Chapter1 Computer Abstractions and Technology

寫過論文的人對abstraction一定不會陌生；不錯(cuò)，本章就是一些綜述的內(nèi)容，對于熱衷技術(shù)的人而言肯定乏味的很，但作為一個(gè)行業(yè)管理者而言，可謂字字珠璣啊；

1.1 introduction

首先介紹了計(jì)算機(jī)科學(xué)在廣闊領(lǐng)域的發(fā)展，主要是如下領(lǐng)域：

• automobiles
• cell phones
• Human genome project
• world wide web
• search engines

然后總結(jié)傳統(tǒng)計(jì)算機(jī)應(yīng)用及其特點(diǎn)，主要分為3個(gè)領(lǐng)域：

• PCs
• Servers
• Embedded computers

最后歡迎大家來到后PC時(shí)代，介紹后PC時(shí)代的特點(diǎn)：

• 最大特點(diǎn)就是PMD替換PC
• PMD：Personal Mobile Device

那么讀者可以從這本書學(xué)習(xí)到什么呢？答案如下：

• 高級軟件語言（C,JAVA）如何轉(zhuǎn)為硬件能理解的語言，硬件如何執(zhí)行程序；
• 軟硬件接口是什么，軟件如何讓硬件執(zhí)行特定功能；
• 什么決定了程序的性能，怎么提高性能；
• 硬件工程師可利用什么技術(shù)提高性能；
• 硬件工程師可利用什么技術(shù)提高energy efficiency；
• Parallelism的原因及其后續(xù)演進(jìn)；
• 現(xiàn)代計(jì)算機(jī)架構(gòu)中個(gè)great ideas；

1.2 Eight Great Ideas in Computer Architecture.

過去60年，計(jì)算機(jī)架構(gòu)出現(xiàn)了8個(gè)great ideas

• Design for Moore's Law
• Use Abstraction to Simplify Design
• Make the Common Case Fast
• Performance via parallelism
• Performance via Pipelining
• Performance via Prediction
• Hierarchy of Memories
• Dependability via Redundancy

1.3 Below Your Program

Application software在system software之上，而system software又在hardware之上；

有兩種system software：

1. operating system
2. compiler
   將高級語言編寫的程序翻譯成機(jī)器能執(zhí)行的指令；

From a High-Level Language to the Language of Hardware
機(jī)器所能理解的語言叫做
instruction：通常為binary形式的，如1001010100101110，此指令為兩個(gè)數(shù)相加；

1.4 Under the Covers

以ipad2為例，從LCD到CPU的硬件簡介。

1.5 Technologies for Buildig Processors and Memory

簡單介紹半導(dǎo)體工藝
the cost of an integrated circuit can be expressed in 3 simple equations:
Cost per die=Cost per water/(Dies per wafer x yield)
Dies per wafer=wafer area /Die area
Yield=1/(1+(Defects per area x Die area/2))^2

1.6 Performance

• Defining Performance
• Measuring Performance
  • time is the measure of computer performance，時(shí)間有不同的定義方式：
    • wall clock time 經(jīng)過時(shí)間
    • response time 響應(yīng)時(shí)間
    • elapsed time 運(yùn)行時(shí)間
  • CPU execution time/CPU time
    • user CPU time:單純執(zhí)行程序的時(shí)間
    • system CPU time：為了執(zhí)行程序而調(diào)用操作系統(tǒng)的時(shí)間；
    • 二者很難精確區(qū)分；
  • 一般我們喜歡用elapsed time來評價(jià)一個(gè)程序的性能，其實(shí)elapsed time和CPU time還是有區(qū)別的
    • we will use the term system performance to refer to elapsed time on an unloaded system and CPU performance to refer to user CPU time;
  • 如一些server對IO很依賴，從而性能需要評價(jià)軟硬件綜合性能，而一些application則可能只關(guān)注throughput或者response time，或者兩者的組合。所以為提高performance，你必須要知道哪些方面構(gòu)成了performance matric matters，從而方便找到performance的瓶頸。
• CPU Performance and its Factors
  • 減少應(yīng)用程序的時(shí)鐘個(gè)數(shù)或者提高時(shí)鐘頻率
• Instruction Performance
  • 時(shí)鐘周期數(shù)=指令個(gè)數(shù)指令平均時(shí)鐘個(gè)數(shù)
  • 指令平均時(shí)鐘個(gè)數(shù)：clock cycles per instrunction,簡稱CPI
    • 每個(gè)指令執(zhí)行的時(shí)鐘個(gè)數(shù)不同，CPI為所有指令執(zhí)行的時(shí)鐘個(gè)數(shù)的平均數(shù)；
    • CPI provides one way of comparing two different implementations of the identical instruction set architecture，since the number of instructions executed for a program will, of course, be the same.
• The Clasic CPU Performance Equation
  • CPU time=Instrunction count CPIClock cycle time(or 1/Clock rate)
  • 注意IPC=1/CPI，IPC: instructions per clock
  • 注意現(xiàn)在很多CPU有變頻特點(diǎn)，如Intel i7可以將頻率提升10%直到CPU太熱在降回來，這種技術(shù)被稱為turbo技術(shù)；

1.7 The Power Wall

首先介紹了8代Intel CPU的頻率和功耗走勢圖；需要注意的是在Pentium4（2001）時(shí)，頻率達(dá)到3.6GHz,功耗達(dá)到103M，起后頻率與功耗都有降低；

其次介紹單個(gè)晶體管的動(dòng)態(tài)功耗(焦耳和瓦特角度）：
Energy1/2Capacitive load Voltages²
Power1/2Capacitive load Voltages² Frequency switched

Frequency switched和時(shí)鐘頻率相關(guān)；
with regard to the Figure, how could clock rates grow by a factor of 1000 while power increased by only a factor of 30?(頻率有1000倍增長而功耗只有30倍增長)
Energy and thus power can be reduced by lowering the voltage，每代CPU都采用這種技術(shù)，即每代電壓減少15%（0.0225）

In 20 years, voltages have gone from 5V to 1V, which is why the increase in power is only 30 times.(0.0225x1000=22.530)

現(xiàn)代的問題是，繼續(xù)降低電壓會使晶體管too leaky（服務(wù)器芯片40%功耗來源于leakage）

to try to adress the power problem, designers have already attached large devices to increase cooling. and they turn off parts of the chip that are not used in a given clock cycle.
盡管需要高昂的降溫設(shè)備（300W功耗），這種方案還是應(yīng)用于PC和server，而PMD則不需要；

Power is a challenge for integrated circuits for 2 reasons:

1. power must be brought in and distributed around the chip;現(xiàn)代芯片可能需要數(shù)百個(gè)ground和power的引腳
2. power is dissipated as heat and must be removed.(功耗消耗為熱量需要更貴的散熱設(shè)備)

1.8 The Sea Change: The Switch from Uniprocessors to Multiprocessors

為了解決power問題，引入了multiprocessor的概念，這就要求現(xiàn)代程序員必須要考慮重新編寫已有的程序以適應(yīng)multiprocers；

目前看程序員的轉(zhuǎn)型還很少，期待未來的改變；
For parallel programming, the challenges include scheduling,load balancing,time for synchronization and overhead for communication between the parties .

后面介紹各個(gè)章節(jié)為parallel revolution而引入的內(nèi)容；

1.9 Real Stuff: Benchmarking the Intel Core i7

每個(gè)章節(jié)的結(jié)束會舉個(gè)實(shí)例來復(fù)習(xí)本章內(nèi)容，第一章結(jié)束介紹Benchmark程序如何對不同的CPU進(jìn)行評價(jià)；

如今的Benchmark大多出自SPEC：System Performance Evaluation Cooperative
感興趣可以看看，具體不介紹了；

1.10 Fallacies and Pitfalls

每章的結(jié)束也會有謬論與陷阱這一小節(jié)；

1.11 Concluding Remarks

總結(jié)本章所有內(nèi)容。

1.12 Historical Perspective and Further Reading

online reading