[Bug target/114077] New: ICE in do_SUBST at combine.cc with aarch64 crosscompiler

[Bug target/114077] New: ICE in do_SUBST at combine.cc with aarch64 crosscompiler

[pheeck at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=114077

            Bug ID: 114077
           Summary: ICE in do_SUBST at combine.cc with aarch64
                    crosscompiler
           Product: gcc
           Version: 14.0
            Status: UNCONFIRMED
          Keywords: ice-on-valid-code, needs-bisection
          Severity: normal
          Priority: P3
         Component: target
          Assignee: unassigned at gcc dot gnu.org
          Reporter: pheeck at gcc dot gnu.org
  Target Milestone: ---
              Host: x86_64-linux
            Target: aarch64-gnu-linux

While compiling the gcc.dg/asan/pr107317.c testsuite testcase using aarch64
crosscompiler:

aarch64-linux-gnu-gcc
/home/worker/buildworker/tiber-option-juggler/build/gcc/testsuite/gcc.dg/asan/pr107317.c
-fno-optimize-sibling-calls -fharden-control-flow-redundancy -mabi=ilp32 -Ofast

we hit this ICE:

during RTL pass: combine
/home/worker/buildworker/tiber-option-juggler/build/gcc/testsuite/gcc.dg/asan/pr107317.c:
In function ‘foo’:
/home/worker/buildworker/tiber-option-juggler/build/gcc/testsuite/gcc.dg/asan/pr107317.c:13:1:
internal compiler error: in do_SUBST, at combine.cc:698
   13 | }
      | ^
0x78f6e1 do_SUBST
       
/home/worker/buildworker/tiber-gcc-trunk-aarch64/build/gcc/combine.cc:698
0x199342c subst
       
/home/worker/buildworker/tiber-gcc-trunk-aarch64/build/gcc/combine.cc:5583
0x199337d subst
       
/home/worker/buildworker/tiber-gcc-trunk-aarch64/build/gcc/combine.cc:5536
0x1996348 try_combine
       
/home/worker/buildworker/tiber-gcc-trunk-aarch64/build/gcc/combine.cc:3302
0x199afd4 combine_instructions
       
/home/worker/buildworker/tiber-gcc-trunk-aarch64/build/gcc/combine.cc:1264
0x199afd4 rest_of_handle_combine
       
/home/worker/buildworker/tiber-gcc-trunk-aarch64/build/gcc/combine.cc:15091
0x199afd4 execute
       
/home/worker/buildworker/tiber-gcc-trunk-aarch64/build/gcc/combine.cc:15135
Please submit a full bug report, with preprocessed source (by using
-freport-bug).
Please include the complete backtrace with any bug report.
See <https://gcc.gnu.org/bugs/> for instructions.

There's an open PR for an ICE with roughly the same stacktrace: pr81874. Maybe
it's the same problem.

The testcase:

/* PR middle-end/107317 */
/* { dg-do compile { target ilp32 } } */
/* { dg-options "-fsanitize=address -ffat-lto-objects" } */

void bar (float *, float *); 

void
foo (void)      /* { dg-error "exceeds maximum" } */
{
  float a[400000000];
  float b[200000000];
  bar (a, b); 
}

The compiler configuration:

Using built-in specs.
COLLECT_GCC=/home/worker/cross/bin/aarch64-linux-gnu-gcc
COLLECT_LTO_WRAPPER=/home/worker/cross/libexec/gcc/aarch64-linux-gnu/14.0.1/lto-wrapper
Target: aarch64-linux-gnu
Configured with:
/home/worker/buildworker/tiber-gcc-trunk-aarch64/build/configure
--enable-languages=c,c++,fortran,rust,m2 --disable-bootstrap
--disable-libsanitizer --disable-multilib --enable-checking=release
--prefix=/home/worker/cross --target=aarch64-linux-gnu
--with-as=/usr/bin/aarch64-suse-linux-as
Thread model: posix
Supported LTO compression algorithms: zlib zstd
gcc version 14.0.1 20240222 (experimental)
37127ed975e09813eaa2d1cf1062055fce45dd16 (GCC)

[Bug target/114077] ICE in do_SUBST at combine.cc with aarch64 crosscompiler with -fharden-control-flow-redundancy and -mabi=ilp32

[pinskia at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=114077

--- Comment #1 from Andrew Pinski <pinskia at gcc dot gnu.org> ---
I get a different ICE:
t5.c: In function ‘foo’:                                                       
                                                                               
                                                                             |
t5.c:9:1: internal compiler error: in decompose, at rtl.h:2298                 
                                                                               
                                                                             |
    9 | foo (void)      /* { dg-error "exceeds maximum" } */                   
                                                                               
                                                                             |
      | ^~~                                                                    
                                                                               
                                                                             |
0x4e82fc wi::int_traits<std::pair<rtx_def*, machine_mode> >::decompose(long*,
unsigned int, std::pair<rtx_def*, machine_mode> const&)
        ../../gcc/rtl.h:2298
0x4e82fc wide_int_ref_storage<false,
true>::wide_int_ref_storage<std::pair<rtx_def*, machine_mode>
>(std::pair<rtx_def*, machine_mode> const&)
        ../../gcc/wide-int.h:1089
0x4e82fc generic_wide_int<wide_int_ref_storage<false, true>
>::generic_wide_int<std::pair<rtx_def*, machine_mode> >(std::pair<rtx_def*,
machine_mode> const&)
        ../../gcc/wide-int.h:847
0x4e82fc simplify_const_unary_operation(rtx_code, machine_mode, rtx_def*,
machine_mode)
        ../../gcc/simplify-rtx.cc:1991
0xd46d92 simplify_context::simplify_unary_operation(rtx_code, machine_mode,
rtx_def*, machine_mode)
        ../../gcc/simplify-rtx.cc:889
0x90aea8 simplify_unary_operation(rtx_code, machine_mode, rtx_def*,
machine_mode)
        ../../gcc/rtl.h:3486
0x90aea8 convert_memory_address_addr_space_1(scalar_int_mode, rtx_def*,
unsigned char, bool, bool)
        ../../gcc/explow.cc:330
0x90b0c9 convert_memory_address_addr_space_1(scalar_int_mode, rtx_def*,
unsigned char, bool, bool)
        ../../gcc/explow.cc:375
0x90b1b9 convert_memory_address_addr_space(scalar_int_mode, rtx_def*, unsigned
char)
        ../../gcc/explow.cc:429
0x90b1b9 memory_address_addr_space(machine_mode, rtx_def*, unsigned char)
        ../../gcc/explow.cc:443
0x930da4 expand_expr_real_1(tree_node*, rtx_def*, machine_mode,
expand_modifier, rtx_def**, bool)
        ../../gcc/expr.cc:11720
0x94148a expand_expr(tree_node*, rtx_def*, machine_mode, expand_modifier)
        ../../gcc/expr.h:316
0x94148a expand_assignment(tree_node*, tree_node*, bool)
        ../../gcc/expr.cc:6397
0x7fab67 expand_gimple_stmt_1
        ../../gcc/cfgexpand.cc:3992
0x7fab67 expand_gimple_stmt
        ../../gcc/cfgexpand.cc:4071
0x7ffc40 expand_gimple_basic_block
        ../../gcc/cfgexpand.cc:6127
0x801d06 execute
        ../../gcc/cfgexpand.cc:6866

[Bug target/114077] ICE in do_SUBST at combine.cc with aarch64 crosscompiler with -fharden-control-flow-redundancy and -mabi=ilp32

[pinskia at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=114077

--- Comment #2 from Andrew Pinski <pinskia at gcc dot gnu.org> ---
Oh yes because I am building without `--enable-checking=release` .

[Bug target/114077] ICE in do_SUBST at combine.cc with aarch64 crosscompiler with -fharden-control-flow-redundancy and -mabi=ilp32

[pinskia at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=114077

--- Comment #3 from Andrew Pinski <pinskia at gcc dot gnu.org> ---
#5  0x000000000110cb3f in simplify_const_unary_operation (code=ZERO_EXTEND,
mode=E_DImode, op=0x7ffff76e0440, op_mode=E_SImode) at
../../gcc/simplify-rtx.cc:2131
2131              result = wide_int::from (op0, width, UNSIGNED);
(gdb) p op0
$1 = {<std::__pair_base<rtx_def*, machine_mode>> = {<No data fields>}, first =
0x7ffff76e0440, second = E_SImode}
(gdb) p width                                                                  
                                                                               
                                                                             |
$2 = 64
(gdb) p op0.first                                                              
                                                                               
                                                                             |
$3 = (rtx_def *) 0x7ffff76e0440
(gdb) p debug_rtx(op0.first)
(const_int -2400000008 [0xffffffff70f2e7f8])
$4 = void

[Bug target/114077] ICE in do_SUBST at combine.cc with aarch64 crosscompiler with -fharden-control-flow-redundancy and -mabi=ilp32

[pinskia at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=114077

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

           What    |Removed                     |Added
----------------------------------------------------------------------------
           See Also|https://gcc.gnu.org/bugzill |
                   |a/show_bug.cgi?id=81874     |

--- Comment #4 from Andrew Pinski <pinskia at gcc dot gnu.org> ---
It has nothing to do with PR 81874 .
The issue happens much earlier. During expand, should have rejected this as the
arrays being too large.

[Bug target/114077] ICE in do_SUBST at combine.cc with aarch64 crosscompiler with -fharden-control-flow-redundancy and -mabi=ilp32

[pinskia at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=114077

--- Comment #5 from Andrew Pinski <pinskia at gcc dot gnu.org> ---
This fixes patch causes the code to be rejected:
```
diff --git a/gcc/function.cc b/gcc/function.cc
index 5ffd438475e..7a0f7faa2d7 100644
--- a/gcc/function.cc
+++ b/gcc/function.cc
@@ -242,7 +242,7 @@ frame_offset_overflow (poly_int64 offset, tree func)
 {
   poly_uint64 size = FRAME_GROWS_DOWNWARD ? -offset : offset;
   unsigned HOST_WIDE_INT limit
-    = ((HOST_WIDE_INT_1U << (GET_MODE_BITSIZE (Pmode) - 1))
+    = ((HOST_WIDE_INT_1U << (GET_MODE_BITSIZE (ptr_mode) - 1))
        /* Leave room for the fixed part of the frame.  */
        - 64 * UNITS_PER_WORD);


```
Though I am not 100% sure if it is correct.