Комментарий Программного комитета:
Главный в The D Language Foundation и автор кучи книг: ссылки на его книжки — http://erdani.com/index.php/articles.
Optimization — making code faster — is an essential ingredient of modern computing. The speed of light is limited, so we're unable to increase computer clock speed anymore; in other news, the matter is composed of discrete atoms so we can't make circuits much smaller than they already are. That means more speed for essential tasks — ranging from speech recognition to self-driving cars to General Artificial Intelligence — is to be found in people's minds.
This talk discusses two new examples at the opposite ends of the spectrum of optimization techniques. First, a silicon-oriented technique: replacing branches with arithmetic. Modern CPUs deal better with straight-line code than with "surprising" branches, and this quirk can be exploited in a principled manner. Second, a (surprise!) classic algorithm can be adjusted to take into account data distribution statistics and locality, for better theoretical and practical performance.