SparseMap: Loop Mapping for Sparse CNNs on Streaming Coarse-grained Reconfigurable Array

TL;DR

SparseMap is a novel mapping method for sparse CNNs on streaming CGRA that significantly reduces cache operations and internal dependencies, improving throughput and efficiency.

Contribution

It introduces an efficient I/O data management and scheduling approach that minimizes cache and dependency issues in sparse CNN acceleration on CGRA.

Findings

Reduces 92.5% cache operations (COPs)

Reduces 46.0% multi-cycle internal dependencies (MCIDs)

Maintains or improves initiation interval (II)

Abstract

Streaming coarse-grained reconfgurable array (CGRA) is a promising architecture for data/computing-intensive applications because of its fexibility, high throughput and efcient memory system. However,when accelerating sparse CNNs, the irregular input data demands inside sparse CNNs would cause excessive caching operations (COPs) and multi-cycle internal dependencies (MCIDs) between operations, declining the throughput of the streaming CGRA. We propose a mapping method for sparse CNNs onto streaming CGRA, SparseMap, which incorporates an efcient I/O data management along with operation scheduling and binding, to reduce the COPs and MCIDs, thereby ensuring the optimal throughput of streaming CGRA.The experimental results show SparseMap reduces 92.5% COPs and 46.0 % MCIDs while achieves the same or even smaller initiation interval (II) compared to previous works.