ARLGFeb 2, 2022

Efficient Memory Partitioning in Software Defined Hardware

arXiv:2202.01261v3
Originality Incremental advance
AI Analysis

This addresses efficiency challenges for programmers using software-defined hardware to implement complex algorithms, representing a strong specific gain rather than a broad paradigm shift.

The paper tackles the problem of automatically partitioning on-chip arrays in software-defined hardware compilers, introducing a system that uses ML cost modeling to generate partitioning schemes that reduce logic resources by 40.3%, FFs by 78.3%, BRAMs by 54.9%, and DSPs by 100% compared to state-of-the-art methods.

As programmers turn to software-defined hardware (SDH) to maintain a high level of productivity while programming hardware to run complex algorithms, heavy-lifting must be done by the compiler to automatically partition on-chip arrays. In this paper, we introduce an automatic memory partitioning system that can quickly compute more efficient partitioning schemes than prior systems. Our system employs a variety of resource-saving optimizations and an ML cost model to select the best partitioning scheme from an array of candidates. We compared our system against various state-of-the-art SDH compilers and FPGAs on a variety of benchmarks and found that our system generates solutions that, on average, consume 40.3% fewer logic resources, 78.3% fewer FFs, 54.9% fewer Block RAMs (BRAMs), and 100% fewer DSPs.

Foundations

The foundational work for this paper's niche, ranked by how specifically the neighbourhood builds on it — not by global fame.

Your Notes