How not to lie with statistics: the correct way to summarize benchmark results
Communications of the ACM - The MIT Press scientific computation series
A portable global optimizer and linker
PLDI '88 Proceedings of the ACM SIGPLAN 1988 conference on Programming Language design and Implementation
Code scheduling and register allocation in large basic blocks
ICS '88 Proceedings of the 2nd international conference on Supercomputing
An approach to ordering optimizing transformations
PPOPP '90 Proceedings of the second ACM SIGPLAN symposium on Principles & practice of parallel programming
Register allocation with instruction scheduling
PLDI '93 Proceedings of the ACM SIGPLAN 1993 conference on Programming language design and implementation
Dependence-conscious global register allocation
Proceedings of the international conference on Programming languages and system architectures
An introduction to genetic algorithms
An introduction to genetic algorithms
An approach for exploring code improving transformations
ACM Transactions on Programming Languages and Systems (TOPLAS)
The SimpleScalar tool set, version 2.0
ACM SIGARCH Computer Architecture News
Optimizing for reduced code space using genetic algorithms
Proceedings of the ACM SIGPLAN 1999 workshop on Languages, compilers, and tools for embedded systems
Postpass Code Optimization of Pipeline Constraints
ACM Transactions on Programming Languages and Systems (TOPLAS)
Introduction to algorithms
A Feasibility Study in Iterative Compilation
ISHPC '99 Proceedings of the Second International Symposium on High Performance Computing
Compiler optimization-space exploration
Proceedings of the international symposium on Code generation and optimization: feedback-directed and runtime optimization
Phase coupling and constant generation in an optimizing microcode compiler
MICRO 15 Proceedings of the 15th annual workshop on Microprogramming
Fast searches for effective optimization phase sequences
Proceedings of the ACM SIGPLAN 2004 conference on Programming language design and implementation
Finding effective compilation sequences
Proceedings of the 2004 ACM SIGPLAN/SIGBED conference on Languages, compilers, and tools for embedded systems
ACME: adaptive compilation made efficient
LCTES '05 Proceedings of the 2005 ACM SIGPLAN/SIGBED conference on Languages, compilers, and tools for embedded systems
Using Machine Learning to Focus Iterative Optimization
Proceedings of the International Symposium on Code Generation and Optimization
Exhaustive Optimization Phase Order Space Exploration
Proceedings of the International Symposium on Code Generation and Optimization
On the decidability of phase ordering problem in optimizing compilation
Proceedings of the 3rd conference on Computing frontiers
MiBench: A free, commercially representative embedded benchmark suite
WWC '01 Proceedings of the Workload Characterization, 2001. WWC-4. 2001 IEEE International Workshop
Cole: compiler optimization level exploration
Proceedings of the 6th annual IEEE/ACM international symposium on Code generation and optimization
Practical exhaustive optimization phase order exploration and evaluation
ACM Transactions on Architecture and Code Optimization (TACO)
| Hi-index | 0.00 |
The problem of finding the most effective set and ordering of optimization phases to generate the best quality code is a fundamental issue in compiler optimization research. Unfortunately, the exorbitantly large phase order search spaces in current compilers make both exhaustive as well as heuristic approaches to search for the ideal optimization phase combination impractical in most cases. In this paper we show that one important reason existing search techniques are so expensive is because they make no attempt to exploit well-known independence relationships between optimization phases to reduce the search space, and correspondingly improve the search time. We explore the impact of two complementary techniques to prune typical phase order search spaces. Our first technique studies the effect of implicit application of cleanup phases, while the other partitions the set of phases into mutually independent groups and develops new multi-stage search algorithms that substantially reduce the search time with no effect on best delivered code performance. Together, our techniques prune the exhaustive phase order search space size by 89%, on average, (96.75% total search space reduction) and show immense potential at making iterative phase order searches more feasible and practical. The pruned search space enables us to find a small set of distinct phase sequences that reach near-optimal phase ordering performance for all our benchmark functions as well as to improve the behavior of our genetic algorithm based heuristic search.