Use separate lookup tables instead of trying to reuse the fli.s
table.

We were missing the 2 denormal cases for fli.h. We also had an issue
where fli.d was only checking 8 bits of the 11 bit exponent.

We fail to use fli.h for the 2 denormal values.
We use fli.d for some values where the value is larger than a float
can represent due to truncating the exponent to 8 bits without checking
if it fits in 8 bits.

This change generalizes the OS version macro for all darwin OSes. The OS-specific OS version macros are still defined to preserve compatibility.

Warn when an rvalue reference function paramter is never moved
from within the function body.

Reviewed By: carlosgalvezp

Differential Revision: https://reviews.llvm.org/D141569

[RISCV] Add overrides of isLoadFromStackSlot/isStoreFromStackSlot signatures that don't have MemBytes.

D145471 added overrides of the other signature to return MemBytes,
but shouldn't have removed these overrides.

These signatures will now call the MemBytes signature and ignore
the MemBytes. This matches X86.

D102894 introduced common code for the emission of ELF attributes. Our
implementation in RISC-V predates this, and basically copies the Arm
logic at the time. This patch removes that duplication and uses the
shared logic instead.

Differential Revision: https://reviews.llvm.org/D145570

These implementations both match the TargetLoweringBase.isZExtFree implementation

Differential Revision: https://reviews.llvm.org/D145875

Preliminary cleanup before adding some additional legality and value tracking handling.

Post ISel, LDS variables are absolute values. Representing them as
such is simpler than the frame recalculation currently used to build assembler
tables from their addresses.

This is a precursor to lowering dynamic/external LDS accesses from non-kernel
functions.

Reviewed By: arsenm

Differential Revision: https://reviews.llvm.org/D144221

Simplify the use of the basic_string and basic_string_view types by
providing default template arguments.

Depends on D145311

Reviewed By: PiotrZSL

Differential Revision: https://reviews.llvm.org/D145724

Improve the generic <string> header by adding the size() method so that
it can be used to replace the custom implementation in the
readability-container-size-empty check.

This requires fixing an incorrect comparison of a std::wstring with a
char string literal.

Unfortunately, removing the custom basic_string implementation means
fixing the line numbers for many of the checks.

Depends on D145312

Reviewed By: PiotrZSL

Differential Revision: https://reviews.llvm.org/D145313

Improve the generic <string> header by adding another constructor,
std::basic_string::empty and operator!= overload set so that it can be
used to replace the custom implementation in the
readability-string-compare check.

Depends on D145311

Reviewed By: PiotrZSL

Differential Revision: https://reviews.llvm.org/D145312

Try to more aggressively narrow masks of extended values.

This is mainly for cases where the mask is trying to zero out any_extended upper bits, assuming we can zext/trunc the values for free.

This catches a few actual missed folds, as well as helps canonicalize a number of other cases which were being caught in isel etc.

Differential Revision: https://reviews.llvm.org/D145866

[X86] and-shift.ll - add gnux32 test coverage to ensure the X32 ABI correctly narrows the i64 shifts

This patch changes the lowering for the following fp builtins to
emit calls to the new aarch64.sve.###.u intrinsics.
  svabd_x, svabd_n_x
  svadd_x, svadd_n_x
  svdiv_x, svdiv_n_x
  svdivr_x, svdivr_n_x
  svmad_x, svmad_n_x
  svmax_x, svmax_n_x
  svmaxnm_x, svmaxnm_n_x
  svmin_x, svmin_n_x
  svminnm_x, svminnm_n_x
  svmla_x, svmla_n_x
  svmls_x, svmls_n_x
  svmsb_x, svmsb_n_x
  svmul_x, svmul_n_x
  svmulx_x, svmulx_n_x
  svnmad_x, svnmad_n_x
  svnmla_x, svnmla_n_x
  svnmls_x, svnmls_n_x
  svnmsb_X, svnmsb_n_x
  svsub_x, svsub_n_x
  svsubr_x, svsubr_n_x

Depends on D143765.

Differential Revision: https://reviews.llvm.org/D143767

D143767 will change the intrinsics used to lower floating-point
svadd_x, svmul_x and svsub_x builtins. This will result in the
combines added as part of D140200 to no longer fire in all cases.
This patch extends the existing combines for contraction to cover
fadd_u, fmul_u and fsub_u intrinsics.

Differential Revision: https://reviews.llvm.org/D144413

The test currently crashes as:

AddressSanitizer: CHECK failed: asan_poisoning.cpp:38 "((AddrIsAlignedByGranularity(addr))) != (0)"

Main stack address/size don't have to be aligned on asan shadow granularity.
Align stack bottom.

Reviewed By: melver, vitalybuka

Differential Revision: https://reviews.llvm.org/D145799

This came about while investigating ways to handle D145468 in a more generic manner, which involves trying harder to fold and(zext(x),c) -> zext(and(x,c))

Alive2: https://alive2.llvm.org/ce/z/7fXtDt (generic fold)

Differential Revision: https://reviews.llvm.org/D145855

This commit introduces the LLVM's visibility attribute and adds it to
both globals and functions.

Furthermore, this commit ensures that "thread_local" is printed in the
correct place and adds a test for that.

Reviewed By: gysit

Differential Revision: https://reviews.llvm.org/D145790

This reverts commit 6590a372.

This reverts commit 59848b9e, as it
causes some failures in AIX-related Lit tests.

The current implementation of `-p` does not claim the argument once it
is passed. Since it pushes `-pg` directly, it is only ever referred to
again when linking. As a result, when compiling with `-S`, the compiler
warns that `-p` goes unused even though that is not the case.

With this patch, if both `-p` and `-pg` are passed, the argument that is
passed second will take precedence. `-p` will still throw an error on
unsupported platforms, regardless of precedence.

This revision includes a test case, which has been placed in
`clang/test/Driver/zos-profiling-error.c`. As a result,
`zos-profiling-error.c` has been renamed to `ibm-profiling.c`. This
revision also passes `clang/test/Driver/aix-ld.c`.

Differential Revision: https://reviews.llvm.org/D145021

The high number of 10000 threads was choking bot builders running on boards with
very small memory.

libc++ clang-cl tests failed after that commit

Look at https://buildkite.com/llvm-project/libcxx-ci/builds/20490

Reviewed By: #libc

Differential Revision: https://reviews.llvm.org/D145858

This reverts commit b00aaab7.

On 64-bit target, when doing i64 BR_CC where one of the comparison operands is a
constant zero, try to fold the compare and BPcc into a BPr instruction.

For all integers, EQ and NE comparison are available, additionally for signed
integers, GT, GE, LT, and LE is also available.

Reviewed By: arsenm

Differential Revision: https://reviews.llvm.org/D142461

Integrate the BranchRelaxation pass to help with relaxing out-of-range
conditional branches.

This is mostly of concern for SPARCv9, which uses conditional branches with
much smaller range than its v8 counterparts.
(Some large autogenerated code, such as the ones generated by TableGen, already
hits this limitation when building in Release)

Reviewed By: arsenm

Differential Revision: https://reviews.llvm.org/D142458

This reverts commit 43ae4b62.

This was intended to be practically NFC in terms of the overall
opt pipeline, but there is experimental data showing that code
changes occurred here:
https://llvm-compile-time-tracker.com/compare.php?from=772aa05452f8ff90a47168e6801cda2acb5a1873&to=43ae4b62b2671cf73e691c0b53324cd39405cd51&stat=size-text

This is the follow-up to D144199 and suggestion from D144045.
We make use of loop info explicit via InstCombine pass parameter
rather than semi-arbitrary via caching.

The only InstCombine transform that uses LoopInfo currently is a
GEP fold in visitGEPOfGEP(), so that shows up as a failure in the
dedicated test for the fold as well as several LoopVectorizer tests
that run extra passes.

I don't see any pass manager regression tests that actually check
for pass options, but this is intended to be NFC for the pass
pipeline behavior - we only try to use loop info where it would
have been used before via caching .

Differential Revision: https://reviews.llvm.org/D144274

HIP runtime is the language runtime of HIP. When users need
to specify rpath, they usually need to specify rpath for
both compiler-rt and HIP runtime. It seems redundant
to have separate options. Therefore make --offload-add-rpath
an alias to -frtlib-add-rpath.

Reviewed by: Fangrui Song, Artem Belevich

Differential Revision: https://reviews.llvm.org/D145393

Remove duplication in abseil-redundant-strcat-calls check tests,
by using dummy <string> header file string.

Depends on D145310

Reviewed By: PiotrZSL

Differential Revision: https://reviews.llvm.org/D145311

When warning would be emitted in constructor for virtual base class
initialization.

Fixes: https://github.com/llvm/llvm-project/issues/31187

Reviewed By: carlosgalvezp

Differential Revision: https://reviews.llvm.org/D144206

The inliner pass performs canonicalization when created programtically, run with `mlir-opt` with default options, or when explicitly specified. However, when running the pipeline resulting from a `-dump-pass-pipeline` on a default inline pass, the canonicalization is not performed as part of the inlining. This is because the default value for the `default-pipeline` option of the inline pass is an empty string, and this is selected during the dumping. When `InlinerPass::initializeOptions` detects the empty string, it sets the `defaultPipeline` to `nullptr`, which was previously set to canonicalize in the `InlinerPass` constructor, thus the canonicalization is not performed.

The added test checks if the inline pass performs canonicalization by default, and that the dumped `default-pipeline` is set to `canonicalize`.

Fixes: https://github.com/llvm/llvm-project/issues/60960

Reviewed By: rriddle

Differential Revision: https://reviews.llvm.org/D145066

This library is linked into libc-hdrgen, which is compiled with
DISABLE_LLVM_LINK_LLVM_DYLIB.  This option avoids linking with
libLLVM.so when LLVM_LINK_LLVM_DYLIB is specified.  Unfortunately,
this can cause wierd linkage errors where symbols are defined
multiple times, with one definition coming from static linkage
and another definition coming from libLLVM.so.  This is usually
apparent as a link error with options defined multiple times.

This patch adds DISABLE_LLVM_LINK_LLVM_DYLIB to this library, to
get consistent linkage of libc-hdrgen.

Differential Revision: https://reviews.llvm.org/D145839

Fix issue [#61260](https://github.com/llvm/llvm-project/issues/61260)

Reviewed By: awarzynski

Differential Revision: https://reviews.llvm.org/D145845