Big-endian targets need to save Z8-Z15 in the same order as
the registers would appear for D8-D15, because the layout is
mandated by the EH ABI. BE targets therefore use ST1D instead
of the normal STR for those registers (but not for others).
That difference is already tested elsewhere and isn't important
for the SME tests. This patch therefore restricts the affected
tests to LE.
gcc/testsuite/
* gcc.target/aarch64/sme/call_sm_switch_5.c: Restrict tests that
contain Z8-Z23 saves to little-endian.
* gcc.target/aarch64/sme/call_sm_switch_8.c: Likewise.
* gcc.target/aarch64/sme/locally_streaming_1.c: Likewise.
The .cfi scans in these tests failed for *-elf targets because
those targets don't enable .eh_frame info by default.
gcc/testsuite/
* gcc.target/aarch64/sme/call_sm_switch_1.c: Add -funwind-tables.
* gcc.target/aarch64/sme/call_sm_switch_3.c: Likewise.
* gcc.target/aarch64/sme/call_sm_switch_5.c: Likewise.
gcc/fortran/ChangeLog:
PR fortran/105543
* resolve.cc (resolve_symbol): For a CLASS-valued function having a
RESULT clause, ensure that attr.class_ok is set for its symbol as
well as for its resolved result variable.
gcc/testsuite/ChangeLog:
PR fortran/105543
* gfortran.dg/contiguous_13.f90: New test.
Mention Objective-C++ here to be consistent with the surrounding C/ObjC
lines.
gcc/ChangeLog:
* doc/invoke.texi (-fpermissive): Mention ObjC++ for -Wnarrowing.
This patch implements built-in trait for std::remove_pointer.
gcc/cp/ChangeLog:
* cp-trait.def: Define __remove_pointer.
* semantics.cc (finish_trait_type): Handle CPTK_REMOVE_POINTER.
gcc/testsuite/ChangeLog:
* g++.dg/ext/has-builtin-1.C: Test existence of __remove_pointer.
* g++.dg/ext/remove_pointer.C: New test.
Signed-off-by: Ken Matsui <kmatsui@gcc.gnu.org>
As Patrick suggested elsewhere, let's move this into the default case.
gcc/cp/ChangeLog:
* parser.cc (cp_parser_simple_type_specifier): Move trait
handling to default label.
This patch accepts the use of built-in trait identifiers when they are
actually not used as traits. Specifically, we check if the subsequent
token is '(' for ordinary built-in traits or is '<' only for the special
__type_pack_element built-in trait. If those identifiers are used
differently, the parser treats them as normal identifiers. This allows
us to accept code like: struct __is_pointer {};.
gcc/cp/ChangeLog:
* parser.cc (cp_lexer_lookup_trait): Rename to ...
(cp_lexer_peek_trait): ... this. Handle a subsequent token for
the corresponding built-in trait.
(cp_lexer_lookup_trait_expr): Rename to ...
(cp_lexer_peek_trait_expr): ... this.
(cp_lexer_lookup_trait_type): Rename to ...
(cp_lexer_peek_trait_type): ... this.
(cp_lexer_next_token_is_decl_specifier_keyword): Call
cp_lexer_peek_trait_type.
(cp_parser_simple_type_specifier): Likewise.
(cp_parser_primary_expression): Call cp_lexer_peek_trait_expr.
Signed-off-by: Ken Matsui <kmatsui@gcc.gnu.org>
Since RID_MAX soon reaches 255 and all built-in traits are used
approximately once in a C++ translation unit, this patch removes
all RID values for built-in traits and uses the identifier node to
look up the specific trait. Rather than holding traits as keywords,
we set all trait identifiers as cik_trait, which is a new
cp_identifier_kind. As cik_reserved_for_udlit was unused and
cp_identifier_kind is 3 bits, we replaced the unused field with the new
cik_trait. Also, the later patch handles a subsequent token to the
built-in identifier so that we accept the use of non-function-like
built-in trait identifiers.
gcc/c-family/ChangeLog:
* c-common.cc (c_common_reswords): Remove all mappings of
built-in traits.
* c-common.h (enum rid): Remove all RID values for built-in
traits.
gcc/cp/ChangeLog:
* cp-objcp-common.cc (names_builtin_p): Remove all RID value
cases for built-in traits. Check for built-in traits via
the new cik_trait kind.
* cp-tree.h (enum cp_trait_kind): Set its underlying type to
addr_space_t.
(struct cp_trait): New struct to hold trait information.
(cp_traits): New array to hold a mapping to all traits.
(cik_reserved_for_udlit): Rename to ...
(cik_trait): ... this.
(IDENTIFIER_ANY_OP_P): Exclude cik_trait.
(IDENTIFIER_TRAIT_P): New macro to detect cik_trait.
* lex.cc (cp_traits): Define its values, declared in cp-tree.h.
(init_cp_traits): New function to set cik_trait and
IDENTIFIER_CP_INDEX for all built-in trait identifiers.
(cxx_init): Call init_cp_traits function.
* parser.cc (cp_lexer_lookup_trait): New function to look up a
built-in trait by IDENTIFIER_CP_INDEX.
(cp_lexer_lookup_trait_expr): Likewise, look up an
expression-yielding built-in trait.
(cp_lexer_lookup_trait_type): Likewise, look up a type-yielding
built-in trait.
(cp_keyword_starts_decl_specifier_p): Remove all RID value cases
for built-in traits.
(cp_lexer_next_token_is_decl_specifier_keyword): Handle
type-yielding built-in traits.
(cp_parser_primary_expression): Remove all RID value cases for
built-in traits. Handle expression-yielding built-in traits.
(cp_parser_trait): Handle cp_trait instead of enum rid.
(cp_parser_simple_type_specifier): Remove all RID value cases
for built-in traits. Handle type-yielding built-in traits.
Co-authored-by: Patrick Palka <ppalka@redhat.com>
Signed-off-by: Ken Matsui <kmatsui@gcc.gnu.org>
This patch adds uaddv/usubv support on the H8 port to speed up those pesky
builtin-overflow tests. It's a variant of something I'd been running for a
while -- the major change between the old approach I'd been using and this
patch is this version does not expose the CC register until after reload to be
consistent with the rest of the H8 port.
The general approach is to first clear the GPR that's going to hold the
overflow status, perform the arithmetic operation (add/sub), then use addx to
move the overflow indicator (in the C bit) into the GPR holding the overflow
status.
That's a significant improvement over the mess of logicals that's generated by
the generic code.
Handling signed overflow is possible and something I'll probably port to this
scheme at some point. It's a bit more complex because we can't trivially move
the bit from CCR into the right position in a GPR and other quirks of the H8.
This has been regression tested on the H8 without problems. Pushing to the
trunk.
gcc/
* config/h8300/addsub.md (uaddv<mode>4, usubv<mode>4): New expanders.
(uaddv): New define_insn_and_split plus post-reload pattern.
Inspired by Roger's work on the ARC port, this patch provides a
define_and_split pattern to optimize sign extended bitfields starting at
position 0 using an approach that doesn't require shifting.
It then builds on that to provide another define_and_split pattern to support
arbitrary signed bitfield extractions -- it uses a right logical shift to move
the bitfield into position 0, then the specialized pattern above to sign extend
the MSB of the field through the rest of the register.
This is often, but certainly not always, better than a two shift approach. The
code uses the sizes of the sequences to select between the two shift approach
and single shift with extension from an arbitrary location approach.
There's certainly further improvements that could be made here, but I think
we're getting the bulk of the improvements already.
Regression tested on the H8 port without errors. Installing on the trunk.
gcc/
* config/h8300/h8300-protos.h (use_extvsi): Prototype.
* config/h8300/combiner.md: Two new define_insn_and_split patterns
to implement signed bitfield extractions.
* config/h8300/h8300.cc (use_extvsi): New function.
Various approaches are used to optimize extracting a sign extended single bit
bitfield. The length computation of 10 bytes was conservatively correct, but
inaccurate.
In particular when the bit we want is in the low half word we don't need the
move high half to low half instruction. Account for that in the length
computation.
This was spotted when looking at regressions in the generalized signed bitfield
extraction pattern.
This has been regression tested on the H8 port.
gcc/
* config/h8300/combiner.md (single bit signed bitfield extraction): Fix
length computation when the bit we want is in the low half word.
This fixes the length computation for logical shifts on the H8/SX.
The H8/SX has a richer set of logical shifts compared to early parts in the H8
family. It has special 2 byte instructions for shifts by power of two
immediate values as well as a special 4 byte shift by other immediate values.
These were never accounted for (AFIACT) in the length computation for shifts.
Until now that's mostly just affected branch shortening. But an upcoming patch
uses instruction lengths to select between two potential sequences and getting
these lengths wrong will cause it to miss optimization opportunities on the
H8/SX.
gcc
* config/h8300/h8300.cc (compute_a_shift_length): Fix computation
of logical shifts on the H8/SX.
This function is never called when param_l1_cache_line_size is 0,
but it uses int and unsigned int variables to hold alignment in
bits, so for large param_l1_cache_line_size it is zero and e.g.
DECL_ALIGN () % param_align_bits can divide by zero.
Looking at the code, the function uses tree_fits_uhwi_p on the trees
before converting them using tree_to_uhwi to int variables, which
looks just wrong, either it would need to punt if it doesn't fit
into those and also check for overflows during the computation,
or use unsigned HOST_WIDE_INT for all of this. That also fixes
the division by zero, as param_l1_cache_line_size maximum is INT_MAX,
that multiplied by 8 will always fit.
2023-12-09 Jakub Jelinek <jakub@redhat.com>
PR tree-optimization/112887
* tree-ssa-phiopt.cc (hoist_adjacent_loads): Change type of
param_align, param_align_bits, offset1, offset2, size2 and align1
variables from int or unsigned int to unsigned HOST_WIDE_INT.
* gcc.dg/pr112887.c: New test.
This testcase got fixed with
r14-6132-g50f2a3370d177f8fe9bea0461feb710523e048a2 .
I'm just adding a testcase so that it doesn't reappear.
2023-12-09 Jakub Jelinek <jakub@redhat.com>
PR tree-optimization/112924
* gcc.dg/pr112924.c: New test.
As noted in the PR, we support both features required for the 202110L
value, so we should define it with that value.
libstdc++-v3/ChangeLog:
PR libstdc++/111826
* include/bits/version.def (format): Update value.
* include/bits/version.h: Regenerate.
* testsuite/std/format/functions/format.cc:
What I implemented in r14-6199-g45630fbcf7875b does not match what I
proposed for LWG 4016, and it imposes additional, unwanted requirements
on the emplace and insert member functions of the container being
populated.
libstdc++-v3/ChangeLog:
PR libstdc++/112876
* include/std/ranges (ranges::to): Do not try to use an iterator
returned by the container's emplace or insert member functions.
* testsuite/std/ranges/conv/1.cc (Cont4::emplace, Cont4::insert):
Use the iterator parameter. Do not return an iterator.
driver:finalize used by JIT clears the mdswitches pointer; if it was
allocated before, that leaks the memory.
2023-12-09 Costas Argyris <costas.argyris@gmail.com>
Jakub Jelinek <jakub@redhat.com>
PR driver/93019
* gcc.cc (driver::finalize): Call XDELETEVEC on mdswitches before
clearing it.
Signed-off-by: Costas Argyris <costas.argyris@gmail.com>
There is another thing I wonder about: with -Wno-attributes= we are
supposed to ignore the attributes altogether, but we are actually still
warning about them when we emit these generic warnings about ignoring
all attributes which appertain to this and that (perhaps with some
exceptions we first remove from the attribute chain), like:
void foo () { [[foo::bar]]; }
with -Wattributes -Wno-attributes=foo::bar
Shouldn't we call some helper function in cases like this and warn
not when std_attrs (or how the attribute chain var is called) is non-NULL,
but if it is non-NULL and contains at least one non-attribute_ignored_p
attribute?
I've kept warnings for cases where the C++ standard says explicitly any
attributes aren't ok -
"If an attribute-specifier-seq appertains to a friend declaration, that
declaration shall be a definition."
or
https://eel.is/c++draft/dcl.type.elab#3
or
https://eel.is/c++draft/temp.spec#temp.explicit-3
For some changes I haven't figured out how could I cover it in the
testsuite.
Note, C uses a different strategy, it has c_warn_unused_attributes
function which warns about all the attributes one by one unless they
are ignored (or allowed in certain position).
Though that is just a single diagnostic wording, while C++ FE just warns
that there are some ignored attributes and doesn't name them individually
(except for namespace and using namespace) and uses different wordings in
different spots.
2023-12-09 Jakub Jelinek <jakub@redhat.com>
gcc/
* attribs.h (any_nonignored_attribute_p): Declare.
* attribs.cc (any_nonignored_attribute_p): New function.
gcc/cp/
* parser.cc (cp_parser_statement, cp_parser_expression_statement,
cp_parser_declaration, cp_parser_asm_definition): Don't diagnose
ignored attributes if !any_nonignored_attribute_p.
* decl.cc (grokdeclarator): Likewise.
* name-lookup.cc (handle_namespace_attrs, finish_using_directive):
Don't diagnose ignoring of attr_ignored_p attributes.
gcc/testsuite/
* g++.dg/warn/Wno-attributes-1.C: New test.
PR112932 let me notice there is a bug of current VLS mode misalign pattern.
Adapt it same as VLA mode.
Commited as it is obvious fix.
PR target/112932
gcc/ChangeLog:
* config/riscv/vector.md (movmisalign<mode>): Fix VLSmode bugs.
gcc/testsuite/ChangeLog:
* gcc.target/riscv/rvv/autovec/vls/misalign-1.c: Ditto.
* gcc.target/riscv/rvv/autovec/pr112932.c: New test.
These tests were recently xfailed on ilp32 targets though
passing on almost all ilp32 targets (known exceptions: ia32
and some arm subtargets). They've been changed around too
much to remain useful.
PR testsuite/112786
* gcc.dg/tree-ssa/scev-3.c, gcc.dg/tree-ssa/scev-4.c,
gcc.dg/tree-ssa/scev-5.c: Remove.
The emutls pass requires PROP_ssa, but if the strubm pass (or any
other pre-SSA pass) issues errors, all of the build_ssa_passes are
skipped, so the property is not set, but emutls still attempts to run,
on targets that use it, despite earlier errors, so it hits the
unsatisfied requirement.
Adjust emutls to be skipped in case of earlier errors.
for gcc/ChangeLog
* tree-emutls.cc: Include diagnostic-core.h.
(pass_ipa_lower_emutls::gate): Skip if errors were seen.
For decltype((x)) within a lambda where x is not captured, we dubiously
require that the lambda has a capture default, unlike for decltype(x).
But according to [expr.prim.id.unqual]/3 we should just ignore the lambda
in this case. This patch narrowly fixes this issue by disabling the
capture_decltype handling and falling back to the ordinary handling when
the innermost lambda has no capture-default. In fact, we can restrict
the special handling to only by-copy lambdas since that's what
[expr.prim.id.unqual]/3 is concerned with; for by-ref implicit captures
both code paths should give the same result anyway.
During review some other issues were discovered which are documented in
a new FIXME.
PR c++/83167
gcc/cp/ChangeLog:
* semantics.cc (capture_decltype): Inline into its only caller ...
(finish_decltype_type): ... here. Update nearby comment to refer
to recent standard. Add FIXME. Restrict uncaptured variable type
transformation to happen only for lambdas with a by-copy
capture-default.
gcc/testsuite/ChangeLog:
* g++.dg/cpp0x/lambda/lambda-decltype4.C: New test.
gcc/analyzer/ChangeLog:
* region-model.cc (contains_uninit_p): Only check for
svalues that the infoleak warning can handle.
gcc/testsuite/ChangeLog:
* gcc.dg/plugin/infoleak-uninit-size-1.c: New test.
* gcc.dg/plugin/infoleak-uninit-size-2.c: New test.
* gcc.dg/plugin/plugin.exp: Add the new tests.
Signed-off-by: David Malcolm <dmalcolm@redhat.com>
PR112875 test ran into a wrong assert (gcc_unreachable) in elimination
in a debug insn. The insn seems ok. So I change the assertion.
To be more accurate I made it the same as analogous reload pass code.
gcc/ChangeLog:
PR rtl-optimization/112875
* lra-eliminations.cc (lra_eliminate_regs_1): Change an assert.
Add ASM_OPERANDS case.
gcc/testsuite/ChangeLog:
PR rtl-optimization/112875
* gcc.target/i386/pr112875.c: New test.
When working on the previous patch I put [[]] [[]] asm (""); into a
testcase, but was surprised it wasn't parsed.
The problem is that when cp_parser_std_attribute_spec returns NULL, it
can mean 2 different things, one is that the next token(s) are neither
[[ nor alignas (in that case the caller should break from the loop),
or when we parsed something like [[]] - it was valid attribute specifier,
but didn't specify any attributes in it.
The following patch fixes that by using a magic value of void_list_node
for the case where the first tokens are neither [[ nor alignas and so
where cp_parser_std_attribute_spec_seq should stop iterating to differentiate
it from NULL_TREE which stands for some attribute specifier has been parsed,
but it didn't contain any (or any valid) attributes.
2023-12-08 Jakub Jelinek <jakub@redhat.com>
* parser.cc (cp_parser_std_attribute_spec): Return void_list_node
rather than NULL_TREE if token is neither CPP_OPEN_SQUARE nor
RID_ALIGNAS CPP_KEYWORD.
(cp_parser_std_attribute_spec_seq): For attr_spec == void_list_node
break, for attr_spec == NULL_TREE continue.
* g++.dg/cpp0x/gen-attrs-79.C: New test.
The following testcase is miscompiled because two ubsan instrumentations
run into each other.
The first one is the shift instrumentation. Before the C++ FE calls
it, it wraps the 2 shift arguments with cp_save_expr, so that side-effects
in them aren't evaluated multiple times. And, ubsan_instrument_shift
itself uses unshare_expr on any uses of the operands to make sure further
modifications in them don't affect other copies of them (the only not
unshared ones are the one the caller then uses for the actual operation
after the instrumentation, which means there is no tree sharing).
Now, if there are side-effects in the first operand like say function
call, cp_save_expr wraps it into a SAVE_EXPR, and ubsan_instrument_shift
in this mode emits something like
if (..., SAVE_EXPR <foo ()>, SAVE_EXPR <op1> > const)
__ubsan_handle_shift_out_of_bounds (..., SAVE_EXPR <foo ()>, ...);
and caller adds
SAVE_EXPR <foo ()> << SAVE_EXPR <op1>
after it in a COMPOUND_EXPR. So far so good.
If there are no side-effects and cp_save_expr doesn't create SAVE_EXPR,
everything is ok as well because of the unshare_expr.
We have
if (..., SAVE_EXPR <op1> > const)
__ubsan_handle_shift_out_of_bounds (..., ptr->something[i], ...);
and
ptr->something[i] << SAVE_EXPR <op1>
where ptr->something[i] is unshared.
In the testcase below, the !x->s[j] ? 1 : 0 expression is wrapped initially
into a SAVE_EXPR though, and unshare_expr doesn't unshare SAVE_EXPRs nor
anything used in them for obvious reasons, so we end up with:
if (..., SAVE_EXPR <!(bool) VIEW_CONVERT_EXPR<const struct S *>(x)->s[j] ? 1 : 0>, SAVE_EXPR <op1> > const)
__ubsan_handle_shift_out_of_bounds (..., SAVE_EXPR <!(bool) VIEW_CONVERT_EXPR<const struct S *>(x)->s[j] ? 1 : 0>, ...);
and
SAVE_EXPR <!(bool) VIEW_CONVERT_EXPR<const struct S *>(x)->s[j] ? 1 : 0> << SAVE_EXPR <op1>
So far good as well. But later during cp_fold of the SAVE_EXPR we find
out that VIEW_CONVERT_EXPR<const struct S *>(x)->s[j] ? 0 : 1 is actually
invariant (has TREE_READONLY set) and so cp_fold simplifies the above to
if (..., SAVE_EXPR <op1> > const)
__ubsan_handle_shift_out_of_bounds (..., (bool) VIEW_CONVERT_EXPR<const struct S *>(x)->s[j] ? 0 : 1, ...);
and
((bool) VIEW_CONVERT_EXPR<const struct S *>(x)->s[j] ? 0 : 1) << SAVE_EXPR <op1>
with the s[j] ARRAY_REFs and other expressions shared in between the two
uses (and obviously the expression optimized away from the COMPOUND_EXPR in
the if condition.
Then comes another ubsan instrumentation at genericization time,
this time to instrument the ARRAY_REFs with strict bounds checking,
and replaces the s[j] in there with s[.UBSAN_BOUNDS (0B, SAVE_EXPR<j>, 8), SAVE_EXPR<j>]
As the trees are shared, it does that just once though.
And as the if body is gimplified first, the SAVE_EXPR<j> is evaluated inside
of the if body and when it is used again after the if, it uses a potentially
uninitialized value of j.1 (always uninitialized if the shift count isn't
out of bounds).
The following patch fixes that by unshare_expr unsharing the folded argument
of a SAVE_EXPR if we've folded the SAVE_EXPR into an invariant and it is
used more than once.
2023-12-08 Jakub Jelinek <jakub@redhat.com>
PR sanitizer/112727
* cp-gimplify.cc (cp_fold): If SAVE_EXPR has been previously
folded, unshare_expr what is returned.
* c-c++-common/ubsan/pr112727.c: New test.
This adds a sanity check to cp_parser_expression_statement similar to
the one in finish_expr_stmt added by r6-6795-g0fd9d4921f7ba2, which
effectively downgrades accepts-invalid/wrong-code bugs like this one
into ice-on-invalid/ice-on-valid ones.
PR c++/112658
gcc/cp/ChangeLog:
* parser.cc (cp_parser_expression_statement): If the statement
is error_mark_node, make sure we've seen_error().
When cp_build_c_cast commits to an erroneous const_cast, we neglect to
replay errors from build_const_cast_1 which can lead to us incorrectly
accepting (and "miscompiling") the cast, or triggering the assert in
finish_expr_stmt.
This patch fixes this oversight. This was the original fix for the ICE
in PR112658 before r14-5941-g305a2686c99bf9 made us accept the testcase
there after all. I wasn't able to come up with an alternate testcase for
which this fix has an effect anymore, but below is a reduced version of
the PR112658 testcase (accepted ever since r14-5941) for good measure.
PR c++/112658
PR c++/94264
gcc/cp/ChangeLog:
* typeck.cc (cp_build_c_cast): If we're committed to a const_cast
and the result is erroneous, call build_const_cast_1 a second
time to issue errors. Use complain=tf_none instead of =false.
gcc/testsuite/ChangeLog:
* g++.dg/cpp0x/initlist-array20.C: New test.
This patch adds vectorized strcmp and strncmp implementations and
tests. Similar to strlen, expansion is still guarded by
-minline-str(n)cmp.
gcc/ChangeLog:
PR target/112109
* config/riscv/riscv-protos.h (expand_strcmp): Declare.
* config/riscv/riscv-string.cc (riscv_expand_strcmp): Add
strategy handling and delegation to scalar and vector expanders.
(expand_strcmp): Vectorized implementation.
* config/riscv/riscv.md: Add TARGET_VECTOR to strcmp and strncmp
expander.
gcc/testsuite/ChangeLog:
* gcc.target/riscv/rvv/autovec/builtin/strcmp-run.c: New test.
* gcc.target/riscv/rvv/autovec/builtin/strcmp.c: New test.
* gcc.target/riscv/rvv/autovec/builtin/strncmp-run.c: New test.
* gcc.target/riscv/rvv/autovec/builtin/strncmp.c: New test.
This patch implements a vectorized strlen by re-using and slightly
adjusting the rawmemchr implementation. Rawmemchr returns the address
of the needle while strlen returns the difference between needle address
and start address.
As before, strlen expansion is guarded by -minline-strlen.
While testing with -minline-strlen I encountered a vsetvl problem in
memcpy-chk.c where we didn't insert a vsetvl at the proper spot (after
a setjmp). This needs to be fixed separately and I figured I'd post
this patch as-is.
gcc/ChangeLog:
PR target/112109
* config/riscv/riscv-protos.h (expand_rawmemchr): Add strlen
parameter.
* config/riscv/riscv-string.cc (riscv_expand_strlen): Call
rawmemchr.
(expand_rawmemchr): Add strlen handling.
* config/riscv/riscv.md: Add TARGET_VECTOR to strlen expander.
gcc/testsuite/ChangeLog:
* gcc.target/riscv/rvv/autovec/builtin/strlen-run.c: New test.
* gcc.target/riscv/rvv/autovec/builtin/strlen.c: New test.
early-ra's likely_operand_match_p didn't handle relaxed and special
memory constraints, which meant that the pass wasn't able to match
LD1RQ instructions to their constraints, and so backed out of
trying to allocate. This patch fixes that by switching the sense
of the match: does the rtx seem appropriate for the constraint?,
rather than: does the constraint seem appropriate for the rtx?
Also, I came across a case that needed more general equivalence
detection. Previously we would only record equivalences after
the last definition of the source register, but it's worth trying
to handle cases where the destination register's live range is
restricted to a block, and the next definition of the source
occurs only after the end of the destination register's live range.
The patch also fixes a cut-&-pasto that Alex noticed (thanks).
gcc/
* config/aarch64/aarch64-early-ra.cc (allocno_info::chain_next):
Put into an enum with...
(allocno_info::last_def_point): ...new member variable.
(allocno_info::m_current_bb_point): New member variable.
(likely_operand_match_p): Switch based on get_constraint_type,
rather than based on rtx code. Handle relaxed and special memory
constraints.
(early_ra::record_copy): Allow the source of an equivalence to be
assigned to more than once.
(early_ra::record_allocno_use): Invalidate any previous equivalence.
Initialize last_def_point.
(early_ra::record_allocno_def): Set last_def_point.
(early_ra::valid_equivalence_p): New function, split out from...
(early_ra::record_copy): ...here. Use last_def_point to handle
source registers that have a later definition.
(make_pass_aarch64_early_ra): Fix comment.
gcc/testsuite/
* gcc.target/aarch64/sme/strided_2.c: New test.
