Efficient Automatic Scheduling of Imaging and Vision Pipelines for the GPU

Luke Anderson, MIT CSAIL

Andrew Adams, Adobe

Karima Ma, MIT CSAIL

Tzu-Mao Li, MIT CSAIL and UC San Diego

Tian Jin, MIT CSAIL

Jonathan Ragan-Kelley, MIT CSAIL


Abstract

We present a new algorithm to quickly generate high-performance GPU implementations of complex imaging and vision pipelines, directly from high-level Halide algorithm code. It is fully automatic, requiring no schedule templates or hand-optimized kernels. We address the scalability challenge of extending search-based automatic scheduling to map large real-world programs to the deep hierarchies of memory and parallelism on GPU architectures in reasonable compile time. We achieve this using (1) a two-phase search algorithm that first ‘freezes’ decisions for the lowest cost sections of a program, allowing relatively more time to be spent on the important stages, (2) a hierarchical sampling strategy that groups schedules based on their structural similarity, then samples representatives to be evaluated, allowing us to explore a large space with few samples, and (3) memoization of repeated partial schedules, amortizing their cost over all their occurrences. We guide the process with an efficient cost model combining machine learning, program analysis, and GPU architecture knowledge. We evaluate our method’s performance on a diverse suite of real-world imaging and vision pipelines. Our scalability optimizations lead to average compile time speedups of 49x (up to 530x). We find schedules that are on average 1.7x faster than existing automatic solutions (up to 5x), and competitive with what the best human experts were able to achieve in an active effort to beat our automatic results.

Publication

Efficient Automatic Scheduling of Imaging and Vision Pipelines for the GPU
Luke Anderson, Andrew Adams, Karima Ma, Tzu-Mao Li, Tian Jin, Jonathan Ragan-Kelley
Proceedings of the ACM on Programming Languages (OOPSLA 2021)

Files

OOPSLA 2021 Paper

BibTeX

Acknowledgments

This work was partially funded by Toyota Research Institute, NSF awards CCF-1723445 and CCF- 1846502, and DARPA agreement HR00112090017.