Replace some custom matrix diagnostic kinds with the more generic
err_builtin_invalid_arg_type introduced in D111985.

Reviewed By: aaron.ballman, erichkeane

Differential Revision: https://reviews.llvm.org/D112532

RewritePatterns.td/RewritePatterns.inc is used only by the
FIROps.cpp file. This patch move this file logically in the Dialect
folder together with FIRDialet, FIROps, FIRTypes ...
It also rename it to CanonicalizationPatterns.td.

This patch is part of the upstreaming effort from fir-dev branch.

Reviewed By: schweitz

Differential Revision: https://reviews.llvm.org/D112522

This is a test-only failure. The test wrongly assumes that this gets us
a tagged pointer:
```
NSObject* num1 = @7;
assert(isTaggedPtr(num1));
```

However, on newer deployment targets that have “const data support” we
get a “normal” pointer to constant object.

Radar-Id: rdar://problem/83217293

Polynomial approximation can be extented to support N-d vectors.
N-dimensional vectors are useful when vectorizing operations on N-dimensional
tiles. Before lowering to LLVM these vectors are usually unrolled or flattened
to 1-dimensional vectors.

Differential Revision: https://reviews.llvm.org/D112566

With https://reviews.llvm.org/rG14c9207063bb00823a5126131e50c93f6e288bd3, the build is broken with -DMLIR_INCLUDE_TESTS=OFF. This patch fixes the build and we may want to do a better fix to the layering in a followup.

Differential Revision: https://reviews.llvm.org/D112560

There is an issue where the AST code does not compute the correct SourceRange
for a ObjCObjectPointer.

From Richard Smith (ie @zygoloid) in discord:

I think the problem is that we set an invalid location for the * (because there
isn't one): https://github.com/llvm/llvm-project/blob/main/clang/lib/Sema/SemaType.cpp#L1121
And then we use the default getLocalSourceRangeImpl for a PointerLikeTypeLoc
that just assumes the * location is the type's end location:
https://github.com/llvm/llvm-project/blob/main/clang/include/clang/AST/TypeLoc.h#L1293
Possibly we should be special-casing that here:
https://github.com/llvm/llvm-project/blob/main/clang/lib/AST/TypeLoc.cpp#L228

My change:

introduces a AST dump test to show the issue in the first commit
special cases ObjCObjectPointerType in the second commit to correctly compute
the end location

address an test error on window os , exclude the test llvm/test/tools/llvm-readobj/XCOFF/xcoff-auxiliary-header.test from
windows OS.
http://45.33.8.238/win/47662/step_11.txt
for
https://reviews.llvm.org/D82549

This reverts commit 5ec83226.

It broke a number of bots, e.g.:
https://lab.llvm.org/buildbot/#/builders/52/builds/11811

Reviewed By: melver

Differential Revision: https://reviews.llvm.org/D112563

Mark LWG2731 as complete. The type alias `mutex_type` is only provided if
`scoped_lock` is given one mutex type and it has been implemented that
way since the beginning of Clang 5 it seems. There already are tests for
verifying existence (and lack thereof) for `mutex_type` type alias
depending on the number of mutex types, so there is nothing to
do for this LWG issue.

Reviewed By: Quuxplusone, Mordante, #libc

Differential Revision: https://reviews.llvm.org/D112462

Added a notification in the placeholder section. While writing things
like preciate of an attribute, we may embed certain placeholder in the C
expression. Note that the type of the placeholder is only guaranteed to
be the base type like mlir::Type, it's better not to use the derived
type which is based on the implementation.

Reviewed By: mehdi_amini

Differential Revision: https://reviews.llvm.org/D112396

Change-Id: Idcce321c825ecc6b3a111a683e24dc10015f6872

This patch refactors the shared_ptr methods from being defined out-of-line
to being defined inline in the class, like what we do for all new code in
the library. The benefits of doing that are that code is not as scattered
around and is hence easier to understand, and it avoids a ton of duplication
due to SFINAE checks. Defining the method where it is declared also removes
the possibility for mismatched attributes.

As a fly-by change, this also:

- Adds a few _LIBCPP_HIDE_FROM_ABI attributes
- Uses __enable_if_t instead of enable_if as a function argument, to match
  the style that we use everywhere else.

Differential Revision: https://reviews.llvm.org/D112478

These are now redundant after https://reviews.llvm.org/D106872

Change-Id: I82edfedf1d45cac4e3368d77ce3a48c78e342c19

Use RdxDesc->getOpcode instead of getUnderlingInstr()->getOpcode.
Move the code which finds Kind and IsOrdered to be outside the for loop
since neither of these change with the vector part.

Differential Revision: https://reviews.llvm.org/D112547

It seems like protobuf crashed the `std::string` checker.
Somehow it acquired `UnknownVal` as the sole `std::string` constructor
parameter, causing a crash in the `castAs<Loc>()`.

This patch addresses this.

Reviewed By: martong

Differential Revision: https://reviews.llvm.org/D112551

1.Combining kind min/max of Vector reduction op has been changed to
  minf/maxf, minsi/maxsi, and minui/maxui. Modify getVectorReductionOp
  accordingly.
2.Add min/max to supported reductions.

Reviewed By: dcaballe, nicolasvasilache

Differential Revision: https://reviews.llvm.org/D112246

This patch implements __builtin_elementwise_max and
__builtin_elementwise_min, as specified in D111529.

Reviewed By: aaron.ballman

Differential Revision: https://reviews.llvm.org/D111985

We were previously always emitting the GOT into `__DATA_CONST`, even for
target platforms where it should end up in `__DATA`.

I stumbled onto this while trying to use the `class-dump` tool -- with
the wrong segment names, it fails to locate the ObjC runtime info and
therefore fails to dump any classes.

Reviewed By: #lld-macho, oontvoo

Differential Revision: https://reviews.llvm.org/D112500

Before this change, we would add "using std::vector<int>" instead of
just "using std::vector;", which would not even compile.

Fixes https://github.com/clangd/clangd/issues/904

Differential Revision: https://reviews.llvm.org/D112530

We make assumption that:
getDeclForComment(getDeclForComment(X)) == getDeclForComment(X)
but this is not true if you have a template
instantionation of a template instantiation, which is the case when, for
example, you have a <=> operator in a templated class.

This fix makes getDeclForComment() call itself recursively to ensure
this property is always true.

Fixes https://github.com/clangd/clangd/issues/901

Differential Revision: https://reviews.llvm.org/D112527

This is inconsistent with the code and comments in
RecursiveASTVisitor.h and is not a correct behavior:

https://github.com/llvm/llvm-project/blob/main/clang/include/clang/AST/RecursiveASTVisitor.h#L1879-L1880

> We shouldn't traverse D->getTypeForDecl(); it's a result of
> declaring the type, not something that was written in the source.

Fixes the failures in 1c2e249f

Verify that SecondII is not replaced.

It was discovered that an extra register COPY remained when expanding a
(variable length) memory operation with a loop and there was another use of
the involved address register(s) afterwards.

A simple fix for this is to COPY the address registers before the loop and
use that new vreg instead.

Review: Ulrich Weigand

Differential Revision: https://reviews.llvm.org/D112065

The final reduction nodes should not be reordered, the order does not
matter for reductions. Also, it might be profitable to vectorize smaller
reduction trees, reduction cost may compensate small tree cost.

Part of D111574

Differential Revision: https://reviews.llvm.org/D112467

Summary:

The patch supports parsing the xcoff object file auxiliary header with llvm-readobj with option "auxiliary-headers"

the format of auxiliary header as
https://www.ibm.com/support/knowledgecenter/en/ssw_aix_72/filesreference/XCOFF.html#XCOFF__fyovh386shar

Reviewers: James Henderson, Jason Liu, Hubert Tong, Esme yi, Sean Fertile.

Differential Revision: https://reviews.llvm.org/D82549

This patch specifies a set of vector builtins for Clang, as discussed on
cfe-dev:
https://lists.llvm.org/pipermail/cfe-dev/2021-September/068999.html
https://lists.llvm.org/pipermail/cfe-dev/2021-October/069070.html

Reviewed By: scanon

Differential Revision: https://reviews.llvm.org/D111529

This was forgotten when setting up the from-scratch configs.

Differential Revision: https://reviews.llvm.org/D112460

All tsan interceptors check for initialization and/or initialize things
as necessary lazily, so we can pretend everything is initialized in the
COMMON_INTERCEPTOR_NOTHING_IS_INITIALIZED check to avoid double-checking
for initialization (this is only necessary for sanitizers that don't
handle initialization on common grounds).

Reviewed By: melver

Differential Revision: https://reviews.llvm.org/D112446

Print PC of the previous lock, not the current one.
The current one will be printed during unwind.

Reviewed By: melver

Differential Revision: https://reviews.llvm.org/D112533

I was under the impression that `return false;` in the
RecursiveASTVisitor stops the traversal for the subtree but it appears
that it stops the whole tree traversal, so this change introduces a bug
where `ReferencedLocationCrawler` will not collect any symbols past an
enum.

This is a follow-up on D112209.

CSR_AMDGPU_VGPRs_24_255 and CSR_AMDGPU_VGPRs_32_255 are not used
anywhere, so remove them.

Differential Revision: https://reviews.llvm.org/D112535

add support for ROCR_VISIBLE_DEVICES similar to name and purpose
as CUDA_VISIBLE_DEVICES

Differential Revision: https://reviews.llvm.org/D112503