Teresa Johnson [Wed, 27 Apr 2016 13:28:35 +0000 (13:28 +0000)]
[ThinLTO] Use valueid instead of bitcode offsets in combined index file
Summary:
With the removal of support for lazy parsing of combined index summary
records (e.g. r267344), we no longer need to include the summary record
bitcode offset in the VST entries for definitions. Change the combined
index format to be similar to the per-module index format in using value
ids to cross-reference from the summary record to the VST entry (rather
than the summary record bitcode offset to cross-reference in the other
direction).
The visible changes are:
1) Add the value id to the combined summary records
2) Remove the summary offset from the combined VST records, which has
the following effects:
- No longer need the VST_CODE_COMBINED_GVDEFENTRY record, as all
combined index VST entries now only contain the value id and
corresponding GUID.
- No longer have duplicate VST entries in the case where there are
multiple definitions of a symbol (e.g. weak/linkonce), as they all
have the same value id and GUID.
An implication of #2 above is that in order to hook up an alias to the
correct aliasee based on the value id of the aliasee recorded in the
combined index alias record, we need to scan the entries in the index
for that GUID to find the one from the same module (i.e. the case where
there are multiple entries for the aliasee). But the reader no longer
has to maintain a special map to hook up the alias/aliasee.
Reviewers: joker.eph
Subscribers: joker.eph, llvm-commits
Differential Revision: http://reviews.llvm.org/D19481
llvm-svn: 267712
Haojian Wu [Wed, 27 Apr 2016 13:23:39 +0000 (13:23 +0000)]
Remove explain-config testcase that detect hard-coded check.
Remove it for now, we need to better way to figure out how to test it.
llvm-svn: 267711
Dmitry Vyukov [Wed, 27 Apr 2016 12:59:35 +0000 (12:59 +0000)]
tsan: fix build
error: implicit declaration of function 'abort' is invalid in C99
llvm-svn: 267710
Dmitry Vyukov [Wed, 27 Apr 2016 12:56:16 +0000 (12:56 +0000)]
tsan: fix darwin Go build
Ifdef out global variables with destructors.
This requires runtime support that is not provided by Go runtime
(in particular _dso_handle symbol).
llvm-svn: 267709
Artur Pilipenko [Wed, 27 Apr 2016 12:51:01 +0000 (12:51 +0000)]
NFC. Introduce Value::getPointerDerferecnceableBytes
Extract a part of isDereferenceableAndAlignedPointer functionality to Value::getPointerDerferecnceableBytes. Currently it's a NFC, but in future I'm going to accumulate all the logic about value dereferenceability in this function similarly to Value::getPointerAlignment function (D16144).
Reviewed By: reames
Differential Revision: http://reviews.llvm.org/D17572
llvm-svn: 267708
Johannes Doerfert [Wed, 27 Apr 2016 12:49:11 +0000 (12:49 +0000)]
[FIX] Propagate execution domain of invariant loads
If the base pointer of an invariant load is is loaded conditionally, that
condition needs to hold for the invariant load too. The structure of the
program will imply this for domain constraints but not for imprecisions in
the modeling. To this end we will propagate the execution context of base
pointers during code generation and thus ensure the derived pointer does
not access an invalid base pointer.
llvm-svn: 267707
Jonathan Coe [Wed, 27 Apr 2016 12:48:25 +0000 (12:48 +0000)]
Expose cxx constructor and method properties through libclang and python bindings.
Summary:
I have exposed the following function through libclang and the clang.cindex python bindings:
clang_CXXConstructor_isConvertingConstructor,
clang_CXXConstructor_isCopyConstructor,
clang_CXXConstructor_isDefaultConstructor,
clang_CXXConstructor_isMoveConstructor,
clang_CXXMethod_isDefaulted
I need (some of) these methods for a C++ code model I am building in Python to drive a code generator.
Reviewers: compnerd, skalinichev
Subscribers: cfe-commits
Differential Revision: http://reviews.llvm.org/D15469
llvm-svn: 267706
Rafael Espindola [Wed, 27 Apr 2016 12:47:30 +0000 (12:47 +0000)]
Reduce usage of isRelRelative.
It is now used only for relocations that only set the low bits inside a
page. Everything else is handled by getRelExpr.
I will send a another review renaming and better documenting
isRelRelative.
llvm-svn: 267705
Pavel Labath [Wed, 27 Apr 2016 12:43:37 +0000 (12:43 +0000)]
Remove flaky decorator from three tests on linux
The flakyness is no longer reproducible, and the tests seem to be passing reliably now.
llvm-svn: 267704
Dmitry Vyukov [Wed, 27 Apr 2016 12:30:48 +0000 (12:30 +0000)]
tsan: change tsan/Go interface for obtaining the current Processor
Current interface assumes that Go calls ProcWire/ProcUnwire
to establish the association between thread and proc.
With the wisdom of hindsight, this interface does not work
very well. I had to sprinkle Go scheduler with wire/unwire
calls, and any mistake leads to hard to debug crashes.
This is not something one wants to maintian.
Fortunately, there is a simpler solution. We can ask Go
runtime as to what is the current Processor, and that
question is very easy to answer on Go side.
Switch to such interface.
llvm-svn: 267703
Rafael Espindola [Wed, 27 Apr 2016 12:25:22 +0000 (12:25 +0000)]
Represent TOC relative relocations as GOTREL.
That way we only need to subtract the offset is relocateOne.
llvm-svn: 267702
Rafael Espindola [Wed, 27 Apr 2016 12:21:27 +0000 (12:21 +0000)]
Create a .got when PPC64 uses a TOC.
This simplifies the logic for computing the value of the toc base.
llvm-svn: 267701
Haojian Wu [Wed, 27 Apr 2016 12:17:04 +0000 (12:17 +0000)]
Fix a crash in cppcoreguidelines-pro-type-member-init when checking a type with a template parameter as a base class.
Summary: Fixed a crash in cppcoreguidelines-pro-type-member-init when encountering a type that uses one of its template parameters as a base when compiling for C++98.
Patch by Michael Miller!
Reviewers: aaron.ballman, alexfh, hokein
Subscribers: cfe-commits
Differential Revision: http://reviews.llvm.org/D19539
llvm-svn: 267700
Simon Pilgrim [Wed, 27 Apr 2016 12:04:44 +0000 (12:04 +0000)]
[InstCombine][SSE] Regenerated vector shift tests
llvm-svn: 267699
Rafael Espindola [Wed, 27 Apr 2016 11:54:07 +0000 (11:54 +0000)]
Simplify handling of R_PPC64_TOC. NFC.
llvm-svn: 267698
Alexander Kornienko [Wed, 27 Apr 2016 11:45:14 +0000 (11:45 +0000)]
clang-tidy -list-checks should exit with non-zero code when no checks are enabled.
llvm-svn: 267697
Michael Zuckerman [Wed, 27 Apr 2016 11:43:14 +0000 (11:43 +0000)]
[Clang][BuiltIn][AVX512] Adding intrinsics without mask for VBROADCAST and VPBROADCAST instruction set .
Differential Revision: http://reviews.llvm.org/D19196
llvm-svn: 267696
Alexey Bataev [Wed, 27 Apr 2016 11:38:05 +0000 (11:38 +0000)]
[OPENMP] Fix crash on initialization of classes with no init clause in
declare reductions.
If reduction clause is applied to instance of class with user-defined
reduction operation without initialization clause, it may cause a crash.
Patch fixes this issue.
llvm-svn: 267695
Zlatko Buljan [Wed, 27 Apr 2016 11:31:44 +0000 (11:31 +0000)]
[mips][microMIPS] Add CodeGen support for SUBU16, SUB, SUBU, DSUB and DSUBU instructions
Differential Revision: http://reviews.llvm.org/D16676
llvm-svn: 267694
Zlatko Buljan [Wed, 27 Apr 2016 11:02:23 +0000 (11:02 +0000)]
[mips][microMIPS] Add CodeGen support for SLL16, SRL16, SLL, SLLV, SRA, SRAV, SRL and SRLV instructions
Differential Revision: http://reviews.llvm.org/D17989
llvm-svn: 267693
Artur Pilipenko [Wed, 27 Apr 2016 11:00:48 +0000 (11:00 +0000)]
isSafeToLoadUnconditionally support queries without a context
This is required to use this function from isSafeToSpeculativelyExecute
Reviewed By: hfinkel
Differential Revision: http://reviews.llvm.org/D16231
llvm-svn: 267692
Vassil Vassilev [Wed, 27 Apr 2016 10:46:06 +0000 (10:46 +0000)]
[modules] Fix Decl's Used invariant.
The Decl::isUsed has a value for every decl. In non-module builds it is very
difficult (but possible) to break this invariant but when we walk up the redecl
chain we find the neccessary information.
When deserializing the decls from a module it is much more difficult to update
correctly this invariant. The patch centralizes the information whether a decl
is used in the canonical decl marking the entire entity as being used.
Fixes https://llvm.org/bugs/show_bug.cgi?id=27401
Patch by Cristina Cristescu and me.
Thanks to Richard Smith who helped to debug and understand the issue!
Reviewed by Richard Smith.
llvm-svn: 267691
Michael Zuckerman [Wed, 27 Apr 2016 10:44:15 +0000 (10:44 +0000)]
[Clang][BuiltIn][AVX512]Adding intrinsics for vmovntdqa vmovntpd vmovntps instruction set
Differential Revision: http://reviews.llvm.org/D19529
llvm-svn: 267690
Artur Pilipenko [Wed, 27 Apr 2016 10:42:29 +0000 (10:42 +0000)]
Use DL preferred alignment for alloca in Value::getPointerAlignment
Teach Value::getPointerAlignment that allocas with no explicit alignment are aligned to preferred alignment of the allocated type.
Reviewed By: hfinkel
Differential Revision: http://reviews.llvm.org/D17569
llvm-svn: 267689
Pavel Labath [Wed, 27 Apr 2016 10:40:52 +0000 (10:40 +0000)]
Fix compiler warnings in SymbolFilePDBTests
llvm-svn: 267688
Haojian Wu [Wed, 27 Apr 2016 10:22:31 +0000 (10:22 +0000)]
Don't search compilation database in explain-check test.
Reviewers: alexfh
Subscribers: cfe-commits
Differential Revision: http://reviews.llvm.org/D19583
llvm-svn: 267687
Simon Pilgrim [Wed, 27 Apr 2016 09:53:09 +0000 (09:53 +0000)]
[InstCombine][SSE] Added DemandedBits tests for MOVMSK instructions
MOVMSK zeros the upper bits of the gpr - we should be able to use this.
llvm-svn: 267686
George Rimar [Wed, 27 Apr 2016 09:24:03 +0000 (09:24 +0000)]
Removed dead code. NFC.
llvm-svn: 267685
George Rimar [Wed, 27 Apr 2016 09:16:28 +0000 (09:16 +0000)]
[ELF] - Align sections file offsets correctly.
System V ABI 4.1 specifies that program header's p_vaddr should equal p_offset, modulo p_align.
(https://docs.oracle.com/cd/E19683-01/816-1386/chapter6-83432/index.html).
It was possible to violate this using the linkerscript.
Patch fixes the issue.
Differential revision: http://reviews.llvm.org/D19416
llvm-svn: 267684
Haojian Wu [Wed, 27 Apr 2016 09:15:01 +0000 (09:15 +0000)]
[ClangTidy] Add an 'explain-checks' option to diagnose where each checks comes from.
Reviewers: alexfh
Subscribers: cfe-commits
Differential Revision: http://reviews.llvm.org/D18694
llvm-svn: 267683
Dmitry Vyukov [Wed, 27 Apr 2016 08:41:31 +0000 (08:41 +0000)]
tsan: fix windows build
warning: '__sanitizer::uptr __sanitizer::GetPreferredBase(const char*)' defined but not used
llvm-svn: 267682
Dmitry Vyukov [Wed, 27 Apr 2016 08:39:32 +0000 (08:39 +0000)]
tsan: fix windows build
llvm-svn: 267681
Dmitry Vyukov [Wed, 27 Apr 2016 08:34:55 +0000 (08:34 +0000)]
tsan: add missing file
llvm-svn: 267680
Dmitry Vyukov [Wed, 27 Apr 2016 08:28:08 +0000 (08:28 +0000)]
tsan: fix compiler warning
tsan_debugging.cc: In function ‘void* __tsan_get_current_report()’:
tsan_debugging.cc:61:18: warning: cast from type ‘const __tsan::ReportDesc*’
to type ‘void*’ casts away qualifiers [-Wcast-qual]
return (void *)rep;
llvm-svn: 267679
Dmitry Vyukov [Wed, 27 Apr 2016 08:23:02 +0000 (08:23 +0000)]
tsan: split thread into logical and physical state
This is reincarnation of http://reviews.llvm.org/D17648 with the bug fix pointed out by Adhemerval (zatrazz).
Currently ThreadState holds both logical state (required for race-detection algorithm, user-visible)
and physical state (various caches, most notably malloc cache). Move physical state in a new
Process entity. Besides just being the right thing from abstraction point of view, this solves several
problems:
Cache everything on P level in Go. Currently we cache on a mix of goroutine and OS thread levels.
This unnecessary increases memory consumption.
Properly handle free operations in Go. Frees are issue by GC which don't have goroutine context.
As the result we could not do anything more than just clearing shadow. For example, we leaked
sync objects and heap block descriptors.
This will allow to get rid of libc malloc in Go (now we have Processor context for internal allocator cache).
This in turn will allow to get rid of dependency on libc entirely.
Potentially we can make Processor per-CPU in C++ mode instead of per-thread, which will
reduce resource consumption.
The distinction between Thread and Processor is currently used only by Go, C++ creates Processor per OS thread,
which is equivalent to the current scheme.
llvm-svn: 267678
Alexey Bataev [Wed, 27 Apr 2016 07:56:03 +0000 (07:56 +0000)]
[OPENMP] Fix for codegen of captured variables in inlined directives.
Currently there is a problem with codegen of inlined directives inside
lambdas, it may cause a crash during codegen because of incorrect
capturing of variables. Patch fixes this problem.
llvm-svn: 267677
Ekaterina Romanova [Wed, 27 Apr 2016 07:14:02 +0000 (07:14 +0000)]
Updated doxygen comments for intrinsics.
(1) Removed \code.. \endcode tags around the instruction name. This matches the doxygen format for all other intrinsics.
(2) Did a better formatting for the comments (to fit into 80 columns more compactly).
llvm-svn: 267676
Adam Nemet [Wed, 27 Apr 2016 05:59:51 +0000 (05:59 +0000)]
Fixed sphinx warning from r267672
llvm-svn: 267675
Mohit K. Bhakkad [Wed, 27 Apr 2016 05:49:42 +0000 (05:49 +0000)]
[Compiler-rt][CFI] Enabling CFI for MIPS64
Reviewers: eugenis
Subscribers: jaydeep, sagar, Sanitizers
Differential Revision: http://reviews.llvm.org/D19531
llvm-svn: 267674
Simon Atanasyan [Wed, 27 Apr 2016 05:31:28 +0000 (05:31 +0000)]
[ELF][MIPS] Remove getMipsGpAddr(). NFC
llvm-svn: 267673
Adam Nemet [Wed, 27 Apr 2016 05:28:18 +0000 (05:28 +0000)]
[LoopDist] Add llvm.loop.distribute.enable loop metadata
Summary:
D19403 adds a new pragma for loop distribution. This change adds
support for the corresponding metadata that the pragma is translated to
by the FE.
As part of this I had to rethink the flag -enable-loop-distribute. My
goal was to be backward compatible with the existing behavior:
A1. pass is off by default from the optimization pipeline
unless -enable-loop-distribute is specified
A2. pass is on when invoked directly from opt (e.g. for unit-testing)
The new pragma/metadata overrides these defaults so the new behavior is:
B1. A1 + enable distribution for individual loop with the pragma/metadata
B2. A2 + disable distribution for individual loop with the pragma/metadata
The default value whether the pass is on or off comes from the initiator
of the pass. From the PassManagerBuilder the default is off, from opt
it's on.
I moved -enable-loop-distribute under the pass. If the flag is
specified it overrides the default from above.
Then the pragma/metadata can further modifies this per loop.
As a side-effect, we can now also use -enable-loop-distribute=0 from opt
to emulate the default from the optimization pipeline. So to be precise
this is the new behavior:
C1. pass is off by default from the optimization pipeline
unless -enable-loop-distribute or the pragma/metadata enables it
C2. pass is on when invoked directly from opt
unless -enable-loop-distribute=0 or the pragma/metadata disables it
Reviewers: hfinkel
Subscribers: joker.eph, mzolotukhin, llvm-commits
Differential Revision: http://reviews.llvm.org/D19431
llvm-svn: 267672
Vaivaswatha Nagaraj [Wed, 27 Apr 2016 05:25:09 +0000 (05:25 +0000)]
[Cloning] cloneLoopWithPreheader(): add assert to ensure no sub-loops
Summary:
cloneLoopWithPreheader() does not update LoopInfo for sub-loop of
the original loop being cloned. Add assert to ensure no sub-loops for loop being cloned.
Reviewers: anemet, ashutosh.nema, hfinkel
Subscribers: mzolotukhin, llvm-commits
Differential Revision: http://reviews.llvm.org/D15922
llvm-svn: 267671
Craig Topper [Wed, 27 Apr 2016 05:17:00 +0000 (05:17 +0000)]
[Support][X86] Add a few more Intel model numbers to getHostCPUName for airmont and knl.
llvm-svn: 267670
Craig Topper [Wed, 27 Apr 2016 05:16:58 +0000 (05:16 +0000)]
[Support][X86] Change the case values in the Intel family 6 code to hex so its easier to compare with Intel's docs. NFC
llvm-svn: 267669
Mehdi Amini [Wed, 27 Apr 2016 05:11:44 +0000 (05:11 +0000)]
Revert "Support "preserving" the summary information when using setModule() API in LTOCodeGenerator"
This reverts commit r267665.
ASAN shows that there is a use of undefined value.
From: Mehdi Amini <mehdi.amini@apple.com>
llvm-svn: 267668
Jason Molenda [Wed, 27 Apr 2016 04:50:51 +0000 (04:50 +0000)]
Committing patch from <Michael Woerister <michaelwoerister@posteo.net>
to use the default clang C/C++ expression parser when debugging
Rust programs. Ideally there would be a rust language plugin to
support their language natively, but until then this will get simple
variable display to work.
http://reviews.llvm.org/D19545
llvm-svn: 267667
Craig Topper [Wed, 27 Apr 2016 04:40:03 +0000 (04:40 +0000)]
[Support][X86] Add a couple more Broadwell CPU models numbers to getHostCPUName.
llvm-svn: 267666
Mehdi Amini [Wed, 27 Apr 2016 04:24:10 +0000 (04:24 +0000)]
Support "preserving" the summary information when using setModule() API in LTOCodeGenerator
Another attempt at r267655...
From: Mehdi Amini <mehdi.amini@apple.com>
llvm-svn: 267665
Mehdi Amini [Wed, 27 Apr 2016 03:34:28 +0000 (03:34 +0000)]
Revert "Support "preserving" the summary information when using setModule() API in LTOCodeGenerator"
This reverts commit r267657, r267656, and r267655.
The test does not pass on multiple bots, I'm unsure why yet but let's unbreak them.
From: Mehdi Amini <mehdi.amini@apple.com>
llvm-svn: 267664
Rui Ueyama [Wed, 27 Apr 2016 03:04:56 +0000 (03:04 +0000)]
Simplify. NFC.
llvm-svn: 267663
Evgeny Stupachenko [Wed, 27 Apr 2016 03:04:54 +0000 (03:04 +0000)]
The patch fixes PR27392.
Summary:
It is incorrect to compare TripCount (which is BECount + 1)
with extraiters (or Count) to check if we should enter unrolled
loop or not, because TripCount can potentially overflow
(when BECount is max unsigned integer).
While comparing BECount with (Count - 1) is overflow safe and
therefore correct.
Reviewer: hfinkel
Differential Revision: http://reviews.llvm.org/D19256
From: Evgeny Stupachenko <evstupac@gmail.com>
llvm-svn: 267662
Philip Reames [Wed, 27 Apr 2016 03:03:15 +0000 (03:03 +0000)]
[LVI] Delete stale and misleading comment.
llvm-svn: 267661
Chuang-Yu Cheng [Wed, 27 Apr 2016 02:59:28 +0000 (02:59 +0000)]
[ppc64] fix bug in prologue that mfocrf's cr operand should be explict state instead of implicit
This fixes PR27414
Reviewers: kbarton mgrang tjablin
http://reviews.llvm.org/D19255
llvm-svn: 267660
Rui Ueyama [Wed, 27 Apr 2016 02:58:27 +0000 (02:58 +0000)]
Remove unnecessary trailing semicolons.
Since this semicolon existed in an early test file,
it has spread to many files.
llvm-svn: 267659
Ahmed Bougacha [Wed, 27 Apr 2016 01:51:38 +0000 (01:51 +0000)]
[X86] Set AddPristinesAndCSRs to FixupBW LivePhysRegs. NFC.
We run after PEI, so we need to AddPristinesAndCSRs.
In practice, that makes no difference here, because we only ask about
liveness of super-registers of defined GR8/GR16 registers, so they
can't be pristine. Still, it's the correct thing to do.
Thanks to Quentin for noticing!
Follow-up to r267495.
llvm-svn: 267658
Mehdi Amini [Wed, 27 Apr 2016 01:49:11 +0000 (01:49 +0000)]
Fix the test from r267656: Support "preserving" the summary information when using setModule() API in LTOCodeGenerator
From: Mehdi Amini <mehdi.amini@apple.com>
llvm-svn: 267657
Mehdi Amini [Wed, 27 Apr 2016 01:47:46 +0000 (01:47 +0000)]
Add a test for r267655: Support "preserving" the summary information when using setModule() API in LTOCodeGenerator
From: Mehdi Amini <mehdi.amini@apple.com>
llvm-svn: 267656
Mehdi Amini [Wed, 27 Apr 2016 01:46:48 +0000 (01:46 +0000)]
Support "preserving" the summary information when using setModule() API in LTOCodeGenerator
From: Mehdi Amini <mehdi.amini@apple.com>
llvm-svn: 267655
Marshall Clow [Wed, 27 Apr 2016 01:46:43 +0000 (01:46 +0000)]
Fix = that should have been == in test. Thanks to STL@microsoft for the catch
llvm-svn: 267654
Sanjoy Das [Wed, 27 Apr 2016 01:44:31 +0000 (01:44 +0000)]
Fix typo in comment; NFC
llvm-svn: 267653
Ahmed Bougacha [Wed, 27 Apr 2016 01:35:29 +0000 (01:35 +0000)]
[X86] Don't assume that MMX extractelts are from index 0.
It's probably the case for all 3 MMX users out there, but with
hand-crafted IR, you can trigger selection failures. Fix that.
llvm-svn: 267652
Ahmed Bougacha [Wed, 27 Apr 2016 01:35:25 +0000 (01:35 +0000)]
[X86] Re-enable MMX i32 extractelt combine.
This effectively adds back the extractelt combine removed by r262358:
the direct case can still occur (because x86_mmx is special, see
r262446), but it's the indirect case that's now superseded by the
generic combine.
llvm-svn: 267651
Cong Hou [Wed, 27 Apr 2016 01:29:18 +0000 (01:29 +0000)]
Detects the SAD pattern on X86 so that much better code will be emitted once the pattern is matched.
Differential revision: http://reviews.llvm.org/D14840
llvm-svn: 267649
Philip Reames [Wed, 27 Apr 2016 01:02:25 +0000 (01:02 +0000)]
[LVI] Add a comment explaining a subtle piece of code
Or at least, I didn't understand the implications the first several times I read it it.
llvm-svn: 267648
Adam Nemet [Wed, 27 Apr 2016 00:52:48 +0000 (00:52 +0000)]
[Docs] Try to clarify the concept of domains for noalias scope
Summary:
This tries to anchor down the concept of domains a bit better. I had
trouble initially relating this to anything. Also talking to David
Majnemer on IRC suggested that I wasn't the only one.
Reviewers: hfinkel
Subscribers: llvm-commits, majnemer
Differential Revision: http://reviews.llvm.org/D18799
llvm-svn: 267647
Mehdi Amini [Wed, 27 Apr 2016 00:32:13 +0000 (00:32 +0000)]
ThinLTO: do not promote GlobalVariable that have a specific section.
Differential Revision: http://reviews.llvm.org/D18298
From: Mehdi Amini <mehdi.amini@apple.com>
llvm-svn: 267646
Matt Arsenault [Wed, 27 Apr 2016 00:32:09 +0000 (00:32 +0000)]
SLSR: Use UnknownAddressSpace instead of 0 for pure arithmetic.
In the case where isLegalAddressingMode is used for cases
not related to addressing modes, such as pure adds and muls,
it should not be using address space 0. LSR already passes -1
as the address space in these cases.
llvm-svn: 267645
Mehdi Amini [Wed, 27 Apr 2016 00:32:02 +0000 (00:32 +0000)]
LTOCodeGenerator: turns linkonce(_odr) into weak_(odr) when present "MustPreserve" set
Summary:
If the linker requested to preserve a linkonce function, we should
honor this even if we drop all uses.
Reviewers: dexonsmith
Subscribers: llvm-commits, joker.eph
Differential Revision: http://reviews.llvm.org/D19527
From: Mehdi Amini <mehdi.amini@apple.com>
llvm-svn: 267644
Adam Nemet [Wed, 27 Apr 2016 00:31:03 +0000 (00:31 +0000)]
[LoopDist] Split main class. NFC
This splits out the per-loop functionality from the Pass class.
With this the fact whether the loop is forced-distribute with the new
metadata/pragma can be cached in the per-loop class rather than passed
around.
llvm-svn: 267643
Philip Reames [Wed, 27 Apr 2016 00:30:55 +0000 (00:30 +0000)]
[LVI] Reduce compile time by lazily scanning blocks if needed
When encountering a non-local pointer, LVI would eagerly scan the block for dereferences of the given object to prove the pointer to be non null. That's all well and good, but *then* we'd go recurse through our input blocks. As a result, we could end up scanning each and every block we traverse, even if the final definition was obviously non null or we found a constant value somewhere up the chain. The previous code papered over this by using the isKnownNonNull routine from value tracking. This made the duplication less painful in the common case.
Instead, we know do the block scan only *after* we've gotten the recursive results back. This lets us stop scanning individual blocks as soon as we've determined it to be non-null in any predecessor block and use our usual merge rules to propagate that information cheaply through successor blocks. For a pointer which can be found non-null, this does strictly less work and sometimes substaintially so.
Note that the case where we *can't* prove something non-null is still the really expensive case. We end up scanning each and every block looking for a dereference and never end up finding one.
llvm-svn: 267642
Quentin Colombet [Wed, 27 Apr 2016 00:16:29 +0000 (00:16 +0000)]
[MachineInstrBundle] Actually set the PartialDeadDef flag only when the register
is defined!
The users were checking the proper thing (Defined + PartialDeadDef), but the
information may have been wrong for other use cases, so fix that.
llvm-svn: 267641
Peter Collingbourne [Wed, 27 Apr 2016 00:05:06 +0000 (00:05 +0000)]
ELF: Merge UndefinedBitcode and UndefinedElf. NFC.
Differential Revision: http://reviews.llvm.org/D19566
llvm-svn: 267640
Peter Collingbourne [Wed, 27 Apr 2016 00:05:03 +0000 (00:05 +0000)]
ELF: Re-implement -u directly and remove CanKeepUndefined flag.
The semantics of the -u flag are to load the lazy symbol named by the flag. We
were previously relying on this behavior falling out of symbol resolution
against a synthetic undefined symbol, but that didn't quite give us the
correct behavior, so we needed a flag to mark symbols created with -u so
we could treat them specially in the writer. However, it's simpler and less
error prone to implement the required behavior directly and remove the flag.
This fixes an issue where symbols loaded with -u would receive hidden
visibility even when the definition in an object file had wider visibility.
Differential Revision: http://reviews.llvm.org/D19560
llvm-svn: 267639
Quentin Colombet [Tue, 26 Apr 2016 23:55:41 +0000 (23:55 +0000)]
[MachineInstrBundle] Update the comment for PhysRegInfo::DeadDef.
I missed read the comment when I commited r267621 and thought the
comment did not need update. Matthias kindly proved me wrong.
Fixing that.
llvm-svn: 267638
Peter Collingbourne [Tue, 26 Apr 2016 23:52:44 +0000 (23:52 +0000)]
ELF: Move code to where it is used, and related cleanups. NFC.
Differential Revision: http://reviews.llvm.org/D19490
llvm-svn: 267637
Andrew Kaylor [Tue, 26 Apr 2016 23:49:41 +0000 (23:49 +0000)]
Add optimization bisect opt-in calls for SystemZ passes
Differential Revision: http://reviews.llvm.org/D19562
llvm-svn: 267636
Andrew Kaylor [Tue, 26 Apr 2016 23:44:31 +0000 (23:44 +0000)]
Add optimization bisect opt-in calls for NVPTX passes
Differential Revision: http://reviews.llvm.org/D19518
llvm-svn: 267635
Quentin Colombet [Tue, 26 Apr 2016 23:44:14 +0000 (23:44 +0000)]
[X86] Make sure it is safe to clobber EFLAGS, if need be, when choosing
the prologue.
Do not use basic blocks that have EFLAGS live-in as prologue if we need
to realign the stack. Realigning the stack uses AND instruction and this
clobbers EFLAGS.
An other alternative would have been to save and restore EFLAGS around
the stack realignment code, but this is likely inefficient.
Fixes PR27531.
llvm-svn: 267634
Adrian Prantl [Tue, 26 Apr 2016 23:42:43 +0000 (23:42 +0000)]
Module debugging: Add an assertion.
llvm-svn: 267633
Richard Smith [Tue, 26 Apr 2016 23:40:43 +0000 (23:40 +0000)]
PR27513: When determining which declaration to put into an exported lookup
table for a module / PCH, never map from a normal declaration of a class to an
injected-class-name declaration (or vice versa). Those declarations live in
distinct lookup tables and should not be confused.
We really shouldn't be using a CXXRecordDecl to represent an
injected-class-name in the first place; I've filed PR27532 so we don't forget.
llvm-svn: 267632
Justin Bogner [Tue, 26 Apr 2016 23:39:29 +0000 (23:39 +0000)]
PM: Port Reassociate to the new pass manager
llvm-svn: 267631
Adrian Prantl [Tue, 26 Apr 2016 23:37:38 +0000 (23:37 +0000)]
Module debugging: Also correctly handle typedef'd foward-declared members.
Thanks again to Richard Smith for pointing this out.
llvm-svn: 267630
Mitch Bodart [Tue, 26 Apr 2016 23:36:38 +0000 (23:36 +0000)]
[X86] Replace -mcpu with -mattr in several tests
Differential Revision: http://reviews.llvm.org/D19568
llvm-svn: 267629
Justin Bogner [Tue, 26 Apr 2016 23:32:00 +0000 (23:32 +0000)]
Reassociate: Convert another functor into a lambda. NFC
Also move the explanatory comment with it.
llvm-svn: 267628
Philip Reames [Tue, 26 Apr 2016 23:27:33 +0000 (23:27 +0000)]
[LVI] Cut short search if we know we can't return a useful result
Previously we were recursing on our operands for unary and binary operators regardless of whether we knew how to reason about the operator in question. This has the effect of doing a potentially large amount of work, only to throw it away. By checking whether the operation is one LVI can handle, we can cut short the search and return the (overdefined) answer more quickly. The quality of the results produced should not change.
llvm-svn: 267626
Sunil Srivastava [Tue, 26 Apr 2016 23:19:00 +0000 (23:19 +0000)]
Check 'r' and 'y specifiers of freebsd_kernel_printf format strings on PS4
This is an addendum to r229921.
llvm-svn: 267625
Sanjay Patel [Tue, 26 Apr 2016 23:15:48 +0000 (23:15 +0000)]
[SimplifyCFG] propagate branch metadata when creating select
llvm-svn: 267624
Quentin Colombet [Tue, 26 Apr 2016 23:14:32 +0000 (23:14 +0000)]
[X86] Teach the expansion of copy instructions how to do proper liveness.
When the simple analysis provided by MachineBasicBlock::computeRegisterLiveness
fails, fall back on the LivePhysReg utility.
llvm-svn: 267623
Quentin Colombet [Tue, 26 Apr 2016 23:14:29 +0000 (23:14 +0000)]
[MachineBasicBlock] Take advantage of the partially dead information.
Thanks to that information we wouldn't lie on a register being live whereas it
is not.
llvm-svn: 267622
Quentin Colombet [Tue, 26 Apr 2016 23:14:24 +0000 (23:14 +0000)]
[MachineInstrBundle] Improvement the recognition of dead definitions.
Now, it is possible to know that partial definitions are dead definitions and
recognize that clobbered registers are also dead.
llvm-svn: 267621
Philip Reames [Tue, 26 Apr 2016 23:10:35 +0000 (23:10 +0000)]
[LVI] Apply transfer rule for overdefine inputs for binary operators
As pointed out by John Regehr over in http://reviews.llvm.org/D19485, LVI was being incredibly stupid about applying its transfer rules. Rather than gathering local facts from the expression itself, it was simply giving up entirely if one of the inputs was overdefined. This greatly impacts the precision of the overall analysis and makes it far more fragile as well.
This patch builds on 267609 which did the same thing for unary casts.
llvm-svn: 267620
Jingyue Wu [Tue, 26 Apr 2016 22:59:25 +0000 (22:59 +0000)]
[NVPTX] Fix some usages of CodeGenOpt::None.
NVPTXLowerKernelArgs is required for correctness, so it should not be guarded
by CodeGenOpt::None.
NVPTXPeephole is optimization only, so it should be skipped when
CodeGenOpt::None.
llvm-svn: 267619
Philip Reames [Tue, 26 Apr 2016 22:52:30 +0000 (22:52 +0000)]
[LVI] A better fix for the assertion error introduced by 267609
Essentially, I was using the wrong size function. For types which were sized, but not primitive, I wasn't getting a useful size for the operand and failed an assert. I fixed this, and also added a guard that the input is a sized type. Test case is for the original mistake. I'm not sure how to actually exercise the sized type check.
llvm-svn: 267618
Philip Reames [Tue, 26 Apr 2016 22:31:53 +0000 (22:31 +0000)]
[LVI] Speculative fix for assertion seen in clang bots
I'll clean this up and add a test case shortly. I want to make sure this does actually fix the bots; if not, I'll revert.
llvm-svn: 267617
Adrian McCarthy [Tue, 26 Apr 2016 22:25:40 +0000 (22:25 +0000)]
Fix TestRegisterVariables.py on Windows
Use __attribute__((regparm(x))) to ensure the compiler enregisters at least some arguments when calling functions.
Differential Revision: http://reviews.llvm.org/D19548
llvm-svn: 267616
Sanjay Patel [Tue, 26 Apr 2016 22:23:38 +0000 (22:23 +0000)]
[LowerExpectIntrinsic] make default likely/unlikely ratio bigger
We need the default ratio to be sufficiently large that it triggers transforms
based on block frequency info (BFI) and plays well with the recently introduced
BranchProbability used by CGP.
Differential Revision: http://reviews.llvm.org/D19435
llvm-svn: 267615
Justin Bogner [Tue, 26 Apr 2016 22:22:18 +0000 (22:22 +0000)]
Reassociate: Simplify using lambdas. NFC
llvm-svn: 267614
Rafael Espindola [Tue, 26 Apr 2016 22:00:04 +0000 (22:00 +0000)]
Return R_GOTREL for R_MIPS_GPREL*.
This lets Writer.cpp know that they are got relative, which will allow
further simplifications.
llvm-svn: 267613
Adrian Prantl [Tue, 26 Apr 2016 21:58:23 +0000 (21:58 +0000)]
Module debugging: Add testcase for member classes of class template specializations.
llvm-svn: 267612
Adrian Prantl [Tue, 26 Apr 2016 21:58:18 +0000 (21:58 +0000)]
Module debugging: Use the definition to determine module-defined types.
Follow-up to r267464. Thanks to Richard Smith for pointing this out!
llvm-svn: 267611