Probabilistic Circuit Implementation Based on P-Bits Using the Intrinsic Random Property of RRAM and P-Bit...

Area of Science:

• Computer Engineering
• Emerging Computational Paradigms
• Hardware Acceleration

Background:

• Probabilistic computing offers efficient solutions for complex optimization problems intractable for traditional digital methods.
• Resistive random-access memory (RRAM) is a promising technology for novel computing architectures.
• Invertible logic and p-circuits are key components in probabilistic computing.

Purpose of the Study:

• To propose a true random number generator (TRNG) based on RRAM for probabilistic computing.
• To develop a standard p-bit cell by integrating RRAM with a piecewise linear activation function.
• To demonstrate resource optimization in probabilistic circuits through a p-bit multiplexing strategy.

Main Methods:

• Design of a TRNG using RRAM and a piecewise linear activation function to create a p-bit cell.
• Implementation of a p-bit multiplexing strategy to enhance resource utilization.
• Field-programmable gate array (FPGA) implementation of invertible probabilistic circuits (AND gates, adders, multipliers).

Main Results:

• The proposed RRAM-based p-bit cell forms a functional component of probabilistic circuits.
• P-bit multiplexing effectively reduces the number of required p-bits and improves resource efficiency.
• FPGA implementation of invertible circuits demonstrates significant savings in hardware resource consumption compared to traditional methods.

Conclusions:

• The RRAM-based probabilistic computing approach offers a hardware-efficient solution for optimization problems.
• The developed p-bit cell and multiplexing strategy are effective in reducing resource overhead.
• This work validates the potential of probabilistic circuits for resource-constrained applications on FPGAs.