harmony-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "Xiao-Feng Li" <...@apache.org>
Subject Re: [drlvm][gc]Mark bit design
Date Wed, 23 Apr 2008 10:26:12 GMT
On Wed, Apr 23, 2008 at 5:15 PM, lyon <ustcliao@gmail.com> wrote:
> hi, Xiao-Feng Li
>  In your blog article "[Harmony GC Internal] Mark bit design" ,
>  you write that "They might not know each other, and continue to scan the
>  slots of the marked object."
>  (The third paragraph). If they continue to scan the slots of the marked
>  object, they would push the same
>  p_ref into the collector tracestack.Is this the redundant work? Will
>  this situation cause to be wrong?

It's redundant, but not wrong. In my experience, the redundancy is negligible.

>  And, according to some details with partial-forward ,you ues the AGE_BIT
>  to indicate the objects are
>  uncopied objects.
>  I find out another method. the old copy of a forwarded object has
>  FORWARD_BIT set. The those live
>  object left uncopied was also set FORWARD_BIT(The original method is
>  that uncopied obj was set MARK_BIT )
>  But The those live object left uncopied was not forwarded to mos
>  really,They still keep the original place.
>  The Mark_bit of the remained objs will be not set. Then the bit flip.
>  Those uncopied objects will have the clean FORWARD_BIT(original mark_bit
>  is not set ) automatically,
>  I think it solve the problem,
>  but I do not know whether it has affected other algorithms in the next
>  collection. Is this method feasible

It's ok to use any of following bits: MARK_BIT, FORWARD_BIT, or
AGE_BIT, as long as the algorithm is well designed.

>  thank you.
>  liaoyin


View raw message