[Bug target/119374] [12/13/14/15 regression] armv7ve+neon-vfpv4: Miscompilation with -O2

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

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=119374

--- Comment #2 from Andrew Pinski <pinskia at gcc dot gnu.org> ---

;; shift_36 = 19 - _8;

(insn 20 19 0 (set (reg/v:SI 133 [ shiftD.6157 ])
        (minus:SI (const_int 19 [0x13])
            (reg:SI 121 [ _8 ]))) "/app/example.cpp":33:11 discrim 1 -1
     (nil))


;; vect_yx_40.8_78 = vect_yy_34.7_80 >> shift_36;

(insn 34 29 35 (set (reg:V4SI 162)
        (neg:V4SI (reg:V4SI 158))) "/app/example.cpp":39:5 -1
     (nil))

(insn 35 34 36 (set (reg:V4SI 161 [ vect_yx_40.8_78 ])
        (unspec:V4SI [
                (reg:V4SI 142 [ vect_yy_34.7D.6181 ])
                (reg:V4SI 162)
            ] VSHLQ_S)) "/app/example.cpp":39:5 -1
     (expr_list:REG_EQUAL (ashiftrt:V4SI (reg:V4SI 142 [ vect_yy_34.7D.6181 ])
            (reg:V4SI 158))
        (nil)))



Confirmed, I have seen this before.