Mojo function

grouped_matmul_sm100_persistent

grouped_matmul_sm100_persistent[c_type: DType, c_shape: DimList, a_type: DType, a_shape: DimList, b_type: DType, b_shape: DimList, transpose_b: Bool, *, config: MatmulConfig[a_type, b_type, c_type, transpose_b], cta_group: Int = 1, num_pipeline_stages: Optional[UInt] = None, a_swizzle: TensorMapSwizzle = 3, b_swizzle: TensorMapSwizzle = 3](c: NDBuffer[c_type, 2, MutableAnyOrigin, c_shape], a: NDBuffer[a_type, 2, MutableAnyOrigin, a_shape], a_offsets: NDBuffer[DType.uint32, 1, MutableAnyOrigin], max_num_tokens_per_expert: Int, b: NDBuffer[b_type, 3, MutableAnyOrigin, b_shape], expert_ids: NDBuffer[DType.int32, 1, MutableAnyOrigin], num_active_experts: Int, ctx: DeviceContext)