nojhan/paradiseo
1
0
Fork 0
Code Issues Pull requests Projects Releases Packages Wiki Activity

git-svn-id: svn://scm.gforge.inria.fr/svnroot/paradiseo@2530 331e1502-861f-0410-8da2-ba01fb791d7f

Browse source
This commit is contained in:
boufaras 2011-10-26 14:40:27 +00:00
commit 16e2331c2b
1 changed files with 71 additions and 27 deletions
Download patch file Download diff file Expand all files Collapse all files

102
branches/ParadisEO-GPU/README.txt
View file

@ -1,37 +1,81 @@
This package contains the source for ParadisEO-GPU problems.
PARADISEO-GPU README FILE
=======================================================================
check latest news at http://paradiseo.gforge.inria.fr/
=======================================================================
# Step 1 - Configuration
------------------------
Edit the "install.cmake" file by entering the FULL PATH of :
-"ParadisEO_PATH" where Paradiseo directory has been installed in your host.
-"ParadisEO-GPU_PATH" where ParadisEO-GPU package has been decompressed in your host.
-"CUDA_PATH" where CUDA has been installed in your host.
-"NVIDIA_PATH" where NVIDIA has been installed in your host.
Welcome to ParadisEO-GPU, the reusable design and implementation of parallel meta-
heuristics on Graphics Processing Units (GPU).
The latest news about ParadisEO-GPU can be found on the gforge repository at
http://paradiseo.gforge.inria.fr/index.php?n=Doc.API
In case of any problem, please e-mail us at
paradiseo-help@lists.gforge.inria.fr
# Step 2 - Build process
------------------------
ParadisEO is assumed to be compiled. To download ParadisEO, please visit http://paradiseo.gforge.inria.fr/.
Go to the ParadisEO-GPU/build/ directory and lunch cmake:
(Unix) > cmake .. -DENABLE_CMAKE_TESTING=TRUE -DCMAKE_BUILD_TYPE=Debug
=======================================================================
BUILDING PARADISEO-PEO
=======================================================================
The basic installation procedure goes the following.
To compile paradiseo-gpu in the default directory,
go to paradiseo-gpu/build/ and run:
> cmake ..
> make
// for an easy-use of the provided lessons
> make install
// optional (if the documentation is not already available)
> make doc
To compile paradiseo-peo anywhere else, simply run:
> cmake $(GPU) -Dconfig=<path to the install.cmake file>
> make
// for an easy-use of the provided lessons
> make install
// optional (if the documentation is not already available)
> make doc
To clean everything, simply run
> make clean
# Step 3 - Compilation
----------------------
In the ParadisEO-GPU/build/ directory:
(Unix) > make
===================================================================
DIRECTORY STRUCTURE
===================================================================
After unpacking the archive file, you should end up with the following
structure:
# Step 4 - Execution
---------------------
A toy example is given to test the components. You can run these tests as following.
To define problem-related components for your own problem, please refer to the tutorials available on the website : http://paradiseo.gforge.inria.fr/.
In the ParadisEO-GPU/build/ directory:
(Unix) > ctest -D ExperimentalStart -D ExperimentalBuild -D ExperimentalTest -D ExperimentalSubmit
.../ The main PARADISEO-GPU directory, created when unpacking.
|
+-- build BUILD directory that contains libraries and executable files.
|
+-- src SOURCE directory Contains most PARADISEO-GPU .h files.
|
+-- doc DOCUMENTATION directory (generated by Doxygen).
| |
| +- html HTML files - start at index.html.
| |
| +- latex latex files - use to generate Postcript doc.
| |
| +- man Unix man format documentation.
|
|
+-- tutorial APPLICATIONS
|
+-- INSTANCES The instances problem directory (QAP instances ...)
|
+-- OneMax How make One max problem with parallel evalution of neighborhood(bit flipping) with hamming distances =1
|
+-- KswapOneMax How make One max problem with parallel evalution of neighborhood(bit flipping) with hamming distances >=1
|
+-- PPP_GPU How make Permuted Perceptron Problem with parallel evalution of neighborhood on GPU
|
+-- QAP_GPU How make Quadratic Assignement Problem with parallel evaluation of neighborhood on GPU
|
+-- QAP_CPU How make Quadratic Assignement Problem with sequentiel evaluation of neighborhood on CPU
In the directory "tutorial", there is an example of One Max problem which illustrate how to use this package.
# Documentation
---------------
The API-documentation is available in doc/html/index.html
===================================================================
NOTES
===================================================================
Mailing list : paradiseo-help@lists.gforge.inria.fr