Wesley Pang, Gregory Jun, Feiyang Liu, Deming Chen. TileFuse: Fused Mixed-Precision Kernels for Quantized LLM Inference on AMD XDNA2 NPUs. In 34th IEEE Annual International Symposium on Field-Programmable Custom Computing Machines, FCCM 2026, Atlanta, GA, USA, May 13-16, 2026. pages 277, IEEE, 2026. [doi]
Abstract is missing.