MPI as an abstraction for software-hardware interaction for HPRCs

Abstract

Abstract-High Performance Reconfigurable Computers (HPRCs) consist of one or more standard microprocessors tightly coupled with one or more reconfigurable FPGAs. HPRCs have been shown to provide good speedups and good cost/performance ratios, but not necessarily ease of use, leading to a slow acceptance of this technology. HPRCs introduce new design challenges, such as the lack of portability across platforms, incompatibilities with legacy code, users reluctant to change their code base, a prolonged learning curve, and the need for a system-level Hardware/Software co-design development flow. This paper presents the evolution and current work on TMD-MPI, which started as an MPI-based programming model for Multiprocessor Systems-on-Chip implemented in FPGAs, and has now evolved to include multiple X86 processors. TMD-MPI is shown to address current design challenges in HPRC usage, suggesting that the MPI standard has enough syntax and semantics to program these new types of parallel architectures. Also presented is the TMD-MPI Ecosystem, which consists of research projects and tools that are developed around TMD-MPI to further improve HPRC usability.

A Design of Pipeline Chain Algorithm Based on Circuit Switching for MPI Broadcast Communication System

“…비 록 메시지 패싱이 일반적으로 HPC에서 사용되고 있지만, 최근 임베디드 시스템에서의 FPGA(Field Programmable Gate Arrays)를 이용하여 MPI를 보 다 효율적으로 구현할 수 있어 임베디드 애플리케 이션(embedded application)에서의 관심이 높아지고 있다 [7,8] . 또한 디지털 셋탑 박스(set-top box)와 모 바일 기기에서 등에서 HD(High Definition) 급의 화질을 요구하기 때문에 임베디드 애플리케이션 분 야에서의 관심이 높다 [9] .…”

Integrating High-Level Synthesis into MPI

“…This section provides a brief overview of our previous work on ArchES-MPI [4]. ArchES-MPI was developed to provide a well-known programming model for specifying and developing HPRC applications by adopting MPI [1], one of the most popular and effective standards in HPC.…”

“…This allows the host processor to manage system-wide communication via MPI, while the hardware accelerator is implemented separately. Our previous work on TMD-MPI [3], [4] attempts to remove this separation between programming models by providing an implementation of MPI intended for HPRC software and hardware.…”

SIRC: An Extensible Reconfigurable Computing Communication API

“…However, the focus of these projects has largely been on creating programming abstractions for developers to easily model the parallelism or communication relationships between their hardware and software modules (e.g. multithreading [1] or message passing [2]). While appropriate for large-scale computations in which the HW/SW co-design problem is complex, these programming models may be overly complicated and introduce unnecessary overhead in smaller-scale systems in which the computation is not split over multiple hardware modules and the division between what should be done in hardware versus software is clear.…”