opensubscriber
   Find in this group all groups
 
Unknown more information…

m : mit-scheme-devel@gnu.org 21 September 2011 • 5:47AM -0400

Re: [MIT-Scheme-devel] Cleaning weak hash tables in secondary GC daemon?
by Taylor R Campbell

REPLY TO AUTHOR
 
REPLY TO GROUP




   Date: Tue, 20 Sep 2011 14:21:22 -0700
   From: Matt Birkholz <matt@birk...>

   I find trigger-secondary-gc-daemons! is called only in gc-clean,
   which is called only in runtime/make.scm and disk-save.

GC-FINISH also calls GC-CLEAN when there are fewer than 4096 words
remaining in the heap after garbage collection.  When this happens,
Scheme will abort to the nearest REPL and exclaim `out of memory!'.

Ideally, in this case, Scheme would instead expand the heap, rather
than aborting; and, independently, the OS could tell Scheme when
physical memory is scarce, in which case Scheme could try to shrink
the heap by running secondary GC daemons, aborting to the nearest
REPL, &c.  But no OS does this, and we don't have a scalable GC, so
this strategy doesn't work and would perform badly anyway.

_______________________________________________
MIT-Scheme-devel mailing list
MIT-Scheme-devel@gnu....
https://lists.gnu.org/mailman/listinfo/mit-scheme-devel

Bookmark with:

Delicious   Digg   reddit   Facebook   StumbleUpon

Related Messages

opensubscriber is not affiliated with the authors of this message nor responsible for its content.