[avr-libc-dev] Re: Inline assembler and compiler optimization

[David Brown]

Weddington, Eric wrote:
>  
>
> > -----Original Message-----
> > From: 
> > address@hidden 
> > [mailto:address@hidden
> > org] On Behalf Of David Brown
> > Sent: Tuesday, March 17, 2009 4:19 PM
> > To: address@hidden
> > Subject: [avr-libc-dev] Re: Inline assembler and compiler optimization
>
> > Googling for "gcc memory_barrier" suggests that it is a name 
> > used in the
> > RTL patterns.
>
> Hmm. Need to look more into the machine description then.
>  
> > The linux kernel uses names such as mb(), rmb() and wmb(), which are a
> > bit too short.  On the architectures I looked at, these are defined to
> > be "barrier()", which for gcc is in turn defined as a volatile memory
> > clobber and for the intel compiler it is defined as the
> > __memory_barrier() builtin.
>  
> Well that's certainly one route. We could perhaps add it as a gcc builtin.
>
> Thanks for doing this research, David. :-)

Thank Google :-)

I would imagine that a builtin involves more work than just #defining 
__memory_barrier() as a memory clobber.  It would be better, however, if 
gcc could have a consistent set of memory barrier builtins (including 
the read and write barriers used by Linux) that could be used across all 
targets.  I don't know how much influence you have in the gcc world at 
large...

Either way, a "#define memory_barrier() __memory_barrier()" in one of 
the headers would be useful to give a slightly more friendly name to users.

mvh.,

David