On 2/17/21 12:04 PM, Jeroen Dobbelaere wrote:>> I'm confused.
>>
>> Are you interested in adding `willreturn` to a function via clang?
> I think so (also getting a little confused :( ) ..
> What should the behavior be of a '__attribute__((const))' function
?
> Is the progress guaranteed ?
>
> See https://www.godbolt.org/z/GoPYhM
>
>
> // extern "C" ...
> extern int ptrfun1(int, int*, int*) __attribute__((const));
>
> int b[5];
>
> int foo() {
> int a[10];
> int index = ptrfun1(5, &a[0], &b[0]);
> a[index]=10;
> return a[index];
> }
>
> clang-11 is able to optimize this away
>
> clang-trunc is mixed:
> - for this case, the call will not optimize it away. (as far as I see,
since D94106)
> - But if you do not use the return value upfront, it will be optimized
away.
> ...
> int index = 3;
> ptrfun1(5, &a[0], &b[0]);
> ...
>
> If a '__attribute__((const))' function is allowed to not progress,
then not all llvm passes are aware of this
> with the current mapping on llvm attributes and imho, we'll need an
attribute to indicate that we want
> to have progress. Or, clang should map '__attribute__((const))' to
'readnone willreturn'.
FWIW, I think removal is correct because you run it as C++ program.
This is "just" a phase ordering issue. Run O3 again and trunk happily
removes it: https://www.godbolt.org/z/95qeah
That said, I agree, we should check why this is happening and what to do
about it.
Now wrt. the attribute lowering (const/pure) I think we could add
willreturn iff the standard is C++11 or newer, but
not otherwise. For C++ before 11 and C we can always have infinite loops
with constant conditions, IIRC.
~ Johannes
>
> Greetings,
>
> Jeroen Dobbelaere
>
>
>> -----Original Message-----
>> From: Johannes Doerfert <johannesdoerfert at gmail.com>
>> Sent: Wednesday, February 17, 2021 18:11
>> To: Jeroen Dobbelaere <dobbel at synopsys.com>; llvm-dev at
lists.llvm.org; Florian
>> Hahn <florian_hahn at apple.com>
>> Subject: Re: willreturn
>>
>> I'm confused.
>>
>> Are you interested in adding `willreturn` to a function via clang?
>>
>>
>> On 2/17/21 11:05 AM, Jeroen Dobbelaere wrote:
>>> After the '[local] treat calls that may not return as being
alive' patch,
>>
https://urldefense.com/v3/__https://reviews.llvm.org/D94106__;!!A4F2R9G_pg!LFI
>> Ix70hH4oLQtVkzqKiChp_3Ubjedy7AJEChQFkMZjrjdF1WwkKgf1qBaJL0GzkK45Vdo4q$
>>> it seems not to be possible in clang to indicate that a 'C'
function or an
>> 'extern "C"' function will progress.
>>> At least, I did not find an attribute that indicates this. (
>> __attribute__((const)) is not sufficient)
>>> Is there already a clang attribute to indicate this ?
>>>
>>> Thanks,
>>>
>>> Jeroen Dobbelaere
>>>