For the complete documentation index, see llms.txt. Markdown versions of all pages are available by appending .md to any URL (e.g. /max/get-started.md).
Mojo module
matmul_swiglu
CPU entry point for fused GEMM+SwiGLU on SM100.
Creates TMA descriptors for A, B, and C output, then launches the SwiGLU kernel. Output uses a 3-phase epilogue: st_matrix to full SMEM, SwiGLU pass to half SMEM, TMA store to GMEM.
Was this page helpful?
Thank you! We'll create more content like this.
Thank you for helping us improve!