-
-
Notifications
You must be signed in to change notification settings - Fork 442
Building HPX on Stampede
Because the $HOME directory quotas on Stampede are so small (5GB), we recommend installing HPX in the $WORK directory.
mkdir $WORK/local_install
module swap intel/15.0.2 gcc/4.9.1
module load boost/1.55.0
module load cmake/3.1.0
wget https://www.open-mpi.org/software/hwloc/v1.11/downloads/hwloc-1.11.2.tar.gz
tar -xzvf hwloc-1.11.2.tar.gz
cd hwloc-1.11.2
./configure --prefix=$WORK/local_install
make
make install
download jemalloc-3.6.0.sh
chmod +x jemalloc-3.6.0.sh
./jemalloc-3.6.0.sh $WORK/local_install
wget http://stellar.cct.lsu.edu/files/hpx_0.9.99.tar.gz
tar -xzvf hpx_0.9.99.tar.gz
Download this build script: build_hpx_release_stampede.sh
Make it executable:
chmod +x build_hpx_release_stampede.sh
And finally, run it:
./build_hpx_release_stampede.sh
This script assumes that you've downloaded HPX in your $WORK
directory. Change any paths as needed.
In order to enable support for some c++ features we need, we have to specify some extra compiler flags to the intel compiler:
-DCMAKE_CXX_FLAGS:STRING=-gcc-name=/opt/apps/gcc/4.9.1/bin/gcc -gxx-name=/opt/apps/gcc/4.9.1/bin/g++
And because the correct c libraries used are not part of the LD_LIBRARY_PATH
when the Intel modules are loaded, we have to manually add the path to the correct libraries before running anything built:
export LD_LIBRARY_PATH=/opt/apps/gcc/4.9.1/lib64/:$LD_LIBRARY_PATH
- HPX Resource Guide
- HPX Source Code Structure and Coding Standards
- Improvement of the HPX core runtime
- How to Get Involved in Developing HPX
- How to Report Bugs in HPX
- Known issues in HPX V1.0.0
- HPX continuous integration build configurations
- How to run HPX on various Cluster environments
- Google Summer of Code
- Google Season of Documentation
- Documentation Projects
- Planning and coordination