Jim Ingham [Tue, 12 Mar 2019 19:27:39 +0000 (19:27 +0000)]
Check the result of creating a node from __next_ in the std::list formatter.
There's a single report of a crash coming from this current_sp being NULL. I don't
have a repro case, and I couldn't get it to happen by hand-corrupting a list. We
always get an error instead. So I don't have a test case. But checking for null
is clearly right here.
<rdar://problem/
48503320>
llvm-svn: 355957
Jim Ingham [Tue, 12 Mar 2019 19:25:29 +0000 (19:25 +0000)]
Re-enable this test, the underlying bug was fixed and the test now passes.
llvm-svn: 355956
Sanjay Patel [Tue, 12 Mar 2019 19:20:45 +0000 (19:20 +0000)]
[x86] scalarize extractelement 0 of FP vselect
llvm-svn: 355955
Peter Collingbourne [Tue, 12 Mar 2019 19:19:23 +0000 (19:19 +0000)]
ELF: Simplify. NFCI.
We don't need to take a slice of SectionCommands in addOrphanSections()
because it is not modified until the end of the function.
Differential Revision: https://reviews.llvm.org/D59239
llvm-svn: 355954
Philip Reames [Tue, 12 Mar 2019 19:12:33 +0000 (19:12 +0000)]
[CodeGen] Add MMOs to statepoint nodes during SelectionDAG
The existing statepoint lowering code does something odd; it adds machine memory operands post instruction selection. This was copied from the stackmap/patchpoint implementation, but appears to be non-idiomatic.
This change is largely NFC. It moves the MMO creation logic into SelectionDAG building. It ends up not quite being NFC because the size of the stack slot is reflected in the MMO. The old code blindly used pointer size for the MMO size, which appears to have always been incorrect for larger values. It just happened nothing actually relied on the MMOs, so it worked out okay.
For context, I'm planning on removing the MOVolatile flag from these in a future commit, and then removing the MOStore flag from deopt spill slots in a separate one. Doing so is motivated by a small test case where we should be able to better schedule spill slots, but don't do so due to a memory use/def implied by the statepoint.
Differential Revision: https://reviews.llvm.org/D59106
llvm-svn: 355953
Alexey Bataev [Tue, 12 Mar 2019 18:52:33 +0000 (18:52 +0000)]
[OPENMP 5.0]Initial support for 'allocator' clause.
Added parsing/sema analysis/serialization/deserialization for the
'allocator' clause of the 'allocate' directive.
llvm-svn: 355952
Jim Ingham [Tue, 12 Mar 2019 18:48:58 +0000 (18:48 +0000)]
Fix the project for r355939 (ASTUtils.{h,c})
llvm-svn: 355951
Duncan P. N. Exon Smith [Tue, 12 Mar 2019 18:38:04 +0000 (18:38 +0000)]
Modules: Add LangOptions::CacheGeneratedPCH
Add an option to cache the generated PCH in the ModuleCache when
emitting it. This protects clients that build PCHs and read them in the
same process, allowing them to avoid race conditions between parallel
jobs the same way that Clang's implicit module build system does.
rdar://problem/
48740787
llvm-svn: 355950
Teresa Johnson [Tue, 12 Mar 2019 18:28:05 +0000 (18:28 +0000)]
[SCEV] Use depth limit for trunc analysis
Summary:
This fixes an extremely long compile time caused by recursive analysis
of truncs, which were not previously subject to any depth limits unlike
some of the other ops. I decided to use the same control used for
sext/zext, since the routines analyzing these are sometimes mutually
recursive with the trunc analysis.
Reviewers: mkazantsev, sanjoy
Subscribers: sanjoy, jdoerfert, llvm-commits
Tags: #llvm
Differential Revision: https://reviews.llvm.org/D58994
llvm-svn: 355949
Jinsong Ji [Tue, 12 Mar 2019 18:27:09 +0000 (18:27 +0000)]
Set useful flags for vector imm setting instructions
Vector imm setting instructions like XXLXORz/XXLXORspz/XXLXORdpz
Should behave like LI8.
We should set corresponding flags to allow rematerialization and other
opts in LICM, RA, Scheduling etc.
Differential Revision: https://reviews.llvm.org/D58645
llvm-svn: 355948
Craig Topper [Tue, 12 Mar 2019 18:20:25 +0000 (18:20 +0000)]
[SanitizerCoverage] Avoid splitting critical edges when destination is a basic block containing unreachable
This patch adds a new option to SplitAllCriticalEdges and uses it to avoid splitting critical edges when the destination basic block ends with unreachable. Otherwise if we split the critical edge, sanitizer coverage will instrument the new block that gets inserted for the split. But since this block itself shouldn't be reachable this is pointless. These basic blocks will stick around and generate assembly, but they don't end in sane control flow and might get placed at the end of the function. This makes it look like one function has code that flows into the next function.
This showed up while compiling the linux kernel with clang. The kernel has a tool called objtool that detected the code that appeared to flow from one function to the next. https://github.com/ClangBuiltLinux/linux/issues/351#issuecomment-
461698884
Differential Revision: https://reviews.llvm.org/D57982
llvm-svn: 355947
Eli Friedman [Tue, 12 Mar 2019 18:14:16 +0000 (18:14 +0000)]
[RISCV][MC] Find matching pcrel_hi fixup in more cases.
If a symbol points to the end of a fragment, instead of searching for
fixups in that fragment, search in the next fragment.
Fixes spurious assembler error with subtarget change next to "la"
pseudo-instruction, or expanded equivalent.
Alternate proposal to fix the problem discussed in
https://reviews.llvm.org/D58759.
Testcase by Ana Pazos.
Differential Revision: https://reviews.llvm.org/D58943
llvm-svn: 355946
Jinsong Ji [Tue, 12 Mar 2019 17:55:32 +0000 (17:55 +0000)]
[NFC][PowerPC] Update testcases using utils/update_llc_test_checks.py
llvm-svn: 355945
George Burgess IV [Tue, 12 Mar 2019 17:48:53 +0000 (17:48 +0000)]
Add a creduce script for clang crashes
This CL adds a script that calls C-Reduce on an input file and given the
clang crash script, which is used to generate an interestingness test
for C-Reduce.
Patch by Amy Huang!
Differential Revision: https://reviews.llvm.org/D59118
llvm-svn: 355944
Adrian McCarthy [Tue, 12 Mar 2019 17:40:51 +0000 (17:40 +0000)]
Correcting some comments in PdbIndex.cpp [NFC]
ICF can cause multiple symbols to start at the same virtual address.
I plan to handle this shortly, but I wanted to correct the comment for
now.
Deleted an obsolete comment about adjusting the offset for the magic
number at the beginning of the debug info stream. This adjustment is
handled at a lower level now.
llvm-svn: 355943
Jason Liu [Tue, 12 Mar 2019 17:33:07 +0000 (17:33 +0000)]
Test commit: add a blank line in test case ppc64-dq-expr.s
llvm-svn: 355942
Jonas Devlieghere [Tue, 12 Mar 2019 17:31:04 +0000 (17:31 +0000)]
[Reproducers] Add a test to ensure we can reuse the reproducer dir.
Yesterday I noticed a reproducer test failing after making a local
change. Removing the reproducer directory solved the issue. Add a test
case that detects this.
llvm-svn: 355941
Jonas Devlieghere [Tue, 12 Mar 2019 17:10:28 +0000 (17:10 +0000)]
[Reproducers] Stop recording instead of deallocating
The command interpreter holds a pointer to a DataRecorder. After
generating the reproducer, we deallocated all the DataRecorders, causing
the command interpreter to hold a non-null reference to an invalid
object.
This patch changes the behavior of the command provider to stop the
DataRecorders when a reproducer is generated, rather than deallocating
them.
llvm-svn: 355940
Raphael Isemann [Tue, 12 Mar 2019 17:09:33 +0000 (17:09 +0000)]
Add ability to import std module into expression parser to improve C++ debugging
Summary:
This patch is the MVP version of importing the std module into the expression parser to improve C++ debugging.
What happens in this patch is that we inject a `@import std` into our expression source code. We also
modify our internal Clang instance for parsing this expression to work with modules and debug info
at the same time (which is the main change in terms of LOC). We implicitly build the `std` module on the first use. The
C++ include paths for building are extracted from the debug info, which means that this currently only
works if the program is compiled with `-glldb -fmodules` and uses the std module. The C include paths
are currently specified by LLDB.
I enabled the tests currently only for libc++ and Linux because I could test this locally. I'll enable the tests
for other platforms once this has landed and doesn't break any bots (and I implemented the platform-specific
C include paths for them).
With this patch we can now:
* Build a libc++ as a module and import it into the expression parser.
* Read from the module while also referencing declarations from the debug info. E.g. `std::abs(local_variable)`.
What doesn't work (yet):
* Merging debug info and C++ module declarations. E.g. `std::vector<CustomClass>` doesn't work.
* Pretty much anything that involves the ASTImporter and templated code. As the ASTImporter is used for saving the result declaration, this means that we can't
call yet any function that returns a non-trivial type.
* Use libstdc++ for this, as it requires multiple include paths and Clang only emits one include path per module. Also libstdc++ doesn't support Clang modules without patches.
Reviewers: aprantl, jingham, shafik, friss, davide, serge-sans-paille
Reviewed By: aprantl
Subscribers: labath, mgorny, abidh, jdoerfert, lldb-commits
Tags: #c_modules_in_lldb, #lldb
Differential Revision: https://reviews.llvm.org/D58125
llvm-svn: 355939
James Henderson [Tue, 12 Mar 2019 17:00:25 +0000 (17:00 +0000)]
[yaml2obj]Allow explicit symbol indexes in relocations and emit error for bad names
Prior to this change, the "Symbol" field of a relocation would always be
assumed to be a symbol name, and if no such symbol existed, the
relocation would reference index 0. This confused me when I tried to use
a literal symbol index in the field: since "0x1" was not a known symbol
name, the symbol index was set as 0. This change falls back to treating
unknown symbol names as integers, and emits an error if the name is not
found and the string is not an integer.
Note that the Symbol field is optional, so if a relocation doesn't
reference a symbol, it shouldn't be specified. The new error required a
number of test updates.
Reviewed by: grimar, ruiu
Differential Revision: https://reviews.llvm.org/D58510
llvm-svn: 355938
Nikita Popov [Tue, 12 Mar 2019 16:57:25 +0000 (16:57 +0000)]
[SDAG] Expand pow2 mulo using shifts
Expand MULO with constant power of two operand into a shift. The
overflow is checked with (x << shift) >> shift == x, where the right
shift will be logical for umulo and arithmetic for smulo (with
exception for multiplications by signed_min).
Differential Revision: https://reviews.llvm.org/D59041
llvm-svn: 355937
Jonas Devlieghere [Tue, 12 Mar 2019 16:44:18 +0000 (16:44 +0000)]
[Reproducers] Support capturing a reproducer without an explicit path.
Tablegen doesn't support options that are both flags and take values as
an argument. I noticed this when doing the tablegen rewrite, but forgot
that that affected the reproducer --capture flag.
This patch makes --capture a flag and adds --capture-path to specify a
path for the reproducer. In reality I expect this to be mostly used for
testing, but it could be useful nonetheless.
Differential revision: https://reviews.llvm.org/D59238
llvm-svn: 355936
Craig Topper [Tue, 12 Mar 2019 16:35:30 +0000 (16:35 +0000)]
[X86] Arrange more CPU features to inherit from earlier CPUs. NFCI
This makes SandyBridge inherit back to Westmere/Nehalem.
Make bdver1-4 inherit from each other and btver2 inherit from btver1.
llvm-svn: 355935
Haojian Wu [Tue, 12 Mar 2019 16:11:46 +0000 (16:11 +0000)]
[clang-tidy] NOLINT support for "clang-diagnostic-*".
Reviewers: alexfh, aaron.ballman
Reviewed By: alexfh
Subscribers: xazax.hun, cfe-commits
Tags: #clang
Differential Revision: https://reviews.llvm.org/D59255
llvm-svn: 355934
Simon Pilgrim [Tue, 12 Mar 2019 16:00:59 +0000 (16:00 +0000)]
Regenerate sign_extend.ll test.
This will change as part of the fix for the regressions in D58017.
llvm-svn: 355933
Simon Pilgrim [Tue, 12 Mar 2019 15:58:28 +0000 (15:58 +0000)]
[DAGCombine] Pull out repeated demanded bitmask generation. NFCI.
llvm-svn: 355932
Stefan Granitz [Tue, 12 Mar 2019 15:54:35 +0000 (15:54 +0000)]
Revert "[CMake] Avoid clang-tablegen-targets dependency when building sphinx docs (experimental)"
This reverts commit
511066858d44101703d61eded9abf8caff0f9fe0.
This turned out unnecessary to fix the bot.
llvm-svn: 355931
Kristina Brooks [Tue, 12 Mar 2019 15:44:18 +0000 (15:44 +0000)]
[Docs] Add note about legacy PM to Ch4 of tutorial
Add a note about legacy FunctionPassManager to the LLVM tutorial.
It seems to confuse some people, worth adding a warning to the tutorial
to elaborate and suggest using `llvm::legacy::FunctionPassManager` for
now. Not a perfect solution but hopefully will avoid confusion
in the meantime.
Differential Revision: https://reviews.llvm.org/D59258
llvm-svn: 355930
James Henderson [Tue, 12 Mar 2019 15:42:38 +0000 (15:42 +0000)]
[llvm-cxxfilt]Add test to show that empty lines can be handled
I recently discovered a bug in llvm-cxxfilt introduced in r353743 but
was fixed later incidentally due to r355031. Specifically, llvm-cxxfilt
was attempting to call .back() on an empty string any time there was a
new line in the input. This was causing a crash in my debug builds only.
This patch simply adds a test that explicitly tests that llvm-cxxfilt
handles empty lines correctly. It may pass under release builds under
the broken behaviour, but it fails at least in debug builds.
Reviewed by: mattd
Differential Revision: https://reviews.llvm.org/D58785
llvm-svn: 355929
James Henderson [Tue, 12 Mar 2019 15:37:34 +0000 (15:37 +0000)]
[FileCheck]Remove assertions that prevent matching an empty string at file start before CHECK-NEXT/SAME
This patch removes two assertions that were preventing writing of a test
that checked an empty line followed by some text. For example:
CHECK: {{^$}}
CHECK-NEXT: foo()
The assertion was because the current location the CHECK-NEXT was
scanning from was the start of the buffer. A similar issue occurred with
CHECK-SAME. These assertions don't protect against anything, as there is
already an error check that checks that CHECK-NEXT/EMPTY/SAME don't
appear first in the checks, and the following code works fine if the
pointer is at the start of the input.
Reviewed by: probinson, thopre, jdenny
Differential Revision: https://reviews.llvm.org/D58784
llvm-svn: 355928
Louis Dionne [Tue, 12 Mar 2019 15:32:00 +0000 (15:32 +0000)]
[CMake] Tell libc++ that we're using compiler-rt on Apple platforms
Reviewers: beanz, arphaman, EricWF
Subscribers: dberris, mgorny, jkorous, dexonsmith, jdoerfert, cfe-commits
Tags: #clang
Differential Revision: https://reviews.llvm.org/D58921
llvm-svn: 355927
Tim Northover [Tue, 12 Mar 2019 15:22:23 +0000 (15:22 +0000)]
CodeGenPrep: preserve inbounds attribute when sinking GEPs.
Targets can potentially emit more efficient code if they know address
computations never overflow. For example ILP32 code on AArch64 (which only has
64-bit address computation) can ignore the possibility of overflow with this
extra information.
llvm-svn: 355926
Xing GUO [Tue, 12 Mar 2019 14:58:01 +0000 (14:58 +0000)]
fix break tests after rL355922. NFCI
llvm-svn: 355925
Liang Zou [Tue, 12 Mar 2019 14:48:32 +0000 (14:48 +0000)]
[format] \t => ' '
Summary:
1. \t => ' '
2. test commit access
Reviewers: Higuoxing, liangdzou
Reviewed By: Higuoxing, liangdzou
Subscribers: kristina, llvm-commits
Tags: #llvm
Differential Revision: https://reviews.llvm.org/D59243
llvm-svn: 355924
Kostya Kortchinsky [Tue, 12 Mar 2019 14:46:31 +0000 (14:46 +0000)]
[scudo][standalone] Implement checksumming functions
Summary:
This CL implements the checksumming functions. This departs from the
current Scudo code in one aspect: the software version is no longer
CRC32 but a BSD checksum. This is because the software CRC32 was too
impactful in terms of performances, the BSD checksum has no array
lookup which is better (and saves 1KB of data).
As with the current version, we only flip the CRC compiler flag for
a single compilation unit by default, to allow for a library compiled
with HW CRC32 to work on a system without HW CRC32.
There is some platform & hardware specific code in those files, but
since departs from a mere platform split, it felt right to me to have
it that way.
Reviewers: morehouse, eugenis, vitalybuka, hctim, mcgrathr
Reviewed By: morehouse
Subscribers: mgorny, delcypher, jfb, jdoerfert, #sanitizers, llvm-commits
Tags: #llvm, #sanitizers
Differential Revision: https://reviews.llvm.org/D59116
llvm-svn: 355923
Xing GUO [Tue, 12 Mar 2019 14:30:13 +0000 (14:30 +0000)]
[llvm-readobj] Print symbol version when dumping relocations (PR31564)
Summary: This helps resolve https://bugs.llvm.org/show_bug.cgi?id=31564
Reviewers: jhenderson, grimar
Reviewed By: jhenderson
Subscribers: rupprecht, llvm-commits
Tags: #llvm
Differential Revision: https://reviews.llvm.org/D59175
llvm-svn: 355922
Fangrui Song [Tue, 12 Mar 2019 14:20:22 +0000 (14:20 +0000)]
[SimplifyLibCalls] Simplify optimizePuts
The code might intend to replace puts("") with putchar('\n') even if the
return value is used. It failed because use_empty() was used to guard
the whole block. While returning '\n' (putchar('\n')) is technically
correct (puts is only required to return a nonnegative number on
success), doing this looks weird and there is really little benefit to
optimize puts whose return value is used. So don't do that.
llvm-svn: 355921
Jinsong Ji [Tue, 12 Mar 2019 14:01:29 +0000 (14:01 +0000)]
[NFC][PowerPC]Assert when trying to generate directmove below P8.
This was found when we generated COPY from G8RC to F8RC in
EmitInstrWithCustomInserter without checking proper architecture,
we silently generated mtvsrd, which require P8 and up.
This is a NFC patch to add assert when we call copyPhysReg, in case
someone accidentally generate COPY between G8RC to F8RC for P7 and
below.
llvm-svn: 355920
Louis Dionne [Tue, 12 Mar 2019 13:54:37 +0000 (13:54 +0000)]
[pstl] Deprecate non-CMake based build
All of LLVM builds with CMake, so it doesn't make sense to maintain a
separate set of build files just for pstl.
Thanks to Thomas Rodgers for the patch.
Differential Revision: https://reviews.llvm.org/D59111
llvm-svn: 355919
Louis Dionne [Tue, 12 Mar 2019 13:48:25 +0000 (13:48 +0000)]
[pstl] Properly extract the version number from pstl_config.h
Previously, we'd be performing math on `#define PSTL_VERSION NNN` instead
of just `NNN`. It seems like older CMakes didn't complain, but newer
CMakes do complain because it doesn't make sense.
llvm-svn: 355918
Fangrui Song [Tue, 12 Mar 2019 13:44:42 +0000 (13:44 +0000)]
[XRay][docs] Fix option name
llvm-svn: 355917
Anastasia Stulova [Tue, 12 Mar 2019 12:46:56 +0000 (12:46 +0000)]
[PR41007][OpenCL] Allow printf in C++ mode.
As for OpenCL C, we need to allow using printf and toolchain variadic
functions (prefixed by "__") in C++ mode.
Differential Revision: https://reviews.llvm.org/D59219
llvm-svn: 355915
Eugene Leviant [Tue, 12 Mar 2019 12:41:06 +0000 (12:41 +0000)]
[llvm-objcopy] Remove unneeded checks. NFC
Differential revision: https://reviews.llvm.org/D59081
llvm-svn: 355914
Simon Pilgrim [Tue, 12 Mar 2019 11:51:59 +0000 (11:51 +0000)]
Revert rL355906: [SLP] Remove redundancy of performing operand reordering twice: once in buildTree() and later in vectorizeTree().
This is a refactoring patch that removes the redundancy of performing operand reordering twice, once in buildTree() and later in vectorizeTree().
To achieve this we need to keep track of the operands within the TreeEntry struct while building the tree, and later in vectorizeTree() we are just accessing them from the TreeEntry in the right order.
This patch is the first in a series of patches that will allow for better operand reordering across chains of instructions (e.g., a chain of ADDs), as presented here: https://www.youtube.com/watch?v=gIEn34LvyNo
Patch by: @vporpo (Vasileios Porpodas)
Differential Revision: https://reviews.llvm.org/D59059
........
Reverted due to buildbot failures that I don't have time to track down.
llvm-svn: 355913
Simon Pilgrim [Tue, 12 Mar 2019 11:31:06 +0000 (11:31 +0000)]
Try to fix SLPVectorizer BoUpSLP::BoEdgeInfo::dump visibility on non-debug builds
llvm-svn: 355912
Kristof Umann [Tue, 12 Mar 2019 11:22:30 +0000 (11:22 +0000)]
Revert "[analyzer] Fix function macro crash"
Buildbot breaks when LLVm is compiled with memory sanitizer.
WARNING: MemorySanitizer: use-of-uninitialized-value
#0 0xa3d16d8 in getMacroNameAndPrintExpansion(blahblah)
lib/StaticAnalyzer/Core/PlistDiagnostics.cpp:903:11
llvm-svn: 355911
George Rimar [Tue, 12 Mar 2019 11:10:29 +0000 (11:10 +0000)]
[LLD][ELF] - Show symbols visibility in "undefined symbol..." error messages.
This teaches LLD to report visibility when showing undefined symbol errors
and fixes https://bugs.llvm.org/show_bug.cgi?id=40770.
Differential revision: https://reviews.llvm.org/D58577
llvm-svn: 355909
Sam Parker [Tue, 12 Mar 2019 11:06:15 +0000 (11:06 +0000)]
[ARM][NFC] Delete original smlad tests
Because I don't understand svn.
llvm-svn: 355908
Sam Parker [Tue, 12 Mar 2019 11:01:11 +0000 (11:01 +0000)]
[ARM][NFC] Move smlad tests
Created a test/CodeGen/ARM/ParallelDSP folder.
llvm-svn: 355907
Simon Pilgrim [Tue, 12 Mar 2019 10:51:51 +0000 (10:51 +0000)]
[SLP] Remove redundancy of performing operand reordering twice: once in buildTree() and later in vectorizeTree().
This is a refactoring patch that removes the redundancy of performing operand reordering twice, once in buildTree() and later in vectorizeTree().
To achieve this we need to keep track of the operands within the TreeEntry struct while building the tree, and later in vectorizeTree() we are just accessing them from the TreeEntry in the right order.
This patch is the first in a series of patches that will allow for better operand reordering across chains of instructions (e.g., a chain of ADDs), as presented here: https://www.youtube.com/watch?v=gIEn34LvyNo
Patch by: @vporpo (Vasileios Porpodas)
Differential Revision: https://reviews.llvm.org/D59059
llvm-svn: 355906
Fangrui Song [Tue, 12 Mar 2019 10:31:52 +0000 (10:31 +0000)]
[SimplifyLibCalls] Fix comments about fputs, memchr, and s[n]printf. NFC
llvm-svn: 355905
Eugene Leviant [Tue, 12 Mar 2019 10:10:29 +0000 (10:10 +0000)]
[CGP] Fix UB when GEP is bound to trivial PHINode
Differential revision: https://reviews.llvm.org/D59140
llvm-svn: 355904
Kristof Umann [Tue, 12 Mar 2019 10:03:32 +0000 (10:03 +0000)]
[analyzer] Fix function macro crash
When there is a functor-like macro which is passed as parameter to another
"function" macro then its parameters are not listed at the place of expansion:
#define foo(x) int bar() { return x; }
#define hello(fvar) fvar(0)
hello(foo)
int main() { 1 / bar(); }
Expansion of hello(foo) asserted Clang, because it expected an l_paren token in
the 3rd line after "foo", since it is a function-like token.
Patch by Tibor Brunner!
Differential Revision: https://reviews.llvm.org/D57893
llvm-svn: 355903
David Stuttard [Tue, 12 Mar 2019 09:52:58 +0000 (09:52 +0000)]
[AMDGPU] Add support for immediate operand for S_ENDPGM
Summary:
Add support for immediate operand in S_ENDPGM
Change-Id: I0c56a076a10980f719fb2a8f16407e9c301013f6
Reviewers: alexshap
Subscribers: qcolombet, arsenm, kzhuravl, jvesely, wdng, nhaehnle, yaxunl, tpr, t-tye, eraman, arphaman, Petar.Avramovic, llvm-commits
Tags: #llvm
Differential Revision: https://reviews.llvm.org/D59213
llvm-svn: 355902
Sjoerd Meijer [Tue, 12 Mar 2019 09:48:02 +0000 (09:48 +0000)]
[TTI] Enable analysis of clib functions in getIntrinsicCosts. NFCI.
This is addressing the issue that we're not modeling the cost of clib functions
in TTI::getIntrinsicCosts and thus we're basically addressing this fixme:
// FIXME: This is wrong for libc intrinsics.
To enable analysis of clib functions, we not only need an intrinsic ID and
formal arguments, but also the actual user of that function so that we can e.g.
look at alignment and values of arguments. So, this is the initial plumbing to
pass the user of an intrinsinsic on to getCallCosts, which queries
getIntrinsicCosts.
Differential Revision: https://reviews.llvm.org/D59014
llvm-svn: 355901
Simon Tatham [Tue, 12 Mar 2019 09:28:19 +0000 (09:28 +0000)]
[TableGen] Allow 2^63-1 and 2^63-2 as int literals.
These two values correspond to the 'Empty' and 'Tombstone' special
keys defined by DenseMapInfo<int64_t>, which means that neither one
can be used as a key in DenseMap<int64_t, anything>. Hence, if you try
to use either of those values as an int literal, IntInit::get() fails
an assertion when it tries to insert them into its static cache of
int-literal objects.
Fixed by replacing the DenseMap with a std::map, which doesn't intrude
on the space of legal values of the key type.
Reviewers: nhaehnle, hfinkel, javedabsar, efriedma
Reviewed By: efriedma
Subscribers: fhahn, efriedma, hiraditya, llvm-commits
Tags: #llvm
Differential Revision: https://reviews.llvm.org/D59016
llvm-svn: 355900
Alex Bradbury [Tue, 12 Mar 2019 09:26:53 +0000 (09:26 +0000)]
[RISCV] Add test cases for the lp64 ABI
These are closely modeled on similar tests for the ilp32 ABI. Like those
tests, we group together tests that should be common cross lp64, lp64+lp64f,
and lp64+lp64f+lp64d ABIs.
llvm-svn: 355899
Raphael Isemann [Tue, 12 Mar 2019 07:45:04 +0000 (07:45 +0000)]
Correctly look up declarations in inline namespaces
Summary:
This patch marks the inline namespaces from DWARF as inline and also ensures that looking
up declarations now follows the lookup rules for inline namespaces.
Reviewers: aprantl, shafik, serge-sans-paille
Reviewed By: aprantl
Subscribers: eraman, jdoerfert, lldb-commits
Tags: #c_modules_in_lldb, #lldb
Differential Revision: https://reviews.llvm.org/D59198
llvm-svn: 355897
Raphael Isemann [Tue, 12 Mar 2019 07:40:54 +0000 (07:40 +0000)]
Fix git-llvm crashing when trying to remove directory while cleaning
Summary:
I'm trying to push D59198 but it seems that `git-llvm push` can't handle the fact
that I add a new directory in the patch:
```
> git llvm push -n
Pushing 1 commit:
e7c0a9bd136 Correctly look up declarations in inline namespaces
Traceback (most recent call last):
File "llvm/utils/git-svn//git-llvm", line 431, in <module>
args.func(args)
File "llvm/utils/git-svn//git-llvm", line 385, in cmd_push
clean_svn(svn_root)
File "llvm/utils/git-svn//git-llvm", line 201, in clean_svn
os.remove(os.path.join(svn_repo, filename))
IsADirectoryError: [Errno 21] Is a directory: '.git/llvm-upstream-svn/lldb/trunk/packages/Python/lldbsuite/test/expression_command/inline-namespace'
```
This patch just uses shutil to delete the directory instead of trying to use `os.remove`
which only works for files.
Reviewers: mehdi_amini, jlebar
Reviewed By: jlebar
Subscribers: llvm-commits
Tags: #llvm
Differential Revision: https://reviews.llvm.org/D59236
llvm-svn: 355896
Kristina Brooks [Tue, 12 Mar 2019 07:08:19 +0000 (07:08 +0000)]
Very minor typo. NFC
Typo `we we're` => `we were` in the pass EarlyCSE
Patch by liangdzou (Liang ZOU)
Differential Revision: https://reviews.llvm.org/D59241
llvm-svn: 355895
Eli Friedman [Tue, 12 Mar 2019 02:20:01 +0000 (02:20 +0000)]
Fix test to unconditionally create a GNU-format archive.
On Darwin targets, llvm-ar creates a Darwin format archive by default,
which ld.lld can't read, so it was printing an unexpected error.
llvm-svn: 355894
Peter Collingbourne [Tue, 12 Mar 2019 02:18:03 +0000 (02:18 +0000)]
ELF: Remove dead code. NFCI.
The Live bit is already set to false by SectionBase.
Differential Revision: https://reviews.llvm.org/D59052
llvm-svn: 355893
Peter Collingbourne [Tue, 12 Mar 2019 02:17:01 +0000 (02:17 +0000)]
llvm-objcopy: Remove unused field. NFCI.
Differential Revision: https://reviews.llvm.org/D59126
llvm-svn: 355892
Petr Hosek [Tue, 12 Mar 2019 02:12:48 +0000 (02:12 +0000)]
[Driver] Support object files in addition to static and shared libraries in compiler-rt
This change introduces support for object files in addition to static
and shared libraries which were already supported which requires
changing the type of the argument from boolean to an enum.
Differential Revision: https://reviews.llvm.org/D56044
llvm-svn: 355891
Aaron Smith [Tue, 12 Mar 2019 02:00:39 +0000 (02:00 +0000)]
[DebugInfo] Add test cases for FlagNonTrivial
Summary:
This is a test case to go with D44406 which added FlagNonTrivial to mark that a C++ record is non-trivial to support CodeView debug emission.
While it looks like FlagTypePassByValue can imply triviality and FlagTypePassByReference can imply non-triviality that is not true. Some non-trivial cases use a combination of FlagNonTrivial and FlagTypePassByValue instead of FlagTypePassByReference. See the test cases and D44406 for discussion.
Reviewers: dblaikie, rnk, zturner
Reviewed By: dblaikie
Subscribers: jdoerfert, llvm-commits
Tags: #llvm
Differential Revision: https://reviews.llvm.org/D59010
llvm-svn: 355890
Sanjoy Das [Tue, 12 Mar 2019 01:31:44 +0000 (01:31 +0000)]
Reland "Relax constraints for reduction vectorization"
Change from original commit: move test (that uses an X86 triple) into the X86
subdirectory.
Original description:
Gating vectorizing reductions on *all* fastmath flags seems unnecessary;
`reassoc` should be sufficient.
Reviewers: tvvikram, mkuper, kristof.beyls, sdesmalen, Ayal
Reviewed By: sdesmalen
Subscribers: dcaballe, huntergr, jmolloy, mcrosier, jlebar, bixia, llvm-commits
Tags: #llvm
Differential Revision: https://reviews.llvm.org/D57728
llvm-svn: 355889
Eli Friedman [Tue, 12 Mar 2019 01:24:39 +0000 (01:24 +0000)]
[ELF] Print a better error for an archive containing a non-ELF file.
Hopefully gives a more readable error message for the most obvious
mistake.
Differential Revision: https://reviews.llvm.org/D59170
llvm-svn: 355888
Jonas Devlieghere [Tue, 12 Mar 2019 00:49:10 +0000 (00:49 +0000)]
[CMake] Avoid clang-tablegen-targets dependency when building sphinx docs (experimental)
Proposal to fix bot
http://lab.llvm.org:8011/builders/lldb-sphinx-docs/builds/1564/steps/cmake-configure/logs/stdio
Patch by: Stefan Gränitz
Differential revision: https://reviews.llvm.org/D59232
llvm-svn: 355887
Rui Ueyama [Tue, 12 Mar 2019 00:24:34 +0000 (00:24 +0000)]
Include an archive file name in an error message for a corrupted file.
Differential Revision: https://reviews.llvm.org/D59212
llvm-svn: 355886
Evgeniy Stepanov [Tue, 12 Mar 2019 00:19:09 +0000 (00:19 +0000)]
[msan] Fix BMI2 detection in msan tests.
llvm-svn: 355885
Nathan Lanza [Mon, 11 Mar 2019 23:30:58 +0000 (23:30 +0000)]
Add a case in SymbolFile{Native,}PDB::TranslateLanguage for Swift
Summary: see above
Reviewers: compnerd
Differential Revision: https://reviews.llvm.org/D59230
llvm-svn: 355883
Nathan Lanza [Mon, 11 Mar 2019 23:27:59 +0000 (23:27 +0000)]
Add Swift enumerator value for CodeView::SourceLanguage
Summary:
Swift now generates PDBs for debugging on Windows. llvm and lldb
need a language enumerator value too properly handle the output
emitted by swiftc.
Subscribers: jdoerfert, llvm-commits
Tags: #llvm
Differential Revision: https://reviews.llvm.org/D59231
llvm-svn: 355882
David Blaikie [Mon, 11 Mar 2019 23:10:33 +0000 (23:10 +0000)]
Hexagon RDF: Replace function template (plus explicit specializations) with non-template overloads
For the design in question, overloads seem to be a much simpler and less subtle solution.
This removes ODR issues, and errors of the kind where code that uses the
specialization in question will accidentally and erroneously specialize
the primary template. This only "works" by accident; the program is
ill-formed NDR.
(Found with -Wundefined-func-template.)
Patch by Thomas Köppe!
Differential Revision: https://reviews.llvm.org/D58998
llvm-svn: 355880
Jonas Devlieghere [Mon, 11 Mar 2019 23:09:09 +0000 (23:09 +0000)]
[Reproducers] Make ReproducerInstrumentation a textual header
The RECORD macro is context sensitive because it depends on the
LLDB_GET_INSTRUMENTATION_DATA. This updates the modulemap to mark that
header as textual.
llvm-svn: 355879
Nico Weber [Mon, 11 Mar 2019 23:02:18 +0000 (23:02 +0000)]
lld-link: Only print demangled symbol names by default
This makes lld-link's output a bit more concise. Since most developers can't
read mangled names, this should make the output a bit easier to understand as
well. It also makes lld-link's output consistent with ld.lld's output.
(link.exe prints both demangled and mangled names; lld-link used to match
link.exe output but now no longer does.)
For people working on toolchains, add a `/demangle:no` flag that makes lld-link
print the mangled name instead of the demangled name. (If desired, people could
pipe that through `demumble -b` to get the old behavior of both demangled and
mangled output.)
Differential Revision: https://reviews.llvm.org/D58132
llvm-svn: 355878
Eric Fiselier [Mon, 11 Mar 2019 22:55:21 +0000 (22:55 +0000)]
Allow optional to tolerate being used with a nested class.
When Clang tries to complete a type containing `std::optional` it
considers the `in_place_t` constructor with no arguments which checks
if the value type is default constructible. If the value type is a
nested class type, then this check occurs too early and poisons the
is_default_constructible trait.
This patch makes optional deduce `in_place_t` so we can prevent
this early SFINAE evaluation. Technically this could break people
doing weird things with the in_place_t tag, but that seems less
important than making the nested class case work.
llvm-svn: 355877
Rong Xu [Mon, 11 Mar 2019 22:51:38 +0000 (22:51 +0000)]
[PGO] Add options for context-sensitive PGO
Add lld options for CSPGO (context-sensitive PGO).
Differential Revision: https://reviews.llvm.org/D56675
llvm-svn: 355876
Alex Langford [Mon, 11 Mar 2019 22:49:36 +0000 (22:49 +0000)]
Rewrite comment to be clearer
llvm-svn: 355875
Sanjoy Das [Mon, 11 Mar 2019 22:37:31 +0000 (22:37 +0000)]
Revert "Relax constraints for reduction vectorization"
This reverts commit r355868. Breaks hexagon.
llvm-svn: 355873
Craig Topper [Mon, 11 Mar 2019 22:29:00 +0000 (22:29 +0000)]
[X86] Remove ProcModel and ProcFeatures tablegen classes. Move all feature lists into a ProcessorFeatures class.
ProcFeatures was a class that just concatenated two feature lists together and gave it a name. We used it to inherit features between CPUs.
ProcModel took a two CPU feature lists and concatenated them before deferring to ProcessorModel. This was to allow inherited features and specific features to be passed to each CPU.
Both of these allowed for only very rigid CPU inheritance rules.
With this patch we now store all of the lists we were using for inheritance in one object and do any list oncatenation we want there. Then we just pass whatever list we want from this class into the ProcessorModel class for each CPU.
Hopefully this gives us more flexibility to build up feature lists in whatever ways we think make sense. Perhaps untangling ISA flags and tuning flags.
I've only touched the CPUs that were directly affected by the removal of the ProcModel and ProcFeatures classes. We should move more of the feature lists into ProcessorFeatures.
llvm-svn: 355872
Jessica Paquette [Mon, 11 Mar 2019 22:18:01 +0000 (22:18 +0000)]
Recommit "[GlobalISel][AArch64] Add selection support for G_EXTRACT_VECTOR_ELT"
After r355865, we should be able to safely select G_EXTRACT_VECTOR_ELT without
running into any problematic intrinsics.
Also add a fix for lane copies, which don't support index 0.
llvm-svn: 355871
Evgeniy Stepanov [Mon, 11 Mar 2019 21:50:10 +0000 (21:50 +0000)]
Remove ASan asm instrumentation.
Summary: It is incomplete and has no users AFAIK.
Reviewers: pcc, vitalybuka
Subscribers: srhines, kubamracek, mgorny, krytarowski, eraman, hiraditya, jdoerfert, #sanitizers, llvm-commits, thakis
Tags: #sanitizers, #llvm
Differential Revision: https://reviews.llvm.org/D59154
llvm-svn: 355870
Alex Bradbury [Mon, 11 Mar 2019 21:41:22 +0000 (21:41 +0000)]
[RISCV] Do a sign-extension in a compare-and-swap of 32 bit in RV64A
AtomicCmpSwapWithSuccess is legalised into an AtomicCmpSwap plus a comparison.
This requires an extension of the value which, by default, is a
zero-extension. When we later lower AtomicCmpSwap into a PseudoCmpXchg32 and then expanded in
RISCVExpandPseudoInsts.cpp, the lr.w instruction does a sign-extension.
This mismatch of extensions causes the comparison to fail when the compared
value is negative. This change overrides TargetLowering::getExtendForAtomicOps
for RISC-V so it does a sign-extension instead.
Differential Revision: https://reviews.llvm.org/D58829
Patch by Ferran Pallarès Roca.
llvm-svn: 355869
Sanjoy Das [Mon, 11 Mar 2019 21:36:41 +0000 (21:36 +0000)]
Relax constraints for reduction vectorization
Summary:
Gating vectorizing reductions on *all* fastmath flags seems unnecessary;
`reassoc` should be sufficient.
Reviewers: tvvikram, mkuper, kristof.beyls, sdesmalen, Ayal
Reviewed By: sdesmalen
Subscribers: dcaballe, huntergr, jmolloy, mcrosier, jlebar, bixia, llvm-commits
Tags: #llvm
Differential Revision: https://reviews.llvm.org/D57728
llvm-svn: 355868
Alex Bradbury [Mon, 11 Mar 2019 21:35:26 +0000 (21:35 +0000)]
[RISCV] Allow fp as an alias of s0
The RISC-V Assembly Programmer's Manual defines fp as another alias of x8.
However, our tablegen rules only recognise s0. This patch adds fp as another
alias of x8. GCC also accepts fp.
Differential Revision: https://reviews.llvm.org/D59209
Patch by Ferran Pallarès Roca.
llvm-svn: 355867
Jonas Devlieghere [Mon, 11 Mar 2019 21:32:20 +0000 (21:32 +0000)]
[Reproducers] Reinterpret cast to void*
Apparently the log_append variant added in r355863 is considered
ambiguous. At this point I'm out of ideas so a good old reinterpret cast
will have to do. If anybody has a better idea I'd be happy to hear it.
llvm-svn: 355866
Jessica Paquette [Mon, 11 Mar 2019 20:51:17 +0000 (20:51 +0000)]
[GlobalISel][AArch64] Always fall back on aarch64.neon.addp.*
Overloaded intrinsics aren't necessarily safe for instruction selection. One
such intrinsic is aarch64.neon.addp.*.
This is a temporary workaround to ensure that we always fall back on that
intrinsic. Eventually this will be replaced with a proper solution.
https://bugs.llvm.org/show_bug.cgi?id=40968
Differential Revision: https://reviews.llvm.org/D59062
llvm-svn: 355865
Alex Bradbury [Mon, 11 Mar 2019 20:43:29 +0000 (20:43 +0000)]
[RISCV][NFC] Convert some MachineBaiscBlock::iterator(MI) to MI.getIterator()
llvm-svn: 355864
Jonas Devlieghere [Mon, 11 Mar 2019 20:31:21 +0000 (20:31 +0000)]
[Reproducers] Implement log_append for function pointers.
Changing the type in the DUMMY macro to void* doesn't actually fix the
build error, because the argument type is deducted from the template (as
opposed to when serializing through the instrumentation framework, where
this would matter). Instead I've added a proper instance of log_append
that takes function pointers and logs their address.
llvm-svn: 355863
Nico Weber [Mon, 11 Mar 2019 20:23:40 +0000 (20:23 +0000)]
Remove esan.
It hasn't seen active development in years, and it hasn't reached a
state where it was useful.
Remove the code until someone is interested in working on it again.
Differential Revision: https://reviews.llvm.org/D59133
llvm-svn: 355862
Med Ismail Bennani [Mon, 11 Mar 2019 20:23:34 +0000 (20:23 +0000)]
Fix some comment typos.
Signed-off-by: Med Ismail Bennani <medismail.bennani@gmail.com>
llvm-svn: 355861
Nikita Popov [Mon, 11 Mar 2019 20:22:13 +0000 (20:22 +0000)]
[SDAG][AArch64] Legalize VECREDUCE
Fixes https://bugs.llvm.org/show_bug.cgi?id=36796.
Implement basic legalizations (PromoteIntRes, PromoteIntOp,
ExpandIntRes, ScalarizeVecOp, WidenVecOp) for VECREDUCE opcodes.
There are more legalizations missing (esp float legalizations),
but there's no way to test them right now, so I'm not adding them.
This also includes a few more changes to make this work somewhat
reasonably:
* Add support for expanding VECREDUCE in SDAG. Usually
experimental.vector.reduce is expanded prior to codegen, but if the
target does have native vector reduce, it may of course still be
necessary to expand due to legalization issues. This uses a shuffle
reduction if possible, followed by a naive scalar reduction.
* Allow the result type of integer VECREDUCE to be larger than the
vector element type. For example we need to be able to reduce a v8i8
into an (nominally) i32 result type on AArch64.
* Use the vector operand type rather than the scalar result type to
determine the action, so we can control exactly which vector types are
supported. Also change the legalize vector op code to handle
operations that only have vector operands, but no vector results, as
is the case for VECREDUCE.
* Default VECREDUCE to Expand. On AArch64 (only target using VECREDUCE),
explicitly specify for which vector types the reductions are supported.
This does not handle anything related to VECREDUCE_STRICT_*.
Differential Revision: https://reviews.llvm.org/D58015
llvm-svn: 355860
Alexey Bataev [Mon, 11 Mar 2019 19:51:42 +0000 (19:51 +0000)]
[OPENMP]Fix codegen for declare target link in target regions.
If the declare target link global is used in the target region
indirectly (used in the inner parallel, teams, etc. regions), we may
miss this variable and it leads to incorrect codegen.
llvm-svn: 355858
Nico Weber [Mon, 11 Mar 2019 19:30:13 +0000 (19:30 +0000)]
gn build: Merge r355777
llvm-svn: 355857
Nico Weber [Mon, 11 Mar 2019 19:14:22 +0000 (19:14 +0000)]
gn build: Merge r355835
llvm-svn: 355856
Nico Weber [Mon, 11 Mar 2019 19:11:15 +0000 (19:11 +0000)]
gn build: Merge r355834
llvm-svn: 355855
Jonas Paulsson [Mon, 11 Mar 2019 19:00:37 +0000 (19:00 +0000)]
[RegAlloc] Avoid compile time regression with multiple copy hints.
As a fix for https://bugs.llvm.org/show_bug.cgi?id=40986 ("excessive compile
time building opencollada"), this patch makes sure that no phys reg is hinted
more than once from getRegAllocationHints().
This handles the case were many virtual registers are assigned to the same
physreg. The previous compile time fix (r343686) in weightCalcHelper() only
made sure that physical/virtual registers are passed no more than once to
addRegAllocationHint().
Review: Dimitry Andric, Quentin Colombet
https://reviews.llvm.org/D59201
llvm-svn: 355854
Shoaib Meenai [Mon, 11 Mar 2019 18:53:57 +0000 (18:53 +0000)]
[clang] Add install targets for API headers
Add an install target for clang's API headers, which allows them to be
included in distributions. The install rules already existed, but they
lacked a component and a target, making them only accessible via a full
install. These headers are useful for writing clang-based tooling, for
example. They're the clang equivalent to the llvm-headers target and
complement the clang-libraries target.
Differential Revision: https://reviews.llvm.org/D58317
llvm-svn: 355853
Brian Gesiak [Mon, 11 Mar 2019 18:31:28 +0000 (18:31 +0000)]
[coroutines][PR40979] Ignore unreachable uses across suspend points
Summary:
Depends on https://reviews.llvm.org/D59069.
https://bugs.llvm.org/show_bug.cgi?id=40979 describes a bug in which the
-coro-split pass would assert that a use was across a suspend point from
a definition. Normally this would mean that a value would "spill" across
a suspend point and thus need to be stored in the coroutine frame. However,
in this case the use was unreachable, and so it would not be necessary
to store the definition on the frame.
To prevent the assert, simply remove unreachable basic blocks from a
coroutine function before computing spills. This avoids the assert
reported in PR40979.
Reviewers: GorNishanov, tks2103
Reviewed By: GorNishanov
Subscribers: EricWF, jdoerfert, llvm-commits, lewissbaker
Tags: #llvm
Differential Revision: https://reviews.llvm.org/D59068
llvm-svn: 355852
Michael Trent [Mon, 11 Mar 2019 18:29:25 +0000 (18:29 +0000)]
Detect malformed LC_LINKER_COMMANDs in Mach-O binaries
Summary:
llvm-objdump can be tricked into reading beyond valid memory and
segfaulting if LC_LINKER_COMMAND strings are not null terminated. libObject
does have code to validate the integrity of the LC_LINKER_COMMAND struct,
but this validator improperly assumes linker command strings are null
terminated.
The solution is to report an error if a string extends beyond the end of
the LC_LINKER_COMMAND struct.
Reviewers: lhames, pete
Reviewed By: pete
Subscribers: rupprecht, llvm-commits
Tags: #llvm
Differential Revision: https://reviews.llvm.org/D59179
llvm-svn: 355851