A Mixed-signal Computing-in-Memory Solver for Integer Constrained Least Square Problems
An 8T-SRAM-based Mixed-Signal Computing-In-Memory (MS-CIM) is demonstrated in 65nm CMOS for solving Integer-Constrained Least Square problems. The test-chip consumes 21.76mW (at 1.2V) and requires 720us to solve a 20-input 16-output Integer Constrained Least Square (ICLS) problem generating solutions with quality ranking in the top 99.98 percentile.
Authors
- Sudarshan Sharma (ORCID: https://orcid.org/0009-0004-0949-4577)
- Wen Wang (ORCID: https://orcid.org/0000-0002-2810-5109)
- Sheng Zhang (ORCID: https://orcid.org/0000-0003-2304-0956)
- N. M. Anisur Rahman
- Narasimha Vasishta Kidambi
- Daehyun Kim (ORCID: https://orcid.org/0000-0002-5582-3579)
- Sramana Mukhopadhyay (ORCID: https://orcid.org/0000-0002-4588-3068)
Institutions
- Georgia Institute of Technology (US)
Publication Details
- Published
- 2025-05-25
- DOI
- https://doi.org/10.1109/iscas56072.2025.11043550
- Primary Topic
- Advanced Memory and Neural Computing
- Type
- article
- Field-Weighted Citation Impact
- 0.00