harmony-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "Xiao-Feng Li" <xiaofeng...@gmail.com>
Subject Re: [doc] updated Harmony GC quick hacking guide
Date Sun, 20 Apr 2008 07:43:57 GMT
On Sat, Apr 19, 2008 at 4:49 PM, Prakhar Goyal <prakhargoyal@gmail.com> wrote:
> Xiao-Feng,
>
>  Thanks for the earlier reply.
>
>  I read your notes on parallel garbage collection given on your blog and
>  found it very interesting.
>
>  Mapping collector optimizes the phases further. Does it really affect the
>  optimization very much ?

Mapping Collector has not been implemented in Harmony. So I don't have
any data on its performance. Since it's used for major collection
(compacting), for applications that do not cause many major
collections, it may not be really beneficial.

>  I also followed the essentials by reading the file:
>  http://people.apache.org/~xli/presentations/parallel_garbage_collection.pdf
>
>  I really want to contribute to the ongoing work in GC. Kindly notify me of
>  the same, if any possibility appears.
>
>  I would be highly interested in joining as a contributor to this project in
>  Google Summer of Code.

GSoC is going to announce the supported projects. It looks like this
project is unlikely to be supported. Well, you are welcome to
contribute anytime, since Harmony is open to everybody.

>  Thanks & with regards,
>  Prakhar Goyal
>  Dept of Computer Science
>  IIT Bombay
>
>
>
>  On Thu, Apr 17, 2008 at 6:14 AM, Xiao-Feng Li <xiaofeng.li@gmail.com> wrote:
>
>  > Prakhar, thanks for your interests in this semispace GC in Harmony.
>  > Reply inlined below:
>  >
>  > On Wed, Apr 16, 2008 at 10:18 PM, Prakhar Goyal <prakhargoyal@gmail.com>
>  > wrote:
>  > > Xiao-Feng,
>  > >
>  > >  I found the material given on
>  > >
>  > http://xiao-feng.blogspot.com/2008/02/harmony-gc-internal-semi-space-garbage.html
>  > >
>  > >  very interesting. I further question the choice of age threshold as 1.
>  > We
>  > >  can further develop it further with age threshold different from1 so as
>  > to
>  > >  be more efficient. In that case, the factors of evaluation need to be
>  > >  computed.
>  >
>  > Age 1 might be not perfect, but it's adequate in my evaluations of
>  > typical benchmarks. That means, most objects cannot survive the second
>  > round of collection. To have bigger age threshold may not bring
>  > obvious benefit. For example, 100MB new objects may have only 5MB
>  > survivals after the first collection, and then 1MB after the second
>  > round. This is small enough for MOS to accommodate mostly, and to
>  > avoid major collections largely.
>  >
>  > >  Also, I would like to ask if the semi-space implementation is complete
>  > in
>  > >  all aspects. If not, I would be happy to know about remaining aspects,
>  > and I
>  > >  will soon start working on implementing them.
>  >
>  > Hmm, let me think. Then I will give you a reply.
>  >
>  > >  Regards,
>  > >  Prakhar Goyal
>  > >  IIT Bombay.
>  > >
>  > >  On Tue, Apr 15, 2008 at 11:47 AM, Senaka Fernando <senakafdo@gmail.com>
>  > >  wrote:
>  > >
>  > >
>  > >
>  > >  > Xiao-Feng,
>  > >  >
>  > >  > Thanks for this valuable presentation. This helps me a lot!
>  > >  >
>  > >  > Regards,
>  > >  > Senaka
>  > >  >
>  > >  > On Tue, Apr 15, 2008 at 4:16 AM, Xiao-Feng Li <xiaofeng.li@gmail.com>
>  > >  > wrote:
>  > >  >
>  > >  > > I made some updates, esp. on the contract between VM and GC. The
>  > >  > > contract is the key for GC porting.
>  > >  > >
>  > >  > > http://people.apache.org/~xli/presentations/harmony_gc_source.pdf<http://people.apache.org/%7Exli/presentations/harmony_gc_source.pdf>
>  > <http://people.apache.org/%7Exli/presentations/harmony_gc_source.pdf>
>  > >  > <http://people.apache.org/%7Exli/presentations/harmony_gc_source.pdf>
>  > >  > >
>  > >  > > Thanks,
>  > >  > > xiaofeng
>  > >  > >
>  > >  > > --
>  > >  > > http://xiao-feng.blogspot.com
>  > >  > >
>  > >  >
>  > >
>  >
>  >
>  >
>  > --
>  > http://xiao-feng.blogspot.com
>  >
>



-- 
http://xiao-feng.blogspot.com

Mime
View raw message