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 module

bmm

Provides batched matrix multiplication (BMM) for CPU and GPU targets.

comptime values​

elementwise_epilogue_type​

comptime elementwise_epilogue_type = def[c_type: DType, width: SIMDLength, *, alignment: Int = Int(1)](Coord[*?], SIMD[c_type, width]) capturing thin -> None

Fused elementwise epilogue over the ORIGINAL (uncollapsed) buffer rank.

Batched matmul collapses the batch dims to run a 2-D kernel, so this does not receive the kernel's own coordinate -- unlike linalg.utils.elementwise_epilogue_type, which is always rank 2. The wrappers below un-collapse it back to the caller's own rank. The one exception is _bmm_sm100_blockwise_scaled_fp8_kernel, which rebuilds a rank-3 (batch, m, n) coordinate and so accepts a rank-3 caller only.

logger​

comptime logger = Logger(stdout, prefix=String(""), source_location=False)

Functions​

Was this page helpful?