> There's a point there, but on the other hand it doesn't seem sensible to 
> cripple ourselves as more and more people move to newer platforms. I 
> guess the answer would involve balancing (a) does it cause tremendous 
> gains on a pentium, (b) does it make a crippling difference a 386, and
> (c) how many people are still using 386s anyway compared to 486, p5? I 
> suspect the answers to (a) and (b) are no; I have no idea about c.
> 
> nd

Hmm, I always understood '-m486' was a performance increase on a 486, but
did not change the performance on a 386.  Basically that the code generator
chose instructions that were faster on the 486, but equivalent on a 386.
The only con was a slight increase in the size of the generated code.
(this is from memory, but I'm fairly sure it is correct)

-Andrew