A Multicast-Capable AXI Crossbar for Many-Core Machine Learning Accelerators
Student Contest:
Yes
Affiliation Type:
Academia
Keywords:
AI accelerators, on-chip networks, multicast communication, AXI
Abstract:
To keep up with the growing computational requirements of machine learning workloads, many-core accelerators integrate an ever-increasing number of processing elements, putting the efficiency of memory and interconnect subsystems to the test. In this work, we present the design of a multicast-capable AXI crossbar, with the goal of enhancing data movement efficiency in massively-parallel machine learning accelerators. We propose a lightweight, yet flexible, multicast implementation, resulting in a 9.3% area overhead on an 8-to-8 crossbar, and bounded by 21.6% on an asymptotically-large crossbar. To demonstrate the flexibility and end-to-end impact of our design, we integrate our extension in an open-source 288-core accelerator. We evaluate the performance impact on a typical microbenchmark. We further report tangible performance improvements on a key computational kernel for machine learning workloads, matrix multiplication, measuring a 30% speedup on our reference system.
