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_prefill_sparse_utils
SM100 (B200) sparse MLA prefill shared surface (BF16 + FP8 KV variants).
Holds the config, base shared-memory layout, QK/SV MMA operand structs, and
MLAPrefillSparseCommon — the dtype-agnostic machinery (Q-load prologue, the
QK/SV MMA pipeline mma with its fp8_active flag, cp_q_from_smem_to_tmem,
_raw_indices_to_tma_rows, kv_valid_producer) instantiated by both the
BF16-KV kernel (mla_prefill_sparse.mojo) and the FP8-KV kernel
(mla_prefill_sparse_kv_fp8.mojo). Mirrors the sparse-decode split
(mla_decode_utils.mojo + mla_decode_sparse_kv_{bf16,fp8}.mojo).
Structs
Was this page helpful?
Thank you! We'll create more content like this.
Thank you for helping us improve!