Difference between "-march=k6-3" and "-march=k6-2"?

[Jan Hubicka]

> In GCC v3.2.x, what is the difference in code generation between the 
> "-march=k6-3" and "-march=k6-2" options?
> 
> My understanding is that the only difference between these CPUs is that 
> the K6-3 has an L2 cache, where the K6-2 does not.  Is the difference 
> between these -march option just exploiting that cache, or is there more 
> to the code generation than that?
It is there just for completeness.  K6-1 differs from K6-2 in the
support for 3dNOW and it seemed to be confusing to choose just the
submodels that are different from each other and let user to guess which
one is equivalent to modes he has on the table.

Honza
> 
> Thanks.