[Bug rtl-optimization/83304] [8 Regression] FAIL: gcc.c-torture/execute/pr61725.c -O3 -fomit-frame-pointer -funroll-loops -fpeel-loops -ftracer -finline-functions

[segher at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=83304

--- Comment #9 from Segher Boessenkool <segher at gcc dot gnu.org> ---
At some point during combine we have

insn_cost 4 for    72: r127:SI=0xffffffffffffff81
insn_cost 4 for     9: r130:SI=ctz(r125:SI)
insn_cost 4 for    69: cc:CC=cmp(r125:SI,0)
insn_cost 4 for    70: r131:SI={(cc:CC!=0)?r130:SI:0xffffffffffffffff}
      REG_DEAD r130:SI
insn_cost 12 for    16: r133:SI=cc:CC==0
      REG_DEAD cc:CC
                    19: {r134:SI=r131:SI!=0xffffffffffffffff;clobber cc:CC;}
      REG_UNUSED cc:CC
      REG_DEAD r131:SI

and then it does

Trying 70 -> 19:
   70: r131:SI={(cc:CC!=0)?r130:SI:0xffffffffffffffff}
      REG_DEAD r130:SI
   19: {r134:SI=r131:SI!=0xffffffffffffffff;clobber cc:CC;}
      REG_UNUSED cc:CC
      REG_DEAD r131:SI
Failed to match this instruction:
(parallel [
        (set (reg:SI 134)
            (ne:SI (reg:CC 100 cc)
                (const_int 0 [0])))
        (clobber (reg:CC 100 cc))
    ])
Successfully matched this instruction:
(set (reg:SI 134)
    (ne:SI (reg:CC 100 cc)
        (const_int 0 [0])))
allowing combination of insns 70 and 19
original costs 4 + 16 = 20
replacement cost 12
deferring deletion of insn with uid = 70.
deferring deletion of insn with uid = 9.
modifying insn i3    19: r134:SI=cc:CC!=0
      REG_DEAD cc:CC
deferring rescan insn with uid = 19.

but the REG_DEAD note on insn 16 is left, although now 19 uses it too.