Author: Lei Zhou
Title: Implementation of the Partitioned Global Address Space Model for Multi-GPU Systems
Abstract:
The Partitioned Global Address Space (PGAS) approach is a promising
programming model in high performance parallel computing that combines
the advantages of distributed memory systems and shared memory
systems. The PGAS model has been used on a variety of hardware
platforms in the form of PGAS programming languages like Unified
Parallel C, Chapel and Fortress. However, in spite of the increasing
adoption in distributed and shared memory systems, the extension of
the PGAS model to accelerator platforms is still not well supported.
To exploit the immense computational power of multi-GPU systems, this
work proposes the design of a Partitioned Global Address Space model
for multi-GPU systems. Several issues of combining logically separated
GPU memories on multiple graphic cards are addressed. Furthermore, the
execution model of modern GPU architectures is studied, based on which
a task creation mechanism supporting heterogeneous platforms and load
balancing is proposed. The proposed model is implemented as an
extension of DASH, a C++ template library that realizes the PGAS
concept via operator overloading, on multi-GPU systems. Experimental
results suggest promising programmability and performance of the model
and its implementation.