Short manual of triangleramsey
------------------------------
Author: Jan Goedgebeur (jan.goedgebeur@ugent.be)
In collaboration with: Gunnar Brinkmann (gunnar.brinkmann@ugent.be)
Triangleramsey is a program for the efficient generation of maximal triangle-free graphs.
Triangleramsey can also be used to generate triangle Ramsey graphs for a given testgraph efficiently.
This program has been tested on Linux and Mac OS X.
Installation
------------
The latest version of triangleramsey can be obtained from http://caagt.ugent.be/triangleramsey/
First download, extract and configure nauty from http://cs.anu.edu.au/~bdm/nauty/
Triangleramsey requires nauty 2.5 (or a more recent version).
Then copy the following files to the triangleramsey directory:
naugraph.c nautil.c nauty.c nauty.h naurng.c naurng.h schreier.c schreier.h sorttemplates.c
You can make a triangleramsey binary using "make".
If you want to generate graphs with more than 32 vertices, make a binary using "make 64bit".
This "64-bit binary" can also be used to generate graphs with <= 32 vertices, but is (slightly) slower than the regular binary.
Usage
-----
An overview of all options can also be found by executing "./triangleramsey -h".
Usage: ./triangleramsey [options]
At least the number of vertices be specified.
The generated graphs are written to stdout and are encoded in multicode format (see Appendix A).
Possible options are:
n: No graphs will be output (i.e. the graphs are only counted).
file : Starts the generation process from the graphs in ,
else the generation is started from P3.
ramsey : Generates ramseygraphs for the graph in .
write_all_ramseygraphs: Can only be used in combination with option 'ramsey'.
Generates all mtf ramseygraphs for the testgraph instead of
stopping when one ramseygraph was found.
When this option is used, the ramseygraphs are written to
a file instead of to stdout (unless if option 'n' is used).
mod : Splits the generation in (more or less equally big) parts.
Here part will be executed.
Splitting the generation causes a small overhead,
so the sum of the timings for the small parts will be slightly more
than the time needed to run the same case without modulo.
But this overhead is usually negligible compared to the
total execution time.
The normal rules for modulo calculation apply. So m 0 2 will give
the same result as m 0 4 and m 2 4 combined.
Examples
--------
"./triangleramsey 18 outputlevel 17"
Generates all maximal triangle-free graphs with at most 18 vertices and outputs the graphs with at least 17 vertices to stdout.
"./triangleramsey 21 mod 10 200"
Splits the generation of all maximal triangle-free graphs with 21 vertices into 200 (more or less equally big) parts and executes part 10.
"./triangleramsey 25 ramsey testgraph.mc"
Tries to generate a triangle Ramsey graph with 25 vertices for the graph in "testgraph.mc.
The program stops as soon as a Ramsey graph was found. The program writes 1 to stdout if no Ramsey graph was found.
"./triangleramsey 25 ramsey testgraph.mc write_all_ramseygraphs"
Generates all mtf triangle Ramsey graphs with 25 vertices for the graph in "testgraph.mc.
The Ramsey graphs are written to "Ramseygraphs_25_testgraph.mc".
The program writes 1 to stdout if no Ramsey graphs were found.
"./triangleramsey 25 ramsey testgraph.mc write_all_ramseygraphs file inputgraphs.mc"
Generates all mtf triangle Ramsey graphs with 25 vertices for the graph in "testgraph.mc.
The construction is started from the graphs in inputgraphs.mc instead of from P3.
Don't hesitate to contact me at jan.goedgebeur@ugent.be if you have any further questions or suggestions.
Changelog
---------
2012-06-28: First release.
2017-07-06: Fixed rare bug wrt "write_all_ramseygraphs" option reported by Manfred Scheucher.
Appendix A: definition of the multicode format
----------------------------------------------
This code is binary. The first entry is the number of vertices.
Vertices are numbered 1,...,n. For each vertex x there is a list of
neighbours with higher labels than x, followed by a zero.
The last list is always empty (there are no neighbours of n with a higher number than n),
therefore the last "list" is not followed by a zero.
After the last byte the next graph follows.
The codelength of a graph in multicode is number of vertices + number of edges.