Roman Gareev [Sun, 21 Aug 2016 11:09:19 +0000 (11:09 +0000)]
Simplify AccFuncMap to vector<> AccessFunctions
getAccessFunctions() is dead code and the 'BB' argument
of getOrCreateAccessFunctions() is not used. This patch deletes
getAccessFunctions and transforms AccFuncMap into
a std::vector<std::unique_ptr<MemoryAccess>> AccessFunctions.
Reviewed-by: Tobias Grosser <tobias@grosser.es>
Differential Revision: https://reviews.llvm.org/D23759
llvm-svn: 279394
Guy Blank [Sun, 21 Aug 2016 08:02:27 +0000 (08:02 +0000)]
[AVX512][FastISel] Do not use K registers in TEST instructions
In some cases, FastIsel was emitting TEST instruction with K reg input, which is illegal.
Changed to using KORTEST when dealing with K regs.
Differential Revision: https://reviews.llvm.org/D23163
llvm-svn: 279393
Asaf Badouh [Sun, 21 Aug 2016 07:56:47 +0000 (07:56 +0000)]
[X86][AVX512F] minor fix of the parameter names
add "__" prefix
Bug 28842 https://llvm.org/bugs/show_bug.cgi?id=29040
Differential Revision: https://reviews.llvm.org/D23753
llvm-svn: 279392
Duncan P. N. Exon Smith [Sun, 21 Aug 2016 00:08:10 +0000 (00:08 +0000)]
ARM: Avoid dereferencing end() in ARMFrameLowering::emitEpilogue
This fixes the crash from PR29072, where the MachineBasicBlock::iterator
wasn't being properly checked against MachineBasicBlock::end() before
iterating. This was another bug exposed by the new
ilist::iterator::operator*() assertion from r279314.
This testcase is poor quality. bugpoint couldn't reduce any further,
and I haven't had time to dig into what's going on so I can't invent a
better one. I didn't even get good CHECK lines in: this is just a
crasher.
I'm committing anyway since this is a real crash with an obvious fix,
but I'll leave PR29072 open and ask an ARM maintainer to help improve
the testcase.
llvm-svn: 279391
Todd Fiala [Sat, 20 Aug 2016 23:24:02 +0000 (23:24 +0000)]
Fix broken macOS LLDB Xcode build from r279314
llvm-svn: 279390
Simon Pilgrim [Sat, 20 Aug 2016 23:09:57 +0000 (23:09 +0000)]
[X86][SSE] Regenerate 32-bit buildvector test
llvm-svn: 279389
Simon Pilgrim [Sat, 20 Aug 2016 22:00:53 +0000 (22:00 +0000)]
[X86][SSE] Regenerate subvector extraction widening test
llvm-svn: 279388
Simon Pilgrim [Sat, 20 Aug 2016 21:56:33 +0000 (21:56 +0000)]
[X86] Regenerate fp truncate tests
llvm-svn: 279387
Simon Pilgrim [Sat, 20 Aug 2016 21:37:30 +0000 (21:37 +0000)]
Regenerate test
llvm-svn: 279386
Simon Pilgrim [Sat, 20 Aug 2016 21:35:45 +0000 (21:35 +0000)]
Regenerate test
llvm-svn: 279385
Simon Pilgrim [Sat, 20 Aug 2016 21:07:41 +0000 (21:07 +0000)]
[X86][XOP] Tweak vpermil2pd test to stop it being combined away
llvm-svn: 279384
Vitaly Buka [Sat, 20 Aug 2016 20:23:50 +0000 (20:23 +0000)]
[asan] Minimize code size by using __asan_set_shadow_* for large blocks
Summary:
We can insert function call instead of multiple store operation.
Current default is blocks larger than 64 bytes.
Changes are hidden behind -asan-experimental-poisoning flag.
PR27453
Differential Revision: https://reviews.llvm.org/D23711
llvm-svn: 279383
Simon Pilgrim [Sat, 20 Aug 2016 20:21:27 +0000 (20:21 +0000)]
Wdocumentation fix
llvm-svn: 279382
Simon Pilgrim [Sat, 20 Aug 2016 20:04:18 +0000 (20:04 +0000)]
Use SDValue::getOpcode() helper instead of via SDValue::getNode()
llvm-svn: 279381
Vitaly Buka [Sat, 20 Aug 2016 18:34:39 +0000 (18:34 +0000)]
[asan] Initialize __asan_set_shadow_* callbacks
Summary:
Callbacks are not being used yet.
PR27453
Reviewers: kcc, eugenis
Subscribers: llvm-commits
Differential Revision: https://reviews.llvm.org/D23634
llvm-svn: 279380
Vitaly Buka [Sat, 20 Aug 2016 18:34:36 +0000 (18:34 +0000)]
[asan] Optimize store size in FunctionStackPoisoner::poisonRedZones
Summary: Reduce store size to avoid leading and trailing zeros.
Reviewers: kcc, eugenis
Subscribers: llvm-commits
Differential Revision: https://reviews.llvm.org/D23648
llvm-svn: 279379
Artem Dergachev [Sat, 20 Aug 2016 17:35:53 +0000 (17:35 +0000)]
[analyzer] Use faster hashing (MD5) in CloneDetector.
This replaces the old approach of fingerprinting every AST node into a string,
which avoided collisions and was simple to implement, but turned out to be
extremely ineffective with respect to both performance and memory.
The collisions are now dealt with in a separate pass, which no longer causes
performance problems because collisions are rare.
Patch by Raphael Isemann!
Differential Revision: https://reviews.llvm.org/D22515
llvm-svn: 279378
Vitaly Buka [Sat, 20 Aug 2016 17:22:36 +0000 (17:22 +0000)]
[asan] Rename asan-instrument-allocas -> asan-instrument-dynamic-allocas
Summary: Depends on D23707.
Subscribers: kubabrecka
Differential Revision: https://reviews.llvm.org/D23709
llvm-svn: 279377
Vitaly Buka [Sat, 20 Aug 2016 17:22:27 +0000 (17:22 +0000)]
[asan] Cleanup instrumentation of dynamic allocas
Summary:
Extract instrumenting dynamic allocas into separate method.
Rename asan-instrument-allocas -> asan-instrument-dynamic-allocas
Differential Revision: https://reviews.llvm.org/D23707
llvm-svn: 279376
Simon Pilgrim [Sat, 20 Aug 2016 17:07:38 +0000 (17:07 +0000)]
[X86][SSE] Added vector interleave test (PR21281)
llvm-svn: 279375
Benjamin Kramer [Sat, 20 Aug 2016 16:51:33 +0000 (16:51 +0000)]
[Sema] Don't crash on scanf on forward-declared enums.
This is valid in GNU C, which allows pointers to incomplete enums. GCC
just pretends that the underlying type is 'int' in those cases, follow
that behavior.
llvm-svn: 279374
Vitaly Buka [Sat, 20 Aug 2016 16:48:24 +0000 (16:48 +0000)]
[asan] Add support of lifetime poisoning into ComputeASanStackFrameLayout
Summary:
We are going to combine poisoning of red zones and scope poisoning.
PR27453
Reviewers: kcc, eugenis
Subscribers: llvm-commits
Differential Revision: https://reviews.llvm.org/D23623
llvm-svn: 279373
Duncan P. N. Exon Smith [Sat, 20 Aug 2016 14:58:31 +0000 (14:58 +0000)]
Move unittests/Support/IteratorTest.cpp to unittests/ADT/
This testing stuff from ADT, not Support. Fix the file location.
llvm-svn: 279372
Vassil Vassilev [Sat, 20 Aug 2016 14:50:22 +0000 (14:50 +0000)]
PR28423: Compare primary declaration contexts.
In certain cases (mostly coming from modules), Sema's idea of the StdNamespace
does not point to the first declaration of namespace std.
Patch by Cristina Cristescu!
Reviewed by Richard Smith.
llvm-svn: 279371
Matthew Simpson [Sat, 20 Aug 2016 14:49:02 +0000 (14:49 +0000)]
Reapply "[SLP] Initialize VectorizedValue when gathering"
The test case included in r279125 exposed existing undefined behavior in the
SLP vectorizer that it did not introduce. This patch reapplies the original
patch, but modifies the test case to avoid hitting the undefined behavior. This
allows us to close PR28330 while keeping the UBSan bot happy. The undefined
behavior the original test uncovered will be addressed in a follow-on patch.
Reference: https://llvm.org/bugs/show_bug.cgi?id=28330
llvm-svn: 279370
Matthew Simpson [Sat, 20 Aug 2016 14:10:06 +0000 (14:10 +0000)]
[SLP] Add command line option for minimum tree size (NFC)
llvm-svn: 279369
Ed Schouten [Sat, 20 Aug 2016 10:54:51 +0000 (10:54 +0000)]
Add R_386_TLS_LE as a relocation having an implicit addend.
TLS on i386 in non-PIE/PIC code seems broken right now, because we don't
properly add the addend encoded in the instruction to the resulting
offset when processing R_386_TLS_LE relocations.
Extend one of the existing tests for TLS on i686 to use an addend.
PR: https://llvm.org/bugs/show_bug.cgi?id=29068
Reviewed by: ruiu
Differential Revision: https://reviews.llvm.org/D23741
llvm-svn: 279368
Artem Dergachev [Sat, 20 Aug 2016 10:06:59 +0000 (10:06 +0000)]
[analyzer] Make CloneDetector consider macro expansions.
So far macro-generated code was treated by the CloneDetector as normal code.
This caused that some macros where reported as false-positive clones because
large chunks of code coming from otherwise concise macro expansions were treated
as copy-pasted code.
This patch ensures that macros are treated in the same way as literals/function
calls. This prevents macros that expand into multiple statements
from being reported as clones.
Patch by Raphael Isemann!
Differential Revision: https://reviews.llvm.org/D23316
llvm-svn: 279367
Artem Dergachev [Sat, 20 Aug 2016 09:57:21 +0000 (09:57 +0000)]
[analyzer] Make CloneDetector consider template arguments.
For example, code samples `isa<Stmt>(S)' and `isa<Expr>(S)'
are no longer considered to be clones.
Patch by Raphael Isemann!
Differential Revision: https://reviews.llvm.org/D23555
llvm-svn: 279366
Vitaly Buka [Sat, 20 Aug 2016 07:09:39 +0000 (07:09 +0000)]
Revert "[SLP] Initialize VectorizedValue when gathering" to fix ubsan bot.
This reverts commit r279125.
https://reviews.llvm.org/D23410
llvm-svn: 279363
Chandler Carruth [Sat, 20 Aug 2016 07:05:13 +0000 (07:05 +0000)]
[CMake/ASan] Skip using libedit if ASan is enabled -- it leaks memory.
Note that you'll have to clear this variable from your CMake cache for
it to take effect.
llvm-svn: 279362
Xinliang David Li [Sat, 20 Aug 2016 05:28:42 +0000 (05:28 +0000)]
[Profile] add test with large counts
llvm-svn: 279361
Chandler Carruth [Sat, 20 Aug 2016 04:57:28 +0000 (04:57 +0000)]
[PM] Introduce an abstraction for all the analyses over a particular IR
unit for use in the PreservedAnalyses set.
This doesn't have any important functional change yet but it cleans
things up and makes the analysis substantially more efficient by
avoiding querying through the type erasure for every analysis.
I also think it makes it much easier to reason about how analyses are
preserved when walking across pass managers and across IR unit
abstractions.
Thanks to Sean and Mehdi both for the comments and suggestions.
Differential Revision: https://reviews.llvm.org/D23691
llvm-svn: 279360
Matt Arsenault [Sat, 20 Aug 2016 04:17:17 +0000 (04:17 +0000)]
amdgpu: Fix default case value for get_local_size
llvm-svn: 279359
Manman Ren [Sat, 20 Aug 2016 03:00:54 +0000 (03:00 +0000)]
Revert r279351 and r279357 due to bot failures
llvm-svn: 279358
Manman Ren [Sat, 20 Aug 2016 02:28:15 +0000 (02:28 +0000)]
Fix windows bot
llvm-svn: 279357
Teresa Johnson [Sat, 20 Aug 2016 01:24:07 +0000 (01:24 +0000)]
[gold/ThinLTO] Restore ThinLTO file management in gold plugin
Summary:
The gold-plugin changes added along with the new LTO API in r278338 had
the effect of removing the management of the PluginInputFile that
ensured the files weren't released back to gold until the backend
threads were complete. Add back the old file handling.
Fixes PR29020.
Reviewers: mehdi_amini
Subscribers: mehdi_amini, llvm-commits, hjl.tools
Differential Revision: https://reviews.llvm.org/D23721
llvm-svn: 279356
Teresa Johnson [Sat, 20 Aug 2016 01:22:10 +0000 (01:22 +0000)]
[gold] Fix new gold test to specify emulation mode
Add emulation mode option for new test added in r279023.
llvm-svn: 279355
Todd Fiala [Sat, 20 Aug 2016 00:54:03 +0000 (00:54 +0000)]
added Xcode lldb-gtest-for-debugging target
The lldb-gtest target is for CI and runs the tests as
part of the build phase. It does not support debugging
the gtests from Xcode, though, due to the run happening
during the build phase.
This change adds a lldb-gtest-for-debugging target that
can be used to debug gtests.
llvm-svn: 279354
Enrico Granata [Sat, 20 Aug 2016 00:48:11 +0000 (00:48 +0000)]
Add logic to the ObjC runtime in LLDB to extract the pointer values of the two singleton (pairtons?) instances of __NSCFBoolean that represent true and false
This is useful because that knowledge will in turn allow no-code-running formatting of boolean NSNumbers; but that's a commit that will have to wait Monday..
llvm-svn: 279353
Evgeniy Stepanov [Sat, 20 Aug 2016 00:38:55 +0000 (00:38 +0000)]
[msan] Disable prlimit test on glibc < 2.13.
llvm-svn: 279352
Manman Ren [Sat, 20 Aug 2016 00:04:21 +0000 (00:04 +0000)]
[NFC] Add a class ObjCProtocolQualifiers to wrap APIs for ObjC protocol list.
This is in preparation of adding a new type class ObjCTypeParamType that
can take protocol qualifiers. ObjCProtocolQualifiers will be shared between
ObjCObjectType and ObjCTypeParamType.
rdar://
24619481
rdar://
25060179
Differential Revision: http://reviews.llvm.org/D23078
llvm-svn: 279351
Matt Arsenault [Sat, 20 Aug 2016 00:01:21 +0000 (00:01 +0000)]
amdgcn: Fix get_local_size IR return type
llvm-svn: 279350
Mehdi Amini [Fri, 19 Aug 2016 23:54:40 +0000 (23:54 +0000)]
[LTO] Add the ability to test -thinlto-emit-imports-files through llvm-lto2
Summary:
Start bringing llvm-lto2 to a level where we can test the LTO API
a bit deeper.
Reviewers: tejohnson
Subscribers: llvm-commits, mehdi_amini
Differential Revision: https://reviews.llvm.org/D23681
llvm-svn: 279349
Vedant Kumar [Fri, 19 Aug 2016 23:15:35 +0000 (23:15 +0000)]
[Driver] Remove unused #include of llvm/Support/config.h
This is a follow-up to r279112 (which removed the need for the header)
and in the same spirit as r279035 (which attempted to un-break
out-of-tree builds).
llvm-svn: 279348
Matthias Braun [Fri, 19 Aug 2016 23:03:28 +0000 (23:03 +0000)]
MachineFunction: Add llvm_unreachable for missing properties
Most compilers should give you a warning anyway though.
llvm-svn: 279346
Jim Ingham [Fri, 19 Aug 2016 22:58:26 +0000 (22:58 +0000)]
Remove a test that depends on knowing all compiler's register allocation schemes.
This test was using a condition that would compare a variable against the register that would hold
it. It was failing with clang on arm64 because clang put the variable on the stack.
This is not a supportable way to write tests.
llvm-svn: 279345
Krzysztof Parzyszek [Fri, 19 Aug 2016 22:57:23 +0000 (22:57 +0000)]
Reset "undef" flag when coalescing subregister into whole register
llvm-svn: 279344
Matt Arsenault [Fri, 19 Aug 2016 22:49:39 +0000 (22:49 +0000)]
amdgcn: Correct return types to be size_t
llvm-svn: 279343
Tim Northover [Fri, 19 Aug 2016 22:40:08 +0000 (22:40 +0000)]
GlobalISel: support legalization of G_FCONSTANTs
llvm-svn: 279341
Tim Northover [Fri, 19 Aug 2016 22:40:00 +0000 (22:40 +0000)]
GlobalISel: teach legalizer how to handle integer constants.
llvm-svn: 279340
Sanjay Patel [Fri, 19 Aug 2016 22:33:26 +0000 (22:33 +0000)]
[InstCombine] use m_APInt to allow icmp (shl X, Y), C folds for splat constant vectors, part 1
This is a partial enablement (move the ConstantInt guard down) because there are many
different folds here and one of the later ones will require reworking 'isSignBitCheck'.
llvm-svn: 279339
Matthias Braun [Fri, 19 Aug 2016 22:31:45 +0000 (22:31 +0000)]
MachineFunction: Cleanup/simplify MachineFunctionProperties::print()
- Always compile print() regardless of LLVM_ENABLE_DUMP. (We usually
only gard dump() functions with that).
- Only show the set properties to reduce output clutter.
- Remove the unused variant that even shows the unset properties.
- Fix comments
llvm-svn: 279338
Matthias Braun [Fri, 19 Aug 2016 22:31:42 +0000 (22:31 +0000)]
MachineFunction: Make LastProperty an alias of the last property
This avoids unnecessary cases in switch statements covering all
properties.
llvm-svn: 279337
Chris Bieneman [Fri, 19 Aug 2016 22:22:35 +0000 (22:22 +0000)]
[CMake] Support building on OS X without Xcode installation
This should resolve PR23162. This patch has two parts.
First we need to check the error code from xcodebuild when querying for SDKs, second if the OS X SDK is not discovered, we ensure that /usr/include exists and use / as the OS X sysroot.
llvm-svn: 279336
Daniel Berlin [Fri, 19 Aug 2016 22:18:38 +0000 (22:18 +0000)]
Partially revert 279331, as we modify this instruction in the loop
llvm-svn: 279335
Chris Bieneman [Fri, 19 Aug 2016 22:17:48 +0000 (22:17 +0000)]
[CMake] Add check-compiler-rt target for runtimes build
Durning standalone builds (which includes runtimes builds) we want to create a target named check-compiler-rt. Additionally we also create check-all if it doesn't already exist as a convienence target that depends on check-compiler-rt.
This allows us to generate a single check target that invokes lit for all test suites in the runtimes projects, while avoiding name collision of check-all and not breaking existing workflows.
llvm-svn: 279334
Chris Bieneman [Fri, 19 Aug 2016 22:17:46 +0000 (22:17 +0000)]
[CMake] Add top-level install-compiler-rt target
This is a wrapper target of all the component install targets. This wrapper target will be used by the new LLVM runtimes directory to connect top-level targets to the runtime project target.
llvm-svn: 279333
Vitaly Buka [Fri, 19 Aug 2016 22:12:58 +0000 (22:12 +0000)]
Revert "[asan] Add support of lifetime poisoning into ComputeASanStackFrameLayout"
This reverts commit r279020.
Speculative revert in hope to fix asan test on arm.
llvm-svn: 279332
Daniel Berlin [Fri, 19 Aug 2016 22:06:23 +0000 (22:06 +0000)]
Convert some depth first traversals to depth_first
llvm-svn: 279331
Devin Coughlin [Fri, 19 Aug 2016 22:04:45 +0000 (22:04 +0000)]
[www] Add nullability questions to analyzer FAQ.
llvm-svn: 279330
Tim Shen [Fri, 19 Aug 2016 21:52:42 +0000 (21:52 +0000)]
[CallGraph] Use decltype instead of pointer_to_unary_function. NFC.
Reviewers: dblaikie
Subscribers: cfe-commits
Differential Revision: https://reviews.llvm.org/D23726
llvm-svn: 279329
Tim Shen [Fri, 19 Aug 2016 21:52:34 +0000 (21:52 +0000)]
[CallGraph] Use decltype instead of pointer_to_unary_function. NFC.
Reviewers: dblaikie
Subscribers: llvm-commits
Differential Revision: https://reviews.llvm.org/D23725
llvm-svn: 279328
Lang Hames [Fri, 19 Aug 2016 21:27:16 +0000 (21:27 +0000)]
[lldb] Use OrcMCJITReplacement rather than MCJIT as the underlying JIT for LLDB
expression evaluation.
OrcMCJITReplacement is a reimplementation of MCJIT using ORC components, and
provides an easy upgrade path to ORC for existing MCJIT clients. There should be
no functional changes resulting from this switch.
llvm-svn: 279327
Tim Shen [Fri, 19 Aug 2016 21:20:13 +0000 (21:20 +0000)]
[GraphTraits] Make nodes_iterator dereference to NodeType*/NodeRef
Currently nodes_iterator may dereference to a NodeType* or a NodeType&. Make them all dereference to NodeType*, which is NodeRef later.
Differential Revision: https://reviews.llvm.org/D23704
Differential Revision: https://reviews.llvm.org/D23705
llvm-svn: 279326
Krzysztof Parzyszek [Fri, 19 Aug 2016 21:12:52 +0000 (21:12 +0000)]
[Packetizer] Add debugging code to stop packetization after N instructions
llvm-svn: 279325
Krzysztof Parzyszek [Fri, 19 Aug 2016 21:07:35 +0000 (21:07 +0000)]
[Hexagon] Avoid register dependencies on indirect branches in packetizer
Do not packetize the instruction setting the branch address with the
indirect branch itself.
llvm-svn: 279324
Tim Shen [Fri, 19 Aug 2016 21:04:45 +0000 (21:04 +0000)]
[ADT] add pointer_iterator, the opposite of pointee_iterator
Differential Revision: https://reviews.llvm.org/D23703
llvm-svn: 279323
Vedant Kumar [Fri, 19 Aug 2016 21:00:40 +0000 (21:00 +0000)]
Revert "[lldb][cmake] Remove libclang as an lldbBase dependency (NFCI)"
This reverts commit r279296. Including LLDBDependencies breaks the
netbsd lldb bot because it exposes LLDB_USED_LIBS, which causes
lldb_link_common_libs to run to completion in unintended sites, which
results in a malformed call to target_link_libraries.
http://lab.llvm.org:8011/builders/lldb-amd64-ninja-netbsd7/builds/5989
Thanks to Chris Bieneman for figuring this out!
llvm-svn: 279322
Kostya Serebryany [Fri, 19 Aug 2016 20:57:09 +0000 (20:57 +0000)]
[libFuzzer] fix the non-debug build warnings
llvm-svn: 279321
Tim Northover [Fri, 19 Aug 2016 20:48:23 +0000 (20:48 +0000)]
GlobalISel: translate floating-point round/extend
llvm-svn: 279320
Tim Northover [Fri, 19 Aug 2016 20:48:16 +0000 (20:48 +0000)]
GlobalISel: translate floating-point comparisons
llvm-svn: 279319
Vitaly Buka [Fri, 19 Aug 2016 20:48:09 +0000 (20:48 +0000)]
[asan] Temporarily mark test as broken on Windows
llvm-svn: 279318
Justin Lebar [Fri, 19 Aug 2016 20:46:45 +0000 (20:46 +0000)]
[NVPTX] Switch nvptx-use-infer-addrspace to true.
Summary:
This switches us to use a different, more powerful algorithm for address
space inference. I've tested this locally and it seems to work great.
Once we're more confident in it, we can remove the old pass altogether.
Reviewers: jingyue
Subscribers: llvm-commits, tra, jholewinski
Differential Revision: https://reviews.llvm.org/D23694
llvm-svn: 279317
Kate Stone [Fri, 19 Aug 2016 20:46:37 +0000 (20:46 +0000)]
Reverted premature changes to .clang-format
llvm-svn: 279316
Kate Stone [Fri, 19 Aug 2016 20:44:07 +0000 (20:44 +0000)]
Moved #include for lldb-python.h to a distinct group with a reminder comment
declaring that it must be first. Failure to do so results in build failures
on macOS due to subtle header conflicts.
llvm-svn: 279315
Duncan P. N. Exon Smith [Fri, 19 Aug 2016 20:40:12 +0000 (20:40 +0000)]
Reapply "ADT: Remove UB in ilist (and use a circular linked list)"
This reverts commit r279053, reapplying r278974 after fixing PR29035
with r279104.
Note that r279312 has been committed in the meantime, and this has been
rebased on top of that. Otherwise it's identical to r278974.
Note for maintainers of out-of-tree code (that I missed in the original
message): if the new isKnownSentinel() assertion is firing from
ilist_iterator<>::operator*(), this patch has identified a bug in your
code. There are a few common patterns:
- Some IR-related APIs htake an IRUnit* that might be nullptr, and pass
in an incremented iterator as an insertion point. Some old code was
using "&*++I", which in the case of end() only worked by fluke. If
the IRUnit in question inherits from ilist_node_with_parent<>, you can
use "I->getNextNode()". Otherwise, use "List.getNextNode(*I)".
- In most other cases, crashes on &*I just need to check for I==end()
before dereferencing.
- There's also occasional code that sends iterators into a function, and
then starts calling I->getOperand() (or other API). Either check for
end() before the entering the function, or early exit.
Note for if the static_assert with HasObsoleteCustomization is firing
for you:
- r278513 has examples of how to stop using custom sentinel traits.
- r278532 removed ilist_nextprev_traits since no one was using it. See
lld's r278469 for the only migration I needed to do.
Original commit message follows.
----
This removes the undefined behaviour (UB) in ilist/ilist_node/etc.,
mainly by removing (gutting) the ilist_sentinel_traits customization
point and canonicalizing on a single, efficient memory layout. This
fixes PR26753.
The new ilist is a doubly-linked circular list.
- ilist_node_base has two ilist_node_base*: Next and Prev. Size-of: two
pointers.
- ilist_node<T> (size-of: two pointers) is a type-safe wrapper around
ilist_node_base.
- ilist_iterator<T> (size-of: two pointers) operates on an
ilist_node<T>*, and downcasts to T* on dereference.
- ilist_sentinel<T> (size-of: two pointers) is a wrapper around
ilist_node<T> that has some extra API for list management.
- ilist<T> (size-of: two pointers) has an ilist_sentinel<T>, whose
address is returned for end().
The new memory layout matches ilist_half_embedded_sentinel_traits<T>
exactly. The Head pointer that previously lived in ilist<T> is
effectively glued to the ilist_half_node<T> that lived in
ilist_half_embedded_sentinel_traits<T>, becoming the Next and Prev in
the ilist_sentinel_node<T>, respectively. sizeof(ilist<T>) is now the
size of two pointers, and there is never any additional storage for a
sentinel.
This is a much simpler design for a doubly-linked list, removing most of
the corner cases of list manipulation (add, remove, etc.). In follow-up
commits, I intend to move as many algorithms as possible into a
non-templated base class (ilist_base) to reduce code size.
Moreover, this fixes the UB in ilist_iterator/getNext/getPrev
operations. Previously, ilist_iterator<T> operated on a T*, even when
the sentinel was not of type T (i.e., ilist_embedded_sentinel_traits and
ilist_half_embedded_sentinel_traits). This added UB to all operations
involving end(). Now, ilist_iterator<T> operates on an ilist_node<T>*,
and only downcasts when the full type is guaranteed to be T*.
What did we lose? There used to be a crash (in some configurations) on
++end(). Curiously (via UB), ++end() would return begin() for users of
ilist_half_embedded_sentinel_traits<T>, but otherwise ++end() would
cause a nice dependable nullptr dereference, crashing instead of a
possible infinite loop. Options:
1. Lose that behaviour.
2. Keep it, by stealing a bit from Prev in asserts builds.
3. Crash on dereference instead, using the same technique.
Hans convinced me (because of the number of problems this and r278532
exposed on Windows) that we really need some assertion here, at least in
the short term. I've opted for #3 since I think it catches more bugs.
I added only a couple of unit tests to root out specific bugs I hit
during bring-up, but otherwise this is tested implicitly via the
extensive usage throughout LLVM.
Planned follow-ups:
- Remove ilist_*sentinel_traits<T>. Here I've just gutted them to
prevent build failures in sub-projects. Once I stop referring to them
in sub-projects, I'll come back and delete them.
- Add ilist_base and move algorithms there.
- Check and fix move construction and assignment.
Eventually, there are other interesting directions:
- Rewrite reverse iterators, so that rbegin().getNodePtr()==&*rbegin().
This allows much simpler logic when erasing elements during a reverse
traversal.
- Remove ilist_traits::createNode, by deleting the remaining API that
creates nodes. Intrusive lists shouldn't be creating nodes
themselves.
- Remove ilist_traits::deleteNode, by (1) asserting that lists are empty
on destruction and (2) changing API that calls it to take a Deleter
functor (intrusive lists shouldn't be in the memory management
business).
- Reconfigure the remaining callback traits (addNodeToList, etc.) to be
higher-level, pulling out a simple_ilist<T> that is much easier to
read and understand.
- Allow tags (e.g., ilist_node<T,tag1> and ilist_node<T,tag2>) so that T
can be a member of multiple intrusive lists.
llvm-svn: 279314
Reid Kleckner [Fri, 19 Aug 2016 20:22:39 +0000 (20:22 +0000)]
Revert "[SimplifyCFG] Rewrite SinkThenElseCodeToEnd"
This reverts commit r279229. It breaks intrinsic function calls in
diamonds.
llvm-svn: 279313
Duncan P. N. Exon Smith [Fri, 19 Aug 2016 20:17:23 +0000 (20:17 +0000)]
Reapply "ADT: Tidy up ilist_traits static asserts, NFC"
This spiritually reapplies r279012 (reverted in r279052) without the
r278974 parts. The differences:
- Only the HasGetNext trait exists here, so I've only cleaned up (and
tested) it. I still added HasObsoleteCustomization since I know
this will be expanding when r278974 is reapplied.
- I changed the unit tests to use static_assert to catch problems
earlier in the build.
- I added negative tests for the type traits.
Original commit message follows.
----
Change the ilist traits to use decltype instead of sizeof, and add
HasObsoleteCustomization so that additions to this list don't
need to be added in two places.
I suspect this will now work with MSVC, since the trait tested in
r278991 seems to work. If for some reason it continues to fail on
Windows I'll follow up by adding back the #ifndef _MSC_VER.
llvm-svn: 279312
Tim Northover [Fri, 19 Aug 2016 20:09:15 +0000 (20:09 +0000)]
GlobalISel: translate floating-point constants
llvm-svn: 279311
Tim Northover [Fri, 19 Aug 2016 20:09:11 +0000 (20:09 +0000)]
GlobalISel: translate float/int conversion instructions.
llvm-svn: 279310
Tim Northover [Fri, 19 Aug 2016 20:09:07 +0000 (20:09 +0000)]
GlobalISel: support translating select instructions.
llvm-svn: 279309
Tim Northover [Fri, 19 Aug 2016 20:09:03 +0000 (20:09 +0000)]
GlobalISel: fix insert/extract to work on ConstantExprs too.
No tests yet unfortunately (ConstantFolding reduces all supported constants to
ConstantInts before we get to translation). Soon.
llvm-svn: 279308
Tim Northover [Fri, 19 Aug 2016 20:09:01 +0000 (20:09 +0000)]
GlobalISel: fix stale comment
llvm-svn: 279307
Tim Northover [Fri, 19 Aug 2016 20:08:55 +0000 (20:08 +0000)]
GlobalISel: translate insertvalue instructions.
This adds a G_INSERT instruction, which technically makes G_SEQUENCE redundant
(it's equivalent to a G_INSERT into an IMPLICIT_DEF). We'll leave G_SEQUENCE
for now though: it's likely to be far more common as it's a fundamental part of
legalization, so avoiding the mess and bloat of the extra IMPLICIT_DEFs is
probably worthwhile.
llvm-svn: 279306
Tom Stellard [Fri, 19 Aug 2016 19:59:18 +0000 (19:59 +0000)]
MachineScheduler: Add constructor functions for the DAGMutations
Summary: This way they can be re-used by target-specific schedulers.
Reviewers: atrick, MatzeB, kparzysz
Subscribers: kparzysz, llvm-commits, MatzeB
Differential Revision: https://reviews.llvm.org/D23678
llvm-svn: 279305
Krzysztof Parzyszek [Fri, 19 Aug 2016 19:36:35 +0000 (19:36 +0000)]
[Hexagon] Add RUN line to test
llvm-svn: 279304
Krzysztof Parzyszek [Fri, 19 Aug 2016 19:29:15 +0000 (19:29 +0000)]
[Hexagon] Fix subesthetic indentation
llvm-svn: 279303
Krzysztof Parzyszek [Fri, 19 Aug 2016 19:17:28 +0000 (19:17 +0000)]
[Hexagon] Allow i1 values for 'r' constraint in inline-asm
llvm-svn: 279302
Simon Pilgrim [Fri, 19 Aug 2016 19:07:10 +0000 (19:07 +0000)]
[CostModel][X86] Added sub, or, and, fadd and fsub costs and missing 512-bit mul costs
llvm-svn: 279301
Sanjay Patel [Fri, 19 Aug 2016 19:03:07 +0000 (19:03 +0000)]
[InstCombine] remove an icmp fold that is already handled by InstSimplify
Specifically, this is done near the end of "SimplifyICmpInst" using
computeKnownBits() as the broader solution. There are even vector
tests (yay!) for this in test/Transforms/InstSimplify/compare.ll.
I considered putting an assert here instead of just deleting, but
then we could assert every possible fold in InstSimplify in
InstCombine, so...less is more?
llvm-svn: 279300
Richard Smith [Fri, 19 Aug 2016 18:57:17 +0000 (18:57 +0000)]
Add missing #include found by modules build.
llvm-svn: 279298
Krzysztof Parzyszek [Fri, 19 Aug 2016 18:46:13 +0000 (18:46 +0000)]
[Hexagon] Do not cache alloca instructions during isel
They can be deleted or replicated, so the cache may become outdated.
They only need to be visited once during frame lowering, so just scan
the function instead.
llvm-svn: 279297
Vedant Kumar [Fri, 19 Aug 2016 18:41:50 +0000 (18:41 +0000)]
[lldb][cmake] Remove libclang as an lldbBase dependency (NFCI)
It's pulling in all kinds of things it doesn't need (e.g, clang-tidy!).
Eliminating this dependency removes 1056 dependencies from the
'CommandObjectFrame.cpp.o' target and 454 dependencies from the 'lldb'
target. On my machine, this shaves 7 minutes off of a clean build of
lldb.
Thanks to Zachary Turner for pointing out some issues with an earlier
version of this patch!
Differential Revision: https://reviews.llvm.org/D22987
llvm-svn: 279296
Chandler Carruth [Fri, 19 Aug 2016 18:36:06 +0000 (18:36 +0000)]
[PM] Re-instate r279227 and r279228 with a fix to the way the templating
was done to hopefully appease MSVC.
As an upside, this also implements the suggestion Sanjoy made in code
review, so two for one! =]
I'll be watching the bots to see if there are still issues.
llvm-svn: 279295
Tim Northover [Fri, 19 Aug 2016 18:32:14 +0000 (18:32 +0000)]
GlobalISel: improve representation of G_SEQUENCE and G_EXTRACT
First, make sure all types involved are represented, rather than being implicit
from the register width.
Second, canonicalize all types to scalar. These operations just act in bits and
don't care about vectors.
Also standardize spelling of Indices in the MachineIRBuilder (NFC here).
llvm-svn: 279294
Simon Pilgrim [Fri, 19 Aug 2016 18:24:10 +0000 (18:24 +0000)]
[CostModel][X86] Added some AVX512 and 512-bit vector cost tests
llvm-svn: 279291
Kyle Butt [Fri, 19 Aug 2016 18:17:06 +0000 (18:17 +0000)]
Revert "IfConversion: Rescan diamonds."
This reverts commit
bfd62a4b4465dd21811bf615c3b04c30ddb09f7b.
llvm-svn: 279289
Kyle Butt [Fri, 19 Aug 2016 18:17:04 +0000 (18:17 +0000)]
Revert "CodeGen: If Convert blocks that would form a diamond when tail-merged."
This reverts commit
0fda93481c4231c06b838ef476c0c404c51ff875.
llvm-svn: 279288
Tim Northover [Fri, 19 Aug 2016 18:09:41 +0000 (18:09 +0000)]
GlobalISel: allow extractvalue to extract an aggregate.
llvm-svn: 279287