A Learned Performance Model for Tensor Processing Units

Part of Proceedings of Machine Learning and Systems 3 pre-proceedings (MLSys 2021)

Bibtex »Paper »

Bibtek download is not availble in the pre-proceeding


Sam Kaufman, Phitchaya Phothilimthana, Yanqi Zhou, Charith Mendis, Sudip Roy, Amit Sabne, Mike Burrows


Accurate hardware performance models are critical to efficient code generation. They can be used by compilers to make heuristic decisions, by superoptimizers as a minimization objective, or by autotuners to find an optimal configuration for a specific program. However, they are difficult to develop because contemporary processors are complex, and the recent proliferation of deep learning accelerators has increased the development burden. We demonstrate a method of learning performance models from a corpus of tensor computation graph programs for Tensor Processing Unit (TPU) instances. We show that our learned model outperforms a heavily-optimized analytical performance model on two tasks—tile-size selection and operator fusion—and that it helps an autotuner discover faster programs in a setting where access to TPUs is limited or expensive.