Benchmark Analyzer

Format requirements: The .csv file must be in the following format:

Each column header must be a label for one test run. (i.e. the name of the program or a version number, version 6 below) All columns with blank headers will be ignored.
Each row header must be a label for one particular test. The label must consist of different aspects of the test separated by colons. (xSize10:ySize10:methodFloat below)

Sample analysis: Consider this sample set of benchmarks of a piece of software:

	version 5	version 6
xSize10:ySize10:methodFloat	10.3	6.7
xSize100:ySize10:methodFloat	98.8	95.4
xSize10:ySize100:methodFloat	97.4	64.7
xSize100:ySize100:methodFloat	1002.2	945.1
xSize10:ySize10:methodInt	5.3	3.4
xSize100:ySize10:methodInt	48.5	47.7
xSize10:ySize100:methodInt	49.5	32.9
xSize100:ySize100:methodInt	501.1	489.5

If we assume each data point is an average of 5 samples, we get this sample analysis. Note that we can clearly see that for x size=10 we get a good speedup over version 5, but for x size=100 we don't.

About this project: This project uses R to generate the box plots and decision trees, and uses RPy to interface with R from Python.

Code files:

analyzebenchmarks.py - main Python script that generates the report.
doanalysis.cgi - CGI script wrapper around analyzebenchmarks.py