Efficient tensor core-based GPU kernels for structured sparsity under reduced precision2021-11-130 OA AI DOI:10.1145/3458817.3476182原文链接原文求助分享收藏摘要 En