platform/upstream/llvm.git
2 years agoAdding an Nvidia representative to security group
Nikhil Gupta [Wed, 18 Aug 2021 10:05:08 +0000 (12:05 +0200)]
Adding an Nvidia representative to security group

Adding Nikhil Gupta as an Nvidia representative to the llvm security
group.

Differential Revision: https://reviews.llvm.org/D106906

2 years agoPre-commit frem test in PowerPC
Qiu Chaofan [Wed, 18 Aug 2021 09:52:53 +0000 (17:52 +0800)]
Pre-commit frem test in PowerPC

2 years agoFix a typo around 'builtin'
Qiu Chaofan [Wed, 18 Aug 2021 09:42:12 +0000 (17:42 +0800)]
Fix a typo around 'builtin'

2 years ago[OpenCL] C++ for OpenCL version 2021 introduced to command line.
Justas Janickas [Thu, 12 Aug 2021 15:20:11 +0000 (16:20 +0100)]
[OpenCL] C++ for OpenCL version 2021 introduced to command line.

Introduces language standard `lang_openclcpp2021` and allows
`clc++2021` as a version flag for `-cl-std` in command line.
Defines macros related to C++ for OpenCL version 2021.

C++ for OpenCL version 2021 has been proposed in an RFC:
https://lists.llvm.org/pipermail/cfe-dev/2021-August/068593.html

Differential Revision: https://reviews.llvm.org/D108038

2 years agoMark tsan cxa_guard_acquire test as unsupported on Darwin
Azharuddin Mohammed [Wed, 18 Aug 2021 05:45:05 +0000 (22:45 -0700)]
Mark tsan cxa_guard_acquire test as unsupported on Darwin

This test started failing after recent commit `636428c727cd`. Marking it
as unsupported to get the bot back to green until we get it sorted out.

2 years agoAArch64: compare correct type for multi-valued SDNode.
Tim Northover [Wed, 18 Aug 2021 08:26:03 +0000 (09:26 +0100)]
AArch64: compare correct type for multi-valued SDNode.

If Orig produces more than one value (rare) with different types (rarer) then
we need to make sure we check against the one that Orig actually represents,
not just the first type.

Unfortunately because of the combination of things that need to happen I wasn't
able to produce a test.

2 years ago[JILink][ELF] Include binding and visibility values in error messages.
Lang Hames [Wed, 18 Aug 2021 08:03:47 +0000 (18:03 +1000)]
[JILink][ELF] Include binding and visibility values in error messages.

This should make it easier to track down JITLink errors for unrecognized
binding or visibility types, e.g.
https://lab.llvm.org/buildbot#builders/112/builds/8599.

2 years ago[ORC] Remove unused headers.
Lang Hames [Wed, 18 Aug 2021 06:28:21 +0000 (16:28 +1000)]
[ORC] Remove unused headers.

2 years ago[gn build] Port d1ec581ebfca
LLVM GN Syncbot [Wed, 18 Aug 2021 08:10:40 +0000 (08:10 +0000)]
[gn build] Port d1ec581ebfca

2 years ago[clangd] IncludeCleaner as a library: Find all references to symbols in the file
Kirill Bobyrev [Wed, 18 Aug 2021 08:08:08 +0000 (10:08 +0200)]
[clangd] IncludeCleaner as a library: Find all references to symbols in the file

This is the first patch in an ongoing attempt of Include Cleaner: unused/missing
headere diagnostics, an IWYU-like functionality implementation for clangd. The
work is split into (mostly) distinct and parallelizable pieces:

- Finding all referenced locations (this patch).
- Finding all referenced locations of macros.
- Building IncludeGraph and marking headers as unused, used and directly used.
- Making use of the introduced library and add an option to use in clangd.

---

* Adding support for standard library headers (possibly through mapping
  genfiles).

Based on https://reviews.llvm.org/D100540.

Reviewed By: sammccall

Differential Revision: https://reviews.llvm.org/D105426

2 years ago[flang] Add default implementation for SYSTEM_CLOCK
Diana Picus [Mon, 12 Jul 2021 13:45:44 +0000 (13:45 +0000)]
[flang] Add default implementation for SYSTEM_CLOCK

Add an implementation for the runtime functions related to SYSTEM_CLOCK.
As with CPU_TIME, this is based on std::clock(), which should be
available everywhere, but it is highly recommended to add
platform-specific implementations for systems where std::clock() behaves
poorly (e.g. POSIX).

The documentation for std::clock() doesn't specify a maximum value and
in fact wrap around behaviour is non-conforming.  Therefore, this
implementation of SYSTEM_CLOCK is not guaranteed to wrap around either,
and after std::clock reaches its maximum value we will likely just
return failure rather than wrap around. If this happens often on your
system, please add a new platform-specific implementation.

We define COUNT_MAX as either the maximum value that can be stored in
a std::clock_t or in a 64-bit integer (whichever is smaller), and
COUNT_RATE as CLOCKS_PER_SEC. For POSIX systems, the value of
CLOCKS_PER_SEC is hardcoded to 10^6 and irrelevant for the values
returned by std::clock.

Differential Revision: https://reviews.llvm.org/D105969

2 years ago[AArch64][GlobalISel] Add support for selection of s8:fpr = G_UNMERGE <8 x s8>
Amara Emerson [Wed, 18 Aug 2021 07:18:26 +0000 (00:18 -0700)]
[AArch64][GlobalISel] Add support for selection of s8:fpr = G_UNMERGE <8 x s8>

2 years ago[CMake] Fix recompile all .inc files with LLVM_OPTIMIZED_TABLEGEN in Visual Studio.
Daniil Fukalov [Wed, 11 Aug 2021 12:51:43 +0000 (15:51 +0300)]
[CMake] Fix recompile all .inc files with LLVM_OPTIMIZED_TABLEGEN in Visual Studio.

After some moment VS solution generated with LLVM_OPTIMIZED_TABLEGEN started to
generate all .inc files for each build. The reason was it had
"<path to native tablegen>/llvm-tblgen" without .exe as a dependency.

Differential Revision: https://reviews.llvm.org/D107898

2 years agoRevert "Allow rematerialization of virtual reg uses"
Petr Hosek [Wed, 18 Aug 2021 07:12:41 +0000 (00:12 -0700)]
Revert "Allow rematerialization of virtual reg uses"

This reverts commit 877572cc193a470f310eec46a7ce793a6cc97c2f which
introduced PR51516.

2 years ago[InstrProfiling][NFC] Initialize MadeChange variable
Petr Hosek [Wed, 18 Aug 2021 06:31:50 +0000 (23:31 -0700)]
[InstrProfiling][NFC] Initialize MadeChange variable

This addresses an issue introduced in 389dc94d4be7 which triggers
a crash on Windows.

2 years ago[Test][AggressiveInstCombine] Add one more test for shift truncation
Anton Afanasyev [Wed, 18 Aug 2021 06:09:05 +0000 (09:09 +0300)]
[Test][AggressiveInstCombine] Add one more test for shift truncation

Add test for which `OrigBitWidth != SrcBitWidth`
(https://reviews.llvm.org/D108091#2950131)

2 years agoUse installed llvm-lit.py instead of lit.py PR-51072
Yaron Keren [Tue, 17 Aug 2021 17:15:39 +0000 (20:15 +0300)]
Use installed llvm-lit.py instead of lit.py PR-51072

Three tests fail when building and testing LLVM from the Visual C++ environment
since they use the repo version of lit.py that do not have local customization
builtin_parameters = { 'build_mode' : 'Release' }
https://bugs.llvm.org/show_bug.cgi?id=51072

Reviewed By: dyung

Differential Revision: https://reviews.llvm.org/D108085

2 years ago[AMDGPU] Set wait state for meta instructions to zero
Christudasan Devadasan [Tue, 17 Aug 2021 04:26:36 +0000 (00:26 -0400)]
[AMDGPU] Set wait state for meta instructions to zero

It looked more reasonable to set the wait state to
zero for all non-instructions. With that we can avoid
the special handling for them in `getWaitStatesSince`
and `AdvanceCycle`. This NFC patch makes the handling
more generic.

2 years ago[AggressiveInstCombine] Fix unsigned overflow
Anton Afanasyev [Wed, 18 Aug 2021 05:11:54 +0000 (08:11 +0300)]
[AggressiveInstCombine] Fix unsigned overflow

Fix issue reported here: https://reviews.llvm.org/D108091#2950930

2 years ago[gn build] Port e256445bfff1
LLVM GN Syncbot [Wed, 18 Aug 2021 05:02:05 +0000 (05:02 +0000)]
[gn build] Port e256445bfff1

2 years ago[ORC-RT][ORC] Introduce ELF/*nix Platform and runtime support.
Lang Hames [Wed, 18 Aug 2021 00:47:21 +0000 (10:47 +1000)]
[ORC-RT][ORC] Introduce ELF/*nix Platform and runtime support.

This change adds support to ORCv2 and the Orc runtime library for static
initializers, C++ static destructors, and exception handler registration for
ELF-based platforms, at present Linux and FreeBSD on x86_64. It is based on the
MachO platform and runtime support introduced in bb5f97e3ad1.

Patch by Peter Housel. Thanks very much Peter!

Reviewed By: lhames

Differential Revision: https://reviews.llvm.org/D108081

2 years ago[WebAssembly] Autogenerate checks for simd-conversions.ll
Thomas Lively [Wed, 18 Aug 2021 04:35:22 +0000 (21:35 -0700)]
[WebAssembly] Autogenerate checks for simd-conversions.ll

In preparation for adding more tests more simply.

Differential Revision: https://reviews.llvm.org/D108264

2 years ago[NFC][llvm-xray] add a llvm-xray convert option `no-demangle`
Xu Mingjie [Wed, 18 Aug 2021 04:17:09 +0000 (12:17 +0800)]
[NFC][llvm-xray] add a llvm-xray convert option `no-demangle`

When option `--symbolize` is true, llvm-xray convert will demangle function
name on default. This patch adds a llvm-xray convert option `no-demangle` to
determine whether to demangle function name when symbolizing function ids from
the input log.

Reviewed By: MaskRay, smeenai

Differential Revision: https://reviews.llvm.org/D108019

2 years ago[NFC] More get/removeAttribute() cleanup
Arthur Eubanks [Wed, 18 Aug 2021 03:25:16 +0000 (20:25 -0700)]
[NFC] More get/removeAttribute() cleanup

2 years ago[NFC] Cleanup more AttributeList::addAttribute()
Arthur Eubanks [Tue, 17 Aug 2021 01:24:22 +0000 (18:24 -0700)]
[NFC] Cleanup more AttributeList::addAttribute()

2 years ago[NFC] Cleanup attribute methods in Function
Arthur Eubanks [Tue, 17 Aug 2021 00:35:02 +0000 (17:35 -0700)]
[NFC] Cleanup attribute methods in Function

2 years ago[NFC] Migrate some callers away from Function/AttributeLists methods that take an...
Arthur Eubanks [Tue, 17 Aug 2021 00:12:49 +0000 (17:12 -0700)]
[NFC] Migrate some callers away from Function/AttributeLists methods that take an index

These methods can be confusing.

2 years ago[NFC] Replace Function handling of attributes with less confusing calls
Arthur Eubanks [Mon, 16 Aug 2021 23:29:20 +0000 (16:29 -0700)]
[NFC] Replace Function handling of attributes with less confusing calls

To avoid magic constants and confusing indexes.

2 years ago[mlir][ods] Fix packing in OperandOrAttribute
Jacques Pienaar [Wed, 18 Aug 2021 03:55:48 +0000 (20:55 -0700)]
[mlir][ods] Fix packing in OperandOrAttribute

Wrong combiner was used which led to information loss.

2 years ago[CFE][X86] Enable complex _Float16 support
Wang, Pengfei [Wed, 18 Aug 2021 01:06:54 +0000 (09:06 +0800)]
[CFE][X86] Enable complex _Float16 support

Support complex _Float16 on X86 in C/C++ following the latest X86 psABI. (https://gitlab.com/x86-psABIs)

Reviewed By: LuoYuanke

Differential Revision: https://reviews.llvm.org/D105331

2 years ago[clang] enable sanitizers for hexagon
Brian Cain [Mon, 16 Aug 2021 23:59:42 +0000 (16:59 -0700)]
[clang] enable sanitizers for hexagon

2 years ago[compiler-rt] Add hexagon support to profile
Brian Cain [Tue, 17 Aug 2021 19:41:59 +0000 (12:41 -0700)]
[compiler-rt] Add hexagon support to profile

Adds build support for hexagon linux to profile

2 years ago[compiler-rt] Add hexagon support to safestack
Brian Cain [Tue, 17 Aug 2021 19:41:27 +0000 (12:41 -0700)]
[compiler-rt] Add hexagon support to safestack

Adds build support for hexagon linux to safestack.

2 years ago[compiler-rt] Add hexagon support to CFI
Brian Cain [Tue, 17 Aug 2021 19:40:51 +0000 (12:40 -0700)]
[compiler-rt] Add hexagon support to CFI

Adds build support for hexagon linux to cfi.

2 years ago[sanitizer] Add hexagon support to scudo
Brian Cain [Mon, 16 Aug 2021 23:38:09 +0000 (16:38 -0700)]
[sanitizer] Add hexagon support to scudo

Adds build support for hexagon linux to scudo, scudo-standalone

2 years ago[sanitizer] Add hexagon support to ubsan
Brian Cain [Mon, 16 Aug 2021 23:44:57 +0000 (16:44 -0700)]
[sanitizer] Add hexagon support to ubsan

Adds build support for hexagon linux to ubsan.

2 years ago[sanitizer] Add hexagon support to asan
Brian Cain [Mon, 16 Aug 2021 23:33:21 +0000 (16:33 -0700)]
[sanitizer] Add hexagon support to asan

Adds build support for hexagon linux to asan.

2 years ago[sanitizer] Add hexagon support to lsan
Brian Cain [Mon, 16 Aug 2021 23:28:32 +0000 (16:28 -0700)]
[sanitizer] Add hexagon support to lsan

Adds build support for hexagon linux to lsan.

2 years ago[sanitizer] Add hexagon support to sanitizer-common
Brian Cain [Tue, 17 Aug 2021 21:38:38 +0000 (14:38 -0700)]
[sanitizer] Add hexagon support to sanitizer-common

Adds build support for hexagon linux to sanitizer common.

2 years ago[lldb] Include arm64 in affected_by_radar_34562999
Jonas Devlieghere [Wed, 18 Aug 2021 01:30:48 +0000 (18:30 -0700)]
[lldb] Include arm64 in affected_by_radar_34562999

The same issue impacts arm64, both on-device and on Apple Silicon.

2 years ago[lldb] Extend isAArch64 to arm64 and arm64e
Jonas Devlieghere [Wed, 18 Aug 2021 01:05:46 +0000 (18:05 -0700)]
[lldb] Extend isAArch64 to arm64 and arm64e

This fixes TestMemoryTag on Apple Silicon.

2 years ago[CVP] processSwitch: Remove default case when switch cover all possible values.
Jun Ma [Thu, 15 Jul 2021 10:37:40 +0000 (18:37 +0800)]
[CVP] processSwitch: Remove default case when switch cover all possible values.

Differential Revision: https://reviews.llvm.org/D106056

2 years ago[NFC][CVP] Add one switch testcase
Jun Ma [Tue, 10 Aug 2021 06:52:00 +0000 (14:52 +0800)]
[NFC][CVP] Add one switch testcase

2 years ago[RegAlloc] Remove addAllocPriorityToGlobalRanges hook
Qiu Chaofan [Wed, 18 Aug 2021 02:12:19 +0000 (10:12 +0800)]
[RegAlloc] Remove addAllocPriorityToGlobalRanges hook

It was introduced in 1a6dc92 and only enabled on PowerPC/AMDGPU. That
should be enabled for all targets.

Reviewed By: arsenm

Differential Revision: https://reviews.llvm.org/D108010

2 years ago[DAGCombiner] Teach isKnownToBeAPowerOfTwo handle SPLAT_VECTOR
jacquesguan [Wed, 18 Aug 2021 01:48:33 +0000 (09:48 +0800)]
[DAGCombiner] Teach isKnownToBeAPowerOfTwo handle SPLAT_VECTOR

Make DAGCombine turn mul by power of 2 into shl for scalable vector.

Reviewed By: frasercrmck

Differential Revision: https://reviews.llvm.org/D107883

2 years ago[flang][OpenMP] Add semantic check for target nesting
PeixinQiao [Wed, 18 Aug 2021 01:40:52 +0000 (09:40 +0800)]
[flang][OpenMP] Add semantic check for target nesting

This patch implements the following check for TARGET construct:
```
OpenMP Version 5.0 Target construct restriction: If a target update,
target data, target enter data, or target exit data construct is
encountered during execution of a target region, the behavior is
unspecified.
```

Also add one test case for the check.

Reviewed By: kiranchandramohan, clementval

Differential Revision: https://reviews.llvm.org/D106165

2 years ago[X86] AVX512FP16 instructions enabling 3/6
Wang, Pengfei [Wed, 18 Aug 2021 01:03:01 +0000 (09:03 +0800)]
[X86] AVX512FP16 instructions enabling 3/6

Enable FP16 conversion instructions.

Ref.: https://software.intel.com/content/www/us/en/develop/download/intel-avx512-fp16-architecture-specification.html

Reviewed By: LuoYuanke

Differential Revision: https://reviews.llvm.org/D105265

2 years ago[runtimeunroll] Support multiple exits to latch exit w/epilogue loop
Philip Reames [Wed, 18 Aug 2021 00:50:58 +0000 (17:50 -0700)]
[runtimeunroll] Support multiple exits to latch exit w/epilogue loop

This patch extends the runtime unrolling infrastructure to support unrolling a loop with multiple exiting blocks branching to the same exit block used by the latch. It intentionally does not include a cost model change to enable this functionality unless appropriate force flags are used.

I decided to restrict this to the epilogue case. Given the changes ended up being pretty generic, we may be able to unblock the prolog case too, but I want to do that in a separate change to reduce the amount of code we all have to understand at one time.

Differential Revision: https://reviews.llvm.org/D107381

2 years ago[LLDB][GUI] Add Breakpoints window
Omar Emara [Tue, 17 Aug 2021 23:58:05 +0000 (16:58 -0700)]
[LLDB][GUI] Add Breakpoints window

This patch adds a breakpoints window that lists all breakpoints and
breakpoints locations. The window is implemented as a tree, where the
first level is the breakpoints and the second level is breakpoints
locations.

The tree delegate was hardcoded to only draw when there is a process,
which is not necessary for breakpoints, so the relevant logic was
abstracted in the TreeDelegateShouldDraw method.

Reviewed By: clayborg

Differential Revision: https://reviews.llvm.org/D107386

2 years ago[test] Remove a redundant test line
Philip Reames [Wed, 18 Aug 2021 00:33:12 +0000 (17:33 -0700)]
[test] Remove a redundant test line

This was made redundant when I removed -instcombine from output in 70ffd65c, but I didn't notice.  nikic pointed that out in review of D107381

2 years ago[llvm-profgen] Move profiled binary loading out of PerfReader
wlei [Tue, 17 Aug 2021 22:53:31 +0000 (15:53 -0700)]
[llvm-profgen] Move profiled binary loading out of PerfReader

Change to use unique pointer of profiled binary to unblock asan.

At same time, I realized we can decouple to move the profiled binary loading out of PerfReader, so I made some other related refactors.

Reviewed By: hoy

Differential Revision: https://reviews.llvm.org/D108254

2 years ago[LLDB][GUI] Refactor form drawing using subsurfaces
Omar Emara [Tue, 17 Aug 2021 23:06:46 +0000 (16:06 -0700)]
[LLDB][GUI] Refactor form drawing using subsurfaces

This patch adds a new method SubSurface to the Surface class. The method
returns another surface that is a subset of this surface. This is
important to further abstract away drawing from the ncurses objects. For
instance, fields could previously be drawn on subpads only but can now
be drawn on any surface. This is needed to create the file search
dialogs and similar functionalities.

There is an opportunity to refactor window drawing in general using
surfaces, but we shall consider this separately later.

Differential Revision: https://reviews.llvm.org/D107761

2 years ago[LLDB] Fix off by one logging placeholders in ClangASTSource::layoutRecordType()
Shafik Yaghmour [Tue, 17 Aug 2021 23:29:00 +0000 (16:29 -0700)]
[LLDB] Fix off by one logging placeholders in ClangASTSource::layoutRecordType()

D72391 Added some additional information to the logging but in this case instead of using
placeholder 2 and 3 they used 3 and 4.

2 years ago[Modules] Do not remove failed modules after the control block phase
Ben Barham [Tue, 17 Aug 2021 22:53:40 +0000 (15:53 -0700)]
[Modules] Do not remove failed modules after the control block phase

Reading modules first reads each control block in the chain and then all
AST blocks.

The first phase is intended to find recoverable errors, eg. an out of
date or missing module. If any error occurs during this phase, it is
safe to remove all modules in the chain as no references to them will
exist.

While reading the AST blocks, however, various fields in ASTReader are
updated with references to the module. Removing modules at this point
can cause dangling pointers which can be accessed later. These would be
otherwise harmless, eg. a binary search over `GlobalSLocEntryMap` may
access a failed module that could error, but shouldn't crash. Do not
remove modules in this phase, regardless of failures.

Since this is the case, it also doesn't make sense to return OutOfDate
during this phase, so remove the two cases where this happens.

When they were originally added these checks would return a failure when
the serialized and current path didn't match up. That was updated to an
OutOfDate as it was found to be hit when using VFS and overriding the
umbrella. Later on the path was changed to instead be the name as
written in the module file, resolved using the serialized base
directory. At this point the check is really only comparing the name of
the umbrella and only works for frameworks since those don't include
`Headers/` in the name (which means the resolved path will never exist)

Given all that, it seems safe to ignore this case entirely for now.
This makes the handling of an umbrella header/directory the same as
regular headers, which also don't check for differences in the path
caused by VFS.

Resolves rdar://79329355

Differential Revision: https://reviews.llvm.org/D107690

2 years ago[hwasan] [NFC] pull out helper function.
Florian Mayer [Tue, 3 Aug 2021 10:07:52 +0000 (11:07 +0100)]
[hwasan] [NFC] pull out helper function.

Reviewed By: hctim

Differential Revision: https://reviews.llvm.org/D107334

2 years ago[hwasan] Support malloc in atfork.
Evgenii Stepanov [Sat, 14 Aug 2021 00:09:10 +0000 (17:09 -0700)]
[hwasan] Support malloc in atfork.

Before this change we were locking the StackDepot in the fork()
interceptor. This results in a deadlock when allocator functions are
used in a pthread_atfork() callback.

Instead, set up a pthread_atfork() callback at init that locks/unlocks
both StackDepot and the allocator. Since our callback is set up very
early, the pre-fork callback is executed late, and both post-fork ones
are executed early, which works perfect for us.

Differential Revision: https://reviews.llvm.org/D108063

2 years agoLoopNest Analysis expansion to return instructions that prevent a Loop
Mark Danial [Tue, 17 Aug 2021 22:23:43 +0000 (22:23 +0000)]
LoopNest Analysis expansion to return instructions that prevent a Loop
Nest from being perfect

Expand LoopNestAnalysis to return the full list of instructions that
cause a loop nest to be imperfect. This is useful for other passes to
know if they should continue for in the inner loops.
Added New function getInterveningInstructions
that returns a small vector with the instructions that prevent a loop
for being perfect. Also added a couple of helper functions to reduce
code duplication.

Reviewed By: Whitney

Differential Revision: https://reviews.llvm.org/D107773

2 years ago[flang] Fix the vector version of EOSHIFT with a BOUNDARY argument
Peter Steinfeld [Tue, 17 Aug 2021 21:57:42 +0000 (14:57 -0700)]
[flang] Fix the vector version of EOSHIFT with a BOUNDARY argument

When the vector version of EOSHIFT was called, the BOUNDARY argument was being
ignored.  I fixed that and added a test that would not pass without this fix.

Differential Revision: https://reviews.llvm.org/D108249

2 years ago[libc++abi] Fix possible infinite loop in itanium demangler
Mikhail Borisov [Tue, 17 Aug 2021 22:10:57 +0000 (18:10 -0400)]
[libc++abi] Fix possible infinite loop in itanium demangler

A libfuzzer run has discovered some inputs for which the demangler does
not terminate. When minimized, it looks like this: _Zcv1BIRT_EIS1_E

Deciphered:

_Z
cv    - conversion operator

      * result type
 1B   - "B"
 I    - template args begin
  R   - reference type              <.
   T_ - forward template reference   |  *
 E    - template args end            |  |
                                     |  |
      * parameter type               |  |
 I    - template args begin          |  |
  S1_ - substitution #1              * <'
 E    - template args end

The reason is: template-parameter refs in conversion operator result type
create forward-references, while substitutions are instantly resolved via
back-references. Together these can create a reference loop. It causes an
infinite loop in ReferenceType::collapse().

I see three possible ways to avoid these loops:

1. check if resolving a forward reference creates a loop and reject the
   invalid input (hard to traverse AST at this point)
2. check if a substitution contains a malicious forward reference and
   reject the invalid input (hard to traverse AST at this point;
   substitutions are quite common: may affect performance; hard to
   clearly detect loops at this point)
3. detect loops in ReferenceType::collapse() (cannot reject the input)

This patch implements (3) as seemingly the least-impact change. As a
side effect, such invalid input strings are not rejected and produce
garbage, however there are already similar guards in
`if (Printing) return;` checks.

Fixes https://llvm.org/PR51407

Differential Revision: https://reviews.llvm.org/D107712

2 years ago[libc++abi][NFC] Move PODSmallVector definition to the top of ItaniumDemangle.h
Mikhail Borisov [Tue, 17 Aug 2021 22:06:53 +0000 (18:06 -0400)]
[libc++abi][NFC] Move PODSmallVector definition to the top of ItaniumDemangle.h

This change is needed to clean the non-relevant parts of diff
from https://reviews.llvm.org/D107712

Differential Revision: https://reviews.llvm.org/D107771

2 years ago(NFC) clang-format hwasan/hwasan_linux.cpp
Evgenii Stepanov [Tue, 17 Aug 2021 17:57:25 +0000 (10:57 -0700)]
(NFC) clang-format hwasan/hwasan_linux.cpp

Differential Revision: https://reviews.llvm.org/D108224

2 years agoRevert "Simplify setting up LLVM as bazel external repo"
Geoffrey Martin-Noble [Tue, 17 Aug 2021 21:45:43 +0000 (14:45 -0700)]
Revert "Simplify setting up LLVM as bazel external repo"

This breaks the rules for creating repository dependencies when invoked
from another workspace. See comment on https://reviews.llvm.org/D107714

This reverts commit 934f084ad42231d225bd0eee3342f4b5f7e1ce6e.

Differential Revision: https://reviews.llvm.org/D108245

2 years ago[GlobalOpt] Check stored once value's type before setting global initializer
Arthur Eubanks [Tue, 17 Aug 2021 19:00:45 +0000 (12:00 -0700)]
[GlobalOpt] Check stored once value's type before setting global initializer

In the provided test case, we were trying to set the global's
initializer to `i32* null` when the global's value type was `@0`.

Reviewed By: rnk

Differential Revision: https://reviews.llvm.org/D108232

2 years ago[cmake] Properly support target properties.
Stephen Neuendorffer [Mon, 16 Aug 2021 00:05:25 +0000 (17:05 -0700)]
[cmake] Properly support target properties.

It's sometimes useful to use these directives when dealing with
external projects:
target_link_directories
target_link_libraries
target_include_directories

However, under certain circumstances,
llvm_add_library can generate multiple targets.  We need to transfer
these properties to the new targets.  Note that using a generator
expression is necessary because these properties will only be set
after llvm_add_library is called.

Differential Revision: https://reviews.llvm.org/D108098

2 years ago[libc] Add strtol, strtoul, and strtoull
Michael Jones [Thu, 12 Aug 2021 21:34:02 +0000 (21:34 +0000)]
[libc] Add strtol, strtoul, and strtoull

Updates the internal string conversion function so that it
uses the new Limits.h added in a previous commit for max and min values,
and has a templated type. This makes implementing the other strto*
functions very simple.

Reviewed By: sivachandra

Differential Revision: https://reviews.llvm.org/D107999

2 years ago[X86] EmitInstrWithCustomInserter - silence uninitialized variable warnings. NFCI.
Simon Pilgrim [Tue, 17 Aug 2021 21:00:54 +0000 (22:00 +0100)]
[X86] EmitInstrWithCustomInserter - silence uninitialized variable warnings. NFCI.

Consistently use llvm_unreachable for the default case in the inner switch statements.

2 years agoAdd missing nullptr check
Adrian Prantl [Tue, 17 Aug 2021 20:58:40 +0000 (13:58 -0700)]
Add missing nullptr check

Unfortunatley the IR Verifier doesn't reject debug intrinsics that
have nullptr as arguments, so coro::salvageDebugInfo for now also
needs to deal with them.

rdar://81979541

2 years ago[LICM] Drop -licm-n2-threshold option
Nikita Popov [Tue, 17 Aug 2021 20:34:33 +0000 (22:34 +0200)]
[LICM] Drop -licm-n2-threshold option

This was a diagnostic option used to demonstrate a weakness in
the AST-based LICM implementation. This problem does not exist
in the MSSA-based LICM implementation, which has been enabled
for a long time now. As such, this option is no longer relevant.

2 years ago[llvm-readobj] Refactor ELFDumper::printAttributes()
Jozef Lawrynowicz [Tue, 17 Aug 2021 20:28:31 +0000 (13:28 -0700)]
[llvm-readobj] Refactor ELFDumper::printAttributes()

The current implementation of printAttributes makes it fiddly to extend
attribute support for new targets.

By refactoring the code so all target specific variables are
initialized in a switch/case statement, it becomes simpler to extend
attribute support for new targets.

Reviewed By: jhenderson, MaskRay

Differential Revision: https://reviews.llvm.org/D107968

2 years ago[lldb] Fix#2 of DW_AT_ranges DW_FORM_sec_offset not using DW_AT_rnglists_base (used...
Jan Kratochvil [Tue, 17 Aug 2021 20:16:43 +0000 (22:16 +0200)]
[lldb] Fix#2 of DW_AT_ranges DW_FORM_sec_offset not using DW_AT_rnglists_base (used by GCC)

Fix D98289 so that it works even for 2nd..nth compilation unit
(.debug_rnglists).

Reviewed By: dblaikie, ikudrin

Differential Revision: https://reviews.llvm.org/D106466

2 years agoRemove unused imports.
Dan Albert [Tue, 17 Aug 2021 20:03:27 +0000 (13:03 -0700)]
Remove unused imports.

2 years agoSimplify setting up LLVM as bazel external repo
Christian Sigg [Sun, 8 Aug 2021 13:19:33 +0000 (15:19 +0200)]
Simplify setting up LLVM as bazel external repo

Only require one intermediate repository instead of two.
Fewer parameters in llvm_config.
Remove bazel_skylib dependency.

Reviewed By: goncharov

Differential Revision: https://reviews.llvm.org/D107714

2 years ago[mlir][linalg] Don't drop existing attributes when creating ops
Lei Zhang [Tue, 17 Aug 2021 19:44:31 +0000 (15:44 -0400)]
[mlir][linalg] Don't drop existing attributes when creating ops

Reviewed By: mravishankar

Differential Revision: https://reviews.llvm.org/D108219

2 years agoSimplify testcase from c411c1b
Weverything [Tue, 17 Aug 2021 19:07:18 +0000 (12:07 -0700)]
Simplify testcase from c411c1b

2 years ago[PassBuilder] Use loop-mssa for licm
Nikita Popov [Mon, 16 Aug 2021 18:58:31 +0000 (20:58 +0200)]
[PassBuilder] Use loop-mssa for licm

Currently specifying -licm or -passes=licm will implicitly create
-passes=loop(licm). This does not match the intended default (used
by the legacy PM and by the default pipeline) of using the
MemorySSA-based LICM implementation. As I plan to drop the non-MSSA
implementation, this will stop working entirely...

This special-cases licm to create a loop-mssa manager instead. At
this point it's still possible to use -passes='loop(licm)' to opt
into the AST-based implementation.

Differential Revision: https://reviews.llvm.org/D108155

2 years ago[llvm-profgen] Clean up code dealing with multiple binaries
wlei [Thu, 12 Aug 2021 01:01:37 +0000 (18:01 -0700)]
[llvm-profgen] Clean up code dealing with multiple binaries

As we decided to support only one binary each time, this patch cleans up the related code dealing with multiple binaries. We can use `llvm-profdata` to merge profile from multiple binaries.

Reviewed By: hoy, wenlei

Differential Revision: https://reviews.llvm.org/D108002

2 years ago[OpenMP] Fix the usage of sscanf on MinGW
Martin Storsjö [Tue, 17 Aug 2021 10:15:49 +0000 (13:15 +0300)]
[OpenMP] Fix the usage of sscanf on MinGW

KMP_SSCANF only evaluates to sscanf_s within
    #if KMP_OS_WINDOWS && KMP_MSVC_COMPAT
so we need to pass the sscanf_s specific parameters within a similar
condition.

Differential Revision: https://reviews.llvm.org/D108196

2 years ago[sanitizers] Fix building on 32 bit Windows after 7256c05ecb7
Martin Storsjö [Mon, 16 Aug 2021 08:50:10 +0000 (11:50 +0300)]
[sanitizers] Fix building on 32 bit Windows after 7256c05ecb7

On 32 bit, 'long' (which is 32 bit on Windows) is used as base
type for SIZE_T and similar.

Differential Revision: https://reviews.llvm.org/D108191

2 years ago[InstCombine] add TODO about another min/max fold; NFC
Sanjay Patel [Tue, 17 Aug 2021 17:46:08 +0000 (13:46 -0400)]
[InstCombine] add TODO about another min/max fold; NFC

Suggested in post-commit for d0975b7cb0e1

2 years ago[InstCombine] add tests for min/max with 'not' ops; NFC
Sanjay Patel [Tue, 17 Aug 2021 15:38:10 +0000 (11:38 -0400)]
[InstCombine] add tests for min/max with 'not' ops; NFC

2 years ago[RISCV] Use RISCV::RVVBitsPerBlock for RGK_ScalableVector in getRegisterBitWidth.
Craig Topper [Tue, 17 Aug 2021 17:52:34 +0000 (10:52 -0700)]
[RISCV] Use RISCV::RVVBitsPerBlock for RGK_ScalableVector in getRegisterBitWidth.

I might be wrong, but I think this is should be width of the known
min size we use for scalable vectors. It shouldn't scale with
minimum vlen.

Reviewed By: frasercrmck

Differential Revision: https://reviews.llvm.org/D107945

2 years ago[flang] Add missing call to BeginReadingRecord()
peter klausler [Fri, 13 Aug 2021 18:57:58 +0000 (11:57 -0700)]
[flang] Add missing call to BeginReadingRecord()

NAMELIST input needs to start with a call to BeginReadingRecord().
Internal unit input doesn't care (so unit tests were passing), but
external unit input does need the call and will assert without it.

Differential Revision: https://reviews.llvm.org/D108051

2 years agoSelectionDAGBuilder::visitInlineAsm - don't dereference dyn_cast<> results.
Simon Pilgrim [Tue, 17 Aug 2021 17:20:04 +0000 (18:20 +0100)]
SelectionDAGBuilder::visitInlineAsm - don't dereference dyn_cast<> results.

dyn_cast<> can return nullptr if the cast is illegal, use cast<> instead which will assert that the cast is correct.

Fixes static analyser warning.

2 years ago[AArch64] AArch64DAGToDAGISel::tryReadRegister/tryWriteRegister - don't dereference...
Simon Pilgrim [Tue, 17 Aug 2021 17:15:33 +0000 (18:15 +0100)]
[AArch64] AArch64DAGToDAGISel::tryReadRegister/tryWriteRegister - don't dereference dyn_cast<> results.

dyn_cast<> can return nullptr if the cast is illegal, use cast<> instead which will assert that the cast is correct.

Fixes static analyser warnings.

2 years ago[ARM] ARMDAGToDAGISel::tryReadRegister/tryWriteRegister - don't dereference dyn_cast...
Simon Pilgrim [Tue, 17 Aug 2021 17:13:59 +0000 (18:13 +0100)]
[ARM] ARMDAGToDAGISel::tryReadRegister/tryWriteRegister - don't dereference dyn_cast<> results.

dyn_cast<> can return nullptr if the cast is illegal, use cast<> instead which will assert that the cast is correct.

Fixes static analyser warnings.

2 years ago[libc++] Update the version of CMake in the Docker image
Louis Dionne [Tue, 17 Aug 2021 17:35:50 +0000 (13:35 -0400)]
[libc++] Update the version of CMake in the Docker image

2 years agoRevert "[tsan] Fix GCC 8.3 build after D107911"
Marco Elver [Tue, 17 Aug 2021 17:26:19 +0000 (19:26 +0200)]
Revert "[tsan] Fix GCC 8.3 build after D107911"

This reverts commit 797fe59e6b9512652c0ae5a6b69a3c6f5a573fcd.

The use of "EventType type : 3" is replicated for all Event structs and
therefore was still present. As a result this still caused failures on
older GCCs (9.2 or 8.3 or earlier).

The particular bot that was failing due to buggy GCC was fixed by
fef39cc472a773fae4761deaab1c701024ad13ec.

Therefore, no reason to keep the workaround around; revert it.

Reviewed By: vitalybuka

Differential Revision: https://reviews.llvm.org/D108192

2 years agoscan-build-py: Force the opening in utf-8
Sylvestre Ledru [Tue, 17 Aug 2021 17:17:30 +0000 (19:17 +0200)]
scan-build-py: Force the opening in utf-8

It fails on ubuntu bionic otherwise with:
```
scan-build-py-14: Run 'scan-view /tmp/scan-build-2021-08-09-09-14-36-765350-nx9s888s' to examine bug reports.
scan-build-py-14: Internal error.
Traceback (most recent call last):
  File "/usr/lib/llvm-14/lib/libscanbuild/__init__.py", line 125, in wrapper
    return function(*args, **kwargs)
  File "/usr/lib/llvm-14/lib/libscanbuild/analyze.py", line 72, in scan_build
    number_of_bugs = document(args)
  File "/usr/lib/llvm-14/lib/libscanbuild/report.py", line 35, in document
    for bug in read_bugs(args.output, html_reports_available):
  File "/usr/lib/llvm-14/lib/libscanbuild/report.py", line 282, in read_bugs
    for bug in parser(bug_file):
  File "/usr/lib/llvm-14/lib/libscanbuild/report.py", line 421, in parse_bug_html
    for line in handler.readlines():
  File "/usr/lib/python3.6/encodings/ascii.py", line 26, in decode
    return codecs.ascii_decode(input, self.errors)[0]
UnicodeDecodeError: 'ascii' codec can't decode byte 0xc3 in position 3360: ordinal not in range(128)
scan-build-py-14: Please run this command again and turn on verbose mode (add '-vvvv' as argument).
```

I guess it is caused by a problem in Python 3.6

Reviewed By: phosek, isthismyaccount

Differential Revision: https://reviews.llvm.org/D107887

2 years agohwasan: Move stack ring buffer initialization before InitStackAndTls.
Peter Collingbourne [Tue, 17 Aug 2021 03:13:42 +0000 (20:13 -0700)]
hwasan: Move stack ring buffer initialization before InitStackAndTls.

D104248 moved the call to GetThreadStackAndTls to before the
initialization of the ring buffer TLS slot. As a result, if libc
is instrumented we crash in pthread_getattr_np which is called from
__sanitizer::GetThreadStackTopAndBottom.

Fix the problem by moving the stack ring buffer initialization before
the call to InitStackAndTls.

Differential Revision: https://reviews.llvm.org/D108184

2 years ago[flang] Fix regression from recent runtime input fix
peter klausler [Mon, 16 Aug 2021 20:15:01 +0000 (13:15 -0700)]
[flang] Fix regression from recent runtime input fix

A recent runtime I/O change[1] was meant to improve the handling of
input from external files missing a terminal newline on their last
records; the change was "triggered" by the wrong circumstances and
causing reads that should have pulled more data into the buffer to be
treated as EOFs.  So fix that, and also don't retain input data
in the buffer once an input record has been finished unless it's
known that list-directed or NAMELIST input of a repeated input item
may need to backspace a non-positionable external unit to return
to the beginning of the repeated item.

[1] 6578893a0453384346f149479f8574dfff977ace

Differential Revision: https://reviews.llvm.org/D108164

2 years ago[AMDGPU] Fix lowering of AMDGPU::G_CTTZ_ZERO_UNDEF to AMDGPU::G_AMDGPU_FFBL_B32
Simon Pilgrim [Tue, 17 Aug 2021 16:57:01 +0000 (17:57 +0100)]
[AMDGPU] Fix lowering of AMDGPU::G_CTTZ_ZERO_UNDEF to AMDGPU::G_AMDGPU_FFBL_B32

As mentioned on D107474, there was a copy+paste typo repeating G_CTLZ_ZERO_UNDEF that coverity reported as dead code.

Differential Revision: https://reviews.llvm.org/D108210

2 years ago[VP] Add vector-predicated reduction intrinsics
Fraser Cormack [Mon, 14 Jun 2021 08:42:00 +0000 (09:42 +0100)]
[VP] Add vector-predicated reduction intrinsics

This patch adds vector-predicated ("VP") reduction intrinsics corresponding to
each of the existing unpredicated `llvm.vector.reduce.*` versions. Unlike the
unpredicated reductions, all VP reductions have a start value. This start value
is returned when the no vector element is active.

Support for expansion on targets without native vector-predication support is
included.

This patch is based on the ["reduction
slice"](https://reviews.llvm.org/D57504#1732277) of the LLVM-VP reference patch
(https://reviews.llvm.org/D57504).

Reviewed By: craig.topper

Differential Revision: https://reviews.llvm.org/D104308

2 years ago[libc][NFC] Disable double precision cos, sin and tan on Windows.
Siva Chandra Reddy [Tue, 17 Aug 2021 16:42:34 +0000 (16:42 +0000)]
[libc][NFC] Disable double precision cos, sin and tan on Windows.

The current x86_64 implementations do not build on the windows bot
machine. We will enable them back after fixing the problem.

2 years ago[OpenMP][NFC] Add option to print module after OpenMPOpt for debugging
Joseph Huber [Mon, 16 Aug 2021 17:06:54 +0000 (13:06 -0400)]
[OpenMP][NFC] Add option to print module after OpenMPOpt for debugging

This patch adds an extra option to print the module after running one of
the OpenMPOpt passes if debugging is enabled. This makes it much easier
to inspect the effects of this pass when doing debugging.

Reviewed By: jdoerfert

Differential Revision: https://reviews.llvm.org/D108146

2 years ago[SCEVExpander] Stop hoisting IR when reusing phis
Philip Reames [Tue, 17 Aug 2021 16:26:59 +0000 (09:26 -0700)]
[SCEVExpander] Stop hoisting IR when reusing phis

his is a fix for PR43678, and is an alternate patch to D105723.

The basic issue we're running into is that LSR + SCEVExpander are moving the very instruction whose operand we're in the process of expanding. This breaks the subtle and ill-documented invariant which let LSR work. (Full story can be found here: https://reviews.llvm.org/D105723#2878473)

Rather than attempting a fix, this change just removes the optimization entirely. The code is entirely untested, and removing it appears to have no impact I can find.  This code was added back in 2014 by 1e12f8563d4b7 with a single test which does not seem to actually test the hoisting logic.

From a philosophical standpoint, it also seems very strange to have the expander implementing optimizations which should live in a dedicated transform pass.

Differential Revision: https://reviews.llvm.org/D106178

2 years ago[analyzer] Adjust JS code of analyzer's HTML report for IE support.
Denys Petrov [Tue, 3 Aug 2021 16:09:00 +0000 (19:09 +0300)]
[analyzer] Adjust JS code of analyzer's HTML report for IE support.

Summary: Change and replace some functions which IE does not support. This patch is made as a continuation of D92928 revision. Also improve hot keys behavior.

Differential Revision: https://reviews.llvm.org/D107366

2 years agoFix 2: [MCParser] Correctly handle CRLF line ends when consuming line comments
Tozer [Tue, 17 Aug 2021 16:09:44 +0000 (17:09 +0100)]
Fix 2: [MCParser] Correctly handle CRLF line ends when consuming line comments

Fixes an issue with revision 5c6f748c and ad40cb88.

Adds an mcpu argument to the test command, preventing an invalid default
CPU from being used on some platforms.

2 years ago[llvm-objdump] -T: print symbol versions
Fangrui Song [Tue, 17 Aug 2021 16:10:50 +0000 (09:10 -0700)]
[llvm-objdump] -T: print symbol versions

Similar to D94907 (llvm-nm -D).

The output will match GNU objdump 2.37.
Older versions don't use ` (version)` for undefined symbols.

Reviewed By: jhenderson

Differential Revision: https://reviews.llvm.org/D108097

2 years agoAllow setting attributes in build method generated by YAML-gen.
MaheshRavishankar [Tue, 17 Aug 2021 15:59:13 +0000 (08:59 -0700)]
Allow setting attributes in build method generated by YAML-gen.

Reviewed By: gysit

Differential Revision: https://reviews.llvm.org/D108182

2 years ago[Object] Move llvm-nm's symbol version utility to ELFObjectFile::readDynsymVersions
Fangrui Song [Tue, 17 Aug 2021 16:06:39 +0000 (09:06 -0700)]
[Object] Move llvm-nm's symbol version utility to ELFObjectFile::readDynsymVersions

The utility can be reused by llvm-objdump -T.

Reviewed By: jhenderson

Differential Revision: https://reviews.llvm.org/D108096