This diff adds support for TF32 as a Builtin floating point type. This
supplements the recent addition of the TF32 semantic to the LLVM APFloat class
by extending usage to MLIR.

https://reviews.llvm.org/D151923

More information on the TF32 type can be found here:

https://blogs.nvidia.com/blog/2020/05/14/tensorfloat-32-precision-format/

Reviewed By: jpienaar

Differential Revision: https://reviews.llvm.org/D153705

This accomplishes a few minor things:
- Removed unnecessary uses of `this->`
- Removed an unnecessary std::string allocation.
- Removed some nesting to improve readability using early returns where
  it makes sense.
- Replaced `strtoul` with `llvm::to_integer` which avoids another
  std::string allocation.
- Removed braces from single statement conditions, removed
  else-after-returns.

Differential Revision: https://reviews.llvm.org/D154534

Summary: No need to create a ConstString, `GetName` already returns a StringRef.

Reviewers: JDevlieghere, mib, jasonmolenda

Subscribers:

Differential Revision: https://reviews.llvm.org/D154386

Summary:

  Adding a new option -traceback-table to print out the traceback info of xcoff ojbect file.

Reviewers: James Henderson, Fangrui Song, Stephen Peckham, Xing Xue

Differential Revision: https://reviews.llvm.org/D89049

The next-gen plugins didn't correctly configure tests and were never
actually being run. Since deleting the old plugin we stopped getting
`libomptarget` tests. This patch fixes the issue and allows the targets
to be built

Reviewed By: JonChesterfield

Differential Revision: https://reviews.llvm.org/D154619

Review: https://reviews.llvm.org/D153883

This is the way most targets do it for a simple mapping.

We can't do this for all builtins due to type overloading of the IR intrinsics.

Reviewed By: asb

Differential Revision: https://reviews.llvm.org/D154567

This is required to bring CaptureTracking in line with the new
semantics from D154051, as gep inbounds p, 0 is now always non-poison.

There are many ways in which the inbounds special case could be
preserved: If the index is known non-zero, or there is an inbounds
chain down to an identified object, etc. However, I have opted to
drop the special case entirely, as it appears to be low value:
In cases where we can determine such things (e.g. the affected test
cases) we would end up removing the compare via isGEPKnownNonNull()
logic anyway.

Differential Revision: https://reviews.llvm.org/D154054

Previously we added both `-GR-` and `-fno-rtti` if RTTI was disabled.
When building with clang 16.x, that caused this error in part of the build:
```
clang-16: error: argument unused during compilation: '-G R-' [-Werror,-Wunused-command-line-argument]
```
I think the strange message is because clang is seeing `R-` as the argument
to `-G`, which is a valid clang option.

`-GR-` is an alternate syntax for the `/GR-` option for MSVC
(the dash means disable RTTI):
https://learn.microsoft.com/en-us/cpp/build/reference/gr-enable-run-time-type-information?view=msvc-170

This error is sort of fixed by cd18efb6
but not intentionally. Also, we'd have to wait for 17.x to benefit from that.

The proper fix here is to only add `-GR-` if we are building with MSVC
or the MSVC-like clang-cl, and add `-fno-rtti` if not.

Reviewed By: #libc, simon_tatham, michaelplatings, ldionne

Differential Revision: https://reviews.llvm.org/D154582

This ensures that the __Fortran_builtins module is always used
as an intrinsic module by __Fortran_type_info, which avoids issues
when, for instance, the intrinsic modules dir is present in the
include paths.

Fixes https://github.com/llvm/llvm-project/issues/63592

Reviewed By: klausler

Differential Revision: https://reviews.llvm.org/D154452

Fixes #63700

This matches the data type of the intrinsics. This case be seen
from the removal of sext and trunc instructions from the IR.

Reviewed By: asb

Differential Revision: https://reviews.llvm.org/D154577

We should look through implicit casts before determining the type of the arguments, and only allow explicit cast to _Nonnull to suppress warning

```
void foo(NSString *_Nonnull);

foo((NSString * _Nonnull)nil); // no-warning
id obj = nil;
foo(obj); // should warning here (implicit cast id to NSString *_Nonnull)

```

Reviewed By: xazax.hun, steakhal

Differential Revision: https://reviews.llvm.org/D154221

ParseStatus is slightly more convenient to use due to implicit
conversion from bool, which allows to do something like:
```
  return Error(L, "msg");
```
when with MatchOperandResultTy it had to be:
```
  Error(L, "msg");
  return MatchOperand_ParseFail;
```
It also has more appropriate name since parse* methods are not only for
parsing operands.

Reviewed By: MaskRay

Differential Revision: https://reviews.llvm.org/D154319

The itanium ABI for certain platforms requires a minimum alignments for
member function pointers to reserve certain bits for distinguishing
virtual and non-virtual functions.

Our implementation of this however depends on the alignment of the
function involved, which may however not reflect the true alignment of
function pointers on certain targets for which the alignment is
independent of the function (e.g. AIX). Worse, the 2-byte alignment
we use may be less than the ABI minimum for the target, and in the case
we are using explicit sections will result in invalid codegen.

This patch attempts to correct this situation by considering the target
alignment of function pointers as part of making the decision about
whether we need to adjust the function alignment to conform to the ABI.
Targets which do not provide the function ptr alignment information
will return a value of 1 when queried and will conservatively retain
the old alignment.

Differential Revision: https://reviews.llvm.org/D147184

Consider a setup without a system-wide installation of lit.
Instead you pass the path to lit like this:

```
cmake ...
-DLLVM_EXTERNAL_LIT=<PATH_TO_LIT_BINARY> ...
```

Then you will run into this error:

```
ninja: error: unknown target 'check-lld'
```

I have a buildbot builder that fails with this message. Here's the
passage that triggers this error:

https://github.com/llvm/llvm-zorg/blob/d3bfd5ccbceb542098c350e4d071ceceac6854cb/zorg/buildbot/builders/annotated/standalone-build.sh#L194-L239

By using `LLVM_EXTERNAL_LIT` instead of `LLVM_LIT` we fix this problem.

See
[here](https://llvm.org/docs/GettingStarted.html#stand-alone-builds) for
a description:

> Both the LLVM_ROOT and LLVM_EXTERNAL_LIT options are required to do stand-alone builds for all sub-projects. Additional required options for each sub-project can be found in the table below.

Differential Revision: https://reviews.llvm.org/D154599

Differential Revision: https://reviews.llvm.org/D154580

All this does is add uses of reserved registers, which
aren't tracked by anything. Saves a loop info computation.

The sign bit has no impact on the exponent, so strip these away. Saves
on the source modifier encoding cost. I left the GlobalISel handling
until there's a resolution to issue #62628.

We should do this in instcombine too, but legalization should be
introducing more frexps than it currently is where this would occur.

Patterns already exist for 64-bit that I've simply copied and
converted to include the necessary truncation.

Differential Revision: https://reviews.llvm.org/D154350

Verify that the correct number of `scalable_sizes` was provided.

Differential Revision: https://reviews.llvm.org/D154600

Test for https://github.com/llvm/llvm-project/issues/47927

The replacement op insertion point was off by one.

Differential Revision: https://reviews.llvm.org/D154608

D154070 has added /dwodir to lld/COFF to tells LTO backend to create dwo
directory and files. This patch makes clang to emit /dwodir to lld when
user specify -gsplit-dwarf with LTO. This behavior is simiar to DWARF
fission with LTO for ELF.

A simple use case:
$clang-cl -c -flto -gdwarf main.c -o main.o
$clang-cl -c -flto -gdwarf a.c -o a.o
$clang-cl -flto -fuse-ld=lld -gdwarf -gsplit-dwarf main.o a.o

This'll generate a dwo file: main.exe_dwo/0.dwo

Reviewed By: mstorsjo, MaskRay, hans

Differential Revision: https://reviews.llvm.org/D154295

The original plan was to turn this into its own pattern, but one of the difficulties was deeming when splitting the vector is required.
`SplitIntegerStores` essentially already did that by checking for field overlap.
Therefore, it was renamed to `SplitStores` and extended to splitting stores with values of vector and integer type.

The vector splitting is done in a simple manner by simply using `extractelement` to get each vector element. Subsequent pattern applications are responsible for further cleaning up the output and making it type-consistent.

Worst case, if the code cannot be transformed into a type-consistent form (due to e.g. the code explicitly doing partial writes to elements or similar), we might needlessly do a vector split.

Differential Revision: https://reviews.llvm.org/D154583

Jeremy Morse noted in D154451 that LLVM doesn't drop a `dbg.value` when its value is being removed, but rather sets the operand to `undef`. This preserves the debug info and gives allows the debugger to instead inform the user that the variable has been optimized out, rather than not displaying the variable at all.

This patch fixes that mistake done in the previous revision by mirroring that behaviour in MLIR as well.

Differential Revision: https://reviews.llvm.org/D154476

Add test coverage for demanded bit analysis of shl.

Added tests on range checks with non-strick predicate:
    * N - IV > limit
    * IV - N < limit
    * IV + N < limit

Also added tests with known to be non-negative N

Differential Revision: https://reviews.llvm.org/D154593

Extended BPFCheckAndAdjustIR pass with sinkMinMax() transformation
that undoes LICM hoistMinMax pass.

The undo transformation converts the following patterns:

    x < min(a, b) -> x < a && x < b
    x > min(a, b) -> x > a || x > b
    x < max(a, b) -> x < a || x < b
    x > max(a, b) -> x > a && x > b

Where 'a' or 'b' is a constant.
Also supports `sext min(...) ...` and `zext min(...) ...`.

Differential Revision: https://reviews.llvm.org/D147990

Include the function name + analysis when printing the analysis for
testing.

With the semantics change from D154051, it is no longer valid to
fold gep inbounds undef to poison (unless we know the index is
non-zero). Fold it to undef instead.

Differential Revision: https://reviews.llvm.org/D154215

Re-apply eda47fdd after implementing __truediv__ for TensorUse.

[MLIR][Linalg] Add more arith named ops to linalg

Following up the 'add' named op, here are the remaining basic arithmetic
and maths, including a 'div_unsigned' for integer unsigned values. In the
same pattern as 'matmul_unsigned', the simply named 'div' assumes signed
values and the '_unsigned' variation handles the unsigned values.

It's a bit odd, but there doesn't seem to be a easy way to restrict to
specific types to make 'div_unsigned' only work with integers in the
structured ops framework.

Same as 'add', these have strict semantics regarding casts.

Unary math ops will need some massaging, so I split these ones for now
as I continue working on them.

Differential Revision: https://reviews.llvm.org/D154524