gcc.git
4 years agoFix latch PHI arg lookup in vectorizable_reduction for double-reduction
Richard Biener [Tue, 20 Oct 2020 11:15:45 +0000 (13:15 +0200)]
Fix latch PHI arg lookup in vectorizable_reduction for double-reduction

We were using the wrong loop to figure the latch arg of a
double-reduction PHI.  Which isn't a problem in case ->dest_idx
match up with the outer loop edges - but that's of course not guaranteed.

2020-10-20  Richard Biener  <rguenther@suse.de>

* tree-vect-loop.c (vectorizable_reduction): Use the correct
loops latch edge for the PHI arg lookup.

4 years agoopenmp: Implement support for OMP_TARGET_OFFLOAD environment variable
Kwok Cheung Yeung [Tue, 20 Oct 2020 11:15:59 +0000 (04:15 -0700)]
openmp: Implement support for OMP_TARGET_OFFLOAD environment variable

This implements support for the OMP_TARGET_OFFLOAD environment variable
introduced in the OpenMP 5.0 standard, which controls how offloading
is handled.  It may be set to MANDATORY (abort if offloading cannot be
performed), DISABLED (no offloading to devices) or DEFAULT (offload to
device if possible, fall back to host if not).

2020-10-20  Kwok Cheung Yeung  <kcy@codesourcery.com>

libgomp/
* env.c (gomp_target_offload_var): New.
(parse_target_offload): New.
(handle_omp_display_env): Print value of OMP_TARGET_OFFLOAD.
(initialize_env): Parse OMP_TARGET_OFFLOAD.
* libgomp.h (gomp_target_offload_t): New.
(gomp_target_offload_var): New.
* libgomp.texi (OMP_TARGET_OFFLOAD): New section.
* target.c (resolve_device): Generate error if device not found and
offloading is mandatory.
(gomp_target_fallback): Generate error if offloading is mandatory.
(GOMP_target): Add argument in call to gomp_target_fallback.
(GOMP_target_ext): Likewise.
(gomp_target_data_fallback): Generate error if offloading is mandatory.
(GOMP_target_data): Add argument in call to gomp_target_data_fallback.
(GOMP_target_data_ext): Likewise.
(gomp_target_task_fn): Add argument in call to gomp_target_fallback.
(gomp_target_init): Return early if offloading is disabled.

4 years agolibstdc++: Define noop coroutine details private and inline [PR 95917]
Jonathan Wakely [Tue, 20 Oct 2020 10:19:58 +0000 (11:19 +0100)]
libstdc++: Define noop coroutine details private and inline [PR 95917]

This moves the __noop_coro_frame type, the __noop_coro_fr global
variable, and the __dummy_resume_destroy function from namespace scope,
replacing them with private members of the specialization
coroutine_handle<noop_coroutine_promise>.

The function and variable are also declared inline, so that they
generate no code unless used.

libstdc++-v3/ChangeLog:

PR libstdc++/95917
* include/std/coroutine (__noop_coro_frame): Replace with
noop_coroutine_handle::__frame.
(__dummy_resume_destroy): Define inline in __frame.
(__noop_coro_fr): Replace with noop_coroutine_handle::_S_fr
and define as inline.
* testsuite/18_support/coroutines/95917.cc: New test.

4 years agolibstdc++: Remove inheritance from std::coroutine_handle<> [LWG 3460]
Jonathan Wakely [Tue, 20 Oct 2020 10:18:35 +0000 (11:18 +0100)]
libstdc++: Remove inheritance from std::coroutine_handle<> [LWG 3460]

This removes the coroutine_handle<> base class from the primary template
and the noop_coroutine_promise explicit specialization. To preserve the
API various members are added, as they are no longer inherited from the
base class.

I've also tweaked some indentation and formatting, and replaced
subclause numbers from the standard with stable names like
[coroutine.handle.con].

libstdc++-v3/ChangeLog:

* include/std/coroutine (coroutine_handle<_Promise>): Remove
base class. Add constructors, conversions, accessors etc. as
proposed for LWG 3460.
(coroutine_handle<noop_coroutine_promise>): Likewise.
* testsuite/18_support/coroutines/lwg3460.cc: New test.

4 years agoMSP430: Support a memory operand for op1 of andneghi3
Jozef Lawrynowicz [Tue, 20 Oct 2020 10:26:20 +0000 (11:26 +0100)]
MSP430: Support a memory operand for op1 of andneghi3

This fixes an ICE caused by an unrecognizeable insn generated when
compiling gcc.c-torture/execute/pr97386-1.c at -O0.

gcc/ChangeLog:

* config/msp430/msp430.md (andneghi3): Allow general operand for
op1 and update output assembler template.

4 years agocollect-utils.c, lto-wrapper + mkoffload: Improve -save-temps filename
Tobias Burnus [Tue, 20 Oct 2020 10:14:03 +0000 (12:14 +0200)]
collect-utils.c, lto-wrapper + mkoffload: Improve -save-temps filename

gcc/ChangeLog:

* collect-utils.c (collect_execute, fork_execute): Add at-file suffix
argument.
* collect-utils.h (collect_execute, fork_execute): Update prototype.
* collect2.c (maybe_run_lto_and_relink, do_link, main, do_dsymutil):
Update calls by passing NULL.
* config/i386/intelmic-mkoffload.c (compile_for_target,
generate_host_descr_file, prepare_target_image, main): Likewise.
* config/gcn/mkoffload.c (compile_native, main): Pass at-file suffix.
* config/nvptx/mkoffload.c (compile_native, main): Likewise.
* lto-wrapper.c (compile_offload_image): Likewise.

4 years agoFix compilation on older systems
Vitaly Buka [Tue, 20 Oct 2020 05:44:34 +0000 (22:44 -0700)]
Fix compilation on older systems

Cherry-pick upstream commit f97ca48b1cbbf5da065e94271cb3af4f1c907dd4.

Fixes https://bugs.llvm.org/show_bug.cgi?id=47896

4 years agoSpecial case shifting by zero in operator_rshift::op1_range.
Aldy Hernandez [Tue, 20 Oct 2020 06:53:21 +0000 (08:53 +0200)]
Special case shifting by zero in operator_rshift::op1_range.

gcc/ChangeLog:

* range-op.cc (operator_rshift::op1_range): Special case
shifting by zero.

4 years ago[Ada] Refine type for type constraining routines
Piotr Trojanek [Sun, 26 Jul 2020 13:24:29 +0000 (15:24 +0200)]
[Ada] Refine type for type constraining routines

gcc/ada/

* sem_ch3.adb (Constrain_Decimal, Constrain_Enumeration,
Constrain_Float, Constrain_Integer, Constrain_Ordinary_Fixed):
Refine parameter type from Node_Id to Entity_Id.

4 years ago[Ada] Prevent crashes when pretty-printing freeze nodes from gdb
Piotr Trojanek [Fri, 24 Jul 2020 10:52:00 +0000 (12:52 +0200)]
[Ada] Prevent crashes when pretty-printing freeze nodes from gdb

gcc/ada/

* sprint.adb (po): Set Dump_Freeze_Null to False; align colons.
(ps): Likewise.

4 years ago[Ada] AI12-0339: Empty function for Aggregate aspect of Ada containers
Ed Schonberg [Thu, 23 Jul 2020 15:51:39 +0000 (11:51 -0400)]
[Ada] AI12-0339: Empty function for Aggregate aspect of Ada containers

gcc/ada/

* sinfo.ads, sinfo.adb: The flag Box_Present can appear in
Iterated_Element_Association nodes.
* sem_aggr.adb (Resolve_Aggregate): Call
Resolve_Container_Aggregate when type of context has
corresponding aspect.
* sem_type.adb (Covers): In Ada_2020 an aggregate is compatible
with a type that carries the corresponding aspect.
* exp_ch3.adb (Make_Controlling_Function_Wrappers): Do not
create declarations and bodies for inherited primitive functions
of null extensions that dispatch on result, when current scope
includes an immediately visible non-overloadable homonym of the
function.
* libgnat/a-cborse.adb, libgnat/a-cborse.ads,
libgnat/a-cbhase.ads, libgnat/a-cbhase.adb,
libgnat/a-cborma.adb, libgnat/a-cborma.ads,
libgnat/a-cbhama.adb, libgnat/a-cbhama.ads,
libgnat/a-cbdlli.adb, libgnat/a-cbdlli.ads,
libgnat/a-convec.ads, libgnat/a-ciorse.ads,
libgnat/a-cihase.ads, libgnat/a-cihase.adb,
libgnat/a-ciorma.ads, libgnat/a-cihama.ads,
libgnat/a-cihama.adb, libgnat/a-cidlli.ads,
libgnat/a-cidlli.adb, libgnat/a-coinve.adb,
libgnat/a-cobove.adb, libgnat/a-cobove.ads,
libgnat/a-convec.adb, libgnat/a-coinve.ads,
libgnat/a-coorse.ads, libgnat/a-cohase.adb,
libgnat/a-cohase.ads, libgnat/a-coorma.ads,
libgnat/a-cohama.adb, libgnat/a-cohama.ads,
libgnat/a-cdlili.ads: Add primitive function Empty for use in
aspect Aggregate, and add corresponding body or expression
function.

4 years ago[Ada] Add support for Unreferenced aspect on formal parameters
Arnaud Charlet [Tue, 21 Jul 2020 09:45:31 +0000 (05:45 -0400)]
[Ada] Add support for Unreferenced aspect on formal parameters

gcc/ada/

* aspects.adb (Has_Aspect_Specifications_Flag): Add
N_Parameter_Specification.
* par-ch13.adb (Aspect_Specifications_Present): Also handle case
of an unknown aspect on the last formal parameter (terminated by
a Tok_Right_Paren). Minor reformatting.
* par-ch6.adb (P_Formal_Part): Scan aspects on formal
parameters.
* par.adb: Fix typos.
* sem_ch6.adb (Process_Formals): Add processing of aspects and
in particular Unreferenced aspect for now.
* sinfo.ads: Allow ASPECT_SPECIFICATIONS on a
PARAMETER_SPECIFICATION.
* doc/gnat_rm/implementation_defined_aspects.rst
(Aspect Unreferenced): Update documentation.
* gnat_rm.texi: Regenerate.

4 years ago[Ada] Refine result type of Get_Accessibility
Piotr Trojanek [Thu, 23 Jul 2020 21:30:53 +0000 (23:30 +0200)]
[Ada] Refine result type of Get_Accessibility

gcc/ada/

* sem_util.ads, sem_util.adb (Get_Accessibility): Refine result
type from Node_Id to Entity_Id.

4 years ago[Ada] Replace low-level Ekind membership tests with high-level Is_Formal
Piotr Trojanek [Thu, 23 Jul 2020 21:26:20 +0000 (23:26 +0200)]
[Ada] Replace low-level Ekind membership tests with high-level Is_Formal

gcc/ada/

* einfo.adb, exp_attr.adb, sem_ch13.adb, sem_util.adb: Use
Is_Formal where possible.

4 years ago[Ada] Implement AI12-0280 Making 'Old more flexible
Steve Baird [Thu, 21 May 2020 21:42:53 +0000 (14:42 -0700)]
[Ada] Implement AI12-0280 Making 'Old more flexible

gcc/ada/

* sem_util.ads: Declare a new package, Old_Attr_Util, which in
turn declares two more packages, Conditional_Evaluation and
Indirect_Temps. Conditional_Evaluation provides a predicate for
deciding whether a given 'Old attribute reference is eligible
for conditional evaluation and, in the case where it is
eligible, a function that constructs the Boolean-valued
condition that is to be evaluated at run time in deciding
whether to evaluate the attribute prefix.  Indirect_Temps
provides support for declaring a temporary which is only
initialized conditionally; more specifically, an access type and
a variable of that type are declared (unconditionally) and then
the variable is (conditionally) initialized with an allocator.
The existence of the access type and the pointer variable is
hidden from clients, except that a predicate,
Is_Access_Type_For_Indirect_Temp, is provided for identifying
such access types. This is needed because we want such an access
type to be treated like a "normal" access type (specifically
with respect to finalization of allocated objects). Other parts
of the compiler treat access types differently if
Comes_From_Source is False, or if the secondary stack storage
pool is used; this predicate is used to disable this special
treatment.
* sem_attr.adb (Uneval_Old_Msg): Improve message text to reflect
Ada202x changes.
(Analyze_Attribute): A previously-illegal 'Old attribute
reference is accepted in Ada2020 if it is eligible for
conditional evaluation.
* sem_res.adb (Valid_Conversion): Do not treat a rewritten 'Old
attribute like other rewrite substitutions. This makes a
difference, for example, in the case where we are generating the
expansion of a membership test of the form "Saooaaat'Old in
Named_Access_Type"; in this case Valid_Conversion needs to
return True (otherwise the expansion will be False - see the
call site in exp_ch4.adb).
* exp_attr.adb (Expand_N_Attribute_Reference): When expanding a
'Old attribute reference, test for the case where the reference
is eligible for conditional evaluation. In that case, use the
new "indirect temporary" mechanism provided by Sem_Util.
* exp_prag.adb
(Expand_Attributes_In_Consequence.Expand_Attributes): If
Sem_Util.Indirect_Temp_Needed indicates that there could be
correctness problems associated with the old expansion scheme
for dealing with 'Old attributes in contract cases consequences,
then we use the new "indirect temporary" mechanism provided by
Sem_Util instead. We do not want to do this unconditionally.
* sem_util.adb: Provide a body for the new Old_Attr_Util
package. Further work is needed in several areas for
correctness:
- The function Is_Repeatedly_Evaluated does not deal with
container aggregates yet.
- The function Is_Known_On_Entry does not deal with interactions
with the Global aspect.
Each area where more work is needed is indicated with a "???"
comment in the code; a more detailed description can be found
there. Some optimization opportunties are similarly indicated
with a "???" comment.
* exp_ch3.adb (Freeze_Type): In deciding whether to generate
expansion for the list controller of an access type, take the
predicate Is_Access_Type_For_Indirect_Temp into account. If the
predicate yields True, then generate the expansion.
* exp_util.adb (Build_Allocate_Deallocate_Proc): We don't
normally finalize allocated objects that are allocated on the
secondary stack. Add an exception to this rule if the predicate
Is_Access_Type_For_Indirect_Temp yields True.  As a result of
this exception, we have to deal with the fact that gigi expects
a different parameter profile if we are using the secondary
stack pool; the Pool and Alignment parameters must be omitted in
this case.

4 years ago[Ada] Spurious error in current instance used as formal package
Javier Miranda [Thu, 23 Jul 2020 09:55:16 +0000 (05:55 -0400)]
[Ada] Spurious error in current instance used as formal package

gcc/ada/

* sem_ch12.adb (Install_Parents_Of_Generic_Context,
Remove_Parents_Of_Generic_Context): New subprograms.
(Instantiate_Package_Body): Adding assertions to ensure that
installed parents are properly removed.

4 years ago[Ada] Make attribute Update an obsolescent feature
Claire Dross [Tue, 21 Jul 2020 09:14:02 +0000 (11:14 +0200)]
[Ada] Make attribute Update an obsolescent feature

gcc/ada/

* sem_attr.adb (Analyze_Attribute): Emit a warning on 'Update
when Warn_On_Obsolescent_Feature is set to True.

4 years ago[Ada] CUDA: build procedure calls instead of function calls
Richard Kenner [Thu, 23 Jul 2020 08:07:24 +0000 (10:07 +0200)]
[Ada] CUDA: build procedure calls instead of function calls

gcc/ada/

* gnat_cuda.adb (Build_Register_Function_Call): Make procedure
call instead of function, rename to
Build_Register_Procedure_Call.
(Build_CUDA_Init_Proc): Make procedure call instead of function.

4 years ago[Ada] Crash on cond expression as actual for anonymous access formal
Justin Squirek [Wed, 22 Jul 2020 12:13:24 +0000 (08:13 -0400)]
[Ada] Crash on cond expression as actual for anonymous access formal

gcc/ada/

* exp_ch6.adb (Expand_Branch): Properly anticipate expansion of
conditional expressions producing object declarations in
addition to assignment statements, and rename formal.

4 years ago[Ada] Fixes for pretty command-line GNATprove output with -gnatdF
Yannick Moy [Wed, 22 Jul 2020 07:14:54 +0000 (09:14 +0200)]
[Ada] Fixes for pretty command-line GNATprove output with -gnatdF

gcc/ada/

* errout.adb (Write_Source_Code_Line): Adopt display closer to
GCC format.
(Output_Messages): Deal specially with info messages.
* erroutc.adb (Prescan_Message): Fix bug leading to check
messages being considered as error messages in pretty output
mode.

4 years ago[Ada] Hang on cond expression as actual for anonymous access formal
Justin Squirek [Tue, 21 Jul 2020 19:06:37 +0000 (15:06 -0400)]
[Ada] Hang on cond expression as actual for anonymous access formal

gcc/ada/

* exp_ch6.adb (Expand_Call_Helper): Properly handle the case
where the condition of a conditional expression has been
optimized out when calculating the value of an extra
accessibility parameter.

4 years ago[Ada] gnatpp: Fix documentation of threshold switches
Bob Duff [Mon, 20 Jul 2020 17:55:25 +0000 (13:55 -0400)]
[Ada] gnatpp: Fix documentation of threshold switches

gcc/ada/

* doc/gnat_ugn/gnat_utility_programs.rst: Change "_" to "-".

4 years ago[Ada] New warning on not fully initialized box aggregate
Arnaud Charlet [Thu, 16 Jul 2020 16:38:10 +0000 (12:38 -0400)]
[Ada] New warning on not fully initialized box aggregate

gcc/ada/

* sem_aggr.adb (Resolve_Aggregate): Warn on not fully
initialized box aggregate.
* sem_aggr.ads: Fix typo.
* sem_res.adb (Resolve_Actuals): Fix typo in error message
format marking it incorrectly as a continuation message.
* sem_elab.adb (Check_Internal_Call_Continue): Similarly, add
missing primary message in case of a call to an actual generic
subprogram.
* sem_warn.adb (Check_References): Do not warn on read but never
assigned variables if the type is partially initialized.
* libgnat/a-except.ads, libgnat/a-ststun.ads,
libgnat/g-sechas.ads, libgnat/a-cbdlli.ads,
libgnat/a-cfdlli.ads, libgnat/a-cobove.ads,
libgnat/a-cohata.ads, libgnat/a-crbltr.ads,
libgnat/a-cbmutr.ads, libgnat/a-crdlli.ads,
libgnat/a-cbsyqu.ads: Address new warning.
* doc/gnat_ugn/building_executable_programs_with_gnat.rst:
Update doc on -gnatwv.
* gnat_ugn.texi: Regenerate.

gcc/testsuite/

* gnat.dg/opt11.adb: Add new expected warning.

4 years ago[Ada] Inlining nonstatic calls to static expression functions
Gary Dismukes [Fri, 17 Jul 2020 22:28:52 +0000 (18:28 -0400)]
[Ada] Inlining nonstatic calls to static expression functions

gcc/ada/

* sem_ch6.adb (Analyze_Expression_Function): Mark static
expression functions as inlined.

4 years ago[Ada] Crash on precondition, discriminant and protected objects
Arnaud Charlet [Thu, 16 Jul 2020 12:27:54 +0000 (08:27 -0400)]
[Ada] Crash on precondition, discriminant and protected objects

gcc/ada/

* contracts.adb (Is_Prologue_Renaming): This function was
missing support for E_Constant which can also be generated in
protected objects.

4 years ago[Ada] Issue with gnatbind -V switch and Ada 2012
Arnaud Charlet [Fri, 17 Jul 2020 09:41:03 +0000 (05:41 -0400)]
[Ada] Issue with gnatbind -V switch and Ada 2012

gcc/ada/

* bindgen.adb (Gen_Bind_Env_String): Generate Ada 2012 compatible
strings. Code cleanup.

4 years ago[Ada] Fix comments as volatility properties can apply to objects
Yannick Moy [Fri, 17 Jul 2020 07:47:44 +0000 (09:47 +0200)]
[Ada] Fix comments as volatility properties can apply to objects

gcc/ada/

* sem_util.adb, sem_util.ads: Comment fix.

4 years ago[Ada] Spurious discriminant check on "for of" loop
Arnaud Charlet [Mon, 13 Jul 2020 09:23:17 +0000 (05:23 -0400)]
[Ada] Spurious discriminant check on "for of" loop

gcc/ada/

* sem_ch8.adb (Check_Constrained_Object): Suppress discriminant
checks when the type has default discriminants and comes from
expansion of a "for of" loop.

4 years ago[Ada] Flexible AST node structure
Bob Duff [Tue, 14 Jul 2020 21:10:19 +0000 (17:10 -0400)]
[Ada] Flexible AST node structure

gcc/ada/

* atree.ads: Make Default_Node a constant.  Remove the
modification of Comes_From_Source, and use a separate flag for
that.  Change Sloc to 0; it always overwritten, and never left
as the No_Location value.
(Print_Statistics): Move to spec so we can call it from
gnat1drv.
(Num_Nodes): Rename to clarify that this is approximate.
Correct comment: nodes and entities are never deleted, the count
is never decremented, and this is not used by Xref.
(Initialize): Correct comment: Error_List is not created here.
Other minor naming and comment changes.
* atree.adb (Extend_Node, New_Copy, New_Entity, New_Node):
Streamline these. Simplify and improve efficiency.  Move code
from Allocate_Initialize_Node to these, where it can be executed
unconditionally.  Take advantage of automatic zeroing of the
Nodes table.
(Allocate_Initialize_Node): Remove this. It was an efficiency
bottleneck, and somewhat complicated, because it was called from
4 places, and had all sorts of conditionals to check where it
was called from. Better to move most of that code to the call
sites, where it can be executed (or not) unconditionally.
(Allocate_New_Node): New procedure to partly replace
Allocate_Initialize_Node (called from just 2 of those 4 places).
(Comes_From_Source_Default): New flag written/read by
Set_Comes_From_Source_Default/Get_Comes_From_Source_Default.
This allows us to make Default_Node into a constant with
all-zeros value.
(Set_Paren_Count_Of_Copy): New procedure to avoid duplicated
code.
(Report): New procedure to encapsulate the call to the reporting
procedure.
(Atree_Private_Part): We now need a body for this package, to
contain package body Nodes.
(Approx_Num_Nodes_And_Entities): Was Num_Nodes.  For efficiency,
compute the answer from Nodes.Last. That way we don't need to
increment a counter on every node creation. Other minor naming
and comment changes.
* gnat1drv.adb: Call Atree.Print_Statistics if -gnatd.A switch
was given.  Add comment documenting the new order dependency (we
must process the command line before calling Atree.Initialize).
* debug.adb: Document -gnatd.A.
* einfo.adb, sinfo.adb: Remove useless Style_Checks pragmas.
* nlists.ads (Allocate_List_Tables): Inline makes node creation
a little faster.
* nlists.adb (Initialize): Remove local constant E, which didn't
seem to add clarity.
* treepr.adb (Print_Init): Use renamed
Approx_Num_Nodes_And_Entities function.
* types.ads: Change the Low and High bounds as described above.
* types.h: Change Low and High bounds to match types.ads.
* sem_ch8.adb, namet.adb, namet.ads: Move the computation of
Last_Name_Id from sem_ch8 to namet, and correct it to not assume
Name_Ids are positive.
* ali.adb, ali-util.adb, bindo-writers.adb, exp_dist.adb,
fmap.adb, fname-uf.adb, osint.adb: Fix various hash functions to
avoid assuming the various ranges are positive. Note that "mod"
returns a nonnegative result when the second operand is
positive. "rem" can return negative values in that case (in
particular, if the first operand is negative, which it now is).
* switch-c.adb: Allow switch -gnaten to control the value of
Nodes_Size_In_Meg.
* doc/gnat_ugn/building_executable_programs_with_gnat.rst:
Remove traling whitespaces.
* opt.ads (Nodes_Size_In_Meg): New Variable.

4 years ago[Ada] Fix internal error on bit-aligned component of function call
Eric Botcazou [Thu, 16 Jul 2020 09:07:32 +0000 (11:07 +0200)]
[Ada] Fix internal error on bit-aligned component of function call

gcc/ada/

* exp_util.adb (Remove_Side_Effects): Always generate a renaming
that is handled by the front-end in the case of an indexed or a
selected component whose prefix has a nonstandard representation.

4 years ago[Ada] Document Ada.Task_Initialization
Pat Rogers [Mon, 13 Jul 2020 20:31:54 +0000 (15:31 -0500)]
[Ada] Document Ada.Task_Initialization

gcc/ada/

* doc/gnat_rm/the_gnat_library.rst: Add Ada.Task_Initialization.
* gnat_rm.texi: Regenerate.

4 years ago[Ada] Display source code pointing at locations in messages for GNATprove
Yannick Moy [Fri, 10 Jul 2020 14:52:03 +0000 (16:52 +0200)]
[Ada] Display source code pointing at locations in messages for GNATprove

gcc/ada/

* errout.adb: (Error_Msg-Internal): Pass the location for a line
insertion if any in the message.
(Output_Messages: Add display of source code lines if -gnatdF is
set.
(Write_Source_Code_Line): Code clean up.
* erroutc.adb (Prescan_Message): Apply prescan for continuation
lines when -gnatdF is set, and record presence of line
insertion.
* erroutc.ads (Has_Insertion_Line): New global for prescan.
(Error_Msg_Object): Add field to record line insertion if
present.
* errutil.adb (Error_Msg): Pass no location for Insertion_Sloc.

4 years ago[Ada] Remove extra validity check in case statement
Arnaud Charlet [Mon, 13 Jul 2020 12:14:20 +0000 (08:14 -0400)]
[Ada] Remove extra validity check in case statement

gcc/ada/

* exp_ch5.adb (Expand_N_Case_Statement): Do not generate
validity check when possible.

4 years ago[Ada] Ada_2020: Further code cleanup for container aggregates
Ed Schonberg [Mon, 13 Jul 2020 20:29:56 +0000 (16:29 -0400)]
[Ada] Ada_2020: Further code cleanup for container aggregates

gcc/ada/

* exp_aggr.adb (Expand_Iterated_Component): Reorganize code to
ensure that Loop_Id is properly initialized on all paths, and
remove code duplication.

4 years ago[Ada] Propagate predicate function to a full view of a private subtype
Piotr Trojanek [Mon, 13 Jul 2020 13:22:27 +0000 (15:22 +0200)]
[Ada] Propagate predicate function to a full view of a private subtype

gcc/ada/

* sem_ch3.adb (Analyze_Subtype_Declaration): Propagate predicate
function to full view of the created type as well, if it was
created.

4 years ago[Ada] Code clean up
Arnaud Charlet [Mon, 13 Jul 2020 16:28:31 +0000 (18:28 +0200)]
[Ada] Code clean up

gcc/ada/

* exp_attr.adb (Expand_N_Attribute_Reference): Merge handling of
Simple_Storage_Pool and Storage_Pool.

4 years ago[Ada] Support for new aspect Subprogram_Variant on recursive subprograms
Piotr Trojanek [Mon, 13 Jul 2020 10:42:18 +0000 (12:42 +0200)]
[Ada] Support for new aspect Subprogram_Variant on recursive subprograms

gcc/ada/

* aspects.ads: Introduce Subprogram_Variant aspect with the
following properties: GNAT-specific, with mandatory expression,
not a representation aspect, never delayed.
* contracts.adb (Expand_Subprogram_Contract): Mention new aspect
in the comment.
(Add_Contract_Item): Support addition of pragma
Subprogram_Variant to N_Contract node.
(Analyze_Entry_Or_Subprogram_Contract): Mention new aspect in
the comment; add pragma Subprogram_Variant to N_Contract node.
(Build_Postconditions_Procedure): Adapt call to
Insert_Before_First_Source_Declaration, which is now reused in
expansion of new aspect.
(Process_Contract_Cases_For): Also process Subprogram_Variant,
which is stored in N_Contract node together with Contract_Cases.
* contracts.ads (Analyze_Entry_Or_Subprogram_Contract): Mention
new aspect in the comment.
(Analyze_Entry_Or_Subprogram_Body_Contract): Likewise.
* einfo.adb (Get_Pragma): Support retrieval of new pragma.
* einfo.ads (Get_Pragma): Likewise.
* exp_ch6.adb (Check_Subprogram_Variant): New routine for
emitting call to check Subprogram_Variant expressions at run
time.
(Expand_Call_Helper): Check Subprogram_Variant expressions at
recursive calls.
* exp_prag.adb (Make_Op): Moved from expansion of pragma
Loop_Variant to Exp_Util, so it is now reused for expansion of
pragma Subprogram_Variant.
(Process_Variant): Adapt call to Make_Op after moving it to
Exp_Util.
(Expand_Pragma_Subprogram_Variant): New routine.
* exp_prag.ads (Expand_Pragma_Subprogram_Variant): Likewise.
* exp_util.adb (Make_Variant_Comparison): Moved from Exp_Prag
(see above).
* exp_util.ads (Make_Variant_Comparison): Likewise.
* inline.adb (Remove_Aspects_And_Pragmas): Handle aspect/pragma
Subprogram_Variant just like similar contracts.
* par-prag.adb (Prag): Likewise.
* sem.adb (Insert_Before_First_Source_Declaration): Moved from
Contracts (see above).
* sem.ads (Insert_Before_First_Source_Declaration): Likewise.
* sem_ch12.adb: Mention new aspect in the comment about
"Implementation of Generic Contracts", just like similar aspects
are mentioned there.
* sem_ch13.adb (Insert_Pragma): Mention new aspect in the
comment, because this routine is now used for Subprogram_Variant
just like for other similar aspects.
(Analyze_Aspect_Specifications): Mention new aspect in comments;
it is handled just like aspect Contract_Cases.
(Check_Aspect_At_Freeze_Point): Do not expect aspect
Subprogram_Variant just like we don't expect aspect
Contract_Cases.
* sem_prag.adb (Ensure_Aggregate_Form): Now also used for pragma
Subprogram_Variant, so update comment.
(Analyze_Pragma): Add initial checks for pragma
Subprogram_Variant.
(Analyze_Subprogram_Variant_In_Decl_Part): New routine with
secondary checks on the new pragma.
(Sig_Flags): Handle references within pragma Subprogram_Variant
expression just like references in similar pragma
Contract_Cases.
(Is_Valid_Assertion_Kind): Handle Subprogram_Variant just like
other similar contracts.
* sem_prag.ads (Analyze_Subprogram_Variant_In_Decl_Part): New
routine.
* sem_res.adb (Same_Or_Aliased_Subprograms): Moved to Sem_Util,
so it can be reused for detection of recursive calls where
Subprogram_Variant needs to be verified.
* sem_util.adb (Is_Subprogram_Contract_Annotation): Handle new
Subprogram_Variant annotation just like other similar
annotations.
(Same_Or_Aliased_Subprograms): Moved from Sem_Res (see above).
* sem_util.ads (Is_Subprogram_Contract_Annotation): Mention new
aspect in the comment.
(Same_Or_Aliased_Subprograms): Moved from Sem_Res (see above).
* sinfo.ads (N_Contract): Document handling of
Subprogram_Variant.
* snames.ads-tmpl: Add name for the internally generated
procedure with checks for Subprogram_Variant expression, name
for the new aspect and new pragma corresponding to aspect
Subprogram_Variant.

4 years ago[Ada] Minor reorder subprogram declarations alphabetically
Piotr Trojanek [Mon, 13 Jul 2020 09:47:16 +0000 (11:47 +0200)]
[Ada] Minor reorder subprogram declarations alphabetically

gcc/ada/

* exp_util.ads: Reorder declaration.

4 years agotree-optimization/97496 - avoid SLP externs in patterns
Richard Biener [Tue, 20 Oct 2020 06:14:54 +0000 (08:14 +0200)]
tree-optimization/97496 - avoid SLP externs in patterns

I forgot to guard the promotion to external for the case where the
def is in a pattern.

2020-10-20  Richard Biener  <rguenther@suse.de>

PR tree-optimization/97496
* tree-vect-slp.c (vect_get_and_check_slp_defs): Guard extern
promotion with not in pattern.

* gcc.dg/vect/bb-slp-pr97496.c: New testcase.

4 years agoAdd { target int128 } to gcc.dg/pr97488.c
Aldy Hernandez [Tue, 20 Oct 2020 06:39:02 +0000 (02:39 -0400)]
Add { target int128 } to gcc.dg/pr97488.c

__int128 does not exist on 32-bit targets.

gcc/testsuite/ChangeLog:

* gcc.dg/pr97488.c: Add target int128 predicate.

4 years agoIBM Z: Emit vector alignment hints for strlen
Stefan Schulze Frielinghaus [Sun, 18 Oct 2020 17:53:49 +0000 (19:53 +0200)]
IBM Z: Emit vector alignment hints for strlen

In case the vectorized version of strlen is used, then each memory
access inside the loop is 16-byte aligned.  Thus add this kind of
information so that vector alignment hints can later on be emitted.

gcc/ChangeLog:

* config/s390/s390.c (s390_expand_vec_strlen): Add alignment
for memory access inside loop.

4 years agoDaily bump.
GCC Administrator [Tue, 20 Oct 2020 00:16:29 +0000 (00:16 +0000)]
Daily bump.

4 years agoUse precision and sign to compare types for ranges
Andrew MacLeod [Mon, 19 Oct 2020 23:04:40 +0000 (19:04 -0400)]
Use precision and sign to compare types for ranges

Sanity check ranges by comparing just SIGN and PRECISION.

gcc/
PR tree-optimization/97360
* gimple-range.h (range_compatible_p): New.
* gimple-range-gori.cc (is_gimple_logical_p): Use range_compatible_p.
(range_is_either_true_or_false): Ditto.
(gori_compute::outgoing_edge_range_p): Cast result to the correct
type if necessary.
(logical_stmt_cache::cacheable_p): Use range_compatible_p.
* gimple-range.cc (gimple_ranger::calc_stmt): Check range_compatible_p
before casting the range.
(gimple_ranger::range_on_exit): Use range_compatible_p.
(gimple_ranger::range_on_edge): Ditto.

gcc/testsuite/
* gcc.dg/pr97360-2.c: New test.

4 years agotestsuite: Enable and adjust powerpc vec-extract and vsu testcases.
David Edelsohn [Mon, 19 Oct 2020 22:31:07 +0000 (18:31 -0400)]
testsuite: Enable and adjust powerpc vec-extract and vsu testcases.

This patch remove unnecessary target requirements from many vec-extract
testcases and enables the vsu testcases on AIX.

gcc/testsuite/ChangeLog:

* gcc.target/powerpc/sad-vectorize-1.c: Remove target.
* gcc.target/powerpc/sad-vectorize-2.c: Remove target.
* gcc.target/powerpc/vec-extract-v16qi-df.c: Remove target.
* gcc.target/powerpc/vec-extract-v16qi.c: Remove target.
* gcc.target/powerpc/vec-extract-v16qiu-df.c: Remove target.
* gcc.target/powerpc/vec-extract-v16qiu.c: Remove target.
* gcc.target/powerpc/vec-extract-v2df.c: Remove target.
* gcc.target/powerpc/vec-extract-v2di.c: Require lp64.
* gcc.target/powerpc/vec-extract-v4sf.c: Remove target.
* gcc.target/powerpc/vec-extract-v4si-df.c: Remove target.
* gcc.target/powerpc/vec-extract-v4si.c: Remove target.
* gcc.target/powerpc/vec-extract-v4siu-df.c: Remove target.
* gcc.target/powerpc/vec-extract-v4siu.c: Remove target.
* gcc.target/powerpc/vec-extract-v8hi-df.c: Remove target.
* gcc.target/powerpc/vec-extract-v8hi.c: Remove target.
* gcc.target/powerpc/vec-extract-v8hiu-df.c: Remove target.
* gcc.target/powerpc/vec-extract-v8hiu.c: Remove target.
* gcc.target/powerpc/vec-init-1.c: Remove target.
* gcc.target/powerpc/vec-init-2.c: Require lp64.
* gcc.target/powerpc/vec-init-3.c: Require lp64.
* gcc.target/powerpc/vec-init-4.c: Remove target.
* gcc.target/powerpc/vec-init-5.c: Remove target.
* gcc.target/powerpc/vec-init-6.c: Require lp64.
* gcc.target/powerpc/vec-init-7.c: Require lp64.
* gcc.target/powerpc/vec-init-8.c: Remove target.
* gcc.target/powerpc/vec-init-9.c: Require lp64.
* gcc.target/powerpc/vec-setup-double.c: Remove target.
* gcc.target/powerpc/vec-setup-long.c: Remove target.
* gcc.target/powerpc/vsu/vec-xl-len-13.c: Correct expected warnings.
* gcc.target/powerpc/vsu/vsu.exp: Enable on AIX.

4 years agolibstdc++: Implement std::make_unique_for_overwrite
Jonathan Wakely [Mon, 19 Oct 2020 21:11:39 +0000 (22:11 +0100)]
libstdc++: Implement std::make_unique_for_overwrite

This is the std::unique_ptr part of P1020R1 (as amended by P1973R1) for
C++20. The std::shared_ptr part still needs to be done.

libstdc++-v3/ChangeLog:

* include/bits/unique_ptr.h (make_unique_for_overwrite): Define
for C++20.
* testsuite/20_util/unique_ptr/creation/array_neg.cc: Remove
unused header. Adjust standard reference.
* testsuite/20_util/unique_ptr/creation/for_overwrite.cc: New test.
* testsuite/20_util/unique_ptr/creation/for_overwrite__neg.cc: New test.

4 years agotestsuite: Enable and adjust powerpc fold-vec-extract/insert testcases.
David Edelsohn [Mon, 19 Oct 2020 13:50:44 +0000 (09:50 -0400)]
testsuite: Enable and adjust powerpc fold-vec-extract/insert testcases.

This patch enables a number of testcases on non-Linux targets, adjusts
the target requirements, and updates the expected results.

gcc/testsuite/ChangeLog:

* gcc.target/powerpc/fold-vec-extract-char.p7.c: Remove target.
* gcc.target/powerpc/fold-vec-extract-char.p8.c: Remove target.
* gcc.target/powerpc/fold-vec-extract-char.p9.c: Remove target.
Expect 3 addi for ilp32.
* gcc.target/powerpc/fold-vec-extract-double.p7.c: Remove target.
Add -mbig-endian for Linux.
* gcc.target/powerpc/fold-vec-extract-double.p8.c: Remove target.
* gcc.target/powerpc/fold-vec-extract-float.p7.c: Remove target.
* gcc.target/powerpc/fold-vec-extract-float.p8.c: Remove target.
* gcc.target/powerpc/fold-vec-extract-float.p9.c: Require lp64.
* gcc.target/powerpc/fold-vec-extract-int.p7.c: Remove target.
* gcc.target/powerpc/fold-vec-extract-int.p8.c: Remove target.
* gcc.target/powerpc/fold-vec-extract-int.p9.c: Remove target.
Expect 3 addi for ilp32.
* gcc.target/powerpc/fold-vec-extract-longlong.p7.c: Remove target.
Expect 4 addi for ilp32.
* gcc.target/powerpc/fold-vec-extract-longlong.p8.c: Remove target.
Expect 4 addi for ilp32.
* gcc.target/powerpc/fold-vec-extract-longlong.p9.c: Require lp64.
* gcc.target/powerpc/fold-vec-extract-short.p7.c: Remove target.
* gcc.target/powerpc/fold-vec-extract-short.p8.c: Remove target.
* gcc.target/powerpc/fold-vec-extract-short.p9.c: Require lp64.
* gcc.target/powerpc/fold-vec-insert-char-p8.c: Remove target.
* gcc.target/powerpc/fold-vec-insert-char-p9.c: Remove target.
* gcc.target/powerpc/fold-vec-insert-double.c: Remove target.
* gcc.target/powerpc/fold-vec-insert-float-p8.c: Remove target.
* gcc.target/powerpc/fold-vec-insert-float-p9.c: Remove target.
* gcc.target/powerpc/fold-vec-insert-int-p8.c: Remove target.
* gcc.target/powerpc/fold-vec-insert-int-p9.c: Remove target.
Require 8 addi for ilp32.
* gcc.target/powerpc/fold-vec-insert-longlong.c: Remove target.
* gcc.target/powerpc/fold-vec-insert-short-p8.c: Remove target.
* gcc.target/powerpc/fold-vec-insert-short-p9.c: Remove target.
* gcc.target/powerpc/fold-vec-select-double.c: Remove target.
* gcc.target/powerpc/fold-vec-select-float.c: Remove target.
* gcc.target/powerpc/fold-vec-splats-int.c: Require lp64.
* gcc.target/powerpc/fold-vec-splats-longlong.c: Require lp64.

4 years agocoroutines: Emit error for invalid promise return types [PR97438].
Iain Sandoe [Sat, 17 Oct 2020 10:21:11 +0000 (11:21 +0100)]
coroutines: Emit error for invalid promise return types [PR97438].

At one stage, use cases were proposed for allowing the promise
type to contain both return_value and return_void.  That was
not accepted into C++20, so we should reject it as per the PR.

gcc/cp/ChangeLog:

PR c++/97438
* coroutines.cc (struct coroutine_info): Add a field to
record that we emitted a promise type error.
(coro_promise_type_found_p): Check for the case that the
promise type contains both return_void and return_value.
Emit an error if so, with information about the wrong
type methods.

gcc/testsuite/ChangeLog:

PR c++/97438
* g++.dg/coroutines/pr97438.C: New test.

4 years agocplxlower: Avoid a transform when looking at a default definition
Martin Jambor [Mon, 19 Oct 2020 17:21:10 +0000 (19:21 +0200)]
cplxlower: Avoid a transform when looking at a default definition

In PR 97456, IPA-SRA triggers a bug in tree-complex.c where it
converts:

 <bb 2>
   a$_M_value_21 = COMPLEX_EXPR <ISRA.18_10(D), ISRA.18_10(D)>;

(where ISRA.18 is IPA-SRA created PARM_DECL with DECL_IGNORED_P set,
which is why it only happens with IPA-SRA) into:

  <bb 2>
    a$_M_value_21 = COMPLEX_EXPR <a$_M_value$real_10(D), a$_M_value$real_10(D)>;

i.e. it replaces two uses of the parameter default-def with two
uninitialized default-defs of a new variable - all in hope to produce
code with better debug info.

This patch fixes it by avoiding the transform when the SSA_NAME to be
replaced is a default definition.

gcc/ChangeLog:

2020-10-19  Martin Jambor  <mjambor@suse.cz>

PR tree-optimization/97456
* tree-complex.c (set_component_ssa_name): Do not replace ignored decl
default definitions with new component vars.  Reorder if conditions.

gcc/testsuite/ChangeLog:

2020-10-19  Martin Jambor  <mjambor@suse.cz>

PR tree-optimization/97456
* gcc.dg/tree-ssa/pr97456.c: New test.

4 years agolibstdc++: Use reserved name for C++20 attribute
Jonathan Wakely [Mon, 19 Oct 2020 16:57:14 +0000 (17:57 +0100)]
libstdc++: Use reserved name for C++20 attribute

Although the compiler supports the [[no_unique_address]] attribute, it's
not a reserved name prior to C++20, so we can't use it in std::tuple.
Use [[__no_unique_address__]] instead.

libstdc++-v3/ChangeLog:

* include/std/tuple (_Head_base<Idx, Head, true>): Use reserved
form of __no_unique_address__ attribute because
no_unique_address is not reserved prior to C++20.

4 years agolibstdc++: Optimize container node-handle type for size
Jonathan Wakely [Mon, 19 Oct 2020 16:56:54 +0000 (17:56 +0100)]
libstdc++: Optimize container node-handle type for size

The use of std::optional in _Node_handle makes the node handle types for
associative and unordered containers larger than necessary. It also
greatly increases the amount of code included, as <optional> is quite
large.

The boolean flag that records whether the std::optional contains a value
is redundant, because the _Node_handle::_M_ptr member provides the same
information. If the node handle has a non-null pointer it also has an
allocator, and not otherwise. By replacing std::optional with a custom
union type (and using _M_ptr to tell which union member is active) all
node handle sizes can be reduced by sizeof(allocator_type::pointer).

This makes the node handle types incompatible with previous releases, so
must be done before the C++17 ABI is fixed for GCC 11.

libstdc++-v3/ChangeLog:

* include/bits/node_handle.h (_Node_handle_common): Replace
std::optional with custom type.
* testsuite/20_util/variant/exception_safety.cc: Add missing
header include.

4 years agopreprocessor: Fix non-fn fn-like macro at EOF [PR97471]
Nathan Sidwell [Mon, 19 Oct 2020 14:57:50 +0000 (07:57 -0700)]
preprocessor: Fix non-fn fn-like macro at EOF [PR97471]

We inject EOF tokens between macro argument lists, but had
confused/stale logic in the non-fn invocation.  Renamed the magic
'eof' token, as it's now only used for macro argument termination.
Always rewind the non-OPEN_PAREN token.

libcpp/
* internal.h (struct cpp_reader): Rename 'eof' field to 'endarg'.
* init.c (cpp_create_reader): Adjust.
* macro.c (collect_args): Use endarg for separator.  Always rewind
in the not-fn case.
gcc/testsuite/
* c-c++-common/cpp/pr97471.c: New.

4 years agors6000: correct BE vextract_fp_from_short[hl] vperm masks
David Edelsohn [Mon, 19 Oct 2020 12:59:46 +0000 (08:59 -0400)]
rs6000: correct BE vextract_fp_from_short[hl] vperm masks

xvcvhpsp instruction converts a vector of bfloat16 half precision to single
precision.  The intrinsics vextract_fp_from_shorth and
vextract_fp_from_shortl select the high or low four elements of a
half precision vector to convert.  The intrinsics use vperm to select
the appropriate portion of the half precision vector and redistribute
the values for the xvcvhpsp instruction.  The big endian versions of the
masks for the intrinsics were initialized wrong.  This patch replaces the
masks with the correct values.  This corrects the failure of
builtins-3-p9-runnable.c testcase on big endian systems.

Bootstrapped powerpc-ibm-aix7.2.3.0 Power9.

gcc/ChangeLog:

* config/rs6000/vsx.md (vextract_fp_from_shorth):  Fix vals_be.
(vextract_fp_from_shortl) Same.

4 years agoGracefully handle right shifts larger than the precision.
Aldy Hernandez [Mon, 19 Oct 2020 10:18:46 +0000 (06:18 -0400)]
Gracefully handle right shifts larger than the precision.

gcc/ChangeLog:

PR tree-optimization/97488
* range-op.cc (operator_lshift::op1_range): Handle large right shifts.

gcc/testsuite/ChangeLog:

* gcc.dg/pr97488.c: New test.

4 years agoIPA: fix one more UBSAN error
Martin Liska [Mon, 19 Oct 2020 08:52:53 +0000 (10:52 +0200)]
IPA: fix one more UBSAN error

It fixes:

/home/marxin/Programming/gcc2/gcc/ipa-modref-tree.h:482:22: runtime error: load of value 255, which is not a valid value for type 'bool'
    #0 0x18e5df3 in modref_tree<int>::merge(modref_tree<int>*, vec<modref_parm_map, va_heap, vl_ptr>*) /home/marxin/Programming/gcc2/gcc/ipa-modref-tree.h:482
    #1 0x18dc180 in ipa_merge_modref_summary_after_inlining(cgraph_edge*) /home/marxin/Programming/gcc2/gcc/ipa-modref.c:1779
    #2 0x18c1c72 in inline_call(cgraph_edge*, bool, vec<cgraph_edge*, va_heap, vl_ptr>*, int*, bool, bool*) /home/marxin/Programming/gcc2/gcc/ipa-inline-transform.c:492
    #3 0x4a3589c in inline_small_functions /home/marxin/Programming/gcc2/gcc/ipa-inline.c:2216
    #4 0x4a3b230 in ipa_inline /home/marxin/Programming/gcc2/gcc/ipa-inline.c:2697
    #5 0x4a3d902 in execute /home/marxin/Programming/gcc2/gcc/ipa-inline.c:3096
    #6 0x1edf831 in execute_one_pass(opt_pass*) /home/marxin/Programming/gcc2/gcc/passes.c:2509
    #7 0x1ee26af in execute_ipa_pass_list(opt_pass*) /home/marxin/Programming/gcc2/gcc/passes.c:2936
    #8 0x103f31b in ipa_passes /home/marxin/Programming/gcc2/gcc/cgraphunit.c:2700
    #9 0x103fb40 in symbol_table::compile() /home/marxin/Programming/gcc2/gcc/cgraphunit.c:2777
    #10 0x104092b in symbol_table::finalize_compilation_unit() /home/marxin/Programming/gcc2/gcc/cgraphunit.c:3022
    #11 0x235723b in compile_file /home/marxin/Programming/gcc2/gcc/toplev.c:485
    #12 0x235fff9 in do_compile /home/marxin/Programming/gcc2/gcc/toplev.c:2321
    #13 0x23605fc in toplev::main(int, char**) /home/marxin/Programming/gcc2/gcc/toplev.c:2460
    #14 0x4e2b93b in main /home/marxin/Programming/gcc2/gcc/main.c:39
    #15 0x7ffff6f0ae09 in __libc_start_main ../csu/libc-start.c:314
    #16 0x9a0be9 in _start (/home/marxin/Programming/gcc2/objdir/gcc/cc1+0x9a0be9)

gcc/ChangeLog:

* ipa-modref.c (compute_parm_map): Clear vector.

4 years agotree-optimization/97486 - avoid edge insertion in SLP vectorizing
Richard Biener [Mon, 19 Oct 2020 09:30:46 +0000 (11:30 +0200)]
tree-optimization/97486 - avoid edge insertion in SLP vectorizing

This avoids edge inserting and eventual splitting during BB SLP
vectorization for now.

2020-10-19  Richard Biener  <rguenther@suse.de>

PR tree-optimization/97486
* tree-vect-slp.c (vect_slp_function): Split after stmts
ending a BB.

* gcc.dg/vect/bb-slp-pr97486.c: New testcase.

4 years agolibstdc++: Remove unused header
Jonathan Wakely [Mon, 19 Oct 2020 10:18:42 +0000 (11:18 +0100)]
libstdc++: Remove unused header

This header was not part of the patch submitted and reviewed, and was
not added to include/Makefile.am so never gets installed anyway. It
appears to have been committed by mistake, so let's remove it.

libstdc++-v3/ChangeLog:

* include/precompiled/expc++.h: Removed.

4 years agodoc: Add closing parenthesis to -ffat-lto-objects docs
Jonathan Wakely [Mon, 19 Oct 2020 09:24:03 +0000 (10:24 +0100)]
doc: Add closing parenthesis to -ffat-lto-objects docs

gcc/ChangeLog:

* doc/invoke.texi (OPptimize Options): Add missing closing
parenthesis.

4 years ago[Ada] Ada_2020: Implement Key_Expression for named container aggregates
Ed Schonberg [Fri, 10 Jul 2020 15:13:57 +0000 (11:13 -0400)]
[Ada] Ada_2020: Implement Key_Expression for named container aggregates

gcc/ada/

* par-ch4.adb: (P_Aggregate_Or_Paren_Expr): Recognize
Iterated_Element_Component.
(P_Iterated_Component_Association): Rebuild node as an Iterated_
Element_Association when Key_Expression is present, and attach
either the Loop_Parameter_Specification or the
Iterator_Specification to the new node.
* sem_aggr.adb: (Resolve_Container_Aggregate):
Resolve_Iterated_Association handles bota Iterated_Component_
and Iterated_Element_Associations, in which case it analyzes and
resoles the orresponding Key_Expression.
* exp_aggr.adb (Expand_Iterated_Component): If a Key_Expression
is present, use it as the required parameter in the call to the
insertion routine for the destination container aggregate. Call
this routine for both kinds of Iterated_Associations.

4 years ago[Ada] Missing check on array concatenation
Arnaud Charlet [Fri, 10 Jul 2020 12:21:53 +0000 (08:21 -0400)]
[Ada] Missing check on array concatenation

gcc/ada/

* exp_ch4.adb (Expand_Concatenate): Enable needed range checks.

4 years ago[Ada] Suppress warnings in expansion of "for ... of" loops
Bob Duff [Fri, 10 Jul 2020 11:37:33 +0000 (07:37 -0400)]
[Ada] Suppress warnings in expansion of "for ... of" loops

gcc/ada/

* exp_ch6.adb (Make_Build_In_Place_Call_In_Object_Declaration):
Set the Warnings_Off flag on the pointer object used in the
expansion of iterators and similar.

4 years ago[Ada] Use alternate stack for signal handling on PowerPC/Linux
Eric Botcazou [Fri, 10 Jul 2020 08:28:45 +0000 (10:28 +0200)]
[Ada] Use alternate stack for signal handling on PowerPC/Linux

gcc/ada/

* Makefile.rtl (PowerPC/Linux): Use s-taspri__posix.ads instead
of s-taspri__posix-noaltstack.ads for s-taspri.ads.

4 years ago[Ada] Remove excessive guards in building predicate functions
Piotr Trojanek [Thu, 9 Jul 2020 20:16:46 +0000 (22:16 +0200)]
[Ada] Remove excessive guards in building predicate functions

gcc/ada/

* sem_ch13.adb (Add_Call): Remove excessive condition and
unnecessary call to Set_Has_Predicates.

4 years ago[Ada] Alternative display of multi-line messages for GNATprove
Yannick Moy [Thu, 9 Jul 2020 09:52:49 +0000 (11:52 +0200)]
[Ada] Alternative display of multi-line messages for GNATprove

gcc/ada/

* debug.adb: Use debug switch -gnatdF for this alternative
display of messages.
* errout.adb (Output_Messages): Alternative display when -gnatdF
is used.
* erroutc.adb (Output_Msg_Text): Likewise.

4 years ago[Ada] AI12-0352: Early derivation and equality of untagged types
Arnaud Charlet [Thu, 9 Jul 2020 12:26:37 +0000 (08:26 -0400)]
[Ada] AI12-0352: Early derivation and equality of untagged types

gcc/ada/

* sem_ch6.adb (Check_Untagged_Equality): Check for AI12-0352.

4 years ago[Ada] Compiler abort on in_out function parameter with type invariant
Ed Schonberg [Wed, 8 Jul 2020 13:04:16 +0000 (09:04 -0400)]
[Ada] Compiler abort on in_out function parameter with type invariant

gcc/ada/

* exp_ch6.adb (Add_View_Conversion_Invariants): Do not insert
generated invariant checks when the call is a function call.
These tests are properly inserted in the code in procedure
Insert_Post_Call_Actions, which takes care of finding the proper
insertion point for the checks.
(Insert_Post_Call_Actions): Add question marks to indicate
possible gap in handling function calls that appear as aggregate
components.

4 years ago[Ada] Wrong freezing for expression function with contracts in a generic
Arnaud Charlet [Thu, 9 Jul 2020 09:36:18 +0000 (05:36 -0400)]
[Ada] Wrong freezing for expression function with contracts in a generic

gcc/ada/

* contracts.adb (Process_Preconditions_For): Do not exclude
instances.
* sem_ch4.adb (Analyze_Quantified_Expression): Disable spurious
warning on internally generated variables.

4 years ago[Ada] Implement initialization of CUDA runtime
Ghjuvan Lacambre [Thu, 30 Jan 2020 10:47:00 +0000 (11:47 +0100)]
[Ada] Implement initialization of CUDA runtime

gcc/ada/

* debug.adb: Document -gnatd_c flag as being used for CUDA.
* gnat_cuda.ads: New file.
* gnat_cuda.adb: New file.
* rtsfind.ads: Add Interfaces_C_Strings package and
RE_Fatbin_Wrapper, RE_Register_Fat_Binary,
RE_Register_Fat_Binary_End, RE_Register_Function, RE_Chars_Ptr,
RE_New_Char_Array entities.
* rtsfind.adb: Create new Interfaces_C_Descendant subtype,
handle it.
* sem_ch7.adb (Analyze_Package_Body_Helper): Call CUDA init
procedure.
* sem_prag.adb (Analyze_Pragma): Call Add_Cuda_Kernel procedure.
* gcc-interface/Make-lang.in (GNAT_ADA_OBJS): Add gnat_cuda.o.

4 years ago[Ada] Expanded names in ghost assignments
Bob Duff [Tue, 7 Jul 2020 21:29:44 +0000 (17:29 -0400)]
[Ada] Expanded names in ghost assignments

gcc/ada/

* ghost.adb (Whole_Object_Ref): New function to compute the name
of the whole object.
(Mark_And_Set_Ghost_Assignment): Rewrite to use
Whole_Object_Ref.  We need to partly analyze the left-hand side
in order to distinguish expanded names and record components.
* lib-xref.ads, lib-xref.adb (Deferred_References): Move table
to body, and add Defer_Reference to update the table, avoiding
duplicates.
(Generate_Reference): Avoid duplicates.
* sem_ch8.ads, sem_ch8.adb (Find_Direct_Name): Remove _OK
parameters, which are no longer needed. Ignore errors in
Ignore_Errors mode.
* sem_util.ads, sem_util.adb (Preanalyze_Without_Errors): Make
this public, so we can call it from Ghost.
* errout.ads, scng.adb, sem_prag.adb: Minor.

4 years ago[Ada] Do not replace bounds for packed arrays that depend on discriminants
Ghjuvan Lacambre [Mon, 6 Jul 2020 13:56:28 +0000 (15:56 +0200)]
[Ada] Do not replace bounds for packed arrays that depend on discriminants

gcc/ada/

* exp_attr.adb (Expand_N_Attribute_Reference): Check if type
depends on discriminant.

4 years ago[Ada] Ada2020: AI12-0304 Put_Image attrs of lang-def types
Bob Duff [Tue, 7 Jul 2020 18:36:50 +0000 (14:36 -0400)]
[Ada] Ada2020: AI12-0304 Put_Image attrs of lang-def types

gcc/ada/

* libgnat/a-coinve.adb, libgnat/a-cidlli.adb (Put_Image): Call
Iterate.

4 years ago[Ada] ACATS 4.1R - Exception missed
Arnaud Charlet [Tue, 7 Jul 2020 13:09:32 +0000 (09:09 -0400)]
[Ada] ACATS 4.1R - Exception missed

gcc/ada/

* sem_aggr.adb (Resolve_Record_Aggregate): Properly apply
subtype constraints when using a Default_Value.
* freeze.adb: Fix typo.

4 years ago[Ada] Reject use of Relaxed_Initialization on scalar/access param or result
Yannick Moy [Mon, 6 Jul 2020 12:58:28 +0000 (14:58 +0200)]
[Ada] Reject use of Relaxed_Initialization on scalar/access param or result

gcc/ada/

* sem_ch13.adb (Analyze_Aspect_Relaxed_Initialization): Fix bug
where a call to Error_Msg_N leads to crash due to
Error_Msg_Name_1 being removed by the call, while a subsequent
call to Error_Msg_N tries to use it. The variable
Error_Msg_Name_1 should be restored prior to the next call. Also
add checking for the new rules.

4 years ago[Ada] No range check on fixed point to integer conversion
Arnaud Charlet [Mon, 29 Jun 2020 08:22:35 +0000 (04:22 -0400)]
[Ada] No range check on fixed point to integer conversion

gcc/ada/

* checks.adb (Apply_Type_Conversion_Checks): Minor code clean
up.
* exp_ch4.adb (Discrete_Range_Check): Optimize range checks.
Update comments.
(Expand_N_Type_Conversion): Generate range check when rewriting
a type conversion if needed. Add assertion.
* exp_ch6.adb (Expand_Simple_Function_Return): Minor code clean
up.
* sem_res.adb (Resolve_Type_Conversion): Apply range check when
needed.  Update comments.

4 years ago[Ada] Clarify protection offered by preconditions on Ada.Text_IO units
Yannick Moy [Tue, 7 Jul 2020 09:19:47 +0000 (11:19 +0200)]
[Ada] Clarify protection offered by preconditions on Ada.Text_IO units

gcc/ada/

* libgnat/a-textio.ads: Update top-level comment.

4 years ago[Ada] Fix typo in error message about contract cases
Piotr Trojanek [Fri, 3 Jul 2020 13:42:53 +0000 (15:42 +0200)]
[Ada] Fix typo in error message about contract cases

gcc/ada/

* sem_prag.adb (Analyze_Contract_Cases_In_Decl_Part): Fix typo.

4 years ago[Ada] Simplify detection of by-copy types
Piotr Trojanek [Thu, 2 Jul 2020 21:39:06 +0000 (23:39 +0200)]
[Ada] Simplify detection of by-copy types

gcc/ada/

* exp_ch6.adb (Expand_Actuals): Simplify condition for by-copy
types.

4 years ago[Ada] Clarify current design of Errout wrt global variable usage
Yannick Moy [Mon, 6 Jul 2020 16:17:34 +0000 (18:17 +0200)]
[Ada] Clarify current design of Errout wrt global variable usage

gcc/ada/

* errout.ads: Add comment regarding lack of preservation of
Errout state across successive calls to the API.

4 years ago[Ada] Ada2020: AI12-0304 Put_Image attrs of lang-def types
Bob Duff [Thu, 2 Jul 2020 17:32:40 +0000 (13:32 -0400)]
[Ada] Ada2020: AI12-0304 Put_Image attrs of lang-def types

gcc/ada/

* exp_put_image.adb (Build_Elementary_Put_Image_Call): Use the
base type to recognize various cases of access types.
* libgnat/a-cbdlli.adb, libgnat/a-cbdlli.ads, libgnat/a-cbhama.adb,
libgnat/a-cbhama.ads, libgnat/a-cbhase.adb, libgnat/a-cbhase.ads,
libgnat/a-cbmutr.adb, libgnat/a-cbmutr.ads, libgnat/a-cborma.adb,
libgnat/a-cborma.ads, libgnat/a-cborse.adb, libgnat/a-cborse.ads,
libgnat/a-cdlili.adb, libgnat/a-cdlili.ads, libgnat/a-cidlli.adb,
libgnat/a-cidlli.ads, libgnat/a-cihama.adb, libgnat/a-cihama.ads,
libgnat/a-cihase.adb, libgnat/a-cihase.ads, libgnat/a-cimutr.adb,
libgnat/a-cimutr.ads, libgnat/a-ciorma.adb, libgnat/a-ciorma.ads,
libgnat/a-ciormu.adb, libgnat/a-ciormu.ads, libgnat/a-ciorse.adb,
libgnat/a-ciorse.ads, libgnat/a-coboho.adb, libgnat/a-coboho.ads,
libgnat/a-cobove.adb, libgnat/a-cobove.ads, libgnat/a-cohama.adb,
libgnat/a-cohama.ads, libgnat/a-cohase.adb, libgnat/a-cohase.ads,
libgnat/a-coinho.adb, libgnat/a-coinho.ads,
libgnat/a-coinho__shared.adb, libgnat/a-coinho__shared.ads,
libgnat/a-coinve.adb, libgnat/a-coinve.ads, libgnat/a-comutr.adb,
libgnat/a-comutr.ads, libgnat/a-coorma.adb, libgnat/a-coorma.ads,
libgnat/a-coormu.adb, libgnat/a-coormu.ads, libgnat/a-coorse.adb,
libgnat/a-coorse.ads, libgnat/a-strunb.adb, libgnat/a-strunb.ads,
libgnat/a-strunb__shared.adb, libgnat/a-strunb__shared.ads:
Implement Put_Image attibute.
* libgnat/a-stteou.ads, libgnat/s-putima.ads,
libgnat/a-stouut.ads, libgnat/a-stoubu.adb: Make
Ada.Strings.Text_Output, Ada.Strings.Text_Output.Utils, and
System.Put_Images Pure, so they can be with'ed by Pure units
that should have Put_Image defined.
* libgnat/a-stouut.adb: Add missing column adjustments, and
remove a redundant one.
* libgnat/s-putima.adb (Put_Arrow): New routine to print an
arrow. Avoids adding a with clause to some containers.

4 years ago[Ada] Ada2020: matching parentheses and brackets
Bob Duff [Thu, 2 Jul 2020 22:16:57 +0000 (18:16 -0400)]
[Ada] Ada2020: matching parentheses and brackets

gcc/ada/

* par-ch4.adb (P_Aggregate_Or_Paren_Expr): Require matching
parens or brackets.
* par.adb, par-tchk.adb (T_Right_Bracket): New procedure to give
an error on missing ].

4 years ago[Ada] Crash in expression function defined in protected object
Javier Miranda [Mon, 6 Jul 2020 12:04:01 +0000 (08:04 -0400)]
[Ada] Crash in expression function defined in protected object

gcc/ada/

* sem_ch8.adb (Find_Direct_Name): Do not trust in the decoration
of the Entity attribute in constants associated with
discriminals of protected types.

4 years ago[Ada] AI12-0170: Abstract subprogram calls in class-wide precond exprs
Gary Dismukes [Sun, 5 Jul 2020 17:29:02 +0000 (13:29 -0400)]
[Ada] AI12-0170: Abstract subprogram calls in class-wide precond exprs

gcc/ada/

* sem_disp.adb (Check_Dispatching_Context): When the enclosing
subprogram is abstract, bypass early return if the call is
tag-indeterminate, to continue with the later error checking.

4 years ago[Ada] Private type unexpectedly visible
Javier Miranda [Sun, 5 Jul 2020 10:43:30 +0000 (06:43 -0400)]
[Ada] Private type unexpectedly visible

gcc/ada/

* sem_ch7.adb (Uninstall_Declarations): Uninstall the
declaration of a subtype declaration defined in the private part
of a package.

4 years ago[Ada] Clean up support of square brackets
Arnaud Charlet [Thu, 2 Jul 2020 10:10:28 +0000 (06:10 -0400)]
[Ada] Clean up support of square brackets

gcc/ada/

* par-ch4.adb (P_Aggregate_Or_Paren_Expr): Simplify code since
we are always under -gnatX if we encounter a Tok_Left_Bracket.
* scng.adb (Scan): [] is an aggregate under -gnatX and a wide
character otherwise.

4 years ago[Ada] Remove useless initialization and refine type of a local variable
Piotr Trojanek [Thu, 2 Jul 2020 12:09:06 +0000 (14:09 +0200)]
[Ada] Remove useless initialization and refine type of a local variable

gcc/ada/

* exp_ch6.adb (Expand_Call_Helper): Cleanup.

4 years ago[Ada] Refine type of a local variable
Piotr Trojanek [Thu, 2 Jul 2020 12:06:58 +0000 (14:06 +0200)]
[Ada] Refine type of a local variable

gcc/ada/

* exp_ch6.adb (Is_Direct_Deep_Call): Refine type from Node_Id to
Entity_Id.

4 years ago[Ada] Simplify membership test for operators
Piotr Trojanek [Thu, 2 Jul 2020 12:05:30 +0000 (14:05 +0200)]
[Ada] Simplify membership test for operators

gcc/ada/

* exp_ch6.adb (May_Fold): Detect all operators, i.e. both binary
and unary ones.

4 years ago[Ada] Simplify repeated calls with membership test
Piotr Trojanek [Thu, 2 Jul 2020 12:04:07 +0000 (14:04 +0200)]
[Ada] Simplify repeated calls with membership test

gcc/ada/

* inline.adb (Expand_Inlined_Call): Simplify repeated calls to
Nkind.

4 years agoHandle right shifts by zero in range-ops.
Aldy Hernandez [Mon, 19 Oct 2020 07:48:27 +0000 (09:48 +0200)]
Handle right shifts by zero in range-ops.

If the shift amount in operator_lshift::op1_range was zero, an invalid range
of [1, 0] was being created.

gcc/ChangeLog:

PR tree-optimization/97467
* range-op.cc (operator_lshift::op1_range): Handle shifts by 0.

gcc/testsuite/ChangeLog:

* gcc.dg/pr97467.c: New test.

4 years agotree-optimization/97466 - remove spurious assert
Richard Biener [Mon, 19 Oct 2020 09:07:36 +0000 (11:07 +0200)]
tree-optimization/97466 - remove spurious assert

This removes an assertion that was supposed to be only for temporary
debugging.  I've also re-indented the code which I missed as well.

2020-10-19  Richard Biener  <rguenther@suse.de>

PR tree-optimization/97466
* tree-vect-slp.c (vect_get_and_check_slp_defs): Remove
spurious assert, re-indent.

4 years ago[PATCH] fold x << (n % C) to x << (n & C-1) if C meets power2
guojiufu [Mon, 19 Oct 2020 05:48:19 +0000 (13:48 +0800)]
[PATCH] fold x << (n % C) to x << (n & C-1) if C meets power2

This patch fixes PR66552 which is also as:
https://gcc.gnu.org/pipermail/gcc-patches/2020-February/540930.html
which requests to optimizes (x shift (n mod C)) to
(x shift (n bit_and (C - 1))) when C is a constant and power of two.

gcc/ChangeLog
2020-10-19  Li Jia He  <helijia@gcc.gnu.org>

PR tree-optimization/66552
* match.pd (x << (n % C) -> x << (n & C-1)): New simplification.

gcc/testsuite/ChangeLog
2020-10-19  Li Jia He  <helijia@gcc.gnu.org>

PR tree-optimization/66552
* gcc.dg/pr66552.c: New testcase.

4 years agoSimplify comparison GIMPLE IL verification
Richard Biener [Mon, 19 Oct 2020 08:41:19 +0000 (10:41 +0200)]
Simplify comparison GIMPLE IL verification

There's an old extra allowance for same-mode pointer comparison
for which I don't see any good reason today where the only special-case
of pointers in useless_type_conversion_p is that of function/method
pointers vs. non-function/method pointers.

2020-10-19  Richard Biener  <rguenther@suse.de>

* tree-cfg.c (verify_gimple_comparison): Drop special-case
for pointer comparison.

4 years agolibsanitizer: Add recently added commit to LOCAL_PATCHES.
Martin Liska [Mon, 19 Oct 2020 07:05:34 +0000 (09:05 +0200)]
libsanitizer: Add recently added commit to LOCAL_PATCHES.

libsanitizer/ChangeLog:

* LOCAL_PATCHES: Add one commit.

4 years agoDaily bump.
GCC Administrator [Mon, 19 Oct 2020 00:16:22 +0000 (00:16 +0000)]
Daily bump.

4 years agorevamp ada.numerics.aux
Alexandre Oliva [Sun, 18 Oct 2020 20:19:53 +0000 (17:19 -0300)]
revamp ada.numerics.aux

Instead of mapping elementary functions for all types to a single
type, use the intrinsics available for the various base types.

A new Ada.Numerics.Aux_Generic_Float is introduced to explicitly
dispatch, based on the 'Digits attribute of the base type, to the
various newly-added Aux_Short_Float, Aux_Float, Aux_Long_Float, or
Aux_Long_Long_Float.

The Aux_Short_Float unit is implemented in terms of the Aux_Float one,
and the others rely on the elementary functions from the C Math
library for float, double and long double types, respectively.

An Aux_Linker_Options is added, and units that import intrinsics from
libm/libc depend on it to provide the "-lm" linker option if needed.
The option is provided by default, but there is an alternate version
that doesn't, that is used for vxworks targets.

The Aux variant that used to open-code Sin and Cos for the ancient
ppc-darwin, because of insufficient precision in libc, is dropped,
along with the alternate dummy body for Aux.  Both are presumed no
longer needed.

The original Ada.Numerics.Aux is retained, for backward compatibility,
as a wrapper for a newly-added Aux_Compat, that renames
Aux_Long_Float, except on x86, in which an alternate version renames
Aux_Long_Long_Float.

Generic_Elementary_Functions and Generic_Complex_Types are adjusted to
use Aux_Generic_Float, avoiding the type conversions and inefficiencies of
computing results in higher precision than requested.

Generic_Complex_Elementary_Functions is adjusted to enable an
additional instance of the sincos optimization, even without -gnatn.

for  gcc/ada/ChangeLog

* Makefile.rtl (GNATRTL_NONTASKING_OBJS): Compile Ada.Numerics
child units Aux_Generic_Float, Aux_Long_Long_Float, Aux_Long_Float,
Aux_Float, Aux_Short_Float, Aux_Compat, and Aux_Linker_Options.
(X86_TARGET_PAIRS): Drop dummy body for Aux.  Use x86 version
of Aux_Compat.
(X86_64_TARGET_PAIRS): Likewise.
(LIBGNAT_TARGET_PAIRS): On VxWorks, select the nolibm
variants.  Drop the darwin version of Aux.  Drop the redundant
libc-x86 numaux variants on x86* kfreebsd variants.
* libgnat/a-nagefl.ads: New Aux_Generic_Float.
* libgnat/a-naliop.ads: New Aux_Linker_Options.
* libgnat/a-naliop__nolibm.ads: New.
* libgnat/a-nallfl.ads: New Aux_Long_Long_Float.
* libgnat/a-nalofl.ads: New Aux_Long_Float.
* libgnat/a-nuaufl.ads: New Aux_Float.
* libgnat/a-nashfl.ads: New Aux_Short_Float.
* libgnat/a-ngcefu.adb (Exp): Factor out the Im (X) passed to
Sin and Cos in the Complex variant too.
* libgnat/a-ngcoty.adb: Switch to Aux_Generic_Float.  Drop
redundant conversions.
* libgnat/a-ngelfu.adb: Likewise.
* libgnat/a-nuauco.ads: New Aux_Compat.
* libgnat/a-nuauco__x86.ads: New.
* libgnat/a-numaux.ads: Replace with Compat wrapper.
* libgnat/a-numaux__darwin.adb: Remove.
* libgnat/a-numaux__darwin.ads: Remove.
* libgnat/a-numaux__dummy.adb: Remove.
* libgnat/a-numaux__libc-x86.ads: Remove.
* libgnat/a-numaux__vxworks.ads: Remove.

4 years agolibsanitizer, Darwin, Bootstrap : Fix bootstrap on Darwin <= 15.
Iain Sandoe [Sun, 18 Oct 2020 07:48:58 +0000 (08:48 +0100)]
libsanitizer, Darwin, Bootstrap : Fix bootstrap on Darwin <= 15.

The latest upstream merge for libsanitizer introduces code that makes
use of some macro values that are not available in SDKs for versions
of Darwin <= 15 (macOS 10.11).

Add definitions for these where they are not present.

libsanitizer/ChangeLog:

* sanitizer_common/sanitizer_mac.h: Ensure that TARGET_OS_
macros are defined where the macOS SDK does not contain
them.
(TARGET_OS_OSX, TARGET_OS_IOS, TARGET_OS_TV, TARGET_OS_WATCH):
Define where needed.

4 years agoPR libfortran/97063 - Wrong result for vector (step size is negative) * matrix
Harald Anlauf [Sun, 18 Oct 2020 18:15:26 +0000 (20:15 +0200)]
PR libfortran/97063 - Wrong result for vector (step size is negative) * matrix

The MATMUL intrinsic provided a wrong result for rank-1 times rank-2 array
when a negative stride was used for addressing the elements of the rank-1
array, because a check on strides was erroneously placed before the check
on the rank.  Interchange order of checks.

libgfortran/ChangeLog:

* m4/matmul_internal.m4: Move check for rank-1 times rank-2 before
checks on strides for rank-2 times rank-2.
* generated/matmul_c10.c: Regenerated.
* generated/matmul_c16.c: Likewise.
* generated/matmul_c4.c: Likewise.
* generated/matmul_c8.c: Likewise.
* generated/matmul_i1.c: Likewise.
* generated/matmul_i16.c: Likewise.
* generated/matmul_i2.c: Likewise.
* generated/matmul_i4.c: Likewise.
* generated/matmul_i8.c: Likewise.
* generated/matmul_r10.c: Likewise.
* generated/matmul_r16.c: Likewise.
* generated/matmul_r4.c: Likewise.
* generated/matmul_r8.c: Likewise.
* generated/matmulavx128_c10.c: Likewise.
* generated/matmulavx128_c16.c: Likewise.
* generated/matmulavx128_c4.c: Likewise.
* generated/matmulavx128_c8.c: Likewise.
* generated/matmulavx128_i1.c: Likewise.
* generated/matmulavx128_i16.c: Likewise.
* generated/matmulavx128_i2.c: Likewise.
* generated/matmulavx128_i4.c: Likewise.
* generated/matmulavx128_i8.c: Likewise.
* generated/matmulavx128_r10.c: Likewise.
* generated/matmulavx128_r16.c: Likewise.
* generated/matmulavx128_r4.c: Likewise.
* generated/matmulavx128_r8.c: Likewise.

gcc/testsuite/ChangeLog:

* gfortran.dg/matmul_20.f90: New test.

4 years agoDaily bump.
GCC Administrator [Sun, 18 Oct 2020 00:16:23 +0000 (00:16 +0000)]
Daily bump.

4 years agotestsuite: simplify target requirements for various Power9 testcases.
David Edelsohn [Sat, 17 Oct 2020 20:58:43 +0000 (16:58 -0400)]
testsuite: simplify target requirements for various Power9 testcases.

This patch removes unnnecessary or simplifies requirements for various
Power9 testcase.

gcc/testsuite/ChangeLog:

* gcc.target/powerpc/p9-dimode1.c: Remove target.
* gcc.target/powerpc/p9-dimode2.c: Remove target.
* gcc.target/powerpc/p9-fpcvt-1.c: Remove target.
* gcc.target/powerpc/p9-fpcvt-2.c: Require lp64.
* gcc.target/powerpc/p9-minmax-1.c: Remove target.
* gcc.target/powerpc/p9-minmax-2.c: Remove target.
* gcc.target/powerpc/p9-minmax-3.c: Remove target.
* gcc.target/powerpc/p9-splat-1.c: Require lp64.
* gcc.target/powerpc/p9-splat-2.c: Remove target.
* gcc.target/powerpc/p9-splat-3.c: Remove target.
* gcc.target/powerpc/p9-splat-4.c: Require lp64.
* gcc.target/powerpc/p9-vbpermd.c: Require lp64.
* gcc.target/powerpc/p9-vneg.c: Require lp64.
* gcc.target/powerpc/p9-vparity.c: Require lp64.
* gcc.target/powerpc/p9-vpermr.c: Require LE.
* gcc.target/powerpc/p9-xxbr-1.c: Remove target.
* gcc.target/powerpc/p9-xxbr-2.c: Require lp64.
* gcc.target/powerpc/p9-xxbr-3.c: Require lp64.

4 years agolibstdc++: Fix visitor return type diagnostics [PR97449]
Ville Voutilainen [Sat, 17 Oct 2020 19:08:50 +0000 (22:08 +0300)]
libstdc++: Fix visitor return type diagnostics [PR97449]

libstdc++-v3/ChangeLog:

PR libstdc++/97449
* include/std/variant
(__gen_vtable_impl<>::_S_apply_single_alt):
Diagnose visitor return type mismatches here..
(__gen_vtable_impl</*base case*/>::_S_apply):
..not here.

4 years agotestsuite: Enable builtins-3-p9.c on BE targets.
David Edelsohn [Thu, 15 Oct 2020 22:50:05 +0000 (18:50 -0400)]
testsuite: Enable builtins-3-p9.c on BE targets.

This patch removes the unnecessary "le" target requirement
of the builtins-3-p9.c testcase.

gcc/testsuite/ChangeLog:

2020-10-16  David Edelsohn  <dje.gcc@gmail.com>

* gcc.target/powerpc/builtins-3-p9.c: Remove le.