When the scavenger is not allowed to spill, the only difference between
forward and backward should be the heuristics used to pick an available
register. Forwards scavenging tries to pick a register that can be used
again later in the BB; backwards scavenging tries to pick one that can
be used earlier.

Backwards scavenging is preferred because it does not rely on accurate
kill flags.

Differential Revision: https://reviews.llvm.org/D151323

A buildbot has failed on the absence of the blank line at the end of the bullet list.

The type alias template is not diagnosed when instantiating an expected non-type template argument in a dependent scope, causing ICE.

Besides that, the diagnostic message has been updated to account for the fact that the function template  is not the only non-type template.

Fixes #62533

Reviewed By: #clang-language-wg, erichkeane

Differential Revision: https://reviews.llvm.org/D151062

The interop API routines try to invoke external entries, but we did
not have support for KMP_DLSYM_NEXT on Windows. Also added proper
guards for STUB build.

Differential Revision: https://reviews.llvm.org/D149892

Correct link is clang-tidy/checks/performance/no-automatic-move

Fix bad link to documentation.

This allows subclasses (such as the ErrorCheckingTrackingListener in IREE) to produce better error messages.

Differential Revision: https://reviews.llvm.org/D151322

Only the last instruction was remapped before.

We were only handling `const T&`/`T&&` ctor pairs, and we were missing uref-based ctors.

Differential Revision: https://reviews.llvm.org/D151092

The following ops currently have the same finger print, even though they are different:
```
func.func @test() {
  "test.foo"() ({
    "test.bar"() : () -> ()
  }) : () -> ()
}
```
And:
```
func.func @test() {
  "test.bar"() : () -> ()
  "test.foo"() ({ }) : () -> ()
}
```

The SHA1 hash used in OperationFingerPrint is order-sensitive, but the ops are hashed in the same order (post-order traversal), so the hash is the same. Switching to pre-order traversal does not solve the issue; a similar example, where IR differs just in its nesting structure, can be constructed.

The problem is solved by hashing the parent op pointer. (Alternatively, a traversal over the IR that hashes scope markers (`{}`) could be used.)

Differential Revision: https://reviews.llvm.org/D151306

Reviewed By: gchatelet

Differential Revision: https://reviews.llvm.org/D151256

This patch makes `-Wshadow` warn about the shadowings by static local variables.

Fixes https://github.com/llvm/llvm-project/issues/62850
Differential Revision: https://reviews.llvm.org/D151214

Reland: [clang][AST] Print name instead of type when diagnosing uninitialized subobject in constexpr variables

This patch improves the diagnostic on uninitialized subobjects in constexpr variables by modifying the diagnostic message to display the subobject's name instead of its type.

Fixes https://github.com/llvm/llvm-project/issues/58601
Differential Revision: https://reviews.llvm.org/D146358

The initial bring-up of the Transform dialect relied on PDL to provide
the default handle type (`!pdl.operation`) and the matching capability.
Both are now provided natively by the Transform dialect removing the
reason to have a hard dependency on the PDL dialect and its interpreter.
Move PDL-related transform operations into a separate extension.

This requires us to introduce a dialect state extension mechanism into
the Transform dialect so it no longer needs to know about PDL constraint
functions that may be injected by extensions similarly to operations and
types. This mechanism will be reused to connect pattern application
drivers and the Transform dialect.

This completes the restructuring of the Transform dialect to remove
overrilance on PDL.

Note to downstreams: flow that are using `!pdl.operation` with Transform
dialect operations will now require `transform::PDLExtension` to be
applied to the transform dialect in order to provide the transform
handle type interface for `!pdl.operation`.

Reviewed By: springerm

Differential Revision: https://reviews.llvm.org/D151104

We only expect these ctor / dtor functions to be called with a single
thread. Add the appropriate attributes to indicate this to the backend.

Reviewed By: arsenm

Differential Revision: https://reviews.llvm.org/D151153

This mimicks the implementation of the libstdcpp std::unique_ptr
formatter.

This has been attempted several years ago in
`0789722d` but was reverted in
`e7dd3972`.

The difference to the original patch is that we now maintain
a `$$dereference$$` member and we only store weak pointers
to the other children inside the synthetic frontend. This is
what the libc++ formatters do to prevent the recursion mentioned
in the revert commit.

Add log, log2, log10, log1p and their unit tests to bazel layout.

Reviewed By: gchatelet

Differential Revision: https://reviews.llvm.org/D151252

Instead of blindly traversing the use-def chain of constant arguments,
compute known constants along the way. Stop as soon as a user cannot
be replaced by a constant. Keep it light-weight by handling some basic
instruction types.

Differential Revision: https://reviews.llvm.org/D150464

Fixes part of the issue in https://github.com/llvm/llvm-project/issues/62013

Reviewed By: psoni2628

Differential Revision: https://reviews.llvm.org/D151206

Add RegionBranchOpIntefface to scf.forall and scf.parallel op to make analysis trace through subregions.

Differential Revision: https://reviews.llvm.org/D151287

Consecutive concats that happen on the same axis can be folded into a
single, bigger concat. This patch implements this folding by
implementing the tosa::ConcatOp::fold method.

Differential Revision: https://reviews.llvm.org/D151210

D151036 adds an assertions that prohibits iterating over sub- and
super-registers of a null register. This is already the case when
iterating over register units of a null register, and worked by
accident for sub- and super-registers.
The only place where the assertion is currently triggering is in
CriticalAntiDepBreaker::ScanInstruction. Other places are changed
in case new assertions are added and should be harmless otherwise.

Differential Revision: https://reviews.llvm.org/D151288

Do not open code creation of an interceptor alias to another
interceptor. Instead, use INTERCEPTOR() + ALIAS() + WRAP.

Reviewed By: dvyukov

Differential Revision: https://reviews.llvm.org/D151219

The __interceptor_*setjmp() definitions appear to have been defined for
the purpose of TSAN_INTERCEPT(), but on non-Mac systems, it seems
TSAN_INTERCEPT() isn't even being used anymore for setjmp.

Remove them. Nothing should call them anyway (due to CHECK-fail), so
having the linker fail is better than failing at runtime.

Reviewed By: dvyukov

Differential Revision: https://reviews.llvm.org/D151218

Use ASM_WRAPPER_NAME to produce the name of the __interceptor_*
functions.

NFC.

Reviewed By: dvyukov

Differential Revision: https://reviews.llvm.org/D151217

Most uses of ALIAS() are in conjunction with WRAPPER_NAME().

Simplify the code and just make ALIAS() turn its argument into a string
(similar to Linux kernel's __alias macro). This in turn allows removing
WRAPPER_NAME().

NFC.

Reviewed By: dvyukov

Differential Revision: https://reviews.llvm.org/D151216

This commit ensures that the merge_handles operation prints its
`deduplicate` attribute as an optional keyword instead of "unit".

Reviewed By: ftynse

Differential Revision: https://reviews.llvm.org/D151310

This seems to be an issue currently where there are nested/chained GEP/BitCast Pointers.
The patch generates a new GEP for the wider load to avoid dominance problems.

Differential Revision: https://reviews.llvm.org/D150864

For small fixed-length vector copies like
vsetivli   zero, 2, e16, m1, ta, ma
vle16.v    v8, (a0)
vse16.v    v8, (a1)

We can scalarize them if the total vector size < XLEN:
lw a0, 0(a0)
sw a0, 0(a1)

This patch adds a DAG combine to do so, reusing much of the existing
logic in https://reviews.llvm.org/D150717

Reviewed By: reames

Differential Revision: https://reviews.llvm.org/D151103

Reviewed By: reames

Differential Revision: https://reviews.llvm.org/D151211

Only check for overflow on the min/max values, don't also check
for predicates in addition to that.

Make it clear that the leading/trailing zeros handling is only
relevant for the unknown LHS case, which is a fast path to avoid
the full shift amount loop in cases where it would not produce
better results.

I removed the conflict check from computeKnownBitsFromShiftOperator()
in D150648 assuming that this is now handled on the KnownBits side.
However, the nsw handling is still inside ValueTracking, so we
still need to handle conflicts there. Restore the check closer to
where it is relevant.

Fixes https://github.com/llvm/llvm-project/issues/62908.

After 3a223f1e, it's no longer
necessary to put the users into a vector. We can directly iterate
them instead.

And make sure udiv() Exact does not produce conflicts.

This fixes #62880

Differential Revision: https://reviews.llvm.org/D151200