Justin Lebar [Tue, 22 Nov 2016 23:13:49 +0000 (23:13 +0000)]
[StructurizeCFG] Remove unnecessary "using" in class.
Reviewers: arsenm
Subscribers: wdng, llvm-commits
Differential Revision: https://reviews.llvm.org/D26995
llvm-svn: 287719
Justin Lebar [Tue, 22 Nov 2016 23:13:44 +0000 (23:13 +0000)]
[StructurizeCFG] Merge the two constructors into one.
Reviewers: arsenm
Subscribers: wdng, llvm-commits
Differential Revision: https://reviews.llvm.org/D26994
llvm-svn: 287718
Justin Lebar [Tue, 22 Nov 2016 23:13:37 +0000 (23:13 +0000)]
[StructurizeCFG] Use a for-each loop instead of iterators in runOnRegion.
Summary:
Reviewers: arsenm
Subscribers: wdng, llvm-commits
Differential Revision: https://reviews.llvm.org/D26993
llvm-svn: 287717
Justin Lebar [Tue, 22 Nov 2016 23:13:33 +0000 (23:13 +0000)]
[StructurizeCFG] Make hasOnlyUniformBranches a non-member function.
Summary: Lets us get rid of one member variable too.
Reviewers: arsenm
Subscribers: wdng, llvm-commits
Differential Revision: https://reviews.llvm.org/D26992
llvm-svn: 287716
Justin Lebar [Tue, 22 Nov 2016 23:13:29 +0000 (23:13 +0000)]
[CUDA] Note in docs that you need to build with -lcudart on MacOS
-lcudart_static doesn't work. We don't know why.
llvm-svn: 287715
Rui Ueyama [Tue, 22 Nov 2016 23:13:08 +0000 (23:13 +0000)]
Fix memory leak detected by asan.
llvm-svn: 287714
Richard Smith [Tue, 22 Nov 2016 22:55:12 +0000 (22:55 +0000)]
Make diagnostic for use of default member initializer before enclosing class is
complete a little more general; it is produced in other cases than the one that
it previously talked about.
llvm-svn: 287713
Sanjay Patel [Tue, 22 Nov 2016 22:54:36 +0000 (22:54 +0000)]
add and use isBitwiseLogicOp() helper function; NFCI
llvm-svn: 287712
Rui Ueyama [Tue, 22 Nov 2016 22:54:03 +0000 (22:54 +0000)]
Accept -script=<file> in addition to -script <file>.
Fixes PR31126.
llvm-svn: 287711
Dehao Chen [Tue, 22 Nov 2016 22:50:01 +0000 (22:50 +0000)]
Before sample pgo annotation, do not inline a function that has no debug info. (NFC)
If there is no debug info in the callee, inlining it will not help annotator. This avoids infinite loop as reported in PR/31119.
llvm-svn: 287710
Davide Italiano [Tue, 22 Nov 2016 22:11:25 +0000 (22:11 +0000)]
[SCCP] Remove code in visitBinaryOperator (and add tests).
We visit and/or, we try to derive a lattice value for the
instruction even if one of the operands is overdefined.
If the non-overdefined value is still 'unknown' just return and wait
for ResolvedUndefsIn to "plug in" the correct value. This simplifies
the logic a bit. While I'm here add tests for missing cases.
llvm-svn: 287709
Matthias Braun [Tue, 22 Nov 2016 22:09:03 +0000 (22:09 +0000)]
TargetSubtargetInfo: Move implementation to lib/CodeGen; NFC
TargetSubtargetInfo is filled with CodeGen specific interfaces nowadays
(getInstrInfo(), getFrameLowering(), getSelectionDAGInfo()) most of the
tuning flags like enablePostRAScheduler(), getAntiDepBreakMode(),
enableRALocalReassignment(), ... also do not seem to be universal enough
to make sense outside of CodeGen.
Differential Revision: https://reviews.llvm.org/D26948
llvm-svn: 287708
Sanjay Patel [Tue, 22 Nov 2016 22:05:48 +0000 (22:05 +0000)]
[InstCombine] change bitwise logic type to eliminate bitcasts
In PR27925:
https://llvm.org/bugs/show_bug.cgi?id=27925
...we proposed adding this fold to eliminate a bitcast. In D20774, there was
some concern about changing the type of a bitwise op as well as creating
bitcasts that might not be free for a target. However, if we're strictly
eliminating an instruction (by limiting this to one-use ops), then we should
be able to do this in InstCombine.
But we're cautiously restricting the transform for now to vector types to
avoid possible backend problems. A transform to make sure the logic op is
legal for the target should be added to reverse this transform and improve
codegen.
Differential Revision: https://reviews.llvm.org/D26641
llvm-svn: 287707
Simon Pilgrim [Tue, 22 Nov 2016 22:04:50 +0000 (22:04 +0000)]
[X86][AVX512DQ] Add fp <-> int tests for AVX512DQ/AVX512DQ+VL
llvm-svn: 287706
Chandler Carruth [Tue, 22 Nov 2016 21:40:10 +0000 (21:40 +0000)]
[LCG] Add a previously missing assert about the relationship of RefSCCs.
No intended change, everything seems to be in working order already.
llvm-svn: 287705
Rafael Espindola [Tue, 22 Nov 2016 21:37:38 +0000 (21:37 +0000)]
Fix test to not depend on the path size.
llvm-svn: 287704
Tobias Grosser [Tue, 22 Nov 2016 21:31:59 +0000 (21:31 +0000)]
Update to isl-0.17.1-284-gbb38638
Regular maintenance update with only minor changes.
llvm-svn: 287703
Peter Collingbourne [Tue, 22 Nov 2016 21:25:30 +0000 (21:25 +0000)]
LTO: Remove a now-unused InputFile accessor.
llvm-svn: 287702
Rafael Espindola [Tue, 22 Nov 2016 21:12:20 +0000 (21:12 +0000)]
move VerDef finalization before DynStrTab
llvm-svn: 287701
Vyacheslav Klochkov [Tue, 22 Nov 2016 20:52:53 +0000 (20:52 +0000)]
Fixed the lost FastMathFlags in GVN(Global Value Numbering).
Reviewer: Hal Finkel.
Differential Revision: https://reviews.llvm.org/D26952
llvm-svn: 287700
Davide Italiano [Tue, 22 Nov 2016 20:37:37 +0000 (20:37 +0000)]
[LTO] Remove a check on datalayout.
Now that lld switched to lib/LTO, which always calls setDataLayout(),
we don't need this check anymore.
Thanks to Peter for pointing out!
llvm-svn: 287699
Chandler Carruth [Tue, 22 Nov 2016 20:35:32 +0000 (20:35 +0000)]
[LCG] Start using SCC relationship predicates in the unittest.
This mostly gives us nice unittesting of the predicates themselves. I'll
start using them further in subsequent commits to help test the actual
operations performed on the graph.
llvm-svn: 287698
Rui Ueyama [Tue, 22 Nov 2016 20:32:22 +0000 (20:32 +0000)]
Remove PDBFileBuilder::build() and related functions.
PDBFileBuilder supports two different ways to create files.
One is PDBFileBuilder::commit. That function takes a filename
and write a result to the file. The other is PDBFileBuilder::build.
That returns a new PDBFile object.
This patch removes the latter because no one is using it and
in a real life situation we are very unlikely to need it.
Even if you need it, it'd be easy to write a new PDB to a memory
buffer and read it back.
Removing PDBFileBuilder::build enables us to remove other classes
build transitively.
Differential Revision: https://reviews.llvm.org/D26987
llvm-svn: 287697
Kuba Mracek [Tue, 22 Nov 2016 20:24:26 +0000 (20:24 +0000)]
[sanitizer] Fix the dedup_token_length_test.cc testcase to not fail when user's home directory contains "bar"
Differential Revision: https://reviews.llvm.org/D24605
llvm-svn: 287696
Vyacheslav Klochkov [Tue, 22 Nov 2016 20:23:04 +0000 (20:23 +0000)]
Fixed the lost FastMathFlags in Reassociate optimization.
Reviewer: Hal Finkel.
Differential Revision: https://reviews.llvm.org/D26957
llvm-svn: 287695
Tobias Grosser [Tue, 22 Nov 2016 20:21:16 +0000 (20:21 +0000)]
[CodeGen] Add flag to code-generate most memory access expressions
Introduce the new flag -polly-codegen-generate-expressions which forces Polly
to code generate AST expressions instead of using our SCEV based access
expression generation even for cases where the original memory access relation
was not changed and the SCEV based access expression could be code generated
without any issue.
This is an experimental option for better testing the isl ast expression
generation. The default behavior of Polly remains unchanged. We also exclude
a couple of cases for which the AST expression is not yet working.
llvm-svn: 287694
Eric Fiselier [Tue, 22 Nov 2016 20:15:39 +0000 (20:15 +0000)]
Allow libc++ to use modules in C++03.
Libc++ internal uses <atomic> in C++03 code but the module map forbids its use.
This causes the libc++ 'std' module to fail to build in C++03.
This patch removes the requirement to fix this issue.
llvm-svn: 287693
Rui Ueyama [Tue, 22 Nov 2016 20:15:35 +0000 (20:15 +0000)]
Remove '.' from a help message.
llvm-svn: 287692
John McCall [Tue, 22 Nov 2016 20:12:32 +0000 (20:12 +0000)]
Remove variable only used in an assertion.
llvm-svn: 287691
Eric Fiselier [Tue, 22 Nov 2016 20:05:19 +0000 (20:05 +0000)]
Attempt to fix stdint/cstdint modules try 2
llvm-svn: 287690
Marcin Koscielnicki [Tue, 22 Nov 2016 20:03:40 +0000 (20:03 +0000)]
Add test for D21736.
Differential Revision: http://reviews.llvm.org/D21741
llvm-svn: 287689
Marcin Koscielnicki [Tue, 22 Nov 2016 20:03:35 +0000 (20:03 +0000)]
[CodeGen] Insert TargetLibraryInfoWrapperPass before anything else.
Currently, TargetLibraryInfoWrapperPass is inserted by PMBuilder.
However, some passes are inserted manually before the PMBuilder
ones - if any of them happens to use TargetLibraryInfoWrapperPass,
it'll get a default-constructed one, with an unknown target triple.
This happens to InstrProfiling in D21736, breaking it.
Differential Revision: http://reviews.llvm.org/D21737
llvm-svn: 287688
Eric Fiselier [Tue, 22 Nov 2016 20:00:21 +0000 (20:00 +0000)]
Cleanup module.modulemap
llvm-svn: 287687
Paul Robinson [Tue, 22 Nov 2016 19:46:51 +0000 (19:46 +0000)]
Restructure DwarfDebug::beginInstruction(). [NFC]
Will help a pending patch.
Differential Revision: http://reviews.llvm.org/D26982
llvm-svn: 287686
Nico Weber [Tue, 22 Nov 2016 19:38:07 +0000 (19:38 +0000)]
darwin: Unconditionally pass -lto_library, remove -Wliblto warning.
https://reviews.llvm.org/D25932 made it so that clang always checks if
libLTO.dylib is present on disk, even if -flto is not being used. The
motivation for that change was that if a dependency happens to contain bitcode,
ld64 will try to load libLTO without -flto explicitly being enabled. However,
the change had the undesirable side effect of warning if libLTO.dylib doesn't
exist even if it isn't needed.
Change things so that -lto_library is always passes, independent of if it
exists or not. ld64 only looks at this flag if it uses LTO. If the dylib
exists, all is well. If it doesn't, and LTO is not being used, all is well too.
If ld64 does end up using LTO and the dylib does not exist, ld64 will print
something like
ld: could not process llvm bitcode object file, because foo/libLTO.dylib could not be loaded file 'test.o' for architecture x86_64
https://reviews.llvm.org/D26984
llvm-svn: 287685
Shoaib Meenai [Tue, 22 Nov 2016 19:36:26 +0000 (19:36 +0000)]
[Triple] Add Facebook vendor
Add a compiler vendor for Facebook, to enable future vendor-specific
behavior.
Differential Revision: https://reviews.llvm.org/D25136
llvm-svn: 287684
Rui Ueyama [Tue, 22 Nov 2016 19:24:52 +0000 (19:24 +0000)]
Fix build breakage.
We cannot have MipsRldMap class and In<ELFT>::MipsRldMap.
Renamed the class.
llvm-svn: 287683
Chandler Carruth [Tue, 22 Nov 2016 19:23:31 +0000 (19:23 +0000)]
[LCG] Add utilities to compute parent and ascestor relationships between
SCCs.
These will be fairly expensive routines to call and might be abused in
real code, but are quite useful when debugging or in asserts and are
reasonable and well formed properties to query.
I've used one of them in an assert that was requested in a code review
here. In subsequent commits I'll start using these routines more
heavily, for example in unittests etc. But this at least gets the
groundwork in place.
Differential Revision: https://reviews.llvm.org/D25506
llvm-svn: 287682
Simon Dardis [Tue, 22 Nov 2016 19:17:23 +0000 (19:17 +0000)]
[mips] seb, seh instruction aliases
Add the single operand form.
Reviewers: vkalintiris
Differential Revision: https://reviews.llvm.org/D26961
llvm-svn: 287681
Andrew Kaylor [Tue, 22 Nov 2016 19:16:04 +0000 (19:16 +0000)]
Add IntrInaccessibleMemOnly property for intrinsics
Differential Revision: https://reviews.llvm.org/D26485
llvm-svn: 287680
Nemanja Ivanovic [Tue, 22 Nov 2016 19:02:07 +0000 (19:02 +0000)]
[PowerPC] Emit VMX loads/stores for aligned ops to avoid adding swaps on LE
This patch corresponds to review:
https://reviews.llvm.org/D26861
It also fixes PR30730.
Committing on behalf of Lei Huang.
llvm-svn: 287679
Simon Pilgrim [Tue, 22 Nov 2016 18:55:49 +0000 (18:55 +0000)]
[CostModel][X86] Updated sitofp/uitofp scalar/vector cost tests
Better coverage of all legal types + special cases.
Removed old fptoui tests which are all handled in fptoui.ll
llvm-svn: 287678
Meador Inge [Tue, 22 Nov 2016 18:01:50 +0000 (18:01 +0000)]
[ELF] Allow `ASSERT` in output section descriptions
GNU LD allows `ASSERT` commands to be in output section descriptions.
Note that LD also mandates that `ASSERT` commands in this context must
end with a semicolon.
llvm-svn: 287677
Simon Pilgrim [Tue, 22 Nov 2016 17:50:06 +0000 (17:50 +0000)]
[X86][SSE] Combine UNPCKL(FHADD,FHADD) -> FHADD for v2f64 shuffles.
This occurs during UINT_TO_FP v2f64 lowering.
We can easily generalize this to other horizontal ops (FHSUB, PACKSS, PACKUS) as required - we are doing something similar with PACKUS in lowerV2I64VectorShuffle
llvm-svn: 287676
Eugene Leviant [Tue, 22 Nov 2016 17:49:14 +0000 (17:49 +0000)]
[ELF] Convert .rld_map to input section
Differential revision: https://reviews.llvm.org/D26958
llvm-svn: 287675
Benjamin Kramer [Tue, 22 Nov 2016 17:29:45 +0000 (17:29 +0000)]
[clang-rename] Prune away AST nodes more correctly and effectively when looking for a point
Due to the way the preprocessor works nodes can be half in a macro or a
different file. This means checking the file name of the start location
of a Decl is not a correct way of checking if the entire Decl is in that
file. Remove that flawed assumption and replace it with a more effective
check: If the point we're looking for is *inside* of the begin and end
location of a Decl, look inside.
This should make clang-rename more reliable (for example macro'd namespaces
threw it off before, as seen in the test case) and maybe a little faster
by pruning off more stuff that the RecursiveASTVisitor doesn't have to
drill into.
llvm-svn: 287649
Zachary Turner [Tue, 22 Nov 2016 17:10:15 +0000 (17:10 +0000)]
Re-add "demonstrate new Args API"
This fixes the build breakage due to the use of C++14.
llvm-svn: 287647
Vasileios Kalintiris [Tue, 22 Nov 2016 16:43:49 +0000 (16:43 +0000)]
[mips] Add support for unaligned load/store macros.
Add missing unaligned store macros (ush/usw) and fix the exisiting
implementation of the unaligned load macros in order to generate
identical expansions with the GNU assembler.
llvm-svn: 287646
Tim Northover [Tue, 22 Nov 2016 16:17:20 +0000 (16:17 +0000)]
CodeGen: simplify TargetMachine::getSymbol interface. NFC.
No-one actually had a mangler handy when calling this function, and
getSymbol itself went most of the way towards getting its own mangler
(with a local TLOF variable) so forcing all callers to supply one was
just extra complication.
llvm-svn: 287645
Zvi Rackover [Tue, 22 Nov 2016 15:33:28 +0000 (15:33 +0000)]
[X86] Change lowerBuildVectorToBitOp() to take a BuildVectorSDNode. NFC.
llvm-svn: 287644
Zvi Rackover [Tue, 22 Nov 2016 15:17:52 +0000 (15:17 +0000)]
[X86] Remove dead code from LowerVectorBroadcast
Summary: Splat vectors are canonicalized to BUILD_VECTOR's so the code can be simplified. NFC-ish.
Reviewers: craig.topper, delena, RKSimon, andreadb
Subscribers: RKSimon, llvm-commits
Differential Revision: https://reviews.llvm.org/D26678
llvm-svn: 287643
Chad Rosier [Tue, 22 Nov 2016 14:25:02 +0000 (14:25 +0000)]
[AArch64] Set the max interleave factor for Falkor.
llvm-svn: 287642
Chad Rosier [Tue, 22 Nov 2016 14:12:09 +0000 (14:12 +0000)]
[AArch64] Maximize 80-column. NFC.
llvm-svn: 287640
Eric Liu [Tue, 22 Nov 2016 13:46:42 +0000 (13:46 +0000)]
Make llvm::Error generated from replacement interfaces more specific.
Summary:
The new error information contains the type of error (e.g. overlap or bad file path)
and the replacement(s) that is causing the error. This enables us to resolve some errors.
For example, for insertion at the same location conflict, we need to know the
existing replacement which conflicts with the new replacement in order to calculate
the new position to be insert before/after the existing replacement (for merging).
Reviewers: klimek, bkramer
Subscribers: djasper, cfe-commits
Differential Revision: https://reviews.llvm.org/D26853
llvm-svn: 287639
Simon Pilgrim [Tue, 22 Nov 2016 13:27:29 +0000 (13:27 +0000)]
Fix line endings
llvm-svn: 287638
Benjamin Kramer [Tue, 22 Nov 2016 13:13:33 +0000 (13:13 +0000)]
[wasm] hack around test failure after r287553.
This test is very brittle as small changes to block layout break the
check patterns. Hack around a change one more time.
llvm-svn: 287637
Sam McCall [Tue, 22 Nov 2016 11:40:25 +0000 (11:40 +0000)]
Improve detection of global vs local variables
Summary:
Improve detection of global vs local variables.
Currently when a global variable is optimized out or otherwise has an unknown
location (DW_AT_location is empty) it gets reported as local.
I added two new heuristics:
- if a mangled name is present, the variable is global (or static)
- if DW_AT_location is present but invalid, the variable is global (or static)
Subscribers: lldb-commits
Differential Revision: https://reviews.llvm.org/D26908
llvm-svn: 287636
Simon Pilgrim [Tue, 22 Nov 2016 11:29:19 +0000 (11:29 +0000)]
[SelectionDAG] ComputeNumSignBits of TRUNCATE operations
Add basic ComputeNumSignBits support for TRUNCATE ops for cases where the source's number of sign bits overlaps with the truncated size.
Improves X86 SIGN_EXTEND_IN_REG vector cases which were needlessly sign extending boolean vector results.
Differential Revision: https://reviews.llvm.org/D26851
llvm-svn: 287635
Dmitry Vyukov [Tue, 22 Nov 2016 11:09:35 +0000 (11:09 +0000)]
tsan: fix build after 287632
1. Mac does not have MAP_32BIT.
Define it to 0 if it is not defined.
2. We are lacking nolibc stub for ListOfModules::init.
Add it.
llvm-svn: 287634
Axel Naumann [Tue, 22 Nov 2016 10:00:23 +0000 (10:00 +0000)]
Missing initializer.
llvm-svn: 287633
Dmitry Vyukov [Tue, 22 Nov 2016 09:49:11 +0000 (09:49 +0000)]
tsan: switch libignore from /proc/self/maps to dl_iterate_phdr
/proc/self/maps can't be read atomically, this leads to episodic
crashes in libignore as it thinks that a module is loaded twice.
See the new test for an example.
dl_iterate_phdr does not have this problem.
Switch libignore to dl_iterate_phdr.
llvm-svn: 287632
Omair Javaid [Tue, 22 Nov 2016 09:47:00 +0000 (09:47 +0000)]
Fix build failure on Linux and BSD by reverting r287597
Linux and BSD builds failing after this changes from rev 287597.
llvm-svn: 287631
Coby Tayree [Tue, 22 Nov 2016 09:30:29 +0000 (09:30 +0000)]
[AVX512][inline-asm] Fix AVX512 inline assembly instruction resolution when the size qualifier of a memory operand is not specified explicitly.
This commit handles cases where the size qualifier of an indirect memory reference operand in Intel syntax is missing (e.g. "vaddps xmm1, xmm2, [a]").
GCC will deduce the size qualifier for AVX512 vector and broadcast memory operands based on the possible matches:
"vaddps xmm1, xmm2, [a]" matches only “XMMWORD PTR” qualifier.
"vaddps xmm1, xmm2, [a]{1to4}" matches only “DWORD PTR” qualifier.
This is different from the current behavior of LLVM, which deduces the size qualifier based on the size of the memory operand.
For "vaddps xmm1, xmm2, [a]"
"char a;" will imply "BYTE PTR" qualifier
"short a;" will imply "WORD PTR" qualifier.
This commit aligns LLVM to GCC’s behavior.
This is the LLVM part of the review.
The Clang part of the review: https://reviews.llvm.org/D26587
Differential Revision: https://reviews.llvm.org/D26586
llvm-svn: 287630
Dmitry Vyukov [Tue, 22 Nov 2016 08:59:17 +0000 (08:59 +0000)]
tsan: mark cur_thread_placeholder definition as initial-exec
See https://sourceware.org/bugzilla/show_bug.cgi?id=20805 and
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=78294 for context.
Previously we marked only declaration as initial-exec. But compilers treat
initial-exec attribute somewhat differently. Mark definition as well.
llvm-svn: 287629
Adam Nemet [Tue, 22 Nov 2016 07:35:19 +0000 (07:35 +0000)]
Rename option to -lto-pass-remarks-output
The new option -pass-remarks-output broke LLVM_LINK_LLVM_DYLIB because
of the duplicate option name with opt.
llvm-svn: 287628
Adam Nemet [Tue, 22 Nov 2016 07:35:14 +0000 (07:35 +0000)]
Rename option to -lto-pass-remarks-output
The new option -pass-remarks-output broke LLVM_LINK_LLVM_DYLIB because
of the duplicate option name with opt.
llvm-svn: 287627
Tobias Grosser [Tue, 22 Nov 2016 07:18:57 +0000 (07:18 +0000)]
[test] Simplify test case by removing unreferenced instructions [NFC]
Drop instructions that do not influence the memory impact of a basic block.
They are not needed to reproduce the original bug (verified) and will cause
random test noise if we would decide to only model the instructions that
have visible side-effects.
llvm-svn: 287626
Tobias Grosser [Tue, 22 Nov 2016 07:06:59 +0000 (07:06 +0000)]
[test] Ensure important basic blocks in test case have side effects
Add two store instructions at the end of basic blocks that are required to
reproduce the original bug to ensure we always process and model these basic
blocks. This makes this test case stable even in case we would decide to bail
out early of basic blocks which do not modify the global state. Also add
additional check lines to verify how we model the basic block.
llvm-svn: 287625
Craig Topper [Tue, 22 Nov 2016 07:00:06 +0000 (07:00 +0000)]
[TableGen][ISel] When factoring ScopeMatcher, if the child of the ScopeMatcher we're working on is also a ScopeMatcher, merge all its children into the one we're working on.
There were several cases in X86 where we were unable to fully factor a ScopeMatcher but created nested ScopeMatchers for some portions of it. Then we created a SwitchType that split it up and further factored it so that we ended up with something like this:
SwitchType
Scope
Scope
Sequence of matchers
Some other sequence of matchers
EndScope
Another sequence of matchers
EndScope
...Next type
This change turns it into this:
SwitchType
Scope
Sequence of matchers
Some other sequence of matchers
Another sequence of matchers
EndScope
...Next type
Several other in-tree targets had similar nested scopes like this. Overall this doesn't save many bytes, but makes the isel output a little more regular.
llvm-svn: 287624
Tobias Grosser [Tue, 22 Nov 2016 06:28:08 +0000 (06:28 +0000)]
test: add more details to non-affine test case
We add CHECK lines to this test case to make it easier to see the difference
between affine and non-affine memory accesses. We also change the test case to
use a parameteric index expression as otherwise our range analysis will
understand that the non-affine memory access can only access input[1],
which makes it difficult to see that the memory access is in-fact modeled as
non-affine access.
llvm-svn: 287623
Craig Topper [Tue, 22 Nov 2016 05:31:43 +0000 (05:31 +0000)]
[X86] Remove alternate CodeGenOnly version of (v)movq that declared the load size as i128mem. Change all uses to the use the i64mem version.
I'm sure this caused the load size to misprint in Intel syntax output. We were also inconsistent about which patterns used which instruction between VEX and EVEX.
There are two different reg/reg versions of movq, one from a GPR and one from the lower 64-bits of an XMM register. This changes the loading folding table to use the single i64mem memory form for folding both cases. But we need to use TB_NO_REVERSE to prevent a duplicate entry in the unfolding table.
llvm-svn: 287622
Craig Topper [Tue, 22 Nov 2016 04:57:34 +0000 (04:57 +0000)]
[AVX-512] Add support for commuting VPERMT2(B/W/D/Q/PS/PD) to/from VPERMI2(B/W/D/Q/PS/PD).
Summary:
The index and one of the table operands can be swapped by changing the opcode to the other version. Neither of these operands are the one that can load from memory so this can't be used to increase memory folding opportunities.
We need to handle the unmasked forms and the kz forms. Since the load operand isn't being commuted we can commute the load and broadcast instructions too.
Reviewers: igorb, delena, Ayal, Farhana, RKSimon
Subscribers: llvm-commits
Differential Revision: https://reviews.llvm.org/D25652
llvm-svn: 287621
Rui Ueyama [Tue, 22 Nov 2016 04:33:01 +0000 (04:33 +0000)]
Inline small function. NFC.
llvm-svn: 287620
Saleem Abdulrasool [Tue, 22 Nov 2016 04:32:54 +0000 (04:32 +0000)]
MC: ensure that we have a section before accessing it
We would attempt to access the symbol section without ensuring that the symbol
was not absolute. When the assembler referenced relocation is not evaluated to
the absolute, but when we record the relocation, we would query the section.
Because the symbol is absolute, it does not have a section associated with it,
triggering an assertion. Just be more careful about the access of the section.
Addresses PR31064!
llvm-svn: 287619
Artem Dergachev [Tue, 22 Nov 2016 04:29:23 +0000 (04:29 +0000)]
[analyzer] Fix a crash on accessing a field within a literal-initialized union.
Because in case of unions we currently default-bind compound values in the
store, this quick fix avoids the crash for this case.
Patch by Ilya Palachev and independently by Alexander Shaposhnikov!
Differential Revision: https://reviews.llvm.org/D26442
llvm-svn: 287618
Rui Ueyama [Tue, 22 Nov 2016 04:28:39 +0000 (04:28 +0000)]
Do not save unused pointers to In<ELFT>.
llvm-svn: 287617
Rui Ueyama [Tue, 22 Nov 2016 04:17:12 +0000 (04:17 +0000)]
Remove default definition no one uses.
llvm-svn: 287616
Rui Ueyama [Tue, 22 Nov 2016 04:13:09 +0000 (04:13 +0000)]
Convert MipsOptionsSection to SyntheticSection.
llvm-svn: 287615
Rui Ueyama [Tue, 22 Nov 2016 03:57:08 +0000 (03:57 +0000)]
Convert MipsReginfoSection to SyntheticSection.
llvm-svn: 287614
Rui Ueyama [Tue, 22 Nov 2016 03:57:06 +0000 (03:57 +0000)]
Convert MipsAbiFlagsSection to SyntheticSection.
llvm-svn: 287613
Craig Topper [Tue, 22 Nov 2016 03:51:53 +0000 (03:51 +0000)]
[AVX-512] Add support for changing the element size of PALIGNR/VALIGND/VALIGNQ shuffles if they feed a vselect with a different type
Summary:
Shuffle lowering widens the element size of a shuffle if elements are contiguous. This is sometimes help because wider element types have more shuffle options. If the shuffle is one of the arguments to a vselect this shuffle widening can introduce a bitcast between the vselect and the shuffle. This will prevent isel from selecting a masked operation. If the shuffle can be written equally efficiently with a different element size to match the vselect type we should change the shuffle type to allow masking.
This patch does this conversion for all VALIGND/VALIGNQ sizes. It also supports turning 128-bit PALIGNR into VALIGND/VALIGNQ. This fixes the case shown in PR31018.
I plan to add support for more operations in future patches.
Reviewers: RKSimon, zvi, delena
Subscribers: llvm-commits
Differential Revision: https://reviews.llvm.org/D26902
llvm-svn: 287612
Peter Collingbourne [Tue, 22 Nov 2016 03:38:40 +0000 (03:38 +0000)]
Object: Make SymbolicFile::symbol_{begin,end}() virtual and remove unnecessary wrappers.
llvm-svn: 287611
Chandler Carruth [Tue, 22 Nov 2016 03:27:43 +0000 (03:27 +0000)]
[ADT] Add initializer list support to SmallPtrSet so that sets can be
easily initialized with some initial values.
llvm-svn: 287610
Ed Maste [Tue, 22 Nov 2016 02:25:41 +0000 (02:25 +0000)]
Turn on LLDB_EDITLINE_USE_WCHAR on FreeBSD
llvm-svn: 287609
Stanislav Mekhanoshin [Tue, 22 Nov 2016 01:42:34 +0000 (01:42 +0000)]
[AMDGPU] Fix multiple vreg definitions in si-lower-control-flow
Differential Revision: https://reviews.llvm.org/D26939
llvm-svn: 287608
Rui Ueyama [Tue, 22 Nov 2016 01:36:19 +0000 (01:36 +0000)]
Remove redundant assignment.
llvm-svn: 287607
Rui Ueyama [Tue, 22 Nov 2016 01:31:32 +0000 (01:31 +0000)]
Convert BuildId a derived class of SyntheticSection.
Some synthetic sections are not derived calsses of SyntehticSection.
They are derived directly from InputSection. For consistencly, we should
use SyntheticSection.
llvm-svn: 287606
Rui Ueyama [Tue, 22 Nov 2016 01:10:34 +0000 (01:10 +0000)]
Remove a parameter from getOutputLoc and rename for readability. NFC.
llvm-svn: 287605
Peter Collingbourne [Tue, 22 Nov 2016 01:03:40 +0000 (01:03 +0000)]
Analysis: gep inbounds (gep inbounds (...)) is inbounds.
Differential Revision: https://reviews.llvm.org/D26441
llvm-svn: 287604
Rui Ueyama [Tue, 22 Nov 2016 00:54:15 +0000 (00:54 +0000)]
Merge BuildId subclasses.
We had five different BuildId subclasses for five different types
of build-ids. They can simply be merged to a single class.
llvm-svn: 287603
Peter Collingbourne [Tue, 22 Nov 2016 00:43:30 +0000 (00:43 +0000)]
Fix -Asserts build, and add some more test cases.
llvm-svn: 287602
Peter Collingbourne [Tue, 22 Nov 2016 00:21:43 +0000 (00:21 +0000)]
Sema, CodeGen: Ensure that an implicit copy ctor is available more often under the Microsoft C++ ABI.
This is needed because whether the constructor is deleted can control whether
we pass structs by value directly.
To fix this properly we probably want a more direct way for CodeGen to ask
whether the constructor was deleted.
Fixes PR31049.
Differential Revision: https://reviews.llvm.org/D26822
llvm-svn: 287600
Richard Smith [Mon, 21 Nov 2016 23:43:54 +0000 (23:43 +0000)]
Indicate in AST dump whether special member functions are defaulted and trivial.
llvm-svn: 287599
Zachary Turner [Mon, 21 Nov 2016 23:18:13 +0000 (23:18 +0000)]
Fix a bug caught by adding LLVM_NODISCARD to StringRef.
llvm-svn: 287598
Zachary Turner [Mon, 21 Nov 2016 23:18:07 +0000 (23:18 +0000)]
Add the new Args / entry-access API.
The long-term goal here is to get rid of the functions
GetArgumentAtIndex() and GetQuoteCharAtIndex(), instead
replacing them with operator based access and range-based for
enumeration. There are a lot of callsites, though, so the
changes will be done incrementally, starting with this one.
Differential Revision: https://reviews.llvm.org/D26883
llvm-svn: 287597
Zachary Turner [Mon, 21 Nov 2016 23:17:15 +0000 (23:17 +0000)]
Remove LLVM_NODISCARD in one more place.
llvm-svn: 287596
Rui Ueyama [Mon, 21 Nov 2016 23:17:09 +0000 (23:17 +0000)]
Remove useless newlines.
llvm-svn: 287595
Zachary Turner [Mon, 21 Nov 2016 23:02:28 +0000 (23:02 +0000)]
Remove LLVM_NODISCARD from two more StringRef members.
This should be everything.
llvm-svn: 287594
Matt Arsenault [Mon, 21 Nov 2016 22:56:42 +0000 (22:56 +0000)]
DAG: Ignore call site attributes when emitting target intrinsic
A target intrinsic may be defined as possibly reading memory,
but the call site may have additional knowledge that it doesn't read
memory. The intrinsic lowering will expect the pessimistic
assumption of the intrinsic definition, so the chain should
still be used.
llvm-svn: 287593
Geoff Berry [Mon, 21 Nov 2016 22:51:10 +0000 (22:51 +0000)]
[AArch64LoadStoreOptimizer] Don't treat write to XZR/WZR as a clobber.
Summary:
When searching for load/store instructions to pair/merge don't treat
writes to WZR/XZR as clobbers since they don't change the value read
from WZR/XZR (which is always 0).
Reviewers: mcrosier, junbuml, jmolloy, t.p.northover
Subscribers: aemerson, llvm-commits, rengolin
Differential Revision: https://reviews.llvm.org/D26921
llvm-svn: 287592