This reverts commit 88d7508d.
It's reported to break builds when symlinking other projects inside
the `tools` directory.

Update the Python tests (ie tests run via `lldb-dotest -p TestTrace`) to
handle new error introduced in D136610.

Test Plan:
`lldb-dotest -p TestTrace`

Differential Revision: https://reviews.llvm.org/D136801

The fix for the unitest case introduced a dependency on the
MC library causing a failure in:

  https://lab.llvm.org/buildbot/#/builders/121/builds/24567
  clang-ppc64le-multistage/stage1
  undefined reference to symbol 'llvm::TargetRegistry::lookupTarget'

Added:
- MC to the LLVM_LINK_COMPONENTS list.

Reviewed By: jryans

Differential Revision: https://reviews.llvm.org/D136837

Differential Revision: https://reviews.llvm.org/D136836

[fixed test to work with reverse iteration]

The `FunctionSpecialization` pass has support for specialising
functions, which are called with literal arguments. This functionality
is disabled by default and is enabled with the option
`-function-specialization-for-literal-constant` .  There are a few
issues with the implementation, though:

* even with the default, the pass will still specialise based on
   floating-point literals

* even when it's enabled, the pass will specialise only for the `i1`
    type (or `i2` if all of the possible 4 values occur, or `i3` if all
    of the possible 8 values occur, etc)

The reason for this is incorrect check of the lattice value of the
function formal parameter. The lattice value is `overdefined` when the
constant range of the possible arguments is the full set, and this is
the reason for the specialisation to trigger. However, if the set of
the possible arguments is not the full set, that must not prevent the
specialisation.

This patch changes the pass to NOT consider a formal parameter when
specialising a function if the lattice value for that parameter is:

* unknown or undef
* a constant
* a constant range with a single element

on the basis that specialisation is pointless for those cases.

Is also changes the criteria for picking up an actual argument to
specialise if the argument is:

* a LLVM IR constant
* has `constant` lattice value
 has `constantrange` lattice value with a single element.

Reviewed By: ChuanqiXu

Differential Revision: https://reviews.llvm.org/D135893

Change-Id: Iea273423176082ec51339aa66a5fe9fea83557ee

Move `cmake_policy()` settings from `llvm/CMakeLists.txt` into a shared
`cmake/modules/CMakePolicy.cmake`.  Include it from all relevant
projects that support standalone builds, in order to ensure that
the policies are consistently set whether they are built in-tree
or stand-alone.

Differential Revision: https://reviews.llvm.org/D136572

`AffineMap.dropResult` erases one result from the array and it changes indexing. Calling `dropResult` is a loop with increasing indexes does not produce a desired result.

Differential Revision: https://reviews.llvm.org/D136833

We should not delete block predecessors (via replacing successors
of terminators) while iterating them, otherwise we may skip some
of them. Instead, save predecessors to a separate vector and iterate
over it.

This change adds memory space support to tensor.pad. (tensor.generate and tensor.from_elements do not support memory spaces yet.)

The memory space is inferred from the buffer of the source tensor.

Instead of lowering tensor.pad to tensor.generate + tensor.insert_slice, it is now lowered to bufferization.alloc_tensor (with the correct memory space) + linalg.map + tensor.insert_slice.

Memory space support for the remaining two tensor ops is left for a later point, as this requires some more design discussions.

Differential Revision: https://reviews.llvm.org/D136265

This should have been part of D136767.

There is no memref equivalent of tensor.generate. The purpose of this change is to avoid creating scf.parallel loops during bufferization.

Differential Revision: https://reviews.llvm.org/D136767

The writeonly attribute for memset_pattern16 (and other referenced
libcalls) is being added by InferFunctionAttrs nowadays. No need
to special-case it here.

Fixes: https://github.com/llvm/llvm-project/issues/50886

**Adding requires clause to template head** or **constraining the template parameter type** is ineffective because, even though it creates a non-equivalent template head [temp.over.link#6](https://eel.is/c++draft/temp.over.link#6) and hence eligible for overload resolution, `Derived::foo` still [hides any previous using decl](https://github.com/llvm/llvm-project/blob/main/clang/lib/Sema/SemaOverload.cpp#L1283-L1301,).
Clang diverges from gcc here and can be seen more clearly in this example:
```
struct base {
  template <int N, int M>
  int foo() { return 1; };
};

struct bar : public base {
  using base::foo;
  template <int N>
  int foo() { return 2; };
};

int main() {
  bar f;
  f.foo<10, 10>(); // clang previously errored while GCC does not.
}
```
https://godbolt.org/z/v5hnh6czq. We see that `bar::foo` hides `base::foo` because it only differs in the head.

Adding a trailing `requires` to the definition was a nice find. In this case, clang considers them [overloads](https://github.com/llvm/llvm-project/blob/main/clang/lib/Sema/SemaOverload.cpp#L1148-L1152) because of [mismatching requires clause.](https://github.com/llvm/llvm-project/blob/main/clang/lib/Sema/SemaOverload.cpp#L1390-L1405). So both of them make it to the overload resolution (where constrained Derived::foo is rejected then).

---

In this patch, we do not ignore matching the template head (template parameters, type contraints and trailing requires) while considering whether the using decl of base member should be hidden. The return type of a templated function is still not considered as different return types would create ambiguous candidates.

The changed tests looks reasonable and also matches GCC behaviour: https://godbolt.org/z/8KqPEThrY

Note: We are now able to create an ambiguity in case where both base member and derived member specialisations satisfy the constraints (when the constraints are not same). Ideally using-decl should not create ambiguity. I plan to fix this later if it gathers more attention.

Reviewed By: ilya-biryukov, #clang-language-wg

Differential Revision: https://reviews.llvm.org/D136440

This change fixes a bug where a dialect is initialized multiple times. This triggers an assertion when the ops of the dialect are registered (`error: operation named ... is already registered`).

This bug can be triggered as follows:

1. Dialect A depends on dialect B (as per ADialect.td).

2. Somewhere there is an extension of dialect B that depends on dialect A (e.g., it defines external models create ops from dialect A). E.g.:
```
registry.addExtension(+[](MLIRContext *ctx, BDialect *dialect) {
  BDialectOp::attachInterface ...
  ctx->loadDialect<ADialect>();
});
```

3. When dialect A is loaded, its `initialize` function is called twice:

```
     ADialect::ADialect()
        |     |
        |     v
        |   ADialect::initialize()
        v
     getOrLoadDialect<BDialect>()
        |
        v
     (load extension of BDialect)
        |
        v
     ctx->loadDialect<ADialect>()  // user wrote this in the extension
        |
        v
     getOrLoadDialect<ADialect>()  // the dialect is not "fully" loaded yet
        |
        v
     ADialect::ADialect()
        |
        v
     ADialect::initialize()
```

An example of a dialect extension that depends on other dialects is `Dialect/Tensor/Transforms/BufferizableOpInterfaceImpl.cpp`. That particular dialect extension does not trigger this bug. (It would trigger this bug if the SCF dialect would depend on the Tensor dialect.)

This change introduces a new dialect state: dialects that are currently being loaded. Same as dialects that were already fully loaded (and initialized), dialects that are in the process of being loaded are not loaded a second time.

Differential Revision: https://reviews.llvm.org/D136685

For more details about these instructions, please refer to the latest ISE document: https://www.intel.com/content/www/us/en/develop/download/intel-architecture-instruction-set-extensions-programming-reference.html

Initial authored by Liu Chen (@LiuChen3)

Reviewed By: RKSimon

Differential Revision: https://reviews.llvm.org/D135951

This simplifies the BufferizableOpInterface implementation of vector.transfer_write.

Differential Revision: https://reviews.llvm.org/D136348

The unitest and test cases are platform dependent (x86_64)
causing failures in:

  https://lab.llvm.org/buildbot/#/builders/245/builds/146
  https://lab.llvm.org/buildbot/#/builders/188/builds/21397
  No available targets are compatible with triple "x86_64-unknown-unknown".

Added:
- ';REQUIRES: x86-registered-target' to the LIT tests.
- Code to check if the target 'Triple::x86_64' is supported to
  the unittest case.

This is a pre-commit for the fix in D136784.

Reviewed By: SjoerdMeijer

Differential Revision: https://reviews.llvm.org/D136783

Pre-commit test case for D126043

Reviewed By: Meinersbur, #loopoptwg

Differential Revision: https://reviews.llvm.org/D134823

tensor.insert and tensor.insert_slice (as destination style ops) do no longer need to implement the entire BufferizableOpInterface.

Differential Revision: https://reviews.llvm.org/D136347

GDB implemented data_bit_offset in https://sourceware.org/bugzilla/show_bug.cgi?id=12616
which has been present since GDB 8.0.

GCC started using it at GCC 11.

Reviewed By: dblaikie

Differential Revision: https://reviews.llvm.org/D135583

There are a lot of cases for pseudos of the same instruction, here
we just use existed mapping table to map pseudos to real instructions
to reduce cases.

Reviewed By: kito-cheng

Differential Revision: https://reviews.llvm.org/D128271

And update the unittest.

Reviewed By: dblaikie

Differential Revision: https://reviews.llvm.org/D136242

Rename module related things according to the consensus in
https://discourse.llvm.org/t/rfc-unifying-the-terminology-about-modules-in-clang/66054/
to reduce further confusings.

This only renames things I can make sure. It doesn't  mean all the names
in Preprocessor are correct now.

This patch seems to introduce bugs on aarch64.
Reverting while we investigate the root cause.

This reverts commit 02841488.

Clang detects the GCC version from the libdir. However, modern
GCC versions only include the major version in the libdir
(something like lib/gcc/powerpc64le-linux-gnu/12/), not all
version components. For this reason, even though the system has
a supported libstdcxx, it will still fail the check against the
12.1.0 version requirement.

Fix this by doing the same thing we do for patch versions: Assume
that a missing minor version is larger than any specific version.

To allow this to be tested, we need to fix two additional issues:
First, the GCC toolchain directories used for testing need to
contain a crtbegin.o file to be properly detected. The existing
tests actually ended up using a 0.0.0 version, rather the intended
one. Second, we also need to satisfy the glibc version check based
on the dynamic linker. To do so, respect the --dyld-prefix argument
and add the necessary file to the test toolchain directory.

Differential Revision: https://reviews.llvm.org/D136258

This reverts commit c34de60e.

Also allow unranked tensors/memrefs with destination style op outputs.

This allows for a simpler implementation of the BufferizableOpInterface (in a subsequent commit).

Differential Revision: https://reviews.llvm.org/D136346

When looking through phis, BasicAA has to guard against the
possibility that values from two separate cycle iterations are
being compared -- in this case, even though the SSA values may
be the same, they cannot be considered as equal.

This is currently done by keeping a set of VisitedPhiBBs for any
phis we looked through, and then checking whether the relevant
instruction is reachable from one of the phis.

This patch replaces this set with a single flag. If the flag is
set, then we will not assume equality for any instruction part
of a cycle. While this is nominally less accurate, it makes
essentially no difference in practice. Here are the AA stats
for test-suite:

    aa.NumMayAlias  |   3072005 |  3072016
    aa.NumMustAlias |    337858 |   337854
    aa.NumNoAlias   |  13255345 | 13255349

The motivation for the change is to expose the MayBeCrossIteration
flag to AA users, which will allow fixing miscompiles related to
incorrect handling of cross-iteration AA queries.

Differential Revision: https://reviews.llvm.org/D136174

Also add line range examples from #58161.

Differential Revision: https://reviews.llvm.org/D136658

We have currently no need for unranked tensors/memrefs.

Differential Revision: https://reviews.llvm.org/D136588

While profiling InclusionRewriter, it was found that counting lines was
so slow that it took up 20% of the processing time. Surely, calling
memcmp() of size 1 on every substring in the window isn't a good idea.

Use StringRef::find() instead; in the case of N=1 it will forward to
memcmp which is much more optimal. For 2<=N<256 it will run the same
memcmp loop as we have now, which is still suboptimal but at least does
not regress anything.

Differential Revision: https://reviews.llvm.org/D133658

```
%mapped = linalg.map
  ins(%arg0 : tensor<64xf32>)
  outs(%arg1 : tensor<64xf32>)
  (%in: f32) {
    %0 = math.absf %in : f32
    linalg.yield %0 : f32
  }
%reduced = linalg.reduce
  ins(%arg0 : tensor<16x32x64xf32>)
  outs(%arg1 : tensor<16x64xf32>)
  dimensions = [1]
  (%in: f32, %init: f32) {
    %0 = arith.addf %in, %init : f32
    linalg.yield %0 : f32
  }
%transposed = linalg.transpose
  ins(%arg0 : tensor<16x32x64xf32>)
  outs(%arg1 : tensor<32x64x16xf32>)
  permutation = [1, 2, 0]
```

Differential Revision: https://reviews.llvm.org/D136818

When writing a tensor.extract/tensor.insert, the rank of the tensor is implied by the number of specified indices. When extracting from/inserting into an unranked tensor, it should first be casted to a ranked version.

Differential Revision: https://reviews.llvm.org/D136756

We have a downstream project with a command-line utility that operates
pretty much exactly like `opt`. So it would make sense for us to
maintain tests with update_test_checks.py with our custom tool
substituted for `opt`, as this change allows.

Differential Revision: https://reviews.llvm.org/D136329

createFragmentExpression rejects expressions containing certain ops, like
DW_OP_plus, that may cause the expression to compute a value that can't be
split.

Teach createFragmentExpression that the value loaded from an address computed
using those ops is safe to split.

Update a unittest to account for and test this change.

Reviewed By: StephenTozer

Differential Revision: https://reviews.llvm.org/D136243

This reverts commit a678f863.