Computer Systems:

A Programmer’s Perspective
计算机系统
周学海
[email protected]
0551-63492149
中国科学技术大学
 Welcome to ……
• 主讲：周学海([email protected])
• 地点1: 高新区1号学科楼A430A（周五）
• 地点2：西区（高能效智能计算实验室）

智能计算系统
基于NPU的异构计 高性能实时计算
算系统 系统
异构计 时间可预测的异
构计算系统
算系统

可重构计算系统
基于FPGA的异构计
算系统

9/14/2024 xhzhou@USTC 2
 Welcome to ……
• 助教：
姓名 电子邮件
张亦博 [email protected]
陈兴焱 [email protected]

• 课程主页:
– bb.ustc.edu.cn
• QQ群号：564480202

9/14/2024 xhzhou@USTC 3
 第1章 Introduction

• 计算机系统？
– 系统、模型、物理世界
– 模型vs.现实，抽象vs.具体
– 深入理解计算机系统的重要性
• 举例：5方面具体问题

• 本课程主要内容
– Builder-Centric vs. Programming-Centric

9/14/2024 Bryant and O'Hallaron, Computer Systems: A Programmer's Perspective, Third Edition 4
 System
• 系统的概念：
– ISO/IEC/IEEE15288：为了达到一个或多个所描述的目标由相互作
用的要素构成的集合。是 、用于在所定义的环境中提供
产品或服务，以造福于用户和其他利益相关者
– The International Council on Systems Engineering (INCOSE)：
完成所定义目标的要素、子系统或组件的集合，这些要素包括产品、
过程、人员、信息、技术、设施、服务和其他支持要素
• In Summary: 系统并不仅仅是一些要素的简单集合，是一个由
一组相互关联的要素构成的、能够实现某个（些）目标的整体。
Which of the following would be considered a system?

p A box full of headphones

p The houses in a neighborhood
p The list of text messages exchanged between two people
p A jet airplane with components to assist and control the plane

9/14/2024 Bryant and O'Hallaron, Computer Systems: A Programmer's Perspective, Third Edition 5
 Definition of System

9/14/2024 Bryant and O'Hallaron, Computer Systems: A Programmer's Perspective, Third Edition 6
 What are the properties that characterize a system?

• 系统具有层次结构：
– 由若干相互作用要素（部件）组成，
• 系统是一个整体：
– 其性质不仅是它的 ，而且也是 称为emergency（涌现）
• 系统在由其运行环境和其他系统组成的 中运行：
系统的外部视图包含不属于系统但确实与系统交
互的元素。这些元素的集合称为”操作环境或上下文“（包括系统的使用者）
– system boundary (系统边界）：系统内部和外部视图产生了系统边界的概念。
• 系统的功能（functionality）：
– 通常用系统与其操作环境（特别是用户）的交互来表示。
• 系统结构（System Architecture）：
– 系统在其环境中的基本概念或性质。体现在其要素、关系以及其设计和进化的原则中。
• 与系统使用相关的质量属性：
– 例如可维护、可靠、互操作性等。系统有生命周期，从它的初步和概念阶段到它的（退出应用）消亡

9/14/2024 Bryant and O'Hallaron, Computer Systems: A Programmer's Perspective, Third Edition 7
 Model
• 我们所认识的系统：物理世界的实际系统的“心
里表征”

• Model：现实世界系统的一种抽象
– 模型是在物理世界中可以实现的
– 模型是为某一目的而开发的感兴趣系统的
– 一个模型应该解决 的关切/需求，
并对系统的工程具有明确的有用性
• Model的特性：
– Mapping：是实体的一种 。
– Reduction：反映了实体的 特性。
– Pragmatic：模型（模拟）要能完成某些特定的功
能，以反应实体的特性。

9/14/2024 Bryant and O'Hallaron, Computer Systems: A Programmer's Perspective, Third Edition 8
 Computer System
• 计算机系统：自动进行信息存储、处理和交换
的系统
• 设计和理解复杂的计算机系统：模块化、抽象
化、层次化、分级化
• MBSE：Model-based System
Engineering

• 计算机系统可以有不同层次（视角）的抽象，
– 不同层级的抽象
– 不同视角的抽象

现代计算机系统的抽象层次

9/14/2024 Bryant and O'Hallaron, Computer Systems: A Programmer's Perspective, Third Edition 9
 Computer Systems: A Programmer’s
Perspective

• 大多数计算机课程都强调抽象
– Abstract data types
– Asymptotic analysis
• 抽象是有局限性的
– Especially in the presence of bugs
– Need to understand underlying implementations
– Sometimes the abstract interfaces don’t provide the level of control or
performance you need
• 通过学习这门课程，我们可以：
– Become more effective programmers
• Able to find and eliminate bugs efficiently
• Able to understand and tune program performance
9/14/2024 Bryant and O'Hallaron, Computer Systems: A Programmer's Perspective, Third Edition 10
 Great Reality #1
#1: Int，Float并不等价于数学中的整数和实数
• Example1：Is x2 ≥ 0?
– Float’s: Yes!

– Int’s:
• 40000 * 40000 --> 1600000000
• 50000 * 50000 -->
• Example2: Is (x + y) + z = x + (y + z)?
– Unsigned & Signed Int’s: Yes!
– Float’s:
• (1e20 + -1e20) + 3.14 --> 3.14
• 1e20 + (-1e20 + 3.14) --> 0

9/14/2024 Bryant and O'Hallaron, Computer Systems: A Programmer's Perspective, Third Edition 11
 Computer Arithmetic
• 计算机算术运算是离散量间的（有模）运算
– Arithmetic operations have important mathematical properties
• 不能假定计算机的算术运算满足“通常”的数学属性
– Due to finiteness of representations
– Integer operations
• satisfy “ring” properties：Commutativity, associativity, distributivity

– Floating point operations

• satisfy “ordering” properties: Monotonicity, values of signs

• 给我们带来的启示:
– Need to understand which abstractions apply in which contexts
– Important issues for compiler writers and serious application
programmers
9/14/2024 Bryant and O'Hallaron, Computer Systems: A Programmer's Perspective, Third Edition 12
 Code Security Example
/* Kernel memory region holding user-accessible data */
#define KSIZE 1024
char kbuf[KSIZE];

/* Copy at most maxlen bytes from kernel region to user buffer */

int copy_from_kernel(void *user_dest, int maxlen) {
/* Byte count len is minimum of buffer size and maxlen */
int len = KSIZE < maxlen ? KSIZE : maxlen;
memcpy(user_dest, kbuf, len);
return len;
}

• 类似于在FreeBSD的getpeername实现中的代码
• 有许多人都在试图找出程序中的漏洞

9/14/2024 Bryant and O'Hallaron, Computer Systems: A Programmer's Perspective, Third Edition 13
 Typical Usage
/* Kernel memory region holding user-accessible data */
#define KSIZE 1024
char kbuf[KSIZE];

/* Copy at most maxlen bytes from kernel region to user buffer */

int copy_from_kernel(void *user_dest, int maxlen) {
/* Byte count len is minimum of buffer size and maxlen */
int len = KSIZE < maxlen ? KSIZE : maxlen;
memcpy(user_dest, kbuf, len);
return len;
}

#define MSIZE 528

typedef size_t
void getstuff() {
#include <string.h>
char mybuf[MSIZE];
void *memcpy(void *dest, const void
copy_from_kernel(mybuf, MSIZE); &src, size_t n);
printf(“%s\n”, mybuf);
}

9/14/2024 Bryant and O'Hallaron, Computer Systems: A Programmer's Perspective, Third Edition 14
 Malicious Usage
/* Kernel memory region holding user-accessible data */
#define KSIZE 1024
char kbuf[KSIZE];

/* Copy at most maxlen bytes from kernel region to user buffer */

int copy_from_kernel(void *user_dest, ) {
/* Byte count len is minimum of buffer size and maxlen */
int len = KSIZE < maxlen ? KSIZE : maxlen;
memcpy(user_dest, kbuf, len);
return len;
}

#define MSIZE 528

typedef size_t
void getstuff() {
char mybuf[MSIZE]; #include <string.h>
copy_from_kernel(mybuf, -MSIZE); void *memcpy(void *dest, const void
&src, size_t n);
. . .
}

9/14/2024 Bryant and O'Hallaron, Computer Systems: A Programmer's Perspective, Third Edition 15
 Great Reality #2
#2: 你必须理解熟悉汇编语言
• 虽然可能您永远不会用汇编语言编写程序
– Compilers are much better & more patient than you are
• 但是熟悉汇编代码是理解机器底层执行模型(machine-level execution
model）的关键
– Behavior of programs in presence of bugs
• High-level language model breaks down
– Tuning program performance
• Understand optimizations done/not done by the compiler
• Understanding sources of program inefficiency
– Implementing system software
• Compiler has machine code as target
• Operating systems must manage process state
– Creating/fighting malware
• x86 assembly is the language of choice!

9/14/2024 Bryant and O'Hallaron, Computer Systems: A Programmer's Perspective, Third Edition 16
 Assembly Code Example
• Time Stamp Counter
– Special 64-bit register in Intel-compatible machines
– Incremented every clock cycle
– Read with rdtsc instruction （Read Time-Stamp Counter）
• Application
– Measure time required by procedure
– In units of clock cycles

double t;
start_counter();
P();
t = get_counter();
printf("P required %f clock cycles\n", t);

9/14/2024 Bryant and O'Hallaron, Computer Systems: A Programmer's Perspective, Third Edition 17
 Code to Read Counter
• 使用GCC的内联asm编写少量的汇编代码
• 编译器将汇编代码插入生成的机器代码中

static unsigned cyc_hi = 0;

static unsigned cyc_lo = 0;

/* Set *hi and *lo to the high and low order bits
of the cycle counter.
*/
void access_counter(unsigned *hi, unsigned *lo)
{
asm("rdtsc; movl %%edx,%0; movl %%eax,%1"
: "=r" (*hi), "=r" (*lo)
:
: "%edx", "%eax");
}

9/14/2024 Bryant and O'Hallaron, Computer Systems: A Programmer's Perspective, Third Edition 18
 Code to Read Counter
/* Record the current value of the cycle counter. */
void start_counter()
{
access_counter(&cyc_hi, &cyc_lo);
}

/* Number of cycles since the last call to start_counter. */

double get_counter()
{
unsigned ncyc_hi, ncyc_lo;
unsigned hi, lo, borrow;
/* Get cycle counter */
access_counter(&ncyc_hi, &ncyc_lo);
/* Do double precision subtraction */
lo = ncyc_lo - cyc_lo;
borrow = lo > ncyc_lo;
hi = ncyc_hi - cyc_hi - borrow;
return (double) hi * (1 << 30) * 4 + lo;
}
9/14/2024 Bryant and O'Hallaron, Computer Systems: A Programmer's Perspective, Third Edition 19
 Measuring Time
• 比看起来要复杂：差异的来源有很多因素造成的
• 例如：Sum integers from 1 to n

n cycle Cycles/n
100 961 9.61
1000 8,407 8.41
1,000 8,426 8.43
10,000 82,861 8.29
10,000 82,876 8.29
1,000,000 8,419,907 8.42
1,000,000 8,425,181 8.43
1,000,000,000 8,371,2305,591 8.37

9/14/2024 Bryant and O'Hallaron, Computer Systems: A Programmer's Perspective, Third Edition 20
 Great Reality #3 Memory Matters

#3: 计算机中随机读写存储器是一种非物理的抽象概念
（与现实世界有差距）
• 内存不是非受限的（unbounded)
– It must be allocated and managed
– Many applications are memory dominated
• 内存性能不是统一一致的
– Cache and virtual memory effects can greatly affect program
performance
– Adapting program to characteristics of memory system can
lead to major speed improvements
• 尤其内存引用错误非常有害
– Effects are distant in both time and space

9/14/2024 Bryant and O'Hallaron, Computer Systems: A Programmer's Perspective, Third Edition 21
 Memory Referencing Bug Example

double fun(int i)
{
volatile double d[1] = {3.14};
volatile long int a[2];
a[i] = 1073741824; /* Possibly out of bounds */
return d[0];
}

fun(0) ➙ 3.14
fun(1) ➙ 3.14
fun(2) ➙ 3.1399998664856
fun(3) ➙ 2.00000061035156
fun(4) ➙ 3.14, then segmentation fault

• 其结果与特定的体系结构相关

9/14/2024 Bryant and O'Hallaron, Computer Systems: A Programmer's Perspective, Third Edition 22
 Memory Referencing Bug Example
double fun(int i)
{
volatile double d[1] = {3.14};
volatile long int a[2];
a[i] = 1073741824; /* Possibly out of bounds */
return d[0];
}

fun(0) ➙ 3.14
fun(1) ➙ 3.14
fun(2) ➙ 3.1399998664856
fun(3) ➙ 2.00000061035156
fun(4) ➙ 3.14, then segmentation fault
Explanation: Saved State 4
d7 ... d4 3
d3 ... d0 2 Location accessed by
fun(i)
a[1] 1
a[0] 0
9/14/2024 Bryant and O'Hallaron, Computer Systems: A Programmer's Perspective, Third Edition 23
 Memory Referencing Errors
• C和c++不提供任何内存保护
– Out of bounds array references
– Invalid pointer values
– Abuses of malloc/free
• 内存不当引用会导致bug（nasty bugs）
– Whether or not bug has any effect depends on system and compiler
– Action at a distance
• Corrupted object logically unrelated to one being accessed
• Effect of bug may be first observed long after it is generated
• 我们该如何应对?
– Program in Java,Ruby, or ML
– Understand what possible interactions may occur
– Use or develop tools to detect referencing errors (e.g. Valgrind)
– .....
9/14/2024 Bryant and O'Hallaron, Computer Systems: A Programmer's Perspective, Third Edition 24
 Great Reality #4:
#4：算法的不同实现，其性能也会有较大差别
• 渐进复杂度分析中的常数因子也很重要
• 甚至精确的操作数量也不能精确预测性能
– Easily see 10:1 performance range depending on how code written
– Must optimize at multiple levels: algorithm, data representations,
procedures, and loops
• 必须了解系统才能更好地优化性能
– How programs compiled and executed
– How to measure program performance and identify bottlenecks
– How to improve performance without destroying code modularity and
generality

9/14/2024 Bryant and O'Hallaron, Computer Systems: A Programmer's Perspective, Third Edition 25
 Memory System Performance Example

void copyij(int src[2048][2048], void copyji(int src[2048][2048],

int dst[2048][2048]) int dst[2048][2048])
{ {
int i,j; int i,j;
for (i = 0; i < 2048; i++) for (j = 0; j < 2048; j++)
for (j = 0; j < 2048; j++) for (i = 0; i < 2048; i++)
dst[i][j] = src[i][j]; dst[i][j] = src[i][j];
} }

4.3ms 81.8ms

(2.0 GHz Intel Core i7 Haswell)

19+ times slower!

• 存储按照分层结构组织
• 性能取决于访问模式
– Including how step through multi-dimensional array

9/14/2024 Bryant and O'Hallaron, Computer Systems: A Programmer's Perspective, Third Edition 26
 Why The Performance Differs

9/14/2024 Bryant and O'Hallaron, Computer Systems: A Programmer's Perspective, Third Edition 27
 Memory Performance Example
• 矩阵-矩阵乘法的实现
– Multiple ways to nest loops

/* ijk */ /* ikj */
for (i=0; i<n; i++) { for (i=0; i<n; i++) {
for (j=0; j<n; j++) { for (k=0; k<n; k++) {
sum = 0.0; r = a[i][k];
for (k=0; k<n; k++) for (j=0; j<n; j++)
sum += a[i][k] * b[k][j]; c[i][j] += r * b[k][j];
c[i][j] += sum; }
} }
}

9/14/2024 Bryant and O'Hallaron, Computer Systems: A Programmer's Perspective, Third Edition 28
 Example Matrix Multiplication
Matrix-Matrix Multiplication (MMM) on 2 x Core 2 Duo 3 GHz (double precision)
Gflop/s

Best code (K. Goto)

160x

• Standard desktop computer, vendor compiler, using optimization flags

• Both implementations have exactly the same operations count (2n3)
• What is going on?
9/14/2024 Bryant and O'Hallaron, Computer Systems: A Programmer's Perspective, Third Edition 29
 Memory System

9/14/2024 Bryant and O'Hallaron, Computer Systems: A Programmer's Perspective, Third Edition 30
 Great Reality #5
#5: 计算机不仅仅是执行计算任务
• 计算机系统需要输入和输出数据
– I/O系统对程序的可靠性和性能至关重要
• 计算机系统通过网络互联并相互通信
– 许多系统级问题出现在网络环境中
• Concurrent operations by autonomous processes
• Coping with unreliable media
• Cross platform compatibility
• Complex performance issues

9/14/2024 Bryant and O'Hallaron, Computer Systems: A Programmer's Perspective, Third Edition 31
 Hardware Organization (Naïve)

9/14/2024 Bryant and O'Hallaron, Computer Systems: A Programmer's Perspective, Third Edition 32
 第1章 Introduction

• 计算机系统？
– 系统、模型、物理世界
– 模型vs.现实，抽象vs.具体
– 深入理解计算机系统的重要性:
• 举例：5方面具体问题
• 本课程主要内容
– Builder-Centric vs. Programming-Centric

9/14/2024 Bryant and O'Hallaron, Computer Systems: A Programmer's Perspective, Third Edition 33
 Course Perspective

• 大多数系统课程是Builder-Centric
• Computer Architecture
– Design pipelined processor in Verilog
• Operating Systems
– Implement large portions of operating system
• Compilers
– Write compiler for simple language
• Networking
– Implement and simulate network protocols

9/14/2024 Bryant and O'Hallaron, Computer Systems: A Programmer's Perspective, Third Edition 34
 Course Perspective (Cont.)

• 这门课： Programmer-Centric
• 通过更多地了解底层系统，您可以成为更有效率的程
序员
– 编写更可靠和高效的程序
– 在操作系统中加入需要使用钩子的特性
• 允许软件开发者能够更深入地与操作系统交互，定制和
扩展其功能。
• E.g., concurrency, signal handlers

• ......
9/14/2024 Bryant and O'Hallaron, Computer Systems: A Programmer's Perspective, Third Edition 35
 关于这门课……
• 上课地点：G3-113
• 上课时间：2-18周 周五(3,4,5)
• 参考书：
– Randal E. Bryant and David R. O’Hallaron,
“Computer Systems: A Programmer’s
Perspective”, 2015第3版
（http://csapp.cs.cmu.edu/
– 中译本，深入理解计算机系统，龚亦利等译，机
械工业出版社(第3版,2016)
– Samuel P. Harbison III and Guy L. Steele Jr., “C
A Reference Manual 5th Edition”,Prentice Hall,
2002

9/14/2024 Bryant and O'Hallaron, Computer Systems: A Programmer's Perspective, Third Edition 36
 Grading

• 实验 : 50%
• 平时作业: 10%
• 期末考试: 40%
• ......

9/14/2024 Bryant and O'Hallaron, Computer Systems: A Programmer's Perspective, Third Edition 37
 About the Textbook
• 作者:
– Randal E. Bryant: CMU, Professor, ACM Fellow & IEEE Fellow
– O’Hallaron: CMU，Professor
• 教材结构:
1. Introduction
2. Representing and Manipulating Information
3. Machine-Level Representing of Programs
4. Processor Architecture 程序结构和运行
5. Optimizing Program Performance
6. The Memory Hierarchy
7. Linking
8. Exceptional Control Flow 在系统上运行
9. Virtual Memory
10. System-Level I/O
11. Network Programming 程序间的交互和通信
12. Concurrent Programming
9/14/2024 Bryant and O'Hallaron, Computer Systems: A Programmer's Perspective, Third Edition 38
 Course Structure
• Lectures:
– Introduction (chap. 1)
– Information Representing (chap. 2)
– Machine Language (chap. 3)
– Processor Architecture (chap. 4)
– Code Optimization (chap. 5)
– Memory Hierarchy (chap. 6）
– Linking (chap.7)
– Exception Control(chap.8)
– Virtual Memory(Chap.9)
– I/O(Chap.10)
– Network Programming(Chap.11)
– Concurrent Programming(Chap.12)

9/14/2024 Bryant and O'Hallaron, Computer Systems: A Programmer's Perspective, Third Edition 39
 Lecture topics and assignments
• 程序和数据
• 主要内容
– Bits operations, arithmetic, assembly language programs
– Representation of C control and data structures

• 实验
– L1: Manipulating bits
– L2: Defusing a binary bomb
– L3: Hacking a buffer bomb

9/14/2024 Bryant and O'Hallaron, Computer Systems: A Programmer's Perspective, Third Edition 40
 The Memory Hierarchy

• 存储层次
• 主要内容
– Memory technology, memory hierarchy, caches, disks,
locality

• 实验
– L4 (Perflab): Optimize the performance of an application
kernel function.
• Learn how to exploit locality in your programs.

9/14/2024 Bryant and O'Hallaron, Computer Systems: A Programmer's Perspective, Third Edition 41
 Linking and Exceptional Control Flow

• 链接和异常控制流
• 主要内容
– Object files, static and dynamic linking, libraries,
loading
– Hardware exceptions, processes, process control,
Unix signals, nonlocal jumps

9/14/2024 Bryant and O'Hallaron, Computer Systems: A Programmer's Perspective, Third Edition 42
 Virtual memory

• 虚拟存储
• 主要内容
– Virtual memory, address translation, dynamic storage
allocation

• 实验
– L5: Writing your own malloc package
• Get a real feel for systems programming

9/14/2024 Bryant and O'Hallaron, Computer Systems: A Programmer's Perspective, Third Edition 43
 I/O, Networking, and Concurrency

• 系统I/O、网络、并发
• 主要内容
– High level and low-level I/O, network
programming
– Internet services, Web servers
– concurrency, concurrent server design, threads
– I/O multiplexing with select.

9/14/2024 Bryant and O'Hallaron, Computer Systems: A Programmer's Perspective, Third Edition 44
 L1:Data Lab
• 实验要求：
– Students implement simple logical and
arithmetic functions, but using a highly restricted
subset of C.
– For example, they must compute the absolute
value of a number using only bit-level operations.
• 实验目的：
– This lab helps students understand the bit-level
representations of C data types and the bit-level
behavior of the operations on data.

9/14/2024 Bryant and O'Hallaron, Computer Systems: A Programmer's Perspective, Third Edition 45
 L2:Bomb Lab
• 实验要求：
– A "binary bomb" is a program provided to students as an object
code file.
– When run, it prompts the user to type in 6 different strings. If any
of these is incorrect, the bomb ``explodes,'' printing an error
message and logging the event on a grading server.
– Students must ``defuse'' their own unique bomb by
disassembling and reverse engineering the program to
determine what the 6 strings should be.
• 实验目的：
– The lab teaches students to understand assembly language, and
also forces them to learn how to use a debugger. It's also great
fun. A legendary lab among the CMU undergrads.

9/14/2024 Bryant and O'Hallaron, Computer Systems: A Programmer's Perspective, Third Edition 46
 L3: Buffer Lab

• 实验要求：
– Students modify the run-time behavior of a
binary executable by exploiting a buffer overflow
bug.
• 实验目的：
– This lab teaches the students about the stack
discipline and teaches them about the danger of
writing code that is vulnerable to buffer overflow
attacks.

9/14/2024 Bryant and O'Hallaron, Computer Systems: A Programmer's Perspective, Third Edition 47
 L4:Performance Lab

• 实验要求：
– Students must optimize the performance of an
application kernel function such as convolution
or matrix transposition.
• 实验目的：
– This lab provides a very clear demonstration of
the properties of cache memories, and gives
students experience with low-level program
optimization.

9/14/2024 Bryant and O'Hallaron, Computer Systems: A Programmer's Perspective, Third Edition 48
 L5: Malloc Lab

• 实验要求：
– Students implement their own versions of malloc,
free, and realloc.
• 实验目的：
– This lab gives students a clear understanding of
data layout and organization, and requires them
to evaluate different trade-offs between space
and time efficiency.
– One of our favorite labs. When students finish
this one, they really understand pointers!

9/14/2024 Bryant and O'Hallaron, Computer Systems: A Programmer's Perspective, Third Edition 49
 Acknowledgements

• This course was developed and fine-tuned by

Randal E. Bryant and David O’Hallaron. They wrote
The Book!
• http://www.cs.cmu.edu/~./213/schedule.html

9/14/2024 Bryant and O'Hallaron, Computer Systems: A Programmer's Perspective, Third Edition 50