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

tile_gemm

def tile_gemm[N: Int, K: Int, ACTIVE: Bool](lds: LdsBytes, mut acc: StaticTuple[SIMD[.float32, 4], Int(16)], a_base: Int, a_rows: Int, a_scales: Int, sa_stride: Int, m_tile_row: Int, b: Int, b_scales: Int, n_wave: Int, wave: Int)

One tile's K into the wave's accumulators: rows m_tile_row ..< m_tile_row + BM of A, of which the first a_rows exist (the rest read as zeros), columns n_wave ..< n_wave + WN of B. sa_stride is the row stride of A's scales. ACTIVE == False: the wave owns no columns of this tile (past N), and only fills its share of A and meets the barriers.

Was this page helpful?