Fig. 3: In-memory computing hardware implementation. | Nature Communications

Fig. 3: In-memory computing hardware implementation.

From: Computing high-degree polynomial gradients in memory

Fig. 3

a A toy example of CNF Boolean function with M = 2, N = K = 4, which is equivalent to the monomial in Fig. 1a with a1 = 1, a2 = −1, a3 = 1, and a4 = − 1 when represented in a PUBO form. b A prospective in-memory 1T1R memristor circuit implementation of the panel (a) problem for parallel computation of its break values. Each 1T1R memory cell is comprised of a select transistor coupled with a memristor. Note the gate voltages are tied to the word lines in the utilized chip to suppress leakages through unselected memory cells. A text shown in blue corresponds to a specific variable assignment x1 = 1, x2 = 0, x3 = 1, x4 = 0. c Experimental setup details for solving the considered 14-variable 64-clause 3-SAT problem. Two arrays represent 1T1R crossbar circuits of a chip in the integrated CMOS/memristor setup and are used for demonstrating forward and backward passes. The colormaps show the measured conductance of programmed memristors corresponding to the studied 3-SAT problem in the experiment.

Back to article page