MultPIM: Fast Stateful Multiplication for Processing-in-Memory

Processing-in-memory (PIM) seeks to eliminate computation/memory data transfer using devices that support both storage and logic. Stateful logic techniques such as IMPLY, MAGIC FELIX can perform gates within memristive crossbar arrays with massive parallelism. Multiplication via stateful is an active field of research due the wide implications. Recently, RIME has become state-of-the-art algorithm for single-row multiplication by partitions, reducing latency previous 5.1×. In this paper, we begin proposing novel partition-based computation broadcasting shifting data. Then, design in-memory based on carry-save add-shift (CSAS) technique. Finally, develop a full-adder significantly improves (FELIX) design. These contributions constitute MultPIM, multiplier reduces time complexity from quadratic linear-log. For 32-bit numbers, MultPIM additional 4.2× over RIME, while even slightly area overhead. Furthermore, optimize full-precision matrix-vector improve 25.5× FloatPIM multiplication.