On Fri, Mar 22, 2013 at 2:29 PM, Robert Haas <robertmhaas@gmail.com> wrote:
> The "barrier" essentially
> divides up the code into chunks and requires that those chunks be
> optimized independently by the compiler without knowledge of what
> earlier or later chunks are doing
While all this sounds sensible I would love to see a gcc programmer or
llvm programmer actually comment on what they think volatile does and
what they want to implement in the compiler.
I'm a bit worried that we're making assumptions like "things happen in
a specific order" that aren't really justified. In these days of
superscalar execution and multi-level caches things may be weirder
than we're imagining.
--
greg