…th (#97353)

Make the results of the two IsInteroperableIntrinsicType() utility
routines a tri-state std::optional<bool> so that cases where the
character length is simply unknown can be distinguished from those cases
where the length is known and not acceptable. Use this distinction to
not emit a confusing warning about interoperability with C_LOC()
arguments when the length is unknown and might well be acceptable during
execution.

Support the predefined macro __TIMESTAMP__ as interpreted by GCC. It
expands to a character literal with the time of last modification of the
top-level source file in asctime(3) format, e.g. "Tue Jul 4 10:18:05
1776".

…ackages (#98420)

This reduces Sphinx dependencies for building lldb man pages as lldb man
pages don't use markdown.

Section IDs are 64 bit and if a section ID was over 4GB, then the
tabular output of the "target modules dump sections" command would not
align to the column headers. Also if the section type's name was too
long, the output wouldn't algin. This patch fixes this issue.

Old output looked like:
```
(lldb) image dump sections a.out
Sections for '/tmp/a.out' (arm):
  SectID     Type             File Address                             Perm File Off.  File Size  Flags      Section Name
  ---------- ---------------- ---------------------------------------  ---- ---------- ---------- ---------- ----------------------------
  0xffffffffffffffff container        [0x0000000000001000-0x0000000000001010)  rw-  0x00000074 0x00000010 0x00000000 a.out.PT_LOAD[0]
  0x00000001 data             [0x0000000000001000-0x0000000000001010)  rw-  0x00000074 0x00000010 0x00000003 a.out.PT_LOAD[0]..data
  0xfffffffffffffffe container        [0x0000000000001000-0x0000000000001010)  rw-  0x00000084 0x00000000 0x00000000 a.out.PT_TLS[0]
  0x00000002 zero-fill        [0x0000000000001000-0x0000000000001010)  rw-  0x00000084 0x00000000 0x00000403 a.out.PT_TLS[0]..tbss
  0x00000003 regular                                                   ---  0x00000084 0x00000001 0x00000000 a.out..strtab
  0x00000004 regular                                                   ---  0x00000085 0x0000001f 0x00000000 a.out..shstrtab
```
New output looks like:
```
(lldb) image dump sections a.out
Sections for '/tmp/a.out' (arm):
  SectID             Type                   File Address                             Perm File Off.  File Size  Flags      Section Name
  ------------------ ---------------------- ---------------------------------------  ---- ---------- ---------- ---------- ----------------------------
  0xffffffffffffffff container              [0x0000000000001000-0x0000000000001010)  rw-  0x00000074 0x00000010 0x00000000 a.out.PT_LOAD[0]
  0x0000000000000001 data                   [0x0000000000001000-0x0000000000001010)  rw-  0x00000074 0x00000010 0x00000003 a.out.PT_LOAD[0]..data
  0xfffffffffffffffe container              [0x0000000000001000-0x0000000000001010)  rw-  0x00000084 0x00000000 0x00000000 a.out.PT_TLS[0]
  0x0000000000000002 zero-fill              [0x0000000000001000-0x0000000000001010)  rw-  0x00000084 0x00000000 0x00000403 a.out.PT_TLS[0]..tbss
  0x0000000000000003 regular                                                         ---  0x00000084 0x00000001 0x00000000 a.out..strtab
  0x0000000000000004 regular                                                         ---  0x00000085 0x0000001f 0x00000000 a.out..shstrtab
```

This change is an implementation of
#87367 investigation on
supporting IEEE math operations as intrinsics.
Which was discussed in this RFC:
https://discourse.llvm.org/t/rfc-all-the-math-intrinsics/78294

This change adds constraint intrinsics and some lowering cases for
`acos`, `asin`, `atan`, `cosh`, `sinh`, and `tanh`.
The only x86 specific change was for f80.

#70079
#70080
#70081
#70083
#70084
#95966
    
The x86 lowering is going to be done in three pr changes with this being
the first.
A second PR will be put up for Loop Vectorizing and then SLPVectorizer.

The constraint intrinsics is also going to be in multiple parts, but
just 2.
This part covers just the llvm specific changes, part2 will cover clang
specifc changes and legalization for backends than have special
legalization
 requirements like aarch64 and wasm.

f18 current emits an error when an assignment is made to an array
section with a vector subscript, and the array is finalized with a
non-elemental final subroutine. Some other compilers emit this error
because (I think) they want variables to only be finalized in place, not
by a subroutine call involving copy-in & copy-out of the finalized
elements.

Since many other Fortran compilers can handle this case, and there's
nothing in the standards to preclude it, let's downgrade this error
message to a portability warning.

This patch got complicated because the API for the WhyNotDefinable()
utility routine was such that it would return a message only in error
cases, and there was no provision for returning non-fatal messages. It
now returns either nothing, a fatal message, or a non-fatal warning
message, and all of its call sites have been modified to cope.

)

We emit an incorrect error message when !DIR$ IGNORE_TKR appears in a
separate module procedure's interface declaration.

Fixes #98210.

This patch updates the clang-tidy checks for llvm-libc to ensure that
the namespace macro used to declare the libc namespace is updated from
LIBC_NAMESPACE to LIBC_NAMESPACE_DECL which by default has hidden
visibility.

Co-authored-by: Prabhu Rajesakeran <prabhukr@google.com>

Data designators like "a(j:k)" are parsed into array section references,
but once rank and type information is in hand, some of them turn out to
actually be substring references. The code that recognizes these cases
was suffering from a "false positive" in the case of a construct entity
in a SELECT RANK construct due to the use of a predicate member function
(Symbol::IsObjectArray) that only works on ObjectEntityDetails symbols.
Fix the test to use the more general Symbol::Rank() member function.

…98478)

This can cause breakage with user code that does "#define A ...".

Doing so allows one side to fold entirely into the mask applied to the
other recursive call (or a vmerge.vv at worst). This is a generalization
of the existing IsSelect case (both operands are selects), so I removed
that code in the process.

This actually started as an attempt to remove the IsSelect bit as I'd
thought it was fully redundant
with the recursive formulation, but digging into test deltas revealed
that we depended on that
to catch the majority of the identity cases, and that in turn we were
missing some cases where only RHS was an identity.

Reverts #98469

We can't add this dependency

```
   OBJECT_LIBS RTSanitizerCommon
                RTSanitizerCommonLibc
```

safestack is security hardening, and RTSanitizerCommon is too fat for
that.

Module files emitted by this Fortran compiler are valid Fortran source
files. Symbols that are USE-associated into modules are represented in
their module files with USE statements and special comments with hash
codes in them to ensure that those USE statements resolve to the same
modules that were used to build the module when its module file was
generated.

This scheme prevents unchecked module file growth in large applications
by not emitting USE-associated symbols redundantly. This problem can be
especially bad when derived type definitions must be repeated in the
module files of their clients, and the clients of those modules, and so
on. However, this scheme has the disadvantage that clients of modules
must be compiled with dependent modules in the module search path.

This new -fhermetic-module-files option causes module file output to be
free of dependences on any non-intrinsic module files; dependent modules
are instead emitted as part of the module file, rather than being
USE-associated. It is intended for top level library module files that
are shipped with binary libraries when it is not convenient to collect
and ship their dependent module files as well.

Fixes #97398.

When collecting candidates to pre-compute cost for operands of exit
conditions, skip users outside the loop when checking if they are in
ExistInstrs. The users outside the loop should be ignored, as they won't
make a value live in the VPlan.

This fixes a failure when building for X86 with sanitizers on macOS
after b841e2e
(https://green.lab.llvm.org/job/llvm.org/job/clang-stage2-cmake-RgSan/287/)

Atexit needs to be linked into exit on linux since atexit defines
__cxa_finalize. This should probably be fixed a different way but this
works for now.

spec: microsoft/hlsl-specs#263

- `Attr.td` - Define the HLSL loop attribute hints (unroll and loop)
- `AttrDocs.td` - Add documentation for unroll and loop
- `CGLoopInfo.cpp` - Add codegen for HLSL unroll that maps to clang
unroll expectations
- `ParseStmt.cpp` - For statements if HLSL define DeclSpecAttrs via
MaybeParseMicrosoftAttributes
- `SemaStmtAttr.cpp` - Add the HLSL loop unroll handeling

resolves #70114

dxc examples: 
- for loop: https://hlsl.godbolt.org/z/8EK6Pa139
- while loop:  https://hlsl.godbolt.org/z/ebr5MvEcK
- do while: https://hlsl.godbolt.org/z/be8cedoTs 

Documentation:

![Screenshot_20240531_143000](https://github.com/llvm/llvm-project/assets/1802579/9da9df9b-68a6-49eb-9d4f-e080aa2eff7f)

The names in the `end function` were incorrect. Those have been removed.

This is analogous to `HUGE_VAL`.

GCC emits a warning when using the visibility attribute which needs to
be diagnosed and addressed, but this change should unbreak the GCC build
as a temporary workaround.

The issue is tracked as #98548.

We automatically inject this dependency into all object libraries that
use the libc CMake functions, but `libc_diff_test_utils` uses `add_library`
so we need to add this dependency manually.

This matches what is done for FreeBSD.

OpenBSD has a few special program header types, and other such ELF
extensions. Setting the ELFOSABI like so will allow LLD to support them
without needlessly impacting non-OpenBSD ELFs.

Testing strategy matches 06cecdc.

Take two of #98158 / b64c1de, which was
reverted in #98494 / c026135.
Preexisting test is fixed now.

I made a mistake in #98553.
Sorry.

Adding dep to TosaDialect increases binary size unnecessarily

Update availability information added in 1eb7f05. exp10 is available
on iOS >= 7.0 and macOS >= 10.9. On all other platforms, it is available
on any version. Also drop the x86 check, as the availability only
depends on the OS version, not the target platform.

PR: #98542

- if there is an object made there is a space after
- fixed tests.yaml -- spacing between characters issue

Reapplies #92957, fixing an instance where the `template` keyword was
missing prior to a dependent name in `llvm/ADT/ArrayRef.h`. An
_alias-declaration_ is used to work around a bug affecting GCC releases
before 11.1 (see https://gcc.gnu.org/bugzilla/show_bug.cgi?id=94799) which
rejects the use of the `template` keyword prior to the
_nested-name-specifier_ in the class member access.

MSVC has a __cpuidex function implemented to call the underlying cpuid
instruction which accepts a leaf, subleaf, and data array that the
output data is written into. This patch adds this functionality into
clang under the cpuid.h header. This also makes clang match GCC's
behavior. GCC has had __cpuidex in its cpuid.h since 2020.

This is another attempt to land https://reviews.llvm.org/D158348.

If requested, via the -memprof-report-hinted-sizes option, track the
total profiled size of each MIB through the thin link, then report on
the corresponding allocation coldness after all cloning is complete.

To save size, a different bitcode record type is used for the allocation
info when the option is specified, and the sizes are kept separate from
the MIBs in the index.

…#97363)

This patch fixes another place in ProfileData where we have a pointer
to an array of InstrProfValueData and its length separately.

addValueData is a bit unique in that it remaps incoming values in
place before adding them to ValueSites.  AFAICT, no caller of
addValueData uses updated incoming values.  With this patch, we add
value data to ValueSites first and then remaps values there.  This
way, we can take ArrayRef<InstrProfValueData> as a parameter.

This test sets a breakpoint on malloc, as a way to stop early in
dyld's setting up code, before the system libraries are initialized
so we can confirm that we don't fetch the Objective-C class table
before it's initialized.

In macOS 15 (macOS Sonoma), dyld doesn't call malloc any longer,
so this heuristic/trick isn't working.  It does call other things
called *alloc though, so I'm changing this to use a regex breakpoint
on that, to keep the test working.

When implicit data transfer is created, make sure we generate the
`freemem` op on the `allocmem` result value and not the declare op
value.

This was found by the Clang Static Analyzer.

PrintAddressSpaceLayout can accidentally
mmap into the gap.

See the comment in handleTlsRelocation. For TLSDESC=>IE (the TLS symbol
is defined in another DSO), R_RISCV_TLSDESC_{LOAD_LO12,ADD_LO12_I,CALL}
referencing a non-preemptible label uses the `R_RELAX_TLS_GD_TO_LE` code
path.

If there is no TLS section, `getTlsTpOffset` will be called with null
`Out::tlsPhdr`, leading to a null pointer dereference. Since the return
value is used by `RISCV::relocateAlloc` and ignored there, just return
0.

LoongArch TLSDESC doesn't use STT_NOTYPE labels. The `if (..) return 0;`
is a no-op for LoongArch.

This patch is a follow-up to #79239 and fixes some comments.

Pull Request: #98569

For synchronous unwind tables, the call frame information can be
slightly reduced by bundling the `.cfi_negate_ra_state` instruction with
other CFI instructions in the prolog, saving 1 byte per function used
for `DW_CFA_advance_loc`.

This was suggested in
[D156428](https://reviews.llvm.org/D156428#4554317).

Same as X86, , if X's size is BitWidth, then X sdiv 2 can be
expressived as
```
  X += X >> (BitWidth - 1)
  X = X >> 1
```

Fix #97884

#94322 defines .preinit_array to initialize nsan early.
DT_PREINIT_ARRAY can only be used with the main executable. GNU ld would
complain when a DSO has .preinit_array. Therefore,
nsan_preinit.cpp cannot be linked into `libclang_rt.nsan.so` (#98415).

Working with @alexander-shaposhnikov, we noticed that `Nsan-x86_64-Test
--gtest_output=json` without `.preinit_array` will sigsegv. This is
because googletest with the JSON output calls `localtime_r` , which
calls `free(0)` and fails when `REAL(free)` remains uninitialized
(nullptr). This is benign with the default output because malloc/free
are all paired and `REAL(free)(ptr)` is not called.

To fix the unittest failure, `__nsan_init` needs to be called early
(.preinit_array).
`asan/tests/CMakeLists.txt:ASAN_UNITTEST_INSTRUMENTED_LINK_FLAGS` ues
`-fsanitize=address` to ensure `asan_preinit.cpp.o` is linked into the
unittest executable. Port the approach and remove
`NSAN_TEST_RUNTIME_OBJECTS`.

Fix #98523

Pull Request: #98564

And remove spaces around '-' printing ranges.

…unk (#98286)

Most of the time when we coalesce and delete a vsetvli, we shrink the
LiveInterval of its AVL register now that there is one less use. However
there's one edge case we were missing where if we have two vsetvlis with
no users of vl or vtype in between, we coalesced a vsetvli without
shrinking it's AVL.

This fixes it by shrinking the LiveInterval whenever we delete a
vsetvli, and also makes the LiveIntervals consistent in-situ by not
removing the use before shrinking.

This fixes a -verify-machineinstrs assertion in an MIR test case I found
while investigating
#97264 (comment).
I couldn't recreate this at the LLVM IR level, seemingly because
RISCVInsertVSETVLI will just avoid inserting extra vsetvlis that don't
need coalesced.

…fiers (#98023)

Previously, we only pushed the function scope once we entered the
function definition, whereas tryCaptureVariable() requires at least one
function scope available when ParmVarDecls being captured have been
owned by a function. This led to problems parsing the noexcept
specifiers, as the DeclRefExprs inside them were improperly computed.

Fixes #97453

Extract the logic whether to emit a global var based on CUDA/HIP
host/device related attributes to CodeGenModule::shouldEmitCUDAGlobalVar
to be used by other places.

memprof_rtl.cpp calls InitializeShadowMemory() - which
dynamically/"randomly" chooses a base address for the shadow mapping -
prior to InitializeAllocator(). If we are unlucky, the shadow memory may
be mapped in the same region where the allocator wants to be.

This patch fixes the issue by changing the allocator to dynamically
choosing a base address, as suggested by Vitaly. For comparison, HWASan
already dynamically chooses the base addresses for the shadow mapping
and allocator.

The "unlucky" failure was observed on a new buildbot:
https://lab.llvm.org/buildbot/#/builders/66/builds/1361/steps/17/logs/stdio

---------

Co-authored-by: Vitaly Buka <vitalybuka@gmail.com>

…8427)

Fixes #92530.

… CoroBegin (#98546)

…pecific CSRs (#97287)

This PR is a follow-up of PR #96174 which added the framework to resolve
encoding conflicts among vendor specific CSRs. This PR explicitly
enables this only for the RISCV target.

The precedence of `==` operator is superior to `?:` operator. This line
is evaluated as:

```cpp
assert((ArgTranslations.size() == F->isVarArg()) ? 5 : PassthroughArgSize);
```
I guess this is not what is wanted. This causes a warning with gcc:
```
[131/602] Building CXX object lib/Target/AArch64/CMakeFiles/LLVMAArch64CodeGen.dir/AArch64Arm64ECCallLowering.cpp.o
In file included from /usr/include/c++/11/cassert:44,
                 from /home/linux/dev/llvm-project/llvm/include/llvm/Support/CommandLine.h:33,
                 from /home/linux/dev/llvm-project/llvm/lib/Target/AArch64/AArch64Arm64ECCallLowering.cpp:31:
/home/linux/dev/llvm-project/llvm/lib/Target/AArch64/AArch64Arm64ECCallLowering.cpp: In member function ‘llvm::Function* {anonymous}::AArch64Arm64ECCallLowering::buildEntryThunk(llvm::Function*)’:
/home/linux/dev/llvm-project/llvm/lib/Target/AArch64/AArch64Arm64ECCallLowering.cpp:528:50: warning: ‘?:’ using integer constants in boolean context [-Wint-in-bool-context]
  528 |   assert(ArgTranslations.size() == F->isVarArg() ? 5 : PassthroughArgSize);
      |          ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~^~~~~~~~~~~~~~~~~~~~~~~~
```
Add parenthesis to fix the problem.

… by \"

This reverts commit f8b1ca4.
It incorrectly replaces `t` in `blt` to `h`.

```
.altmacro
.macro gen t
  blt 2f
2:
.endm

gen h
```

Fix #98558

Before the patch `fixed-shadow.c` test died with an obscure SEGV, 
because shadow was mapped over libc.so.

Note, FindDynamicShadowStart is expected to select in available region.

This patch adds the following member functions:
- User::setOperand()
- User::replaceUsesOfWith()
- Value::replaceAllUsesWith()
- Value::replaceUsesWithIf()

This is a support data structure that acts as a cache for replacer-like
functions that map values between two domains. The difference compared
to just using a map to cache in-out pairs is that this class is able to
handle replacer logic that is self-recursive (and thus may cause
infinite recursion in the naive case).

This class provides a hook for the user to perform cycle pruning when a
cycle is identified, and is able to perform context-sensitive caching so
that the replacement result for an input that is part of a pruned cycle
can be distinct from the replacement result for the same input when it
is not part of a cycle.

In addition, this class allows deferring cycle pruning until specific
inputs are repeated. This is useful for cases where not all elements in
a cycle can perform pruning. The user still must guarantee that at least
one element in any given cycle can perform pruning. Even if not, an
assertion will eventually be tripped instead of infinite recursion (the
run-time is linearly bounded by the maximum cycle length of its input).

In the .altmacro mode, an argument can be expanded even if not preceded
by \. This behavior is not enabled for Darwin, which uses $
(`isIdentifierChar('$')` is true) for macro expansion.

This is f8b1ca4 with a fix.

…ribute. NFC

Remove mention of sharing with new and old PM. The old PM code is gone.

Argument promotion doesn't handle recursive function calls to promote
arguments. This patch adds functionality to handle self recursive
function calls, i.e. whose SCC size is 1. Due to complexity of
ValueTracking in recursive calls with SCC size greater than 1, we bail
out in such cases.

Co-authored-by: vporpo <vporpodas@google.com>

This patch introduces DeclBase::isInNamedModule API to ease the use
of modules slightly.

This patch updates the function `getReductionPatternCost` to handle the
cost of min/max reductions by `TTI.getMinMaxReductionCost`.

Close #98583

Currently, clang will reject the following code:

```
export module mod;
extern "C++" void func();
export extern "C++" {
    void func();
}
```

while both MSVC and GCC accept it. Although clang's behavior matches the
current wording, from the discussion, the consensus is that we should
accept the above example from the intention. Since the intention to not
allow export redeclaration which is not exported is to make the linkage
clear. But it doesn't matter with the declarations within global module.

This patch fixes:

  compiler-rt/lib/hwasan/hwasan_report.cpp:331:57: error: format
  specifies type 'void *' but the argument has type 'const uptr *'
  (aka 'const unsigned long *') [-Werror,-Wformat-pedantic]

…for variable llvm.memcpy" (#98482)

Reverts #98295, which reverted #97998

The failure in the "InOneWeekend" test of the HIP test suite on
clang-hip-vega20
(https://lab.llvm.org/buildbot/#/builders/123/builds/1498) seems to be
unrelated; I observed it (and a similar failure for the "TheNextWeek"
test in the same suite) intermittently on my system, with and without
the patch applied. (It occurred in 2 out of 50 repeated runs without the
patch and in 1 out of 50 runs with the patch.)

This patch adds a check for the correct number of `loops` results of the
`transform.structured.tile_using_for` Op to the verifier, fixing a
crash.

Fix #98008

…laration" (#98593)

Reverts #98075

bots are broken

Patch [3/x] to fix structured bindings debug info in SROA.

This function computes a fragment, bit-extract operation if needed, and new
constant offset to describe a part of a variable covered by some memory.

This generalises, simplifies, and replaces at::calculateFragmentIntersect. That
version is still used as a wrapper for now though to keep this change NFC.

The new version takes doesn't have a DbgRecord parameter, instead using an
explicit address and address offset. The old version only operates on
dbg_assigns and this change means it can also operate on dbg_declare records
easily, which it will do in a subsequent patch.

The new version has a new out-param OffsetFromLocationInBits which is set to
the difference between the first bit of the variable location and the first
bit of the memory slice. This will be used in a subsequent patch in SROA to
determine the new offset to use in the address expression after splitting an
alloca.

- Add `MachineBlockFrequencyAnalysis`.
- Add `MachineBlockFrequencyPrinterPass`.
- Use `MachineBlockFrequencyInfoWrapperPass` in legacy pass manager.
- `LazyMachineBlockFrequencyInfo::print` is empty, drop it due to new
pass manager migration.

https://llvm.org/docs/CodingStandards.html tells us that we should avoid
evaluating `.end()` each time if possible.

Missing `-mtriple` in test, causes failure on some none x86 default
targets.

…CE_DIR is set (#98464)

This was found during testing of llvm snapshots for Fedora.

This test was looking for an exact string match of the path calculated
by starting with lib/liblldb and with bin/lldb. However when
CLANG_RESOURCE_DIR is set to something e.g. "../lib/clang/19", the way
the initial path is handled is different.

Instead of taking the parent of the parent of the binary, that is
foo/bin/lldb/ -> foo/, it uses the parent of the binary and appends
CLANG_RESOURCE_DIR to that. As CLANG_RESOURCE_DIR is defined as being a
path relative to the parent dir's of the clang binary.

This means that if you start with foo/lib/lidblldb the resulting path is
lib/../lib/clang/19, but if you start with bin/lldb the result is
bin/../lib/clang/19.

I don't want to change the starting path of
DefaultComputeClangResourceDirectory (which is bin/lldb) as I suspect
that's chosen instead of liblldb for good reason.

So the way to make this test work is to check not for exact path matches
but that the "real" (".." resolved) version of the paths are the same.
That way foo/bin/../lib and foo/lib/../lib will be the same.

…6218)

Restrict `DropInnerMostUnitDimsTransfer{Read|Write}` so that it fails
when one of the indices to be dropped could be != 0 and "out of bounds":

```mlir
func.func @negative_example(%arg0: memref<16x1xf32>, %arg1: vector<8x1xf32>, %idx_1: index, %idx_2: index) {
  vector.transfer_write %arg1, %arg0[%idx_1, %idx_2] {in_bounds = [true, false]} : vector<8x1xf32>, memref<16x1xf32>
  return
}
```

This is an edge case that could represent an out-of-bounds access,
though that will depend on the actual value of %i. Importantly, without
this change it would be transformed as follows:

```mlir
func.func @negative_example(%arg0: memref<16x1xf32>, %arg1: vector<8x1xf32>, %arg2: index, %arg3: index) {
  %subview = memref.subview %arg0[0, 0] [16, 1] [1, 1] : memref<16x1xf32> to memref<16xf32, strided<[1]>>
  %0 = vector.shape_cast %arg1 : vector<8x1xf32> to vector<8xf32>
  vector.transfer_write %0, %subview[%arg2] {in_bounds = [true]} : vector<8xf32>, memref<16xf32, strided<[1]>>
  return
}
```

This is incorrect - `%idx_2` is ignored and the "out of bounds" flags is
not propagated. Hence the extra restriction to avoid such cases.

NOTE: This is a follow-up for: #94904

…rict DWARF v2 mode (#98335)

During testing of #96202 we
found that when clang set to DWARF v2 was used to build the test file,
lldb could not tell that the unsigned enum type was in fact unsigned. So
it defaulted to signed and printed the wrong value.

The reason for this is that DWARFv2 does not include DW_AT_type in
DW_TAG_enumeration_type. This was added in DWARF v3:
"The enumeration type entry may also have a DW_AT_type attribute which
refers to the underlying data type used to implement the enumeration.

In C or C++, the underlying type will be the appropriate integral type
determined by the compiler from the properties of the enumeration
literal values."

I noticed that gcc does emit this attribute for DWARF v2 but not when
strict DWARF is requested (more details in
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=16063#c7).

This patch changes to clang to do the same. This will improve the
experience of anyone using tools that can understand the attribute but
for whatever reason are stuck building binaries containing v2 only.

You can see a current clang/gcc comparison here:
https://godbolt.org/z/eG9Kc9WGf

https://reviews.llvm.org/D42734 added the original code that emitted
this for >= v3 only.

To reduce build times move them to TargetCodeGenInfo.

Refactor of #98329

Use Python's builtin enum class instead of writing our own.

This is preparation for passing a strict type check in PR #78114 ,
fixing 927 out of 1341 strict typing errors

---------

Co-authored-by: Jannick Kremer <jannick-kremer@gmx.de>
Co-authored-by: Vlad Serebrennikov <serebrennikov.vladislav@gmail.com>

… logical shift nodes #83840 (#86922)

Resolve #83840

This shares code with WsloopOp (the changes to Wsloop should be NFC).
OpenMPIRBuilder basically implements SECTIONS as a wsloop over a case
statement with each SECTION as a case for a particular loopiv value.

Unfortunately it proved very difficult to share code between these and
ParallelOp. ParallelOp does quite a few things differently (doing more
work inside of the bodygen callback and laying out blocks differently).
Aligning reduction implementations for wsloop and parallel will probably
involve functional changes to both, so I won't attempt that in this
commit.