x86 shared library performance

[Fergus Henderson]

Hi,

I did some benchmarking of Mercury's shared libraries on Linux.
Some of you may be interested in the results.

Cheers,
	Fergus.

----------------------

When compiling with `-O2', I got the following results:
(subject to timing inaccuracies

cqueens	shared time slowdown = 79.00%
crypt	shared time slowdown = 26.00%
deriv	shared time slowdown = 25.00%
nrev	shared time slowdown = 34.00%
poly	shared time slowdown = 32.00%
primes	shared time slowdown = 18.00%
qsort	shared time slowdown = 39.00%
queens	shared time slowdown = 51.00%
query	shared time slowdown = 122.00%
tak	shared time slowdown = 94.00%

However, when compiling with `-O6 --intermodule-optimization',
the slowdowns disappeared into the timing noise.

cqueens	shared time slowdown = -6.00%
crypt	shared time slowdown = 0%
deriv	shared time slowdown = 1.00%
nrev	shared time slowdown = 0%
poly	shared time slowdown = 0%
primes	shared time slowdown = -1.00%
qsort	shared time slowdown = 7.00%
queens	shared time slowdown = 5.00%
tak	shared time slowdown = 0%

The sizes for tak are
				-O2		-O6 --intermodule-optimization
	everything shared	5k		5k
	just libc/libm shared	47k		65k
	everything static	183k		183k

-- 
Fergus Henderson <fjh at cs.mu.oz.au>   |  "I have always known that the pursuit
WWW: <http://www.cs.mu.oz.au/~fjh>   |  of excellence is a lethal habit"
PGP: finger fjh at 128.250.37.3         |     -- the last words of T. S. Garp.