High Performance Im2win and Direct Convolutions using Three Tensor Layouts on SIMD Architectures

Convolution is the core component within deep neural networks and it is computationally intensive and time consuming. Tensor data layouts significantly impact convolution operations in terms of memory access and computational efficiency. Yet, there is still a lack of comprehensive performance characterization on data layouts on SIMD architectures concerning convolution methods. This paper proposes three novel data layouts for im2win convolution: NHWC, CHWN, and CHWN8, and introduces a set of general optimization techniques for both direct and im2win convolutions. We compare the optimized im2win convolution with the direct convolution and PyTorch's im2col-based convolution across the aforementioned layouts on SIMD machines. The experiments demonstrated that the im2win convolution with the new NHWC layout achieved up to 355% performance speedup over NCHW layout. Our optimizations also significantly improve the performance of both im2win and direct convolutions. Our optimized im2win and direct convolutions achieved up to 95% and 94% of machine's theoretical peak performance, respectively.

Im2win: Memory Efficient Convolution On SIMD Architectures

Convolution is the most expensive operation among neural network operations, thus its performance is critical to the overall performance of neural networks. Commonly used convolution approaches, including general matrix multiplication (GEMM)-based convolution and direct convolution, rely on im2col for data transformation or do not use data transformation at all, respectively. However, the im2col data transformation can lead to at least 2$\times$ memory footprint compared to not using data transformation at all, thus limiting the size of neural network models running on memory-limited systems. Meanwhile, not using data transformation usually performs poorly due to nonconsecutive memory access although it consumes less memory. To solve those problems, we propose a new memory-efficient data transformation algorithm, called im2win. This algorithm refactorizes a row of square or rectangle dot product windows of the input image and flattens unique elements within these windows into a row in the output tensor, which enables consecutive memory access and data reuse, and thus greatly reduces the memory overhead. Furthermore, we propose a high-performance im2win-based convolution algorithm with various optimizations, including vectorization, loop reordering, etc. Our experimental results show that our algorithm reduces the memory overhead by average to 41.6% compared to the PyTorch's convolution implementation based on im2col, and achieves average to 3.6$\times$ and 5.3$\times$ speedup in performance compared to the im2col-based convolution and not using data transformation, respectively.

Im2win: An Efficient Convolution Paradigm on GPU

Convolution is the most time-consuming operation in deep neural network operations, so its performance is critical to the overall performance of the neural network. The commonly used methods for convolution on GPU include the general matrix multiplication (GEMM)-based convolution and the direct convolution. GEMM-based convolution relies on the im2col algorithm, which results in a large memory footprint and reduced performance. Direct convolution does not have the large memory footprint problem, but the performance is not on par with GEMM-based approach because of the discontinuous memory access. This paper proposes a window-order-based convolution paradigm on GPU, called im2win, which not only reduces memory footprint but also offers continuous memory accesses, resulting in improved performance. Furthermore, we apply a range of optimization techniques on the convolution CUDA kernel, including shared memory, tiling, micro-kernel, double buffer, and prefetching. We compare our implementation with the direct convolution, and PyTorch's GEMM-based convolution with cuBLAS and six cuDNN-based convolution implementations, with twelve state-of-the-art DNN benchmarks. The experimental results show that our implementation 1) uses less memory footprint by 23.1% and achieves 3.5$\times$ TFLOPS compared with cuBLAS, 2) uses less memory footprint by 32.8% and achieves up to 1.8$\times$ TFLOPS compared with the best performant convolutions in cuDNN, and 3) achieves up to 155$\times$ TFLOPS compared with the direct convolution. We further perform an ablation study on the applied optimization techniques and find that the micro-kernel has the greatest positive impact on performance.