CUDA Application Design and Development

I have found the other books in this series by nvidia quite helpful; I was looking forward to this text because it promised to present an entire application life-cycle. In particular, I was interested in profiling and debugging techniques, which (at least at the time of the purchase) pretty poorly documented.So there was some good information in this book, esp towards the end, however this particular author has some sort of idée fixe about objective functions. They are, perhaps, not a terrible example application, but myself not being well versed in the topic, I found that the author would alternately baby you through basic steps, and then skip over large portions of the theory necessary to understand the algorithms employed.While perhaps not quite fair to include in this book, the same author has written similarly about the Intel Phi architecture, and at a recent conference, his presentation (to experts) could best be described as "bloviating" whilst at the same time being almost completely uninformative.Unfortunately, it is currently the only book I know of that (again) takes you through the lifecycle development of a CUDA-enabled app. Suggestions welcome.

This is indeed a very good book to quickly learn about CUDA. Lot's of examples.Although it does not cover new features. I hope a new revision comes out soon with new features included in the CUDA.If you know nothing about CUDA and you want to get a quick start I still recommend this.

It's a rookie level CUDA book. It includes the info you need to know as a CUDA starter and you can start to program very soon. At meantime, I suggest read David B. Kirk's <<Programming Massively Parallel Processors>> as well, which gives more fundamental understanding of how CUDA works at a higher level.

This books is structured in a very weird way.It keeps showing you convoluted examples before actually explaining all the concepts you need to know before you can understand the example. Before explaining you the architecture, which is key for you to understand the code, it goes through the above process quite a few times.I can't imagine that someone with zero knowledge on the subject would read this cover to back and be somewhat proficient at coding in CUDA.If you are a beginner, I would say that you would be better off looking elsewhere. This book might be a good acquisition for you if you already know a lot of CUDA and want to go through another book.

One of the book I read some but no time to ready all. comparing to other book with more updated informaiton, I rather read new or latest one to worth my time...

One of the worst books on software development I have ever read. Absolutely useless. Authors writing in software area are getting worse and worse.

The book is a mixture of good and bad. On the good, the author knows a great deal about the subject and does a pretty good job of explaining GPU architecture. Chapters 1 and 4-7 are very useful and definitely worth reading for any cuda developers. Chapter 12 is also fun.The bad- I think large parts of the book are straight from other works by the author especially from a series of articles he wrote for Dr. Dobb's. The summary for Chapter 8 even refers to the chapter as "this article". So, often concepts are explained twice or used before they're explained in a later chapter e.g. thrust::reduce is used from chapter 1, finally explained in chapter 8 i believe.Chapters 2 and 3 should be avoided at all costs especially all the code. Read it at your own risk. The explanation of it is fragmented and random design choices and variable naming just adds to the confusion. The code in the other chapters is better. As a rule of thumb, if the code takes up more than 1 page skip it and any discussion of it - there's no benefit to reading it.There's also an extremely uneven assumption of reader knowledge throughout the book - the author at one point spends a page and a half explaining why testing is important. It takes him 3 lines to explain that someone pointed out his code could come up with a number so large that the variable couldn't hold it. Why didn't he just say "overflow" ? This is just after discussing pre-processor directives and simplex methods!So yeah, read the Dr. Dobb's articles if they're freely available. There's some good content in this book but a lot of it can be found in the cuda docs and websites/blogs. Given those resources, I feel this book isn't worth the price.

This is probably not the best first book on Cuda since it is more an overview of the possibilities available fromt he Cuda platform, but it's a great book for pointing users int he right direction once they have a grasp of the basics with Cuda. Generic functions and links are provided along with concise 'get-you-started' programs and snippets, so that the reader is presented with a variety of hooks into video apps, medical imaging technology and so on, without having to search high and low online for either tutorials or inspiration. With this in your armory, it's all about hard-core coding hereon, and no-one's going to do that for you! Brilliant book, my only gripe is that I'd have preferred more full-on A-Z examples.

Seemingly nice book to have, but I have read better ones.Its kind of getting all over the place with somewhat messy examples.Pity.

Ich hatte viele Artikel der Reihe „CUDA, Supercomputing for the Masses“ von Rob Farber auf der Dr. Dobb’s Webseite gelesen und war recht angetan. Das war damals 2009 eine der wenigen Möglichkeiten, etwas anspruchsvolleres über CUDA zu lesen.Der Titel des Buches erweckte in mir die Vorstellung, dass dieses Buch allgemeine Richtlinien und Prinzipen für die Erstellung von CUDA-Applikationen zum Thema hätte.Leider ist das Buch aber eher nur eine Ansammlung von Artikeln mit Themen, die den Autoren anscheinend interessiert haben, die aber in einem Lehrbuch über CUDA nichts verloren haben. Machine Learning mag ja ein ganz interessantes Thema sein, aber dann sollte man das Buch auch „Machine Learning with CUDA“ nennen. Der Autor wird sich hier selbst nicht gerecht.Es gibt dennoch ein paar interessante Teile im Buch. Allerdings muss man sich für diese das Buch nicht extra kaufen, denn sie werden in neueren Büchern besser erklärt. Wenn man aber begeisterter CUDA-Fan ist, kann man das Buch schon lesen und evtl. auch ein paar Anregungen erhalten.Hier meine Kritikpunkte im einzelnen.• Kap. 2 ist viel zu sehr mit der Theorie des Machine Learning’s beschäftigt. Funktoren kann man auch bei einfacheren Algorithmen verwenden.• Makros sollte man nicht in C++ verwenden (S.47)• Ein #include mitten im Code ist auch übel (S.47)• Kap. 8 über die Ausführung von CUDA-Code auf x86-Prozessoren bezieht sich auf ein kommerzielles Produkt der Firma PGI. Das ist eigentlich ein Kapitel Werbung und nicht als solche gekennzeichnet.• Die 4,5 Seiten Beispielcode enthalten sehr viel Mathematik, aber wenig CUDA (S. 57ff).• In Kapitel 9 wird sehr viel OpenGL-Code abgedruckt. Das interessiert in einem CUDA-Buch nicht.• Das Framework GPU Ocelot beschreibt der Autor als „popular, actively maintained“ (S.187). Das ist stark übertrieben, es gab pro Jahr einen Vortrag zu diesem Thema und auch nur jährlich ein Release.• SWAN und MCUDA waren mal Forschungsprojekte und sind laut ihrer Webseiten seit 2010 nicht mehr aktiv.

O Livro é Muito bom....com teoria e prática muito bem organizados mas o meu livro veio com várias páginas parecendo uma xerox, ou uma impressão fraca(pouco apagada). A página que estou anexando na FOto é exatamente como a vejo. Já tive oportunidade de ver outro livro do mesmo e a imagem é bem nínida assim como todas as outras do mesmo livro. Sugiro que a loja verifique os livros antes de enviar. A médida que continuo a leitura noto ainda mais falhas. O Livro tem páginas coloridas(Original) a versão que me foi enviada não tem uma figura colorida ...parece uma xerox mesmo. Uma falta de respeito da amazon.