CUDA是設計用于幫助開發并行程序的計算體系結構。通過與廣泛的軟件平臺相結合,cuda體系結構使程序員可以充分利用圖形處理單元(gpu)的強大能力構建高性能的應用程序。當然,gpu已經在很長時間內用于實現復雜的圖形和游戲應用程序。現在,cuda將這種具有價值的資源帶給在其他領域內從事應用程序開發的程序員,包括科學、工程和財務領域。這些程序員不需要了解圖形編程的相關知識,而只要能夠采用適當擴展的c語言版本進行編程即可。
本書由cuda軟件平臺團隊中的兩位博學成員編寫而成,他們向程序員展示了如何使用這種新的技術,并且通過大量可以運行的示例介紹了cuda開發的每個領域。在簡要介紹cuda平臺和體系結構以及快速指導cudac之后,本書詳細介紹了與每個關鍵的cuda功能相關的技術,以及如何權衡使用這些功能。通過閱讀本書,您將掌握使用每個cudac擴展的時機以及編寫性能極為優越的cuda軟件的方式。
CUDA范例精解:通用GPU編程(影印版)》由清華大學出版社出版。
山德爾(Jason Sanders)是NVIDIA公司CUDA平臺團隊中的博學軟件工程師,他協助開發了早期版本的CUDA系統軟件,并且幫助制定了作為異構計算的行業標準的OpenCL 1.0規范。Jason也在ATI Technologies、Apple和Novell擔任相關職務。 康洛特(Edward Kandrot)是NVIDIA公司CU
foreword
preface
acknowledgments
about the authors
1 why cuda ? why now?
1.1 chapter objectives
1.2 the age of parau. el. processing
1.3 the rise of gpu computing
1.4 cuda
1.5 applications of cuda
1.6 chapter review
2 getting started
3.1 chapter objectives
2.2 deve!.opment environment
2.3 chapter review
3 introduction to cuda c
3.1 chapter objectives
3.2 a first program
3.3 querying devices
3.4 using device properties
3.5 chapter review
4 parallel programming in cuda c
4.1 chapter objectives
4.2 cuda para[tel programming
4.3 chapter review
5 thread cooperation
5.1 chapter objectives
5.2 splitting parallel blocks
5.3 shared memory and synchronization
5.4 chapter review
6 constant memory and events
6.1 chapter objectives
6.2 constant memory
6.3 measuring performance with events
6.4 chapter review
7 texture memory
7.1 chapter objectives
7.2 texture memory overview
7.3 simulating heat transfer
7.4 chapter review
8 graphics interoperability
8.1 chapter objectives
8.2 graphics interoperation
8.3 gpu ripple with graphics interoperability
8.4 heat transfer with graphics interop
8.5 directx interoperability
8.6 chapter' review
9 atomics
9.1 chapter objectives
9.2 compute capability
9.3 atomic operations overview
9.4computing histograms
9.5 chapter review
10 streams
10.1 chapter objectives
10.2 page-locked host memory
10.3 cuda streams
10.4 using a single cuda stream
10.5 using multipte cuda streams
10.6 gpu work scheduling
10.7 using multiple cuda streams effectively
10.8 chapter review
11 cuda c on multiple gpus
11.1 chapter objectives
11.2 zero-copy host memory
11.3 using multiple gpus
11.4 portable pinned memory
11.5 chapter review
12 the final countdown
12.1 chapter objectives
12.2 cuda tools
12.3 written resources
12.4 code resources
12.5 chapter review
a advanced atomics
a.1 dot product revisited
a.2 impl. ementing a hash tabte
a.3 appendix review
index
In recent years, however, manufacturers have been forced to l,ook for al,terna-tives to this traditional, source of increased computational, power. Because ofvarious fundamental- l,imitations in the fabrication of integrated circuits, it is nol-onger feasibl.e to rel.y on upward-spiral,ing processor cl,ock speeds as a meansfor extracting additional power from existing architectures. Because of power andheat restrictions as wel,l, as a rapidl,y approaching physical- l,imit to transistor size,researchers and manufacturers have begun to l,ook el.sewhere.Outside the woHd of consumer computing, supercomputers have for decadesextracted massive performance gains in simil,ar ways. The performance of aprocessor used in a supercomputer has cl,imbed astronomical,l,y, simil.ar to theimprovements in the personal- computer CPU. However, in addition to dramaticimprovements in the performance of a singl,e processor, supercomputer manu-facturers have al,so extracted massive leaps in performance by steadily increasingthe number of processors. It is not uncommon for the fastest supercomputers tohave tens or hundreds of thousands of processor cores working in tandem.In the search for additional, processing power for personat computers, theimprovement in supercomputers raises a very good question: Rather than sol,el,yl,ooking to increase the performance of a single processing core, why not putmore than one in a personal- computer? In this way, personal- computers coul,dcontinue to improve in performance without the need for continuing increases inprocessor clock speed.
"對于處理基于圖形加速器的計算系統的人員來說,本書是必不可少的讀物。" ——Jack Dongarra博士(田納西大學特聘教授和橡樹嶺國家實驗室杰出研究員)作序推薦
簡單易懂的書 建議GPU入門的人購買
cuda的經典書籍,值得購買~
實用的教材
男朋友說還不錯
商品很好,隔了這么久才來好評,不錯不錯
我英文一般 看起來有點吃力 外國人寫書比中國人強那是一定的
還可以,就是簡單了點,不過老外能把問題講清楚,還是比較推薦。。。。。。
介紹cuda為數不多的書,正在看,目前還可以