CUTEv2: Unified and Configurable Matrix Extension for Diverse CPU Architectures with Minimal Design Overhead

TL;DR

This paper introduces a unified, configurable matrix extension for CPUs that reduces design overhead, enhances AI workload performance, and is adaptable across diverse architectures with open-source implementation.

Contribution

It presents a decoupled, flexible matrix extension architecture supporting mixed-precision operations and asynchronous abstraction, integrated into open-source CPU platforms.

Findings

Matrix unit utilization exceeds 90% on GEMM workloads.

Achieves up to 2.31x speedup on AI models like Llama3.

Design occupies only 0.53 mm² in 14nm CMOS.

Abstract

Matrix extensions have emerged as an essential feature in modern CPUs to address the surging demands of AI workloads. However, existing designs often incur substantial hardware and software design overhead. Tight coupling with the CPU pipeline complicates integration across diverse CPUs, while fine-grained synchronous instructions hinder the development of high-performance kernels. This paper proposes a unified and configurable CPU matrix extension architecture. By decoupling matrix units from the CPU pipeline, the design enables low-overhead integration while maintaining close coordination with existing compute and memory resources. The configurable matrix unit supports mixed-precision operations and adapts to diverse compute demands and memory bandwidth constraints. An asynchronous matrix multiplication abstraction with flexible granularity conceals hardware details, simplifies matrix-vector overlap, and supports a unified software stack. The architecture is integrated into four open-source CPU RTL platforms and evaluated on representative AI models. Matrix unit utilization under GEMM workloads exceeds 90% across all platforms. When configured with compute throughput and memory bandwidth comparable to Intel AMX, our design achieves speedups of 1.57x, 1.57x, and 2.31x on ResNet, BERT, and Llama3, with over 30% of the gains attributed to overlapped matrix-vector execution. A 4 TOPS@2GHz matrix unit occupies only 0.53 mm\textsuperscript{2} in 14nm CMOS. These results demonstrate strong cross-platform adaptability and effective hardware-software co-optimization, offering a practical matrix extension for the open-source community.