In the first part of this talk, I will present an autotuner for production machine learning (ML) compilers that can tune both graph-level and subgraph-level optimizations at multiple compilation stages. We demonstrate how to incorporate machine learning techniques such as a learned cost model to reduce autotuning time. Our learned cost model has high accuracy and outperforms a heavily-optimized analytical performance model. I will outline how we deploy the XLA autotuner at datacenter scale to automatically tune the most heavily-used production models in Google’s fleet everyday. The deployed tile size and flag autotuners have been saving approximately 2% of fleetwide TPU compute time.
While the first part of the talk focuses on the efficiency of ML workloads on compute resources at Google, the second part of the talk investigates the efficiency of Google’s storage system. In particular, we introduce Disk of Theseus, a methodology to accurately reconstruct disk I/O traces by carefully combining subsampled I/O traces collected from multiple disks. We show that our synthesized traces are accurate compared to real metrics being collected. We then demonstrate how Disk of Theseus enables diverse counterfactual I/O trace synthesis and analyses of hypothetical policy and hardware changes using three case studies: (1) deploying new disks of an imagined capacity, (2) spinning down HDDs storing cold data, and (3) using low-RPM HDDs.
Phitchaya Phothilimthana (Mangpo) is a research scientist at Google Brain, where she leads Machine Learning for Machine Learning Compilers effort (one of Google Brain moonshots in 2020). Her research interests include compilers, machine learning for systems, program synthesis, and energy-aware computing. Mangpo received an undergraduate degree in Computer Science from MIT and PhD from UC Berkeley. Mangpo was a recipient of Microsoft Research PhD Fellowship and Qualcomm Innovation Fellowship. In high school, Mangpo was the first female student to represent Thailand at IOI, where she received bronze and silver medals in 2006 and 2007 respectively.