Abstract:The proposal of compute-in-memory (CIM) is a breakthrough for the traditional von Neumann architecture to achieve efficient computing research. This architecture has unique advantages in the computing field thanks to supporting multi-line computing and without data transmission between processor and memory. In this paper, an in-memory computing structure based on 9T SRAM unit is proposed, which can both operate on memory and computing mode. Compared with the previous works, thanks to the redundant units, the computational structure can directly complete XNOR operations and storage of the whole SRAM array in only one cycle, without the need of extra digital logic circuits (such as AND, OR circuits), which can significantly improve the parallelism of the computation. Meanwhile, the architecture can map the XNOR logical operation into the binary multiplication, and then add up the one-bit multiplication results through the addition tree, thus realizing the binary convolution calculation. A 64-bit × 64-bit (4 Kb) SRAM array with the proposed scheme is designed and simulated in 55 nm CMOS technology. Simulation results show excellent stability and write yields in SRAM memory mode at an operating frequency of 200 MHz. In computing mode, the SRAM array power consumption for logical operation is 52.68 fJ/bit at 1.2 V supply voltage. At a minimum supply voltage of 0.8 V, the power consumption is only 5.58 fJ/bit, with an energy efficiency 179.21 TOPS/W.

An STT-MRAM Based Reconfigurable Computing-in-memory Architecture for General Purpose Computing

Design of an Area-Efficient Computing in Memory Platform Based on STT-MRAM

A 6T-3M SOT-MRAM for In-Memory Computing with Reconfigurable Arithmetic Operations.

An STT-MRAM Based in Memory Architecture for Low Power Integral Computing

In-Memory Computing Paradigms by Exploiting the Intrinsic Interactions Between STT and SOT

Reconfigurable Bit-Serial Operation Using Toggle SOT-MRAM for High-Performance Computing in Memory Architecture

A 9T-SRAM Based Computing-in-memory with Redundant Unit and Digital Operation for Boolean Logic and MAC

A Reconfigurable 4T2R ReRAM Computing In-Memory Macro for Efficient Edge Applications

Architecture-circuit-technology Co-Optimization for Resistive Random Access Memory-Based Computation-in-memory Chips

A6T-3MSOT-MRAMforin-memorycomputingwith reconfigurable arithmetic operations

Reconfigurable Logic Operation Scheme Based on Gate-Tunable CBRAM

A CFMB STT-MRAM-Based Computing-in-Memory Proposal with Cascade Computing Unit for Edge AI Devices

Computing-in-Memory Paradigm Based on STT-MRAM with Synergetic Read/Write-Like Modes

Heterogeneous Instruction Set Architecture for RRAM-enabled In-memory Computing.

Memory Compiler for RRAM In-Memory Computation

Spintronic Memory-Based Reconfigurable Computing

An 8T SRAM Array with Configurable Word Lines for In-Memory Computing Operation

Reconfigurable Processing in Memory Architecture Based on Spin Orbit Torque

A High-Reliability and Low-Power Computing-in-memory Implementation Within STT-MRAM

Exploring a SOT-MRAM Based In-Memory Computing for Data Processing.

Configurable in-memory computing architecture based on dual-port SRAM