US 12,105,985 B2
Processing data in-memory with memory devices having a crossbar array structure
Iason Giannopoulos, Thalwil (CH); Navaneeth Rameshan, Zurich (CH); Vara Sudananda Prasad Jonnalagadda, Wallisellen (CH); and Abu Sebastian, Adliswil (CH)
Assigned to International Business Machines Corporation, Armonk, NY (US)
Filed by International Business Machines Corporation, Armonk, NY (US)
Filed on Mar. 21, 2022, as Appl. No. 17/655,634.
Prior Publication US 2023/0297268 A1, Sep. 21, 2023
Int. Cl. G06F 3/06 (2006.01)
CPC G06F 3/0655 (2013.01) [G06F 3/0604 (2013.01); G06F 3/0679 (2013.01)] 25 Claims
OG exemplary drawing
 
1. A computer-implemented method of processing data in-memory comprising:
providing a memory device having a crossbar array structure including N input lines and M output lines, arranged in N rows and M columns, which are interconnected at cross-points defining N×M cells, wherein
the cells include respective memory elements, whereby the crossbar array structure includes a set of N×M memory elements, which decomposes into a first subset of K×M memory elements and a second subset of L×M memory elements, in accordance with a first subset of K rows and a second subset of L rows of the N rows, respectively, where N≥2, M≥2, 1≤L<M, and K+L=N, and
target data are stored across the first subset of K×M memory elements, whereas operand data are stored across the second subset of L×M memory elements;
applying electrical signals to at least two input lines corresponding to at least two rows, the latter including at least one of the K rows and at least one of the L rows, to obtain output signals in output of the M output lines, the output signals depending on target values and operand values, in accordance with data stored across said at least two rows; and
while reading out the output signals, performing a transformation operation, in-memory, on the target values corresponding to data stored in at least one of the subsets of K rows based on the operand values corresponding to data stored in at least one of the subsets of L rows, to obtain transformed data.