Re: [i386, patch, RFC] HLE support in GCC

[Andi Kleen <andi at firstfloor dot org>]

On Wed, Apr 11, 2012 at 03:12:44PM +0200, Jakub Jelinek wrote:
> On Wed, Apr 11, 2012 at 03:06:35PM +0200, Andi Kleen wrote:
> > Do you really imply ACQUIRE/RELEASE with HLE_ACQUIRE/RELEASE now? I don't
> > see that in the code. I think that's really required, otherwise the optimizer
> > will do the wrong thing and move memory references outside the region.
> 
> IMHO the separate bits for HLE_ACQUIRE/RELEASE are desirable, you could very
> well use __ATOMIC_ACQ_REL or __ATOMIC_SEQ_CST if you want stronger memory

That would still work, right? If you have multiple the stronger wins.

> barrier, still coupled with either __ATOMIC_HLE_ACQUIRE or
> __ATOMIC_HLE_RELEASE...

If the compiler barrier is not implied I can just see lots of buggy code that 
subtly breaks with higher optimizer levels, compiler updates etc. Of course 
it will be all gccs fault.

-Andi

-- 
ak@linux.intel.com -- Speaking for myself only.