Kevin Enderby [Fri, 21 Oct 2016 18:22:35 +0000 (18:22 +0000)]
For llvm-objdump for Mach-O files add printing of
the ARM_THREAD_STATE in the same format as
otool-classic(1) on darwin.
Also remove an extra space in printing the initprot to make
the output match otool-classic(1) on darwin.
rdar://
28851457
llvm-svn: 284852
Jim Ingham [Fri, 21 Oct 2016 18:18:25 +0000 (18:18 +0000)]
Add the new stdcpp formatters to the Xcode project.
llvm-svn: 284851
Eli Friedman [Fri, 21 Oct 2016 18:08:02 +0000 (18:08 +0000)]
[SCEVAffinator] Make precise modular math more correct.
Integer math in LLVM IR is modular. Integer math in isl is
arbitrary-precision. Modeling LLVM IR math correctly in isl requires
either adding assumptions that math doesn't actually overflow, or
explicitly wrapping the math. However, expressions with the "nsw" flag
are special; we can pretend they're arbitrary-precision because it's
undefined behavior if the result wraps. SCEV expressions based on IR
instructions with an nsw flag also carry an nsw flag (roughly; actually,
the real rule is a bit more complicated, but the details don't matter
here).
Before this patch, SCEV flags were also overloaded with an additional
function: the ZExt code was mutating SCEV expressions as a hack to
indicate to checkForWrapping that we don't need to add assumptions to
the operand of a ZExt; it'll add explicit wrapping itself. This kind of
works... the problem is that if anything else ever touches that SCEV
expression, it'll get confused by the incorrect flags.
Instead, with this patch, we make the decision about whether to
explicitly wrap the math a bit earlier, basing the decision purely on
the SCEV expression itself, and not its users.
Differential Revision: https://reviews.llvm.org/D25287
llvm-svn: 284848
Enrico Granata [Fri, 21 Oct 2016 18:03:55 +0000 (18:03 +0000)]
I hadn't fixed all the instances of the old marker for scripted format strings; do so now
llvm-svn: 284847
Rui Ueyama [Fri, 21 Oct 2016 17:50:26 +0000 (17:50 +0000)]
Use FileCheck's -strict-whitespace option instead of sed.
llvm-svn: 284846
Mandeep Singh Grang [Fri, 21 Oct 2016 17:29:10 +0000 (17:29 +0000)]
[polly] Change SmallPtrSet which are being iterated into SmallSetVector
Summary: Otherwise the lack of an iteration order results in non-determinism in codegen.
Reviewers: _jdoerfert, zinob, grosser
Tags: #polly
Differential Revision: https://reviews.llvm.org/D25863
llvm-svn: 284845
Sanjay Patel [Fri, 21 Oct 2016 17:24:26 +0000 (17:24 +0000)]
[DAG] fold negation of sign-bit
0 - X --> 0, if the sub is NUW
0 - X --> 0, if X is 0 or the minimum signed value and the sub is NSW
0 - X --> X, if X is 0 or the minimum signed value
This is the DAG equivalent of:
https://reviews.llvm.org/rL284649
plus the fold for the NUW case which already existed in InstSimplify.
Note that we miss a vector fold because of a deficiency in the DAG version of
computeKnownBits().
llvm-svn: 284844
Artem Belevich [Fri, 21 Oct 2016 17:15:46 +0000 (17:15 +0000)]
Removed unused function argument. NFC.
Differential Revision: https://reviews.llvm.org/D25839
llvm-svn: 284843
Sanjay Patel [Fri, 21 Oct 2016 16:56:29 +0000 (16:56 +0000)]
[x86] add tests for potential negation folds
These are the backend equivalents for the tests added in r284627.
The patterns may emerge late, so we should have folds for these in the DAG too.
llvm-svn: 284842
Krzysztof Parzyszek [Fri, 21 Oct 2016 16:38:29 +0000 (16:38 +0000)]
[Hexagon] Handle spills of partially defined double vector registers
After register allocation it is possible to have a spill of a register
that is only partially defined. That in itself it fine, but creates a
problem for double vector registers. Stores of such registers are pseudo
instructions that are expanded into pairs of individual vector stores,
and in case of a partially defined source, one of the stores may use
an entirely undefined register. To avoid this, track the defined parts
and only generate actual stores for those.
llvm-svn: 284841
Derek Schuff [Fri, 21 Oct 2016 16:38:07 +0000 (16:38 +0000)]
[WebAssembly] Fix for 0xc call_indirect changes
Summary:
Need to reorder the operands to have the callee as the last argument.
Adds a pseudo-instruction, and a pass to lower it into a real
call_indirect.
This is the first of two options for how to fix the problem.
Reviewers: dschuff, sunfish
Subscribers: jfb, beanz, mgorny, llvm-commits
Differential Revision: https://reviews.llvm.org/D25708
llvm-svn: 284840
Abderrazek Zaafrani [Fri, 21 Oct 2016 16:28:27 +0000 (16:28 +0000)]
Set the vectorizer MaxInterleaveFactor for Exynos.
llvm-svn: 284839
Reid Kleckner [Fri, 21 Oct 2016 16:26:27 +0000 (16:26 +0000)]
Fix -Wunused-variable warning in libFuzzer
llvm-svn: 284838
Simon Pilgrim [Fri, 21 Oct 2016 16:22:16 +0000 (16:22 +0000)]
[X86][SSE] Regenerated sext/zext constant folding tests and added i686 tests
llvm-svn: 284837
Reid Kleckner [Fri, 21 Oct 2016 16:09:20 +0000 (16:09 +0000)]
Remove unnecessary x86 backend requirements from OpenMP tests
Clang can generate LLVM IR for x86 without a registered x86 backend.
llvm-svn: 284836
Simon Pilgrim [Fri, 21 Oct 2016 16:07:51 +0000 (16:07 +0000)]
[X86] Use DAG::getBuildVector helper wrapper where possible. NFCI.
llvm-svn: 284835
Konstantin Zhuravlyov [Fri, 21 Oct 2016 16:02:35 +0000 (16:02 +0000)]
[MachineMemOperand][AtomicSDNode] Remove getSuccessOrdering()
Differential Revision: https://reviews.llvm.org/D25786
llvm-svn: 284834
Simon Pilgrim [Fri, 21 Oct 2016 15:51:24 +0000 (15:51 +0000)]
[X86][SSE] Regenerated chained pmovsx store tests and added i686 tests
llvm-svn: 284833
Abderrazek Zaafrani [Fri, 21 Oct 2016 15:24:08 +0000 (15:24 +0000)]
Test commit
llvm-svn: 284832
Tamas Berghammer [Fri, 21 Oct 2016 15:05:03 +0000 (15:05 +0000)]
Fix incorrect header order introduced in rL284830
llvm-svn: 284831
Tamas Berghammer [Fri, 21 Oct 2016 15:02:44 +0000 (15:02 +0000)]
Add data formatter for libstdc++ unique_ptr
Differential revision: https://reviews.llvm.org/D25734
llvm-svn: 284830
Tamas Berghammer [Fri, 21 Oct 2016 15:02:38 +0000 (15:02 +0000)]
Add data formatter for libstdc++ tuple
Differential revision: https://reviews.llvm.org/D25733
llvm-svn: 284829
Tamas Berghammer [Fri, 21 Oct 2016 15:02:32 +0000 (15:02 +0000)]
Improve the libstdc++ smart pointer formatters
* Display the strong/weak count in the summary
* Display the pointed object as a synthetic member
* Create synthetic children for weak/strong count
Differential revision: https://reviews.llvm.org/D25726
llvm-svn: 284828
Artur Pilipenko [Fri, 21 Oct 2016 15:02:21 +0000 (15:02 +0000)]
[LVI] Fix a bug with a guard being the very first instruction in a BB not taken into account
While looking for guards use reverse iterator and scan up to rend() not to begin()
llvm-svn: 284827
Sanjay Patel [Fri, 21 Oct 2016 14:58:30 +0000 (14:58 +0000)]
fix variable names; NFCI
Because we're just 'or-ing' these 2 variables later in the code, I
don't think there's a logical bug here, but of course the string with
"no size" is the one that should have the size suffix stripped off.
llvm-svn: 284826
Artem Tamazov [Fri, 21 Oct 2016 14:49:22 +0000 (14:49 +0000)]
[AMDGPU][mc] Fix ds_min/max[_rtn]_f32 - extra source operand removed.
Fixes Bug 28215. Lit tests updated.
Differential Revision: https://reviews.llvm.org/D25837
llvm-svn: 284825
Sanjay Patel [Fri, 21 Oct 2016 14:36:58 +0000 (14:36 +0000)]
[DAG] use SDNode flags 'nsz' to enable fadd/fsub with zero folds
As discussed in D24815, let's start the process of killing off the broken fast-math global
state housed in TargetOptions and eliminate the need for function-level fast-math attributes.
Here we enable two similar folds that are possible when we don't care about signed-zero:
fadd nsz x, 0 --> x
fsub nsz 0, x --> -x
Note that although the test cases include a 'sin' function call, I'm side-stepping the
FMF-on-calls question (and lack of support in the DAG) for now. It's not needed for these
tests - isNegatibleForFree/GetNegatedExpression just look through a ISD::FSIN node.
Also, when we create an FNEG node and propagate the Flags of the FSUB to it, this doesn't
actually do anything today because Flags are silently dropped for any node that is not a
binary operator.
Differential Revision: https://reviews.llvm.org/D25297
llvm-svn: 284824
Simon Pilgrim [Fri, 21 Oct 2016 13:00:47 +0000 (13:00 +0000)]
[X86][AVX2] Begun generalizing lowering to VPERMD/VPERMPS in preparation for AVX512 support.
llvm-svn: 284823
Simon Pilgrim [Fri, 21 Oct 2016 12:51:16 +0000 (12:51 +0000)]
Wdocumentation fix
llvm-svn: 284822
Simon Pilgrim [Fri, 21 Oct 2016 12:14:24 +0000 (12:14 +0000)]
[X86][AVX512] Add mask/maskz writemask support to subvector broadcast shuffle decode comments
llvm-svn: 284821
Peter Smith [Fri, 21 Oct 2016 11:25:33 +0000 (11:25 +0000)]
[ELF] Add sh_link field to .ARM.exidx sections for relocatable links
When doing a relocatable link the .ARM.exidx sections with the
SHF_LINK_ORDER flag set need to set the sh_link field to the executable
section they describe. We find the appropriate OutputSection by
following the sh_link field of the .ARM.exidx InputSections.
The getOutputSectionName() function rules make sure that when there are
multiple .ARM.exidx InputSections in an OutputSection they all have the
same sh_link field.
Differential revision: https://reviews.llvm.org/D25825
llvm-svn: 284820
Pavel Labath [Fri, 21 Oct 2016 11:14:04 +0000 (11:14 +0000)]
Add TestMultipleHits.py
This tests that lldb handles the situation when a single instruction triggers
multiple watchpoint hits. It currently fails on arm due to what appears to be a
lldb-server bug (pr30758).
llvm-svn: 284819
John Brawn [Fri, 21 Oct 2016 11:08:48 +0000 (11:08 +0000)]
[LoopUnroll] Keep the loop test only on the first iteration of max-or-zero loops
When we have a loop with a known upper bound on the number of iterations, and
furthermore know that either the number of iterations will be either exactly
that upper bound or zero, then we can fully unroll up to that upper bound
keeping only the first loop test to check for the zero iteration case.
Most of the work here is in plumbing this 'max-or-zero' information from the
part of scalar evolution where it's detected through to loop unrolling. I've
also gone for the safe default of 'false' everywhere but howManyLessThans which
could probably be improved.
Differential Revision: https://reviews.llvm.org/D25682
llvm-svn: 284818
Pavel Labath [Fri, 21 Oct 2016 10:52:11 +0000 (10:52 +0000)]
Revert "Fix a race condition between "ephemeral watchpoint disable/enable" and continue in commands."
This reverts commit r284795, as it breaks watchpoint handling on arm (and
presumable all architectures that report watchpoint hits without executing the
tripping instruction).
There seems to be something fundamentally wrong with this patch: it uses
process_sp->AddPreResumeAction to re-enable the watchpoint, but the whole point
of the step-over-watchpoint logic (which AFAIK is the only user of this class) is
to disable the watchpoint *after* we resume to do the single step.
I have no idea how to fix this except by reverting the offending patch.
llvm-svn: 284817
Simon Pilgrim [Fri, 21 Oct 2016 10:50:52 +0000 (10:50 +0000)]
[X86][AVX] Add 32-bit target tests for vector lzcnt/tzcnt to demonstrate missed folding opportunities
llvm-svn: 284816
Benjamin Kramer [Fri, 21 Oct 2016 10:19:37 +0000 (10:19 +0000)]
Remove non-existing file from modulemap.
This picked up a builtin header if it happened to be available.
llvm-svn: 284815
Bjorn Pettersson [Fri, 21 Oct 2016 09:53:42 +0000 (09:53 +0000)]
[AArch64] Corrected spill size for DDD register class. NFCI
Summary:
The spill size was incorrectly set to 196 bits,
which isn't a multiple of 8. This problem was detected when
experimenting with asserts that the spill size should be a
multiple of the byte size.
New corrected value for the spill size is set to 192 bits.
Note that tablegen (RegisterInfoEmitter) will divide the
size set in the RegisterClass definition by 8. So this
change should not have any impact on the tablegen output
(trunc(192/8) == trunc(196/8) == 24 bytes).
Reviewers: t.p.northover
Subscribers: llvm-commits, aemerson, rengolin
Differential Revision: https://reviews.llvm.org/D25818
llvm-svn: 284814
Benjamin Kramer [Fri, 21 Oct 2016 09:51:41 +0000 (09:51 +0000)]
Fix WebAssembly test after r284757.
The change to MachineSink shuffles code around, disable it.
llvm-svn: 284813
Benjamin Kramer [Fri, 21 Oct 2016 09:15:57 +0000 (09:15 +0000)]
[Support] Fix AlignOf test on i386-linux.
On i386 alignof(double) = 8 is not the same as alignof(struct { double
}) = 4. This used to be not an issue because the old implementation
always measured alignment inside of structs. Wrap a dummy struct around
the test to avoid this issue.
llvm-svn: 284812
Renato Golin [Fri, 21 Oct 2016 08:03:49 +0000 (08:03 +0000)]
Revert "DR583, DR1512: Implement a rewrite to C++'s 'composite pointer type' rules."
This reverts commit r284800, as it failed all ARM/AArch64 bots.
llvm-svn: 284811
George Rimar [Fri, 21 Oct 2016 07:46:24 +0000 (07:46 +0000)]
[ELF] Add DebugInfoDWARF dependency
rL284708 introduces a link error when building with BUILD_SHARED_LIBS:
undefined reference to `llvm::DWARFContext::parseCompileUnits()'
undefined reference to `llvm::DWARFContextInMemory::DWARFContextInMemory(
llvm::object::ObjectFile const&,
llvm::LoadedObjectInfo const*)'
The functions are available in libDebugInfoDWARF, from llvm.
Patch by Visoiu Mistrih Francis
Differential revision: https://reviews.llvm.org/D25843
llvm-svn: 284810
Simon Atanasyan [Fri, 21 Oct 2016 07:22:30 +0000 (07:22 +0000)]
[ELF][MIPS] Put local GOT entries accessed via a 16-bit index first
Some MIPS relocations used to access GOT entries are able to manipulate
16-bit index. The other ones like R_MIPS_CALL_HI16/LO16 can handle
32-bit indexes. 16-bit relocations are generated by default. The 32-bit
relocations are generated by -mxgot flag passed to compiler. Usually
these relocation are not mixed in the same code but files like crt*.o
contain 16-bit relocations so even if all "user's" code compiled with
-mxgot flag a few 16-bit relocations might come to the linking phase.
Now LLD does not differentiate local GOT entries accessed via a 16-bit
and 32-bit indexes. That might lead to relocation's overflow if 16-bit
entries are allocated to far from the beginning of the GOT.
The patch introduces new "part" of MIPS GOT dedicated to the local GOT
entries accessed by 32-bit relocations. That allows to put local GOT
entries accessed via a 16-bit index first and escape relocation's overflow.
Differential revision: https://reviews.llvm.org/D25833
llvm-svn: 284809
Craig Topper [Fri, 21 Oct 2016 05:55:40 +0000 (05:55 +0000)]
[AVX-512] Add tests to show opportunities for commuting vpermi2/vpermt2 instructions.
Commuting will be added in a future commit.
llvm-svn: 284808
Martin Probst [Fri, 21 Oct 2016 05:11:38 +0000 (05:11 +0000)]
clang-format: [JS] Fix template string ASI.
Summary:
Previously, automatic semicolon insertion would add an unwrapped line
when a template string contained a line break.
var x = `foo${
bar}`;
Would be formatted with `bar...` on a separate line and no indent.
Reviewers: djasper
Subscribers: cfe-commits, klimek
Differential Revision: https://reviews.llvm.org/D25675
llvm-svn: 284807
Rui Ueyama [Fri, 21 Oct 2016 04:52:13 +0000 (04:52 +0000)]
Simplify. NFC.
llvm-svn: 284806
Rui Ueyama [Fri, 21 Oct 2016 04:52:11 +0000 (04:52 +0000)]
Add comments.
llvm-svn: 284805
Rui Ueyama [Fri, 21 Oct 2016 04:32:46 +0000 (04:32 +0000)]
Simplify by merging a lambda with addSymbol. NFC.
llvm-svn: 284804
Rui Ueyama [Fri, 21 Oct 2016 03:23:23 +0000 (03:23 +0000)]
Do not create .shstrtab and use .strtab instead.
In-memory ELF object files created here are consumed immediately,
so as long as the generated files are syntactically correct, we don't
care about the details.
llvm-svn: 284803
Richard Smith [Fri, 21 Oct 2016 03:15:03 +0000 (03:15 +0000)]
Don't try to use !Previous.empty() as a proxy for "Is this a redeclaration?" --
we don't collapse that down to a single entry if it's not a redeclaration.
Instead, set the Redeclaration bit on the Declarator to indicate whether a
function is a redeclaration (which may not have been linked into the
redeclaration chain if it's a dependent context friend).
Fixes a rejects-valid; see testcase.
llvm-svn: 284802
Bruno Cardoso Lopes [Fri, 21 Oct 2016 03:14:27 +0000 (03:14 +0000)]
[ModuleMap] Add more module entries to cover some non modular headers
These modules are necessary on Darwin to allow modules with
'no_undeclared_includes' (introduced in clang r284797) to work properly
while using libc++ headers.
Patch extracted from a suggested module.modulemap from Richard Smith!
llvm-svn: 284801
Richard Smith [Fri, 21 Oct 2016 02:36:37 +0000 (02:36 +0000)]
DR583, DR1512: Implement a rewrite to C++'s 'composite pointer type' rules.
This has two significant effects:
1) Direct relational comparisons between null pointer constants (0 and nullopt)
and pointers are now ill-formed. This was always the case for C, and it
appears that C++ only ever permitted by accident. For instance, cases like
nullptr < &a
are now rejected.
2) Comparisons and conditional operators between differently-cv-qualified
pointer types now work, and produce a composite type that both source
pointer types can convert to (when possible). For instance, comparison
between 'int **' and 'const int **' is now valid, and uses an intermediate
type of 'const int *const *'.
Clang previously supported #2 as an extension.
We do not accept the cases in #1 as an extension. I've tested a fair amount of
code to check that this doesn't break it, but if it turns out that someone is
relying on this, we can easily add it back as an extension.
llvm-svn: 284800
Jason Molenda [Fri, 21 Oct 2016 02:32:08 +0000 (02:32 +0000)]
Guard GetEscapedHostname against a nullptr hostname.
This can happen if you debug an iOS corefile on
a mac, where PlatformPOSIX::GetHostname ends up
not providing a hostname because we're working
with a platform of remote-ios.
llvm-svn: 284799
Bruno Cardoso Lopes [Fri, 21 Oct 2016 01:49:14 +0000 (01:49 +0000)]
[Driver][Darwin] Pass -no_deduplicate to ld64
Recent versions of ld64 run a deduplicate pass, which is on by default.
Disable the pass by using -no_deduplicate in certain condition and
enhance total compile time.
rdar://problem/
25455336
llvm-svn: 284798
Bruno Cardoso Lopes [Fri, 21 Oct 2016 01:41:56 +0000 (01:41 +0000)]
[Modules] Add 'no_undeclared_includes' module map attribute
The 'no_undeclared_includes' attribute should be used in a module to
tell that only non-modular headers and headers from used modules are
accepted.
The main motivation behind this is to prevent dep cycles between system
libraries (such as darwin) and libc++.
Patch by Richard Smith!
llvm-svn: 284797
Davide Italiano [Fri, 21 Oct 2016 01:37:02 +0000 (01:37 +0000)]
Revert "[GVN/PRE] Hoist global values outside of loops."
There's no agreement about this patch. I personally find the
PRE machinery of the current GVN hard enough to reason about
that I'm not sure I'll try to land this again, instead of working
on the rewrite).
llvm-svn: 284796
Jim Ingham [Fri, 21 Oct 2016 00:06:38 +0000 (00:06 +0000)]
Fix a race condition between "ephemeral watchpoint disable/enable" and continue in commands.
Also, watchpoint commands, like breakpoint commands, need to run in async mode.
This was causing intermittent failures in TestWatchpointCommandPython.py, which is now solid.
llvm-svn: 284795
Reid Kleckner [Thu, 20 Oct 2016 23:30:39 +0000 (23:30 +0000)]
Revert 9 changes from r284793, they still fail on some bots
llvm-svn: 284794
Reid Kleckner [Thu, 20 Oct 2016 23:11:45 +0000 (23:11 +0000)]
Remove 24 instances of 'REQUIRES: shell'
Tests fall into one of the following categories:
- The requirement was unnecessary
- Additional quoting was required for backslashes in paths (see "sed -e
's/\\/\\\\/g'") in the sanitizer tests.
- OpenMP used 'REQUIRES: shell' as a proxy for the test failing on
Windows. Those tests fail there reliably, so use XFAIL instead.
I tried not to remove shell requirements that were added to suppress
flaky test failures, but if I screwed up, we can add it back as needed.
llvm-svn: 284793
Jim Ingham [Thu, 20 Oct 2016 22:50:00 +0000 (22:50 +0000)]
Add an API to remove an action from the Process PreResumeActions.
llvm-svn: 284792
Jim Ingham [Thu, 20 Oct 2016 22:49:06 +0000 (22:49 +0000)]
Remove an unnecessary and incorrect check for num locations of a breakpoint
by grubbing the break list output. If you pass a number of locations into
the run_break_* functions, they will check that this is right for you.
llvm-svn: 284791
Rui Ueyama [Thu, 20 Oct 2016 22:29:22 +0000 (22:29 +0000)]
Attempt to fix buildbots.
llvm-svn: 284790
Keno Fischer [Thu, 20 Oct 2016 22:15:56 +0000 (22:15 +0000)]
Fix cross-endianness RuntimeDyld relocation for ARM
rL284780 fixed the PREL31 relocation and added a test for it. Being
the first such test for ARM relocations, it exposed incorrect endianness
assumptions (causing buildbot failures on big-endian hosts). Fix that by
using the same helpers used for the x86 case.
llvm-svn: 284789
Enrico Granata [Thu, 20 Oct 2016 22:10:07 +0000 (22:10 +0000)]
frame.script became script.frame a while ago; fix up the docs
llvm-svn: 284788
Enrico Granata [Thu, 20 Oct 2016 22:05:21 +0000 (22:05 +0000)]
This debugging message has been left in the code for years, until one day it randomly hit on some corrupted memory
It is misleading to users in its current form, and only interesting to me - remove it
rdar://
28812568
llvm-svn: 284787
Rui Ueyama [Thu, 20 Oct 2016 21:57:06 +0000 (21:57 +0000)]
Move code from InputFile to ELFCreator to decouple the two files.
Now that only one non-member function is exported from ELFCreator.h.
All the details are handled internally in ELFCreator.cpp file.
llvm-svn: 284786
Richard Smith [Thu, 20 Oct 2016 21:53:09 +0000 (21:53 +0000)]
Re-commit r284753, reverted in r284778, with a fix for PR30749.
Original commit message:
[c++1z] Teach composite pointer type computation how to compute the composite
pointer type of two function pointers with different noexcept specifications.
While I'm here, also teach it how to merge dynamic exception specifications.
llvm-svn: 284785
Li Huang [Thu, 20 Oct 2016 21:38:39 +0000 (21:38 +0000)]
[SCEV] Add a threshold to restrict number of mul operands to be inlined into SCEV
This is to avoid inlining too many multiplication operands into a SCEV, which could
take exponential time in the worst case.
Reviewers: Sanjoy Das, Mehdi Amini, Michael Zolotukhin
Differential Revision: https://reviews.llvm.org/D25794
llvm-svn: 284784
Alexander Shaposhnikov [Thu, 20 Oct 2016 21:20:35 +0000 (21:20 +0000)]
[clang] Remove FileEntry copy-constructor
Code cleanup: address FIXME in the file
include/clang/Basic/FileManager.h and remove
copy-constructor of the class FileEntry.
Test plan: make check-clang
Differential revision: https://reviews.llvm.org/D22712
llvm-svn: 284782
Reid Kleckner [Thu, 20 Oct 2016 21:17:28 +0000 (21:17 +0000)]
Revert "Disable swiftcall test on windows: More brutal way to appease windows bots"
This reverts commit r284174. The tests pass for me locally. It must have
been a 2015 only crash.
Fixes PR30699
llvm-svn: 284781
Keno Fischer [Thu, 20 Oct 2016 21:15:29 +0000 (21:15 +0000)]
Fix PREL31 relocation on ARM
Summary:
This is a 31bits relative relocation instead of a 32bits absolute relocation.
Reviewers: t.p.northover, peter.smith, rengolin
Subscribers: aemerson, llvm-commits, samparker
Differential Revision: https://reviews.llvm.org/D25069
llvm-svn: 284780
Michael Kuperstein [Thu, 20 Oct 2016 21:04:31 +0000 (21:04 +0000)]
[X86] Enable interleaved memory access by default
This lets the loop vectorizer generate interleaved memory accesses on x86.
Differential Revision: https://reviews.llvm.org/D25350
llvm-svn: 284779
Hans Wennborg [Thu, 20 Oct 2016 20:54:32 +0000 (20:54 +0000)]
Revert r284753 "[c++1z] Teach composite pointer type computation how to compute the composite"
It caused PR30749.
llvm-svn: 284778
Reid Kleckner [Thu, 20 Oct 2016 20:53:20 +0000 (20:53 +0000)]
Fix off-by-one error in PPCaching.cpp token annotation assertion
This assert is intended to defend against backtracking into the middle
of a sequence of tokens that is being replaced with an annotation, but
it's OK if we backtrack to the exact position of the start of the
annotation sequence. Use a <= comparison instead of <.
Fixes PR25946
llvm-svn: 284777
Rui Ueyama [Thu, 20 Oct 2016 20:50:00 +0000 (20:50 +0000)]
llvm-dwarfdump is now used in a test, so add a depedency on the tool.
llvm-svn: 284776
Michal Gorny [Thu, 20 Oct 2016 20:45:40 +0000 (20:45 +0000)]
[Driver] Refactor DetectDistro() parameters to take VFS ref only. NFC
Refactor the DetectDistro() function to take a single vfs::FileSystem
reference only, instead of Driver and llvm::Triple::ArchType.
The ArchType parameter was not used anyway, and Driver was only used to
obtain the VFS.
Aside to making the API simpler and more transparent, it makes it
easier to add unit tests for the function in the future -- since
the tests would need only to provide an appropriate VFS.
Differential Revision: https://reviews.llvm.org/D25819
llvm-svn: 284774
Mike Aizatsky [Thu, 20 Oct 2016 20:33:54 +0000 (20:33 +0000)]
adding new symbols to llvm-symbolizer symbol test
llvm-svn: 284773
Daniel Berlin [Thu, 20 Oct 2016 20:13:45 +0000 (20:13 +0000)]
[MSSA] Avoid unnecessary use walks when calling getClobberingMemoryAccess
Summary:
This allows us to mark when uses have been optimized.
This lets us avoid rewalking (IE when people call getClobberingAccess on everything), and also
enables us to later relax the requirement of use optimization during updates with less cost.
Reviewers: george.burgess.iv
Subscribers: llvm-commits
Differential Revision: https://reviews.llvm.org/D25172
llvm-svn: 284771
Michal Gorny [Thu, 20 Oct 2016 20:13:35 +0000 (20:13 +0000)]
[Driver] Parse Debian version as integer when possible. NFC
Replace the string matching for /etc/debian_version with split
integer/string matching algorithm. When the file contains 'major.minor'
version number, parse the major version as integer and use a switch
clause to match it. Otherwise, attempt 'codename/sid' matching using
a StringSwitch.
Differential Revision: https://reviews.llvm.org/D25696
llvm-svn: 284770
Kevin Enderby [Thu, 20 Oct 2016 20:10:30 +0000 (20:10 +0000)]
Another additional error check for invalid Mach-O files for the
load commands that use the MachO::twolevel_hints_command type
which includes only the LC_TWOLEVEL_HINTS load command.
This is not used in llvm libObject code or in llvm tool code. But
does appear in one of the binary test files. While this load command is
obsolete it is easier to add code for it in libObject than edit or change
the binary test case.
llvm-svn: 284769
Reid Kleckner [Thu, 20 Oct 2016 20:02:24 +0000 (20:02 +0000)]
Add a space to work around a Windows CPython / MSys quoting bug
Summary:
The rules for quoting the command line that a subprocess receives are
user space conventions implemented by the C runtime. Python's quoting
rules are implemented here:
https://github.com/python/cpython/blob/
c30098c8c6014f3340a369a31df9c74bdbacc269/Lib/subprocess.py#L725
The result is that the final command line C string computed by Python is
'echo \"'. Mingw doesn't appear to interpret that backslash as escaping
the quote because it is not already inside a quoted region. As a result,
our echo command prints a single backslash instead of a quote.
The whole issue can be sidestepped by adding a space a forcing Python to
put the argument to echo in double quotes.
Reviewers: inglorion, ruiu
Subscribers: llvm-commits
Differential Revision: https://reviews.llvm.org/D25841
llvm-svn: 284768
Vedant Kumar [Thu, 20 Oct 2016 18:44:14 +0000 (18:44 +0000)]
[CodeGen] Devirtualize calls to methods marked final in a derived class
If we see a virtual method call to Base::foo() but can infer that the
object is an instance of Derived, and that 'foo' is marked 'final' in
Derived, we can devirtualize the call to Derived::foo().
Differential Revision: https://reviews.llvm.org/D25813
llvm-svn: 284766
Konstantin Zhuravlyov [Thu, 20 Oct 2016 18:34:58 +0000 (18:34 +0000)]
[AMDGPU] Handle R_AMDGPU_ABS64 relocation
This is needed for the following case (OpenCL example):
__global int Var = 0;
__global int* Ptr[] = {&Var};
...
Differential Revision: https://reviews.llvm.org/D25815
llvm-svn: 284764
Zachary Turner [Thu, 20 Oct 2016 18:34:37 +0000 (18:34 +0000)]
Fix case of file include path.
llvm-svn: 284763
Zachary Turner [Thu, 20 Oct 2016 18:31:19 +0000 (18:31 +0000)]
[CodeView] Refactor serialization to use StreamInterface.
This was all using ArrayRef<>s before which presents a problem
when you want to serialize to or deserialize from an actual
PDB stream. An ArrayRef<> is really just a special case of
what can be handled with StreamInterface though (e.g. by using
a ByteStream), so changing this to use StreamInterface allows
us to plug in a PDB stream and get all the record serialization
and deserialization for free on a MappedBlockStream.
Subsequent patches will try to remove TypeTableBuilder and
TypeRecordBuilder in favor of class that operate on
Streams as well, which should allow us to completely merge
the reading and writing codepaths for both types and symbols.
Differential Revision: https://reviews.llvm.org/D25831
llvm-svn: 284762
Richard Smith [Thu, 20 Oct 2016 18:29:25 +0000 (18:29 +0000)]
[c++1z] Fix assertion failure when using the wrong number of bindings for a
struct with unnamed bitfields.
llvm-svn: 284761
Konstantin Zhuravlyov [Thu, 20 Oct 2016 18:22:36 +0000 (18:22 +0000)]
[AMDGPU] Make note record name a static const member of target streamer
Differential Revision: https://reviews.llvm.org/D25746
llvm-svn: 284760
Konstantin Zhuravlyov [Thu, 20 Oct 2016 18:12:38 +0000 (18:12 +0000)]
[AMDGPU] Emit constant address space data in .rodata section and use relocations instead of fixups (amdhsa only)
Differential Revision: https://reviews.llvm.org/D25693
llvm-svn: 284759
Rui Ueyama [Thu, 20 Oct 2016 18:07:11 +0000 (18:07 +0000)]
Remove unused #include.
llvm-svn: 284758
Dehao Chen [Thu, 20 Oct 2016 18:06:52 +0000 (18:06 +0000)]
Using branch probability to guide critical edge splitting.
Summary:
The original heuristic to break critical edge during machine sink is relatively conservertive: when there is only one instruction sinkable to the critical edge, it is likely that the machine sink pass will not break the critical edge. This leads to many speculative instructions executed at runtime. However, with profile info, we could model the splitting benefits: if the critical edge has 50% taken rate, it would always be beneficial to split the critical edge to avoid the speculated runtime instructions. This patch uses profile to guide critical edge splitting in machine sink pass.
The performance impact on speccpu2006 on Intel sandybridge machines:
spec/2006/fp/C++/444.namd 25.3 +0.26%
spec/2006/fp/C++/447.dealII 45.96 -0.10%
spec/2006/fp/C++/450.soplex 41.97 +1.49%
spec/2006/fp/C++/453.povray 36.83 -0.96%
spec/2006/fp/C/433.milc 23.81 +0.32%
spec/2006/fp/C/470.lbm 41.17 +0.34%
spec/2006/fp/C/482.sphinx3 48.13 +0.69%
spec/2006/int/C++/471.omnetpp 22.45 +3.25%
spec/2006/int/C++/473.astar 21.35 -2.06%
spec/2006/int/C++/483.xalancbmk 36.02 -2.39%
spec/2006/int/C/400.perlbench 33.7 -0.17%
spec/2006/int/C/401.bzip2 22.9 +0.52%
spec/2006/int/C/403.gcc 32.42 -0.54%
spec/2006/int/C/429.mcf 39.59 +0.19%
spec/2006/int/C/445.gobmk 26.98 -0.00%
spec/2006/int/C/456.hmmer 24.52 -0.18%
spec/2006/int/C/458.sjeng 28.26 +0.02%
spec/2006/int/C/462.libquantum 55.44 +3.74%
spec/2006/int/C/464.h264ref 46.67 -0.39%
geometric mean +0.20%
Manually checked 473 and 471 to verify the diff is in the noise range.
Reviewers: rengolin, davidxl
Subscribers: llvm-commits
Differential Revision: https://reviews.llvm.org/D24818
llvm-svn: 284757
Chris Bieneman [Thu, 20 Oct 2016 18:01:19 +0000 (18:01 +0000)]
Re-landing a cleaned up implementation of r284550
This time it should actually work. The previous implementaiton was not
getting the linker or compiler flag set correctly in all the right
situations. By moving the check down and basing it of whether or not CXX
is set I we can have the logic to add the flags exist only once for the
linker and once for the compiler instead of duplicating it.
llvm-svn: 284756
Simon Pilgrim [Thu, 20 Oct 2016 18:00:35 +0000 (18:00 +0000)]
[CostModel][X86] Fixed AVX1/AVX512 sdiv/udiv uniformconst costs for 256/512 bit integer vectors
We weren't checking for uniform const costs before the general cost, resulting in very high estimates.
llvm-svn: 284755
Ekaterina Romanova [Thu, 20 Oct 2016 17:59:15 +0000 (17:59 +0000)]
Add more doxygen comments to emmintrin.h's intrinsics.
With this patch, 75% of the intrinsics in this file will be documented now. The patches for the rest of the intrisics in this file will be send out later.
The doxygen comments are automatically generated based on Sony's intrinsics document.
I got an OK from Eric Christopher to commit doxygen comments without prior code review upstream. This patch was internally reviewed by Yunzhong Gao.
llvm-svn: 284754
Richard Smith [Thu, 20 Oct 2016 17:57:33 +0000 (17:57 +0000)]
[c++1z] Teach composite pointer type computation how to compute the composite
pointer type of two function pointers with different noexcept specifications.
While I'm here, also teach it how to merge dynamic exception specifications.
llvm-svn: 284753
Pirama Arumuga Nainar [Thu, 20 Oct 2016 17:56:36 +0000 (17:56 +0000)]
Fix *_EXTEND_VECTOR_INREG legalization
Summary:
While promoting *_EXTEND_VECTOR_INREG nodes whose inputs are already
promoted, perform the appropriate sign extension for the promoted node
before doing the *_EXTEND_VECTOR_INREG operation. If not, the undefined
high-order bits of the promoted operand may (a) be garbage inc ase of
zext) or (b) contribute the wrong sign-bit (in case of sext)
Updated the promote-vec3.ll test after this change. The diff shows
explicit zeroing in case of zext and intermediate sign extension in case
of sext.
Reviewers: RKSimon
Subscribers: llvm-commits, srhines
Differential Revision: https://reviews.llvm.org/D25790
llvm-svn: 284752
Simon Atanasyan [Thu, 20 Oct 2016 17:53:59 +0000 (17:53 +0000)]
[ELF][MIPS] Use getMipsLocalEntriesNum() everywhere we need number of `local` GOT entries. NFC
llvm-svn: 284751
Simon Atanasyan [Thu, 20 Oct 2016 17:53:55 +0000 (17:53 +0000)]
[ELF][MIPS] Separate calculation of MIPS GOT index and offset of the corresponding part of the GOT. NFC
MIPS GOT consists of some parts: local, global, TLS entries. This change
separates calculation of MIPS GOT index and offset of the corresponding
part of the GOT. That makes code a bit clear and allow to extend number
of parts in the future.
llvm-svn: 284750
Renato Golin [Thu, 20 Oct 2016 17:41:08 +0000 (17:41 +0000)]
[clang-cl] Fix test that shouldn't be running on non-x86
The clang-cl test required x86-registered-target but it defaulted to the
host's triple and AArch64 still doesn't support COFF, so the test failed.
The triple was "aarch64-pc-windows-msvc18.0.0" with ObjectFormat equals
llvm::Triple::COFF, failing assertion:
Assertion `(TT.isOSBinFormatELF() || TT.isOSBinFormatMachO()) &&
"Only expect Darwin and ELF targets"
in AArch64MCTargetDesc.cpp:78.
Making the test only run on Windows hosts obviously fixes the problem.
llvm-svn: 284749
Simon Pilgrim [Thu, 20 Oct 2016 17:16:38 +0000 (17:16 +0000)]
[CostModel][X86] Added tests for sdiv/udiv costs for uniform const and uniform const power-of-2
Shows poor costings in AVX1/AVX512BW for certain vector types
llvm-svn: 284748
Andrey Churbanov [Thu, 20 Oct 2016 17:14:17 +0000 (17:14 +0000)]
Fixed memory leak mistakenly introduced by https://reviews.llvm.org/D23115
Differential Revision: http://reviews.llvm.org/D25510
llvm-svn: 284747
Sanjay Patel [Thu, 20 Oct 2016 16:55:45 +0000 (16:55 +0000)]
[Target] remove TargetRecip class; 2nd try
This is a retry of r284495 which was reverted at r284513 due to use-after-scope bugs
caused by faulty usage of StringRef.
This version also renames a pair of functions:
getRecipEstimateDivEnabled()
getRecipEstimateSqrtEnabled()
as suggested by Eric Christopher.
original commit msg:
[Target] remove TargetRecip class; move reciprocal estimate isel functionality to TargetLowering
This is a follow-up to https://reviews.llvm.org/D24816 - where we changed reciprocal estimates to be function attributes
rather than TargetOptions.
This patch is intended to be a structural, but not functional change. By moving all of the
TargetRecip functionality into TargetLowering, we can remove all of the reciprocal estimate
state, shield the callers from the string format implementation, and simplify/localize the
logic needed for a target to enable this.
If a function has a "reciprocal-estimates" attribute, those settings may override the target's
default reciprocal preferences for whatever operation and data type we're trying to optimize.
If there's no attribute string or specific setting for the op/type pair, just use the target
default settings.
As noted earlier, a better solution would be to move the reciprocal estimate settings to IR
instructions and SDNodes rather than function attributes, but that's a multi-step job that
requires infrastructure improvements. I intend to work on that, but it's not clear how long
it will take to get all the pieces in place.
Differential Revision: https://reviews.llvm.org/D25440
llvm-svn: 284746