Tree [ab304b] master /

File Date Author Commit
 autoparse 2010-12-26 listmg listmg [618aaf] Moved the autoparse to its own directory.
 doc 2010-12-25 listmg listmg [50e9f3] Added a tech note on the fileparser.
 example 2010-12-18 listmg listmg [19f606] Merge branch 'dev/fileparser' into develop
 src 2010-12-26 listmg listmg [618aaf] Moved the autoparse to its own directory.
 util 2010-12-04 David Car David Car [58b2dc] Merge branch 'dev/fileparser' into develop
 .gitignore 2010-10-17 David Car David Car [c260ad] * Resolved merge conflicts in Makefile.
 CHANGELOG.html 2010-10-20 listmg listmg [c99dd0] Modified changelog.
 Makefile 2011-09-29 Sur Huang Sur Huang [ab304b] * Modified tar target in Makefile to get rid of...
 README 2010-12-19 listmg listmg [635a4c] Merge branch 'develop' of ssh://
 TODOS 2009-10-26 David Car David Car [d1db77] * Added CHANGELOG and TODOS
 VERSION 2010-12-18 listmg listmg [90416e] Merge branch 'develop' into dev/fileparser

Read Me

David Car (
Michael List (

A Fortran compiler supporting the ISO_C_BINDINGS intrinsic module.  Currently,
the following should work:

gfortran > 4.3
ifort > 10.0
pgfortran latest

Also, of course you need the NVIDIA CUDA SDK.  I'm using version 2.3 at this time.

None at the moment



The CUDA.F90 module contains ISO_C_BINDINGS interfaces for various CUDA SDK
functions.  Compiling that will give you the cuda.o object file and cuda.mod
module.  Simply "use" the cuda module in your code.  See the main.F90 for an

To build the example test, do:

make library
make ARCH=64 library
make ARCH=64

This will build in emulation mode.  To build the code for the device, edit the
Makefile NVCCFLAGS variable and remove -deviceemu.  It should then look like:


Then do:

make clean
make clean
make ARCH=64

You'll find a "test" executable in the examples directory.  Run that.  Its a
simple vector add.  If it fails you may wish to decrease the size of the
allocated vectors (they are quite large for low or mid range cards).