持续更新中…

dense gemm

1.mxfp8 × mxfp8 → BF16

https://github.com/haowen-han/cutlass/tree/mxfp8_sm100_hhw/hhw_mxfp8_gemm

2.nvfp4 × nvfp4 → BF16

https://github.com/haowen-han/cutlass/tree/mxfp8_sm100_hhw/hhw_nvfp4_gemm

3.MXFP8 × MXFP4 → BF16

https://github.com/haowen-han/cutlass/tree/fp8_fp4_sm100_hhw/hhw_mxfp4_mxfp8_gemm