Re: [gimple] assignments to volatile

[Nathan Sidwell <nathan@codesourcery.com>]

On 06/25/10 17:41, Mike Stump wrote:
> On Jun 24, 2010, at 11:35 PM, Nathan Sidwell wrote:
>> On 06/24/10 16:42, Mike Stump wrote:
>>
>>> Sorry for not expounding.  I thought it would be obvious I was talking about well formed code...  First case would be something like:
>>>
>>> volatile int i, j, k;
>>> volatile int vi;
>>> void foo(volatile int&   j) {
>>>    foo(i = k);
>>> }
>>>
>>> here, we need to bind j to the address of i.  The second case is something like:
>>
>> Correct.  The patch does not change this behaviour -- the C++ front end already has to explicitly break out the lvalue from the assignment to conform to the semantics of TREEs.  The current G++ behaviour is not to re-read the assigned-to value.
>
> g++ seems not to share your position:
>
> 	movl	_i, %eax
> 	movl	%eax, (%esp)
>
> volatile int i, j, k;
> volatile int vi;
> void foo(int j) {
>    foo(i = k);
> }
>
> GNU C++ (GCC) version 4.6.0 20100620 (experimental) [trunk revision 161045] (x86_64-apple-darwin10)
> 	compiled by GNU C version 4.2.1 (Apple Inc. build 5659), GMP version 5.0.1
>
> ?  Certainly I must misunderstand something, can you spot it?

You have changed the testcase to be passing by value from passing by reference. 
  Why do you think those would be equivalent?

nathan
-- 
Nathan Sidwell    ::   http://www.codesourcery.com   ::         CodeSourcery