The Case for Polymorphic Registers in Dataflow Computing

Catalin Bogdan Ciobanu*, Georgi Gaydadjiev, Christian Pilato, Donatella Sciuto

*Corresponding author for this work

Research output: Contribution to journalArticleAcademicpeer-review

2 Citations (Scopus)
139 Downloads (Pure)

Abstract

Heterogeneous systems are becoming increasingly popular, delivering high performance through hardware specialization. However, sequential data accesses may have a negative impact on performance. Data parallel solutions such as Polymorphic Register Files (PRFs) can potentially accelerate applications by facilitating high-speed, parallel access to performance-critical data. This article shows how PRFs can be integrated into dataflow computational platforms. Our semi-automatic, compiler-based methodology generates customized PRFs and modifies the computational kernels to efficiently exploit them. We use a separable 2D convolution case study to evaluate the impact of memory latency and bandwidth on performance compared to a state-of-the-art NVIDIA Tesla C2050GPU. We improve the throughput upto 56.17X and show that the PRF-augmented system outperforms the GPU for for 9 x 9 or larger mask sizes, even in bandwidth-constrained systems.

Original languageEnglish
Pages (from-to)1185-1219
Number of pages35
JournalInternational journal of parallel programming
Volume46
Issue number6
DOIs
Publication statusPublished - Dec-2018
Externally publishedYes

Keywords

  • Dataflow computing
  • Parallel memory accesses
  • Polymorphic register file
  • Bandwidth
  • Vector lanes
  • Convolution
  • High performance computing
  • High-level synthesis
  • ARCHITECTURE

Fingerprint

Dive into the research topics of 'The Case for Polymorphic Registers in Dataflow Computing'. Together they form a unique fingerprint.

Cite this