Building an ADDR_EXPR uses the canonical type to build the pointer
type, but then, as we dereference it, we lose track of lax alignment
known to apply to the dereferenced object. This might not be a
problem in general, but it is when the compiler implicitly introduces
address taking and dereferencing, as it does for asm statements, and
as it may do in some loop optimizations.
From: Richard Biener <rguenther@suse.de>
for gcc/ChangeLog
PR middle-end/87054
* gimplify.c (gimplify_expr): Retain alignment of
addressable lvalue in dereference.
From: Alexandre Oliva <oliva@adacore.com>
for gcc/testsuite/ChangeLog
PR middle-end/87054
* gcc.dg/pr87054.c: New.
From-SVN: r264450
Back when we had the logic to output is_stmt but never exercised it,
it didn't matter that we didn't test for assembler support for it.
But there are still assemblers out there that do not support it, so
now that we enable the formerly latent is_stmt logic, we'd better make
sure the assembler can deal with it.
for gcc/ChangeLog
PR bootstrap/87013
* configure.ac: Check for .loc is_stmt support.
* configure, config.in: Rebuilt.
* dwarf2out.c (dwarf2out_source_line): Skip is_stmt
if not supported.
From-SVN: r264449
These options have been deprecated for many years, supplanted by -misel
and -mno-isel. This patch finally removes them.
* config/rs6000/rs6000.opt (misel=no, misel=yes): Delete.
* doc/invoke.texi (RS/6000 and PowerPC Options): Delete -misel=yes and
-misel=no.
From-SVN: r264447
gcc/fortran/
* dump-parse-tree.c (show_omp_clauses): Add missing omp list_types
and reorder the switch cases to match the enum in gfortran.h.
From-SVN: r264446
My patch of 2016-08-26 to avoid calling a trivial default constructor
introduced TARGET_EXPRs initialized with void_node to express trivial
initialization. But when this shows up in a VEC_INIT_EXPR, we weren't
prepared to handle it. Fixed by handling it explicitly in
cxx_eval_vec_init_1.
* constexpr.c (cxx_eval_vec_init_1): Handle trivial initialization.
From-SVN: r264442
PEELING_FOR_GAPS now means "peel one iteration for the epilogue",
in much the same way that PEELING_FOR_ALIGNMENT > 0 means
"peel that number of iterations for the prologue". We weren't
taking this into account when deciding whether we needed to peel
further scalar iterations beyond the iterations for "gaps" and
"alignment".
Only the first test failed before the patch. The other two
are just for completeness.
2018-09-20 Richard Sandiford <richard.sandiford@arm.com>
gcc/
PR tree-optimization/87288
* tree-vect-loop.c (vect_analyze_loop_2): Take PEELING_FOR_GAPS
into account when determining PEELING_FOR_NITERS.
gcc/testsuite/
PR tree-optimization/87288
* gcc.dg/vect/pr87288-1.c: New test.
* gcc.dg/vect/pr87288-2.c: Likewise,
* gcc.dg/vect/pr87288-3.c: Likewise.
From-SVN: r264440
Epilogue loop vectorisation skips vect_enhance_data_refs_alignment
since it doesn't make sense to version or peel the epilogue loop
(that will already have happened for the main loop). But this means
that it also fails to check whether the accesses are suitably aligned
for the new vector subarch.
We don't seem to carry alignment information from the (potentially
peeled or versioned) main loop to the epilogue loop, which would be
good to fix at some point. I think we want this patch regardless,
since there's no guarantee that the alignment requirements are the
same for every subarch.
2018-09-20 Richard Sandiford <richard.sandiford@arm.com>
gcc/
PR tree-optimization/86877
* tree-vect-loop.c (vect_analyze_loop_2): Call
vect_verify_datarefs_alignment.
gcc/testsuite/
PR tree-optimization/86877
* gfortran.dg/vect/vect-8-epilogue.F90: New test.
From-SVN: r264439
/cp
2018-09-19 Paolo Carlini <paolo.carlini@oracle.com>
PR c++/87324
* parser.c (cp_parser_initializer_list): Assign error_mark_node
to the index upon error.
/testsuite
2018-09-19 Paolo Carlini <paolo.carlini@oracle.com>
PR c++/87324
* g++.dg/cpp0x/desig5.C: New.
From-SVN: r264428
2018-09-19 Paul Thomas <pault@gcc.gnu.org>
PR fortran/84109
* trans-expr.c (gfc_trans_assignment_1): The rse.pre for the
assignment of deferred character intrinsic elemental function
results to a realocatable lhs must not be added to the exterior
block if they are array valued but must go to the loop body.
2018-09-19 Paul Thomas <pault@gcc.gnu.org>
PR fortran/84109
* gfortran.dg/elemental_function_3.f90 : New test.
From-SVN: r264427
This code in try_combine uses the wrong mode. This fails (with RTL
checking) in trunk, but not in any released branches.
PR rtl-optimization/86902
* combine.c (try_combine): When changing the CC mode used, don't change
an unrelated mode in other_insn to that new CC mode.
From-SVN: r264426
PR c++/87357 - missing -Wconversion warning
* decl.c (grok_op_properties): Remove diagnostic parts mentioning
a conversion to a reference to void. Use
same_type_ignoring_top_level_qualifiers_p rather than comparing types
directly.
* g++.dg/warn/Wconversion5.C: New test.
From-SVN: r264425
The vec_select operator is documented to require a const_int for the lane
selector operand, but GCN has an instruction that can select the lane at
runtime, so it seems reasonable to remove this restriction.
This patch simply replaces assertions that the operand is constant with early
exits from the optimizers. I think it's reasonable that vec_select with a
non-constant operand cannot be optimized, yet.
Also included is the necessary documentation tweak.
2018-09-19 Andrew Stubbs <ams@codesourcery.com>
gcc/
* doc/rtl.texi: Adjust vec_select description.
* simplify-rtx.c (simplify_binary_operation_1): Allow VEC_SELECT to use
non-constant selectors.
From-SVN: r264423
Assertions should be used to check preconditions that users must meet,
not to check whether the implementation is correct.
* include/bits/regex_automaton.tcc (_StateSeq<_TraitsT>::_M_clone()):
Remove __glibcxx_assert statements and use map::find instead of
map::operator[].
From-SVN: r264422
Use the STLUR instruction introduced in Armv8.4-a.
This instruction has the store-release semantic like STLR but can take a
9-bit unscaled signed immediate offset.
Example test case:
```
void
foo ()
{
int32_t *atomic_vals = calloc (4, sizeof (int32_t));
atomic_store_explicit (atomic_vals + 1, 2, memory_order_release);
}
```
Before patch generates
```
foo:
stp x29, x30, [sp, -16]!
mov x1, 4
mov x0, x1
mov x29, sp
bl calloc
mov w1, 2
add x0, x0, 4
stlr w1, [x0]
ldp x29, x30, [sp], 16
ret
```
After patch generates
```
foo:
stp x29, x30, [sp, -16]!
mov x1, 4
mov x0, x1
mov x29, sp
bl calloc
mov w1, 2
stlur w1, [x0, 4]
ldp x29, x30, [sp], 16
ret
```
We introduce a new feature flag to indicate the presence of this instruction.
The feature flag is called AARCH64_ISA_RCPC8_4 and is included when targeting
armv8.4 architecture.
We also introduce an "arch" attribute to be checked called "rcpc8_4" after this
feature flag.
gcc/
2018-09-19 Matthew Malcomson <matthew.malcomson@arm.com>
* config/aarch64/aarch64-protos.h
(aarch64_offset_9bit_signed_unscaled_p): New declaration.
* config/aarch64/aarch64.md (arches): New "rcpc8_4" attribute value.
(arch_enabled): Add check for "rcpc8_4" attribute value of "arch".
* config/aarch64/aarch64.h (AARCH64_FL_RCPC8_4): New bitfield.
(AARCH64_FL_FOR_ARCH8_4): Include AARCH64_FL_RCPC8_4.
(AARCH64_FL_PROFILE): Move index so flags are ordered.
(AARCH64_ISA_RCPC8_4): New flag.
* config/aarch64/aarch64.c (offset_9bit_signed_unscaled_p): Renamed
to aarch64_offset_9bit_signed_unscaled_p.
* config/aarch64/atomics.md (atomic_store<mode>): Allow offset
and use stlur.
* config/aarch64/constraints.md (Ust): New constraint.
* config/aarch64/predicates.md.
(aarch64_9bit_offset_memory_operand): New predicate.
(aarch64_rcpc_memory_operand): New predicate.
gcc/testsuite/
2018-09-19 Matthew Malcomson <matthew.malcomson@arm.com>
* gcc.target/aarch64/atomic-store.c: New.
From-SVN: r264421
2018-09-18 François Dumont <fdumont@gcc.gnu.org>
PR libstdc++/87135
* src/c++11/hashtable_c++0x.cc:
(_Prime_rehash_policy::_M_next_bkt): Return a prime no smaller than
requested size, but not necessarily greater.
(_Prime_rehash_policy::_M_need_rehash): Rehash only if target size is
strictly greater than next resize threshold.
* testsuite/23_containers/unordered_map/modifiers/reserve.cc: Adapt test
to validate that there is no rehash as long as number of insertion is
lower or equal to the reserved number of elements.
From-SVN: r264413
2018-09-18 Thomas Koenig <tkoenig@gcc.gnu.org>
PR fortran/29550
* gfortran.h (gfc_expr): Add external_blas flag.
* frontend-passes.c (matrix_case): Add case A2TB2T.
(optimize_namespace): Handle flag_external_blas by
calling call_external_blas.
(get_array_inq_function): Add argument okind. If
it is nonzero, use it as the kind of argument
to be used.
(inline_limit_check): Remove m_case argument, add
limit argument instead. Remove assert about m_case.
Set the limit for inlining from the limit argument.
(matmul_lhs_realloc): Handle case A2TB2T.
(inline_matmul_assign): Handle inline limit for other cases with
two rank-two matrices. Remove no-op calls to inline_limit_check.
(call_external_blas): New function.
* trans-intrinsic.c (gfc_conv_intrinsic_funcall): Do not add
argument to external BLAS if external_blas is already set.
2018-09-18 Thomas Koenig <tkoenig@gcc.gnu.org>
PR fortran/29550
* gfortran.dg/inline_matmul_13.f90: Adjust count for
_gfortran_matmul.
* gfortran.dg/inline_matmul_16.f90: Likewise.
* gfortran.dg/promotion_2.f90: Add -fblas-matmul-limit=1. Scan
for dgemm instead of dgemm_. Add call to random_number to make
standard conforming.
* gfortran.dg/matmul_blas_1.f90: New test.
* gfortran.dg/matmul_bounds_14.f: New test.
* gfortran.dg/matmul_bounds_15.f: New test.
* gfortran.dg/matmul_bounds_16.f: New test.
* gfortran.dg/blas_gemm_routines.f: New test / additional file for
preceding tests.
From-SVN: r264412
2018-09-18 Thomas Koenig <tkoenig@gcc.gnu.org>
PR fortran/29550
* gfortran.h (gfc_expr): Add external_blas flag.
* frontend-passes.c (matrix_case): Add case A2TB2T.
(optimize_namespace): Handle flag_external_blas by
calling call_external_blas.
(get_array_inq_function): Add argument okind. If
it is nonzero, use it as the kind of argument
to be used.
(inline_limit_check): Remove m_case argument, add
limit argument instead. Remove assert about m_case.
Set the limit for inlining from the limit argument.
(matmul_lhs_realloc): Handle case A2TB2T.
(inline_matmul_assign): Handle inline limit for other cases with
two rank-two matrices. Remove no-op calls to inline_limit_check.
(call_external_blas): New function.
* trans-intrinsic.c (gfc_conv_intrinsic_funcall): Do not add
argument to external BLAS if external_blas is already set.
2018-09-18 Thomas Koenig <tkoenig@gcc.gnu.org>
PR fortran/29550
* gfortran.dg/inline_matmul_13.f90: Adjust count for
_gfortran_matmul.
* gfortran.dg/inline_matmul_16.f90: Likewise.
* gfortran.dg/promotion_2.f90: Add -fblas-matmul-limit=1. Scan
for dgemm instead of dgemm_. Add call to random_number to make
standard conforming.
* gfortran.dg/matmul_blas_1.f90: New test.
* gfortran.dg/matmul_bounds_14.f: New test.
* gfortran.dg/matmul_bounds_15.f: New test.
* gfortran.dg/matmul_bounds_16.f: New test.
* gfortran.dg/blas_gemm_routines.f: New test / additional file for
preceding tests.
From-SVN: r264411
2018-09-18 Paul Thomas <pault@gcc.gnu.org>
PR fortran/87239
* trans-expr.c (gfc_trans_assignment_1): The rse.pre for the
assignment of deferred character elemental function results to
a realocatable lhs must not be added to the exterior block but
must go to the loop body.
2018-09-18 Paul Thomas <pault@gcc.gnu.org>
PR fortran/87239
* gfortran.dg/elemental_function_2.f90 : New test.
From-SVN: r264409
P1064R0 - Allowing Virtual Function Calls in Constant Expressions
* call.c (build_over_call): No longer check if we're outside a template
function.
* class.c (build_vtbl_initializer): Build vtable's constructor with
indexes.
* constexpr.c (cxx_eval_constant_expression): Don't ignore _vptr's
initializer. Handle OBJ_TYPE_REF.
(potential_constant_expression_1): Handle OBJ_TYPE_REF.
* decl.c (maybe_commonize_var): Bail out for any DECL_ARTIFICIAL.
(initialize_artificial_var): Mark the variable as constexpr.
(grokdeclarator): Change error to pedwarn. Only warn when
pedantic and not C++2a.
* gimple-fold.c (gimple_get_virt_method_for_vtable): Adjust assert.
* g++.dg/cpp0x/constexpr-virtual5.C: Adjust dg-error.
* g++.dg/cpp2a/constexpr-virtual1.C: New test.
* g++.dg/cpp2a/constexpr-virtual2.C: New test.
* g++.dg/cpp2a/constexpr-virtual3.C: New test.
* g++.dg/cpp2a/constexpr-virtual4.C: New test.
* g++.dg/cpp2a/constexpr-virtual5.C: New test.
* g++.dg/cpp2a/constexpr-virtual6.C: New test.
* g++.dg/cpp2a/constexpr-virtual7.C: New test.
* g++.dg/cpp2a/constexpr-virtual8.C: New test.
* g++.dg/cpp2a/constexpr-virtual9.C: New test.
* g++.dg/diagnostic/virtual-constexpr.C: Skip for C++2a. Use
-pedantic-errors. Adjust dg-error.
From-SVN: r264408
2018-09-18 Paul Thomas <pault@gcc.gnu.org>
PR fortran/87336
* trans-array.c (gfc_get_array_span): Try to get the element
length of incomplete types. Return NULL_TREE otherwise.
(gfc_conv_expr_descriptor): Only set the 'span' field if the
above does not return NULL_TREE. Set 'span' field if possible
for all new descriptors.
2018-09-18 Paul Thomas <pault@gcc.gnu.org>
PR fortran/87336
* gfortran.dg/pointer_array_10.f90 : New test.
* gfortran.dg/assign_10.f90 : Increase 'parm' count to 20.
* gfortran.dg/transpose_optimization_2.f90 : Increase 'parm'
count to 72.
From-SVN: r264405
This comment is quite cryptic and very out-of-date by now. Committing.
* config/rs6000/rs6000.md: Remove old "Cygnus sibcall" comment.
From-SVN: r264403
Combine will put CLOBBER (with a non-void mode) anywhere in a pattern
to poison it. reg_overlap_mentioned_p did not handle this. This patch
fixes that.
PR rtl-optimization/86882
* rtlanal.c (reg_overlap_mentioned_p): Handle CLOBBER.
From-SVN: r264400
The deleter only needs to be invocable when the unique_ptr destructor
and reset member function are instantiated. In other contexts it might
not be possible to pass unique_ptr<T, D>::pointer to the deleter, if
that requires a derived-to-base conversion from T* and T is incomplete.
* include/bits/unique_ptr.h (__uniq_ptr_impl): Remove static assertion
checking invocable condition.
(unique_ptr::~unique_ptr, unique_ptr::reset): Restore static assertion
here, where types must be complete. Pass pointer to deleter as an
rvalue.
* testsuite/20_util/unique_ptr/requirements/incomplete.cc: New test.
From-SVN: r264399
* config/i386/i386.md (*<code>extend<mode>xf2): Macroize insn from
*<code>extendsfxf2 and *<code>extenddfxf2 using MODEF mode iterator.
From-SVN: r264398
The changes to invoke.texi in r242433 left some unwanted spaces that
texi2pod.pl interprets as verbatim formatting. There are also some
grammatical errors due to the removal of references to GCJ, where the
G++ driver is referred to in the plural.
PR other/87353
* doc/invoke.texi (Link Options): Fix formatting and grammar.
From-SVN: r264395
This new test has some difficulties on the fabsl function.
On arm this is because we don't support the _Float128 type which the test uses.
This is handled in the patch by requiring a float128 target selector.
On aarch64-none-elf, a Newlib target, it fails because fabsl is not available.
long double support is known to be incomplete in newlib, and the fabsl function is not available
for targets where long double is larger than a double.
Therefore this patch skips the test on such targets.
* gcc.dg/warn-abs-1.c: Require float128 target.
Skip if large_long_double newlib target.
From-SVN: r264392
These two tests started failing after commit r264335 that adjusted the cutoff point at which the diagnostic suggestions machinery decides a suggestion is meaningful.
For these tests it means we no longer suggest anything as an alternative to "armv8-a-typo" as an "arch=" pargma value. We do still list the valid options, we just don't prefer one particular value over the others.
When I first wrote this test it wasn't with a particular architecture suggestion in mind, but rather to test that the suggestion machinery is being sanely invoked.
So this patch changes the dg-message check to treat the "did you mean...?" hunk as optional (in case the heuristics in the suggestions machinery change again).
With this patch the two tests PASS again on aarch64.
* gcc.target/aarch64/spellcheck_1.c:
Make architecture suggestion optional.
* gcc.target/aarch64/spellcheck_4.c:
Likewise.
From-SVN: r264390
As described in https://gcc.gnu.org/ml/gcc-patches/2018-09/msg00963.html this test generates UXTW instructions with -mabi=ilp32
because the foo* functions take pointers and store results into them. In ILP32 the callee clears the top bits with a UXTW.
This trips the scan-assembler-not UXTW test that checks that the zero_extend form of the BFXIL pattern is used, which it is.
This patch avoids this problem by not passing pointers to the results, but instead using global variables for which the foo* functions
will synthesise the address using ADRP, avoiding the UXTW instructions.
With this patch the test PASSes fully with -mabi=ilp32 and still PASSes on LP64.
* gcc.target/aarch64/combine_bfxil.c: Avoid passing pointers to
functions.
From-SVN: r264389
* config/mmix/mmix.c (TARGET_INIT_LIBFUNCS): Override with...
(mmix_init_libfuncs): New function: make __builtin_ffs expand
to __ffsdi2.
From-SVN: r264381
