[Bug middle-end/91603] New: Unaligned access in expand_assignment

[bernd.edlinger at hotmail dot de]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=91603

            Bug ID: 91603
           Summary: Unaligned access in expand_assignment
           Product: gcc
           Version: 10.0
            Status: UNCONFIRMED
          Severity: normal
          Priority: P3
         Component: middle-end
          Assignee: unassigned at gcc dot gnu.org
          Reporter: bernd.edlinger at hotmail dot de
  Target Milestone: ---

Currently this ICEs due to the middle-end sanitization:

$ cat test.c
typedef __simd64_int32_t int32x2_t;
typedef __attribute__((aligned (1))) int32x2_t unalignedvec;

unalignedvec a = {11, 13};

void foo(unalignedvec *);

void test()
{
  unalignedvec x = a;
  foo (&x);
  a = x;
}
$ arm-linux-gnueabihf-gcc -O3 -S test.c
during RTL pass: expand
test.c: In function 'test':
test.c:10:16: internal compiler error: in gen_movv2si, at
config/arm/vec-common.md:30
   10 |   unalignedvec x = a;
      |                ^
0x7bb33c gen_movv2si(rtx_def*, rtx_def*)
        ../../gcc-trunk/gcc/config/arm/vec-common.md:30
0xa4a807 insn_gen_fn::operator()(rtx_def*, rtx_def*) const
        ../../gcc-trunk/gcc/recog.h:318
0xa4a807 emit_move_insn_1(rtx_def*, rtx_def*)
        ../../gcc-trunk/gcc/expr.c:3694
0xa4ab94 emit_move_insn(rtx_def*, rtx_def*)
        ../../gcc-trunk/gcc/expr.c:3790
0xa522bf store_expr(tree_node*, rtx_def*, int, bool, bool)
        ../../gcc-trunk/gcc/expr.c:5855
0xa52bfd expand_assignment(tree_node*, tree_node*, bool)
        ../../gcc-trunk/gcc/expr.c:5441
0xa52bfd expand_assignment(tree_node*, tree_node*, bool)
        ../../gcc-trunk/gcc/expr.c:4982
0x934adf expand_gimple_stmt_1
        ../../gcc-trunk/gcc/cfgexpand.c:3777
0x934adf expand_gimple_stmt
        ../../gcc-trunk/gcc/cfgexpand.c:3875
0x93a451 expand_gimple_basic_block
        ../../gcc-trunk/gcc/cfgexpand.c:5915
0x93c1b6 execute
        ../../gcc-trunk/gcc/cfgexpand.c:6538
Please submit a full bug report,
with preprocessed source if appropriate.
Please include the complete backtrace with any bug report.
See <https://gcc.gnu.org/bugs/> for instructions.

But it is actually a genuine wrong code bug:
with the assertions removed, it generates this code:

test:
        @ args = 0, pretend = 0, frame = 8
        @ frame_needed = 0, uses_anonymous_args = 0
        push    {r4, r6, r7, lr}
        movw    r4, #:lower16:.LANCHOR0
        movt    r4, #:upper16:.LANCHOR0
        sub     sp, sp, #8
        mov     r0, sp
        vldr    d16, [r4]   <= unaligned access (will trap)
        vstr    d16, [sp]
        bl      foo
        ldrd    r6, [sp]
        uxtb    ip, r6
        ubfx    r2, r6, #16, #8
        lsr     lr, r6, #24
        uxtb    r0, r7
        lsr     r1, r7, #24
        strb    ip, [r4]
        ubfx    r3, r7, #16, #8
        strb    r2, [r4, #2]
        ubfx    ip, r6, #8, #8
        ubfx    r2, r7, #8, #8
        strb    lr, [r4, #3]
        strb    ip, [r4, #1]
        strb    r0, [r4, #4]
        strb    r2, [r4, #5]
        strb    r3, [r4, #6]
        strb    r1, [r4, #7]
        add     sp, sp, #8
        @ sp needed
        pop     {r4, r6, r7, pc}