ptg22232966
HISTORICAL CONTEXT
11
importance of CUDA became apparent, it also became clear that sharing the
appreciation of CUDA was the right thing to do. The result was a special topics
course on applications of CUDA offered to students from a wide range of depart-
ments with no prerequisites other than an introductory computing class. The
writing of this book follows the third offering of the course, which has evolved to
include system setup, exploration of CUDA Samples, a very fast introduction to
C programming, CUDA’s model for parallelism—along with the few things
beyond basic C needed to implement it—and a survey of useful CUDA features
including atomic functions, shared memory, and interactive graphics. All of that
happens in the first half of a 10-week quarter, and the remaining time is dedi-
cated to projects and presentations. Each student in the class gives a short pre-
sentation of their project topic in week 5 and then a final presentation in week 10
of what they have accomplished. Class meetings in between are a combination of
consulting hours and presentations (by instructor or guest speakers) on experi-
ences creating and using CUDA-powered applications. In terms of the contents
of the book, the course translates roughly as follows:
• System setup and initial CUDA Sample run: Appendices A and B, Chapter 1.
• Brief introduction to C: Appendix C.
• CUDA, its model for parallelism, and language essentials: Chapters 2 and 3.
• Interactive graphics: Chapter 4.
• Shared memory: First part of Chapter 5.
• Atomic functions: First part of Chapter 6.
• Project: Turn the students loose to explore the remainder of Chapters 5–9
for project inspiration, encourage active writing of code, and provide as much
help as possible getting over hurdles that arise.
While the organization of the book has some relation to the course, the content
is largely the result of a mostly unguided foray into the CUDA world, which turns
out to be a big place where it is not hard to get lost. We have filtered through a
lot of material including books, sample codes, research papers, and seminar
presentations, with the goal of getting down to the gist of what engineers from
outside the field would really need to know to get a handle on what can be done
with a powerful parallel computing tool like CUDA and understand how they can
actually make use of that tool themselves. It is our sincere hope that the result
is a book that can fulfill the dual purposes of enabling individual learning and
我的内容管理
展开
