Iain Sandoe [Sat, 25 Jun 2022 08:58:35 +0000 (09:58 +0100)]
testsuite, Darwin: Fix darwin-comm-1.c error messages for Darwin <= 10.
When amending the allowed alignment size to accommodate the larger values
permitted by newer tools, we retained the object file limit of 2^15 for
Darwin versions <= 10, since that is what the native tools expect there.
This triggers a different diagnostic path with a distinct error message,
which is checked in the revised test here.
Roger Sayle [Sat, 25 Jun 2022 08:35:45 +0000 (09:35 +0100)]
Iterating cprop_hardreg... Third time's a charm.
This middle-end patch proposes the "hard register constant propagation"
pass be performed up to three times on each basic block (up from the
current two times) if the second pass successfully made changes.
The motivation for three passes is to handle the "swap idiom" (i.e.
t = x; x = y; y = t;" sequences) that get generated by register allocation
(reload).
Consider the x86_64 test case for __int128 addition recently discussed
on gcc-patches. With that proposed patch, the input to the cprop_hardreg
pass looks like:
Because a successful propagation was found, cprop_hardreg then runs
a second pass/sweep on affected basic blocks (using worklist), and
on this second pass notices that the second instruction, %rdi := %rsi,
may now be propagated (%rsi was killed in the before the first transform),
and after a second pass, we now end up with:
which is the current behaviour on mainline. However, a third and final
pass would now notice that the first insn, "%r8 := %rdi" is also now
eliminable, and a third iteration would produce optimal code:
The patch below creates two worklists, and alternates between them on
sucessive passes, populating NEXT with the basic block id's of blocks
that were updated during the current pass over the CURR worklist.
It should be noted that this a regression fix; GCC 4.8 generated
optimal code with two moves (whereas GCC 12 required 5 moves, up
from GCC 11's 4 moves).
2022-06-25 Roger Sayle <roger@nextmovesoftware.com>
Richard Biener <rguenther@suse.de>
gcc/ChangeLog
* regcprop.cc (pass_cprop_hardreg::execute): Perform a third
iteration over each basic block that was updated by the second
iteration.
Xi Ruoyao [Fri, 24 Jun 2022 06:27:18 +0000 (14:27 +0800)]
intl: stop using fgrep for exgettext
fgrep has been deprecated in favor of grep -F for a long time, and the
next grep release (3.8 or 4.0) will print a warning of fgrep is used.
And, the fgrep command in exgettext is no longer useful after we
migrated from SVN to Git. Remove the fgrep command so we won't see the
warning.
Jason Merrill [Fri, 24 Jun 2022 18:40:12 +0000 (14:40 -0400)]
c++: Include -Woverloaded-virtual in -Wall [PR87729]
This seems like a good warning to have in -Wall, as requested. But as
pointed out in PR20423, some users want a warning only when a derived
function doesn't override any base function. So let's put that lesser
version in -Wall (and -Woverloaded-virtual=1) while leaving the semantics
for the existing option the same.
PR c++/87729
PR c++/20423
gcc/c-family/ChangeLog:
* c.opt (Woverloaded-virtual): Add levels, include in -Wall.
testsuite: Skip btf-bitfields-1.c if int is less than 32-bits
This test spuriously fails on AVR with:
error: width of 'bitfield_c' exceeds its type
8-bit and 16-bit microcontrollers do not seem to be the target audience
for BTF file format. So the least intrusive fix is to simply skip the
test for them.
gcc/testsuite/ChangeLog:
* gcc.dg/debug/btf/btf-bitfields-1.c: Skip if int is less than
32-bits.
Harald Anlauf [Fri, 24 Jun 2022 20:21:39 +0000 (22:21 +0200)]
Fortran: fix checking of arguments to UNPACK when MASK is a variable [PR105813]
gcc/fortran/ChangeLog:
PR fortran/105813
* check.cc (gfc_check_unpack): Try to simplify MASK argument to
UNPACK so that checking of the VECTOR argument can work when MASK
is a variable.
gcc/testsuite/ChangeLog:
PR fortran/105813
* gfortran.dg/unpack_vector_1.f90: New test.
Dimitar Dimitrov [Fri, 13 May 2022 19:46:59 +0000 (22:46 +0300)]
testsuite: Handle default_packed for gcc.dg/builtin-object-size-20.c
The gcc.dg/builtin-object-size-20.c test case assumes that the target
inserts padding between structure members. Obviously it fails for
targets which pack structures by default.
Split the cases into two tests, so that the ones requiring structure
padding can be skipped for default_packed targets.
gcc/testsuite/ChangeLog:
* gcc.dg/builtin-object-size-20.c: Remove cases which
work on default_packed targets.
* gcc.dg/builtin-object-size-22.c: New test with the cases
removed above.
Epiphany, PRU, ARC and NDS32 may predefine __big_endian__ and
__little_endian__ macros. This leads to spurious warnings like:
gcc.dg/sso/memcpy-1.c:7: warning: "__little_endian__" redefined
Fix by renaming the macros in the test.
gcc/testsuite/ChangeLog:
* gcc.dg/sso/memcpy-1.c (__big_endian__, __little_endian__):
Rename macros to avoid conflicts with predefined ones.
Dimitar Dimitrov [Sun, 15 May 2022 14:30:52 +0000 (17:30 +0300)]
testsuite: Remove reliance on argc in lto/pr101868_0.c
Some embedded targets do not pass any argv arguments. When argc is
zero, this causes spurious failures for lto/pr101868_0.c. Fix by
following the strategy in r0-114701-g2c49569ecea56d. Use a volatile
variable instead of argc to inject a runtime value into the test.
I validated the following:
- No changes in testresults for x86_64-pc-linux-gnu.
- The spurious failures are fixed for PRU target.
- lto/pr101868_0.c still fails on x86_64-pc-linux-gnu, if
the PR/101868 fix (r12-2254-gfedcf3c476aff7) is reverted.
PR tree-optimization/101868
gcc/testsuite/ChangeLog:
* gcc.dg/lto/pr101868_0.c (zero): New volatile variable.
(main): Use it instead of argc.
Iain Buclaw [Thu, 23 Jun 2022 16:24:07 +0000 (18:24 +0200)]
d: Add `@register' attribute to compiler and library.
The `@register` attribute specifies that a local or `__gshared` variable
is to be given a register storage-class in the C sense of the term, and
will be placed into a register named `registerName`.
The variable needs to boiled down to a data type that fits the target
register. It also cannot have either thread-local or `extern` storage.
It is an error to take the address of a register variable.
PR d/105413
gcc/d/ChangeLog:
* d-attribs.cc (d_handle_register_attribute): New function.
(d_langhook_attribute_table): Add register attribute.
* d-codegen.cc (d_mark_addressable): Error if taken address of
register variable.
(build_frame_type): Error if register variable has non-local
references.
* d-tree.h (d_mark_addressable): Add complain parameter.
* decl.cc (get_symbol_decl): Mark register varibles DECL_REGISTER.
Error when register variable declared thread-local or extern.
* expr.cc (ExprVisitor::visit (IndexExp *)): Don't complain about
marking register vectors as addressable in an ARRAY_REF.
Iain Buclaw [Fri, 24 Jun 2022 17:48:55 +0000 (19:48 +0200)]
d: Construct indexes of ARRAY_TYPE using ARRAY_REF.
This is a small simplification over `((T *)&array)[index]', which also
allows eliding an unneccesary marking of TREE_ADDRESSABLE when the array
expression is a parameter or variable declaration.
Iain Buclaw [Wed, 22 Jun 2022 17:11:20 +0000 (19:11 +0200)]
tilegx: Fix infinite loop in gen-mul-tables generator
Since around GCC 10, the condition `j < (INTMAX_MAX / 10)' will get
optimized into `j != 922337203685477580', which will result in an
infinite loop for certain inputs of `j'.
Copy the condition already used by the -DTILEPRO generator code, which
doesn't fall into this trap.
gcc/ChangeLog:
* config/tilepro/gen-mul-tables.cc (tilegx_emit): Adjust loop
condition to avoid overflow.
Ian Lance Taylor [Fri, 24 Jun 2022 00:11:05 +0000 (17:11 -0700)]
compiler: use bool for comma-ok if not already boolean
If a comma-ok variable already has a type, and that type is not a
boolean type, then set the type of the temporary variable to bool.
Otherwise we may try to convert an unnamed bool type to an interface
type, which will fail. But we don't want to always use bool, because
the type of the comma-ok variable may be a named bool type, in
which case the assignment would fail (or need an explicit conversion).
David Malcolm [Fri, 24 Jun 2022 17:44:48 +0000 (13:44 -0400)]
analyzer: consolidate call_string instances
ana::call_string is a wrapper around an auto_vec of callsites, leading
to non-trivial copying when copying around call_string instances, e.g.
in ana::program_point.
This patch consolidates call_string instances within the
region_model_manager: it now owns the root/empty call_string, and
each call_string instance tracks its children, lazily creating them on
demand, so that the call_string instances form a tree-like hierarchy in
memory. Doing this requires passing the region_model_manager to the
various program_point factory methods, so that they can get at the root
call_string.
Instances of call_string become immutable (apart from their internal
cache for looking up their children); operations that previously
modified them now return the call_string for the result of the
operation.
I wasn't able to observe any performance impact of this, but it
simplifies call_string and program_point management, and thus I hope
will make it easier to improve call summarization. In particular,
region_model_manager::log_stats will now print a hierarchical dump of
all the call_string instances used in the analysis (in -fdump-analyzer
and -fdump-analyzer-stderr).
gcc/analyzer/ChangeLog:
* call-string.cc: Add includes of "analyzer/analyzer.h"
and "analyzer/analyzer-logging.h".
(call_string::call_string): Delete copy ctor.
(call_string::operator=): Delete.
(call_string::operator==): Delete.
(call_string::hash): Delete.
(call_string::push_call): Make const, returning the resulting
call_string.
(call_string::pop): Delete.
(call_string::cmp_ptr_ptr): New.
(call_string::validate): Assert that m_parent is non-NULL, or
m_elements is empty.
(call_string::call_string): Move default ctor here from
call-string.h and reimplement. Add ctor taking a parent
and an element.
(call_string::~call_string): New.
(call_string::recursive_log): New.
* call-string.h (call_string::call_string): Move default ctor's
defn to call-string.cc. Delete copy ctor. Add ctor taking a
parent and an element.
(call_string::operator=): Delete.
(call_string::operator==): Delete.
(call_string::hash): Delete.
(call_string::push_call): Make const, returning the resulting
call_string.
(call_string::pop): Delete decl.
(call_string::get_parent): New.
(call_string::cmp_ptr_ptr): New decl.
(call_string::get_top_of_stack): New.
(struct call_string::hashmap_traits_t): New.
(class call_string): Add friend class region_model_manager. Add
DISABLE_COPY_AND_ASSIGN.
(call_string::~call_string): New decl.
(call_string::recursive_log): New decl.
(call_string::m_parent): New field.
(call_string::m_children): New field.
* constraint-manager.cc (selftest::test_many_constants): Pass
model manager to program_point::origin.
* engine.cc (exploded_graph::exploded_graph): Likewise.
(exploded_graph::add_function_entry): Likewise for
program_point::from_function_entry.
(add_tainted_args_callback): Likewise.
(exploded_graph::maybe_process_run_of_before_supernode_enodes):
Update for change to program_point.get_call_string.
(exploded_graph::process_node): Likewise.
(class function_call_string_cluster): Convert m_cs from a
call_string to a const call_string &.
(struct function_call_string): Likewise.
(pod_hash_traits<function_call_string>::hash): Use pointer_hash
for m_cs.
(pod_hash_traits<function_call_string>::equal): Update for change
to m_cs.
(root_cluster::add_node): Update for change to
function_call_string.
(viz_callgraph_node::dump_dot): Update for change to call_string.
* exploded-graph.h (per_call_string_data::m_key): Convert to a
reference.
(struct eg_call_string_hash_map_traits): Delete.
(exploded_graph::call_string_data_map_t): Remove traits class.
* program-point.cc: Move include of "analyzer/call-string.h" to
after "analyzer/analyzer-logging.h".
(program_point::print): Update for conversion of m_call_string to
a pointer.
(program_point::to_json): Likewise.
(program_point::push_to_call_stack): Update for immutability of
call strings.
(program_point::pop_from_call_stack): Likewise.
(program_point::hash): Use pointer hashing for m_call_string.
(program_point::get_function_at_depth): Update for change to
m_call_string.
(program_point::validate): Update for changes to call_string.
(program_point::on_edge): Likewise.
(program_point::origin): Move here from call-string.h. Add
region_model_manager param and use it to get empty call string.
(program_point::from_function_entry): Likewise.
(selftest::test_function_point_ordering): Likewise.
(selftest::test_function_point_ordering): Likewise.
* program-point.h (program_point::program_point): Update for
change to m_call_string.
(program_point::get_call_string): Likewise.
(program_point::get_stack_depth): Likewise.
(program_point::origin): Add region_model_manager param, and move
defn to call-string.cc.
(program_point::from_function_entry): Likewise.
(program_point::empty): Drop call_string.
(program_point::deleted): Likewise.
(program_point::program_point): New private ctor.
(program_point::m_call_string): Convert from call_string to const
call_string *.
* program-state.cc (selftest::test_program_state_merging): Update
for call_string changes.
(selftest::test_program_state_merging_2): Likewise.
* region-model-manager.cc
(region_model_manager::region_model_manager): Construct
m_empty_call_string.
(region_model_manager::log_stats): Log the call strings.
* region-model.cc (assert_region_models_merge): Pass the
region_model_manager when creating program_point instances.
(selftest::test_state_merging): Likewise.
(selftest::test_constraint_merging): Likewise.
(selftest::test_widening_constraints): Likewise.
(selftest::test_iteration_1): Likewise.
* region-model.h (region_model_manager::get_empty_call_string):
New.
(region_model_manager::m_empty_call_string): New.
* sm-signal.cc (register_signal_handler::impl_transition): Update
for changes to call_string.
Signed-off-by: David Malcolm <dmalcolm@redhat.com>
Will Schmidt [Thu, 23 Jun 2022 20:50:57 +0000 (15:50 -0500)]
[PATCH,RS6000 3/5] Rework the RS6000_BTM defines, continued.
The RS6000_BTM_<xxxx> definitions are mostly unused after
the rs6000 builtin code was reworked. This cleans
up the remaining RS6000_BTM_<xxxx> references by replacing
them with their OPTION_MASK_<xxxx> equivalents.
This patch removes the defines
RS6000_BTM_MODULO, RS6000_BTM_ALTIVEC, RS6000_BTM_CMPB,
RS6000_BTM_VSX, RS6000_BTM_P8_VECTOR, RS6000_BTM_P9_VECTOR,
RS6000_BTM_P9_MISC, RS6000_BTM_CRYPTO, RS6000_BTM_HTM,
RS6000_BTM_FRE.
Will Schmidt [Thu, 23 Jun 2022 20:48:21 +0000 (15:48 -0500)]
[PATCH,RS6000 2/5] Rework the RS6000_BTM defines.
The RS6000_BTM_<xxxx> definitions are mostly unused after the rs6000
builtin code was reworked. The remaining references can be replaced
with the OPTION_MASK_<xxxx> and MASK_<xxxx> equivalents.
This patch remvoes the defines:
RS6000_BTM_FRES, RS6000_BTM_FRSQRTE, RS6000_BTM_FRSQRTES,
RS6000_BTM_POPCNTD, RS6000_BTM_CELL, RS6000_BTM_DFP,
RS6000_BTM_HARD_FLOAT, RS6000_BTM_LDBL128, RS6000_BTM_64BIT,
RS6000_BTM_POWERPC64, RS6000_BTM_FLOAT128, RS6000_BTM_FLOAT128_HW
RS6000_BTM_MMA, RS6000_BTM_P10.
I note that the BTM -> OPTION_MASK mappings are not always 1-to-1.
in particular the BTM_FRES and BTM_FRSQRTE values were both mapped to
OPTION_MASK_PPC_GFXOPT, while the BTM_FRE and BTM_FRSQRTES both mapped
to OPTION_MASK_POPCNTB. In total I spent quite a bit of time
double-checking these since it looked like copy/paste errors. I split
some of these changes out into a subsequent patch to limit the amount
of potential confusion in any particular patch.
Xi Ruoyao [Fri, 24 Jun 2022 07:02:23 +0000 (15:02 +0800)]
libstdc++: use grep -E instead of egrep in scripts
egrep has been deprecated in favor of grep -E for a long time, and the
next grep release (3.8 or 4.0) will print a warning of egrep is used.
Stop using egrep so we won't see the warning.
grep's from GNU, BSD (including Mac OS X), AIX, BusyBox all support -E
and -F. Solaris grep doesn't support -E, but extract_symvers.in already
contains a special case for Solaris and doxygen documentation generation
is already broken on non-GNU.
libstdc++-v3/ChangeLog:
* scripts/extract_symvers.in: Use grep -E instead of egrep.
* scripts/run_doxygen: Likewise.
Jan Hubicka [Fri, 24 Jun 2022 11:52:44 +0000 (13:52 +0200)]
Fix stmt_kills_ref_p WRT external throws
Add missing check to stmt_kills_ref_p for case that function
is terminated by EH before call return value kills the ref. In the PR
I tried to construct testcase but I don't know how to do that until I
annotate EH code with fnspec attributes which I will do in separate patch
and add a testcase.
PR ipa/106057
* tree-ssa-alias.cc (stmt_kills_ref_p): Check for external throw.
Roger Sayle [Fri, 24 Jun 2022 06:15:08 +0000 (07:15 +0100)]
PR target/105930: Split *xordi3_doubleword after reload on x86.
This patch addresses PR target/105930 which is an ia32 stack frame size
regression in high-register pressure XOR-rich cryptography functions
reported by Linus Torvalds. The underlying problem is once the limited
number of registers on the x86 are exhausted, the register allocator
has to decide which to spill, where some eviction choices lead to much
poorer code, but these consequences are difficult to predict in advance.
The patch below, which splits xordi3_doubleword and iordi3_doubleword
after reload (instead of before), significantly reduces the amount of
spill code and stack frame size, in what might appear to be an arbitrary
choice.
My explanation of this behaviour is that the mixing of pre-reload split
SImode instructions and post-reload split DImode instructions is
confusing some of the heuristics used by reload. One might think
that splitting early gives the register allocator more freedom to
use available registers, but in practice the constraint that double
word values occupy consecutive registers (when ultimately used as a
DImode value) is the greater constraint. Instead, I believe in this
case, the pseudo registers used in memory addressing, appear to be
double counted for split SImode instructions compared to unsplit
DImode instructions. For the reduced test case in comment #13, this
leads to %eax being used to hold the long-lived argument pointer "v",
blocking the use of the ax:dx pair for processing double word values.
The important lines are at the very top of the assembly output:
GCC 11 [use %ecx to address memory, require a 24-byte stack frame]
sub esp, 24
mov ecx, DWORD PTR [esp+40]
GCC 12 [use %eax to address memory, require a 44-byte stack frame]
sub esp, 44
mov eax, DWORD PTR [esp+64]
2022-06-24 Roger Sayle <roger@nextmovesoftware.com>
Uroš Bizjak <ubizjak@gmail.com>
gcc/ChangeLog
PR target/105930
* config/i386/i386.md (*<any_or>di3_doubleword): Split after
reload. Use rtx_equal_p to avoid creating memory-to-memory moves,
and emit NOTE_INSN_DELETED if operand[2] is zero (i.e. with -O0).
Alexandre Oliva [Fri, 24 Jun 2022 02:20:53 +0000 (23:20 -0300)]
libstdc++: check for openat
rtems6.0 has fdopendir, and fcntl.h defines AT_FDCWD and declares
openat, but there's no openat in libc. Adjust dir-common.h to not
assume ::openat just because of AT_FDCWD.
for libstdc++-v3/ChangeLog
* acinclude.m4 (GLIBCXX_CHECK_FILESYSTEM_DEPS): Check for
openat.
* configure, config.h.in: Rebuilt.
* src/filesystem/dir-common.h (openat): Use ::openat if
_GLIBCXX_HAVE_OPENAT.
* src/filesystem/dir.cc (dir_and_pathname): Use dirfd if
_GLIBCXX_HAVE_OPENAT.
Alexandre Oliva [Fri, 24 Jun 2022 02:20:52 +0000 (23:20 -0300)]
libstdc++: xfail rename tests on rtems
::rename on RTEMS does not meet several POSIX requirements, despite
compliance with C and C++ standards. ::std::filesystem::rename, in
turn, has requirements borrowed from POSIX, so it would have to be a
lot more than a simple wrapper around ::rename on RTEMS, and even then
fall short.
Until RTEMS reimplements ::rename for POSIX compliance, expect
filesystem rename tests to fail on it.
for libstdc++-v3/ChangeLog
* testsuite/27_io/filesystem/operations/rename.cc: xfail on
rtems.
* testsuite/experimental/filesystem/operations/rename.cc:
Likewise.
Alexandre Oliva [Fri, 24 Jun 2022 02:20:51 +0000 (23:20 -0300)]
libstdc++: testsuite: skip fs last_write_time tests if dummy
The last_write_time functions are defined in ways that are useful, or
that fail immediately, depending on various macros. When they fail
immediately, the filesystem last_write_time.cc tests fail noisily, but
the fail is entirely expected.
Define NO_LAST_WRITE_TIME in the last_write_time.cc tests, according
to the macros that select implementations of last_write_time, and use
it through the new dg-require-target-fs-lwt to skip tests that are
expected to fail.
for libstdc++-v3/ChangeLog
* testsuite/util/testsuite_fs.h (NO_LAST_WRITE_TIME): Define
when appropriate.
* testsuite/lib/libstdc++.exp
(check_v3_target_fs_last_write_time): New.
* testsuite/lib/dg-options.exp (dg-require-target-fs-lwt):
New.
* testsuite/27_io/filesystem/operations/last_write_time.cc:
Skip the test if the features are unavailable.
* testsuite/experimental/filesystem/operations/last_write_time.cc:
Likewise.
Alexandre Oliva [Fri, 24 Jun 2022 02:20:51 +0000 (23:20 -0300)]
libstdc++: testsuite: skip fs space tests on dummy implementations
The do_space function is defined in ways that are useful, or that fail
immediately, depending on various macros. When it fails immediately,
the filesystem space.cc tests fail noisily, but the fail is entirely
expected.
Define NO_SPACE in testsuite_fs.h, according to the macros that select
implementations of do_space, and use it to skip tests that are
expected to fail, through a new dg-require.
Several filesystem tests expect to be able to create symlinks even
when !defined (_GLIBCXX_HAVE_SYMLINK), and fail predictably, reducing
the amount of testing of other filesystem features.
They are already skipped for mingw targets. I've extended the
skipping to other targets in which _GLIBCXX_HAVE_SYMLINK is undefined,
through a new NO_SYMLINKS macro in testsuite_fs.h that guards
skippable portions of tests, and dg-require-target-fs-symlinks for
tests that would be reduced to nothing.
for libstdc++-v3/ChangeLog
* testsuite/util/testsuite_fs.h (NO_SYMLINKS): Define on
mingw and when create_symlink is a dummy.
* testsuite/27_io/filesystem/operations/symlink_status.cc:
Drop mingw xfail.
(test01, test02): Don't create symlinks when NO_SYMLINKS is
defined.
* testsuite/27_io/filesystem/operations/canonical.cc (test03):
Likewise.
* testsuite/27_io/filesystem/operations/copy.cc (test02):
Likewise.
* testsuite/27_io/filesystem/operations/create_directories.cc
(test04): Likewise.
* testsuite/27_io/filesystem/operations/create_directory.cc
(test01): Likewise.
* testsuite/27_io/filesystem/operations/permissions.cc
(test03, test04): Likewise.
* testsuite/27_io/filesystem/operations/remove.cc (test01):
Likewise.
* testsuite/27_io/filesystem/operations/remove_all.cc (test01):
Likewise.
* testsuite/27_io/filesystem/operations/rename.cc
(test_symlinks): Likewise.
* testsuite/27_io/filesystem/operations/weakly_canonical.cc
(test01): Likewise.
* testsuite/experimental/filesystem/iterators/recursive_directory_iterator.cc
(test06): Likewise.
* testsuite/experimental/filesystem/operations/copy.cc
(test01): Likewise.
* testsuite/experimental/filesystem/operations/create_directories.cc
(test04): Likewise.
* testsuite/experimental/filesystem/operations/create_directory.cc
(test01): Likewise.
* testsuite/experimental/filesystem/operations/permissions.cc
(test03, test04): Likewise.
* testsuite/experimental/filesystem/operations/remove.cc
(test01): Likewise.
* testsuite/experimental/filesystem/operations/remove_all.cc
(test01): Likewise.
* testsuite/experimental/filesystem/operations/rename.cc
(test01): Likewise.
* testsuite/lib/libstdc++.exp
(v3_check_preprocessor_condition): Add optional inc parameter.
Add it to the test program after include bits/c++config.h.
(check_v3_target_fs_symlinks): New.
* testsuite/lib/dg-options.exp
(dg-require-target-fs-symlinks): New.
* testsuite/27_io/filesystem/operations/read_symlink.cc:
Replace mingw xfail with require target-fs-symlinks.
* testsuite/experimental/filesystem/operations/read_symlink.cc:
Likewise.
Alexandre Oliva [Fri, 24 Jun 2022 02:20:48 +0000 (23:20 -0300)]
Introduce -nostdlib++ option
Using g++ to link without libstdc++, as in g++.dg/abi/pure-virtual1.C,
is error prone, because there's no way to tell g++ to drop libstdc++
without also dropping libc and any other libraries that the target
implicitly links in.
This has often led to the need for manual adjustments to this
testcase.
I figured adding support for -nostdlib++, even though redundant, makes
some sense. One could presumably use gcc rather than g++ for linking,
for the same effect, but sometimes changing the link command is harder
than adding an option, as in our testsuite.
Since clang already had an option with this effect, we've adopted the
same spelling.
for gcc/ChangeLog
* common.opt (nostdlib++): New.
* doc/invoke.texi (-nostdlib++): Document it.
Alexandre Oliva [Fri, 24 Jun 2022 02:20:47 +0000 (23:20 -0300)]
libstdc++: testsuite: avoid predictable mkstemp
This patch was originally meant to reduce the likelihood that
nonexistent_path() returns the same pathname for from and to.
It was prompted by a target system with a non-random implementation of
mkstemp, that returns a predictable sequence of filenames and selects
the first one that isn't already taken.
That turned out not to be enough: nonexistent_path adds a suffix to
the filename chosen by mkstemp and removes the file it created, so
mkstemp may very well insist on the same basename, and the case that
doesn't use mkstemp doesn't even check whether the file already
exists.
Anyway, by the time I realized this wasn't enough, I'd already
implemented some of the changes, and I figured I might as well
contribute them, even though they don't really solve any problem, and
even if they did, they'd be just a partial solution.
for libstdc++-v3/ChangeLog
* testsuite/27_io/filesystem/operations/copy.cc (test02):
Select TO after creating FROM.
(test03, test04): Likewise.
* testsuite/experimental/filesystem/operations/copy.cc
(test02, test03, test04): Likewise.
Though sleep, nanosleep and clock_nanosleep are all POSIX cancellation
points, not all target systems follow this POSIX requirement.
30_threads/thread/native_handle/cancel.cc will run until it times out
on such systems.
Rather than failing a C++ library test because of a limitation of the
target system, this patch gives the test a chance to successfully
exercise the features it intends to exercise, by introducing a
cancellation point in a loop that would otherwise run indefinitely on
systems exhibiting this limitation.
for libstdc++-v3/ChangeLog
* testsuite/30_threads/thread/native_handle/cancel.cc: Add an
explicit cancellation point in case sleep_for lacks one.
Alexandre Oliva [Fri, 24 Jun 2022 02:20:45 +0000 (23:20 -0300)]
aarch64: testsuite: symbol-range fallback to compile
On some of our embedded aarch64 targets, RAM size is too small for
this test to fit. It doesn't look like this test requires linking,
and if it does, the -tiny version may presumably get most of the
coverage without going overboard in target system requirements.
Still, linking may be useful, so introduce a two_plus_gigs effective
target, that checks for the ability to link a program with 2GiB of
sbss, and use that to select whether to link or just compile
symbol-range.c.
* lib/target-supports.exp
(check_effective_target_two_plus_gigs): New.
* gcc.target/aarch64/symbol-range.c: Link only on
two_plus_gigs targets, compile otherwise.
Jason Merrill [Thu, 23 Jun 2022 18:41:19 +0000 (14:41 -0400)]
c++: designated init cleanup [PR105925]
build_aggr_conv expects to run after reshape_init, which will usually have
filled out all the CONSTRUCTOR indexes; there's no reason to limit using
those to the case where the user gave an explicit designator.
PR c++/105925
gcc/cp/ChangeLog:
* call.cc (build_aggr_conv): Don't depend on
CONSTRUCTOR_IS_DESIGNATED_INIT.
Jason Merrill [Thu, 23 Jun 2022 20:04:02 +0000 (16:04 -0400)]
c++: anon union designated init [PR105925]
This testcase was failing because CONSTRUCTOR_IS_DESIGNATED_INIT wasn't
getting set on the introduced CONSTRUCTOR for the anonymous union, and
build_aggr_conv uses that flag to decide whether to pay attention to the
indexes of the CONSTRUCTOR. So set the flag when we see a designator rather
than relying on copying it from another CONSTRUCTOR.
This avoids some redundant errors on desig4.C because we stop setting
CONSTRUCTOR_IS_DESIGNATED_INIT on _Complex CONSTRUCTORs where it's
nonsense.
PR c++/105925
gcc/cp/ChangeLog:
* decl.cc (reshape_init_array_1): Set
CONSTRUCTOR_IS_DESIGNATED_INIT here.
(reshape_init_class): And here.
(reshape_init): Not here.
gcc/testsuite/ChangeLog:
* g++.dg/ext/desig4.C: Remove extra errors.
* g++.dg/cpp2a/desig26.C: New test.
Patrick Palka [Thu, 23 Jun 2022 20:36:43 +0000 (16:36 -0400)]
c++: constexpr folding in unevaluated context [PR105931]
Changing the type of N from int to unsigned in decltype82.C (from r13-986-g0ecb6b906f215e) reveals another spot where we perform constexpr
evaluation in an unevaluated context for sake of warnings, this time
from the call to shorten_compare in cp_build_binary_op, which calls
fold_for_warn.
We could (and probably should) suppress the shorten_compare warnings
when in an unevaluated context, but there's probably other callers of
fold_for_warn that are similarly affected. So this patch takes the
approach of directly suppressing fold_for_warn when in an unevaluated
context.
PR c++/105931
gcc/cp/ChangeLog:
* expr.cc (fold_for_warn): Don't fold when in an unevaluated
context.
Patrick Palka [Thu, 23 Jun 2022 20:18:55 +0000 (16:18 -0400)]
c++: context completion in lookup_template_class [PR105982]
The below testcase demonstrates that completion of the substituted
context during lookup_template_class can end up registering the desired
specialization for us in more cases than r13-1045-gcb7fd1ea85feea
anticipated. In particular this can happen for a non-dependent
specialization of a nested class as well.
For this testcase, during overload resolution with A's guides, we
substitute the deduced argument T=int into the TYPENAME_TYPE B::C,
during which we call lookup_template_class for A<T>::B with T=int,
which completes A<int> for the first time, which recursively registers
the desired specialization of B already. The parent call to
lookup_template_class then tries to register the same specialization,
triggering an ICE.
This patch fixes this by making lookup_template_class determine more
directly whether we need to recheck the specializations table after
completion of the context -- when and only when the call to complete_type
had an effect.
PR c++/105982
gcc/cp/ChangeLog:
* pt.cc (lookup_template_class): After calling complete_type for
the substituted context, check the table again iff the type was
previously incomplete and complete_type made it complete.
David Malcolm [Thu, 23 Jun 2022 18:59:24 +0000 (14:59 -0400)]
diagnostics: add ability to associate diagnostics with rules from coding standards
gcc/ChangeLog:
* common.opt (fdiagnostics-show-rules): New option.
* diagnostic-format-json.cc (diagnostic_output_format_init_json):
Fix up context->show_rules.
* diagnostic-format-sarif.cc
(diagnostic_output_format_init_sarif): Likewise.
* diagnostic-metadata.h (diagnostic_metadata::rule): New class.
(diagnostic_metadata::precanned_rule): New class.
(diagnostic_metadata::add_rule): New.
(diagnostic_metadata::get_num_rules): New.
(diagnostic_metadata::get_rule): New.
(diagnostic_metadata::m_rules): New field.
* diagnostic.cc (diagnostic_initialize): Initialize show_rules.
(print_any_rules): New.
(diagnostic_report_diagnostic): Call it.
* diagnostic.h (diagnostic_context::show_rules): New field.
* doc/invoke.texi (-fno-diagnostics-show-rules): New option.
* opts.cc (common_handle_option): Handle
OPT_fdiagnostics_show_rules.
* toplev.cc (general_init): Set up global_dc->show_rules.
gcc/testsuite/ChangeLog:
* gcc.dg/plugin/diagnostic-test-metadata.c: Expect " [STR34-C]" to
be emitted at the "gets" call.
* gcc.dg/plugin/diagnostic_plugin_test_metadata.c
(pass_test_metadata::execute): Associate the "gets" diagnostic
with a rule named "STR34-C".
Signed-off-by: David Malcolm <dmalcolm@redhat.com>
Jonathan Wakely [Thu, 23 Jun 2022 13:25:49 +0000 (14:25 +0100)]
libstdc++: Properly remove temporary directories in filesystem tests
Although these tests use filesystem::remove_all to clean up, that fails
because it uses recursive_directory_iterator which is intentionally
bodged by the custom readdir defined in the test.
Just use POSIX rmdir to clean up. We don't need to use _rmdir or _wrmdir
for Windows, because we'll never reach test02() on targets where the
custom readdir doesn't interpose the one from libc.
libstdc++-v3/ChangeLog:
* testsuite/27_io/filesystem/iterators/error_reporting.cc: Use
rmdir to remove directories.
* testsuite/experimental/filesystem/iterators/error_reporting.cc:
Likewise.
Jason Merrill [Thu, 23 Jun 2022 04:24:22 +0000 (00:24 -0400)]
c++: -Waddress and if constexpr [PR94554]
Like we avoid various warnings for seemingly tautological expressions when
substituting a template, we should avoid warning for the implicit conversion
to bool in an if statement. I considered also doing this for the conditions
in loop expressions, but that seems unnecessary, as a loop condition is
unlikely to be a constant.
The change to finish_if_stmt_cond isn't necessary since dependent_operand_p
looks through IMPLICIT_CONV_EXPR, but makes it more constent with
e.g. build_x_binary_op that determines the type of an expression and then
builds it using the original operands.
PR c++/94554
gcc/cp/ChangeLog:
* pt.cc (dependent_operand_p): Split out from...
(tsubst_copy_and_build): ...here.
(tsubst_expr) [IF_STMT]: Use it.
* semantics.cc (finish_if_stmt_cond): Keep the pre-conversion
condition in the template tree.
Jason Merrill [Thu, 23 Jun 2022 03:50:23 +0000 (23:50 -0400)]
c++: -Waddress and value-dependent expr [PR105885]
We already suppress various warnings for code that would be tautological if
written directly, but not when it's the result of template substitution. It
seems we need to do this for -Waddress as well.
PR c++/105885
gcc/cp/ChangeLog:
* pt.cc (tsubst_copy_and_build): Also suppress -Waddress for
comparison of dependent operands.
Nathan Sidwell [Wed, 22 Jun 2022 12:54:30 +0000 (05:54 -0700)]
c++: Prune unneeded macro locations
This implements garbage collection on locations within macro
expansions, when streaming out a CMI. When doing the reachability
walks, we now note which macro locations we need and then only write
those locations. The complication here is that every macro expansion
location has an independently calculated offset. This complicates
writing, but reading remains the same -- the macro locations of a CMI
continue to form a contiguous block.
For std headers this reduced the number of macro maps by 40% and the
number of locations by 16%. For a GMF including iostream, it reduced
it by 80% and 60% respectively.
Ordinary locations are still transformed en-mass. They are somewhat
more complicated to apply a similar optimization to.
testsuite: Compile slsr-39.c without vectorisation
The fix for PR106019 regressed slsr-39.c for -m32 -march=cascadelake
because we are now able to vectorise the code. (Whether the code model
should be allowing that is a different question -- the vectorised code
looked worse to me.)
The test runs at -O2 and predates vectorisation being enabled at -O2,
so this patch just adds -fno-tree-vectorize.
gcc/testsuite/
* gcc.dg/tree-ssa/slsr-39.c: Force vectorization off.
Jason Merrill [Thu, 23 Jun 2022 01:29:16 +0000 (21:29 -0400)]
c++: tweak deduction with auto template parms
While looking at PR105964 I noticed that we were unnecessarily repeating
the deduction loop because of seeing a non-type parameter with type 'auto'.
It is indeed dependent, but not on any other deductions.
gcc/cp/ChangeLog:
* pt.cc (type_unification_real): An auto tparm can't
be affected by other deductions.
Jason Merrill [Wed, 22 Jun 2022 22:19:11 +0000 (18:19 -0400)]
c++: dependence of baselink [PR105964]
helper<token>::c isn't dependent just because we haven't deduced its return
type yet. type_dependent_expression_p already knows how to deal with that
for bare FUNCTION_DECL, but needs to learn to look through a BASELINK.
PR c++/105964
gcc/cp/ChangeLog:
* pt.cc (type_dependent_expression_p): Look through BASELINK.
Jason Merrill [Wed, 22 Jun 2022 18:57:21 +0000 (14:57 -0400)]
c++: class scope function lookup [PR105908]
In r12-1273 for PR91706, I removed the code in get_class_binding that
stripped BASELINK. This testcase demonstrates that we still need to strip
it in outer_binding before putting the overload set in IDENTIFIER_BINDING,
for compatibility with bindings added directly for declarations.
Nathan Sidwell [Wed, 22 Jun 2022 14:51:44 +0000 (07:51 -0700)]
c++: Remove ifdefed code
The only reason I chose to use DECL_UID on this hash table was to make
it stable against ASLR and perturbations due to other allocations.
It's not required for correctness, as the comment mentions the
equality fn uses pointer identity.
Jason Merrill [Fri, 10 Jun 2022 20:35:21 +0000 (16:35 -0400)]
ubsan: default to trap on unreachable at -O0 and -Og [PR104642]
When not optimizing, we can't do anything useful with unreachability in
terms of code performance, so we might as well improve debugging by turning
__builtin_unreachable into a trap. I think it also makes sense to do this
when we're explicitly optimizing for the debugging experience.
In the PR richi suggested introducing an -funreachable-traps flag for this.
This functionality is already implemented as -fsanitize=unreachable
-fsanitize-trap=unreachable, and we want to share the implementation, but it
does seem useful to have a separate flag that isn't affected by the various
sanitization controls. -fsanitize=unreachable takes priority over
-funreachable-traps if both are enabled.
Jakub observed that this would slow down -O0 by default from running the
sanopt pass, so this revision avoids the need for sanopt by rewriting calls
introduced by the compiler immediately, and calls written by the user at
fold time. Many of the calls introduced by the compiler are also rewritten
immediately to ubsan calls when not trapping, which fixes ubsan-8b.C;
previously the call to f() was optimized away before sanopt. But this early
rewriting isn't practical for uses of __builtin_unreachable in
devirtualization and such, so sanopt rewriting is still done for
non-trapping sanitize.
PR c++/104642
gcc/ChangeLog:
* common.opt: Add -funreachable-traps.
* doc/invoke.texi (-funreachable-traps): Document it.
* opts.cc (finish_options): Enable at -O0 or -Og.
* tree.cc (build_common_builtin_nodes): Add __builtin_trap.
(builtin_decl_unreachable, build_builtin_unreachable): New.
* tree.h: Declare them.
* ubsan.cc (sanitize_unreachable_fn): Factor out.
(ubsan_instrument_unreachable): Use
gimple_build_builtin_unreachable.
* ubsan.h (sanitize_unreachable_fn): Declare.
* gimple.cc (gimple_build_builtin_unreachable): New.
* gimple.h: Declare it.
* builtins.cc (expand_builtin_unreachable): Add assert.
(fold_builtin_0): Call build_builtin_unreachable.
* sanopt.cc: Don't run for just SANITIZE_RETURN
or SANITIZE_UNREACHABLE when trapping.
* cgraphunit.cc (walk_polymorphic_call_targets): Use new
unreachable functions.
* gimple-fold.cc (gimple_fold_call)
(gimple_get_virt_method_for_vtable)
* ipa-fnsummary.cc (redirect_to_unreachable)
* ipa-prop.cc (ipa_make_edge_direct_to_target)
(ipa_impossible_devirt_target)
* ipa.cc (walk_polymorphic_call_targets)
* tree-cfg.cc (pass_warn_function_return::execute)
(execute_fixup_cfg)
* tree-ssa-loop-ivcanon.cc (remove_exits_and_undefined_stmts)
(unloop_loops)
* tree-ssa-sccvn.cc (eliminate_dom_walker::eliminate_stmt):
Likewise.
* g++.dg/ubsan/return-8a.C: New test.
* g++.dg/ubsan/return-8b.C: New test.
* g++.dg/ubsan/return-8d.C: New test.
* g++.dg/ubsan/return-8e.C: New test.
When dr_may_alias_p is called without a loop context, it tries
to use the tree-affine interface to calculate the difference
between the two addresses and use that difference to check whether
the gap between the accesses is known at compile time. However, as the
example in the PR shows, this doesn't expand SSA_NAMEs and so can easily
be defeated by things like reassociation.
One fix would have been to use aff_combination_expand to expand the
SSA_NAMEs, but we'd then need some way of maintaining the associated
cache. This patch instead reuses the innermost_loop_behavior fields
(which exist even when no loop context is provided).
It might still be useful to do the aff_combination_expand thing too,
if an example turns out to need it.
gcc/
PR tree-optimization/106019
* tree-data-ref.cc (dr_may_alias_p): Try using the
innermost_loop_behavior to disambiguate non-loop queries.
gcc/testsuite/
PR tree-optimization/106019
* gcc.dg/vect/bb-slp-pr106019.c: New test.
Alexandre Oliva [Wed, 22 Jun 2022 02:11:02 +0000 (23:11 -0300)]
libstdc++: eh_globals: gthreads: reset _S_init before deleting key
Clear __eh_globals_init's _S_init in the dtor before deleting the
gthread key.
This ensures that, in case any code involved in deleting the key
interacts with eh_globals, the key that is being deleted won't be
used, and the non-thread-specific eh_globals fallback will.
Alexandre Oliva [Wed, 22 Jun 2022 02:11:01 +0000 (23:11 -0300)]
libstdc++: testsuite: call sched_yield for nonpreemptive targets
As in the gcc testsuite, systems without preemptive multi-threading
require sched_yield calls to be placed at points in which a context
switch might be needed to enable the test to complete.
Alexandre Oliva [Wed, 22 Jun 2022 02:11:00 +0000 (23:11 -0300)]
libstdc++: testsuite: work around bitset namespace pollution
rtems6 declares a global struct bitset in a header file included
indirectly by sys/types.h, that ambiguates the unqualified references
to bitset after "using namespace std" in the testsuite.
Work around the namespace pollution with using declarations of
std::bitset.
for libstdc++-v3/ChangeLog
* testsuite/23_containers/bitset/cons/dr1325-2.cc: Work around
global struct bitset.
* testsuite/23_containers/bitset/ext/15361.cc: Likewise.
* testsuite/23_containers/bitset/input/1.cc: Likewise.
* testsuite/23_containers/bitset/to_string/1.cc: Likewise.
* testsuite/23_containers/bitset/to_string/dr396.cc: Likewise.
Alexandre Oliva [Wed, 22 Jun 2022 02:10:59 +0000 (23:10 -0300)]
testsuite: outputs.exp: cleanup before running tests
Use the just-added dry-run infrastructure to clean up files that may
have been left over by interrupted runs of outputs.exp, which used to
lead to spurious non-repeatable (self-fixing) failures.
for gcc/testsuite/ChangeLog
* gcc.misc-tests/outputs.exp: Clean up left-overs first.
Alexandre Oliva [Wed, 22 Jun 2022 02:10:58 +0000 (23:10 -0300)]
testsuite: outputs.exp: test for skip_atsave more thoroughly
The presence of -I or -L flags in link command lines changes the
driver's, and thus the linker's behavior, WRT naming files with
command-line options. With such flags, the driver creates .args.0 and
.args.1 files, whereas without them it's the linker (collect2, really)
that creates .ld1_args.
I've hit some fails on a target system that doesn't have -I or -L
flags in the board config file, but it does add some of them
implicitly with configured-in driver self specs. Alas, the test in
outputs.exp doesn't catch that, so we proceed to run rather than
skip_atsave tests.
I've reworked the outest procedure to allow dry runs and to return
would-have-been pass/fail results as lists, so we can now test whether
certain files are created and use that to configure the actual test
runs.
for gcc/testsuite/ChangeLog
* gcc.misc-tests/outputs.exp (outest): Introduce quiet mode,
create and return lists of passes and fails. Use it to catch
skip_atsave cases where -L flags are implicitly added by
driver self specs.
Alexandre Oliva [Wed, 22 Jun 2022 02:10:58 +0000 (23:10 -0300)]
c++: testsuite: require lto_incremental in pr90990_0.C
Other LTO tests that use -r require the lto_incremental effective
target. I suppose pr90990_0.C is missing it due to an oversight.
This patch arranges for this test to also be skipped on
non-lto_incremental targets.
Roger Sayle [Tue, 21 Jun 2022 23:08:56 +0000 (00:08 +0100)]
PR target/105991: Recognize PLUS and XOR forms of rldimi in rs6000.md.
This patch addresses PR target/105991 where a change to prefer representing
shifts and adds at the tree-level as multiplications, causes problems for
the rldimi patterns in the powerpc backend. The issue is that rs6000.md
models this pattern using IOR, and some variants that have the equivalent
PLUS or XOR in the RTL fail to match some *rotl<mode>4_insert patterns.
This is fixed in this patch by adding a define_insn_and_split to locally
canonicalize the PLUS and XOR forms to the backend's preferred IOR form.
An alternative fix might be for the RTL optimizers to define a canonical
form for these plus_xor_ior equivalent expressions, but the logical
choice might be plus (which may appear in an addressing mode), and such
a change may require a number of tweaks to update various backends
(i.e. a more intrusive change than the one proposed here).
Many thanks for Marek Polacek for bootstrapping and regression testing
this change without problems.
2022-06-22 Roger Sayle <roger@nextmovesoftware.com>
Marek Polacek <polacek@redhat.com>
Segher Boessenkool <segher@kernel.crashing.org>
Kewen Lin <linkw@linux.ibm.com>
gcc/ChangeLog
PR target/105991
* config/rs6000/rs6000.md (rotl<mode>3_insert_3): Check that
exact_log2 doesn't return -1 (or zero).
(plus_xor): New code iterator.
(*rotl<mode>3_insert_3_<code>): New define_insn_and_split.
gcc/testsuite/ChangeLog
PR target/105991
* gcc.target/powerpc/pr105991.c: New test case.
Jakub Jelinek [Tue, 21 Jun 2022 15:51:08 +0000 (17:51 +0200)]
libgomp: Fix up target-31.c test [PR106045]
The i variable is used inside of the parallel in:
#pragma omp simd safelen(32) private (v)
for (i = 0; i < 64; i++)
{
v = 3 * i;
ll[i] = u1 + v * u2[0] + u2[1] + x + y[0] + y[1] + v + h[0] + u3[i];
}
where i is predetermined linear (so while inside of the body
it is safe, private per SIMD lane var) the final value is written to
the shared variable, and in:
for (i = 0; i < 64; i++)
if (ll[i] != u1 + 3 * i * u2[0] + u2[1] + x + y[0] + y[1] + 3 * i + 13 + 14 + i)
#pragma omp atomic write
err = 1;
which is a normal loop and so it isn't in any way privatized there.
So we have a data race, fixed by adding private (i) clause to the
parallel.
2022-06-21 Jakub Jelinek <jakub@redhat.com>
Paul Iannetta <piannetta@kalrayinc.com>
Arjun Shankar [Tue, 21 Jun 2022 10:12:11 +0000 (12:12 +0200)]
match.pd: Remove "+ 0x80000000" in int comparisons [PR94899]
Expressions of the form "X + CST < Y + CST" where:
* CST is an unsigned integer constant with only the MSB set, and
* X and Y's types have integer conversion ranks <= CST's
can be simplified to "(signed) X < (signed) Y".
This is because, assuming a 32-bit signed numbers,
(unsigned) INT_MIN + 0x80000000 is 0, and
(unsigned) INT_MAX + 0x80000000 is UINT_MAX.
i.e. the result increases monotonically with signed input.
This means:
((signed) X < (signed) Y) iff (X + 0x80000000 < Y + 0x80000000)
gcc/
PR tree-optimization/94899
* match.pd (X + C < Y + C -> (signed) X < (signed) Y, if C is
0x80000000): New simplification.
gcc/testsuite/
* gcc.dg/pr94899.c: New test.
Jakub Jelinek [Tue, 21 Jun 2022 09:40:16 +0000 (11:40 +0200)]
ifcvt: Don't introduce trapping or faulting reads in noce_try_sign_mask [PR106032]
noce_try_sign_mask as documented will optimize
if (c < 0)
x = t;
else
x = 0;
into x = (c >> bitsm1) & t;
The optimization is done if either t is unconditional
(e.g. for
x = t;
if (c >= 0)
x = 0;
) or if it is cheap. We already check that t doesn't have side-effects,
but if t is conditional, we need to punt also if it may trap or fault,
as we make it unconditional.
I've briefly skimmed other noce_try* optimizations and didn't find one that
would suffer from the same problem.
2022-06-21 Jakub Jelinek <jakub@redhat.com>
PR rtl-optimization/106032
* ifcvt.cc (noce_try_sign_mask): Punt if !t_unconditional, and
t may_trap_or_fault_p, even if it is cheap.
Jakub Jelinek [Tue, 21 Jun 2022 09:38:59 +0000 (11:38 +0200)]
expand: Fix up expand_cond_expr_using_cmove [PR106030]
If expand_cond_expr_using_cmove can't find a cmove optab for a particular
mode, it tries to promote the mode and perform the cmove in the promoted
mode.
The testcase in the patch ICEs on arm because in that case we pass temp which
has the promoted mode (SImode) as target to expand_operands where the
operands have the non-promoted mode (QImode).
Later on the function uses paradoxical subregs:
if (GET_MODE (op1) != mode)
op1 = gen_lowpart (mode, op1);
if (GET_MODE (op2) != mode)
op2 = gen_lowpart (mode, op2);
to change the operand modes.
The following patch fixes it by passing NULL_RTX as target if it has
promoted mode.
2022-06-21 Jakub Jelinek <jakub@redhat.com>
PR middle-end/106030
* expr.cc (expand_cond_expr_using_cmove): Pass NULL_RTX instead of
temp to expand_operands if mode has been promoted.
Xionghu Luo [Thu, 9 Jun 2022 07:46:30 +0000 (15:46 +0800)]
if-to-switch: Don't skip the first condition bb when find_conditions in if-to-switch [PR105740]
The if condition is at last of first bb, so side effect statement in first BB
doesn't matter, then the first if condition could also be folded to switch
table.
gcc/ChangeLog:
PR target/105740
* gimple-if-to-switch.cc (find_conditions): Don't skip the first
condition bb.
gcc/testsuite/ChangeLog:
PR target/105740
* gcc.dg/tree-ssa/if-to-switch-11.c: New test.
Signed-off-by: Xionghu Luo <xionghuluo@tencent.com>
tree-object-size: Don't let error_mark_node escape for ADDR_EXPR [PR105736]
The addr_expr computation does not check for error_mark_node before
returning the size expression. This used to work in the constant case
because the conversion to uhwi would end up causing it to return
size_unknown, but that won't work for the dynamic case.
Modify the control flow to explicitly return size_unknown if the offset
computation returns an error_mark_node.
gcc/ChangeLog:
PR tree-optimization/105736
* tree-object-size.cc (addr_object_size): Return size_unknown
when object offset computation returns an error.
gcc/testsuite/ChangeLog:
PR tree-optimization/105736
* gcc.dg/builtin-dynamic-object-size-0.c (TV4): New struct.
(val3): New variable.
(test_pr105736): New test.
(main): Call it.
Iain Sandoe [Sun, 19 Jun 2022 09:31:01 +0000 (10:31 +0100)]
testsuite, asan: Avoid color in asan test output.
The presence of the color markers in the some of the asan tests
appears to confuse the dg-output matching (possibly a platform
TCL or termios bug) on some Darwin platforms.
Since the color is not being tested, switch it off (makes the log
files easier to read too). This fixes a large number of spurious
test fails on AVX512 Darwin19.
Iain Sandoe [Sun, 19 Jun 2022 09:21:28 +0000 (10:21 +0100)]
testsuite, Darwin: Allow for two CTOR bodies in array61 test.
For targets without alias support, we emit two essentially identical function
bodies into the gimple (complete and base CTORs). So this test needs to allow
for that when the target does not support aliases. The target support alias
test does not seem to be usable in the context of a single scan-tree-dump so
the fix here uses the target designation.
Note that the array has 10 elements, so that if the test were failing (because
we were emitting 10 inits instead of a loop) the count would be expected to
exceed 2, on Darwin and 1 where there's alias support.
Richard Earnshaw [Mon, 20 Jun 2022 15:18:36 +0000 (16:18 +0100)]
arm: more testsutie fallout for mve move-immediate changes
Unfortunately, there is more fall-out in the testsuite for my changes
to use MVE move-immediate operations instead of literal pool loads.
Fixed as follows:
Iain Sandoe [Sun, 19 Jun 2022 09:12:47 +0000 (10:12 +0100)]
testsuite: Require init_priority target support in a test.
The attr-cdtor-1 test fails on targets without init priority since the
diagnostic emitted concerns the absence of support. Disable the test
on such targets.
Richard Biener [Mon, 20 Jun 2022 11:40:50 +0000 (13:40 +0200)]
middle-end/106027 - fix types in needle folding
The fold_to_nonsharp_ineq_using_bound folding ends up creating invalid
typed IL which confuses later foldings. The following fixes that.
2022-06-20 Richard Biener <rguenther@suse.de>
PR middle-end/106027
* fold-const.cc (fold_to_nonsharp_ineq_using_bound): Use the
type of the prevailing comparison for the new comparison type.
(fold_binary_loc): Use proper types for the A < X && A + 1 > Y
to A < X && A >= Y folding.
Kewen Lin [Mon, 20 Jun 2022 12:44:21 +0000 (07:44 -0500)]
vect: Respect slp decision when applying suggested uf [PR105940]
This follows Richi's suggestion in PR105940, it aims to avoid
inconsistent slp decision between when the suggested unroll
factor is worked out and when the suggested unroll factor is
applied.
If the previous slp decision is true when the suggested unroll
factor is worked out, when we are applying unroll factor we
don't need to start over with slp off if the analysis with slp
on fails. On the other hand, if the previous slp decision is
false when the suggested unroll factor is worked out, when we
are applying unroll factor we can skip the slp handlings.
Function vect_is_simple_reduction saves reduction chains for
subsequent slp analyses, we have to disable this early otherwise
there is an ICE in vectorizable_reduction for below:
if (REDUC_GROUP_FIRST_ELEMENT (stmt_info))
gcc_assert (slp_node
&& REDUC_GROUP_FIRST_ELEMENT (stmt_info)
== stmt_info);
PR tree-optimization/105940
gcc/ChangeLog:
* tree-vect-loop.cc (vect_analyze_loop_2): Add new parameter
slp_done_for_suggested_uf and adjust with it accordingly.
(vect_analyze_loop_1): Add new variable slp_done_for_suggested_uf,
pass it down to vect_analyze_loop_2 for the initial analysis and
applying suggested unroll factor.
(vect_is_simple_reduction): Add parameter slp and adjust with it.
(vect_analyze_scalar_cycles_1): Add parameter slp and pass down.
(vect_analyze_scalar_cycles): Likewise.
yulong [Wed, 8 Jun 2022 02:19:21 +0000 (10:19 +0800)]
RISC-V: Fix a bug that is the CMO builtins are missing parameter
We changed builtins format about zicbom and zicboz subextensions and modified test cases.
diff with the previous version:
1.We modified the FUNCTION_TYPE from RISCV_VOID_FTYPE_SI/DI to RISCV_VOID_FTYPE_VOID_PTR.
2.We added a new RISCV_ATYPE_VOID_PTR in riscv-builtins.cc and a new DEF_RISCV_FTYPE (1, (VOID, VOID_PTR)) in riscv-ftypes.def.
3.We deleted DEF_RISCV_FTYPE (1, (VOID, SI/DI)).
4.We modified the input parameters of the test cases.
Thanks, Simon and Kito.
gcc/ChangeLog:
* config/riscv/riscv-builtins.cc (RISCV_ATYPE_VOID_PTR): New.
* config/riscv/riscv-cmo.def (RISCV_BUILTIN): Changed the FUNCTION_TYPE
of RISCV_BUILTIN.
* config/riscv/riscv-ftypes.def (0): Remove unused.
(1): New.
gcc/testsuite/ChangeLog:
* gcc.target/riscv/cmo-zicbom-1.c: modified the input parameters.
* gcc.target/riscv/cmo-zicbom-2.c: modified the input parameters.
* gcc.target/riscv/cmo-zicboz-1.c: modified the input parameters.
* gcc.target/riscv/cmo-zicboz-2.c: modified the input parameters.
projects / thirdparty / gcc.git / log
