llvm dev - Jul 2009 - [LLVMdev] ARM cross compiling causes segmentation fault

I tried a couple of options (-mcpu=arm1136j-s, -mcpu=arm1136jf-s,
-march=armv6, ...) to let the compile know the specific ARM processor, but
the same issue is still there. I tried to take a look at .s file in /tmp
directory, but it's already cleaned up. Is it because I enabled the
optimization option when I compiled llvm?

Regards,
Won

On Wed, Jul 8, 2009 at 1:28 PM, Dale Johannesen <dalej at apple.com>
wrote:
>
> On Jul 8, 2009, at 11:04 AMPDT, Won J Jeon wrote:
>
> > Hello,
> >
> > I built a llvm toolchain for ARM and I'm trying to build a simple
> > application on ARM using LLVM. When I compiled the code, it has the
> > following message and the binary code has a segmentation fault when
> > it is executed.
> >
> > /tmp/ccJPjMe4.s: Assembler messages:
> > /tmp/ccJPjMe4.s:4250: rdhi, rdlo and rm must all be different
> > /tmp/ccJPjMe4.s:4589: rdhi, rdlo and rm must all be different
> > /tmp/ccJPjMe4.s:4694: rdhi, rdlo and rm must all be different
> > /tmp/ccJPjMe4.s:4899: rdhi, rdlo and rm must all be different
> > /tmp/ccJPjMe4.s:4915: rdhi, rdlo and rm must all be different
>
> Make sure you've told the compiler the correct version of ARM;
> different generations of the architecture have different restrictions
> of this sort.
> Look at the .s file to see what instruction it's complaining about.
> (MUL had such a restriction in early versions of ARM, but I'd expect
> the message to read differently if that's it.)
> If you've told the compiler the right version of ARM and you think the
> compiler is wrong, file a bug with reproducible example.
>
> _______________________________________________
> LLVM Developers mailing list
> LLVMdev at cs.uiuc.edu         http://llvm.cs.uiuc.edu
> http://lists.cs.uiuc.edu/mailman/listinfo/llvmdev
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL:
<http://lists.llvm.org/pipermail/llvm-dev/attachments/20090708/25bece08/attachment.html>

On Jul 8, 2009, at 12:52 PMPDT, Won J Jeon wrote:
> I tried a couple of options (-mcpu=arm1136j-s, -mcpu=arm1136jf-s, - 
> march=armv6, ...) to let the compile know the specific ARM  
> processor, but the same issue is still there. I tried to take a look  
> at .s file in /tmp directory, but it's already cleaned up. Is it  
> because I enabled the optimization option when I compiled llvm?
Use -S to keep the assembly file around.

You don't seem to be very familiar with compilers, please try to  
figure things like this out from the documentation first.
> Regards,
> Won
>
> On Wed, Jul 8, 2009 at 1:28 PM, Dale Johannesen <dalej at apple.com>
> wrote:
>
> On Jul 8, 2009, at 11:04 AMPDT, Won J Jeon wrote:
>
> > Hello,
> >
> > I built a llvm toolchain for ARM and I'm trying to build a simple
> > application on ARM using LLVM. When I compiled the code, it has the
> > following message and the binary code has a segmentation fault when
> > it is executed.
> >
> > /tmp/ccJPjMe4.s: Assembler messages:
> > /tmp/ccJPjMe4.s:4250: rdhi, rdlo and rm must all be different
> > /tmp/ccJPjMe4.s:4589: rdhi, rdlo and rm must all be different
> > /tmp/ccJPjMe4.s:4694: rdhi, rdlo and rm must all be different
> > /tmp/ccJPjMe4.s:4899: rdhi, rdlo and rm must all be different
> > /tmp/ccJPjMe4.s:4915: rdhi, rdlo and rm must all be different
>
> Make sure you've told the compiler the correct version of ARM;
> different generations of the architecture have different restrictions
> of this sort.
> Look at the .s file to see what instruction it's complaining about.
> (MUL had such a restriction in early versions of ARM, but I'd expect
> the message to read differently if that's it.)
> If you've told the compiler the right version of ARM and you think the
> compiler is wrong, file a bug with reproducible example.
>
> _______________________________________________
> LLVM Developers mailing list
> LLVMdev at cs.uiuc.edu         http://llvm.cs.uiuc.edu
> http://lists.cs.uiuc.edu/mailman/listinfo/llvmdev
>
> _______________________________________________
> LLVM Developers mailing list
> LLVMdev at cs.uiuc.edu         http://llvm.cs.uiuc.edu
> http://lists.cs.uiuc.edu/mailman/listinfo/llvmdev
-------------- next part --------------
An HTML attachment was scrubbed...
URL:
<http://lists.llvm.org/pipermail/llvm-dev/attachments/20090708/cfb8d249/attachment.html>

Run llvm-gcc with "-save-temps" and it will keep the .s file for you  
to look at.

On Jul 8, 2009, at 12:52 PM, Won J Jeon wrote:
> I tried a couple of options (-mcpu=arm1136j-s, -mcpu=arm1136jf-s, - 
> march=armv6, ...) to let the compile know the specific ARM  
> processor, but the same issue is still there. I tried to take a look  
> at .s file in /tmp directory, but it's already cleaned up. Is it  
> because I enabled the optimization option when I compiled llvm?
>
> Regards,
> Won
>
> On Wed, Jul 8, 2009 at 1:28 PM, Dale Johannesen <dalej at apple.com>
> wrote:
>
> On Jul 8, 2009, at 11:04 AMPDT, Won J Jeon wrote:
>
> > Hello,
> >
> > I built a llvm toolchain for ARM and I'm trying to build a simple
> > application on ARM using LLVM. When I compiled the code, it has the
> > following message and the binary code has a segmentation fault when
> > it is executed.
> >
> > /tmp/ccJPjMe4.s: Assembler messages:
> > /tmp/ccJPjMe4.s:4250: rdhi, rdlo and rm must all be different
> > /tmp/ccJPjMe4.s:4589: rdhi, rdlo and rm must all be different
> > /tmp/ccJPjMe4.s:4694: rdhi, rdlo and rm must all be different
> > /tmp/ccJPjMe4.s:4899: rdhi, rdlo and rm must all be different
> > /tmp/ccJPjMe4.s:4915: rdhi, rdlo and rm must all be different
>
> Make sure you've told the compiler the correct version of ARM;
> different generations of the architecture have different restrictions
> of this sort.
> Look at the .s file to see what instruction it's complaining about.
> (MUL had such a restriction in early versions of ARM, but I'd expect
> the message to read differently if that's it.)
> If you've told the compiler the right version of ARM and you think the
> compiler is wrong, file a bug with reproducible example.
>
> _______________________________________________
> LLVM Developers mailing list
> LLVMdev at cs.uiuc.edu         http://llvm.cs.uiuc.edu
> http://lists.cs.uiuc.edu/mailman/listinfo/llvmdev
>
> _______________________________________________
> LLVM Developers mailing list
> LLVMdev at cs.uiuc.edu         http://llvm.cs.uiuc.edu
> http://lists.cs.uiuc.edu/mailman/listinfo/llvmdev
-------------- next part --------------
An HTML attachment was scrubbed...
URL:
<http://lists.llvm.org/pipermail/llvm-dev/attachments/20090708/ab442549/attachment.html>

Thanks. I could take a look at the lines and all of them have smull
instruction like 'smull r0, r1, r0, r1'.

Won

On Wed, Jul 8, 2009 at 2:54 PM, Dale Johannesen <dalej at apple.com>
wrote:
>
>  On Jul 8, 2009, at 12:52 PMPDT, Won J Jeon wrote:
>
>  I tried a couple of options (-mcpu=arm1136j-s, -mcpu=arm1136jf-s,
> -march=armv6, ...) to let the compile know the specific ARM processor, but
> the same issue is still there. I tried to take a look at .s file in /tmp
> directory, but it's already cleaned up. Is it because I enabled the
> optimization option when I compiled llvm?
>
>
> Use -S to keep the assembly file around.
>
> You don't seem to be very familiar with compilers, please try to figure
> things like this out from the documentation first.
>
>
>  Regards,
> Won
>
> On Wed, Jul 8, 2009 at 1:28 PM, Dale Johannesen <dalej at apple.com>
wrote:
>
>>
>> On Jul 8, 2009, at 11:04 AMPDT, Won J Jeon wrote:
>>
>> > Hello,
>> >
>> > I built a llvm toolchain for ARM and I'm trying to build a
simple
>> > application on ARM using LLVM. When I compiled the code, it has
the
>> > following message and the binary code has a segmentation fault
when
>> > it is executed.
>> >
>> > /tmp/ccJPjMe4.s: Assembler messages:
>> > /tmp/ccJPjMe4.s:4250: rdhi, rdlo and rm must all be different
>> > /tmp/ccJPjMe4.s:4589: rdhi, rdlo and rm must all be different
>> > /tmp/ccJPjMe4.s:4694: rdhi, rdlo and rm must all be different
>> > /tmp/ccJPjMe4.s:4899: rdhi, rdlo and rm must all be different
>> > /tmp/ccJPjMe4.s:4915: rdhi, rdlo and rm must all be different
>>
>> Make sure you've told the compiler the correct version of ARM;
>> different generations of the architecture have different restrictions
>> of this sort.
>> Look at the .s file to see what instruction it's complaining about.
>> (MUL had such a restriction in early versions of ARM, but I'd
expect
>> the message to read differently if that's it.)
>> If you've told the compiler the right version of ARM and you think
the
>> compiler is wrong, file a bug with reproducible example.
>>
>> _______________________________________________
>> LLVM Developers mailing list
>> LLVMdev at cs.uiuc.edu         http://llvm.cs.uiuc.edu
>> http://lists.cs.uiuc.edu/mailman/listinfo/llvmdev
>>
>
> _______________________________________________
> LLVM Developers mailing list
> LLVMdev at cs.uiuc.edu         http://llvm.cs.uiuc.edu
> http://lists.cs.uiuc.edu/mailman/listinfo/llvmdev
>
>
>
> _______________________________________________
> LLVM Developers mailing list
> LLVMdev at cs.uiuc.edu         http://llvm.cs.uiuc.edu
> http://lists.cs.uiuc.edu/mailman/listinfo/llvmdev
>
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL:
<http://lists.llvm.org/pipermail/llvm-dev/attachments/20090708/2ae6d687/attachment.html>