[MLton-devel] Porting from Moscow ML

Stephen Weeks MLton@mlton.org
Tue, 24 Sep 2002 23:08:56 -0700

> I've now finished porting my prover, and, as promised, I've extracted
> a benchmark for MLton:

Great.  Thanks.  I see similar results on my 1.6Ghz machine

real	17m14.164s
user	16m52.420s
sys	0m12.670s

The gc-summary stats are also interesting.  They show an allocation
rate of over 150M/s.

GC type		time ms	 number		  bytes	      bytes/sec
-------------	-------	-------	---------------	---------------
copying		218,240	 21,088	 22,791,002,548	    104,430,911	
mark-compact	      0	      0	              0	              0	
minor		      0	      0	              0	              0	
total GC time: 219,150 ms (21.4%)
max pause: 280 ms
total allocated: 157,193,968,992 bytes
max live: 24,054,568 bytes
max semispace: 190,545,920 bytes
max stack size: 41,984 bytes
marked cards: 0
minor scanned: 0 bytes
minor skipped: 0 bytes

In any case, the following would help us to add it to our benchmark

* make it use only the standard basis library -- avoid, e.g.,
  MLton.eq. If we don't do this, we can't compare with other
* make it run for something closer to a minute or less.  Actually
  probably less is important, since I assume the other compilers will
  be slower.
* make sure all the computation is encapsulated in a single toplevel
  function, main().  That makes it easy to build with differnt

This sf.net email is sponsored by:ThinkGeek
Welcome to geek heaven.
MLton-devel mailing list