PyDac: a distributed runtime system and programming model for a heterogeneous many-core architecture

Search for this publication on Google Scholar
Huang, B. (2014). PyDac: a distributed runtime system and programming model for a heterogeneous many-core architecture. Unc Charlotte Electronic Theses And Dissertations.
Download PDF

Analytics

86 views ◎
44 downloads ⇓


Abstract

Heterogeneous many-core architectures that consist of big, fast cores and small, energy-efficient cores are very promising for future high-performance computing (HPC) systems. These architectures offer a good balance between single-threaded performance and multithreaded throughput. Such systems impose challenges on the design of programming model and runtime system. Specifically, these challenges include (a) how to fully utilize the chip's performance, (b) how to manage heterogeneous, unreliable hardware resources, and (c) how to generate and manage a large amount of parallel tasks.This dissertation proposes and evaluates a Python-based programming framework called PyDac. PyDac supports a two-level programming model. At the high level, a programmer creates a very large number of tasks, using the divide-and-conquer strategy. At the low level, tasks are written in imperative programming style. The runtime system seamlessly manages the parallel tasks, system resilience, and inter-task communication with architecture support. PyDac has been implemented on both an field-programmable gate array (FPGA) emulation of an unconventional heterogeneous architecture and a conventional multicore microprocessor. To evaluate the performance, resilience, and programmability of the proposed system, several micro-benchmarks were developed. We found that (a) the PyDac abstracts away task communication and achieves programmability, (b) the micro-benchmarks are scalable on the hardware prototype, but (predictably) serial operation limits some micro-benchmarks, and (c) the degree of protection versus speed could be varied in redundant threading that is transparent to programmers.

Details
Author

Huang, Bin

Title

PyDac: a distributed runtime system and programming model for a heterogeneous many-core architecture

Physical Description

1 online resource (88 pages) : PDF

Date

2014

Degree Granting Institution

University of North Carolina at Charlotte

Abstract

Heterogeneous many-core architectures that consist of big, fast cores and small, energy-efficient cores are very promising for future high-performance computing (HPC) systems. These architectures offer a good balance between single-threaded performance and multithreaded throughput. Such systems impose challenges on the design of programming model and runtime system. Specifically, these challenges include (a) how to fully utilize the chip's performance, (b) how to manage heterogeneous, unreliable hardware resources, and (c) how to generate and manage a large amount of parallel tasks.This dissertation proposes and evaluates a Python-based programming framework called PyDac. PyDac supports a two-level programming model. At the high level, a programmer creates a very large number of tasks, using the divide-and-conquer strategy. At the low level, tasks are written in imperative programming style. The runtime system seamlessly manages the parallel tasks, system resilience, and inter-task communication with architecture support. PyDac has been implemented on both an field-programmable gate array (FPGA) emulation of an unconventional heterogeneous architecture and a conventional multicore microprocessor. To evaluate the performance, resilience, and programmability of the proposed system, several micro-benchmarks were developed. We found that (a) the PyDac abstracts away task communication and achieves programmability, (b) the micro-benchmarks are scalable on the hardware prototype, but (predictably) serial operation limits some micro-benchmarks, and (c) the degree of protection versus speed could be varied in redundant threading that is transparent to programmers.

Genre

doctoral dissertations

Subjects--Topics

Computer engineering

Degree

Ph.D.

Keywords

Emulation
FPGA
Many-Core Architecture
Programming Model
Python
Runtime System

Subject Area

Electrical Engineering

Advisor(s)

Sass, Ronald

Degree Note

Thesis (Ph.D.)--University of North Carolina at Charlotte, 2014.

Rights Statement

This Item is protected by copyright and/or related rights. You are free to use this Item in any way that is permitted by the copyright and related rights legislation that applies to your use. For other uses you need to obtain permission from the rights-holder(s). For additional information, see http://rightsstatements.org/page/InC/1.0/.

Rights Holder Information

Copyright is held by the author unless otherwise indicated.

Identifier

Huang_uncc_0694D_10641

Permalink

http://hdl.handle.net/20.500.13093/etd:1474