XSP: Across-Stack Profiling and Analysis of Machine Learning Models on GPUs
This addresses the problem for hardware designers and system providers in optimizing systems for ML model performance, though it is incremental as it builds on distributed tracing and automated analysis.
The paper tackles the challenge of profiling machine learning models across the hardware/software stack by proposing XSP, an across-stack profiling design that provides a holistic view of execution, and demonstrates its effectiveness by analyzing 65 state-of-the-art models to uncover insights difficult to obtain with existing tools.
There has been a rapid proliferation of machine learning/deep learning (ML) models and wide adoption of them in many application domains. This has made profiling and characterization of ML model performance an increasingly pressing task for both hardware designers and system providers, as they would like to offer the best possible system to serve ML models with the target latency, throughput, cost, and energy requirements while maximizing resource utilization. Such an endeavor is challenging as the characteristics of an ML model depend on the interplay between the model, framework, system libraries, and the hardware (or the HW/SW stack). Existing profiling tools are disjoint, however, and only focus on profiling within a particular level of the stack, which limits the thoroughness and usefulness of the profiling results. This paper proposes XSP - an across-stack profiling design that gives a holistic and hierarchical view of ML model execution. XSP leverages distributed tracing to aggregate and correlates profile data from different sources. XSP introduces a leveled and iterative measurement approach that accurately captures the latencies at all levels of the HW/SW stack in spite of the profiling overhead. We couple the profiling design with an automated analysis pipeline to systematically analyze 65 state-of-the-art ML models. We demonstrate that XSP provides insights which would be difficult to discern otherwise.