This is the mail archive of the
gcc-patches@gcc.gnu.org
mailing list for the GCC project.
Re: [PATCH] Eliminate TREE_INVARIANT
Mark Mitchell wrote:
Richard Guenther wrote:
Not really, but maybe the FEs depend on things like
const int i = 5;
&a[i]
being an invariant address?
It's good to check that things work OK with -O0, but I think that if the
FEs depend on this, that's broken. TREE_INVARIANT is about what GCC can
prove; not what any language standard says. I think the FEs should be
using language-specific flags that correspond to language-specific
concepts (like C++'s "integral constant expressions").
Is this an approval? :-) (The patch obviously needs a GWP).
Paolo