Barbara Chapman (1) (1954–)

The first part of the book (up to chapter 4) is an explanation of the OpenMP standard. The second part (chapters 5-7) contains information which is related to parallel programming in general. There is also a shorter third part which covers details of OpenMP.

The first part is solid and mostly readable. It seems to regularly borrow from the standard (which is fine) but also has a tendency, to refer to the standard for the details; the problem that I found with this was that the standard was not really more detailed than the text in the book (maybe I should have used the OpenMP 2.5 standard, but Visual C++ is based on OpenMP 2.0). Nevertheless, this part would deserve 4 stars.

Unfortunately, the quality of the later parts have significantly lower quality. The analysis of performance in chapter 5 is lacking: The results show good throughput until a certain problem size, but then for very big problems the throughput degrades severely. This severe performance hit is never analysed. The whole discussion is about the moderate sized problems.
The source code in the higher chapters often contains oddly nested braces, the code in Chapter 8 even contains lots of syntax errors. Also I found the higher chapters quite repetitious in an annoying way and I feel that I wasted quite some time with them.

Finally, I found that the memory model received far to little attention, although a later chapter devotes 3 or so pages to it.… (more)