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

SM100 (Blackwell) MLA prefill kernels.

Provides dense and sparse multi-latent attention (MLA) prefill entry points for NVIDIA SM100 GPUs, including the generic (non-blockscale), blockwise-scale, and sparse (with optional FP8 KV cache) variants.

Functions

• ​mla_sm100_prefill: Dense MLA prefill dispatcher for SM100 (Blackwell).
• ​mla_sm100_prefill_sparse: Sparse MLA prefill (DSv3.2 absorbed shape, BF16, SM100).
• ​mla_sm100_prefill_sparse_fp8: FP8 KV-cache variant of mla_sm100_prefill_sparse.

Was this page helpful?

Thank you! We'll create more content like this.

Thank you for helping us improve!

😔 What went wrong?

Some code doesn’t work

It includes inaccurate information

It's missing information I need

It was difficult to understand

Other

Submit