`ubsan_minimal` makes use of the `_sanitizer::atomic_load` function.
This function uses the `DCHECK` macro which in debug builds will use
the `_sanitizer::CheckFailed` function.

This function is part of `sanitizer_common` but `ubsan_minimal` doesn't
use this so the implementation is missing which leads to link failures
on macOS when trying to link `libclang_rt.ubsan_minimal_osx_dynamic.dylib`.
This is in contrast to the BFD linker on Linux which doesn't seem to care
about the missing symbol.

A basic implementation of `_sanitizer::CheckFailed` has been added to
the `ubsan_minimal` debug build to avoid the link error. The
implementation could definitely be improved but I don't know which
functions can be used in this context so I decided to restrict myself to
functions only being used in `ubsan_minimal` already.

llvm-svn: 326032

[X86][SSE] combineSubToSubus - begun generalizing to work with any type sizes with SplitBinaryOpsAndApply

llvm-svn: 326030

llvm-svn: 326029

Enable multiple COPY hints to eliminate more COPYs during register allocation.

Note that this is something all targets should do, see
https://reviews.llvm.org/D38128.

Review: James Y Knight
llvm-svn: 326028

This commit indents each level by two space characters, e.g.

#if defined(CONDITION)
#  define _LIBCPP_NAME VALUE
#else
#  define _LIBCPP_NAME VALUE
#endif

The simple #ifndef, #define, and #endif sequences are not indented, e.g.

#ifndef _LIBCPP_NAME
#define _LIBCPP_NAME ...
#endif

llvm-svn: 326027

llvm-svn: 326026

Summary:
Interceptors initialization may need to allocate memory. So if we initialize too
early we can crash in non initialized allocator.

Reviewers: pcc, eugenis

Subscribers: llvm-commits

Differential Revision: https://reviews.llvm.org/D43669

llvm-svn: 326025

llvm-svn: 326024

All use declarations need to be directly placed in the top-level module
anyway, knowing the submodule doesn't really help. The header that has
the offending #include can easily be seen in the diagnostics source
location.

Review: https://reviews.llvm.org/D43673
llvm-svn: 326023

[X86] Remove __builtin_ia32_permvarsf256_mask and __builtin_ia32_permvarsi256_mask and use the avx2 unmasked versions and a select instead.

llvm-svn: 326022

Replace if() with a switch(). Because random changes in the code seem to
suppress the crash.

Story so far:
r325966 - Crash introduced.
r325969 - Speculative fix had no effect.
r325978 - Tried to bisect the offending function, crash suddenly disappeared.
r326016 - After another random change in the code, bug appeared again.

llvm-svn: 326021

llvm-svn: 326020

When a lifetime-extended temporary is on a branch of a conditional operator,
materialization of such temporary occurs after the condition is resolved.

This change allows us to understand, by including the MaterializeTemporaryExpr
in the construction context, the target for temporary materialization in such
cases.

Differential Revision: https://reviews.llvm.org/D43483

llvm-svn: 326019

llvm-svn: 326018

llvm-svn: 326017

In order to bind a temporary to a const lvalue reference, a no-op cast is added
to make the temporary itself const, and only then the reference is taken
(materialized). Skip the no-op cast when looking for the construction context.

Differential Revision: https://reviews.llvm.org/D43481

llvm-svn: 326016

When a constructor of a temporary with a single argument is treated
as a functional cast expression, skip the functional cast expression
and provide the correct construction context for the temporary.

Differential Revision: https://reviews.llvm.org/D43480

llvm-svn: 326015

When constructing a temporary that is going to be lifetime-extended through a
MaterializeTemporaryExpr later, CFG elements for the respective constructor
can now be queried to obtain the reference to that MaterializeTemporaryExpr
and therefore gain information about lifetime extension.

This may produce multi-layered construction contexts when information about
both temporary destruction and lifetime extension is available.

Differential Revision: https://reviews.llvm.org/D43477

llvm-svn: 326014

llvm-svn: 326013

llvm-svn: 326012

V_SUBBREV_U32 is a commute opcode for V_SUBB_U32. However, when
we try to commute V_SUBB_U32 in order to shrink it we do not then
process V_SUBBREV_U32 and it stay VOP3. This is fixed.

Differential Revision: https://reviews.llvm.org/D43699

llvm-svn: 326011

Summary:
The setlocale(3) function reloads the ctype(3) arrays from
external files. This happens behind the scenes in the internals
of libc (citrus library, runes functions etc).

ctype(3) functions like isspace(3) can be provided with two
variations on NetBSD: inlined or via a global symbol in libc:

```
#if defined(_NETBSD_SOURCE) && !defined(_CTYPE_NOINLINE) && \
    !defined(__cplusplus)
#include <sys/ctype_inline.h>
#else
#include <sys/ctype_bits.h>
#endif
```

The in-lined versions are de-facto array lookup operations.

```
#define isspace(c)      ((int)((_ctype_tab_ + 1)[(c)] & _CTYPE_S))
```

After setting setlocale(3) the ctype(3) arrays (_ctype_tab_,
_toupper_tab_, _tolower_tab_) are reload behind the scenes
and they are required to be marked as initialized.

Set them initialized inside the common setlocale(3) interceptor.

The arrays are of size of 257 elements: 0..255 + 1 (EOF).

This corrects errors on NetBSD/amd64 in applications
prebuilt with MSan.

Sponsored by <The NetBSD Foundation>

Reviewers: vitalybuka, dvyukov, joerg

Reviewed By: vitalybuka

Subscribers: llvm-commits, kubamracek, #sanitizers

Tags: #sanitizers

Differential Revision: https://reviews.llvm.org/D42020

llvm-svn: 326008

Summary:
There are applications out there which allocate more than 1 << 18 large chunks
of memory (those handled by LargeMmapAllocator, aka secondary allocator).

For 64 bits, secondary allocator stores allocated chunks in a growing on
demand region of memory, growing in blocks of 128K, up to 1 << 20 chunks total.

Sanitizer internal allocator's secondary uses fixed size array storing up
to 1 << 15 chunks (down to 256K from 2Mb of memory used for that array).

Nothing is changed for 32 bits, chunks are still stored in the fixed size
array (up to 1 << 15 chunks).

Reviewers: eugenis

Subscribers: kubamracek, delcypher, #sanitizers, llvm-commits

Differential Revision: https://reviews.llvm.org/D43693

llvm-svn: 326007

- an ambiguous reference to Optional<T> in llvm-dwarfdump.cpp (fixed
  with an explicit prefix).
- a missing base class initialization in Entry copy constructor (fixed
  by using the implicitly default constructor, which is possible after
  some changes which were done during review).

llvm-svn: 326006

This diff fixes the name of the argument of 
setSymTab and makes setSymTab/setStrTab private 
(to make the public interface a bit cleaner).

Test plan: make check-all

Differential revision: https://reviews.llvm.org/D43661

llvm-svn: 326005

Summary:
Add a llc command line option and WebAssembly architecture feature for
exception handling.

Reviewers: dschuff

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

Differential Revision: https://reviews.llvm.org/D43683

llvm-svn: 326004

Summary:
This patch implements the name lookup functionality of the .debug_names
accelerator table and hooks it up to "llvm-dwarfdump -find". To make the
interface of the two kinds of accelerator tables more consistent, I've
created an abstract "DWARFAcceleratorTable::Entry" class, which provides
a consistent interface to access the common functionality of the table
entries (such as getting the die offset, die tag, etc.). I've also
modified the apple table to vend entries conforming to this interface.

Reviewers: JDevlieghere, aprantl, probinson, dblaikie

Subscribers: vleschuk, clayborg, echristo, llvm-commits

Differential Revision: https://reviews.llvm.org/D43067

llvm-svn: 326003

This addresses a compiler warning.

llvm-svn: 326002

The lldb-server unit tests don't test the right thing when the debug
server in use is copied from somewhere else. This can lead to spurious
test failures.

Disable these unit tests when an external debug server is in use.

Fixes llvm.org/PR36494.

llvm-svn: 326001

StartingAccess is a MemoryUseOrDef. No need to check again.

llvm-svn: 326000

This portion can be matched by other patterns. We don't need it to make the larger pattern valid. It's sufficient to have a v1i1 mask input without caring where it came from.

llvm-svn: 325999

llvm-svn: 325998

When indirect field is initialized with another field, you have
MemberExpr with CXXThisExpr that corresponds to the field's immediate
anonymous parent. But 'this' was referring to the non-anonymous parent.
So when we were building LValue Designator, it was incorrect as it had
wrong starting point. Usage of such designator would cause unexpected
APValue changes and crashes.

The fix is in adjusting 'this' for indirect fields from non-anonymous
parent to the field's immediate parent.

Discovered by OSS-Fuzz:
https://bugs.chromium.org/p/oss-fuzz/issues/detail?id=4985

rdar://problem/36359187

Reviewers: rsmith, efriedma

Reviewed By: rsmith

Subscribers: cfe-commits, jkorous-apple

Differential Revision: https://reviews.llvm.org/D42498

llvm-svn: 325997

LLVM has supported a new target feature "alu32" which could be enabled or
disabled by "-mattr=[+|-]alu32" when using llc.

This patch link Clang with it, so it could be also done by passing related
options to Clang, for example:

  -Xclang -target-feature -Xclang +alu32

Signed-off-by: Jiong Wang <jiong.wang@netronome.com>
Reviewed-by: Yonghong Song <yhs@fb.com>
llvm-svn: 325996

llvm-svn: 325995

This ensures that any availability attributes are attached to the
category before the availability for the referenced protocols is checked.

rdar://37829755

llvm-svn: 325994

This patch test disassembler output for load/store instructions when
-mattr=+alu32 specified for which we want to use "w" register format.

Also, this patch extended the existing insn-unit.s and insn-unit-32.s to
make sure disassemblers for all other instructions are not affected.

Signed-off-by: Jiong Wang <jiong.wang@netronome.com>
Reviewed-by: Yonghong Song <yhs@fb.com>
llvm-svn: 325993

This patch adds some unit tests for 32-bit subregister support.
We want to make sure ALU32, subregister load/store and new peephole
optimization are truely enabled once -mattr=+alu32 specified.

Signed-off-by: Jiong Wang <jiong.wang@netronome.com>
Reviewed-by: Yonghong Song <yhs@fb.com>
llvm-svn: 325992

This pass performs peephole optimizations to cleanup ugly code sequences at
MachineInstruction layer.

Currently, the only optimization in this pass is to eliminate type
promotion
sequences for zero extending 32-bit subregisters to 64-bit registers.

If the compiler could prove the zero extended source come from 32-bit
subregistere then it is safe to erase those promotion sequece, because the
upper half of the underlying 64-bit registers were zeroed implicitly
already.

Signed-off-by: Jiong Wang <jiong.wang@netronome.com>
Reviewed-by: Yonghong Song <yhs@fb.com>
llvm-svn: 325991

When -mattr=+alu32 passed to the disassembler, use decoder namespace for
32-bit subregister.

This is to disassemble load and store instructions in preferred B format
as described in previous commit:

      w = *(u8 *) (r + off) // BPF_LDX | BPF_B
      w = *(u16 *)(r + off) // BPF_LDX | BPF_H
      w = *(u32 *)(r + off) // BPF_LDX | BPF_W

      *(u8 *) (r + off) = w // BPF_STX | BPF_B
      *(u16 *)(r + off) = w // BPF_STX | BPF_H
      *(u32 *)(r + off) = w // BPF_STX | BPF_W

NOTE: all other instructions should still use the default decoder
      namespace.

Signed-off-by: Jiong Wang <jiong.wang@netronome.com>
Reviewed-by: Yonghong Song <yhs@fb.com>
llvm-svn: 325990