|
|
|
|
| Re: UAD-1 Multiband to be released tonight or tomorrow [message #61807 is a reply to message #61796] |
Fri, 23 December 2005 09:39   |
Tony Benson
Messages: 453
Registered: June 2006
|
Senior Member |
|
|
g, "Well it sure is pretty but
it
>> sure is slower too." At the time it was attributed to the new GUI but
now
>> that there's more evidence out there maybe that should be re-thought.
The
>> user space and kernel space on *nix machines is usually quite distinct,
so
>> do you think maybe Apple made some kernel changes in Intel OS X? I don't
>> have access to the software, so I can't say for sure if they've actually
>> gone monolithic but from the benchmarks that would be a *highly* logical
>> guess. If they haven't gone fully monolithic I would expect that they
lifted
>> some more code from a BSD licensed *nix to vastly improve the current
kernel
>> space. I'd still be on monolithic but it almost has to be one or the other.
>>
>>
>> Or maybe I'm just peddling urban myths . . .
>
>Maybe. :^) If the kernel has changed for Intel, it's likely it will
>change for PPC. We'll see.
Why is that likely? Why would Apple tell all of their PPC users that they've
been sucking up CPU cycles all of these years by using faulty kernel technology?
Why would they spend time on the kernel space of an operating system they've
already decided to drop in the end? Why not just talk about how "optimized"
the new X86 version of OS X is and make everyone buy new hardware?
TCB
>Cheers,
> -Jamie
> http://www.JamieKrutz.com
>
>
>> TCB
>>
>>
>>>OTOH, the Opteron looked pretty good, too. It will be very interesting
>>>to look at the upcoming Intel dual Yonah processor with OSX, in
>>>(rumored) another couple of weeks, and see how that combination does.
>>>This is all a moving target.
>>>
>>>In the other link, the main bias is the same as yours: OSX is not open
>>>source from top to bottom. True. Beyond that complaint, with his
>>>specific statistical software he finds better performance under Linux
>>>when testing on a couple of older G5 boxes. Not sure if this says much
>>>abou
|
|
|
|
| Re: UAD-1 Multiband to be released tonight or tomorrow [message #61812 is a reply to message #61807] |
Fri, 23 December 2005 10:39 |
erlilo
Messages: 405
Registered: June 2005
|
Senior Member |
|
|
erformance computing (threading,
kernel
> >> access, etc.). I'm told that I'm passing around "urban myths" and
demands
> >> for proof are made. After the proof is given I'm told that they just
reflecty
> >> my "bias" as someone who likes Debian.
> >
> >Not "just." You clearly explained your reasons for disliking commercial
>
> >operating systems and your reasons for liking open source. I have no
> >problem with that. It's a bias, but a supportable one.
>
> Well, my bias doesn't make OS X perform 1/10th as well as Yellow Dog using
> the same hardware when crunching MySQL queries that spawn more than 40
threads
> or so. OS X does that all on its own.
>
> >
> >> OK, that's one way to explain things, and a way that I would guess is
> very
> >> appealing to someone in for a few grand worth of G5 hardware that is
leaning
> >> into the headwind of kernel design that looked great on the chalkboard
> in
> >> Comp Sci 310: Mach Kernel Development but Linus saw and said, "Wow,
what
> >> a great way to destroy my database queries!"
> >
> >Linus is brilliant, of course.
> >
> >
> > > However, let's look at another
> >> explanation. Apple is plannning on moving to Intel hardware. The
developer
> >> versions for Intel hardware seem distressingly fast in comparison to
supposedly
> >> superior chip design from IBM. Everyone who owned an early G4 with OS
> 9 on
> >> it remembers installing OS X and thinking, "Well it sure is pretty but
> it
> >> sure is slower too." At the time it was attributed to the new GUI but
> now
> >> that there's more evidence out there maybe that should be re-thought.
> The
> >> user space and kernel space on *nix machines is usually quite distinct,
> so
> >> do you think maybe Apple made some kernel changes in Intel OS X? I
don't
> >> have access to the software, so I can't say for sure if they've
actually
> >> gone monolithic but from the benchmarks that would be a *highly*
logical
> >> guess. If they haven't gone fully monolithic I would expect that they
> lifted
> >> some more code from a BSD licensed *nix to vastly improve the current
> kernel
> >> space. I'd still be on monolithic but it almost has to be one or the
other.
> >>
> >>
> >> Or maybe I'm just peddling urban myths . . .
> >
> >Maybe. :^) If the kernel has changed for Intel, it's likely it will
> >change for PPC. We'll see.
>
> Why is that li
|
|
|
|
The PARIS Forums
Search
Help
Members
Register
Login
Home
