Re: 2.4.19pre3aa2

[Andrea Arcangeli <andrea@suse.de>]

On Thu, Mar 14, 2002 at 05:16:20PM +0100, Dave Jones wrote:
> On Thu, Mar 14, 2002 at 05:12:59PM +0100, Andrea Arcangeli wrote:
>  > Correct. I think the CONFIG option isn't worthwhile in the first place
>  > and this is why I only left the CONFIG_M686 knowing most smp kernels are
>  > compiled that way.  4096bytes of virtual vmallc space and some houndred
>  > bytes of bytecode doesn't worth the config option. If something the
>  > CONFIG_F00F would be more a documentation effort 8).
>  
>  nononono! CONFIG_FOOF is a derived symbol from whatever CONFIG_Mx8x
>  is set. Much in the way we derive CONFIG_X86_WP_WORKS_OK, CONFIG_X86_PPRO_FENCE
>  and freinds.. 

Of course I perfectly understood what you meant. I designed the current
way of doing the CONFIG_X86 some years back so I know what you mean.

Just re-read the previous email with that in mind. What I meant is that
4096bytes of virtual vmalloc space doesn't worth a CONFIG_F00F IMHO.
While CONFIG_F00F isn't wasted user time because the user won't see it,
it still clobbers the source code, but nevertheless it is better than
the current halfway broken #ifdef CONFIG_M686 and that's why I said if
somebody bothers to verymicrooptimize I'm ok, I just won't microoptimize
myself and I'd drop CONFIG_M686 instead.

Andrea