Hello, Mark.

On Thu, 11 May 2000, you wrote:
> Is this the type of kludges Linux should have in the kernel because a few people
> want to run
> configurations that Intel does not even support?

This is a reasonable criticism.  However, note that Intel does not support
Celerons in an SMP environment at all... lots of people use them.  I guess my
main reason for wanting this patch was that Win2k works just fine (for
Microsoft's definition of 'just fine') in this configuration.

>   In looking at the patch, every
> call to
> do_gettimeoffset() will now go through a procvar, instead of being called directly.
>  It's not
> a significiant impact, but it is an impact none-the-less.  ANY impact to support
> this type

As I pointed out before, I'm really out of my league when doing kernel patches.
 I would really like some feedback here.  Can you think of a way of doing this
more efficiently?  Because if I could get the efficiency up, I think this would
answer your concerns.  Also, in the case that the unpatched kernel decides not
to use TSC (i.e. TSC is not supported by the CPU), is it not true that all
calls already go through a procvar?

> of configuration is too much (in my opinion). Some things belong off-stream.  If
> someone
> is running this configuration, they should patch their own kernel to not impact the
> rest of us.

Once again, let me summarise:  Win2k allows this configuration to work.  This
is why I created the patch.  However, your concerns are valid.  Looking at the
patch, can you think of any way to make it as efficient as an unpatched kernel?

I am quite happy to stick my patch somewhere for 'officially unofficial'
patches and let people grab it if they need it.  My major concern with this is
that I would then be responsible for keeping the patch up-to-date with all new
kernels and I'm not sure I'd be capable of doing this.

> ....wait a sec, as I put on flame-resistant clothing  ;-)

<g>  I am very interested to know what other people think.  I'm also very very
interested to know how much inefficiency we get by going through a procvar. 
I'm not interested in flame-wars, though.

-- 
Christopher Thompson  http://hypocrite.org/
hypocrite.org will be down for several hours
on Saturday.  Email to me may be delayed
during that time.
-
Linux SMP list: FIRST see FAQ at http://www.irisa.fr/prive/dmentre/smp-howto/
To Unsubscribe: send "unsubscribe linux-smp" to [EMAIL PROTECTED]

Reply via email to