Re: Add restrict annotations to all functions that require it

[Alejandro Colomar via Libc-alpha <libc-alpha@sourceware.org>]

[-- Attachment #1.1: Type: text/plain, Size: 1997 bytes --]

Hi Joseph!

On 12/7/22 00:57, Joseph Myers wrote:
> On Sat, 3 Dec 2022, Alejandro Colomar via Libc-alpha wrote:
> 
>> For example, this would be nonsense (and I honestly don't know what the
>> compiler would understand from it):
>>
>> void broken(int *restrict a, int *b, int *c);
>> void broken(int *a, int *restrict b, int *c);
> 
> Qualifiers on parameters in a function declarator that is not part of a
> definition of that function have no semantic effect; restrict on function
> parameters only has a semantic effect in the definition of the function.
> 

Having restrict in the prototype may not change the semantics, but it is 
meaningful to the warnings:


$ cat restrict.c
void broken(int *restrict a, int *b, int *c);
void broken(int *a, int *restrict b, int *c);

void f(int *restrict p, int *restrict q, int *restrict r)
{
	broken(p, p, q);
	broken(p, q, p);
	broken(p, q, r);
}

$ cc -Wall -Wextra -S restrict.c
restrict.c: In function ‘f’:
restrict.c:6:9: warning: passing argument 1 to ‘restrict’-qualified parameter 
aliases with argument 2 [-Wrestrict]
     6 |         broken(p, p, q);
       |         ^~~~~~
restrict.c:7:9: warning: passing argument 1 to ‘restrict’-qualified parameter 
aliases with argument 3 [-Wrestrict]
     7 |         broken(p, q, p);
       |         ^~~~~~


By not having semantic effect you mean that the compiler (optimizer) won't take 
advantage of those if the prototype with restrict is seen in the translation 
unit where the function definition is compiled, given that the function 
definition doesn't specify restrict?

Whatever is the semantic result, I believe GCC should have a warning for 
mismatching restrict qualifiers.  The above doubt is reasonable, and a warning 
would help make sure that it's not relevant, because code should't do that 
normally, and with the help of the warning, that can be enforced.

Cheers,

Alex

-- 
<http://www.alejandro-colomar.es/>

[-- Attachment #2: OpenPGP digital signature --]
[-- Type: application/pgp-signature, Size: 833 bytes --]