For the complete documentation index, see llms.txt. Markdown versions of all pages are available by appending .md to any URL (e.g. /get-started.md).

Mojo package

sm100_structured

SM100 Structured Kernels - Blackwell matmul implementation.

Packages

• ​block_scaled: Block-scaled matmul kernel for SM100.
• ​blockwise_fp8: Blockwise FP8 matmul kernel for SM100.
• ​blockwise_fp8_1d2d: Blockwise FP8 1D2D grouped matmul kernel for SM100.
• ​default: Default SM100 matmul kernel - Standard FP8/BF16 warp-specialized implementation.
• ​fused_swiglu: SM100 Fused GEMM+SwiGLU kernel: BF16 matmul with SwiGLU in the epilogue.
• ​grouped_block_scaled: Grouped block-scaled matmul kernel for SM100.
• ​grouped_block_scaled_1d1d: Grouped block-scaled matmul with 1D-1D tensor layout for SM100.
• ​structured_kernels: Shared library components for SM100 structured kernels.