JK> CB> BUT, most 486 optimizors still tend to do the old style of 486
JK> CB> performance.  And the pentium type optimizations are rarely anti-486
JK>What do you mean for old/new style of 486???
The newer 486 cpus have different cycle timings than the older ones.
As a quick example, my newer Cyrix 486 can do a 16 bit mul in 3 cycles
and a 32 bit mul in 7 cycles.  But, the classic Intel 486 requires 13
cycles minimum and can go up to 42 cycles.  (There are of course other
examples of both faster and slower instructions.) That means sections of
code that take 'x' cycles on mine, will take 'y' cycles on somebody
elses. The difference may be better or worse.
There are enough cases from one brand of 486 to another, and old vs.
newer 486 that the 486s can run differently than expected.  The
difference isn't as much as there is between a 386 vs. any 486, but it
can still be quite a bit of difference.
Also, many 486 optimizers only do limited code reorganization,
RISCification of the instructions, etc.  Pentium class optimizers tend
to be much more aggressive about the order of instructions because they
have to be.  Most of those same type optimizations will also be
effective on the 486 too (provided you can be sure it's not actually
generating any of the few Pentium specific instructions.)
--- QScan/PCB v1.19b / 01-0162
---------------