Photonic SRAM In-Memory Computing Tensor Core
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current computing systems face challenges in achieving high-speed data-intensive computing, particularly in integrating data storage and processing to eliminate bottlenecks between storage and processing units.
Innovation Solution
The development of a photonic SRAM-based in-memory computing tensor core that incorporates optical memory cells, waveguides, optical filters, and electro-optical circuitry to perform operations such as multiplication, convolution, and transposition directly within the memory cell.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Speed
If data storage and processing are integrated in the same circuit, then data processing speed is improved, but device complexity increases
Solution Approach 1:
The patent merges data storage and processing functions into a single integrated circuit. Optical memory cells store data while photonic compute elements perform mathematical operations directly on the stored data without requiring data transfer between separate storage and processing units, thereby eliminating the bottleneck between storage and processing while maintaining high processing speed.
Solution Approach 2:
The patent replaces traditional electronic computing with photonic computing. Optical signals carry data through waveguides and undergo mathematical operations via photonic components such as optical modulators and detectors, substituting electronic signal processing with optical signal processing to achieve higher speed and reduced complexity.
2Adaptability or versatility
If photonic SRAM-based in-memory computing is implemented, then computational capability is improved, but manufacturing complexity increases
Solution Approach 1:
The patent segments the computing system into modular components: optical memory cells for data storage, photonic compute elements for mathematical operations, waveguides for signal transmission, and electro-optical converters for signal conversion. This modular segmentation allows each component to be manufactured and tested independently, reducing overall manufacturing complexity while maintaining high computational capability.
Applied Scientific Principles
This section explains which scientific principles are used to turn an abstract innovation direction into a practical engineering solution.
Function Achieved in This Case
This approach enables high-speed computing by performing mathematical operations within the memory cell, thereby increasing data processing speeds and extending processor capabilities.
Implementation Method 1
one or more optical resonators, such as ring resonators or micro ring resonators
Implementation Method 2
filter modulators, such as electro-optical modulators for one or more ring resonators
Implementation Method 3
one or more waveguides, including waveguides optically coupled to one or more optical resonators
Implementation Method 4
electrical connections to one or more optical sources, such as a photo diode, which may supply electrical signals for controlling output of the one or more optical sources
Data Source
AI summary
Provided is memory circuit comprising a plurality of ring resonators optically coupled to one or more waveguides and electrically coupled to one or more diodes, where the diodes are optically coupled to the one or more waveguides, which is configured to provide optical SRAM or another type of memory with in-memory computing. Further provided is an array of said memory circuits.