Computer Architecture Overview

Principles

Definition

Computer architecture is the fundamental discipline concerned with the design, organization, and implementation of computer systems. It investigates the principles and methods for structuring hardware components and their interactions to achieve desired performance, efficiency, and functionality. This field encompasses defining the instruction set architecture (ISA), the microarchitecture, and the system-level design, considering key metrics such as performance, power consumption, cost, and reliability. Its significance lies in enabling the creation of efficient, powerful, and specialized computing hardware that underpins technological advancement.

Ontological type

Core Subfields

Performance Metrics

Design Methodologies

Key developments

Key Figures

Scalable Parallelism Paradigm era

John L. Hennessy [1] was a leading figure associated with Stanford University [3] and the University of California, Berkeley [4] during the Scalable Parallelism Paradigm era. Hennessy [1]'s central contribution in this era is exemplified by the Stanford Dash multiprocessor [6], which demonstrated scalable parallelism and informed subsequent high-throughput design and evaluation practices. Aman Gupta [2] was associated with Stanford University [3] and the University of Minnesota [5] during this era. Gupta [2]'s contributions through the Stanford Dash multiprocessor [6] illustrate collaborative approaches to scalable architectures and helped drive software–hardware co-design emphasis in this period.

Heterogeneous Reconfigurable Microarchitectures era

Ian Foster [1], affiliated with Northwestern University [3] and the University of Chicago [4], emerged as a central figure in advancing heterogeneous reconfigurable microarchitectures during this era. His notable contribution, Globus: a Metacomputing Infrastructure Toolkit [7], enabled scalable, cross-institution resource coordination and reproducible experimentation across heterogeneous platforms, addressing the era's emphasis on programmable fabrics and configurable execution resources. Carl Kesselman [2], associated with University of California, San Diego [5] and the California Institute of Technology [6], likewise played a pivotal role in this period. His work on Globus [7] helped extend metacomputing and distributed resource management to heterogeneous environments, enabling researchers to assemble flexible execution fabrics and tailor resources to workload characteristics, a cornerstone of the era.

Memory-Centric Heterogeneous Acceleration era

Jeremy C. Smith [1] is linked with Rockefeller University [3] and Harvard University [4] during this era. The GROMACS paper [7] demonstrates multi-level parallelism enabling scalable molecular simulations from laptops to supercomputers, underscoring the era’s emphasis on memory-aware, data-local execution. Shekhar Borkar [2] is associated with University of California, Berkeley [5] and University of Minnesota [6] during this era. The future of microprocessors [8] outlines a vision for scalable, energy-aware CPUs and memory-aware accelerators that foreshadow heterogeneous, memory-centric architectures.