Parameterized derived types are handled in a special way and start with 'Pdt'.
If the 'P' is not uppercase, gfc_get_derived_type (which calls
gfc_get_module_backend_decl) does not find the existing declaration and
builds a new type. The middle end then sees those types as being different
and nonalising, creating an endless loop for pdt_14.f03.
gcc/fortran/ChangeLog:
PR fortran/97652
* module.c (mio_symbol): Fix symbol name for pdt_type.
Convert
x < (short) ((unsigned short)x + const)
to
x <= SHORT_MAX – const
(and similarly for other integral types) if const is not 0.
gcc/
PR tree-optimization/97223
* match.pd (overflow detection and optimization): Handle conversions.
gcc/testsuite/
* gcc.dg/no-strict-overflow-4.c: Adjust expected output.
gcc/ChangeLog:
* common/config/riscv/riscv-common.c (riscv_implied_info):
Add static and const.
(riscv_subset_list::handle_implied_ext): Add const due to
riscv_implied_info changed to const.
- Disable kasan if target is unsupported and -fasan-shadow-offset= is not
given, no matter `--param asan-stack=1` is given or not.
- Moving KASAN option checking testcase to gcc.dg, those testcase could be
useful for all other target which not support asan.
- Verifed on riscv and x86.
gcc/ChangeLog:
PR target/96307
* toplev.c (process_options): Remove param_asan_stack checking for kasan
option checking.
gcc/testsuite/ChangeLog:
PR target/96307
* gcc.dg/pr96307.c: New.
* gcc.target/riscv/pr96260.c: Move this test case from here to ...
* gcc.dg/pr96260.c: ... here.
* gcc.target/riscv/pr91441.c: Move this test case from here to ...
* gcc.dg/pr91441.c: ... here.
* lib/target-supports.exp (check_effective_target_no_fsanitize_address):
New proc.
Here's a small deficiency in decltype(auto). [dcl.type.auto.deduct]/5:
If the placeholder-type-specifier is of the form decltype(auto), [...]
the type deduced for T is determined [...] as though E had been the operand
of the decltype. So:
int &&i = 0;
decltype(auto) j = i; // should behave like int &&j = i; error
We deduce j's type in do_auto_deduction via finish_decltype_type which
takes an 'id' argument. Currently we compute 'id' as false, because
stripped_init is *i (a REFERENCE_REF_P). But it seems to me we should
rather set 'id' to true here, by looking through the REFERENCE_REF_P,
so that finish_decltype_type DTRT.
gcc/cp/ChangeLog:
PR c++/78209
* pt.c (do_auto_deduction): If init is REFERENCE_REF_P, use its
first operand.
gcc/testsuite/ChangeLog:
PR c++/78209
* g++.dg/cpp1y/decltype-auto1.C: New test.
This PR asks that we add a warning option for an existing (very old)
warning, so that it can be disabled selectively. clang++ uses
-Wexceptions for this, so I added this new option rather than using
e.g. -Wnoexcept.
gcc/c-family/ChangeLog:
PR c++/97675
* c.opt (Wexceptions): New option.
gcc/cp/ChangeLog:
PR c++/97675
* except.c (check_handlers_1): Use OPT_Wexceptions for the
warning. Use inform for the second part of the warning.
gcc/ChangeLog:
PR c++/97675
* doc/invoke.texi: Document -Wexceptions.
gcc/testsuite/ChangeLog:
PR c++/97675
* g++.old-deja/g++.eh/catch10.C: Adjust dg-warning.
* g++.dg/warn/Wexceptions1.C: New test.
* g++.dg/warn/Wexceptions2.C: New test.
This patch implements the -Wvexing-parse warning to warn about the
sneaky most vexing parse rule in C++: the cases when a declaration
looks like a variable definition, but the C++ language requires it
to be interpreted as a function declaration. This warning is on by
default (like clang++). From the docs:
void f(double a) {
int i(); // extern int i (void);
int n(int(a)); // extern int n (int);
}
Another example:
struct S { S(int); };
void f(double a) {
S x(int(a)); // extern struct S x (int);
S y(int()); // extern struct S y (int (*) (void));
S z(); // extern struct S z (void);
}
You can find more on this in [dcl.ambig.res].
I spent a fair amount of time on fix-it hints so that GCC can recommend
various ways to resolve such an ambiguity. Sometimes that's tricky.
E.g., suggesting default-initialization when the class doesn't have
a default constructor would not be optimal. Suggesting {}-init is also
not trivial because it can use an initializer-list constructor if no
default constructor is available (which ()-init wouldn't do). And of
course, pre-C++11, we shouldn't be recommending {}-init at all.
I also uncovered a bug in cp_parser_declarator, where we were setting
*parenthesized_p to true despite the comment saying the exact opposite.
gcc/c-family/ChangeLog:
PR c++/25814
* c.opt (Wvexing-parse): New option.
gcc/cp/ChangeLog:
PR c++/25814
* cp-tree.h (enum cp_tree_index): Add CPTI_EXPLICIT_VOID_LIST.
(explicit_void_list_node): Define.
(PARENTHESIZED_LIST_P): New macro.
(struct cp_declarator): Add function::parens_loc.
* decl.c (cxx_init_decl_processing): Initialize explicit_void_list_node.
(grokparms): Also break when explicit_void_list_node.
* parser.c (make_call_declarator): New location_t parameter. Use it
to set declarator->u.function.parens_loc.
(cp_parser_lambda_declarator_opt): Pass UNKNOWN_LOCATION to
make_call_declarator.
(warn_about_ambiguous_parse): New function.
(cp_parser_init_declarator): Call warn_about_ambiguous_parse.
(cp_parser_declarator): Set *parenthesized_p to false rather than to
true.
(cp_parser_direct_declarator): Create a location for the function's
parentheses and pass it to make_call_declarator.
(cp_parser_parameter_declaration_clause): Return explicit_void_list_node
for (void).
(cp_parser_parameter_declaration_list): Set PARENTHESIZED_LIST_P
in the parameters tree.
gcc/ChangeLog:
PR c++/25814
* doc/invoke.texi: Document -Wvexing-parse.
gcc/testsuite/ChangeLog:
PR c++/25814
* g++.dg/cpp2a/fn-template16.C: Add a dg-warning.
* g++.dg/cpp2a/fn-template7.C: Likewise.
* g++.dg/lookup/pr80891-5.C: Likewise.
* g++.dg/lto/pr79050_0.C: Add extern.
* g++.dg/lto/pr84805_0.C: Likewise.
* g++.dg/parse/pr58898.C: Add a dg-warning.
* g++.dg/template/scope5.C: Likewise.
* g++.old-deja/g++.brendan/recurse.C: Likewise.
* g++.old-deja/g++.jason/template4.C: Likewise.
* g++.old-deja/g++.law/arm4.C: Likewise.
* g++.old-deja/g++.mike/for2.C: Likewise.
* g++.old-deja/g++.other/local4.C: Likewise.
* g++.old-deja/g++.pt/crash3.C: Likewise.
* g++.dg/warn/Wvexing-parse.C: New test.
* g++.dg/warn/Wvexing-parse2.C: New test.
* g++.dg/warn/Wvexing-parse3.C: New test.
* g++.dg/warn/Wvexing-parse4.C: New test.
* g++.dg/warn/Wvexing-parse5.C: New test.
* g++.dg/warn/Wvexing-parse6.C: New test.
* g++.dg/warn/Wvexing-parse7.C: New test.
libstdc++-v3/ChangeLog:
PR c++/25814
* testsuite/20_util/reference_wrapper/lwg2993.cc: Add a dg-warning.
* testsuite/25_algorithms/generate_n/87982_neg.cc: Likewise.
fix range-ops equal/not_equal to not reuse the result range as intermediary.
value_query::value routines should pasa multi-range in as some other rangeop
routines build into this result, so we may need better precision.
gcc/
PR tree-optimization/97725
* range-op.cc (operator_equal::fold_range): Use new tmp value.
(operator_not_equal::fold_range): Ditto.
* value-query.cc (range_query::value_of_expr): Use int_range_max
not a value_range.
(range_query::value_on_edge): Ditto.
(range_query::value_of_stmt): Ditto.
gcc/testsuite/
* gcc.dg/pr97725.c: New.
The relational operators for std::optional were using the wrong types
in the declval expressions used to constrain them. Instead of using
const lvalues they were using non-const rvalues, which meant that a type
might satisfy the constraints but then give an error when the function
body was instantiated.
libstdc++-v3/ChangeLog:
PR libstdc++/96269
* include/std/optional (operator==, operator!=, operator<)
(operator>, operator<=, operator>=): Fix types used in
SFINAE constraints.
* testsuite/20_util/optional/relops/96269.cc: New test.
Pick an alternate static chain regnum and document that
the port needs to be configured to issue -ffixed-r18 by
default. The current definition of TARGET_OS_USES_R18 is
not honored by the common part of the back-end.
2020-11-04 Olivier Hainque <hainque@adacore.com>
gcc/
* config/aarch64/aarch64-vxworks.h (TARGET_OS_USES_R18):
Remove definition.
(STATIC_CHAIN_REGNUM): Redefine to 9.
As described in the PR, the recursive_directory_iterator constructor
calls advance(ec), but ec is a pointer so it calls _Dir::advance(bool).
The intention was to either call advance() or advance(*ec) depending
whether the pointer is null or not.
This fixes the bug and renames the parameter to ecptr to make similar
mistakes less likely in future.
libstdc++-v3/ChangeLog:
PR libstdc++/97731
* src/filesystem/dir.cc (recursive_directory_iterator): Call the
right overload of _Dir::advance.
* testsuite/experimental/filesystem/iterators/97731.cc: New test.
This fixes some multiple definition errors caused by the changes for
PR libstdc++/90295. The previous solution for inlining the members of
std::exception_ptr but still exporting them from the library was to
suppress the 'inline' keyword on those functions when compiling
libsupc++/eh_ptr.cc, so they get defined in that file. That produces ODR
violations though, because there are now both inline and non-inline
definitions in the library, due to the use of std::exception_ptr in
other files sucg as src/c++11/future.cc.
The new solution is to define all the relevant members as 'inline'
unconditionally, but use __attribute__((used)) to cause definitions to
be emitted in libsupc++/eh_ptr.cc as before. This doesn't quite work
however, because PR c++/67453 means the attribute is ignored on
constructors and destructors. As a workaround, the old solution
(conditionally inline) is still used for those members, but they are
given the always_inline attribute so that they aren't emitted in
src/c++11/future.o as inline definitions.
libstdc++-v3/ChangeLog:
PR libstdc++/97729
* include/std/future (__basic_future::_M_get_result): Use
nullptr for null pointer constant.
* libsupc++/eh_ptr.cc (operator==, operator!=): Remove
definitions.
* libsupc++/exception_ptr.h (_GLIBCXX_EH_PTR_USED): Define
macro to conditionally add __attribute__((__used__)).
(operator==, operator!=, exception_ptr::exception_ptr())
(exception_ptr::exception_ptr(const exception_ptr&))
(exception_ptr::~exception_ptr())
(exception_ptr::operator=(const exception_ptr&))
(exception_ptr::swap(exception_ptr&)): Always define as
inline. Add macro to be conditionally "used".
The test uses -fpic and doesn't query the target support
for that option otherwise, resulting in failure on configurations
not supporting -fpic such as VxWorks for kernel mode.
2020-11-03 Olivier Hainque <hainque@adacore.com>
gcc/testsuite/
* gcc.dg/sms-12.c: Add dg-require-effective-target fpic.
The change moves the definitions of PROBE_STACK_FIRST_REG
and PROBE_STACK_SECOND_REG to a more appropriate place for such
items (here, in aarch64.md as suggested by Richard), and adjusts
their value from r9/r10 to r10/r11 to free r9 for a possibly
more general purpose (e.g. as a static chain at least on targets
which have a private use of r18, such as Windows or Vxworks).
2020-11-07 Olivier Hainque <hainque@adacore.com>
gcc/
* config/aarch64/aarch64.md: Define PROBE_STACK_FIRST_REGNUM
and PROBE_STACK_SECOND_REGNUM constants, designating r10/r11.
Replacements for the PROBE_STACK_FIRST/SECOND_REG constants in
aarch64.c.
* config/aarch64/aarch64.c (PROBE_STACK_FIRST_REG): Remove.
(PROBE_STACK_SECOND_REG): Remove.
(aarch64_emit_probe_stack_range): Adjust to the _REG -> _REGNUM
suffix update for PROBE_STACK register numbers.
They say third time is the charm.. It looks like the testcase
disables the cost model and so AArch64 we end up being able to
do the permute but on x86 we can't. However when analyzing the
testcase I didn't disable the cost model hence the difference.
So I now guard the testcase on vect_load_lanes as there's not a
"can do any permute" test directive and load lanes is what I will
be fixing up next year so this should catch it.
gcc/testsuite/ChangeLog:
* gcc.dg/vect/slp-11b.c: Guard statements.
Sometimes the overflow flag will leak into the IL. Drop it while
creating ranges.
There are various places we could plug this. This patch just plugs things
at get_tree_range which is the entry point for ranges from tree expressions.
It fixes the PR, and probably fixes the ranger entirely, but we may need
to revisit this.
For example, I looked to see if there were other places that created
ranges with TREE_OVERFLOW set, and there are various. For example,
the following code pattern appears multiple times in vr-values.c:
else if (is_gimple_min_invariant (op0))
vr0.set (op0);
This can pick up TREE_OVERFLOW from the IL if present. However, the
ranger won't see them so we're good.
At some point we should audit all this. Or perhaps just nuke all
TREE_OVERFLOW's at irange::set.
For now, this will do.
gcc/ChangeLog:
PR tree-optimization/97721
* gimple-range.cc (get_tree_range): Drop overflow from constants.
gcc/testsuite/ChangeLog:
* gcc.dg/pr97721.c: New test.
This change fixes a bug in the i386 backend when adding
-fzero-call-used-regs=all on a target that has no x87
registers.
When there is no x87 registers available, we should not
zero stack registers.
gcc/ChangeLog:
PR target/97715
* config/i386/i386.c (zero_all_st_registers): Return
earlier when the FPU is disabled.
gcc/testsuite/ChangeLog:
PR target/97715
* gcc.target/i386/zero-scratch-regs-32.c: New test.
- Add a missing 'explicit' to a basic_stringbuf constructor.
- Set up the get/put area pointers in the constructor from strings using
different allocator types.
- Remove public basic_stringbuf::__sv_type alias.
- Do not construct temporary basic_string objects with a
default-constructed allocator.
Also, change which basic_string constructor is used, as a minor
compile-time optimization. Constructing from a basic_string_view
requires more work from the compiler, so just use a pointer and length.
libstdc++-v3/ChangeLog:
* include/std/sstream (basic_stringbuf(const allocator_type&):
Add explicit.
(basic_stringbuf(const basic_string<C,T,SA>&, openmode, const A&)):
Call _M_stringbuf_init. Construct _M_string from pointer and length
to avoid constraint checks for string view.
(basic_stringbuf::view()): Make __sv_type alias local to the
function.
(basic_istringstream(const basic_string<C,T,SA>&, openmode, const A&)):
Pass string to _M_streambuf instead of constructing a temporary
with the wrong allocator.
(basic_ostringstream(const basic_string<C,T,SA>&, openmode, const A&)):
Likewise.
(basic_stringstream(const basic_string<C,T,SA>&, openmode, const A&)):
Likewise.
* src/c++20/sstream-inst.cc: Use string_view and wstring_view
typedefs in explicit instantiations.
* testsuite/27_io/basic_istringstream/cons/char/1.cc: Add more
tests for constructors.
* testsuite/27_io/basic_ostringstream/cons/char/1.cc: Likewise.
* testsuite/27_io/basic_stringbuf/cons/char/1.cc: Likewise.
* testsuite/27_io/basic_stringbuf/cons/char/2.cc: Likewise.
* testsuite/27_io/basic_stringbuf/cons/wchar_t/1.cc: Likewise.
* testsuite/27_io/basic_stringbuf/cons/wchar_t/2.cc: Likewise.
* testsuite/27_io/basic_stringstream/cons/char/1.cc: Likewise.
The following fixes SLP vectorization of stores that were
pattern recognized. Since in SLP vectorization pattern analysis
happens after dataref group analysis we have to adjust the groups
with the pattern stmts. This has some effects down the pipeline
and exposes cases where we looked at the wrong pattern/non-pattern
stmts.
2020-11-05 Richard Biener <rguenther@suse.de>
* tree-vect-data-refs.c (vect_slp_analyze_node_dependences):
Use the original stmts.
(vect_slp_analyze_node_alignment): Use the pattern stmt.
* tree-vect-slp.c (vect_fixup_store_groups_with_patterns):
New function.
(vect_slp_analyze_bb_1): Call it.
* gcc.dg/vect/bb-slp-69.c: New testcase.
This optimizes sequential permutes. i.e. if there are two permutes back to back
this function applies the permute of the parent to the child and removed the
parent.
This relies on the materialization point calculation in optimize SLP.
This allows us to remove useless permutes such as
ldr q0, [x0, x3]
ldr q2, [x1, x3]
trn1 v1.4s, v0.4s, v0.4s
trn2 v0.4s, v0.4s, v0.4s
trn1 v0.4s, v1.4s, v0.4s
mov v1.16b, v3.16b
fcmla v1.4s, v0.4s, v2.4s, #0
fcmla v1.4s, v0.4s, v2.4s, #90
str q1, [x2, x3]
from the sequence the vectorizer puts out and give
ldr q0, [x0, x3]
ldr q2, [x1, x3]
mov v1.16b, v3.16b
fcmla v1.4s, v0.4s, v2.4s, #0
fcmla v1.4s, v0.4s, v2.4s, #90
str q1, [x2, x3]
instead.
gcc/ChangeLog:
* tree-vect-slp.c (vect_slp_tree_permute_noop_p): New.
(vect_optimize_slp): Optimize permutes.
(vectorizable_slp_permutation): Fix typo.
My previous patch accidentally enabled some tests on x86 because my target
selector foo was weak.. This now properly only runs them on AArch64.
gcc/testsuite/ChangeLog:
* gcc.dg/vect/slp-11b.c: Update testcase.
* gcc.dg/vect/slp-perm-6.c: Update target selector.
The change to clear the external_die_map slot after creating
the concrete instance DIE broke abstract origin processing which
tried to make sure to have those point to the early abstract instance
and not the concrete instance. The following restores this by
eventually following the abstract origin link in the concrete instance.
2020-11-05 Richard Biener <rguenther@suse.de>
PR debug/97718
* dwarf2out.c (add_abstract_origin_attribute): Make sure to
point to the abstract instance.
This patch stores the SLP instance kind in the SLP instance so that we can use
it later when detecting load/store lanes support.
This also changes the load/store lane support check to only check if the SLP
kind is a store. This means that in order for the load/lanes to work all
instances must be of kind store.
gcc/ChangeLog:
* tree-vect-loop.c (vect_analyze_loop_2): Check kind.
* tree-vect-slp.c (vect_build_slp_instance): New.
(enum slp_instance_kind): Move to...
* tree-vectorizer.h (enum slp_instance_kind): .. Here
(SLP_INSTANCE_KIND): New.
This patch is to make vector CTOR with char/short leverage direct
move instructions when they are available. With one constructed
test case, it can speed up 145% for char and 190% for short on P9.
Tested SPEC2017 x264_r at -Ofast on P9, it gets 1.61% speedup
(but based on unexpected SLP see PR96789).
Bootstrapped/regtested on powerpc64{,le}-linux-gnu P8 and
powerpc64le-linux-gnu P9.
gcc/ChangeLog:
PR target/96933
* config/rs6000/rs6000.c (rs6000_expand_vector_init): Use direct move
instructions for vector construction with char/short types.
* config/rs6000/rs6000.md (p8_mtvsrwz_v16qisi2): New define_insn.
(p8_mtvsrd_v16qidi2): Likewise.
gcc/testsuite/ChangeLog:
PR target/96933
* gcc.target/powerpc/pr96933-1.c: New test.
* gcc.target/powerpc/pr96933-2.c: New test.
* gcc.target/powerpc/pr96933-3.c: New test.
* gcc.target/powerpc/pr96933-4.c: New test.
* gcc.target/powerpc/pr96933.h: New test.
* gcc.target/powerpc/pr96933-run.h: New test.
This moves the code that checks for load/store lanes further in the pipeline and
places it after slp_optimize. This would allow us to perform optimizations on
the SLP tree and only bail out if we really have a permute.
With this change it allows us to handle permutes such as {1,1,1,1} which should
be handled by a load and replicate.
This change however makes it all or nothing. Either all instances can be handled
or none at all. This is why some of the test cases have been adjusted.
gcc/ChangeLog:
* tree-vect-slp.c (vect_analyze_slp_instance): Moved load/store lanes
check to ...
* tree-vect-loop.c (vect_analyze_loop_2): ..Here
gcc/testsuite/ChangeLog:
* gcc.dg/vect/slp-11b.c: Update output scan.
* gcc.dg/vect/slp-perm-6.c: Likewise.
Go programs expect to be able to get reliable backtrace information
with correct file/line information, but -fipa-icf-functions breaks
that because it merges together distinct functions which should have
distinct file/line info.
* go-lang.c (go_langhook_post_options): Disable
-fipa-icf-functions if it was not explicitly enabled.
Add a timestamp to supplement the global range cache to detect when a value
may become stale.
gcc/
PR tree-optimization/97515
* gimple-range-cache.h (class ranger_cache): New prototypes plus
temporal cache pointer.
* gimple-range-cache.cc (struct range_timestamp): New.
(class temporal_cache): New.
(temporal_cache::temporal_cache): New.
(temporal_cache::~temporal_cache): New.
(temporal_cache::get_timestamp): New.
(temporal_cache::set_dependency): New.
(temporal_cache::temporal_value): New.
(temporal_cache::current_p): New.
(temporal_cache::set_timestamp): New.
(temporal_cache::set_always_current): New.
(ranger_cache::ranger_cache): Allocate the temporal cache.
(ranger_cache::~ranger_cache): Free temporal cache.
(ranger_cache::get_non_stale_global_range): New.
(ranger_cache::set_global_range): Add a timestamp.
(ranger_cache::register_dependency): New. Add timestamp dependency.
* gimple-range.cc (gimple_ranger::range_of_range_op): Add operand
dependencies.
(gimple_ranger::range_of_phi): Ditto.
(gimple_ranger::range_of_stmt): Check if global range is stale, and
recalculate if so.
gcc/testsuite/
* gcc.dg/pr97515.c: Check listing for folding of entire function.
As noted in PR 96817 this new test fails if the library is built without
futexes. That's expected of course, but we might as well fail more
obviously than a deadlock that eventually times out.
libstdc++-v3/ChangeLog:
* testsuite/18_support/96817.cc: Fail fail if the library is
configured to not use futexes.
I forgot to cost vectorized PHIs. Scalar PHIs are just costed
as scalar_stmt so the following costs vector PHIs as vector_stmt.
2020-11-04 Richard Biener <rguenther@suse.de>
* tree-vectorizer.h (vectorizable_phi): Adjust prototype.
* tree-vect-stmts.c (vect_transform_stmt): Adjust.
(vect_analyze_stmt): Pass cost_vec to vectorizable_phi.
* tree-vect-loop.c (vectorizable_phi): Do costing.
This properly sets the abnormal flag when vectorizing live lanes
when the original scalar was live across an abnormal edge.
2020-11-04 Richard Biener <rguenther@suse.de>
PR tree-optimization/97709
* tree-vect-loop.c (vectorizable_live_operation): Set
SSA_NAME_OCCURS_IN_ABNORMAL_PHI when necessary.
* gcc.dg/vect/bb-slp-pr97709.c: New testcase.
The following patch generalizes the x ? 1 : 0 -> (int) x optimization
to handle also left shifts by constant.
During x86_64-linux and i686-linux bootstraps + regtests it triggered
in 1514 unique non-LTO -m64 cases (sort -u on log mentioning
filename, function name and shift count) and 1866 -m32 cases.
Unfortunately, the patch regresses (before the tests have been adjusted):
+FAIL: gcc.dg/tree-ssa/ssa-ccp-11.c scan-tree-dump-times optimized "if " 0
+FAIL: gcc.dg/vect/bb-slp-pattern-2.c -flto -ffat-lto-objects scan-tree-dump-times slp1 "optimized: basic block" 1
+FAIL: gcc.dg/vect/bb-slp-pattern-2.c scan-tree-dump-times slp1 "optimized: basic block" 1
and in both cases it actually results in worse code.
> > We'd need some optimization that would go through all PHI edges and
> > compute if some use of the phi results don't actually compute a constant
> > across all the PHI edges - 1 & 0 and 0 & 1 is always 0.
> PRE should do this, IMHO only optimizing it at -O2 is fine.
> > Similarly, in the slp vectorization test there is:
> > a[0] = b[0] ? 1 : 7;
> note this, carefully avoiding the already "optimized" b[0] ? 1 : 0 ...
> So the option is to put : 7 in the 2, 4 an 8 case as well. The testcase
> wasn't added for any real-world case but is artificial I guess for
> COND_EXPR handling of invariants.
> But yeah, for things like SLP it means we eventually have to
> implement reverse transforms for all of this to make the lanes
> matching. But that's true anyway for things like x + 1 vs. x + 0
> or x / 3 vs. x / 2 or other simplifications we do.
2020-11-04 Jakub Jelinek <jakub@redhat.com>
PR tree-optimization/97690
* tree-ssa-phiopt.c (conditional_replacement): Also optimize
cond ? pow2p_cst : 0 as ((type) cond) << cst.
* gcc.dg/tree-ssa/phi-opt-22.c: New test.
* gcc.dg/tree-ssa/ssa-ccp-11.c: Use -O2 instead of -O1.
* gcc.dg/vect/bb-slp-pattern-2.c (foo): Use ? 2 : 7, ? 4 : 7 and
? 8 : 7 instead of ? 2 : 0, ? 4 : 0, ? 8 : 0.
