User Tools

Site Tools


cluster:161

Differences

This shows you the differences between two versions of the page.

Link to this comparison view

Both sides previous revision Previous revision
Next revision
Previous revision
Next revision Both sides next revision
cluster:161 [2017/08/07 13:48]
hmeij07
cluster:161 [2019/08/05 16:25]
hmeij07 [lammps-11Aug17]
Line 1: Line 1:
 \\ \\
 **[[cluster:0|Back]]** **[[cluster:0|Back]]**
 +
 +==== lammps-11Aug17 ====
 +
 +and now lammps-22Aug18
 +
 +Update: 
 +  * n78/gtx1080 lammps 11aug17 (centos7, mpich3/mpic++, cuda 8/sm_61, /usr/local)
 +      * [[cluster:164|GTX 1080 Ti]]
 +  * n37/k20 lammps 22aug18 (centos7, openmpi 1.8.4/mpic++, cuda 9.2/sm_35, /usr/local) 
 +      * [[cluster:173|K20 Redo Usage]]
 +  * n33/k20 lammps 11aug17 + 22aug18 (centos6, openmpi 1.8.4, cuda 5/sm35, /share/apps/CENTOS6)
 +      * this page
 +  * n33/k20 older lammps (centos6, mvapich, cuda 5/sm_??, /cm/shared/apps)
 +      * /home/hmeij/sharptail/run.gpu or gromacs.sub files
 +
 + --- //[[hmeij@wesleyan.edu|Henk]] 2018/09/20 09:30//
 +
 +<code>
 +N78
 +# devel packages for libjpeg, blas, lapack, boost
 +
 +which nvcc
 +export PATH=/usr/local/mpich-3.1.4/bin:$PATH
 +export LD_LIBRARY_PATH=/usr/local/mpich-3.1.4/lib:$LD_LIBRARY_PATH
 +which mpicxx
 +
 +# untar stable.gz
 +cd lib/gpu
 +vi Makefile.linux.double
 +# edits
 +CUDA_HOME = /usr/local/cuda
 +# best fitting ARCH for GTX 1080 with CUDA 8
 +CUDA_ARCH = -arch=sm_61
 +http://arnon.dk/matching-sm-architectures-arch-and-gencode-for-various-nvidia-cards/
 +for GTX1080 on n78
 +CUDA_PRECISION = -D_DOUBLE_DOUBLE
 +make -f Makefile.linux.double
 +ls -lrt
 +# new files: libgpu.a and Makefile.lammps
 +
 +cd ../../src/
 +vi MAKE/Makefile.mpi
 +# edits
 +LMP_INC =       -DLAMMPS_GZIP -DLAMMPS_MEMALIGN=64 -DLAMMPS_JPEG
 +JPG_INC =       -I/usr/include
 +JPG_PATH =      -L/usr/lib64
 +JPG_LIB =       -ljpeg
 +
 +make yes-rigid
 +make yes-gpu
 +make yes-colloid
 +make yes-class2
 +make yes-kspace
 +make yes-misc
 +make yes-molecule
 +
 +make purge
 +make package-upate
 +
 +which mpicxx
 +make mpi
 +mv lmp_mpi /usr/local/lammps-11Aug17/lmp_mpi-double-double-with-gpu[-rigid]
 +
 +# repeat from scratch for SINGLE_SINGLE and SINGLE_DOUBLE
 +
 +
 +</code>
  
 ==== Lammps CPU/GPU ==== ==== Lammps CPU/GPU ====
Line 17: Line 84:
 <code> <code>
  
-on n33, a gpu node+N33, a gpu node
  
-  mkdir /tmp/laamps; cd !$+  mkdir /tmp/lammps; cd !$
   gunzip lammps-stable.tar.gz    gunzip lammps-stable.tar.gz 
   tar xvf lammps-stable.tar    tar xvf lammps-stable.tar 
Line 132: Line 199:
 # need both /cm/shared... and /cm/local ... # need both /cm/shared... and /cm/local ...
  
-# the malaysan approach also works, see install file+# the malaysian approach also works, see install file
  
  
Line 268: Line 335:
 # stage the data # stage the data
 scp -r ~/lammps/colloid-gpu/* . scp -r ~/lammps/colloid-gpu/* .
-# feed the wrapper+# feed the binary
 /share/apps/CENTOS6/lammps/31Mar17/lmp_mpi -h /share/apps/CENTOS6/lammps/31Mar17/lmp_mpi -h
 /share/apps/CENTOS6/lammps/31Mar17/lmp_mpi \ /share/apps/CENTOS6/lammps/31Mar17/lmp_mpi \
cluster/161.txt ยท Last modified: 2020/07/16 17:16 by hmeij07