[Bug target/64009] New: ICE when compiling pr48335-2.c with armeb-linux-gnueabi-gcc

[chenshanyaoboy at gmail dot com]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=64009

            Bug ID: 64009
           Summary: ICE when compiling pr48335-2.c with
                    armeb-linux-gnueabi-gcc
           Product: gcc
           Version: 4.9.3
            Status: UNCONFIRMED
          Severity: major
          Priority: P3
         Component: target
          Assignee: unassigned at gcc dot gnu.org
          Reporter: chenshanyaoboy at gmail dot com

armeb-linux-gnueabi-gcc -c -O2  testsuite/gcc.dg/pr48335-2.c 
pr48335-2.c: In function 'f5':
pr48335-2.c:58:1: error: unrecognizable insn:
 }
 ^
(insn 8 7 9 2 (set (zero_extract:SI (subreg:SI (reg/v:SF 112 [ s+4 ]) 0)
            (const_int 16 [0x10])
            (const_int -8 [0xfffffffffffffff8]))
        (reg/v:SI 113 [ x ])) pr48335-2.c:55 -1
     (nil))
pr48335-2.c:58:1: internal compiler error: in extract_insn, at recog.c:2202

small example extracted from pr48335-2.c
cat ice.c

typedef short U __attribute__((may_alias, aligned (1)));

struct S
{
  _Complex float d __attribute__((aligned (8)));
};

void bar(struct S);

void f5 (int x)
{
  struct S s = { .d = 0.0f };
  ((U *)((char *) &s.d + 1))[3] = x;
  __real__ s.d *= 7.0;
  bar (s);
}

This ICE just happened in big-endian mode, it's ok  in little-endian.
aarch64 also have the same problem, when using aarch64-linux-gnu-gcc, it
reported:
aarch64-linux-gnu-gcc -w -c -O1 ice.c -mbig-endian
/tmp/cc9vId7S.s: Assembler messages:
/tmp/cc9vId7S.s:12: Error: immediate value out of range 0 to 31 at operand 3 --
`bfi w1,w0,-8,16'