[Issue 5623] Slow GC with large heaps

[d-bugmail at puremagic.com]

http://d.puremagic.com/issues/show_bug.cgi?id=5623



--- Comment #4 from David Simcha <dsimcha at yahoo.com> 2011-02-20 14:48:45 PST ---
(In reply to comment #3)
> A very simple benchmark, in D and Java. You may use the D version with and
> without your patch, and then the Java version too, and show the three timings,
> with N = 15 or 18 or more:
> 
> http://codepad.org/yMGK34cb
> http://codepad.org/DIOeYn6p

I didn't both(In reply to comment #3)
> A very simple benchmark, in D and Java. You may use the D version with and
> without your patch, and then the Java version too, and show the three timings,
> with N = 15 or 18 or more:
> 
> http://codepad.org/yMGK34cb
> http://codepad.org/DIOeYn6p

Using n = 18:

D (with patch):  62 seconds
D (without patch):  68 seconds
Java:  6 seconds

I'm not sure what this told us that we didn't already know, though.  We already
knew Java's GC is much better than D's.  My patch is meant to address issues
with large object allocation, not small object.  (Large is defined as at least
one full page.  If no large objects are allocated for the duration of the
program, then my patch has virtually no effect.)  This benchmark does tons of
small object allocation and no large object allocation.  The small difference
between with and without patch may even just be noise.

-- 
Configure issuemail: http://d.puremagic.com/issues/userprefs.cgi?tab=email
------- You are receiving this mail because: -------