[Bug target/70873] [7 Regressio] 20% performance regression at 482.sphinx3 after r235442 with -O2 -m32 on Haswell.

[ubizjak at gmail dot com]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=70873

--- Comment #27 from Uroš Bizjak <ubizjak at gmail dot com> ---
(In reply to H.J. Lu from comment #26)

> But when this splitter fails, no other splitters will be tried.

Bah. This is clearly an implementation bug in the split pass. I don't think we
have to work around it, the infrastructure bug should be fixed instead. Can you
please open a new bug about it?

I propose to go ahead and commit my patch, which is already accumulating
several semi-related fixes and improvements. The patch will probably expose an
optimization problem in the splitting pass, so we will have a testcase to
illustrate the problem.