Difference between revisions of "Benchmarking: Quantum Espresso 5.0.2 GPU"
Jump to navigation
Jump to search
m |
m (→Requirements) |
||
| (5 intermediate revisions by the same user not shown) | |||
| Line 3: | Line 3: | ||
- Intel compilers <br> | - Intel compilers <br> | ||
- Intel MKL (whatever version above 10) <br> | - Intel MKL (whatever version above 10) <br> | ||
| − | - Open MPI | + | - Open MPI |
== Download == | == Download == | ||
| Line 23: | Line 23: | ||
module load openmpi/1.6.3-gpu-intel-compilers | module load openmpi/1.6.3-gpu-intel-compilers | ||
| − | Configure | + | 4. Configure |
./configure --enable-parallel --enable-openmp --enable-cuda \ | ./configure --enable-parallel --enable-openmp --enable-cuda \ | ||
--with-gpu-arch=35 --with-cuda-dir=/shared/apps/cuda/cuda-5.0 \ | --with-gpu-arch=35 --with-cuda-dir=/shared/apps/cuda/cuda-5.0 \ | ||
--enable-magma --enable-phigemm | --enable-magma --enable-phigemm | ||
| − | + | 5. Build <br> | |
Move to espresso top directory | Move to espresso top directory | ||
cd .. | cd .. | ||
make -f Makefile.gpu pw-gpu | make -f Makefile.gpu pw-gpu | ||
| + | |||
| + | == Run == | ||
| + | 1. Login to a GPU compute node <br> | ||
| + | Load modules | ||
| + | module load cuda/5.0 | ||
| + | module load openmpi/1.6.3-gpu-intel-compilers | ||
| + | module load gpu/QE/5.0.2-GPU-Intel-OpenMPI | ||
| + | 2. Run | ||
| + | mpirun -np 2 pw-gpu.x | ||
Latest revision as of 10:03, 7 March 2013
Requirements
- CUDA 5.0
- Intel compilers
- Intel MKL (whatever version above 10)
- Open MPI
Download
1. Traverse to the install directory Download QE 5.0.2 and untar it
wget http://qe-forge.org/gf/download/frsrelease/116/403/espresso-5.0.2.tar.gz tar zxf espresso-5.0.2.tar.gz
2. Change directory to espresso-5.0.2 Download QE 5.0.2 GPU plugin untar
wget http://qe-forge.org/gf/download/frsrelease/119/419/QE-GPU-5.0.2-build2.tar.gz tar zxf QE-GPU-5.0.2-build2.tar.gz cd GPU
Configure and build
3. Load modules
module load cuda/5.0 module load intel/compilers/13.0.1 module load openmpi/1.6.3-gpu-intel-compilers
4. Configure
./configure --enable-parallel --enable-openmp --enable-cuda \ --with-gpu-arch=35 --with-cuda-dir=/shared/apps/cuda/cuda-5.0 \ --enable-magma --enable-phigemm
5. Build
Move to espresso top directory
cd .. make -f Makefile.gpu pw-gpu
Run
1. Login to a GPU compute node
Load modules
module load cuda/5.0 module load openmpi/1.6.3-gpu-intel-compilers module load gpu/QE/5.0.2-GPU-Intel-OpenMPI
2. Run
mpirun -np 2 pw-gpu.x