[Bug tree-optimization/59249] Jump threading makes if-conversion and following vectorization impossible.

[Bug tree-optimization/59249] Jump threading makes if-conversion and following vectorization impossible.

[rguenth at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=59249

Richard Biener <rguenth at gcc dot gnu.org> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|UNCONFIRMED                 |NEW
   Last reconfirmed|                            |2013-11-22
     Ever confirmed|0                           |1

--- Comment #1 from Richard Biener <rguenth at gcc dot gnu.org> ---
Btw, the issue is that we only have critical predecessors for inserting code
on.
The fix is to re-organize if-conversion to either split a random edge when
that happens or to re-organize it to compute edge predicates rather than
basic-block predicates.

[Bug tree-optimization/59249] if-conversion doesn't handle basic-blocks with only critical predecessor edges

[rguenth at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=59249

--- Comment #2 from Richard Biener <rguenth at gcc dot gnu.org> ---
Btw, see PR57521 for how the code is different from that on the 4.8 branch.

[Bug tree-optimization/59249] if-conversion doesn't handle basic-blocks with only critical predecessor edges

[bmei at broadcom dot com]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=59249

--- Comment #3 from Bingfeng Mei <bmei at broadcom dot com> ---
Richard, I am not sure I understand about how to split edge.

     BB4 
     / \
    /   \
  BB5    |
   |\    |
   | \   |
   |  \  |
   |   BB6
   |   /
   |  /
   BB7


Compiler (HEAD) complains "only critical predecessors of BB6" (its predcessor
BB5 has more than one successor). Do you suggest to split edge between BB5 &
BB6 and insert an empty BB? 

In the email thread, you blame poor implementation of tree-level if-conversion.
But RTL-level CE passes cannot handle that too.

[Bug tree-optimization/59249] if-conversion doesn't handle basic-blocks with only critical predecessor edges

[bmei at broadcom dot com]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=59249

--- Comment #4 from Bingfeng Mei <bmei at broadcom dot com> ---
Even I split one critical predecessor edge, predicate of BB6 is still ORed
result of two conditions from BB4 & BB5. ORing two conditions results in a
sequence of statements that cannot be vectorized. Vectorizer complains of
"bit-precision arithmetic not supported" because of boolean operations.

Not sure how to transform the code except reverting back to a form similar to
pre jump-threading.

[Bug tree-optimization/59249] if-conversion doesn't handle basic-blocks with only critical predecessor edges

[pinskia at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=59249

Andrew Pinski <pinskia at gcc dot gnu.org> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|NEW                         |RESOLVED
           See Also|                            |https://gcc.gnu.org/bugzill
                   |                            |a/show_bug.cgi?id=56541
      Known to work|                            |7.1.0
      Known to fail|7.1.0                       |6.1.0
         Resolution|---                         |FIXED
   Target Milestone|---                         |7.0

--- Comment #5 from Andrew Pinski <pinskia at gcc dot gnu.org> ---
Fixed in GCC 7, I suspect by r7-529-g1cfcd39eee915a758 (aka PR 56541).