Marek Polacek [Wed, 13 Jan 2021 18:12:14 +0000 (13:12 -0500)]
c++: Crash when deducing template arguments [PR98659]
maybe_instantiate_noexcept doesn't expect to see error_mark_node, but
the new callsite I introduced in r11-6476 can pass error_mark_node to
it. So cope.
gcc/cp/ChangeLog:
PR c++/98659
* pt.c (maybe_instantiate_noexcept): Return false if FN is
error_mark_node.
gcc/testsuite/ChangeLog:
PR c++/98659
* g++.dg/template/deduce8.C: New test.
Marek Polacek [Fri, 15 Jan 2021 03:14:38 +0000 (22:14 -0500)]
c++: ICE with USING_DECL redeclaration [PR98687]
My recent patch that introduced push_using_decl_bindings didn't
handle USING_DECL redeclaration, therefore things broke. This patch
amends that by breaking out a part of finish_nonmember_using_decl
out to a separate function, push_using_decl_bindings, and calling it.
It needs an overload, because name_lookup is only available inside
of name-lookup.c.
gcc/cp/ChangeLog:
PR c++/98687
* name-lookup.c (push_using_decl_bindings): New, broken out of...
(finish_nonmember_using_decl): ...here.
* name-lookup.h (push_using_decl_bindings): Update declaration.
* pt.c (tsubst_expr): Update the call to push_using_decl_bindings.
gcc/testsuite/ChangeLog:
PR c++/98687
* g++.dg/lookup/using64.C: New test.
* g++.dg/lookup/using65.C: New test.
Martin Sebor [Tue, 19 Jan 2021 22:10:30 +0000 (15:10 -0700)]
PR middle-end/98664 - inconsistent -Wfree-nonheap-object for inlined calls to system headers
gcc/ChangeLog:
PR middle-end/98664
* tree-ssa-live.c (remove_unused_scope_block_p): Keep scopes for
all functions, even if they're not declared artificial or inline.
* tree.c (tree_inlined_location): Use macro expansion location
only if scope traversal fails to expose one.
gcc/testsuite/ChangeLog:
PR middle-end/98664
* gcc.dg/Wvla-larger-than-4.c: Adjust expected output.
* gcc.dg/plugin/diagnostic-test-inlining-3.c: Same.
* g++.dg/warn/Wfree-nonheap-object-5.C: New test.
* gcc.dg/Wfree-nonheap-object-4.c: New test.
Patrick Palka [Tue, 19 Jan 2021 21:20:00 +0000 (16:20 -0500)]
c++: Always check access during late-parsing of members [PR58993]
This patch removes a vestigial use of dk_no_check from
cp_parser_late_parsing_for_member, which ideally should have been
removed as part of the PR41437 patch that improved access checking
inside templates. This allows us to correctly reject f1 and f2 in
the testcase access34.C below (whereas before we'd only reject f3).
Additional testing revealed a new access issue when late-parsing a hidden
friend within a class template. In the testcase friend68.C below, we're
tripping over the checking assert from friend_accessible_p(f, S::j, S, S)
during lookup of j in x.j (for which type_dependent_object_expression_p
returns false, which is why we're doing the lookup at parse time). The
reason for the assert failure is that DECL_FRIENDLIST(S) contains f but
DECL_BEFRIENDING_CLASSES(f) is empty, and so friend_accessible_p (which
looks at DECL_BEFRIENDING_CLASSES) wants to return false, but is_friend
(which looks at DECL_FRIENDLIST) returns true.
For sake of symmetry one would expect that DECL_BEFRIENDING_CLASSES(f)
contains S, but add_friend avoids updating DECL_BEFRIENDING_CLASSES when
the class type (S in this case) is dependent, for some reason.
This patch works around this issue by making friend_accessible_p
consider the DECL_FRIEND_CONTEXT of the access scope. Thus we sidestep
the DECL_BEFRIENDING_CLASSES / DECL_FRIENDLIST asymmetry issue while
correctly validating the x.j access at parse time.
A earlier version of this patch checked friend_accessible_p instead of
protected_accessible_p in the DECL_FRIEND_CONTEXT hunk below, but this
had the side effect of making us accept the ill-formed testcase friend69.C
below (ill-formed because the hidden friend g is not actually a member
of A, so g doesn't have access to B's members despite B befriending A).
gcc/cp/ChangeLog:
PR c++/41437
PR c++/58993
* search.c (friend_accessible_p): If scope is a hidden friend
defined inside a dependent class, consider access from the
class.
* parser.c (cp_parser_late_parsing_for_member): Don't push a
dk_no_check access state.
gcc/testsuite/ChangeLog:
PR c++/41437
PR c++/58993
* g++.dg/opt/pr87974.C: Adjust.
* g++.dg/template/access34.C: New test.
* g++.dg/template/friend68.C: New test.
* g++.dg/template/friend69.C: New test.
Marek Polacek [Fri, 8 Jan 2021 20:48:41 +0000 (15:48 -0500)]
c++: ICE when late parsing noexcept/NSDMI [PR98333]
Since certain members of a class are a complete-class context
[class.mem.general]p7, we delay their parsing untile the whole class has
been parsed. For instance, NSDMIs and noexcept-specifiers. The order
in which we perform this delayed parsing matters; we were first parsing
NSDMIs and only they did we parse noexcept-specifiers. That turns out
to be wrong: since NSDMIs may use noexcept-specifiers, we must process
noexcept-specifiers first. Otherwise we'll ICE in code that doesn't
expect to see DEFERRED_PARSE.
This doesn't just shift the problem, noexcept-specifiers can use members
with a NSDMI just fine, and I've also tested a similar test with this
member function:
In other words, the accesses can alias only if XB == YB at runtime.
However, this doesn't cope correctly with section anchors.
For example, if XB is an anchor symbol and YB is at offset
XO from the anchor, then:
[XB + XO, XB + XO + XS)
overlaps
[YB, YB + YS)
whatever the value of XO is. In other words, when doing the
alias check for two symbols whose local definitions are in
the same block, we should apply the known difference between
their block offsets to the intersection test above.
gcc/
PR rtl-optimization/92294
* alias.c (compare_base_symbol_refs): Take an extra parameter
and add the distance between two symbols to it. Enshrine in
comments that -1 means "either 0 or 1, but we can't tell
which at compile time".
(memrefs_conflict_p): Update call accordingly.
(rtx_equal_for_memref_p): Likewise. Take the distance between symbols
into account.
Will Schmidt [Wed, 13 Jan 2021 19:48:30 +0000 (13:48 -0600)]
[PATCH, rs6000] Update pr88233.c test (pr91799)
Hi,
This is a follow-up fix to clean up pr91799. Per review of test results,
it appears that the combination of target and dg-require stanzas is
not sufficient to properly limit the test to 64-bit only on darwin.
This adds an additional dg-require clause to limit the test to 64-bit
environments.
Tested on power7 and power8 using assorted variations of
make -k check-gcc-c "RUNTESTFLAGS=powerpc.exp=pr88233.c
--target_board=unix/'{-mcpu=power7,-mcpu=power6,-mcpu=power8}''{-m32,-m64}'"
PR target/91799
2021-01-19 Will Schmidt <will_schmidt@vnet.ibm.com>
Kyrylo Tkachov [Tue, 19 Jan 2021 15:36:55 +0000 (15:36 +0000)]
aarch64: Remove testing of saturation cumulative QC bit
Since we don't guarantee the ordering of the QC flag in FPSR in the
saturation intrinsics, we shouldn't be testing for it.
I want to relax the flags for some of the builtins to enable more
optimisation but that triggers the QC flag tests in
advsimd-intrinsics.exp.
We don't implement the saturation flag access intrinsics in aarch64
anyway and we don't want to.
Richard Biener [Tue, 19 Jan 2021 12:40:39 +0000 (13:40 +0100)]
middle-end/98638 - avoid SSA reference to stmts after SSA deconstruction
Since SSA names do leak into global tree data structures like
TYPE_SIZE or in this case GFC_DECL_SAVED_DESCRIPTOR because of
frontend bugs we have to be careful to wipe references to the
CFG when we deconstruct SSA form because we now do ggc_free that.
2021-01-19 Richard Biener <rguenther@suse.de>
PR middle-end/98638
* tree-ssanames.c (fini_ssanames): Zero SSA_NAME_DEF_STMT.
Tobias Burnus [Tue, 19 Jan 2021 10:57:34 +0000 (11:57 +0100)]
OpenMP/Fortran: Fixes for {use,is}_device_ptr
gcc/fortran/ChangeLog:
PR fortran/98476
* openmp.c (resolve_omp_clauses): Change use_device_ptr
to use_device_addr for unless type(c_ptr); check all
list item for is_device_ptr.
Martin Jambor [Tue, 19 Jan 2021 10:28:48 +0000 (11:28 +0100)]
ipa-sra: Do not remove return values needed because of non-call EH
IPA-SRA already contains a check to figure out that an otherwise dead
parameter is actually required because of non-call exceptions, but it
is not present at the equivalent spot where SRA figures out whether
the return statement is used for anything useful. This patch adds
that condition there.
Unfortunately, even though this patch should be good enough for any
normal (I'd even say reasonable) use of the compiler, it hints that
when the user manually switches all sorts of DCE, IPA-SRA would
probably leave behind problematic statements manipulating what
originally were return values, just like it does for parameters (PR
93385). Fixing this properly might unfortunately be a separate issue
from the mentioned bug because the LHS of a call is changed during
call redirection and the caller often is not a clone. But I'll see
what I can do.
Meanwhile, the patch below has been bootstrapped and tested on x86_64.
gcc/ChangeLog:
2021-01-18 Martin Jambor <mjambor@suse.cz>
PR ipa/98690
* ipa-sra.c (ssa_name_only_returned_p): New parameter fun. Check
whether non-call exceptions allow removal of a statement.
(isra_analyze_call): Pass the appropriate function to
ssa_name_only_returned_p.
Geng Qi [Mon, 18 Jan 2021 06:09:27 +0000 (14:09 +0800)]
RISC-V: The 'multilib-generator' enhancement.
Think about this case:
./multilib-generator rv32imc-ilp32-rv32imac,rv32imacxthead-f
Here are 2 problems:
1. A unexpected 'xtheadf' extension was made.
2. The arch 'rv32imac' was not be created.
This modification fix these two, and also sorts 'multi-letter'.
gcc/ChangeLog:
* config/riscv/arch-canonicalize (longext_sort): New function for
sorting 'multi-letter'.
* config/riscv/multilib-generator: Adjusting the loop of 'alt' in
'alts'. The 'arch' may not be the first of 'alts'.
(_expand_combination): Add underline for the 'ext' without '*'.
This is because, a single-letter extension can always be treated well
with a '_' prefix, but it cannot be separated out if it is appended
to a multi-letter.
Ian Lance Taylor [Mon, 18 Jan 2021 22:38:10 +0000 (14:38 -0800)]
libbacktrace: use correct directory/filename for DWARF 5
PR debug/98716
* dwarf.c (read_v2_paths): Allocate zero entry for dirs and
filenames.
(read_line_program): Remove parameter u, change caller. Don't
subtract one from dirs and filenames index.
(read_function_entry): Don't subtract one from filenames index.
PPC64 can generate jumps with clobbered pseudo-regs and a BB with
such jump can have abnormal output edges. IRA hits an assert when trying
to split abnormal critical edge to deal with asm goto output reloads
later. The patch just skips splitting abnormal edges. It is assumed
that asm-goto with output reloads can not be in BB with output abnormal edges.
Patrick Palka [Mon, 18 Jan 2021 21:41:46 +0000 (16:41 -0500)]
c++: Add CTAD + pack expansion testcase
After r11-6614 made cp_walk_subtrees walk into the template of a CTAD
placeholder, we now correctly accept the below testcase. We used to
reject it because find_parameter_packs_r would fail to find the
parameter pack Ts inside the CTAD placeholder within the pack expansion.
Jakub Jelinek [Mon, 18 Jan 2021 18:13:44 +0000 (19:13 +0100)]
widening_mul: Fix up signed multiplication overflow check handling [PR98727]
I forgot one line, which means that if the second operand of the multiplication
isn't constant, it would be just the same as the first one.
2021-01-18 Jakub Jelinek <jakub@redhat.com>
PR tree-optimization/98727
* tree-ssa-math-opts.c (match_arith_overflow): Fix up computation of
second .MUL_OVERFLOW operand for signed multiplication with overflow
checking if the second operand of multiplication is not constant.
This patch extends the "malloc" checker within -fanalyzer to use
these attributes. It is based on an earlier patch:
'RFC: add "deallocated_by" attribute for use by analyzer'
https://gcc.gnu.org/pipermail/gcc-patches/2020-October/555544.html
which added a different attribute. The patch needed a lot of reworking
to support multiple deallocators per allocator.
My hope was that this would provide a minimal level of markup that would
support library-checking without requiring lots of further markup.
I attempted to use this to detect a memory leak within a Linux
driver (CVE-2019-19078), by adding the attribute to mark these fns:
extern struct urb *usb_alloc_urb(int iso_packets, gfp_t mem_flags);
extern void usb_free_urb(struct urb *urb);
where there is a leak of a "urb" on an error-handling path.
Unfortunately I ran into the problem that there are various other fns
that take "struct urb *" and the analyzer conservatively assumes that a
urb passed to them might or might not be freed and thus stops tracking
state for them.
Hence this will only detect issues for the simplest cases (without
adding another attribute).
gcc/analyzer/ChangeLog:
* analyzer.h (is_std_named_call_p): New decl.
* diagnostic-manager.cc (path_builder::get_sm): New.
(state_change_event_creator::state_change_event_creator): Add "pb"
param.
(state_change_event_creator::on_global_state_change): Don't consider
state changes affecting other state_machines.
(state_change_event_creator::on_state_change): Likewise.
(state_change_event_creator::m_pb): New field.
(diagnostic_manager::add_events_for_eedge): Pass pb to visitor
ctor.
* region-model-impl-calls.cc
(region_model::impl_deallocation_call): New.
* region-model.cc: Include "attribs.h".
(region_model::on_call_post): Handle fndecls referenced by
__attribute__((deallocated_by(FOO))).
* region-model.h (region_model::impl_deallocation_call): New decl.
* sm-malloc.cc: Include "stringpool.h" and "attribs.h". Add
leading comment.
(class api): Delete.
(enum resource_state): Update comment for change from api to
deallocator and deallocator_set.
(allocation_state::allocation_state): Drop api param. Add
"deallocators" and "deallocator".
(allocation_state::m_api): Drop field in favor of...
(allocation_state::m_deallocators): New field.
(allocation_state::m_deallocator): New field.
(enum wording): Add WORDING_DEALLOCATED.
(struct deallocator): New.
(struct standard_deallocator): New.
(struct custom_deallocator): New.
(struct deallocator_set): New.
(struct custom_deallocator_set): New.
(struct standard_deallocator_set): New.
(struct deallocator_set_map_traits): New.
(malloc_state_machine::m_malloc): Drop field
(malloc_state_machine::m_scalar_new): Likewise.
(malloc_state_machine::m_vector_new): Likewise.
(malloc_state_machine::m_free): New field
(malloc_state_machine::m_scalar_delete): Likewise.
(malloc_state_machine::m_vector_delete): Likewise.
(malloc_state_machine::deallocator_map_t): New typedef.
(malloc_state_machine::m_deallocator_map): New field.
(malloc_state_machine::deallocator_set_cache_t): New typedef.
(malloc_state_machine::m_custom_deallocator_set_cache): New field.
(malloc_state_machine::custom_deallocator_set_map_t): New typedef.
(malloc_state_machine::m_custom_deallocator_set_map): New field.
(malloc_state_machine::m_dynamic_sets): New field.
(malloc_state_machine::m_dynamic_deallocators): New field.
(api::api): Delete.
(deallocator::deallocator): New ctor.
(deallocator::hash): New.
(deallocator::dump_to_pp): New.
(deallocator::cmp): New.
(deallocator::cmp_ptr_ptr): New.
(standard_deallocator::standard_deallocator): New ctor.
(deallocator_set::deallocator_set): New ctor.
(deallocator_set::dump): New.
(custom_deallocator_set::custom_deallocator_set): New ctor.
(custom_deallocator_set::contains_p): New.
(custom_deallocator_set::maybe_get_single): New.
(custom_deallocator_set::dump_to_pp): New.
(standard_deallocator_set::standard_deallocator_set): New ctor.
(standard_deallocator_set::contains_p): New.
(standard_deallocator_set::maybe_get_single): New.
(standard_deallocator_set::dump_to_pp): New.
(start_p): New.
(class mismatching_deallocation): Update for conversion from api
to deallocator_set and deallocator.
(double_free::emit): Use %qs.
(class use_after_free): Update for conversion from api to
deallocator_set and deallocator.
(malloc_leak::describe_state_change): Only emit "allocated here" on
a start->nonnull transition, rather than on other transitions to
nonnull.
(allocation_state::dump_to_pp): Update for conversion from api to
deallocator_set.
(allocation_state::get_nonnull): Likewise.
(malloc_state_machine::malloc_state_machine): Likewise.
(malloc_state_machine::~malloc_state_machine): New.
(malloc_state_machine::add_state): Update for conversion from api
to deallocator_set.
(malloc_state_machine::get_or_create_custom_deallocator_set): New.
(malloc_state_machine::maybe_create_custom_deallocator_set): New.
(malloc_state_machine::get_or_create_deallocator): New.
(malloc_state_machine::on_stmt): Update for conversion from api
to deallocator_set. Handle "__attribute__((malloc(FOO)))", and
the special attribute set on FOO.
(malloc_state_machine::on_allocator_call): Update for conversion
from api to deallocator_set. Add "returns_nonnull" param and use
it to affect which state to transition to.
(malloc_state_machine::on_deallocator_call): Update for conversion
from api to deallocator_set.
gcc/ChangeLog:
* attribs.h (fndecl_dealloc_argno): New decl.
* builtins.c (call_dealloc_argno): Split out second half of
function into...
(fndecl_dealloc_argno): New.
* doc/extend.texi (Common Function Attributes): Document the
interaction between the analyzer and the malloc attribute.
* doc/invoke.texi (Static Analyzer Options): Likewise.
gcc/testsuite/ChangeLog:
* gcc.dg/analyzer/attr-malloc-1.c: New test.
* gcc.dg/analyzer/attr-malloc-2.c: New test.
* gcc.dg/analyzer/attr-malloc-4.c: New test.
* gcc.dg/analyzer/attr-malloc-5.c: New test.
* gcc.dg/analyzer/attr-malloc-6.c: New test.
* gcc.dg/analyzer/attr-malloc-CVE-2019-19078-usb-leak.c: New test.
* gcc.dg/analyzer/attr-malloc-misuses.c: New test.
Jonathan Wakely [Mon, 18 Jan 2021 14:23:13 +0000 (14:23 +0000)]
libstdc++: Only test writing to wostream if supported [PR 98725]
libstdc++-v3/ChangeLog:
PR libstdc++/98725
* testsuite/20_util/unique_ptr/io/lwg2948.cc: Do not try to
write to a wide character stream if wide character support is
disabled in the library.
Richard Biener [Mon, 18 Jan 2021 14:18:15 +0000 (15:18 +0100)]
testsuite/97494 - adjust gcc.dg/vect/slp-11b.c
Support for loop SLP splitting exposed that slp-11b.c has
folding that breaks SLP discovery which isn't what was intended
when the testcase was written. The following makes it SLP-able
and "only" run into the issue that a load permutation is required.
And tries to adjust the target selectors accordingly.
Christophe Lyon [Mon, 18 Jan 2021 13:55:04 +0000 (13:55 +0000)]
[arm,testsuite]: Fix options for vceqz_p64.c and vceqzq_p64.c
These two tests need:
dg-require-effective-target arm_crypto_ok
dg-add-options arm_crypto
because they use intrinsics that need -mfpu=crypto-neon-fp-armv8.
Jakub Jelinek [Mon, 18 Jan 2021 10:28:17 +0000 (11:28 +0100)]
libstd++: : Add workaround for as Error: file number less than one error [PR98708]
As mentioned in the PR, since the switch to DWARF5 by default instead of
DWARF4, gcc fails to build when configured against recent binutils.
The problem is that cxx11-ios_failure* is built in separate steps,
-S compilation (with -g -O2) followed by some sed and followed by
-c -g -O2 -g0 assembly. When gcc is configured against recent binutils
and DWARF5 is the default, we emit .file 0 "..." directive on which the
assembler then fails (unless --gdwarf-5 is passed to it, but we don't want
that generally because on the other side older assemblers don't like -g*
passed to it when invoked on *.s file with compiler generated debug info.
I hope the bug will be fixed soon on the binutils side, but it would be nice
to have a workaround.
The following patch is one of the possibilities, another one is to do that
but add configure check for whether it is needed,
essentially
echo 'int main () { return 0; }' > conftest.c
${CXX} ${CXXFLAGS} -g -O2 -S conftest.c -o conftest.s
${CXX} ${CXXFLAGS} -g -O2 -g0 -c conftest.s -o conftest.o
and if the last command fails, we need that -gno-as-loc-support.
Or yet another option would be I think do a different check, whether
${CXX} ${CXXFLAGS} -g -O2 -S conftest.c -o conftest.s
${CXX} ${CXXFLAGS} -g -O2 -c conftest.s -o conftest.o
works and if yes, don't add the -g0 to cxx11-ios_failure*.s assembly.
Jakub Jelinek [Mon, 18 Jan 2021 06:18:46 +0000 (07:18 +0100)]
libgomp: Don't access gomp_sem_t as int using atomics unconditionally
This patch introduces gomp_sem_getcount wrapper, which uses sem_getvalue
for POSIX and atomic loads for linux futex and accel. rtems for now
remains broken.
2021-01-18 Jakub Jelinek <jakub@redhat.com>
* config/linux/sem.h (gomp_sem_getcount): New function.
* config/posix/sem.h (gomp_sem_getcount): New function.
* config/posix/sem.c (gomp_sem_getcount): New function.
* config/accel/sem.h (gomp_sem_getcount): New function.
* task.c (task_fulfilled_p): Use gomp_sem_getcount.
(omp_fulfill_event): Likewise.
David Edelsohn [Mon, 18 Jan 2021 00:33:04 +0000 (19:33 -0500)]
testsuite: powerpc fold-vec and sse updates.
Recent code generation changes have affected the count of some instructions.
This patch updates the instruction count for fold-vec-extract on P7 and P8.
Also, some of SSE emulation intrinsics only work on LE systems.
gcc/testsuite/ChangeLog:
* gcc.target/powerpc/fold-vec-extract-char.p7.c: Adjust addi count.
* gcc.target/powerpc/fold-vec-extract-double.p7.c: Same.
* gcc.target/powerpc/fold-vec-extract-float.p7.c: Same.
* gcc.target/powerpc/fold-vec-extract-float.p8.c: Same.
* gcc.target/powerpc/fold-vec-extract-int.p7.c: Same.
* gcc.target/powerpc/fold-vec-extract-int.p8.c: Same.
* gcc.target/powerpc/fold-vec-extract-short.p7.c: Same.
* gcc.target/powerpc/fold-vec-extract-short.p8.c: Same.
* gcc.target/powerpc/sse-andnps-1.c: Restrict to LE.
* gcc.target/powerpc/sse-movhps-1.c: Restrict to LE.
* gcc.target/powerpc/sse-movlps-1.c: Restrict to LE.
* gcc.target/powerpc/sse2-andnpd-1.c: Restrict to LE.
David Edelsohn [Sun, 17 Jan 2021 23:18:56 +0000 (18:18 -0500)]
testsuite: Skip DWARF 5 testcases on AIX.
AIX does not support DWARF 5.
This patch skips the DWARF 5-specific testcases.
gcc/testsuite/ChangeLog:
* g++.dg/debug/dwarf2/inline-ns-2.C: Skip on AIX.
* g++.dg/debug/dwarf2/inline-var-2.C: Skip on AIX.
* g++.dg/debug/dwarf2/inline-var-3.C: Skip on AIX.
* g++.dg/debug/dwarf2/lang-cpp11.C: Skip on AIX.
* g++.dg/debug/dwarf2/lang-cpp14.C: Skip on AIX.
* g++.dg/debug/dwarf2/lang-cpp17.C: Skip on AIX.
* g++.dg/debug/dwarf2/lang-cpp20.C: Skip on AIX.
* gcc.dg/debug/dwarf2/inline6.c: Skip on AIX.
* gcc.dg/debug/dwarf2/lang-c11.c: Skip on AIX.
* gcc.dg/debug/dwarf2/pr41445-7.c: Skip on AIX.
* gcc.dg/debug/dwarf2/pr41445-8.c: Skip on AIX.
Martin Jambor [Sun, 17 Jan 2021 21:31:09 +0000 (22:31 +0100)]
ipa: Adjust cgraph verifier to materialization on demand (PR 98222)
after switching to materialization of clones on demand, the verifier
can happen to see edges leading to a clone of a materialized clone.
This means its clone_of is NULL and former_clone_of needs to be
checked in order to verify that the callee is a clone of the original
decl, which it did not do and reported edges to pointing to a wrong
place.
Fixed with the following patch, which has been pre-approved by Honza.
Bootstrapped and tested on x86_64-linux, pushed to master.
Martin
gcc/ChangeLog:
2021-01-15 Martin Jambor <mjambor@suse.cz>
PR ipa/98222
* cgraph.c (clone_of_p): Check also former_clone_of as we climb
the clone tree.
gcc/c/
* c-parser.c (c_parser_omp_clause_detach): New.
(c_parser_omp_all_clauses): Handle PRAGMA_OMP_CLAUSE_DETACH clause.
(OMP_TASK_CLAUSE_MASK): Add mask for PRAGMA_OMP_CLAUSE_DETACH.
* c-typeck.c (c_finish_omp_clauses): Handle PRAGMA_OMP_CLAUSE_DETACH
clause. Prevent use of detach with mergeable and overriding the
data sharing mode of the event handle.
gcc/cp/
* parser.c (cp_parser_omp_clause_detach): New.
(cp_parser_omp_all_clauses): Handle PRAGMA_OMP_CLAUSE_DETACH.
(OMP_TASK_CLAUSE_MASK): Add mask for PRAGMA_OMP_CLAUSE_DETACH.
* pt.c (tsubst_omp_clauses): Handle OMP_CLAUSE_DETACH clause.
* semantics.c (finish_omp_clauses): Handle OMP_CLAUSE_DETACH clause.
Prevent use of detach with mergeable and overriding the data sharing
mode of the event handle.
gcc/fortran/
* dump-parse-tree.c (show_omp_clauses): Handle detach clause.
* frontend-passes.c (gfc_code_walker): Walk detach expression.
* gfortran.h (struct gfc_omp_clauses): Add detach field.
(gfc_c_intptr_kind): New.
* openmp.c (gfc_free_omp_clauses): Free detach clause.
(gfc_match_omp_detach): New.
(enum omp_mask1): Add OMP_CLAUSE_DETACH.
(enum omp_mask2): Remove OMP_CLAUSE_DETACH.
(gfc_match_omp_clauses): Handle OMP_CLAUSE_DETACH for OpenMP.
(OMP_TASK_CLAUSES): Add OMP_CLAUSE_DETACH.
(resolve_omp_clauses): Prevent use of detach with mergeable and
overriding the data sharing mode of the event handle.
* trans-openmp.c (gfc_trans_omp_clauses): Handle detach clause.
* trans-types.c (gfc_c_intptr_kind): New.
(gfc_init_kinds): Initialize gfc_c_intptr_kind.
* types.def
(BT_FN_VOID_OMPFN_PTR_OMPCPYFN_LONG_LONG_BOOL_UINT_PTR_INT): Rename
to...
(BT_FN_VOID_OMPFN_PTR_OMPCPYFN_LONG_LONG_BOOL_UINT_PTR_INT_PTR):
...this. Add extra argument.
Directly using the corresponding -mcustom-insn=N options for this
floating-point unit leads to a combinatorial explosion in the potential
count of multilibs which may break the build.
gcc/
* config/nios2/nios2.c (NIOS2_FPU_CONFIG_NUM): Adjust value.
(nios2_init_fpu_configs): Provide register values for new
-mcustom-fpu-cfg=fph2 option variant.
* doc/invoke.texi (-mcustom-fpu-cfg=fph2): Document new option
variant.
Sebastian Huber [Sat, 16 Jan 2021 08:00:23 +0000 (09:00 +0100)]
nios2: Remove custom instruction warnings
Do not warn if custom instructions are not used due to missing
optimization flags. This prevents build errors with -Werror which
cannot be disabled via a dedicated warning option.
One reason to remove these warnings is to enable a multilib for the
"Nios II Floating Point Hardware 2 Component". For example, the
libatomic target library in GCC is built with -Werror and the warnings
removed by this patch resulted in errors like:
cc1: error: switch '-mcustom-fmins' has no effect unless '-ffinite-math-only' is specified [-Werror]
cc1: error: switch '-mcustom-fmaxs' has no effect unless '-ffinite-math-only' is specified [-Werror]
cc1: error: switch '-mcustom-round' has no effect unless '-fno-math-errno' is specified [-Werror]
Jakub Jelinek [Sat, 16 Jan 2021 08:21:52 +0000 (09:21 +0100)]
match.pd: Optimize ((cst << x) & 1) [PR96669]
While we had a ((1 << x) & 1) != 0 to x == 0 optimization already,
this patch adds ((cst << x) & 1) optimization too, this time the
second constant must be 1 though, not some power of two, but the first
one can be any constant. If it is even, the result is false, if it is
odd, the result is x == 0.
2021-01-16 Jakub Jelinek <jakub@redhat.com>
PR tree-optimization/96669
* match.pd ((CST << x) & 1 -> x == 0): New simplification.
* gcc.dg/tree-ssa/pr96669-1.c: Adjust regexp.
* gcc.dg/tree-ssa/pr96669-2.c: New test.
Jakub Jelinek [Sat, 16 Jan 2021 08:17:38 +0000 (09:17 +0100)]
cd_dce: Return TODO_update_address_taken from last cd_dce [PR96271]
On the following testcase, handle_builtin_memcmp in the strlen pass folds
the memcmp into comparison of two MEM_REFs. But nothing triggers updating
of addressable vars afterwards, so even when the parameters are no longer
address taken, we force the parameters to stack and back anyway.
This patch causes TODO_update_address_taken to happen right before last forwprop
pass (at the end of last cd_dce), so after strlen1 too.
2021-01-16 Jakub Jelinek <jakub@redhat.com>
PR tree-optimization/96271
* passes.def: Pass false argument to first two pass_cd_dce
instances and true to last instance. Add comment that
last instance rewrites no longer addressed locals.
* tree-ssa-dce.c (pass_cd_dce): Add update_address_taken_p member and
initialize it.
(pass_cd_dce::set_pass_param): New method.
(pass_cd_dce::execute): Return TODO_update_address_taken from
last cd_dce instance.
Jerry DeLisle [Fri, 15 Jan 2021 21:48:42 +0000 (13:48 -0800)]
fortran: Fixes a bug in ISO_Fortran_binding.c.
libgfortran/ChangeLog:
* runtime/ISO_Fortran_binding.c (CFI_establish): Fixed signed
char arrays. Signed char or uint8_t arrays would cause
crashes unless an element size is specified.
gcc/testsuite/ChangeLog:
* gfortran.dg/iso_fortran_binding_uint8_array.f90: New test.
* gfortran.dg/iso_fortran_binding_uint8_array_driver.c: New test.
Nathan Sidwell [Fri, 15 Jan 2021 19:38:43 +0000 (11:38 -0800)]
c++: Fix qualified array-type construction [PR 98538]
This was an assert that was too picky. The reason I had to alter
array construction was that on stream in, we cannot dynamically determine
a type's dependentness. Thus on stream out of the 'problematic' types,
we save the dependentness for reconstruction. Fortunately the paths into
cp_build_qualified_type_real from streamin with arrays do have the array's
dependentess set as needed.
PR c++/98538
gcc/cp/
* tree.c (cp_build_qualified_type_real): Propagate an array's
dependentness to the copy, if known.
gcc/testsuite/
* g++.dg/template/pr98538.C: New.
Jakub Jelinek [Fri, 15 Jan 2021 20:12:14 +0000 (21:12 +0100)]
match.pd: Generalize the PR64309 simplifications [PR96669]
The following patch generalizes the PR64309 simplifications, so that instead
of working only with constants 1 and 1 it works with any two power of two
constants, and works also for right shift (in that case it rules out the
first one being negative, as it is arithmetic shift then).
2021-01-15 Jakub Jelinek <jakub@redhat.com>
PR tree-optimization/96669
* match.pd (((1 << A) & 1) != 0 -> A == 0,
((1 << A) & 1) == 0 -> A != 0): Generalize for 1s replaced by
possibly different power of two constants and to right shift too.
Jakub Jelinek [Fri, 15 Jan 2021 20:10:44 +0000 (21:10 +0100)]
match.pd: Optimize (x < 0) ^ (y < 0) to (x ^ y) < 0 etc. [PR96681]
This patch simplifies comparisons that test the sign bit xored together.
If the comparisons are both < 0 or both >= 0, then we should xor the operands
together and compare the result to < 0, if the comparisons are different,
we should compare to >= 0.
Alexandre Oliva [Fri, 15 Jan 2021 19:22:54 +0000 (16:22 -0300)]
drop -dumpbase-ext from producer string
The -dumpbase and -dumpdir options are excluded from the producer
string output in debug information, but -dumpbase-ext was not. This
patch excludes it as well.
Jason Merrill [Wed, 13 Jan 2021 18:27:06 +0000 (13:27 -0500)]
c++: Avoid redundant copy in {} init [PR98642]
Here, initializing from { } implies a call to the default constructor for
base. We were then seeing that we're initializing a base subobject, so we
tried to copy the result of that call. This is clearly wrong; we should
initialize the base directly from its default constructor.
This patch does a lot of refactoring of unsafe_copy_elision_p and adds
make_safe_copy_elision that will also try to do the base constructor
rewriting from the last patch.
gcc/cp/ChangeLog:
PR c++/98642
* call.c (unsafe_return_slot_p): Return int.
(init_by_return_slot_p): Split out from...
(unsafe_copy_elision_p): ...here.
(unsafe_copy_elision_p_opt): New name for old meaning.
(build_over_call): Adjust.
(make_safe_copy_elision): New.
* typeck2.c (split_nonconstant_init_1): Elide copy from safe
list-initialization.
* cp-tree.h: Adjust.
Jason Merrill [Wed, 13 Jan 2021 18:27:53 +0000 (13:27 -0500)]
c++: Fix copy elision for base initialization
While working on PR98642 I noticed that in this testcase we were eliding the
copy, calling the complete default constructor to initialize the B base
subobject, and therefore wrongly initializing the non-existent A subobject
of B. The test doesn't care whether the copy is elided or not, but checks
that we are actually calling a base constructor for B.
The patch preserves the elision, but changes the initializer to call the
base constructor instead of the complete constructor.
gcc/cp/ChangeLog:
* call.c (base_ctor_for, make_base_init_ok): New.
(build_over_call): Use make_base_init_ok.
Jason Merrill [Fri, 15 Jan 2021 16:42:00 +0000 (11:42 -0500)]
c++: Fix list-init of array of no-copy type [PR63707]
build_vec_init_elt models initialization from some arbitrary object of the
type, i.e. copy, but in the case of list-initialization we don't do a copy
from the elements, we initialize them directly.
gcc/cp/ChangeLog:
PR c++/63707
* tree.c (build_vec_init_expr): Don't call build_vec_init_elt
if we got a CONSTRUCTOR.
gcc/testsuite/ChangeLog:
PR c++/63707
* g++.dg/cpp0x/initlist-array13.C: New test.
Jakub Jelinek [Fri, 15 Jan 2021 18:26:01 +0000 (19:26 +0100)]
testsuite: Add testcase coverage for already fixed [PR96671]
The fix for this PR didn't come with any test coverage, I've added
tests that make sure we optimize it no matter what order of the x ^ y ^ z
operands is used.
2021-01-15 Jakub Jelinek <jakub@redhat.com>
PR tree-optimization/96671
* gcc.dg/tree-ssa/pr96671-1.c: New test.
* gcc.dg/tree-ssa/pr96671-2.c: New test.
David Malcolm [Fri, 15 Jan 2021 18:26:39 +0000 (13:26 -0500)]
bootstrap: fix failing diagnostic selftest on Windows [PR98696]
In one of the selftests in g:f10960558540636800cf5d3d6355969621fbc17e
I didn't consider that paths can contain backslashes, which happens
for the tempfiles on Windows hosts.
gcc/ChangeLog:
PR bootstrap/98696
* diagnostic.c
(selftest::test_print_parseable_fixits_bytes_vs_display_columns):
Escape the tempfile name when constructing the expected output.
Jakub Jelinek [Fri, 15 Jan 2021 18:20:29 +0000 (19:20 +0100)]
c-family: Improve MEM_REF printing for diagnostics [PR98597]
Ok, here is an updated patch which fixes what I found, and implements what
has been discussed on the mailing list and on IRC, i.e. if the types
are compatible as well as alias sets are same, then it prints
what c_fold_indirect_ref_for_warn managed to create, otherwise it uses
that info for printing offsets using offsetof (except when it starts
with ARRAY_REFs, because one can't have offsetof (struct T[2][2], [1][0].x.y)
The uninit-38.c test (which was the only one I believe which had tests on the
exact spelling of MEM_REF printing) contains mainly changes to have space
before * for pointer types (as that is how the C pretty-printers normally
print types, int * rather than int*), plus what might be considered a
regression from what Martin printed, but it is actually a correctness fix.
When the arg is a pointer with type pointer to VLA with char element type
(let's say the pointer is p), which is what happens in several of the
uninit-38.c tests, omitting the (char *) cast is incorrect, as p + 1
is not the 1 byte after p, but pointer to the end of the VLA.
It only happened to work because of the hacks (which I don't like at all
and are dangerous, DECL_ARTIFICIAL var names with dot inside can be pretty
much anything, e.g. a lot of passes construct their helper vars from some
prefix that designates intended use of the var plus numeric suffix), where
the a.1 pointer to VLA is printed as a which if one is lucky happens to be
a variable with VLA type (rather than pointer to it), and for such vars
a + 1 is indeed &a[0] + 1 rather than &a + 1. But if we want to do this
reliably, we'd need to make sure it comes from VLA (e.g. verify that the
SSA_NAME is defined to __builtin_alloca_with_align and that there exists
a corresponding VAR_DECL with DECL_VALUE_EXPR that has the a.1 variable
in it).
2021-01-15 Jakub Jelinek <jakub@redhat.com>
PR tree-optimization/98597
* c-pretty-print.c: Include options.h.
(c_fold_indirect_ref_for_warn): New function.
(print_mem_ref): Use it. If it returns something that has compatible
type and is TBAA compatible with zero offset, print it and return,
otherwise print it using offsetof syntax or array ref syntax. Fix up
printing if MEM_REFs first operand is ADDR_EXPR, or when the first
argument has pointer to array type. Print pointers using the standard
formatting.
* gcc.dg/uninit-38.c: Expect a space in between type name and asterisk.
Expect for now a (char *) cast for VLAs.
* gcc.dg/uninit-40.c: New test.
Jakub Jelinek [Fri, 15 Jan 2021 18:17:53 +0000 (19:17 +0100)]
openmp: Change the way of building of reduction array type
The PR98597 patch regresses on _Atomic-3.c, as in the C FE building an
array type with qualified elements results in a type incompatible with
when an array type with unqualified elements is qualified afterwards.
This patch adds a workaround for that.
2021-01-15 Jakub Jelinek <jakub@redhat.com>
* c-typeck.c (c_finish_omp_clauses): For reduction build array with
unqualified element type and then call c_build_qualified_type on the
ARRAY_TYPE.
Kyrylo Tkachov [Fri, 15 Jan 2021 17:55:57 +0000 (17:55 +0000)]
[PATCH] aarch64: Implement vmlsl[_high]* intrinsics using builtins
This patch reimplements some more intrinsics using RTL builtins in the
straightforward way.
Thankfully most of the RTL infrastructure is already in place for it.
Nathan Sidwell [Fri, 15 Jan 2021 16:48:54 +0000 (08:48 -0800)]
c++: Fix langspecs with -fsyntax-only [PR98591]
-fsyntax-only is handled specially in the driver and causes it to add
'-o /dev/null' (or a suitable OS-specific variant thereof). PCH is
handled in the language driver. I'd not sufficiently protected the
-fmodule-only action of adding a dummy assembler from the actions of
-fsyntax-only, so we ended up with two -o options.
PR c++/98591
gcc/cp/
* lang-specs.h: Fix handling of -fmodule-only with -fsyntax-only.
aarch64: Add a minipass for fusing CC insns [PR88836]
This patch adds a small target-specific pass to remove redundant SVE
PTEST instructions. There are two important uses of this:
- Removing PTESTs after WHILELOs (PR88836). The original testcase
no longer exhibits the problem due to more recent optimisations,
but it can still be seen in simple cases like the one in the patch.
It also shows up in 450.soplex.
- Removing PTESTs after RDFFRs in ACLE code.
This is just an interim “solution” for GCC 11. I hope to replace
it with something generic and target-independent for GCC 12.
However, the use cases above are very important for performance,
so I'd rather not leave the bug unfixed for yet another release cycle.
Since the pass is intended to be short-lived, I've not added
a command-line option for it. The pass can be disabled using
-fdisable-rtl-cc_fusion if necessary.
Although what the pass does is independent of SVE, it's motivated
only by SVE cases and doesn't trigger for any non-SVE test I've seen.
I've therefore gated it on TARGET_SVE and restricted it to PTEST
patterns.
gcc/
PR target/88836
* config.gcc (aarch64*-*-*): Add aarch64-cc-fusion.o to extra_objs.
* Makefile.in (RTL_SSA_H): New variable.
* config/aarch64/t-aarch64 (aarch64-cc-fusion.o): New rule.
* config/aarch64/aarch64-protos.h (make_pass_cc_fusion): Declare.
* config/aarch64/aarch64-passes.def: Add pass_cc_fusion after
pass_combine.
* config/aarch64/aarch64-cc-fusion.cc: New file.
gcc/testsuite/
PR target/88836
* gcc.target/aarch64/sve/acle/general/ldff1_8.c: New test.
* gcc.target/aarch64/sve/ptest_1.c: Likewise.
Noticed while working on something else that the insn_change_watermark
destructor could call cancel_changes for changes that no longer exist.
The loop in cancel_changes is a nop in that case, but:
num_changes = num;
can mess things up.
I think this would only affect nested uses of insn_change_watermark.
gcc/
* recog.h (insn_change_watermark::~insn_change_watermark): Avoid
calling cancel_changes for changes that no longer exist.
IBM Z: Fix linking to libatomic in target test cases
One of the test cases failed to link because of missing paths to
libatomic. Reuse procedures in lib/atomic-dg.exp to gather these paths.
gcc/testsuite/ChangeLog:
2021-01-15 Marius Hillenbrand <mhillen@linux.ibm.com>
* gcc.target/s390/s390.exp: Call lib atomic-dg.exp to link
libatomic into testcases in gcc.target/s390/md.
* gcc.target/s390/md/atomic_exchange-1.c: Remove no unnecessary
-latomic.
Christophe Lyon [Thu, 15 Oct 2020 17:13:59 +0000 (17:13 +0000)]
arm: Implement vceqq_p64, vceqz_p64 and vceqzq_p64 intrinsics
This patch adds implementations for vceqq_p64, vceqz_p64 and
vceqzq_p64 intrinsics.
vceqq_p64 uses the existing vceq_p64 after splitting the input vectors
into their high and low halves.
vceqz[q] simply call the vceq and vceqq with a second argument equal
to zero.
The added (executable) testcases make sure that the poly64x2_t
variants have results with one element of all zeroes (false) and the
other element with all bits set to one (true).
gcc/testsuite/
PR target/71233
* gcc.target/aarch64/advsimd-intrinsics/p64_p128.c: Add tests for
vceqz_p64, vceqq_p64 and vceqzq_p64.
* gcc.target/arm/simd/vceqz_p64.c: New test.
* gcc.target/arm/simd/vceqzq_p64.c: New test.
Richard Biener [Fri, 15 Jan 2021 13:02:06 +0000 (14:02 +0100)]
testsuite/96147 - scan for vectorized load
This changes gcc.dg/vect/bb-slp-9.c to scan for a vectorized load
instead of a vectorized BB which then correctly captures the
unaligned load we try to test and not some intermediate built
from scalar vector.
2021-01-15 Richard Biener <rguenther@suse.de>
PR testsuite/96147
* gcc.dg/vect/bb-slp-9.c: Scan for a vector load transform.
Richard Biener [Fri, 15 Jan 2021 12:52:12 +0000 (13:52 +0100)]
testsuite/96147 - key scanning on vect_hw_misalign
gcc.dg/vect/slp-45.c failed to key the vectorization capability
scanning on vect_hw_misalign. Since the stores are strided
they cannot be (all) analyzed to be aligned.
2021-01-15 Richard Biener <rguenther@suse.de>
PR testsuite/96147
* gcc.dg/vect/slp-45.c: Key scanning on
vect_hw_misalign.
Richard Biener [Fri, 15 Jan 2021 12:49:04 +0000 (13:49 +0100)]
testsuite/96147 - remove scanning for ! vect_hw_misalign
This removes scanning that's too difficult to get correct for all
targets, leaving the correctness test for them and keeping the
vectorization capability check to vect_hw_misalign targets.
Christophe Lyon [Thu, 15 Oct 2020 17:13:59 +0000 (17:13 +0000)]
arm: Implement vceqq_p64, vceqz_p64 and vceqzq_p64 intrinsics
This patch adds implementations for vceqq_p64, vceqz_p64 and
vceqzq_p64 intrinsics.
vceqq_p64 uses the existing vceq_p64 after splitting the input vectors
into their high and low halves.
vceqz[q] simply call the vceq and vceqq with a second argument equal
to zero.
The added (executable) testcases make sure that the poly64x2_t
variants have results with one element of all zeroes (false) and the
other element with all bits set to one (true).
Richard Biener [Fri, 15 Jan 2021 12:31:28 +0000 (13:31 +0100)]
testsuite/96098 - remove redundant testcase
The testcase morphed in a way no longer testing what it was originally supposed to do and slightly altering it shows the original issue isn't fixed (anymore).
The limit as set as result of PR91403 (and dups) prevents the issue for larger
arrays but the testcase has
double a[128][128];
which results in a group size of "just" 512 (the limit is 4096). Avoiding
the 'BB vectorization with gaps at the end of a load is not supported'
by altering it to do
shows that costing has improved further to not account the dead loads making
the previous test inefficient. In fact the underlying issue isn't fixed
(we do code-generate dead loads).
In fact the vector permute load is even profitable, just the excessive
code-generation issue exists (and is "fixed" by capping it a constant
boundary, just too high for this particular testcase).
The testcase now has "dups", so I'll simply remove it.
The recent changes to error on mixing -march=i386 and -fcf-protection broke
bootstrap. This patch changes lib{atomic,gomp,itm} configury, so that it
only adds -march=i486 to flags if really needed (i.e. when 486 or later isn't
on by default already). Similarly, it will not use ifuncs if -mcx16
(or -march=i686 for 32-bit) is on by default.
2021-01-15 Jakub Jelinek <jakub@redhat.com>
PR target/70454
libatomic/
* configure.tgt: For i?86 and x86_64 determine if -march=i486 needs to
be added through preprocessor check on
__GCC_HAVE_SYNC_COMPARE_AND_SWAP_4. Determine if try_ifunc is needed
based on preprocessor check on __GCC_HAVE_SYNC_COMPARE_AND_SWAP_16
or __GCC_HAVE_SYNC_COMPARE_AND_SWAP_8.
libgomp/
* configure.tgt: For i?86 and x86_64 determine if -march=i486 needs to
be added through preprocessor check on
__GCC_HAVE_SYNC_COMPARE_AND_SWAP_4.
libitm/
* configure.tgt: For i?86 and x86_64 determine if -march=i486 needs to
be added through preprocessor check on
__GCC_HAVE_SYNC_COMPARE_AND_SWAP_4.
Christophe Lyon [Thu, 19 Nov 2020 16:25:48 +0000 (16:25 +0000)]
arm: Auto-vectorization for MVE: vshr
This patch enables MVE vshr instructions for auto-vectorization. New
MVE patterns are introduced that take a vector of constants as second
operand, all constants being equal.
The existing mve_vshrq_n_<supf><mode> is kept, as it takes a single
immediate as second operand, and is used by arm_mve.h.
The vashr<mode>3 and vlshr<mode>3 expanders are moved fron neon.md to
vec-common.md, updated to rely on the normal expansion scheme to
generate shifts by immediate.
Christophe Lyon [Mon, 16 Nov 2020 14:58:19 +0000 (14:58 +0000)]
arm: Auto-vectorization for MVE: vshl
This patch enables MVE vshlq instructions for auto-vectorization.
The existing mve_vshlq_n_<supf><mode> is kept, as it takes a single
immediate as second operand, and is used by arm_mve.h.
We move the vashl<mode>3 insn from neon.md to an expander in
vec-common.md, and the mve_vshlq_<supf><mode> insn from mve.md to
vec-common.md, adding the second alternative fron neon.md.
mve_vshlq_<supf><mode> will be used by a later patch enabling
vectorization for vshr, as a unified version of
ashl3<mode3>_[signed|unsigned] from neon.md. Keeping the use of unspec
VSHLQ enables to generate both 's' and 'u' variants.
It is not clear whether the neon_shift_[reg|imm]<q> attribute is still
suitable, since this insn is also used for MVE.
I kept the mve_vshlq_<supf><mode> naming instead of renaming it to
ashl3_<supf>_<mode> as discussed because the reference in
arm_mve_builtins.def automatically inserts the "mve_" prefix and I
didn't want to make a special case for this.
I haven't yet found why the v16qi and v8hi tests are not vectorized.
With dest[i] = a[i] << b[i] and:
{
int i;
unsigned int i.24_1;
unsigned int _2;
int16_t * _3;
short int _4;
int _5;
int16_t * _6;
short int _7;
int _8;
int _9;
int16_t * _10;
short int _11;
unsigned int ivtmp_42;
unsigned int ivtmp_43;
}
the vectorizer says:
mve-vshl.c:37:96: note: ==> examining statement: _5 = (int) _4;
mve-vshl.c:37:96: note: vect_is_simple_use: operand *_3, type of def: internal
mve-vshl.c:37:96: note: vect_is_simple_use: vectype vector(8) short int
mve-vshl.c:37:96: missed: conversion not supported by target.
mve-vshl.c:37:96: note: vect_is_simple_use: operand *_3, type of def: internal
mve-vshl.c:37:96: note: vect_is_simple_use: vectype vector(8) short int
mve-vshl.c:37:96: note: vect_is_simple_use: operand *_3, type of def: internal
mve-vshl.c:37:96: note: vect_is_simple_use: vectype vector(8) short int
mve-vshl.c:37:117: missed: not vectorized: relevant stmt not supported: _5 = (int) _4;
mve-vshl.c:37:96: missed: bad operation or unsupported loop bound.
mve-vshl.c:37:96: note: ***** Analysis failed with vector mode V8HI
projects / thirdparty / gcc.git / log
