Powerbook

Powerbook

[Kári Davíðsson]

Hi,

This is way off topic but I have been unable to find this on the web so
I ask the people who are most likely to have done this kind of
"benchmark".

I am on the search for a notbook that I will use partly to do coding
under Linux.

How long is a kernel recompile on a Powerbook G4 (550Mhz/667Mhz)?

Thanks and sorry for the noise...

K.D.

** Sent via the linuxppc-dev mail list. See http://lists.linuxppc.org/

Re: Powerbook

[Joern Nettingsmeier]

Kári Davíðsson wrote:
>
> Hi,
>
> This is way off topic but I have been unable to find this on the web so
> I ask the people who are most likely to have done this kind of
> "benchmark".
>
> I am on the search for a notbook that I will use partly to do coding
> under Linux.
>
> How long is a kernel recompile on a Powerbook G4 (550Mhz/667Mhz)?
>
> Thanks and sorry for the noise...

<asbestos mode="full body protection">
if you're in for quick compilation, get an intel machine.
</asbestos>

jörn (runs)



--
Watch out where the huskies go and don't you eat
the yellow snow !
	- Frank Zappa

** Sent via the linuxppc-dev mail list. See http://lists.linuxppc.org/

Re: Powerbook

[Anton Blanchard]

> <asbestos mode="full body protection">
> if you're in for quick compilation, get an intel machine.
> </asbestos>

Of course a PowerPC machine holds the current kernel compile record :)

Anton

** Sent via the linuxppc-dev mail list. See http://lists.linuxppc.org/

Re: Powerbook

[Paul Mackerras]

Joern Nettingsmeier writes:

> <asbestos mode="full body protection">
> if you're in for quick compilation, get an intel machine.
> </asbestos>

No, if you're in for quick compilation, compile a kernel for i386
rather than a kernel for ppc, for some reason gcc is quicker at
producing x86 output than ppc output, i.e. the x86 compilation
involves less computation.

About 18 months ago I did a comparison of a (then) recent dual G4
powermac with a (then) recent dual pentium II system.  I timed all
four combinations of ppc or x86 machine and ppc or x86 output.  The
ppc machine was consistently faster than the x86 machine for the same
computation, although the ppc machine producing ppc output was slower
than the x86 machine producing x86 output.

Paul.

** Sent via the linuxppc-dev mail list. See http://lists.linuxppc.org/

Re: Powerbook

[David Edelsohn]

>>>>> Paul Mackerras writes:

Paul> No, if you're in for quick compilation, compile a kernel for i386
Paul> rather than a kernel for ppc, for some reason gcc is quicker at
Paul> producing x86 output than ppc output, i.e. the x86 compilation
Paul> involves less computation.

	PowerPC GCC is multi-arched by default for both 32-bit and 64-bit,
so many values internally are computed as 64-bit "long long" quantities.
i386 does not default to x86_64, so it uses word-sized ints.

	PowerPC GCC has many more registers to allocate and more
restricted operands for instructions.

	GCC on RISC (e.g., PowerPC) runs the scheduler at two different
phases of the compilation.  GCC on CISC (e.g., i386) runs the scheduler
only once.

	Those are some of the differences I can think of off the top of my
head.

David

** Sent via the linuxppc-dev mail list. See http://lists.linuxppc.org/

RE: Powerbook

[Kári Davíðsson]

> -----Original Message-----
> From: Paul Mackerras [mailto:paulus@samba.org]
> Sent: 3. maí 2002 22:56
> To: Joern Nettingsmeier
> Cc: Kári Davíðsson; linuxppc-dev@lists.linuxppc.org
> Subject: Re: Powerbook
>
>
> Joern Nettingsmeier writes:
>
> > <asbestos mode="full body protection">
> > if you're in for quick compilation, get an intel machine.
> > </asbestos>
>
> No, if you're in for quick compilation, compile a kernel for i386
> rather than a kernel for ppc, for some reason gcc is quicker at
> producing x86 output than ppc output, i.e. the x86 compilation
> involves less computation.

It has something to do with RISC versus CISC architectures, i.e. RISC
offloads much of scheduling (?) to the compiler....
My old trusted Alpha had the same "problem" when I was using it for
development.
It is also very visible if you use cross compiler on an PIII to generate
ppc code (like I am doing now). The same code compiles significantly
slower
if cross compiled rather than natively compiled for the PIII.

K.D.

** Sent via the linuxppc-dev mail list. See http://lists.linuxppc.org/