[Bug tree-optimization/97666] New: [11 Regression] bootstrap fail for powerpc-darwin while building libgfortran after r11-4485

[Bug tree-optimization/97666] New: [11 Regression] bootstrap fail for powerpc-darwin while building libgfortran after r11-4485

[iains at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=97666

            Bug ID: 97666
           Summary: [11 Regression] bootstrap fail for powerpc-darwin
                    while building libgfortran after r11-4485
           Product: gcc
           Version: 11.0
            Status: UNCONFIRMED
          Severity: normal
          Priority: P3
         Component: tree-optimization
          Assignee: unassigned at gcc dot gnu.org
          Reporter: iains at gcc dot gnu.org
  Target Milestone: ---

Created attachment 49483
  --> https://gcc.gnu.org/bugzilla/attachment.cgi?id=49483&action=edit
preprocessed generated source example for libgfortran.

unfortunately, although this does fail for a stage#1 on powerpc-darwin, it does
not reproduce on a cross from x86_64.

Program received signal SIGBUS, Bus error.

0x0192cc68 in vect_is_simple_use (vinfo=0x441499e0, stmt=0x4413c8c0,
slp_node=0x2c6c058 <pch_address_space+3489880>, operand=0, op=0xbfffdca8,
slp_def=0xbfffdce8, dt=0xbfffdcb8, vectype=0xbfffdcb4, def_stmt_info_out=0x0)
    at /src-local/gcc-master-asi/gcc/tree-vect-stmts.c:11344
11344         *vectype = SLP_TREE_VECTYPE (child);
(gdb) bt
#0  0x0192cc68 in vect_is_simple_use (vinfo=0x441499e0, stmt=0x4413c8c0,
slp_node=0x2c6c058 <pch_address_space+3489880>, operand=0, op=0xbfffdca8,
slp_def=0xbfffdce8, dt=0xbfffdcb8, vectype=0xbfffdcb4, def_stmt_info_out=0x0)
    at /src-local/gcc-master-asi/gcc/tree-vect-stmts.c:11344
#1  0x01917cb8 in vectorizable_operation (vinfo=0x441499e0,
stmt_info=0x4413c8c0, gsi=0x0, vec_stmt=0x0, slp_node=0x2c6c058
<pch_address_space+3489880>, cost_vec=0xbfffe1b4) at
/src-local/gcc-master-asi/gcc/tree-vect-stmts.c:5776
#2  0x01929dd0 in vect_analyze_stmt (vinfo=0x441499e0, stmt_info=0x4413c8c0,
need_to_vectorize=0xbfffdf64, node=0x2c6c058 <pch_address_space+3489880>,
node_instance=0x4413cfd0, cost_vec=0xbfffe1b4)
    at /src-local/gcc-master-asi/gcc/tree-vect-stmts.c:10686
#3  0x0197eb00 in vect_slp_analyze_node_operations_1 (vinfo=0x441499e0,
node=0x2c6c058 <pch_address_space+3489880>, node_instance=0x4413cfd0,
cost_vec=0xbfffe1b4) at /src-local/gcc-master-asi/gcc/tree-vect-slp.c:3385
#4  0x0197f194 in vect_slp_analyze_node_operations (vinfo=0x441499e0,
node=0x2c6c058 <pch_address_space+3489880>, node_instance=0x4413cfd0,
visited_set=..., visited_vec=..., cost_vec=0xbfffe1b4)
    at /src-local/gcc-master-asi/gcc/tree-vect-slp.c:3532
#5  0x0197f148 in vect_slp_analyze_node_operations (vinfo=0x441499e0,
node=0x2c6c00c <pch_address_space+3489804>, node_instance=0x4413cfd0,
visited_set=..., visited_vec=..., cost_vec=0xbfffe1b4)
    at /src-local/gcc-master-asi/gcc/tree-vect-slp.c:3525
#6  0x0197fa30 in vect_slp_analyze_operations (vinfo=0x441499e0) at
/src-local/gcc-master-asi/gcc/tree-vect-slp.c:3722
#7  0x019428a4 in vect_analyze_loop_2 (loop_vinfo=0x441499e0,
fatal=@0xbfffe934: false, n_stmts=0xbfffe90c) at
/src-local/gcc-master-asi/gcc/tree-vect-loop.c:2338
#8  0x01944294 in vect_analyze_loop (loop=0x29be2d8 <pch_address_space+680664>,
shared=0xbfffea88) at /src-local/gcc-master-asi/gcc/tree-vect-loop.c:2799
#9  0x01999e10 in try_vectorize_loop_1 (simduid_to_vf_htab=@0xbfffebc8: 0x0,
num_vectorized_loops=0xbfffebc0, loop=0x29be2d8 <pch_address_space+680664>,
loop_vectorized_call=0x0, loop_dist_alias_call=0x0)
    at /src-local/gcc-master-asi/gcc/tree-vectorizer.c:995
#10 0x0199a5ec in try_vectorize_loop (simduid_to_vf_htab=@0xbfffebc8: 0x0,
num_vectorized_loops=0xbfffebc0, loop=0x29be2d8 <pch_address_space+680664>) at
/src-local/gcc-master-asi/gcc/tree-vectorizer.c:1148
#11 0x0199a830 in vectorize_loops () at
/src-local/gcc-master-asi/gcc/tree-vectorizer.c:1230
#12 0x01765f74 in pass_vectorize::execute (this=0x44109b70, fun=0x4473c270) at
/src-local/gcc-master-asi/gcc/tree-ssa-loop.c:414
#13 0x012ad464 in execute_one_pass (pass=0x44109b70) at
/src-local/gcc-master-asi/gcc/passes.c:2517
#14 0x012ad874 in execute_pass_list_1 (pass=0x44109b70) at
/src-local/gcc-master-asi/gcc/passes.c:2606
#15 0x012ad8b4 in execute_pass_list_1 (pass=0x44109630) at
/src-local/gcc-master-asi/gcc/passes.c:2607
#16 0x012ad8b4 in execute_pass_list_1 (pass=0x441089b0) at
/src-local/gcc-master-asi/gcc/passes.c:2607
#17 0x012ad944 in execute_pass_list (fn=0x4473c270, pass=0x44108830) at
/src-local/gcc-master-asi/gcc/passes.c:2617
#18 0x00a1eaa4 in cgraph_node::expand (this=<error reading variable: Could not
find the frame base for "cgraph_node::expand()".>) at
/src-local/gcc-master-asi/gcc/cgraphunit.c:1829
#19 0x00a1f39c in expand_all_functions () at
/src-local/gcc-master-asi/gcc/cgraphunit.c:1997
#20 0x00a2032c in symbol_table::compile (this=<error reading variable: Could
not find the frame base for "symbol_table::compile()".>) at
/src-local/gcc-master-asi/gcc/cgraphunit.c:2361
#21 0x00a2096c in symbol_table::finalize_compilation_unit (this=<error reading
variable: Could not find the frame base for
"symbol_table::finalize_compilation_unit()".>) at
/src-local/gcc-master-asi/gcc/cgraphunit.c:2542
#22 0x01488688 in compile_file () at /src-local/gcc-master-asi/gcc/toplev.c:485
#23 0x0148cff0 in do_compile () at /src-local/gcc-master-asi/gcc/toplev.c:2321
#24 0x0148d58c in toplev::main (this=<error reading variable: Could not find
the frame base for "toplev::main(int, char**)".>, argc=<error reading variable:
Could not find the frame base for "toplev::main(int, char**)".>,
    argv=<error reading variable: Could not find the frame base for
"toplev::main(int, char**)".>) at /src-local/gcc-master-asi/gcc/toplev.c:2460
#25 0x01c89534 in main (argc=37, argv=0xbffff1ec) at
/src-local/gcc-master-asi/gcc/main.c:39

====  I don't really know what I'm looking at here - so these are somewhat
random prints of values.

gdb) p *vectype
$1 = (tree) 0x0
(gdb) p child
$2 = (slp_tree) 0x0
(gdb) up
#1  0x01913cd4 in vectorizable_operation (vinfo=0x4403d9b0,
stmt_info=0x44041330, gsi=0x0, vec_stmt=0x0, slp_node=0x2c61058
<pch_address_space+3489880>, cost_vec=0xbfffe1a8) at
/src-local/gcc-master-asi/gcc/tree-vect-stmts.c:5800
5800      if (!vect_is_simple_use (vinfo, stmt_info, slp_node,

(gdb) p *vinfo
$4 = {kind = vec_info::loop, shared = 0xbfffea88, stmt_vec_infos = {m_vec =
0x44041420}, stmt_vec_info_ro = true, slp_instances = {<vec<_slp_instance*,
va_heap, vl_ptr>> = {m_vec = 0x44041bf0}, <No data fields>}, base_alignments =
{
    m_table = {m_entries = 0x4403d830, m_size = 13, m_n_elements = 5,
m_n_deleted = 0, m_searches = 10, m_collisions = 0, m_size_prime_index = 1,
m_ggc = false, m_sanitize_eq_and_hash = true,
      static m_gather_mem_stats = <optimized out>}}, grouped_stores =
{<vec<_stmt_vec_info*, va_heap, vl_ptr>> = {m_vec = 0x0}, <No data fields>},
target_cost_data = 0x44029280, used_vector_modes = {m_table = {m_entries =
0x4403d940,
      m_size = 13, m_n_elements = 1, m_n_deleted = 0, m_searches = 38,
m_collisions = 0, m_size_prime_index = 1, m_ggc = false, m_sanitize_eq_and_hash
= true, static m_gather_mem_stats = <optimized out>}}, vector_mode =
E_V16QImode}

(gdb) p *stmt_info
$5 = {type = undef_vec_info_type, live = false, in_pattern_p = false,
pattern_stmt_p = false, vectorizable = true, stmt = 0x299e9f4
<pch_address_space+596468>, vectype = 0x295c600 <pch_address_space+325120>,
vec_stmts = {m_vec = 0x0},
  dr_aux = {dr = 0x0, stmt = 0x0, misalignment = -2, target_alignment =
{<poly_int_pod<1u, unsigned long long>> = {coeffs = {0}}, <No data fields>},
base_misaligned = false, base_decl = 0x0, offset = 0x0}, dr_wrt_vec_loop = {
    base_address = 0x0, offset = 0x0, init = 0x0, step = 0x0, base_alignment =
0, base_misalignment = 0, offset_alignment = 0, step_alignment = 0},
loop_phi_evolution_base_unchanged = 0x0, loop_phi_evolution_part = 0x0,
  related_stmt = 0x0, pattern_def_seq = 0x0, simd_clone_info = {m_vec = 0x0},
def_type = vect_internal_def, slp_type = pure_slp, first_element = 0x0,
next_element = 0x0, size = 0, store_count = 0, gap = 0, min_neg_dist = 0,
  relevant = vect_used_only_live, gather_scatter_p = false, strided_p = false,
simd_lane_access_p = 0, memory_access_type = VMAT_INVARIANT,
induc_cond_initial_val = 0x0, reduc_epilogue_adjustment = 0x0, reduc_type =
TREE_CODE_REDUCTION,
  reduc_code = ERROR_MARK, reduc_fn = IFN_LAST, reduc_idx = 1, reduc_def = 0x0,
reduc_vectype_in = 0x0, reduc_vectype = 0x0, force_single_cycle = false,
is_reduc_info = false, min_output_precision = 8, min_input_precision = 0,
  operation_precision = 0, operation_sign = SIGNED, mask_precision = 0,
slp_vect_only_p = false}

(gdb) p *slp_node
$7 = {children = {m_vec = 0x44028840}, stmts = {m_vec = 0x44041bd0}, ops =
{m_vec = 0x0}, representative = 0x44041330, load_permutation = {m_vec = 0x0},
lane_permutation = {m_vec = 0x0}, vectype = 0x295c600
<pch_address_space+325120>,
  vec_stmts = {m_vec = 0x0}, vec_defs = {m_vec = 0x0}, vec_stmts_size = 4,
refcnt = 1, max_nunits = {<poly_int_pod<1u, unsigned long long>> = {coeffs =
{16}}, <No data fields>}, def_type = vect_internal_def, lanes = 4, code =
ERROR_MARK,
  vertex = 1}

[Bug tree-optimization/97666] [11 Regression] bootstrap fail for powerpc-darwin while building libgfortran after r11-4485

[iains at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=97666

Iain Sandoe <iains at gcc dot gnu.org> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
           Keywords|                            |ice-on-valid-code
             Target|                            |powerpc-apple-darwin9
                 CC|                            |rguenth at gcc dot gnu.org
             Status|UNCONFIRMED                 |NEW
     Ever confirmed|0                           |1
               Host|                            |powerpc-apple-darwin9
   Last reconfirmed|                            |2020-11-01

[Bug tree-optimization/97666] [11 Regression] bootstrap fail for powerpc-darwin while building libgfortran after r11-4485

[rguenth at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=97666

Richard Biener <rguenth at gcc dot gnu.org> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
   Target Milestone|---                         |11.0

--- Comment #1 from Richard Biener <rguenth at gcc dot gnu.org> ---
Hmm, in vectorizable_operation () can you try

(gdb) p debug (slp_node)

?

If r11-4485 is the correct bisect point -fdump-tree-slp-details before/after
this rev might also provide a clue ...

[Bug tree-optimization/97666] [11 Regression] bootstrap fail for powerpc-darwin while building libgfortran after r11-4485

[cvs-commit at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=97666

--- Comment #2 from CVS Commits <cvs-commit at gcc dot gnu.org> ---
The master branch has been updated by Richard Biener <rguenth@gcc.gnu.org>:

https://gcc.gnu.org/g:9d1b813d0f7c9a8d80b0aee6eb1418b0afdf0f84

commit r11-4653-g9d1b813d0f7c9a8d80b0aee6eb1418b0afdf0f84
Author: Richard Biener <rguenther@suse.de>
Date:   Tue Nov 3 12:06:19 2020 +0100

    bootstrap/97666 - fix array of bool allocation

    This fixes the bad assumption that sizeof (bool) == 1

    2020-11-03  Richard Biener  <rguenther@suse.de>

            PR bootstrap/97666
            * tree-vect-slp.c (vect_build_slp_tree_2): Scale
            allocation of skip_args by sizeof (bool).

[Bug tree-optimization/97666] [11 Regression] bootstrap fail for powerpc-darwin while building libgfortran after r11-4485

[rguenth at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=97666

Richard Biener <rguenth at gcc dot gnu.org> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|NEW                         |RESOLVED
         Resolution|---                         |FIXED

--- Comment #3 from Richard Biener <rguenth at gcc dot gnu.org> ---
Fixed I think.

[Bug tree-optimization/97666] [11 Regression] bootstrap fail for powerpc-darwin while building libgfortran after r11-4485

[cvs-commit at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=97666

--- Comment #4 from CVS Commits <cvs-commit at gcc dot gnu.org> ---
The master branch has been updated by Richard Biener <rguenth@gcc.gnu.org>:

https://gcc.gnu.org/g:8d887367a7a3bc2cf1ca926899567aeeba915e44

commit r11-4708-g8d887367a7a3bc2cf1ca926899567aeeba915e44
Author: Richard Biener <rguenther@suse.de>
Date:   Wed Nov 4 09:36:05 2020 +0100

    bootstrap/97666 - really fix sizeof (bool) issue

    Pastoed the previous fix too quickly, the following fixes the
    correct spot - the memset, not the allocation.

    2020-11-04  Richard Biener  <rguenther@suse.de>

            PR bootstrap/97666
            * tree-vect-slp.c (vect_build_slp_tree_2): Revert previous
            fix and instead adjust the memset.