This prevents duplicates to be pushed into the stack and hypothetically
should reduce memory footprint on ugly cornercases with multiple repeating
duplicates in 'and' tree.

Fixes https://github.com/clangd/clangd/issues/1082

Differential Revision: https://reviews.llvm.org/D136925

It can be useful to external users as well for detecting if there were any changes in IR between passes.

Differential Revision: https://reviews.llvm.org/D137078

Differential Revision: https://reviews.llvm.org/D137174

In ELF, symbols of type STT_GNU_IFUNC need to be resolved by calling the
function at the symbol's address. This is implemented by adding special
stubs for all symbols of that type.

Differential Revision: https://reviews.llvm.org/D105465

Use DL-aware ConstantFoldCompareInstOperands() API instead of
ConstantExpr API. The practical effect of this is that SCCP can
now fold comparisons that require DL.

This is a fairly large method that is unlikely to benefit from
inlining.

Relative to the previous attempt, this also updates the
ValueLattice unit tests.

-----

Resolve the TODO about incorrect getCompare() behavior. This can
be made more precise (e.g. by materializing the undef value and
performing constant folding on it), but for now just return an
unknown result to fix the correctness issue.

This should be NFC in terms of user-visible behavior, because the
only user of this method (SCCP) was already guarding against
UndefValue results.

Having all these instruction-specific overloads does not seem to
provide any compile-time benefit, so drop them in favor of the
generic methods accepting "const Instruction *". Only leave behind
the per-instruction AAQI overloads, which are part of the internal
implementation.

This op dumps the associated payload IR to stderr. It has proven useful for printf-style debugging.

Differential Revision: https://reviews.llvm.org/D137151

The new framework makes it explicit which processor feature is being
used and allows for easier per platform customization:
 - ARM cpu now uses trivial implementations to reduce code size.
 - Memcmp, Bcmp and Memmove have been optimized for x86
 - Bcmp has been optimized for aarch64.

This is a reland of https://reviews.llvm.org/D135134 (b3f1d58a, 02841488)

Reviewed By: courbet

Differential Revision: https://reviews.llvm.org/D136595

This reverts commit 870fbf8e.

Commit 359bc5c5 caused
 Assertion `isa<To>(Val) && "cast<Ty>() argument of incompatible type!"'
failures in decomposeGEP when the GEP pointer operand is a vector.

Fix is to use DataLayout::getIndexTypeSizeInBits when fetching the
index size, as it will use the scalar type in case of a ptr vector.

Differential Revision: https://reviews.llvm.org/D137185

This patch adds the assembly/disassembly for the following instructions:

ld1q : Gather load quadwords
st1q : Scatter store quadwords

The reference can be found here:
https://developer.arm.com/documentation/ddi0602/2022-09

Differential Revision: https://reviews.llvm.org/D137167

This patch extracts the testing logic from `op_tests.cpp` into
`memory_check_utils.h` so we can reuse it for mem* function integration
tests.

This makes testing consistent and thorough.
For instance this catches a bug that got unnoticed during submission of
D136595 and D135134. Integration test for memcmp was only testing a
single size.

This also leverages ASAN to make sure that data is not read / written
outside permitted boundaries

Differential Revision: https://reviews.llvm.org/D136865

This patch adds the assembly/disassembly for the following
predicate pair instructions:

pext:    Set pair of predicates from predicate-as-counter
whilelt: While incrementing signed scalar less than scalar
whilele: While incrementing signed scalar less than or equal to scalar
whilegt: While incrementing signed scalar greater than scalar
whilege: While incrementing signed scalar greater than or equal to scalar
whilelo: While incrementing unsigned scalar lower than scalar
whilels: While incrementing unsigned scalar lower or same as scalar
whilehs: While decrementing unsigned scalar higher or same as scalar
whilehi: While decrementing unsigned scalar higher than scalar

The reference can be found here:
https://developer.arm.com/documentation/ddi0602/2022-09

Differential Revision: https://reviews.llvm.org/D136759

Flang build on windows with MSVC environment and clang-cl compiler
requires clang_rt.builtin.${target} library. This patch allows us to
locate and include this link library. This is mostly needed for flang
runtime and associated unittests. Clang implicitly uses this library
while calling lld-link via clang driver however a standalone call to
lld-link doesnt link against clang_rt library on its own.

Differential Revision: https://reviews.llvm.org/D137112

c-strings.c and volatile-1.c failing due to alignment issue on WoA. I am
going to mark them as XFAIL for now.

This is a continuation of D136939, adding the default attributes
(nosync, nofree, nocallback and willreturn) to more readnone
X86 intrinsics.

This still doesn't cover all of them...

Differential Revision: https://reviews.llvm.org/D137163

Use the utility to lower Constant<T> that was split from current lowering
in https://reviews.llvm.org/D136955.

The difference in HLFIR is the addition of a fir.declare on constant
outlined in memory so that all the information about them is available.

Lowering to HLFIR is enabled in Brideg::genExprValue to allow testing
of scalar constant lowering.

Differential Revision: https://reviews.llvm.org/D137084

PrivateLinkage and HiddenVisibility are implicitly dso_local.

The logic to calculate the offset of lrw16 is complex and the Value
before adjustFixupValue should in range of 0 <= (Value >> 2) <= 0xfe.

It also influences the relax condition.

.bss without symbol and num following can't be compiled in CSKY target, which is consistent to GCC behavior.
So when --save-temps generates .bss, we should UsesELFSectionDirectiveForBSS to avoid such error.

Replace the use of the deprecated `llvm-config` tool with LLVM's CMake
files for detecting LLVM in standalone builds.

Differential Revision: https://reviews.llvm.org/D137024

The sanitizer reported memory leak issues; the command line
used in the test case is:

llvm-debuginfo-analyzer --attribute=level
                        --print=instructions
                        pr-incorrect-instructions-dwarf-clang.o

When dealing with logical instruction lines associated with
an artificial logical scope, skip the process of finding
their enclosing scope. Just add them to the scope.

Create logical debug lines only if the command line specifies:
  --print=lines or --print=elements or --print=all

Reviewed By: jryans, vitalybuka

Differential Revision: https://reviews.llvm.org/D137156

The dependency scanner relies on the module map filtering logic in `ASTWriter`. The algorithm currently considers all system module maps affecting, which is not only sub-optimal, but can also cause failures when building a module explicitly (see attached test case).

This patch applies the same filtering logic to system module maps.

Reviewed By: Bigcheese

Differential Revision: https://reviews.llvm.org/D136007

The special case for bit extraction pattern is  `((x >> C) & mask) << C`.
It can be combined to `x & (mask << C)` by return true in isDesirableToCommuteWithShift.

Fix: #56427

Reviewed By: dmgreen

Differential Revision: https://reviews.llvm.org/D136014

When a value used in the forOp is defined outside the region but within
the parent warpOp we need to return and distribute the value to pass it
to new operations created within the loop.
Also simplify the lambda interface.

Differential Revision: https://reviews.llvm.org/D137146

The handling for MemIntrinsic is not dependent on optimistic
information, no need to put it in update for now. Added a TODO though.

OpenMP offloading always uses uniform workgroups, see
https://reviews.llvm.org/D135374. The runtime doesn't need to handle
non-uniform workgroups at all either.

Differential Revision: https://reviews.llvm.org/D135444

This reverts commit 82c820b9.

This failed llvm-testsuite in our downstream and a similar issue
was reported by @rogfer01.

In D106876, we stopped serializing module map files that didn't affect compilation of the current module.

However, since each `SourceLocation` is simply an offset into `SourceManager`'s global buffer of concatenated input files in, these need to be adjusted during serialization. Otherwise, they can incorrectly point after the buffer or into subsequent input file.

This patch starts adjusting `SourceLocation`s, `FileID`s and other `SourceManager` offsets in `ASTWriter`.

Reviewed By: dexonsmith

Differential Revision: https://reviews.llvm.org/D136624

This patch delays some `ASTWriter` logic until after we've checked whether the source location entry we're serializing as an affecting file or not.

Depends on D137214.

Reviewed By: dexonsmith

Differential Revision: https://reviews.llvm.org/D137216

This makes the tests more minimal, and in particular it avoids relying on a complete `<cstdlib>`, which may not be available on all platforms.

Differential Revision: https://reviews.llvm.org/D137188

This is a prep-patch for D136624 which allows querying `SourceManager` with raw offsets instead of `SourceLocation`s.

Reviewed By: dexonsmith

Differential Revision: https://reviews.llvm.org/D137215