[Bug rtl-optimization/117938] [15 Regression] wrong code with -O2 --param=max-cse-insns=1 with late-combine (since r15-1735-ge62ea4fb8ffcab)

[pinskia at gcc dot gnu.org via Gcc-bugs]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=117938

--- Comment #12 from Andrew Pinski <pinskia at gcc dot gnu.org> ---
(In reply to Uroš Bizjak from comment #11)
> Created attachment 59822 [details]
> .postreload and .late_combine2 dumps
> 
> pr117938.c.323r.postreload
> pr117938.c.325r.late_combine2

  338: {sp:DI=r13:DI+0x80;clobber flags:CC;}
      REG_ARGS_SIZE 0
successfully matched this instruction to *adddi_1:
(parallel [
        (set (reg/f:DI 7 sp)
            (plus:DI (reg/f:DI 0 ax [262])
                (const_int 128 [0x80])))
        (clobber (reg:CC 17 flags))
    ])
original cost = 4 + 4 + 4 + 4 + 4 + 4 (weighted: 24.000000), replacement cost =
4 + 4 + 4 + 4 + 4 (weighted: 20.000000); keeping replacement
rescanning insn with uid = 190.
rescanning insn with uid = 192.
rescanning insn with uid = 194.
rescanning insn with uid = 196.
rescanning insn with uid = 338.
moving insn 338 after insn 197


so just the insn #s are different; oh because I accidently was using -g.
Anyways what I said still applies here too.