Jakub Jelinek [Wed, 28 Mar 2018 19:15:39 +0000 (21:15 +0200)]
re PR middle-end/82004 (SPEC CPU2017 628.pop2_s miscompare)
PR tree-optimization/82004
* gimple-match-head.c (optimize_pow_to_exp): New function.
* match.pd (pow(C,x) -> exp(log(C)*x)): Wrap with #if GIMPLE.
Don't fold to exp if optimize_pow_to_exp is false.
Jakub Jelinek [Wed, 28 Mar 2018 14:45:21 +0000 (16:45 +0200)]
Fix wrong use-after-scope sanitization for omp variable (PR sanitizer/85081).
2018-03-28 Jakub Jelinek <jakub@redhat.com>
Martin Liska <mliska@suse.cz>
PR sanitizer/85081
* gimplify.c (asan_poison_variable): Don't do the check for
gimplify_omp_ctxp here.
(gimplify_decl_expr): Do it here.
(gimplify_target_expr): Likewise.
2018-03-28 Jakub Jelinek <jakub@redhat.com>
Martin Liska <mliska@suse.cz>
PR sanitizer/85081
* g++.dg/asan/pr85081.C: New test.
Co-Authored-By: Martin Liska <mliska@suse.cz>
From-SVN: r258924
Jakub Jelinek [Wed, 28 Mar 2018 11:59:06 +0000 (13:59 +0200)]
gfortran.h (gfc_dt): Rename default_exp field to dec_ext.
* gfortran.h (gfc_dt): Rename default_exp field to dec_ext.
* ioparm.def (IOPARM_dt_default_exp): Rename to ...
(IOPARM_dt_dec_ext): ... this.
* trans-io.c (build_dt): Adjust for default_exp renaming to
dec_ext and IOPARM_dt_default_exp renaming to IOPARM_dt_dec_ext.
* io.c (match_io): Likewise.
* io/io.h (IOPARM_DT_DEFAULT_EXP): Rename to ...
(IOPARM_DT_DEC_EXT): ... this.
* io/list_read.c (parse_real): Adjust for IOPARM_DT_DEFAULT_EXP
renaming to IOPARM_DT_DEC_EXT.
(read_real): Likewise.
* io/read.c (read_f): Likewise.
Alexandre Oliva [Wed, 28 Mar 2018 05:05:30 +0000 (05:05 +0000)]
[PR c++/84789] adjust testcase for -fconcepts
When compiling with -fconcepts,
cp_parser_template_declaration_after_export calls
cp_parser_template_introduction and that preparses qualified-ids not
preceded by typename in such a way that, when we get to
cp_parser_parse_and_diagnose_invalid_type_name and then
cp_parser_diagnose_invalid_type_name, the nested name specifier no
longer carries the previous template-dependent context, so we don't
stand a chance to suggest the use of 'typename' any more. Thus,
tolerate in the testcase the poorer error messages we get.
for gcc/testsuite/ChangeLog
PR c++/84789
* g++.dg/template/pr84789.C: Adjust for testing with
-fconcepts too.
Alexandre Oliva [Wed, 28 Mar 2018 05:05:14 +0000 (05:05 +0000)]
[PR c++/84973] don't defer output of uninstantiated templates
When an anon struct gets a name through a typedef, we reset its
linkage and that of its members. Member functions may get vague
linkage, which schedules them for deferred output, but we don't want
to add them to the queue if they're uninstantiated templates,
e.g. because the enclosing function is a template. They will be added
as needed when the enclosing template is instantiated.
Alexandre Oliva [Wed, 28 Mar 2018 05:04:59 +0000 (05:04 +0000)]
[PR c++/84968] reject stmt-exprs in noexcept constexprs
We reject extended statement-expressions in template parameters, so we
might as well reject them in constant expressions used in noexcept
specifications.
Michael Meissner [Tue, 27 Mar 2018 23:14:22 +0000 (23:14 +0000)]
re PR target/84914 (PowerPC complex multiply/divide calls the wrong function when -mabi=ieeelongdouble)
[gcc]
2018-03-27 Michael Meissner <meissner@linux.vnet.ibm.com>
PR target/84914
* config/rs6000/rs6000.c (create_complex_muldiv): New helper
function to create the function decl for complex long double
multiply and divide for -mabi=ieeelongdouble.
(init_float128_ieee): Call it.
[gcc/testsuite]
2018-03-27 Michael Meissner <meissner@linux.vnet.ibm.com>
PR target/84914
* gcc.target/powerpc/mulkc-2.c: New tests to make sure complex
long double multiply/divide uses the correct function.
* gcc.target/powerpc/mulkc-3.c: Likewise.
* gcc.target/powerpc/divkc-2.c: Likewise.
* gcc.target/powerpc/divkc-3.c: Likewise.
Jakub Jelinek [Tue, 27 Mar 2018 19:59:30 +0000 (21:59 +0200)]
re PR c++/85061 (ICE with __builtin_offsetof applied to static member)
PR c++/85061
* c-common.c (fold_offsetof_1) <case COMPOUND_EXPR>: Assert that
get_base_address of the second operand is a VAR_P, rather than the
operand itself, and use gcc_checking_assert instead of gcc_assert.
H.J. Lu [Tue, 27 Mar 2018 17:18:51 +0000 (17:18 +0000)]
i386: Insert ENDBR to trampoline for -fcf-protection=branch -mibt
When -fcf-protection=branch -mibt are used, we need to insert ENDBR
to trampoline. TRAMPOLINE_SIZE is creased by 4 bytes to accommodate
4-byte ENDBR instruction.
gcc/
PR target/85044
* config/i386/i386.c (ix86_trampoline_init): Insert ENDBR for
-fcf-protection=branch -mibt.
* config/i386/i386.h (TRAMPOLINE_SIZE): Increased by 4 bytes.
gcc/testsuite/
PR target/85044
* gcc.target/i386/pr85044.c: New test.
Kyrylo Tkachov [Tue, 27 Mar 2018 16:52:10 +0000 (16:52 +0000)]
[AArch64] XFAIL gcc.target/aarch64/store_v2vec_lanes.c for ILP32
The test in question fails for ilp32. The initial analysis I did in the PR for it
is that for ILP32 we generate somewhat different address forms that we'd need to adjust aarch64_classify_address to catch.
Given the optimisation this test checks for was added for GCC 8 it is not a regression, and improving the codegen on ILP32
would be an enhancement rather than a fix. So Richi has asked for it to be marked as XFAIL on ILP32, which is what this
patch does.
Checked that the test still passes on LP64 and appears as XFAIL on -mabi=ilp32.
PR target/83009
* gcc.target/aarch64/store_v2vec_lanes.c: XFAIL for ilp32.
This has been in my patch stack for quite some time. The problem here
was that we weren't handling arm_word_relocations in
arm_valid_symbolic_address and is the surest fix for this
for GCC8 and GCC7.
Regression tested on arm-none-linux-gnueabihf . Applying to
trunk and backporting to GCC-7 in a day or so.
Richard Biener [Tue, 27 Mar 2018 13:26:04 +0000 (13:26 +0000)]
re PR testsuite/82847 (gcc.dg/vect/slp-perm-9.c fail)
2018-03-27 Richard Biener <rguenther@suse.de>
PR testsuite/82847
* lib/target-supports.exp (check_ssse3_available): New function.
(check_effective_target_vect_perm_short): Enable for x86 if
check_ssse3_available.
* config.gcc (aarch64*-*-linux*): New TARGET_DEFAULT_ASYNC_UNWIND_TABLES
* common/config/aarch64/aarch64-common.c (aarch64_optimization_table[]):
Turn on fasynchronous-unwind-tables and funwind-tables.
Paolo Carlini [Tue, 27 Mar 2018 06:52:04 +0000 (06:52 +0000)]
re PR c++/84632 (internal compiler error: tree check: expected record_type or union_type or qual_union_type, have array_type in reduced_constant_expression_p, at cp/constexpr.c:1778)
/cp
2018-03-27 Paolo Carlini <paolo.carlini@oracle.com>
Jason Merrill <jason@redhat.com>
PR c++/84632
* init.c (build_aggr_init): When initializing from array,
reject anything but CONSTRUCTORs and TARGET_EXPRs.
(build_vec_init): Handle separately ARRAY_TYPEs.
/testsuite
2018-03-27 Paolo Carlini <paolo.carlini@oracle.com>
Jason Merrill <jason@redhat.com>
Tom de Vries [Mon, 26 Mar 2018 09:45:49 +0000 (09:45 +0000)]
Fix switch conversion in offloading functions
2018-03-26 Tom de Vries <tom@codesourcery.com>
PR tree-optimization/85063
* omp-general.c (offloading_function_p): New function. Factor out
of ...
* omp-offload.c (pass_omp_target_link::gate): ... here.
* omp-general.h (offloading_function_p): Declare.
* tree-switch-conversion.c (build_one_array): Mark CSWTCH.x variable
with attribute omp declare target for offloading functions.
* testsuite/libgomp.c/switch-conversion-2.c: New test.
* testsuite/libgomp.c/switch-conversion.c: New test.
* testsuite/libgomp.oacc-c-c++-common/switch-conversion-2.c: New test.
* testsuite/libgomp.oacc-c-c++-common/switch-conversion.c: New test.
Jerry DeLisle [Sun, 25 Mar 2018 03:47:28 +0000 (03:47 +0000)]
re PR fortran/69497 (ICE in gfc_free_namespace, at fortran/symbol.c:3701)
2018-03-24 Jerry DeLisle <jvdelisle@gcc.gnu.org>
Dominique d'Humieres <dominiq@gcc.gnu.org>
PR fortran/89497
* symbol.c (gfc_free_namespace): Delete the assert and only if
refs count is equals zero, free the namespace. Otherwise,
something is halfway and other errors will resound.
Thomas Koenig [Sat, 24 Mar 2018 17:37:21 +0000 (17:37 +0000)]
re PR fortran/70068 (ICE: out of memory on involving empty substring)
2018-03-24 Thomas Koenig <tkoenig@gcc.gnu.org>
PR fortran/70068
* expr.c (find_substring_ref): Change types of start, end
and length variables to gfc_charlen_t. Set length to zero
for empty substring.
2018-03-24 Thomas Koenig <tkoenig@gcc.gnu.org>
PR fortran/70068
* gfortran.dg/substr_7.f90: New test.
Use SCEV information when aligning for vectorisation (PR 84005)
This PR is another regression caused by the removal of the simple_iv
check in dr_analyze_innermost for BB analysis. Without splitting out
the step, we weren't able to find an underlying object whose alignment
could be increased.
As with PR81635, I think the simple_iv was only handling one special
case of something that ought to be more general. The more general
thing here is that if the address can be analysed as a scalar
evolution, and if all updates preserve alignment N, it's possible
to align the address to N by increasing the alignment of the base
object to N. That applies also to outer loops, and to both loop
and BB analysis.
I wasn't sure where the new functions ought to live, but tree-data-ref.c
seemed OK since (a) that already does scev analysis on addresses and
(b) you'd want to use dr_analyze_innermost first if you were analysing
a reference.
2018-03-24 Richard Sandiford <richard.sandiford@linaro.org>
gcc/
PR tree-optimization/84005
* tree-data-ref.h (get_base_for_alignment): Declare.
* tree-data-ref.c (get_base_for_alignment_1): New function.
(get_base_for_alignment): Likewise.
* tree-vect-data-refs.c (vect_compute_data_ref_alignment): Use
get_base_for_alignment to find a suitable base object, instead
of always using drb->base_address.
gcc/testsuite/
PR tree-optimization/84005
* gcc.dg/vect/bb-slp-1.c: Make sure there is no message about
failing to force the alignment.
Kyrylo Tkachov [Fri, 23 Mar 2018 16:43:43 +0000 (16:43 +0000)]
[arm] PR target/85026: Fix ldrsh length estimate in Thumb state
This bug has been reported against GCC 7.3.0 but it is latent in all release branches and on trunk.
We underestimate the length of the LRSH instruction in Thumb state.
Unlike other load instructions LDRSH can be encoded in 16 bits only when using a register offset.
In the testcase we have "ldrsh r2, [r4]" being assigned a length of 2, which is wrong.
So we don't calculate branch ranges properly and cause the assembler error.
The fix is to make the unaligned_loadhis insn similar to the *arm_extendqihi_insn insn that outputs an LDRSB.
Just remove the wrong 2-byte alternative. I don't think this is worth inventing a new "register-offset-only" constraint.
This also makes the patch safer for backporting.
Bootstrapped and tested on arm-none-linux-gnueabihf.
PR target/85026
* config/arm/arm.md (unaligned_loadhis): Remove first alternative.
Clean up attributes.
Jakub Jelinek [Fri, 23 Mar 2018 08:55:53 +0000 (09:55 +0100)]
re PR sanitizer/85029 (-fsanitize=undefined internal compiler error: in maybe_optimize_ubsan_ptr_ifn, at sanopt.c:493)
PR sanitizer/85029
* sanopt.c (maybe_optimize_ubsan_ptr_ifn): If DECL_REGISTER (base),
just don't try to optimize it rather than assert it never happens.
Alexandre Oliva [Fri, 23 Mar 2018 01:19:14 +0000 (01:19 +0000)]
[PR c++/84789] do not fail to resolve typename into template-independent
Although resolve_typename_type always takes a template-dependent
type-id, and it usually resolves it to another template-dependent
type-id, it is not correct to require the latter: in declarators,
template-dependent scopes may turn out to name template-independent
types, as in the pr84789-2.C and pr84789-3.C testcases.
The ill-formed testcase pr84789.C trips the same too-strict assert,
and also gets fixed by removing the assertion on the simplified scope.
However, whereas when the dependent type cannot be resolved, we get an
error that suggests 'typename' is missing:
pr84789.C:12:3: error: need ‘typename’ before ‘typename B<T>::A::I::I’
because ‘typename B<T>::A::I’ is a dependent scope
B<T>::A::I::I i;
^~~~
when it can, we got errors that did not point at that possibility,
which may be confusing:
pr84789.C:9:15: error: ‘A::I’ {aka ‘int’} is not a class type
B<T>::A::I::I i; // { dg-error "typename" }
^
pr84789.C:9:15: error: ‘I’ in ‘A::I’ {aka ‘int’} does not name a type
Changing the parser diagnostic code that reports an invalid type name
so that it does not attempt to reparse the name as a declarator gets
us the superior diagnostic of a missing 'typename' keyword.
for gcc/cp/ChangeLog
PR c++/84789
* pt.c (resolve_typename_type): Drop assert that stopped
simplification to template-independent types. Add assert to
verify the initial scope is template dependent.
* parser.c (cp_parser_parse_and_diagnose_invalid_type_name):
Reparse the id expression as a type-name, not a declarator.
Alexandre Oliva [Fri, 23 Mar 2018 01:19:01 +0000 (01:19 +0000)]
[PR c++/84729] reject parenthesized array init
A parenthesized initializer was only accepted when new()ing an array in
permissive mode. We were not careful, however, to convert the
TREE_LIST initializer to the array element type in this extension.
Instead of fixing it, converting the initializer to the base type
after turning the TREE_LIST initializer to a compound_expr, we disable
this deprecated extension.
for gcc/cp/ChangeLog
PR c++/84729
* init.c (build_vec_init): Error at parenthesized array init.
Alexandre Oliva [Fri, 23 Mar 2018 01:18:48 +0000 (01:18 +0000)]
Disable auto_is_implicit_function_template_parm_p while parsing attributes
We don't want functions to become implicit templates just because of
random uses of auto in unexpected places. Disabling the special
handling of auto while parsing attributes, for example, makes for
more sensible errors.
for gcc/cp/ChangeLog
PR c++/84610
PR c++/84642
PR c++/84942
* cp-tree.h (temp_override): New template class, generalizing
a cleanup that was only used...
* parser.c (cp_parser_parameter_declaration_clause):
... here for auto_is_implicit_function_template_parm_p.
(cp_parser_gnu_attributes_opt): Use it here as well.
(cp_parser_std_attribute): Likewise.
Steven G. Kargl [Thu, 22 Mar 2018 21:42:07 +0000 (21:42 +0000)]
re PR fortran/84922 (fortran reports inconsistency in rank of arguments in interface and contained procedures)
2018-03-22 Steven G. Kargl <kargl@gcc.gnu.org>
PR fortran/84922
* decl.c (get_proc_name): If the MODULE prefix appears in interface
body, then it must appear on the contained subroutine or function.
While here, fix nearby mis-indented code.
2018-03-22 Steven G. Kargl <kargl@gcc.gnu.org
PR fortran/84922
* gfortran.dg/interface_42.f90: New test.
* gfortran.dg/interface_43.f90: New test.
Kelvin Nilsen [Thu, 22 Mar 2018 20:18:35 +0000 (20:18 +0000)]
rs6000-builtin.def: Remove various BU_ALTIVEC_X macro expansions for definition of...
gcc/ChangeLog:
2018-03-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
* config/rs6000/rs6000-builtin.def: Remove various BU_ALTIVEC_X
macro expansions for definition of ST_INTERNAL_<mode> and
LD_INTERNAL_<mode> builtins.
* config/rs6000/rs6000-protos.h (rs6000_address_for_altivec):
Remove prototype.
* config/rs6000/rs6000.c (altivec_expand_ld_builtin): Delete this
function.
(altivec_expand_st_builtin): Likewise.
(altivec_expand_builtin): Remove calls to deleted functions.
(rs6000_address_for_altivec): Delete this function.
* config/rs6000/vector.md: Remove expands for
vector_altivec_load_<mode> and vector_altivec_store_<mode>.
Sudakshina Das [Thu, 22 Mar 2018 17:24:41 +0000 (17:24 +0000)]
[ARM][PR target/84826] Fix ICE in extract_insn, at recog.c:2304 on
arm-linux-gnueabi
The ICE in the bug report was happening because the macro
USE_RETURN_INSN (FALSE) was returning different values at different points
in the compilation. This was internally occurring because the function
arm_compute_static_chain_stack_bytes () which was dependent on
arm_r3_live_at_start_p () was giving a different value after the cond_exec
instructions were created in ce3 causing the liveness of r3 to escape up
to the start block.
The function arm_compute_static_chain_stack_bytes () should really only
compute the value once duringepilogue/prologue stage. This pass introduces
a new member 'static_chain_stack_bytes' to the target definition of the
struct machine_function which gets calculated in expand_prologue and is the
value that is returned by arm_compute_static_chain_stack_bytes () beyond that.
Kelvin Nilsen [Thu, 22 Mar 2018 15:42:39 +0000 (15:42 +0000)]
re PR target/84760 (Finish implementation of __builtin_altivec_lvx_v1ti)
gcc/ChangeLog:
2018-03-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
PR target/84760
* doc/extend.texi: Add four new prototypes for vec_ld.
* config/rs6000/rs6000-builtin.def (LVX_V1TI): Reorder symbol
definitions for more logical presentation.
* config/rs6000/rs6000-c.c: (altivec_overloaded_builtins): Add
entries for V1TI variants of __builtin_altivec_ld builtin.
* config/rs6000/rs6000.c: (altivec_expand_lv_builtin): Add test and
handling of V1TI variant of LVX icode pattern.
(altivec_expand_builtin): Add case for ALTIVEC_BUILTIN_LVX_V1TI.
(rs6000_gimple_fold_builtin): Likewise.
(altivec_init_builtins): Add code to define
__builtin_altivec_lvx_v1ti function.
gcc/testsuite/ChangeLog:
2018-03-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
PR target/84760
* gcc.target/powerpc/altivec-ld-1.c: New test.
Jakub Jelinek [Thu, 22 Mar 2018 08:32:12 +0000 (09:32 +0100)]
re PR sanitizer/85018 (Many sanitizer tests ICE since r258681)
PR sanitizer/85018
* dwarf2asm.c (dw2_output_indirect_constant_1): Set
DECL_INITIAL (decl) to decl at the end.
* varasm.c (use_blocks_for_decl_p): Revert the 2018-03-20 change,
adjust the comment.
Jason Merrill [Thu, 22 Mar 2018 03:53:19 +0000 (23:53 -0400)]
PR c++/81311 - wrong C++17 overload resolution.
* call.c (build_user_type_conversion_1): Remove C++17 code.
(conv_binds_ref_to_prvalue): New.
(build_over_call): Handle C++17 copy elision.
(build_special_member_call): Only do C++17 copy elision here if the
argument is already the right type.
Joseph Myers [Wed, 21 Mar 2018 22:29:37 +0000 (22:29 +0000)]
Adjust __builtin_tgmath handling of integer arguments to _FloatN narrowing macros.
When adding __builtin_tgmath to support a better tgmath.h
implementation, I noted that further changes might be needed regarding
the TS 18661 functions that round their results to a narrower type,
because of unresolved issues with how the corresponding type-generic
macros are defined in TS 18661.
The resolution of those issues is still in flux, but the latest
version does indeed require something slightly different from
__builtin_tgmath. It specifies that integer arguments to type-generic
macros such as f32xadd are treated as _Float64 not double - which was
also present in earlier versions of the resolution - but then it also
specifies different handling for _Float64 arguments and double
arguments, which wasn't in earlier versions. Specifically, in the
latest version
<http://www.open-std.org/jtc1/sc22/wg14/www/docs/n2213.pdf>, f32xadd
with _Float64 arguments would call f32xaddf64, while f32xadd with
double arguments would call f32xaddf64x. Since integer arguments are
converted directly to the argument type of the selected function (not
to double / _Float64x unless that ends up as the argument type), this
is a user-visible difference in semantics that means __builtin_tgmath
actually needs to implement treating integer arguments as _Float64 in
this case (the rest of the latest semantics can then be implemented in
the header, with a few inline functions there).
To avoid releasing with the older version of the __builtin_tgmath
semantics that doesn't work with the latest proposed DR#13 resolution,
this patch implements a rule in __builtin_tgmath that maps integer
types to _Float64 (respectively _Complex _Float64 for complex integer
types) where all the specified functions return the same _FloatN or
_FloatNx type. This does not affect any existing uses of
__builtin_tgmath in glibc's or GCC's tgmath.h since I haven't yet
added any of these type-generic macros to glibc when adding the
corresponding narrowing functions.
Bootstrapped with no regressions on x86_64-pc-linux-gnu.
* doc/extend.texi (__builtin_tgmath): Document when complex
integer types are treated as _Complex _Float64.
gcc/c:
* c-parser.c (c_parser_postfix_expression): For __builtin_tgmath
where all functions return the same _FloatN or _FloatNx type,
treat integer types as _Float64 instead of double.
gcc/testsuite:
* gcc.dg/builtin-tgmath-3.c: New test.
projects / thirdparty / gcc.git / log
