3.4 bootstrap failure on powerpc-apple-darwin7.2.0

3.4 bootstrap failure on powerpc-apple-darwin7.2.0

[Timothy J. Wood]

../gcc/configure --prefix="$PREFIX" --enable-altivec  
--enable-languages="c,c++,objc"
make

/Users/bungi/Source/GNU/gcc/gcc-3.4/build/gcc/xgcc  
-B/Users/bungi/Source/GNU/gcc/gcc-3.4/build/gcc/  
-B/Users/bungi/Source/GNU/gcc/gcc-3.4/install/powerpc-apple- 
darwin7.2.0/bin/  
-B/Users/bungi/Source/GNU/gcc/gcc-3.4/install/powerpc-apple- 
darwin7.2.0/lib/ -isystem  
/Users/bungi/Source/GNU/gcc/gcc-3.4/install/powerpc-apple-darwin7.2.0/ 
include -isystem  
/Users/bungi/Source/GNU/gcc/gcc-3.4/install/powerpc-apple-darwin7.2.0/ 
sys-include -O2  -DIN_GCC    -W -Wall -Wwrite-strings  
-Wstrict-prototypes -Wmissing-prototypes -Wold-style-definition   
-isystem ./include  -mlong-double-128 -g -DHAVE_GTHR_DEFAULT  
-DIN_LIBGCC2 -D__GCC_FLOAT_NOT_NEEDED  -I. -I -I../../gcc/gcc  
-I../../gcc/gcc/ -I../../gcc/gcc/../include -I../intl -fexceptions -c  
../../gcc/gcc/unwind-dw2.c -o libgcc/./unwind-dw2.o
/var/tmp//ccec9Kuq.s:4837:stvx vector instruction is optional for the  
PowerPC (not allowed without -force_cpusubtype_ALL option)
/var/tmp//ccec9Kuq.s:4840:stvx vector instruction is optional for the  
PowerPC (not allowed without -force_cpusubtype_ALL option)
/var/tmp//ccec9Kuq.s:4843:stvx vector instruction is optional for the  
PowerPC (not allowed without -force_cpusubtype_ALL option)
/var/tmp//ccec9Kuq.s:4846:stvx vector instruction is optional for the  
PowerPC (not allowed without -force_cpusubtype_ALL option)
...

   Am I supposed to do something to get '-force_cpusubtype_ALL' (or  
maybe -faltivec) passed to xgcc when bootstrapping, or is this an  
actual bug?

-tim

Re: 3.4 bootstrap failure on powerpc-apple-darwin7.2.0

[Andrew Pinski]

On Jan 17, 2004, at 20:39, Timothy J. Wood wrote:

>
> ../gcc/configure --prefix="$PREFIX" --enable-altivec 
> --enable-languages="c,c++,objc"


--enable-altivec is really only for ppc linux to enable the altivec ABI 
by default which is
done for PPC Darwin by default already.

If you remove that you should be able to bootstrap just fine, note that 
the syntax for
altivec differs for FSF than what the original version of MOT's PEM.

Thanks,
Andrew Pinski

Re: 3.4 bootstrap failure on powerpc-apple-darwin7.2.0

[Timothy J. Wood]

On Jan 18, 2004, at 8:07 AM, Andrew Pinski wrote:

>
> On Jan 17, 2004, at 20:39, Timothy J. Wood wrote:
>
>>
>> ../gcc/configure --prefix="$PREFIX" --enable-altivec 
>> --enable-languages="c,c++,objc"
>
>
> --enable-altivec is really only for ppc linux to enable the altivec 
> ABI by default which is
> done for PPC Darwin by default already.

   Thanks; could someone update the documentation in install.texi to say 
that this should not be passed for PPC Darwin?

> If you remove that you should be able to bootstrap just fine, note 
> that the syntax for
> altivec differs for FSF than what the original version of MOT's PEM.

   I discovered this last night too, but then Altivec doesn't seem work. 
  I thought it was because I didn't have --enable-altivec, but I guess 
either something is busted or I'm doing something (else) wrong.  I'll 
send out a new mail on that to change the subject line :)

-tim

Re: 3.4 bootstrap failure on powerpc-apple-darwin7.2.0

[Ziemowit Laski]

On 18 Jan 2004, at 8.07, Andrew Pinski wrote:

>
> On Jan 17, 2004, at 20:39, Timothy J. Wood wrote:
>
>>
>> ../gcc/configure --prefix="$PREFIX" --enable-altivec 
>> --enable-languages="c,c++,objc"
>
>
> --enable-altivec is really only for ppc linux to enable the altivec 
> ABI by default which is
> done for PPC Darwin by default already.

Actually, in the course of my recent AltiVec struggles, I failed to 
discover any purpose whatsoever for the '--enable-altivec' config 
switch -- perhaps Aldy can enlighten us.

--Zem

Re: 3.4 bootstrap failure on powerpc-apple-darwin7.2.0

[Andrew Pinski]

On Jan 18, 2004, at 15:10, Ziemowit Laski wrote:
>> --enable-altivec is really only for ppc linux to enable the altivec 
>> ABI by default which is
>> done for PPC Darwin by default already.

> Actually, in the course of my recent AltiVec struggles, I failed to 
> discover any purpose whatsoever for the '--enable-altivec' config 
> switch -- perhaps Aldy can enlighten us.


Doing a simple search using google, I found when this was added:
http://gcc.gnu.org/ml/gcc-patches/2001-12/msg02123.html

It is exactly does what I said it does is to change the ABI by default.

Thanks,
Andrew Pinski

Re: 3.4 bootstrap failure on powerpc-apple-darwin7.2.0

[Ziemowit Laski]

On 18 Jan 2004, at 15.18, Andrew Pinski wrote:

> On Jan 18, 2004, at 15:10, Ziemowit Laski wrote:
>>> --enable-altivec is really only for ppc linux to enable the altivec 
>>> ABI by default which is
>>> done for PPC Darwin by default already.
>
>> Actually, in the course of my recent AltiVec struggles, I failed to 
>> discover any purpose whatsoever for the '--enable-altivec' config 
>> switch -- perhaps Aldy can enlighten us.
>
>
> Doing a simple search using google, I found when this was added:
> http://gcc.gnu.org/ml/gcc-patches/2001-12/msg02123.html
>
> It is exactly does what I said it does is to change the ABI by default.

Indeed, except I fail to see why this can't be done at run-time, when 
the user specifies '-maltivec'.

--Zem

Re: 3.4 bootstrap failure on powerpc-apple-darwin7.2.0

[Andrew Pinski]

On Jan 18, 2004, at 15:35, Ziemowit Laski wrote:
> Indeed, except I fail to see why this can't be done at run-time, when 
> the user specifies '-maltivec'.


It can done at runtime with -mabi=altivec but the configure option is 
so they do not have to
add it every time.

Thanks,
Andrew Pinski

Re: 3.4 bootstrap failure on powerpc-apple-darwin7.2.0

[Dale Johannesen]

On Jan 17, 2004, at 8:39 PM, Timothy J. Wood wrote:
> ../gcc/configure --prefix="$PREFIX" --enable-altivec 
> --enable-languages="c,c++,objc"
> make
>
> /Users/bungi/Source/GNU/gcc/gcc-3.4/build/gcc/xgcc ... 
> libgcc/./unwind-dw2.o
> /var/tmp//ccec9Kuq.s:4837:stvx vector instruction is optional for the 
> PowerPC (not allowed without -force_cpusubtype_ALL option)
>
>   Am I supposed to do something to get '-force_cpusubtype_ALL' (or 
> maybe -faltivec) passed to xgcc when bootstrapping, or is this an 
> actual bug?

It seems desirable to me to be able to enable the compiler to generate 
Altivec code without
actually doing so in the bootstrap.  This bootstrapped compiler would 
not run on a G3.

Re: 3.4 bootstrap failure on powerpc-apple-darwin7.2.0

[Segher Boessenkool]

On 19-jan-04, at 0:18, Andrew Pinski wrote:
>>> --enable-altivec is really only for ppc linux to enable the altivec 
>>> ABI by default which is
>>> done for PPC Darwin by default already.
>
>> Actually, in the course of my recent AltiVec struggles, I failed to 
>> discover any purpose whatsoever for the '--enable-altivec' config 
>> switch -- perhaps Aldy can enlighten us.
>
> Doing a simple search using google, I found when this was added:
> http://gcc.gnu.org/ml/gcc-patches/2001-12/msg02123.html
>
> It is exactly does what I said it does is to change the ABI by default.

Another way to do this is to configure for a 
powerpc-whatever-linux-gnualtivec
target.  Maybe this second way should go?


Segher