Monomorphism-based CGRA Mapping via Space and Time Decoupling

TL;DR

This paper introduces a novel CGRA mapping method that decouples space and time exploration, significantly speeding up compilation while maintaining high-quality mappings, especially for large CGRAs.

Contribution

We propose a space-time decoupled mapping approach using SMT and monomorphism-based search, improving scalability and compilation speed for large CGRAs.

Findings

Achieves comparable mapping quality to state-of-the-art methods.

Reduces compilation time by approximately 10^5 times for large CGRAs.

Effective for large-scale CGRA configurations.

Abstract

Coarse-Grain Reconfigurable Arrays (CGRAs) provide flexibility and energy efficiency in accelerating compute-intensive loops. Existing compilation techniques often struggle with scalability, unable to map code onto large CGRAs. To address this, we propose a novel approach to the mapping problem where the time and space dimensions are decoupled and explored separately. We leverage an SMT formulation to traverse the time dimension first, and then perform a monomorphism-based search to find a valid spatial solution. Experimental results show that our approach achieves the same mapping quality of state-of-the-art techniques while significantly reducing compilation time, with this reduction being particularly tangible when compiling for large CGRAs. We achieve approximately $10^5\times$ average compilation speedup for the benchmarks evaluated on a $20\times 20$ CGRA.