CUDA Fortran for Scientists and Engineers: Best Practices for Efficient CUDA Fortran Programming

I have a few thousand hours of experience working with CUDA Fortran and can say that this book is perfect for the advance and beginner CUDA Fortran programmers alike. The concepts are laid out succinctly and clearly. The book includes codes written for tests and performance feedback related to each topic with detailed discussion that will be very helpful for learning CUDA Fortran and fully understanding the CUDA GPU architectures. The book is helpful also in that it includes methods for using CUDA cards up through the Tesla K20. This was important for me as the Tesla K20 has, up until the past couple of months, had poor literature describing best practices and how to take advantage of the card's capabilities. The biggest offering this book brings to the table is that you won't have to waste time searching around online trying to find half-baked information in forums and/or papers.I also recommend "The CUDA Handbook" as another resource to understand programming with GPUs.

Very useful! The approach is very didactic!

Very good book!

I utilized this book for my first explicit FEM code with cuda fortran(Tesla K20c). With easy-to-understand descriptions, I could understand many concepts in detail. The important thing is I have been able to write cuda fortran code and accelerate my original sequential CPU code. I hope more and more practical topics such as how to accelerate Conjugate Gradient method used in the various numerical simulator would be covered in the next revision of this book!! I love this book! Great thanks to the author!

CUDA Fortran - Excellent reference for Scientists

Beim Stöbern in diesem Buch habe ich gesehen, dass die Autoren wertvolle Optimierungstipps geben. Da bin ich neugierig geworden. Ich selber bin kein Fortran-Programmierer, kann daher das Buch aus dieser Perspektive nicht beurteilen. Aber für mich C++-Entwickler bietet dieses Buch viele wertvolle Hinweise über die Optimierung von CUDA-Kerneln. Als Beispiel kann hier die mathematisch präzise Analyse der "Fused Multiply Add (FMA)”-Operation in Abschnitt 5.2.1 genannt werden. Das Buch wurde von zwei sehr kompetenten Autoren geschrieben und ist sehr solide. Die Autoren arbeiten beide bei NVIDIA, verfügen daher über Insider-Informationen.Das Buch ist leider schon ein wenig veraltet, die aktuelle Maxwell-Architektur ist erst 2014 erschienen. Im Buch wird die Kepler-Architektur bis zur Compute Capability 3.5 behandelt.Einen Punkt Abzug gibt es aber, doch, weil im Anhang auf 61 Seiten Source-Code ausgedruckt wurde. Das ist doch heute nicht mehr zeitgemäß, den kann man sich herunterladen.

It is really a good book for a person working GPU calculations. There are lots of details and example codes about how to write the Fortran codes with CUDA fortran.

I was very excited to see PGI had brought Fortran to the new generation of GPU devices (well the Nvidia ones), but the PGI compiler did not prove to be very easy to use and I was especially frustrated when some of PGIs own example codes did not run on my Alienware with its two GTX560M cards.Then out can this book and my enthusiasm is reborne!The examples in this book, that I have tried so far, do work! The level of explanation works for me, but I fear it may be just out of reach for, say, an undergraduate interested in using GPU cards to execute code for a research project.I applaud this book, but I have not yet worked my way through to the final sections. Well done so far!

Make the most of CUDA on your graphics card with this great little book!