platform/upstream/gcc.git
3 years ago[Ada] Add Interface_Name field to E_Loop_Parameter
Bob Duff [Wed, 10 Feb 2021 12:42:05 +0000 (07:42 -0500)]
[Ada] Add Interface_Name field to E_Loop_Parameter

gcc/ada/

* gen_il-gen-gen_entities.adb (E_Loop_Parameter): Add
Interface_Name field.

3 years ago[Ada] Disable certain checks in predefined units
Bob Duff [Tue, 9 Feb 2021 19:23:37 +0000 (14:23 -0500)]
[Ada] Disable certain checks in predefined units

gcc/ada/

* sem_cat.adb (Check_Non_Static_Default_Expr): Allow nonstatic
expression in predefined unit with pragma Preelaborate.

3 years ago[Ada] Clarify the semantics of signed intrinsic shift operations
Yannick Moy [Fri, 5 Feb 2021 14:19:57 +0000 (15:19 +0100)]
[Ada] Clarify the semantics of signed intrinsic shift operations

gcc/ada/

* doc/gnat_rm/intrinsic_subprograms.rst: More details on shift
operations for signed types. Also add the missing Import and
Convention on the example.
* gnat_rm.texi: Regenerate.

3 years ago[Ada] Small cleanup in exception handling C code
Eric Botcazou [Tue, 9 Feb 2021 11:52:08 +0000 (12:52 +0100)]
[Ada] Small cleanup in exception handling C code

gcc/ada/

* raise-gcc.c: Include <cstdarg> instead of <stdarg.h> in C++.
Include <stdbool.h> and unconditionally <stdlib.h> in C.

3 years ago[Ada] Remove redundant guard from Find_Overlaid_Entity
Piotr Trojanek [Mon, 8 Feb 2021 12:55:05 +0000 (13:55 +0100)]
[Ada] Remove redundant guard from Find_Overlaid_Entity

gcc/ada/

* sem_util.ads (Find_Overlaid_Entity): Simplify comment for
spec.
* sem_util.adb (Find_Overlaid_Entity): Remove defensive code
from body.

3 years ago[Ada] Remove const qualifier on a couple of pointed-to types
Eric Botcazou [Mon, 8 Feb 2021 11:00:19 +0000 (12:00 +0100)]
[Ada] Remove const qualifier on a couple of pointed-to types

gcc/ada/

* argv.c: Add include of <stdlib.h> for the runtime.
(gnat_argv): Change type to char ** and initialize to NULL.
(gnat_envp): Likewise.
* argv-lynxos178-raven-cert.c: Add include of <stdlib.h>.
(gnat_argv): Change type to char ** and initialize to NULL.
(gnat_envp): Likewise.

3 years ago[Ada] Fix bug in if_expressions introduced by var-size nodes changes
Bob Duff [Fri, 5 Feb 2021 18:06:13 +0000 (13:06 -0500)]
[Ada] Fix bug in if_expressions introduced by var-size nodes changes

gcc/ada/

* gen_il-gen.adb (Setter_Needs_Parent): Add missing
Then_Actions.  Fix self-contradictory comment.
* exp_util.adb (Insert_Actions): Minor comment improvments.

3 years ago[Ada] Add support for folding more and/or expressions
Arnaud Charlet [Sat, 30 Jan 2021 16:52:54 +0000 (11:52 -0500)]
[Ada] Add support for folding more and/or expressions

gcc/ada/

* sem_eval.adb (Eval_Logical_Op, Test_Expression_Is_Foldable):
Add support for folding more "and"/"or" expressions.
* exp_util.adb (Side_Effect_Free): Fix handling of membership
tests.

3 years ago[Ada] Replace repeated calls by references to a local object
Piotr Trojanek [Wed, 3 Feb 2021 08:19:20 +0000 (09:19 +0100)]
[Ada] Replace repeated calls by references to a local object

gcc/ada/

* sem_res.adb (Resolve_Actual): Replace repeated calls to
"Etype (F)" with references to "F_Typ", which keeps the results
of exactly that call.

3 years ago[Ada] Variable-sized node types: improve error messages
Bob Duff [Mon, 1 Mar 2021 22:35:17 +0000 (17:35 -0500)]
[Ada] Variable-sized node types: improve error messages

gcc/ada/

* gen_il-gen.adb (To_Bit_Offset): Use 'Base to avoid overflow in
computations in Last_Bit when Offset = 'Last.
(Choose_Offset): Give a better error message when we run out of
fields.  In particular, point out that
Gen_IL.Internals.Bit_Offset'Last needs to be increased.

3 years ago[Ada] Variable-sized node types -- cleanup
Bob Duff [Thu, 25 Feb 2021 15:38:55 +0000 (10:38 -0500)]
[Ada] Variable-sized node types -- cleanup

gcc/ada/

* atree.ads, einfo-utils.ads, einfo-utils.adb, fe.h, gen_il.adb,
gen_il.ads, gen_il-gen-gen_entities.adb,
gen_il-gen-gen_nodes.adb, sem_ch12.adb, sem_ch3.adb,
sem_util.adb, sinfo-utils.ads, treepr.adb, types.ads: Clean up
??? comments and other comments.
* atree.adb: Clean up ??? comments and other comments.
(Validate_Node): Fix bug: "Off_0 (N) < Off_L (N)"
should be "Off_0 (N) <= Off_L (N)".
* gen_il-gen.adb, gen_il-gen.ads: Clean up ???
comments and other comments.  Add support for getter-specific
and setter-specific preconditions.  Detect the error of putting
a field in the wrong subrange.  Misc cleanup.
(Node_Field vs. Entity_Field): Clean up Nmake.  Improve
comments.
* gen_il-utils.ads: Misc cleanup. Move...
* gen_il-internals.ads: ... here.
* gen_il-utils.adb: Misc cleanup. Move...
* gen_il-internals.adb: ... here.
* gen_il-fields.ads: Move Was_Default_Init_Box_Association,
which was in the wrong subrange.  Add comments.  Misc cleanup.
* gen_il-types.ads: Add Named_Access_Kind.
* sinfo-cn.adb: Clean up ??? comments and other comments.
Remove redundant assertions.
* einfo.ads, sinfo.ads: Clean up ??? comments and other
comments.  Remove all the comments indicating field offsets.
These are obsolete now that Gen_IL computes the offsets
automatically.

3 years ago[Ada] Rename Ada 202* to Ada 2022
Arnaud Charlet [Mon, 1 Mar 2021 10:45:25 +0000 (05:45 -0500)]
[Ada] Rename Ada 202* to Ada 2022

gcc/ada/

* einfo.ads, errout.adb, errout.ads, exp_aggr.adb, exp_ch5.adb,
exp_ch6.adb, exp_ch8.adb, exp_ch9.adb, exp_imgv.adb,
exp_put_image.adb, fe.h, impunit.adb, impunit.ads,
libgnat/a-cobove.ads, libgnat/a-convec.ads, opt.ads,
par-ch12.adb, par-ch3.adb, par-ch4.adb, par-ch5.adb,
par-ch6.adb, par-prag.adb, par-util.adb, scans.ads, scng.adb,
sem_aggr.adb, sem_attr.adb, sem_ch10.adb, sem_ch12.adb,
sem_ch13.adb, sem_ch3.adb, sem_ch5.adb, sem_ch6.adb,
sem_ch8.adb, sem_elab.adb, sem_eval.adb, sem_prag.adb,
sem_res.adb, sem_type.adb, sem_util.adb, sem_util.ads,
sinfo.ads, snames.ads-tmpl, sprint.adb, switch-c.adb, usage.adb,
doc/gnat_ugn/building_executable_programs_with_gnat.rst,
doc/gnat_rm/implementation_defined_aspects.rst,
gcc-interface/trans.c: Update all references to Ada 2020 to Ada
2022. Rename pragma Ada_2020 to Ada_2022.  Update documentation
accordingly.
* gnat_ugn.texi, gnat_rm.texi: Regenerate.

3 years ago[Ada] Avoid inappropriate error messages regarding aggregates and variant parts
Steve Baird [Thu, 18 Feb 2021 01:54:53 +0000 (17:54 -0800)]
[Ada] Avoid inappropriate error messages regarding aggregates and variant parts

gcc/ada/

* sem_util.adb (Gather_Components): Factor the test that was
already being used to govern emitting a pre-Ada_2020 error
message into an expression function,
OK_Scope_For_Discrim_Value_Error_Messages. Call that new
function in two places: the point where the same test was being
performed previously, and in governing emission of a newer
Ada_2020 error message. In both cases, the out-mode parameter
Gather_Components.Report_Errors is set to True even if no error
messages are generated within Gather_Components.
* sem_util.ads: Correct a comment.

3 years ago[Ada] Add more initialization of Stored_Constraint
Richard Kenner [Sun, 14 Feb 2021 13:02:19 +0000 (08:02 -0500)]
[Ada] Add more initialization of Stored_Constraint

gcc/ada/

* sem_ch3.adb (Array_Type_Declaration, Build_Derived_Type):
Reinitialize Stored_Constraint when needed.
(Set_Modular_Size): Likewise.
* atree.adb: (Check_Vanishing_Fields): Add node id to debugging
information.

3 years ago[Ada] Fix bug in subtype of private type with invariants
Bob Duff [Sat, 13 Feb 2021 21:43:22 +0000 (16:43 -0500)]
[Ada] Fix bug in subtype of private type with invariants

gcc/ada/

* sem_util.adb (Propagate_Invariant_Attributes): Call
Set_Has_Own_Invariants on the base type, because these are
Base_Type_Only. The problem is that the base type of a type is
indeed a base type when Set_Base_Type is called, but then the
type is mutated into a subtype in rare cases.
* atree.ads, atree.adb (Is_Entity): Export. Correct subtype of
parameter in body.
* gen_il-gen.adb: Improve getters so that "Pre => ..." can refer
to the value of the field. Put Warnings (Off) on some with
clauses that are not currently used, but might be used by such
Pre's.

3 years ago[Ada] Robust switching from incomplete to access types
Piotr Trojanek [Wed, 24 Feb 2021 18:39:21 +0000 (19:39 +0100)]
[Ada] Robust switching from incomplete to access types

gcc/ada/

* sem_ch3.adb (Access_Type_Declaration): Add comments to explain
the ordering of Mutate_Kind and Set_Directly_Designated_Type;
remove temporary setting of Ekind to E_Access_Type for building
_master objects, since now the Ekind is already set to its final
value. Move repeated code into Setup_Access_Type routine and use
it so that Process_Subtype is executed before mutating the kind
of the type entity.
* gen_il-gen-gen_entities.adb (Gen_Entities): Remove
Directly_Designated_Type from E_Void, E_Private_Record,
E_Limited_Private_Type and Incomplete_Kind; now it only belongs
to Access_Kind entities.
* sem_util.adb: Minor reformatting.

3 years agoexpr: Fix up VEC_PACK_TRUNC_EXPR expansion [PR101046]
Jakub Jelinek [Tue, 15 Jun 2021 09:36:47 +0000 (11:36 +0200)]
expr: Fix up VEC_PACK_TRUNC_EXPR expansion [PR101046]

The following testcase ICEs, because we have a mode mismatch.
VEC_PACK_TRUNC_EXPR's operands have different modes from the result
(same vector mode size but twice as large element),
but we were passing non-NULL subtarget with the mode of the result
to the expansion of its arguments, so the VEC_PERM_EXPR in one of the
operands which had V8SImode operands and result had V16HImode target.

Fixed by clearing the subtarget if we are changing mode.

2021-06-15  Jakub Jelinek  <jakub@redhat.com>

PR target/101046
* expr.c (expand_expr_real_2) <case VEC_PACK_FIX_TRUNC_EXPR,
case VEC_PACK_TRUNC_EXPR>: Clear subtarget when changing mode.

* gcc.target/i386/pr101046.c: New test.

3 years agoHandle multiple latches in irreducible region mark
Richard Biener [Mon, 14 Jun 2021 13:36:57 +0000 (15:36 +0200)]
Handle multiple latches in irreducible region mark

The following makes irreducible region discovery handle multiple latches.

2021-06-14  Richard Biener  <rguenther@suse.de>

* cfgloopanal.c (mark_irreducible_loops): Use a dominance
check to identify loop latches.
* cfgloop.c (verify_loop_structure): Likewise.
* loop-init.c (apply_loop_flags): Allow marked irreducible
regions even with multiple latches.
* predict.c (rebuild_frequencies): Simplify.

3 years agoAssert we have irreducible regions marked in the threader
Richard Biener [Tue, 15 Jun 2021 08:27:47 +0000 (10:27 +0200)]
Assert we have irreducible regions marked in the threader

This adds an assert that would fire if any thread path registry
user runs into the check that relies on such regions marked.

2021-06-15  Richard Biener  <rguenther@suse.de>

* tree-ssa-threadupdate.c
(jump_thread_path_registry::mark_threaded_blocks): Assert we
have marked irreducible regions.

3 years agotestsuite: Fix Wattributes test cases for s390 and add new tests.
Robin Dapp [Tue, 15 Jun 2021 07:06:15 +0000 (09:06 +0200)]
testsuite: Fix Wattributes test cases for s390 and add new tests.

There are several FAILs because we have an s390-specific check for a
warning which is not necessary anymore.  Remove it.

Add a new test case that expects a warning about conflicting function
alignment.  This would fail on s390 before but most likely on other
targets as well so it can be a target-independent test.

Also, add a test to verify that we do not emit a note when specifying
conflicting alignment for the same declaration.  Need to explicitly
handle every dg-note because handling one disables dg-note pruning.

gcc/testsuite/ChangeLog:

* c-c++-common/Wattributes.c: Remove s390-specific check and add
new tests.
* gcc.dg/Wattributes-6.c: Likewise.

3 years agoc-family: Copy DECL_USER_ALIGN even if DECL_ALIGN is similar.
Robin Dapp [Tue, 15 Jun 2021 07:06:02 +0000 (09:06 +0200)]
c-family: Copy DECL_USER_ALIGN even if DECL_ALIGN is similar.

When re-declaring a function with differing attributes DECL_USER_ALIGN
is usually not merged/copied when DECL_ALIGN is similar.  On s390 this
will cause a warning message not to be shown.  Similarly, we warned
about the wrong alignment when short-circuiting an alignment initialization in
common_handle_aligned_attribute ().

Fix this by copying DECL_USER_ALIGN even if DECL_ALIGN is similar as
well as getting rid of the short-circuited initialization.

gcc/c-family/ChangeLog:

* c-attribs.c (common_handle_aligned_attribute): Remove short
circuit and dead code.

gcc/c/ChangeLog:

* c-decl.c (merge_decls): Copy DECL_USER_ALIGN if DECL_ALIGN is
similar.

gcc/cp/ChangeLog:

* decl.c (duplicate_decls): Likewise.

3 years agors6000: Fix test case failures by PR100085 [PR101020]
Xionghu Luo [Tue, 15 Jun 2021 01:00:11 +0000 (20:00 -0500)]
rs6000: Fix test case failures by PR100085 [PR101020]

Tested pass, committing as obvious.

gcc/testsuite/ChangeLog:

2021-06-14  Xionghu Luo  <luoxhu@linux.ibm.com>

PR target/101020
* gcc.target/powerpc/float128-call.c: Adjust counts for be and le.

3 years agoDaily bump.
GCC Administrator [Tue, 15 Jun 2021 00:16:37 +0000 (00:16 +0000)]
Daily bump.

3 years agoTeach compute_objsize about placement new [PR100876].
Martin Sebor [Mon, 14 Jun 2021 22:34:48 +0000 (16:34 -0600)]
Teach compute_objsize about placement new [PR100876].

Resolves:
PR c++/100876 - -Wmismatched-new-delete should understand placement new when it's not inlined

gcc/ChangeLog:

PR c++/100876
* builtins.c (gimple_call_return_array): Check for attribute fn spec.
Handle calls to placement new.
(ndecl_dealloc_argno): Avoid placement delete.

gcc/testsuite/ChangeLog:

PR c++/100876
* g++.dg/warn/Wmismatched-new-delete-4.C: New test.
* g++.dg/warn/Wmismatched-new-delete-5.C: New test.
* g++.dg/warn/Wstringop-overflow-7.C: New test.
* g++.dg/warn/Wfree-nonheap-object-6.C: New test.
* g++.dg/analyzer/placement-new.C: Prune out expected warning.

3 years agolibstdc++: Fix noexcept-specifier for ranges::empty
Jonathan Wakely [Mon, 14 Jun 2021 21:42:05 +0000 (22:42 +0100)]
libstdc++: Fix noexcept-specifier for ranges::empty

Signed-off-by: Jonathan Wakely <jwakely@redhat.com>
libstdc++-v3/ChangeLog:

* include/bits/ranges_base.h (ranges::empty): Check whether
conversion to bool can throw.
* testsuite/std/ranges/access/empty.cc: Check for correct
noexcept-specifier.

3 years agors6000: MMA builtin usage ICEs when used in a #pragma omp parallel and using -fopenmp...
Peter Bergner [Mon, 14 Jun 2021 21:55:18 +0000 (16:55 -0500)]
rs6000: MMA builtin usage ICEs when used in a #pragma omp parallel and using -fopenmp [PR100777]

Using an MMA builtin within an openmp parallel code block, leads to an SSA
verification ICE on the temporaries we create while expanding the MMA builtins
at gimple time.  The solution is to use create_tmp_reg_or_ssa_name(), which
knows when to create either an SSA or register temporary.

2021-06-14  Peter Bergner  <bergner@linux.ibm.com>

gcc/
PR target/100777
* config/rs6000/rs6000-call.c (rs6000_gimple_fold_mma_builtin): Use
create_tmp_reg_or_ssa_name().

gcc/testsuite/
PR target/100777
* gcc.target/powerpc/pr100777.c: New test.

3 years agoLimit new value calculations to first order effects.
Andrew MacLeod [Mon, 14 Jun 2021 19:33:59 +0000 (15:33 -0400)]
Limit new value calculations to first order effects.

When utilzing poor values during propagation, we mostly care about values that
were undefined/processed directly used in calcualting the SSA_NAME being
processed.  2nd level derivations of such poor values rarely affect the
inital calculation.  Leave them to when they are directly encountered.

* gimple-range-cache.cc (ranger_cache::ranger_cache): Adjust.
(ranger_cache::enable_new_values): Set to specified value and
return the old value.
(ranger_cache::disable_new_values): Delete.
(ranger_cache::fill_block_cache): Disable non 1st order derived
poor values.
* gimple-range-cache.h (ranger_cache): Adjust prototypes.
* gimple-range.cc (gimple_ranger::range_of_expr): Adjust.

3 years agolibstdc++: Fix common_reference for non-reference results [PR100894]
Jonathan Wakely [Mon, 14 Jun 2021 19:31:00 +0000 (20:31 +0100)]
libstdc++: Fix common_reference for non-reference results [PR100894]

The result of COMMON-REF(A&, B&&) where they have no common reference
type should not be a reference. The implementation of COMMON-REF fails
to check that the result is a reference, so is well-formed when it
shouldn't be. This means that common_reference uses that result when it
shouldn't.

The fix is to reject the result of COMMON-REF(A, B) if it's not a
reference, so that common_reference falls through to the next case,
which uses COND-RES, which yields a non-reference result.

Signed-off-by: Jonathan Wakely <jwakely@redhat.com>
libstdc++-v3/ChangeLog:

PR libstdc++/100894
* include/std/type_traits (__common_ref_impl<X&, Y&>): Only
use the type if it's a reference.
* testsuite/20_util/common_reference/100894.cc: New test.

3 years agoi386: Split V2HImode *punpckwd to SSE instruction [PR101058]
Uros Bizjak [Mon, 14 Jun 2021 18:56:18 +0000 (20:56 +0200)]
i386: Split V2HImode *punpckwd to SSE instruction [PR101058]

V2HImode *punpckwd should not be split to the insn that depends on
TARGET_MMX_WITH_SSE, since the later is disabled on 32bit targets.

Also return true early from ix86_vectorize_vec_perm_const when testing
with V2HI mode.  *punpckwd can be used to implement all permutations.

2021-06-14  Uroš Bizjak  <ubizjak@gmail.com>

gcc/
PR target/101058
* config/i386/i386-expand.c (ix86_vectorize_vec_perm_const):
Return true early when testing with V2HImode.
* config/i386/mmx.md (*punpckwd): Split to sse2_pshuflw_1.

gcc/testsuite/

PR target/101058
* gcc.target/i386/pr101058.c: New test.

3 years agoc-family: Add fix-it suggestions for more <stdlib.h> names [PR101052]
Jonathan Wakely [Mon, 14 Jun 2021 10:38:11 +0000 (11:38 +0100)]
c-family: Add fix-it suggestions for more <stdlib.h> names [PR101052]

PR c++/101052

gcc/c-family/ChangeLog:

* known-headers.cc (get_stdlib_header_for_name): Add known
headers for EXIT_FAILURE, EXIT_SUCCESS, abort, atexit, calloc,
exit, and getenv.

gcc/testsuite/ChangeLog:

* g++.dg/spellcheck-stdlib.C: Add checks for <cstdlib> names.
* gcc.dg/spellcheck-stdlib.c: Likewise.

3 years agoarm: Auto-vectorization for MVE: add pack/unpack patterns
Christophe Lyon [Thu, 3 Jun 2021 14:35:50 +0000 (14:35 +0000)]
arm: Auto-vectorization for MVE: add pack/unpack patterns

This patch adds vec_unpack<US>_hi_<mode>, vec_unpack<US>_lo_<mode>,
vec_pack_trunc_<mode> patterns for MVE.

It does so by moving the unpack patterns from neon.md to
vec-common.md, while adding them support for MVE. The pack expander is
derived from the Neon one (which in turn is renamed into
neon_quad_vec_pack_trunc_<mode>).

The patch introduces mve_vec_unpack<US>_lo_<mode> and
mve_vec_unpack<US>_hi_<mode> which are similar to their Neon
counterparts, except for the assembly syntax.

The patch introduces mve_vec_pack_trunc_lo_<mode> to avoid the need for a
zero-initialized temporary, which is needed if the
vec_pack_trunc_<mode> expander calls @mve_vmovn[bt]q_<supf><mode>
instead.

With this patch, we can now vectorize the 16 and 8-bit versions of
vclz and vshl, although the generated code could still be improved.
For test_clz_s16, we now generate
        vldrh.16        q3, [r1]
        vmovlb.s16   q2, q3
        vmovlt.s16   q3, q3
        vclz.i32  q2, q2
        vclz.i32  q3, q3
        vmovnb.i32      q1, q2
        vmovnt.i32      q1, q3
        vstrh.16        q1, [r0]
which could be improved to
        vldrh.16        q3, [r1]
vclz.i16 q1, q3
        vstrh.16        q1, [r0]
if we could avoid the need for unpack/pack steps.

For reference, clang-12 generates:
vldrh.s32       q0, [r1]
vldrh.s32       q1, [r1, #8]
vclz.i32        q0, q0
vstrh.32        q0, [r0]
vclz.i32        q0, q1
vstrh.32        q0, [r0, #8]

2021-06-11  Christophe Lyon  <christophe.lyon@linaro.org>

gcc/
* config/arm/mve.md (mve_vec_unpack<US>_lo_<mode>): New pattern.
(mve_vec_unpack<US>_hi_<mode>): New pattern.
(@mve_vec_pack_trunc_lo_<mode>): New pattern.
(mve_vmovntq_<supf><mode>): Prefix with '@'.
* config/arm/neon.md (vec_unpack<US>_hi_<mode>): Move to
vec-common.md.
(vec_unpack<US>_lo_<mode>): Likewise.
(vec_pack_trunc_<mode>): Rename to
neon_quad_vec_pack_trunc_<mode>.
* config/arm/vec-common.md (vec_unpack<US>_hi_<mode>): New
pattern.
(vec_unpack<US>_lo_<mode>): New.
(vec_pack_trunc_<mode>): New.

gcc/testsuite/
* gcc.target/arm/simd/mve-vclz.c: Update expected results.
* gcc.target/arm/simd/mve-vshl.c: Likewise.
* gcc.target/arm/simd/mve-vec-pack.c: New test.
* gcc.target/arm/simd/mve-vec-unpack.c: New test.

3 years agoC/C++: Fix unused set var warning with omp_clause_affinity [PR100913]
Tobias Burnus [Mon, 14 Jun 2021 14:49:24 +0000 (16:49 +0200)]
C/C++: Fix unused set var warning with omp_clause_affinity [PR100913]

PR c/100913
gcc/c/ChangeLog:

* c-parser.c (c_parser_omp_clause_affinity): No need to set iterator
var in the error case.

gcc/cp/ChangeLog:

* parser.c (cp_parser_omp_clause_affinity): No need to set iterator
var in the error case.

3 years agotree-optimization/100934 - properly mark irreducible regions for DOM
Richard Biener [Mon, 14 Jun 2021 12:57:26 +0000 (14:57 +0200)]
tree-optimization/100934 - properly mark irreducible regions for DOM

The jump threading code requires marked irreducible regions for the
purpose of validating jump threading paths but DOM fails to provide
that resulting in mised number of iteration upper bounds clearing.

2021-06-14  Richard Biener  <rguenther@suse.de>

PR tree-optimization/100934
* tree-ssa-dom.c (pass_dominator::execute): Properly
mark irreducible regions.

* gcc.dg/torture/pr100934.c: New testcase.

3 years agolibstdc++: Only run Filesystem TS test if supported
Jonathan Wakely [Mon, 14 Jun 2021 13:22:07 +0000 (14:22 +0100)]
libstdc++: Only run Filesystem TS test if supported

libstdc++-v3/ChangeLog:

* testsuite/experimental/filesystem/path/native/conv_c++23.cc:
Add dg-require-filesystem-ts directive.

3 years agolibstdc++: Fix std::any constraints [PR101034]
Jonathan Wakely [Mon, 14 Jun 2021 13:18:33 +0000 (14:18 +0100)]
libstdc++: Fix std::any constraints [PR101034]

PR libstdc++/101034

libstdc++-v3/ChangeLog:

* include/std/any (any(in_place_t<T>, initializer_list<U>, A&&...))
(any::emplace<T>(initializer_list<U>, A&&...)): Fix constraint
to use lvalue.
* testsuite/20_util/any/cons/101034.cc: New test.

3 years agodocs: Fix -Wno-cpp note
Martin Liska [Mon, 14 Jun 2021 14:07:10 +0000 (16:07 +0200)]
docs: Fix -Wno-cpp note

gcc/ChangeLog:

* doc/invoke.texi: Put r{...} on the same line as @item.

3 years agodocs: add missing newline
Martin Liska [Mon, 14 Jun 2021 14:04:13 +0000 (16:04 +0200)]
docs: add missing newline

gcc/ChangeLog:

* doc/invoke.texi: Add missing newline.

3 years agodocs: remove extra '+' character in option listing.
Martin Liska [Mon, 14 Jun 2021 13:36:38 +0000 (15:36 +0200)]
docs: remove extra '+' character in option listing.

gcc/ChangeLog:

* doc/invoke.texi: Remove '+' charasters.

3 years agolibstdc++: Add explicit -std=gnu++17 option to test
Jonathan Wakely [Mon, 14 Jun 2021 12:17:40 +0000 (13:17 +0100)]
libstdc++: Add explicit -std=gnu++17 option to test

This test has no -std option so when the testsuite is run with
-std=gnu++20 or later, this test will use that. The recent addition of
no_unique_address will cause it to FAIL, because that's a reserved word
after C++17. Add an explicit option, so that this test alays uses
exactly C++17.

libstdc++-v3/ChangeLog:

* testsuite/17_intro/headers/c++2017/all_attributes.cc: Add
-std=gnu++17 option.

3 years agolibstdc++: Implement LWG 3465 for std::compare_partial_order_fallback [PR101056]
Jonathan Wakely [Mon, 14 Jun 2021 11:30:52 +0000 (12:30 +0100)]
libstdc++: Implement LWG 3465 for std::compare_partial_order_fallback [PR101056]

libstdc++-v3/ChangeLog:

PR libstdc++/101056
* libsupc++/compare (compare_partial_order_fallback): Add
constraint using reversed parameter order, as per LWG 3465.
* testsuite/18_support/comparisons/algorithms/fallback.cc:
Adjust expected result.

3 years agolibstdc++: Change [cmp.alg] assertions to constraints
Jonathan Wakely [Mon, 14 Jun 2021 11:25:43 +0000 (12:25 +0100)]
libstdc++: Change [cmp.alg] assertions to constraints

This moves the same_as<decay_t<_Tp>, decay_t<_Up>> checks from the
[cmp.alg] function bodies into their constraints.

Also add a test for the compare_xxx_order_fallback algorithms.

libstdc++-v3/ChangeLog:

* libsupc++/compare (__decayed_same_as): New helper concept.
(strong_order, weak_order, partial_order): Constrain with new
concept instead of using static_assert.
(compare_strong_order_fallback, compare_weak_order_fallback)
(compare_partial_order_fallback): Likewise. Do not deduce return
types. Remove redundant if-constexpr checks.
* testsuite/18_support/comparisons/algorithms/fallback.cc: New test.

3 years agoFortran: resolve.c - remove '*XCNEW' based nullifying
Tobias Burnus [Mon, 14 Jun 2021 12:36:20 +0000 (14:36 +0200)]
Fortran: resolve.c - remove '*XCNEW' based nullifying

gcc/fortran/ChangeLog:

* resolve.c (resolve_variable): Remove *XCNEW used to
nullify nullified memory.

3 years agoarc: Add --with-fpu support for ARCv2 cpus
Claudiu Zissulescu [Mon, 14 Jun 2021 12:33:17 +0000 (15:33 +0300)]
arc: Add --with-fpu support for ARCv2 cpus

Support for a compile-time default FPU. The --with-fpu configuration
option is ignored if -mfpu compiler option is specified. The FPU
options are only available for ARCv2 cpus.

gcc/
2021-06-14  Claudiu Zissulescu  <claziss@synopsys.com>

* config.gcc (arc): Add support for with_cpu option.
* config/arc/arc.h (OPTION_DEFAULT_SPECS): Add fpu.

Signed-off-by: Claudiu Zissulescu <claziss@synopsys.com>
3 years agoDo not check if SMS succeeds on powerpc
Aaron Sawdey [Fri, 11 Jun 2021 18:49:18 +0000 (13:49 -0500)]
Do not check if SMS succeeds on powerpc

These tests have become unstable and SMS either succeeds or doesn't
depending on things like changes in instruction latency. Removing
the scan-rtl-dump-times checks for powerpc*-*-*.

gcc/testsuite

* gcc.dg/sms-1.c: Remove scan-rtl-dump-times check.
* gcc.dg/sms-2.c: Remove scan-rtl-dump-times check.
* gcc.dg/sms-3.c: Remove scan-rtl-dump-times check.
* gcc.dg/sms-4.c: Remove scan-rtl-dump-times check.
* gcc.dg/sms-6.c: Remove scan-rtl-dump-times check.
* gcc.dg/sms-8.c: Remove scan-rtl-dump-times check.
* gcc.dg/sms-10.c: Remove scan-rtl-dump-times check.

3 years agolibstdc++: Use reserved name for attribute [PR101055]
Jonathan Wakely [Mon, 14 Jun 2021 09:58:15 +0000 (10:58 +0100)]
libstdc++: Use reserved name for attribute [PR101055]

The no_unique_address attribute is not a reserved name until C++20, so
to use it in C++11/14/17 modes we should use the __no_unique_address_
form. We already use that form when using the attribute, but not in the
__has_cpp_attribute check.

libstdc++-v3/ChangeLog:

PR libstdc++/101055
* include/std/tuple: Use reserved form of attribute name.
* testsuite/17_intro/headers/c++2011/all_attributes.cc: Add
check for no_unique_address.
* testsuite/17_intro/headers/c++2014/all_attributes.cc:
Likewise.
* testsuite/17_intro/headers/c++2017/all_attributes.cc:
Likewise.

3 years agotree-optimization/101031 - fix strlen opt invalidation logic
Richard Biener [Mon, 14 Jun 2021 07:37:24 +0000 (09:37 +0200)]
tree-optimization/101031 - fix strlen opt invalidation logic

strlen opt uses ao_ref_init_from_ptr_and_size to prepare alias
queries to invalidate its knowledge about strings.  It constrains
the size using the number of known-nonzero chars and adds one
for a terminating nul - without knowing whether such nul exists
or even fits the object.  The latter is now a problem since the
oracle disambiguates an access of size two (as built so) against
a store to a plain char variable (where a terminating nul does not
fit).  The fix is to instead increment max_size but leave size to
the number of chars we know are accessed.

2021-06-14  Richard Biener  <rguenther@suse.de>

PR tree-optimization/101031
* tree-ssa-strlen.c (maybe_invalidate): Increment max_size
instead of size when accounting for a possibly string
terminating nul.

* gcc.dg/torture/pr101031.c: New testcase.

3 years agoevrp: fix AddressSanitizer: alloc-dealloc-mismatch
Martin Liska [Mon, 14 Jun 2021 08:22:36 +0000 (10:22 +0200)]
evrp: fix AddressSanitizer: alloc-dealloc-mismatch

The fixed error is:

==21166==ERROR: AddressSanitizer: alloc-dealloc-mismatch (operator new [] vs operator delete) on 0x60300000d900
    #0 0x7367d7 in operator delete(void*, unsigned long) /home/marxin/BIG/buildbot/buildworker/marxinbox-gcc-asan/build/libsanitizer/asan/asan_new_delete.cpp:172
    #1 0x3b82e6e in pointer_equiv_analyzer::~pointer_equiv_analyzer() /home/marxin/BIG/buildbot/buildworker/marxinbox-gcc-asan/build/gcc/gimple-ssa-evrp.c:161
    #2 0x3b83387 in hybrid_folder::~hybrid_folder() /home/marxin/BIG/buildbot/buildworker/marxinbox-gcc-asan/build/gcc/gimple-ssa-evrp.c:517
    #3 0x3b83387 in execute_early_vrp /home/marxin/BIG/buildbot/buildworker/marxinbox-gcc-asan/build/gcc/gimple-ssa-evrp.c:686
    #4 0x1790611 in execute_one_pass(opt_pass*) /home/marxin/BIG/buildbot/buildworker/marxinbox-gcc-asan/build/gcc/passes.c:2567
    #5 0x1792003 in execute_pass_list_1 /home/marxin/BIG/buildbot/buildworker/marxinbox-gcc-asan/build/gcc/passes.c:2656
    #6 0x1792029 in execute_pass_list_1 /home/marxin/BIG/buildbot/buildworker/marxinbox-gcc-asan/build/gcc/passes.c:2657
    #7 0x179209f in execute_pass_list(function*, opt_pass*) /home/marxin/BIG/buildbot/buildworker/marxinbox-gcc-asan/build/gcc/passes.c:2667
    #8 0x178a5f3 in do_per_function_toporder(void (*)(function*, void*), void*) /home/marxin/BIG/buildbot/buildworker/marxinbox-gcc-asan/build/gcc/passes.c:1773
    #9 0x1792fac in do_per_function_toporder(void (*)(function*, void*), void*) /home/marxin/BIG/buildbot/buildworker/marxinbox-gcc-asan/build/gcc/plugin.h:191
    #10 0x1792fac in execute_ipa_pass_list(opt_pass*) /home/marxin/BIG/buildbot/buildworker/marxinbox-gcc-asan/build/gcc/passes.c:3001
    #11 0xc525fc in ipa_passes /home/marxin/BIG/buildbot/buildworker/marxinbox-gcc-asan/build/gcc/cgraphunit.c:2154
    #12 0xc525fc in symbol_table::compile() /home/marxin/BIG/buildbot/buildworker/marxinbox-gcc-asan/build/gcc/cgraphunit.c:2289
    #13 0xc5a096 in symbol_table::compile() /home/marxin/BIG/buildbot/buildworker/marxinbox-gcc-asan/build/gcc/cgraphunit.c:2269
    #14 0xc5a096 in symbol_table::finalize_compilation_unit() /home/marxin/BIG/buildbot/buildworker/marxinbox-gcc-asan/build/gcc/cgraphunit.c:2537
    #15 0x1a7a17c in compile_file /home/marxin/BIG/buildbot/buildworker/marxinbox-gcc-asan/build/gcc/toplev.c:482
    #16 0x69c758 in do_compile /home/marxin/BIG/buildbot/buildworker/marxinbox-gcc-asan/build/gcc/toplev.c:2210
    #17 0x69c758 in toplev::main(int, char**) /home/marxin/BIG/buildbot/buildworker/marxinbox-gcc-asan/build/gcc/toplev.c:2349
    #18 0x6a932a in main /home/marxin/BIG/buildbot/buildworker/marxinbox-gcc-asan/build/gcc/main.c:39
    #19 0x7ffff7820b34 in __libc_start_main ../csu/libc-start.c:332
    #20 0x6aa5fd in _start (/home/marxin/BIG/buildbot/buildworker/marxinbox-gcc-asan/objdir/gcc/cc1+0x6aa5fd)

0x60300000d900 is located 0 bytes inside of 32-byte region [0x60300000d900,0x60300000d920)
allocated by thread T0 here:
    #0 0x735ab7 in operator new[](unsigned long) /home/marxin/BIG/buildbot/buildworker/marxinbox-gcc-asan/build/libsanitizer/asan/asan_new_delete.cpp:102
    #1 0x3b82dac in pointer_equiv_analyzer::pointer_equiv_analyzer(gimple_ranger*) /home/marxin/BIG/buildbot/buildworker/marxinbox-gcc-asan/build/gcc/gimple-ssa-evrp.c:156

gcc/ChangeLog:

* gimple-ssa-evrp.c (pointer_equiv_analyzer::~pointer_equiv_analyzer): Use delete[].

3 years agoPick up global ranges in ranger after inlining.
Aldy Hernandez [Sun, 13 Jun 2021 14:20:33 +0000 (16:20 +0200)]
Pick up global ranges in ranger after inlining.

Ranger was not picking up global ranges because doing so could remove
__builtin_unreachable calls too early to the detriment of LTO.  However,
we can safely remove these calls after inlining.  This patch removes the
restriction and allows ranger to pick up global ranges under these
circumstances.

Tested on x86-64 Linux.

gcc/ChangeLog:

* value-query.cc (gimple_range_global): Call get_range_global
if called after inlining.

3 years ago[PATCH] config: Backport "Rely less on internal symbols" (serial 68) to gettext.m4
Michael Forney [Mon, 14 Jun 2021 02:30:09 +0000 (22:30 -0400)]
[PATCH] config: Backport "Rely less on internal symbols" (serial 68) to gettext.m4

intl
* configure: Regenerated.

3 years ago[PATCH] config: Backport "Rely less on internal symbols" (serial 69) to gettext.m4
Michael Forney [Mon, 14 Jun 2021 02:24:06 +0000 (22:24 -0400)]
[PATCH] config: Backport "Rely less on internal symbols" (serial 69) to gettext.m4

config

* gettext.m4 (AM_GNU_GETTEXT): Skip checks for the internal
symbols _nl_msg_cat_cntr, _nl_domain_bindings, and
_nl_expand_alias, if __GNU_GETTEXT_SUPPORTED_REVISION is defined.
Backport of gettext serial 68 patch.

3 years agoDaily bump.
GCC Administrator [Mon, 14 Jun 2021 00:16:35 +0000 (00:16 +0000)]
Daily bump.

3 years agoi386: Improve variable permutation insn avoidance [PR101021]
Uros Bizjak [Sun, 13 Jun 2021 19:50:51 +0000 (21:50 +0200)]
i386: Improve variable permutation insn avoidance [PR101021]

Emit constant permutation insn directly from expand_vec_perm_shufb.

2021-06-13  Uroš Bizjak  <ubizjak@gmail.com>

gcc/
PR target/101021
* config/i386/i386-expand.c (expand_vec_perm_pshufb):
Emit constant permutation insn directly from here.

3 years agoc: adjust [[maybe_unused]] testcase
Jason Merrill [Sun, 13 Jun 2021 15:34:38 +0000 (11:34 -0400)]
c: adjust [[maybe_unused]] testcase

Another testcase update needed for my r12-1405 commit.

gcc/testsuite/ChangeLog:

* gcc.dg/c2x-attr-maybe_unused-1.c: Expect no warnings.

3 years agouse range based for loops to iterate over vec<>
Trevor Saunders [Sat, 12 Jun 2021 03:49:22 +0000 (23:49 -0400)]
use range based for loops to iterate over vec<>

This changes users of FOR_EACH_VEC_ELT to use range based for loops,
where the index variables are otherwise unused.  As such the index
variables are all deleted, producing shorter and simpler code.

Signed-off-by: Trevor Saunders <tbsaunde@tbsaunde.org>
gcc/analyzer/ChangeLog:

* call-string.cc (call_string::call_string): Use range based for
to iterate over vec<>.
(call_string::to_json): Likewise.
(call_string::hash): Likewise.
(call_string::calc_recursion_depth): Likewise.
* checker-path.cc (checker_path::fixup_locations): Likewise.
* constraint-manager.cc (equiv_class::equiv_class): Likewise.
(equiv_class::to_json): Likewise.
(equiv_class::hash): Likewise.
(constraint_manager::to_json): Likewise.
* engine.cc (impl_region_model_context::on_svalue_leak):
Likewise.
(on_liveness_change): Likewise.
(impl_region_model_context::on_unknown_change): Likewise.
* program-state.cc (sm_state_map::set_state): Likewise.
* region-model.cc (test_canonicalization_4): Likewise.

gcc/ChangeLog:

* attribs.c (find_attribute_namespace): Iterate over vec<> with
range based for.
* auto-profile.c (afdo_find_equiv_class): Likewise.
* gcc.c (do_specs_vec): Likewise.
(do_spec_1): Likewise.
(driver::set_up_specs): Likewise.
* gimple-loop-jam.c (any_access_function_variant_p): Likewise.
* gimple-ssa-store-merging.c (compatible_load_p): Likewise.
(imm_store_chain_info::try_coalesce_bswap): Likewise.
(imm_store_chain_info::coalesce_immediate_stores): Likewise.
(get_location_for_stmts): Likewise.
* graphite-poly.c (print_iteration_domains): Likewise.
(free_poly_bb): Likewise.
(remove_gbbs_in_scop): Likewise.
(free_scop): Likewise.
(dump_gbb_cases): Likewise.
(dump_gbb_conditions): Likewise.
(print_pdrs): Likewise.
(print_scop): Likewise.
* ifcvt.c (cond_move_process_if_block): Likewise.
* lower-subreg.c (decompose_multiword_subregs): Likewise.
* regcprop.c (pass_cprop_hardreg::execute): Likewise.
* sanopt.c (sanitize_rewrite_addressable_params): Likewise.
* sel-sched-dump.c (dump_insn_vector): Likewise.
* store-motion.c (store_ops_ok): Likewise.
(store_killed_in_insn): Likewise.
* timevar.c (timer::named_items::print): Likewise.
* tree-cfgcleanup.c (cleanup_control_flow_pre): Likewise.
(cleanup_tree_cfg_noloop): Likewise.
* tree-data-ref.c (dump_data_references): Likewise.
(print_dir_vectors): Likewise.
(print_dist_vectors): Likewise.
(dump_data_dependence_relations): Likewise.
(dump_dist_dir_vectors): Likewise.
(dump_ddrs): Likewise.
(create_runtime_alias_checks): Likewise.
(free_subscripts): Likewise.
(save_dist_v): Likewise.
(save_dir_v): Likewise.
(invariant_access_functions): Likewise.
(same_access_functions): Likewise.
(access_functions_are_affine_or_constant_p): Likewise.
(find_data_references_in_stmt): Likewise.
(graphite_find_data_references_in_stmt): Likewise.
(free_dependence_relations): Likewise.
(free_data_refs): Likewise.
* tree-inline.c (copy_debug_stmts): Likewise.
* tree-into-ssa.c (dump_currdefs): Likewise.
(rewrite_update_phi_arguments): Likewise.
* tree-ssa-propagate.c (clean_up_loop_closed_phi): Likewise.
* tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr):
Likewise.
(vect_slp_analyze_node_dependences): Likewise.
(vect_slp_analyze_instance_dependence): Likewise.
(vect_record_base_alignments): Likewise.
(vect_get_peeling_costs_all_drs): Likewise.
(vect_peeling_supportable): Likewise.
* tree-vectorizer.c (vec_info::~vec_info): Likewise.
(vec_info::free_stmt_vec_infos): Likewise.

gcc/cp/ChangeLog:

* constexpr.c (cxx_eval_call_expression): Iterate over vec<>
with range based for.
(cxx_eval_store_expression): Likewise.
(cxx_eval_loop_expr): Likewise.
* decl.c (wrapup_namespace_globals): Likewise.
(cp_finish_decl): Likewise.
(cxx_simulate_enum_decl): Likewise.
* parser.c (cp_parser_postfix_expression): Likewise.

3 years ago[committed] More improvements to H8 logicals for test/compare elimination
Jeff Law [Sun, 13 Jun 2021 15:09:38 +0000 (11:09 -0400)]
[committed] More improvements to H8 logicals for test/compare elimination

gcc/
* config/h8300/logical.md (<code>qi3_1<cczn>): New pattern.
(andqi3_1<cczn>): Removed.
(<ors>qi3_1): Do not split for IOR/XOR a single bit.
(H8/SX bit logicals): Split out from other patterns.
* config/h8300/multiply.md (mulqihi3_const<cczn>): Renamed from
mulqihi3_const_clobber_flags.
(mulqihi3<cczn>, mulhisi3_const<cczn>, mulhisi3<cczn>): Similarly

3 years agox86: Replace ix86_red_zone_size with ix86_red_zone_used
H.J. Lu [Fri, 11 Jun 2021 14:31:29 +0000 (07:31 -0700)]
x86: Replace ix86_red_zone_size with ix86_red_zone_used

Add red_zone_used to machine_function to track if red zone is used.
When expanding function prologue, set red_zone_used to true if red
zone is used.

gcc/

PR target/101023
* config/i386/i386.c (ix86_expand_prologue): Set red_zone_used
to true if red zone is used.
(ix86_output_indirect_jmp): Replace ix86_red_zone_size with
ix86_red_zone_used.
* config/i386/i386.h (machine_function): Add red_zone_used.
(ix86_red_zone_size): Removed.
(ix86_red_zone_used): New.
* config/i386/i386.md (peephole2 patterns): Replace
ix86_red_zone_size with ix86_red_zone_used.

gcc/testsuite/

PR target/101023
* g++.target/i386/pr101023a.C: New test.
* g++.target/i386/pr101023b.C: Likewise.

3 years agocontrib/gcc-changelog: Check that PR in subject is in changelog
Tobias Burnus [Sun, 13 Jun 2021 05:46:54 +0000 (07:46 +0200)]
contrib/gcc-changelog: Check that PR in subject is in changelog

This patch checks that a '[PRnnnn]' and '(PRnnnn)' also appears as PR in the
changelog part of the commit message.  And it does likewise for 'PR comp/nnnn'
except that then also the component name is checked.  (Note that the reverse
is permitted, i.e. PR(s) only appearing in the changelog.)
To avoid false positives, PR numbers in the subject line are ignored,
if 'revert' appears.
Additionally, reject commits with a nonempty second line.

contrib/ChangeLog:

* gcc-changelog/git_commit.py (pr_regex): Add ?P<pr> for group('pr').
(subject_pr_regex, subject_pr2_regex): New.
(GitInfo.__init__, GitCommit.parse_changelog): Check subject PRs.
* gcc-changelog/git_email.py (SUBJECT_PREFIX, subject_patch_regex): New.
(GitEmail.__init__): Parse 'Subject:' and pass it to GitInfo.
* gcc-changelog/test_email.py (test_pr_only_in_subject,
test_wrong_pr_comp_in_subject, test_copyright_years): New.
* gcc-changelog/test_patches.txt (0030-PR-c-92746, pr-check1.patch):
Update to avoid triggering the new check.
(0001-rs6000-Support-doubleword, pr-wrong-comp.patch,
copyright-years.patch): New.

3 years agoDaily bump.
GCC Administrator [Sun, 13 Jun 2021 00:16:35 +0000 (00:16 +0000)]
Daily bump.

3 years agotestsuite: fix AIX testsuite failures
David Edelsohn [Fri, 11 Jun 2021 00:12:08 +0000 (20:12 -0400)]
testsuite: fix AIX testsuite failures

* g++.dg/ext/builtin-shufflevector-2.C: Ignore psabi warning.
* gcc.dg/uninit-pr93100.c: Skip on AIX.
* gcc.target/powerpc/pr100085.c: Require int128 and float128.

3 years agoc-family: don't warn for [[maybe_unused]] on data member
Jason Merrill [Fri, 11 Jun 2021 20:10:50 +0000 (16:10 -0400)]
c-family: don't warn for [[maybe_unused]] on data member

The C++17 standard (and C2x) says that [[maybe_unused]] may be applied to a
non-static data member, so we shouldn't warn about it.  And I don't see a
reason not to handle a FIELD_DECL the same as any other decl, by setting
TREE_USED on it.  It doesn't look like anything yet cares about that flag on
a FIELD_DECL, but setting it shouldn't hurt.

gcc/c-family/ChangeLog:

* c-attribs.c (handle_unused_attribute): Handle FIELD_DECL.

gcc/testsuite/ChangeLog:

* g++.dg/ext/attrib62.C: No longer warn.
* g++.dg/diagnostic/maybe_unused1.C: New test.

gcc/ChangeLog:

* doc/extend.texi (unused variable attribute): Applies to
structure fields as well.

3 years agoc++: constexpr and array[0] [PR101029]
Jason Merrill [Fri, 11 Jun 2021 20:55:30 +0000 (16:55 -0400)]
c++: constexpr and array[0] [PR101029]

build_vec_init_elt exits early if we're initializing a zero-element array,
so build_vec_init needs to do the same to avoid trying to instantiate things
after we've already started throwing important bits away.

PR c++/101029

gcc/cp/ChangeLog:

* init.c (build_vec_init): Shortcut [0] case.

gcc/testsuite/ChangeLog:

* g++.dg/ext/array4.C: New test.

3 years agoc++: speed up looking up the current class
Jason Merrill [Wed, 9 Jun 2021 21:48:14 +0000 (17:48 -0400)]
c++: speed up looking up the current class

While looking at template instantiation tracing, I noticed that we were
frequently looking up a particular class template instance while
instantiating it.  This patch shortcuts that lookup, and speeds up compiling
stdc++.h with my (checking/unoptimized) compiler by about 3%.

gcc/cp/ChangeLog:

* pt.c (lookup_template_class_1): Shortcut current_class_type.

3 years agoDaily bump.
GCC Administrator [Sat, 12 Jun 2021 00:16:27 +0000 (00:16 +0000)]
Daily bump.

3 years agoFix a typo in an AutoFDO error string
Eugene Rozenfeld [Fri, 11 Jun 2021 23:28:55 +0000 (16:28 -0700)]
Fix a typo in an AutoFDO error string

gcc/ChangeLog:

* auto-profile.c (read_profile): fix a typo in an error string

3 years agoFix effective target for check-builtin-vec_rlnm-runnable.c test
Carl Love [Fri, 11 Jun 2021 19:49:27 +0000 (14:49 -0500)]
Fix effective target for check-builtin-vec_rlnm-runnable.c test

The effective target for a Power 9 runnable test should be
p9vector_hw.

2021-06-11  Carl Love  <cel@us.ibm.com>

gcc/testsuite/ChangeLog

* gcc.target/powerpc/check-builtin-vec_rlnm-runnable.c
(dg-require-effective-target): Change target to p9vector_hw.

3 years agoFor 'OMP_CLAUSE' in 'dump_generic_node', dump the whole OMP clause chain
Thomas Schwinge [Fri, 11 Jun 2021 13:37:33 +0000 (15:37 +0200)]
For 'OMP_CLAUSE' in 'dump_generic_node', dump the whole OMP clause chain

... instead of just the first clause.

gcc/
* tree-pretty-print.h (dump_omp_clauses): Add 'bool = true'
default argument.
* tree-pretty-print.c (dump_omp_clauses): Update.
(dump_generic_node) <OMP_CLAUSE>: Use it.
gcc/testsuite/
* gcc.dg/gomp/simd-clones-2.c: Enhance.

3 years agoc++: Substitute into function parms in lexical order [PR96560]
Patrick Palka [Fri, 11 Jun 2021 20:00:52 +0000 (16:00 -0400)]
c++: Substitute into function parms in lexical order [PR96560]

This makes tsubst_arg_types substitute into a function's parameter types
in left-to-right instead of right-to-left order, in accordance with DR 1227.

DR 1227
PR c++/96560

gcc/cp/ChangeLog:

* pt.c (tsubst_arg_types): Rearrange so that we substitute into
TYPE_ARG_TYPES in forward order while short circuiting
appropriately.  Adjust formatting.

gcc/testsuite/ChangeLog:

* g++.dg/template/sfinae-dr1227.C: New test.

3 years agolibstdc++: Fix filesystem::path comparisons for C++23
Jonathan Wakely [Fri, 11 Jun 2021 18:18:11 +0000 (19:18 +0100)]
libstdc++: Fix filesystem::path comparisons for C++23

In C++23 there is a basic_string_view(Range&&) constructor, which is
constrained to take a range (specifically, a contiguous_range). When the
filesystem::path comparison operators call lhs.compare(rhs) the overload
taking a string_view is considered, which means checking whether path
satisfies the range concept. That satisfaction result changes depending
whether path::iterator is complete, which is ill-formed; no diagnostic
required. To avoid the problem, this change ensures that the overload
resolution is performed in a context where path::iterator is complete
and the range concept is satisfied. (The result of overload resolution
is always that the compare(const path&) overload is the best match, but
we still have to consider the compare(basic_string_view<value_type>) one
to decide if it even participates in overload resolution).

For std::filesystem::path we can't define the comparison operators later
in the file, because they are hidden friends, so a new helper is
introduced that gets defined when everything else is complete.

For std::experimental::filesystem::path we can just move the definitions
of the comparison operators later in the file.

Signed-off-by: Jonathan Wakely <jwakely@redhat.com>
libstdc++-v3/ChangeLog:

* include/bits/fs_path.h (operator==, operator<=>): Use new
_S_compare function.
(path::_S_compare): New function to call path::compare in a
context where path::iterator is complete.
* include/experimental/bits/fs_path.h (operator<, operator==):
Define after path::iterator is complete.
* testsuite/27_io/filesystem/path/native/conv_c++23.cc: New
test.
* testsuite/experimental/filesystem/path/native/conv_c++23.cc:
New test.

3 years agod: foreach over a tuple doesn't work on 16-bit targets (PR100999)
Iain Buclaw [Fri, 11 Jun 2021 17:33:07 +0000 (19:33 +0200)]
d: foreach over a tuple doesn't work on 16-bit targets (PR100999)

Improves semantic passes in the front-end around the `foreach' and
`static foreach' statements to be more resilient to compiling in a
minimal D runtime environment.  Checking of the index type has been
improved as well so now there won't be needless compiler errors when
using 8 or 16-bit integers as index types when the size fits the
expected loop range.

gcc/d/ChangeLog:

PR d/100999
* dmd/MERGE: Merge upstream dmd 7a3808254.

libphobos/ChangeLog:

PR d/100999
* src/MERGE: Merge upstream phobos 55bb17543.

3 years agoarm: Fix polymorphic variants failing with undefined reference to `__ARM_undef` error.
Srinath Parvathaneni [Fri, 11 Jun 2021 17:11:58 +0000 (18:11 +0100)]
arm: Fix polymorphic variants failing with undefined reference to `__ARM_undef` error.

This patch fixes the issue mentioned in PR101016, which is mve polymorphic variants
failing at linking with undefined reference to "__ARM_undef" error.

gcc/ChangeLog:

2021-06-11  Srinath Parvathaneni  <srinath.parvathaneni@arm.com>

PR target/101016
* config/arm/arm_mve.h (__arm_vld1q): Change __ARM_mve_coerce(p0,
int8_t const *) to __ARM_mve_coerce1(p0, int8_t *) in the argument for
the polymorphic variants matching code.
(__arm_vld1q_z): Likewise.
(__arm_vld2q): Likewise.
(__arm_vld4q): Likewise.
(__arm_vldrbq_gather_offset): Likewise.
(__arm_vldrbq_gather_offset_z): Likewise.

gcc/testsuite/ChangeLog:

2021-06-11  Srinath Parvathaneni  <srinath.parvathaneni@arm.com>

PR target/101016
* gcc.target/arm/mve/intrinsics/pr101016.c: New test.

3 years ago[PATCH] PR tree-optimization/96392 Optimize x+0.0 if x is an integer
Roger Sayle [Fri, 11 Jun 2021 16:42:14 +0000 (17:42 +0100)]
[PATCH] PR tree-optimization/96392 Optimize x+0.0 if x is an integer

Doh! Wrong patch version.  Sorry for the inconvenience.

2020-06-11  Roger Sayle  <roger@nextmovesoftware.com>

gcc/ChangeLog
PR tree-optimization/96392
* fold-const.h (tree_expr_maybe_real_minus_zero_p): Fix prototype.

3 years ago[PATCH] PR tree-optimization/96392 Optimize x+0.0 if x is an integer
Roger Sayle [Fri, 11 Jun 2021 16:15:38 +0000 (17:15 +0100)]
[PATCH] PR tree-optimization/96392 Optimize x+0.0 if x is an integer

The patch implements a missed optimization enhancement.  Under usual
IEEE rules, x+0.0 can't be simplified to x when x might potentially
be an IEEE minus zero (-0.0).  The current logic in the middle-end
checks whether the type of x should honor signed zeros, but with this
patch we introduce tree_expr_maybe_real_minus_zero_p that allows us
to confirm that the value can't possibly be -0.0, for example, the result
of a conversion from an integer type, or the result of fabs (or has a
type that doesn't honor signed zero).

Whilst modifying match.pd, I also converted some additional folding
transformations from "testing the type" to "testing the value".

2020-06-10  Roger Sayle  <roger@nextmovesoftware.com>

gcc/ChangeLog
PR tree-optimization/96392
* fold-const.c (fold_real_zero_addition_p): Take both arguments
of the addition or subtraction, not just the zero.  Use this
other argument in tests for signaling NaNs and signed zeros.
(tree_expr_maybe_real_minus_zero_p): New predicate.
* fold-const.h (fold_real_zero_addition_p): Update prototype.
(tree_expr_maybe_real_minus_zero_p): New function prototype.
* match.pd: Update calls to fold_real_zero_addition_p.
Replace HONOR_NANS with tree_expr_maybe_nan_p.
Replace HONOR_SIGNED_ZEROS with tree_expr_maybe_real_minus_zero_p.
Replace HONOR_SNANS with tree_expr_maybe_signaling_nan_p.
* tree-ssa-reassoc.c (eliminate_using_constants): Update
call to fold_real_zero_addition_p.

gcc/testsuite/ChangeLog
PR tree-optimization/96392
* gcc.dg/pr96392.c: New test.

3 years agotree-optimization/101025 - fix store-motion dependence checking
Richard Biener [Fri, 11 Jun 2021 10:06:08 +0000 (12:06 +0200)]
tree-optimization/101025 - fix store-motion dependence checking

This plugs a hole in store-motion where it fails to perform dependence
checking on conditionally executed but not store-motioned refs.

2021-06-11  Richard Biener  <rguenther@suse.de>

PR tree-optimization/101025
* tree-ssa-loop-im.c (sm_seq_valid_bb): Make sure to process
all refs that require dependence checking.

* gcc.dg/torture/pr101025.c: New testcase.

3 years agoarm: Fix the mve multilib for the broken cmse support (pr99939).
Srinath Parvathaneni [Fri, 11 Jun 2021 14:56:37 +0000 (15:56 +0100)]
arm: Fix the mve multilib for the broken cmse support (pr99939).

The current CMSE support in the multilib build for
"-march=armv8.1-m.main+mve -mfloat-abi=hard -mfpu=auto" is broken
as specified in PR99939 and this patch fixes the issue.

gcc/testsuite/ChangeLog:

2021-06-11  Srinath Parvathaneni  <srinath.parvathaneni@arm.com>

PR target/99939
* gcc.target/arm/cmse/cmse-18.c: Add separate scan-assembler
directives check for target is v8.1-m.main+mve or not before
comparing the assembly output.
* gcc.target/arm/cmse/cmse-20.c: New test.

libgcc/ChangeLog:

2021-06-11  Srinath Parvathaneni  <srinath.parvathaneni@arm.com>

PR target/99939
* config/arm/cmse_nonsecure_call.S: Add __ARM_FEATURE_MVE
macro.
* config/arm/t-arm: To link cmse.o and cmse_nonsecure_call.o
on passing -mcmse option.

3 years agoc++: Add C++23 consteval if support - P1938R3 [PR100974]
Jakub Jelinek [Fri, 11 Jun 2021 13:50:34 +0000 (15:50 +0200)]
c++: Add C++23 consteval if support - P1938R3 [PR100974]

The following patch implements consteval if support.
There is a new IF_STMT_CONSTEVAL_P flag on IF_STMT and IF_COND is
boolean_false_node to match the non-manifestly constant evaluation
behavior, while constexpr evaluation special-cases it.  Perhaps cleaner
would be to set the condition to __builtin_is_constant_evaluated () call
but we need the IF_STMT_CONSTEVAL_P flag anyway and the IL would be larger.

And I'm not changing the libstdc++ side, where perhaps we could change
std::is_constant_evaluated definition for
 #ifdef __cpp_if_consteval
case to if consteval { return true; } else { return false; }
but we need to keep it defined to __builtin_is_constant_evaluated ()
for C++20 or older.

2021-06-11  Jakub Jelinek  <jakub@redhat.com>

PR c++/100974
gcc/c-family/
* c-cppbuiltin.c (c_cpp_builtins): Predefine __cpp_if_consteval for
-std=c++2b for P1938R3 consteval if support.
gcc/cp/
* cp-tree.h (struct saved_scope): Add consteval_if_p
member.  Formatting fix for the discarded_stmt comment.
(in_consteval_if_p, IF_STMT_CONSTEVAL_P): Define.
* parser.c (cp_parser_lambda_expression): Temporarily disable
in_consteval_if_p when parsing lambda body.
(cp_parser_selection_statement): Parse consteval if.
* decl.c (struct named_label_entry): Add in_consteval_if member.
(level_for_consteval_if): New function.
(poplevel_named_label_1, check_previous_goto_1, check_goto): Handle
consteval if.
* constexpr.c (cxx_eval_builtin_function_call): Clarify in comment
why CP_BUILT_IN_IS_CONSTANT_EVALUATED needs to *non_constant_p
for !ctx->manifestly_const_eval.
(cxx_eval_conditional_expression): For IF_STMT_CONSTEVAL_P evaluate
condition as if it was __builtin_is_constant_evaluated call.
(potential_constant_expression_1): For IF_STMT_CONSTEVAL_P always
recurse on both branches.
* cp-gimplify.c (genericize_if_stmt): Genericize IF_STMT_CONSTEVAL_P
as the else branch.
* pt.c (tsubst_expr) <case IF_STMT>: Copy IF_STMT_CONSTEVAL_P.
Temporarily set in_consteval_if_p when recursing on
IF_STMT_CONSTEVAL_P then branch.
(tsubst_lambda_expr): Temporarily disable
in_consteval_if_p when instantiating lambda body.
* call.c (immediate_invocation_p): Return false when
in_consteval_if_p.
gcc/testsuite/
* g++.dg/cpp23/consteval-if1.C: New test.
* g++.dg/cpp23/consteval-if2.C: New test.
* g++.dg/cpp23/consteval-if3.C: New test.
* g++.dg/cpp23/consteval-if4.C: New test.
* g++.dg/cpp23/consteval-if5.C: New test.
* g++.dg/cpp23/consteval-if6.C: New test.
* g++.dg/cpp23/consteval-if7.C: New test.
* g++.dg/cpp23/consteval-if8.C: New test.
* g++.dg/cpp23/consteval-if9.C: New test.
* g++.dg/cpp23/consteval-if10.C: New test.
* g++.dg/cpp23/feat-cxx2b.C: Add __cpp_if_consteval tests.

3 years agoanalyzer: tweak priority of callstrings in worklist::key_t::cmp
David Malcolm [Fri, 11 Jun 2021 13:30:33 +0000 (09:30 -0400)]
analyzer: tweak priority of callstrings in worklist::key_t::cmp

While debugging another issue I noticed that the analyzer could fail to
merge nodes for control flow in which one path had called a function
and another path hadn't:

        BB
       /  \
      /    \
 fn call   no fn call
      \    /
       \  /
     join BB

The root cause was that the worklist sort function wasn't prioritizing
call strings, and thus it was fully exploring the "no function called"
path to the exit BB, and only then exploring the "within the function call"
parts of the "funcion called" path.

This patch prioritizes call strings when sorting the worklist so that
the nodes with deeper call strings are processed before those with shallower
call strings, thus allowing such nodes to be merged at the joinpoint.

gcc/analyzer/ChangeLog:
* engine.cc (worklist::key_t::cmp): Move sort by call_string to
before SCC.

gcc/testsuite/ChangeLog:
* gcc.dg/analyzer/loop-0-up-to-n-by-1-with-iter-obj.c: Update
expected number of enodes after the loop.
* gcc.dg/analyzer/paths-8.c: New test.

Signed-off-by: David Malcolm <dmalcolm@redhat.com>
3 years agotree-optimization/101028 - fix endless SLP reassoc discovery
Richard Biener [Fri, 11 Jun 2021 11:36:26 +0000 (13:36 +0200)]
tree-optimization/101028 - fix endless SLP reassoc discovery

This fixes a missing clearing of mismatched lanes from the
fatal fail path in SLP reassoc discovery in the most conservative
way.

2021-06-11  Richard Biener  <rguenther@suse.de>

PR tree-optimization/101028
* tree-vect-slp.c (vect_build_slp_tree_2): When SLP
reassoc discovery fails fatally, mark appropriate lanes
in matches[] so.

* gcc.dg/pr101028.c: New testcase.

3 years agotree-optimization/101026 - fix SLP re-association
Richard Biener [Fri, 11 Jun 2021 11:33:17 +0000 (13:33 +0200)]
tree-optimization/101026 - fix SLP re-association

Since we cannot yet encode the operation in the SLP node itself
but need a representative stmt require an existing one for now
to avoid the need to build a fake GIMPLE stmt.

2021-06-11  Richard Biener  <rguenther@suse.de>

PR tree-optimization/101026
* tree-vect-slp.c (vect_build_slp_tree_2): Make sure we
have a representative for the associated chain nodes.

* gfortran.dg/pr101026.f: New testcase.

3 years agosimplify-rtx: Fix up simplify_logical_relational_operation for vector IOR [PR101008]
Jakub Jelinek [Fri, 11 Jun 2021 10:59:43 +0000 (12:59 +0200)]
simplify-rtx: Fix up simplify_logical_relational_operation for vector IOR [PR101008]

simplify_relational_operation callees typically return just const0_rtx
or const_true_rtx and then simplify_relational_operation attempts to fix
that up if the comparison result has vector mode, or floating mode,
or punt if it has scalar mode and vector mode operands (it doesn't know how
exactly to deal with the scalar masks).
But, simplify_logical_relational_operation has a special case, where
it attempts to fold (x < y) | (x >= y) etc. and if it determines it is
always true, it just returns const_true_rtx, without doing the dances that
simplify_relational_operation does.
That results in an ICE on the following testcase, where such folding happens
during expansion (of debug stmts into DEBUG_INSNs) and we ICE because
all of sudden a VOIDmode rtx appears where it expects a vector (V4SImode)
rtx.

The following patch fixes that by moving the adjustement into a separate
helper routine and using it from both simplify_relational_operation and
simplify_logical_relational_operation.

2021-06-11  Jakub Jelinek  <jakub@redhat.com>

PR rtl-optimization/101008
* simplify-rtx.c (relational_result): New function.
(simplify_logical_relational_operation,
simplify_relational_operation): Use it.

* gcc.dg/pr101008.c: New test.

3 years agoi386: Fix up *vec_concat<mode>_0_1 [PR101007]
Jakub Jelinek [Fri, 11 Jun 2021 10:58:22 +0000 (12:58 +0200)]
i386: Fix up *vec_concat<mode>_0_1 [PR101007]

On Fri, Apr 23, 2021 at 12:53:58PM +0800, Hongtao Liu via Gcc-patches wrote:
> -(define_insn "*vec_concatv4si_0"
> -  [(set (match_operand:V4SI 0 "register_operand"       "=v,x")
> -     (vec_concat:V4SI
> -       (match_operand:V2SI 1 "nonimmediate_operand" "vm,?!*y")
> -       (match_operand:V2SI 2 "const0_operand"       " C,C")))]
> +(define_insn "*vec_concat<mode>_0"
> +  [(set (match_operand:VI124_128 0 "register_operand"       "=v,x")
> +     (vec_concat:VI124_128
> +       (match_operand:<ssehalfvecmode> 1 "nonimmediate_operand" "vm,?!*y")
> +       (match_operand:<ssehalfvecmode> 2 "const0_operand"       " C,C")))]
>    "TARGET_SSE2"
>    "@
>     %vmovq\t{%1, %0|%0, %1}
> @@ -22154,6 +22157,24 @@ (define_insn "avx_vec_concat<mode>"
>     (set_attr "prefix" "maybe_evex")
>     (set_attr "mode" "<sseinsnmode>")])
>
> +(define_insn_and_split "*vec_concat<mode>_0"
> +  [(set (match_operand:V 0 "register_operand")
> +     (vec_select:V
> +       (vec_concat:<ssedoublevecmode>
> +         (match_operand:V 1 "nonimmediate_operand")
> +         (match_operand:V 2 "const0_operand"))
> +       (match_parallel 3 "movq_parallel"
> +         [(match_operand 4 "const_int_operand")])))]
> +  "ix86_pre_reload_split ()"
> +  "#"
> +  "&& 1"
> +  [(set (match_dup 0)
> +     (vec_concat:V (match_dup 1) (match_dup 5)))]
> +{
> +  operands[1] = gen_lowpart (<ssehalfvecmode>mode, operands[1]);
> +  operands[5] = CONST0_RTX (<ssehalfvecmode>mode);
> +})

This regressed the following testcase with -msse -mno-sse2.
The define_insn_and_split splits the permutation into *vec_concat<mode>_0
or *vec_concatv2di_0 insns which both have TARGET_SSE2 in their
conditions (for the former you can see it above), but the
define_insn_and_split matches always when the V mode's condition do,
which for V16QI/V8HI/V4SI/V2DI/V4SF modes is always (well, when those
modes are valid, which is TARGET_SSE).

2021-06-11  Jakub Jelinek  <jakub@redhat.com>

PR target/101007
* config/i386/sse.md (*vec_concat<mode>_0_1): Require TARGET_SSE2.

* gcc.target/i386/sse-pr101007.c: New test.

3 years agoi386: Try to avoid variable permutation instruction [PR101021]
Uros Bizjak [Fri, 11 Jun 2021 10:31:42 +0000 (12:31 +0200)]
i386: Try to avoid variable permutation instruction [PR101021]

Some permutations can be implemented without costly PSHUFB instruction, e.g.:

{ 8,9,10,11,12,13,14,15, 0,1,2,3,4,5,6,7 } with PALIGNR,

{ 0,1,2,3, 4,5,6,7, 4,5,6,7, 12,13,14,15 } with PSHUFD,

{ 0,1, 2,3, 2,3, 6,7, 8,9,10,11,12,13,14,15 } with PSHUFLW and

{ 0,1,2,3,4,5,6,7, 8,9, 10,11, 10,11, 14,15 } with PSHUFHW.

All these instructions have constant shuffle control mask and do not
need to load shuffle mask from a memory to a temporary XMM register.

2021-06-11  Uroš Bizjak  <ubizjak@gmail.com>

gcc/
PR target/101021
* config/i386/i386-expand.c (expand_vec_perm_pshufb): Return
false if the permutation can be implemented with constant
permutation instruction in wider mode.
(canonicalize_vector_int_perm): Move above expand_vec_perm_pshufb.
Handle V8QImode and V4HImode.

gcc/testsuite/

PR target/101021
* gcc.target/i386/pr101021-1.c: New test.
* gcc.target/i386/pr101021-2.c: Ditto.

3 years agoIntroduce -Wcoverage-invalid-line-number
Martin Liska [Tue, 1 Jun 2021 13:13:18 +0000 (15:13 +0200)]
Introduce -Wcoverage-invalid-line-number

PR gcov-profile/100788

gcc/ChangeLog:

* common.opt: Add new option.
* coverage.c (coverage_begin_function): Emit warning instead on
the internal compiler error.
* doc/invoke.texi: Document the option.
* toplev.c (process_options): Enable it by default.

gcc/testsuite/ChangeLog:

* gcc.dg/pr100788.c: New test.

3 years agoAdd missing ChangeLog entry.
Martin Liska [Fri, 11 Jun 2021 09:36:58 +0000 (11:36 +0200)]
Add missing ChangeLog entry.

3 years agomiddle-end/101009 - fix distance vector recording
Richard Biener [Fri, 11 Jun 2021 07:33:58 +0000 (09:33 +0200)]
middle-end/101009 - fix distance vector recording

This fixes recording of distance vectors in case the DDR has just
constant equal indexes.  In that case we expect distance vectors
with zero distances to be recorded which is what was done when
any distance was computed for affine indexes.

2021-06-11  Richard Biener  <rguenther@suse.de>

PR middle-end/101009
* tree-data-ref.c (build_classic_dist_vector_1): Make sure
to set *init_b to true when we encounter a constant equal
index pair.
(compute_affine_dependence): Also dump the actual DR_REF.

* gcc.dg/torture/pr101009.c: New testcase.

3 years agoDaily bump.
GCC Administrator [Fri, 11 Jun 2021 09:09:28 +0000 (09:09 +0000)]
Daily bump.

3 years agogcc-changelog: Ignore one more commit.
Martin Liska [Fri, 11 Jun 2021 08:13:54 +0000 (10:13 +0200)]
gcc-changelog: Ignore one more commit.

contrib/ChangeLog:

* gcc-changelog/git_update_version.py: Ignore commit that
violates rules and was somehow pushed.

3 years agors6000: Support more short/char to float conversion
Kewen Lin [Fri, 11 Jun 2021 07:43:40 +0000 (02:43 -0500)]
rs6000: Support more short/char to float conversion

For some cases that when we load unsigned char/short values from
the appropriate unsigned char/short memories and convert them to
double/single precision floating point value, there would be
implicit conversions to int first.  It makes GCC not leverage the
P9 instructions lxsibzx/lxsihzx.  This patch is to add the related
define_insn_and_split to support this kind of scenario.

Bootstrapped/regtested on powerpc64le-linux-gnu P9 and
powerpc64-linux-gnu P8.

gcc/ChangeLog:

* config/rs6000/rs6000.md
(floatsi<SFDF:mode>2_lfiwax_<QHI:mode>_mem_zext): New
define_insn_and_split.

gcc/testsuite/ChangeLog:

* gcc.target/powerpc/p9-fpcvt-3.c: New test.

3 years agoUse stablesort for sorting association chain
Richard Biener [Fri, 11 Jun 2021 06:04:41 +0000 (08:04 +0200)]
Use stablesort for sorting association chain

This should preserve the original association order as much as possible
for the initial SLP discovery attempt and also improve consistency.

2021-06-11  Richard Biener  <rguenther@suse.de>

* tree-vect-slp.c (vect_build_slp_tree_2): Use stablesort
to sort operands of the associative chain.

3 years agoExpose stable sort algorithm to gcc_sort_r and add vec::stablesort
Richard Biener [Thu, 10 Jun 2021 09:03:55 +0000 (11:03 +0200)]
Expose stable sort algorithm to gcc_sort_r and add vec::stablesort

This makes it possible to apply GCCs stable sort algorithm to vec<>
and also use it with the qsort_r compatible interface.

2021-06-10  Richard Biener  <rguenther@suse.de>

* system.h (gcc_stablesort_r): Declare.
* sort.cc (gcc_sort_r): Support stable sort.
(gcc_stablesort_r): Define.
* vec.h (vec<>::stablesort): Add.

3 years agoc++: Extend std::is_constant_evaluated in if warning [PR100995]
Marek Polacek [Wed, 9 Jun 2021 19:18:39 +0000 (15:18 -0400)]
c++: Extend std::is_constant_evaluated in if warning [PR100995]

Jakub pointed me at
<http://www.open-std.org/jtc1/sc22/wg21/docs/papers/2021/p1938r3.html#compiler-warnings>
which shows that our existing warning could be extended to handle more
cases.  This patch implements that.

A minor annoyance was handling macros, in libstdc++ we have

  reference operator[](size_type __pos) {
      __glibcxx_assert(__pos <= size());
      ...
  }

wherein __glibcxx_assert expands to

  if (__builtin_is_constant_evaluated() && !bool(__pos <= size())
    ...

but I'm of a mind to not warn on that.

Once consteval if makes it in, we should tweak this warning one more
time.

PR c++/100995

gcc/cp/ChangeLog:

* constexpr.c (maybe_constexpr_fn): New.
* cp-tree.h (maybe_constexpr_fn): Declare.
* semantics.c (find_std_constant_evaluated_r): New.
(maybe_warn_for_constant_evaluated): New.
(finish_if_stmt_cond): Call it.

gcc/testsuite/ChangeLog:

* g++.dg/cpp2a/is-constant-evaluated9.C: Add dg-warning.
* g++.dg/cpp2a/is-constant-evaluated12.C: New test.

3 years agoc++: matching deduced template template parameters [PR67829]
Patrick Palka [Thu, 10 Jun 2021 22:31:21 +0000 (18:31 -0400)]
c++: matching deduced template template parameters [PR67829]

During deduction, when the template of the argument for a bound ttp
is a template template parameter, we need to consider the
TEMPLATE_TEMPLATE_PARAMETER for matching rather than the TEMPLATE_DECL
thereof, because the canonical form of a template template parameter as
a template argument is the former tree, not the latter.

PR c++/67829

gcc/cp/ChangeLog:

* pt.c (unify) <case BOUND_TEMPLATE_TEMPLATE_PARM>: When
the TEMPLATE_DECL of a BOUND_TEMPLATE_TEMPLATE_PARM argument is
a template template parameter, adjust to the
TEMPLATE_TEMPLATE_PARAMETER before falling through.

gcc/testsuite/ChangeLog:

* g++.dg/template/ttp34.C: New test.
* g++.dg/template/ttp34a.C: New test.
* g++.dg/template/ttp34b.C: New test.

3 years agoc++: normalization of non-templated return-type-req [PR100946]
Patrick Palka [Thu, 10 Jun 2021 22:31:18 +0000 (18:31 -0400)]
c++: normalization of non-templated return-type-req [PR100946]

Here the satisfaction cache is conflating the satisfaction value of the
two return-type-requirements because the corresponding constrained
'auto's have level 2, but they capture an empty current_template_parms.
This ultimately causes the satisfaction cache to think the type
constraint doesn't depend on the deduced type of the expression.

When normalizing the constraints on an 'auto', the assumption made by
normalize_placeholder_type_constraints is that the level of the 'auto'
is one greater than the depth of the captured current_template_parms, an
assumption which is not holding here.  So this patch just makes n_p_t_c
adjust the normalization context appropriately in this situation.

PR c++/100946

gcc/cp/ChangeLog:

* constraint.cc (normalize_placeholder_type_constraints): When
normalizing a non-templated return-type-requirement, add a dummy
level to initial_parms.

gcc/testsuite/ChangeLog:

* g++.dg/cpp2a/concepts-return-req3.C: New test.

3 years agoi386: Add V8QI and other 64bit vector permutations [PR89021]
Peter Bergner [Thu, 10 Jun 2021 18:54:12 +0000 (13:54 -0500)]
i386: Add V8QI and other 64bit vector permutations [PR89021]

In addition to V8QI permutations, several other missing permutations are
added for 64bit vector modes for TARGET_SSSE3 and TARGET_SSE4_1 targets.

2021-06-10  Uroš Bizjak  <ubizjak@gmail.com>

gcc/
PR target/89021
* config/i386/i386-expand.c (ix86_split_mmx_punpck):
Handle V2SF mode.  Emit SHUFPS to fixup unpack-high for V2SF mode.
(expand_vec_perm_blend): Handle 64bit modes for TARGET_SSE4_1.
(expand_vec_perm_pshufb): Handle 64bit modes for TARGET_SSSE3.
(expand_vec_perm_pblendv): Handle 64bit modes for TARGET_SSE4_1.
(expand_vec_perm_interleave2): Handle 64bit modes.
(expand_vec_perm_even_odd_pack): Handle V8QI mode.
(expand_vec_perm_even_odd_1): Ditto.
(ix86_vectorize_vec_perm_const): Ditto.
* config/i386/i386.md (UNSPEC_PSHUFB): Move from ...
* config/i386/sse.md: ... here.
* config/i386/mmx.md (*vec_interleave_lowv2sf):
New insn_and_split pattern.
(*vec_interleave_highv2sf): Ditto.
(mmx_pshufbv8qi3): New insn pattern.
(*mmx_pblendw): Ditto.

3 years agolibgo: update to Go1.16.5 release
Ian Lance Taylor [Thu, 10 Jun 2021 19:37:34 +0000 (12:37 -0700)]
libgo: update to Go1.16.5 release

Reviewed-on: https://go-review.googlesource.com/c/gofrontend/+/326772

3 years agors6000: Add new __builtin_vsx_build_pair and __builtin_mma_build_acc built-ins
Peter Bergner [Thu, 10 Jun 2021 18:54:12 +0000 (13:54 -0500)]
rs6000: Add new __builtin_vsx_build_pair and __builtin_mma_build_acc built-ins

The __builtin_vsx_assemble_pair and __builtin_mma_assemble_acc built-ins
currently assign their first source operand to the first VSX register
in a pair/quad, their second operand to the second register in a pair/quad, etc.
This is not endian friendly and forces the user to generate different calls
depending on endianness.  In agreement with the POWER LLVM team, we've
decided to lightly deprecate the assemble built-ins and replace them with
"build" built-ins that automatically handle endianness so the same built-in
call and be used for both little-endian and big-endian compiles.  We are not
removing the assemble built-ins, since there is code in the wild that use
them, but we are removing their documentation to encourage the use of the
new "build" variants.

gcc/
* config/rs6000/rs6000-builtin.def (build_pair): New built-in.
(build_acc): Likewise.
* config/rs6000/rs6000-call.c (mma_expand_builtin): Swap assemble
source operands in little-endian mode.
(rs6000_gimple_fold_mma_builtin): Handle VSX_BUILTIN_BUILD_PAIR.
(mma_init_builtins): Likewise.
* config/rs6000/rs6000.c (rs6000_split_multireg_move): Handle endianness
ordering for the MMA assemble and build source operands.
* doc/extend.texi (__builtin_vsx_build_acc, __builtin_mma_build_pair):
Document.
(__builtin_mma_assemble_acc, __builtin_mma_assemble_pair): Remove
documentation.

gcc/testsuite/
* gcc.target/powerpc/mma-builtin-4.c (__builtin_vsx_build_pair): Add
tests.  Update expected counts.
* gcc.target/powerpc/mma-builtin-5.c (__builtin_mma_build_acc): Add
tests.  Update expected counts.

3 years agod: Fix ICE in TypeInfoDeclaration, at dmd/declaration.c (PR100967)
Iain Buclaw [Thu, 10 Jun 2021 17:59:23 +0000 (19:59 +0200)]
d: Fix ICE in TypeInfoDeclaration, at dmd/declaration.c (PR100967)

Generate a stub TypeInfo class even if the root Object class is missing.
The front-end will take care of issuing an error and abort the
compilation when running semantic on constructed TypeInfo objects.

The errors issued by the code generation pass relating to missing or
disabled RTTI has been consolidated into a single function, so that a
meaningful error will be emitted before the front-end terminates.

gcc/d/ChangeLog:

PR d/100967
* d-frontend.cc (getTypeInfoType): Move TypeInfo checks to
check_typeinfo_type and call new function.
* d-tree.h (check_typeinfo_type): Declare.
* typeinfo.cc: Include dmd/scope.h.
(create_frontend_tinfo_types): Generate front-end types even if Object
is missing.
(build_typeinfo): Move TypeInfo checks to check_typeinfo_type and call
new function.
(check_typeinfo_type): New function.

gcc/testsuite/ChangeLog:

PR d/100967
* gdc.dg/pr100967.d: New test.

3 years agors6000: Fix *TItype_ppc
Segher Boessenkool [Thu, 10 Jun 2021 15:14:49 +0000 (15:14 +0000)]
rs6000: Fix *TItype_ppc

The *TItype_ppc definitions are guarded by _ARCH_PPC64, so all
declarations using it should do so as well.

2021-06-10  Segher Boessenkool  <segher@kernel.crashing.org>
libgcc/
* config/rs6000/quad-float128.h: Guard all uses of [U]TItype_ppc by
_ARCH_PPC64 .

3 years agoUse auto_vec in ssa_equiv_stack.
Aldy Hernandez [Thu, 10 Jun 2021 07:20:30 +0000 (09:20 +0200)]
Use auto_vec in ssa_equiv_stack.

There is a mismatch between the new and the delete for the
ssa_equiv_stack class.  The correct idiom should have been delete[].
It has been pointed out that perhaps a better alternative is to use
an auto_vec which does everything automatically.  Plus, it is more
consistent with m_stack which is already an auto_vec.

This patch fixes the issue in PR100984.

Tested on x86-64 Linux.

gcc/ChangeLog:

        PR tree-optimization/100984
* gimple-ssa-evrp.c  (ssa_equiv_stack): Use auto_vec for
replacements table.
(ssa_equiv_stack::~ssa_equiv_stack): Remove.

3 years agoUse memory loads and extensions to eliminate redundant test/compare insns
Jeff Law [Thu, 10 Jun 2021 13:57:51 +0000 (09:57 -0400)]
Use memory loads and extensions to eliminate redundant test/compare insns

gcc/

* config/h8300/h8300.c (select_cc_mode): Handle MEM.  Use
REG_P.
* config/h8300/extensions.md: Replace _clobber_flags patterns
with <cczn>.

3 years agos390: Allow more vcond_mask patterns.
Robin Dapp [Thu, 10 Jun 2021 12:18:28 +0000 (14:18 +0200)]
s390: Allow more vcond_mask patterns.

Change vcond_mask iterator as to allow the corresponding int mode for
the condition/mask so e.g. boolean conditions become possible:

  vtarget = bool_cond ? vsource1 : vsource2.

gcc/ChangeLog:

* config/s390/vector.md (vcond_mask_<mode><mode>): Change to
(vcond_mask_<mode><tointvec>): this.

gcc/testsuite/ChangeLog:

* gcc.target/s390/vector/vcond-mixed-double.c: New test.
* gcc.target/s390/vector/vcond-mixed-float.c: New test.