Accelerating SpMM Kernel with Cache-First Edge Sampling for Graph Neural Networks
This addresses a critical inference time bottleneck for GNN users, enabling faster processing for large-scale or more complex models, though it is incremental as it optimizes an existing operator rather than introducing a new paradigm.
The paper tackles the performance bottleneck of sparse-dense matrix multiplication (SpMM) in graph neural networks (GNNs), which accounts for 95% of inference time on GPUs, by introducing ES-SpMM, a cache-first edge sampling mechanism and codesigned kernel that speeds up SpMM by up to 4.35x with no accuracy loss and 45.3x with less than 1% accuracy loss.
Graph neural networks (GNNs), an emerging deep learning model class, can extract meaningful representations from highly expressive graph-structured data and are therefore gaining popularity for wider ranges of applications. However, current GNNs suffer from the poor performance of their sparse-dense matrix multiplication (SpMM) operator, even when using powerful GPUs. Our analysis shows that 95% of the inference time could be spent on SpMM when running popular GNN models on NVIDIA's advanced V100 GPU. Such SpMM performance bottleneck hinders GNNs' applicability to large-scale problems or the development of more sophisticated GNN models. To address this inference time bottleneck, we introduce ES-SpMM, a cache-first edge sampling mechanism and codesigned SpMM kernel. ES-SpMM uses edge sampling to downsize the graph to fit into GPU's shared memory. It thus reduces the computation cost and improves SpMM's cache locality. To evaluate ES-SpMM's performance, we integrated it with a popular GNN framework, DGL, and tested it using representative GNN models and datasets. Our results show that ES-SpMM outperforms the highly optimized cuSPARSE SpMM kernel by up to 4.35x with no accuracy loss and by 45.3x with less than a 1% accuracy loss.