Jason Merrill [Tue, 6 Aug 2019 14:07:59 +0000 (10:07 -0400)]
PR c++/91378 - ICE with noexcept and auto return type.
Here, since the call to g is not type-dependent, we call mark_used on it to
determine its return type. This also wants to instantiate the
noexcept-expression. But since nothing in maybe_instantiate_noexcept was
calling push_to_top_level, we substituted b.i with processing_template_decl
set, so we left it unresolved for later access checking. As a result, the
type of C::g<int> remained instantiation-dependent, leading to an ICE in
type_dependent_expression_p on the assert that the type of a function
template with no dependent template arguments must be non-dependent.
Jakub Jelinek [Tue, 6 Aug 2019 07:26:32 +0000 (09:26 +0200)]
tree.h (OMP_CLAUSE_LASTPRIVATE_TASKLOOP_IV): Rename to ...
* tree.h (OMP_CLAUSE_LASTPRIVATE_TASKLOOP_IV): Rename to ...
(OMP_CLAUSE_LASTPRIVATE_LOOP_IV): ... this. Adjust comment.
* gimplify.c (gimple_add_tmp_var): In SIMD contexts, turn addressable
new vars into GOVD_PRIVATE rather than GOVD_LOCAL.
(gimplify_omp_for): Don't do C++ random access iterator clause
adjustments on combined constructs from OMP_LOOP. For OMP_LOOP,
don't predetermine the artificial iterator in case of C++ random
access iterators as lastprivate, but private. For OMP_LOOP, force
bind expr around simd body and force for_pre_body before the
construct. Use OMP_CLAUSE_LASTPRIVATE_LOOP_IV instead of
OMP_CLAUSE_LASTPRIVATE_TASKLOOP_IV.
(gimplify_omp_loop): Add firstprivate clauses on OMP_PARALLEL for
diff var of C++ random access iterators. Handle
OMP_CLAUSE_FIRSTPRIVATE. For OMP_CLAUSE_LASTPRIVATE_LOOP_IV, if
not outermost also add OMP_CLAUSE_FIRSTPRIVATE, and in both cases
clear OMP_CLAUSE_LASTPRIVATE_LOOP_IV on the lastprivate clause
on the OMP_FOR and OMP_DISTRIBUTE constructs if any.
* omp-low.c (lower_rec_input_clauses): For
OMP_CLAUSE_LASTPRIVATE_LOOP_IV on simd copy construct the private
variables instead of default constructing them.
(lower_lastprivate_clauses): Use OMP_CLAUSE_LASTPRIVATE_LOOP_IV
instead of OMP_CLAUSE_LASTPRIVATE_TASKLOOP_IV and move the
is_taskloop_ctx check from the assert to the guarding condition.
gcc/cp/
* parser.c (cp_parser_omp_for_loop): For OMP_LOOP, ignore parallel
clauses and predetermine iterator as lastprivate.
* semantics.c (handle_omp_for_class_iterator): Use
OMP_CLAUSE_LASTPRIVATE_LOOP_IV instead of
OMP_CLAUSE_LASTPRIVATE_TASKLOOP_IV, set it for lastprivate also
on OMP_LOOP construct. If a clause is missing for class iterator
on OMP_LOOP, add firstprivate clause, and if there is private
clause, turn it into firstprivate too.
(finish_omp_for): Formatting fix. For OMP_LOOP, adjust
OMP_CLAUSE_LASTPRIVATE_LOOP_IV clause CP_CLAUSE_INFO, so that it
uses copy ctor instead of default ctor.
* cp-gimplify.c (cp_gimplify_expr): Handle OMP_LOOP like
OMP_DISTRIBUTE etc.
(cp_fold_r): Likewise.
(cp_genericize_r): Likewise.
(cxx_omp_finish_clause): Also finish lastprivate clause with
OMP_CLAUSE_LASTPRIVATE_LOOP_IV flag.
* pt.c (tsubst_omp_clauses): Handle OMP_CLAUSE_BIND.
(tsubst_omp_for_iterator): For OMP_LOOP, ignore parallel
clauses and predetermine iterator as lastprivate.
* constexpr.c (potential_constant_expression_1): Handle OMP_LOOP
like OMP_DISTRIBUTE etc.
libgomp/
* testsuite/libgomp.c++/loop-13.C: New test.
* testsuite/libgomp.c++/loop-14.C: New test.
* testsuite/libgomp.c++/loop-15.C: New test.
Tom Honermann [Mon, 5 Aug 2019 19:34:23 +0000 (19:34 +0000)]
PR c++/88095, CTAD for literal operator templates per P0732
This patch fixes PR c++/88095: class nontype template parameter UDL string
literals doesn't accepts deduction placeholder
It also addresses a latent issue; literal operator templates with template
parameter packs of literal class type were previously accepted. The patch
corrects this and adds a test (udlit-class-nttp-neg.C).
This fix is needed for one of the char8_t remediation approaches documented
in P1423, and may be helpful for existing code bases impacted by the char8_t
changes adopted via P0482 for C++20.
gcc/cp/ChangeLog:
2019-08-02 Tom Honermann <tom@honermann.net>
* parser.c (cp_parser_template_declaration_after_parameters): Enable
class template argument deduction for non-type template parameters
in literal operator templates.
gcc/testsuite/ChangeLog:
2019-08-02 Tom Honermann <tom@honermann.net>
PR c++/88095
* g++.dg/cpp2a/udlit-class-nttp-ctad.C: New test.
* g++.dg/cpp2a/udlit-class-nttp-ctad-neg.C: New test.
* g++.dg/cpp2a/udlit-class-nttp-ctad-neg2.C: New test.
* g++.dg/cpp2a/udlit-class-nttp.C: New test.
* g++.dg/cpp2a/udlit-class-nttp-neg.C: New test.
* g++.dg/cpp2a/udlit-class-nttp-neg2.C: New test.
* c-opts.c (c_common_post_options): Enable -Wcomma-subscript by
default for C++2a, unless -Wno-deprecated.
* c.opt (Wcomma-subscript): New warning.
* parser.c (cp_parser_postfix_open_square_expression): Warn about uses
of a comma operator within a subscripting expression.
(cp_parser_skip_to_closing_square_bracket_1): New function, made out
of...
(cp_parser_skip_to_closing_square_bracket): ...this.
* doc/invoke.texi: Document -Wcomma-subscript.
* g++.dg/cpp2a/comma1.C: New test.
* g++.dg/cpp2a/comma2.C: New test.
* g++.dg/cpp2a/comma3.C: New test.
* g++.dg/cpp2a/comma4.C: New test.
Adding SVE intrinsics on top of the existing AArch64 intrinsics blows
the 12-bit function_code in tree_function_decl. That bitfield has no
spare bits, but it comes at the end of the structure and is preceded
by a pointer, so on LP64 hosts there's currently a 32-bit hole at end.
This patch therefore makes function_code an independent field and
moves the bitfield to the 32-bit hole.
I wondered about instead making function_code 16 bits, so that the
patch leaves 28 spare bits instead of just 12. That seemed a bit
short-term though; I can't guarantee that we won't blow 16 bits once
the SVE2 functions are added...
If we run out of bits again, we can start chomping from the top
of the enum. E.g. 24 bits should surely be enough, but there's
no point paying the overhead of the masking until we need it.
2019-08-05 Richard Sandiford <richard.sandiford@arm.com>
gcc/
* tree-core.h (tree_function_decl): Make function_code an
independent field. Group the remaining bitfields into bytes
and move decl_type so that it contines to be at a byte boundary.
Leave 12 bits for future expansion.
This patch folds IFN_MASK_LOAD and IFN_MASK_STOREs to normal accesses
if the mask is all-true. This can happen for fully-masked loops that
didn't actually need to be (which could be handled by the vectoriser
instead), or for unrolled fully-masked loops whose first iteration is
guaranteed to operate on a full vector. It's also useful when the
accesses are generated directly by intrinsics (to follow for SVE).
2019-08-05 Richard Sandiford <richard.sandiford@arm.com>
gcc/
* gimple-fold.c (gimple_fold_mask_load_store_mem_ref)
(gimple_fold_mask_load, gimple_fold_mask_store): New functions.
(gimple_fold_call): Use them to fold IFN_MASK_LOAD and
IFN_MASK_STORE.
gcc/testsuite/
* gcc.target/aarch64/sve/mask_load_1.c: New test.
Martin Liska [Mon, 5 Aug 2019 12:53:01 +0000 (14:53 +0200)]
Handle new operators with no arguments in DCE.
2019-08-05 Martin Liska <mliska@suse.cz>
PR c++/91334
* tree-ssa-dce.c (propagate_necessity): Handle new operators
with not arguments.
(eliminate_unnecessary_stmts): Likewise.
2019-08-05 Martin Liska <mliska@suse.cz>
PR c++/91334
* g++.dg/torture/pr91334.C: New test.
PR middle-end/91169
* fold-const.c (get_array_ctor_element_at_index): Create
offset_ints according to the sign of the index type and treat
that as signed if it is obviously so.
* gcc.target/i386/avx-loadu2-m128-1.c: New test.
* gcc.target/i386/avx-loadu2-m128-2.c: New test.
* gcc.target/i386/avx-loadu2-m128d-1.c: New test.
* gcc.target/i386/avx-loadu2-m128d-2.c: New test.
* gcc.target/i386/avx-loadu2-m128i-1.c: New test.
* gcc.target/i386/avx-loadu2-m128i-2.c: New test.
* gcc.target/i386/avx-storeu2-m128-1.c: New test.
* gcc.target/i386/avx-storeu2-m128-2.c: New test.
* gcc.target/i386/avx-storeu2-m128d-1.c: New test.
* gcc.target/i386/avx-storeu2-m128d-2.c: New test.
* gcc.target/i386/avx-storeu2-m128i-1.c: New test.
* gcc.target/i386/avx-storeu2-m128i-2.c: New test.
Kito Cheng [Mon, 5 Aug 2019 03:32:38 +0000 (03:32 +0000)]
RISC-V: Promote type correctly for libcalls
- argument and return value for libcall won't promote at
default_promote_function_mode_always_promote, however we expect it
should sign-extend as normal function.
- Witout this patch, this test case will fail at -march=rv64i -mabi=lp64.
- The implementation of riscv_promote_function_mode is borrowed from MIPS.
gcc/ChangeLog
* config/riscv/riscv.c (riscv_promote_function_mode): New.
(TARGET_PROMOTE_FUNCTION_MODE): Use riscv_promote_function_mode.
Steven G. Kargl [Sun, 4 Aug 2019 15:52:55 +0000 (15:52 +0000)]
re PR fortran/88227 (ICE in gfc_convert_boz, at fortran/target-memory.c:788)
2019-08-04 Steven G. Kargl <kargl@gcc.gnu.org>
PR fortran/88227
* check.c (oct2bin): New function. Convert octal string to binary.
(hex2bin): New function. Convert hexidecimal string to binary.
(bin2real): New function. Convert binary string to REAL. Use
oct2bin and hex2bin.
(gfc_boz2real): Use fallback conversion bin2real.
Alexandre Oliva [Fri, 2 Aug 2019 18:46:51 +0000 (18:46 +0000)]
rework Ada EH Machine_Occurrence deallocation
Introduce exception handler ABI #1 to ensure single release, no access
after release of reraised Machine_Occurrences, and no failure to
re-reraise a Machine_Occurrence.
Unlike Ada exceptions, foreign exceptions do not get a new
Machine_Occurrence upon reraise, but each handler would delete the
exception upon completion, normal or exceptional, save for the case of
a 'raise;' statement within the handler, that avoided the delete by
clearing the exception pointer that the cleanup would use to release
it. The cleared exception pointer might then be used by a subsequent
reraise within the same handler. Get_Current_Excep.all would also
expose the Machine_Occurrence to reuse by Reraise_Occurrence, even for
native exceptions.
Under ABI #1, Begin_Handler_v1 claims responsibility for releasing an
exception by saving its cleanup and setting it to Claimed_Cleanup.
End_Handler_v1 restores the cleanup and runs it, as long as it isn't
still Claimed_Cleanup (which indicates an enclosing handler has
already claimed responsibility for releasing it), and as long as the
same exception is not being propagated up (the next handler of the
propagating exception will then claim responsibility for releasing
it), so reraise no longer needs to clear the exception pointer, and it
can just propagate the exception, just like Reraise_Occurrence.
ABI #1 is fully interoperable with ABI #0, i.e., exception handlers
that call the #0 primitives can be linked together with ones that call
the #1 primitives, and they will not misbehave. When a #1 handler
claims responsibility for releasing an exception, even #0 reraises
dynamically nested within it will refrain from releasing it. However,
when a #0 handler is a handler of a foreign exception that would have
been responsible for releasing it with #1, a Reraise_Occurrence of
that foreign or other Machine_Occurrence-carrying exception may still
cause the exception to be released multiple times, and to be used
after it is first released, even if other handlers of the foreign
exception use #1.
for gcc/ada/ChangeLog
* libgnat/a-exexpr.adb (Begin_Handler_v1, End_Handler_v1): New.
(Claimed_Cleanup): New.
(Begin_Handler, End_Handler): Document.
* gcc-interface/trans.c (gigi): Switch to exception handler
ABI #1.
(Exception_Handler_to_gnu_gcc): Save the original cleanup
returned by begin handler, pass it to end handler, and use
EH_ELSE_EXPR to pass a propagating exception to end handler.
(gnat_to_gnu): Leave the exception pointer alone for reraise.
(add_cleanup): Handle EH_ELSE_EXPR, require it by itself.
Martin Liska [Fri, 2 Aug 2019 09:23:56 +0000 (11:23 +0200)]
Properly detect working jobserver in gcc driver.
2019-08-02 Martin Liska <mliska@suse.cz>
PR lto/91313
* gcc.c (driver::maybe_run_linker): Call detect_jobserver
to detect working job server.
(driver::detect_jobserver): Test whether jobserver
is active from GCC driver. That will prevent situation where
GCC is invoked from a LD plugin and the linker already uses
file descriptors suggested by make. That leads to a wrong
detection.
* gcc.h (driver): Add detect_jobserver.
* lto-wrapper.c (jobserver_active_p): Simplify sscanf by
not scanning for --jobserver-auth prefix.
Jakub Jelinek [Fri, 2 Aug 2019 08:28:31 +0000 (10:28 +0200)]
re PR tree-optimization/91201 (SIMD not generated for horizontal sum of bytes in array)
PR tree-optimization/91201
* config/i386/i386-expand.c (ix86_expand_vector_extract): For elt == 0
V16QImode extraction without sse4.1 try to use V4SImode lowpart
extraction.
* gcc.target/i386/sse2-pr91201-3.c: New test.
* gcc.target/i386/sse2-pr91201-4.c: New test.
* gcc.target/i386/sse2-pr91201-5.c: New test.
* gcc.target/i386/sse2-pr91201-6.c: New test.
PR c++/90947
* c-c++-common/array-1.c: New test.
* g++.dg/abi/mangle73.C: New test.
* g++.dg/cpp2a/nontype-class23.C: New test.
* g++.dg/init/array53.C: New test.
re PR c++/90590 (enumeration value not handled in switch warning for std::ios_base::seek_dir)
PR c++/90590
* c-warn.c (c_do_switch_warnings): Suppress warning for enumerators
with reserved names that are in a system header.
* c-c++-common/pr90590-1.c: New test.
* c-c++-common/pr90590-1.h: New test.
* c-c++-common/pr90590-2.c: New test.
* c-c++-common/pr90590-2.h: New test.
Jason Merrill [Wed, 31 Jul 2019 18:50:00 +0000 (14:50 -0400)]
PR c++/90538 - multiple expansions of capture packs
Previously, with init-capture the type of the closure field was a
DECLTYPE_TYPE of the initializer. But since each time we tsubst a lambda we
get a different lambda, that meant that if the initializer is a lambda, we'd
end up with different closure types in the field and initializer after
substitution (PR 87322). We dealt with this by remembering the lambda
instantiation within each pack expansion element, using
local_specialization_stack to separate the elements. But that broke this
testcase, because it lost lambda capture proxies that also use
local_specializations.
So, this patch removes the local_specializations changes from that patch and
fixes 87322 differently, by giving init-capture fields 'auto' type and doing
deduction later. There's a bit of a kludge to get the right number of
fields by pretending that 'auto...' uses the parameter packs from the
initializer, but it does the trick.
* cp-tree.h (DECLTYPE_FOR_INIT_CAPTURE): Remove.
* lambda.c (add_capture): Copy parameter packs from init.
(lambda_capture_field_type): Always use auto for init-capture.
* pt.c (uses_parameter_packs): Return tree.
(tsubst) [DECLTYPE_TYPE]: Remove init-capture handling.
(gen_elem_of_pack_expansion_instantiation): Don't push
local_specialization_stack.
(prepend_one_capture): New.
(tsubst_lambda_expr): Use it. Don't touch local_specializations.
(do_auto_deduction): Avoid redundant error.
Jonathan Wakely [Wed, 31 Jul 2019 16:40:39 +0000 (17:40 +0100)]
Implement "P0631R4 Math Constants" for C++20
The values of the constants are taken from Glibc where the equivalent
constant exists, or by rounding the actual constant to the same number
of digits as the Glibc constants have.
P0631R4 Math Constants
* include/Makefile.am: Add new header.
* include/Makefile.in: Regenerate.
* include/precompiled/stdc++.h: Include new header.
* include/std/numbers: New header.
* include/std/version (__cpp_lib_math_constants): Define.
* testsuite/26_numerics/numbers/1.cc: New test.
* testsuite/26_numerics/numbers/2.cc: New test.
* testsuite/26_numerics/numbers/3.cc: New test.
* testsuite/26_numerics/numbers/nonfloat_neg.cc: New test.
Jonathan Wakely [Wed, 31 Jul 2019 14:38:26 +0000 (15:38 +0100)]
PR libstdc++/91308 fix constraints on unique_ptr assignment
PR libstdc++/91308
* include/bits/unique_ptr.h (unique_ptr::__safe_conversion_up): Remove
constraints on deleter that should only apply to the constructor.
(unique_ptr<T[], D>::__safe_conversion_up): Likewise.
(unique_ptr<T[], D>::unique_ptr(unique_ptr<U, D>&&)): Restore
constraints on deleter here.
* testsuite/20_util/unique_ptr/assign/91308.cc: New test.
Jakub Jelinek [Wed, 31 Jul 2019 14:32:24 +0000 (16:32 +0200)]
re PR c/91192 (non-deterministic ICE on invalid)
PR c/91192
* c-parser.c (c_parser_sizeof_expression): Call set_c_expr_source_range
even if finish is UNKNOWN_LOCATION, just use start as finish in that
case.
Matt Thomas [Wed, 31 Jul 2019 14:11:16 +0000 (14:11 +0000)]
config.gcc (hppa*-*-netbsd*): New target.
gcc/ChangeLog:
* config.gcc (hppa*-*-netbsd*): New target.
* config/pa/pa-netbsd.h: New file.
* config/pa/pa32-netbsd.h: New file.
libgcc/ChangeLog:
* config.host (hppa*-*-netbsd*): New case.
* config/pa/t-netbsd: New file.
Co-Authored-By: Matthew Green <mrg@eterna.com.au> Co-Authored-By: Maya Rashish <coypu@sdf.org> Co-Authored-By: Nick Hudson <nick@nthcliff.demon.co.uk>
From-SVN: r273933
Richard Biener [Wed, 31 Jul 2019 10:55:27 +0000 (10:55 +0000)]
vr-values.h (vr_values::swap_vr_value): New.
2019-07-31 Richard Biener <rguenther@suse.de>
* vr-values.h (vr_values::swap_vr_value): New.
(vr_values::free_value_range): likewise.
* vr-values.c (vr_values::swap_vr_value): Implement.
* gimple-ssa-evrp-analyze.h (evrp_range_analyzer::pop_value_range):
Do not return a range or take a var.
(evrp_range_analyzer::stack): Change back to recording a non-const
value_range *.
* gimple-ssa-evrp-analyze.c
(evrp_range_analyzer::record_ranges_from_stmt): Free unused
value-range.
(evrp_range_analyzer::pop_to_marker): Adjust.
(evrp_range_analyzer::push_value_range): Use new swap_vr_value.
(evrp_range_analyzer::pop_value_range): Likewise. Free the
no longer needed value-range.
Jakub Jelinek [Wed, 31 Jul 2019 09:22:48 +0000 (11:22 +0200)]
re PR tree-optimization/91201 (SIMD not generated for horizontal sum of bytes in array)
PR tree-optimization/91201
* config/i386/sse.md (reduc_plus_scal_v16qi): New expander.
(REDUC_PLUS_MODE): Add V32QImode for TARGET_AVX and V64QImode for
TARGET_AVX512F.
(reduc_plus_scal_<mode>): Improve formatting by introducing
a temporary.
* gcc.target/i386/sse2-pr91201.c: New test.
* gcc.target/i386/avx2-pr91201.c: New test.
* gcc.target/i386/avx512bw-pr91201.c: New test.
This patch enables the new Transactional Memory Extension announced recently
as part of Arm's new architecture technologies.
We introduce a new optional extension "tme" to enable this. The following
instructions are part of the extension:
* tstart <Xt>
* ttest <Xt>
* tcommit
* tcancel #<imm>
We have also added ACLE intrinsics for the instructions.
projects / thirdparty / gcc.git / log
