[libomptarget][amdgcn] Implement missing symbols in deviceRTL

Malloc, wtime are stubs. Malloc needs a hostrpc implementation which is
a work in progress, wtime needs some experimentation to find out the
multiplier to get a time in seconds as documentation is scarce.

Reviewed By: ronlieb

Differential Revision: https://reviews.llvm.org/D89725

Revert "Fix compiler-rt build on Windows after D89640"

This reverts commit a7acee89.
This reverts commit d09b0891.

Reason: breaks Linux / x86_64 build.

Update `Lexer` / `Lexer::Lexer` to use `MemoryBufferRef` instead of
`MemoryBuffer*`. Callers that were acquiring a `MemoryBuffer*` via
`SourceManager::getBuffer` were updated, such that if they checked
`Invalid` they use `getBufferOrNone` and otherwise `getBufferOrFake`.

Differential Revision: https://reviews.llvm.org/D89398

Now, convert-shape-to-std doesn't internally create memrefs, which was
previously a bit of a layering violation. The conversion to memrefs
should logically happen as part of bufferization.

Differential Revision: https://reviews.llvm.org/D89669

The framework was not tracking ops created in any regions of the cloned
op.

Differential Revision: https://reviews.llvm.org/D89668

It's unfortunate that this requires adding a dependency on scf dialect
to std bufferization (and hence all of std transforms). This is a bit
perilous. We might want a lib/Transforms/Bufferize/ with a separate
bufferization library per dialect?

Differential Revision: https://reviews.llvm.org/D89667

Add bufferizations for extract_element and tensor_from_elements.

Differential Revision: https://reviews.llvm.org/D89594

Measure amount of high-level or fixed-cost operations performed during
building/loading modules and during header search. High-level operations
like building a module or processing a .pcm file are motivated by
previous issues where clang was re-building modules or re-reading .pcm
files unnecessarily. Fixed-cost operations like `stat` calls are tracked
because clang cannot change how long each operation takes but it can
perform fewer of such operations to improve the compile time.

Also tracking such stats over time can help us detect compile-time
regressions. Added stats are more stable than the actual measured
compilation time, so expect the detected regressions to be less noisy.

On relanding drop stats in MemoryBuffer.cpp as their value is pretty low
but affects a lot of clients and many of those aren't interested in
modules and header search.

rdar://problem/55715134

Reviewed By: aprantl, bruno

Differential Revision: https://reviews.llvm.org/D86895

GFX10 enables third addressing mode for flat scratch instructions,
an ST mode. In that mode both register operands are omitted and
only swizzled offset is used in addition to flat_scratch base.

Differential Revision: https://reviews.llvm.org/D89501

Renamed ThreadIntelPT to TreaceThread, making it a top-level class. I noticed that this class can and shuld work for any trace plugin and there's nothing intel-pt specific in it.
With that TraceThread change, I was able to move most of the json file parsing logic to the base class TraceSessionFileParser, which makes adding new plug-ins easier.

This originally was part of https://reviews.llvm.org/D89283

Differential Revision: https://reviews.llvm.org/D89408

Before the change attempt to link libLTO.so against shared
LLVM library failed as:

```
[ 76%] Linking CXX shared library ../../lib/libLTO.so
... /usr/bin/cmake -E cmake_link_script CMakeFiles/LTO.dir/link.txt --verbose=1
c++ -o ...libLTO.so.12git ...ibLLVM-12git.so
ld: CMakeFiles/LTO.dir/lto.cpp.o: in function `llvm::InitializeAllTargetInfos()':
include/llvm/Config/Targets.def:31: undefined reference to `LLVMInitializeVETargetInfo'
```

It happens because on linux llvm build system sets default
symbol visibility to "hidden". The fix is to set visibility
back to "default" for exported APIs with LLVM_EXTERNAL_VISIBILITY.

Bug: https://bugs.llvm.org/show_bug.cgi?id=47847

Reviewed By: simoll

Differential Revision: https://reviews.llvm.org/D89633

This recommits 7f1f89ec and
40df06cd with bug fixes for
memory sanitizer failure and Tensile build failure.

PartialDiagnostic misses some functions compared to DiagnosticBuilder.

This patch refactors DiagnosticBuilder and PartialDiagnostic, extracts
the common functionality so that the streaming << operators are
shared.

Differential Revision: https://reviews.llvm.org/D84362

Port pass to NPM and update tests in DebugInfo/Generic.

Differential Revision: https://reviews.llvm.org/D89730

The main tricky thing here is forward-declaring the enum:
we have to specify it's underlying data type.

In particular, this avoids the danger of switching over the SCEVTypes,
but actually switching over an integer, and not being notified
when some case is not handled.

I have updated most of such switches to be exaustive and not have
a default case, where it's pretty obvious to be the intent,
however not all of them.

Makes diff in next commit more readable

If we switch over an enum, compiler can easily issue a diagnostic
if some case is not handled. However with an if cascade that isn't so.
Experimental evidence suggests new behavior to be superior.

D18714 introduced writeonly attribute:

"Also start using the attribute for memset, memcpy, and memmove intrinsics,
and remove their special-casing in BasicAliasAnalysis."

But actually, writeonly was not attached to memmove - oversight, it seems.

So let's add it. As we can see, this helps DSE to eliminate redundant stores.

Reviewed By: jdoerfert

Differential Revision: https://reviews.llvm.org/D89724

Move the functions to the helper header and keep the arch specific
logic there.

Differential Revision: https://reviews.llvm.org/D89681

Use fs::path as variable type instead of std::string, when the input
potentially is a path, as they can't be implicitly converted back to
string.

Differential Revision: https://reviews.llvm.org/D89674

Use .string() instead of .native() in places where we want to combine
paths with std::string.

Convert some methods to take a fs::path as parameter instead of
std::string, for cases where they are called with paths as
parameters (which can't be implicitly converted to std::string if
the path's string_type is wstring).

Differential Revision: https://reviews.llvm.org/D89530

This fixes/silences a few failures on libstdc++ on linux.

Differential Revision: https://reviews.llvm.org/D89676

Differential Revision: https://reviews.llvm.org/D89673

Differential Revision: https://reviews.llvm.org/D89675

Mark this as a libcpp specific test; the standard doesn't say that
this method should be noexcept.

Differential Revision: https://reviews.llvm.org/D89677

The standard doesn't declare this overload as noexcept, but doesn't
either say that it strictly cannot be noexcept either. The function
doesn't throw on errors that are signaled via error_code, but the
standard says that it may throw a bad_alloc.

This fixes an error with libstdc++ on linux.

Differential Revision: https://reviews.llvm.org/D89678

This makes them more readable in llvm-lit's output on failures.

This only applies the change on the filesystem test subdir.

Differential Revision: https://reviews.llvm.org/D89680

Differential Revision: https://reviews.llvm.org/D89702

Update clang/lib/Format and clang/lib/Rewrite to use a `MemoryBufferRef`
from `getBufferOrFake` instead of `MemoryBuffer*` from `getBuffer`.

No functionality change here, since the call sites weren't checking if
the buffer was valid.

Differential Revision: https://reviews.llvm.org/D89406

Summary:
Initializer merging generates pretty inefficient code for large allocas
that also happens to trigger an exponential algorithm somewhere in
Machine Instruction Scheduler. See https://bugs.llvm.org/show_bug.cgi?id=47867.

This change adds an upper limit for the alloca size. The default limit
is selected such that worst case size of memtag-generated code is
similar to non-memtag (but because of the ISA quirks, this case is
realized at the different value of alloca size, ex. memset inlining
triggers at sizes below 512, but stack tagging instructions are 2x
shorter, so limit is approx. 256).

We could try harder to emit more compact code with initializer merging,
but that would only affect large, sparsely initialized allocas, and
those are doing fine already.

Reviewers: vitalybuka, pcc

Subscribers: llvm-commits

The NPM runs SpeculateAroundPHIs which breaks critical edges, causing a
branch we check for to not directly jump back to the same block.

This enables these transforms for vectors:
(ctpop x) u< 2 -> (x & x-1) == 0
(ctpop x) u> 1 -> (x & x-1) != 0
(ctpop x) == 1 --> (x != 0) && ((x & x-1) == 0)
(ctpop x) != 1 --> (x == 0) || ((x & x-1) != 0)

All enabled if CTPOP isn't Legal. This differs from the scalar
behavior where the first two are done unconditionally and the
last two are done if CTPOP isn't Legal or Custom. The Legal
check produced better results for vectors based on X86's
custom handling. Might be worth re-visiting scalars here.

I disabled the looking through truncate for vectors. The
code that creates new setcc can use the same result VT as the
original setcc even if we truncated the input. That may work
work for most scalars, but definitely wouldn't work for vectors
unless it was a vector of i1.

Fixes or at least improves PR47825

Reviewed By: spatel

Differential Revision: https://reviews.llvm.org/D89346

We have pseudo instructions we use for bitcasts between these types.
We have them in the load folding table, but not the store folding
table. This adds them there so they can be used for stack spills.

I added an exact size check so that we don't fold when the stack slot
is larger than the GPR. Otherwise the upper bits in the stack slot
would be garbage. That would be fine for Eli's test case in PR47874,
but I'm not sure its safe in general.

A step towards fixing PR47874. Next steps are to change the ADDSSrr_Int
pseudo instructions to use FR32 as the second source register class
instead of VR128. That will keep the coalescer from promoting the
register class of the bitcast instruction which will make the stack
slot 4 bytes instead of 16 bytes.

Reviewed By: RKSimon

Differential Revision: https://reviews.llvm.org/D89656

AA computes the correct result for phi/a1 aliasing, while BatchAA
produces an incorrect result depening on which queries have been
performed beforehand.