Skip to content

Domain-specific framework for performance analysis of parallel programs

Notifications You must be signed in to change notification settings

thu-pacman/PerFlow

 
 

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

74 Commits
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 

Repository files navigation

PerFlow

Domain-specific framework for performance analysis of parallel programs

Dependency and Build

PerFlow is dependent on

  • Dyninst
  • Boost Boost will be installed automatically with Dyninst.
  • PAPI
  • igraph
  • cmake >= 3.16

Dyninst and PAPI need user to build themselves. igraph has been integrated into PerFlow as submodule.

git submodule update --init

There are two ways to build PerFlow. One is to build dependency from source and specify their location when building PerFlow, the other is to use spack to build these.

Build Dependency from Source

cmake .. -DBOOST_ROOT=/path_to_your_boost_install_dir -DDyninst_DIR=/path_to_your_dyninst_install_dir/lib/cmake/Dyninst -DPAPI_PREFIX=/path_to_your_papi_install_dir

# You should make sure that there is `DyninstConfig.cmake` in /path_to_your_dyninst_install_dir/lib/cmake/Dyninst
# And there is `include` `lib` in /path_to_your_papi_install_dir
# And there is `include` `lib` in /path_to_your_boost_install_dir, `boost` in /path_to_your_boost_install_dir/include

Note that if dyninst is built from source, the boost will be downloaded and installed automatically with it, in the install directory of dyninst.

In this case, the cmake commands will be like

cmake .. -DBOOST_ROOT=/path_to_your_dyninst_install_dir  -DDyninst_DIR=/path_to_your_dyninst_install_dir/lib/cmake/Dyninst -DPAPI_PREFIX=/path_to_your_papi_install_dir

Build Dependency from Spack

The recommended way to build Dyninst (with Boost) and PAPI is to use Spack

spack install dyninst # boost will be installed at the same time
spack install papi

# before building PerFlow
spack load dyninst # boost will be loaded at the same time
spack load papi

# build
mkdir build
cd build
cmake ..

About

Domain-specific framework for performance analysis of parallel programs

Resources

Stars

Watchers

Forks

Packages

No packages published

Languages

  • C++ 88.0%
  • Python 9.1%
  • CMake 2.2%
  • Other 0.7%