Jonathan Wakely [Sat, 4 May 2019 14:35:25 +0000 (15:35 +0100)]
Fix std::hash<std::error_condition>
The hash value should be based on the identity (i.e. address) of the
error_category member, not its object representation (i.e. underlying
bytes).
* include/std/system_error (error_code): Remove friend declaration
for hash<error_code>.
(hash<error_code>::operator()): Use public member functions to access
value and category.
(hash<error_condition>::operator()): Use address of category, not
its object representation.
* src/c++11/compatibility-c++0x.cc (hash<error_code>::operator()):
Use public member functions to access value and category.
* testsuite/19_diagnostics/error_condition/hash.cc: New test.
* include/bits/hashtable.h (_Hashtable<>::rehash): Review comment.
* include/bits/hashtable_policy.h
(_Prime_rehash_policy::_M_bkt_for_elements): Use __builtin_ceill.
(_Power2_rehash_policy::_M_bkt_for_elements): Likewise.
(_Power2_rehash_policy::_M_next_bkt): Enforce returning a result not
smaller than input value rather than always greater. Preserve
_M_next_resize if called with 0 input. Use __builtin_floorl.
(_Power2_rehash_policy::_M_need_rehash): Rehash only if number of
elements + number of insertions is greater than _M_next_resize. Start
with 11 buckets if not told otherwise. Use __builtin_floorl.
(_Rehash_base<>::reserve): Use rehash policy _M_bkt_for_elements.
* src/c++11/hashtable_c++0x.cc (_Prime_rehash_policy::_M_next_bkt):
Preserve _M_next_resize if called with 0 input. Use __builtin_floorl.
(_Prime_rehash_policy::_M_need_rehash): Start with 11 buckets if not
told otherwise. Use __builtin_floorl.
* testsuite/23_containers/unordered_set/hash_policy/71181.cc: Adapt test
to also validate _Power2_rehash_policy.
* testsuite/23_containers/unordered_set/hash_policy/power2_rehash.cc:
Adapt.
Armv6 has support for unaligned accesses to memory. However, the
thumb1 code patterns were trying to use the 32-bit code constraints.
One failure mode from this was that the patterns are designed to be
compatible with conditional execution and this was then causing an
assert in the compiler.
The unaligned_loadhis pattern is only used for expanding extv, which
in turn is only enabled for systems supporting thumb2. Given that
there is no simple expansion for a thumb1 sign-extending load (the
instruction has no immediate offset form and requires two registers in
the address) it seems simpler to just disable this for thumb1.
Fixed thusly:
PR target/89400
* config/arm/arm.md (unaligned_loadsi): Add variant for thumb1.
Restrict 'all' variant to 32-bit configurations.
(unaligned_loadhiu): Likewise.
(unaligned_storehi): Likewise.
(unaligned_storesi): Likewise.
(unaligned_loadhis): Disable when compiling for thumb1.
Richard Biener [Fri, 3 May 2019 10:47:21 +0000 (10:47 +0000)]
tree-vect-stmts.c (get_group_load_store_type): Avoid peeling for gaps by loading only lower halves of vectors if possible.
2019-05-03 Richard Biener <rguenther@suse.de>
* tree-vect-stmts.c (get_group_load_store_type): Avoid
peeling for gaps by loading only lower halves of vectors
if possible.
(vectorizable_load): Likewise.
Jakub Jelinek [Fri, 3 May 2019 07:32:06 +0000 (09:32 +0200)]
re PR tree-optimization/90303 (ICE in hash_odr_name with fastcall attribute starting with r267359)
PR tree-optimization/90303
* ipa-devirt.c (obj_type_ref_class, get_odr_type): Don't use
TYPE_CANONICAL for TYPE_STRUCTURAL_EQUALITY_P types in !in_lto_p mode.
Richard Biener [Fri, 3 May 2019 07:07:28 +0000 (07:07 +0000)]
re PR c++/89698 (Run-time error due to optimization of field access after cast at -Os/-O2 and higher)
2019-05-03 Richard Biener <rguenther@suse.de>
PR tree-optimization/89698
* gimple-fold.c (canonicalize_constructor_val): Early out
for constants, handle unfolded INTEGER_CSTs as they appear in
C++ virtual table ctors.
Jonathan Wakely [Thu, 2 May 2019 21:23:38 +0000 (22:23 +0100)]
PR libstdc++/90314 fix non-equivalent declarations of std::swap
In order to use the _GLIBCXX_NOEXCEPT_IF macro for an expression
containing commas I enclosed it in parentheses, so the preprocessor
wouldn't treat it as two arguments to the function-like macro. Clang
gives an error because now the noexcept-specifier noexcept((C)) is not
equivalent to the noexcept(C) one on the declaration of swap in
<type_traits>.
Instead of requiring extra parentheses around the expression, redefine
_GLIBCXX_NOEXCEPT_IF as a variadic macro (even though supporting that in
C++98 is a GNU extension).
PR libstdc++/90314
* include/bits/c++config (_GLIBCXX_NOEXCEPT_IF): Use variadic macro.
* include/bits/move.h (swap): Remove extra parentheses.
rs6000: Make debug regnums independent of internal ones
Where we use "internal GCC register numbers" in debug info, that
defines an ABI, so we cannot change those numbers. But we want to
change the internal numbers, and sometimes we do that without
remembering this gotcha anyway; so let's make everything independent
of the internal numbers.
For those registers that are not recognised here (we still have MQ for
example, but also the GCC-internal frame pointer and arg pointer
registers), this just returns the internal register number. This is a
bit worrying: that number could be the same as that for a register we
validly want to have in debug info. I first had a gcc_unreachable ()
for that, but this does now work because dwarf2cfi calls
rs6000_dbx_register_number for every internal register. Then I just
returned 0 for the internal regs, but that causes various regression
tests to fail. So now I return the internal register number again,
as it was before; but this needs to be fixed.
* config/rs6000/rs6000.c (rs6000_dbx_register_number): Do not use
the internal register number, for any "real" register.
triggered with the RISC-V lp64 multilib in a GCC build configured with
`--enable-multilib --enable-languages=all --target=riscv64-linux-gnu'.
This is due to unconditional explicit use of F extension instructions
within inline assembly, to access IEEE exception flags. The use of
these instructions is not allowed when building for a soft-float ABI.
Correct the problem by wrapping said inline assembly into a conditional
such that if `D_SoftFloat' is true, then reads from IEEE exception flags
return 0 and writes are ignored instead, complementing r270522
("libphobos: Add D support for RISC-V Linux"), which is an updated
version of <https://gcc.gnu.org/ml/gcc-patches/2019-04/msg00325.html>,
where the problematic code has originated from.
Jonathan Wakely [Thu, 2 May 2019 15:44:59 +0000 (16:44 +0100)]
Improve API docs for mathematical special functions
This prevents "Mathematical Special Functions" appearing in the
top-level menu of the generated HTML docs, and adds "TR1" to the title
for the TR1 docs, to avoid duplicate titles.
* include/bits/specfun.h: Improve docs.
* include/tr1/cmath: Likewise. Fix nesting of preprocessor conditions
and namespaces.
Jonathan Wakely [Thu, 2 May 2019 15:44:46 +0000 (16:44 +0100)]
Adjust Doxygen processing of pb_ds containers
Several of the pb_ds headers are intended to be included multiple times,
within the definition of various class templates. The including files
define macros like PB_DS_CLASS_C_DEC and PB_DS_GEN_POS before including
these headers.
In some cases the types defined in the headers are actually nested types
within other classes, and so should not have been documented as though
they are declared in the global namespace, as in:
https://gcc.gnu.org/onlinedocs/gcc-8.3.0/libstdc++/api/a12028.html
In other cases the headers provide inline member function definitions,
but when processed by Doxygen the class name "PB_DS_CLASS_C_DEC" is not
recognised.
This patch makes Doxygen ignore definitions that only make sense when
included in the right context with the right macros defined.
Jonathan Wakely [Thu, 2 May 2019 15:44:16 +0000 (16:44 +0100)]
Update libstdc++ Doxygen config
The GROUP_NESTED_COMPOUNDS option means that types nested inside inline
namespaces or other classes will be automatically added to a Doxygen
group, e.g. this actually works as intended:
/**
* @defgroup chrono Time
* @ingroup utilities
*
* Classes and functions for time.
* @{
*/
Currently chrono::duration and chrono::time_point are not added to the
"chrono" group. They would need an explicit @ingroup tag added to them
individually. With GROUP_NESTED_COMPOUNDS=YES they get added to the
enclosing group.
The SORT_BY_SCOPE_NAME option means that the list of classes will sort
by class name, not the full qualified-id. Currently the alphabetical
Class List for classes beginning with 'c' looks like:
i.e. the list is sorted by the namespaces first, then the class names.
This is not helpful when you don't know which namespace a class might be
in, and inline namespaces with reserved names are not hidden (see
https://github.com/doxygen/doxygen/issues/5914 for a feature request to
allow that).
This allows you to find a class by name more easily.
Also add PREDEFINED macros so that __attribute__ and various macros like
_GLIBCXX_NO_DISCARD, _GLIBCXX14_CONSTEXPR don't appear in the generated
docs.
* doc/doxygen/user.cfg.in: Regenerate with Doxygen 1.8.14 and set
GROUP_NESTED_COMPOUNDS=YES and SORT_BY_SCOPE_NAME=NO. Add various
_GLIBCXX_xxx macros and __attribute__(X) to PREDEFINED macros that
Doxygen expands.
Iain Sandoe [Thu, 2 May 2019 14:55:21 +0000 (14:55 +0000)]
darwin, testsuite, fix a failing test.
The istantiate2.C test has started to fail since Darwin's impl. of
this part of the ABI was fixed. It now emits the same output as
other platforms (and clang).
2019-05-02 Iain Sandoe <iain@sandoe.co.uk>
* g++.dg/ext/instantiate2.C: Remove special-caseing for Darwin.
Richard Biener [Thu, 2 May 2019 14:08:08 +0000 (14:08 +0000)]
re PR tree-optimization/89653 (Missing vectorization of loop containing std::min/std::max and temporary)
2019-05-02 Richard Biener <rguenther@suse.de>
PR tree-optimization/89653
* tree-ssa-loop.c (pass_data_tree_loop_init): Execute
update-address-taken before the pass.
* passes.def (pass_tree_loop_init): Put comment before it.
This patch does two things. For the general vectoriser, it adds support to
perform fully masked reductions over expressions that don't support masking.
This is achieved by using VEC_COND_EXPR where possible. At the moment this is
implemented for DOT_PROD_EXPR only, but the framework is there to extend it to
other expressions.
Related to that, this patch adds support to vectorize dot product using SVE. It
also uses the new functionality to ensure that the resulting loop is masked.
Given this input code:
uint32_t
dotprod (uint8_t *restrict x, uint8_t *restrict y, int n)
{
uint32_t sum = 0;
for (int i = 0; i < n; i++)
{
sum += x[i] * y[i];
}
* config/aarch64/aarch64-sve.md (<sur>dot_prod<vsi2qi>): Taken from SVE
ACLE branch.
* config/aarch64/iterators.md: Copied Vetype_fourth, VSI2QI and vsi2qi from
SVE ACLE branch.
* tree-vect-loop.c (use_mask_by_cond_expr_p): New function to check if a
VEC_COND_EXPR be inserted to emulate a conditional internal function.
(build_vect_cond_expr): Emit the VEC_COND_EXPR.
(vectorizable_reduction): Use the functions above to vectorize in a
fully masked loop codes that don't have a conditional internal
function.
Martin Liska [Thu, 2 May 2019 08:43:25 +0000 (10:43 +0200)]
Enhance target and target_clone error messages.
2019-05-02 Martin Liska <mliska@suse.cz>
* cgraphclones.c: Call valid_attribute_p with 1 for
target_clone.
* config/i386/i386-c.c (ix86_pragma_target_parse): Use 0 as
it's for target attribute.
* config/i386/i386-protos.h (ix86_valid_target_attribute_tree):
Add new boolean argument.
* config/i386/i386.c (ix86_valid_target_attribute_inner_p):
Likewise.
(ix86_valid_target_attribute_tree): Pass target_clone_attr
to ix86_valid_target_attribute_inner_p.
(ix86_valid_target_attribute_p): Pass flags argument to
ix86_valid_target_attribute_inner_p.
(get_builtin_code_for_version): Use 0 as it's target attribute.
2019-05-02 Martin Liska <mliska@suse.cz>
Martin Liska [Thu, 2 May 2019 08:16:12 +0000 (10:16 +0200)]
Postpone print of --help=* option.
2019-05-02 Martin Liska <mliska@suse.cz>
* gcc.c (process_command): Add dummy file only
if n_infiles == 0.
* opts-global.c (decode_options): Pass lang_mask.
* opts.c (print_help): New function.
(finish_options): Print --help if help_option_argument
is set.
(common_handle_option): Factor out content of OPT__help_
into print_help.
* opts.h (finish_options): Add new argument.
Martin Liska [Thu, 2 May 2019 07:57:38 +0000 (09:57 +0200)]
Prefer to use strlen call instead of inline expansion (PR target/88809).
2019-05-02 Martin Liska <mliska@suse.cz>
PR target/88809
* config/i386/i386.c (ix86_expand_strlen): Use strlen call.
With -minline-all-stringops use inline expansion using 4B loop.
* doc/invoke.texi: Document the change of
-minline-all-stringops.
2019-05-02 Martin Liska <mliska@suse.cz>
PR target/88809
* gcc.target/i386/pr88809.c: New test.
* gcc.target/i386/pr88809-2.c: New test.
A direct interface is an interface whose data word contains the
actual data value, instead of a pointer to it. The gc toolchain
creates a direct interface if the value is pointer shaped, that
includes pointers (including unsafe.Pointer), functions, channels,
maps, and structs and arrays containing a single pointer-shaped
field. In gccgo, we only do this for pointers. This CL unifies
direct interface types with gc. This reduces allocations when
converting such types to interfaces.
Our method functions used to always take pointer receivers, to
make interface calls easy. Now for direct interface types, their
value methods will take value receivers. For a pointer to those
types, when converted to interface, the interface data contains
the pointer. For that interface to call a value method, it will
need a wrapper method that dereference the pointer and invokes
the value method. The wrapper method, instead of the actual one,
is put into the itable of the pointer type.
In the runtime, adjust funcPC for the new layout of interfaces of
functions.
Previously, each time we do an interface conversion for which the
method table is not known at compile time, we allocate a new
method table.
This CL ports the mechanism of itab caching from the gc runtime,
adapted to our itab representation and method finding mechanism.
With the cache, we reuse the same itab for the same (interface,
concrete) type pair. This reduces allocations in interface
conversions.
Unlike the gc runtime, we don't prepopulate the cache with
statically allocated itabs, as currently we don't have a way to
find them. This means we don't deduplicate run-time allocated
itabs with compile-time allocated ones. But that is not too bad
-- it is just a cache anyway.
As now itabs are never freed, it is also possible to drop the
write barrier for writing the first word of an interface header.
I'll leave this optimization for the future.
PR fortran/60144
* match.c (gfc_match_parens): Change the location for missing ')'.
(gfc_match_if): Detect a missing '('. Remove the spurious named
constant error. Change the wording of some errors.
(gfc_match_else): Change the wording of an error.
(gfc_match_elseif): Detect a missing '('. Improve the matching
process to get a better syntax analysis.
Jonathan Wakely [Wed, 1 May 2019 00:08:36 +0000 (01:08 +0100)]
PR libstdc++/61761 fix std::proj for targets without C99 cproj
The current generic implementation of __complex_proj used when cproj is
not available calculates the wrong projection, giving a different result
than given by C99's cproj.
When C99 cproj is not available but isinf and copysign are, use those to
give correct results for float, double and long double. Otherwise, and
for other specializations of std::complex, just use a generic version
that returns its argument, and so doesn't support infinities.
We might want to consider adding additional overloads of __complex_proj
to support extended types such as _Float64x, _Float128 etc.
PR libstdc++/61761
* include/std/complex (__complex_proj): Return parameter unchanged.
[_GLIBCXX_USE_C99_COMPLEX] (__complex_proj): Change overloads for
floating-point types to take std::complex arguments.
[_GLIBCXX_USE_C99_MATH_TR1] (__complex_proj): Add overloads for
floating-point types.
* testsuite/26_numerics/complex/proj.cc: New test.
This keeps coming up repeatedly and the ACLE has finally added
__ARM_FEATURE_ATOMICS for the LSE feature in GCC. This is now part of
the latest ACLE release
(https://developer.arm.com/docs/101028/latest/5-feature-test-macros)
I know it's late for GCC-9 but this is a simple macro which need not
wait for another year.
Ok for trunk and to backport to all release branches ?
Jonathan Wakely [Tue, 30 Apr 2019 10:39:59 +0000 (11:39 +0100)]
Fix filesystem::path tests
The root_path.cc test had some debugging macros left in accidentally, so
didn't FAIL correctly if an assertion failed.
The string-char8_t.cc tests didn't compile on Windows.
* testsuite/27_io/filesystem/path/decompose/root_path.cc: Remove
macros accidentally left in.
* testsuite/27_io/filesystem/path/native/string-char8_t.cc: Remove
unnecessary -lstdc++fs option. Fix test for mingw.
* testsuite/experimental/filesystem/path/native/string-char8_t.cc:
Fix test for mingw.
Martin Liska [Tue, 30 Apr 2019 09:19:41 +0000 (11:19 +0200)]
Refactor gimple_fold_builtin_memory_op function.
2019-04-30 Martin Liska <mliska@suse.cz>
* gimple-fold.c (gimple_fold_builtin_memory_op): Change endp
into built_in_function enum. Remove code for endp == 2 and
use BUILT_IN_* constants.
(gimple_fold_builtin): Call the function with fcode.
projects / thirdparty / gcc.git / log
