[MLton-devel] card-size self-compile experiments

Stephen Weeks MLton@mlton.org
Wed, 14 Aug 2002 17:46:41 -0700


Here are the results of doing self compiles with versions of the
MLton using different card sizes.  There are execution times, as well
as statistics obtained while scanning for intergenerational pointers

marked cards
	the number of marked cards seen
bytes scanned
	the number of bytes of objects scanned for pointers
bytes skipped 
	the number of bytes of objects skipped when starting from
	an object that starts on a card boundary and skipping to 
	a marked card
	
			10^6	10^6 	10^6
card	mutator	GC 	marked	bytes	bytes
size	time(s)	time(s)	cards	scanned	skipped
-----	-------	-------	-------	------- -------
  16	191.9	 90.3    29.9	  815	  457
  32	190.7	 80.2	 23.6	1,084	  571
  64	189.8	 75.9	 17.5	1,388	  632
 128	189.3	 74.0	 11.6	1,734	  692
 256	188.8	 73.9	  7.3	2,101	  772
 512	188.6	 74.9	  4.4	2,503	  990
1024	188.5	 76.2	  2.6	2,927	1,047

These numbers show that 256 seems like a reasonable choice for card
size.

For a point of reference, here are some stats that apply to all the
self compiles

21,000 * 10^6 bytes allocated
 2,000 * 10^6 bytes copied (minor)
 1,000 * 10^6 bytes copied (major)



-------------------------------------------------------
This sf.net email is sponsored by: Dice - The leading online job board
for high-tech professionals. Search and apply for tech jobs today!
http://seeker.dice.com/seeker.epl?rel_code=31
_______________________________________________
MLton-devel mailing list
MLton-devel@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/mlton-devel