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

Abstract

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.

A Task-driven Parallel Code Generation Scheme for Coarse Grain Parallelization on Android Platform

“…Then, each worker-thread binding to a core starts to execute a macrotask in Fork/Join framework with work-stealing. In code generation of the task-driven execution, the first phase detects the inter-macrotask parallelism according to the layer-unified coarse grain parallelization [6], [7], which defines macrotasks hierarchically, analyzes the earliest executable conditions in Table 1 and represents them as a macro-task-graph (MTG) [6], [10]. The second phase generates the task-driven parallel code that can execute the macrotasks on multicores in a dynamic scheduling manner.…”

“…On the other hand, the parallel code generation scheme for the layer-unified coarse grain parallelization has been proposed [7]. This scheme can generate a parallel Java code with the original scheduler implemented by Runnable interface, but the generated parallel code does not adopt the task-driven execution style and can not cope with Java Fork/Join framework.…”

“…Particularly, in order to enhance coarse grain parallelism extraction, the layer-unified execution control scheme [6], [7] has been proposed, which can utilize coarse grain task parallelism among different layers.…”