Reproduce “FPnew: An Open-Source Multi-Format Floating-Point Unit Architecture for Energy-Proportional Transprecision Computing” on OpenResearch
Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.
Universal app for Apple silicon and Intel. Requires macOS 11 or later.
About this paper
Stefan Mach, Fabian Schuiki, Florian Zaruba, Luca Benini
FPnew is a configurable open-source floating-point unit that supports multiple standard and custom formats, packed SIMD operations, and mixed-format fused multiply-add. It can multiply FP16 values and accumulate in FP32. In an RI5CY kernel, this reports 22 correct bits with zero relative error against the exact FP16-rounded result at 0.63 of FP32 system energy; energy is post-layout simulated, not silicon-measured.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.