>> i'm not a friend of x86 ;) but sometimes something has to be done... 
>> so i went to the URL posted above and set a few CFLAGS in mk.conf as 
>> i did for amiga, sparc, sparc64, macppc, etc etc according to gcc's 
>> manual and the hardware i wanted NetBSD to run on.
>>
>> it runs seemless, stable and a-okay :) (as well as the other 
>> architectures do)
>>
>> intel built the P2 as a kind of Pentium Pro (P6 architecture), but 
>> wanted it to be more successful in the market. in the meantime, the 
>> P55C (Pentium MMX) was introduced, so they took a PPro Core, stripped 
>> the expensive full-speed L2 cache (as well as that phat ceramic 
>> package with two dies ;), replaced by a not-so-well solution (more 
>> waitstates, half clockspeed IIRC, etc.) but added MMX (like the P55C 
>> had already).
>>
>> i wonder why gcc does diffenciate between (flags) 'pentiumpro' and 
>> 'pentium2', but does not (altho this would be logically correct, 
>> because there is that clear difference between P2 and PPro, with and 
>> w/o MMX) imply the -mmmx flag for pentium2 automatically...
>
> 	For gcc 3.3.3 (shipping with NetBSD 2.0) pentium2 implies
> 	mmx. From gnu/dist/gcc/gcc/config/i386/i386.c:
> 	      {"pentiumpro", PROCESSOR_PENTIUMPRO, 0},
> 	      {"pentium2", PROCESSOR_PENTIUMPRO, PTA_MMX},
>
> 	You may want to look at pkgsrc/devel/cpuflags to 'automatically'
> 	set compiler flags. It seems to indicate that pentium2 was
> 	introduced in gcc 3.1

cpuflags is mandatory and 'heavily' used by me. ;)

but sometimes one should not rely only on software (i.e. cpuflags in 
this case), although it is useful.

your quote from gnu/dist/gcc/gcc/config/i386/i386.c convinces me; 
nevertheless, this should be clear w/o consulting this file, reading 
the documentation should really clarify this. well, this is not up to 
the NetBSD folk...

-- 
:x!

timo