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_utils

Structs

• ​CvtToMMAPipeline:
• ​MLAConfig:
• ​MLAKVLayouts: Comptime layout and size metadata for MLA K/V tiles.
• ​MLAPositionSummary:
• ​SM100MLA:
• ​TMAtoCvtPipeline:

Functions

• ​cvt_block_fp8_to_bf16_with_scale: TileTensor overload — standalone implementation using .ptr and comptime static_shape/static_stride directly.
• ​select_mla_prefill_config: Selects the supported SM100 MLA-prefill config for these dims.
• ​split_smem: Split a shared memory tensor into two TileTensors at the boundary of first_size elements.