Previously we were calling glrRecover() ad-hoc at the end of input.
Two main problems with this:
- glrRecover() on two separate code paths is inelegant
- We may have to recover several times in succession (e.g. to exit from
nested scopes), so we need a loop at end-of-file
Having an actual shift action for an EOF terminal allows us to handle
both concerns in the main shift/recover/reduce loop.
This revealed a recovery design bug where recovery could enter a loop by
repeatedly choosing the same parent to identically recover from.
Addressed this by allowing each node to be used as a recovery base once.
Differential Revision: https://reviews.llvm.org/D130550
The patch mainly focuses on the no warnings for -Wtautological-compare.
It work fine for the positive numbers but doesn't for the negative
numbers. This is because the warning explicitly checks for an
IntegerLiteral AST node, but -1 is represented by a UnaryOperator with
an IntegerLiteral sub-Expr.
Fixes#42918
Differential Revision: https://reviews.llvm.org/D130510
Whilst writing a patch to add extra tail-folding RUN lines to
existing tests I noticed a few areas where they can be
cleaned up a little:
1. scalable-reductions.ll: fmin_fast does not mark fcmp as fast.
2. sve-inductions-unusual-types.ll: remove direct references to
SSA variable names.
3. sve-strict-fadd-cost.ll: don't force vector width so we see
costs for different VFs in one go. This will be important for
the follow-on patch.
4. sve-vector-reverse.ll,vector-reverse-mask4.ll: add noalias
keyword to simplify IR.
4. sve-widen-gep.ll,sve-widen-phi.ll: regenerate using script.
These changes will make the subsequent patch adding RUN lines much
easier to review!
Differential Revision: https://reviews.llvm.org/D132219
This aligns the ``heap[]`` array in ``fallback_malloc.cpp`` to ensure
that it can be safely cast to ``heap_node*``, and also adjusts the
allocation algorithm to ensure that every allocated block has the
alignment requested by ``__attribute__((aligned))``, by putting the
block's ``heap_node`` header 4 bytes before an aligned address.
Patch originally by Eric Fiselier: this is an updated version of
D12669, which was never landed.
Reviewed By: ldionne, #libc_abi
Differential Revision: https://reviews.llvm.org/D129842
Implement negating and inverting values. Also implement
IntegralToBoolean casts so the operations are easier to test.
Differential Revision: https://reviews.llvm.org/D132098
Harmonize the status and error values of Communication::Read() when
running with and without read thread. Prior to this change, Read()
would return eConnectionStatusSuccess if read thread was enabled
and the read timed out or reached end-of-file, rather than
the respective states that are returned if read thread was disabled.
Now, it correctly returns eConnectionStatusTimedOut
and eConnectionStatusEndOfFile, and sets the error respectively.
Sponsored by: The FreeBSD Foundation
Differential Revision: https://reviews.llvm.org/D132217
This was discussed on Discord with the consensus that we should rename the macros.
Reviewed By: ldionne, Mordante, var-const, avogelsgesang, jloser, #libc
Spies: libcxx-commits
Differential Revision: https://reviews.llvm.org/D131498
Certain idioms are ignored by -Wunused in header files only.
The current "is a header" check assumes that if headers are the main file, we're
building a PCH or a module or something. However in tools we may be parsing the
header in its own right, but still want to treat it as a header.
Fixes https://github.com/clangd/vscode-clangd/issues/360
Differential Revision: https://reviews.llvm.org/D129642
Our GLR uses lookahead: only perform reductions that might be consumed by the
shift immediately following. However when shift fails and so reduce is followed
by recovery instead, this restriction is incorrect and leads to missing heads.
In turn this means certain recovery strategies can't be made to work. e.g.
```
ns := NAMESPACE { namespace-body } [recover=Skip]
ns-body := namespace_opt
```
When `namespace { namespace {` is parsed, we can recover the inner `ns` (using
the `Skip` strategy to ignore the missing `}`). However this `namespace` will
not be reduced to a `namespace-body` as EOF is not in the follow-set, and so we
are unable to recover the outer `ns`.
This patch fixes this by tracking which heads were produced by constrained
reduce, and discarding and rebuilding them before performing recovery.
This is a prerequisite for the `Skip` strategy mentioned above, though there are
some other limitations we need to address too.
Reviewed By: hokein
Differential Revision: https://reviews.llvm.org/D130523
Expose these as variables as that's what the standard calls them (and D131175).
To make this work, we also fix a bug in SelectionTree: PredefinedExpr has
an implicit/invisible StringLiteral, and SelectionTree should not traverse
implicit things.
Reviewed By: ckandeler
Differential Revision: https://reviews.llvm.org/D132135
This patch fixes the following two bugs in `PPCInstrInfo::isSignOrZeroExtended` helper, which is used from sign-/zero-extension elimination in PPCMIPeephole pass.
- Registers defined by load with update (e.g. LBZU) were identified as already sign or zero-extended. But it is true only for the first def (loaded value) and not for the second def (i.e. updated pointer).
- Registers defined by ORIS/XORIS were identified as already sign-extended. But, it is not true for sign extension depending on the immediate (while it is ok for zero extension).
To handle the first case, the parameter for the helpers is changed from `MachineInstr` to a register number to distinguish first and second defs. Also, this patch moves the initialization of PPCMIPeepholePass to allow mir test case.
Reviewed By: nemanjai
Differential Revision: https://reviews.llvm.org/D40554
This patch replaces svget, svset and svcreate aarch64 intrinsics for tuple
types with the generic llvm-ir intrinsics extract/insert vector
Differential Revision: https://reviews.llvm.org/D131547
This patch switches the MainLoop class to use the WSAEventSelect
mechanism to wait for multiple sockets to become readable. The
motivation for doing that is that this allows us to wait for other kinds
of events as well (as long as they can be converted to WSAEvents). This
will allow us to avoid (abstract away) pipe-based multiplexing
mechanisms in the generic code, since pipes cannot be combined with
sockets on windows.
Since the windows implementation will now look completely different than
the posix (file descriptor-based) implementations, I have split the
MainLoop class into two (MainLoopPosix and MainLoopWindows), with the
common code going into MainLoopBase.
Differential Revision: https://reviews.llvm.org/D131159
Fix for the problem with displaying options `-fsyntax-only` in clang and flang-new in help
Fix https://github.com/llvm/llvm-project/issues/57033
Before:
``` $ clang -help | grep syntax
-objcmt-migrate-property-dot-syntax
Enable migration of setter/getter messages to property-dot syntax
```
After:
```
$ clang -help | grep syntax
-fsyntax-only Run the preprocessor, parser and semantic analysis stages
-objcmt-migrate-property-dot-syntax
Enable migration of setter/getter messages to property-dot syntax
```
Reviewed By: vzakhari, awarzynski, MaskRay, alexiprof
Differential Revision: https://reviews.llvm.org/D131808
It's unclear to me why this wasn't tagged this way already in
87fe0709d4 / D114612 where the
feature flag executor-has-no-bash was added, as this test did exist
in its current form already at that time.
Differential Revision: https://reviews.llvm.org/D131446
Increase the timeout tolerance if TEST_IS_EXECUTED_IN_A_SLOW_ENVIRONMENT
is set, similarly to how it's done in a couple other tests.
Use `std::this_thread::yield();` instead of busylooping. When multiple
threads are busylooping, it's plausible that not all threads even get
started running before the timeout runs out.
This makes the threading tests succeed if run in Windows runners on
Github Actions.
Differential Revision: https://reviews.llvm.org/D131483
This makes its role clearer. It's plausible that one may want to manually
define TEST_IS_EXECUTED_IN_A_SLOW_ENVIRONMENT when running the tests in
some environments - in particular, it seems to be necessary to use the
higher tolerance timeouts if running the tests on Windows runners
on Github Actions.
Also add the descriptive comment in one file where it was missing.
Differential Revision: https://reviews.llvm.org/D131484
The PATH is set in order to be able to find the tested DLL at runtime.
When linking statically, it's not necessary to set the PATH.
Setting PATH in the executor has the downside that it clears the
existing path (it's not prepended/appended to it), which means
that the executed tools can't find other tools - which sets the
executor-has-no-bash flag.
By removing the unnecessary setting of PATH, we have a properly
working bash even when wrapped by the executor, which gets rid
of the executor-has-no-bash flag in this test configuration, which
makes 9 more testcases be executed.
Differential Revision: https://reviews.llvm.org/D131481
Lambdas are implemented as regular classes internally,
and the captured variables end up as members there.
Do not diagnose those - the check should cover only
regular classes and structs.
Differential Revision: https://reviews.llvm.org/D131780
The commit of af2d11b1d5 missed a case where
the value of a suggested module needed to be reset to nullptr. Fixed thus
and added a testcase to cover the circumstance.
This change adds support for lowering llvm.prefetch directly using
GlobalISel. Currently, llvm.prefetch falls back to SelectionDAG.
This Change:
- Adds an AArch64-specific G_PREFETCH generic instruction, to be used
where AArch64ISD::PREFETCH is used in SelectionDAG.
- Adds the GINodeEquiv so patterns are translated over to GlobalISel
automatically.
- Corrects the AArch64Prefetch patterns to use a target immediate, which
is needed to get the patterns to translate across correctly.
- Translates the SelectionDAG legalisation of the prefetch intrinsic
into the corresponding GlobalISel legalisation.
Differential Revision: https://reviews.llvm.org/D132043
The current llvm.prefetch intrinsic docs state "The rw, locality and
cache type arguments must be constant integers."
This change:
- Makes arg 3 (cache type) an ImmArg
- Improves the verifier error messages to reference the incorrect
argument.
- Fixes two tests which contradict the docs.
This is needed as the lowering to GlobalISel is different for ImmArgs
compared to other constants. The non-ImmArgs create a G_CONSTANT MIR
instruction, the for ImmArgs the constant is put directly on the
intrinsic's MIR instruction as an immediate.
Differential Revision: https://reviews.llvm.org/D132042
Guillaume Chatelet