Skip to content

Installation

Jump to bottom
Abdullah edited this page Feb 27, 2026 · 19 revisions

Installation Guide

This guide covers installing GraphBrew on Linux systems.

🚀 Quick Install (One-Click)

The fastest way to get started - automatically checks dependencies, builds, and downloads test graphs:

git clone https://github.com/UVA-LavaLab/GraphBrew.git
cd GraphBrew

# Check what dependencies are missing
python3 scripts/graphbrew_experiment.py --check-deps

# Auto-install missing dependencies (needs sudo)
python3 scripts/graphbrew_experiment.py --install-deps

# Run full pipeline
python3 scripts/graphbrew_experiment.py --full --size small

This will:

  • Verify all system dependencies (Boost, g++, libnuma, etc.)
  • Build all binaries automatically
  • Download benchmark graphs from SuiteSparse
  • Run the complete experiment pipeline

System Requirements

Minimum Requirements

  • OS: Linux (Ubuntu 20.04+ recommended)
  • Compiler: GCC 7+ with C++17 support
  • Memory: 8GB RAM (16GB+ recommended for large graphs)
  • Disk: 50GB+ for benchmark graphs

Automatic Dependency Check

GraphBrew can check and install dependencies automatically:

# Check all dependencies
python3 scripts/graphbrew_experiment.py --check-deps

# Sample output:
# GraphBrew Dependency Status:
# ==================================================
#   ✓ make             GNU Make 4.3
#   ✓ c++ compiler     g++ 11.4.0
#   ✓ boost            v1.74
#   ✓ numa             found
#   ✓ tcmalloc         found
#   ✓ python           v3.10.12
# ==================================================
#   All required dependencies satisfied!

# Install missing dependencies (Ubuntu/Debian/Fedora/macOS)
python3 scripts/graphbrew_experiment.py --install-deps

Manual Dependency Installation

If automatic installation doesn't work, install manually:

# Ubuntu/Debian
sudo apt-get update
sudo apt-get install -y \
    build-essential \
    g++ \
    libboost-all-dev \
    libnuma-dev \
    google-perftools \
    python3 \
    python3-pip

# Fedora/RHEL
sudo dnf install -y \
    gcc-c++ \
    make \
    boost-devel \
    numactl-devel \
    gperftools \
    python3 \
    python3-pip

# Arch Linux
sudo pacman -S \
    base-devel \
    boost \
    numactl \
    gperftools

# macOS (with Homebrew)
xcode-select --install
brew install gcc boost google-perftools

Optional Dependencies

# For visualization and analysis (optional)
pip3 install numpy matplotlib pandas

Manual Build

Clone the Repository

git clone https://github.com/UVA-LavaLab/GraphBrew.git
cd GraphBrew

Basic Build

# Build all benchmarks
make all

# Build with cache simulation support
make sim

# Or build specific benchmarks
make pr    # PageRank
make bfs   # Breadth-First Search
make cc    # Connected Components

Build Options

Option Description Example
RABBIT_ENABLE=0 Disable Rabbit Order (enabled by default) RABBIT_ENABLE=0 make all
DEBUG=1 Build with debug symbols DEBUG=1 make all
SANITIZE=1 Enable address sanitizer SANITIZE=1 make all

Rabbit Order and Boost 1.58

Rabbit Order (algorithm 8) has two variants:

  • csr (default): Native CSR implementation - faster, no external dependencies
  • boost: Original Boost-based implementation - requires Boost 1.58.0

Note: The csr variant is used by default and does not require Boost. Only install Boost 1.58 if you need the original boost variant.

The boost variant requires Boost 1.58.0 specifically for compatibility. System package managers often install newer versions which may cause issues.

Automatic Installation (Recommended)

# Download, compile, and install Boost 1.58.0 to /opt/boost_1_58_0
python3 scripts/graphbrew_experiment.py --install-boost

This automatically:

  1. Downloads Boost 1.58.0 source (~73MB)
  2. Runs bootstrap.sh to configure
  3. Compiles with b2 using all CPU cores
  4. Installs to /opt/boost_1_58_0 with proper include/ and lib/ structure

Note: Compilation takes 10-30 minutes depending on your system.

Manual Installation

# Download Boost 1.58.0
wget https://archives.boost.io/release/1.58.0/source/boost_1_58_0.tar.gz
tar -xzf boost_1_58_0.tar.gz
cd boost_1_58_0

# Configure
./bootstrap.sh --prefix=/opt/boost_1_58_0

# Compile and install (uses all CPU cores)
cpuCores=$(nproc)
sudo ./b2 --with=all -j $cpuCores install

# Cleanup
cd .. && rm -rf boost_1_58_0 boost_1_58_0.tar.gz

Verify Boost Installation

# Check Boost version at GraphBrew expected path
cat /opt/boost_1_58_0/include/boost/version.hpp | grep "BOOST_LIB_VERSION"
# Should show: #define BOOST_LIB_VERSION "1_58"

# Or use the dependency checker
python3 scripts/graphbrew_experiment.py --check-deps

Building with Rabbit Order

# Default build (RabbitOrder enabled, requires Boost 1.58 + libnuma + tcmalloc)
make all

# Build WITHOUT RabbitOrder (no Boost/libnuma/tcmalloc needed)
RABBIT_ENABLE=0 make all

Note: RABBIT_ENABLE=1 is the Makefile default. Building with RabbitOrder requires Boost headers (1.58+), libnuma, and google-perftools (tcmalloc).

Verify Installation

# Check if binaries were created
ls -la bench/bin/

# Run a quick test
./bench/bin/pr -g 12 -o 0 -n 1

Expected output:

Generate Time:       0.02xxx
Build Time:          0.01xxx
...
Trial Time:          0.00xxx

Python Environment

Core scripts require only Python 3.8+ standard library. Optional: pip3 install numpy matplotlib pandas for visualization.

python3 scripts/graphbrew_experiment.py --help  # Verify scripts work

Directory Structure After Build

See Code-Architecture for full directory layout. Key paths:

  • bench/bin/ — Compiled binaries (pr, bfs, cc, etc.)
  • bench/bin_sim/ — Cache simulation binaries
  • scripts/lib/ — Core Python modules
  • results/data/ — Runtime model storage (perceptron weights, DT, hybrid, kNN models in adaptive_models.json), auto-created on first run by ensure_prerequisites()
  • results/ — Experiment outputs

Troubleshooting Build Issues

Missing Boost Headers

# Error: fatal error: boost/range/algorithm.hpp: No such file or directory
sudo apt-get install libboost-all-dev

Compiler Version Too Old

# Check GCC version
g++ --version

# Install newer GCC
sudo apt-get install g++-11
export CXX=g++-11
make clean && make all

OpenMP Not Found

# Error: -fopenmp not supported
sudo apt-get install libomp-dev

Out of Memory During Build

# Reduce parallel jobs
make -j2 all  # Instead of default parallel

Next Steps

← Back to Home | Quick Start →

Clone this wiki locally