How to solve the current memory access and data transfer bottlenecks: at the processor architecture or at the compiler level?

Francky Catthoor

Nikil D. Dutt

Christos Kozyrakis Stanford

Design, Automation and Test in Europe Conference and Exhibition (DATE), 2000


Abstract

Current processor architectures are increasingly dominated by data-access bottlenecks in caches, system buses, and main-memory subsystems. This invited hot-topic paper examines whether future solutions should rely primarily on processor architecture, compiler and synthesis technology, or a combination of both. It discusses explicitly parallel architectures, memory-aware compilation, and the interaction between hardware and software techniques for improving memory performance.