Skip to content

Latest commit

 

History

History
96 lines (73 loc) · 2.28 KB

README.md

File metadata and controls

96 lines (73 loc) · 2.28 KB

Python Installation

From PyPI (once published)

pip install sharedmap

From source

# Clone the repository
git clone https://github.com/HenningWoydt/SharedMap.git
cd SharedMap

# Build dependencies first
./build.sh

# Install Python package
pip install .

Usage

import sharedmap as sm

# Define your graph in CSR format
v_weights = [1, 1, 1, 1, 1, 1, 1, 1]
adj_ptrs = [0, 3, 6, 8, 12, 16, 19, 21, 22]
adj_weights = [1] * 22
adj = [1, 2, 3, 0, 3, 4, 0, 3, 0, 1, 2, 5, 1, 5, 6, 7, 3, 4, 6, 4, 5, 4]

# Define hardware hierarchy and distances
hierarchy = [2, 2]  # 2 nodes, 2 cores per node
distance = [1, 10]  # distance within node: 1, between nodes: 10

# Partition the graph
comm_cost, partition = sm.partition_graph(
    v_weights, adj_ptrs, adj_weights, adj,
    hierarchy, distance,
    imbalance=0.03,
    n_threads=1,
    seed=0
)

print(f"Communication Cost: {comm_cost}")
print(f"Partition: {partition}")

API Reference

partition_graph()

Main function for graph partitioning.

Parameters:

  • v_weights: Vertex weights (array of integers >= 1)
  • adj_ptrs: CSR adjacency pointers (n+1 integers)
  • adj_weights: Edge weights (array of integers >= 1)
  • adj: CSR adjacency list
  • hierarchy: Hierarchy levels (e.g., [4, 8, 6])
  • distance: Communication distances (e.g., [1, 10, 100])
  • imbalance: Maximum allowed imbalance (default: 0.03)
  • n_threads: Number of threads (default: 1)
  • seed: Random seed (default: 0)
  • strategy: Thread distribution strategy (default: Strategy.NB_LAYER)
  • parallel_alg: Parallel algorithm (default: Algorithm.MTKAHYPAR_QUALITY)
  • serial_alg: Serial algorithm (default: Algorithm.KAFFPA_STRONG)
  • verbose: Print statistics (default: False)

Returns:

  • comm_cost: Communication cost J(C, D, PI)
  • partition: Partition assignment for each vertex (numpy array)

Enums

Strategy:

  • Strategy.NAIVE
  • Strategy.LAYER
  • Strategy.QUEUE
  • Strategy.NB_LAYER

Algorithm:

  • Algorithm.KAFFPA_FAST
  • Algorithm.KAFFPA_ECO
  • Algorithm.KAFFPA_STRONG
  • Algorithm.MTKAHYPAR_DEFAULT
  • Algorithm.MTKAHYPAR_QUALITY
  • Algorithm.MTKAHYPAR_HIGHEST_QUALITY

validate_input()

Validates input parameters before partitioning.

Returns: bool - True if input is valid, False otherwise.