Emit now takes a std::unique_ptr<MaterializationResponsibility> instead
of a MaterializationResponsibility directly.

This should fix:
http://green.lab.llvm.org/green/view/LLDB/job/lldb-cmake-standalone/

The write_cmake_config() here still looks busted, but at least
the value that's explicitly set is now set correctly.

Define Fortran::Semantics::Scope::GetName in the header so it is available
to Fortran::Evaluate::Tool::AttachDeclaration without a circular dependency
introduced in 82edd428.

Reviewed By: tskeith

Differential Revision: https://reviews.llvm.org/D87505

Addressed some CR issues pointed out in D87111. Formatting and other nits.
The original Diff D87111 - Add an option for unrolling loops up to a factor.

Reviewed By: bondhugula

Differential Revision: https://reviews.llvm.org/D87313

Differential Revision: https://reviews.llvm.org/D87414

This reverts commit 324a5320.

On closer examination of at least one of the test diffs,
this does not appear to be correct in all cases. Even the
existing 'nsw' creation may be wrong based on this example:
https://alive2.llvm.org/ce/z/uL4Hw9
https://alive2.llvm.org/ce/z/fJMKQS

According to LLVM coding standards:
https://llvm.org/docs/CodingStandards.html#name-types-functions-variables-and-enumerators-properly

Differential Revision: https://reviews.llvm.org/D87498

There's no signed wrap if both geps have 'inbounds':
https://alive2.llvm.org/ce/z/nZkQTg
https://alive2.llvm.org/ce/z/7qFauh

These were supposed to be in 0caeaff1 and D83054, but a fat-fingered
error when git-adding missed them. Ooops.

Fix by directly using INT_MAX and INT32_MAX.

Patch by: @nullptr.cpp (Yang Fan)

Differential Revision: https://reviews.llvm.org/D87347

Based on the discussion in D82598#2171312. Thanks @NoQ!

D82598 is titled "Get rid of statement liveness, because such a thing doesn't
exist", and indeed, expressions express a value, non-expression statements
don't.

if (a && get() || []{ return true; }())
    ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ has a value
    ~ has a value
    ~~~~~~~~~~ has a value
                  ~~~~~~~~~~~~~~~~~~~~ has a value
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ doesn't have a value

That is simple enough, so it would only make sense if we only assigned symbolic
values to expressions in the static analyzer. Yet the interface checkers can
access presents, among other strange things, the following two methods:

ProgramState::BindExpr(const Stmt *S, const LocationContext *LCtx, SVal V,
                       bool Invalidate=true)
ProgramState::getSVal(const Stmt *S, const LocationContext *LCtx)

So, what gives? Turns out, we make an exception for ReturnStmt (which we'll
leave for another time) and ObjCForCollectionStmt. For any other loops, in order
to know whether we should analyze another iteration, among other things, we
evaluate it's condition. Which is a problem for ObjCForCollectionStmt, because
it simply doesn't have one (CXXForRangeStmt has an implicit one!). In its
absence, we assigned the actual statement with a concrete 1 or 0 to indicate
whether there are any more iterations left. However, this is wildly incorrect,
its just simply not true that the for statement has a value of 1 or 0, we can't
calculate its liveness because that doesn't make any sense either, so this patch
turns it into a GDM trait.

Fixing this allows us to reinstate the assert removed in
https://reviews.llvm.org/rG032b78a0762bee129f33e4255ada6d374aa70c71.

Differential Revision: https://reviews.llvm.org/D86736

This reverts commit 3eb141e5.

This uses __builtin_mul_overflow which is not available everywhere.

This patch introduces a new ConstraintSystem class, that maintains a set
of linear constraints and uses Fourier–Motzkin elimination to eliminate
constraints to check if there are solutions for the system.

It also adds a convert-constraint-log-to-z3.py script, which can parse
the debug output of the constraint system and convert it to a python
script that feeds the constraints into Z3 and checks if it produces the
same result as the LLVM implementation. This is for verification
purposes.

Reviewed By: spatel

Differential Revision: https://reviews.llvm.org/D84544

Target triples may contain a dash in the platform name (e.g.
"aarch64-arm-none-eabi"). Account for it when splitting the triple
into components.

Reviewed By: ldionne, #libc

Differential Revision: https://reviews.llvm.org/D87508

Add a new index page to be the Flang documentation mainpage instead of
Overview.md, which jumps straight into the compiler Design. The index file
needs to be in .rst format to use the toctree directive to create table of
contents.

Also use the sphinx_markdown_tables extension to generate html tables form
markdown.

A number of additional style changes to the existing docs were needed to make
this work well:
 * Convert all headings to the # style, which works better with toctree's
   titlesonly option. Ensure that there is only one top-level heading per
   document.
 * Add a title to documents that don't have one for rendering on the index.
 * Convert the grammar docs from .txt to .md. for better rendering
 * Fixed broken link to a section in another document - sphinx does not seem to
   support anchor links in markdown files.

Depends on D87226

Reviewed By: sameeranjoshi

Differential Revision: https://reviews.llvm.org/D87242

This test was using a very odd combination of cortex-m7 and Neon. I have
changed it to thumbv7em only.

Fixes Transformer's `Range` parser to handle `char` in a platform-independent way.

Differential Revision: https://reviews.llvm.org/D87409

In short, macro expansions handled the case where a variadic parameter mapped to
multiple arguments, but not the other way around. An internal ticket was
submitted that demonstrated that we fail an assertion. Macro expansion so far
worked by lexing the source code token-by-token and using the Preprocessor to
turn these tokens into identifiers or just get their proper spelling, but what
this counter intuitively doesn't do, is actually expand these macros, so we have
to do the heavy lifting -- in this case, figure out what __VA_ARGS__ expands
into. Since this case can only occur in a nested macro, the information we
gathered from the containing macro does contain this information. If a parameter
resolves to __VA_ARGS__, we need to temporarily stop getting our tokens from the
lexer, and get the tokens from what __VA_ARGS__ maps to.

Differential Revision: https://reviews.llvm.org/D86135

This revision refactors and cleans up a bunch of things to simplify StructuredOpInterface
before work can proceed on Linalg on tensors:
- break out pieces of the StructuredOps trait that are part of the StructuredOpInterface,
- drop referenceIterators and referenceIndexingMaps that end up being more confusing than useful,
- drop NamedStructuredOpTrait

In https://reviews.llvm.org/rG257b29715bb27b7d9f6c3c40c481b6a4af0b37e5,
the definition of OptTable::Info::Flags was changed from `unsigned
short` to `unsigned int`, but the definition/declaration of
OptTable::findByPrefix wasn't updated to reflect that.

This patch updates findByPrefix accordingly.

This was landed but reverted in 5b9c2b1b due to asan picking up a memory
leak. This is fixed in the change to InstrRefBasedImpl.cpp. Original
commit message follows:

[LiveDebugValues][NFC] Add instr-ref tests, adapt old tests

This patch adds a few tests in DebugInfo/MIR/InstrRef/ of interesting
behaviour that the instruction referencing implementation of
LiveDebugValues has. Mostly, these tests exist to ensure that if you
give the "-experimental-debug-variable-locations" command line switch,
the right implementation runs; and to ensure it behaves the same way as
the VarLoc LiveDebugValues implementation.

I've also touched roughly 30 other tests, purely to make the tests less
rigid about what output to accept. DBG_VALUE instructions are usually
printed with a trailing !debug-location indicating its scope:

  !debug-location !1234

However InstrRefBasedLDV produces new DebugLoc instances on the fly,
meaning there sometimes isn't a numbered node when they're printed,
making the output:

  !debug-location !DILocation(line: 0, blah blah)

Which causes a ton of these tests to fail. This patch removes checks for
that final part of each DBG_VALUE instruction. None of them appear to
be actually checking the scope is correct, just that it's present, so
I don't believe there's any loss in coverage here.

Differential Revision: https://reviews.llvm.org/D83054

Uses of Module are all implicit to PassInfoMixin<> so we can guarantee PassManager.h to handle it for us.

We only need this if DONT_GET_PLUGIN_LOADER_OPTION isn't defined.

See https://bugs.llvm.org/show_bug.cgi?id=47488 , rGfb109c42 is
optimizing out part of this test.

The AAPCS64 specifies the internal type is used for c++ mangling. For
bfloat16 it was defined as `BFloat16` when it should be `Bfloat16`, i.e.
lowercase 'f'.

For more information, see:

https://github.com/ARM-software/abi-aa/blob/master/aapcs64/aapcs64.rst#appendix-support-for-advanced-simd-extensions

Reviewed By: stuij

Differential Revision: https://reviews.llvm.org/D87463

Summary:

This is the first patch implementing the new Flang driver as outlined in [1],
[2] & [3]. It creates Flang driver (`flang-new`) and Flang frontend driver
(`flang-new -fc1`). These will be renamed as `flang` and `flang -fc1` once the
current Flang throwaway driver, `flang`, can be replaced with `flang-new`.

Currently only 2 options are supported: `-help` and `--version`.

`flang-new` is implemented in terms of libclangDriver, defaulting the driver
mode to `FlangMode` (added to libclangDriver in [4]). This ensures that the
driver runs in Flang mode regardless of the name of the binary inferred from
argv[0].

The design of the new Flang compiler and frontend drivers is inspired by it
counterparts in Clang [3]. Currently, the new Flang compiler and frontend
drivers re-use Clang libraries: clangBasic, clangDriver and clangFrontend.

To identify Flang options, this patch adds FlangOption/FC1Option enums.
Driver::printHelp is updated so th...

The element types of scalable vectors are defined in terms of stdint
types in the ACLE. This patch fixes the mapping to builtin types for the
ILP32 ABI when creating VLS types with the arm_sve_vector_bits, where
the mapping is as follows:

  int32_t -> LongTy
  int64_t -> LongLongTy
  uint32_t -> UnsignedLongTy
  uint64_t -> UnsignedLongLongTy

This is implemented by leveraging getBuiltinVectorTypeInfo which is
target agnostic since it calls ASTContext::getIntTypeForBitwidth for
integer types. The element type for svfloat16_t is changed from
Float16Ty to HalfTy when creating VLS types since this is what is used
elsewhere.

For more information, see:

https://github.com/ARM-software/abi-aa/blob/master/aapcs64/aapcs64.rst#types-varying-by-data-model
https://github.com/ARM-software/abi-aa/blob/master/aapcs64/aapcs64.rst#appendix-support-for-scalable-vectors

Reviewed By: efriedma

Differential Revision: https://reviews.llvm.org/D87358

These are only ever used as references in the called functions, so just pass the original reference instead of copying it.

Without this, future optimizer improvements can optimize the entire
function to "return 0".