[SPARC] Errata workaround for GRLIB-TN-0013

This patch provides a workaround for the errata described in GRLIB-TN-0013.

If the workaround is enabled it will:

* Prevent div and sqrt instructions in the delay slot.

* Insert NOPs to prevent the sequence (div/sqrt) -> (two or three floating
  point operations or loads) -> (div/sqrt).

* Not insert NOPs if any of the floating point operations have a dependency
  on the destination register of the first (div/sqrt).

* Not insert NOPs if one of the floating point operations is a (div/sqrt).

* Insert NOPs to prevent (div/sqrt) followed by a branch.

It is applicable to GR712RC, UT700, and UT699.

2017-11-29  Daniel Cederman  <cederman@gaisler.com>

gcc/
* config/sparc/sparc.c (fpop_reg_depend_p): New function.
(div_sqrt_insn_p): New function.
(sparc_do_work_around_errata): Insert NOP instructions to
prevent sequences that could trigger the TN-0013 errata for
certain LEON3 processors.
(pass_work_around_errata::gate): Also test sparc_fix_lost_divsqrt.
(sparc_option_override): Set sparc_fix_lost_divsqrt appropriately.
* config/sparc/sparc.md (fix_lost_divsqrt): New attribute.
(in_branch_delay): Prevent div and sqrt in delay slot if
fix_lost_divsqrt.
* config/sparc/sparc.opt (sparc_fix_lost_divsqrt): New variable.

From-SVN: r255237

[SPARC] Errata workaround for GRLIB-TN-0010

This patch provides a workaround for the errata described in GRLIB-TN-0010.

If the workaround is enabled it will:

 * Insert a NOP between load instruction and atomic
   instruction (swap, ldstub, casa).

 * Insert a NOP at branch target if load in delay slot
   and atomic instruction at branch target.

It is applicable to UT700.

2017-11-29  Daniel Cederman  <cederman@gaisler.com>

gcc/
* config/sparc/sparc.c (atomic_insn_p): New function.
(sparc_do_work_around_errata): Insert NOP instructions to
prevent sequences that could trigger the TN-0010 errata for
UT700.
* config/sparc/sync.md (atomic_compare_and_swap_leon3_1): Make
instruction referable in atomic_insns_p.

From-SVN: r255236

[SPARC] Errata workaround for GRLIB-TN-0011

This patch provides a workaround for the errata described in GRLIB-TN-0011.

If the workaround is enabled it will:

 * Insert .align 16 before atomic instructions (swap, ldstub, casa).

It is applicable to GR712RC.

2017-11-29  Daniel Cederman  <cederman@gaisler.com>

gcc/
* config/sparc/sync.md (swapsi): 16-byte align if sparc_fix_gr712rc.
(atomic_compare_and_swap_leon3_1): Likewise.
(ldstub): Likewise.

From-SVN: r255235

[SPARC] Errata workaround for GRLIB-TN-0012

This patch provides a workaround for the errata described in GRLIB-TN-0012.

If the workaround is enabled it will:

 * Prevent any floating-point operation from being placed in the
   delay slot of an annulled integer branch.

 * Place a NOP at the branch target of an integer branch if it is
   a floating-point operation or a floating-point branch.

It is applicable to GR712RC.

2017-11-29  Daniel Cederman  <cederman@gaisler.com>

gcc/
* config/sparc/sparc.c (fpop_insn_p): New function.
(sparc_do_work_around_errata): Insert NOP instructions to
prevent sequences that could trigger the TN-0012 errata for
GR712RC.
(pass_work_around_errata::gate): Also test sparc_fix_gr712rc.
* config/sparc/sparc.md (fix_gr712rc): New attribute.
(in_branch_annul_delay): Prevent floating-point instructions
in delay slot of annulled integer branch.

From-SVN: r255234

re PR tree-optimization/83202 (Try joining operations on consecutive array elements during tree vectorization)

2017-11-29  Richard Biener  <rguenther@suse.de>

PR tree-optimization/83202
* tree-vect-slp.c (scalar_stmts_set_t): New typedef.
(bst_fail): Use it.
(vect_analyze_slp_cost_1): Add visited set, do not account SLP
nodes vectorized to the same stmts multiple times.
(vect_analyze_slp_cost): Allocate a visited set and pass it down.
(vect_analyze_slp_instance): Adjust.
(scalar_stmts_to_slp_tree_map_t): New typedef.
(vect_schedule_slp_instance): Add a map recording the SLP node
representing the vectorized stmts for a set of scalar stmts.
Avoid code-generating redundancies.
(vect_schedule_slp): Allocate map and pass it down.

* gcc.dg/vect/costmodel/x86_64/costmodel-pr83202.c: New testcase.

From-SVN: r255233

Fix PR number 83817->83187

From-SVN: r255232

[PATCH] complex type canonicalization

https://gcc.gnu.org/ml/gcc-patches/2017-11/msg02453.html
PR c++/83817
* tree.c (build_complex_type): Fix canonicalization.  Only fill in
type if it is new.

PR c++/83187
* g++.dg/opt/pr83187.C: New.

From-SVN: r255231

[AArch64] Fix ICE due to store_pair_lanes

The recently added store_pair_lanes causes ICEs in output_operand.
This is due to aarch64_classify_address treating it like a 128-bit STR
rather than a STP.  The valid immediate offsets don't fully overlap,
causing it to return false.  Eg. offset 264 is a valid 8-byte STP offset
but not a valid 16-byte STR offset since it isn't a multiple of 16.

The original instruction isn't passed in the printing code, so the context
is unclear.  The solution is to add a new operand formatting specifier
which is used for LDP/STP instructions like this.  This, like the Uml
constraint that applies to store_pair_lanes, uses PARALLEL when calling
aarch64_classify_address so that it knows it is an STP.
Also add the 'z' specifier for future use by load/store pair instructions.

    gcc/
* config/aarch64/aarch64.c (aarch64_print_operand): Add new
cases for printing LDP/STP memory addresses.
(aarch64_print_address_internal): Renamed from
aarch64_print_operand_address, added parameter, add Pmode check.
(aarch64_print_ldpstp_address): New function for LDP/STP addresses.
(aarch64_print_operand_address): Indirect to
aarch64_print_address_internal.
* config/aarch64/aarch64-simd.md (store_pair_lanes): Use new
'y' operand output specifier.

From-SVN: r255230

re PR middle-end/83185 (ICE with -fsanitize=address in build_simple_mem_ref_loc, at tree.c:4696)

PR middle-end/83185
* tree.c (build_simple_mem_ref_loc): Handle
get_addr_base_and_unit_offset returning a MEM_REF.

* gcc.dg/asan/pr83185.c: New test.

From-SVN: r255229

re PR middle-end/80929 (Division with constant no more optimized to mult highpart)

PR middle-end/80929
* rtlanal.c (seq_cost): For non-single_set insns try to use insn_cost.

From-SVN: r255228

fix merge conflict in libgfortran/ChangeLog

From-SVN: r255227

re PR target/80819 (Useless store to the stack  in _mm_set_epi64x with SSE4 -mno-avx)

PR target/80819
* config/i386/sse.md (vec_concatv2di): Remove * from (=Yr,0,*rm)
alternative.

* gcc.target/i386/pr80819-1.c: New test.
* gcc.target/i386/pr80819-2.c: New test.

From-SVN: r255226

re PR libfortran/83168 (FAIL: gfortran.dg/fmt_f0_2.f90 with a sanitized libgfortran)

2017-11-28  Jerry DeLisle  <jvdelisle@gcc.gnu.org>

PR libgfortran/83168
* io/write.c (select_string): Bump size by one to avoid
overrun.

From-SVN: r255225

Daily bump.

From-SVN: r255224

Fix some incorrectly formatted ChangeLog entries.

From-SVN: r255220

Handle very long lines when printing fix-it hints

When fixing PR c/82050 I noticed a bug in how we print fix-it hints
for very long lines: we weren't taking into account the x-offset for
the line when printing the fix-it hint.

This could lead to output where instead of printing:

foo.c:14:3944: error: etc
  = foo.field
        ^~~~~
        replacement

where the lines have been offset to start printing at about column 3900,
the "replacement" line was erroneously *not* offset, and was thus
prefixed by thousands of spaces, leading to large whitespace gaps in
the output, and the replacement failing to line up with the source to be
replaced.

Fixed thusly.

gcc/ChangeLog:
* diagnostic-show-locus.c (layout::print_trailing_fixits): Handle
m_x_offset.
(layout::move_to_column): Likewise.

gcc/testsuite/ChangeLog:
* gcc.dg/plugin/diagnostic-test-show-locus-bw.c
(test_very_wide_line): Update expected output to include a
fix-it hint.
* gcc.dg/plugin/diagnostic-test-show-locus-color.c
(test_very_wide_line): Likewise.
* gcc.dg/plugin/diagnostic_plugin_test_show_locus.c
(test_show_locus): Add a fix-it hint to "test_very_wide_line".

From-SVN: r255219

re PR sanitizer/81275 (-fsanitize=thread produce incorrect -Wreturn-type warning)

PR sanitizer/81275
* cp-tree.h (SWITCH_STMT_ALL_CASES_P): Define.
(SWITCH_STMT_NO_BREAK_P): Define.
(note_break_stmt, note_iteration_stmt_body_start,
note_iteration_stmt_body_end): Declare.
* decl.c (struct cp_switch): Add has_default_p, break_stmt_seen_p
and in_loop_body_p fields.
(push_switch): Clear them.
(pop_switch): Set SWITCH_STMT_CANNOT_FALLTHRU_P if has_default_p
and !break_stmt_seen_p.  Assert in_loop_body_p is false.
(note_break_stmt, note_iteration_stmt_body_start,
note_iteration_stmt_body_end): New functions.
(finish_case_label): Set has_default_p when both low and high
are NULL_TREE.
* parser.c (cp_parser_iteration_statement): Use
note_iteration_stmt_body_start and note_iteration_stmt_body_end
around parsing iteration body.
* pt.c (tsubst_expr): Likewise.
* cp-objcp-common.c (cxx_block_may_fallthru): Return false for
SWITCH_STMT which contains no BREAK_STMTs, contains a default:
CASE_LABEL_EXPR and where SWITCH_STMT_BODY isn't empty and
can't fallthru.
* semantics.c (finish_break_stmt): Call note_break_stmt.
* cp-gimplify.c (genericize_switch_stmt): Copy SWITCH_STMT_ALL_CASES_P
bit to SWITCH_ALL_CASES_P.  Assert that if SWITCH_STMT_NO_BREAK_P then
the break label is not TREE_USED.

* g++.dg/warn/pr81275-1.C: New test.
* g++.dg/warn/pr81275-2.C: New test.
* g++.dg/warn/pr81275-3.C: New test.
* c-c++-common/tsan/pr81275.c: Skip for C++ and -O2.

From-SVN: r255218

re PR sanitizer/81275 (-fsanitize=thread produce incorrect -Wreturn-type warning)

PR sanitizer/81275
* tree.c (block_may_fallthru): Return false if SWITCH_ALL_CASES_P
is set on SWITCH_EXPR and !block_may_fallthru (SWITCH_BODY ()).
c/
* c-typeck.c (c_finish_case): Set SWITCH_ALL_CASES_P if
c_switch_covers_all_cases_p returns true.
c-family/
* c-common.c (c_switch_covers_all_cases_p_1,
c_switch_covers_all_cases_p): New functions.
* c-common.h (c_switch_covers_all_cases_p): Declare.
testsuite/
* c-c++-common/tsan/pr81275.c: New test.

From-SVN: r255217

PR 53796 Improve INQUIRE(RECL=...) handling

The current F2018 draft (N2137) specifies behavior of the RECL=
specifier in the INQUIRE statement, where it previously was left as
undefined. Namely:

- If the unit is not connected, RECL= should be given the value -1.
- If the unit is connected with stream access, RECL= should be given
  the value -2.

Further, as PR 53796 describes, the handling of RECL= is poor in other
ways as well. When the recl is set to the maximum possible
(GFC_INTEGER_8_HUGE / LLONG_MAX), which it does by default except for
preconnected units, and when INQUIRE(RECL=) is used with a 4 byte
integer, the value is truncated and the 4 byte value is thus
-1. Fixing this to generate an error is a lot of work, as currently
the truncation is done by the frontend, the library sees only an 8
byte value with no indication that the frontend is going to copy it to
a 4 byte one. Instead, this patch does a bit twiddling trick such that
the truncated 4 byte value is GFC_INTEGER_4_HUGE while still being
0.99999999 * GFC_INTEGER_8_HUGE which is large enough for all
practical purposes.

Finally, the patch removes GFORTRAN_DEFAULT_RECL which was used only
for preconnected units, and instead uses the same approach as describe
above.

Regtested on x86_64-pc-linux-gnu, Ok for trunk.

gcc/fortran/ChangeLog:

2017-11-28  Janne Blomqvist  <jb@gcc.gnu.org>

PR fortran/53796
* gfortran.texi: Remove mentions of GFORTRAN_DEFAULT_RECL.

libgfortran/ChangeLog:

2017-11-28  Janne Blomqvist  <jb@gcc.gnu.org>

PR fortran/53796
* io/inquire.c (inquire_via_unit): Set recl to -1 for unconnected
units.
* io/io.h (default_recl): New variable.
* io/open.c (new_unit): Set recl to default_recl for sequential,
-2 for stream access.
* io/transfer.c (read_block_form): Test against default_recl
instead of DEFAULT_RECL.
(write_block): Likewise.
* io/unit.c (init_units): Calculate max_offset, default_recl.
* libgfortran.h (DEFAULT_RECL): Remove.
* runtime/environ.c: Remove GFORTRAN_DEFAULT_RECL.

gcc/testsuite/ChangeLog:

2017-11-28  Janne Blomqvist  <jb@gcc.gnu.org>

PR fortran/53796
* gfortran.dg/inquire_recl_f2018.f90: New test.

From-SVN: r255215

Reject fix-it hints for various awkward boundary cases (PR c/82050)

PR c/82050 reports a failed assertion deep within diagnostic_show_locus's
code for printing fix-it hints.

The root cause is a fix-it hint suggesting a textual replacement,
where the affected column numbers straddle the LINE_MAP_MAX_COLUMN_NUMBER
boundary, so that the start of the range has a column number, but the
end of the range doesn't.

The fix is to verify that the column numbers are sane when adding fix-it
hints to a rich_location, rejecting fix-it hints where they are not.

libcpp/ChangeLog:
PR c/82050
* include/line-map.h (LINE_MAP_MAX_COLUMN_NUMBER): Move here.
* line-map.c (LINE_MAP_MAX_COLUMN_NUMBER): ...from here.
(rich_location::maybe_add_fixit): Reject fix-it hints in which
the start column exceeds the next column.

From-SVN: r255214

[PR 82808] Use proper result types for arithmetic jump functions

2017-11-28  Prathamesh Kulkarni  <prathamesh.kulkarni@linaro.org>
    Martin Jambor  <mjambor@suse.cz>

PR ipa/82808
* tree.h (expr_type_first_operand_type_p): Declare
* tree.c (expr_type_first_operand_type_p): New function.
* ipa-prop.h (ipa_get_type): Allow i to be out of bounds.
(ipa_value_from_jfunc): Adjust declaration.
* ipa-cp.c (ipa_get_jf_pass_through_result): New parameter RES_TYPE.
Use it as result type for arithmetics, unless it is NULL in which case
be more conservative.
(ipa_value_from_jfunc): New parameter PARM_TYPE, pass it to
ipa_get_jf_pass_through_result.
(propagate_vals_across_pass_through): Likewise.
(propagate_scalar_across_jump_function): New parameter PARM_TYPE, pass
is to propagate_vals_across_pass_through.
(propagate_constants_across_call): Pass PARM_TYPE to
propagate_scalar_across_jump_function.
(find_more_scalar_values_for_callers_subset): Pass parameter type to
ipa_value_from_jfunc.
(cgraph_edge_brings_all_scalars_for_node): Likewise.
* ipa-fnsummary.c (evaluate_properties_for_edge): Renamed parms_info
to caller_parms_info, pass parameter type to ipa_value_from_jfunc.
* ipa-prop.c (try_make_edge_direct_simple_call): New parameter
target_type, pass it to ipa_value_from_jfunc.
(update_indirect_edges_after_inlining): Pass parameter type to
try_make_edge_direct_simple_call.

testsuite/
* gcc.dg/ipa/pr82808.c: New test.

Co-Authored-By: Martin Jambor <mjambor@suse.cz>
From-SVN: r255212

* gimple-ssa-evrp-analyze.c
(evrp_range_analyzer::record_ranges_from_phis): Only use SCEV to
refine ranges if scev_initialized_p returns true.
* vr-values.c (vr_values::extract_range_from_phi_node): Likewise.

From-SVN: r255211

* g++.dg/cpp0x/fntmpdefarg2a.C: Move expected warning.

From-SVN: r255210

Enable VBMI2 support [5/7]

gcc/
* config/i386/avx512vbmi2intrin.h (_mm512_shrdi_epi16,
_mm512_mask_shrdi_epi16, _mm512_maskz_shrdi_epi16, _mm512_shrdi_epi32,
_mm512_mask_shrdi_epi32, _mm512_maskz_shrdi_epi32, _mm512_shrdi_epi64,
_mm512_mask_shrdi_epi64, _mm512_maskz_shrdi_epi64): New intrinsics.
* config/i386/avx512vbmi2vlintrin.h (_mm256_shrdi_epi16,
_mm256_mask_shrdi_epi16, _mm256_maskz_shrdi_epi16,
_mm256_mask_shrdi_epi32, _mm256_maskz_shrdi_epi32, _mm256_shrdi_epi32,
_mm256_mask_shrdi_epi64, _mm256_maskz_shrdi_epi64, _mm256_shrdi_epi64,
_mm_mask_shrdi_epi16, _mm_maskz_shrdi_epi16, _mm_shrdi_epi16,
_mm_mask_shrdi_epi32, _mm_maskz_shrdi_epi32, _mm_shrdi_epi32,
_mm_mask_shrdi_epi64, _mm_maskz_shrdi_epi64, _mm_shrdi_epi64): Ditto.
* config/i386/i386-builtin.def (__builtin_ia32_vpshrd_v32hi,
__builtin_ia32_vpshrd_v32hi_mask, __builtin_ia32_vpshrd_v16hi,
__builtin_ia32_vpshrd_v16hi_mask, __builtin_ia32_vpshrd_v8hi,
__builtin_ia32_vpshrd_v8hi_mask, __builtin_ia32_vpshrd_v16si,
__builtin_ia32_vpshrd_v16si_mask, __builtin_ia32_vpshrd_v8si,
__builtin_ia32_vpshrd_v8si_mask, __builtin_ia32_vpshrd_v4si,
__builtin_ia32_vpshrd_v4si_mask, __builtin_ia32_vpshrd_v8di,
__builtin_ia32_vpshrd_v8di_mask, __builtin_ia32_vpshrd_v4di,
__builtin_ia32_vpshrd_v4di_mask, __builtin_ia32_vpshrd_v2di,
__builtin_ia32_vpshrd_v2di_mask): New builtins.
* config/i386/sse.md (vpshrd_<mode><mask_name>): New pattern.

gcc/testsuite/
* gcc.target/i386/avx-1.c: Handle new intrinsics.
* gcc.target/i386/sse-13.c: Ditto.
* gcc.target/i386/sse-23.c: Ditto.
* gcc.target/i386/avx512f-vpshrdd-2.c: New.
* gcc.target/i386/avx512f-vpshrdq-2.c: Ditto.
* gcc.target/i386/avx512vl-vpshrd-1.c: Ditto.
* gcc.target/i386/avx512vl-vpshrdd-2.c: Ditto.
* gcc.target/i386/avx512vl-vpshrdq-2.c: Ditto.

From-SVN: r255209

Enable VBMI2 support [4/7]

gcc/
config/i386/avx512vbmi2intrin.h (_mm512_shldi_epi16,
_mm512_mask_shldi_epi16, _mm512_maskz_shldi_epi16, _mm512_shldi_epi32,
_mm512_mask_shldi_epi32, _mm512_maskz_shldi_epi32, _mm512_shldi_epi64,
_mm512_mask_shldi_epi64, _mm512_maskz_shldi_epi64): New intrinsics.
config/i386/avx512vbmi2vlintrin.h (_mm256_shldi_epi16,
_mm256_mask_shldi_epi16, _mm256_maskz_shldi_epi16,
_mm256_mask_shldi_epi32, _mm256_maskz_shldi_epi32, _mm256_shldi_epi32,
_mm256_mask_shldi_epi64, _mm256_maskz_shldi_epi64, _mm256_shldi_epi64,
_mm_mask_shldi_epi16, _mm_maskz_shldi_epi16, _mm_shldi_epi16,
_mm_mask_shldi_epi32, _mm_maskz_shldi_epi32, _mm_shldi_epi32,
_mm_mask_shldi_epi64, _mm_maskz_shldi_epi64, _mm_shldi_epi64): Ditto.
config/i386/i386-builtin-types.def (V32HI_FTYPE_V32HI_V32HI_INT,
V32HI_FTYPE_V32HI_V32HI_INT_V32HI_INT, V16SI_FTYPE_V16SI_V16SI_INT,
V16SI_FTYPE_V16SI_V16SI_INT_V16SI_INT,
V8DI_FTYPE_V8DI_V8DI_INT_V8DI_INT, V8SI_FTYPE_V8SI_V8SI_INT_V8SI_INT,
V16HI_FTYPE_V16HI_V16HI_INT_V16HI_INT,
V4DI_FTYPE_V4DI_V4DI_INT_V4DI_INT,
V8HI_FTYPE_V8HI_V8HI_INT_V8HI_INT,
V4SI_FTYPE_V4SI_V4SI_INT_V4SI_INT,
V2DI_FTYPE_V2DI_V2DI_INT_V2DI_INT): New types.
config/i386/i386-builtin.def (__builtin_ia32_vpshld_v32hi,
__builtin_ia32_vpshld_v32hi_mask, __builtin_ia32_vpshld_v16hi,
__builtin_ia32_vpshld_v16hi_mask, __builtin_ia32_vpshld_v8hi,
__builtin_ia32_vpshld_v8hi_mask, __builtin_ia32_vpshld_v16si,
__builtin_ia32_vpshld_v16si_mask, __builtin_ia32_vpshld_v8si,
__builtin_ia32_vpshld_v8si_mask, __builtin_ia32_vpshld_v4si,
__builtin_ia32_vpshld_v4si_mask, __builtin_ia32_vpshld_v8di,
__builtin_ia32_vpshld_v8di_mask, __builtin_ia32_vpshld_v4di,
__builtin_ia32_vpshld_v4di_mask, __builtin_ia32_vpshld_v2di,
__builtin_ia32_vpshld_v2di_mask): New builtins.
config/i386/i386.c (ix86_expand_args_builtin): Handle new types.
config/i386/sse.md (vpshld_<mode><mask_name>): New pattern.

gcc/testsuite/
gcc.target/i386/avx-1.c: Handle new intrinics.
gcc.target/i386/sse-13.c: Ditto.
gcc.target/i386/sse-23.c: Ditto.
gcc.target/i386/avx512f-vpshld-1.c: New test.
gcc.target/i386/avx512f-vpshldd-2.c: Ditto.
gcc.target/i386/avx512f-vpshldq-2.c: Ditto.
gcc.target/i386/avx512vl-vpshld-1.c: Ditto.
gcc.target/i386/avx512vl-vpshldd-2.c: Ditto.
gcc.target/i386/avx512vl-vpshldq-2.c: Ditto.

From-SVN: r255208

re PR fortran/83021 (gfortran segfault in polymorphic assignment)

2017-11-28  Paul Thomas  <pault@gcc.gnu.org>

PR fortran/83021
* resolve.c (resolve_component): Only escape for use assciated
vtypes if the current namespace has no proc_name and is most
particularly block data.

From-SVN: r255202

re PR tree-optimization/80776 (-Wformat-overflow false positive for %d on integer bounded by __builtin_unreachable)

2017-11-28  Richard Biener  <rguenther@suse.de>

PR tree-optimization/80776
* gimple-ssa-evrp-analyze.h (evrp_range_analyzer::set_ssa_range_info):
Declare.
* gimple-ssa-evrp-analyze.c (evrp_range_analyzer::set_ssa_range_info):
New function.
(evrp_range_analyzer::record_ranges_from_incoming_edges):
If the incoming edge is an effective fallthru because the other
edge only reaches a __builtin_unreachable () then record ranges
derived from the controlling condition in SSA info.
(evrp_range_analyzer::record_ranges_from_phis): Use set_ssa_range_info.
(evrp_range_analyzer::record_ranges_from_stmt): Likewise.

* gcc.dg/pr80776-1.c: New testcase.
* gcc.dg/pr80776-2.c: Likewise.

From-SVN: r255201

Makefile.in (SELFTEST_FLAGS): Use nul instead of /dev/null on mingw build hosts.

2017-11-28  Olivier Hainque  <hainque@adacore.com>

* Makefile.in (SELFTEST_FLAGS): Use nul instead of /dev/null
on mingw build hosts.

From-SVN: r255200

Add std::to_address for C++2a (P0653R2)

2017-11-28  Glen Joseph Fernandes  <glenjofe@gmail.com>

* include/bits/ptr_traits.h (to_address): Implement to_address.
* testsuite/20_util/to_address/1.cc: New test.

From-SVN: r255199

re PR debug/81307 (g++.dg/debug/debug9.C -gstabs FAILs)

PR debug/81307
* dbxout.c (dbx_block_with_cold_children): Fix function comment.

From-SVN: r255198

re PR tree-optimization/83141 (SRA and memcpy folding interact badly generating wrong-code)

2017-11-28  Richard Biener  <rguenther@suse.de>

PR middle-end/83141
* gimple-fold.c (gimple_fold_builtin_memory_op): For aggregate
copies generated from memcpy use a character array as reference
type.

From-SVN: r255197

re PR c/83117 (FAIL: gcc.target/x86_64/abi/ms-sysv/ms-sysv.c (test for excess errors))

PR c/83117
* gcc.target/x86_64/abi/ms-sysv/gen.cc (make_do_tests_decl): Drop
const from do_test_{u,v}*.

From-SVN: r255196

Remove Cilk Plus support.

* Makefile.def (target_modules): Remove libcilkrts.
* Makefile.in: Ditto.
* configure: Ditto.
* configure.ac: Ditto.

contrib/
* contrib/gcc_update: Ditto.

gcc/
* Makefile.in (cilkplus.def, cilk-builtins.def, c-family/cilk.o,
c-family/c-cilkplus.o, c-family/array-notation-common.o,
cilk-common.o, cilk.h, cilk-common.c): Remove.
* builtin-types.def
(BT_FN_INT_PTR_PTR_PTR_FTYPE_BT_INT_BT_PTR_BT_PTR_BT_PTR): Remove.
* builtins.c (is_builtin_name): Remove cilkplus condition.
(BUILT_IN_CILK_DETACH, BUILT_IN_CILK_POP_FRAME): Remove.
* builtins.def (DEF_CILK_BUILTIN_STUB, DEF_CILKPLUS_BUILTIN,
cilk-builtins.def, cilkplus.def): Remove.
* cif-code.def (CILK_SPAWN): Remove.
* cilk-builtins.def: Delete.
* cilk-common.c: Ditto.
* cilk.h: Ditto.
* cilkplus.def: Ditto.
* config/darwin.h (fcilkplus): Delete.
* cppbuiltin.c: Ditto.
* doc/extend.texi: Remove cilkplus doc.
* doc/generic.texi: Ditto.
* doc/invoke.texi: Ditto.
* doc/passes.texi: Ditto.
* gcc.c (fcilkplus): Remove.
* gengtype.c (cilk.h): Remove.
* gimple-pretty-print.c (dump_gimple_omp_for): Remove cilkplus support.
* gimple.h (GF_OMP_FOR_KIND_CILKFOR, GF_OMP_FOR_KIND_CILKSIMD): Remove.
* gimplify.c (gimplify_return_expr, maybe_fold_stmt,
gimplify_call_expr, is_gimple_stmt, gimplify_modify_expr,
gimplify_scan_omp_clauses, gimplify_adjust_omp_clauses,
gimplify_omp_for, gimplify_expr): Remove cilkplus conditions.
* ipa-fnsummary.c (ipa_dump_fn_summary, compute_fn_summary,
inline_read_section): Ditto.
* ipa-inline-analysis.c (cilk.h): Remove.
* ira.c (ira_setup_eliminable_regset): Remove cilkplus support.
* lto-wrapper.c (merge_and_complain, append_compiler_options,
append_linker_options): Remove condition for fcilkplus.
* lto/lto-lang.c (cilk.h): Remove.
(lto_init): Remove condition for fcilkplus.
* omp-expand.c (expand_cilk_for_call): Delete.
(expand_omp_taskreg, expand_omp_for_static_chunk,
expand_omp_for): Remove cilkplus
conditions.
(expand_cilk_for): Delete.
* omp-general.c (omp_extract_for_data): Remove cilkplus support.
* omp-low.c (scan_sharing_clauses, create_omp_child_function,
execute_lower_omp, diagnose_sb_0): Ditto.
* omp-simd-clone.c (simd_clone_clauses_extract): Ditto.
* tree-core.h (OMP_CLAUSE__CILK_FOR_COUNT_): Delete.
* tree-nested.c: Ditto.
* tree-pretty-print.c (dump_omp_clause): Remove cilkplus support.
(dump_generic_node): Ditto.
* tree.c (OMP_CLAUSE__CILK_FOR_COUNT_): Delete.
* tree.def (cilk_simd, cilk_for, cilk_spawn_stmt,
cilk_sync_stmt): Delete.
* tree.h (CILK_SPAWN_FN, EXPR_CILK_SPAWN): Delete.

gcc/c-family/
* array-notation-common.c: Delete.
* c-cilkplus.c: Ditto.
* c-common.c (_Cilk_spawn, _Cilk_sync, _Cilk_for): Remove.
* c-common.def (ARRAY_NOTATION_REF): Remove.
* c-common.h (RID_CILK_SPAWN, build_array_notation_expr,
build_array_notation_ref, C_ORT_CILK, c_check_cilk_loop,
c_validate_cilk_plus_loop, cilkplus_an_parts,
cilk_ignorable_spawn_rhs_op,
cilk_recognize_spawn): Remove.
* c-gimplify.c (CILK_SPAWN_STMT): Remove.
* c-omp.c: Remove CILK_SIMD check.
* c-pragma.c: Ditto.
* c-pragma.h: Remove CILK related pragmas.
* c-pretty-print.c (c_pretty_printer::postfix_expression): Remove
ARRAY_NOTATION_REF condition.
(c_pretty_printer::expression): Ditto.
* c.opt (fcilkplus): Remove.
* cilk.c: Delete.

gcc/c/
* Make-lang.in (c/c-array-notation.o): Remove.
* c-array-notation.c: Delete.
* c-decl.c: Remove cilkplus condition.
* c-parser.c (c_parser_cilk_simd, c_parser_cilk_for,
c_parser_cilk_verify_simd, c_parser_array_notation,
c_parser_cilk_clause_vectorlength, c_parser_cilk_grainsize,
c_parser_cilk_simd_fn_vector_attrs,
c_finish_cilk_simd_fn_tokens): Delete.
(c_parser_declaration_or_fndef): Remove cilkplus condition.
(c_parser_direct_declarator_inner): Ditto.
(CILK_SIMD_FN_CLAUSE_MASK): Delete.
(c_parser_attributes, c_parser_compound_statement,
c_parser_statement_after_labels, c_parser_if_statement,
c_parser_switch_statement, c_parser_while_statement,
c_parser_do_statement, c_parser_for_statement,
c_parser_unary_expression, c_parser_postfix_expression,
c_parser_postfix_expression_after_primary,
c_parser_pragma, c_parser_omp_clause_name, c_parser_omp_all_clauses,
c_parser_omp_for_loop,
c_finish_omp_declare_simd): Remove cilkplus support.
* c-typeck.c (build_array_ref, build_function_call_vec,
convert_arguments, lvalue_p, build_compound_expr, c_finish_return,
c_finish_if_stmt, c_finish_loop,
build_binary_op): Remove cilkplus support.

gcc/cp/
* Make-lang.in (cp/cp-array-notation.o, cp/cp-cilkplus.o): Delete.
* call.c (convert_for_arg_passing, build_cxx_call): Remove cilkplus.
* constexpr.c (potential_constant_expression_1): Ditto.
* cp-array-notation.c: Delete.
* cp-cilkplus.c: Ditto.
* cp-cilkplus.h: Ditto.
* cp-gimplify.c (cp_gimplify_expr, cp_fold_r, cp_genericize): Remove
cilkplus condition.
* cp-objcp-common.c (ARRAY_NOTATION_REF): Delete.
* cp-tree.h (cilkplus_an_triplet_types_ok_p): Delete.
* decl.c (grokfndecl, finish_function): Remove cilkplus condition.
* error.c (dump_decl, dump_expr): Remove ARRAY_NOTATION_REF condition.
* lambda.c (cp-cilkplus.h): Remove.
* parser.c (cp_parser_cilk_simd, cp_parser_cilk_for,
cp_parser_cilk_simd_vectorlength): Delete.
(cp_debug_parser, cp_parser_ctor_initializer_opt_and_function_body,
cp_parser_postfix_expression, cp_parser_postfix_open_square_expression,
cp_parser_statement, cp_parser_jump_statement,
cp_parser_direct_declarator,
cp_parser_late_return_type_opt, cp_parser_gnu_attribute_list,
cp_parser_omp_clause_name, cp_parser_omp_clause_aligned,
cp_parser_omp_clause_linear, cp_parser_omp_all_clauses,
cp_parser_omp_flush, cp_parser_omp_for_cond, cp_parser_omp_for_incr,
cp_parser_omp_for_loop_init, cp_parser_omp_for_loop,
cp_parser_omp_declare_simd): Remove cilkplus support.
(CILK_SIMD_FN_CLAUSE_MASK, cp_parser_late_parsing_cilk_simd_fn_info,
cp_parser_cilk_grainsize): Remove.
(cp_parser_pragma, c_parse_file): Remove cilkplus support.
(cp_parser_cilk_simd_vectorlength, cp_parser_cilk_simd_linear,
cp_parser_cilk_simd_clause_name, cp_parser_cilk_simd_all_clauses,
cp_parser_cilk_simd, cp_parser_cilk_for): Remove.
* parser.h (IN_CILK_SIMD_FOR, IN_CILK_SPAWN): Remove.
* pt.c (tsubst_attribute, tsubst_expr, tsubst_copy_and_build): Remove
cilkplus support.
* semantics.c (finish_goto_stmt, begin_while_stmt, finish_do_body,
finish_init_stmt, finish_switch_cond, simplify_aggr_init_expr,
finish_omp_clauses, finish_omp_clauses,
finish_omp_for): Remove cilkplus support.
* tree.c (lvalue_kind): Remove ARRAY_NOTATION_REF conditon.
* typeck.c (cp_build_array_ref, cp_build_compound_expr,
check_return_expr): Remove cilkplus support.

gcc/testsuite/
* c-c++-common/attr-simd-3.c: Delete.
* c-c++-common/cilk-plus/AN/an-if.c: Delete.
* c-c++-common/cilk-plus/AN/array_test1.c: Delete.
* c-c++-common/cilk-plus/AN/array_test2.c: Delete.
* c-c++-common/cilk-plus/AN/array_test_ND.c: Delete.
* c-c++-common/cilk-plus/AN/builtin_fn_custom.c: Delete.
* c-c++-common/cilk-plus/AN/builtin_fn_mutating.c: Delete.
* c-c++-common/cilk-plus/AN/builtin_func_double.c: Delete.
* c-c++-common/cilk-plus/AN/builtin_func_double2.c: Delete.
* c-c++-common/cilk-plus/AN/comma_exp.c: Delete.
* c-c++-common/cilk-plus/AN/conditional.c: Delete.
* c-c++-common/cilk-plus/AN/decl-ptr-colon.c: Delete.
* c-c++-common/cilk-plus/AN/dimensionless-arrays.c: Delete.
* c-c++-common/cilk-plus/AN/exec-once.c: Delete.
* c-c++-common/cilk-plus/AN/exec-once2.c: Delete.
* c-c++-common/cilk-plus/AN/fn_ptr-2.c: Delete.
* c-c++-common/cilk-plus/AN/fn_ptr.c: Delete.
* c-c++-common/cilk-plus/AN/fp_triplet_values.c: Delete.
* c-c++-common/cilk-plus/AN/gather-scatter-errors.c: Delete.
* c-c++-common/cilk-plus/AN/gather_scatter.c: Delete.
* c-c++-common/cilk-plus/AN/if_test.c: Delete.
* c-c++-common/cilk-plus/AN/if_test_errors.c: Delete.
* c-c++-common/cilk-plus/AN/misc.c: Delete.
* c-c++-common/cilk-plus/AN/n-ptr-test.c: Delete.
* c-c++-common/cilk-plus/AN/parser_errors.c: Delete.
* c-c++-common/cilk-plus/AN/parser_errors2.c: Delete.
* c-c++-common/cilk-plus/AN/parser_errors3.c: Delete.
* c-c++-common/cilk-plus/AN/parser_errors4.c: Delete.
* c-c++-common/cilk-plus/AN/pr57457-2.c: Delete.
* c-c++-common/cilk-plus/AN/pr57457.c: Delete.
* c-c++-common/cilk-plus/AN/pr57490.c: Delete.
* c-c++-common/cilk-plus/AN/pr57541-2.c: Delete.
* c-c++-common/cilk-plus/AN/pr57541.c: Delete.
* c-c++-common/cilk-plus/AN/pr57577.c: Delete.
* c-c++-common/cilk-plus/AN/pr58942.c: Delete.
* c-c++-common/cilk-plus/AN/pr61191.c: Delete.
* c-c++-common/cilk-plus/AN/pr61455-2.c: Delete.
* c-c++-common/cilk-plus/AN/pr61455.c: Delete.
* c-c++-common/cilk-plus/AN/pr61962.c: Delete.
* c-c++-common/cilk-plus/AN/pr61963.c: Delete.
* c-c++-common/cilk-plus/AN/pr62008.c: Delete.
* c-c++-common/cilk-plus/AN/pr63884.c: Delete.
* c-c++-common/cilk-plus/AN/rank_mismatch.c: Delete.
* c-c++-common/cilk-plus/AN/rank_mismatch2.c: Delete.
* c-c++-common/cilk-plus/AN/rank_mismatch3.c: Delete.
* c-c++-common/cilk-plus/AN/sec_implicit.c: Delete.
* c-c++-common/cilk-plus/AN/sec_implicit2.c: Delete.
* c-c++-common/cilk-plus/AN/sec_implicit_ex.c: Delete.
* c-c++-common/cilk-plus/AN/sec_reduce_ind_same_value.c: Delete.
* c-c++-common/cilk-plus/AN/sec_reduce_max_min_ind.c: Delete.
* c-c++-common/cilk-plus/AN/sec_reduce_return.c: Delete.
* c-c++-common/cilk-plus/AN/side-effects-1.c: Delete.
* c-c++-common/cilk-plus/AN/test_builtin_return.c: Delete.
* c-c++-common/cilk-plus/AN/test_sec_limits.c: Delete.
* c-c++-common/cilk-plus/AN/tst_lngth.c: Delete.
* c-c++-common/cilk-plus/AN/vla.c: Delete.
* c-c++-common/cilk-plus/CK/Wparentheses-1.c: Delete.
* c-c++-common/cilk-plus/CK/cilk-for-2.c: Delete.
* c-c++-common/cilk-plus/CK/cilk-for-3.c: Delete.
* c-c++-common/cilk-plus/CK/cilk-fors.c: Delete.
* c-c++-common/cilk-plus/CK/cilk_for_errors.c: Delete.
* c-c++-common/cilk-plus/CK/cilk_for_grain.c: Delete.
* c-c++-common/cilk-plus/CK/cilk_for_grain_errors.c: Delete.
* c-c++-common/cilk-plus/CK/cilk_for_ptr_iter.c: Delete.
* c-c++-common/cilk-plus/CK/compound_cilk_spawn.c: Delete.
* c-c++-common/cilk-plus/CK/concec_cilk_spawn.c: Delete.
* c-c++-common/cilk-plus/CK/errors.c: Delete.
* c-c++-common/cilk-plus/CK/fib.c: Delete.
* c-c++-common/cilk-plus/CK/fib_init_expr_xy.c: Delete.
* c-c++-common/cilk-plus/CK/fib_no_return.c: Delete.
* c-c++-common/cilk-plus/CK/fib_no_sync.c: Delete.
* c-c++-common/cilk-plus/CK/invalid_spawns.c: Delete.
* c-c++-common/cilk-plus/CK/invalid_sync.c: Delete.c
* c-c++-common/cilk-plus/CK/nested_cilk_for.c: Delete.
* c-c++-common/cilk-plus/CK/no_args_error.c: Delete.
* c-c++-common/cilk-plus/CK/pr59631.c: Delete.
* c-c++-common/cilk-plus/CK/pr60197-2.c: Delete.
* c-c++-common/cilk-plus/CK/pr60197.c: Delete.
* c-c++-common/cilk-plus/CK/pr60469.c: Delete.
* c-c++-common/cilk-plus/CK/pr60586.c: Delete.
* c-c++-common/cilk-plus/CK/pr63307.c: Delete.
* c-c++-common/cilk-plus/CK/pr69826-1.c: Delete.
* c-c++-common/cilk-plus/CK/pr69826-2.c: Delete.
* c-c++-common/cilk-plus/CK/pr79428-4.c: Delete.
* c-c++-common/cilk-plus/CK/pr79428-7.c: Delete.
* c-c++-common/cilk-plus/CK/spawn_in_return.c: Delete.
* c-c++-common/cilk-plus/CK/spawnee_inline.c: Delete.
* c-c++-common/cilk-plus/CK/spawner_inline.c: Delete.
* c-c++-common/cilk-plus/CK/spawning_arg.c: Delete.
* c-c++-common/cilk-plus/CK/steal_check.c: Delete.
* c-c++-common/cilk-plus/CK/sync_wo_spawn.c: Delete.
* c-c++-common/cilk-plus/CK/test__cilk.c: Delete.
* c-c++-common/cilk-plus/CK/varargs_test.c: Delete.
* c-c++-common/cilk-plus/PS/Wparentheses-1.c: Delete.
* c-c++-common/cilk-plus/PS/body.c: Delete.
* c-c++-common/cilk-plus/PS/clauses1.c: Delete.
* c-c++-common/cilk-plus/PS/clauses2.c: Delete.
* c-c++-common/cilk-plus/PS/clauses3.c: Delete.
* c-c++-common/cilk-plus/PS/clauses4.c: Delete.
* c-c++-common/cilk-plus/PS/for1.c: Delete.
* c-c++-common/cilk-plus/PS/for2.c: Delete.
* c-c++-common/cilk-plus/PS/for3.c: Delete.
* c-c++-common/cilk-plus/PS/pr69363.c: Delete.
* c-c++-common/cilk-plus/PS/reduction-1.c: Delete.
* c-c++-common/cilk-plus/PS/reduction-2.c: Delete.
* c-c++-common/cilk-plus/PS/reduction-3.c: Delete.
* c-c++-common/cilk-plus/PS/run-1.c: Delete.
* c-c++-common/cilk-plus/PS/safelen.c: Delete.
* c-c++-common/cilk-plus/PS/vectorlength-2.c: Delete.
* c-c++-common/cilk-plus/PS/vectorlength-3.c: Delete.
* c-c++-common/cilk-plus/PS/vectorlength.c: Delete.
* c-c++-common/cilk-plus/SE/ef_error.c: Delete.
* c-c++-common/cilk-plus/SE/ef_error2.c: Delete.
* c-c++-common/cilk-plus/SE/ef_error3.c: Delete.
* c-c++-common/cilk-plus/SE/ef_test.c: Delete.
* c-c++-common/cilk-plus/SE/ef_test2.c: Delete.
* c-c++-common/cilk-plus/SE/vlength_errors.c: Delete.
* g++.dg/cilk-plus/AN/array_function.c: Delete.
* g++.dg/cilk-plus/AN/array_test1_tplt.c: Delete.
* g++.dg/cilk-plus/AN/array_test2_tplt.c: Delete.
* g++.dg/cilk-plus/AN/array_test_ND_tplt.c: Delete.
* g++.dg/cilk-plus/AN/braced_list.c: Delete.
* g++.dg/cilk-plus/AN/builtin_fn_custom_tplt.c: Delete.
* g++.dg/cilk-plus/AN/builtin_fn_mutating_tplt.c: Delete.
* g++.dg/cilk-plus/AN/fp_triplet_values_tplt.c: Delete.
* g++.dg/cilk-plus/AN/postincr_test.c: Delete.
* g++.dg/cilk-plus/AN/preincr_test.c: Delete.
* g++.dg/cilk-plus/CK/catch_exc.c: Delete.
* g++.dg/cilk-plus/CK/cf3.c: Delete.
* g++.dg/cilk-plus/CK/cilk-for-tplt.c: Delete.
* g++.dg/cilk-plus/CK/const_spawn.c: Delete.
* g++.dg/cilk-plus/CK/fib-opr-overload.c: Delete.
* g++.dg/cilk-plus/CK/fib-tplt.c: Delete.
* g++.dg/cilk-plus/CK/for1.c: Delete.
* g++.dg/cilk-plus/CK/lambda_spawns.c: Delete.
* g++.dg/cilk-plus/CK/lambda_spawns_tplt.c: Delete.
* g++.dg/cilk-plus/CK/pr60586.c: Delete.
* g++.dg/cilk-plus/CK/pr66326.c: Delete.
* g++.dg/cilk-plus/CK/pr68001.c: Delete.
* g++.dg/cilk-plus/CK/pr68997.c: Delete.
* g++.dg/cilk-plus/CK/pr69024.c: Delete.
* g++.dg/cilk-plus/CK/pr69048.c: Delete.
* g++.dg/cilk-plus/CK/pr69267.c: Delete.
* g++.dg/cilk-plus/CK/pr80038.c: Delete.
* g++.dg/cilk-plus/CK/stl_iter.c: Delete.
* g++.dg/cilk-plus/CK/stl_rev_iter.c: Delete.
* g++.dg/cilk-plus/CK/stl_test.c: Delete.
* g++.dg/cilk-plus/cilk-plus.exp
* g++.dg/cilk-plus/ef_test.C: Delete.
* g++.dg/cilk-plus/for.C: Delete.
* g++.dg/cilk-plus/for2.C: Delete.
* g++.dg/cilk-plus/for3.C: Delete.
* g++.dg/cilk-plus/for4.C: Delete.
* g++.dg/cilk-plus/pr60967.C: Delete.
* g++.dg/cilk-plus/pr69028.C: Delete.
* g++.dg/cilk-plus/pr70565.C: Delete.
* g++.dg/pr57662.C: Delete.
* gcc.dg/cilk-plus/cilk-plus.exp
* gcc.dg/cilk-plus/for1.c: Delete.
* gcc.dg/cilk-plus/for2.c: Delete.
* gcc.dg/cilk-plus/jump-openmp.c: Delete.
* gcc.dg/cilk-plus/jump.c: Delete.
* gcc.dg/cilk-plus/pr69798-1.c: Delete.
* gcc.dg/cilk-plus/pr69798-2.c: Delete.
* gcc.dg/cilk-plus/pr78306.c: Delete.
* gcc.dg/cilk-plus/pr79116.c: Delete.
* gcc.dg/graphite/id-28.c: Delete.
* lib/cilk-plus-dg.exp: Delete.
* lib/target-supports.exp (cilkplus_runtime): Delete.

Co-Authored-By: Sebastian Peryt <sebastian.peryt@intel.com>
From-SVN: r255195

[Patch AArch64] Fixup floating point division with -march=armv8-a+nosimd

The canonical examples is :

    double
    foo (double x, double y)
      {
        return x / y;
      }

    with -march=armv8-a+nosimd

generates a function that calls __divdf3. Ofcourse on AArch64 we don't
have any software floating point and this causes issues.

There is also a problem in +nosimd that has existed since the dawn of
time in the port with respect to long doubles (128 bit floating
point), here the ABI and the compiler expect the presence of the SIMD
unit as these parameters are passed in the vector registers. Thus
while +nosimd tries to prevent the use of SIMD instructions in the
compile we don't get this right as we end up using ldr qN / str qN
instructions and even there I think things go wrong in a simple
example that I tried.

Is that sufficient to consider marking +nosimd as deprecated in GCC-8
and remove this in a future release ?

That is not a subject for this patch but something separate but I
would like to put this into trunk and the release
branches. Bootstrapped and regression tested on my aarch64 desktop.

Ok ?

From-SVN: r255194

re PR rtl-optimization/81020 (wrong code with -O -fno-tree-bit-ccp -fno-tree-coalesce-vars -fno-tree-vrp)

PR rtl-optimization/81020
* gcc.dg/pr81020.c: New test.

From-SVN: r255193

tree.def (SWITCH_EXPR): Change from 3 operand to 2 operand tree.

* tree.def (SWITCH_EXPR): Change from 3 operand to 2 operand tree.
Adjust comment.
* tree.h (SWITCH_LABELS): Remove.
* gimplify.c (gimplify_switch_expr): Don't test SWITCH_LABELS,
assert SWITCH_BODY is non-NULL.
* tree-pretty-print.c (dump_generic_node): Remove SWITCH_LABELS
handling.
* tree.c (block_may_fallthru): Always return true; for SWITCH_EXPR.
c/
* c-typeck.c (c_start_case): Build SWITCH_EXPR using build2 instead
of build3.
cp/
* cp-gimplify.c (genericize_switch_stmt): Build SWITCH_EXPR using
build2_loc instead of build3_loc.
ada/
* gcc-interface/trans.c (Case_Statement_to_gnu): Build SWITCH_EXPR
using build2 instead of build3.
jit/
* jit-playback.c (add_switch): Build SWITCH_EXPR using build2 instead
of build3.  Formatting fixes.  Adjust funciton comment.
fortran/
* trans-decl.c (gfc_trans_entry_master_switch): Build SWITCH_EXPR
using fold_build2_loc instead of fold_build3_loc.
* trans-io.c (io_result): Likewise.
* trans-stmt.c (gfc_trans_integer_select, gfc_trans_character_select):
Likewise.
go/
* go-gcc.cc (Gcc_backend::switch_statement): Build SWITCH_EXPR using
build2_loc instead of build3_loc.
brig/
* brigfrontend/brig-branch-inst-handler.cc
(brig_branch_inst_handler::operator): Build SWITCH_EXPR using build2
instead of build3.

From-SVN: r255192

re PR tree-optimization/80788 (ICE in set_value_range, at tree-vrp.c:252)

PR tree-optimization/80788
* match.pd (X +- C1 CMP C2 -> X CMP C2 -+ C1): If res
has TREE_OVERFLOW set, call drop_tree_overflow.

* gcc.dg/pr80788.c: New test.

From-SVN: r255191

re PR target/83158 (gcc.target/i386/pr78057.c fail)

2017-11-28  Richard Biener  <rguenther@suse.de>

PR tree-optimization/83158
* tree-vrp.c (intersect_ranges): Prefer ~[0, 0] in a few more
cases.

From-SVN: r255190

* MAINTAINERS (write after approval): Add myself.

From-SVN: r255189

rs6000: Improve comparison rtx_cost (PR81288)

The current rs6000 rtx_cost for comparisons against 0 is very high if
TARGET_ISEL && !TARGET_MFCRF, much higher than for reg-reg comparisons,
much higher than a load of 0 and such a reg-reg-comparison.  This leads
to infinite recursion in CSE (see PR81288).

This patch removes the too-high cost, also simplifying this code.

PR 81288/target
* config/rs6000/rs6000.c (rs6000_rtx_costs): Do not handle
TARGET_ISEL && !TARGET_MFCRF differently.  Simplify code.

From-SVN: r255188

rs6000: Improve scc isel

If we have a negative condition we can use a literal 0 in the isel,
instead of having to load it into a register.  If the condition is from
a comparison with an immediate we can change e.g. LT to LE and adjust
the immediate, saving a li instruction.

* config/rs6000/rs6000.md (<code><GPR:mode><GPR2:mode>2_isel): Change
LT/GT/LTU/GTU to LE/GE/LEU/GEU where possible.

From-SVN: r255186

Daily bump.

From-SVN: r255185

PR c++/83058 - ICE on C++ code with negative array index: in warn_placement_new_too_small

gcc/cp/ChangeLog:

PR c++/83058
* init.c (warn_placement_new_too_small): Use offset_int instead of
HOST_WIDE_INT.

gcc/testsuite/ChangeLog:

PR c++/83058
* g++.dg/warn/Wplacement-new-size-5.C: New test.

From-SVN: r255182

re PR c++/81888 (Structured bindings stopped working)

PR c++/81888
* parser.c (cp_parser_decomposition_declaration): Reject just
BRACE_ENCLOSED_INITIALIZER_P initializers with nelts != 1 rather
than all such CONSTRUCTORs, and only if is_direct_init is true.

* g++.dg/cpp1z/decomp30.C: Add a test for structured binding with
= {} and = { a, a } initializers.
* g++.dg/cpp1z/decomp31.C: New test.

From-SVN: r255180

pt.c (primary_template_specialization_p): Rename from

* pt.c (primary_template_specialization_p): Rename from

primary_template_instantiation_p.  Don't check
DECL_TEMPLATE_INSTANTIATION.
* call.c, cp-tree.h, decl2.c: Adjust.

From-SVN: r255179

[gcc]
2017-11-27  Michael Meissner  <meissner@linux.vnet.ibm.com>

PR middle_end/82333
* varasm.c (compare_constant): Take the mode of the constants into
account when comparing floating point constants.

[gcc/testsuite]
2017-11-27  Michael Meissner  <meissner@linux.vnet.ibm.com>

PR middle_end/82333
* gcc.target/powerpc/pr82333.c: New test.

From-SVN: r255177

Add missing noexcept in string_view::find_first_of declaration

2017-11-27  Stephan Bergmann  <sbergman@redhat.com>

* include/std/string_view (basic_string_view::find_first_of): Add
missing noexcept.

From-SVN: r255176

hash-set.h (DEFINE_DEBUG_HASH_SET): Remove static qualifier from explicit instantiation of debug_helper.

* hash-set.h (DEFINE_DEBUG_HASH_SET): Remove static qualifier
from explicit instantiation of debug_helper.
* vec.h (DEFINE_DEBUG_VEC): Ditto.

From-SVN: r255175

Fix std::unordered_map deduction guide

2017-11-27  Stephan Bergmann  <sbergman@redhat.com>

* include/bits/unordered_map.h (unordered_map): Fix deduction guide.
* include/debug/unordered_map (unordered_map): Likewise.

From-SVN: r255174

gimple-fold.c (gimple_fold_builtin_memory_op): Remove dead code, refactor a bit.

2017-11-27  Richard Biener  <rguenther@suse.de>

* gimple-fold.c (gimple_fold_builtin_memory_op): Remove dead code,
refactor a bit.

From-SVN: r255172

tree.c (wide_int_to_tree): Free discarded INTEGER_CST.

2017-11-27  Richard Biener  <rguenther@suse.de>

* tree.c (wide_int_to_tree): Free discarded INTEGER_CST.
(type_hash_canon): Also clear int_cst_hash_table entry for
TYPE_MIN/MAX_VALUE.
(build_nonstandard_integer_type): Hash all TYPE_MAX_VALUEs.

From-SVN: r255171

extend.texi: Add -A suffix (ARMv8*-A, ARMv7-A).

2017-11-27  Tamar Christina  <tamar.christina@arm.com>

* doc/extend.texi: Add -A suffix (ARMv8*-A, ARMv7-A).
* doc/invoke.texi: Add -A suffix (ARMv8*-A, ARMv7-A).
* doc/sourcebuild.texi: Add -A suffix (ARMv8*-A, ARMv7-A).

From-SVN: r255169

re PR c++/81675 (attribute(noreturn) of destructor in :? not honored)

PR c++/81675
* cp-gimplify.c (cp_fold) <case COND_EXPR>: Don't return immediately
for VOID_TYPE_P COND_EXPRs, instead fold the operands and if op0 is
INTEGER_CST, ensure that both op1 and op2 are non-NULL and fall
through into normal folding, otherwise just rebuild x if any op
changed.

* g++.dg/warn/pr81675.C: New test.

From-SVN: r255167

Fix UB in hash-map.h

bootstrap-ubsan shows:
  gcc/hash-map.h:277:19: runtime error: member access within null pointer of type 'struct hash_map'

Fix the issue by returning early.

From-SVN: r255166

cfgloop.h (struct loop): Document usage of USHRT_MAX for unroll.

* cfgloop.h (struct loop): Document usage of USHRT_MAX for unroll.
* loop-unroll.c (decide_unroll_constant_iterations): Implement it.
(decide_unroll_runtime_iterations): Likewise.
(decide_unroll_stupid): Likewise.

From-SVN: r255165

Fix code generation for buildtin_longjmp with CET.

According to the description of  inssp instruction from Intel CET it
adusts the shadow stack pointer (ssp) only by value in the range of
[0..255]. As a number of adjustment could be greater than 255 there
should be a loop generated to adjust ssp.

gcc/
* config/i386/i386.md: Add a loop with incssp.
* testsuite/gcc.target/i386/cet-sjlj-1.c: Fix test.
* testsuite/gcc.target/i386/cet-sjlj-4.c: Likewise.

From-SVN: r255164

[PR 81248] Fix ipa-sra size check

2017-11-27  Martin Jambor  <mjambor@suse.cz>

PR tree-optimization/81248
* tree-sra.c (splice_param_accesses): Remove size check.
(decide_one_param_reduction): Fix size check.
* gimple-pretty-print.c (dump_profile): Silence warning.
* params.def (PARAM_IPA_SRA_PTR_GROWTH_FACTOR): Adjust description.

testsuite/
* g++.dg/ipa/pr81248.C: New test.
* gcc.dg/tree-ssa/ssa-pre-31.c: Disable IPA-SRA.
* gcc/testsuite/gcc.dg/ipa/ipcp-cstagg-2.c: Likewise.

From-SVN: r255163

null-8.C (bar): New function.

* g++.dg/ubsan/null-8.C (bar): New function.
(foo): Use bar instead of __builtin_printf.

From-SVN: r255162

re PR debug/81307 (g++.dg/debug/debug9.C -gstabs FAILs)

PR debug/81307
* dbxout.c (lastlineno): New variable.
(dbx_debug_hooks): Use dbxout_switch_text_section as
switch_text_section debug hook.
(dbxout_function_end): Switch to current_function_section
rather than function_section.  If crtl->has_bb_partition,
output just one N_FUN, depending on in_cold_section_p.
(dbxout_source_line): Remember last lineno in lastlineno.
(dbxout_switch_text_section): New function.
(dbxout_function_decl): Adjust dbxout_block caller.
(dbx_block_with_cold_children): New function.
(dbxout_block): Return true if any LBRAC/RBRAC have been
emitted.  Use dbx_block_with_cold_children at depth == 0
in second partition.  Add PARENT_BLOCKNUM argument, pass
it optionally adjusted to children.  Output LBRAC/RBRAC
around recursive call only if the block is in the current
partition, if not and anything was output, emit empty
range LBRAC/RBRAC.
* final.c (final_scan_insn): Compute cold_function_name
before calling switch_text_section debug hook.  Call
that hook even if dwarf2out_do_frame if not emitting
dwarf debug info.

* g++.dg/debug/debug9.C: Remove -fno-reorder-blocks-and-partition
workaround.

From-SVN: r255161

re PR target/83100 (powerpc: internal compiler error: in get_variable_section, at varasm.c:1150 with -fdata-sections)

PR target/83100
* varasm.c (bss_initializer_p): Return true for DECL_COMMON
TREE_READONLY decls.

* gcc.dg/pr83100-1.c: New test.
* gcc.dg/pr83100-2.c: New test.
* gcc.dg/pr83100-3.c: New test.
* gcc.dg/pr83100-4.c: New test.

From-SVN: r255160

Fix PR82488 - signed integer overflow in expr.c

bootstrap-ubsan shows:
 gcc/expr.c:4103:17: runtime error: signed integer overflow: 0 - -9223372036854775808 cannot be represented in type 'long int'

Fix by handling the saw_unknown case earlier.

PR rtl-optimization/82488
* expr.c (fixup_args_size_notes): Avoid signed integer overflow.

diff --git a/gcc/expr.c b/gcc/expr.c
index ee07de5aaa44..e9d8555c9452 100644
--- a/gcc/expr.c
+++ b/gcc/expr.c
@@ -4100,10 +4100,13 @@ fixup_args_size_notes (rtx_insn *prev, rtx_insn *last, int end_args_size)
       if (STACK_GROWS_DOWNWARD)
  this_delta = -(unsigned HOST_WIDE_INT) this_delta;

-      args_size -= this_delta;
+      if (saw_unknown)
+ args_size = INT_MIN;
+      else
+ args_size -= this_delta;
     }

-  return saw_unknown ? INT_MIN : args_size;
+  return args_size;
 }

 #ifdef PUSH_ROUNDING
--
Markus

From-SVN: r255159

Daily bump.

From-SVN: r255158

i386.c (processor_target_table): Add skylake_cost for skylake-avx512.

* config/i386/i386.c (processor_target_table): Add skylake_cost for
skylake-avx512.
* config/i386/x86-tune-costs.h (skylake_memcpy, skylake_memset,
skylake_cost): New.

* config/i386/driver-i386.c (host_detect_local_cpu):
Detect skylake-avx512.

* config.gcc: Add -march=cannonlake.
* config/i386/driver-i386.c (host_detect_local_cpu): Detect cannonlake.
* config/i386/i386-c.c (ix86_target_macros_internal): Handle cannonlake.
* config/i386/i386.c (processor_costs): Add m_CANNONLAKE.
(PTA_CANNONLAKE): New.
(processor_target_table): Add cannonlake.
(ix86_option_override_internal): Ditto.
(fold_builtin_cpu): Ditto.
(get_builtin_code_for_version): Handle cannonlake.
(M_INTEL_COREI7_CANNONLAKE): New.
* config/i386/i386.h (TARGET_CANNONLAKE, PROCESSOR_CANNONLAKE): New.
* doc/invoke.texi: Add -march=cannonlake.

gcc/testsuite/

* gcc.target/i386/funcspec-56.inc: Handle new march.
* g++.dg/ext/mv16.C: Ditto.

libgcc/

* config/i386/cpuinfo.c (get_intel_cpu): Handle cannonlake.
* config/i386/cpuinfo.h (processor_subtypes): Add
INTEL_COREI7_CANNONLAKE.

From-SVN: r255155

Plugin support on Windows/MinGW

config/ChangeLog:
2017-11-14 Boris Kolpackov  <boris@codesynthesis.com>

* gcc-plugin.m4: Add support for MinGW.

gcc/ChangeLog:
2017-11-14 Boris Kolpackov  <boris@codesynthesis.com>

* plugin.c (add_new_plugin): Use platform-specific library extensions.
(try_init_one_plugin): Alternative implementation for MinGW.
* Makefile.in (plugin_implib): New.
(gengtype-lex.c): Fix broken AIX workaround.
* configure: Regenerate.
* doc/plugins.texi: Document support for MinGW.

gcc/c/ChangeLog:
2017-11-14 Boris Kolpackov  <boris@codesynthesis.com>

* Make-lang.in (c.install-plugin): Install backend import library.

gcc/cp/ChangeLog:
2017-11-14 Boris Kolpackov  <boris@codesynthesis.com>

* Make-lang.in (c++.install-plugin): Install backend import library.

libcc1/ChangeLog:
2017-11-14 Boris Kolpackov  <boris@codesynthesis.com>

* configure: Regenerate.

From-SVN: r255154

Daily bump.

From-SVN: r255153

re PR rtl-optimization/81553 (ICE in immed_wide_int_const, at emit-rtl.c:607)

PR rtl-optimization/81553
* combine.c (simplify_if_then_else): In (if_then_else COND (OP Z C1) Z)
to (OP Z (mult COND (C1 * STORE_FLAG_VALUE))) optimization, if OP
is a shift where C1 has different mode than the whole shift, use C1's
mode for MULT rather than the shift's mode.

* gcc.c-torture/compile/pr81553.c: New test.

From-SVN: r255150

structret1.C (FrameworkObject::action): Return a value.

* g++.dg/abi/structret1.C (FrameworkObject::action): Return a
value.

From-SVN: r255149

re PR target/82848 (ICE in expand_expr_real_2, at expr.c:8715)

PR target/82848
* config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Don't fold
builtins not enabled in the currently selected ISA.

* gcc.target/powerpc/pr82848.c: New test.

From-SVN: r255148

Daily bump.

From-SVN: r255147

re PR fortran/81304 (Bogus warning with -Wsurprising and -fopenmp: Type specified for intrinsic function 'min' / 'max')

PR fortran/81304
* trans-openmp.c (gfc_trans_omp_array_reduction_or_udr): Set
attr.implicit_type in intrinsic_sym to avoid undesirable warning.

* testsuite/libgomp.fortran/pr81304.f90: New test.

From-SVN: r255144

Factor out division by squares

This patch implements the some of the division optimizations discussed in
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=71026.

The division reciprocal optimization now handles divisions by squares:

     x / (y * y) -> x  * (1 / y) * (1 / y)

This requires at least one more division by y before it triggers - the
3 divisions of (1/ y) are then CSEd into a single division.  Overall
this changes 1 division into 1 multiply, which is generally much faster.

2017-11-24  Jackson Woodruff  <jackson.woodruff@arm.com>

    gcc/
PR tree-optimization/71026
* tree-ssa-math-opts (is_division_by_square, is_square_of): New.
(insert_reciprocals): Change to insert reciprocals before a division
by a square and to insert the square of a reciprocal.
(execute_cse_reciprocals_1): Change to consider division by a square.
(register_division_in): Add importance parameter.

    testsuite/
PR tree-optimization/71026
* gfortran.dg/extract_recip_1.f: New test.
* gcc.dg/extract_recip_3.c: New test.
* gcc.dg/extract_recip_4.c: New test.

From-SVN: r255141

re PR tree-optimization/82402 (error: SSA_NAME_OCCURS_IN_ABNORMAL_PHI should be set)

2017-11-24  Richard Biener  <rguenther@suse.de>

PR tree-optimization/82402
* tree-vect-loop-manip.c (create_lcssa_for_virtual_phi): Properly
set SSA_NAME_OCCURS_IN_ABNORMAL_PHI.

* gcc.dg/torture/pr82402.c: New testcase.

From-SVN: r255140

Simplify 0 - ptr.

2017-11-24  Marc Glisse  <marc.glisse@inria.fr>

* match.pd (0-ptr): New transformation.

From-SVN: r255139

re PR bootstrap/83015 (bootstrap comparison failure on ia64)

PR bootstrap/83015
* ipa-inline.c (inline_small_functions): Set current badnes correctly
when skipping checking.

From-SVN: r255138

Whitespace fix for libgfortran/ChangeLog

From-SVN: r255137

re PR tree-optimization/83128 (Unable to optimize {m,c}alloc when strings builtin are used)

2017-11-24  Richard Biener  <rguenther@suse.de>

PR tree-optimization/83128
* tree-ssa-sccvn.c (fully_constant_vn_reference_p): Handle STRING_CSTs.
(vn_reference_lookup_3): Likewise.

* gcc.dg/tree-ssa/ssa-fre-62.c: New testcase.

From-SVN: r255136

PR 36313 Replace int with gfc_charlen_type, take 3

Still some prototypes that didn't match the implementation.

2017-11-24  Janne Blomqvist  <jb@gcc.gnu.org>
        PR fortran/36313
        * m4/maxloc2s.m4: Replace int with gfc_charlen_type, take 3.
        * m4/minloc2s.m4: Likewise.
        * generated/maxloc2_16_s1.c: Regenerated.
        * generated/maxloc2_16_s4.c: Regenerated.
        * generated/maxloc2_4_s1.c: Regenerated.
        * generated/maxloc2_4_s4.c: Regenerated.
        * generated/maxloc2_8_s1.c: Regenerated.
        * generated/maxloc2_8_s4.c: Regenerated.
        * generated/minloc2_16_s1.c: Regenerated.
        * generated/minloc2_16_s4.c: Regenerated.
        * generated/minloc2_4_s1.c: Regenerated.
        * generated/minloc2_4_s4.c: Regenerated.
        * generated/minloc2_8_s1.c: Regenerated.
        * generated/minloc2_8_s4.c: Regenerated.

From-SVN: r255135

re PR sanitizer/83014 (ICE in pretty-print with -fsanitize=bounds)

PR sanitizer/83014
* ubsan.c (ubsan_type_descriptor): Use pp_unsigned_wide_integer
instead of pp_printf with HOST_WIDE_INT_PRINT_DEC.  Avoid calling
tree_to_uhwi twice.

* gcc.dg/ubsan/pr83014.c: New test.

From-SVN: r255134

tree-object-size.c (pass_through_call): Use gimple_call_return_flags ERF_RETURN*ARG* for builtins other than...

* tree-object-size.c (pass_through_call): Use gimple_call_return_flags
ERF_RETURN*ARG* for builtins other than BUILT_IN_ASSUME_ALIGNED,
check for the latter with gimple_call_builtin_p.  Do not handle
BUILT_IN_STPNCPY_CHK which is not a pass through call.

* gcc.dg/builtin-object-size-18.c: New test.

From-SVN: r255133

Add missing gcc/config.gcc hunks from r255121.

From-SVN: r255132

Add missing hunks from r255121.

From-SVN: r255131

Daily bump.

From-SVN: r255130

Add [[nodiscard]] attribute to std::launder

* libsupc++/new (launder): Add nodiscard attribute.
* testsuite/18_support/launder/nodiscard.cc: New test.

From-SVN: r255127

[ARM] Fix Dot Product NEON intrinsics

2017-11-24  Christophe Lyon  <christophe.lyon@linaro.org>

* config/arm/arm_neon.h: Fix pragma GCC push_options before
vdot_u32.

From-SVN: r255126

re PR ada/83091 (ICE with -g in arm-eabi compilation)

PR ada/83091
* gcc-interface/decl.c (gnat_to_gnu_entity): Do not build a variant
type for the implementation type of a packed array.

From-SVN: r255125

Add [[nodiscard]] attribute to C++17 components

* include/bits/fs_path.h (path::empty): Add nodiscard attribute.
* include/bits/range_access.h (empty): Likewise.
* include/std/string_view (basic_string_view::empty): Likewise.
* testsuite/21_strings/basic_string_view/capacity/empty_neg.cc: New
test.
* testsuite/24_iterators/range_access_cpp17_neg.cc: New test.
* testsuite/27_io/filesystem/path/query/empty_neg.cc: New test.

From-SVN: r255124

PR libstdc++/83134 Ensure std::__not_ converts B::value to bool

PR libstdc++/83134
* include/std/type_traits (__not_): Explicitly convert to bool.
* testsuite/20_util/declval/requirements/1_neg.cc: Adjust dg-error.
* testsuite/20_util/logical_traits/83134.cc: New test.
* testsuite/20_util/make_signed/requirements/typedefs_neg.cc: Adjust
dg-error.
* testsuite/20_util/make_unsigned/requirements/typedefs_neg.cc:
Likewise.

From-SVN: r255122

Enable VBMI2 support [3/7]

gcc/
config/i386/avx512vbmi2intrin.h (_mm512_mask_expand_epi8,
_mm512_maskz_expand_epi8, _mm512_mask_expandloadu_epi8,
_mm512_maskz_expandloadu_epi8, _mm512_mask_expand_epi16,
_mm512_maskz_expand_epi16, _mm512_mask_expandloadu_epi16,
_mm512_maskz_expandloadu_epi16): New intrinsics.
config/i386/avx512vbmi2vlintrin.h (_mm_mask_expand_epi8,
_mm_maskz_expand_epi8, _mm_mask_expandloadu_epi8,
_mm_maskz_expandloadu_epi8, _mm_mask_expand_epi16,
_mm_maskz_expand_epi16, _mm_mask_expandloadu_epi16,
_mm_maskz_expandloadu_epi16, _mm256_mask_expand_epi16,
_mm256_maskz_expand_epi16, _mm256_mask_expandloadu_epi16,
_mm256_maskz_expandloadu_epi16, _mm256_mask_expand_epi8,
_mm256_maskz_expand_epi8, _mm256_mask_expandloadu_epi8,
_mm256_maskz_expandloadu_epi8): New intrinsics.
config/i386/i386-builtin-types.def (V64QI_FTYPE_PCV64QI_V64QI_UDI,
V32HI_FTYPE_PCV32HI_V32HI_USI, V32QI_FTYPE_PCV32QI_V32QI_USI,
V16HI_FTYPE_PCV16HI_V16HI_UHI, V16QI_FTYPE_PCV16QI_V16QI_UHI,
V8HI_FTYPE_PCV8HI_V8HI_UQI): New types.
config/i386/i386.c (ix86_expand_special_args_builtin): Use new types.
config/i386/sse.md (VI248_VLBW): New iterator.
(expand<mode>_mask, expand<mode>_maskz): New patterns.

gcc/testsuite/
gcc.target/i386/avx512f-vpexpandb-1.c: New test.
gcc.target/i386/avx512f-vpexpandb-2.c: Ditto.
gcc.target/i386/avx512f-vpexpandw-1.c: Ditto.
gcc.target/i386/avx512f-vpexpandw-2.c: Ditto.
gcc.target/i386/avx512vl-vpexpandb-1.c: Ditto.
gcc.target/i386/avx512vl-vpexpandb-2.c: Ditto.
gcc.target/i386/avx512vl-vpexpandw-1.c: Ditto.
gcc.target/i386/avx512vl-vpexpandw-2.c: Ditto.

From-SVN: r255121

Enable VBMI2 support [2/7]

gcc/
config.gcc (avx512vbmi2intrin.h, avx512vbmi2vlintrin): New headers.
config/i386/avx512vbmi2intrin.h (_mm512_mask_compress_epi8,
_mm512_maskz_compress_epi8, _mm512_mask_compressstoreu_epi8,
_mm512_mask_compress_epi16, _mm512_maskz_compress_epi16,
_mm512_mask_compressstoreu_epi16): New.
config/i386/avx512vbmi2vlintrin.h (_mm_mask_compress_epi8,
_mm_maskz_compress_epi8, _mm256_mask_compressstoreu_epi16,
_mm_mask_compress_epi16, _mm_maskz_compress_epi16,
_mm256_mask_compress_epi16, _mm256_maskz_compress_epi16,
_mm_mask_compressstoreu_epi8, _mm_mask_compressstoreu_epi16,
_mm256_mask_compress_epi8, _mm256_maskz_compress_epi8,
_mm256_mask_compressstoreu_epi8): New.
config/i386/i386-builtin-types.def (VOID_FTYPE_PV64QI_V64QI_UDI,
VOID_FTYPE_PV32HI_V32HI_USI, VOID_FTYPE_PV32QI_V32QI_USI,
VOID_FTYPE_PV16QI_V16QI_UHI, VOID_FTYPE_PV16HI_V16HI_UHI,
VOID_FTYPE_PV8HI_V8HI_UQI): New types.
config/i386/i386-builtin.def (__builtin_ia32_compressqi512_mask,
__builtin_ia32_compresshi512_mask, __builtin_ia32_compressqi256_mask,
__builtin_ia32_compressqi128_mask, __builtin_ia32_compresshi256_mask,
__builtin_ia32_compresshi128_mask,
__builtin_ia32_compressstoreuqi512_mask,
__builtin_ia32_compressstoreuhi512_mask,
__builtin_ia32_compressstoreuqi256_mask,
__builtin_ia32_compressstoreuqi128_mask,
__builtin_ia32_compressstoreuhi256_mask,
__builtin_ia32_compressstoreuhi128_mask): New builtins.
config/i386/i386.c (ix86_init_mmx_sse_builtins): Create special args
array for flags2.
(ix86_expand_special_args_builtin): Handle new types.
(s4fma_expand): Handle new builtin array.
config/i386/immintrin.h: Include new headers.
config/i386/sse.md (VI12_AVX512VLBW): New iterator.
(compress<mode>_mask, compressstore<mode>_mask): New patterns.

gcc/testsuite/
gcc.target/i386/avx512-check.h: Handle AVX512VBMI2 bit.
gcc.target/i386/avx512f-vpcompressb-1.c: New test.
gcc.target/i386/avx512f-vpcompressb-2.c: Ditto.
gcc.target/i386/avx512f-vpcompressw-1.c: Ditto.
gcc.target/i386/avx512f-vpcompressw-2.c: Ditto.
gcc.target/i386/avx512vl-vpcompressb-1.c: Ditto.
gcc.target/i386/avx512vl-vpcompressb-2.c: Ditto.
gcc.target/i386/avx512vl-vpcompressw-1.c: Ditto.
gcc.target/i386/avx512vl-vpcompressw-2.c: Ditto.
gcc.target/i386/i386.exp (check_effective_target_avx512vbmi2): New.

From-SVN: r255119

rs6000.c (rs6000_gimple_fold_builtin): Formatting fixes.

* config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Formatting
fixes.  Declare temp and g variables at the top in order to avoid
{} in most of the cases.

From-SVN: r255118

parser.c (cp_parser_omp_declare): Change return type to bool from void, return true for declare simd.

* parser.c (cp_parser_omp_declare): Change return type to bool from
void, return true for declare simd.
(cp_parser_pragma): Return cp_parser_omp_declare returned value
rather than always false.

From-SVN: r255117

c-parser.c (c_parser_omp_declare_simd): Reject declare simd in pragma_stmt context.

* c-parser.c (c_parser_omp_declare_simd): Reject declare simd in
pragma_stmt context.

* gcc.dg/gomp/declare-simd-1.c (f9): Remove.
* gcc.dg/gomp/declare-simd-5.c: New test.

From-SVN: r255116

re PR fortran/81841 (THREADPRIVATE (OpenMP) wrongly rejected in BLOCK DATA)

PR fortran/81841
* parse.c (parse_spec): Allow ST_OMP_THREADPRIVATE inside of
BLOCK DATA.

* libgomp.fortran/pr81841.f90: New test.

From-SVN: r255115

pr65240-1.C: Add -Wno-return-type to dg-options.

* g++.dg/pr65240-1.C: Add -Wno-return-type to dg-options.
* g++.dg/pr65240-2.C: Likewise.
* g++.dg/pr65240-3.C: Likewise.
* g++.dg/pr65240-4.C: Likewise.

From-SVN: r255114

Simplify ptr - 0

2017-11-23  Marc Glisse  <marc.glisse@inria.fr>

* match.pd (ptr-0): New transformation.

From-SVN: r255113

This patch adds support for modelling the varying costs of different addressing modes.

This patch adds support for modelling the varying costs of
different addressing modes. The generic cost table treats
all addressing modes as having equal cost.

gcc/ChangeLog:

2017-11-23  Charles Baylis  <charles.baylis@linaro.org>

        * config/arm/arm-protos.h (enum arm_addr_mode_op): New.
        (struct addr_mode_cost_table): New.
        (struct tune_params): Add field addr_mode_costs.
        * config/arm/arm.c (generic_addr_mode_costs): New.
        (arm_slowmul_tune): Initialise addr_mode_costs field.
        (arm_fastmul_tune): Likewise.
        (arm_strongarm_tune): Likewise.
        (arm_xscale_tune): Likewise.
        (arm_9e_tune): Likewise.
        (arm_marvell_pj4_tune): Likewise.
        (arm_v6t2_tune): Likewise.
        (arm_cortex_tune): Likewise.
        (arm_cortex_a8_tune): Likewise.
        (arm_cortex_a7_tune): Likewise.
        (arm_cortex_a15_tune): Likewise.
        (arm_cortex_a35_tune): Likewise.
        (arm_cortex_a53_tune): Likewise.
        (arm_cortex_a57_tune): Likewise.
        (arm_exynosm1_tune): Likewise.
        (arm_xgene1_tune): Likewise.
        (arm_cortex_a5_tune): Likewise.
        (arm_cortex_a9_tune): Likewise.
        (arm_cortex_a12_tune): Likewise.
        (arm_cortex_a73_tune): Likewise.
        (arm_v7m_tune): Likewise.
        (arm_cortex_m7_tune): Likewise.
        (arm_v6m_tune): Likewise.
        (arm_fa726te_tune): Likewise.
        (arm_mem_costs): Use table lookup to calculate cost of addressing
        mode.

From-SVN: r255112

[ARM] Refactor costs calculation for MEM.

This patch moves the calculation of costs for MEM into a
separate function, and reforms the calculation into two
parts. Firstly any additional cost of the addressing mode
is calculated, and then the cost of the memory access itself
is added.

In this patch, the calculation of the cost of the addressing
mode is omitted, to be added in a subsequent patch.

gcc/ChangeLog:

<date>  Charles Baylis  <charles.baylis@linaro.org>

        * config/arm/arm.c (arm_mem_costs): New function.
        (arm_rtx_costs_internal): Use arm_mem_costs.

gcc/testsuite/ChangeLog:

<date>  Charles Baylis  <charles.baylis@linaro.org>

        * gcc.target/arm/addr-modes-float.c: New test.
        * gcc.target/arm/addr-modes-int.c: New test.
        * gcc.target/arm/addr-modes.h: New header.

From-SVN: r255111

re PR fortran/36313 ([F03] {MIN,MAX}{LOC,VAL} should accept character arguments)

2017-11-23  Thomas Koenig  <tkoenig@gcc.gnu.org>

PR fortran/36313
* m4/maxloc2s.m4: Replace int for character length by
gfc_charlen_type.
* m4/minloc2s.m4: Likewise.
* generated./maxloc2_16_s1.c: Regenerated.
* generated./maxloc2_16_s4.c: Regenerated.
* generated./maxloc2_4_s1.c: Regenerated.
* generated./maxloc2_4_s4.c: Regenerated.
* generated./maxloc2_8_s1.c: Regenerated.
* generated./maxloc2_8_s4.c: Regenerated.
* generated./minloc2_16_s1.c: Regenerated.
* generated./minloc2_16_s4.c: Regenerated.
* generated./minloc2_4_s1.c: Regenerated.
* generated./minloc2_4_s4.c: Regenerated.
* generated./minloc2_8_s1.c: Regenerated.
* generated./minloc2_8_s4.c: Regenerated.

From-SVN: r255109

Fix .debug_rnglists generation with -gdwarf-5 -gsplit-dwarf.

Early debug broke generation of .debug_rnglists when using both -gdwarf5
and -gsplit-dwarf. It introduces a generation for init_sections_and_labels,
but doesn't account for the generation of up to 4 unique ranges labels,
two created in init_sections_and_labels and two in output_rnglists.
Fix this by passing generation to output_rnglists and creating 4 unique
labels per generation.

Without this fix using -gdwarf-5 -gsplit-dwarf could result in:
  Error: symbol `.Ldebug_ranges2' is already defined

gcc/ChangeLog:

       * dwarf2out.c (init_sections_and_labels): Use generation to create
       unique ranges_section_label and ranges_base_label. Return generation.
       (output_rnglists): Add generation argument. Use generation to create
       unique ranges labels.
       (dwarf2out_finish): Get generation from init_sections_and_labels
       and pass generation to output_rnglists.

From-SVN: r255107

generic.texi (ANNOTATE_EXPR): Document 3rd operand.

* doc/generic.texi (ANNOTATE_EXPR): Document 3rd operand.
* cfgloop.h (struct loop): Add unroll field.
* function.h (struct function): Add has_unroll bitfield.
* gimplify.c (gimple_boolify) <ANNOTATE_EXPR>: Deal with unroll kind.
(gimplify_expr) <ANNOTATE_EXPR>: Propagate 3rd operand.
* loop-init.c (pass_loop2::gate): Return true if cfun->has_unroll.
(pass_rtl_unroll_loops::gate): Likewise.
* loop-unroll.c (decide_unrolling): Tweak note message.  Skip loops
for which loop->unroll==1.
(decide_unroll_constant_iterations): Use note for consistency and
take loop->unroll into account.  Return early if loop->unroll is set.
Fix thinko in existing test.
(decide_unroll_runtime_iterations): Use note for consistency and
take loop->unroll into account.
(decide_unroll_stupid): Likewise.
* lto-streamer-in.c (input_cfg): Read loop->unroll.
* lto-streamer-out.c (output_cfg): Write loop->unroll.
* tree-cfg.c (replace_loop_annotate_in_block) <annot_expr_unroll_kind>:
        New case.
(replace_loop_annotate) <annot_expr_unroll_kind>: Likewise.
(print_loop): Print loop->unroll if set.
* tree-core.h (enum annot_expr_kind): Add annot_expr_unroll_kind.
* tree-inline.c (copy_loops): Copy unroll and set cfun->has_unroll.
* tree-pretty-print.c (dump_generic_node) <annot_expr_unroll_kind>:
New case.
* tree-ssa-loop-ivcanon.c (try_unroll_loop_completely): Bail out if
loop->unroll is set and smaller than the trip count.  Otherwise bypass
entirely the heuristics if loop->unroll is set.  Remove dead note.
Fix off-by-one bug in other note.
(try_peel_loop): Bail out if loop->unroll is set.  Fix formatting.
(tree_unroll_loops_completely_1): Force unrolling if loop->unroll
is greater than 1.
(tree_unroll_loops_completely): Make static.
(pass_complete_unroll::execute): Use correct type for variable.
(pass_complete_unrolli::execute): Fix formatting.
* tree.def (ANNOTATE_EXPR): Add 3rd operand.
ada/
* gcc-interface/trans.c (gnat_gimplify_stmt) <LOOP_STMT>: Pass 3rd
operand to ANNOTATE_EXPR and also pass unrolling hints.
c/
* c-parser.c (c_parser_while_statement): Pass 3rd operand to
ANNOTATE_EXPR.
(c_parser_do_statement): Likewise.
(c_parser_for_statement): Likewise.
cp/
* pt.c (tsubst_expr) <ANNOTATE_EXPR>: Recurse on 3rd operand.
* semantics.c (finish_while_stmt_cond): Pass 3rd operand to
ANNOTATE_EXPR.
(finish_do_stmt): Likewise.
(finish_for_cond): Likewise.
fortran/
* trans-stmt.c (gfc_trans_forall_loop): Pass 3rd operand to
ANNOTATE_EXPR.

From-SVN: r255106