Memory Macro BIST for Fault Detection in Compute-in-Memory
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Performing Built-In Self-Test (BIST) in Compute in Memory (CIM) macros is challenging due to the complexity of CIM operations and the need for efficient energy consumption.
Innovation Solution
A BIST circuit is designed to perform BIST on CIM macros by generating input vectors and weights, performing MAC operations, and comparing output data with expected signatures to identify faults.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Adaptability or versatility
If discrete logic circuits are used to implement MAC operations, then the CIM can perform deep learning computations, but the energy consumption increases and circuit implementation becomes large
Solution Approach 1:
The patent merges the storage function and compute function into a single integrated memory macro. Weight values are stored in memory cells that also perform MAC operations with input vectors, eliminating the need for separate discrete logic circuits. This integration reduces overall circuit area and energy consumption while maintaining deep learning computational capabilities.
Solution Approach 2:
The memory cells in the CIM macro serve multiple functions: they store weight values, perform multiplication with input vectors, accumulate results, and participate in built-in self-test operations. This multi-functionality eliminates the need for dedicated compute circuits, reducing energy consumption and circuit complexity while achieving versatile deep learning operations.
2Adaptability or versatility
If discrete logic circuits are used to implement MAC operations, then the CIM can perform deep learning computations, but the circuit implementation becomes large
Solution Approach 1:
The patent merges the storage function and compute function into a single integrated memory macro. Weight values are stored in memory cells that also perform MAC operations with input vectors, eliminating the need for separate discrete logic circuits. This integration reduces overall circuit area and energy consumption while maintaining deep learning computational capabilities.
3Reliability
If BIST is performed in CIM macros, then fault identification is achieved, but the complexity of CIM operations makes BIST difficult
Solution Approach 1:
The CIM macro performs built-in self-test operations using its own integrated resources. Test patterns are generated within the macro, MAC operations are executed using the same compute units, and results are compared using the memory cells themselves. This self-service approach eliminates the need for external test equipment and simplifies the BIST process while maintaining reliable fault identification.
Solution Approach 2:
The memory cells in the CIM macro serve multiple functions: they store weight values, perform multiplication with input vectors, accumulate results, and participate in built-in self-test operations. This multi-functionality eliminates the need for dedicated compute circuits, reducing energy consumption and circuit complexity while achieving versatile deep learning operations.
Data Source
AI summary
Performing a built-in self-test (BIST) on a memory macro includes generating a plurality of input vectors such that at least one input vector of the plurality of input vectors is transmitted to the memory macro in each of a plurality of cycles, receiving in each of the plurality of cycles, an output data from the memory macro. The output data is generated by the memory macro in response to processing the at least one input vector. The BIST also includes comparing the output data in each of the plurality of cycles with a signature value and determining whether the memory macro is normal or faulty based upon the comparison.