This is the mail archive of the
gcc-patches@gcc.gnu.org
mailing list for the GCC project.
Re: Simple commit_edge_insertions fix
> On Sun, May 14, 2000 at 10:43:01PM +0200, Jan Hubicka wrote:
> > OK. Only what we would misshandle is potential function doing both
> > jump and set of some pseudo.
> > We don't have such beasts currently I guess.
>
> Yes we do. See decrement-and-branch. That's not an unconditional
Duh :)
> jump though, which is what the code in commit_edge_insertions was
> looking for.
OK.
Can you clarify for me what will happen here? Will such edge appear
as abnormal or normal edge. In case it is normal what will happen
if I insert to it? I see it can be handled by forcing the edge split
(in case we can't emit it at the beggining of destination),
but I can't see any test that would discover such thinks in
commit_edge_insertions.
Maybe this is just non-issue here, since we don't emit such nasty
code depending on changed registers.
I see your point about sibling calls.
I missunderstood our notion about exit edge here. I expected that
the code emited is expected to end before call in same way as
it ends before return in normal edges.
I wasn't thiking much about this issues.
Honza
>
>
> r~