[MLton-devel] cvs commit: mark compact GC and backend/codegen changes

[Stephen Weeks]

> What's the right way to test the mark compact GC?  I'm currently running
> regressions with "-native true -gc-check first" (I tried "-gc-check
> every", but it's way to slow on some of them), with a runtime recompiled
> with gc.c's VERIFY_MARK=TRUE.  It's looking good, although I'm skipping
> all cont, int-inf, thread, signals, and world regressions for now.

Great!  VERIFY_MARK is exactly the right thing to use.  With that on,
every GC does a mark compact and then a copying gc, verifying that the
amount live is the same for both GCS.

I am about to do a checkin that integrates the mark-compact gc,
automatically switches between mark-compact and cheney, and resizes
the heap for both.  In the checkin, I've gotten rid of VERIFY_MARK,
and the auto-switcher only uses mark-compact when the heap gets large,
but there is an obvious "if" statement to change that will cause
mark-compact to be used on every GC.  I may even add a runtime
option to do so.  I'll do a commit email when it's in.


-------------------------------------------------------
This sf.net email is sponsored by:ThinkGeek
Two, two, TWO treats in one.
http://thinkgeek.com/sf
_______________________________________________
MLton-devel mailing list
MLton-devel@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/mlton-devel