ISD::ROTL/ROTR rotation values are guaranteed to act as a modulo amount, so for power-of-2 bitwidths we only need the lowest bits.

Differential Revision: https://reviews.llvm.org/D76201

Summary:
[Clang] Attribute to allow defining undef global variables

Initializing global variables is very cheap on hosted implementations. The
C semantics of zero initializing globals work very well there. It is not
necessarily cheap on freestanding implementations. Where there is no loader
available, code must be emitted near the start point to write the appropriate
values into memory.

At present, external variables can be declared in C++ and definitions provided
in assembly (or IR) to achive this effect. This patch provides an attribute in
order to remove this reason for writing assembly for performance sensitive
freestanding implementations.

A close analogue in tree is LDS memory for amdgcn, where the kernel is
responsible for initializing the memory after it starts executing on the gpu.
Uninitalized variables in LDS are observably cheaper than zero initialized.

Patch is loosely based on the cuda __shared__ and opencl __local variable
implementation which also produces undef global variables.

Reviewers: kcc, rjmccall, rsmith, glider, vitalybuka, pcc, eugenis, vlad.tsyrklevich, jdoerfert, gregrodgers, jfb, aaron.ballman

Reviewed By: rjmccall, aaron.ballman

Subscribers: Anastasia, aaron.ballman, davidb, Quuxplusone, dexonsmith, cfe-commits

Tags: #clang

Differential Revision: https://reviews.llvm.org/D74361

It fixes an ambiguity issue in case of a user has a custom policy and
calls a version of exclusive_scan with binary operation.

Differential Revision: https://reviews.llvm.org/D62719

template_id annotation when parsing a type.

Functions include their arguments in the use-list. Changed function
values mean that the result of the function changed. We only need
to update the call sites with the new function result and do not
have to propagate the call arguments.

To do so, this patch splits up the visitCallSite into handleCallResult
and handleCallArguments and updates markUsersAsChanged to only update
call results for functions.

Reviewers: efriedma, davide

Reviewed By: efriedma

Differential Revision: https://reviews.llvm.org/D75846

Summary:
There seems to be a race condition between the pipe closing and the child process death. Likely these two events are not atomic on some versions of linux.

With the removal of `WNOHANG` we eliminate the race condition, however if the child closes the pipe intentionally then it could result in the test runner hanging. I find this situation less likely, where as I experience failures locally with this race condition rather consistently.

Reviewers: sivachandra, abrachet

Reviewed By: sivachandra, abrachet

Subscribers: MaskRay, jfb, tschuett, libc-commits

Tags: #libc-project

Differential Revision: https://reviews.llvm.org/D76267

The current implementation of binomial_distribution is not guaranteed to
converge for certain extreme configurations of the engine and distribution.
This is due to a mistake in the implementation of the algorithm from the
given reference paper. The algorithm in the paper is guaranteed to
terminate but has redundant statements. The current implementation
simplified away the redundancy into a while loop, but it excludes the
return condition of the case where a good sample cannot be returned for
the particular sample being used from the uniform distribution, which is
what causes the infinite loop. This change guarantees termination by
recognizing that a good sample cannot be returned and returning 0 after
breaking the loop. This is also in contrast to the paper because the
return value as specified in the paper violates basic checks in at least
a subset of the extreme cases where the current implementation fails to
terminate. This default return value of 0 is satisfactory for the
extreme case known so far.

Since this is only meant to affect extreme cases where the algorithm
does not terminate anyways, the behavior is expected to remain exactly
the same for all non-extreme cases that have been terminating so far.

Fixes https://llvm.org/PR44847

Differential Revision: https://reviews.llvm.org/D74997

[DwarfDebug] Fix an assertion error when emitting call site info that combines two DW_OP_stack_values

When compiling

```
struct S {
  float w;
};
void f(long w, long b);
void g(struct S s) {
  int w = s.w;
  f(w, w*4);
}
```

I get Assertion failed: ((!CombinedExpr || CombinedExpr->isValid()) && "Combined debug expression is invalid").

That's because we combine two epxressions that both end in DW_OP_stack_value:

```
(lldb) p Expr->dump()
!DIExpression(DW_OP_LLVM_convert, 32, DW_ATE_signed, DW_OP_LLVM_convert, 64, DW_ATE_signed, DW_OP_stack_value)
(lldb) p Param.Expr->dump()
!DIExpression(DW_OP_constu, 4, DW_OP_mul, DW_OP_LLVM_convert, 32, DW_ATE_signed, DW_OP_LLVM_convert, 64, DW_ATE_signed, DW_OP_stack_value)
(lldb) p CombinedExpr->isValid()
(bool) $0 = false
(lldb) p CombinedExpr->dump()
!DIExpression(4097, 32, 5, 4097, 64, 5, 16, 4, 30, 4097, 32, 5, 4097, 64, 5, 159, 159)
```

I believe that in this particular case combining two stack values is
safe, but I didn't want to sink the special handling into
DIExpression::append() because I do want everyone to think about what
they are doing.

Patch by Adrian Prantl.

Fixes PR45181.
rdar://problem/60383095

Differential Revision: https://reviews.llvm.org/D76164

RDF is designed to be target agnostic. Therefore it would be useful to have it available for other targets, such as X86.

Based on a previous patch by Krzysztof Parzyszek

Differential Revision: https://reviews.llvm.org/D75932

This rule can help add targets to generate special object files like the
crt1.o on linux. Also, it can be used to add specially compiled object
stubs which are to be linked into the entrypoint objects.

Reviewers: abrachet

Differential Revision: https://reviews.llvm.org/D76271

Discovered in a downstream, it is often useful to have slightly
different semantics for an attribute based on its namespace, however our
spelling infrastructure doesn't consider namespace when deciding to
elide the enum list.  The result is that the solution for a case where
an attribute has slightly different semantics based on a namespace
requires checking against the integer value, which is fragile.

This patch makes us always emit the spelling enum if there is more than
1 and we're generating the header.

Differential Revision: https://reviews.llvm.org/D76289

Follow-on suggested in:
D75961

Summary:
This revision adds a new hook, `notifyMatchFailure`, that allows for notifying the rewriter that a match failure is coming with the provided reason. This hook takes as a parameter a callback that fills a `Diagnostic` instance with the reason why the match failed. This allows for the rewriter to decide how this information can be displayed to the end-user, and may completely ignore it if desired(opt mode). For now, DialectConversion is updated to include this information in the debug output.

Differential Revision: https://reviews.llvm.org/D76203

Summary:
DataLayout::getTypeAllocSize() return TypeSize. For cases where the
scalable property doesn't matter, we should explicitly call getKnownMinSize()
to avoid implicit type conversion to uint64_t, which is not valid for scalable
vector type.

Reviewers: sdesmalen, efriedma, apazos, reames

Reviewed By: efriedma

Subscribers: tschuett, hiraditya, rkruppe, psnobl, llvm-commits

Tags: #llvm

Differential Revision: https://reviews.llvm.org/D76260

According to the standard, The event-handle will be considered as if it was specified on a firstprivate clause.

I believe we were previously calculating a pointer info with the scalar base and an offset of 0. But that's not really where the gather is pointing. The offset is a function of the indices of the GEP we looked through.

Also set the size of the MachineMemOperand to UnknownSize

Differential Revision: https://reviews.llvm.org/D76157

Fix to get the AST we generate for function templates closer to what clang generates and expects.
We fix which FuntionDecl we are passing to CreateFunctionTemplateSpecializationInfo and we strip
template parameters from the name when creating the FunctionDecl and FunctionTemplateDecl.

These two fixes together fix asserts and ambiguous lookup issues for several cases which are added to the already existing small function template test.
This fixes issues with overloads, overloads and ADL, variadic function templates and templated operator overloads.

Differential Revision: https://reviews.llvm.org/D75761

Summary:
The former are like:

libcxx/include/typeinfo:322:11: warning: definition of implicit copy constructor for 'bad_cast' is deprecated because it has a user-declared destructor [-Wdeprecated-copy-dtor]
  virtual ~bad_cast() _NOEXCEPT;
          ^
libcxx/include/typeinfo:344:11: note: in implicit copy constructor for 'std::bad_cast' first required here
    throw bad_cast();
          ^

Fix these by adding an explicitly defaulted copy constructor.

The latter are like:

libcxx/include/codecvt:105:37: warning: dynamic exception specifications are deprecated [-Wdeprecated-dynamic-exception-spec]
    virtual int do_encoding() const throw();
                                    ^~~~~~~

Fix these by using the _NOEXCEPT macro instead.

Reviewers: EricWF, mclow.lists, ldionne, #libc

Reviewed By: EricWF, #libc

Subscribers: dexonsmith, libcxx-commits

Tags: #libc

Differential Revision: https://reviews.llvm.org/D76150

Summary:
Was accidentally removed by commit af64948e when it overrode
TargetInfo::convertConstraint.

Fixes: pr/45225

Reviewers: eli.friedman, sdesmalen

Reviewed By: sdesmalen

Subscribers: echristo, sdesmalen, kristof.beyls, cfe-commits, kmclaughlin, srhines

Tags: #clang

Differential Revision: https://reviews.llvm.org/D76297

This reverts commit 1f93b162.

This patch adds llvm.fence. I tried not to change the syntax much.

syntax:

LLVM IR
`fence [syncscope("<target-scope>")] <ordering>`

MLIR LLVM Dialect

`llvm.fence [syncscope("<target-scope>")] <ordering>`

example:
LLVM IR: `fence syncscope("agent") seq_cst`
MLIR: `llvm.fence syncscope("agent") seq_cst`

Differential Revision: https://reviews.llvm.org/D75645

We cannot move wwm over exec copies because the exec register needs an exact exec mask.

Differential Revision: https://reviews.llvm.org/D76232

Summary: The following change is to allow the machine outlining can be applied for Nth times, where N is specified by the compiler option. By default the value of N is 1. The motivation is that the repeated machine outlining can further reduce code size.  Please refer to the presentation "Improving Swift Binary Size via Link Time Optimization" in LLVM Developers' Meeting in 2019.

Reviewers: aschwaighofer, tellenbach, paquette

Reviewed By: paquette

Subscribers: tellenbach, hiraditya, llvm-commits, jinlin

Tags: #llvm

Differential Revision: https://reviews.llvm.org/D71027

Differential Revision: https://reviews.llvm.org/D76262

Summary:
Conservatively escaping everything is bad in coc.nvim which shows the markdown
to the user, and we have reports of it causing problems for other parsers.

Fixes https://github.com/clangd/clangd/issues/301

Reviewers: kadircet

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

Tags: #clang

Differential Revision: https://reviews.llvm.org/D75687

Under certain circumstances we'll end up in the position where the negated shift amount will get truncated to the type specified getScalarShiftAmountTy(), so we need to test for a truncated version of the shift amount as well.

This allows us to remove half of the remaining patterns tested for by X86ISelLowering's combineOrShiftToFunnelShift.

As discussed on PR45118, getInlinedAtScope() shouldn't ever return null. So we can simplify the logic to an assertion and remove all other null tests.

Reviewers: probinson, jyknight

Reviewed By: probinson

Subscribers: llvm-commits

Tags: #llvm

Differential Revision: https://reviews.llvm.org/D76074

Differential Revision: https://reviews.llvm.org/D76039

Summary: courbet

Subscribers: llvm-commits

Tags: #llvm

Differential Revision: https://reviews.llvm.org/D76274

Summary: Prevent InstCombine from removing llvm.assume for which the arguement is true when they have operand bundles with usefull information.

Reviewers: jdoerfert, nikic, lebedev.ri

Reviewed By: jdoerfert

Subscribers: hiraditya, llvm-commits

Tags: #llvm

Differential Revision: https://reviews.llvm.org/D76147

Summary: This change enable the divergence driven selection for the SEXT DAG opcode.

Reviewers: vpykhtin, rampitec

Reviewed By: vpykhtin

Subscribers: arsenm, kzhuravl, jvesely, wdng, nhaehnle, yaxunl, dstuttard, tpr, t-tye, hiraditya, kerbowa, llvm-commits

Differential Revision: https://reviews.llvm.org/D76230