Commit Graph

312089 Commits

Author SHA1 Message Date
Petar Avramovic 3e0da146ac [MIPS GlobalISel] Improve selection of constants
Certain 32 bit constants can be generated with a single instruction
instead of two. Implement materialize32BitImm function for MIPS32.

Differential Revision: https://reviews.llvm.org/D59369

llvm-svn: 356238
2019-03-15 07:07:50 +00:00
Fangrui Song 76a7ecb3ae [ELF] De-template wrapSymbols, isReserved and addGotEntry. NFC
llvm-svn: 356237
2019-03-15 06:58:23 +00:00
Yonghong Song cacac05aca [BPF] do not generate unused local/global types
The kernel currently has a limit for # of types to be 64KB and
the size of string subsection to be 64KB. A simple bcc tool
runqlat.py generates:
  . the size of ~33KB type section, roughly ~10K types
  . the size of ~17KB string section

The majority type is from the types referenced by local
variables in the bpf program. For example, the kernel "task_struct"
itself recursively brings in ~900 other types.
This patch did the following optimization to avoid generating
unused types:
  . do not generate types for local variables unless they are
    function arguments.
  . do not generate types for external globals.

If an external global is not used in the program, llvm
already removes it from IR, so global variable saving is
typical small. For runqlat.py, only one variable "llvm.used"
is the external global.

The types for locals and external globals can be added back
once there is a usage for them.

After the above optimization, the runqlat.py generates:
  . the size of ~1.5KB type section, roughtly 500 types
  . the size of ~0.7KB string section

UPDATE:
  resubmitted the patch after previous revert with
  the following fix:
  use Global.hasExternalLinkage() to test "external"
  linkage instead of using Global.getInitializer(),
  which will assert on external variables.

Signed-off-by: Yonghong Song <yhs@fb.com>
llvm-svn: 356234
2019-03-15 05:51:25 +00:00
Yonghong Song bf3a279bce Revert "[BPF] do not generate unused local/global types"
This reverts commit r356232.

Reason: test failure with ASSERT on enabled build.
llvm-svn: 356233
2019-03-15 05:02:19 +00:00
Yonghong Song 5664d4c8ca [BPF] do not generate unused local/global types
The kernel currently has a limit for # of types to be 64KB and
the size of string subsection to be 64KB. A simple bcc tool
runqlat.py generates:
  . the size of ~33KB type section, roughly ~10K types
  . the size of ~17KB string section

The majority type is from the types referenced by local
variables in the bpf program. For example, the kernel "task_struct"
itself recursively brings in ~900 other types.
This patch did the following optimization to avoid generating
unused types:
  . do not generate types for local variables unless they are
    function arguments.
  . do not generate types for external globals.

If an external global is not used in the program, llvm
already removes it from IR, so global variable saving is
typical small. For runqlat.py, only one variable "llvm.used"
is the external global.

The types for locals and external globals can be added back
once there is a usage for them.

After the above optimization, the runqlat.py generates:
  . the size of ~1.5KB type section, roughtly 500 types
  . the size of ~0.7KB string section

Signed-off-by: Yonghong Song <yhs@fb.com>
llvm-svn: 356232
2019-03-15 04:42:01 +00:00
Richard Trieu ef1e06df6f Remove an assert in template pack deduction during nested instantiation.
llvm-svn: 356231
2019-03-15 04:26:02 +00:00
Fangrui Song 639058d7c9 [ELF] Add a test for large .bss.rel.ro
Reviewers: ruiu, pcc, espindola

Reviewed By: pcc

Subscribers: emaste, arichardson, llvm-commits

Tags: #llvm

Differential Revision: https://reviews.llvm.org/D59404

llvm-svn: 356230
2019-03-15 03:37:42 +00:00
Fangrui Song f4646a0bb1 [ELF] Fix 3 tests after D56828
arm-plt-reloc.s arm-thumb-plt-reloc.s: update offset calculations

pack-dyn-relocs-loop.s: this test is very sensitive to exact section
offsets and sizes. If we comment out the following two lines in
SyntheticSections.cpp, we should reproduce `ld.lld: error: thunk
creation not converged` caused by oscillation of the section size.

  if (RelocData.size() < OldSize)
    RelocData.append(OldSize - RelocData.size(), 0);

Use -z norelro to counteract the layout change (to be more specific,
we have to place .dynamic below .foo so that offset(foo) remains 0x10004)

llvm-svn: 356229
2019-03-15 03:20:13 +00:00
Aaron Puchert 2f741868ed Add missing override specifier [NFC]
This should fix a -Winconsistent-missing-override warning that is only
visible when Z3 is enabled.

llvm-svn: 356228
2019-03-15 02:30:07 +00:00
Jorge Gorbe Moya 717b1c804b [lldb-vscode] Fix dangling pointer in request_evaluate.
SBError::GetCString() returns a pointer to a string owned by the SBError
object. The code here was calling GetCString on a temporary and using
the returned pointer after the temporary was destroyed.

Differential Revision: https://reviews.llvm.org/D59400

llvm-svn: 356227
2019-03-15 01:46:50 +00:00
Fangrui Song e8710ef1fb [ELF] Split RW PT_LOAD on the PT_GNU_RELRO boundary
Summary:
Based on Peter Collingbourne's suggestion in D56828.

Before D56828: PT_LOAD(.data PT_GNU_RELRO(.data.rel.ro .bss.rel.ro) .bss)
Old:           PT_LOAD(PT_GNU_RELRO(.data.rel.ro .bss.rel.ro) .data .bss)
New:           PT_LOAD(PT_GNU_RELRO(.data.rel.ro .bss.rel.ro)) PT_LOAD(.data. .bss)

The new layout reflects the runtime memory mappings.
By having two PT_LOAD segments, we can utilize the NOBITS part of the
first PT_LOAD and save bytes for .bss.rel.ro.

.bss.rel.ro is currently small and only used by copy relocations of
symbols in read-only segments, but it can be used for other purposes in
the future, e.g. if a relro section's statically relocated data is all
zeros, we can move it to .bss.rel.ro.

Reviewers: espindola, ruiu, pcc

Reviewed By: ruiu

Subscribers: nemanjai, jvesely, nhaehnle, javed.absar, kbarton, emaste, arichardson, llvm-commits

Tags: #llvm

Differential Revision: https://reviews.llvm.org/D58892

llvm-svn: 356226
2019-03-15 01:29:57 +00:00
Davide Italiano 71560b5e2b [DataFormatters] Remove LLDB_DISABLE_PYTHON from FormatManager.
And DataVisualization. One step closer.

llvm-svn: 356225
2019-03-15 00:45:57 +00:00
Marshall Clow 5bcca9ffd1 Mark vector::operator[] and front/back as noexcept. We already do this for string and string_view. This should give better codegen inside of noexcept functions. Add tests for op[]/front/back/at, because apparently we had none.
llvm-svn: 356224
2019-03-15 00:29:35 +00:00
Artem Dergachev f2192b204f [analyzer] RetainCount: A function isn't a CFRetain if it takes no arguments.
Don't crash when a function has a name that starts with "CF" and ends with
"Retain" but takes 0 arguments. In particular, don't try to treat it as if
it returns its first argument.

These problems are inevitable because the checker is naming-convention-based,
but at least we shouldn't crash.

Differential Revision: https://reviews.llvm.org/D59123

llvm-svn: 356223
2019-03-15 00:26:17 +00:00
Artem Dergachev 06451368d2 [analyzer] Support C++17 aggregates with bases without constructors.
RegionStore now knows how to bind a nonloc::CompoundVal that represents the
value of an aggregate initializer when it has its initial segment of sub-values
correspond to base classes.

Additionally, fixes the crash from pr40022.

Differential Revision: https://reviews.llvm.org/D59054

llvm-svn: 356222
2019-03-15 00:22:59 +00:00
Sam Clegg 3d70a2b7d1 [WebAssembly] Remove unused load/store patterns that use texternalsym
Differential Revision: https://reviews.llvm.org/D59395

llvm-svn: 356221
2019-03-15 00:20:13 +00:00
Yan Zhang 787a773498 Fixed global constant/variable naming check on C++ class for ObjC++ files.
Subscribers: cfe-commits

Tags: #clang

Differential Revision: https://reviews.llvm.org/D59283

llvm-svn: 356220
2019-03-15 00:17:41 +00:00
Matt Arsenault 1d83670dbd AMDGPU: Remove intrinsic operand assert
Before r355981, this was under LLVM_DEBUG. I don't think the assert is
quite right, but this really should be a verifier check. Instcombine
should not be asserting on this sort of thing.

llvm-svn: 356219
2019-03-14 23:45:09 +00:00
Sanjay Patel 2c9275a790 [CGP] add another bailout for degenerate code (PR41064)
This is almost the same as:
rL355345
...and should prevent any potential crashing from examples like:
https://bugs.llvm.org/show_bug.cgi?id=41064
...although the bug was masked by:
rL355823
...and I'm not sure how to repro the problem after that change.

llvm-svn: 356218
2019-03-14 23:14:31 +00:00
Paul Robinson 96c1f2cd6c Tighten up tests that use -debugify as a shortcut. NFC
These now verify that a given instruction has a specific source
location, rather than any old location. We want to make sure we
propagate the correct locations from one instruction to another.

llvm-svn: 356217
2019-03-14 23:09:17 +00:00
Eli Friedman fb26c329af [MC] Sort FDEs by the associated CIE before emitting them.
This isn't necessary according to the DWARF standard, but it matches the
.eh_frame sections emitted by other tools in practice, and the Android
libunwindstack rejects .eh_frame sections where an FDE refers to a CIE
other than the closest previous CIE. So match the other tools and also
sort accordingly.

I consider this a bug in libunwindstack, but it's easy enough to emit
a compatible .eh_frame section for compatibility with installed
operating systems.

Differential Revision: https://reviews.llvm.org/D58266

llvm-svn: 356216
2019-03-14 23:08:19 +00:00
Matt Arsenault bc6d07ca46 MIR: Allow targets to serialize MachineFunctionInfo
This has been a very painful missing feature that has made producing
reduced testcases difficult. In particular the various registers
determined for stack access during function lowering were necessary to
avoid undefined register errors in a large percentage of
cases. Implement a subset of the important fields that need to be
preserved for AMDGPU.

Most of the changes are to support targets parsing register fields and
properly reporting errors. The biggest sort-of bug remaining is for
fields that can be initialized from the IR section will be overwritten
by a default initialized machineFunctionInfo section. Another
remaining bug is the machineFunctionInfo section is still printed even
if empty.

llvm-svn: 356215
2019-03-14 22:54:43 +00:00
Jessica Paquette 7d6784f522 [AArch64][GlobalISel] Add isel support for G_UADDO on s32s and s64s
This adds instruction selection support for G_UADDO on s32s and s64s.

Also
- Add an instruction selection test
- Update the arm64-xaluo.ll test to show that we generate the correct assembly

Differential Revision: https://reviews.llvm.org/D58734

llvm-svn: 356214
2019-03-14 22:54:29 +00:00
Amara Emerson d61b89be8d [AArch64][GlobalISel] Implement selection for G_UNMERGE of vectors to vectors.
This re-uses the previous support for extract vector elt to extract the
subvectors.

Differential Revision: https://reviews.llvm.org/D59390

llvm-svn: 356213
2019-03-14 22:48:18 +00:00
Amara Emerson 2ff2298c3e [AArch64][GlobalISel] Add some support for G_CONCAT_VECTORS.
Handles concatenating 2 x v2s32 and 2 x v4s16

Differential Revision: https://reviews.llvm.org/D59390

llvm-svn: 356212
2019-03-14 22:48:15 +00:00
Marshall Clow 3068500aa7 XFAIL this debug-mode test that I just broke. Eric has a patch out for review (D59166) that rewrites this test completely, so I'm not going to bother fixing it.
llvm-svn: 356211
2019-03-14 22:24:32 +00:00
Davide Italiano 3772796aaa [ObjCLanguage] Remove LLDB_DISABLE_PYTHON markers from the formatters.
llvm-svn: 356210
2019-03-14 22:12:59 +00:00
Marshall Clow 5f6a5ac19c Add noexcept to operator[] for array and deque. This is an extension. We already do this for string and string_view. This should give better codegen inside of noexcept functions.
llvm-svn: 356209
2019-03-14 21:56:57 +00:00
Jason Liu 7f7867b05a Reland the rest of "Add AIX Target Info"
llvm-svn 356197 relanded previously failing test case max_align.c.
This commit will reland the rest of llvm-svn 356060 commit.

Differential Revision: https://reviews.llvm.org/D59048

llvm-svn: 356208
2019-03-14 21:54:30 +00:00
Jordan Rupprecht 12ed01dcf9 [llvm-strip] Hook up (unimplemented) --only-keep-debug
For ELF, we accept but ignore --only-keep-debug. Do the same for llvm-strip.

COFF does implement this, so update the test that it is supported.

llvm-svn: 356207
2019-03-14 21:51:42 +00:00
Matt Arsenault 0b31b24c13 AMDGPU: Correct type for waitcnt debug flag
llvm-svn: 356206
2019-03-14 21:23:59 +00:00
Adrian Prantl 07b97492d4 Add test I forgot to git-add in r356163.
llvm-svn: 356205
2019-03-14 21:23:52 +00:00
Sam Clegg 595a97a200 Line wrap README file
llvm-svn: 356204
2019-03-14 21:09:14 +00:00
Nikita Popov 48eb21ee5f [InstCombine] Add tests for range-based saturing math overflow; NFC
Tests for cases where overflow can be determined, but not based on
known bits.

llvm-svn: 356203
2019-03-14 21:06:46 +00:00
Julian Lettner 4d4f64cddd [NFC][TSan] Move libdispatch tests into their own subfolder
Remove 'gcd' file prefix. GCD stands for Grand Central Dispatch, which
is another name for libdispatch.
https://apple.github.io/swift-corelibs-libdispatch/

Remove `REQUIRE: dispatch` from tests.

Also rename lit feature 'dispatch' -> 'libdispatch' to be more explicit
what this is about.

Reviewed By: kubamracek

Differential Revision: https://reviews.llvm.org/D59341

llvm-svn: 356202
2019-03-14 20:59:41 +00:00
Julian Lettner e0e02444bc [TSan][libdispatch] Remove dependency on system headers
Including <dispatch/dispatch.h> and <Blocks.h> transitively pulls in
other system headers. Let's try to avoid that.

Blocks.h: compiler-rt already includes a blocks runtime. Just use the
header file that comes with it.

dispatch.h: Declare the bare minimum required for our implementation,
i.e., everything needed to define the interceptors, but not the
interceptors themselves. See tsan_dispatch_defs.h. I spotted a few other
places in compile-rt, where we declare libdispatch types. Maybe this
file can be moved to sanitizer_common if we deem it useful enough.

tsan_libdispatch.cc now compiles on Linux/Clang (requires support for
-fblocks). Linking still requires some manual configuration.

Reviewed By: kubamracek

Differential Revision: https://reviews.llvm.org/D59145

llvm-svn: 356201
2019-03-14 20:59:37 +00:00
Pete Couperus 9fd1848823 [ARC] Add more load/store variants.
On ARC ISA, general format of load instruction is this:

    LD<zz><.x><.aa><.di> a, [b,c]
And general format of store is this:
    ST<zz><.aa><.di> c, [b,s9]
Where:

<zz> is data size field and can be one of
  <empty> (bits 00) - Word (32-bit), default behavior
  B             (bits 01) - Byte
  H             (bits 10) - Half-word (16-bit)

 <.x> is data extend mode:
  <empty> (bit 0) - If size is not Word(32-bit), then data is zero extended
  X       (bit 1) - If size is not Word(32-bit), then data is sign extended

 <.aa> is address write-back mode:
  <empty> (bits 00) - no write-back
  .AW  (bits 01) - Preincrement, base register updated pre memory transaction
  .AB  (bits 10) - Postincrement, base register updated post memory transaction

 <.di> is cache bypass mode:
  <empty> (bit 0) - Cached memory access, default mode
  .DI     (bit 1) - Non-cached data memory access

  This patch adds these load/store instruction variants to the ARC backend.

Patch By Denis Antrushin! <denis@synopsys.com>

Differential Revision: https://reviews.llvm.org/D58980

llvm-svn: 356200
2019-03-14 20:50:54 +00:00
Nico Weber 51fe000d8a gn build: Add build files for clang-doc
Differential Revision: https://reviews.llvm.org/D59379

llvm-svn: 356199
2019-03-14 20:41:19 +00:00
Alexey Bataev 0d8fcdf11a [OPENMP]Fix crash for the ordered(n) clause.
If the doacross lop construct is used and the loop counter is declare
outside of the loop, the compiler might crash trying to get the address
of the loop counter. Patch fixes this problem.

llvm-svn: 356198
2019-03-14 20:36:00 +00:00
Jason Liu c4420b00f1 Reland part of "Add AIX Target Info"
This patch reland the test case max_align.c which is failing at
Windows and PS4 platform in the previous commit.

Differential Revision: https://reviews.llvm.org/D59048

original llvm-svn: 356060

llvm-svn: 356197
2019-03-14 20:27:39 +00:00
Eli Friedman 4af1c26502 [CodeGen] Consider tied operands when adjusting inline asm operands.
The constraint "0" in the following asm did not consider the its
relationship with "=y" when try to replace the type of the operands.

asm ("nop" : "=y"(Mu8_1 ) : "0"(Mu8_0 ));

Patch by Xiang Zhang.

Differential Revision: https://reviews.llvm.org/D56990

llvm-svn: 356196
2019-03-14 19:46:51 +00:00
Sanjay Patel 38f07b1966 [InstCombine] remove duplicate tests
These got accidentally doubled with rL356191.

llvm-svn: 356195
2019-03-14 19:41:21 +00:00
Louis Dionne 72b8e77a49 [pstl] By default, disable the parallel policies
Since we don't have any non-trivial PSTL backend that doesn't require
TBB yet, enabling the parallel policies by default breaks people that
try to build all of LLVM without having an installation of TBB. Since
this is unacceptable, parallel policies are disabled by default.

We can re-enable it once we have a backend that does not require anything
beyond what C++ already requires. For example, we could have a simple
backend that uses std::thread by default or something along those lines,
with the understanding that vendors would use their own (more efficient)
backend.

llvm-svn: 356194
2019-03-14 19:33:58 +00:00
Sunil Srivastava 6823c823d1 Handle consecutive-double-quotes in Windows argument parsing
Windows command line argument processing treats consecutive double quotes
as a single double-quote. This patch implements this functionality.

Differential Revision: https://reviews.llvm.org/D58662

llvm-svn: 356193
2019-03-14 19:26:04 +00:00
Sanjay Patel de1d5d3675 [InstCombine] canonicalize funnel shift constant shift amount to be modulo bitwidth
The shift argument is defined to be modulo the bitwidth, so if that argument
is a constant, we can always reduce the constant to its minimal form to allow
better CSE and other follow-on transforms.

We need to be careful to ignore constant expressions here, or we will likely
infinite loop. I'm adding a general vector constant query for that case.

Differential Revision: https://reviews.llvm.org/D59374

llvm-svn: 356192
2019-03-14 19:22:08 +00:00
Sanjay Patel 6e86216531 [InstCombine] add tests for funnel shift constant shift amount mod bitwidth; NFC
llvm-svn: 356191
2019-03-14 19:22:00 +00:00
Zachary Turner bb26a7eb11 Return llvm::Error and llvm::Expected from DWARF parsing code.
The goal here is to improve our error handling and error recovery while
parsing DWARF, while at the same time getting us closer to being able to
merge LLDB's DWARF parser with LLVM's. To this end, I've udpated several
of the low-level parsing functions in LLDB to return llvm::Error and
llvm::Expected.

For now, this only updates LLDB parsing functions and not LLVM. In some
ways, this actually gets us *farther* from parity with the two
interfaces, because prior to this patch, at least the parsing interfaces
were the same (i.e. they all just returned bools, and now with this
patch they're diverging). But, I chose to do this for two primary
reasons.

LLDB has error logging code engrained deep within some of its parsing
functions. We don't want to lose this logging information, but obviously
LLVM has no logging mechanism at all. So if we're to merge the
interfaces, we have to find a way to still allow LLDB to properly report
parsing errors while not having the reporting code be inside of LLVM.
LLDB (and indeed, LLVM) overload the meaning of the false return value
from all of these extraction functions to mean both "We reached the null
entry at the end of a list of items, therefore everything was
successful" as well as "something bad and unrecoverable happened during
parsing". So you would have a lot code that would do something like:
while (foo.extract(...)) {
  ...
}

But when the loop stops, why did it stop? Did it stop because it
finished parsing, or because there was an error? Because of this, in
some cases we don't always know whether it is ok to proceed, or how to
proceed, but we were doing it anyway.

In this patch, I solve the second problem by introducing an
enumeration called DWARFEnumState which has two values MoreItems and
Complete. Both of these indicate success, but the latter indicates
that we reached the null entry. Then, I return this value instead of
bool, and convey parsing failure separately.

To solve the first problem (and convey parsing failure) these
functions now return either llvm::Error or llvm::Expected<DWARFEnumState>.
Having this extra bit of information allows us to properly convey all 3 of
"error, bail out", "success, call this function again", and "success,
don't call this function again".

In subsequent patches I plan to extend this pattern to the rest of the
parsing interfaces, which will ultimately get all of the log statements
and error reporting out of the low level parsing code and into the high
level parsing code (e.g. SymbolFileDWARF, DWARFASTParserClang, etc).

Eventually, these same changes will have to be backported to LLVM's
DWARF parser, but diverging in the short term is the easiest way to
converge in the long term.

Differential Revision: https://reviews.llvm.org/D59370

llvm-svn: 356190
2019-03-14 19:05:55 +00:00
Alina Sbirlea 2d7458a351 [MemorySSA] Remove redundant walker assignment [NFC].
Subscribers: llvm-commits
llvm-svn: 356189
2019-03-14 18:45:17 +00:00
Alexandre Ganea 3e60ee9f10 [LLD][COFF] Add /summary to print statistics
/summary prints information about the data (OBJ/LIB/PDB) processed by LLD. The goal is have an estimate about the inputs and outputs, to better understand where the timings go.

Differential Revision: https://reviews.llvm.org/D58599

llvm-svn: 356188
2019-03-14 18:45:08 +00:00
Erik Pilkington 3689caebec [Sema] Fix a use-after-free of a _Nonnull ParsedAttr
We were allocating the implicit attribute in the declarator's attribute pool,
but putting into the declaration specifier's ParsedAttributesView. If there are
multiple declarators, then we'll use the attribute from the declaration
specifier after clearing out the declarators attribute pool. Fix this by
allocating the attribute in the declaration specifier's pool.

rdar://48529718

Differential revision: https://reviews.llvm.org/D59327

llvm-svn: 356187
2019-03-14 18:38:02 +00:00