[llvm-bugs] [Bug 24738] Loop vectorizer introduces bad "undef"

[via llvm-bugs]

https://llvm.org/bugs/show_bug.cgi?id=24738

Mattias Eriksson <mattias.v.eriksson at ericsson.com> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|RESOLVED                    |REOPENED
         Resolution|WORKSFORME                  |---

--- Comment #7 from Mattias Eriksson <mattias.v.eriksson at ericsson.com> ---
(In reply to comment #6)
> Ah, I see now, I thought opt was failing. Do you by any chance have less
> simplified example? I can see that this might be an issue, but I suspect
> that the rootcause might be somewhere else and to verify it it would be
> super helpful to have a test that really exposes a bug (in the current
> reproducer the undef value doesn't matter as we never take the corresponding
> edge).

Unfortunately I do not have saved the original problem. I remember that I found
it in randomized testing and that it was rare. I also remember that it was a
real wrong-code error that I had to spend a day or so to track down.

I will try to make an effort in finding a "real" error again by running some
more randomized testing during the weekend.

-- 
You are receiving this mail because:
You are on the CC list for the bug.
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.llvm.org/pipermail/llvm-bugs/attachments/20160429/791adc32/attachment.html>