Re: [boost] [shared_ptr] possible bug in sp_counted_base_cw_ppc.hpp

From: Rene Rivera (grafik.list_at_[hidden])
Date: 2005-04-11 11:01:31

Peter Dimov wrote:
> Joaquín Mª López Muñoz wrote:
> 
>> I think that the functions
>>
>> atomic_increment
>> atomic_decrement
>> atomic_conditional_increment
>>
>> should be marked as inline, as they are defined in a header file. This
>> is causing multiple definition problems, see for instance
>>
>> http://tinyurl.com/6eub2
> 
> Yes, you are right. I'd fix it but I don't know whether "inline" needs 
> to go before or after "asm". ;-) Rene?

Sorry.. it took a while, waiting for the CVS update. And the answer is 
"neither". Function level assembly can't be inlined. At least that's 
what the compilers says :-)

Which means changing to use statement level assembly. So it would have 
to be something like this...

inline void atomic_increment( register long * pw )
{
     asm
     {
loop:

     lwarx   r4, 0, r3
     addi    r4, r4, 1
     stwcx.  r4, 0, r3
     bne-    loop
     }
}

Which when applied to the other functions, compiles without warnings. 
And since all the tests pass, I checked in the changes to CVS.

-- 
-- Grafik - Don't Assume Anything
-- Redshift Software, Inc. - http://redshift-software.com
-- rrivera/acm.org - grafik/redshift-software.com
-- 102708583/icq - grafikrobot/aim - Grafik/jabber.org