Add a new directive `either` to specify the operands can be matched in either order
Reviewed By: jpienaar, Mogball
Differential Revision: https://reviews.llvm.org/D110666
This is consistent with what we do for other operands that are required
to be constants.
I don't think this results in any real changes. The pattern match
code for isel treats ConstantSDNode and TargetConstantSDNode the same.
This is in preparation for only invalidating analyses on changed
functions.
Reviewed By: asbirlea
Differential Revision: https://reviews.llvm.org/D113303
A new tool that compares TargetLibraryInfo's opinion of the availability
of library function calls against the functions actually exported by a
specified set of libraries. Can be helpful in verifying the correctness
of TLI for a given target, and avoid mishaps such as had to be addressed
in D107509 and 94b4598d.
The tool currently supports ELF object files only, although it's unlikely
to be hard to add support for other formats.
Differential Revision: https://reviews.llvm.org/D111358
instruction the key points to is deleted
Use weak value handles for both the key and the value. The entry is
invalid if either value handle is null.
This fixes an assertion failure in BasicAAResult::alias that is caused
by UnderlyingObjCPtrCache returning a wrong value.
I don't have a test case for this patch that fails reliably.
rdar://83984790
getting the tls base address. unlike linux arm64, the tpidr_el0 returns always 0 (aka unused)
thus using tpidrro_el0 instead clearing up the cpu id encoded in the lower bits.
Reviewed-By: yln
Differential Revision: https://reviews.llvm.org/D112866
Currently, LOAD_STACK_GUARD on ARM is only implemented for Mach-O targets, and
other targets rely on the generic support which may result in spilling of the
stack canary value or address, or may cause it to be kept in a callee save
register across function calls, which means they essentially get spilled as
well, only by the callee when it wants to free up this register.
So let's implement LOAD_STACK GUARD for other targets as well. This ensures
that the load of the stack canary is rematerialized fully in the epilogue.
This code was split off from
D112768: [ARM] implement support for TLS register based stack protector
for which it is a prerequisite.
Reviewed By: nickdesaulniers
Differential Revision: https://reviews.llvm.org/D112811
At this point, every supported compiler that claims a -std=c++17 mode
should also support `if constexpr`. This was an issue for GCC 5
and GCC 6, but hasn't been an issue since GCC 7. (Our current
minimum supported GCC version, IIUC, is GCC 10 or 11.)
Differential Revision: https://reviews.llvm.org/D113348
Merge definition visibility the same way we do for other decls. Without
the fix the added test emits `-Wobjc-method-access` as it cannot find a
visible protocol. Make this warning `-Werror` so the test would fail
when protocol visibility regresses.
rdar://83600696
Differential Revision: https://reviews.llvm.org/D111860
- CUDA cannot associate memory space with pointer types. Even though Clang could add extra attributes to specify the address space explicitly on a pointer type, it breaks the portability between Clang and NVCC.
- This change proposes to assume the address space from a pointer from the assumption built upon target-specific address space predicates, such as `__isGlobal` from CUDA. E.g.,
```
foo(float *p) {
__builtin_assume(__isGlobal(p));
// From there, we could assume p is a global pointer instead of a
// generic one.
}
```
This makes the code portable without introducing the implementation-specific features.
Note that NVCC starts to support __builtin_assume from version 11.
Reviewed By: arsenm
Differential Revision: https://reviews.llvm.org/D112041
Generate static function for matching the type/attribute to reduce the
memory footprint.
Reviewed By: Mogball
Differential Revision: https://reviews.llvm.org/D110199
When an array's shape involves references to symbols that are not
invariant in a scope -- the classic example being a dummy array
with an explicit shape involving other dummy arguments -- the
compiler was creating shape expressions that referenced those
symbols. This might be valid if those symbols are somehow
captured and copied at each entry point to a subprogram, and
the copies referenced in the shapes instead, but that's not
the case.
This patch introduces a new expression predicate IsScopeInvariantExpr(),
which defines a class of expressions that contains constant expressions
(in the sense that the standard uses that term) as well as references
to items that may be safely accessed in a context-free way throughout
their scopes. This includes dummy arguments that are INTENT(IN)
and not VALUE, descriptor inquiries into descriptors that cannot
change, and bare LEN type parameters within the definitions of
derived types. The new predicate is then used in shape analysis
to winnow out results that would have otherwise been contextual.
Differential Revision: https://reviews.llvm.org/D113309
This is needed for the paths to work when using forward slashes;
this fixes the DirectoryWatcherTests unit tests.
Also allocate missing space for the null terminator, which seems to
have been missing all along (writing the terminator out of bounds).
Differential Revision: https://reviews.llvm.org/D113264
This fixes the following clang VFS tests, if `windows_slash` is the
default style:
Clang :: VFS/implicit-include.c
Clang :: VFS/relative-path.c
Clang-Unit :: Frontend/./FrontendTests.exe/CompilerInstance.DefaultVFSOverlayFromInvocation
Also clarify a couple references to `Style::windows` into
`Style::windows_backslash`, to make it clearer that each of them are
opinionated in different directions (even if it doesn't matter for
calls to e.g. `is_absolute`).
Differential Revision: https://reviews.llvm.org/D113272
InstCombine converts range tests of the form (X > C1 && X < C2) or
(X < C1 || X > C2) into checks of the form (X + C3 < C4) or
(X + C3 > C4). It is possible to express all range tests in either
of these forms (with different choices of constants), but currently
neither of them is considered canonical. We may have equivalent
range tests using either ult or ugt.
This proposes to canonicalize all range tests to use ult. An
alternative would be to canonicalize to either ult or ugt depending
on the specific constants involved -- e.g. in practice we currently
generate ult for && style ranges and ugt for || style ranges when
going through the insertRangeTest() helper. In fact, the "clamp like"
fold was relying on this, which is why I had to tweak it to not
assume whether inversion is needed based on just the predicate.
Proof: https://alive2.llvm.org/ce/z/_SP_rQ
Differential Revision: https://reviews.llvm.org/D113366
Component::CreatePointerDescriptor unconditionally expects a
vector of subscripts to be passed as an argument, but is called
from NAMELIST input with a null pointer. Make that argument
a nullable pointer, move it to the end of the argument list,
and give it a default value of nullptr.
Differential Revision: https://reviews.llvm.org/D113312
Previously, jumps to labels in constructs from exterior statements
would elicit only a warning. Upgrade these to errors unless the
branch into the construct would enter into only DO, IF, and SELECT CASE
constructs, whose interiors don't scope variables or have other
set-up/tear-down semantics. Branches into these "safe" constructs
are still errors if they're nested in an unsafe construct that doesn't
also enclose the exterior branch statement.
Differential Revision: https://reviews.llvm.org/D113310
I am planning to upstream MachOObjectFile code to support Darwin
chained fixups. In order to test the new parser features we need a way
to produce correct (and incorrect) chained fixups. Right now the only
tool that can produce them is the Darwin linker. To avoid having to
check in binary files, this patch allows obj2yaml to print a hexdump
of the raw LINKEDIT and DATA segment, which both allows to
bootstrap the parser and enables us to easily create malformed inputs
to test error handling in the parser.
This patch adds two new options to obj2yaml:
-raw-data-segment
-raw-linkedit-segment
Differential Revision: https://reviews.llvm.org/D113234
A CHECK() in semantics is triggering when analyzing a program
with an undefined derived type pointer because the CHECK is
expecting a new error message to have been issued in a function
but not allowing for the case that a diagnostic could have been
produced earlier. Adjust the predicate.
Differential Revision: https://reviews.llvm.org/D113307
This provides better support for `LambdaCapture`s by making them first-
class and allowing them to be bindable. In addition, this implements several
`LambdaCapture`-related matchers. This does not update how lambdas are
traversed. As a result, something like trying to match `lambdaCapture()` by
itself will not work - it must be used as an inner matcher.
Reviewed By: aaron.ballman, sammccall
Differential Revision: https://reviews.llvm.org/D112491
The value of EPSILON() was off by one power of two due to my
misreading of the standard; this patch corrects it.
Confirmed by checking other Fortran compilers, which all agree.
Differential Revision: https://reviews.llvm.org/D113308
[NFC] As part of using inclusive language within the llvm project, this patch
replaces master with main in file/directory paths in llvm LIT tests.
Reviewed By: bjope
Differential Revision: https://reviews.llvm.org/D113190
Extension of D112504. Lower amdgpu printf to `__llvm_omp_vprintf`
which takes the same const char*, void* arguments as cuda vprintf and also
passes the size of the void* alloca which will be needed by a non-stub
implementation of `__llvm_omp_vprintf` for amdgpu.
This removes the amdgpu link error on any printf in a target region in favour
of silently compiling code that doesn't print anything to stdout.
The exact set of changes to check-openmp probably needs revision before commit
Reviewed By: jdoerfert
Differential Revision: https://reviews.llvm.org/D112680
The existing code has unnecessary logic to indirectly pass
errors through function calls. This diff gets rid of the fluff.
Test Plan: Existing unit tests
Reviewed By: jhenderson, drodriguez, alexander-shaposhnikov
Differential Revision: https://reviews.llvm.org/D113301
Add pad_const field to tosa.pad.
Add builders to enable optional construction of pad_const in pad op.
Update documentation of tosa.clamp to match spec wording.
Signed-off-by: Suraj Sudhir <suraj.sudhir@arm.com>
Reviewed By: rsuderman
Differential Revision: https://reviews.llvm.org/D113322
When processing the devious NAMELIST input
&group logarray = t t t
= 666 /
for LOGICAL::logarray(3) and INTEGER::t, the runtime library
needs to do some look-ahead on the input stream to make sure
that the last "t" on the first line is a truth value rather than
an item name -- which in this case it is. This look-ahead
was implemented in a previous patch but only worked for internal
input cases; this patch implements look-ahead capabilities for
input from an external file, too (and also adjusts repeated
list-directed input items to use this infrastructure, too).
Differential Revision: https://reviews.llvm.org/D113311
Please note that the updated documentation reflects the design that we are
working towards, but haven't implemented competely yet. We took this approach
in order to provide a more holisitic and complete overview of the design.
In particular, this document assumes that Flang's frontend and compiler
driver can already generate code. This is still work-in-progress and is
being developed on the `fir-dev` branch in the F18 repository [1].
[1] https://github.com/flang-compiler/f18-llvm-project/tree/fir-dev
Differential Revision: https://reviews.llvm.org/D111573
All phi-like recipes should be at the beginning of a VPBasicBlock with
no other recipes in between. Ensure that the recurrence-splicing recipe
is not added between phi-like recipes, but after them.
Reviewed By: Ayal
Differential Revision: https://reviews.llvm.org/D111301
These and MULHS/MULHU both default to Legal. Targets need to set
the ones they don't support to Expand.
I think MULHS/MULHU likely has priority in most places so this
change probably isn't directly testable. I found it while looking
at disabling MULHS/MULHU for nxvXi64 as required for Zve64x.
Reviewed By: frasercrmck
Differential Revision: https://reviews.llvm.org/D113325
Declarative attribute and type formats with assembly formats. Define an
`assemblyFormat` field in attribute and type defs with a `mnemonic` to
generate a parser and printer.
```tablegen
def MyAttr : AttrDef<MyDialect, "MyAttr"> {
let parameters = (ins "int64_t":$count, "AffineMap":$map);
let mnemonic = "my_attr";
let assemblyFormat = "`<` $count `,` $map `>`";
}
```
Use `struct` to define a comma-separated list of key-value pairs:
```tablegen
def MyType : TypeDef<MyDialect, "MyType"> {
let parameters = (ins "int":$one, "int":$two, "int":$three);
let mnemonic = "my_attr";
let assemblyFormat = "`<` $three `:` struct($one, $two) `>`";
}
```
Use `struct(*)` to capture all parameters.
Reviewed By: rriddle
Differential Revision: https://reviews.llvm.org/D111594
This avoids crashes when the pattern is applied to ops with
tensor semantics.
Reviewed By: ThomasRaoux
Differential Revision: https://reviews.llvm.org/D113415
When targeting a specific CPU with scalable vectorization, the knowledge
of that particular CPU's vscale value can be used to tune the cost-model
and make the cost per lane less pessimistic.
If the target implements 'TTI.getVScaleForTuning()', the cost-per-lane
is calculated as:
Cost / (VScaleForTuning * VF.KnownMinLanes)
Otherwise, it assumes a value of 1 meaning that the behavior
is unchanged and calculated as:
Cost / VF.KnownMinLanes
Reviewed By: kmclaughlin, david-arm
Differential Revision: https://reviews.llvm.org/D113209
Chia-hung Duan