Requirements:
- GNU Make 3.81+
- JDK 1.7+

for windows install cygwin with make+bash features

To compile:
  make 
  (if you run in trouble its usually because of CRLF dirt) E.g. scalac did not run because it was CRLF

To run

  bash run-bench.sh
  
Parameters of the default run are hardcoded in run-bench.sh

To update results of subset of serializers

  bash run-bench.sh serializer1,serializer2,..
  
How this works:
  
  the project was somewhat messy, i probably contributed to that .. but its easier to manage now.
  I removed the config files to avoid problems when serializers rename or are un/commented. Additionally
  each benchmark runs in a separate VM.
  
  1) the run-bench.sh uses class BenchmarkExporter.java to get a ',' separated list of all serializers to run.
     All benchmarks registered in the BenchnarkRunner.java class a run be default.
  2) after each bench finished, the mk-stats.sh script aggregates the results found in ./results/raw to ./results/stats.txt
     the the class StatsCruncher.java creates a Textile style report.
     To modify the report output, directly edit the StatsCruncher.java . You are welcome to enhance this using
     some templating/configuration files.
     You can copy the results/report.texfile directly to github wiki in textile-mode.
  3) You can run the mk-stats.sh script again without having to re-run the benchmarks if you want to add
     custom query charts.
  4) to filter out certain libraries, do not uncomment them in BenchMarkRunner.java (except they fail), better
     edit StatsCruncher.java to exclude them.