One of the main reason for preferring C++ over simpler, higher-level programming languages is that C++ allows the development of complex software in a way that makes more efficient use of hardware resources than when using these other languages. The language does not guarantee efficient code automatically, but provides a toolchest that aids programmers in the pursuit of efficiency. Sloppy C++ code may be no more efficient than higher-level implementations of the same algorithms, but a good C++ programmer with knowledge of the language can write software that is efficient from the first cut and then optimize the code further. This book provides techniques for writing efficient code and for optimizing existing software.
Often, there is no single solution to a programming problem that is optimal for all cases. Thus, optimization generally does not mean writing optimally performing software; rather, it means incrementally changing (refactoring) software to increase its performance, bringing it closer to the optimum.
Such optimization requires that the software source is written in a sufficiently modular way that performance critical parts can be isolated. With suitably written code, it then requires the use of tools, libraries, knowledge and time to change those parts in a way that increases the overall execution speed of the software.
Nowadays, many optimizations are already performed by compilers and are no longer the programmer's burden. This book discusses higher-level optimizations that present-day compilers are not (yet) able to perform.
This book is aimed at readers already familiar with the C++ language, who want to use it to develop high performance application software or software libraries.
Almost all of the optimization techniques presented are platform independent and, therefore, there will be few references to specific operating systems, processor architectures or compilers. However, some of the presented techniques may be ineffective or inapplicable in some combinations of operating system/processor/compiler.