The original Green Marl README is stored in README.GM.md

How to run

This guide assumes that you already downloaded and installed shoal.

1. Clone the repository: git clone [email protected]:libshoal/Green-Marl.git gm in the same folder then the Shoal library (i.e. libshoal/ and gm/ are in the same directory)
2. Make Shoal available in Green Marl: cd gm && ln -s ../libshoal shoal
3. Install Green Marl's dependencies as listed in README.GM.md, Section 3-1.
4. Configure GM_TOP in setup.mk to point to Green Marl's root directory
5. Build the Green Marl compiler make compiler -j24 - this includes Shoal's compiler extensions
6. Build the runtime library, libgm, $(cd apps/output_cpp/gm_graph/avro-c-1.7.2 && make), followed by cd apps/output_cpp/gm_graph; make; cd ../../..
7. Compile pagerank: make sk_pagerank
8. Run pagerank: scripts/run.sh pagerank $(nproc) ours soc-LiveJournal1. Graphs have to be stored in ../graphs/ relative to the Green Marl directory.

Acquiring a graph

In the paper, we used the Twitter and LiveJournal graphs as workload for evaluation.

Twitter

ToDo

LiveJournal

A smaller graph is LiveJournal, which can be acquired from here.

The unzipped version of the graph is a plain-text representation. For Green Marl, we convert this to a binary representation with the integrated converter:

apps/output_cpp/bin/gm_format_converter soc-LiveJournal1.txt soc-LiveJournal1.bin schema -GMInputFormat=EDGE

The md5 checksum for our version of the LiveJournal graph is: 8447a521d15781478a46e97ee2fcac60

Note on 4M-pages

Shoal supports 4M pages. See the shoal documentation on how to setup machines to use these.