[plt-scheme] Updated Benchmarks

From: Brent Fulgham (bfulg at pacbell.net)
Date: Sun Jan 21 15:13:48 EST 2007

-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1

I have re-run the shootout benchmarks using the 369.5 version of  
MzScheme, and found the following rankings (compared to straight C).   
The first column in each group is CPU time, the second is memory  
used.  Negative values mean worse performance compared to C:


Benchmark				369.3									369.5
======================================================================== 
===============
chameneos				-23		-57								-13.4		-84
cheap-concurrency		-10.2	-7.3								-5.9			-8.7
fannkuch				-28		-1.5								-24			+1.8
fasta					-18		-10.2							-13.8		-10.2
mandelbrot				-140	-11.6							-80			-11.6	
n-body					-145	-11.3							-74			-11.7	
nsieve					-2.0		-1.4								-2.0			+1.8
nsieve-bits				-5.4		-4.8								-5.5			-5.2
partial-sums				-8.9		-13.3							-6.1			-11.8
pidigits					-24		-8.6								-6.2			-14.0
recursive				-21		-10.8							-18			-11.0
reverse-complement		-8.6		-3.9								-9.5			-7.1	
startup					-27	 	1.4								-27	
sum-file					-12.8	-12.6							-15			-13.2


So, it seems that in almost all cases the performance has  
significantly improved, at the cost of some memory use in some  
cases.  The only regression I see is in 'reverse-compliment' and 'sum- 
file', both of which are heavily string and file-I/O dependent.

Thanks,

- -Brent	
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.4.6 (Darwin)

iD8DBQFFs8mAzGDdrzfvUpURAqRRAJ9SLevu9Kftu3IHFifnY5SSslHnLgCfTc8y
V0Wt0Hy3p4YHfrnFpLgnbE8=
=Tip/
-----END PGP SIGNATURE-----


Posted on the users mailing list.