Optimizing GPU Memory Transactions for Convolution Operations

摘要

Convolution computation is a common operation in deep neural networks (DNNs) and is often responsible for performance bottlenecks during training and inferencing. Existing approaches for accelerating convolution operations aim to reduce computational complexity. However, these strategies often increase the memory footprint with extra memory accesses, thereby leaving much room for performance improvement. This paper presents a novel approach to optimize memory access for convolution operations, specifically targeting GPU execution. Our approach leverages two optimization techniques to reduce the number of memory operations for convolution operations performed on the width and height dimensions. For convolution computations on the width dimension, we exploit shuffle instructions to exchange the overlapped columns of the input for reducing the number of memory transactions. For convolution operations on the height dimension, we multiply each overlapped row of the input with multiple rows of a filter to compute multiple output elements to improve the data locality of row elements. We apply our approach to 2D and multi-channel 2D convolutions on an NVIDIA 2080Ti GPU. For 2D convolution, our approach delivers over faster performance than the state-of-the-art image processing libraries. For multi-channel 2D convolutions, we obtain up to speedups over the quickest algorithm of cuDNN. We apply our approach to 2D and multi-channel 2D convolutions on an NVIDIA 2080Ti GPU. For 2D convolution, our approach delivers over $mathbf{2}imes$ faster performance than the state-of-the-art image processing libraries. For multi-channel 2D convolutions, we obtain up to $mathbf{1.3}imes$ speedups over the quickest algorithm of cuDNN.