Re: [dataflow]: PATCH COMMITTED to make arm-elf work.

[Kenneth Zadeck <zadeck at naturalbridge dot com>]

Richard Earnshaw wrote:
> On Fri, 2007-02-02 at 11:54 -0500, Kenneth Zadeck wrote:
>   
>> Paolo Bonzini wrote:
>>     
>>> Kenneth Zadeck wrote:
>>>       
>>>> The arm-elf is now regression free on the dataflow branch.  This patch
>>>> has been bootstrapped and regression tested on x86-64, x86-32, ppc and
>>>> ia-64 linux.
>>>>
>>>> Most of the changes to make the arm work were not actually in the arm
>>>> port.
>>>>
>>>> 1) ASM_INPUT with no parameters is no longer a acceptable substitute for
>>>> gen_blockage.  Gen_blockage, with no parameters is now required on all
>>>> platforms.
>>>>         
>>> Does this affect user code too?  Would it make sense to expand asm("")
>>> as a blockage instruction?
>>>
>>> Thanks,
>>>
>>> Paolo
>>>       
>> i think you need to ask someone else about this.  if asm_input with no
>> parms is really supposed to be volatile just because it has a pretty
>> face, then there are things that will have to be changed.  However,
>> rhyolite and stevenb did not think that this was the case.
>>     
>
> I think the docs say that any asm with no inputs or outputs is treated
> as volatile; however, I think this is handled in the front-ends (and
> they thus become volatile before the middle-end sees it).
>
> R.
>
>   
If that is true, then all is well.  What does not work is that if the
back end(s) generate a asm_input and just expect that the volatile fairy
is going to magically confer volatile on them as was done in the three
places that we changed to gen_blockage.