This patch allows C-style casting between fixed-size and scalable
vectors. This kind of cast was previously blocked by the compiler, but
it should be allowed.
Differential Revision: https://reviews.llvm.org/D91262
According to ELF v2 ABI, both IEEE 128-bit and IBM extended floating
point variables should be quad-word (16 bytes) aligned. Previously, only
vector types are considered aligned as quad-word on PowerPC.
This patch will fix incorrectness of IEEE 128-bit float argument in
va_arg cases.
Reviewed By: rjmccall
Differential Revision: https://reviews.llvm.org/D91596
Summary:
This patch adds support for passing in the original delcaration name in the source file to the libomptarget runtime. This will allow the runtime to provide more intelligent debugging messages. This patch takes the original expression parsed from the OpenMP map / update clause and provides a textual representation if it was explicitly mapped, otherwise it takes the name of the variable declaration as a fallback. The information in passed to the runtime in a global array of strings that matches the existing ident_t source location strings using ";name;filename;column;row;;"
Reviewers: jdoerfert
Differential Revision: https://reviews.llvm.org/D89802
In some cases, when deserializing a `CXXMethodDecl` of a `CXXSpecializationTemplateDecl`,
the call to `FunctionDecl::setPure()` happens before the `DefinitionData` member has been
populated (which appears to happen lower down in a `mergeRedeclarable` call), causing a
crash (https://reviews.llvm.org/P8228).
This diff fixes this by deferring the `FunctionDecl::setPure()` till after the `DefinitionData` has
been filled in.
Reviewed By: lxfind
Differential Revision: https://reviews.llvm.org/D86853
This patch creates a SystemZ folder in clang/test/CodeGen to contain systemz-related lit tests.
Reviewed By: muiez
Differential Revision: https://reviews.llvm.org/D91628
Adding features in OpenMP 5.1 specification, as documented in feature change history, to the 5.1 table. I alphabetized the rows of the table according to the category. For deprecating master construct, I just used 'other' as the category.
Reviewed By: jdoerfert
Differential Revision: https://reviews.llvm.org/D90802
The compiler should treat array subscript with base pointer as a first
pointer in complex data, it is used only for member expression with base
pointer.
Differential Revision: https://reviews.llvm.org/D91660
Fix a crash when evaluating a constexpr function which contains
recovery-exprs. https://bugs.llvm.org/show_bug.cgi?id=46837
Would be nice to have constant expression evaluator support general template
value-dependent expressions, but it requires more work.
This patch is a good start I think, to handle the error-only
value-dependent expressions.
Differential Revision: https://reviews.llvm.org/D84637
Baremetal toolchain is not adding sysroot/lib to the library
search path. This is forcing the user to do it manually. This commit
fixes this shortcoming by adding the sysroot/lib to library search path
if sysroot is not empty.
Reviewed By: jroelofs
Differential Revision: https://reviews.llvm.org/D91559
Matrix types in memory are represented as arrays, but accessed through
vector pointers, with the alignment specified on the access operation.
For inline assembly, update pointer arguments to use vector pointers.
Otherwise there will be a mis-match if the matrix is also an
input-argument which is represented as vector.
Reviewed By: nickdesaulniers
Differential Revision: https://reviews.llvm.org/D91631
- The new option, -arcmt-action, is a simple enum based option.
- The driver is modified to translate the existing -ccc-acmt-* options accordingly
Depends on D83298
Reviewed By: Bigcheese
Original patch by Daniel Grumberg.
Differential Revision: https://reviews.llvm.org/D83315
As with precompiled headers, it's useful for indexers to be able to
continue through compiler errors in dependent modules.
Resolves rdar://69816264
Reviewed By: akyrtzi
Differential Revision: https://reviews.llvm.org/D91580
The tests don't specify a triple in some cases, since they shouldn't be
necessary, so I've updated the tests to detect via macro when they are
running on win32 to give the slightly altered diagnostic.
If the variable is implicitly firstprivatized in the inner task-based
region, it also must be firstprivatized in outer task-based regions.
Previously firstprivates were captured in tasks but later it was
optimized to reduce the memory usage. But still need to mark such
variables as implicit firstprivate in outer tasks.
Differential Revision: https://reviews.llvm.org/D91627
Don't match Stmt or Decl nodes not spelled in the source when using
TK_IgnoreUnlessSpelledInSource. This prevents accidental modification
of source code at incorrect locations.
Differential Revision: https://reviews.llvm.org/D90984
Update the ASTNodeTraverser to dump only nodes spelled in source. There
are only a few which need to be handled, but Decl nodes for which
isImplicit() is true are handled together.
Update the RAV instances used in ASTMatchFinder to ignore the nodes too.
As with handling of template instantiations, it is necessary to allow
the RAV to process the implicit nodes because they need to be visitable
before the first traverse() matcher is encountered. An exception to
this is in the MatchChildASTVisitor, because we sometimes wish to make a
node matchable but make its children not-matchable. This is the case
for defaulted CXXMethodDecls for example.
Extend TransformerTests to illustrate the kinds of problems that can
arise when performing source code rewriting due to matching implicit
nodes.
This change accounts for handling nodes not spelled in source when using
direct matching of nodes, and when using the has() and hasDescendant()
matchers. Other matchers such as
cxxRecordDecl(hasMethod(cxxMethodDecl())) still succeed for
compiler-generated methods for example after this change. Updating the
implementations of hasMethod() and other matchers is for a follow-up
patch.
Differential Revision: https://reviews.llvm.org/D90982
If the data member pointer is mapped, the compiler tries to optimize the
mapping of such data by discarding explicit mapping flags and trying to
emit combined data instead. In some cases, this optimization is not
quite correctly implemented and it leads to a program crash at the
runtime. Instead, if the data member is mapped, just emit it as is and
do not emit combined mapping flags for it.
Differential Revision: https://reviews.llvm.org/D91552
This patch adds a new test cases which uses a matrix value as memory
inline assembly argument. Currently the pointer element type does not
match the vector type.
Lax vector conversions was behaving incorrectly for implicit casts
between scalable and fixed-length vector types. For example, this:
#include <arm_sve.h>
#define N __ARM_FEATURE_SVE_BITS
#define FIXED_ATTR __attribute__((arm_sve_vector_bits(N)))
typedef svfloat32_t fixed_float32_t FIXED_ATTR;
void allowed_depending() {
fixed_float32_t fs32;
svfloat64_t s64;
fs32 = s64;
}
... would fail because the vectors have differing lane sizes. This patch
implements the correct behaviour for
-flax-vector-conversions={none,all,integer}. Specifically:
- -flax-vector-conversions=none prevents all lax vector conversions
between scalable and fixed-sized vectors.
- -flax-vector-conversions=integer allows lax vector conversions between
scalable and fixed-size vectors whose element types are integers.
- -flax-vector-conversions=all allows all lax vector conversions between
scalable and fixed-size vectors (including those with floating point
element types).
The implicit conversions are implemented as bitcasts.
Reviewed By: fpetrogalli
Differential Revision: https://reviews.llvm.org/D91067
In the wake of https://reviews.llvm.org/D89559, we discovered that a
couple of tests (the ones modified below to have additional triple
versions) would fail on Win32, for 1 of two reasons. We seem to not
have a win32 buildbot anymore, so the triple is to make sure this
doesn't get broken in the future.
First, two of the three 'note-candidate' functions weren't appropriately
skipping the remaining conversion functions.
Second, in 1 situation (note surrogate candidates) we actually print the
type of the conversion operator. The two tests that ran into that
needed updating to make sure it printed the proper one in the win32
case.
opencl-c.h disables all extensions at its end, but opencl-c-base.h
does not, and that causes any inclusion of only opencl-c-base.h to
leave some extensions (such as cl_khr_fp16) enabled. This affects the
-fdeclare-opencl-builtins option for example.
This violates the OpenCL Extension Specification which specifies that
"The initial state of the compiler is as if the directive #pragma
OPENCL EXTENSION all : disable was issued".
Fix by disabling all extensions at the end of opencl-c-base.h and
enable extensions inside opencl.h which relied on opencl-c-base.h
enabling the cl_khr_fp16/64 extensions.
Differential Revision: https://reviews.llvm.org/D91429
The option -fproc-stat-report=<file> makes driver to generate child
process resource comsumption report. In the report executable name was
not quoted and it made parsing the report more difficult. With this
change the executable name is surrounded by quotation marks.
Add an option -munsafe-fp-atomics for AMDGPU target.
When enabled, clang adds function attribute "amdgpu-unsafe-fp-atomics"
to any functions for amdgpu target. This allows amdgpu backend to use
unsafe fp atomic instructions in these functions.
Differential Revision: https://reviews.llvm.org/D91546
arguments.
* Adds 'nonnull' and 'dereferenceable(N)' to 'this' pointer arguments
* Gates 'nonnull' on -f(no-)delete-null-pointer-checks
* Introduces this-nonnull.cpp and microsoft-abi-this-nullable.cpp tests to
explicitly test the behavior of this change
* Refactors hundreds of over-constrained clang tests to permit these
attributes, where needed
* Updates Clang12 patch notes mentioning this change
Reviewed-by: rsmith, jdoerfert
Differential Revision: https://reviews.llvm.org/D17993
- In certain cases, a generic pointer could be assumed as a pointer to
the global memory space or other spaces. With a dedicated target hook
to query that address space from a given value, infer-address-space
pass could infer and propagate that to all its users.
Differential Revision: https://reviews.llvm.org/D91121
In order to support attribute((constructor)) and attribute((destructor)),
which is used by various LLVM non-C++ runtime components, AIX will include
crti[_64].o and -bcdtors for C language link invocations by default.
Differential Revision: https://reviews.llvm.org/D91361
This patch adds a new pass to add !annotation metadata for entries in
@llvm.global.anotations, which is generated using
__attribute__((annotate("_name"))) on functions in Clang.
This has been discussed on llvm-dev as part of
RFC: Combining Annotation Metadata and Remarks
http://lists.llvm.org/pipermail/llvm-dev/2020-November/146393.html
Reviewed By: thegameg
Differential Revision: https://reviews.llvm.org/D91195
Adapt the declarations of `svpattern` and `svprfop` to the most recent
one defined in section "5. Enum declarations" of the SVE ACLE
specifications [1].
The signature of the intrinsics using these enums have been changed
accordingly.
A test has been added to make sure that `svpattern` and `svprfop` are
not typedefs.
[1] https://developer.arm.com/documentation/100987/latest, version
00bet6
Reviewed By: joechrisellis
Differential Revision: https://reviews.llvm.org/D91333
Original commit message: "
Move the test compiler setup in a common place. NFCI
This patch reduces the copy paste in the unittest/CodeGen folder by moving the
common compiler setup phase in a header file.
Differential revision: https://reviews.llvm.org/D91061
"
This patch includes a fix for the memory leaks pointed out by @vitalybuka
This patch updates Clang's IRGen to add !annotation nodes with an
"auto-init" annotation to all stores for auto-initialization.
As discussed in 'RFC: Combining Annotation Metadata and Remarks'
(http://lists.llvm.org/pipermail/llvm-dev/2020-November/146393.html)
this allows using optimization remarks to track down where auto-init
code was inserted (and not removed by optimizations).
There are a few cases in the tests where !annotation gets dropped by
optimizations. Those optimizations will be updated in subsequent
patches.
This patch is based on a patch by Francis Visoiu Mistrih.
Reviewed By: thegameg, paquette
Differential Revision: https://reviews.llvm.org/D91417
This enables automatically parsing and generating CC1 arguments for options where two flags control the same field, e.g. -fexperimental-new-pass-manager and -fno-experimental new pass manager.
Reviewed By: Bigcheese, dexonsmith
Original patch by Daniel Grumberg.
Differential Revision: https://reviews.llvm.org/D83071
parameters.
It appears that LLVM isn't able to generate a DW_AT_const_value for a
constant of class type, but if it could, we'd match GCC's debug info in
this case, and in the interim we no longer crash.
Reviewed here: https://reviews.llvm.org/D91409 by Aaron.
Highlights of the review:
- avoid an underlying type for enums
- avoid enum bit fields (MSVC packing anomalies) and favor static_casts to unsigned bit-fields
Patch by Thorsten Schuett <schuett@gmail.com> w some minor fixes in SemaType.cpp where a couple asserts had to be repaired to deal with lack of implicit coversion to int.
Thanks Thorsten!
As Richard Smith pointed out in the review of D90123, both the C and C++
standard call it lvalue and rvalue, so let's stick to the same spelling
in Clang.
When an overloaded member function has a ref-qualifier, like:
class X {
void f() &&;
void f(int) &;
};
we would print strange notes when the ref-qualifier doesn't fit the value
category:
X x;
x.f();
X().f(0);
would both print a note "no known conversion from 'X' to 'X' for object
argument" on their relevant overload instead of pointing out the
mismatch in value category.
At first I thought the solution is easy: just use the FailureKind member
of the BadConversionSequence struct. But it turns out that we weren't
properly setting this for function arguments. So I went through
TryReferenceInit to make sure we're doing that right, and found a number
of notes in the existing tests that improved as well.
Fixes PR47791.
Reviewed By: rsmith
Differential Revision: https://reviews.llvm.org/D90123
This patch reduces the copy paste in the unittest/CodeGen folder by moving the
common compiler setup phase in a header file.
Differential revision: https://reviews.llvm.org/D91061
Clang supports compiling CUDA source files,
CUDA header files may contain CUDA specific code
that is why they have special extension, which
can be recognized by nvcc (CUDA compiler driver)
as CUDA source file.
Format them by default as well.
Reviewed By: MyDeveloperDay
Patch By: tomilov
Differential Revision: https://reviews.llvm.org/D90780
See discussion in https://bugs.llvm.org/show_bug.cgi?id=45073 / https://reviews.llvm.org/D66324#2334485
the implementation is known-broken for certain inputs,
the bugreport was up for a significant amount of timer,
and there has been no activity to address it.
Therefore, just completely rip out all of misexpect handling.
I suspect, fixing it requires redesigning the internals of MD_misexpect.
Should anyone commit to fixing the implementation problem,
starting from clean slate may be better anyways.
This reverts commit 7bdad08429,
and some of it's follow-ups, that don't stand on their own.
The code has a few sequence that looked like:
Ops.push_back(Ops[0]);
Ops.erase(Ops.begin());
And are equivalent to:
std::rotate(Ops.begin(), Ops.begin() + 1, Ops.end());
The latter has the advantage of never reallocating the vector, which
would be a bug in the original code as push_back would read from the
memory it deallocated.
Since RPATH initialization was disabled for the runtime libraries to
avoid overwriting RPATH unconditionally we need to explicity set up it
for the Win to Arm Linux cross builds.
See some details here: https://reviews.llvm.org/D91099
Make it required. Since it's a module pass, optnone won't test it, so
extend the clang test to also use opt-bisect now that it's supported.
14/16 check-dfsan tests failed with NPM enabled, now all pass.
Reviewed By: leonardchan
Differential Revision: https://reviews.llvm.org/D91385
Clean up the logic for `err_fe_{pch,module,ast}_file_modified` to use a
`select` like other ASTReader diagnostics. There should be no
functionality change here, just a cleanup.
Differential Revision: https://reviews.llvm.org/D91367
This RUN line was added as a temporary measure to undo the damage done
by D79655. Everyone's build system should be fine by now.
Reviewed By: tlively
Differential Revision: https://reviews.llvm.org/D91448
All three diagnostics have a select between "PCH", "module", and "AST"
in the text. The most generic of these is "AST", so rename them from
`err_module_...` to `err_ast_...`.
Differential Revision: https://reviews.llvm.org/D91436
This logic seems easier to follow without the `Error()` helper, and
checking `DiagnosticsEngine::isDiagnosticInFlight` just once up front.
Differential Revision: https://reviews.llvm.org/D91366
Just skip (non-bitfield) zero-sized fields, like we do with empty bases.
The class->struct conversion in the test is because -std=c++20 else deletes some default methods
due to non-accessible base dtors otherwise.
As a side-effect of writing the test, I discovered that D76801 did an ABI breaking change of sorts
for Objective-C's @encode. But it's been in for a while, so I'm not sure if we want to row back on
that or now.
Fixes PR48048.
Differential Revision: https://reviews.llvm.org/D90622
`TD->getTemplatedDecl()` might not be a DeclContext variant, which can
trigger an assertion inside `isa<>`.
Differential Revision: https://reviews.llvm.org/D91380
Merge existing marhsalling info kinds and add some primitives to
express flag options that contribute to a bitfield.
Depends on D82574
Original patch by Daniel Grumberg.
Reviewed By: Bigcheese
Differential Revision: https://reviews.llvm.org/D82860
The new option `-fproc-stat-info=<file>` can be used to generate report
about used memory and execution tile of each stage of compilation.
Documentation for this option can be found in `UserManual.rst`. The
option can be used in parallel builds.
Differential Revision: https://reviews.llvm.org/D78903
- If an aggregate argument is indirectly accessed within kernels, direct
passing results in unpromotable `alloca`, which degrade performance
significantly. InferAddrSpace pass is enhanced in
[D91121](https://reviews.llvm.org/D91121) to take the assumption that
generic pointers loaded from the constant memory could be regarded
global ones. The need for the coercion on aggregate arguments is
mitigated.
Differential Revision: https://reviews.llvm.org/D89980
This was motivated by changes to llvm's `not --crash` disabling symbolization
but I ended up removing `not` from the script entirely because it
returns differently depending on whether clang "crashes" or exits for some
other reason. The script had to choose between calling `not` and `not --crash`
and sometimes it was wrong.
The script also now disables symbolization when we don't read the stack
trace because symbolizing is kind of slow.
Differential Revision: https://reviews.llvm.org/D91372
VE needs to support integrated assembler and "nas". This "nas"
doesn't recognize ".sigaddr" pseudo mnemonics, so need to disable
it. This patch disable it on VE by default. Also add a regression
test for that.
Reviewed By: simoll
Differential Revision: https://reviews.llvm.org/D91350
This removes lots of duplicated code which was necessary before
https://reviews.llvm.org/D89158.
Now we can use PassBuilder::runRegisteredEPCallbacks().
This is mostly sanitizers.
There is likely more that can be done to simplify, but let's start with this.
Reviewed By: ychen
Differential Revision: https://reviews.llvm.org/D90870
Need to check if there are map types for the components before trying to
access them when trying to modify type mappings for combined partial
mappings.
Differential Revision: https://reviews.llvm.org/D91370
to synchronize with tools/clang-format/git-clang-format
tra: Keeping them in sync does have a minor benefit of not raising a question why the two maps are different.
Differential Revision: https://reviews.llvm.org/D91034
There are two factory functions used to create a semantic attribute,
Create() and CreateImplicit(). CreateImplicit() does not need to
specify the source range of the attribute since it's an implicitly-
generated attribute. The same logic does not apply to Create(), so
this removes the default argument from those declarations to avoid
accidentally creating a semantic attribute without source location
information.
As of D80952 we are disabling strict floating point on all hosts except
those that are explicitly listed as supported. Use of strict floating point
on other hosts requires use of the -fexperimental-strict-floating-point
flag. This is to avoid bugs like "https://bugs.llvm.org/show_bug.cgi?id=45329"
(which has an incorrect link in the previous review).
In the review for D80952 I was asked to mark the -fexperimental option as
a MarshallingInfoFlag. This patch does exactly that.
Differential Revision: https://reviews.llvm.org/D88987
We want to allow using MMA on P10 CPU only. This patch prevents the use of MMA
with the -mmma option on P9 CPUs and earlier.
Differential Revision: https://reviews.llvm.org/D91200
For dllexported default constructors with default arguments, we export
default constructor closures which pass in the default args. (See D8331
for a good explanation.)
For templates, that means those default args must be instantiated even
if the function isn't called. That is done by the
InstantiateDefaultCtorDefaultArgs() function, but it wasn't done for
explicit specializations, causing asserts (see bug).
Differential revision: https://reviews.llvm.org/D91089
We have option -mabi=ieeelongdouble to set current long double to
IEEEquad semantics. Like what GCC does, we need to define
__LONG_DOUBLE_IEEE128__ macro in this case, and __LONG_DOUBLE_IBM128__
if using PPCDoubleDouble.
Reviewed By: steven.zhang
Differential Revision: https://reviews.llvm.org/D90208
Non-mechanical changes:
- Added FIXME to StringLiteral to cover multi-token string literals.
- LiteralExpression::getLiteralToken() is gone. (It was never called)
This is because we don't codegen methods in Alternatives
It's conceptually suspect if we consider multi-token string literals, though.
Differential Revision: https://reviews.llvm.org/D91277
Some targets may add required passes via
TargetMachine::registerPassBuilderCallbacks(). We need to run those even
under -O0. As an example, BPFTargetMachine adds
BPFAbstractMemberAccessPass, a required pass.
This also allows us to clean up BackendUtil.cpp (and out-of-tree Rust
usage of the NPM) by allowing us to share added passes like coroutines
and sanitizers between -O0 and other optimization levels.
Since callbacks may end up not adding passes, we need to check if the
pass managers are empty before adding them, so PassManager now has an
isEmpty() function. For example, polly adds callbacks but doesn't always
add passes in those callbacks, so this is necessary to keep
-debug-pass-manager tests' output from changing depending on if polly is
enabled or not.
Tests are a continuation of those added in
https://reviews.llvm.org/D89083.
Reviewed By: asbirlea, Meinersbur
Differential Revision: https://reviews.llvm.org/D89158
except where they are necessary to disambiguate the target.
This substantially improves diagnostics from the standard library,
which are otherwise full of `::__1::` noise.
arguments of types by default.
This somewhat improves the worst-case printing of types like
std::string, std::vector, etc., where many irrelevant default arguments
can be included in the type as printed if we've lost the type sugar.
Avoid requiring an actual MemoryBuffer in ComputePreambleBounds, when
a MemoryBufferRef will do just fine.
Differential Revision: https://reviews.llvm.org/D90890
The Itanium CXX ABI grammer has been extended to support parameterized
vendor extended types [1].
This patch updates Clang's mangling for matrix types to use the new
extension.
[1] b359d28971
Reviewed By: rjmccall
Differential Revision: https://reviews.llvm.org/D91253
Support a vector register constraint in inline asm of clang.
Add a regression test also.
Reviewed By: simoll
Differential Revision: https://reviews.llvm.org/D91251
This reverts commit 09248a5d25.
Some builds are broken. I suspect a `static constexpr` in a class missing a
definition out of class (required pre-c++17).
Noticed while fixing unused prefix warnings - there isn't actually any diff in the loop unrolled ir between old/new pass managers any more, so the broken checks were superfluous
Similar to the previous patch, this doesn't convert *all* the classes that
could be converted. It also doesn't enforce any new invariants etc.
It *does* include some data we don't use yet: specific token types that are
allowed and optional/required status of sequence items. (Similar to Dmitri's
prototype). I think these are easier to add as we go than later, and serve
a useful documentation purpose.
Differential Revision: https://reviews.llvm.org/D90659
`-###` has always been supported in the new flang driver. This patch
merely makes sure that it's included when printing the help screen (i.e.
`flang-new -help`).
Merge existing marhsalling info kinds and add some primitives to
express flag options that contribute to a bitfield.
Depends on D82574
Reviewed By: Bigcheese
Differential Revision: https://reviews.llvm.org/D82860
This defines two node archetypes with trivial class definitions:
- Alternatives: the generated abstract classes are trivial as all
functionality is via LLVM RTTI
- Unconstrained: this is a placeholder, I think all of these are going to be
Lists but today they have no special accessors etc, so we just say
"could contain anything", and migrate them one-by-one to Sequence later.
Compared to Dmitri's prototype, Nodes.td looks more like a class hierarchy and
less like a grammar. (E.g. variants list the Alternatives parent rather than
vice versa).
The main reasons for this:
- the hierarchy is an important part of the API we want direct control over.
- e.g. we may introduce abstract bases like "loop" that the grammar doesn't
care about in order to model is-a concepts that might make refactorings
more expressive. This is less natural in a grammar-like idiom.
- e.g. we're likely to have to model some alternatives as variants and others
as class hierarchies, the choice will probably be based on natural is-a
relationships.
- it reduces the cognitive load of switching from editing *.td to working with
code that uses the generated classes
Differential Revision: https://reviews.llvm.org/D90543
In JavaScript breaking before a `@tag` in a comment puts it on a new line, and
machinery that parses these comments will fail to understand such comments.
This adapts clang-format to not break before `@`. Similar functionality exists
for not breaking before `{`.
Reviewed By: mprobst
Differential Revision: https://reviews.llvm.org/D91078
Since these are scoped enumerators, they have to be prefixed by DeclaratorContext, so lets remove Context from the name, and return some characters to the multiverse.
Patch was reviewed here: https://reviews.llvm.org/D91011
Thank you to aaron, bruno, wyatt and barry for indulging me.
This header has long lacked a standard multiple inclusion guard
like other headers have, for no apparent reason. The GCC header
of the same name likewise lacks one up through release 10.1, but
trunk GCC (release 11, and perhaps future 10.x) has fixed it
(see https://gcc.gnu.org/bugzilla/show_bug.cgi?id=96238).
Reviewed By: phosek
Differential Revision: https://reviews.llvm.org/D91226
This enables a method sending an autorelease message to an object and
returning the object in MRR to avoid adding the object to an autorelease
pool if a call to objc_retainAutoreleasedReturnValue in the caller
function accepts the hand off of the retain count.
rdar://problem/50678052
Differential Revision: https://reviews.llvm.org/D91111
The original bug was discovered in T75057860. Clang front-end emits an AST that looks like this for an co_await expression:
|- ExprWithCleanups
|- -CoawaitExpr
|- -MaterializeTemporaryExpr ... Awaiter
...
|- -CXXMemberCallExpr ... .await_ready
...
|- -CallExpr ... __builtin_coro_resume
...
|- -CXXMemberCallExpr ... .await_resume
...
ExprWithCleanups is responsible for cleaning up (including calling dtors) for the temporaries generated in the wrapping expression).
In the above structure, the __builtin_coro_resume part (which corresponds to the code for the suspend case in the co_await with symmetric transfer), the pseudocode looks like this:
__builtin_coro_resume(
awaiter.await_suspend(
from_address(
__builtin_coro_frame())).address());
One of the temporaries that's generated as part of this code is the coroutine handle returned from awaiter.await_suspend() call. The call returns a handle which is a prvalue (since it's a returned value on the fly). In order to call the address() method on it, it needs to be converted into an xvalue. Hence a materialized temp is created to hold it. This temp will need to be cleaned up eventually. Now, since all cleanups happen at the end of the entire co_await expression, which is after the <coro.suspend> suspension point, the compiler will think that such a temp needs to live across suspensions, and need to be put on the coroutine frame, even though it's only used temporarily just to call address() method.
Such a phenomena not only unnecessarily increases the frame size, but can lead to ASAN failures, if the coroutine was already destroyed as part of the await_suspend() call. This is because if the coroutine was already destroyed, the frame no longer exists, and one can not store anything into it. But if the temporary object is considered to need to live on the frame, it will be stored into the frame after await_suspend() returns.
A fix attempt was done in https://reviews.llvm.org/D87470. Unfortunately it is incorrect. The reason is that cleanups in Clang works more like linearly than nested. There is one current state indicating whether it needs cleanup, and an ExprWithCleanups resets that state. This means that an ExprWithCleanups must be capable of cleaning up all temporaries created in the wrapping expression, otherwise there will be dangling temporaries cleaned up at the wrong place.
I eventually found a walk-around (https://reviews.llvm.org/D89066) that doesn't break any existing tests while fixing the issue. But it targets the final co_await only. If we ever have a co_await that's not on the final awaiter and the frame gets destroyed after suspend, we are in trouble. Hence we need a proper fix.
This patch is the proper fix. It does the folllowing things to fully resolve the issue:
1. The AST has to be generated in the order according to their nesting relationship. We should not generate AST out of order because then the code generator would incorrectly track the state of temporaries and when a cleanup is needed. So the code in buildCoawaitCalls is reorganized so that we will be generating the AST for each coawait member call in order along with their child AST.
2. await_ready() call is wrapped with an ExprWithCleanups so that temporaries in it gets cleaned up as early as possible to avoid living across suspension.
3. await_suspend() call is wrapped with an ExprWithCleanups if it's not a symmetric transfer. In the case of a symmetric transfer, in order to maintain the musttail call contract, the ExprWithCleanups is wraaped before the resume call.
4. In the end, we mark again that it needs a cleanup, so that the entire CoawaitExpr will be wrapped with a ExprWithCleanups which will clean up the Awaiter object associated with the await expression.
Differential Revision: https://reviews.llvm.org/D90990
In C++11 standard, to become implicitly movable, the expression in return
statement should be a non-volatile automatic object. CWG1579 changed the rule
to require that the expression only needs to be an automatic object. C++14
standard and C++17 standard kept this rule unchanged. C++20 standard changed
the rule back to require the expression be a non-volatile automatic object.
This should be a typo in standards, and VD should be non-volatile.
Differential Revision: https://reviews.llvm.org/D88295
The value_type is a const pointer, which makes the iteator non-copyable.
Before the patch, the normal usage like below was illegal:
```
auto It = lookupresult.begin();
...
It = lookupresult.end(); // the copy is not allowed.
```
Differential Revision: https://reviews.llvm.org/D91158
Jan Svoboda