From GPU

Contents 1 FLAGON: Fortran-9X Library for GPU Numerics 1.1 Introduction 1.2 Downloading 1.3 Content 1.4 Functionalities 1.5 Development 1.6 Installation Instructions

[edit] FLAGON: Fortran-9X Library for GPU Numerics

The wiki has moved to http://flagon.wiki.sourceforge.net/

[edit] Introduction

FLAGON is a library for programming NVIDIA CUDA from Fortran 95 (and is currently implemented currently on Intel Fortran on Linux and Windows). It was originally developed as a proprietary product by Fantalgo, LLC in 2007 under the internal name DevObject, but was then released in to the public domain under the LGPL. A poster paper describing the package was presented Supercomputing 2007. Subsequently researchers at UMD have added to its functionality, and updated it to the latest release of CUDA. The package was publicly announced on the GPGPU and CUDA forums in August 2008.

The developers of FLAGON are Yuancheng Luo, Nail A. Gumerov (who conceived the project and developed the initial version), Kate Despain, Bill Dorland and Ramani Duraiswami.

[edit] Downloading

For those who want to jump in to the code, the current build can be obtained from our SVN repository at

svn co https://flagon.svn.sourceforge.net/svnroot/flagon flagon

An older version is at svn://spare.cscamm.umd.edu/DevObject/DevObjectRuntime

However, download from the older version is deprecated

[edit] Content

User Manual Intro

Announcements

[edit] Functionalities

Basic Functions

Device Functions

Custom Functions

CUBLAS Functions

CUFFT Functions

[edit] Development

Technical Developments

Future Developments

Example Applications

[edit] Installation Instructions

Installation instructions for Windows XP (32 and 64) are at FLAGONBuildingGuide.doc

Instructions for Linux will be posted shortly.