Whoops. When un-disabling visibility support for mmix, I missed that
some of the newly enabled tests were FAILs, for not emitting .hidden
for references to external declarations. This takes care of
gcc.dg/visibility-14.c .. -19.c, and gcc.dg/visibility-23.c.
gcc:
* config/mmix/mmix.h (ASM_OUTPUT_EXTERNAL): Define to
default_elf_asm_output_external.
In PR ipa/96291 the test contained an SCC with one
unoptimized function. This tricked ipa-cp into NULL dereference.
has_undead_caller_from_outside_scc_p() did not take into account
that unoptimized funtions don't have IPA summary analysis. And
dereferenced NULL pointer causing an ICE.
gcc/
PR ipa/96291
* ipa-cp.c (has_undead_caller_from_outside_scc_p): Consider
unoptimized callers as undead.
gcc/testsuite/
PR ipa/96291
* gcc.dg/lto/pr96291_0.c: New testcase.
* gcc.dg/lto/pr96291_1.c: Support file.
* gcc.dg/lto/pr96291_2.c: Likewise.
* gcc.dg/lto/pr96291.h: Likewise.
This patch implements several constant folding optimizations
for __builtin_parity and friends. We canonicalize popcount(x)&1
as parity(x) in gimple, and potentially convert back again when
we expand to RTL. parity(~x) is simplified to parity(x), which
is true for all integer modes with an even number of bits.
But probably most usefully, parity(x)^parity(y) can be simplified
to a parity(x^y), requiring only a single libcall or popcount.
This patch optimizes popcount and parity of an argument known to have
at most a single bit set, to be that single bit. Hence, popcount(x&8)
is simplified to (x>>3)&1. This generalizes the existing optimization
of popcount(x&1) being simplified to x&1, which is cleaned up with
this patch.
2020-07-28 Roger Sayle <roger@nextmovesoftware.com>
Richard Biener <rguenther@suse.de>
gcc/ChangeLog
* match.pd (popcount(x)&1 -> parity(x)): New simplification.
(parity(~x) -> parity(x)): New simplification.
(parity(x)^parity(y) -> parity(x^y)): New simplification.
(parity(x&1) -> x&1): New simplification.
(popcount(x) -> x>>C): New simplification.
gcc/testsuite/ChangeLog
* gcc.dg/fold-popcount-5.c: New test.
* gcc.dg/fold-parity-1.c: Likewise.
* gcc.dg/fold-parity-2.c: Likewise.
* gcc.dg/fold-parity-3.c: Likewise.
* gcc.dg/fold-parity-4.c: Likewise.
* gcc.dg/fold-parity-5.c: Likewise.
I discovered we were not setting DECL_CONTEXT in a few cases, and
grokfndecl's control flow wasn't making it clear that we were doing it
in all cases.
gcc/cp/
* cp-gimplify.c (cp_genericize_r): Set IMPORTED_DECL's context.
* cp-objcp-common.c (cp_pushdecl): Set decl's context.
* decl.c (grokfndecl): Make DECL_CONTEXT setting clearer.
fixup_type_variants was almost doing all that finish_struct needs.
May as well make it do it all.
gcc/cp/
* class.c (fixup_type_variants): Copy TYPE_SIZE and
TYPE_SIZE_UINIT.
(finish_struct): Call it.
The following testcase ICEs, because cp_lexer_safe_previous_token calls
cp_lexer_previous_token and that ICEs, because all tokens in the lexer
buffer before the current one (CPP_EOF) have been purged.
cp_lexer_safe_previous_token is used in the context where it is ok if it
punts, so the patch changes the function so that it doesn't assert there is
some previous token, but instead returns NULL like in other cases where it
punts.
In addition to this, in the last hunk it does a micro-optimization, don't
call the potentially expensive function if it will not need the result,
instead check the least expensive condition first.
And the middle hunk is a similar change from Mark's version of the patch,
to use the safe variant in there because it is again just about a hint
and it is better not to provide the hint than to ICE, though we don't have a
testcase that would ICE.
2020-07-28 Jakub Jelinek <jakub@redhat.com>
Mark Wielaard <mark@klomp.org>
PR c++/96328
* parser.c (cp_lexer_safe_previous_token): Don't call
cp_lexer_previous_token, instead inline it by hand and return NULL
instead of failing assertion if all previous tokens until the first
one are purged.
(cp_parser_error_1): Optimize - only call cp_lexer_safe_previous_token
if token->type is CPP_NAME. Use cp_lexer_safe_previous_token instead
of cp_lexer_previous_token for the missing_token_desc != RT_NONE
case too.
* g++.dg/diagnostic/pr96328.C: New test.
Co-Authored-By: Mark Wielaard <mark@klomp.org>
For (invalid) newline inside of the raw string literal delimiter, doing
continue means we skip the needed processing of newlines. Instead of
duplicating that, this patch just doesn't continue for those.
2020-07-28 Jakub Jelinek <jakub@redhat.com>
PR preprocessor/96323
* lex.c (lex_raw_string): For c == '\n' don't continue after reporting
an prefix delimiter error.
* c-c++-common/cpp/pr96323.c: New test.
Add support for 16-bits shifts and for sign extension from 8 bits to
16 bits.
This patch has been tested on nvptx-none with no new regressions.
2020-07-28 Roger Sayle <roger@nextmovesoftware.com>
Tom de Vries <tdevries@suse.de>
gcc/ChangeLog:
* config/nvptx/nvptx.md (extendqihi2): New instruction.
(ashl<mode>3, ashr<mode>3, lshr<mode>3): Support HImode.
gcc/testsuite/ChangeLog:
* gcc.target/nvptx/cvt.c: New test.
* gcc.target/nvptx/shift16.c: New test.
We can always use the final map to get the return value, rather than
conditionally only when there was an immediate line directive.
libcpp/
* init.c (cpp_read_main_file): Always use the last map for
the return value.
We used to consider range size on insertion but on unique keys container
not all range values might be inserted resulting in over-sizing. In this
case we just consider user reservation and if none then the container will
adapt to actually inserted elements.
libstdc++-v3/ChangeLog:
* include/bits/hashtable.h
(_Hashtable<>(_InputIterator, _InputIterator, size_t, const _H1&,
const _H2&, const _Hash&, const _Equal&, const _ExtractKey&,
const allocator_type&, true_type)): New.
(_Hashtable<>(_InputIterator, _InputIterator, size_t, const _H1&,
const _H2&, const _Hash&, const _Equal&, const _ExtractKey&,
const allocator_type&, false_type)): New.
(_Hashtable<>(_InputIterator, _InputIterator, size_t, const _H1&,
const _H2&, const _Hash&, const _Equal&, const _ExtractKey&,
const allocator_type&)): Delegate to latters.
(operator=(initializer_list<value_type>)): Rehash if too small.
(_M_insert(_Arg&&, const _NodeGenerator&, true_type)): Remove
size_t len parameter.
* include/bits/hashtable_policy.h (_Insert_base<>::_M_insert_range):
Do not try to get input range distance.
* testsuite/23_containers/unordered_set/cons/bucket_hint.cc: New test.
* testsuite/23_containers/unordered_set/modifiers/insert.cc: New test.
2020-07-28 Jakub Jelinek <jakub@redhat.com>
* g++.dg/cpp2a/nontype-subob1.C: Allow s1, i1 or x1 instead of l1 for
targets with short, int or long long ssize_t.
The following testcase ICEs in maybe_warn_rdwr_sizes. The problem is that
the caller uses its fndecl and fntype variables to fill up rdwr_map, and
the fntype in that case is a prototype with the access attribute and all
the checks needed for that performed. But the maybe_warn_rdwr_sizes
function tries to rediscover fndecl/fntype itself and does it differently
from how the caller did (for fndecl get_callee_fndecl and fntype from that
FUNCTION_DECL, otherwise sets fntype to CALL_EXPR_FN's type).
On the testcase, get_callee_fndecl does find a FUNCTION_DECL because
it does STRIP_NOPS in between.
Instead of trying to rediscover those, this patch just passes them down,
like is done in several other functions.
2020-07-28 Jakub Jelinek <jakub@redhat.com>
PR middle-end/96335
* calls.c (maybe_warn_rdwr_sizes): Add FNDECL and FNTYPE arguments,
instead of trying to rediscover them in the body.
(initialize_argument_information): Adjust caller.
* gcc.dg/pr96335.c: New test.
This patch is to refactor the existing peel_iters_prologue and
peel_iters_epilogue cost model handlings, by following the structure
below suggested by Richard Sandiford:
- calculate peel_iters_prologue
- calculate peel_iters_epilogue
- add costs associated with peel_iters_prologue
- add costs associated with peel_iters_epilogue
- add costs related to branch taken/not_taken.
Bootstrapped/regtested on aarch64-linux-gnu.
gcc/ChangeLog:
* tree-vect-loop.c (vect_get_known_peeling_cost): Factor out some code
to determine peel_iters_epilogue to...
(vect_get_peel_iters_epilogue): ...this new function.
(vect_estimate_min_profitable_iters): Refactor cost calculation on
peel_iters_prologue and peel_iters_epilogue.
This ports https://golang.org/cl/227163 to the Go frontend.
This is a step toward moving up to the go1.15rc1 release.
Original CL description:
cmd/compile,runtime: pass only ptr and len to some runtime calls
Some runtime calls accept a slice, but only use ptr and len.
This change modifies most such routines to accept only ptr and len.
After this change, the only runtime calls that accept an unnecessary
cap arg are concatstrings and slicerunetostring.
Neither is particularly common, and both are complicated to modify.
Negligible compiler performance impact. Shrinks binaries a little.
There are only a few regressions; the one I investigated was
due to register allocation fluctuation.
Passes 'go test -race std cmd', modulo golang/go#38265 and golang/go#38266.
Wow, does that take a long time to run.
file before after Δ %
compile 19655024 19655152 +128 +0.001%
cover 5244840 5236648 -8192 -0.156%
dist 3662376 3658280 -4096 -0.112%
link 6680056 6675960 -4096 -0.061%
pprof 14789844 14777556 -12288 -0.083%
test2json 2824744 2820648 -4096 -0.145%
trace 11647876 11639684 -8192 -0.070%
vet 8260472 8256376 -4096 -0.050%
total 115163736 115118808 -44928 -0.039%
For golang/go#36890
Reviewed-on: https://go-review.googlesource.com/c/gofrontend/+/245099
This avoids requiring an init function to initialize the variable.
This can only be done if x is a static initializer.
The go1.15rc1 runtime package relies on this optimization.
The package has a variable "var maxSearchAddr = maxOffAddr".
The maxSearchAddr variable is used by code that runs before package
initialization is complete.
Reviewed-on: https://go-review.googlesource.com/c/gofrontend/+/245098
These FAILs for mmix showed up as regressions for me due to a
flaw in the btest-gcc.sh test-results-accounting: a bug was
recently fixed regarding the naming of dump-files so the names
are again correct. To wit, parts of the tests that were
UNRESOLVED, due to missing dump-files, and ignored in the
presence of other parts (execution, excess errors) PASSing,
became FAIL, trumping the PASSing parts of the tests.
As in a recent patch, the variables are "privatized" using
ASM_PN_FORMAT for MMIX and the lines to match look like:
y::0_1 = { i }
y::0_1, points-to NULL, points-to vars: { D.1465 } (nonlocal, escaped)
instead of e.g. for cris-elf:
y.0_1 = { i }
y.0_1, points-to NULL, points-to vars: { D.1433 } (nonlocal, escaped)
Also checked that the general pattern still matches for cris-elf.
gcc/testsuite:
* gcc.dg/torture/pr39074-2.c: Adjust for mmix.
* gcc.dg/torture/pr39074.c, gcc.dg/torture/pta-callused-1.c: Ditto.
The logic to figure out where a missing #include should be inserted
uses pointer equality to check filenames -- the routine even says so.
But cpplib makes no such guarantee. It happens to be true for input
that it preprocesses[* see line zero below], but is not true for
source that has already been preprocessed -- all those '# ...' line
directives produce disctinct filename strings. That renders using
-fdirectives-only as a prescanning stage (as I understand some people
do), broken.
This patch changes to string comparisons, and explicitly rejects any
line-zero location map that occurs at the beginning of a file. The
very first map of a file has a different string to the remaining maps,
and we never tripped on that because of the pointer comparison. The
second testcase deploys -save-temps to cause an intermediate
preprocessed output that is read back.
gcc/c-family/
* c-common.c (try_to_locate_new_include_insertion_point): Use
strcmp, not pointer equality.
gcc/testsuite/
* g++.dg/lookup/missing-std-include-10.h: New.
* g++.dg/lookup/missing-std-include-10.C: New.
* g++.dg/lookup/missing-std-include-11.C: New.
The as-base type never got a name. For modules I needed to give it a
name to serialize properly, and it's useful when debugging the
compiler, so we may as well have it on trunk. There's also a bug
where its fields can have NSDMIs from the main class. This happens to
be silent on trunk, but can be a GC leak where we retain a deferred
parse node there. (On modules it blows up, because we're not prepared
to serialize deferred parse nodes, as they should never survive
parsing.
gcc/cp/
* cp-tree.h (enum cp_tree_index): Add CPTI_AS_BASE_IDENTIFIER.
(as_base_identifier): Define.
* decl.c (initialize_predifined_identifiers): Initialize as_base
identifier.
* class.c (layout_class_type): Name the as-base type. Zap
NSDMI its fields may have.
Resolves:
PR tree-optimization/84079 - missing -Warray-bounds taking the address of past-the-end element of a multidimensional array
gcc/ChangeLog:
PR tree-optimization/84079
* gimple-array-bounds.cc (array_bounds_checker::check_addr_expr):
Only allow just-past-the-end references for the most significant
array bound.
gcc/testsuite/ChangeLog:
PR tree-optimization/84079
* gcc.dg/Warray-bounds-62.c: New test.
This patch makes the -falign-foo=0 work as described in the
documentation. Thanks for all the suggestions.
Changelog:
2020-07-27 Hu Jiangping <hujiangping@cn.fujitsu.com>
PR driver/96247
* opts.c (check_alignment_argument): Set the -falign-Name
on/off flag on and set the -falign-Name string value null,
when the command-line specified argument is zero.
We were scanning only function literals with closures, but not all
function literals have closures.
Discovered because compiler failed building 1.15rc1, as there is a
function literal in the runtime package (p1 in hexdumpWords) that has
no closure and, without escape analysis, was forcing a variable to the
heap which is not permitted in the runtime.
Reviewed-on: https://go-review.googlesource.com/c/gofrontend/+/244802
Attach and detach operations are not supposed to affect structural or
dynamic reference counts for OpenACC. Previously they did so, which led to
subtle problems in some circumstances. We can avoid reference-counting
attach/detach operations by extending and slightly repurposing the
do_detach field in target_var_desc. It is now called is_attach to better
reflect its new role.
2020-07-27 Julian Brown <julian@codesourcery.com>
Thomas Schwinge <thomas@codesourcery.com>
libgomp/
* libgomp.h (struct target_var_desc): Rename do_detach field to
is_attach.
* oacc-mem.c (goacc_exit_datum_1): Add assert. Don't set finalize for
GOMP_MAP_FORCE_DETACH. Update checking to use is_attach field.
(goacc_enter_data_internal): Don't affect reference counts
for attach mappings.
(goacc_exit_data_internal): Don't affect reference counts for detach
mappings.
* target.c (gomp_map_vars_existing): Don't affect reference counts for
attach mappings.
(gomp_map_vars_internal): Set renamed is_attach flag unconditionally to
mark attach mappings.
(gomp_unmap_vars_internal): Use is_attach flag to prevent affecting
reference count for attach mappings.
* testsuite/libgomp.oacc-c-c++-common/mdc-refcount-1.c: New test.
* testsuite/libgomp.oacc-c-c++-common/mdc-refcount-2.c: New test.
* testsuite/libgomp.oacc-c-c++-common/mdc-refcount-2.c: New test.
* testsuite/libgomp.oacc-fortran/deep-copy-6-no_finalize.F90: Mark
test as shouldfail.
* testsuite/libgomp.oacc-fortran/deep-copy-6.f90: Adjust to fail
gracefully in no-finalize mode.
Co-Authored-By: Thomas Schwinge <thomas@codesourcery.com>
Also fix the tests that fail on targets without uselocale.
libstdc++-v3/ChangeLog:
* src/c++17/floating_from_chars.cc (from_chars_impl): Ensure
that FE_NEAREST is used.
* testsuite/20_util/from_chars/4.cc: Do not use if constexpr in
a { target c++14 } test.
[!_GLIBCXX_HAVE_USELOCALE]: Disable all tests.
* testsuite/20_util/from_chars/5.cc [!_GLIBCXX_HAVE_USELOCALE]:
Likewise.
* testsuite/20_util/from_chars/6.cc: New test.
gcc/ChangeLog:
PR tree-optimization/96058
* expr.c (string_constant): Build string_constant only
for a type that has same precision as char_type_node
and is an integral type.
I didn't find anything using EDGE_DFS_BACK in RTL infrastructure
used by var-tracking or var-tracking itself, the following removes
this computation.
2020-07-24 Richard Biener <rguenther@suse.de>
* var-tracking.c (variable_tracking_main_1): Remove call
to mark_dfs_back_edges.
gcc/ChangeLog:
PR tree-optimization/96128
* tree-vect-generic.c (expand_vector_comparison): Do not expand
vector comparison with VEC_COND_EXPR.
gcc/testsuite/ChangeLog:
PR tree-optimization/96128
* gcc.target/s390/vector/pr96128.c: New test.
Mixing -fcf-protection and -fcf-protection=none objects are allowed.
Linker just merges -fcf-protection values from all input objects.
Add -fcf-protection=check for the final link with LTO. An error is
issued if LTO object files are compiled with different -fcf-protection
values. Otherwise, -fcf-protection=check is ignored at the compile
time. Without explicit -fcf-protection at link time, -fcf-protection
values from LTO object files are merged at the final link.
gcc/
PR bootstrap/96203
* common.opt: Add -fcf-protection=check.
* flag-types.h (cf_protection_level): Add CF_CHECK.
* lto-wrapper.c (merge_and_complain): Issue an error for
mismatching -fcf-protection values with -fcf-protection=check.
Otherwise, merge -fcf-protection values.
* doc/invoke.texi: Document -fcf-protection=check.
gcc/testsuite/
PR bootstrap/96203
* gcc.target/i386/pr96203-1.c: New test.
* gcc.target/i386/pr96203-2.c: Likewise.
gcc/ada/
* sem_res.adb (Resolve_Actuals): Restrict the check on matching
aliased components to view conversions of array types that are
not placed in an instance. In such case at runtime an object is
created.
* sem_util.ads (Is_Actual_In_Out_Parameter, Is_View_Conversion):
New subprograms.
* sem_util.adb (Is_Actual_In_Out_Parameter, Is_View_Conversion):
New subprograms.
gcc/ada/
* exp_ch6.adb (Expand_Call_Helper): Modify addition of the extra
accessibility parameter to take into account the extra
accessibility of formals within the calling subprogram.
gcc/ada/
* errout.ads, errout.adb (Error_Msg_Ada_2020_Feature): New
procedure analogous to Error_Msg_Ada_2012_Feature.
* sem_attr.adb (Analyze_Image_Attribute): Use
Error_Msg_Ada_2012_Feature and Error_Msg_Ada_2020_Feature to
indicate that Object'Image is allowed in Ada 2012, and that
'Image is allowed for any type in Ada 2020.
gcc/ada/
* libgnat/a-strunb.adb (Sum, Mul, Saturated_Sum, Saturated_Mul):
New routines. Use them when resulting string size more that
length of the strings in parameters.
(Unbounded_Slice): Use "- 1" instead of "+ 1" in opposite side
of condition to avoid overflow.
* libgnat/a-strunb__shared.adb (Sum, Mul): New routines.
(Allocate): New routine with 2 parameters. Use routine above
when resulting string size more that length of the strings in
parameters.
(Aligned_Max_Length): Do not try to align to more than Natural'Last.
(Unbounded_Slice): Use "- 1" instead of "+ 1" in opposite side
of condition to avoid overflow.
gcc/ada/
* sem_prag.adb (Analyze_Pragma, Pragma_Max_Entry_Queue_Length):
Refine error message to indicate that the pragma must apply to
an entry declaration, not just an entry.
gcc/ada/
* exp_ch6.adb (Make_Build_In_Place_Iface_Call_In_Allocator):
Revert previous patch, and add a missing type conversion to
displace the pointer to the allocated object to reference the
target dispatch table.
gcc/ada/
* sem_res.adb (Resolve_Actuals): Restore restrictive check on
view conversions which required matching value of
Has_Aliased_Components of formals and actuals. Required to avoid
the regression of ACATS b460005.
gcc/ada/
* sem_ch12.adb (Instantiate_Package_Body): Add commentary for a
nesting issue with parent handling and private view switching.
(Switch_View): Do not skip specific private-dependent subtypes.
Hans-Peter Nilsson
GCC Administrator
Sergei Trofimovich
Roger Sayle
Nathan Sidwell
Jakub Jelinek
Tom de Vries
François Dumont
Kewen Lin
Ian Lance Taylor
Joseph Myers
Martin Sebor
Hu Jiangping
Julian Brown
Jonathan Wakely
Martin Liska
Richard Biener
H.J. Lu
Alexandre Oliva
Javier Miranda
Arnaud Charlet
Justin Squirek
Bob Duff
Ghjuvan Lacambre
Dmitriy Anisimkov
Gary Dismukes
Eric Botcazou