Re: [PATCH 1/6] new timeofday core subsystem for -mm (v.B3)

[Ulrich Windl]

On 18 Jun 2005 at 14:02, Roman Zippel wrote:

> Hi,
> 
> On Fri, 17 Jun 2005, john stultz wrote:
> 
> > o Uses nanoseconds as the kernel's base time unit
> 
> Maybe I missed it, but was there ever a conclusive discussion about the 
> perfomance impact this has?
> I see lots of new u64 variables. I'm especially interested how this code 
> scales down to small and slow machines, where such a precision is absolute 
> overkill. How do these patches change current and possibly common time 
> operations?

Hi all!

I had the impression that for slow and small machines every recent Linux 
distribution is overkill. Whenever I complained every relied "Harddisks are cheap, 
memory is cheap, get a new CPU". I do understand your doubts however. For my 
personal experience with my PPSkit patches, I found out that my ols 386/SX @16MHz 
failed to receive all serial characters when I timestamped each of them using my 
new clock routines. However a 486@33MHz would do (it had better serial UART chips, 
too). I would not thing my code was terribly efficient, because I tried to make it 
"right" first.

And even the 386 had limited support for 64 bit operations. Since the 486 a 32bit 
add is specified as using 1 CPU cycle (most likely in the optimal case). So doing 
one more would not harm that much.

Basically, either the new clock system has to be optional (a maintenance nightmare 
most likely), or you'll have to require a specific amount of performance for the 
latest software. If you cannot fulfill the requirements, you'll have to stick with 
an older release of the software.

Maybe let's try to make it as good (correct and efficient (and understandable) as 
good as we can.

Regards,
Ulrich

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