Sudakshina Das [Mon, 9 Dec 2019 10:22:01 +0000 (10:22 +0000)]
[GCC, Vect] Fix costing for vector shifts
While looking at the vectorization for following example, we realized that
even though vectorizable_shift function was distinguishing vector shifted
by vector from vector shifted by scalar, while modelling the cost it would
always add the cost of building a vector constant despite not needing it for
vector shifted by scalar.
This patch fixes this by using scalar_shift_arg to determine whether we need
to build a vector for the second operand or not. This causes the test case
below to now vectorize.
gcc/ChangeLog:
2019-12-09 Sudakshina Das <sudi.das@arm.com>
Richard Sandiford <richard.sandiford@arm.com>
* tree-vect-stmt.c (vectorizable_shift): Condition ndts for
vect_model_simple_cost call on scalar_shift_arg.
gcc/testsuite/ChangeLog:
2019-12-09 Sudakshina Das <sudi.das@arm.com>
* gcc.dg/vect/vect-shift-5.c: New test.
Co-Authored-By: Richard Sandiford <richard.sandiford@arm.com>
From-SVN: r279114
Jakub Jelinek [Mon, 9 Dec 2019 10:13:18 +0000 (11:13 +0100)]
re PR tree-optimization/92834 (misssed SLP vectorization in LightPixel)
PR tree-optimization/92834
* match.pd (A - ((A - B) & -(C cmp D)) -> (C cmp D) ? B : A,
A + ((B - A) & -(C cmp D)) -> (C cmp D) ? B : A): New simplifications.
PR tree-optimization/89007
* tree-vect-patterns.c (vect_recog_average_pattern): If there is no
target support available, generate code to distribute rshift over plus
and add a carry.
testsuite/
* gcc.target/aarch64/sve/pr89007-1.c: New test.
* gcc.target/aarch64/sve/pr89007-2.c: Likewise.
Jonathan Wakely [Mon, 9 Dec 2019 09:58:56 +0000 (09:58 +0000)]
libstdc++: fix buffer overflow in path::operator+= (PR92853)
When concatenating a path ending in a root-directory onto another path,
we added an empty filename to the end of the path twice, but only
reserved space for one. That meant the second write went past the end of
the allocated buffer.
PR libstdc++/92853
* src/c++17/fs_path.cc (filesystem::path::operator+=(const path&)):
Do not process a trailing directory separator twice.
* testsuite/27_io/filesystem/path/concat/92853.cc: New test.
* testsuite/27_io/filesystem/path/concat/path.cc: Test more cases.
Hongtao Liu [Mon, 9 Dec 2019 04:16:24 +0000 (04:16 +0000)]
Enable mask movement for VCOND_EXPR under avx512f for
128/256-bit vector when integer mask is available.
Changelog
gcc/
PR target/92686
* config/i386/sse.md
(*<avx512>_cmp<mode>3<mask_scalar_merge_name><round_saeonly_name>,
*<avx512>_cmp<mode>3<mask_scalar_merge_name>,
*<avx512>_ucmp<mode>3<mask_scalar_merge_name>,
*<avx512>_ucmp<mode>3<mask_scalar_merge_name>): New.
* config/i386/i386.c (ix86_print_operand): New operand substitution.
* config/i386/i386-expand.c (ix86_valid_mask_cmp_mode):
New function.
(ix86_expand_sse_cmp): Relax condition for integer mask from
512-bit vector to all 128/256/512-bit vector. Delete code gen
for avx512f compare patterns since we have generic pattern now.
(ix86_expand_sse_movcc): Adjust condition and codegen for
maskcmp.
(ix86_expand_int_sse_cmp): Don't canonicalize the comparison
when corresponding vector compare is available.
Jan Hubicka [Sun, 8 Dec 2019 17:02:30 +0000 (18:02 +0100)]
Fix overflows in -fprofile-reorder-functions
This patch fixes three sissues with -fprofile-reorder-functions:
1) First is that tp_first_run is stored as 32bit integer while it can easily
overflow (and does so during Firefox profiling).
2) Second problem is that flag_profile_functions can
not be tested w/o function context.
The changes to expand_all_functions makes it to work on mixed units by
first outputting all functions w/o -fprofile-reorder-function (or with no
profile info) and then outputting in first_run order
3) LTO partitioner was mixing up order by tp_first_run and by order.
for no_reorder we definitly want to order via first, while for everything
else we want to roder by second.
I have also merged duplicated comparators since they are bit fragile into
tp_first_run_node_cmp.
I originaly started to look into this because of undefined symbols with
Firefox PGO builds. These symbols went away with fixing these bug but I am not
quite sure how. it is possible that there is another problem in lto_blanced_map
but even after reading the noreorder code few times carefuly I did not find it.
Other explanation would be that our new qsort with broken comparator due to
overflow can actualy remove some entries in the array, but that sounds bit
crazy.
Bootstrapped/regested x86_64-linux.
* cgraph.c (cgraph_node::dump): Make tp_first_run 64bit.
* cgraph.h (cgrpah_node): Likewise.
(tp_first_run_node_cmp): Deeclare.
* cgraphunit.c (node_cmp): Rename to ...
(tp_first_run_node_cmp): ... this; export; watch for 64bit overflows;
clear tp_first_run for no_reorder and !flag_profile_reorder_functions.
(expand_all_functions): Collect tp_first_run and normal functions to
two vectors so the other functions remain sorted. Do not check for
flag_profile_reorder_functions it is function local flag.
* profile.c (compute_value_histograms): Update tp_first_run printing.
* lto-partition.c (node_cmp): Turn into simple order comparsions.
(varpool_node_cmp): Remove.
(add_sorted_nodes): Use node_cmp.
(lto_balanced_map): Use tp_first_run_node_cmp.
Jan Hubicka [Sun, 8 Dec 2019 14:02:43 +0000 (15:02 +0100)]
Silence overactive sanity check with -fpartial-profile-training
do_estimate_edge_time tests that cached and real values matches. This test
is not working precisely for global profiles because of roundoff issues when
profile of clones is subtracted from profile of offline body. This is checked
by presence of ipa counter. This breaks with partial profile training because
we turn IPA profiles to local when they drop to 0.
Marek Polacek [Sat, 7 Dec 2019 17:35:54 +0000 (17:35 +0000)]
PR c++/91678 - wrong error with decltype and location wrapper.
Compiling this testcase results in a bogus "invalid cast" error; this occurs
since the introduction of location wrappers in finish_id_expression.
Here we are parsing the decltype expression via cp_parser_decltype_expr which
can lead to calling various fold_* and c-family routines. They use
non_lvalue_loc, but that won't create a NON_LVALUE_EXPR wrapper around a location
wrapper.
So before the location wrappers addition cp_parser_decltype_expr would return
NON_LVALUE_EXPR <c>. Now it returns VIEW_CONVERT_EXPR<float *>(c), but the
STRIP_ANY_LOCATION_WRAPPER immediately following it strips the location wrapper,
and suddenly we don't know whether we have an lvalue anymore. And that's sad
because then decltype produces the wrong type, causing nonsense errors.
Eric Botcazou [Sat, 7 Dec 2019 12:20:44 +0000 (12:20 +0000)]
re PR middle-end/90840 (ICE in simplify_subreg, at simplify-rtx.c:6441)
PR middle-end/90840
* expr.c (expand_assignment): In the case of a CONCAT on the LHS, make
sure to pass a valid inner mode in calls to simplify_gen_subreg.
While trying out Dennis's Armv8.6-A patch, I noticed that texi2pod.pl
didn't handle the new @multitable correctly. There were two problems:
(1) @multitables nested in other @tables inherited the @item type from
the enclosing @table. Since the new @multitable is in a @table @samp,
we applied @samp markup to the @multitable @items. This in turn
meant that it captured the @tab separator in the @item markup.
Fixed by pushing an empty item code onto the stack.
(2) We didn't handle @headitem. Fixed by enclosing it in italics,
like we do for section headings. This causes it to be underlined
in the man output.
2019-12-07 Richard Sandiford <richard.sandiford@arm.com>
contrib/
* texi2pod.pl: Handle @headitems in @multitables, printing them
in italics. Push an empty item code onto the stack.
Jakub Jelinek [Fri, 6 Dec 2019 23:43:45 +0000 (00:43 +0100)]
re PR c++/92831 (CWG1299 extend_ref_init_temps_1 punts on COND_EXPRs)
PR c++/92831
* call.c (build_conditional_expr_1): For ?: with omitted middle
operand use cp_stabilize_reference if arg1 is glvalue_p rather than
just if it is lvalue_p.
Jakub Jelinek [Fri, 6 Dec 2019 20:16:27 +0000 (21:16 +0100)]
PR c++/92831 - CWG 1299, not extending temporary lifetime for ?:
* cp-tree.h (extend_ref_init_temps): Add a new argument with NULL
default arg.
* call.c (set_up_extended_ref_temp): Add COND_GUARD argument, pass it
down to extend_ref_init_temps. Before pushing cleanup, if COND_GUARD
is non-NULL, create a bool temporary if needed, initialize to false
and guard the cleanup with the temporary being true.
(extend_ref_init_temps_1): Add COND_GUARD argument, pass it down
to recursive calls and set_up_extended_ref_temp. Handle COND_EXPR.
(extend_ref_init_temps): Add COND_GUARD argument, pass it down to
recursive calls and to extend_ref_init_temps_1.
[C++] Pass type uses through the verify_type_context hook
This patch makes the C++ frontend work with the verify_type_context hook.
We need some new type contexts for features that don't exist in C, but
otherwise the patch is very similar to the C one.
TCTX_CAPTURE_BY_COPY could really be treated as an instance of
TCTX_FIELD, but the error message is better if we split it out.
2019-12-06 Richard Sandiford <richard.sandiford@arm.com>
gcc/cp/
* decl.c (start_decl_1): Use verify_type_context to check whether
the target allows variables of a particular type to have static
or thread-local storage duration.
(check_array_initializer): Use verify_type_context to check whether
the target allows a particular type to be used as an array element.
(create_array_type_for_decl): Likewise.
(cp_finish_decl): Use verify_type_context to check whether
the target allows static member variables of a particular type.
(grokdeclarator): Likewise. Also use verify_type_context to check
whether the target allows non-static member variables of a particular
type.
* except.c: Include target.h.
(is_admissible_throw_operand_or_catch_parameter): Use
verify_type_context to check whether the target allows particular
types to be thrown and caught.
* typeck2.c (add_exception_specifier): Likewise.
* init.c (build_new_1): Use verify_type_context to check whether
the target allows particular types to be dynamically allocated.
(build_vec_delete_1, build_delete): Use verify_type_context to check
whether the target allows particular types to be deleted.
* lambda.c (add_capture): Use verify_type_context to check
whether the target allows particular types to be captured by copy.
* pt.c: Include target.h.
(instantiate_class_template_1): Use verify_type_context to check
whether the target allows non-static member variables of a particular
type.
* typeck.c (cxx_alignof_expr): Use verify_type_context to check
whether the target allows the alignment of a particular type
to be measured.
(pointer_diff, cp_build_unary_op): Use verify_type_context to check
whether the target allows arithmetic involving pointers to particular
types.
gcc/testsuite/
* g++.dg/ext/sve-sizeless-1.C: New test.
* g++.dg/ext/sve-sizeless-2.C: Likewise.
Avoid quadratic behaviour in prune_runtime_alias_test_list
prune_runtime_alias_test_list used ordered_remove to remove a merged
alias pair, which made the function quadratic when many aliases could
be removed.
I had a testcase in which these memmoves accounted for an impressive
85% of compile time. The fact that we had so many probably shows
a deeper problem, but still, it's easy to remove as we go.
2019-12-06 Richard Sandiford <richard.sandiford@arm.com>
gcc/
* tree-data-ref.c (prune_runtime_alias_test_list): Exit early
for empty vectors. Avoid using ordered_remove and instead
shuffle the vector as we go.
Richard Biener [Fri, 6 Dec 2019 10:25:08 +0000 (10:25 +0000)]
genmatch.c (enum tree_code): Remove CONVERT{0,1,2} and VIEW_CONVERT{0,1,2}.
2019-12-06 Richard Biener <rguenther@suse.de>
* genmatch.c (enum tree_code): Remove CONVERT{0,1,2} and
VIEW_CONVERT{0,1,2}.
(expr::opt_grp): Add and initialize.
(lower_opt_convert): Rename to ...
(lower_opt): ... and work on opt_grp, simply switching operations
from being optional to being present or not.
(has_opt_convert): Rename to ...
(has_opt): ... and adjust.
(parser::parse_operation): Return the optional opt_grp,
remove special-casing of conditional operations and more generally
parse [digit]'?'.
(parser::parse_expr): Stick on the parsed opt_grp and perform
rough verification.
(parser::parse_for): Remove now unnecessary code.
(main): Likewise.
* doc/match-and-simplify.texi: Mention ? now works on all
unary operations and also match predicates.
Richard Biener [Fri, 6 Dec 2019 07:53:15 +0000 (07:53 +0000)]
re PR tree-optimization/92819 (Worse code generated on avx2 due to simplify_vector_constructor)
2019-12-06 Richard Biener <rguenther@suse.de>
PR tree-optimization/92819
* match.pd (VEC_PERM_EXPR -> BIT_INSERT_EXPR): Handle inserts
into the last lane. For two-element vectors try inserting
into the last lane when inserting into the first fails.
* gcc.target/i386/pr92819-1.c: New testcase.
* gcc.target/i386/pr92803.c: Adjust.
Disable --eh-frame-hdr with -pie or -shared on nios2-linux-gnu.
The nios2 ABI doesn't have appropriate relocations to support
--eh-frame-hdr with -pie or -shared, so we need to override the
default specs that add that link option. Otherwise we get linker
warnings about invalid FDE encodings.
Jan Hubicka [Thu, 5 Dec 2019 20:53:39 +0000 (21:53 +0100)]
Fix g++.dg/torture/pr59226.C
this patch fixes ICE in g++.dg/torture/pr59226.C which was triggered by
new comdat_local sanity check. What happens here is that function gets
inlined into its own thunk which makes it !comdat_local_p but the updating
code does not notice since thunk calls comdat local alias of the function
itself and we look at alias target rather than original callee.
This also shows that we miss optimization here. Currently we will not inline
thunk out of its comdat local group w/o inlining function it is associated with
into it.
We should teach inline_call to reoslve edges to aliases while inlining and
relax calls_comdat_local flag. But this needs bit more work, so I fix the
ICE first.
* ipa-inline-transform.c (inline_call): Fix maintenatnce of comdat_local
This patch implements C++20 P1331, allowing trivial default initialization in
constexpr contexts.
* c-cppbuiltin.c (c_cpp_builtins): Adjust the value of __cpp_constexpr.
* class.c (trivial_default_constructor_is_constexpr): Return true in
C++20.
* constexpr.c (cx_check_missing_mem_inits): Allow missing field
initializers in C++20.
(cxx_eval_call_expression): Don't clear CONSTRUCTOR_NO_CLEARING for
constexpr constructors in C++20.
(reduced_constant_expression_p): Don't set FIELD for union and array
types. Skip empty class fields without initializers.
* decl.c (check_for_uninitialized_const_var): Permit trivial default
initialization in constexpr.
(next_initializable_field): Don't skip vptr fields.
* method.c (walk_field_subobs): Still consider a constructor that
doesn't initialize all the members constexpr.
* g++.dg/cpp0x/constexpr-array6.C: Adjust dg-error.
* g++.dg/cpp0x/constexpr-ctor.C: Likewise.
* g++.dg/cpp0x/constexpr-diag3.C: Likewise.
* g++.dg/cpp0x/constexpr-diag4.C: Likewise.
* g++.dg/cpp0x/constexpr-ex3.C: Likewise.
* g++.dg/cpp0x/constexpr-template2.C: Likewise.
* g++.dg/cpp0x/constexpr-union2.C: Likewise.
* g++.dg/cpp0x/lambda/lambda-mangle.C: Rip out a piece of code ...
* g++.dg/cpp0x/lambda/lambda-mangle6.C: ... and put it here.
* g++.dg/cpp0x/pr79118.C: Adjust dg-error.
* g++.dg/cpp1y/constexpr-83921-3.C: Likewise.
* g++.dg/cpp1y/constexpr-neg1.C: Likewise.
* g++.dg/cpp1z/constexpr-lambda12.C: Likewise.
* g++.dg/cpp1z/feat-cxx1z.C: Use -std=c++17.
* g++.dg/cpp2a/constexpr-init1.C: New test.
* g++.dg/cpp2a/constexpr-init2.C: New test.
* g++.dg/cpp2a/constexpr-init3.C: New test.
* g++.dg/cpp2a/constexpr-init4.C: New test.
* g++.dg/cpp2a/constexpr-init5.C: New test.
* g++.dg/cpp2a/constexpr-init6.C: New test.
* g++.dg/cpp2a/constexpr-init7.C: New test.
* g++.dg/cpp2a/constexpr-init8.C: New test.
* g++.dg/cpp2a/constexpr-init9.C: New test.
* g++.dg/cpp2a/constexpr-init10.C: New test.
* g++.dg/cpp2a/constexpr-init11.C: New test.
* g++.dg/cpp2a/constexpr-init12.C: New test.
* g++.dg/cpp2a/constexpr-init13.C: New test.
* g++.dg/cpp2a/constexpr-init14.C: New test.
* g++.dg/cpp2a/constexpr-init15.C: New test.
* g++.dg/cpp2a/constexpr-try5.C: Adjust dg-error.
* g++.dg/cpp2a/feat-cxx2a.C: Test __cpp_constexpr.
* g++.dg/cpp2a/lambda-mangle.C: New test.
* g++.dg/debug/dwarf2/pr44641.C: Skip for c++2a.
* g++.dg/ext/stmtexpr21.C: Adjust dg-error.
Co-Authored-By: Jakub Jelinek <jakub@redhat.com>
From-SVN: r279019
Jan Hubicka [Thu, 5 Dec 2019 18:12:51 +0000 (19:12 +0100)]
cgraphclones.c (localize_profile): New function.
* cgraphclones.c (localize_profile): New function.
(cgraph_node::create_clone): Use it for partial profiles.
* common.opt (fprofile-partial-training): New flag.
* doc/invoke.texi (-fprofile-partial-training): Document.
* ipa-cp.c (update_profiling_info): For partial profiles do not
set function profile to zero.
* profile.c (compute_branch_probabilities): With partial profile
watch if edge count is zero and turn all probabilities to guessed.
(compute_branch_probabilities): For partial profiles do not apply
profile when entry count is zero.
* tree-profile.c (tree_profiling): Only do value_profile_transformations
when profile is read.
Tobias Burnus [Thu, 5 Dec 2019 15:18:39 +0000 (15:18 +0000)]
OpenMP] Fix use_device_… with absent optional arg
gcc/fortran/
* trans-openmp.c (gfc_omp_is_optional_argument,
gfc_omp_check_optional_argument): Handle type(c_ptr),value which uses a
hidden argument for the is-present check.
gcc/
* omp-low.c (lower_omp_target): For use_device_ptr/use_derice_addr
and Fortran's optional arguments, unconditionally add the is-present
condition before the libgomp call.
libgomp/
* testsuite/libgomp.fortran/use_device_ptr-optional-2.f90: Add
'type(c_ptr), value' test case. Conditionally map the per-value
passed arguments.
Check for bitwise identity when encoding VECTOR_CSTs (PR 92768)
This PR shows that we weren't checking for bitwise-identical values
when trying to encode a VECTOR_CST, so -0.0 was treated the same as
0.0 for -fno-signed-zeros. The patch adds a new OEP flag to select
that behaviour.
2019-12-05 Richard Sandiford <richard.sandiford@arm.com>
gcc/
PR middle-end/92768
* tree-core.h (OEP_BITWISE): New flag.
* fold-const.c (operand_compare::operand_equal_p): Handle it.
* tree-vector-builder.h (tree_vector_builder::equal_p): Pass it.
gcc/testsuite/
PR middle-end/92768
* gcc.dg/pr92768.c: New test.
Jonathan Wakely [Thu, 5 Dec 2019 13:50:08 +0000 (13:50 +0000)]
libstdc++: Fix bug in std::span test
The previous commit fixed the std::span constructors from const arrays,
revealing a bug in this test.
* testsuite/23_containers/span/lwg3255.cc: Fix test. Constructing a
span of non-const elements should not be possible from a const array
or an array of const elements.
libstdc++: Implement P1872R0 and P1394R0 for std::span
This also fixes a bug in the implementation of LWG 3255, which causes:
FAIL: 23_containers/span/lwg3255.cc (test for excess errors)
That's because the test was wrong and verified the buggy behaviour. That
will be fixed in the following commit.
Implement P1872R0 and P1394R0 for std::span
* include/bits/range_access.h (__adl_begin, __adl_end): Remove.
(sentinel_t, range_value_t, range_reference_t)
(range_rvalue_reference_t, __forwarding_range, disable_sized_range)
(output_range, input_range, forward_range, bidirectional_range)
(random_access_range, contiguous_range, common_range): Move here from
<ranges>, to make this the "ranges lite" internal header.
* include/std/ranges: Move basic aliases and concepts to
<bits/range_access.h>.
* include/std/span: Use concepts and ranges:: calls instead of
enable_if and friends.
* include/std/type_traits: Add __is_array_convertible trait.
Richard Biener [Thu, 5 Dec 2019 13:02:57 +0000 (13:02 +0000)]
re PR tree-optimization/92818 (Typo in vec_perm -> bit_insert pattern)
2019-12-05 Richard Biener <rguenther@suse.de>
PR middle-end/92818
* tree-ssa-forwprop.c (simplify_vector_constructor): Improve
heuristics on what don't care element to choose.
* match.pd (VEC_PERM_EXPR -> BIT_INSERT_EXPR): Fix typo.
Jonathan Wakely [Thu, 5 Dec 2019 12:46:50 +0000 (12:46 +0000)]
libstdc++: Define std::lexicographical_compare_three_way for C++20
* include/bits/stl_algobase.h (lexicographical_compare_three_way):
Define for C++20.
* testsuite/25_algorithms/lexicographical_compare_three_way/1.cc: New
test.
* testsuite/25_algorithms/lexicographical_compare_three_way/
constexpr.cc: New test.
Jakub Jelinek [Thu, 5 Dec 2019 09:04:24 +0000 (10:04 +0100)]
re PR target/92791 (ICE in extract_insn, at recog.c:2311 since r278645)
PR target/92791
* config/i386/i386.md (movstrict<mode>): Move test for
TARGET_PARTIAL_REG_STALL and not optimizing for size from
expander's condition to the body - FAIL; in that case.
Jakub Jelinek [Thu, 5 Dec 2019 09:03:34 +0000 (10:03 +0100)]
re PR fortran/92781 (ICE in convert_nonlocal_reference_op, at tree-nested.c:1065)
PR fortran/92781
* trans-decl.c (gfc_get_symbol_decl): If sym->backend_decl is
current_function_decl, add length to current rather than parent
function and expect DECL_CONTEXT (length) to be current_function_decl.
Jonathan Wakely [Thu, 5 Dec 2019 00:42:06 +0000 (00:42 +0000)]
libstdc++: Implement spaceship for std::array (P1614R2)
As done for std::pair, this defines operator<=> as a non-member function
template and does not alter operator==, as expected to be proposed as
the resolution to an unpublished LWG issue.
Instead of calling std::lexicographical_compare_three_way the <=>
overload is implemented by hand to take advantage of the fact the
element types and array sizes are known to be the same.
* include/bits/cpp_type_traits.h (__is_byte<char8_t>): Add
specialization.
* include/std/array (operator<=>): Likewise.
* testsuite/23_containers/array/comparison_operators/constexpr.cc:
Test three-way comparisons and arrays of unsigned char.
* testsuite/23_containers/array/tuple_interface/get_neg.cc: Adjust
dg-error line numbers.
Joseph Myers [Wed, 4 Dec 2019 23:26:10 +0000 (23:26 +0000)]
Fix C handling of use of lvalues of incomplete types (PR c/36941, PR c/88827).
Bug 88827 points out that GCC should not be rejecting C code that
dereferences a pointer to an incomplete type in the case that uses &*
to take the address of the resulting lvalue, because no constraint is
violated in that case (other than for C90 when the incomplete type is
unqualified void, which we already handle correctly) and as the lvalue
never gets converted to an rvalue there is no undefined behavior
either.
This means that the diagnostic for such a dereference is bogus and
should be removed; if the lvalue gets converted to an rvalue, there
should be an appropriate error later for the use of the incomplete
type. In most cases, there is, but bug 36941 points out the lack of a
diagnostic when the incomplete (non-void) type gets cast to void
(where a diagnostic seems appropriate for this undefined behavior as a
matter of quality of implementation).
This patch removes the bogus diagnostic (and C_TYPE_ERROR_REPORTED
which was only used in the code that is removed - only that one, bogus
diagnostic had this duplicate suppression, not any of the other, more
legitimate diagnostics for use of incomplete types) and makes
convert_lvalue_to_rvalue call require_complete_type for arguments not
of void types, so that all relevant code paths (possibly except some
for ObjC) get incomplete types diagnosed. It's possible that this
makes some other checks for incomplete types obsolete, but no attempt
is made to remove any such checks.
Bootstrapped with no regressions for x86_64-pc-linux-gnu.
PR c/36941
PR c/88827
gcc/c:
* c-typeck.c (convert_lvalue_to_rvalue): Call
require_complete_type for arguments not of void types.
(build_indirect_ref): Do not diagnose dereferencing pointers to
incomplete types.
* c-tree.h (C_TYPE_ERROR_REPORTED): Remove.
Peter Bergner [Wed, 4 Dec 2019 19:53:26 +0000 (19:53 +0000)]
Do not define builtins that overload disabled builtins.
PR bootstrap/92661
* config/rs6000/rs6000-c.c (struct altivec_builtin_types): Move to
rs6000.h.
(altivec_overloaded_builtins): Move to rs6000-call.c.
* config/rs6000/rs6000.h (struct altivec_builtin_types): Moved from
rs6000-c.c.
* config/rs6000/rs6000-call.c (rs6000_builtin_info): Make static.
(altivec_overloaded_builtins): Moved from rs6000-c.c.
(rs6000_common_init_builtins): Do no define builtins that overload
builtins that have been disabled.
Wilco Dijkstra [Wed, 4 Dec 2019 15:40:41 +0000 (15:40 +0000)]
[ARM] Improve max_cond_insns setting for Cortex cores
To enable cores to use the correct max_cond_insns setting, use the core-specific
tuning when a CPU/tune is selected unless -mrestrict-it is explicitly set.
On Cortex-A57 this gives 1.1% performance gain on SPECINT2006 as well as a
0.4% codesize reduction.
gcc/
* config/arm/arm.c (arm_option_override_internal):
Use max_cond_insns from CPU tuning unless -mrestrict-it is used.
Wilco Dijkstra [Wed, 4 Dec 2019 14:45:59 +0000 (14:45 +0000)]
[AArch64] Add support for fused compare and branch
Add support for fused compare with branch. Rename the existing
AARCH64_FUSE_CMP_BRANCH to ALU_BRANCH, and AARCH64_FUSE_ALU_BRANCH
to ALU_CBZ to make it clear what is being fused.
gcc/
* config/aarch64/aarch64.c
(thunderxt88_tunings): Use AARCH64_FUSE_ALU_BRANCH.
(thunderx_tunings): Likewise.
(tsv110_tunings): Use AARCH64_FUSE_ALU_BRANCH and AARCH64_FUSE_ALU_CBZ.
(thunderx2t99_tunings): Likewise.
(aarch_macro_fusion_pair_p): Add support for AARCH64_FUSE_CMP_BRANCH.
* config/aarch64/aarch64-fusion-pairs.def: Add ALU_CBZ fusion.
Richard Biener [Wed, 4 Dec 2019 13:21:39 +0000 (13:21 +0000)]
tree-ssa-sccvn.c (vn_reference_lookup_3): Properly guard empty CTOR and memset partial-def registering.
2019-12-04 Richard Biener <rguenther@suse.de>
* tree-ssa-sccvn.c (vn_reference_lookup_3): Properly guard
empty CTOR and memset partial-def registering. Take advantage
of fancy offset analysis in memset handling.
projects / thirdparty / gcc.git / log
