Directive-Based, High-Level Programming and Optimizations for High-Performance Computing with FPGAs

Abstract

Reconfigurable architectures like Field Programmable Gate Arrays (FPGAs) have been used for accelerating computations from several domains because of their unique combination of flexibility, performance, and power efficiency. However, FPGAs have not been widely used for high-performance computing, primarily because of their programming complexity and difficulties in optimizing performance. In this paper, we present a directive-based, high-level optimization framework for high-performance computing with FP-GAs, built on top of an OpenACC-to-FPGA translation framework called OpenARC. We propose directive extensions and corresponding compile-time optimization techniques to enable the compiler to generate more efficient FPGA hardware configuration files. Empirical evaluation of the proposed framework on an Intel Stratix V with five OpenACC benchmarks from various application domains shows that FPGA-specific optimizations can lead to significant increases in performance across all tested applications. We also demonstrate that applying these high-level directive-based optimizations can allow OpenACC applications to perform similarly to lower-level OpenCL applications with hand-written FPGA-specific optimizations, and offer runtime and power performance benefits compared to CPUs and GPUs.

Performance Evaluation of Compiler Optimizations in FPGA Accelerators

“…The kernel generated from the annotated region is already optimized. A follow-up publication [Lambert et al 2018], by the same group, adds new optimizations to the compiler. Unlike the work of Lloyd et al, the solution does not require host and device compiler integration because the compiler has control of the device code that is produced.…”

FCNNLib: A Flexible Convolution Algorithm Library for Deep Learning on FPGAs

“…As a result, hardware accelerators such as GPUs, FPGAs, and customized ASICs have been employed to accelerate DNNs [4], [5], [6], [7], [8], [9], [10], [11]. Among them, FPGAs emerges as a promising solution owing to its high available parallelism and flexibility [12], [13], [14], [15], [16], [17].…”