Reviewed By: bondhugula

Differential Revision: https://reviews.llvm.org/D116672

This patch fixes:

  mlir/lib/Dialect/Linalg/ComprehensiveBufferize/ModuleBufferization.cpp:635:23:
  error: comparison of integers of different signs: 'int' and 'size_t'
  (aka 'unsigned long') [-Werror,-Wsign-compare]

This will allow linking in the callbacks directly instead of using PLT.

Reviewed By: vitalybuka

Differential Revision: https://reviews.llvm.org/D116182

We can either check the opcode or number of operands or use
ISD::isVPOpcode inside the methods.

In some places I've used number of operands figuring that it is
cheaper than isVPOpcode. I've included isVPOpcode in an assert to
verify.

Reviewed By: frasercrmck

Differential Revision: https://reviews.llvm.org/D116578

These 3 switches map LMUL enum to instruction names. These follow
a regular pattern. Use a macro to reduce the number of source code
lines.

Reviewed By: arcbbb

Differential Revision: https://reviews.llvm.org/D116631

stdcall name mangling requires a suffix with the number equal to the sum of the byte count of all parameter types. In the case of a function prototype that has a parameter type of an incomplete type it is impossible to get the size of the type. While such a function is not callable or able to be defined in the TU, it may still be mangled when generating debug info, which would previously lead to a crash.
This patch fixes that by simply bailing out of the loop and using the so far accumulated byte count. This matches GCCs behaviour as well: https://github.com/gcc-mirror/gcc/blob/bc8d6c60137f8bbf173b86ddf31b15d7ba2a33dd/gcc/config/i386/winnt.c#L203

Fixes https://github.com/llvm/llvm-project/issues/52782

Differential Revision: https://reviews.llvm.org/D116020

The naming has come up as a source of confusion in several recent reviews.  onlyWritesMemory is consist with onlyReadsMemory which we use for the corresponding readonly case as well.

lld-link used to consistently print all /verbose output to stdout, and that was
an intentional decision: https://reviews.llvm.org/rG4bce7bcc88f3

https://reviews.llvm.org/rGe6e206d4b4814 added message() and log(),
and back then `log()` morally was just `if (verbose) message(...)`
and message() wrote to stdout.

So that change moved most /verbose-induced writes to outs() to
log(). Except for the one in printDiscardedMessage(), since
the check for `verbose` for that one is in the caller, in
Writer::createSections():

    if (config->verbose)
      sc->printDiscardedMessage();

Later, https://reviews.llvm.org/D41033 changed log() to write to
stderr. That moved lld-link from writing all its /verbose output
to stdout to writing almost all of its /verbose output to stderr --
except for printDiscardedMessage() output.

This change moves printDiscardedMessage() to call log() as well,
so that all /verbose output once again consistently goes to the same
stream.

Differential Revision: https://reviews.llvm.org/D116667

This change extends the addressing mode optimization
pass to HVX vgather. This is specifically intended to
resolve compiler not generating indexed addresses for
vgather stores to vtcm. Changed the vgather pseudo
instructions to accept an immediate operand and handled
addition of appropriate immediate operand in addressing
mode optimization pass.

The fix in D116381 makes an existing exception message wrong. This
improves the message and fixes the associated unit tests.

Note other message can be also be improved, but that will be done later.
Changing these messages may cause merge conflicts with other patches
that are under review or WIP.

Depends on D116381

Reviewed By: #libc, Quuxplusone, ldionne

Differential Revision: https://reviews.llvm.org/D116495

@CaseyCarter reported that the tests for the std-format-spec rejects leading
zeroes for precision, which the Standard does not require. The Standard allows
them. Only for precision, not for the width or an arg-id.

Fixes the precision parser and adds some test for the arg-id since they
were missing.

Reviewed By: ldionne, #libc

Differential Revision: https://reviews.llvm.org/D116381

Fix typo (leftt -> left)

Reviewed By: curdeius

Differential Revision: https://reviews.llvm.org/D116658

Differential Revision: https://reviews.llvm.org/D116665

Main use of these is in the standard library, where they generally clutter up
the index.

Certain macros are also common, we don't touch indexing of macros in this patch.

Differential Revision: https://reviews.llvm.org/D115301

There is no need to inspect the ReturnOp of the called function.

This change also refactors the bufferization of CallOps in such a way that `lookupBuffer` is called only a single time. This is important for a later change that fixes CallOp bufferization. (There is currently a TODO among the test cases.)

Note: This change modifies a test case but is marked as NFC. There is no change of functionality, but FuncOps with empty bodies are now reported with a different error message.

Differential Revision: https://reviews.llvm.org/D116446

So far, only the custom dialect types are exposed.

The build and packaging is same as for Linalg and SparseTensor, and in
need of refactoring that is beyond the scope of this patch.

Reviewed By: stellaraccident

Differential Revision: https://reviews.llvm.org/D116605

mapped_file_region::dontNeedImpl added in D116366 calls madvise, which
causes problems for z/OS and AIX.

For z/OS, we don't have either madvise, so treat this as a no-op, same
as Windows does.

For AIX, it doesn't have any effect, doesn't have a standardized
signature, and it needs certain feature test macros (i.e. _ALL_SOURCE)
we don't set by default for LLVM on AIX, so just make it a no-op too.

Differential Revision: https://reviews.llvm.org/D116603

Reviewed By: nicolasvasilache

Differential Revision: https://reviews.llvm.org/D116546

Such CallOps were not handled properly. When computing the new result types (and replacement values) of a CallOp, non-tensor return values were not accounted for.

Differential Revision: https://reviews.llvm.org/D116445

Previously, the Python bindings for the Linalg dialect relied on the internal
implementation of core bindings. Most of that functionality was moved, and the
remaining one does not need access to the implementation: it used to accept a
dialect pointer as argument, but it can always be extracted from the operation
that it also accepts; operations are available through PybindAdaptors in an
opaque way. Change the bindings in that direction.

This enables the decoupling of the Linalg dialect Python extension from the
core IR Python extension.

Reviewed By: nicolasvasilache

Differential Revision: https://reviews.llvm.org/D116649

This change simplifies BufferizationState. Having `rewriter` in BufferizationState could be confusing to users because a rewriter is also passed to each `bufferize` function and it is not obvious (by looking at the API) that these two rewriters are the same.

Differential Revision: https://reviews.llvm.org/D116444

This family of instructions includes CPYF (copy forward), CPYB (copy
backward), SET (memset) and SETG (memset + initialise MTE tags), with
some sub-variants to indicate whether address translation is done in a
privileged or unprivileged way. For the copy instructions, you can
separately specify the read and write translations (so that kernels
can safely use these instructions in syscall handlers, to memcpy
between the calling process's user-space memory map and the kernel's
own privileged one).

The unusual thing about these instructions is that they write back to
multiple registers, because they perform an implementation-defined
amount of copying each time they run, and write back to _all_ the
address and size registers to indicate how much remains to be done
(and the code is expected to loop on them until the size register
becomes zero). But this is no problem in LLVM - you just define each
instruction to have multiple outputs, multiple inputs, and a set of
constraints tying their register numbers together appropriately.

This commit introduces a special subtarget feature called MOPS (after
the name the spec gives to the CPU id field), which is a dependency of
the top-level 8.8-A feature, and uses that to enable most of the new
instructions. The SETMG instructions also depend on MTE (and the test
checks that).

Differential Revision: https://reviews.llvm.org/D116157

Other tools take their tool name from argv[0] for use in output
messages. This change makes llvm-strings consistent with other tools
rather than using a hard coded value.

Differential Revision: https://reviews.llvm.org/D116604

Because declarators nest inside-out, we logically need to claim tokens for
parent declarators logically before child ones.
This is the ultimate reason we had problems with DeclaratorDecl, ArrayType etc.

However actually changing the order of traversal is hard, especially for nodes
that have both declarator and non-declarator children.
Since there's only a few TypeLocs corresponding to declarators, we just
have them claim the exact tokens rather than rely on nesting.

This fixes handling of complex declarators, like
`int (*Fun(OuterT^ype))(InnerType);`.

This avoids the need for the DeclaratorDecl early-claim hack, which is
removed.
Unfortunately the DeclaratorDecl early-claims were covering up an AST
anomaly around CXXConstructExpr, so we need to fix that up too.

Based on D116623 and D116618

Differential Revision: https://reviews.llvm.org/D116630

LICM checks that nested ops depend only on values defined outside
before performing hoisting.
However, it specifically omits to check for terminators which can
lead to SSA violations.
This revision fixes the incorrect behavior.

Differential Revision: https://reviews.llvm.org/D116657

VPWidenMemoryInstructionRecipe is a VPValue, so this can be passed
directly, instead of relying on getVPSingleValue.

Differential Revision: https://reviews.llvm.org/D116648

The check should not trigger on lvalue/rvalue overload pairs:

```
struct S {
  S(const A& a) : a(a) {}
  S(A&& a) : a(std::move(a)) {}

  A a;
}
```

Differential Revision: https://reviews.llvm.org/D116535

std::make_reverse_iterator is a C++14 feature, gcc has it since GCC 5.1.

We should bail out if the index is >= the size, not > the size.

Fixes https://github.com/llvm/llvm-project/issues/53002.