The unseen logic diff occurs because MayFoldLoad() is defined like this:

static bool MayFoldLoad(SDValue Op) {
  return Op.hasOneUse() && ISD::isNormalLoad(Op.getNode());
}

The test diffs here all seem ok to me on screen/paper, but it's hard to know
if that will lead to universally better perf for all targets. For example,
if a target implements broadcast from mem as multiple uops, we would have to
weigh the potential reduction of instructions and register pressure vs.
possible increase in number of uops. I don't know if we can make a truly
informed decision on this at compile-time.

The motivating case that I'm looking at in PR42024:
https://bugs.llvm.org/show_bug.cgi?id=42024
...resembles the diff in extract-concat.ll, but we're not going to change the
larger example there without at least 1 other fix.

Differential Revision: https://reviews.llvm.org/D74088

Rather than mixing creation of new instructions and in-place
modification here, create a new log2 intrinsic. This should be
NFC apart from worklist order changes.

These show incorrect duplication of instructions.

Converting a pointer to an integer whose result cannot represented in the
integer type is undefined behavior is C and prohibited in C++. C++ already
has a diagnostic when casting. This adds a diagnostic for C.

Since this diagnostic uses the range of the conversion it also modifies
int-to-pointer-cast diagnostic to use a range.

Fixes PR8718: No warning on casting between pointer and non-pointer-sized int

Differential Revision: https://reviews.llvm.org/D72231

Try to fix -Wnon-virtual-dtor warnings that cause build failure
on clang-pcc64le-rhel.

Related llvm-dev thread:
http://lists.llvm.org/pipermail/llvm-dev/2020-February/138951.html

This patch moves the IRBuilder from templating over the constant
folder and inserter towards making both of these virtual.
There are a couple of motivations for this:

1. It's not possible to share code between use-sites that use
different IRBuilder folders/inserters (short of templating the code
and moving it into headers).
2. Methods currently defined on IRBuilderBase (which is not templated)
do not use the custom inserter, resulting in subtle bugs (e.g.
incorrect InstCombine worklist management). It would be possible to
move those into the templated IRBuilder, but...
3. The vast majority of the IRBuilder implementation has to live
in the header, because it depends on the template arguments.
4. We have many unnecessary dependencies on IRBuilder.h,
because it is not easy to forward-declare. (Significant parts of
the backend depend on it via TargetLowering.h, for example.)

This patch addresses the issue by making the following changes:

* IRBuilderDefaultInserter::InsertHelper becomes virtual.
  IRBuilderBase accepts a reference to it.
* IRBuilderFolder is introduced as a virtual base class. It is
 implemented by ConstantFolder (default), NoFolder and TargetFolder.
  IRBuilderBase has a reference to this as well.
* All the logic is moved from IRBuilder to IRBuilderBase. This means
  that methods can in the future replace their IRBuilder<> & uses
  (or other specific IRBuilder types) with IRBuilderBase & and thus
  be usable with different IRBuilders.
* The IRBuilder class is now a thin wrapper around IRBuilderBase.
  Essentially it only stores the folder and inserter and takes care
  of constructing the base builder.

What this patch doesn't do, but should be simple followups after this change:

* Fixing use of the inserter for creation methods originally defined
  on IRBuilderBase.
* Replacing IRBuilder<> uses in arguments with IRBuilderBase, where useful.
* Moving code from the IRBuilder header to the source file.

From the user perspective, these changes should be mostly transparent:
The only thing that consumers using a custom inserted may need to do is
inherit from IRBuilderDefaultInserter publicly and mark their InsertHelper
as public.

Differential Revision: https://reviews.llvm.org/D73835

The current code has following issues:
1) It has a duplicated logic part.
2) This logic relies on unwrapOrError calls, but if we want to convert
   them to warnings, we will need to change all of them what is hard to do
   because of the duplication.

In this patch I've created a new method that returns Expected<> what allows
now to catch all errors in a single place and remove the code duplication.

Note: this change is itself a refactor NFC. It does not change the current logic
anyhow. It prepares the code for the follow-up(s).

Differential revision: https://reviews.llvm.org/D74545

Summary: This matches the similar feature on operation definitions.

Reviewers: jpienaar, antiagainst

Reviewed By: jpienaar, antiagainst

Subscribers: mehdi_amini, burmako, shauheen, antiagainst, nicolasvasilache, arpith-jacob, mgester, lucyrfox, liufengdb, Joonsoo, llvm-commits

Tags: #llvm

Differential Revision: https://reviews.llvm.org/D74438

Summary: getEncodedSourceLocation can be very costly to compute, especially if the input line becomes very long. This revision inlines some of the verification of a few `getChecked` methods to avoid the materialization of an encoded source location.

Differential Revision: https://reviews.llvm.org/D74587

Some buildbots signaled a problem in this method when the
llvm::function_ref was copied and reused after 1228d42d. To
eliminate the problem we avoid copying the llvm::function_ref and
instead we pass it as a const reference.

After b18cb9c4, clang would sometimes prefer the host C++ includes
(e.g. in /usr/include/c++/v1) before those specified via --sysroot.
While this behavior may be desirable on Linux, it is not so on FreeBSD,
where we make extensive use of --sysroot during the build of the base
system.  In that case, clang must *not* search outside the sysroot,
except for its own internal headers.

Add an override addLibCxxIncludePaths() to restore the old behavior,
which is to simply append /usr/include/c++/v1 to the specified sysroot.
While here, apply clang-format to the FreeBSD specific toolchain files.

Fixes PR44923.

[NFC] Move ValidTextRanges out of DwarfTransformer and into GsymCreator and unify address is not in GSYM errors so all strings match.

libc++ is careful to not fracture overload sets. When one overload
is visible to a user, all of them should be. Anything less causes
subtle bugs and ODR violations.

Previously, in order to support ::abs and ::div being supplied by
both <cmath> and <cstdlib> we had to do awful things that make
<math.h> and <stdlib.h> have header cycles and be non-modular.
This really breaks with modules.

Specifically the problem was that in C++ ::abs introduces overloads
for floating point numbers, these overloads forward to ::fabs,
which are defined in math.h. Therefore ::abs needed to be in math.h
too. But this required stdlib.h to include math.h and math.h to
include stdlib.h.

To avoid these problems the definitions have been moved to stddef.h
(which math includes), and the floating point overloads of ::abs
have been changed to call __builtin_fabs, which both Clang and GCC
support.

graph.

Split off shuffle combine tests that use AVX512F intrinsics, so we can test it with/without AVX512BW support.

Currently we always return true, when markConstantRange is used on an
object already containing a constant range. If NewR is equal to the
existing constant range however, nothing changes and we should return
false.

I also went ahead and added a clarifying comment and improved the
assertion.

Reviewers: efriedma, davide, nikic

Reviewed By: efriedma

Differential Revision: https://reviews.llvm.org/D73240

Refactors matchShuffleAsBitRotate to allow use by both lowerShuffleAsBitRotate and matchUnaryPermuteShuffle.

This patch prepares ValueLatticeElement to be used by SCCP, by:
* making the mark* functions public
* make the mark* functions return a bool indicating if the value has changed.

Reviewers: efriedma, davide, mssimpso, nikic

Reviewed By: nikic

Differential Revision: https://reviews.llvm.org/D60581

This allows it to work properly with masked inc/dec for avx512. Those
would have a vselect as the root node so didn't get a chance to call
combineIncDecVector.

This also simplifies the logic because we don't have to manage
the topological ordering.

This includes a fix for cases where things get marked as overdefined in
ResolvedUndefsIn, but we later discover a constant. To avoid crashing,
we consistently bail out on overdefined values in the visitors. This is
similar to the previous behavior with forcedconstant.

This reverts the revert commit 02b72f56.

So far, all non-COFF-related Emit* functions have been de-capitalized.

Reported by @lbenes (Luke Benes)

I was reported that with commit:
https://github.com/llvm/llvm-project/commit/d3963051c490

gcc-9.2 is giving the warning below.
This should help (I have no gcc 9.2 to test).

[ 57%] Building CXX object tools/obj2yaml/CMakeFiles/obj2yaml.dir/elf2yaml.cpp.o
/llvm/tools/obj2yaml/elf2yaml.cpp: In instantiation of ‘llvm::Expected<llvm::ELFYAML::Object*>
{anonymous}::ELFDumper<ELFT>::dump() [with ELFT = llvm::object::ELFType<llvm::support::little, false>]’:
/llvm/tools/obj2yaml/elf2yaml.cpp:1218:31:   required from ‘llvm::Error elf2yaml(llvm::raw_ostream&,
const llvm::object::ELFFile<ELFT>&) [with ELFT = llvm::object::ELFType<llvm::support::little, false>]’
/llvm/tools/obj2yaml/elf2yaml.cpp:1231:47:   required from here
/llvm/tools/obj2yaml/elf2yaml.cpp:207:41: warning: comparison of integer expressions of different
signedness: ‘llvm::support::detail::packed_endian_specific_integral<unsigned int, llvm::support::little, 1>::value_type’ {aka ‘unsigned int’} and ‘int’ [-Wsign-compare]
  207 |     if (!SymTab || SymTabShndx->sh_link != SymTab - Sections.begin())
/llvm/tools/obj2yaml/elf2yaml.cpp: In instantiation of ‘llvm::Expected<llvm::ELFYAML::Object*>
{anonymous}::ELFDumper<ELFT>::dump() [with ELFT = llvm::object::ELFType<llvm::support::big, false>]’:
...

As noted on D74621, the bswap intrinsic has a self imposed limitation that the type's bitwidth must be divisible by 16, but there's no reason that APInt::byteSwap must have the same limitation, given that it can already handle any byte width.

This commit fixes the broken tests after
commit b6a9fe20
on the expensive check builder:
http://lab.llvm.org:8011/builders/llvm-clang-x86_64-expensive-checks-ubuntu/builds/2884

We can just byteSwap the entire uint64_t VAL and then shift down into place like we do for the multi-word case.

rGf0181cc7 added specific tests up to i64, this adds a general loop to test some basic byte moves for larger APInts.

This hopefully fixes up the expensive checks bot.

Summary:
This patch is extracted from D74308.

It patches all usages of WithColor::error() and WithColor::warning
in DebugInfoDWARF library.

Depends on D74481

Reviewers: jhenderson, dblaikie, probinson, aprantl, JDevlieghere

Reviewed By: JDevlieghere

Subscribers: hiraditya, llvm-commits

Tags: #llvm

Differential Revision: https://reviews.llvm.org/D74635