This re-applies r258525, and this time adds it to LOCAL_PATCHES.
libsanitizer/
* LOCAL_PATCHES: Add r258525.
* sanitizer_common/sanitizer_stacktrace.cc
(BufferedStackTrace::FastUnwindStack): Use the correct frame offset
for PowerPC SYSV ABI.
Richard Biener [Mon, 5 Nov 2018 14:55:53 +0000 (14:55 +0000)]
re PR tree-optimization/87873 (ICE: verify_gimple failed (error: incompatible types in PHI argument 0))
2018-11-05 Richard Biener <rguenther@suse.de>
PR tree-optimization/87873
* tree-ssa-loop-manip.h (split_loop_exit_edge): Add copy_constants_p
argument.
* tree-ssa-loop-manip.c (split_loop_exit_edge): Likewise.
* tree-vect-loop.c (vect_transform_loop): When splitting the
loop exit also create forwarder PHIs for constants.
* tree-vect-loop-manip.c (slpeel_duplicate_current_defs_from_edges):
Handle constant to_arg, add extra checking we match up the correct
PHIs.
Jakub Jelinek [Mon, 5 Nov 2018 10:28:19 +0000 (11:28 +0100)]
re PR tree-optimization/87859 (store-merging miscompilation of mesa)
PR tree-optimization/87859
* gimple-ssa-store-merging.c (struct merged_store_group): Add
only_constants and first_nonmergeable_order members.
(merged_store_group::merged_store_group): Initialize them.
(merged_store_group::do_merge): Clear only_constants member if
adding something other than INTEGER_CST store.
(imm_store_chain_info::coalesce_immediate_stores): Don't merge
stores with order >= first_nonmergeable_order. Use
merged_store->only_constants instead of always recomputing it.
Set merged_store->first_nonmergeable_order if we've skipped any
stores. Attempt to merge overlapping INTEGER_CST stores that
we would otherwise skip.
* gcc.dg/store_merging_24.c: New test.
* gcc.dg/store_merging_25.c: New test.
Xuepeng Guo [Mon, 5 Nov 2018 09:51:09 +0000 (09:51 +0000)]
re PR target/87853 (_mm_cmpgt_epi8 broken with -funsigned-char)
2018-11-05 Xuepeng Guo <xuepeng.guo@intel.com>
PR target/87853
* config/i386/emmintrin.h (__v16qs): New to cope with option
-funsigned-char.
(_mm_cmpeq_epi8): Replace __v16qi with __v16qs.
(_mm_cmplt_epi8): Likewise.
(_mm_cmpgt_epi8): Likewise.
Jason Merrill [Mon, 5 Nov 2018 07:47:05 +0000 (02:47 -0500)]
Implement UDL changes from P0732R2.
Implementing the UDL changes was pretty straightforward; I simplified
cp_parser_userdef_string_literal using the releasing_vec type from mangle.c.
While looking at this, I realized that the string UDL template taking a
character pack that we implemented for C++14 didn't actually make it into
C++14, so I've added a pedwarn for it and no longer suggest it in the
diagnostic about an invalid UDL template.
* cp-tree.h (struct releasing_vec): Move from mangle.c.
Add get_ref method.
* parser.c (cp_parser_userdef_string_literal): Use it. Handle
passing the string to a single template parameter of class type.
(cp_parser_template_declaration_after_parameters): Allow it.
Pedwarn about the character pack template that was proposed but not
accepted for C++14, and don't suggest it.
Jason Merrill [Mon, 5 Nov 2018 07:47:02 +0000 (02:47 -0500)]
Implement P0732R2, class types in non-type template parameters.
There is one significant piece of this that is not implemented yet: the
reliance on defaulted operator<=>, which someone else has been working on.
So, for the moment those lines are commented out of the testcases.
One tricky bit was treating template parameters of classtype as const
lvalues without making their decltype const; for this I used a
VIEW_CONVERT_EXPR wrapper, which previously could only appear in templates
as location wrappers.
The user-defined literal parts of P0732R2 are in the next patch.
Jason Merrill [Mon, 5 Nov 2018 07:46:57 +0000 (02:46 -0500)]
Fix various latent issues revealed by P0732 work.
The initialized_type hunk fixes handling of void AGGR_INIT_EXPRs that call a
non-constructor; an AGGR_INIT_EXPR can have void type if its initialization
semantics are more complicated than just expanding the call.
The cxx_eval_vec_init_1 hunk corrects AGGR_INIT_EXPRs that were
nonsensically built to initialize an object of void type. And the
build_aggr_init_expr hunk makes sure we don't do that again.
The ocp_convert and cxx_eval_outermost_constant_expr hunks deal with making
sure that a constant CONSTRUCTOR has the right type.
* cvt.c (ocp_convert): Don't wrap a CONSTRUCTOR in a NOP_EXPR.
* constexpr.c (initialized_type): Fix AGGR_INIT_EXPR handling.
(cxx_eval_vec_init_1): Correct type of AGGR_INIT_EXPR.
(cxx_eval_outermost_constant_expr): Make sure a CONSTRUCTOR has the
right type. Don't wrap a CONSTRUCTOR if one was passed in.
* tree.c (build_aggr_init_expr): Check for void.
Jason Merrill [Mon, 5 Nov 2018 07:46:52 +0000 (02:46 -0500)]
PR c++/60503 - wrong lambda attribute syntax.
This patch fixes two issues with lambda attribute handling: First, it was in
the wrong place in the grammar. Second, it was treating attributes as
applying to the whole declaration rather than to the function type, as
specified by the standard.
* parser.c (cp_parser_lambda_declarator_opt): Fix attribute
handling.
This introduces a new header that pulls in system includes for use only
in the DMD front-end part of the compiler, fixing up uses of problematic
functions that are prevalent throughout the code.
Commits merged from dmd.
Fix build of the D frontend on the Hurd and KFreeBSD.
Initial patch from Matthias Klose.
https://github.com/dlang/dmd/pull/8893
Don't care about D/C++ compatibility in C++ port.
Fixes build error in https://gcc.gnu.org/PR87788
https://github.com/dlang/dmd/pull/8895
Allow compiling front-end headers with strict warnings.
https://github.com/dlang/dmd/pull/8909
Add root/system.h header for wrapping system includes.
Fixes https://gcc.gnu.org/PR87865
https://github.com/dlang/dmd/pull/8910
Move checkedint to dmd/root.
https://github.com/dlang/dmd/pull/8912
Use rmem instead of libc for malloc() and strdup().
https://github.com/dlang/dmd/pull/8913
Use align(8) for alignment of UnionExp, fixing several BUS errors
due to alignment issues on SPARC.
https://github.com/dlang/dmd/pull/8914
Don't pass NULL pointer as format parameter to errorSupplemental.
https://github.com/dlang/dmd/pull/8916
Bernd Edlinger [Sun, 4 Nov 2018 19:46:08 +0000 (19:46 +0000)]
re PR tree-optimization/87672 (81512c36 causes ICE in bootstrap stage 3 using "-D_FORTIFY_SOURCE=2" (invalid operand in unary operation, incorrect sharing of tree nodes, verify_gimple failed))
Although there's no fundamental reason why shrink wrapping and speculation tracking are incompatible...
Although there's no fundamental reason why shrink wrapping and
speculation tracking are incompatible, a phase-ordering requirement
(we need to do speculation tracking before the final basic block
clean-up) means that the shrink wrapping pass can undo some of the
changes the speculation tracking pass makes. The result is that the
tracking, while still safe is less comprehensive than we really want.
So to keep things simple, and because the tracking code is quite
expensive anyway, it seems best to just disable that pass when we are
tracking speculative execution.
* config/aarch64/aarch64.c (aarch64_override_options): Disable
shrink-wrapping when -mtrack-speculation.
Richard Biener [Fri, 2 Nov 2018 07:53:48 +0000 (07:53 +0000)]
re PR tree-optimization/87776 (Compile time hog during RPO VN)
2018-11-02 Richard Biener <rguenther@suse.de>
PR tree-optimization/87776
* tree-ssa-sccvn.c (do_rpo_vn): Do not mark backedges
executable when iterating but running into rpo-vn-max-loop-depth
and not eliding the iteration.
Marek Polacek [Thu, 1 Nov 2018 22:10:31 +0000 (22:10 +0000)]
Implement P0846R0, ADL and function templates.
* decl.c (grokfndecl): Allow FUNCTION_DECL in assert.
* lex.c (unqualified_fn_lookup_error): Handle TEMPLATE_ID_EXPR.
* parser.c (cp_parser_postfix_expression): Do ADL for a template-name.
(cp_parser_template_id): Give errors if parsing the template argument
list didn't go well. Allow FUNCTION_DECL in assert.
(cp_parser_template_name): Consider a name to refer to a template if
it is an unqualified-id followed by a <. Don't return the identifier
if the decl is a function and dependent.
* pt.c (tsubst_copy) <case OVERLOAD>: Remove assert.
* g++.dg/addr_builtin-1.C: Adjust dg-error.
* g++.dg/cpp2a/fn-template1.C: New test.
* g++.dg/cpp2a/fn-template10.C: New test.
* g++.dg/cpp2a/fn-template11.C: New test.
* g++.dg/cpp2a/fn-template12.C: New test.
* g++.dg/cpp2a/fn-template13.C: New test.
* g++.dg/cpp2a/fn-template14.C: New test.
* g++.dg/cpp2a/fn-template15.C: New test.
* g++.dg/cpp2a/fn-template16.C: New test.
* g++.dg/cpp2a/fn-template2.C: New test.
* g++.dg/cpp2a/fn-template3.C: New test.
* g++.dg/cpp2a/fn-template4.C: New test.
* g++.dg/cpp2a/fn-template5.C: New test.
* g++.dg/cpp2a/fn-template6.C: New test.
* g++.dg/cpp2a/fn-template7.C: New test.
* g++.dg/cpp2a/fn-template8.C: New test.
* g++.dg/cpp2a/fn-template9.C: New test.
* g++.dg/parse/fn-template1.C: New test.
* g++.dg/parse/fn-template2.C: New test.
* g++.dg/parse/template19.C: Adjust dg-error.
* g++.dg/template/pr61745.C: Add target to dg-error.
Thomas Koenig [Thu, 1 Nov 2018 19:50:14 +0000 (19:50 +0000)]
re PR fortran/87782 (runtime error: load of value 1818451807, which is not a valid value for type 'expr_t')
2018-11-01 Thomas Koenig <tkoenig@gcc.gnu.org>
PR fortran/87782
* frontend-passes.c (constant_string_length): If there is a
substring with a length which cannot be reduced to a constant,
return NULL.
Paul Thomas [Thu, 1 Nov 2018 19:36:08 +0000 (19:36 +0000)]
re PR fortran/40196 ([F03] [F08] Type parameter inquiry (str%len, a%kind) and Complex parts (z%re, z%im))
2018-11-01 Paul Thomas <pault@gcc.gnu.org>
PR fortran/40196
* dependency.c (are_identical_variables): Return false if the
inquiry refs are not the same.
(gfc_ref_needs_temporary_p): Break on an inquiry ref.
* dump_parse_tree.c (show_ref): Show the inquiry ref type.
* expr.c (gfc_free_ref_list): Break on an inquiry ref.
(gfc_copy_ref): Copy the inquiry ref types.
(find_inquiry_ref): New function.
(simplify_const_ref, simplify_ref_chain): Call it. Add new arg
to simplify_ref_chain.
(gfc_simplify_expr): Use the new arg in call to
simplify_ref_chain.
(gfc_get_full_arrayspec_from_expr, gfc_is_coarray): Break on
inquiry ref.
(gfc_traverse_expr): Return true for inquiry ref.
* frontend-passes.c (gfc_expr_walker): Break on inquiry ref.
* gfortran.h : Add enums and union member in gfc_ref to
implement inquiry refs.
* intrinsic.c : Fix white nois.
* match.c (gfc_match_assignment): A constant lavlue is an
error.
* module.c : Add DECL_MIO_NAME for inquiry_type and the mstring
for inquiry_types.
(mio_ref): Handle inquiry refs.
* primary.c (is_inquiry_ref): New function.
(gfc_match_varspec): Handle inquiry refs calling new function.
(gfc_variable_attr): Detect inquiry ref for disambiguation
with components.
(caf_variable_attr): Treat inquiry and substring refs in the
same way.
* resolve.c (find_array_spec): ditto.
(gfc_resolve_substring_charlen): If there is neither a charlen
ref not an inquiry ref, return.
(resolve_ref): Handle inqiry refs as appropriate.
(resolve_allocate_expr): Entities with an inquiry ref cannot be
allocated.
* simplify.c (simplify_bound, simplify_cobound): Punt on
inquiry refs.
* trans-array.c (get_array_ctor_var_strlen): Break on inquiry
ref.
*trans-expr.c (conv_inquiry): New function.
(gfc_conv_variable): Retain the last typespec to pass to
conv_inquiry on detecting an inquiry ref.
2018-11-01 Paul Thomas <pault@gcc.gnu.org>
PR fortran/40196
* gfortran.dg/inquiry_part_ref_1.f08: New test.
* gfortran.dg/inquiry_part_ref_2.f90: New test.
* gfortran.dg/inquiry_part_ref_3.f90: New test.
Joseph Myers [Thu, 1 Nov 2018 17:05:17 +0000 (17:05 +0000)]
Avoid stray [] in ia64 assembler test.
Noted in <https://gcc.gnu.org/ml/gcc-patches/2018-11/msg00034.html>.
Fix by adding extra newline in test input is analogous to many such
newlines added in cases where argument to a newly added AC_LANG_SOURCE
had a preprocessor directive on its last line.
* configure.ac (gcc_cv_as_ia64_ltoffx_ldxmov_relocs): Add newline
at end of assembler input text.
* configure: Regenerate.
Martin Liska [Thu, 1 Nov 2018 09:19:31 +0000 (10:19 +0100)]
Make __PRETTY_FUNCTION__-like functions mergeable string csts (PR c++/64266).
2018-11-01 Martin Liska <mliska@suse.cz>
Jason Merrill <jason@redhat.com>
PR c++/64266
PR bootstrap/70422
PR ipa/81277
* cp-tree.h (DECL_FNAME_P): New macro.
* decl.c (cp_make_fname_decl): Set DECL_DECLARED_CONSTEXPR_P,
DECL_VALUE_EXPR, DECL_INITIALIZED_BY_CONSTANT_EXPRESSION_P.
(cp_finish_decl):
* lambda.c (is_capture_proxy): Use DECL_FNAME_P.
* pt.c (tsubst_expr): Handle DECL_PRETTY_FUNCTION_P.
2018-11-01 Martin Liska <mliska@suse.cz>
Jason Merrill <jason@redhat.com>
PR c++/64266
PR bootstrap/70422
PR ipa/81277
* g++.dg/cpp0x/constexpr-__func__2.C: Make it a compilation
test.
* g++.old-deja/g++.ext/pretty4.C: Remove as the run-time
assumptions are not longer valid.
Co-Authored-By: Jason Merrill <jason@redhat.com>
From-SVN: r265711
Alexandre Oliva [Thu, 1 Nov 2018 01:23:08 +0000 (01:23 +0000)]
gOlogy: skip dbranch at -Og
Delayed slot filling moves insns without any regard to variable
location notes, causing the location information in them to become
incorrect.
Fixing that appears to be quite difficult, but filling delay slots is
hardly an essential optimization to run at -Og, so if the user wants
to privilege debuggability, skip delay slot filling.
for gcc/ChangeLog
* opts.c (default_options_table): Do not enable
OPT_fdelayed_branch at -Og.
* doc/invoke.texi (-fdelayed-branch): Document it.
aarch64: Remove early clobber from ATOMIC_LDOP scratch
* config/aarch64/atomics.md (aarch64_atomic_<ATOMIC_LDOP><ALLI>_lse):
The scratch register need not be early-clobber. Document the reason
why we cannot use ST<OP>.
Nathan Sidwell [Wed, 31 Oct 2018 17:12:27 +0000 (17:12 +0000)]
[PATCH] command line macros
https://gcc.gnu.org/ml/gcc-patches/2018-10/msg02062.html
* c-opts.c (c_finish_options): Force command line macro
location. Refactor to avoid repeating main debug hook.
(push_command_line_include): Clarify comment.
Joseph Myers [Wed, 31 Oct 2018 17:03:16 +0000 (17:03 +0000)]
Update GCC to autoconf 2.69, automake 1.15.1 (PR bootstrap/82856).
This patch updates GCC to use autoconf 2.69 and automake 1.15.1.
(That's not the latest automake version, but it's the one used by
binutils-gdb, with which consistency is desirable, and in any case
seems a useful incremental update that should make a future update to
1.16.1 easier.)
The changes are generally similar to the binutils-gdb ones, and are
copied from there where shared files and directories are involved
(there are some further changes to such shared directories, however,
which I'd expect to apply to binutils-gdb once this patch is in GCC).
Largely, obsolete AC_PREREQ calls are removed, while many
AC_LANG_SOURCE calls are added to avoid warnings from aclocal and
autoconf. Multilib support is no longer included in core automake,
meaning that multilib.am needs copying from automake's contrib
directory into the GCC source tree. Autoconf 2.69 has Go support, so
local copies of that support are removed. I hope the D support will
soon be submitted to upstream autoconf so the local copy of that can
be removed in a future update. Changes to how automake generates
runtest calls mean quotes are removed from RUNTEST definitions in five
lib*/testsuite/Makefile.am files (libatomic, libgomp, libitm,
libphobos, libvtv; some others have RUNTEST definitions without
quotes, which are still OK); libgo and libphobos also get
-Wno-override added to AM_INIT_AUTOMAKE so those overrides of RUNTEST
do not generate automake warnings.
Note that the regeneration did not include regeneration of
fixincludes/config.h.in (attempting such regeneration resulted in all
the USED_FOR_TARGET conditionals disappearing; and I don't see
anything in the fixincludes/ directory that would result in such
conditionals being generated, unlike in the gcc/ directory). Also
note that libvtv/testsuite/other-tests/Makefile.in was not
regenerated; that directory is not listed as a subdirectory for which
Makefile.in gets regenerated by calling "automake" in libvtv/, so I'm
not sure how it's meant to be regenerated.
While I mostly fixed warnings should running aclocal / automake /
autoconf, there were various such warnings from automake in the
libgfortran, libgo, libgomp, liboffloadmic, libsanitizer, libphobos
directories that I did not fix, preferring to leave those to the
relevant subsystem maintainers. Specifically, most of those warnings
were of the following form (example from libgfortran):
Makefile.am:48: warning: source file 'caf/single.c' is in a subdirectory,
Makefile.am:48: but option 'subdir-objects' is disabled
automake: warning: possible forward-incompatibility.
automake: At least a source file is in a subdirectory, but the 'subdir-objects'
automake: automake option hasn't been enabled. For now, the corresponding output
automake: object file(s) will be placed in the top-level directory. However,
automake: this behaviour will change in future Automake versions: they
will
automake: unconditionally cause object files to be placed in the same subdirectory
automake: of the corresponding sources.
automake: You are advised to start using 'subdir-objects' option throughout your
automake: project, to avoid future incompatibilities.
I think it's best for the relevant maintainers to add subdir-objects
and do any other associated Makefile.am changes needed. In some cases
the paths in the warnings involved ../; I don't know if that adds any
extra complications to the use of subdir-objects.
I've tested this with native, cross and Canadian cross builds. The
risk of any OS-specific issues should I hope be rather lower than if a
libtool upgrade were included (we *should* do such an upgrade at some
point, but it's more complicated - it involves identifying all our
local libtool changes to see if any aren't included in the upstream
version we update to, and reverting an upstream libtool patch that's
inappropriate for use in GCC); I think it would be better to get this
update into GCC so that people can test in different configurations
and we can fix any issues found, rather than to try to get more and
more testing done before it goes in.
top level:
2018-10-31 Joseph Myers <joseph@codesourcery.com>
PR bootstrap/82856
* multilib.am: New file. From automake.
Merge from binutils-gdb:
2018-06-19 Simon Marchi <simon.marchi@ericsson.com>
* libtool.m4: Use AC_LANG_SOURCE.
* configure.ac: Remove AC_PREREQ, use AC_LANG_SOURCE.
* ar-lib: New file.
* test-driver: New file.
* configure: Re-generate.
config:
2018-10-31 Joseph Myers <joseph@codesourcery.com>
PR bootstrap/82856
* math.m4, tls.m4: Use AC_LANG_SOURCE.
Merge from binutils-gdb:
2018-06-19 Simon Marchi <simon.marchi@ericsson.com>
* override.m4 (_GCC_AUTOCONF_VERSION): Bump from 2.64 to 2.69.
fixincludes:
2018-10-31 Joseph Myers <joseph@codesourcery.com>
gcc:
2018-10-31 Joseph Myers <joseph@codesourcery.com>
PR bootstrap/82856
* configure.ac: Remove AC_PREREQ. Use AC_LANG_SOURCE. Use single
line for second argument of AC_DEFINE_UNQUOTED.
* doc/install.texi (Tools/packages necessary for modifying GCC):
Update to autoconf 2.69 and automake 1.15.1.
* aclocal.m4, config.in, configure: Regenerate.
gnattools:
2018-10-31 Joseph Myers <joseph@codesourcery.com>
gotools:
2018-10-31 Joseph Myers <joseph@codesourcery.com>
PR bootstrap/82856
* config/go.m4: Remove file.
* Makefile.am (ACLOCAL_AMFLAGS): Do not use -I ./config.
* configure.ac: Remove AC_PREREQ. Do not include config/go.m4.
* Makefile.in, aclocal.m4, configure: Regenerate.
intl:
2018-10-31 Joseph Myers <joseph@codesourcery.com>
PR bootstrap/82856
Merge from binutils-gdb:
2018-06-19 Simon Marchi <simon.marchi@ericsson.com>
Nathan Sidwell [Wed, 31 Oct 2018 15:26:28 +0000 (15:26 +0000)]
[6/6] Preprocessor forced macro location
https://gcc.gnu.org/ml/gcc-patches/2018-10/msg02044.html
libcpp/
* internal.h (struct cpp_reader): Rename forced_token_location_p
to forced_token_location and drop its pointerness.
* include/cpplib.h (cpp_force_token_locations): Take location, not
pointer to one.
* init.c (cpp_create_reader): Adjust.
* lex.c (cpp_read_main_file):
Martin Liska [Wed, 31 Oct 2018 14:46:17 +0000 (15:46 +0100)]
Provide extension hint for aarch64 target (PR driver/83193).
2018-10-31 Martin Liska <mliska@suse.cz>
PR driver/83193
* common/config/aarch64/aarch64-common.c (aarch64_parse_extension):
Add new argument invalid_extension.
(aarch64_get_all_extension_candidates): New function.
(aarch64_rewrite_selected_cpu): Add NULL to function call.
* config/aarch64/aarch64-protos.h (aarch64_parse_extension): Add
new argument.
(aarch64_get_all_extension_candidates): New function.
* config/aarch64/aarch64.c (aarch64_parse_arch): Add new
argument invalid_extension.
(aarch64_parse_cpu): Likewise.
(aarch64_print_hint_for_extensions): New function.
(aarch64_validate_mcpu): Provide hint about invalid extension.
(aarch64_validate_march): Likewise.
(aarch64_handle_attr_arch): Pass new argument.
(aarch64_handle_attr_cpu): Provide hint about invalid extension.
(aarch64_handle_attr_isa_flags): Likewise.
2018-10-31 Martin Liska <mliska@suse.cz>
PR driver/83193
* gcc.target/aarch64/spellcheck_7.c: New test.
* gcc.target/aarch64/spellcheck_8.c: New test.
* gcc.target/aarch64/spellcheck_9.c: New test.
Jonathan Wakely [Wed, 31 Oct 2018 12:29:02 +0000 (12:29 +0000)]
PR libstdc++/87822 fix layout change for nested std::pair
The introduction of the empty __pair_base base class for PR 86751
changed the layout of std::pair<std::pair<...>, ...>. The outer pair and
its first member both have a base class of the same type, which cannot
exist at the same address. This causes the first member to be at a
non-zero offset.
The solution is to make the base class depend on the template
parameters, so that each pair type has a different base class type,
which allows the base classes of the outer pair and its first member to
have the same address.
PR libstdc++/87822
* include/bits/stl_pair.h (__pair_base): Change to class template.
(pair): Make base class type depend on template parameters.
* testsuite/20_util/pair/87822.cc: New test.
Richard Biener [Wed, 31 Oct 2018 11:57:33 +0000 (11:57 +0000)]
re PR middle-end/70359 (Code size increase for x86/ARM/others compared to gcc-5.3.0)
2018-10-31 Richard Biener <rguenther@suse.de>
PR middle-end/70359
PR middle-end/86270
* tree-outof-ssa.c (insert_backedge_copies): Restrict
copy generation to useful cases. Place the copy before
the definition of the backedge value when possible.
* gcc.target/i386/pr70359.c: New testcase.
* gcc.target/i386/pr86270.c: Likewise.
* config/arc/arc-arch.h (ARC_TUNE_ARC7XX): New tune value.
* config/arc/arc.c (arc_active_insn): New function.
(check_store_cacheline_hazard): Likewise.
(workaround_arc_anomaly): Use check_store_cacheline_hazard.
(arc_override_options): Disable delay slot scheduler for older
A7.
(arc_store_addr_hazard_p): New implementation, old one renamed to
...
(arc_store_addr_hazard_internal_p): Renamed.
(arc_reorg): Don't combine into brcc instructions which are part
of hardware hazard solution.
* config/arc/arc.md (attr tune): Consider new arc7xx tune value.
(tune_arc700): Likewise.
* config/arc/arc.opt (arc7xx): New tune value.
* config/arc/arc700.md: Improve A7 scheduler.
projects / thirdparty / gcc.git / log
