Preprint
Sep 2026
FlexPosit: Tunable Fractional Precision for LLM Inference Accelerators
FlexPosit is a unified bit-serial systolic array with lightweight per-column decoders, unified Processing Elements (PEs), and a global precision controller, enabling tunable fractional precision while preserving fully regular systolic dataflow.
Yimin Gao, Liangtao Dai, Jun Yin et al.
· 0 citations