LLNL/CallFlow

Name: CallFlow

Owner: Lawrence Livermore National Laboratory

Description: Visualization tool for Calling Context Trees

Created: 2017-11-01 13:54:20.0

Updated: 2018-02-12 16:59:33.0

Pushed: 2017-12-08 04:14:43.0

Homepage:

Size: 3619

Language: JavaScript

GitHub Committers

UserMost Recent Commit# Commits

Other Committers

UserEmailMost Recent Commit# Commits

README

CallFlow 1.0

CallFlow is a visualization tool to view and analyze Calling Context Tree (CCT) data provided by HPCToolkit. CallFlow uses Sankey diagrams to display the CCT hierarchy.

Requirements

You will need to install the latest version of Node.js (v7.7.2 or higher).

Installation

To install, change directory to the src/ folder and run:

pm install
ower install

If you get a libtool related error during libxlmjs build in the first step on Mac OSX, you should try again after removing the macports version of libtool from your path.

To Run

In the src/ directory, run:

ode main.js -d /path/to/datasetinfo.json

Point browser to http://localhost:8500

Inputs

The application requires 3 files with 1 optional file:

  1. Dataset information file: This is a json file that specifies where to look for the data. It requires 3 fields with 1 optional field. The format is as follows:

    
    h" : /path/to/directory/contains/dataset/ (Note the '/' at the end of the path)
    eriment" : Name of the xml file
    eMetric" : Name of the json file that contains the metrics for each call tree node. Detail of the format for this file is bellow.
    fig" : Name of the config file. This is optional. Detail of the format for this file is bellow.
    
    
  2. XML file : This is the xml file for the call tree data.

  3. Node metric file: A json file that contains the metrics for each call tree node Each line in the file contains the following information:

    
     : The call tree node id
    " : An array of inclusive runtimes
    " : An array of exclusive runtimes
    
    
  4. Configuration file: A configuration file to tell the parser how to parse some problematic load modules (this is optional). The format of the file is follows:

    d module name" : {
     : The id of the load module, taken form the xml file
    es" : An array of file prefixes
    ctions" : {
    "name of function" : An array of name prefixes to match under this function
    
    
    
Reference

Any published work that utilizes this software should include the following reference:

Tan Nguyen, Abhinav Bhatele, Peer-Timo Bremer, Todd Gamblin, Martin Schulz,
Wei, David Böhme, and Kwan-Liu Ma. VIPACT: A visualization interface for
yzing calling context trees. In Proceedings of the 3rd Workshop on Visual
ormance Analysis, VPA '16, November 2016. LLNL-CONF-704659.
Copyright

Copyright (c) 2017, Lawrence Livermore National Security, LLC. Produced at the Lawrence Livermore National Laboratory.

Written by Huu Tan Nguyen htpnguyen@ucdavis.edu.

LLNL-CODE-740862. All rights reserved.


This work is supported by the National Institutes of Health's National Center for Advancing Translational Sciences, Grant Number U24TR002306. This work is solely the responsibility of the creators and does not necessarily represent the official views of the National Institutes of Health.