Home * Programming * Optimization

Optimization is about to chose the best element from some set of available alternatives, as referred in mathematical optimization as for instance applied in computer chess to optimize evaluation weights with automated tuning methods, and program optimizations, which is the topic covered on this page. Most importantly there is the algorithmic optimization on design level such as using alpha-beta rather than plain minimax, followed by source code optimizations, and finally Compiler optimizations.

Premature Optimization

As a warning on premature optimization a quote by Donald Knuth ^[1]:

We should forget about small efficiencies, say about 97% of the time: premature optimization is the root of all evil.

Program Optimizations

Program Optimization is a necessary part of a decent chess program. It comes in two forms, compiler-end and program-end. Compiler-end optimization involves using specific flags to get a quick program at the cost of speed of compile and memory usage, whereas program-end optimization involves using inline functions and things like that.

• Avoiding Branches
• Performance Measurement
• Profiling
• SIMD and SWAR Techniques

Compiler-end Optimization

Compilers

Generally, the Intel C Compiler produces faster compiles, though in recent times, the GNU C Compiler is catching up.

Compiler flags

• -O3 is the most aggressive optimization flag, though sometimes this can ruin the performance of the chess program.
• -O2 is the 'safest' optimization flag. It produces a decently fast executable, and won't ruin the performance.
• -O1/-O produces a faster executable than it would normally
• -Os is equivent to -O2 but with size of executable and memory usage in mind.

See also

• Automated Tuning
• Genetic Programming
• NUMA
• Simulated Annealing

Publications

• Bruce W. Leverett (1981). Register allocation in optimizing compilers. Ph.D. thesis, Carnegie Mellon University
• Bruce W. Leverett (1982). Topics in Code Generation and Register Allocation. Carnegie Mellon University
• Bruce W. Leverett (1983). Register allocation in optimizing compilers. UMI Research Press

Forum Posts

• Beginner's guide to graphical profiling by Matthew Lai, CCC, September 10, 2016 » Profiling, Giraffe
  Re: Beginner's guide to graphical profiling by Marco Costalba, CCC, September 10, 2016 » Stockfish
• Reliable speed comparison: some math required by Marco Costalba, CCC, February 27, 2018 » Nodes per second

External Links

• Mathematical optimization from Wikipedia
  Global optimization from Wikipedia
  Optimization problem from Wikipedia
• Program optimization from Wikipedia
• Compiler optimization from Wikipedia
• Software optimization resources. C++ and assembly. Windows, Linux, BSD, Mac OS X by Agner Fog
• Programming Optimization by Paul Hsieh

References

1. ^ Donald Knuth (1974).Structured Programming with go to Statements. ACM Computing Surveys, Vol. 6, No. 4, pdf

What links here?

Up one Level