gcc/ada/
* libgnat/a-cohama.ads: Introduce an equality operator over
cursors.
* libgnat/a-cohase.ads: Ditto.
* libgnat/a-cohama.adb: Add body for "=" over cursors.
(Insert): Do not set the Position component of the cursor that
denotes the inserted element.
* libgnat/a-cohase.adb: Ditto.
vectorizable_reduction had code guarded by:
if (STMT_VINFO_DEF_TYPE (stmt_info) == vect_reduction_def
|| STMT_VINFO_DEF_TYPE (stmt_info) == vect_double_reduction_def)
But that's always true after:
if (STMT_VINFO_DEF_TYPE (stmt_info) != vect_reduction_def
&& STMT_VINFO_DEF_TYPE (stmt_info) != vect_double_reduction_def
&& STMT_VINFO_DEF_TYPE (stmt_info) != vect_nested_cycle)
return false;
if (STMT_VINFO_DEF_TYPE (stmt_info) == vect_nested_cycle)
{
…
return true;
}
(I wasn't sure at first how the empty “else” for the first “if” above
was supposed to work.)
gcc/
* tree-vect-loop.c (vectorizable_reduction): Remove always-true
if condition.
match.pd has a rule to simplify an extension, operation and truncation
back to the original type:
(simplify
(convert (op:s@0 (convert1?@3 @1) (convert2?@4 @2)))
Currently it handles cases in which @2 is an INTEGER_CST, but it
also works for POLY_INT_CSTs.[*]
For INTEGER_CST it doesn't matter whether we test @2 or @4,
but for POLY_INT_CST it is possible to have unfolded (convert …)s.
Originally I saw this leading to some bad ivopts decisions, because
we weren't folding away redundancies from candidate iv expressions.
It's also possible to test the fold directly using the SVE ACLE.
[*] Not all INTEGER_CST rules work for POLY_INT_CSTs, since extensions
don't necessarily distribute over the internals of the POLY_INT_CST.
But in this case that isn't an issue.
gcc/
* match.pd: Simplify an extend-operate-truncate sequence involving
a POLY_INT_CST.
gcc/testsuite/
* gcc.target/aarch64/sve/acle/general/cntb_1.c: New test.
All of the optimizations/transformations mentioned in bugzilla for
PR tree-optimization/40210 are already implemented in mainline GCC,
with one exception. In comment #5, there's a suggestion that
(bswap64(x)>>56)&0xff can be implemented without the bswap as
(unsigned char)x, or equivalently x&0xff.
This patch implements the above optimization, and closely related
variants. For any single bit, (bswap(X)>>C1)&1 can be simplified
to (X>>C2)&1, where bit position C2 is the appropriate permutation
of C1. Similarly, the bswap can eliminated if the desired set of
bits all lie within the same byte, hence (bswap(x)>>8)&255 can
always be optimized, as can (bswap(x)>>8)&123.
Previously,
int foo(long long x) {
return (__builtin_bswap64(x) >> 56) & 0xff;
}
compiled with -O2 to
foo: movq %rdi, %rax
bswap %rax
shrq $56, %rax
ret
with this patch, it now compiles to
foo: movzbl %dil, %eax
ret
2021-07-08 Roger Sayle <roger@nextmovesoftware.com>
Richard Biener <rguenther@suse.de>
gcc/ChangeLog
PR tree-optimization/40210
* match.pd (bswap optimizations): Simplify (bswap(x)>>C1)&C2 as
(x>>C3)&C2 when possible. Simplify bswap(x)>>C1 as ((T)x)>>C2
when possible. Simplify bswap(x)&C1 as (x>>C2)&C1 when 0<=C1<=255.
gcc/testsuite/ChangeLog
PR tree-optimization/40210
* gcc.dg/builtin-bswap-13.c: New test.
* gcc.dg/builtin-bswap-14.c: New test.
This patch adds support for the VDIVSQ, VDIVUQ, VMODSQ, and VMODUQ
instructions to do 128-bit arithmetic.
2021-07-07 Michael Meissner <meissner@linux.ibm.com>
gcc/
PR target/100809
* config/rs6000/rs6000.md (udivti3): New insn.
(divti3): New insn.
(umodti3): New insn.
(modti3): New insn.
gcc/testsuite/
PR target/100809
* gcc.target/powerpc/p10-vdivq-vmodq.c: New test.
I'm working on reimplementing -Wanalyzer-use-of-uninitialized-value, but
I ran into issues with
region_model::add_any_constraints_from_ssa_def_stmt.
This function is from the initial commit of the analyzer and walks the
SSA names finding conditions that were missed due to the GCC 10 era
region_model not retaining useful information on how values were
created; as of GCC 11 the symbolic values contain this information,
and so the conditions can be reconstructed from them instead.
region_model::add_any_constraints_from_ssa_def_stmt is a liability
when tracking uninitialized values as it requires looking up SSA
values when those values may have been purged, thus greatly complicating
detection of uses of uninitialized values.
It's simplest to eliminate it and reimplement the condition-finding
via the makeup of the svalues, which this patch does. Doing so requires
supporting add_condition on svalues rather than just on trees, which
requires some changes to ana::state_machine and its subclasses.
gcc/analyzer/ChangeLog:
* diagnostic-manager.cc (null_assignment_sm_context::get_state):
New overload.
(null_assignment_sm_context::set_next_state): New overload.
(null_assignment_sm_context::get_diagnostic_tree): New.
* engine.cc (impl_sm_context::get_state): New overload.
(impl_sm_context::set_next_state): New overload.
(impl_sm_context::get_diagnostic_tree): New overload.
(impl_region_model_context::on_condition): Convert params from
tree to const svalue *.
* exploded-graph.h (impl_region_model_context::on_condition):
Likewise.
* region-model.cc (region_model::on_call_pre): Move handling of
internal calls to before checking for get_fndecl_for_call.
(region_model::add_constraints_from_binop): New.
(region_model::add_constraint): Split out into a new overload
working on const svalue * rather than tree. Call
add_constraints_from_binop. Drop call to
add_any_constraints_from_ssa_def_stmt.
(region_model::add_any_constraints_from_ssa_def_stmt): Delete.
(region_model::add_any_constraints_from_gassign): Delete.
(region_model::add_any_constraints_from_gcall): Delete.
* region-model.h
(region_model::add_any_constraints_from_ssa_def_stmt): Delete.
(region_model::add_any_constraints_from_gassign): Delete.
(region_model::add_any_constraints_from_gcall): Delete.
(region_model::add_constraint): Add overload decl.
(region_model::add_constraints_from_binop): New decl.
(region_model_context::on_condition): Convert params from tree to
const svalue *.
(noop_region_model_context::on_condition): Likewise.
* sm-file.cc (fileptr_state_machine::condition): Likewise.
* sm-malloc.cc (malloc_state_machine::on_condition): Likewise.
* sm-pattern-test.cc: Include tristate.h, selftest.h,
analyzer/call-string.h, analyzer/program-point.h,
analyzer/store.h, and analyzer/region-model.h.
(pattern_test_state_machine::on_condition): Convert params from tree to
const svalue *.
* sm-sensitive.cc (sensitive_state_machine::on_condition): Delete.
* sm-signal.cc (signal_state_machine::on_condition): Delete.
* sm-taint.cc (taint_state_machine::on_condition): Convert params
from tree to const svalue *.
* sm.cc: Include tristate.h, selftest.h, analyzer/call-string.h,
analyzer/program-point.h, analyzer/store.h, and
analyzer/region-model.h.
(any_pointer_p): Add overload taking const svalue *sval.
* sm.h (any_pointer_p): Add overload taking const svalue *sval.
(state_machine::on_condition): Convert params from tree to
const svalue *. Provide no-op default implementation.
(sm_context::get_state): Add overload taking const svalue *sval.
(sm_context::set_next_state): Likewise.
(sm_context::on_transition): Likewise.
(sm_context::get_diagnostic_tree): Likewise.
* svalue.cc (svalue::all_zeroes_p): New.
(constant_svalue::all_zeroes_p): New.
(repeated_svalue::all_zeroes_p): Convert to vfunc.
* svalue.h (svalue::all_zeroes_p): New decl.
(constant_svalue::all_zeroes_p): New decl.
(repeated_svalue::all_zeroes_p): Convert decl to vfunc.
gcc/testsuite/ChangeLog:
* gcc.dg/analyzer/pattern-test-2.c: Update expected results.
* gcc.dg/plugin/analyzer_gil_plugin.c
(gil_state_machine::on_condition): Remove.
Signed-off-by: David Malcolm <dmalcolm@redhat.com>
The previous MMA patch added some fragile code to initialize its new
built-ins. This patch hardens the initialization.
2021-07-07 Peter Bergner <bergner@linux.ibm.com>
gcc/
* config/rs6000/rs6000-call.c (mma_init_builtins): Use VSX_BUILTIN_LXVP
and VSX_BUILTIN_STXVP.
gcc/ada/
* einfo-utils.adb (Primitive_Operations): Default to returning
Direct_Primitive_Operations in the case of concurrent types
(when Corresponding_Record_Type not present).
* sem_ch9.adb (Analyze_Protected_Type_Declaration): Initialize
Direct_Primitive_Operations to an empty element list.
(Analyze_Task_Type_Declaration): Initialize
Direct_Primitive_Operations to an empty element list.
gcc/ada/
* sem_ch12.adb (Analyze_Subprogram_Instantiation): Mark Anon_Id
intrinsic before calling Analyze_Instance_And_Renamings because
this flag may be propagated to other nodes.
gcc/ada/
* checks.adb (Install_Primitive_Elaboration_Check): Do not
generate elaboration checks for primitives if pragma Pure or
Preelaborate is present. Misc comment fixes, including
referring to the correct kind of check (elaboration, not
accessibility).
* checks.ads, restrict.adb, sem_cat.ads, sinfo.ads: Minor
reformatting and comment fixes.
gcc/ada/
* freeze.adb (Check_Inherited_Conditions): Setting of Ekind,
LSP_Subprogram, and Is_Wrapper needs to happen for null
procedures as well as other wrapper cases, so the code is moved
from the else part in front of the if statement. (Fixes a
latent bug encountered while working on this set of changes.)
* sem_attr.adb (Resolve_Attribute): Report an error for the case
of an Access attribute applied to a primitive of an abstract
type when the primitive has any nonstatic Pre'Class or
Post'Class expressions.
* sem_ch8.adb (Analyze_Subprogram_Renaming): Report an error for
the case of a actual subprogram associated with a nonabstract
formal subprogram when the actual is a primitive of an abstract
type and the primitive has any nonstatic Pre'Class or Post'Class
expressions.
* sem_disp.adb (Check_Dispatching_Context): Remove special
testing for null procedures, and replace it with a relaxed test
that avoids getting an error about illegal calls to abstract
subprograms in cases where RM 6.1.1(7/5) applies in
Pre/Post'Class aspects. Also, remove special test for
Postcondition, which seems to be unnecessary, update associated
comments, and fix a typo in one comment.
(Check_Dispatching_Call): Remove an unneeded return statement,
and report an error for the case of a nondispatching call to a
nonabstract subprogram of an abstract type where the subprogram
has nonstatic Pre/Post'Class aspects.
* sem_util.ads
(Is_Prim_Of_Abst_Type_With_Nonstatic_CW_Pre_Post): New function.
(In_Pre_Post_Condition): Add a flag formal Class_Wide_Only,
defaulted to False, for indicating whether the function should
only test for the node being within class-wide pre- and
postconditions.
* sem_util.adb
(Is_Prim_Of_Abst_Type_With_Nonstatic_CW_Pre_Post): New function
to determine whether a subprogram is a primitive of an abstract
type where the primitive has class-wide Pre/Post'Class aspects
specified with nonstatic expressions.
(In_Pre_Post_Condition): Extend testing to account for the new
formal Class_Wide_Only.
gcc/ada/
* sem_ch12.adb (Check_Shared_Variable_Control_Aspects): Errors
emitted via Check_Volatility_Compatibility are now emitted at
Actual, just like other errors emitted by
Check_Shared_Variable_Control_Aspects.
gcc/ada/
* uname.adb (Get_Body_Name, Get_Parent_Body_Name,
Get_Parent_Spec_Name, Get_Spec_Name, Is_Child_Name,
Is_Body_Name, Is_Spec_Name, Name_To_Unit_Name): Use a local
instead of the global buffer.
gcc/ada/
* sem_ch13.adb (Analyze_Attribute_Definition_Clause): Combine
processing of Size and Value_Size clauses. Ensure that
Value_Size is treated the same as Size, in the cases where both
are allowed (i.e. the prefix denotes a first subtype). Misc
cleanup.
* einfo-utils.adb (Init_Size): Add assertions.
(Size_Clause): Return a Value_Size clause if present, instead of
just looking for a Size clause.
* einfo.ads (Has_Size_Clause, Size_Clause): Change documentation
to include Value_Size.
* sem_ch13.ads, layout.ads, layout.adb: Comment modifications.
gcc/ada/
* exp_dist.adb (Add_RACW_Primitive_Declarations_And_Bodies): Add
TSS_Put_Image to list of predefined primitives that need special
treatment.
(Build_General_Calling_Stubs, Build_Subprogram_Receiving_Stubs):
Remove previous hack for dealing with TSS_Put_Image procedures.
gcc/ada/
* libgnat/g-socket.adb (Get_Socket_Option): Add 500ms only when
Minus_500ms_Windows_Timeout is True.
(Set_Socket_Option): Use "* 1000" instead of "/ 0.001" to
convert to milliseconds.
gcc/ada/
* tbuild.adb (Unchecked_Convert_To): Set the Parent of the new
node to the Parent of the old node.
* tbuild.ads (Unchecked_Convert_To): Document differences
between Convert_To and Unchecked_Convert_To. The previous
documentation claimed they are identical except for the
uncheckedness of the conversion.
gcc/ada/
* gnat1drv.adb (Gnat1drv): Remove flagging of main unit and its
corresponding spec as requiring code generation; now the flags
are set much earlier.
* lib-load.adb (Load_Main_Source): Set Generate_Code flag on the
main unit source.
(Make_Instance_Unit): Copy Generate_Code flag from the main unit
to instance units.
* lib-writ.adb (Write_ALI): Remove redundant condition;
Generate_Code flag is always set for the main unit.
* par-load.adb (Load): Set Generate_Code flag on the main unit's
corresponding spec, if any.
gcc/ada/
* libgnat/g-socket.ads (Option_Name): Add Keep_Alive_Count,
Keep_Alive_Idle, and Keep_Alive_Interval items to enumeration.
(Option_Type): Add Keep_Alive_Count, Keep_Alive_Idle, and
Keep_Alive_Interval alternatives to the case of discriminated
record.
* libgnat/g-socket.adb (Options): Add Keep_Alive_Count,
Keep_Alive_Idle, and Keep_Alive_Interval to items enumerator to
OS constant converter.
(Set_Socket_Option): Process Keep_Alive_Count, Keep_Alive_Idle,
and Keep_Alive_Interval socket options.
(Get_Socket_Option): Idem.
gcc/ada/
* checks.adb, exp_attr.adb, exp_ch4.adb, exp_ch6.adb,
exp_ch9.adb, exp_disp.adb, exp_util.adb, inline.adb,
sem_res.adb: Change all calls to Make_Unchecked_Type_Conversion
to call Unchecked_Convert_To instead. This involves removing
New_Occurrence_Of on the first parameter, because
Unchecked_Convert_To expects a type entity, rather than the name
of one. Also, removed calls to Relocate_Node, because
Unchecked_Convert_To takes care of that.
* sinfo.ads: Change comment to be worded more firmly.
gcc/ada/
* libgnarl/s-tassta.adb (Free_Task): Acquire the Task_Lock
before, rather than after, querying the task's Terminated flag.
Add a corresponding Task_Unlock call.
This adds a testcase for an old fixed PR.
2021-07-07 Richard Biener <rguenther@suse.de>
PR tree-optimization/34195
* gcc.dg/vect/pr34195.c: New testcase.
This improves LIM by recording aggregate copies for disambiguation
purposes instead of as UNANALYZABLE_MEM which will prevent any
invariant or store motion across it. This allows four of the six
references in the loop of the testcase to be promoted.
2021-07-07 Richard Biener <rguenther@suse.de>
PR tree-optimization/99728
* tree-ssa-loop-im.c (gather_mem_refs_stmt): Record
aggregate copies.
(mem_refs_may_alias_p): Add assert we handled aggregate
copies elsewhere.
(sm_seq_valid_bb): Give up when running into aggregate copies.
(ref_indep_loop_p): Handle aggregate copies as never
being invariant themselves but allow other refs to be
disambiguated against them.
(can_sm_ref_p): Do not try to apply store-motion to aggregate
copies.
* g++.dg/opt/pr99728.C: New testcase.
