Mojo function
swapAB_smallM_ceildiv
swapAB_smallM_ceildiv[a_type: DType, b_type: DType, c_type: DType, transpose_b: Bool = True](m: Scalar[DType.uint], pdl_level: PDLLevel) -> MatmulConfig[a_type, b_type, c_type, transpose_b]
Config for m < 41 range with BN = ceildiv(m, 8) * 8 pattern.
Pattern: - BN = ceildiv(m, 8) * 8 (rounds up to next multiple of 8) - stages = 12, cluster = (1,1,1), swapAB = True
Returns:
Was this page helpful?
Thank you! We'll create more content like this.
Thank you for helping us improve!