Quicktree Parallelism

We parallelize the implementation of the Neighbor-Joning (NJ) algorithm based on QuickTree.

Table of Contents

• Quicktree Parallelism
  • Table of Contents
  • Dataset
  • Requirements
  • Build Instructions
  • Run the Program
  • Execution Workflow on DSMLP Server

Dataset

Distance matrix with size 1K, 5K, and 10K are available within the data folder in the repository. Distance matrix in 20K size is available on Google Drive. Please download and move it within the data folder.

Requirements

The following dependencies are required:

• C++ compiler (GCC / Clang)
• CMake
• CUDA Toolkit

Build Instructions

Create a build directory and compile the project:

mkdir -p build
cd build
cmake ..
make -j4

This will generate the executable:

quicktree

Run the Program

Run the program using a distance matrix in PHILIP format as input (output will be printed to the console):

./quicktree -m ../data/matrix_10k.phy

Run the program using a distance matrix in PHILIP format as input (output will be written to the file):

./quicktree -m ../data/matrix_10k.phy -o tree.nwk

The program arguments are as follows:

Argument	Description
-m	input distance matrix (must in PHILIP format)
-o	output phylogenetic tree in Newick format

Execution Workflow on DSMLP Server

1. Clone the project repository in HOME directory, and decompress the data files

    git clone https://github.com/wangly7/quicktree-parallelism.git
    cd quicktree-parallelism/data
    unzip matrix_10k.phy.zip
    cd ..

2. Run the code

   /opt/launch-sh/bin/launch.sh -v a30 -c 8 -g 1 -m 8 -i yatisht/ece213-wi26:latest -f ./quicktree-parallelism/run-commands.sh