View Full Version : Re: The real reason Apple used gcc for Spec


Peter Ammon
07-08-2003, 05:31 AM
no wrote:
> http://www.theregister.co.uk/content/39/31600.html
> "As previously reported, coding for the G5 requires programmers adopt
> the GCC 3.3 compiler.

This is trivially false. If it were so, the G5 would be unable to run
software compiled using previous compilers, which is the vast majority
of Mac software.

> GCC 3.3 supports flags which allow developers to
> generate 970-only code, and to access 64-bit datapaths.


Your post doesn't make this clear, but the "bridge" is in the chip
itself and has nothing to do with the compiler.

> As we say, IBM describes the bridge as temporary: "These resources are
> not to be considered a permanent part of the PowerPC architecture," it
> sternly warns."

IBM's 64 bit POWER processors have been able to run 32 bit programs for
a long long time. See
<http://www-1.ibm.com/servers/aix/os/compatibility/> or
<http://publibn.boulder.ibm.com/doc_link/en_US/a_doc_lib/aixbman/prftungd/prftungd120.htm
>. It's disingenuous to imply otherwise.

>
> The G5 is a very admirable machine, do not get me wrong. So much for
> the claim that Apple could of gotten higher scores with another compiler
> that would of gotten higher scores.

Huh? Non sequiter. How did you get the impression that only code
compiled with gcc 3.3 could run on the G5? Do you think that all
software compiled with other compilers will break if you try to run it
on the G5?

> The performance that was done by a
> "independent third party" even stated that one of the tests was not done
> by them but done by Apple themselves!

Quote, please.

> And surprisingly enough this test
> showed scores lower for the p4 than what has already been gotten using
> the same settings.

No they didn't.

>
> Apple also used a custom malloc library for the spec test for the g5 but
> did not for the p4.

Yes. Nobody knows what impact that had.

>
> I am not saying apple cheated for the g5 score, they got the best
> results they could get.

That's almost certainly false. There are a number of ways of varying
nefarious-ness that Apple could have gotten even better scores.

> But it has been shown they didnt get the best
> results possible for the pc.

The goal is not to get the highest scores possible for each platform,
but to design a test that isolates the variables you are interested in
and controls for the rest.

>
> BTW, what are the quake scores for a DUAL p4 3.2 and a dual xeon (both
> with a 9800 pro video card)

-Peter

Josiah Fizer
07-08-2003, 05:33 AM
On Mon, 07 Jul 2003 21:31:17 -0700, Peter Ammon <pa44[at]cornell.edu>
wrote:

>no wrote:
>> http://www.theregister.co.uk/content/39/31600.html
>> "As previously reported, coding for the G5 requires programmers adopt
>> the GCC 3.3 compiler.
>
>This is trivially false. If it were so, the G5 would be unable to run
>software compiled using previous compilers, which is the vast majority
>of Mac software.
>

Do you understand the difference between coding for the G5 and writing
code that the G5 can run?

<snip>


----== Posted via Newsfeed.Com - Unlimited-Uncensored-Secure Usenet News==----
http://www.newsfeed.com The #1 Newsgroup Service in the World! >100,000 Newsgroups
---= 19 East/West-Coast Specialized Servers - Total Privacy via Encryption =---

Peter Ammon
07-08-2003, 06:06 AM
Josiah Fizer wrote:

> On Mon, 07 Jul 2003 21:31:17 -0700, Peter Ammon <pa44[at]cornell.edu>
> wrote:
>
>
>>no wrote:
>>
>>>http://www.theregister.co.uk/content/39/31600.html
>>>"As previously reported, coding for the G5 requires programmers adopt
>>>the GCC 3.3 compiler.
>>
>>This is trivially false. If it were so, the G5 would be unable to run
>>software compiled using previous compilers, which is the vast majority
>>of Mac software.
>>
>
>
> Do you understand the difference between coding for the G5 and writing
> code that the G5 can run?
>
> <snip>

I take it you're interpreting "coding for the G5" to mean "writing code
specifically optimized for the G5." So is it your contention (or
"no"'s, I suppose) that the people who wrote SPEC were "coding for the
G5?" That's obviously false as well.

-Peter

Beelsebob
07-08-2003, 12:00 PM
Josiah Fizer <jfizer[at]classy.com> wrote in message news:<rbikgv463hglb86bhb7tslfdfkp4np2iap[at]4ax.com>...
> On Mon, 07 Jul 2003 21:31:17 -0700, Peter Ammon <pa44[at]cornell.edu>
> wrote:
>
> >no wrote:
> >> http://www.theregister.co.uk/content/39/31600.html
> >> "As previously reported, coding for the G5 requires programmers adopt
> >> the GCC 3.3 compiler.
> >
> >This is trivially false. If it were so, the G5 would be unable to run
> >software compiled using previous compilers, which is the vast majority
> >of Mac software.
> >
>
> Do you understand the difference between coding for the G5 and writing
> code that the G5 can run?

Yes, I do, and setting the compiler flags to link against the 64bit
memory library falls into the second category - the code will still
run happily on 32 bit machines. The reason for this is that the
library in question is designed to allow 32 bit code to access the
memory range of a 42 bit proccessor (not 64 bit, because that made the
library too complicated). Note that this library has no effect
whatsoever on performance, it simply allows the code to access more
memory.

Bob