[Bug target/77729] aarch64 inserts unneeded uxtb after ldrb, orr ...32

[pinskia at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=77729

Andrew Pinski <pinskia at gcc dot gnu.org> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
           Keywords|                            |missed-optimization
             Status|UNCONFIRMED                 |NEW
   Last reconfirmed|                            |2016-09-25
     Ever confirmed|0                           |1
           Severity|normal                      |enhancement

--- Comment #10 from Andrew Pinski <pinskia at gcc dot gnu.org> ---
This is how we expand it on aarch64:
(insn 10 9 11 (set (reg:QI 81)
        (mem:QI (reg/v/f:DI 80 [ string ]) [0 *string_9(D)+0 S1 A8])) t.c:4 -1
     (nil))

(insn 11 10 12 (set (reg:SI 82)
        (ior:SI (subreg:SI (reg:QI 81) 0)
            (const_int 32 [0x20]))) t.c:4 -1
     (nil))

(insn 12 11 13 (set (reg:SI 83)
        (zero_extend:SI (subreg:QI (reg:SI 82) 0))) t.c:4 -1
     (nil))

(insn 13 12 14 (set (reg:CC 66 cc)
        (compare:CC (reg:SI 83)
            (const_int 116 [0x74]))) t.c:4 -1
     (nil))


-----
(set (reg:SI 83)
    (ior:SI (and:SI (subreg:SI (mem:QI (reg/v/f:DI 80 [ string ]) [0
*string_9(D)+0 S1 A8]) 0)
            (const_int 223 [0xdf]))
        (const_int 32 [0x20])))

Notice how the and there is 223, but really that can be still a zero_extend. 
Basically combine is going funny.

---- CUT ----

Note for 33, orr does not accept 33 so combine does not see 33 and does not
change the and part around the subreg.