IMPORTANT: To view this page as Markdown, append `.md` to the URL (e.g. /get-started.md). For the complete documentation index, see llms.txt.
Skip to main content
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 function

ssd_chunk_cb

def ssd_chunk_cb[kernel_dtype: DType, SLT: TensorLayout, SEngine: TensorEngine, BFn: def[width: Int, alignment: Int](Int, Int, Int) -> SIMD[kernel_dtype, width] & RegisterPassable & ImplicitlyCopyable, CFn: def[width: Int, alignment: Int](Int, Int, Int) -> SIMD[kernel_dtype, width] & RegisterPassable & ImplicitlyCopyable, DSTATE: Int](slots_i: Int32, ngroups_i: Int32, info: TileTensor[.int32, SLT, ImmUntrackedOrigin, Engine=SEngine], cb_out: TileTensor[.float32, SLT, MutUntrackedOrigin, Engine=SEngine], b_fn: BFn, c_fn: CFn)

Computes undecayed fp32 C B^T once per chunk and B/C group.

Grid (slots, ngroups, CHUNK / TILE), block OUT_THREADS. The fragment traversal and increasing-k MMA order match ssd_chunk_output exactly; every fragment the consumer loads, including its masked lanes, is stored.

Was this page helpful?