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
tile_scheduler_splitk
Tile scheduler for split-K matmul kernels on SM100 structured hardware.
Wraps the B200 tile scheduler to partition the K dimension into multiple splits, coordinating per-split workspace reductions and lock-based synchronization across CTAs that contribute to the same output tile.
Structsβ
- β
SchedulerWorkIteratorSplitK: Work iterator for Scheduler warp (split-K) using next-style iteration. - β
TileScheduler: Tile scheduler that partitions the K dimension into multiple splits. - β
WaitAndAdvanceContextSplitK: Context for waiting on CLC barrier and advancing work iterator (Split-K). - β
WorkInfo: Describes a unit of split-K work for a single output tile. - β
WorkIteratorSplitK: Per-warp work iterator for split-K using next-style iteration.
Functionsβ
- β
get_num_tiles: Computes the number of output tiles aligned up to the cluster shape. - β
get_required_locks_buffer_size_bytes: Computes the byte size of the lock buffer needed for split-K reduction.
Was this page helpful?
Thank you! We'll create more content like this.
Thank you for helping us improve!