an attractive choice recently, thanks to its performance improvement as well as its platform independence. Therefore, this paper proposes a parallel Java code generation scheme that realizes coarse grain task parallel processing with layerunified execution control. In this parallel processing, coarse grain tasks of all layers are collectively managed through a dynamic scheduler. In addition, we have developed a prototype parallelizing compiler for Java programs with directives. In performance evaluations, the compiler-generated parallel Java code was confirmed to attain high performance. Concretely, we obtained 7.82 times faster speed-up for the Jacobi program, 7.38 times faster speed-up for the Turb3d program, 6.54 times faster speed-up for the Crypt program, and 6.15 times faster speed-up for the MolDyn program on eight cores of Xeon E5-2660.

show abstract

Parallel Java: A Unified API for Shared Memory and Cluster Parallel Programming in 100% Java

Abstract: Parallel Java is a parallel programming API whose goals are (1)

Cited by 46 publications

References 9 publications

NDetermin: Inferring Nondeterministic Sequential Specifications for Parallelism Correctness

NDetermin: Inferring Nondeterministic Sequential Specifications for Parallelism Correctness

UPC: Large-Scale Memory Efficient Java Primitive Collections

Parallel Java Code Generation for Layer-unified Coarse Grain Task Parallel Processing

Contact Info

Product

Resources

About