Re: [PD] vanilla partitioned convolution abstraction