MCMComm: Hardware-Software Co-Optimization for End-to-End Communication in Multi-Chip-Modules

Fuente: arXiv
Saved in:
Bibliographic Details
Main Authors: Raj, Ritik, Lin, Shengjie, Won, William, Krishna, Tushar
Format: Preprint
Published: 2025
Subjects:
Online Access:
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866916717752483840
author Raj, Ritik
Lin, Shengjie
Won, William
Krishna, Tushar
author_facet Raj, Ritik
Lin, Shengjie
Won, William
Krishna, Tushar
contents Increasing AI computing demands and slowing transistor scaling have led to the advent of Multi-Chip-Module (MCMs) based accelerators. MCMs enable cost-effective scalability, higher yield, and modular reuse by partitioning large chips into smaller chiplets. However, MCMs come at an increased communication cost, which requires critical analysis and optimization. This paper makes three main contributions: (i) an end-to-end, off-chip congestion-aware and packaging-adaptive analytical framework for detailed analysis, (ii) hardware software co-optimization incorporating diagonal links, on-chip redistribution, and non-uniform workload partitioning to optimize the framework, and (iii) using metaheuristics (genetic algorithms, GA) and mixed integer quadratic programming (MIQP) to solve the optimized framework. Experimental results demonstrate significant performance improvements for CNNs and Vision Transformers, showcasing up to 1.58x and 2.7x EdP (Energy delay Product) improvement using GA and MIQP, respectively.
format Preprint
id arxiv_https___arxiv_org_abs_2505_00041
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle MCMComm: Hardware-Software Co-Optimization for End-to-End Communication in Multi-Chip-Modules
Raj, Ritik
Lin, Shengjie
Won, William
Krishna, Tushar
Hardware Architecture
Increasing AI computing demands and slowing transistor scaling have led to the advent of Multi-Chip-Module (MCMs) based accelerators. MCMs enable cost-effective scalability, higher yield, and modular reuse by partitioning large chips into smaller chiplets. However, MCMs come at an increased communication cost, which requires critical analysis and optimization. This paper makes three main contributions: (i) an end-to-end, off-chip congestion-aware and packaging-adaptive analytical framework for detailed analysis, (ii) hardware software co-optimization incorporating diagonal links, on-chip redistribution, and non-uniform workload partitioning to optimize the framework, and (iii) using metaheuristics (genetic algorithms, GA) and mixed integer quadratic programming (MIQP) to solve the optimized framework. Experimental results demonstrate significant performance improvements for CNNs and Vision Transformers, showcasing up to 1.58x and 2.7x EdP (Energy delay Product) improvement using GA and MIQP, respectively.
title MCMComm: Hardware-Software Co-Optimization for End-to-End Communication in Multi-Chip-Modules
topic Hardware Architecture
url https://arxiv.org/abs/2505.00041