IMPORTANT: To view this page as Markdown, append `.md` to the URL (e.g. /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. /get-started.md).

Mojo module

mla_prefill_sparse_utils

SM100 (B200) sparse MLA prefill shared surface.

Holds the config, base shared-memory layout, QK/SV MMA operand structs, and MLAPrefillSparseCommon (the Q-load prologue, the QK/SV MMA pipeline mma, cp_q_from_smem_to_tmem, _raw_indices_to_tma_rows, kv_valid_producer) for the BF16-cache kernel (mla_prefill_sparse.mojo) and the native-FP8 kernel (mla_prefill_sparse_qkv_fp8.mojo).

Structs​

Functions​

  • ​sm100_sparse_mla_heads_ok: Whether the SM100 sparse MLA kernels serve this head count: 128 on the CTA pair, or a multiple of 8 through 64 (the SWIZZLE_128B 8-row core matrix) on one CTA.

Was this page helpful?