Re: Soot 1.2.2 Performance Issues

[Aske Simon Christensen <aske@brics.dk>]

Hi Nathan!

Nathan Gulley wrote:

> I would like to receive input on the various ways you all have been using
> Soot in projects and wheather or not you have encountered any issues with
> its performance.

We have used Soot for loading class files and iterate through the code of all methods to convert it into an internal format for later processing. In our test runs we have used a dual 1GHz Pentium III with 1GB ram.

The total time for loading and traversing all class files ranges from 3.2s for tiny programs to 3.6s for larger programs generated from a couple of thousand lines of Java code. We conclude from these numbers that Soot indeed has a very slow starup time, but once it is up running, the performance seems reasonable.

In our case this slow startup makes little difference, as our processing of the output of this initial conversion takes somewhat more time. But if a single run loads only a few class files to do simple processing on them, it seems that Soot could easily be a major bottleneck.

> I would like to know if anyone else has encountered this performance
> bottleneck or performance bottlenecks in other uses. Also, any advice on
> how to overcome it is welcome.

I don't know if there is any way to make it run faster, apart from doing as much as possible in each run. It seems that even the basic operations incur this delay.

-- 
Aske Simon Christensen  aka  Blueberry