C-SIMD: CORDIC-Driven SIMD Processing Element for Resource-Efficient Multi-Precision Deep Learning Inference

Jan 1, 2026 by Vasundhara Trivedi, Gopal R. Raut, Baker Mohammad, S. Vishvakarma, Akash Kumar (IEEE Access)

DOI 10.1109/ACCESS.2026.3653253

We built C-SIMD, a CORDIC-driven SIMD PE that scalably composes pipelined 8-bit approximate multipliers into configurable 4/8/16/32-bit MACs, squeezing area and power while letting you trade accuracy for complexity across symmetric and asymmetric modes. It fully exploits resources to hit big throughput gains and multi-precision flexibility with minimal accuracy loss, making it a practical, resource-frugal building block for edge DNN inference.

source S2, crossref