platform/upstream/llvm.git
4 years ago[DWARFYAML] Add support for emitting custom range list content.
Xing GUO [Tue, 28 Jul 2020 14:10:44 +0000 (22:10 +0800)]
[DWARFYAML] Add support for emitting custom range list content.

This patch adds support for emitting custom range list content.

We are able to handcraft a custom range list via the following syntax.

```
debug_rnglists:
  - Lists:
      - Entries:
          - Operator: DW_RLE_startx_endx
            Values:   [ 0x1234, 0x1234 ]
      - Content: '1234567890abcdef'
      - Content: 'abcdef1234567890'
```

Reviewed By: jhenderson

Differential Revision: https://reviews.llvm.org/D84618

4 years agoRe-land "[PowerPC] Remove QPX/A2Q BGQ/BGP CNK support"
Jinsong Ji [Tue, 28 Jul 2020 03:02:20 +0000 (03:02 +0000)]
Re-land "[PowerPC] Remove QPX/A2Q BGQ/BGP CNK support"

This reverts commit bf544fa1c3cb80f24d85e84559fb11193846259f.

Fixed the typo in PPCInstrInfo.cpp.

4 years ago[clang-tidy][NFC] Make OptionsView methods as const where missing
Nathan James [Tue, 28 Jul 2020 13:52:32 +0000 (14:52 +0100)]
[clang-tidy][NFC] Make OptionsView methods as const where missing

4 years ago[yaml2obj] - Add a way to override sh_type section field.
Georgii Rymar [Tue, 28 Jul 2020 09:53:06 +0000 (12:53 +0300)]
[yaml2obj] - Add a way to override sh_type section field.

This adds the `ShType` key similar to others `Sh*` keys we have.

My use case is the following. Imagine we have a `SHT_SYMTAB_SHNDX`
section and want to hide it from a dumper. The natural way would be to
do something like:

```
  - Name:    .symtab_shndx
    Type:    [[TYPE=SHT_SYMTAB_SHNDX]]
    Entries: [ 0, 1 ]

```

and then change the TYPE from `SHT_SYMTAB_SHNDX` to something else,
for example to `SHT_PROGBITS`.

But we have a problem: regular sections does not have `Entries` key,
so yaml2obj will be unable to produce a section.

The solution is to introduce a `ShType` key to override the final type.

This is not the first time I am facing the need to change the type. I
was able to invent workarounds or solved issues differently in the past,
but finally came to conclusion that we just should support the `ShType`.

Differential revision: https://reviews.llvm.org/D84738

4 years ago[OpenMP] Fix libomptarget negative tests to expect abort
Joel E. Denny [Tue, 28 Jul 2020 12:57:03 +0000 (08:57 -0400)]
[OpenMP] Fix libomptarget negative tests to expect abort

On runtime failures, D83963 causes the runtime to abort instead of
merely exiting with a non-zero value, but many tests in the
libomptarget test suite still expect the former behavior.  This patch
updates the test suite and was discussed in post-commit comments on
D83963 and D84557.

4 years ago[BPI] Fix memory leak reported by sanitizer bots
Evgeniy Brevnov [Tue, 28 Jul 2020 12:50:40 +0000 (19:50 +0700)]
[BPI] Fix memory leak reported by sanitizer bots

There is a silly mistake where release() is used instead of reset() for free resources of unique pointer.

Reviewed By: ebrevnov

Differential Revision: https://reviews.llvm.org/D84747

4 years agoARM: make Thumb1 instructions non-flag-setting in IT block.
Tim Northover [Mon, 27 Jul 2020 13:37:14 +0000 (14:37 +0100)]
ARM: make Thumb1 instructions non-flag-setting in IT block.

Many Thumb1 instructions are defined to set CPSR if executed outside an IT
block, but leave it alone from inside one. In MachineIR this is represented by
whether an optional register is CPSR or NoReg (0), and affects how the
instructions are printed.

This sets the instruction to the appropriate form during if-conversion.

4 years ago[MLIR][Shape] Support transforming shape.num_elements on tensors
Stephan Herhut [Tue, 28 Jul 2020 11:09:45 +0000 (13:09 +0200)]
[MLIR][Shape] Support transforming shape.num_elements on tensors

The current transformation to shape.reduce does not support tensor values.
This adds the required changes to make that work, including fixing the builder
for shape.reduce.

Differential Revision: https://reviews.llvm.org/D84744

4 years ago[Support] Use InitLLVM in llvm-stress, sancov and TableGen
Anirudh Prasad [Tue, 28 Jul 2020 10:40:47 +0000 (06:40 -0400)]
[Support] Use InitLLVM in llvm-stress, sancov and TableGen

This patch refactors the llvm tools namely, llvm-stress and sancov,
as well as the llvm TableGen utility, to use the new InitLLVM
interface which encapsulates PrettyStackTrace.

This is from https://reviews.llvm.org/D70702, but only for LLVM.

Reviewed-by: Kai
Differential Revision: https://reviews.llvm.org/D83484

4 years ago[Attributor] Add override keyword to the print function of AA
Luofan Chen [Tue, 28 Jul 2020 11:19:23 +0000 (19:19 +0800)]
[Attributor] Add override keyword to the print function of AA

The print() function in the `AbstractAttribute` structure overrides
the function in the `AADepGraphNode`, so we need to mark it as
override.

This should fix a buildbot failure introduced by 5ee07dc.

4 years ago[compiler-rt][cmake] Don't pass --version-script to Illumos ld
Rainer Orth [Tue, 28 Jul 2020 11:21:36 +0000 (13:21 +0200)]
[compiler-rt][cmake] Don't pass --version-script to Illumos ld

Neither the Illumos `ld` nor the Solaris 11.3 one support the `--version-script` and
`z gnu-linker-script-compat` options, which breaks the `compiler-rt` build.

This patch checks for both options instead of hardcoding their use.

Tested on `amd-pc-solaris2.11` (all of Solaris 11.4, 11.3, and Illumos).

Differential Revision: https://reviews.llvm.org/D84559

4 years ago[PowerPC] Split s34imm into two types
Stefan Pintilie [Mon, 27 Jul 2020 17:27:30 +0000 (12:27 -0500)]
[PowerPC] Split s34imm into two types

Currently the instruction paddi always takes s34imm as the type for the
34 bit immediate. However, the PC Relative form of the instruction should
not produce the same fixup as the non PC Relative form.
This patch splits the s34imm type into s34imm and s34imm_pcrel so that two
different fixups can be emitted.

Reviewed By: nemanjai, #powerpc, kamaub

Differential Revision: https://reviews.llvm.org/D83255

4 years ago[BPI][NFC] Consolidate code to deal with SCCs under a dedicated data structure.
Evgeniy Brevnov [Fri, 24 Jul 2020 11:57:10 +0000 (18:57 +0700)]
[BPI][NFC] Consolidate code to deal with SCCs under a dedicated data structure.

In order to facilitate review of D79485 here is a small NFC change which restructures code around handling of SCCs in BPI.

Reviewed By: davidxl

Differential Revision: https://reviews.llvm.org/D84514

4 years ago[SystemZ/ZOS] Implement setLastAccessAndModificationTime()
Kai Nacke [Thu, 2 Jul 2020 12:43:42 +0000 (14:43 +0200)]
[SystemZ/ZOS] Implement setLastAccessAndModificationTime()

The function setLastAccessAndModificationTime() uses function
futimens() or futimes() by default. Both functions are not
available in z/OS, therefore functionality is implemented using
__fchattr() on z/OS.

Reviews by: abhina.sreeskantharajan

Differential Revision: https://reviews.llvm.org/D83945

4 years ago[MLIR][Linalg] Retire C++ DotOp in favor of a linalg-ods-gen'd op
lorenzo chelini [Tue, 28 Jul 2020 10:29:54 +0000 (12:29 +0200)]
[MLIR][Linalg] Retire C++ DotOp in favor of a linalg-ods-gen'd op

- replace DotOp, now that DRR rules have been dropped.

- Capture arguments mismatch in the parser. The number of parsed arguments must
  equal the number of expected arguments.

Reviewed By: ftynse, nicolasvasilache

Differential Revision: https://reviews.llvm.org/D82952

4 years agoAvoid use of std::make_unique in compiler-rt/lib/scudo/standalone/tests/combined_test.cpp
Hans Wennborg [Tue, 28 Jul 2020 10:26:37 +0000 (12:26 +0200)]
Avoid use of std::make_unique in compiler-rt/lib/scudo/standalone/tests/combined_test.cpp

make_unique is a C++14 feature, and this prevents us from building on
Ubuntu Trusty. While we do use a C++14 compatible toolchain for building
in general, we fall back to the system toolchain for building the
compiler-rt tests.

The reason is that those tests get cross-compiled for e.g. 32-bit and
64-bit x86, and while the toolchain provides libstdc++ in those
flavours, the resulting compiler-rt test binaries don't get RPATH set
and so won't start if they're linked with that toolchain.

We've tried linking the test binaries against libstdc++ statically, by
passing COMPILER_RT_TEST_COMPILER_CFLAGS=-static-libstdc++. That mostly
works, but some test targets append -lstdc++ to the compiler invocation.

So, after spending way too much time on this, let's just avoid C++14
here for now.

4 years ago[llvm-readobj] - Add comments and formatting to mips-options-sec.test and mips-reginf...
Georgii Rymar [Tue, 28 Jul 2020 10:31:13 +0000 (13:31 +0300)]
[llvm-readobj] - Add comments and formatting to mips-options-sec.test and mips-reginfo.test. NFCI.

This will allow to extend them (needed for D84651).

4 years ago[llvm-readelf] Symbol index in symbol table printing is not reset
Mikhail Kalashnikov [Tue, 28 Jul 2020 10:06:51 +0000 (13:06 +0300)]
[llvm-readelf] Symbol index in symbol table printing is not reset

Stop using static variables for keeping track of symbol indices.

Bugfix for: https://bugs.llvm.org/show_bug.cgi?id=46777

Differential revision: https://reviews.llvm.org/D84606

4 years ago[Concepts] Fix ast dump for immediately declared constraint.
Haojian Wu [Tue, 28 Jul 2020 09:59:49 +0000 (11:59 +0200)]
[Concepts] Fix ast dump for immediately declared constraint.

Reviewed By: nridge

Differential Revision: https://reviews.llvm.org/D84461

4 years ago[Attributor] Track AA dependency using dependency graph
Luofan Chen [Tue, 28 Jul 2020 10:02:49 +0000 (18:02 +0800)]
[Attributor] Track AA dependency using dependency graph

This patch added dependency graph to the attributor so that we can dump the dependencies between AAs more easily. We can also apply general graph algorithms to the graph, making it easier for us to create deep wrappers.

Reviewed By: jdoerfert

Differential Revision: https://reviews.llvm.org/D78861

4 years ago[clang-format][NFC] Fix a Wdocumentation warning in TokenAnnotator.cpp
Bruno Ricci [Tue, 28 Jul 2020 09:58:34 +0000 (10:58 +0100)]
[clang-format][NFC] Fix a Wdocumentation warning in TokenAnnotator.cpp

4 years ago[Support] Add file lock/unlock functions
Serge Pavlov [Wed, 22 Apr 2020 16:00:12 +0000 (23:00 +0700)]
[Support] Add file lock/unlock functions

This is recommit of f51bc4fb60fb, reverted in 8577595e03fa, because
the function `flock` is not available on Solaris. In this variant
`flock` was replaced with `fcntl`, which is a POSIX function.

New functions `lockFile`, `tryLockFile` and `unlockFile` implement
simple file locking. They lock or unlock entire file. This must be
enough to support simulataneous writes to log files in parallel builds.

Differential Revision: https://reviews.llvm.org/D78896

4 years ago[clang] Pass the NamedDecl* instead of the DeclarationName into many diagnostics.
Bruno Ricci [Mon, 27 Jul 2020 22:22:21 +0000 (23:22 +0100)]
[clang] Pass the NamedDecl* instead of the DeclarationName into many diagnostics.

Background:
-----------
There are two related argument types which can be sent into a diagnostic to
display the name of an entity: DeclarationName (ak_declarationname) or
NamedDecl* (ak_nameddecl) (there is also ak_identifierinfo for
IdentifierInfo*, but we are not concerned with it here).

A DeclarationName in a diagnostic will just be streamed to the output,
which will directly result in a call to DeclarationName::print.

A NamedDecl* in a diagnostic will also ultimately result in a call to
DeclarationName::print, but with two customisation points along the way:

The first customisation point is NamedDecl::getNameForDiagnostic which is
overloaded by FunctionDecl, ClassTemplateSpecializationDecl and
VarTemplateSpecializationDecl to print the template arguments, if any.

The second customisation point is NamedDecl::printName. By default it just
streams the stored DeclarationName into the output but it can be customised
to provide a user-friendly name for an entity. It is currently overloaded by
DecompositionDecl and MSGuidDecl.

What this patch does:
---------------------
For many diagnostics a DeclarationName is used instead of the NamedDecl*.
This bypasses the two customisation points mentioned above. This patches fix
this for diagnostics in Sema.cpp, SemaCast.cpp, SemaChecking.cpp, SemaDecl.cpp,
SemaDeclAttr.cpp, SemaDecl.cpp, SemaOverload.cpp and SemaStmt.cpp.

I have only modified diagnostics where I could construct a test-case which
demonstrates that the change is appropriate (either with this patch or the next
one).

Reviewed By: erichkeane, aaron.ballman

Differential Revision: https://reviews.llvm.org/D84656

4 years ago[clang-format][NFC] Be more careful about the layout of FormatToken.
Bruno Ricci [Mon, 27 Jul 2020 22:19:02 +0000 (23:19 +0100)]
[clang-format][NFC] Be more careful about the layout of FormatToken.

The underlying ABI forces FormatToken to have a lot of padding.

Currently (on x86-64 linux) `sizeof(FormatToken) == 288`. After this patch
`sizeof(FormatToken) == 232`.

No functional changes.

Reviewed By: MyDeveloperDay

Differential Revision: https://reviews.llvm.org/D84306

4 years ago[OpenMP][Tests] Enable nvptx64 testing for most libomptarget tests
Joachim Protze [Tue, 28 Jul 2020 09:08:24 +0000 (11:08 +0200)]
[OpenMP][Tests] Enable nvptx64 testing for most libomptarget tests

Also add $BUILD/lib to the LIBRARY_PATH to fix
https://bugs.llvm.org/show_bug.cgi?id=46836.

Reviewed By: JonChesterfield

Differential Revision: https://reviews.llvm.org/D84557

4 years ago[X86][SSE] Attempt to match OP(SHUFFLE(X,Y),SHUFFLE(X,Y)) -> SHUFFLE(HOP(X,Y))
Simon Pilgrim [Tue, 28 Jul 2020 08:52:38 +0000 (09:52 +0100)]
[X86][SSE] Attempt to match OP(SHUFFLE(X,Y),SHUFFLE(X,Y)) -> SHUFFLE(HOP(X,Y))

An initial backend patch towards fixing the various poor HADD combines (PR34724, PR41813, PR45747 etc.).

This extends isHorizontalBinOp to check if we have per-element horizontal ops (odd+even element pairs), but not in the expected serial order - in which case we build a "post shuffle mask" that we can apply to the HOP result, assuming we have fast-hops/optsize etc.

The next step will be to extend the SHUFFLE(HOP(X,Y)) combines as suggested on PR41813 - accepting more post-shuffle masks even on slow-hop targets if we can fold it into another shuffle.

Differential Revision: https://reviews.llvm.org/D83789

4 years ago[legacyPM] Do not compute preserved analysis if there's no local change
serge-sans-paille [Tue, 28 Jul 2020 08:55:09 +0000 (10:55 +0200)]
[legacyPM] Do not compute preserved analysis if there's no local change

All analysis are preserved if there's no local change, and thanks to
3667d87a33d3c8d4072a41fd84bb880c59347dc0 this property is enforced for all
passes.

Skipping the dependency computation improves the performance when there's a lot
of small functions, where only a few change happen.

Thanks to Nikita Popov who provided this numbers (extract below)

https://llvm-compile-time-tracker.com/compare.php?from=183342c0a9850e60dd7a004b651c83dfb3a7d25e&to=f2f91e6a2743070471cc9471e4e8c646e50c653c&stat=instructions

O3: (number of instructions)
Benchmark               Old             New
kimwitu++               60783M          59968M          (-1.34%)
sqlite3                 73200M          73083M          (-0.16%)
consumer-typeset        52776M          52712M          (-0.12%)
Bullet                  133709M         132940M         (-0.58%)
tramp3d-v4              123864M         123186M         (-0.55%)
mafft                   55534M          55477M          (-0.10%)
ClamAV                  76292M          76164M          (-0.17%)
lencod                  103190M         103061M         (-0.13%)
SPASS                   64068M          63713M          (-0.55%)
7zip                    197332M         196308M         (-0.52%)
geomean                 85750M          85389M          (-0.42%)

Differential Revision: https://reviews.llvm.org/D80707

4 years ago[llvm-readelf] - Do not treat SHT_ANDROID_RELR sections the same as SHT_RELR.
Georgii Rymar [Thu, 23 Jul 2020 10:32:05 +0000 (13:32 +0300)]
[llvm-readelf] - Do not treat SHT_ANDROID_RELR sections the same as SHT_RELR.

Currently, when dumping section headers, llvm-readelf
prints "RELR" for SHT_ANDROID_RELR/SHT_RELR sections.
The behavior was introduced in D47919 and revealed in D84330.

But "SHT_ANDROID_RELR" has a different value from "SHT_RELR".
Also, "SHT_ANDROID_REL/SHT_ANDROID_RELA" are printed as "ANDROID_REL/ANDROID_RELA",
what makes the handling of the "SHT_ANDROID_RELR" inconsistent.

This patch makes llvm-readelf to print "ANDROID_RELR" instead of "RELR".

Differential revision: https://reviews.llvm.org/D84393

4 years ago[clang][cmake] Force CMAKE_LINKER for multistage build in case of BOOTSTRAP_LLVM_ENAB...
Kristina Bessonova [Fri, 29 May 2020 11:14:51 +0000 (13:14 +0200)]
[clang][cmake] Force CMAKE_LINKER for multistage build in case of BOOTSTRAP_LLVM_ENABLE_LLD and MSVC

The issue with LLVM_ENABLE_LLD is that it just passes -fuse-ld=lld
to compiler/linker options which makes sense only for those platforms
where cmake invokes a compiler driver for linking. On Windows (MSVC) cmake
invokes the linker directly and requires CMAKE_LINKER to be specified
otherwise it defaults CMAKE_LINKER to be link.exe.

This patch allows BOOTSTRAP_LLVM_ENABLE_LLD to set CMAKE_LINKER in two cases:
* if building for host Windows,
* if crosscompiling for target Windows.

It also skips adding '-fuse-ld=lld' to make lld-link not warning
about 'unknown argument'.

This fixes build with `clang/cmake/caches/DistributionExample.cmake`
on Windows.

Reviewed By: phosek

Differential Revision: https://reviews.llvm.org/D80873

4 years ago[PowerPC] test case for adding dq form to isLegalAddressingMode, nfc
Chen Zheng [Tue, 28 Jul 2020 06:56:35 +0000 (02:56 -0400)]
[PowerPC] test case for adding dq form to isLegalAddressingMode, nfc

4 years ago[GVN] Rewrite IsValueFullyAvailableInBlock(): no recursion, less false-negatives
Roman Lebedev [Tue, 28 Jul 2020 07:16:52 +0000 (10:16 +0300)]
[GVN] Rewrite IsValueFullyAvailableInBlock(): no recursion, less false-negatives

While this doesn't appear to help with the perf issue being exposed by
D84108, the function as-is is very weird, convoluted, and what's worse,
recursive.

There was no need for `SpeculativelyAvaliableAndUsedForSpeculation`,
tri-state choice is enough. We don't even ever check for that state.

The basic idea here is that we need to perform a depth-first traversal
of the predecessors of the basic block in question, either finding a
preexisting state for the block in a map, or inserting a "placeholder"
`SpeculativelyAvaliable`,

If we encounter an `Unavaliable` block, then we need to give up search,
and back-propagate the `Unavaliable` state to the each successor of
said block, more specifically to the each `SpeculativelyAvaliable`
we've just created.

However, if we have traversed entirety of the predecessors and have not
encountered an `Unavaliable` block, then it must mean the value is fully
available. We could update each inserted `SpeculativelyAvaliable` into
a `Avaliable`, but we don't need to, as assertion excersizes,
because we can assume that if we see an `SpeculativelyAvaliable` entry,
it is actually `Avaliable`, because during the time we've produced it,
if we would have found that it has an `Unavaliable` predecessor,
we would have updated it's successors, including this block,
into `Unavaliable`

Reviewed By: fhahn

Differential Revision: https://reviews.llvm.org/D84181

4 years ago[mlir][NFC] Polish copy removal transform
Ehsan Toosi [Fri, 24 Jul 2020 13:43:59 +0000 (15:43 +0200)]
[mlir][NFC] Polish copy removal transform

Address a few remaining comments in copy removal transform.

Differential Revision: https://reviews.llvm.org/D84529

4 years ago[clang-tidy] Suppress one unittest on macOS.
Artem Dergachev [Thu, 23 Jul 2020 18:37:45 +0000 (11:37 -0700)]
[clang-tidy] Suppress one unittest on macOS.

Possibly a linker bug but I'm in a hurry to fix a buildbot.

Differential Revision: https://reviews.llvm.org/D84453

4 years ago[X86] Detect if EFLAGs is live across XBEGIN pseudo instruction. Add it as livein...
Craig Topper [Tue, 28 Jul 2020 04:11:54 +0000 (21:11 -0700)]
[X86] Detect if EFLAGs is live across XBEGIN pseudo instruction. Add it as livein to the basic blocks created when expanding the pseudo

XBEGIN causes several based blocks to be inserted. If flags are live across it we need to make eflags live in the new basic blocks to avoid machine verifier errors.

Fixes PR46827

Reviewed By: ivanbaev

Differential Revision: https://reviews.llvm.org/D84479

4 years ago[X86] Add support for {disp32} to control size of jmp and jcc instructions in the...
Craig Topper [Tue, 28 Jul 2020 04:11:48 +0000 (21:11 -0700)]
[X86] Add support for {disp32} to control size of jmp and jcc instructions in the assembler

By default we pick a 1 byte displacement and let relaxation enlarge it if necessary. The GNU assembler supports a pseudo prefix to basically pre-relax the instruction the larger size.

I plan to add {disp8} and {disp32} support for memory operands in another patch which is why I've included the parsing code and enum for {disp8} pseudo prefix as well.

Reviewed By: echristo

Differential Revision: https://reviews.llvm.org/D84709

4 years ago[X86] Properly encode a 32-bit address with an index register and no base register...
Craig Topper [Tue, 28 Jul 2020 04:11:42 +0000 (21:11 -0700)]
[X86] Properly encode a 32-bit address with an index register and no base register in 16-bit mode.

In 16-bit mode we can encode a 32-bit address using 0x67 prefix.
We were failing to do this when the index register was a 32-bit
register, the base register was not present, and the displacement
fit in 16-bits.

Fixes PR46866.

4 years agoSupplement instr profile with sample profile.
Wei Mi [Wed, 8 Jul 2020 22:19:44 +0000 (15:19 -0700)]
Supplement instr profile with sample profile.

PGO profile is usually more precise than sample profile. However, PGO profile
needs to be collected from loadtest and loadtest may not be representative
enough to the production workload. Sample profile collected from production
can be used as a supplement -- for functions cold in loadtest but warm/hot
in production, we can scale up the related function in PGO profile if the
function is warm or hot in sample profile.

The implementation contains changes in compiler side and llvm-profdata side.
Given an instr profile and a sample profile, for a function cold in PGO
profile but warm/hot in sample profile, llvm-profdata will either mark
all the counters in the profile to be -1 or scale up the max count in the
function to be above hot threshold, depending on the zero counter ratio in
the profile. The assumption is if there are too many counters being zero
in the function profile, the profile is more likely to cause harm than good,
then llvm-profdata will mark all the counters to be -1 indicating the
function is hot but the profile is unaccountable. In compiler side, if a
function profile with all -1 counters is seen, the function entry count will
be set to be above hot threshold but its internal profile will be dropped.

In the long run, it may be useful to let compiler support using PGO profile
and sample profile at the same time, but that requires more careful design
and more substantial changes to make two profiles work seamlessly. The patch
here serves as a simple intermediate solution.

Differential Revision: https://reviews.llvm.org/D81981

4 years agoDon't form a 'context-independent expr' reference to a member during
Richard Smith [Tue, 28 Jul 2020 00:28:06 +0000 (17:28 -0700)]
Don't form a 'context-independent expr' reference to a member during
name annotation.

Instead, defer forming the member access expression or DeclRefExpr until
we build the use of ClassifyName's result. Just build an
UnresolvedLookupExpr to track the LookupResult until we're ready to
consume it.

This also reverts commit 2f7269b6773de2750f9cd1417ef5f21cd6cf7a91 (other
than its testcase). That change was an attempted workaround for the same
problem.

4 years ago[llvm-readelf] Fix emitting incorrect number of spaces in '--hex-dump'.
Xing GUO [Tue, 28 Jul 2020 01:53:59 +0000 (09:53 +0800)]
[llvm-readelf] Fix emitting incorrect number of spaces in '--hex-dump'.

This patch helps teach llvm-readelf to emit a correct number spaces when
dumping in hex format.

Before this patch, when the hex data doesn't fill the 4th column, some
spaces are missing.

```
Hex dump of section '.sec':
0x00000000 00000000 00000000 00000000 00000000 ................
0x00000010 00000000 00000000 00000000 0000 ..............
```

After this patch:

```
Hex dump of section '.sec':
0x00000000 00000000 00000000 00000000 00000000 ................
0x00000010 00000000 00000000 00000000 0000     ..............
```

Reviewed By: grimar

Differential Revision: https://reviews.llvm.org/D84640

4 years agoTableGen: Check if pattern outputs matches instruction defs
Matt Arsenault [Tue, 28 Jul 2020 00:29:53 +0000 (20:29 -0400)]
TableGen: Check if pattern outputs matches instruction defs

Attempt to fix address sanitizer bots when building ARM.

4 years ago[CFGDiff] Refactor Succ/Pred maps.
Alina Sbirlea [Fri, 24 Jul 2020 23:29:58 +0000 (16:29 -0700)]
[CFGDiff] Refactor Succ/Pred maps.

Summary:
Refactor Succ/Pred maps to have a single map lookup when constructing
children. The preivous desing made sense when used by GraphTraits.
This more closely matches the previous approach in DomTree.

Reviewers: dblaikie

Subscribers: llvm-commits

Tags: #llvm

Differential Revision: https://reviews.llvm.org/D84567

4 years ago[DomTree] Remove dead code.[NFC]
Alina Sbirlea [Mon, 27 Jul 2020 23:41:55 +0000 (16:41 -0700)]
[DomTree] Remove dead code.[NFC]

4 years ago[lldb/ArchSpec] Always match simulator environment in IsEqualTo
Fred Riss [Tue, 28 Jul 2020 00:05:27 +0000 (17:05 -0700)]
[lldb/ArchSpec] Always match simulator environment in IsEqualTo

Summary:
Initially, Apple simulator binarie triples didn't use a `-simulator`
environment and were just differentiated based on the architecture.
For example, `x86_64-apple-ios` would obviously be a simualtor as iOS
doesn't run on x86_64. With Catalyst, we made the disctinction
explicit and today, all simulator triples (even the legacy ones) are
constructed with an environment. This is especially important on Apple
Silicon were the architecture is not different from the one of the
simulated device.

This change makes the simulator part of the environment always part of
the criteria to detect whether 2 `ArchSpec`s are equal or compatible.

Reviewers: aprantl

Subscribers: inglorion, dexonsmith, lldb-commits

Tags: #lldb

Differential Revision: https://reviews.llvm.org/D84716

4 years ago[CMake] Move find_package(ZLIB) to LLVMConfig
Petr Hosek [Mon, 27 Jul 2020 18:55:52 +0000 (11:55 -0700)]
[CMake] Move find_package(ZLIB) to LLVMConfig

This way, downstream projects don't have to invoke find_package(ZLIB)
reducing the amount of boilerplate.

Differential Revision: https://reviews.llvm.org/D84691

4 years ago[OpenMP][Docs] Update `present` modifier status
Joel E. Denny [Mon, 27 Jul 2020 23:22:05 +0000 (19:22 -0400)]
[OpenMP][Docs] Update `present` modifier status

4 years ago[GraphDiff] Use class method getChildren instead of GraphTraits.
Alina Sbirlea [Thu, 16 Jul 2020 22:46:54 +0000 (15:46 -0700)]
[GraphDiff] Use class method getChildren instead of GraphTraits.

Summary:
Use getChildren() method in GraphDiff instead of GraphTraits.

This simplifies the code and allows for refactorigns inside GraphDiff.
All usecase need not have a light-weight/copyable range.
Clean GraphTraits implementation.

Reviewers: dblaikie

Subscribers: hiraditya, llvm-commits, george.burgess.iv

Tags: #llvm

Differential Revision: https://reviews.llvm.org/D84562

4 years agoGlobalISel: Don't fail translate on intrinsics with metadata
Matt Arsenault [Tue, 21 Jul 2020 23:29:37 +0000 (19:29 -0400)]
GlobalISel: Don't fail translate on intrinsics with metadata

4 years agoAMDGPU/GlobalISel: Mark G_ATOMICRMW_{NAND|FSUB} as lower
Matt Arsenault [Sat, 25 Jul 2020 14:17:49 +0000 (10:17 -0400)]
AMDGPU/GlobalISel: Mark G_ATOMICRMW_{NAND|FSUB} as lower

These aren't implemented and we're still relying on the AtomicExpand
pass, but mark these as lower to eliminate a few of the few remaining
no rules defined cases.

4 years agoAMDGPU: global_atomic_csub is not always dereferenceable
Matt Arsenault [Wed, 22 Jul 2020 00:53:57 +0000 (20:53 -0400)]
AMDGPU: global_atomic_csub is not always dereferenceable

4 years ago[llvm] Make ZLIB handling compatible with multi-configuration generators
Jonas Devlieghere [Mon, 27 Jul 2020 22:36:16 +0000 (15:36 -0700)]
[llvm] Make ZLIB handling compatible with multi-configuration generators

The CMAKE_BUILD_TYPE is only meaningful to single-configuration
generators (such as make and Ninja). For multi-configuration generators
like Xcode and MSVC this variable won't be set, resulting in a CMake
error.

4 years agoTableGen/GlobalISel: Allow output instructions with multiple defs
Matt Arsenault [Mon, 13 Jul 2020 12:59:38 +0000 (08:59 -0400)]
TableGen/GlobalISel: Allow output instructions with multiple defs

The DAG behavior allows matchching input patterns with a single result
to the first result of an output instruction that defines multiple
results. The remaining defs are implicitly dead.

This starts to fix using manual selection for AMDGPU add/sub (although
it's still needed, mostly because it's also still needed for
G_PTR_ADD).

4 years ago[llvm][CodeGen] Addressing modes for SVE ldN.
Francesco Petrogalli [Mon, 6 Jul 2020 17:46:59 +0000 (17:46 +0000)]
[llvm][CodeGen] Addressing modes for SVE ldN.

Reviewers: c-rhodes, efriedma, sdesmalen

Subscribers: huihuiz, tschuett, hiraditya, llvm-commits

Tags: #llvm

Differential Revision: https://reviews.llvm.org/D77251

4 years ago[flang] Allow omission of comma in FORMAT(1PE5.2) in runtime
peter klausler [Wed, 22 Jul 2020 23:56:37 +0000 (16:56 -0700)]
[flang] Allow omission of comma in FORMAT(1PE5.2) in runtime

A comma is not required between a scale factor and a following
data edit descriptor (C1302).

Reviewed By: PeteSteinfeld

Differential Revision: https://reviews.llvm.org/D84369

4 years ago[DFSan] Remove unused DataFlowSanitizer vars
Arthur Eubanks [Mon, 27 Jul 2020 21:58:42 +0000 (14:58 -0700)]
[DFSan] Remove unused DataFlowSanitizer vars

Reviewed By: morehouse

Differential Revision: https://reviews.llvm.org/D84704

4 years ago[lldb] NFC: Use early exit in ArchSpec::IsEqualTo
Fred Riss [Mon, 27 Jul 2020 20:51:07 +0000 (13:51 -0700)]
[lldb] NFC: Use early exit in ArchSpec::IsEqualTo

4 years agoFix the move constructor of MMI to move MachineFunctions map
Sridhar Gopinath [Fri, 24 Jul 2020 17:44:48 +0000 (10:44 -0700)]
Fix the move constructor of MMI to move MachineFunctions map

The move constructor of MachineModuleInfo currently does not copy the
MachineFunctions map. This commit fixes this issue.

Patch by Sridhar Gopinath. Thanks!

Differential Revision: https://reviews.llvm.org/D84274

4 years agoRevert "[PowerPC] Remove QPX/A2Q BGQ/BGP CNK support"
Jinsong Ji [Mon, 27 Jul 2020 20:48:44 +0000 (20:48 +0000)]
Revert "[PowerPC] Remove QPX/A2Q BGQ/BGP CNK support"

This reverts commit adffce71538e219aab4eeb024819baa7687262ff.

This is breaking test-suite, revert while investigation.

4 years ago[NFC] Move BitcodeCommon.h from Bitstream to Bitcode
Guillaume Chatelet [Mon, 27 Jul 2020 20:49:01 +0000 (20:49 +0000)]
[NFC] Move BitcodeCommon.h from Bitstream to Bitcode

4 years agoReland "[compiler-rt] [test] Allow expanding lit substitutions recursively"
Sergej Jaskiewicz [Tue, 14 Jul 2020 08:37:27 +0000 (11:37 +0300)]
Reland "[compiler-rt] [test] Allow expanding lit substitutions recursively"

The commit 8372d505082aceb38417e0b561cd32f2e227597b has been reverted
(eafeb8af34946306a7382fa3801cf6e39a1c7226) because it broke asan
tests on green dragon buildbots.

The underlying issue has been fixed in 4dd5c2bee366514cbc3fc4e6da46462bc11a0a3d.

4 years agoRename t2-reduce-size -> thumb2-reduce-size
Arthur Eubanks [Mon, 27 Jul 2020 20:42:13 +0000 (13:42 -0700)]
Rename t2-reduce-size -> thumb2-reduce-size

For readability and consistency with other thumb2 passes like
"thumb2-it".

Reviewed By: arsenm

Differential Revision: https://reviews.llvm.org/D84696

4 years ago[OpenMPOpt] Most SCC's are uninteresting, don't waste time on them (up to 16x faster)
Roman Lebedev [Mon, 27 Jul 2020 20:35:51 +0000 (23:35 +0300)]
[OpenMPOpt] Most SCC's are uninteresting, don't waste time on them (up to 16x faster)

Summary:
This seems obvious in hindsight, but the result is surprising.
I've measured compile-time of `-openmpopt` pass standalone
on RawSpeed unity build, and while there is some OpenMP stuff,
most is not OpenMP. But nonetheless the pass does a lot of costly
preparations before ever trying to look for OpenMP stuff in SCC.

Numbers (n=25): 0.094624s  ->  0.005976s, an -93.68% improvement, or ~16x

Reviewers: jdoerfert

Reviewed By: jdoerfert

Subscribers: yaxunl, hiraditya, guansong, llvm-commits, sstefan1

Tags: #llvm

Differential Revision: https://reviews.llvm.org/D84689

4 years agoUnify the return value of GetByteSize to an llvm::Optional<uint64_t> (NFC-ish)
Adrian Prantl [Sat, 25 Jul 2020 15:27:21 +0000 (08:27 -0700)]
Unify the return value of GetByteSize to an llvm::Optional<uint64_t> (NFC-ish)

This cleanup patch unifies all methods called GetByteSize() in the
ValueObject hierarchy to return an optional, like the methods in
CompilerType do. This means fewer magic 0 values, which could fix bugs
down the road in languages where types can have a size of zero, such
as Swift and C (but not C++).

Differential Revision: https://reviews.llvm.org/D84285

This re-lands the patch with bogus :m_byte_size(0) initalizations removed.

4 years ago[libunwind] Remove old keymgr related logic
Louis Dionne [Fri, 24 Jul 2020 14:13:53 +0000 (10:13 -0400)]
[libunwind] Remove old keymgr related logic

keymgr used to be used on MacOSX <= 10.6, however we don't build libunwind
from scratch for such old systems anymore. Hence, this code isn't useful
anymore.

Differential Revision: https://reviews.llvm.org/D84677

4 years ago[OpenMP] Add more pass-through functions in DeviceTy
Ye Luo [Mon, 27 Jul 2020 20:08:19 +0000 (16:08 -0400)]
[OpenMP] Add more pass-through functions in DeviceTy

Summary:
1. Add DeviceTy::data_alloc, DeviceTy::data_delete, DeviceTy::data_alloc, DeviceTy::synchronize pass-through functions. Avoid directly accessing Device.RTL
2. Fix the type of the first argument of synchronize_ty in rth.h, device id is int32_t which is consistent with other functions.

Reviewers: tianshilei1992, jdoerfert

Reviewed By: tianshilei1992

Subscribers: yaxunl, guansong, sstefan1, openmp-commits

Tags: #openmp

Differential Revision: https://reviews.llvm.org/D84487

4 years ago[lldb/AppleSimulator] Always provide a -simulator environment
Fred Riss [Fri, 24 Jul 2020 16:24:41 +0000 (09:24 -0700)]
[lldb/AppleSimulator] Always provide a -simulator environment

Summary:
This commit is somewhat NFC-ish today as the environment of triples
is not considered when comparing s if one of them is
not set (I plan to change that).

We have made simulator triples unambiguous these days, but the
simulator platforms still advertise triples without the
environment. This wasn't an issue when the sims ran only on
a very different architecure than the real device, but this
has changed with Apple Silicon.

This patch simplifies the way GetSupportedArchitectureAtIndex
is implemented for the sim platforms and adds the environment.
It also trivially adds support for Apple Silicon to those
platforms.

Reviewers: aprantl

Subscribers: lldb-commits

4 years ago[llvm][examples][SimplifyCFG] Fix pass's IR changed reporting
Jon Roelofs [Mon, 27 Jul 2020 19:37:35 +0000 (13:37 -0600)]
[llvm][examples][SimplifyCFG] Fix pass's IR changed reporting

... under the EXPENSIVE_CHECKS build, this fails the assert in the LegacyPM
that verifies whether a pass really did leave the IR alone when it reports no
changes back from its return status.

4 years ago[lldb] Remove CMAKE_VERSION checks now that the minimum version is 3.13.4
Jonas Devlieghere [Mon, 27 Jul 2020 19:30:09 +0000 (12:30 -0700)]
[lldb] Remove CMAKE_VERSION checks now that the minimum version is 3.13.4

4 years ago[PowerPC] Remove QPX/A2Q BGQ/BGP CNK support
Jinsong Ji [Mon, 27 Jul 2020 18:01:40 +0000 (18:01 +0000)]
[PowerPC] Remove QPX/A2Q BGQ/BGP CNK support

Per RFC http://lists.llvm.org/pipermail/llvm-dev/2020-April/141295.html
no one is making use of QPX/A2Q/BGQ/BGP CNK anymore.

This patch remove the support of QPX/A2Q in llvm, BGQ/BGP in clang,
CNK support in openmp/polly.

Reviewed By: hfinkel

Differential Revision: https://reviews.llvm.org/D83915

4 years ago[mlir][AffineToStandard] Make LowerAffine pass Op-agnostic.
MaheshRavishankar [Mon, 27 Jul 2020 19:13:08 +0000 (12:13 -0700)]
[mlir][AffineToStandard] Make LowerAffine pass Op-agnostic.

The LowerAffine psas was a FunctionPass only for legacy
reasons. Making this Op-agnostic allows it to be used from command
line when affine expressions are within operations other than
`std.func`.

Differential Revision: https://reviews.llvm.org/D84590

4 years ago[AIX] Implement AIX special alignment rule about double/long double
Xiangling Liao [Tue, 2 Jun 2020 15:51:58 +0000 (11:51 -0400)]
[AIX] Implement AIX special alignment rule about double/long double

Implement AIX default `power` alignment rule by adding `PreferredAlignment` and
`PreferredNVAlignment` in ASTRecordLayout class.

The patchh aims at returning correct value for `__alignof(x)` and `alignof(x)`
under `power` alignment rules.

Differential Revision: https://reviews.llvm.org/D79719

4 years ago[Attributor][NFC] Add tests to noalias.ll
Shinji Okumura [Mon, 27 Jul 2020 18:50:43 +0000 (03:50 +0900)]
[Attributor][NFC] Add tests to noalias.ll

Summary: Add tests to `noalias.ll` to make changes in D84665 clear

Reviewers: jdoerfert, sstefan1, uenoku, homerdin, baziotis

Subscribers: uenoku, kuter, bbn, llvm-commits

Tags: #llvm

Differential Revision: https://reviews.llvm.org/D84688

4 years ago[clangd] Add more logs and attach tracers to remote index server routines
Kirill Bobyrev [Mon, 27 Jul 2020 18:45:05 +0000 (20:45 +0200)]
[clangd] Add more logs and attach tracers to remote index server routines

Reviewers: kadircet

Reviewed By: kadircet

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

Tags: #clang

Differential Revision: https://reviews.llvm.org/D84499

4 years agoReplace fuzzer::FuzzerDriver's INTERFACE marking with new LLVMRunFuzzerDriver.
Matt Morehouse [Mon, 27 Jul 2020 18:15:35 +0000 (18:15 +0000)]
Replace fuzzer::FuzzerDriver's INTERFACE marking with new LLVMRunFuzzerDriver.

This adds a new extern "C" function that serves the same purpose. This removes the need for external users to depend on internal headers in order to use this feature. It also standardizes the interface in a way that other fuzzing engines will be able to match.

Patch By: IanPudney

Reviewed By: kcc

Differential Revision: https://reviews.llvm.org/D84561

4 years agoRecommit "[libFuzzer] Disable implicit builtin knowledge about memcmp-like functions...
Dokyung Song [Thu, 16 Jul 2020 21:24:06 +0000 (21:24 +0000)]
Recommit "[libFuzzer] Disable implicit builtin knowledge about memcmp-like functions when -fsanitize=fuzzer-no-link is given."

Summary: This patch disables implicit builtin knowledge about memcmp-like functions when compiling the program for fuzzing, i.e., when -fsanitize=fuzzer(-no-link) is given. This allows libFuzzer to always intercept memcmp-like functions as it effectively disables optimizing calls to such functions into different forms. This is done by adding a set of flags (-fno-builtin-memcmp and others) in the clang driver. Individual -fno-builtin-* flags previously used in several libFuzzer tests are now removed, as it is now done automatically in the clang driver.

The patch was once reverted in 8ef9e2bf355d05bc81d8b0fe1e5333eec59a0a91, as this patch was dependent on a reverted commit f78d9fceea736d431e9e3cbca291e3909e3aa46d. This reverted commit was recommitted in 831ae45e3dc609e43ba561af07670a8fe47461ef, so relanding this dependent patch too.

Reviewers: morehouse, hctim

Subscribers: cfe-commits, #sanitizers

Tags: #clang, #sanitizers

Differential Revision: https://reviews.llvm.org/D83987

4 years ago[mlir] [VectorOps] [integration_test] Sparse matrix times vector (DOT version)
aartbik [Mon, 27 Jul 2020 18:22:26 +0000 (11:22 -0700)]
[mlir] [VectorOps] [integration_test] Sparse matrix times vector (DOT version)

Integration test that illustrates the gather operation with a
real-world operation expressed in mostly the Vector dialect.
Uses jagged diagonal storage.

Reviewed By: bondhugula

Differential Revision: https://reviews.llvm.org/D84571

4 years agoPrefix some AArch64/ARM passes with "aarch64-"/"arm-"
Arthur Eubanks [Mon, 27 Jul 2020 17:49:16 +0000 (10:49 -0700)]
Prefix some AArch64/ARM passes with "aarch64-"/"arm-"

 For consistency with other target specific passes.

Reviewed By: arsenm

Differential Revision: https://reviews.llvm.org/D84560

4 years ago[GWP-ASan] Crash handler API returns sizeof(collected trace)
Mitch Phillips [Mon, 27 Jul 2020 17:51:53 +0000 (10:51 -0700)]
[GWP-ASan] Crash handler API returns sizeof(collected trace)

Summary:
Fix up a slight bug with the crash handler API, where we say that we
return the size of the collected trace (instead of the size of the trace
that's returned) when the return buffer is too small, and the result is
truncated.

Also, as a result, patch up a small uninitialized memory bug.

Reviewers: morehouse, eugenis

Reviewed By: eugenis

Subscribers: #sanitizers

Tags: #sanitizers

Differential Revision: https://reviews.llvm.org/D84652

4 years agoUse llvm::is_contained where appropriate (NFC)
Kazu Hirata [Fri, 24 Jul 2020 06:13:44 +0000 (23:13 -0700)]
Use llvm::is_contained where appropriate (NFC)

Summary:
This patch replaces std::find with llvm::is_contained where
appropriate.

Reviewers: efriedma, nhaehnle

Reviewed By: nhaehnle

Subscribers: arsenm, jvesely, nhaehnle, hiraditya, rogfer01, kerbowa, llvm-commits, vkmr

Tags: #llvm

Differential Revision: https://reviews.llvm.org/D84489

4 years ago[StackProtector] Speed up RequiresStackProtector
Nadav Rotem [Mon, 27 Jul 2020 17:06:36 +0000 (10:06 -0700)]
[StackProtector] Speed up RequiresStackProtector

Speed up the method RequiresStackProtector by checking the intrinsic
value of the call. The original code calls getName() that returns an
allocating std::string on each check. This change removes about 96072
std::string instances when compiling sqlite3.c; The function was
discovered with a Facebook-internal performance tool.

Differential Revision: https://reviews.llvm.org/D84620

4 years ago[X86] Add back comment inadvertently lost in 1a1448e6568d9b11f198e510fa9c4cb6b1f4216a.
Craig Topper [Mon, 27 Jul 2020 16:40:26 +0000 (09:40 -0700)]
[X86] Add back comment inadvertently lost in 1a1448e6568d9b11f198e510fa9c4cb6b1f4216a.

4 years ago[flang] Temp Driver - pass the flag to change the default integer kind through to...
AlexisPerry [Mon, 27 Jul 2020 16:57:31 +0000 (09:57 -0700)]
[flang] Temp Driver - pass the flag to change the default integer kind through to F18_FC

fixes BUG 46307

Differential Revision: https://reviews.llvm.org/D84266

4 years ago[NFC][AArch64] Replace some template methods/invocations...
Francesco Petrogalli [Mon, 27 Jul 2020 16:24:18 +0000 (16:24 +0000)]
[NFC][AArch64] Replace some template methods/invocations...

...with the non-template version, as the template version might
increase the size of the compiler build.

Methods affected:

1.`findAddrModeSVELoadStore`
2. `SelectPredicatedStore`

Also, remove the `const` qualifier from the `unsigned` parameters of
the methods to conform with other similar methods in the class.

4 years ago[mlir][Linalg] Enable fusion of std.constant (producer) with
MaheshRavishankar [Mon, 27 Jul 2020 16:48:36 +0000 (09:48 -0700)]
[mlir][Linalg] Enable fusion of std.constant (producer) with
linalg.indexed_generic (consumer) with tensor arguments.

The implementation of fusing std.constant producer with a
linalg.indexed_generic consumer was already in place. It is exposed
with this change. Also cleaning up some of the patterns that implement
the fusion to not be templated, thereby avoiding lot of conditional
checks for calling the right instantiation.

Differential Revision: https://reviews.llvm.org/D84566

4 years ago[X86][SSE] Relax 128-bit restriction on extract_subvector(ext_vector_inreg(X),0)...
Simon Pilgrim [Mon, 27 Jul 2020 16:32:53 +0000 (17:32 +0100)]
[X86][SSE] Relax 128-bit restriction on extract_subvector(ext_vector_inreg(X),0) -> ext_vector_inreg(extract_subvector(X,0)) fold

We only need to ensure that the source is larger than the subvector result type

4 years ago[gn build] Port ee7caa75939
LLVM GN Syncbot [Mon, 27 Jul 2020 16:45:49 +0000 (16:45 +0000)]
[gn build] Port ee7caa75939

4 years ago[gn build] Make syncing to single-line source files work after 1afd889d0b43
Nico Weber [Mon, 27 Jul 2020 16:44:47 +0000 (12:44 -0400)]
[gn build] Make syncing to single-line source files work after 1afd889d0b43

4 years agoReland [llvm-exegesis] Add benchmark latency option on X86 that uses LBR for more...
Vy Nguyen [Mon, 27 Jul 2020 16:38:05 +0000 (12:38 -0400)]
Reland [llvm-exegesis] Add benchmark latency option on X86 that uses LBR for more precise measurements.

        Starting with Skylake, the LBR contains the precise number of cycles between the two
        consecutive branches.
        Making use of this will hopefully make the measurements more precise than the
        existing methods of using RDTSC.

                Differential Revision: https://reviews.llvm.org/D77422

New change: check for existence of field `cycles` in perf_branch_entry before enabling this mode.
This should prevent compilation errors when building for older kernel whose headers don't support it.

4 years ago[gn build] (manually) merge 48c948abeb7
Nico Weber [Mon, 27 Jul 2020 16:30:09 +0000 (12:30 -0400)]
[gn build] (manually) merge 48c948abeb7

4 years ago[TableGen][CGS] Print better errors on overlapping InstRW
Jon Roelofs [Wed, 22 Jul 2020 22:13:20 +0000 (16:13 -0600)]
[TableGen][CGS] Print better errors on overlapping InstRW

Differential Revision: https://reviews.llvm.org/D83588

4 years agoUse INTERFACE_COMPILE_OPTIONS to disable -Wsuggest-override for any target that links...
Logan Smith [Mon, 27 Jul 2020 15:37:01 +0000 (08:37 -0700)]
Use INTERFACE_COMPILE_OPTIONS to disable -Wsuggest-override for any target that links to gtest

This cleans up several CMakeLists.txt's where -Wno-suggest-override was manually specified. These test targets now inherit this flag from the gtest target.

Some unittests CMakeLists.txt's, in particular Flang and LLDB, are not touched by this patch. Flang manually adds the gtest sources itself in some configurations, rather than linking to LLVM's gtest target, so this fix would be insufficient to cover those cases. Similarly, LLDB has subdirectories that manually add the gtest headers to their include path without linking to the gtest target, so those subdirectories still need -Wno-suggest-override to be manually specified to compile without warnings.

Differential Revision: https://reviews.llvm.org/D84554

4 years ago[llvm-exegesis] Check perf_branch_entry for field cycles
Vy Nguyen [Fri, 17 Jul 2020 23:08:48 +0000 (19:08 -0400)]
[llvm-exegesis] Check perf_branch_entry for field cycles

Summary: Follow up to breakages reported in D77422

Reviewers: ondrasej, gchatelet

Tags: #llvm

Differential Revision: https://reviews.llvm.org/D84076

4 years ago[XCOFF][AIX] Handle llvm.used and llvm.compiler.used global array
jasonliu [Mon, 27 Jul 2020 14:04:59 +0000 (14:04 +0000)]
[XCOFF][AIX] Handle llvm.used and llvm.compiler.used global array

For now, just return and do nothing when we see llvm.used and
llvm.compiler.used global array.
Hopefully, we could come up with a good solution later to prevent
linker from eliminating symbols in llvm.used array.

Reviewed By: DiggerLin, daltenty

Differential Revision: https://reviews.llvm.org/D84363

4 years ago[AArch64] fjcvtzs,rmif,cfinv,setf* all clobber nzcv
Jon Roelofs [Tue, 14 Jul 2020 21:45:05 +0000 (15:45 -0600)]
[AArch64] fjcvtzs,rmif,cfinv,setf* all clobber nzcv

Differential Revision: https://reviews.llvm.org/D83818

4 years ago[lit] Don't expand escapes until all substitutions have been applied
Sergej Jaskiewicz [Wed, 15 Jul 2020 15:55:37 +0000 (18:55 +0300)]
[lit] Don't expand escapes until all substitutions have been applied

Otherwise, if a Lit script contains escaped substitutions (like %%p in this test https://github.com/llvm/llvm-project/blob/master/compiler-rt/test/asan/TestCases/Darwin/asan-symbolize-partial-report-with-module-map.cpp#L10), they are unescaped during recursive application of substitutions, and the results are unexpected.

We solve it using the fact that double percent signs are first replaced with #_MARKER_#, and only after all the other substitutions have been applied, #_MARKER_# is replaced with a single percent sign. The only change is that instead of replacing #_MARKER_# at each recursion step, we replace it once after the last recursion step.

Differential Revision: https://reviews.llvm.org/D83894

4 years agoRevert "Re-apply:" Emit DW_OP_implicit_value for Floating point constants""
Amy Kwan [Mon, 27 Jul 2020 14:36:38 +0000 (09:36 -0500)]
Revert "Re-apply:" Emit DW_OP_implicit_value for Floating point constants""

This patch reverts commit `59a76d957a26` as it has caused failure on the
big endian PowerPC buildbots (as well as the SystemZ buildbots).

4 years ago[LLD] [COFF] Fix mingw comdat associativity for leader symbols with a different name
Martin Storsjö [Sat, 25 Jul 2020 09:25:19 +0000 (12:25 +0300)]
[LLD] [COFF] Fix mingw comdat associativity for leader symbols with a different name

For a weak symbol func in a comdat, the actual leader symbol ends up
named like .weak.func.default*. Likewise, for stdcall on i386, the symbol
may be named _func@4, while the section suffix only is "func", which the
previous implementation didn't handle.

This fixes unwinding through weak functions when using
-ffunction-sections in mingw environments.

Differential Revision: https://reviews.llvm.org/D84607

4 years ago[Attributor] Fix qualifier warning of the unittest
Luofan Chen [Mon, 27 Jul 2020 14:28:39 +0000 (22:28 +0800)]
[Attributor] Fix qualifier warning of the unittest

Reviewed By: jdoerfert

Differential Revision: https://reviews.llvm.org/D84532

4 years ago[OpenCL] Fixed missing address space for templated copy constructor.
Anastasia Stulova [Mon, 27 Jul 2020 12:27:21 +0000 (13:27 +0100)]
[OpenCL] Fixed missing address space for templated copy constructor.

Added missing address space for the parameter of copy ctor created
for templated constructor with an R-value reference.

Patch by Ole Strohm (olestrohm)!

Tags: #clang

Differential Revision: https://reviews.llvm.org/D83665

4 years ago[NFC] Replace ".size() < 1" with ".empty()"
Guillaume Chatelet [Mon, 27 Jul 2020 13:54:46 +0000 (13:54 +0000)]
[NFC] Replace ".size() < 1" with ".empty()"

4 years ago[X86][AVX] Fold extract_subvector(truncate(x),0) -> truncate(extract_subvector(x),0)
Simon Pilgrim [Mon, 27 Jul 2020 13:32:32 +0000 (14:32 +0100)]
[X86][AVX] Fold extract_subvector(truncate(x),0) -> truncate(extract_subvector(x),0)

This is currently only supported for VLX targets where the op should be legal.