[MLton-devel] self-compile stats with generational gc

Stephen Weeks MLton@mlton.org
Fri, 16 Aug 2002 17:38:22 -0700


Here are some self-compiles on my three different machines, using my
current working version with auto switching between mark-compact and
cheney copying, as well as automatically turning on/off the
generational gc.

		elapsed	mutator		
RAM  processor	time 	time	gc time	
---- --------	-------	-------	-------	
4G   1.6GHZ	 243.7	190.5	 52.4	
512M 1.6GHz	 313.3	201.5	 62.2	
256M 1.1GHz	2192.7	551.7	482.1

			10^6		10^6		10^6
			bytes	# mark	bytes		bytes
RAM  processor	# copy	copied	compact	m/c'ed	# minor	copied
---- ---------	-------	-------	-------	-------	------- ------
4G   1.6GHZ	50	2,389	 3	  260	  0	    0
512M 1.6GHz	27	1,628	 0	    0	160	1,347
256M 1.1GHz	22	1,150	26	2,550	584	2,601

I am pleased that there isn't much of a slowdown moving from a 4G to a
512M machine.   Going to the 256M machine really hurts, but at least
the self-compile is doable there (which is a new achievement) and
paging only costs a factor of 2.

The auto switching seems reasonable, using almost all copying on the
big machines and lots of mark-compact on the small machine.  The 3
mark-compacts on the 4G machine are due to it being too greedy, making
a semispace of size 1G and then not being able to mmap another
semispace due to fragmentation.


-------------------------------------------------------
This sf.net email is sponsored by: OSDN - Tired of that same old
cell phone?  Get a new here for FREE!
https://www.inphonic.com/r.asp?r=sourceforge1&refcode1=vs3390
_______________________________________________
MLton-devel mailing list
MLton-devel@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/mlton-devel