Skip to content
master
Switch branches/tags
Code

Latest commit

 

Git stats

Files

Permalink
Failed to load latest commit information.
Type
Name
Latest commit message
Commit time
dep
 
 
gui
 
 
qt
 
 
sip
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 

DAGViz

A DAG visualization and analysis tool for task parallel applications based on the computation DAG traces captured from their executions.

DAGViz's GUI was first implemented in C language with the GTK+ toolkit, but then it is changed to Qt5 framework with its Python bindings PyQt5. So there are currently two versions: dagviz-gtk and dagviz-pyqt, and the default one dagviz is symlinked to the newer dagviz-pyqt. Qt-based version is easier to be installed and used on other platforms (e.g., macOS, Windows) rather than the Ubuntu Linux on which it is originally developed.

DAGViz GUI

Besides the dependencies on the GUI toolkits, DAGViz also depends on MassiveThreads for reading the format of the DAG trace files which are recorded by DAG Recorder - a tracing module included in MassiveThreads.

GTK-based version and Qt-based version can be built separately and independently. The newer Qt-based version is recommended. More detailed installation instructions are provided below.

DAGViz-GTK

DAGViz-GTK is written totally in C language with the GTK+ toolkit.

Dependencies

  1. MassiveThreads: https://github.com/massivethreads/massivethreads (Instructions on install here)
  2. GTK+ 3: https://www.gtk.org/download (libgtk-3-dev package on Ubuntu)

There are other optional packages for reading sampling-based traces if necessary:

  • libunwind: for getting call chains Ubuntu 14.04/20.04.1 LTS: libunwind8-dev package

  • libbfd: for dagviz to read binary file. Ubuntu 14.04/20.04.1LTS: binutils-dev package

Quick start

$ git clone git@github.com:massivethreads/dagviz.git
$ cd dagviz
$ make gtk
$ ./dagviz-gtk <*.dag>

DAGViz-PyQt

DAGViz-PyQt is built based on the Qt5 framework, the GUI part is written in Python language with Qt5's Python bindings PyQt5, the visualization part which renders DAG on viewport/canvas is written in C++ (Qt5) in order to better communicate with the compute part which is maintained in C for the purpose of better performance and memory management.

The Python part and C++ part are connected by the SIP tool which is created by the same company of PyQt5.

Dependencies

  1. MassiveThreads: https://github.com/massivethreads/massivethreads
  2. Qt5 v5.9.1: https://www.qt.io/download/
  3. SIP 4.19.3: https://www.riverbankcomputing.com/software/sip/download
  4. PyQt5 v5.9: https://www.riverbankcomputing.com/software/pyqt/download5
  5. Python 3 (python3)

Quick start

$ git clone git@github.com:massivethreads/dagviz.git
$ cd dagviz
$ make qt
$ ./dagviz-pyqt
or
$ make dagviz
$ ./dagviz

Quick troubleshootings

  • sometimes Python.h is not found, try adding its path to appropriate environment variables:

    $ export C_INCLUDE_PATH=/usr/include/python3.5:$C_INCLUDE_PATH
    $ export CPLUS_INCLUDE_PATH=/usr/include/python3.5:$CPLUS_INCLUDE_PATH
  • ensure Qt5's qmake is included in system's path

    QTDIR=$HOME/local/Qt5.9.1/5.9.1/gcc_64
    PATH=$QTDIR/bin:$PATH
  • sip/configure.py: ensure PyQt5's SIP include directory and Qt5's include directory are included in system's include path, otherwise change to correct paths in sip/configure.py, e.g.,

    sip_inc_dir = "$HOME/local/share/sip/PyQt5"
    qt_inc_dir = "$HOME/local/Qt5.9.1/5.9.1/gcc_64/include"
  • ensure PyQt5 modules are included in system's Python path, e.g.,

    $ export PYTHONPATH=$HOME/local/lib/python3.5/dist-packages:$PYTHONPATH
  • On Ubuntu 16.04, some necessary Python and OpenGL packages are:

    $ sudo apt-get install libpython3.5-dev mesa-common-dev libglu1-mesa-dev freeglut3-dev

More detailed notes of installing DAGViz-PyQt's dependencies from sources


MassiveThreads installation:
  • given that MassiveThreads is installed as following

    $ git clone https://github.com/massivethreads/massivethreads.git
    $ cd massivethreads
    $ ./configure --prefix=$HOME/local CFLAGS="-Wall -O3"
    $ make
    $ make install
  • add its paths to environment variables:

    $ export PATH=$HOME/local/bin:$PATH:
    $ export C_INCLUDE_PATH=$HOME/local/include:$C_INCLUDE_PATH
    $ export CPLUS_INCLUDE_PATH=$HOME/local/include$CPLUS_INCLUDE_PATH
    $ export LIBRARY_PATH=$HOME/local/lib:$LIBRARY_PATH
    $ export LD_LIBRARY_PATH=$HOME/local/lib:$LD_LIBRARY_PATH
Qt5 installation:
  • given that Qt5's installation directory is following

    $HOME/local/Qt5.9.1
  • add Qt5's directory and qmake's path to environment variables

    QTDIR=$HOME/local/Qt5.9.1/5.9.1/gcc_64
    PATH=$QTDIR/bin:$PATH
SIP installation:
  • SIP can be installed with following procedure

    $ wget https://sourceforge.net/projects/pyqt/files/sip/sip-4.19.3/sip-4.19.3.tar.gz
    $ tar xf sip-4.19.3.tar.gz
    $ cd sip-4.19.3
    $ python3 configure.py --sysroot=$HOME/local --configuration myconfig.txt
    $ make
    $ make install
  • and the configuration file myconfig.txt has following contents

    # The target Python installation.
    py_platform = linux
    py_inc_dir = /usr/include/python%(py_major).%(py_minor)
    
    # Where SIP will be installed.
    sip_bin_dir = %(sysroot)/bin
    sip_inc_dir = %(sysroot)/include
    sip_module_dir = %(sysroot)/lib/python%(py_major)/dist-packages
    sip_sip_dir = %(sysroot)/share/sip
  • ensure SIP module is included in system's Python path, e.g.,

    $ export PYTHONPATH=$HOME/local/lib/python3/dist-packages:$PYTHONPATH
Python 3:
  • sometimes Python.h is not found, try adding its path to appropriate environment variables:

    $ export C_INCLUDE_PATH=/usr/include/python3.5:$C_INCLUDE_PATH
    $ export CPLUS_INCLUDE_PATH=/usr/include/python3.5:$CPLUS_INCLUDE_PATH
PyQt5 installation:
  • PyQt5 can be installed with following procedure

    $ wget https://sourceforge.net/projects/pyqt/files/PyQt5/PyQt-5.9/PyQt5_gpl-5.9.tar.gz
    $ tar xf PyQt5_gpl-5.9.tar.gz
    $ cd PyQt5_gpl-5.9
    $ python3 configure.py --sysroot=$HOME/local
    $ make -j
    $ make install
  • sometimes Python.h cannot be found despite C_INCLUDE_PATH and CPLUS_INCLUDE_PATH set above, use a configuration file to indicate the path to Python.h

    $ wget https://sourceforge.net/projects/pyqt/files/PyQt5/PyQt-5.9/PyQt5_gpl-5.9.tar.gz
    $ tar xf PyQt5_gpl-5.9.tar.gz
    $ cd PyQt5_gpl-5.9
    $ python3 configure.py --sysroot=$HOME/local --qmake $HOME/local/Qt5.9.1/5.9.1/gcc_64/bin/qmake --sip $HOME/local/bin/sip --sip-incdir $HOME/local/include --sipdir $HOME/local/share/sip/PyQt5 --bindir $HOME/local/bin --destdir $HOME/local/lib/python3/dist-packages --stubsdir $HOME/local/lib/python3/dist-packages/PyQt5 --configuration myconfig.txt
    $ make -j
    $ make install
  • and the configuration file myconfig.txt has following contents

    # The target Python installation.
    py_platform = linux
    py_inc_dir = /usr/include/python%(py_major).%(py_minor)
    qt_shared = True
    
    [Qt 5.9.1]
  • PyQt5 modules have been installed in $HOME/local/lib/python3/dist-packages

About

Visualizer for computational DAG of task parallel executions

Resources

Packages

No packages published