On 24/02/2020 20:08, Andres Freund wrote:
> Hi,
> 
> On 2020-02-24 19:18:26 +0300, Pavel Begunkov wrote:
>> On 24/02/2020 19:02, Jens Axboe wrote:
>>>> Usually doesn't work because of such possible "hackier assignments".
>>>> Ok, I have to go and experiment a bit. Anyway, it probably generates a lot of
>>>> useless stuff, e.g. for req->ctx
>>>
>>> Tried this, and it generates the same code...
>>
>> Maybe it wasn't able to optimise in the first place
>>
>> E.g. for the following code any compiler generates 2 reads (thanks godbolt).
>>
>> extern void foo(int);
>> int bar(const int *v)
>> {
>>     foo(*v);
>>     return *v;
>> }
> 
> Yea, the compiler really can't assume anything for this kind of
> thing.
> a) It's valid C to cast away the const here, as long as it's guaranteed
>    that v isn't pointing to to actually const memory.
> b) foo() could actually have access to *v without the argument,
>    e.g. through a global.
> and even in the case of a const member of a struct, as far as I know
> it's legal to change the values, as long as the allocation isn't const. 

Yep, regular stuff. And that's why I want to find a way to force compilers to
think otherwise. e.g. kind of __attribute__

-- 
Pavel Begunkov