Re: [rvm-research] correctly measuring pause times

-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1

On 01/04/14 14:06, Joe Agajanian wrote:
> Hello,
> 
> My name is Joe Agajanian and I am a student at Harvey Mudd College.
> My team and I have created an incremental garbage collector in MMTK
> and now we are testing the performance. Specifically, we want to
> use the standard practice for measuring mutator pause times and
> throughput similar to the way it is measured in the paper /Ulterior
> Reference Counting: Fast Garbage Collection without a Long Wait /by
> Blackburn and McKinley. In our search for getting these
> measurements here are some of the things we tried:
> 
> * Using the provided statistics in mmtk/utility/statistics. *
> Searching the archives for this mailing list. * Reading the papers
> such as the one mentioned above to see how the pause times were
> measured. * In Plan.java we believe that printPostStats will show
> us throughput but not pause time.

Hi Joe,

I believe Steve used (wrote) the current MMTk statistics mechanism
with -X:gc:printPhaseStats=true in order to do the measurement for the
URC paper.  It may be necessary to increase the size of the phase
array (Stats.MAX_PHASES) for this to be useful, but as long as your
collector is a stop-the-world collector, the existing mechanism should
give enough information.

cheers

> 
> We were unsuccessful with the above attempts of measuring pause
> time so our backup plan is coding the measurement ourselves by
> using the built in timers whenever we call stopAllMutators, but we
> would only want to do this if there wasn't already a standard way
> for measuring pauses so we can get accurate comparisons to existing
> collectors.
> 
> We are guessing that in our search for this answer that we missed
> some guide that probably answers our question already. We would
> greatly appreciate it if someone could point us in the right
> direction. Thank you for taking the time to read our question.
> 
> Cheers, Joe Agajanian
> 
> 
> ------------------------------------------------------------------------------
>
> 
> 
> 
> _______________________________________________ 
> Jikesrvm-researchers mailing list 
> [email protected] 
> https://lists.sourceforge.net/lists/listinfo/jikesrvm-researchers
> 

-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1
Comment: Using GnuPG with Thunderbird - http://www.enigmail.net/

iQEcBAEBAgAGBQJTO6+aAAoJEHm9xBvoMhvaWoIIAMNPFTgxCnMz/xRTQNYDGPAp
RRmeEnnHy6C363IGBad0V82DahW47SugrmRjnpZbDtX2oaG4DLgixc7nEtpXy21e
DGzgSk7u9VBS9tDeYvSjQmyaVsSY9BQ3vsgfhORuptqSBOdfGopglMsrYjF6ddWL
h1VSSN03cMkcsX0lmlm2PAsMzu/swHgfeGe7k+Z6CCN97NSQOy4DemtIiuuHF1c7
AiXcMkUdYr3Jpw4zv+rxg2i0kGYxxVb7XFZtS90+KQIvs6qsQrm1b500GakOBEH6
zgoLeW6otWp9pxvnJOhrVsB+ew2XJU3DEzvpAxaCiKJfk+IIz9O/AwjjNCvsBz4=
=J+5X
-----END PGP SIGNATURE-----

------------------------------------------------------------------------------