提升性能的C++优化技巧：Kurt Guntheroth指南

Performance

需积分: 9 142 浏览量更新于2023-05-25 收藏 5.28MB PDF 举报

身份认证购VIP最低享 7 折!

领优惠券(最高得80元）

资源详情

资源推荐

illustrated using Visual Studio on Windows apply equally to Linux, Mac OS X, or any

other C++ environment. However, the precise timings of different optimizations

depend on the compiler and standard library implementation, and the processor on

which the code is tested. Optimization is an experimental science. Taking optimiza‐

tion advice on faith is fraught with negative surprises.

I am aware that compatibility with various other compilers, and with other Unix and

embedded systems, can be challenging, and I apologize if the code does not compile

on your favorite system. Since this book is not about cross-system compatibility, I

have erred on the side of presenting simple code.

The curly-brace indentation style shown here is not my favorite:

if (bool_condition) {

controlled_statement();

}

However, because it has the advantage of putting the most lines possible on the

printed page, I have chosen to use it for the examples throughout this book.

Using Code Examples

Supplemental material (code examples, sample solutions, etc.) is available for down‐

load at www.guntheroth.com.

This book is here to help you get your job done. In general, if example code is offered

with this book, you may use it in your programs and documentation. You do not

need to contact us for permission unless you’re reproducing a significant portion of

the code. For example, writing a program that uses several chunks of code from this

book does not require permission. Selling or distributing a CD-ROM of examples

from O’Reilly books does require permission. Answering a question by citing this

book and quoting example code does not require permission. Incorporating a signifi‐

cant amount of example code from this book into your product’s documentation does

require permission.

We appreciate, but do not require, attribution. An attribution usually includes the

title, author, publisher, and ISBN. For example: “Optimized C++ by Kurt Guntheroth

If you feel your use of code examples falls outside fair use or the permission given

above, feel free to contact us at permissions@oreilly.com.

Preface | xvii

CHAPTER 1

An Overview of Optimization

The world has a stupendous appetite for computation. Whether the code you are

writing runs on a watch, phone, tablet, workstation, supercomputer, or planet-

spanning network of data centers, there are many programs that must run flat-out all

the time. So it may not be enough merely to accurately convert the cool idea in your

head into lines of code. It may not be enough even to comb the code for defects until

it runs correctly all the time. Your application may be sluggish on the kind of hard‐

ware your customers can afford. Your hardware team may have saddled you with a

tiny processor to meet their power consumption goals. You may be battling a compet‐

itor over throughput or frames per second. Or you may be building out to planetary

scale and just slightly nervous about boiling the oceans. Enter optimization.

This book is about optimization—specifically, optimizing C++ programs, with partic‐

ular reference to patterns of behavior of C++ code. Some of the techniques in this

book are applicable to other programming languages, but I have made no attempt to

explain the techniques in a universal way. Some optimizations that are effective on

C++ code have no effect or are simply not possible in other languages.

This book is about taking correct code that embodies best practices of C++ design,

and changing it into correct code that still embodies good C++ design but also runs

faster and consumes fewer resources on pretty much any computer. Many opportuni‐

ties for optimization occur because some C++ features, used casually, run slowly and

consume many resources. Such code, while correct, is ill-considered, given just a little

general knowledge about modern microprocessor devices or a little thought about the

costs of various C++ constructs. Other optimizations are available because of the fine

control C++ offers over memory management and copying.

This book is not about laboriously coding assembly language subroutines, counting

clock cycles, or learning how many instructions Intel’s latest silicon stylings can dis‐

patch concurrently. There exist developers who work with a single platform (the

Xbox is a good example) for several years, and have the time and the need to master

these dark arts. However, the great majority of developers target phones, tablets, or

PCs, which contain an endless variety of microprocessor chips—some of which have

not yet even been designed. Developers of software embedded into products also face

a variety of processors with widely varying architectures. Attempting to learn pro‐

cessor arcana will make most developers crazy and paralyze them with indecision. I

don’t recommend this path. Processor-dependent optimization is simply not fruitful

for most applications, which must by definition run on a variety of processors.

This book is likewise not about learning the fastest operating system–dependent way

to do some operation on Windows, and again on Linux, OS X, and every embedded

operating system. It is about what you can do in C++, including with the C++ stan‐

dard library. Breaking out of C++ to perform an optimization may make it hard for

peers to review or comment on the optimized code. It’s not an action to be taken

lightly.

This book is about learning how to optimize. Any static catalog of techniques or func‐

tions is doomed, as new algorithms are discovered and new language features become

available. Instead, this book provides several running examples of how code may be

improved incrementally, so that the reader becomes familiar with the code tuning

process and develops the mindset that makes optimization fruitful.

This book is about optimizing the coding process as well. Developers mindful of the

runtime cost of their code can write code that is efficient from the start. With prac‐

tice, writing fast code usually takes no longer than writing slow code.

Finally, this book is about performing miracles; about checking in a change and later

hearing a colleague exclaim in surprise, “Wow, what happened? It just started right

up. Who fixed something?” Optimization is something you can also do to your status

as a developer and your personal pride in your craft.

Optimization Is Part of Software Development

Optimization is a coding activity. In traditional software development processes, opti‐

mization takes place after Code Complete, during the integration and testing phase of

a project, when the performance of a whole program can be observed. In an Agile

development process, one or more sprints may be allocated to optimization after a

feature with performance goals is coded, or as needed to meet performance targets.

The goal of optimization is to improve the behavior of a correct program so that it

also meets customer needs for speed, throughput, memory footprint, power con‐

sumption, and so on. Optimization is thus as important to the development process

as coding features is. Unacceptably poor performance is the same kind of problem for

users as bugs and missing features.

2 | Chapter 1: An Overview of Optimization

One important difference between bug fixing and performance tuning is that perfor‐

mance is a continuous variable. A feature is either coded, or it is not. A bug is either

present or absent. However, performance can be very bad or very good or somewhere

in between. Optimization is also an iterative process in which each time the slowest

part of the program is improved, a new slowest part appears.

Optimization is very much an experimental science, calling for a scientific mindset to

a greater degree than some other coding tasks. To be successful at optimization

requires observing behavior, forming testable hypotheses based on these observa‐

tions, and performing experiments that result in measurements that either support or

refute the hypotheses. Experienced developers often believe they have valid experi‐

ence and valid intuitions about optimal code. But unless they frequently test their

intuitions, they will frequently be wrong. My personal experiences writing the test

programs for this book revealed several results that contradicted my intuitions.

Experimentation, rather than intuition, is a theme of this book.

Optimization Is

Eective

It is difficult for developers to reason about the effects of individual coding decisions

on the overall performance of a large program. Thus, practically all complete pro‐

grams contain significant opportunities for optimization. Even code produced by

experienced teams with plenty of time can often be sped up by a factor of 30%–100%.

For more rushed or less experienced teams, I have seen performance improvements

of 3 to 10 times. A speedup of more than 10x by tweaking code is less likely. However,

selection of a better algorithm or data structure can be the difference between

whether a feature can be deployed or is infeasibly slow.

It’s OK to Optimize

Many treatments of optimization begin with a stern warning: don’t! Don’t optimize,

and if you have to optimize, don’t do it until the end of the project, and don’t do any

more optimizing than you must. For instance, famous computer scientist Donald

Knuth said of optimization:

We should forget about small efficiencies, say about 97 percent of the time: premature

optimization is the root of all evil.

—Donald Knuth, Structured Programming with go to Statements, ACM Computing

Surveys 6(4), December 1974, p268. CiteSeerX: 10.1.1.103.6084

Optimization Is Eective | 3

剩余383页未读，继续阅读

nn123456789

粉丝: 14
资源: 127

会员权益专享

提升性能的C++优化技巧：Kurt Guntheroth指南

Optimized C++ 2016新书 无水印pdf 0分

Optimized C++：Proven.Techniques.for.Heightened.Performance.2016.4.pdf

Optimized C++(O'Reilly,2016)

Link library type specifier "debug" is followed by specifier "optimized" instead of a library name. The first specifier will be ignored.

boost pdf

deallocate (this=0x2397e038, __p=<optimized out>) at /usr/include/c++/4.8.2/ext/new_allocator.h:110

c++ whether

gdb出现<optimized out>

ceres 世界坐标系转图像坐标系 C++

刚刚代码 这个函数 def get_optimized_file_typ 放哪个文件

vs在代码中判断c++优化是否打开

刚刚的这个get_optimized_file_type如何在上传文件的时候使用

cmake的optimized用法

=<optimized out>,

Boot Optimized Storage Solution

value has been optimized out

File "frag-smart2.py", line 289, in <module> ase.io.write("01_1/optimized_POSCAR", box.get_ase_atoms()) AttributeError: 'Mushybox' object has no attribute 'get_ase_atoms'

插入排序（减常数）c++

File "frag-smart2.py", line 289, in <module> ase.io.write("01_1/optimized_POSCAR", box.to_atoms()) AttributeError: 'Mushybox' object has no attribute 'to_atoms'

会员权益专享

最新资源

Optimized C++ 2016新书无水印pdf 0分

刚刚代码这个函数 def get_optimized_file_typ 放哪个文件