Path-Based In-Memory Computing Crossbar Logic
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current digital in-memory computing paradigms rely heavily on expensive WRITE operations during the execution phase, which are energy-inefficient and detrimental to the longevity of non-volatile resistive devices, limiting their ability to deliver deterministic precision for data-intensive applications.
Innovation Solution
A path-based in-memory computing paradigm that performs Boolean logic evaluations using only READ operations during the execution phase, leveraging a one-time compilation phase and synthesizing Boolean functions into a crossbar design with staircase structures to minimize inter-crossbar communication and reduce energy consumption.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If digital in-memory computing paradigms use WRITE operations during the execution phase, then Boolean logic can be evaluated, but energy consumption increases significantly and device lifetime decreases
Solution Approach 1:
The patent applies preliminary action by performing all necessary WRITE operations during a one-time compilation phase before execution. The crossbar design is pre-configured with the Boolean function logic during compilation, so that during the execution phase, only READ operations are needed to evaluate the function. This eliminates the need for repeated WRITE operations during execution, thereby reducing energy consumption and extending device lifetime.
2Measurement precision
If analog in-memory computing is used for matrix-vector multiplication, then high-speed computation with low energy consumption is achieved, but deterministic precision is lost
Solution Approach 1:
The patent substitutes analog in-memory computing with digital in-memory computing using non-volatile memory devices. Instead of relying on analog resistance values that suffer from precision issues, the invention uses digital states (0 and 1) stored in non-volatile memory cells within a crossbar architecture. This digital approach maintains deterministic precision while still achieving low energy consumption because the computation is performed directly in memory without requiring high-precision analog signal processing.
3Use of energy by moving object
If conventional computing architecture is used with bus transfer, then data can be moved between memory and computing units, but energy efficiency deteriorates due to the Von Neumann bottleneck
Solution Approach 1:
The patent merges the storage and computing functions into a single integrated system using a crossbar architecture. The non-volatile memory devices serve both as storage elements and as the basis for logical operations. By combining storage and computation in the same physical location and using the same hardware resources for both functions, the system eliminates the need for data transfer between separate memory and computing units, thereby overcoming the Von Neumann bottleneck and improving both energy efficiency and productivity.
Data Source
AI summary
A system and method for evaluating Boolean functions using in-memory computing comprising a plurality of programmed non-volatile memory devices synthesized in a crossbar design. The evaluation phase of a given Boolean function using the programmed non-volatile memory devices is accomplished using READ operations only.