Re-apply 9fdfb045/r365676 with fixes for PPC and Hexagon.  This involved
moving defaults from TargetTransformInfoImplBase to MCSubtargetInfo.

Rework the TTI cache and software prefetching APIs to prepare for the
introduction of a general system model.  Changes include:

- Marking existing interfaces const and/or override as appropriate
- Adding comments
- Adding BasicTTIImpl interfaces that delegate to a subtarget
  implementation
- Moving the default TargetTransformInfoImplBase implementation to a default
  MCSubtarget implementation

Only a handful of targets use these interfaces currently: AArch64, Hexagon, PPC
and SystemZ.  AArch64 already has a custom subtarget implementation, so its
custom TTI implementation is migrated to use the new facilities in BasicTTIImpl
to invoke its custom subtarget implementation.  The custom TTI implementations
continue to exist for the other targets with this change.  They are not moved
over to subtarget-based implementations.

The end goal is to have the default subtarget implementation defer to the system
model defined by the target.  With this change, the default MCSubtargetInfo
implementation essentially returns the defaults TargetTransformInfoImplBase used
to return.  Existing users of TTI defaults will hit the defaults now in
MCSubtargetInfo.  Targets that define their own custom TTI implementations won't
use the BasicTTIImpl implementations that route to the subtarget.

Once system models are in place for the targets that use these interfaces, their
custom TTI implementations can be removed.

Differential Revision: https://reviews.llvm.org/D63614

llvm-svn: 374205

When a category/extension doesn't repeat a type bound, corresponding
type parameter is substituted with `id` when used as a type argument. As
a result, in the added test case it was causing errors like

> type argument 'T' (aka 'id') does not satisfy the bound ('id<NSCopying>') of type parameter 'T'

We are already checking that type parameters should be consistent
everywhere (see `checkTypeParamListConsistency`) and update
`ObjCTypeParamDecl` to have correct underlying type. And when we use the
type parameter as a method return type or a method parameter type, it is
substituted to the bounded type. But when we use the type parameter as a
type argument, we check `ObjCTypeParamType` that ignores the updated
underlying type and remains `id`.

Fix by desugaring `ObjCTypeParamType` to the underlying type, the same
way we are doing with `TypedefType`.

rdar://problem/54329242

Reviewers: erik.pilkington, ahatanak

Reviewed By: erik.pilkington

Subscribers: jkorous, dexonsmith, ributzka, cfe-commits

Differential Revision: https://reviews.llvm.org/D66696

llvm-svn: 374202

The original patch got reverted because it broke `check-lldb` on a clean
build. This fixes that.

llvm-svn: 374201

Summary:
- [Itanium C++ ABI][1], for certain contexts like default parameter and
  etc., mangling numbering will be local to the particular argument in
  which it appears.
- However, for these cases, the mangle numbering context is allocated per
  expression evaluation stack entry. That causes, for example, two
  lambdas defined/used understand the same default parameter are
  numbered as the same value and, in turn, one of them is not generated
  at all.
- In this patch, an extra mangle numbering context map is maintained in
  the AST context to map taht extra declaration context to its numbering
  context. So that, 2 different lambdas defined/used in the same default
  parameter are numbered differently.

[1]: https://itanium-cxx-abi.github.io/cxx-abi/abi.html

Reviewers: rsmith, eli.friedman

Subscribers: cfe-commits

Tags: #clang

Differential Revision: https://reviews.llvm.org/D68715

llvm-svn: 374200

Currently clang does not save some of the intermediate file generated during device compilation for HIP when -save-temps is specified.

This patch fixes that.

Differential Revision: https://reviews.llvm.org/D68665

llvm-svn: 374198

Summary:
We now have valid files that will return NULL from GetStream().
libedit and the LLDB gui are the only places left that need FILE*
streams.  Both are doing curses-like user interaction that only
make sense with a real terminal anyway, so there is no need to convert
them off of their use of FILE*.   But we should check for null streams
before enabling these features.

Reviewers: JDevlieghere, jasonmolenda, labath

Reviewed By: JDevlieghere, labath

Subscribers: lldb-commits

Tags: #lldb

Differential Revision: https://reviews.llvm.org/D68677

llvm-svn: 374197

(specifying an underlying type for the enum might also be suitable - but
this seems better/as good, since there's a clear expectation this can
contain values other than the actual enumerators of this enum)

llvm-svn: 374196

The lifetime of a ValueObject and all its derivative ValueObjects (children, clones, etc.) is managed by a ClusterManager. These objects are only destroyed when every shared pointer to any of the managed objects in the cluster is destroyed. This means that no object in the cluster can store a shared pointer to another object in the cluster without creating a memory leak of the entire cluster. However, some of the synthetic children front-end implementations do exactly this; this patch fixes that.

Differential Revision: https://reviews.llvm.org/D68641

llvm-svn: 374195

Move progress display to separate file.  Simplify some code paths.
Decouple from other components via progress callback.  Remove unused
`_Display` class.

Reviewed By: serge-sans-paille

Differential Revision: https://reviews.llvm.org/D68525

llvm-svn: 374194

llvm-svn: 374192

Really, we were already 99% of the way there; just needed a couple minor
fixes that affected 64-bit-only builtins.  Based on D61717.

Note that the change to builtin_str changes the type of a few
__builtin_neon_* intrinsics that had the "wrong" type.

Fixes https://bugs.llvm.org/show_bug.cgi?id=43341

Differential Revision: https://reviews.llvm.org/D68683

llvm-svn: 374191

llvm-svn: 374190

Summary:
This clang builtin and corresponding LLVM intrinsic are necessary to
expose the exact semantics of the underlying WebAssembly instruction
to users. LLVM produces a poison value if the dynamic swizzle indices
are greater than the vector size, but the WebAssembly instruction sets
the corresponding output lane to zero. Users who depend on this
behavior can safely use this builtin.

Depends on D68527.

Reviewers: aheejin, dschuff

Subscribers: sbc100, jgravelle-google, hiraditya, sunfish, cfe-commits, llvm-commits

Tags: #clang, #llvm

Differential Revision: https://reviews.llvm.org/D68531

llvm-svn: 374189

Summary:
Adds the new v8x16.swizzle SIMD instruction as specified at
https://github.com/WebAssembly/simd/blob/master/proposals/simd/SIMD.md#swizzling-using-variable-indices.
In addition to adding swizzles as a candidate lowering in
LowerBUILD_VECTOR, also rewrites and simplifies the lowering to
minimize the number of replace_lanes necessary rather than trying to
minimize code size. This leads to more uses of v128.const instead of
splats, which is expected to increase performance.

The new code will be easier to tune once V8 implements all the vector
construction operations, and it will also be easier to add new
candidate instructions in the future if necessary.

Reviewers: aheejin, dschuff

Subscribers: sbc100, jgravelle-google, hiraditya, sunfish, llvm-commits

Tags: #llvm

Differential Revision: https://reviews.llvm.org/D68527

llvm-svn: 374188

as it appears to have broken check-lldb.

This reverts r374184 (git commit 22314179)

llvm-svn: 374187

In particular, the function definition is not marked strictfp despite
containing a function marked strictfp. Also, if any function call is marked
strictfp then all function calls in that function must be marked.

This change to move the one strictfp call to a new properly marked function
meets all the new rules.

Tested with a stricter version of D68233.

Reviewed by:	spatel
Approved by:	spatel
Differential Revision:	https://reviews.llvm.org/D68713

llvm-svn: 374186

llvm-svn: 374185

LLDB has three major testing strategies: unit tests, tests that exercise
the SB API though dotest.py and what we currently call lit tests. The
later is rather confusing as we're now using lit as the driver for all
three types of tests. As most of this grew organically, the directory
structure in the LLDB repository doesn't really make this clear.

The 'lit' tests are part of the root and among these tests there's a
Unit and Suite folder for the unit and dotest-tests. This layout makes
it impossible to run just the lit tests.

This patch changes the directory layout to match the 3 testing
strategies, each with their own directory and their own configuration
file. This means there are now 3 directories under lit with 3
corresponding targets:

 - API (check-lldb-api): Test exercising the SB API.
 - Shell (check-lldb-shell): Test exercising command line utilities.
 - Unit (check-lldb-unit): Unit tests.

Finally, there's still the `check-lldb` target that runs all three test
suites.

Finally, this also renames the lit folder to `test` to match the LLVM
repository layout.

Differential revision: https://reviews.llvm.org/D68606

llvm-svn: 374184

We failed to account for the target register width (max vector factor)
when vectorizing starting from GEPs. This causes vectorization to
proceed to obviously illegal widths as in:
https://bugs.llvm.org/show_bug.cgi?id=43578

For x86, this also means that SLP can produce rogue AVX or AVX512
code even when the user specifies a narrower vector width.

The AArch64 test in ext-trunc.ll appears to be better using the
narrower width. I'm not exactly sure what getelementptr.ll is trying
to do, but it's testing with "-slp-threshold=-18", so I'm not worried
about those diffs. The x86 test is an over-reduction from SPEC h264;
this patch appears to restore the perf loss caused by SLP when using
-march=haswell.

Differential Revision: https://reviews.llvm.org/D68667

llvm-svn: 374183

stack

This patch makes sure that if we tag some memory, we untag that memory before
the function returns/throws via any exit, reachable from the tag operation. For
that we place the untag operation either at:

  a) the lifetime end call for the alloca, if that call post-dominates the
     lifetime start call (where the tag operation is placed), or it (the
     lifetime end call) dominates all reachable exits, otherwise
  b) at the reachable exits

Differential Revision: https://reviews.llvm.org/D68469

llvm-svn: 374182

llvm-svn: 374181

Testing whether a name is mangled or not is extremely cheap and can be
done by looking at the first two characters. Mangled knows how to do
it. On the flip side, many call sites that currently pass in an
is_mangled determination do not know how to correctly do it (for
example, they leave out Swift mangling prefixes).

This patch removes this entry point and just forced Mangled to
determine the mangledness of a string itself.

Differential Revision: https://reviews.llvm.org/D68674

llvm-svn: 374180

Summary:
According the the XCOFF document,
If
Then
XTY_SD
x_scnlen contains the csect length.
XTY_LD
x_scnlen contains the symbol table index of the containing csect.
XTY_CM
x_scnlen contains the csect length.
XTY_ER
x_scnlen contains 0.

Change the SectionLen member name to SectionOrLength is more reasonable.

Authored By: DiggerLin

Reviewed By: hubert.reinterpretcast

Differential Revision: https://reviews.llvm.org/D68650

llvm-svn: 374179

The original patch got reverted because it hit a long-standing legacy
issue on Windows that prevents files from being named `com`. Thanks
Kristina & Jeremy for pointing this out.

llvm-svn: 374178

Summary:
The rule for the moveAllAfterMergeBlocks API si for all instructions
from `From` to have been moved to `To`, while keeping the CFG edges (and
block terminators) unchanged.
Update all the callsites for moveAllAfterMergeBlocks to follow this.

Pending follow-up: since the same behavior is needed everytime, merge
all callsites into one. The common denominator may be the call to
`MergeBlockIntoPredecessor`.

Resolves PR43569.

Reviewers: george.burgess.iv

Subscribers: Prazek, sanjoy.google, llvm-commits

Tags: #llvm

Differential Revision: https://reviews.llvm.org/D68659

llvm-svn: 374177

llvm commit access test succeeded.

llvm-svn: 374175

Testing llvm commit access only.

llvm-svn: 374174

Summary:
Following up on D68471, this CL introduces some `getStats` APIs to
gather statistics in char buffers (`ScopedString` really) instead of
printing them out right away. Ultimately `printStats` will just
output the buffer, but that allows us to potentially do some work
on the intermediate buffer, and can be used for a `mallocz` type
of functionality. This allows us to pretty much get rid of all the
`Printf` calls around, but I am keeping the function in for
debugging purposes.

This changes the existing tests to use the new APIs when required.

I will add new tests as suggested in D68471 in another CL.

Reviewers: morehouse, hctim, vitalybuka, eugenis, cferris

Reviewed By: morehouse

Subscribers: delcypher, #sanitizers, llvm-commits

Tags: #llvm, #sanitizers

Differential Revision: https://reviews.llvm.org/D68653

llvm-svn: 374173

Summary:
r373165 fixed an issue where a templated noexcept member function with a
reference qualifier would be indented more than expected:
```
// Formatting produced with LLVM style with AlwaysBreakTemplateDeclarations: Yes

// before r373165:
struct f {
  template <class T>
      void bar() && noexcept {}
};

// after:
struct f {
  template <class T>
  void bar() && noexcept {}
};

```
The way this is done is that in the AnnotatingParser in
`lib/FormatTokenAnnotator.cpp` the determination of the usage of a `&` or `&&`
(the line in determineTokenType

```
Current.Type = determineStarAmpUsage(...
```
is not performed in some cases anymore, combining with a few additional related
checks afterwards. The net effect of these checks results in the `&` or `&&`
token to start being classified as `TT_Unknown` in cases where before `r373165`
it would be classified as `TT_UnaryOperator` or `TT_PointerOrReference` by
`determineStarAmpUsage`.

This inadvertently caused 2 classes of regressions I'm aware of:

- The address-of `&` after a function assignment would be classified as
  `TT_Unknown`, causing spaces to surround it, disregarding style options:
```
// before r373165:
void (*fun_ptr)(void) = &fun;

// after:
void (*fun_ptr)(void) = & fun;
```

- In cases where there is a function declaration list -- looking macro between
  a template line and the start of the function declaration, an `&` as part of
  the return type would be classified as `TT_Unknown`, causing spaces to
  surround it:
```
// before r373165:
template <class T>
DEPRECATED("lala")
Type& foo();

// after:
template <class T>
DEPRECATED("lala")
Type & foo();
```

In these cases the problems are rooted in the skipping of the classification of
a `&` (and similarly `&&`) by determineStarAmpUsage which effects the formatting
decisions later in the pipeline.

I've looked into the goal of r373165 and noticed that replacing `noexcept` with
`const` in the given example produces no extra indentation with the old code:
```
// before r373165:
struct f {
  template <class T>
  int foo() & const {}
};

struct f {
  template <class T>
      int foo() & noexcept {}
};
```

I investigated how clang-format annotated these two examples differently to
determine the places where the processing of both diverges in the pipeline.
There were two places where the processing diverges, causing the extra indent in
the `noexcept` case:
1. The `const` is annotated as a `TT_TrailingAnnotation`, whereas `noexcept`
   is annotated as `TT_Unknown`. I've updated the `determineTokenType` function
   to account for this by adding a missing `tok:kw_noexcept` to the clause that
   marks a token as `TT_TrailingAnnotation`.
2. The `&` in the second example is wrongly identified as `TT_BinaryOperator`
   in `determineStarAmpUsage`. This is the reason for the extra indentation --
   clang-format gets confused and thinks this is an expression.
   I've updated `determineStarAmpUsage` to check for `tok:kw_noexcept`.

With these two updates in place, the additional parsing introduced by r373165
becomes unnecessary and all added tests pass (with updates, as now clang-format
respects the style configuration for spaces around the `&` in the test
examples).
I've removed these additions and added regression tests for the cases above.

Reviewers: AndWass, MyDeveloperDay

Reviewed By: MyDeveloperDay

Subscribers: cfe-commits

Tags: #clang, #clang-format

Differential Revision: https://reviews.llvm.org/D68695

llvm-svn: 374172

llvm-svn: 374171

Summary: Experiments show that this is the alignment we get (for ELF+Linux), but let's ensure that we have it.

Reviewers: gchatelet

Subscribers: tschuett, llvm-commits

Tags: #llvm

Differential Revision: https://reviews.llvm.org/D68703

llvm-svn: 374170

This adds some extra testing to the existing [su][add/sub]_sat X86 and AArch64
tests and adds equivalent tests for ARM.

llvm-svn: 374169

Summary:
After rL374163, replycallbacks might have a cancellable context, which
will try to access RequestCancellers on destruction. See
http://45.33.8.238/mac/1245/step_7.txt for a sample failure.

Reviewers: sammccall

Subscribers: ilya-biryukov, MaskRay, jkorous, arphaman, usaxena95, cfe-commits, thakis

Tags: #clang

Differential Revision: https://reviews.llvm.org/D68702

llvm-svn: 374168

Reviewers: tra, yaxunl

Subscribers: cfe-commits

Tags: #clang

Differential Revision: https://reviews.llvm.org/D68663

llvm-svn: 374167

When optimising for size and SCEV runtime checks need to be emitted to check
overflow behaviour, the loop vectorizer can run in this assert:

  LoopVectorize.cpp:2699: void llvm::InnerLoopVectorizer::emitSCEVChecks(
  llvm::Loop *, llvm::BasicBlock *): Assertion `!BB->getParent()->hasOptSize()
  && "Cannot SCEV check stride or overflow when opt

We should not generate predicates while optimising for size because
code will be generated for predicates such as these SCEV overflow runtime
checks.

This should fix PR43371.

Differential Revision: https://reviews.llvm.org/D68082

llvm-svn: 374166

llvm-svn: 374165

The `expandLoadImmReal` handles four different and almost non-overlapping
cases: loading a "single" float immediate into a GPR, loading a "single"
float immediate into a FPR, and the same couple for a "double" float
immediate.

It's better to move each `else if` branch into separate methods.

llvm-svn: 374164

llvm-svn: 374163

Some bots complain about missing 'class':

LlvmState.h:70:40: error: declaration of ‘std::unique_ptr<const llvm::TargetMachine> llvm::exegesis::LLVMState::TargetMachine’ [-fpermissive]
   std::unique_ptr<const TargetMachine> TargetMachine;

llvm-svn: 374162

llvm-svn: 374161