[MLton-devel] self compile stats with the new compiler and runtime

Stephen Weeks MLton@mlton.org
Thu, 11 Jul 2002 15:13:48 -0700


Here is some data from a slew of self compiles that I did today.  All
of the numbers were gathered on my 1.6GHz 4G machine.

First, I ran self compiles for a (src) compiler compiled with
-inline-array {false,true} compiling a (dst) compiler using the flag
-inline-array {false,true}.  Here are the total compile times for the
four self-compiles.

	-inline-array
time(s)	src	dst
-------	-----	-----
224.7	false	false
223.4	true	false
229.1	false	true
228.2	true	true

As to code sizes, here are the text + data sizes for each of the two
compiler variants.

code size	-inline-array
---------	-------------
7,211,377	false
7,337,121	true

Not bad, considering that the 2002-04-26 is 7,622,437 bytes.

Finally, to compare mark-compact to copying collection, I did self
compile using only mark-compact collection, and compared that to a
normal self compile (which only uses copying on my 4G machine).

time	gc time	%GC	#GCs
-----	-------	---	----
224.7	 36.3	17%	57
326.3	126.7	41%	47

So, the mark compact is quite a bit slower.  But of course, its real
reason for existence is to avoid paging.  If someone has a 256M
machine that they could run a self compile on, it would be nice to
hear how that does.


-------------------------------------------------------
This sf.net email is sponsored by:ThinkGeek
PC Mods, Computing goodies, cases & more
http://thinkgeek.com/sf
_______________________________________________
MLton-devel mailing list
MLton-devel@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/mlton-devel