כתבה
arXiv cs.AI ·
FP8 הכול שאתה צריך (חלק 2): 3-D FFT מלאה ב-FP64 על FP8-Generation Tensor Cores
FP8 is All You Need (Part 2): Full-FP64 3-D FFT on FP8-Generation Tensor CoresThe Integer-Epilogue Wall and the Minimal Hardware That Would Remove It
GPU של NVIDIA מצליח לבצע FFT 3-D בדיוק FP64 על FP8 tensor cores. החידוש נעשה באמצעות תוכנה שלא דורשת FP64. ה-FP8 layout condition נדרשת לפעולה זו.
תקציר מקורי באנגליתarXiv:2606.23698v3 Announce Type: replace-cross Abstract: The NVIDIA Blackwell Ultra (B300) GPU cuts FP64 vector throughput $\sim 30\times$ while multiplying FP8 tensor throughput. After the recovery of FP64 GEMM via Ozaki Scheme II on FP8 tensor cores and the Tensor-Memory Equilibrium model of the companions ("FP8 is All You Need, Part 1" and "Ozaki 2.5") we ask whether the fifth canonical HPC primitive, the full-FP64 $1024^3$ 3-D FFT, can be carried by the same substrate, and answer with a design and its limit. It is a Bailey six-step transform with no FP64 arithmetic: FP8-tensor DFT GEMMs with fused twiddles, residue-domain Karatsuba combines and exact CRT reconstruction whose bulk is a small GEMM on the FP16 tensor path and whose remainder is a Kulisch fixed-point accumulation with a t
קרא במקור המקורי
arxiv.org
פתח כתבה מקורית