wilson [Fri, 1 Nov 2019 20:35:12 +0000 (20:35 +0000)]
RISC-V: Build soft-float divide routines for -mno-fdiv.
Using -mno-fdiv gives linker errors unless we build the missing divide
routines in libgcc always. There is at least one university project
designing RISC-V parts without FP divide that wants to use the option.
kargl [Fri, 1 Nov 2019 16:27:38 +0000 (16:27 +0000)]
2019-11-01 Steven G. Kargl <kargl@gcc.gnu.org>
PR fortran/90988
* decl.c (gfc_match_private, gfc_match_public): Fixed-form source code
does not require whitespace between PRIVATE (or PUBLIC) and an entity.
2019-11-01 Steven G. Kargl <kargl@gcc.gnu.org>
PR fortran/90988
* gfortran.dg/pr90988_4.f: New test.
rsandifo [Thu, 31 Oct 2019 17:27:02 +0000 (17:27 +0000)]
[AArch64] Split gcc.target/aarch64/sve/vcond_4*
vcond_4.c combined too much stuff into one test, so that when we needed
to add XFAILs for one set of routines, we lost testing of others that
used the same instructions. This patch splits it into four:
vcond_4.c:
The main test for selects between a vector and a scalar, with all
elements having the same size.
After Prathamesh's fix for PR91272, we are now guaranteed to use the
conditionally-loaded vector as the "then" value and the scalar as the
"else" value.
vcond_4_zero.c:
Like vcond_4.c, but comparing with zero.
vcond_4_sel.c:
Tests for selects between two scalars, with all elements having
the same size. The optimisers can legitimately switch the "then"
and "else" order and adjust the comparison to match.
vcond_4_costly.c:
Like vcnod_4.c, but with mixed element sizes. The sequences tested
here will not be the preferred ones once we support mixtures of vector
sizes, but it still tests an important code path.
This means that vcond_4_run.c now only tests what's left in vcond_4.c,
but that seems OK, since the main point was to ensure correct exception
behaviour.
Similarly it means that vcond_5.c only tests what's left in vcond_4.c,
but that too is OK, since the point of the test was to compare the
default handling of each comparison in vcond_4.c with the
-fno-trapping-math equivalent.
2019-10-31 Richard Sandiford <richard.sandiford@arm.com>
gcc/testsuite/
* gcc.target/aarch64/sve/vcond_4.c: Split parts out into...
* gcc.target/aarch64/sve/vcond_4_costly.c,
* gcc.target/aarch64/sve/vcond_4_sel.c,
* gcc.target/aarch64/sve/vcond_4_zero.c: ...these new tests.
* gcc.target/aarch64/sve/vcond_4_run.c: Adjust accordingly.
* gcc.target/aarch64/sve/vcond_5.c: Likewise.
This patch splits gcc.target/aarch64/sve/reduc_strict_3.c into one
test per function, so that it's easier to see what each scan is
matching and also so that we no longer rely on the number of times
that each dump message is printed.
The patch also generalises the tests to work with scalable vectors.
I think the test probably predates support for variable-length
loop-aware SLP.
2019-10-31 Richard Sandiford <richard.sandiford@arm.com>
gcc/testsuite/
* gcc.target/aarch64/sve/reduc_strict_3.c: Split all but the
first function out into...
* gcc.target/aarch64/sve/reduc_strict_4.c,
* gcc.target/aarch64/sve/reduc_strict_5.c,
* gcc.target/aarch64/sve/reduc_strict_6.c,
* gcc.target/aarch64/sve/reduc_strict_7.c,
* gcc.target/aarch64/sve/reduc_strict_8.c,
* gcc.target/aarch64/sve/reduc_strict_9.c: ...these new tests.
Test for scalable vectors instead of 256-bit vectors.
rsandifo [Thu, 31 Oct 2019 17:16:31 +0000 (17:16 +0000)]
[AArch64] Fix build for non-default languages
The SVE PCS support broke go, D and Ada because those languages don't
call TARGET_INIT_BUILTINS. We therefore ended up trying to get the
TYPE_MAIN_VARIANT of a null __SVBool_t.
We shouldn't really need to apply TYPE_MAIN_VARIANT there anyway,
since the ABI-defined types are (and need to be) their own main
variants. This patch asserts for that instead.
2019-10-31 Richard Sandiford <richard.sandiford@arm.com>
gcc/
* config/aarch64/aarch64-sve-builtins.cc (register_builtin_types):
Assert that the type we store in abi_vector_types is its own
main variant.
(svbool_type_p): Don't apply TYPE_MAIN_VARIANT here.
rearnsha [Thu, 31 Oct 2019 16:07:09 +0000 (16:07 +0000)]
[arm] Avoid using negative offsets for 'immediate' addresses when compiling for Thumb2
Thumb2 code now uses the Arm implementation of legitimize_address.
That code has a case to handle addresses that are absolute CONST_INT
values, which is a common use case in deeply embedded targets (eg:
void *p = (void*)0x12345678). Since thumb has very limited negative
offsets from a constant, we want to avoid forming a CSE base that will
then be used with a negative value.
This was reported upstream originally in
https://gcc.gnu.org/ml/gcc-help/2019-10/msg00122.html
rearnsha [Thu, 31 Oct 2019 16:04:53 +0000 (16:04 +0000)]
[arm] Pattern match insns for a + ~b + Carry
On ARM, the SBC instruction is defined as
Ra - Rb - ~C
where C is the carry flag. But -Rb = ~Rb + 1, so this is equivalent to
Ra + ~Rb + 1 - ~C
which then simplifies to
Ra + ~Rb + C
which is essentially an add-with-carry with one operand inverted. We
can define RTL patterns to match this. In thumb2 we can only match
when the operands are both registers, but in Arm state we can also use
RSC to match when Rn is either a constant or a shifted operand.
This overall simplifies some cases of 64-bit arithmetic, for example,
int64_t f (int64_t a, int64_t b) { return a + ~b; }
will now compile to
MVN R2, R2
ADDS R0, R0, R2
SBC R1, R1, R3
* config/arm/arm.md (add_not_cin): New insn.
(add_not_shift_cin): Likewise.
redi [Thu, 31 Oct 2019 13:17:48 +0000 (13:17 +0000)]
Remove PR 92268 workaround and fix new test failures
With the compiler bug fixed we can simplify the __sizable concept to use
a return-type-requirement again. I also realised it was redundantly
re-checking a subset of the sized_sentinel_for requirements.
The compiler fix also revealed bugs in two tests which started to fail
and are fixed by this patch.
* include/bits/range_access.h (__sizable): Rename to __sentinel_size.
Remove workaround for PR c++/92268 and remove redundant requirements
that are already checked by sized_sentinel_for.
* testsuite/std/ranges/access/cend.cc: Fix failures.
* testsuite/std/ranges/access/end.cc: Likewise.
redi [Thu, 31 Oct 2019 13:17:42 +0000 (13:17 +0000)]
Fix handling of arrays in range access customization points
This combines the overloads for arrays with the non-array overloads,
using if-constexpr to choose between the cases, and also enforces the
following:
- ADL should only be used for class types and enumeration types.
- ranges::begin should be ill-formed for rvalue arrays.
- ranges::end should be ill-formed for rvalue arrays, unbounded
arrays, and arrays of incomplete type.
- ranges::size should be ill-formed for unbounded arrays.
* include/bits/range_access.h (ranges::begin): Combine array and
non-array overloads into one function template. Only use ADL for
classes and enums
(ranges::end, ranges::size): Likewise. Make unbounded arrays
ill-formed.
(ranges::rbegin, ranges::rend): Only use ADL for classes and enums.
Reformat _S_noexcept() functions to mirror operator() structure.
* testsuite/std/ranges/access/begin.cc: Check incomplete array.
* testsuite/std/ranges/access/end_neg.cc: New test.
* testsuite/std/ranges/access/size.cc: Check array of incomplete type.
* testsuite/std/ranges/access/size_neg.cc: New test.
jakub [Thu, 31 Oct 2019 10:09:43 +0000 (10:09 +0000)]
* configure.ac: Compute and substitute omp_device_properties and
omp_device_property_deps.
* Makefile.in (generated_files): Add omp-device-properties.h.
(omp-general.o): Depend on omp-device-properties.h.
(omp_device_properties): New make variable.
(omp-device-properties.h, s-omp-device-properties-h,
install-omp-device-properties): New goals.
(install): Depend on install-omp-device-properties for accelerators.
* target.def (TARGET_OMP_DEVICE_KIND_ARCH_ISA): New target hook.
* target.h (enum omp_device_kind_arch_isa): New enum.
* doc/tm.texi.in: Add placeholder for TARGET_OMP_DEVICE_KIND_ARCH_ISA
documentation.
* omp-general.c: Include omp-device-properties.h.
(omp_max_simt_vf): Expect OFFLOAD_TARGET_NAMES to be separated by
colon instead of comma.
(omp_offload_device_kind_arch_isa, omp_maybe_offloaded): New
functions.
(omp_context_selector_matches): Implement device set arch/isa
selectors, improve device set kind selector handling.
* config/i386/i386-options.h (ix86_omp_device_kind_arch_isa): Declare.
* config/i386/i386.c (TARGET_SIMD_CLONE_ADJUST,
TARGET_SIMD_CLONE_USABLE): Formatting fix.
(TARGET_OMP_DEVICE_KIND_ARCH_ISA): Redefine to
ix86_omp_device_kind_arch_isa.
* config/i386/i386-options.c (struct ix86_target_opts): Move type
definition from ix86_target_string to file scope.
(isa2_opts, isa_opts): Moved arrays from ix86_target_string function
to file scope.
(ix86_omp_device_kind_arch_isa): New function.
(ix86_target_string): Moved struct ix86_target_opts, isa2_opts and
isa_opts definitions to file scope.
* config/i386/t-intelmic (omp-device-properties): New goal.
* config/nvptx/t-nvptx (omp-device-properties): Likewise.
* config/nvptx/nvptx.c (nvptx_omp_device_kind_arch_isa): New function.
(TARGET_OMP_DEVICE_KIND_ARCH_ISA): Redefine to
nvptx_omp_device_kind_arch_isa.
* configure: Regenerate.
* doc/tm.texi: Regenerate.
testsuite/
* c-c++-common/gomp/declare-variant-9.c: New test.
* c-c++-common/gomp/declare-variant-10.c: New test.
jakub [Thu, 31 Oct 2019 10:04:47 +0000 (10:04 +0000)]
PR middle-end/92231
* tree.h (fndecl_built_in_p): Use fndecl_built_in_p instead of
DECL_BUILT_IN in comment. Remove redundant ()s around return
argument.
* tree.c (free_lang_data_in_decl): Check if var is FUNCTION_DECL
before calling fndecl_built_in_p.
* gimple-fold.c (gimple_fold_stmt_to_constant_1): Check if
TREE_OPERAND (fn, 0) is a FUNCTION_DECL before calling
fndecl_built_in_p on it.
lto/
* lto-lang.c (handle_const_attribute): Don't call fndecl_built_in_p
on *node that is not FUNCTION_DECL.
testsuite/
* gcc.c-torture/compile/pr92231.c: New test.
avieira [Thu, 31 Oct 2019 09:49:47 +0000 (09:49 +0000)]
[vect]Make vect-epilogues-nomask=1 default
This patch turns epilogue vectorization on by default for all targets. The
following test strategy was chosen for this:
1) Disable epilogue vectorization for all tests that failed due to
scan-tree-dump failures for aarch64, arm and x86_64 inside:
- gcc.dg/vect
- gcc.target/i386
- gcc.target/aarch6
- gfortran.dg/vect
2) Add the Bugzilla's reported testcase but xfail the scan for
"EPILOGUE VECTORIZED" for arm*-*-* as that target does not vectorize the
epilogue.
gcc/ChangeLog:
2019-10-31 Andre Vieira <andre.simoesdiasvieira@arm.com>
* params.def (PARAM_VECT_EPILOGUES_NOMASK): Enable by default.
gcc/testsuite/ChangeLog:
2019-10-31 Andre Vieira <andre.simoesdiasvieira@arm.com>
jason [Thu, 31 Oct 2019 02:31:48 +0000 (02:31 +0000)]
PR c++/84810 - constraints on lambdas
Attached is a patch that adds parsing of the optional requires-clause in a
lambda-expression and lambda-declarator. Additionally, shorthand constraints
from the template-parameter-list are now actually applied and constrain the
synthesized operator().
Previously we were not parsing the requires clauses at all and not saving
the shorthand constraints in the place expected by grokfndecl.
The trailing requires-clause is now also used to suppress synthesis of the
conversion to function pointer for non-capturing non-generic lambdas as per
expr.prim.lambda.closure/7.
This includes a fix to template_class_depth. Previously it was computing the
wrong depth for lambdas in the initializer of a static member of a class
template, exhibited by the concepts-lambda4 test which currently fails on
trunk. The bug was causing grokfndecl to use the constraints from the
template class for the lambda.
gcc/cp/
2019-10-30 Jeff Chapman II <jchapman@lock3software.com>
PR c++/84810 - constraints on lambdas
* lambda.c (maybe_add_lambda_conv_op): Do not synthesize
conversion if the call operator does not satisfy its constraints.
* parser.c (cp_parser_lambda_declarator_opt): Parse
requires-clause on generic lambdas; combine with shorthand
constraints. Parse trailing requires-clause and attach to the
synthesized call operator.
* pt.c (template_class_depth): Only inspect
LAMBDA_TYPE_EXTRA_SCOPE if it is present. This fixes an
incorrect depth calculation for lambdas inside the initializer
of a static data member of a template class.
gcc/testsuite/
2019-10-30 Jeff Chapman II <jchapman@lock3software.com>
PR c++/84810 - constraints on lambdas
* g++.dg/cpp2a/concepts-lambda2.C: New test.
* g++.dg/cpp2a/concepts-lambda3.C: Ditto.
* g++.dg/cpp2a/concepts-lambda4.C: Ditto.
* g++.dg/cpp2a/concepts-pr84810.C: Ditto.
jason [Thu, 31 Oct 2019 02:01:16 +0000 (02:01 +0000)]
PR c++/92268 - hard error satisfying return-type-requirement
Previously we would put the template arguments for the concept-check in a
TEMPLATE_ID and then also pass them to constraints_satisfied_p, which meant
that we would try to normalize the concept-check with the fully instantiated
arguments, leading to sadness. Simply not passing the args to
constraints_satisfied_p fixes the problem.
I also noticed that we weren't detecting substitution failure in the
constraints, but were silently treating it as success.
* constraint.cc (type_deducible_p): Check for substitution failure.
(diagnose_compound_requirement): Adjust diagnostic.
* pt.c (do_auto_deduction): Don't pass cargs to
constraints_satisfied_p.
redi [Wed, 30 Oct 2019 17:42:04 +0000 (17:42 +0000)]
Use if-constexpr instead of overloading for customization point
This combines two of the std::ranges::swap.operator() overloads into a
single function template. Using if-constexpr to choose between
implementations should give the compiler less work to do than using
overloading.
* include/std/concepts (std::ranges::swap): Use a single overload for
the non-array cases, and switch using if-constexpr.
redi [Wed, 30 Oct 2019 15:48:23 +0000 (15:48 +0000)]
Remove some more using-declarations from namespace __gnu_cxx
Similar to some recent patches, this removes using-declarations for
names from namespace std, so that they are not redeclared in __gnu_cxx.
* include/bits/stl_iterator.h (namespace __gnu_cxx): Remove
using-declarations for std::iterator and std::iterator_traits.
(__gnu_cxx::__normal_iterator): Qualify iterator_traits.
* include/ext/algorithm (namespace __gnu_cxx): Remove
using-declarations for std names and qualify those names when used.
Also refer to std::min in parentheses to protect against function-like
macros.
* include/ext/rc_string_base.h: Qualify iterator_traits.
* include/ext/sso_string_base.h: Qualify iterator_traits.
redi [Wed, 30 Oct 2019 15:48:11 +0000 (15:48 +0000)]
Apply C++20 changes to various iterator types
This ensures that __normal_iterator<T*, C> satisfies the
contiguous_iterator concept, by defining the iterator_concept member
type.
Also update vector<bool>'s iterators, reverse_iterator,
istreambuf_iterator and ostreambuf_iterator to meet the C++20
requirements.
PR libstdc++/92272
* include/bits/stl_bvector.h (_Bit_iterator::pointer)
(_Bit_const_iterator::pointer): Define as void for C++20.
* include/bits/stl_iterator.h (reverse_iterator::operator->()): Add
constraints for C++20.
(__normal_iterator::iterator_concept): Define for C++20.
* include/bits/streambuf_iterator.h (istreambuf_iterator::pointer):
Define as void for C++20.
(ostreambuf_iterator::difference_type): Define as ptrdiff_t for C++20.
(ostreambuf_iterator::ostreambuf_iterator()): Add default constructor
for C++20.
* testsuite/23_containers/vector/bool/iterator_c++20.cc: New test.
* testsuite/24_iterators/bidirectional/concept.cc: New test.
* testsuite/24_iterators/bidirectional/tag.cc: New test.
* testsuite/24_iterators/contiguous/concept.cc: New test.
* testsuite/24_iterators/contiguous/tag.cc: New test.
* testsuite/24_iterators/forward/concept.cc: New test.
* testsuite/24_iterators/forward/tag.cc: New test.
* testsuite/24_iterators/input/concept.cc: New test.
* testsuite/24_iterators/input/tag.cc: New test.
* testsuite/24_iterators/istreambuf_iterator/requirements/typedefs.cc:
New test.
* testsuite/24_iterators/ostreambuf_iterator/requirements/typedefs.cc:
New test.
* testsuite/24_iterators/output/concept.cc: New test.
* testsuite/24_iterators/output/tag.cc: New test.
* testsuite/24_iterators/random_access/concept.cc: New test.
* testsuite/24_iterators/random_access/tag.cc: New test.
* testsuite/24_iterators/range_operations/advance_debug_neg.cc: New
test.
* testsuite/24_iterators/random_access_iterator/26020.cc: Move to ...
* testsuite/24_iterators/operations/26020.cc: ... here.
* testsuite/24_iterators/random_access_iterator/
string_vector_iterators.cc: Move to ...
* testsuite/24_iterators/random_access/string_vector_iterators.cc: ...
here.
jozefl [Wed, 30 Oct 2019 14:15:02 +0000 (14:15 +0000)]
2019-10-30 Jozef Lawrynowicz <jozef.l@mittosystems.com>
* config/msp430/msp430.c (msp430_expand_helper): Support expansion of
calls to __mspabi_mpy* functions.
* config/msp430/msp430.md (mulhisi3): New define_expand.
(umulhisi3): New define_expand.
(*mulhisi3_inline): Use old mulhisi3 define_insn.
(*umulhisi3_inline): Use old umulhisi3 define_insn.
jozefl [Wed, 30 Oct 2019 14:01:12 +0000 (14:01 +0000)]
gcc/ChangeLog:
2019-10-30 Jozef Lawrynowicz <jozef.l@mittosystems.com>
* config/msp430/msp430.c (msp430_check_index_not_high_mem): New.
(msp430_check_plus_not_high_mem): New.
(msp430_op_not_in_high_mem): Use new functions to check if the operand
might be in low memory.
Indicate that a 16-bit absolute address is in lower memory.
gcc/testsuite/ChangeLog:
2019-10-30 Jozef Lawrynowicz <jozef.l@mittosystems.com>
* gcc.target/msp430/mlarge-use-430-insn.c: New test.
marxin [Wed, 30 Oct 2019 12:03:00 +0000 (12:03 +0000)]
Remove alias set comparison.
2019-10-30 Martin Liska <mliska@suse.cz>
* ipa-icf-gimple.c (func_checker::compatible_types_p):
Do not compare alias sets. It's handled by operand_equal_p.
2019-10-30 Martin Liska <mliska@suse.cz>
marxin [Wed, 30 Oct 2019 12:02:17 +0000 (12:02 +0000)]
Remove comparison for polymorphic types.
2019-10-30 Martin Liska <mliska@suse.cz>
* ipa-icf-gimple.c (func_checker::func_checker): Do not
initialize m_compare_polymorphic.
(func_checker::compare_decl): Do not compare polymorphic types.
* ipa-icf-gimple.h (m_compare_polymorphic): Remove.
* ipa-icf.c (sem_function::equals_private): Do not call
compare_polymorphic_p.
jakub [Wed, 30 Oct 2019 11:58:03 +0000 (11:58 +0000)]
* cp-tree.h (omp_declare_variant_finalize, build_local_temp): Declare.
* decl.c: Include omp-general.h.
(declare_simd_adjust_this): Add forward declaration.
(omp_declare_variant_finalize_one, omp_declare_variant_finalize): New
function.
(cp_finish_decl, finish_function): Call omp_declare_variant_finalize.
* parser.c (cp_finish_omp_declare_variant): Adjust parsing of the
variant id-expression and propagate enough information to
omp_declare_variant_finalize_one in the attribute so that it can
finalize it.
* class.c (finish_struct): Call omp_declare_variant_finalize.
* tree.c (build_local_temp): No longer static, remove forward
declaration.
* c-c++-common/gomp/declare-variant-2.c: Add a test with , before
match clause.
* c-c++-common/gomp/declare-variant-6.c: Expect diagnostics also from
C++ FE and adjust regexp so that it handles C++ pretty printing of
function names.
* g++.dg/gomp/declare-variant-1.C: New test.
* g++.dg/gomp/declare-variant-2.C: New test.
* g++.dg/gomp/declare-variant-3.C: New test.
* g++.dg/gomp/declare-variant-4.C: New test.
* g++.dg/gomp/declare-variant-5.C: New test.
paolo [Wed, 30 Oct 2019 11:49:21 +0000 (11:49 +0000)]
/cp
2019-10-30 Paolo Carlini <paolo.carlini@oracle.com>
* typeck.c (cp_build_modify_expr): Prefer error + inform to
error + error in one place.
(get_delta_difference_1): Likewise.
(get_delta_difference): Likewise, in two places.
/testsuite
2019-10-30 Paolo Carlini <paolo.carlini@oracle.com>
* g++.dg/conversion/ptrmem2.C: Adjust for error + inform.
* g++.dg/gomp/tpl-atomic-2.C: Likewise.
marxin [Wed, 30 Oct 2019 10:38:52 +0000 (10:38 +0000)]
Use symtab_node::order in LTO sections with body.
2019-10-30 Martin Liska <mliska@suse.cz>
PR lto/91393
PR lto/88220
* cgraph.c (cgraph_node::get_create): Overwrite node->order
from a first_clone in order to get proper LTO section
in LTO stream.
(cgraph_node::get_untransformed_body):
Use lto_get_section_data where symtab_node::order
must be provided.
* cgraphclones.c (cgraph_node::find_replacement):
Update also symbol order.
* ipa-fnsummary.c (ipa_fn_summary_read):
Use new function lto_get_summary_section_data.
* ipa-hsa.c (ipa_hsa_read_summary): Likewise.
* ipa-icf.c (sem_item_optimizer::read_summary):
Likewise.
* ipa-prop.c (ipa_prop_read_jump_functions):
Likewise.
(ipcp_read_transformation_summaries): Likewise.
* ipa-sra.c (ipa_sra_read_summary): Likewise.
* lto-cgraph.c (input_node): Add also order_base.
(input_varpool_node): Likewise.
(input_cgraph_1): Assign the order_base.
(input_cgraph_opt_summary): Use new lto_get_summary_section_data.
* lto-opts.c (lto_write_options): Pass new argument.
* lto-section-in.c (lto_get_section_data): Add new argumente order.
(lto_get_summary_section_data): New.
(lto_get_raw_section_data): Add order argument.
(lto_create_simple_input_block): Likewise.
* lto-section-out.c (lto_destroy_simple_output_block):
Likewise.
* lto-streamer-in.c (lto_input_toplevel_asms):
Use lto_get_summary_section_data.
(lto_input_mode_table): Likewise.
* lto-streamer-out.c (produce_asm): Pass symtab_node::order.
(lto_output_toplevel_asms): Pass new argument.
(copy_function_or_variable): Likewise.
(produce_lto_section):Likewise.
(produce_symtab): Likewise.
(lto_write_mode_table): Likewise.
(produce_asm_for_decls): Likewise.
* lto-streamer.c (lto_get_section_name): Concat symbol name
and symbol order.
* lto-streamer.h (lto_get_section_data): Add order argument.
(lto_get_summary_section_data): New.
(lto_get_raw_section_data): Add order argument.
(lto_get_section_name): Likewise.
* varpool.c (varpool_node::get_constructor): Pass order argument.
2019-10-30 Martin Liska <mliska@suse.cz>
PR lto/91393
PR lto/88220
* lto-common.c (lto_file_finalize): Use lto_get_summary_section_data.
(get_section_data): Add order argument.
2019-10-30 Martin Liska <mliska@suse.cz>
PR lto/91393
PR lto/88220
* gcc.dg/lto/pr91393_0.c: New test.
jakub [Wed, 30 Oct 2019 09:52:01 +0000 (09:52 +0000)]
PR tree-optimization/92262
* tree-ssa-loop-ivopts.c (get_debug_computation_at): Don't unshare
ubase or cbase here.
(remove_unused_ivs): Unshare comp before using it.
rguenth [Wed, 30 Oct 2019 09:21:09 +0000 (09:21 +0000)]
2019-10-30 Richard Biener <rguenther@suse.de>
PR tree-optimization/65930
* tree-vect-loop.c (vect_is_simple_reduction): For reduction
chains also allow a leading and trailing conversion.
* tree-vect-slp.c (vect_get_and_check_slp_defs): Handle
intermediate reduction chains.
(vect_analyze_slp_instance): Likewise. Build a SLP
node for a trailing conversion manually.
markeggleston [Wed, 30 Oct 2019 08:37:29 +0000 (08:37 +0000)]
Suppress warning with -Wno-overwrite-recursive.
The use of -fno-automatic with -frecursive results in a warning implying
that recursion will not work. If all relevant local variable have the
automatic attribute explicitly declared recursion does work and the warning
is redundant.
mpolacek [Tue, 29 Oct 2019 21:06:21 +0000 (21:06 +0000)]
PR c++/90998 - ICE with copy elision in init by ctor and -Wconversion.
After r269667 which introduced joust_maybe_elide_copy, in C++17 we can elide
a constructor if it uses a conversion function that returns a prvalue, and
use the conversion function in its stead.
This eliding means that if we have a candidate that previously didn't have
->second_conv, it can have it after the elision. This confused the
-Wconversion warning because it was assuming that if cand1->second_conv is
non-null, so is cand2->second_conv. Here cand1->second_conv was non-null
but cand2->second_conv remained null, so it crashed in compare_ics.
I checked with clang that both compilers call A::operator B() in C++17 and
B::B(A const &) otherwise.
* call.c (joust): Don't attempt to warn if ->second_conv is null.
mpolacek [Tue, 29 Oct 2019 20:34:43 +0000 (20:34 +0000)]
PR c++/91548 - fix detecting modifying const objects for ARRAY_REF.
This fixes a bogus "modifying a const object" error for an array that actually
isn't declared const. The problem was how I handled ARRAY_REFs here; we
shouldn't look at the ARRAY_REF itself, but at the array its accessing.
* constexpr.c (cxx_eval_store_expression): Don't call
modifying_const_object_p for ARRAY_REF.
* g++.dg/cpp1y/constexpr-tracking-const15.C: New test.
* g++.dg/cpp1y/constexpr-tracking-const16.C: New test.
* g++.dg/cpp1z/constexpr-tracking-const1.C: New test.
redi [Tue, 29 Oct 2019 17:15:04 +0000 (17:15 +0000)]
Minor improvements to testsuite iterator utilities
* testsuite/util/testsuite_iterators.h (BoundsContainer::size()): Add
new member function.
(WritableObject::operator=): Constrain with enable_if when available.
(remove_cv): Use std::remove_if when available.
(test_container::it(int)): Use size().
(test_container::size()): Use BoundsContainer::size().
redi [Tue, 29 Oct 2019 17:14:55 +0000 (17:14 +0000)]
PR libstdc++/92267 fix ABI change in deque iterators
Defaulting the copy constructor on its first declaration made it change
from user-provided (and non-trivial) to implicitly-defined (and
trivial). This caused an ABI incompatibility between GCC 8 and GCC 9,
where functions taking a deque iterator disagree on the argument passing
convention.
PR libstdc++/92267
* include/bits/stl_deque.h (_Deque_iterator(const _Deque_iterator&)):
Do not define as defaulted.
* testsuite/23_containers/deque/types/92267.cc: New test.
avieira [Tue, 29 Oct 2019 13:15:46 +0000 (13:15 +0000)]
[vect]PR 88915: Vectorize epilogues when versioning loops
gcc/ChangeLog:
2019-10-29 Andre Vieira <andre.simoesdiasvieira@arm.com>
PR 88915
* tree-ssa-loop-niter.h (simplify_replace_tree): Change declaration.
* tree-ssa-loop-niter.c (simplify_replace_tree): Add context parameter
and make the valueize function pointer also take a void pointer.
* gcc/tree-ssa-sccvn.c (vn_valueize_wrapper): New function to wrap
around vn_valueize, to call it without a context.
(process_bb): Use vn_valueize_wrapper instead of vn_valueize.
* tree-vect-loop.c (_loop_vec_info): Initialize epilogue_vinfos.
(~_loop_vec_info): Release epilogue_vinfos.
(vect_analyze_loop_costing): Use knowledge of main VF to estimate
number of iterations of epilogue.
(vect_analyze_loop_2): Adapt to analyse main loop for all supported
vector sizes when vect-epilogues-nomask=1. Also keep track of lowest
versioning threshold needed for main loop.
(vect_analyze_loop): Likewise.
(find_in_mapping): New helper function.
(update_epilogue_loop_vinfo): New function.
(vect_transform_loop): When vectorizing epilogues re-use analysis done
on main loop and call update_epilogue_loop_vinfo to update it.
* tree-vect-loop-manip.c (vect_update_inits_of_drs): No longer insert
stmts on loop preheader edge.
(vect_do_peeling): Enable skip-vectors when doing loop versioning if
we decided to vectorize epilogues. Update epilogues NITERS and
construct ADVANCE to update epilogues data references where needed.
* tree-vectorizer.h (_loop_vec_info): Add epilogue_vinfos.
(vect_do_peeling, vect_update_inits_of_drs,
determine_peel_for_niter, vect_analyze_loop): Add or update
declarations.
* tree-vectorizer.c (try_vectorize_loop_1): Make sure to use already
created loop_vec_info's for epilogues when available. Otherwise analyse
epilogue separately.
rguenth [Tue, 29 Oct 2019 09:33:49 +0000 (09:33 +0000)]
2019-10-29 Richard Biener <rguenther@suse.de>
PR tree-optimization/65930
* tree-vect-loop.c (check_reduction_path): Relax single-use
check allowing out-of-loop uses.
(vect_is_simple_reduction): SLP reduction chains cannot have
intermediate stmts used outside of the loop.
(vect_create_epilog_for_reduction): The adjustment might need
to be converted.
(vectorizable_reduction): Annotate live stmts of the reduction
chain with STMT_VINFO_REDUC_DEF.
* tree-vect-stms.c (process_use): Remove no longer true asserts.
projects / thirdparty / gcc.git / log
