Abstract:Energy efficiency of electronic digital processors is primarily limited by the energy consumption of electronic communication and interconnects. The industry is almost unanimously pushing towards replacing both long-haul, as well as local chip interconnects, using optics to drastically increase efficiency. In this paper, we explore what comes after the successful migration to optical interconnects, as with this inefficiency solved, the main source of energy consumption will be electronic digital computing, memory and electro-optical conversion. Our approach attempts to address all these issues by introducing efficient all-optical digital computing and memory, which in turn eliminates the need for electro-optical conversions. Here, we demonstrate for the first time a scheme to enable general purpose digital data processing in an integrated form and present our photonic integrated circuit (PIC) implementation. For this demonstration we implemented a URISC architecture capable of running any classical piece of software all-optically and present a comprehensive architectural framework for all-optical computing to go beyond.

What problem does this paper attempt to address?

The core problem that this paper attempts to solve is the bottleneck in energy efficiency of current electronic digital processors. Specifically, the energy consumption of electronic communication and interconnection is the main factor limiting the energy efficiency of electronic digital processors. Although the industry is striving to improve efficiency by replacing electronic interconnections with optical interconnections, once this problem is solved, the main source of energy consumption will shift to electronic computing, memory, and photoelectric conversion. To solve these problems, this paper proposes a brand - new all - optical general - purpose CPU and optical computer architecture. This architecture aims to further improve energy efficiency by introducing efficient all - optical digital computing and storage to eliminate the need for photoelectric conversion. The following are the specific goals of the paper: 1. **Propose architectural guidelines for optical computing**: Go beyond the von Neumann architecture and show how this method surpasses electronic technology in terms of scalability (Part II). 2. **Clarify common misunderstandings about optical computing**: Include the explanation and correction of long - standing misunderstandings about optical computing (Part II). 3. **Demonstrate the implementation of an all - optical general - purpose 2 - bit demonstration CPU**: Prove the feasibility of all - optical general - purpose digital computing (Part III). ### Key contributions of the paper - **All - optical computing architecture**: The paper introduces a new architecture that combines old and new concepts, such as RWORM, delay - line - based registers, and all - optical LUT (Part III). These designs make all - optical general - purpose computing possible on a photonic platform. - **Reduce energy consumption**: By means of all - optical computing and storage, the need for photoelectric conversion is eliminated, thereby reducing energy consumption. - **Break through the limitations of traditional architectures**: It is shown that optical computing can adopt designs different from traditional electronic architectures, for example, no longer relying on a large number of transistors or specific bit - width requirements. ### Problems solved - **High energy consumption in electronic communication**: Through all - optical interconnection and computing, the energy loss in communication and computing processes is significantly reduced. - **Limitations of existing electronic architectures**: New architectural designs are proposed, enabling optical computing to handle modern workloads without relying on complex instruction sets or a large number of transistors. - **Requirements for high - performance computing**: By optimizing the instruction set and memory access methods, efficient general - purpose computing is achieved, especially for tasks that require a large amount of computing, such as AI. In conclusion, the goal of this paper is to fundamentally solve the energy efficiency problem of electronic digital processors by developing an all - optical general - purpose CPU and optical computer architecture, and to provide a new solution for future high - performance computing.

An All-Optical General-Purpose CPU and Optical Computer Architecture

Advances in Spatial Analog Optical Computing Devices

All-Optical Computing Based on Convolutional Neural Networks

The physics of optical computing

Photonic-Electronic Integrated Circuits for High-Performance Computing and AI Accelerators

In-memory computing on a photonic platform

Integrated electro-optic digital-to-analog link for efficient computing and arbitrary waveform generation

Optical Computing: Status and Perspectives

Integrated photonics modular arithmetic processor

A Parallel Photonic Chip for Nano-Second End-to-end Image Processing, Transmission, and Reconstruction

Silicon-based Optoelectronics for General-Purpose Matrix Computation: a Review

Attojoule Optoelectronics for Low-Energy Information Processing and Communications

A Digital-Optical System-on-package Module Architecture for High Performance Multiprocessors

Computing on Silicon Photonic Platform

Universitet Roadmap on all-optical processing

Reconfigurable Analog Computing Architecture Based on Planar Optical Waveguides

Optical Bio-Inspired Synaptic Devices

Roadmap on all-optical processing

Attojoule Optoelectronics for Low-Energy Information Processing and Communications: a Tutorial Review

Analog Optical Computing for Artificial Intelligence

Implementing optical analog computing and electrooptic Hopfield network by silicon photonic circuits