This is the mail archive of the
mailing list for the GCC project.
Re: Crossjumping fix II
> On Tue, Dec 18, 2001 at 03:46:00PM +0100, Jan Hubicka wrote:
> > It allows insns to match even when number of fake edges does not...
> Err, when would this happen? If there is a fake edge, it had
> better well be the only successor, due to the way they're created.
Hmm, in this case probably yes.
We may have fake edges from unreachable loops, or from call sites, but they are
not created when crossjumping is in progress. I just added them in case
someone in future will use fake edges for something else and not ignoring fake
edges just looked incorrect to me.
Would be the patch OK w/o that hunk?
> And, frankly, I would think they'd better match too.