[Bug rtl-optimization/80301] Sub-optimal code with an array of structs offsetted inside a struct global on x86/x86_64 at -O2

[rguenth at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=80301

Richard Biener <rguenth at gcc dot gnu.org> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
           Keywords|                            |missed-optimization
             Status|UNCONFIRMED                 |NEW
   Last reconfirmed|                            |2017-04-04
          Component|target                      |rtl-optimization
            Version|unknown                     |7.0.1
     Ever confirmed|0                           |1

--- Comment #2 from Richard Biener <rguenth at gcc dot gnu.org> ---
So we fail to combine/fwprop

(insn 9 8 4 2 (parallel [
            (set (reg:DI 92)
                (plus:DI (reg:DI 91 [ index ])
                    (const_int 2 [0x2])))
            (clobber (reg:CC 17 flags))
        ]) "t.c":14 218 {*adddi_1}
     (expr_list:REG_DEAD (reg:DI 91 [ index ])
        (expr_list:REG_UNUSED (reg:CC 17 flags)
            (nil))))

(insn 10 4 11 2 (set (reg:CCZ 17 flags)
        (compare:CCZ (mem:SI (plus:DI (mult:DI (reg:DI 92)
                        (const_int 8 [0x8]))
                    (symbol_ref:DI ("m") [flags 0x2]  <var_decl 0x7f8f2c5b4e10
m>)) [2 m.s[index_4(D)].a+0 S4 A32])
            (reg:SI 88 [ <retval> ]))) "t.c":14 7 {*cmpsi_1}
     (nil))

which is because simply forwarding reg:DI 92 isn't possible and we do not know
how to simplify it into sth valid by associating the 2 into the symbol_ref part
of the outer plus.  For the other MEM things happen to be associated
differently
by CSE1.