* trans.c (gfc_allocate_using_lib,gfc_deallocate_with_status):
Introducing __asm__ __volatile__ ("":::"memory")
after image control statements.
* trans-stmt.c (gfc_trans_sync, gfc_trans_event_post_wait,
gfc_trans_lock_unlock, gfc_trans_critical): Ditto.
* trans-intrinsic.c (gfc_conv_intrinsic_caf_get,
conv_caf_send): Introducing __asm__ __volatile__ ("":::"memory")
after send, before get and around sendget.
Marek Polacek [Wed, 9 Dec 2015 14:58:22 +0000 (14:58 +0000)]
re PR tree-optimization/66949 (ICE at -O2 and -O3 on x86_64-linux-gnu in tree_ssa_phiopt_worker, at tree-ssa-phiopt.c:335)
PR tree-optimization/66949
* tree-ssa-phiopt.c (tree_ssa_phiopt_worker): Don't call
single_non_singleton_phi_for_edges to get the PHI from
factor_out_conditional_conversion. Use NULL_TREE instead of NULL.
(factor_out_conditional_conversion): Adjust declaration. Make it
return the newly-created PHI.
* gcc.dg/torture/pr66949-1.c: New test.
* gcc.dg/torture/pr66949-2.c: New test.
Jakub Jelinek [Wed, 9 Dec 2015 13:42:06 +0000 (14:42 +0100)]
re PR tree-optimization/68786 (Aligned masked store is generated for unaligned pointer)
PR tree-optimization/68786
* tree-if-conv.c: Include builtins.h.
(predicate_mem_writes): Put result of get_object_alignment (ref)
into second argument's value.
* tree-vect-stmts.c (vectorizable_mask_load_store): Put minimum
pointer alignment into second argument's value.
* tree-data-ref.c (get_references_in_stmt): Use value of second
argument for build_aligned_type, and only the type to build
a zero second argument for MEM_REF.
* internal-fn.c (expand_mask_load_optab_fn,
expand_mask_store_optab_fn): Likewise.
Richard Biener [Wed, 9 Dec 2015 13:30:41 +0000 (13:30 +0000)]
re PR tree-optimization/68583 (Missed if-conversion)
2015-12-09 Richard Biener <rguenther@suse.de>
PR tree-optimization/68583
* tree-if-conv.c (if_convertible_phi_p): Drop
flag_tree_loop_if_convert_stores check in favor of the
existing any_mask_load_store check.
(insert_gimplified_predicates): Likewise.
(combine_blocks): Likewise.
(tree_if_conversion): Likewise.
(ifcvt_memrefs_wont_trap): Properly check
flag_tree_loop_if_convert_stores in all places that can end
up introducing store-data-races.
(if_convertible_gimple_assign_stmt_p): Remove restriction
on flag_tree_loop_if_convert_stores for stores we can if-convert
without introducing store-data-races. Force versioning for
all if-converted stores.
Richard Biener [Wed, 9 Dec 2015 08:23:37 +0000 (08:23 +0000)]
tree-vect-stmts.c (vectorizable_load): Set new vinfo only if it was not yet set.
2015-12-09 Richard Biener <rguenther@suse.de>
* tree-vect-stmts.c (vectorizable_load): Set new vinfo only
if it was not yet set.
* tree-vectorizer.h (set_vinfo_for_stmt): Assert we don't
overwrite an existing entry.
Jan Hubicka [Wed, 9 Dec 2015 02:15:05 +0000 (03:15 +0100)]
re PR lto/61886 (LTO breaks fread with _FORTIFY_SOURCE=2)
PR ipa/61886
* lto-streamer.h (lto_symtab_merge_decls, lto_symtab_merge_symbols,
lto_symtab_prevailing_decl): MOve to lto-symtab.h.
* lto-streamer-out.c (DFS::DFS_write_tree_body): Check that
DECL_ABSTRACT_ORIGIN is not error_mark_node.
* lto-symtab.c: Include lto-symtab.h.
(lto_cgraph_replace_node): Do not merge profiles here.
(lto_symtab_merge_p): New function.
(lto_symtab_merge_decls_2): Honor lto_symtab_merge_p.
(lto_symtab_merge_symbols_1): Turn unmerged decls into transparent
aliases.
(lto_symtab_merge_symbols): Do not clear node->aux; we no longer use it.
(lto_symtab_prevailing_decl): Move to lto-symtab.h; rewrite.
* lto.c: Include lto-symtab.h
* lto-symtab.h: New.
Jan Hubicka [Tue, 8 Dec 2015 22:02:23 +0000 (23:02 +0100)]
re PR lto/61886 (LTO breaks fread with _FORTIFY_SOURCE=2)
PR ipa/61886
* varpool.c (varpool_node::get_availability): Recurse only on
weakrefs with definition in the target.
(symbol_table::remove_unreferenced_decls): Keep aliases in the boundary.
Jan Hubicka [Tue, 8 Dec 2015 22:01:41 +0000 (23:01 +0100)]
re PR lto/61886 (LTO breaks fread with _FORTIFY_SOURCE=2)
PR ipa/61886
* ipa-visibility.c (can_replace_by_local_alias): Look through transparent
aliaes; refuse weakrefs.
(update_visibility_by_resolution_info): Skip transparent aliases in the
analysis part
Jan Hubicka [Tue, 8 Dec 2015 22:00:55 +0000 (23:00 +0100)]
re PR lto/61886 (LTO breaks fread with _FORTIFY_SOURCE=2)
PR ipa/61886
* symtab.c (symtab_node::verify_base): Fix thinko in a conditional.
(symtab_node::noninterposable_alias): Do not accept transparent
aliases.
Jan Hubicka [Tue, 8 Dec 2015 20:46:42 +0000 (21:46 +0100)]
re PR lto/61886 (LTO breaks fread with _FORTIFY_SOURCE=2)
PR ipa/61886
* lto-partition.c (add_symbol_to_partition_1): Transparent aliases
are not part of the definition.
(contained_in_symbol): Likewise.
(promote_symbol): When promoting a symbol also promote all transparent
aliases.
(rename_statics): Weakref needs unique name, too.
PR target/68701
* config/i386/i386.c (ix86_option_override_internal): Enable
-maccumulate-outgoing-args when %ebp is fixed due to stack
realignment requirements.
testsuite/ChangeLog:
PR target/68701
* testsuite/gcc.target/i386/pr68701-1.c: New test.
* testsuite/gcc.target/i386/pr68701-2.c: Ditto.
David Malcolm [Tue, 8 Dec 2015 17:03:26 +0000 (17:03 +0000)]
C: fix uninitialized ranges for __alignof__
gcc/c/ChangeLog:
* c-parser.c (c_parser_alignof_expression): Capture location of
closing parenthesis (if any), or of end of unary expression, and
use it to build a src_range for the expression.
gcc/testsuite/ChangeLog:
* gcc.dg/plugin/diagnostic-test-expressions-1.c (test_alignof):
New test function.
David Malcolm [Tue, 8 Dec 2015 16:57:27 +0000 (16:57 +0000)]
PR c/68757: fix uninitialized src_range for various builtins
gcc/c/ChangeLog:
PR c/68757
* c-parser.c (c_parser_get_builtin_args): Add
"out_close_paren_loc" param, and write back to it.
(c_parser_postfix_expression): Capture the closing
parenthesis location for RID_CHOOSE_EXPR,
RID_BUILTIN_CALL_WITH_STATIC_CHAIN, RID_BUILTIN_COMPLEX,
RID_BUILTIN_SHUFFLE and use it to set the source range
for such expressions; within RID_BUILTIN_COMPLEX set
the underlying location.
gcc/testsuite/ChangeLog:
PR c/68757
* gcc.dg/plugin/diagnostic-test-expressions-1.c
(test_builtin_choose_expr): New test function.
(test_builtin_call_with_static_chain): Likewise.
(test_builtin_complex): Likewise.
(test_builtin_shuffle): Likewise.
* lib/target-supports.exp
(check_effective_target_arm_vfp_ok_nocache): New.
(check_effective_target_arm_vfp_ok): Call the new
check_effective_target_arm_vfp_ok_nocache function.
(check_effective_target_arm_fp_ok_nocache): New.
(check_effective_target_arm_fp_ok): New.
(add_options_for_arm_fp): New.
(check_effective_target_arm_crypto_ok_nocache): Require
target_arm_v8_neon_ok instead of arm32.
(check_effective_target_arm_crypto_pragma_ok_nocache): New.
(check_effective_target_arm_crypto_pragma_ok): New.
(add_options_for_arm_vfp): New.
* gcc.target/arm/attr-crypto.c: Use arm_crypto_pragma_ok effective
target. Do not force -mfloat-abi=softfp, use arm_vfp effective
target instead.
* gcc.target/arm/attr-neon-builtin-fail.c: Do not force
-mfloat-abi=softfp, use arm_fp effective target instead.
* gcc.target/arm/attr-neon-fp16.c: Likewise. Remove arm_neon_ok
dependency.
* gcc.target/arm/attr-neon2.c: Do not force -mfloat-abi=softfp,
use arm_vfp effective target instead.
* gcc.target/arm/attr-neon3.c: Likewise.
Steve Ellcey [Mon, 7 Dec 2015 20:37:38 +0000 (20:37 +0000)]
reorg.c (optimize_skip): Do not put frame related instructions in annulled delay slots.
2015-12-07 Steve Ellcey <sellcey@imgtec.com>
* reorg.c (optimize_skip): Do not put frame related instructions
in annulled delay slots.
(steal_delay_list_from_target): Ditto.
(fill_slots_from_thread): Ditto.
Jason Merrill [Mon, 7 Dec 2015 19:34:11 +0000 (14:34 -0500)]
re PR c++/68464 (ICE in valid constexpr function: ../../src/gcc/tree.c:11497)
PR c++/68464
* cp-gimplify.c (cp_fold): Don't assume X has TREE_TYPE.
(cp_genericize): Don't do cp_fold_r here.
(cp_fold_function): New.
* cp-tree.h: Declare it.
* decl.c (finish_function): Call it and the pre-genericize plugin
before NRV processing.
Jason Merrill [Mon, 7 Dec 2015 19:34:04 +0000 (14:34 -0500)]
re PR c++/68170 (Declaring friend template class template in C++1z produces error: specialization of ‘template<class T> class A’ must appear at namespace)
PR c++/68170
* pt.c (maybe_new_partial_specialization): The injected-class-name
is not a new partial specialization.
Jason Merrill [Mon, 7 Dec 2015 19:33:59 +0000 (14:33 -0500)]
g++.exp: Also find -std list in GXX_TESTSUITE_STDS environment variable.
gcc/testsuite/
* lib/g++.exp: Also find -std list in GXX_TESTSUITE_STDS
environment variable.
gcc/cp/
* Make-lang.in (check-c++1z, check-c++-all): Use GXX_TESTSUITE_STDS.
Evandro Menezes [Mon, 7 Dec 2015 19:30:01 +0000 (19:30 +0000)]
Add scheduling model for Exynos M1
gcc/
* config/aarch64/aarch64-cores.def: Use the Exynos M1 sched model.
* config/aarch64/aarch64.md: Include "exynos-m1.md".
* config/arm/arm.md: Likewise.
* config/arm/exynos-m1.md: New file.
Eric Botcazou [Mon, 7 Dec 2015 16:45:45 +0000 (16:45 +0000)]
re PR middle-end/68291 (ICE in emit_move_insn, at expr.c:3540)
PR middle-end/68291
PR middle-end/68292
* cfgexpand.c (set_rtl): Always accept PARALLELs with BLKmode for
SSA names based on RESULT_DECLs.
* function.c (expand_function_start): Do not create BLKmode REGs
for GIMPLE registers when coalescing is enabled.
David Malcolm [Mon, 7 Dec 2015 16:07:00 +0000 (16:07 +0000)]
Fix missing range information for "%q+D" format code
gcc/c-family/ChangeLog:
* c-common.c (c_cpp_error): Update for change to
rich_location::set_range.
gcc/fortran/ChangeLog:
* error.c (gfc_format_decoder): Update for change of
text_info::set_range to text_info::set_location.
gcc/ChangeLog:
* pretty-print.c (text_info::set_range): Rename to...
(text_info::set_location): ...this, converting 2nd param
from source_range to a location_t.
* pretty-print.h (text_info::set_location): Convert
from inline function to external definition.
(text_info::set_range): Delete.
gcc/testsuite/ChangeLog:
* gcc.dg/diagnostic-ranges-1.c: New test file.
* gcc.dg/plugin/diagnostic-test-show-locus-bw.c
(test_percent_q_plus_d): New test function.
* gcc.dg/plugin/diagnostic_plugin_test_show_locus.c
(test_show_locus): Rewrite test code using
rich_location::set_range. Add code to unit-test the "%q+D"
format code.
libcpp/ChangeLog:
* include/line-map.h (rich_location::set_range): Add line_maps *
param; convert param from source_range to source_location. Drop
"overwrite_loc_p" param.
* line-map.c (rich_location::set_range): Likewise, acting as if
"overwrite_loc_p" were true, and getting range from the location.
Kirill Yukhin [Mon, 7 Dec 2015 11:12:41 +0000 (11:12 +0000)]
re PR target/68627 ([i386, AVX-512] Illegal insn generated while compiling spec2k6/437.leslie3d for KNL)
PR target/68627
gcc/
* config/i386/sse.md: Make 'v' alternative work on 'avx512f' ISA only.
Force destination to 512 bits register.
gcc/testsuite/
* gfortran.dg/pr68627.f: New test.
Jason Merrill [Mon, 7 Dec 2015 04:35:08 +0000 (23:35 -0500)]
Fix decltype-call1.C with -std=c++1z.
* parser.h (struct cp_token): Tell GTY that CPP_DECLTYPE uses
tree_check_value.
* parser.c (cp_parser_decltype): Use tree_check_value.
(saved_checks_value): New.
(cp_parser_nested_name_specifier_opt): Use it.
(cp_parser_template_id): Use it.
(cp_parser_simple_type_specifier): Use it.
(cp_parser_pre_parsed_nested_name_specifier): Use it.
Jason Merrill [Mon, 7 Dec 2015 04:34:57 +0000 (23:34 -0500)]
Fix template/ref1.C, nontype15.C, addr-builtin1.C with -std=c++1z.
* parser.c (cp_parser_template_argument): Handle references in
C++1z mode.
* constexpr.c (potential_constant_expression_1): Don't error about
TREE_THIS_VOLATILE on declarations.
[COMPONENT_REF]: Don't consider the object if we're dealing with an
overloaded function.
Jason Merrill [Mon, 7 Dec 2015 04:34:43 +0000 (23:34 -0500)]
PR c++/68597, fix auto9.C and auto-neg1.C with -std=c++1z.
* decl.c (check_tag_decl): Use ds_type_spec in auto diagnostic.
* typeck.c (check_return_expr): Check for inconsistent deduction.
* parser.c (class type_id_in_expr_sentinel): New.
(cp_parser_primary_expression) [RID_VA_ARG]: Use it.
(cp_parser_new_expression): Use it.
(cp_parser_trait_expr): Use it.
(cp_parser_type_id_1): Complain about auto if in_type_id_in_expr_p.
(cp_parser_default_type_template_argument): Check for auto.
(cp_parser_type_id_list): Likewise.
(cp_parser_simple_type_specifier): Allow auto parms if flag_concepts.
* pt.c (do_auto_deduction): Handle erroneous type.
Jason Merrill [Mon, 7 Dec 2015 04:34:36 +0000 (23:34 -0500)]
static_assert8.C: Fix for -std=c++1z.
* g++.dg/cpp0x/static_assert8.C: Fix for -std=c++1z.
* g++.dg/cpp0x/auto3.C: Fix for -std=c++1z.
* g++.dg/cpp1y/feat-cxx14.C: Only run with -std=c++14.
* g++.dg/cpp1y/feat-cxx14-neg.C: Only run with -std=c++14.
* g++.dg/cpp1z/feat-cxx1z.C: Fill in from feat-cxx14.C.
* g++.dg/cpp1z/fold5.C: Don't run with -std=c++1z.
* g++.dg/template/function1.C: Don't run with -std=c++1z.
* g++.dg/template/call4.C: Fix access.
* g++.dg/parse/crash31.C: Remove expected error text.