I fixed some poison-safety violations on related patterns in InstCombine
and noticed that we missed adding nsw/nuw on them, so this adds clauses
to the underlying analysis for that.

We need the undef input restriction to make this safe according to Alive2:
https://alive2.llvm.org/ce/z/48g9K8

Differential Revision: https://reviews.llvm.org/D125500

This reverts commit ad47114a.
See discussion on https://reviews.llvm.org/D124613.

MSVC pragma function tells the compiler to generate calls to functions in the pragma function list, instead of using the builtin. Needs https://reviews.llvm.org/D124701

https://docs.microsoft.com/en-us/cpp/preprocessor/function-c-cpp?view=msvc-170

Reviewed By: aaron.ballman

Differential Revision: https://reviews.llvm.org/D124702

* Set MaxStoresPerMemcpy and MaxStoresPerMemset to 2.

* Optimize stores of replicated values in SystemZ::combineSTORE(). This
  handles the now expanded memory operations and as well some other
  pre-existing cases.

* Reject a big displacement in isLegalAddressingMode() for a vector type.

* Return true from shouldConsiderGEPOffsetSplit().

Reviewed By: Ulrich Weigand

Differential Revision: https://reviews.llvm.org/D122105

When a preprocessor directive is unknown outside of a skipped
conditional block, we give an error diagnostic because we don't know
how to proceed with preprocessing. But when the directive is in a
skipped conditional block, we would not diagnose it on the theory that
the directive may be known to an implementation other than Clang.

Now, for unknown directives inside a skipped conditional block, we
diagnose the unknown directive as a warning if it is sufficiently
similar to a directive specific to preprocessor conditional blocks. For
example, we'll warn about `#esle` and suggest `#else` but we won't warn
about `#progma` because it's not a directive specific to preprocessor
conditional blocks.

Fixes #51598

Differential Revision: https://reviews.llvm.org/D124726

In InnerLoopVectorizer::getOrCreateVectorTripCount there is an
assert that the known minimum value for the VF is a power of 2
when tail-folding is enabled. However, for scalable vectors the
value of vscale may not be a power of 2, which means we have
to worry about the possibility of overflow. I have solved this
problem by adding preheader checks that prevent us from entering
the vector body if the canonical IV would overflow, i.e.

  if ((IntMax - TripCount) < (VF * UF)) ... skip vector loop ...

Differential Revision: https://reviews.llvm.org/D125235

This adds two conjugated folds:

 * A | B -> B if A implies B (https://alive2.llvm.org/ce/z/R6GU4j)
 * A & B -> A if A implies B (https://alive2.llvm.org/ce/z/EGMqyy)

If A and B are icmps themselves, we will usually fold this through
other logic already (though the tests show a couple additional cases
we previously missed). However, isImpliedCond() also supports A
being of the form X & Y, which allows us to handle cases like
(X & Y) | B where X implies B. This addresses the regression from
D125398.

Something that notably doesn't work yet is the (X | Y) & B case.
This is due to an asymmetry in the isImpliedCondition()
implementation that will have to be addressed separately.

Differential Revision: https://reviews.llvm.org/D125530

This patch overhauls how we pick up the ABI library. Instead of setting
ad-hoc flags, it creates interface targets that can be linked against by
the rest of the build, which is easier to follow and extend to support
new ABI libraries.

This is intended to be a NFC change, however there are some additional
simplifications and improvements we can make in the future that would
require a slight behavior change.

Differential Revision: https://reviews.llvm.org/D120727

Before issuing the warning about use of a strict prototype, check if
the declarator is required to have a prototype through some other means
determined at parse time.

This silences false positives in OpenCL code (where the functions are
forced to have a prototype) and block literal expressions.

We commonly want to create either an inbounds or non-inbounds GEP
based on a boolean value, e.g. when preserving inbounds from
existing GEPs. Directly accept such a boolean in the API, rather
than requiring a ternary between CreateGEP and CreateInBoundsGEP.

This change is not entirely NFC, because we now preserve an
inbounds flag in a constant expression edge-case in InstCombine.

A first patch to use the reasoning in ConstraintElimination to simplify
sub with overflow to a regular sub, if the operation is guaranteed to
not overflow.

Reviewed By: spatel

Differential Revision: https://reviews.llvm.org/D125264

That is more accurate, and using a separate class in TableGen seems
appropriate since these are not parts of the text but properties of the
declaration itself.

Reviewed By: gribozavr2

Differential Revision: https://reviews.llvm.org/D125473

That's required to support `\n`, but can also be used for other commands.
We already had the infrastructure in place to parse a varying number of
arguments, we simply needed to generalize it so that it would work not
only for block commands.

This should fix #55319.

Reviewed By: gribozavr2

Differential Revision: https://reviews.llvm.org/D125429

The command traits have a member NumArgs for which all the parsing
infrastructure is in place, but no command was setting it to a value
other than 0. By doing so we get warnings when passing an empty
paragraph to \retval (the first argument is the return value, then comes
the description). We also take \xrefitem along for the ride, although as
the documentation states it's unlikely to be used directly.

Reviewed By: gribozavr2

Differential Revision: https://reviews.llvm.org/D125422

We need to expand special substitutions in four different ways.  This
refactors to only have one conversion from enum to string, and derive
the other 3 needs off that.

The SpecialSubstitution node is derived from the
ExpandedSpecialSubstitution.  While this may seem unintuitive, it
works out quite well, as SpecialSubstitution can then use the former's
getBaseName and remove an unneeded 'basic_' prefix, for those
substitutions that are instantiations (to known typedef).  Similarly
all those instantiations use the same set of template arguments (with
'basic_string', getting an additional 'allocator' arg).

Expansion tests were added in D123134, and remain unchanged.

Reviewed By: MaskRay, dblaikie

Differential Revision: https://reviews.llvm.org/D125257

In d4555698, the name of nano precision timer function has changed from `nano_time` to `nanoTime`, but benchmarks were not updated to reflect that. This change addresses the discrepancy.

Reviewed By: ftynse

Differential Revision: https://reviews.llvm.org/D125217

As noted on D103874, the 'no-op' vector widening intrinsics currently depend on undef shuffle mask indices, moving to poison we need to adjust this to reference a 'undefined' second vector operand, which will be a freeze(undef).

Fixes the `FIXME:` related to adding `forEachTemplateArgument` to the
core AST Matchers library.

Reviewed By: aaron.ballman

Differential Revision: http://reviews.llvm.org/D125383

This patch fixes a bug in areIdsUnique where it ignores the [start, end] range.

No test case is added since there are no use cases through IR from where it
can be tested, and it is hard to create a unittest since we do not currently
have Values in unittests.

Reviewed By: arjunp

Differential Revision: https://reviews.llvm.org/D124735

ArrayRef<Value *> also accepts a single Value *, there's no need
to create separate overloads for this.

If a load with the same offset has already been seen but the load had
a lower alignment, the pass has to check whether the pointer is
dereferenceable and is sufficiently aligned (so, the new alignment must
be taken into account).

This pass is to handle computationally complex operations like
tensor.pad which are not simply lowered to the exact same operation in
the memref dialect.

Differential Revision: https://reviews.llvm.org/D125384

This adds a late Machine Pass to work around a Cortex CPU Erratum
affecting Cortex-A57 and Cortex-A72:
- Cortex-A57 Erratum 1742098
- Cortex-A72 Erratum 1655431

The pass inserts instructions to make the inputs to the fused AES
instruction pairs no longer trigger the erratum. Here the pass errs on
the side of caution, inserting the instructions wherever we cannot prove
that the inputs came from a safe instruction.

The pass is used:
- for Cortex-A57 and Cortex-A72,
- for "generic" cores (which are used when using `-march=`),
- when the user specifies `-mfix-cortex-a57-aes-1742098` or
  `mfix-cortex-a72-aes-1655431` in the command-line arguments to clang.

Reviewed By: dmgreen, simon_tatham

Differential Revision: https://reviews.llvm.org/D119720

Due to an apparent bug in the Doxygen version <1.8.16 used to generate
documentation for MLIR, parts of the navigation (specifically, the lists
of inherited methods for classes) are unusable due to dynsections.js
missing from the output generated by Doxygen. Setting this flag makes
Doxygen always produce the file.

These are the tests for the pass added in
https://reviews.llvm.org/D119720

Reviewed By: dmgreen

Differential Revision: https://reviews.llvm.org/D122747

Most functions do not need a PatternRewriter or ConversionPatternRewriter.

Differential Revision: https://reviews.llvm.org/D125466

No need to expose this as public API anymore.

Differential Revision: https://reviews.llvm.org/D125361

The goal is support tail and mask policy in RVV builtins.
We focus on IR part first.
If the passthru operand is undef, we use tail agnostic, otherwise
use tail undisturbed.

Reviewed By: craig.topper

Differential Revision: https://reviews.llvm.org/D125323

Recommit.

Check for legal zext 'sinks' before inserting a trunc.

Differential Revision: https://reviews.llvm.org/D115451

Previously it built MIR for the results and returned a Register.

This avoids building constants for earlier elements of the vector if
later elements will fail to fold, and allows CSEMIRBuilder::buildInstr
to avoid unconditionally building a copy from the result.

Use a new helper function MachineIRBuilder::buildBuildVectorConstant
to build a G_BUILD_VECTOR of G_CONSTANTs.

Differential Revision: https://reviews.llvm.org/D117758

Simplify existing test and also add it as a codegen test for aarch64.

Program may set the RA_SIGN_STATE pseudo register by expressions.
Libunwind expected only the DW_CFA_AARCH64_negate_ra_state could change the value
of the register which leads to runtime errors on PAC enabled systems.
In the recent version of the aadwarf64[1] a limitation is added[2] to forbid the mixing the
DW_CFA_AARCH64_negate_ra_state with other DWARF Register Rule Instructions.

[1] https://github.com/ARM-software/abi-aa/releases/tag/2022Q1
[2] https://github.com/ARM-software/abi-aa/pull/129

Reviewed By: #libunwind, MaskRay

Differential Revision: https://reviews.llvm.org/D123692