IMPORTANT: To view this page as Markdown, append `.md` to the URL (e.g. /max/get-started.md). For the complete documentation index, see llms.txt.
Skip to main content
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

mla_decode_sparse

Implements the sparse MLA decoding kernel for SM100 (Blackwell) GPUs.

Provides the MLA_SM100_Decode_Sparse struct, a sparse-attention variant of the SM100 MLA decode kernel that loads KV cache tiles via gather4 TMA using per-tile sparse indices, with FP8-to-BF16 conversion in shared memory before UMMA consumption.

Structs

Was this page helpful?