This is the mail archive of the
gcc-patches@gcc.gnu.org
mailing list for the GCC project.
Re: More ifcvt woes
On Sat, May 13, 2000 at 09:37:28PM +0100, Richard Earnshaw wrote:
> Yep, that seems to work equally well; though I don't see why it is wrong
> to check the label counts in this case (is the count accurate or not?)...
The count is accurate, however it leads to what I'd call "superstitious"
coding practices. That is, checking or reverifying various bits of data
at random rather than relying on the CFG to be complete and accurate.
Such habits make it very difficult to reconstruct the logic later.
> Where block 84 only contains a conditional jump (the comparison has been
> cse'd into block 83), and block 85 only contains an unconditional jump (to
> the same location as block 84 conditionally jumps to.
Ug. I don't think ifcvt should be forced to deal with this kind
of nonsense. I wonder why cleanup_cfg didn't do its job...
From your description, it appears that ifcvt didn't do anything wrong,
strictly speaking. Merely that dead-code removal isn't something that
the sanity checks are expecting.
I'll look at this later today or tomorrow.
r~