Book Image

The Art of Writing Efficient Programs

By : Fedor G. Pikus
3 (2)
Book Image

The Art of Writing Efficient Programs

3 (2)
By: Fedor G. Pikus

Overview of this book

The great free lunch of "performance taking care of itself" is over. Until recently, programs got faster by themselves as CPUs were upgraded, but that doesn't happen anymore. The clock frequency of new processors has almost peaked, and while new architectures provide small improvements to existing programs, this only helps slightly. To write efficient software, you now have to know how to program by making good use of the available computing resources, and this book will teach you how to do that. The Art of Efficient Programming covers all the major aspects of writing efficient programs, such as using CPU resources and memory efficiently, avoiding unnecessary computations, measuring performance, and how to put concurrency and multithreading to good use. You'll also learn about compiler optimizations and how to use the programming language (C++) more efficiently. Finally, you'll understand how design decisions impact performance. By the end of this book, you'll not only have enough knowledge of processors and compilers to write efficient programs, but you'll also be able to understand which techniques to use and what to measure while improving performance. At its core, this book is about learning how to learn.
Table of Contents (18 chapters)
1
Section 1 – Performance Fundamentals
7
Section 2 – Advanced Concurrency
11
Section 3 – Designing and Coding High-Performance Programs

Why data sharing is expensive

As we have just seen, concurrent (simultaneous) access of the shared data is a real performance killer. Intuitively, it makes sense: in order to avoid a data race, only one thread can operate on the shared data at any given time. We can accomplish this with a mutex or use an atomic operation if one is available. Either way, when one thread is, say, incrementing the shared variable, all other threads have to wait. Our measurements in the last section confirm it.

However, before taking any action based on observations and experiments, it is critically important to understand precisely what we measured and what can be concluded with certainty.

It is easy to describe what was observed: incrementing a shared variable from multiple threads at the same time does not scale at all and, in fact, is slower than using just one thread. This is true for both atomic shared variables and non-atomic variables guarded by a mutex. We have not tried to measure unguarded...