Re: SMP Extraversion

Linus Torvalds (torvalds@transmeta.com)
Sat, 16 Oct 1999 09:51:12 -0700 (PDT)


On Sat, 16 Oct 1999, Alan Cox wrote:
>
> The big problem one is the 1G/2Gig split, which causes same structures
> wrong offset problems where modules do load. Once you include that you
> end up needing
>
> 1G-uni 2G-uni 1G-smp 2G-smp

Note that I'm probably going to remove the 1G split.

The 3GB user-space is better for users, and with BIGMEM support the
advantages of 2G kernels are much less anyway. I don't see any horribly
compelling reasons for showing that particular difference to users any
more: the 2G split was a hack to avoid doing bigmem, but now...

This is especially true now that I have fairly clean patches from Ingo to
take the BIGMEM stuff up to 64GB - the difference of whether you can
directly map 1GB or 2GB of that physical memory is negligible, as 95% of
the memory will be high anyway on the big boxes.

I kind of like the notion of having SMP and UP modules co-exist nicely,
but I do not enjoy the thought of then adding other random differences
too.. SMP vs UP is kind of fundamental, while 1G vs 2G definitely isn't.

Linus

-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.rutgers.edu
Please read the FAQ at http://www.tux.org/lkml/