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 function
build_sm100_matmul_configs
def build_sm100_matmul_configs[a_type: DType, b_type: DType, c_type: DType, N: Int, K: Int, transpose_b: Bool = True, has_epilogue_tensor: Bool = False, epilogue_is_1d: Bool = False]() -> Set[MatmulConfig[a_type, b_type, c_type, transpose_b]]
Build a set of MatmulConfig instances by sweeping M from 8 to 8192.
Parameters:
- βa_type (
DType):DTypeof the A (left) operand elements; must equalb_type. - βb_type (
DType):DTypeof the B (right) operand elements. - βc_type (
DType):DTypeof the output matrix elements. - βN (
Int): The N dimension (output columns) of the matmul. - βK (
Int): The K dimension (contraction axis) of the matmul. - βtranspose_b (
Bool): Whether the B operand is stored transposed (defaults toTrue). - βhas_epilogue_tensor (
Bool): Whether the kernel uses a TMA epilogue load for an epilogue tensor (defaults toFalse). - βepilogue_is_1d (
Bool): Whether the epilogue tensor is 1D, for example a bias vector (defaults toFalse).
Returns:
Set[MatmulConfig[a_type, b_type, c_type, transpose_b]]: A set of unique MatmulConfig instances covering the swept M range.
Was this page helpful?
Thank you! We'll create more content like this.
Thank you for helping us improve!