notes on current atomics config and usage

[Paolo Carlini]

Paolo Carlini wrote:

> Mark Mitchell wrote:
>
>> Paolo Carlini wrote:
>>
>>> In fact, ideally, from the library point of view, we would like to
>>> *inline* in the headers the builtin atomics and the user should not be
>>> allowed to pass -march=i386 and obtain unresolved symbols at link time
>>> too easily! 
>>
>> I'm not sure what you mean by this; if you use the intrinsics, then the
>> compiler will do the inlining for you?  But, I think the headers should
>> still permit compilation with -march=i386; there should be a macro to
>> check, and libstdc++ should check it. :-)
>
> As pointed out be Benjamin yesterday, there is a problem of 
> consistency between the atomics used inside in the library and those 
> used in the headers. If the user passes -march=i386 and doesn't want 
> unresolved symbols at link time the only ways are: 1- Inline the 
> mutexes, or, 2- Call mutexes inside the library.

Point 2- is not well stated, sorry. Indeed, user code compiled with 
-march=i386 could well call into the library and find atomics 
implemented with builtins, and everything would work fine, if done 
consistently, *but* we would *not* inline, and that was the initial issue.

Paolo.