Summary: The estimation uses AMDGPUTargetLowering::isFNegFree()

Reviewers: rampitec

Reviewed By: rampitec

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

Tags: #llvm

Differential Revision: https://reviews.llvm.org/D82065

LLDB has a lot of code for supporting complex types but we don't have a single
test for it. This adds some basic tests and documents the found bugs.

Summary:
Add patterns to select s_cselect in the isel.

Handle more cases of implicit SCC accesses in si-fix-sgpr-copies
to allow new patterns to work.

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

Tags: #llvm

Differential Revision: https://reviews.llvm.org/D81925

Renamed -force-tail-predication to -force-mve-tail-predication because
that's more descriptive and consistent.

Reviewers: labrinea, dmgreen, pratlucas, LukeGeeson

Reviewed By: dmgreen

Subscribers: kristof.beyls, hiraditya, danielkiss, llvm-commits

Tags: #llvm

Differential Revision: https://reviews.llvm.org/D81486

Summary:
Factor out repetetitive code into helper function and split massive
ExpressionFormat method test into separate test for each method,
removing dead code in passing. Also add a MinInt64 and MaxInt64 checks
when testing getMatchingString.

Reviewers: jhenderson, jdenny, probinson, grimar, arichardson

Reviewed By: jhenderson, grimar

Subscribers: llvm-commits

Tags: #llvm

Differential Revision: https://reviews.llvm.org/D82132

Updated tests disabled by change 5f5f566b.

https://github.com/llvm/llvm-project/commit/5f5f566b265db00f577ead268400d99f34ba9cdd

This patch adds codegen for the following BFloat
operations to the ARM backend:
* concatenation of bf16 vectors
* bf16 vector element extraction
* bf16 vector element insertion
* duplication of a bf16 value into each lane of a vector
* duplication of a bf16 vector lane into each lane

Differential Revision: https://reviews.llvm.org/D81411

The outputs between the direct ast-dump test and the ast-dump test after
deserialization should match modulo a few differences.

For hand-written tests, strip the "<undeserialized declarations>"s and
the "imported"s with sed.

For tests generated with "make-ast-dump-check.sh", regenerate the
output.

Part 1/n.

Some AST dump tests have no "line:" at all. See "ast-dump-wchar.cpp" for
an example.

On AIX, we use __atexit to register dtor functions rather than __cxa_atexit.
So a driver change is needed to default AIX to using -fno-use-cxa-atexit.

Windows platform does not uses __cxa_atexit either. Following its precedent,
we remove the assertion for when -fuse-cxa-atexit is specified by the user,
do not produce a message and silently default to -fno-use-cxa-atexit behavior.

Differential Revision: https://reviews.llvm.org/D82136

1. Provides no piroirity supoort && disables three priority related
   attributes: init_priority, ctor attr, dtor attr;
2. '-qunique' in XL compiler equivalent behavior of emitting sinit
    and sterm functions name using getUniqueModuleId() util function
    in LLVM (currently no support for InternalLinkage and WeakODRLinkage
    symbols);
3. Add testcases to emit IR sample with __sinit80000000, __dtor, and
    __sterm80000000;
4. Temporarily side-steps the need to implement the functionality of
   llvm.global_ctors and llvm.global_dtors arrays. The uses of that
   functionality in this patch (with respect to the name of the functions
   involved) are not representative of how the functionality will be used
   once implemented.

Differential Revision: https://reviews.llvm.org/D74166

Differential Revision: https://reviews.llvm.org/D82045

Summary:
We can't resolve this (if it's a symlink) without further refactoring, but the
current behaviour is just incorrect.

Reviewers: kadircet

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

Tags: #clang

Differential Revision: https://reviews.llvm.org/D82011

Similar to D81937, we might crash when printing a histogram for a GNU hash table
with a 'symndx' index that is larger than the number of dynamic symbols.

This patch adopts and reuses the `getGnuHashTableChains()` helper which performs
a validation of the table. As a side effect the warning reported for
the --gnu-hash-table was improved.

Also with this change we start to report a warning when the histogram is requested for
the GNU hash table, but the dynamic symbols table is empty (size == 0).

Differential revision: https://reviews.llvm.org/D82010

All cases in the test are supported now, it only still failed because an
over-eager regex match not accounting for `, align ` being added to each
load/store now.

Extend the `InheritParentConfig` support introduced in D75184 for the command line option `--config`.
The current behaviour of `--config` is to when set, disable looking for `.clang-tidy` configuration files.
This new behaviour lets you set `InheritParentConfig` to true in the command line to then look for `.clang-tidy` configuration files to be merged with what's been specified on the command line.

Reviewed By: DmitryPolukhin

Differential Revision: https://reviews.llvm.org/D81949

Adds OMPT support for the mutexinoutset dependency

Reviewed by: hbae

Differential Revision: https://reviews.llvm.org/D81890

Without SSE41 we don't have the PCMPEQQ instruction, making cmp-with-zero reductions more complicated than necessary. We can compare as vXi32 (PCMPEQD) and tweak the MOVMSK comparison to test upper/lower DWORD comparisons.

This pre-fixes something that occurs with null tests for vectors of (64-bit) pointers such as in PR35129.

Summary: It was used inside buildCompilerInvocation to speed up stats. But
preambleStatCache doesn't contain stat information performed while
building compiler invocation. So it was an unnecessary optimization.

Furthermore, buildCompilerInvocation in scanPreamble doesn't need to
find gcc installation, include paths and such, as it is only trying to
lex directives. Hence we are passing an empty FS to get rid of any
redundant IO.

Reviewers: sammccall

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

Tags: #clang

Differential Revision: https://reviews.llvm.org/D81719

Summary: Depends on D81998

Reviewers: sammccall

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

Tags: #clang

Differential Revision: https://reviews.llvm.org/D82024

Summary:
Clangd uses FSProvider to get threadsafe views into file systems. This
patch changes naming to make that more explicit.

Depends on D81920

Reviewers: sammccall

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

Tags: #clang

Differential Revision: https://reviews.llvm.org/D81998

Summary:
We've faced a couple of problems when the returned FS didn't have the
proper working directory. New signature makes the API safer against such
problems.

Reviewers: sammccall

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

Tags: #clang

Differential Revision: https://reviews.llvm.org/D81920

If we're permuting ALL the elements of a single vector, then for allof/anyof MOVMSK tests we can avoid the shuffle entirely.

At the moment we use Global ISel by default at -O0, however it is
currently not capable of dealing with scalable vectors for two
reasons:

1. The register banks know nothing about SVE registers.
2. The LLT (Low Level Type) class knows nothing about scalable
   vectors.

For now, the easiest way to avoid users hitting issues when using
the SVE ACLE is to fall back on normal DAG ISel when encountering
instructions that operate on scalable vector types.

I've added a couple of RUN lines to existing SVE tests to ensure
we can compile at -O0. I've also added some new tests to

  CodeGen/AArch64/GlobalISel/arm64-fallback.ll

that demonstrate we correctly fallback to DAG ISel at -O0 when
lowering formal arguments or translating instructions that involve
scalable vector types.

Differential Revision: https://reviews.llvm.org/D81557

The struct store intrinsics in LLVM IR take the individual parts
as arguments, so this patch uses the intrinsics used for `svget`
to break the tuples into individual parts.

Reviewers: c-rhodes, efriedma, ctetreau, david-arm

Reviewed By: efriedma

Tags: #clang

Differential Revision: https://reviews.llvm.org/D81466

Try to avoid calling getVectorNumElements() or relying upon the
TypeSize conversion to uin64_t.

Differential Revision: https://reviews.llvm.org/D81573

This function is going to be added into StackSafety checks.
This patch uses function in ::print implementation to make sure
that it works as expected.

We don't need to lookup InstructionNumbering by number, so
we can use vector with index as assigned number.

Code does not track terminators and do not expose them through interface.
State there is just a state of the last instruction or entry.
So this information is just redundant and doesn't need to be tested.