[Soot-list] Sizes of benchmarks in Dacapo-2006-MR2
Richard Xiao
richardxx at cse.ust.hk
Sun Feb 10 23:43:27 EST 2013
Hi, Zhoulai:
The statistics in my report is obtained with following options:
opts.add("-p");
opts.add("cg");
opts.add("implicit-entry:false");
opts.add("-p");
opts.add("cg");
opts.add("safe-forname:false");
opts.add("-p");
opts.add("cg");
opts.add("safe-newinstance:false");
opts.add("-p");
opts.add("cg.spark");
opts.add("enabled:true");
opts.add("-p");
opts.add("cg.spark");
opts.add("merge-stringbuffer:false");
opts.add("-p");
opts.add("cg.spark");
opts.add("string-constants:false");
opts.add("-p");
opts.add("cg.spark");
opts.add("dump-pag:false");
opts.add("-p");
opts.add("jb");
opts.add("use-original-names:false");
opts.add("-p");
opts.add("jb.ulp");
opts.add("enabled:false");
I don't use the soot option "-process-dir". Instead, I only give the entry
class "dacapo.chart.Main" to Soot.
The LOC is counted just as you did, using
sm.getActiveBody().getUnits().size().
The benchmark in my report is analyzed with the latest version of JDK 1.6.
Any feedback is welcome.
Regards,
Xiao
On Sat, Feb 9, 2013 at 3:58 PM, Zhoulai <zell08v at orange.fr> wrote:
> Thanks.
>
> I agree. The measure of program size certainly should be better defined
> w.r.t the analysis scope. Unfortunately, it may not be trivial to find out
> this info. for some reports. By switching on/off different SOOT options, I
> get different sizes for the same benchmark, which is not surprising.
> However, I don't see which soot options or other kinds of parameters are
> set in some reports with this regard.
>
> For example, I don't see which options are set by Xiao to get 384K LOC
> for 'chart' (p27 in www.cse.ust.hk/~richardxx/papers/geom-tr.pdf )
>
> In the caption, it reads " ... The LOC column counts the lines of code for
> the Jimple IR. ... ".
>
> I count 95337 Jimple units. Admitted, counting Jimple units are different
> from counting Jimple statements (see Eric's earlier reply on my naive
> question). But I still do not see why I get only 1/4 of the published
> report.
>
> I was counting by simple iterating through each individual method of the
> classes in process-dir, using the options below:
> String[] dacapoArgs={
> "-w",
> "-f","n",
> "-p", "cg.spark","enabled:true",
> "-allow-phantom-refs",
> "-process-dir", processDir,
> "-cp", sootClassPath,
>
> "-include",
> "org.apache.",
> "-include",
> "org.w3c.",
>
> "-main-class", className
> };
>
> in which, processDir is the full path of "chart.jar" in dacapo-2006-MR2,
> className is dacapo.chart.Main
>
> Zell.
>
>
> On Fri, Feb 8, 2013 at 7:09 PM, Tony Yan <yan at cse.ohio-state.edu> wrote:
>
>> The stats could differ when different JDK versions are used.
>>
>> Tony
>>
>> On Fri, Feb 8, 2013 at 1:05 PM, Phil Pratt-Szeliga <pcpratts at trifort.org>
>> wrote:
>> > Hello,
>> >
>> > For lines of code count you can try sloccount on debian/ubuntu.
>> >
>> > Phil Pratt-Szeliga
>> > Syracuse University
>> > http://trifort.org/
>> >
>> > On Fri, Feb 8, 2013 at 12:56 PM, Patrick Lam <plam at sable.mcgill.ca>
>> wrote:
>> >> There's a lot of dynamic loading going on, and it would be good to make
>> >> sure that you're including the dynamically-loaded classes.
>> >>
>> >> pat
>> >>
>> >> On 02/08/13 12:35, Zhoulai wrote:
>> >>> Thanks. I did it.
>> >>>
>> >>> Here is the Jimple units number I found
>> >>>
>> >>> bloat: 74552
>> >>> chart 95337
>> >>> eclipse 62025
>> >>> ....
>> >>>
>> >>> But it seems that these numbers are much smaller than what is reported
>> >>> by other researchers. For e.x. Richard Xiao in his work reported 350K
>> >>> Jimple LOC for the benchmark "chart ". This puzzles me because I am
>> >>> using the simplest way to count, i.e., iterating over all classes
>> using
>> >>> -process-dir options, iterating over each methods, then counting the
>> >>> number of units. Although Jimple units are not LOC, I don't see why
>> the
>> >>> numbers I got are so small....
>> >>>
>> >>>
>> >>> Zell
>> >>>
>> >>>
>> >>>
>> >>> On Fri, Feb 8, 2013 at 6:15 PM, Patrick Lam <plam at sable.mcgill.ca
>> >>> <mailto:plam at sable.mcgill.ca>> wrote:
>> >>>
>> >>> I think that what Eric is trying to say here is that he's
>> encouraging
>> >>> you to try to think about how you could figure this out before
>> asking
>> >>> the list. It should be pretty simple to write some code to count
>> sizes.
>> >>>
>> >>> pat
>> >>>
>> >>> On 02/08/13 12:14, Zhoulai wrote:
>> >>> > Thanks for your quick reply.
>> >>> >
>> >>> >
>> >>> > Would you like to give more info. such as "where can we find
>> out
>> >>> these
>> >>> > infos"? Thanks.
>> >>> >
>> >>> > Zell.
>> >>> >
>> >>> >
>> >>> > On Fri, Feb 8, 2013 at 5:51 PM, Eric Bodden
>> >>> <eric.bodden at ec-spride.de <mailto:eric.bodden at ec-spride.de>
>> >>> > <mailto:eric.bodden at ec-spride.de
>> >>> <mailto:eric.bodden at ec-spride.de>>> wrote:
>> >>> >
>> >>> > Hello.
>> >>> >
>> >>> > > I would like to find out the sizes (source or Jimple) of the
>> >>> > benchmarks of
>> >>> > > Dacapo-2006-MR2. Is that possible?
>> >>> >
>> >>> > Yes, it's possible.
>> >>> >
>> >>> > SCNR,
>> >>> > Eric
>> >>> >
>> >>> >
>> >>> >
>> >>> >
>> >>> > _______________________________________________
>> >>> > Soot-list mailing list
>> >>> > Soot-list at sable.mcgill.ca <mailto:Soot-list at sable.mcgill.ca>
>> >>> > http://mailman.cs.mcgill.ca/mailman/listinfo/soot-list
>> >>>
>> >>> _______________________________________________
>> >>> Soot-list mailing list
>> >>> Soot-list at sable.mcgill.ca <mailto:Soot-list at sable.mcgill.ca>
>> >>> http://mailman.cs.mcgill.ca/mailman/listinfo/soot-list
>> >>>
>> >>>
>> >>
>> >> _______________________________________________
>> >> Soot-list mailing list
>> >> Soot-list at sable.mcgill.ca
>> >> http://mailman.cs.mcgill.ca/mailman/listinfo/soot-list
>> >>
>> >>
>> > _______________________________________________
>> > Soot-list mailing list
>> > Soot-list at sable.mcgill.ca
>> > http://mailman.cs.mcgill.ca/mailman/listinfo/soot-list
>>
>>
>>
>> --
>> Dacong (Tony) Yan
>> Ph.D. Student
>> Computer Science and Engineering
>> The Ohio State University, Columbus
>> http://www.cse.ohio-state.edu/~yan
>>
>
>
> _______________________________________________
> Soot-list mailing list
> Soot-list at sable.mcgill.ca
> http://mailman.cs.mcgill.ca/mailman/listinfo/soot-list
>
>
--
Richard Xiao Xiao
PhD Student @ CSE @ Hong Kong University of Science and Technology
www.cse.ust.hk/~richardxx
-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://mailman.cs.mcgill.ca/pipermail/soot-list/attachments/20130211/86cfa387/attachment.html
More information about the Soot-list
mailing list