David B. Kirk

The book can be divided into two parts. Part A (80%) facilitates an introduction into GPU programming with CUDA C. Part B (20%) gives very short introductions into other GPU programming languages or libraries. I had the impression that part B was mainly written by the co-authors mentioned at the beginning of each respective chapter; the reason was possibly that these chapters are not perfectly integrated with the rest of the text. Nevertheless, I found part B quite interesting and appropriate.

The main part A is clearly written in a style that makes it easy and fast to read, but it is not written very carefully. The text contains errors, especially in the code examples. Sometimes it is overly elaborate, almost rambling, especially in the sections that explain what a particular algorithm does.

A deep understanding of computer architecture is not assumed. Many performance-related explanations are provided although, unfortunately, measurements of specific pieces of code are almost completely absent.

Exercises are given at the end of each chapter; but they are not exceptionally inspiring.… (altro)