How do disable generation of LDRD in ARM

[Roger Cruz]

Thanks Andrew.

Yes, I recognized that the code was buggy and it works fine on x86 but not for ARM v7 where alignment was mandatory.  I was hoping that there would be an easy way to cheat by using something like -mfix_cortexm3_ldrd but that didn't work for me.  So I fixed the code with a memcpy which I knew was the right way to do it.  I'll push the changes to the OSS community then.

Roger R. Cruz




On Dec 26, 2012, at 8:00 AM, Andrew Haley <aph@redhat.com> wrote:

> On 12/24/2012 07:53 PM, Roger Cruz wrote:
>> 
>> 
>> I am compiling this piece of code from an open source project that I
>> wish not to have to change.  The problem is that when compiled for
>> ARM, it generates an LDRD instruction, which when executed, causes a
>> bus error since the address in ptr is not doubleword aligned.
> 
> Well, that's going to be tough.  Your code isn't legal C, and a
> compiler is not obligated to generate anything sensible for it.  You
> can't just convert an arbitrary byte pointer to a UINT64* and expect
> working code.  If you lie to the compiler it will bite you.
> 
>> I know I can change the code but I prefer to leave it intact as
>> there may be many more instances of this and I don't wish to
>> maintain the code.
> 
> There's no point talking about "maintenance" because the code is
> already broken.
> 
>> 248    static DWORD64 dwarf2_get_u8(const unsigned char* ptr)
>> 249    {
>> 250        return *(const UINT64*)ptr;
>> 251    }
> 
> This is the right way to do it:
> 
> static DWORD64 dwarf2_get_u8(const unsigned char* ptr)
> {
>  UINT64 poo;
>  memcpy(&poo, ptr, sizeof poo);
>  return poo;
> }
> 
> If GCC can detect that ptr is correctly aligned, it will generate an
> LDRD or LDM instruction, as appropriate.
> 
>> Happy Holidays,
> 
> You too!  :-)
> 
> Andrew.