[Bug rtl-optimization/69447] [5/6 Regression] wrong code with -O2 -fno-schedule-insns and mixed 8/16/32/64bit arithmetics @ armv7a

[rth at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=69447

--- Comment #5 from Richard Henderson <rth at gcc dot gnu.org> ---
This appears to be exposed by lra-remat, but not
the location of the bug.

Before remat:

   Insn 55: point = 3
   ...
   Insn 17: point = 26
 r111: [4..26]

After remat:

    Inserting rematerialization insn before:
   76: {r0:DI=r111:DI-0x2;clobber cc:CC;}

   Insn 55: point = 3
   ...
   Insn 76: point = 7
   ...
   Insn 17: point = 24

So remat hasn't expanded the lifetime of r111 at all,
merely introduced a new use within the existing live
range.  So the failure to keep the top half of the
pseudo live is elsewhere.