It's one type of TT_TrailingReturnArrow.

`IRMover` links in referenced private global values unconditionally, see
`IRLinker::shouldLink()`. If they are part of a non-prevailing comdat,
this leads to duplication of the values. Full and Thin LTO avoid
duplication by changing the linkage of members of non-prevailing comdat
groups to `available_externally`, which was implemented in
https://reviews.llvm.org/D34803 and https://reviews.llvm.org/D135427.
This patch does the same for `Linker`, but limits the effect only to
private members without aliases to minimize interference.

Motivation example:
```
> cat foo.h
inline int foo(int a) { return a + 1; }
> cat bar.cpp
#include "foo.h"
int bar(int a) { return foo(a + 1); }
> cat main.cpp
#include "foo.h"
int bar(int a);
int main(int argc, const char* argv[]) { return bar(argc) + foo(argc); }
> clang++ -c -flto -fprofile-instr-generate main.cpp -o main.o
> clang++ -c -flto -fprofile-instr-generate bar.cpp -o bar.o
> clang++ -fuse-ld=lld -fprofile-instr-generate main.o bar.o -o test1
> ./test1
> llvm-profdata merge --text default.profraw -o -
_Z3fooi
# Counter Values:
2
> llvm-link main.o bar.o -o combined.o
> clang++ -fuse-ld=lld -fprofile-instr-generate combined.o -o test2
> ./test2
> llvm-profdata merge --text default.profraw -o -
_Z3fooi
# Counter Values:
4
```

This patch addresses some examples of bad formatting in Clang. The following commit contains only changes suggested by clang-format: https://github.com/llvm/llvm-project/pull/70349/commits/21689b56d1fc1db0b2263e8049ff656d3757ad36.
I believe it's a net negative on readability, with a couple of particularly bad cases. Highlights:
```diff
-    [[clang::preferred_type(bool)]]
-    mutable unsigned CachedLocalOrUnnamed : 1;
+    [[clang::preferred_type(bool)]] mutable unsigned CachedLocalOrUnnamed : 1;
```
```diff
-    [[clang::preferred_type(TypeDependence)]]
-    unsigned Dependence : llvm::BitWidth<TypeDependence>;
+    [[clang::preferred_type(TypeDependence)]] unsigned Dependence
+        : llvm::BitWidth<TypeDependence>;
```
```diff
-    [[clang::preferred_type(ExceptionSpecificationType)]]
-    unsigned ExceptionSpecType : 4;
+    [[clang::preferred_type(
+        ExceptionSpecificationType)]] unsigned ExceptionSpecType : 4;
```
Style guides doesn't appear to have an opinion on this matter.

Added the simplify function to reduce the size of the constraint system,
referencing the ISL implementation.

Tested it on a simple Benchmark implemented by myself, calling SImplify
before the operation and calling Simplify on the result after Subtract
were tested, respectively.

The Benchmark can be found here:
[benchmark](https://github.com/gilsaia/llvm-project-test-fpl/blob/develop_benchmark/mlir/benchmark/presburger/Benchmark.cpp)

For the case of calling Simplify before each operation, the overall
result is shown in the following figure.

![image](https://github.com/llvm/llvm-project/assets/38588948/7099286e-b9a2-42e0-bc2a-1ed6627ead00)

A comparison of the constraint system sizes and time for each operation
is as follows

![image](https://github.com/llvm/llvm-project/assets/38588948/e5d0e488-f76e-4438-b19e-f6163699c526)

![image](https://github.com/llvm/llvm-project/assets/38588948/119a08de-4ee1-4cde-886c-50a91b502d93)

![image](https://github.com/llvm/llvm-project/assets/38588948/7a8b69ac-6cdb-41ab-9a75-cd016664fa5a)

![image](https://github.com/llvm/llvm-project/assets/38588948/c84b6eb1-62dc-4bae-a771-67d97ebf514a)

![image](https://github.com/llvm/llvm-project/assets/38588948/cdbfa3ed-0155-481e-9273-9d6dba3a2d7b)

![image](https://github.com/llvm/llvm-project/assets/38588948/8c945cff-a0a4-472a-a178-6b6a70a1b16a)

![image](https://github.com/llvm/llvm-project/assets/38588948/0bfe3a2b-3568-4d31-bebf-bd1b3c4e734e)

![image](https://github.com/llvm/llvm-project/assets/38588948/f1a99d56-edf5-45de-a506-512c0584f1d8)

![image](https://github.com/llvm/llvm-project/assets/38588948/ffef3312-6c99-494c-bb52-73aa8df275bb)

![image](https://github.com/llvm/llvm-project/assets/38588948/3e5924a7-8e1f-49d1-bd27-02a2e10a5cc4)

![image](https://github.com/llvm/llvm-project/assets/38588948/cec8be0e-dd19-46fa-88b4-2585d4031c9e)

![image](https://github.com/llvm/llvm-project/assets/38588948/3cb68e89-82c7-4cd2-b6bc-70f15e495ce8)

For the case of calling Simplify on the result after Subtract, the
overall results are as follows

![image](https://github.com/llvm/llvm-project/assets/38588948/be5b9c50-7417-42c8-abbf-8a50f093c3f5)

A comparison of the constraint system sizes and time for subtract is as
follows

![image](https://github.com/llvm/llvm-project/assets/38588948/fafe10ba-f8bd-43cd-b281-aaebf09af0af)

![image](https://github.com/llvm/llvm-project/assets/38588948/24662b40-42fc-47ee-a0a3-1b8b8f5778d2)

This is a follow-up to 910a4bf5.

1. __builtin_clz takes unsigned int, thus for uint16_t
src_rep_t_clz can't use it directly but should subtract 16
(leading 16 bits of the promoted argument are zero).

2. Fix (and simplify) clz_in_sig_frac.

Test plan: ninja check-compiler-rt
(extendhfsf2_test.c and extenddftf2_test.c)

This reverts commit d37b283c.

There was a simple logic bug in the else path. Tests codegen is different with
the fix.

Base on D19403, the exact pragma of distribute is
  `#pragma clang loop distribute`

This requires adding a `NegDelta32` edge kind that cannot be mapped to
existing relocations.

Co-authored-by: Job Noorman <jnoorman@igalia.com>

Reland #67275 with #68953 resolved.

EnumCastOutOfRange checker now reports the name of the enum in the
warning message. Additionally, a note-tag is placed to highlight the
location of the declaration.

This patch adds a verifier to tosa.reverse which checks the axis
argument and input/output tensor ranks for validity. We allow a special
case where `axis == 0 && rank == 0`.

Forgotten in commit 547a5073

[mlir][bytecode] Fix D155919 and enable backward-compatibility and back-deployment between version 5 and version 6 of the bytecode encoding (#70498)

Before D155919, when a dialect was leveraging properties to store
attributes with `usePropertiesForAttributes`, the operand segment sizes
attribute was emitted in the property section in sorted order together
with all the other attributes of an op. After D155919, version 5 of the
bytecode was emitting and parsing operand segment sizes after all the
properties of an op, breaking backward compatibility and back
deployment. This patch fixes the emission ordering and allows to parse
bytecode files emitted before (D155919) with version 5 of MLIR bytecode.
The patch also enables to emit correctly version 5 of MLIR bytecode.

https://github.com/llvm/llvm-project/commit/28b912687900bc0a67cd61c374fce296b09963c4
introduced the path cloning format in the basic-block-sections profile.

This PR validates and applies path clonings. 
A path cloning is valid if all of these conditions hold:
  1. All bb ids in the path are mapped to existing blocks.
2. Each two consecutive bb ids in the path have a successor relationship
in the CFG.
3. The path does not include a block with indirect branches, except
possibly as the last block.
 
Applying a path cloning involves cloning all blocks in the path (except
the first one) and setting up their branches.
Once all clonings are applied, the cluster information is used to guide
block layout in the modified function.

The `LoopLikeOpInterface` already has interface methods to query inits
and iter_args. This commit adds helper functions to query tied
init/iter_arg pairs and removes the corresponding functions for
`scf::ForOp`.

Fixing ```#error "Unknown or unsupported OS"```

Fixes #68966

Opaque pointer cleanup effort. NFC.

Our CODE and LIB are more unified every day!

[OpenMP] Fixes #69905

The corresponding definitions were removed in 7dcbe3d3 and 2a8fa2a8.
Also remove a couple of variables made dead by those changes.

- `(icmp eq/ne (and (add/sub/xor X, P2), P2), P2)`
    -> `(icmp eq/ne (and X, P2), 0)`
- `(icmp eq/ne (and (add/sub/xor X, P2), P2), 0)`
    -> `(icmp eq/ne (and X, P2), P2)`

Folds like this come up with reasonable regularity in odd/even loops.

Proofs: https://alive2.llvm.org/ce/z/45pq2x

Closes #67836

Some upcoming intrinsics will be using these new types

Fixed typo, 80-col

I have a crash when parsing the dyld trie data and I want to ask the
originator to send me the (possibly corrupted) binary that's causing
this. This patch adds some logging to help pinpoint that file.

Introduced with cf0f6a14 (#70020).

https://lab.llvm.org/buildbot/#/builders/237/builds/5145
https://lab.llvm.org/buildbot/#/builders/236/builds/7004
https://lab.llvm.org/buildbot/#/builders/239/builds/4269

Sometimes it passes with Asan but it's extemly slow

```
Slowest Tests:
1388.61s: llvm-libc++-shared.cfg.in :: std/experimental/simd/simd.reference/reference_assignment.pass.cpp
309.42s:  llvm-libc++-shared.cfg.in :: std/containers/sequences/deque/deque.modifiers/insert_size_value.pass.cpp
```

This reverts commit a66051c6.

This seems to have caused issue #70509 so reverting until I have time
to investigate.

This PR adds a function which converts the language to string. This is
intended to be used by the z/OS target, see the patch here:
https://github.com/llvm/llvm-project/pull/68926



---------

Co-authored-by: Yusra Syeda <yusra.syeda@ibm.com>

The instruction should only read the bottom 8 bits of the register, so an
anyext is OK here. Update the comment from zext->anyext to clarify.

Closes #70270 and #70298