Mojo function

block_scaled_matmul

block_scaled_matmul[c_type: DType, a_type: DType, b_type: DType, scales_dtype: DType, //, *, SF_VECTOR_SIZE: Int, transpose_b: Bool = True, target: StringSlice[StaticConstantOrigin] = "cpu", elementwise_lambda_fn: Optional[elementwise_epilogue_type] = None, pdl_level: PDLLevel = PDLLevel()](c_device: NDBuffer[c_type, 2, MutAnyOrigin, c_device.shape], a_device: NDBuffer[a_type, 2, MutAnyOrigin, a_device.shape], b_device: NDBuffer[b_type, 2, MutAnyOrigin, b_device.shape], a_scales_device: NDBuffer[scales_dtype, 5, MutAnyOrigin, a_scales_device.shape], b_scales_device: NDBuffer[scales_dtype, 5, MutAnyOrigin, b_scales_device.shape], tensor_sf: Float32, ctx: DeviceContext)