PyGim: An Efficient Graph Neural Network Library for Real Processing-In-Memory Architectures

Christina Giannoula, Peiming Yang, Ivan Fernandez, Jiacheng Yang, Sankeerth Durvasula, Yu Xin Li, Mohammad Sadrosadati, Juan Gomez Luna, Onur Mutlu, Gennady Pekhimenko

Keywords: machine learning, graph neural networks, sparse matrix-matrix multiplication, library, multicore, processing-in-memory, near-data processing, memory systems, data movement bottleneck, DRAM, benchmarking, real-system characterization, workload characterization

Abstract:

Graph Neural Networks (GNNs) are emerging models to analyze graph-structure data. The GNN execution
involves both compute-intensive and memory-intensive kernels. The memory-intensive kernels dominate
execution time, because they are significantly bottlenecked by data movement between memory and processors.
Processing-In-Memory (PIM) systems can alleviate this data movement bottleneck by placing simple processors
near or inside memory arrays. To this end, we investigate the potential of PIM systems to alleviate the data
movement bottleneck in GNNs, and introduce PyGim, an efficient and easy-to-use GNN library for real PIM
systems. We propose intelligent parallelization techniques for memory-intensive kernels of GNNs tailored
for real PIM systems, and develop an easy-to-use Python API for them. PyGim employs a cooperative GNN
execution, in which the compute- and memory-intensive kernels are executed in processor-centric and
memory-centric computing systems, respectively, to fully exploit the hardware capabilities. PyGim integrates
a lightweight tuner that configures the parallelization strategy of the memory-intensive kernel of GNNs to
provide high system performance, while also enabling high programming ease. We extensively evaluate PyGim
on a real-world PIM system that has 16 PIM DIMMs with 1992 PIM cores connected to a Host CPU. In GNN
inference, we demonstrate that it outperforms prior state-of-the-art PIM works by on average 4.38× (up to
7.20×), and the state-of-the-art PyTorch implementation running on Host (on Intel Xeon CPU) by on average
3.04× (up to 3.44×). PyGim improves energy efficiency by 2.86× (up to 3.68×) and 1.55× (up to 1.75×) over prior
PIM and PyTorch Host schemes, respectively. In memory-intensive kernel of GNNs, PyGim provides 11.6×
higher resource utilization in PIM system than that of PyTorch library (optimized CUDA implementation)
in GPU systems. Our work provides useful recommendations for software, system and hardware designers.
PyGim is publicly and freely available at https://github.com/CMU-SAFARI/PyGim to facilitate the widespread
use of PIM systems in GNNs.

Submission Number: 43