Warning, /geant4/examples/advanced/CaTS/Instructions.md is written in an unsupported language. File is not indexed.
0001 # Prerequisites
0002 Opticks requires Geant4 (10.7.p02), nvidia cuda (11.3) and nvidia Optix (6.5) among other libraries. CaTS in addition will require ROOT. If all these libraries and development headers are available on your machine skip directly to (**Building opticks vs. existing libraries**). On a 'blank' computing system it makes sense to build CLHEP, then Geant4 and finally ROOT assuring that all the necessary development libraries and headers are installed.
0003
0004 # Building CLHEP
0005 The current version of Geant4 10.07.p02 is build on clhep 2.4.4.0.
0006 CLHEP can be found at:
0007 https://proj-clhep.web.cern.ch/proj-clhep/clhep23.html
0008
0009 to build it from scratch using cmake (used cmake version 3.20.5)
0010
0011 cd to the directory where you want to build clhep
0012 wget https://proj-clhep.web.cern.ch/proj-clhep/dist1/clhep-2.4.4.0.tgz
0013 tar xzvf clhep-2.4.4.0.tgz
0014 cd 2.4.4.0/
0015 mkdir CLHEP-build
0016 cd CLHEP-build
0017 cmake -DCMAKE_INSTALL_PREFIX=../CLHEP-install DCLHEP_BUILD_CXXSTD=-std=c++11 ../CLHEP
0018 make -j 8
0019 make install
0020
0021 **Note** the default install directory is /usr/local but one needs root privileges to install it there:
0022
0023 cd to the directory where you want to build clhep
0024 wget https://proj-clhep.web.cern.ch/proj-clhep/dist1/clhep-2.4.4.0.tgz
0025 tar xzvf clhep-2.4.4.0.tgz
0026 cd 2.4.4.0/
0027 mkdir CLHEP-build
0028 cd CLHEP-build
0029 cmake -DCLHEP_BUILD_CXXSTD=-std=c++11 ../CLHEP
0030 make -j 8
0031 sudo make install
0032
0033 # Building Geant4
0034
0035 Geant4 versions are available at:
0036 https://geant4.web.cern.ch/support/download
0037
0038
0039 cd to the directory where you want to install Geant4
0040 wget https://geant4-data.web.cern.ch/releases/geant4.10.07.p02.tar.gz
0041 mkdir geant4.10.07.p02-build
0042 cd geant4.10.07.p02-build
0043 cmake -DCMAKE_INSTALL_PREFIX=../geant4.10.07.p02-install -DGEANT4_BUILD_VERBOSE_CODE=OFF -DGEANT4_INSTALL_DATA=ON -DGEANT4_USE_SYSTEM_CLHEP=ON -DGEANT4_USE_GDML=ON -DGEANT4_USE_SYSTEM_EXPAT=ON -DGEANT4_USE_SYSTEM_ZLIB=ON -DGEANT4_USE_QT=ON -DGEANT4_BUILD_MULTITHREADED=ON -DGEANT4_USE_OPENGL_X11=ON ../geant4.10.07.p02
0044 make -j 8
0045 make install
0046 . ../geant4.10.07.p02-install/bin/geant4.sh
0047
0048
0049 check the output for any error, install any development packages that might be necessary.
0050
0051
0052
0053
0054 # Building ROOT
0055 Instructions how to build ROOT from ssource can be found at:
0056 https://root.cern/install/build_from_source/
0057
0058 # cd to the diretory where you want to install root
0059 git clone --branch latest-stable https://github.com/root-project/root.git root_src
0060 mkdir root-build
0061 cd root-build
0062 cmake -DCMAKE_INSTALL_PREFIX=../root-install ../root
0063 # speed up the make process
0064 new=" -j$(($(grep -c ^processor /proc/cpuinfo) - 1))"
0065 case ":${MAKEFLAGS:=$new}:" in
0066 *:"$new":*) ;;
0067 *) MAKEFLAGS="$MAKEFLAGS:$new" ;;
0068 esac
0069 cmake -DCMAKE_INSTALL_PREFIX=../root-install ../root_src/
0070 cmake --build . --target install
0071
0072 check the output for any error, install any development packages that might be necessary.
0073
0074
0075 . ../root-install/bin/thisroot.sh
0076 root
0077
0078
0079 # Installing CUDA
0080
0081 cuda (11.3) is available at the NVIDIA web site just follow the instruction depending on the system you are using.
0082
0083 https://developer.nvidia.com/cuda-downloads
0084
0085 **Note** this will also install the corresponding NVIDIA graphics driver you might have to reboot.
0086
0087 Another way to obtain cuda is to install the NVIDIA hpc-sdk kit. Which can be found here.
0088 https://developer.nvidia.com/nvidia-hpc-sdk-downloads
0089
0090 The NVIDIA hpc-sdk kit provides an interesting set of tools e.g. nvc++ which allows offloading of parallel algorithms to NVIDIA GPUs.
0091
0092 A good way to check that things are working properly is to build the cuda samples and execute them
0093
0094 # cd to the directory where you want to build the cuda samples. E.g. the commands deviceQueryDrv and deviceQuery provide useful information.
0095 mkdir cuda-test
0096 cd cuda-test
0097 cp -r /usr/local/cuda-11.3/samples .
0098 cd samples/
0099 which nvcc
0100 make
0101 bin/x86_64/linux/release/deviceQuery
0102 bin/x86_64/linux/release/deviceQueryDrv
0103
0104 2 Tools For Monitoring Nvidia GPUs On Linux can be found here:
0105 https://www.linuxuprising.com/2019/06/2-tools-for-monitoring-nvidia-gpus-on.html
0106
0107 # Installing Optix (6.5)
0108
0109 https://developer.nvidia.com/designworks/optix/download
0110
0111 Optix comes with precompiled samples and one might want to try them:
0112
0113 # cd to the Optix installation directory
0114 cd SDK-precompiled-samples
0115 export LD_LIBRARY_PATH=.:$LD_LIBRARY_PATH
0116 # execute e.g.:
0117 ./optixMDLSphere
0118 ./optixSphere
0119 ./optixTutorial
0120 # etc.
0121
0122
0123
0124
0125
0126 # Building opticks vs. existing libraries
0127
0128 This are instructions how to build opticks making use of preinstalled libraries available on the system. These libraries include CLHEP, xerces-c, boost and Geant4.
0129 For geant 4 we use the current version at the time of writing which is Geant4.10.7.p2. We make use of the fact that the om-cmake function of om.bash is sensitive
0130 to CMAKE_PREFIX_PATH envvar so that we can point to the directories where the libraries are installed and void having to rebuild them. In principle just cut and paste the following line to a file change the envars of the different directories to match your system and source the resulting script.
0131
0132 cd to the directory where you want to install Opticks (the WORK_DIR environmental variable will point to this directory).
0133
0134
0135 Here we are using a tagged snapshot of Opticks which can be found in github:
0136
0137 git clone https://github.com/simoncblyth/opticks.git
0138 cd opticks
0139 git checkout tags/v0.1.6 -b v0.1.6-branch
0140 git status
0141
0142 The development version (a. k. a. the latest and greatest) can be found in the following repository
0143
0144 git clone https://bitbucket.org/simoncblyth/opticks.git
0145
0146
0147
0148 change opticks/optickscore/OpticksSwitches.h
0149
0150 so that:
0151
0152 #define WITH_SKIPAHEAD 1
0153
0154 is set.
0155
0156 cat > setup_opticks.sh << +EOF
0157 # ----------------------------------------------------------------------------------------------------------------------
0158 # --- you need to modify the following environmental variables so that point to the specific directories on your system
0159 # ---
0160 export WORK_DIR=/data2/wenzel/gputest_10.7.p02
0161 export OptiX_INSTALL_DIR=/home/wenzel/NVIDIA-OptiX-SDK-6.5.0-linux64
0162 export OPTICKS_COMPUTE_CAPABILITY=75
0163 export CUDA_INSTALL_DIR=/usr/local/cuda-11.3
0164 export CUDA_SAMPLES=${CUDA_INSTALL_DIR}/samples
0165 export G4INSTALL=/data2/wenzel/Geant4.10.07.p02_install
0166 . ${G4INSTALL}/bin/Geant4.sh
0167 export ROOTSYS=/data2/wenzel/root_install
0168 . ${ROOTSYS}/bin/thisroot.sh
0169 # ----------------------------------------------------------------------------------------------------------------------
0170 export LOCAL_BASE=${WORK_DIR}/local
0171 export CMAKE_PREFIX_PATH=${G4INSTALL}:${LOCAL_BASE}/opticks/externals:${OptiX_INSTALL_DIR}:${WORK_DIR}/opticks/cmake/Modules/:${WORK_DIR}/local/opticks:${WORK_DIR}/local/opticks:${WORK_DIR}/local/opticks/externals/
0172 export PYTHONPATH=$WORK_DIR
0173 export OPTICKS_HOME=${WORK_DIR}/opticks
0174 export PATH=${LOCAL_BASE}/bin:${PATH}
0175 export OPTICKS_PREFIX=${WORK_DIR}/local/opticks
0176 export OPTICKS_INSTALL_PREFIX=$LOCAL_BASE/opticks
0177 export OPTICKS_OPTIX_PREFIX=${OptiX_INSTALL_DIR}
0178 export OPTICKS_CUDA_PREFIX=${CUDA_INSTALL_DIR}
0179 export OPTICKS_EMBEDDED_COMMANDLINE_EXTRA="--rngmax 10 --rtx 1 --skipaheadstep 10000"
0180 opticks-(){ . ${OPTICKS_HOME}/opticks.bash && opticks-env $* ; }
0181 op(){ op.sh $* ; }
0182 o(){ cd $(opticks-home) ; hg st ; }
0183 # make sure to add the compiler options
0184 new=" -fPIC"
0185 case ":${CXXFLAGS:=$new}:" in
0186 *:"$new":*) ;;
0187 *) CXXFLAGS="$CXXFLAGS:$new" ;;
0188 esac
0189 new=" -fPIC"
0190 case ":${CFLAGS:=$new}:" in
0191 *:"$new":*) ;;
0192 *) CFLAGS="$CFLAGS:$new" ;;
0193 esac
0194 # speed up the make process
0195 new=" -j$(($(grep -c ^processor /proc/cpuinfo) - 1))"
0196 case ":${MAKEFLAGS:=$new}:" in
0197 *:"$new":*) ;;
0198 *) MAKEFLAGS="$MAKEFLAGS:$new" ;;
0199 esac
0200 # deal with the $LD_LIBRARYPATH
0201 new=${OptiX_INSTALL_DIR}/lib64/
0202 case ":${LD_LIBRARY_PATH:=$new}:" in
0203 *:"$new":*) ;;
0204 *) LD_LIBRARY_PATH="$new:$LD_LIBRARY_PATH" ;;
0205 esac
0206 new=${OPTICKS_HOME}/externals/lib
0207 case ":${LD_LIBRARY_PATH:=$new}:" in
0208 *:"$new":*) ;;
0209 *) LD_LIBRARY_PATH="$new:$LD_LIBRARY_PATH" ;;
0210 esac
0211 new=${CUDA_INSTALL_DIR}/lib64/
0212 case ":${LD_LIBRARY_PATH:=$new}:" in
0213 *:"$new":*) ;;
0214 *) LD_LIBRARY_PATH="$new:$LD_LIBRARY_PATH" ;;
0215 esac
0216 new=${LOCAL_BASE}/opticks/lib/
0217 case ":${LD_LIBRARY_PATH:=$new}:" in
0218 *:"$new":*) ;;
0219 *) LD_LIBRARY_PATH="$new:$LD_LIBRARY_PATH" ;;
0220 esac
0221 opticks-
0222 new=${CUDA_INSTALL_DIR}/bin
0223 case ":${PATH:=$new}:" in
0224 *:"$new":*) ;;
0225 *) PATH="$new:$PATH" ;;
0226 esac
0227 new=${OPTICKS_HOME}/bin/
0228 case ":${PATH:=$new}:" in
0229 *:"$new":*) ;;
0230 *) PATH="$new:$PATH" ;;
0231 esac
0232 new=${OPTICKS_HOME}/ana/
0233 case ":${PATH:=$new}:" in
0234 *:"$new":*) ;;
0235 *) PATH="$new:$PATH" ;;
0236 esac
0237 new=${LOCAL_BASE}/opticks/lib/
0238 case ":${PATH:=$new}:" in
0239 *:"$new":*) ;;
0240 *) PATH="$new:$PATH" ;;
0241 esac
0242 new=${CUDA_SAMPLES}/bin/x86_64/linux/release/
0243 case ":${PATH:=$new}:" in
0244 *:"$new":*) ;;
0245 *) PATH="$new:$PATH" ;;
0246 esac
0247 oinfo-(){
0248 echo 'LD_LIBRARY_PATH:';
0249 echo '================';
0250 echo ${LD_LIBRARY_PATH}| tr : \\n;
0251 echo;
0252 echo 'PATH:';
0253 echo '=====';
0254 echo ${PATH}| tr : \\n;
0255 echo;
0256 echo 'CMAKE_PREFIX_PATH:';
0257 echo '==================';
0258 echo ${CMAKE_PREFIX_PATH}| tr : \\n;
0259 }
0260 dinfo-(){
0261 nvidia-smi;
0262 ${CUDA_SAMPLES}/bin/x86_64/linux/release/deviceQuery
0263 }
0264 +EOF
0265
0266 source setup_opticks.sh
0267 oinfo-
0268 dinfo-
0269 mkdir -p ${WORK_DIR}/local/opticks/externals/
0270 cd ${WORK_DIR}/local/opticks/externals/
0271 ln -s ${OptiX_INSTALL_DIR} OptiX
0272 cd ${WORK_DIR}
0273 opticks-externals-install >& install_ext.log &
0274
0275 scan the log file or any errors and correct them.
0276
0277 cd ${WORK_DIR}
0278 opticks-full >& install_full.log &
0279
0280 scan the log file or any errors and correct them. Before you run opticks-t you want to to create the geocache using e.g. one of the gdml files provided by CaTS
0281 (https://github.com/hanswenzel/CaTS)
0282
0283 geocache-
0284 geocache-create- --gdmlpath ${WORK_DIR}/local/opticks/opticksdata/export/juno1808/g4_00.gdml
0285 export OPTICKS_KEY=`output from above command"
0286 opticks-t
0287
0288 if the geocache-create- command doesn't work:
0289
0290
0291 ${WORK_DIR}/local/opticks/lib/OKX4Test --okx4test --g4codegen --deletegeocache --gdmlpath ${WORK_DIR}/CaTS/gdml/simpleLArTPC.gdml
0292 export OPTICKS_KEY=`output from above command"
0293 opticks-t
0294
0295
0296
0297 # Building CaTS
0298
0299 Instructions for building and running CaTS can be found here:
0300
0301 https://github.com/hanswenzel/CaTS/blob/master/README.md
