This academic project serves as a platform for benchmarking graph algorithms across multiple libraries

1. PageRank:

   • Implementation using Gunrock (Elkin)
   • Implementation using PySpark (Pavlushkin)
   • Implementation using GraphBLAS (Shishin)

2. Single Source Parent BFS:

   • Implementation using Gunrock (Elkin)
   • Implementation using PySpark (Pavlushkin)
   • Implementation using GraphBLAS (Shishin)

3. Multi Source Parent BFS:

   • Implementation using Gunrock (Elkin)
   • Implementation using PySpark (Pavlushkin)
   • Implementation using GraphBLAS (Shishin)

git clone https://github.com/tepa46/graphs_analysis.git

In the /datasets directory, you can find one of the prepared graphs: Email-Enron.

The format is as follows:

node_from_0\tnode_to_0  
node_from_1\tnode_to_1  
...  
node_from_n\tnode_to_n  

You can add your own graphs for benchmarking or use datasets from the Stanford Large Network Dataset Collection.

See benchmarks launch instructions

1. Determine the performance gain provided by GPU multithreading compared to CPU multithreading.
2. Analyze how the runtime of MSBFS increases within the same library when varying the number of starting vertices, relative to SSBFS.

See experimental results