How anudeep512/nsw-parallelisation works
Project type and purpose
• This is a C++‐based research/benchmarking tool that parallelises the NSW (Navigable Small World) graph search algorithm.
• It provides three modes of operation:
– seq.cpp: sequential build & query
– sam.cpp: static allocation of data chunks to threads
– dam.cpp: dynamic allocation of data chunks to threads
File structure
Top‐level items:
• Dataset/ : input text files (100.txt, 1000.txt, …)
• seq.cpp, sam.cpp, dam.cpp : three implementations of the same pipeline
• params.txt : configuration (e.g., graph parameters, thread counts)
• bash_1.sh, bash_2.sh, bash_3.sh : automation scripts for compiling/running each mode
• Reports & presentations (PDFs) for documentation
Architectural implications:
• A clear data‐processing pipeline: Data Loader → Graph Builder → Search Engine → Result Collector
• Two parallelization strategies distinguished only by how work is assigned to threads
README for additional insights
• The project focuses on two parallel strategies (SAM vs. DAM) vs. the baseline sequential version.
• Datasets vary in size to measure scaling and speed‐up.
• Scripts orchestrate compilation and runs with different inputs/parameters.