Mojo module

mha_structured

MHA prefill kernel for gfx950 with structured scheduling.

Supports depth=64, 128, 256. Uses TileTensor throughout for register and SMEM tile management, with TiledMmaOp for MMA dispatch.

Functions

• ​barrier:
• ​block_sync_lds_direct_load:
• ​set_priority:

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