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

block_scaled_smem

Shared memory layout for block-scaled SM100 matmul.

Provides A/B/C tile storage plus scaling factor tile storage (SFA, SFB) following MXFP8 layout conventions. Also includes pipeline barriers and TMEM state.

The tile storage, derived constants, layouts, and accessors are factored into BlockScaledTileCore and shared with GroupedBlockScaledSmem and Grouped1D1DSmem. Each SMEM struct is a thin wrapper that adds the appropriate pipeline bundle.

Structs

• ​BlockScaledSmem: SMEM struct for block-scaled matmul with CLC scheduler pipeline.
• ​BlockScaledTileCore: Core tile storage for block-scaled matmul SMEM structs.

Functions

• ​sf_bk: Compute SF_BK from config.
• ​sf_k_group_size: Compute SF_K_GROUP_SIZE from config.
• ​sfa_dim0: Compute SFA first dimension from config.
• ​sfa_dim1: Compute SFA second dimension from config.
• ​sfb_dim0: Compute SFB first dimension from config.
• ​sfb_dim1: Compute SFB second dimension from config.

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