Feather: Software Emulated FP8 for Older GPUs
Wednesday, June 24, 2026 3:45 PM to 5:15 PM · 1 hr. 30 min. (Europe/Berlin)
Foyer D-G - 2nd Floor
Project Poster
Community EngagementML Systems and FrameworksOptimizing for Energy and Performance
Information
Poster is on display.
Feather is a high-performance emulation library that brings FP8 (E5M2 \& E4M3) precision arithmetic to older GPU architectures (Ampere, Turing, Volta) that lack native hardware support.
By implementing custom bit-packing and optimized Triton kernels, Feather bypasses the memory bandwidth bottleneck, achieving 3x speedup over native PyTorch FP32 \& PyTorch FP16 operations on memory-bound workloads.
Feather is a high-performance emulation library that brings FP8 (E5M2 \& E4M3) precision arithmetic to older GPU architectures (Ampere, Turing, Volta) that lack native hardware support.
By implementing custom bit-packing and optimized Triton kernels, Feather bypasses the memory bandwidth bottleneck, achieving 3x speedup over native PyTorch FP32 \& PyTorch FP16 operations on memory-bound workloads.
Format
on-demandon-site
