This is the mail archive of the
gcc@gcc.gnu.org
mailing list for the GCC project.
Re: Updating the CFG on the fly?
On Tue, Nov 09, 1999 at 04:13:14PM +1300, Michael Hayes wrote:
> OK, I've quickly knocked one up that seems to work. I'd assumed that
> that the equivalent functionality must have been hiding somewhere
> under another guise.
No one else has needed it yet. *shrug*
> However, I noticed that split_edge renumbers
> all the basic blocks after the new block. What is the philosophy
> behind this since it appears that I'll have to recreate all the
> dominator information every tme I do this?
Normally one does not update the CFG in the middle of an
optimization pass, for this very reason. This is exactly
why insert_insn_on_edge doesn't do anything but cache the
info until commit_edge_insertions is called.
So I'll ask again what you are doing that requires you do
create new basic blocks in the middle. You gave me an
abstract description a few days ago and I didn't get it.
Pictures and a concrete example would help.
r~