Understanding the Power of Evolutionary Computation for GPU Code Optimization
Authors:
- 1 November 2022
- conference paper
- Published by Institute of Electrical and Electronics Engineers (IEEE)
- p. 185-198
Abstract
Achieving high performance for GPU codes requires developers to have significant knowledge in parallel programming and GPU architectures, and in-depth understanding of the application. This combination makes it challenging to find performance optimizations for GPU-based applications, especially in scientific computing. This paper shows that significant speedups can be achieved on two quite different scientific workloads using the tool, GEVO, to improve performance over human-optimized GPU code. GEVO uses evolutionary computation to find code edits that improve the runtime of a multiple sequence alignment kernel and a SARS-CoV-2 simulation by 28.9% and 29% respectively. Further, when GEVO begins with an early, unoptimized version of the sequence alignment program, it finds an impressive 30 times speedup–a performance improvement similar to that of the hand-tuned version. This work presents an in-depth analysis of the discovered optimizations, revealing that the primary sources of improvement vary across applications; that most of the optimizations generalize across GPU architectures; and that several of the most important optimizations involve significant code interdependencies. The results showcase the potential of automated program optimization tools to help reduce the optimization burden for scientific computing developers and enhance performance portability for domain-specific accelerators.Keywords
All Related Versions
Funding Information
- National Science Foundation
- National Energy Research Scientific Computing Center
- Lawrence Berkeley National Laboratory
This publication has 38 references indexed in Scilit:
- Grow and Graft a Better CUDA pknotsRG for RNA Pseudoknot Free Energy Calculation2015
- Stochastic optimization of floating-point programs with tunable precisionACM SIGPLAN Notices, 2014
- Software mutational robustnessGenetic Programming and Evolvable Machines, 2013
- BarraCUDA - a fast short read sequence aligner using graphics processing unitsBMC Research Notes, 2012
- Sequencing technologies and genome sequencingJournal of Applied Genetics, 2011
- Synthesis of loop-free programsACM SIGPLAN Notices, 2011
- Z3: An Efficient SMT Solver2008
- Genetic programming as a means for programming computers by natural selectionStatistics and Computing, 1994
- Identification of common molecular subsequencesJournal of Molecular Biology, 1981
- A Deductive Approach to Program SynthesisACM Transactions on Programming Languages and Systems, 1980