Part of Apple's work last week was shoving GCC 3.3 out the door. Now  
earlier last year I had
attempted to compile Squeak with GCC 3.1 with dismal results. However  
tonight I've discovered
that compiling Squeak with GCC 3.3 actually gives good results.
Using:  -O3 -mcpu=750 -funroll-loops -fno-cprop-registers  
-falign-loops=16

gcc (GCC) 3.3 20030304 (Apple Computer, Inc. build 1435)

Gives tinyBenchmarks like so:

'51405622 bytecodes/sec; 1860771 sends/sec'
'51282051 bytecodes/sec; 1873720 sends/sec'
'50673000 bytecodes/sec; 1863347 sends/sec'
'51488334 bytecodes/sec; 1839164 sends/sec'
'50996015 bytecodes/sec; 1876332 sends/sec'
'51405622 bytecodes/sec; 1855642 sends/sec'
'51446945 bytecodes/sec; 1875025 sends/sec'
'50996015 bytecodes/sec; 1854364 sends/sec'
'51446945 bytecodes/sec; 1873720 sends/sec'
'50274941 bytecodes/sec; 1867223 sends/sec'

With GCC 2.95.x (the gcc for the macintosh VM used until now gives:)
-O3 -mcpu=750 -funroll-loops

'51159072 bytecodes/sec; 1766756 sends/sec'
'51159072 bytecodes/sec; 1755239 sends/sec'
'50833995 bytecodes/sec; 1754095 sends/sec'
'50914876 bytecodes/sec; 1767916 sends/sec'
'51159072 bytecodes/sec; 1758678 sends/sec'
'50996015 bytecodes/sec; 1764441 sends/sec'
'51036682 bytecodes/sec; 1767916 sends/sec'
'50713153 bytecodes/sec; 1739364 sends/sec'
'50833995 bytecodes/sec; 1764441 sends/sec'
'51036682 bytecodes/sec; 1760979 sends/sec'

So as one can see the sends per second is a good 6% better!
This is a good thing. It is also reflected in the Mpeg plugin too.  
(better performance).

Thus a 3.5.2b1 VM is pending.

--
======================================================================== 
===
John M. McIntosh <johnmci at smalltalkconsulting.com> 1-800-477-2659
Corporate Smalltalk Consulting Ltd.  http://www.smalltalkconsulting.com
======================================================================== 
===