Product

Ozaki Scheme II

Numerical method that splits double-precision operands into integer residues, multiplies them on low-precision tensor cores and reconstructs the FP64 result through the Chinese Remainder Theorem.

Coverage

Sep 16, 2026
HPCNews
Matsuoka's Revised FP8 Papers Concede the 3-D FFT in Software and Name One Hardware Block That Could Also Lift Emulated DGEMM

Three revised arXiv papers concede the 3-D FFT to hardware and model one converter that could double emulated DGEMM on Rubin. No kernel timing is measured yet.

Sep 16, 2026 · By SCN Staff · 15 min read