Dmitri Gribenko [Wed, 28 Nov 2012 21:40:54 +0000 (21:40 +0000)]
Documentation: improve formatting and remove unneeded empty lines.
llvm-svn: 168817
Michael Ilseman [Wed, 28 Nov 2012 21:21:18 +0000 (21:21 +0000)]
Whoops, fixed bad merge
llvm-svn: 168816
Michael Ilseman [Wed, 28 Nov 2012 21:19:52 +0000 (21:19 +0000)]
Fixed bad test case
llvm-svn: 168815
Michael Ilseman [Wed, 28 Nov 2012 21:17:34 +0000 (21:17 +0000)]
Fast-math: IRBuilder test for creating instructions with fast-math flags
llvm-svn: 168814
Michael Ilseman [Wed, 28 Nov 2012 21:16:19 +0000 (21:16 +0000)]
Fast-math: Extend IRBuilder to have settable FastMathFlags to create instructions with
Also extended IRBuilder's documentation to mention the convenience state for DefaultFPMathTag and FastMathFlags that can be set.
llvm-svn: 168812
Michael Ilseman [Wed, 28 Nov 2012 21:11:25 +0000 (21:11 +0000)]
Fast-math comments and convenience method
llvm-svn: 168811
Jyotsna Verma [Wed, 28 Nov 2012 20:58:14 +0000 (20:58 +0000)]
Define signed const-ext immediate operands and their predicates.
llvm-svn: 168810
Benjamin Kramer [Wed, 28 Nov 2012 20:55:10 +0000 (20:55 +0000)]
ARM: Implement CanLowerReturn so large vectors get expanded into sret.
Fixes 14337.
llvm-svn: 168809
Jim Ingham [Wed, 28 Nov 2012 20:09:03 +0000 (20:09 +0000)]
Cleanup - remove declarations of unimplemented functions.
llvm-svn: 168807
Jakob Stoklund Olesen [Wed, 28 Nov 2012 19:13:06 +0000 (19:13 +0000)]
Make the LiveRegMatrix analysis available to targets.
No functional change, just moved header files.
Targets can inject custom passes between register allocation and
rewriting. This makes it possible to tweak the register allocation
before rewriting, using the full global interference checking available
from LiveRegMatrix.
llvm-svn: 168806
Jordan Rose [Wed, 28 Nov 2012 19:12:44 +0000 (19:12 +0000)]
[analyzer] scan-build: Don't forget to close our temp file for Clang's output.
Also, minor whitespace/indentation fixes.
Patch by Peeter Joot!
llvm-svn: 168805
Jordan Rose [Wed, 28 Nov 2012 19:12:29 +0000 (19:12 +0000)]
[analyzer] scan-build: Treat '.C' files as C++.
Part of PR14443.
llvm-svn: 168804
Eli Bendersky [Wed, 28 Nov 2012 19:00:02 +0000 (19:00 +0000)]
Add backreference matching capabilities to Support/Regex, with
appropriate unit tests. This change in itself is not expected to
affect any functionality at this point, but it will serve as a
stepping stone to improve FileCheck's variable matching capabilities.
Luckily, our regex implementation already supports backreferences,
although a bit of hacking is required to enable it. It supports both
Basic Regular Expressions (BREs) and Extended Regular Expressions
(EREs), without supporting backrefs for EREs, following POSIX strictly
in this respect. And EREs is what we actually use (rightly). This is
contrary to many implementations (including the default on Linux) of
POSIX regexes, that do allow backrefs in EREs.
Adding backref support to our EREs is a very simple change in the
regcomp parsing code. I fail to think of significant cases where it
would clash with existing things, and can bring more versatility to
the regexes we write. There's always the danger of a backref in a
specially crafted regex causing exponential matching times, but since
we mainly use them for testing purposes I don't think it's a big
problem. [it can also be placed behind a flag specific to FileCheck,
if needed].
For more details, see:
* http://lists.cs.uiuc.edu/pipermail/llvmdev/2012-November/055840.html
* http://lists.cs.uiuc.edu/pipermail/llvm-commits/Week-of-Mon-
20121126/156878.html
llvm-svn: 168802
Benjamin Kramer [Wed, 28 Nov 2012 18:35:35 +0000 (18:35 +0000)]
MCJIT depends on JIT.
Unbreaks the CMake shared library build. This is nasty and should be fixed
eventually.
llvm-svn: 168800
Ulrich Weigand [Wed, 28 Nov 2012 18:21:03 +0000 (18:21 +0000)]
Fix initial frame state on powerpc64.
The createPPCMCAsmInfo routine used PPC::R1 as the initial frame
pointer register, but on PPC64 the 32-bit R1 register does not
have a corresponding DWARF number, causing invalid CIE initial
frame state to be emitted. Fix by using PPC::X1 instead.
llvm-svn: 168799
Alexey Samsonov [Wed, 28 Nov 2012 17:34:24 +0000 (17:34 +0000)]
Refactor -fsanitize, -f*-sanitizer arguments parsing. Provide a more careful diagnostic for invalid sets of sanitizers
llvm-svn: 168794
Kostya Serebryany [Wed, 28 Nov 2012 15:01:23 +0000 (15:01 +0000)]
[asan] add DISABLED_BuiltinLongJmpTest
llvm-svn: 168793
Patrik Hägglund [Wed, 28 Nov 2012 14:32:52 +0000 (14:32 +0000)]
Fix warning. [-Wunused-variable]
llvm-svn: 168792
Patrik Hägglund [Wed, 28 Nov 2012 13:30:31 +0000 (13:30 +0000)]
Fix tests with broken datalayout strings.
Buildbot failure at r168785.
llvm-svn: 168791
Dmitry Vyukov [Wed, 28 Nov 2012 13:30:06 +0000 (13:30 +0000)]
tsan: explicitly say that failed to restore the stack
llvm-svn: 168790
Dmitry Vyukov [Wed, 28 Nov 2012 13:01:32 +0000 (13:01 +0000)]
tsan: address several review comments
llvm-svn: 168789
Dmitry Vyukov [Wed, 28 Nov 2012 12:56:52 +0000 (12:56 +0000)]
tsan: add log_path parameter (similar to asan)
remove old log_fileno
llvm-svn: 168788
Dmitry Vyukov [Wed, 28 Nov 2012 12:19:50 +0000 (12:19 +0000)]
tsan: dynamic history size
introduces history_size parameter that can be used to control trace size at startup
llvm-svn: 168786
Patrik Hägglund [Wed, 28 Nov 2012 12:13:12 +0000 (12:13 +0000)]
Add error handling in getInt.
Accordingly, update a testcase with a broken datalayout string.
Also, we never parse negative numbers, because '-' is used as a
separator. Therefore, use unsigned as result type.
llvm-svn: 168785
Dmitry Vyukov [Wed, 28 Nov 2012 10:49:27 +0000 (10:49 +0000)]
tsan: change fast state layout in preparation to dynamic traces
llvm-svn: 168784
Dmitry Vyukov [Wed, 28 Nov 2012 10:35:31 +0000 (10:35 +0000)]
tsan: move traces from tls into dedicated storage at fixed address
helps to reduce tls size (it's weird to have multi-MB tls)
will help with dynamically adjustable trace size
llvm-svn: 168783
Kostya Serebryany [Wed, 28 Nov 2012 10:32:11 +0000 (10:32 +0000)]
[asan] Split AddressSanitizer into two passes (FunctionPass, ModulePass), Clang part.
llvm-svn: 168782
Kostya Serebryany [Wed, 28 Nov 2012 10:31:36 +0000 (10:31 +0000)]
[asan] Split AddressSanitizer into two passes (FunctionPass, ModulePass), LLVM part. This requires a clang part which will follow.
llvm-svn: 168781
Tobias Grosser [Wed, 28 Nov 2012 10:12:21 +0000 (10:12 +0000)]
cmake: Fix installation of include files
Install the configured Config/config.h file, rather than the
Config/config.h.cmake and Config/Config.h.in source files.
Contributed-by: Richard Membarth <richard.membarth@informatik.uni-erlangen.de>
llvm-svn: 168780
Bill Wendling [Wed, 28 Nov 2012 08:41:48 +0000 (08:41 +0000)]
Add back support for reading and parsing 'deplibs'.
This is for backwards compatibility for pre-3.x bc files. The code reads the
code, but does nothing with it.
llvm-svn: 168779
Andy Gibbs [Wed, 28 Nov 2012 08:14:36 +0000 (08:14 +0000)]
Remove obsolete comment missed by r162937
llvm-svn: 168778
Dmitry Vyukov [Wed, 28 Nov 2012 07:44:26 +0000 (07:44 +0000)]
tsan: add description of memory layouts in different configs
llvm-svn: 168777
Richard Smith [Wed, 28 Nov 2012 07:29:00 +0000 (07:29 +0000)]
Teach Lexer::getSpelling about raw string literals. Specifically, if a raw
string literal needs cleaning (because it contains line-splicing in the
encoding prefix or in the ud-suffix), do not clean the section between the
double-quotes -- that's the "raw" bit!
llvm-svn: 168776
Richard Smith [Wed, 28 Nov 2012 06:23:12 +0000 (06:23 +0000)]
Store on the CXXRecordDecl whether the class has, or would have, a copy
constructor/assignment operator with a const-qualified parameter type. The
prior method for determining this incorrectly used overload resolution.
llvm-svn: 168775
Ted Kremenek [Wed, 28 Nov 2012 05:36:28 +0000 (05:36 +0000)]
Remove workaround in RegionStore in r168741 since it is handled more generally by r168757.
llvm-svn: 168774
Andrew Trick [Wed, 28 Nov 2012 05:13:28 +0000 (05:13 +0000)]
misched: Analysis that partitions the DAG into subtrees.
This is a simple, cheap infrastructure for analyzing the shape of a
DAG. It recognizes uniform DAGs that take the shape of bottom-up
subtrees, such as the included matrix multiplication example. This is
useful for heuristics that balance register pressure with ILP. Two
canonical expressions of the heuristic are implemented in scheduling
modes: -misched-ilpmin and -misched-ilpmax.
llvm-svn: 168773
Andrew Trick [Wed, 28 Nov 2012 05:13:24 +0000 (05:13 +0000)]
misched: rename ScheduleDAGILP to ScheduleDFS to prepare for other heuristics.
llvm-svn: 168772
Argyrios Kyrtzidis [Wed, 28 Nov 2012 03:56:16 +0000 (03:56 +0000)]
Introduce ASTUnresolvedSet, an UnresolvedSet-like class, whose contents are
allocated using the allocator associated with an ASTContext.
Use this inside CXXRecordDecl::DefinitionData instead of an UnresolvedSet to
avoid a potential memory leak.
rdar://
12761275
llvm-svn: 168771
Argyrios Kyrtzidis [Wed, 28 Nov 2012 03:56:09 +0000 (03:56 +0000)]
Don't return a pointer to an UnresolvedSetImpl in the CXXRecordDecl interface,
expose only the iterators instead.
llvm-svn: 168770
Richard Smith [Wed, 28 Nov 2012 03:45:24 +0000 (03:45 +0000)]
C++ core issue 1344, PR10618: promote "addition of default argument makes this
a special member" diagnostic from warning to error, and fix the cases where it
produced diagnostics with incorrect wording.
We don't support this as an extension, and we ban it even in C++98 mode. This
breaks too much (for instance, the ABI-specified calling convention for a type
can change if it acquires a copy constructor through the addition of a default
argument).
llvm-svn: 168769
Andrew Trick [Wed, 28 Nov 2012 03:42:49 +0000 (03:42 +0000)]
misched: better alias analysis.
This fixes a hole in the "cheap" alias analysis logic implemented within
the DAG builder itself, regardless of whether proper alias analysis is
enabled. It now handles this pattern produced by LSR+CodeGenPrepare.
%sunkaddr1 = ptrtoint * %obj to i64
%sunkaddr2 = add i64 %sunkaddr1, %lsr.iv
%sunkaddr3 = inttoptr i64 %sunkaddr2 to i32*
store i32 %v, i32* %sunkaddr3
llvm-svn: 168768
Andrew Trick [Wed, 28 Nov 2012 03:42:47 +0000 (03:42 +0000)]
misched: Debug output fix. Use an always valid iterator.
llvm-svn: 168767
Sean Callanan [Wed, 28 Nov 2012 03:23:20 +0000 (03:23 +0000)]
If Clang is looking for an Objective-C method on
a type, and we find it in the origin for that
type, don't look anywhere else; just report it.
<rdar://problem/
12675970>
llvm-svn: 168766
Hal Finkel [Wed, 28 Nov 2012 03:04:10 +0000 (03:04 +0000)]
BBVectorize: Correctly merge SubclassOptionalData
When two instructions are combined into a vector instruction,
the resulting instruction must have the most-conservative flags.
llvm-svn: 168765
Eric Christopher [Wed, 28 Nov 2012 02:49:38 +0000 (02:49 +0000)]
Add brief support for the fission .debug_info.dwo section for
ELF output.
llvm-svn: 168764
Eric Christopher [Wed, 28 Nov 2012 02:49:34 +0000 (02:49 +0000)]
Add comments.
llvm-svn: 168763
Eric Christopher [Wed, 28 Nov 2012 02:49:32 +0000 (02:49 +0000)]
Rearrange ordering of sections.
llvm-svn: 168762
Eric Christopher [Wed, 28 Nov 2012 02:49:28 +0000 (02:49 +0000)]
Move and comment accessor routines.
llvm-svn: 168761
Jakob Stoklund Olesen [Wed, 28 Nov 2012 02:35:17 +0000 (02:35 +0000)]
Remove all references to TargetInstrInfoImpl.
This class has been merged into its super-class TargetInstrInfo.
llvm-svn: 168760
Jakob Stoklund Olesen [Wed, 28 Nov 2012 02:35:13 +0000 (02:35 +0000)]
Move the guts of TargetInstrInfoImpl into the TargetInstrInfo class.
The *Impl class no longer serves a purpose now that the super-class
implementation is in CodeGen.
llvm-svn: 168759
Jakob Stoklund Olesen [Wed, 28 Nov 2012 02:35:09 +0000 (02:35 +0000)]
Move Target{Instr,Register}Info.cpp into lib/CodeGen.
The Target library is not allowed to depend on the large CodeGen
library, but the TRI and TII classes provide abstract interfaces that
require both caller and callee to link to CodeGen.
The implementation files for these classes provide default
implementations of some of the hooks. These methods may need to
reference CodeGen, so they belong in that library.
We already have a number of methods implemented in the
TargetInstrInfoImpl sub-class because of that. I will merge that class
into the parent next.
llvm-svn: 168758
Ted Kremenek [Wed, 28 Nov 2012 01:49:01 +0000 (01:49 +0000)]
Fix another false positive due to a CXX temporary object appearing in a C initializer.
The stop-gap here is to just drop such objects when processing the InitListExpr.
We still need a better solution.
Fixes <rdar://problem/
12755044>.
llvm-svn: 168757
Fariborz Jahanian [Wed, 28 Nov 2012 01:27:44 +0000 (01:27 +0000)]
objective-C arc: Underline the selector when issuing
arc specific diagnostic on the selector. This is objc-arc
part of // rdar://
11303469
llvm-svn: 168756
Andrew Kaylor [Wed, 28 Nov 2012 01:02:06 +0000 (01:02 +0000)]
Moving SectionMemoryManager to MCJIT to avoid cross dependency between JIT and RuntimeDyld
llvm-svn: 168755
Greg Clayton [Wed, 28 Nov 2012 00:44:24 +0000 (00:44 +0000)]
<rdar://problem/
12639603>
Simplify the logging on ObjectFile::~ObjectFile() to not access an classes above the object file (like the module) so we don't crash when logging object lifetimes. The log message contains the "this" pointer value which can be matched up with the constructor log.
llvm-svn: 168754
Eli Bendersky [Wed, 28 Nov 2012 00:27:25 +0000 (00:27 +0000)]
Some grammar fixes
llvm-svn: 168752
Chad Rosier [Wed, 28 Nov 2012 00:21:29 +0000 (00:21 +0000)]
Revert r168630, r168631, and r168633 as these are causing nightly test failures.
llvm-svn: 168751
Daniel Dunbar [Wed, 28 Nov 2012 00:06:11 +0000 (00:06 +0000)]
lit: Bump the version to .3.
llvm-svn: 168750
Daniel Dunbar [Tue, 27 Nov 2012 23:56:28 +0000 (23:56 +0000)]
tests/lit: Change test default parameters to assume local build.
- Also, support overriding them with lit parameters.
llvm-svn: 168749
Enrico Granata [Tue, 27 Nov 2012 23:50:00 +0000 (23:50 +0000)]
Fixing a silly typo in the previous patch
llvm-svn: 168748
Bill Schmidt [Tue, 27 Nov 2012 23:36:26 +0000 (23:36 +0000)]
This patch makes medium code model the default for 64-bit PowerPC ELF.
When the CodeGenInfo is to be created for the PPC64 target machine,
a default code-model selection is converted to CodeModel::Medium
provided we are not targeting the Darwin OS. Defaults for Darwin
are unaffected.
llvm-svn: 168747
Sean Callanan [Tue, 27 Nov 2012 23:34:41 +0000 (23:34 +0000)]
Fixed the debugserver Xcode project to allow
DebugClang builds of LLDB to build a properly
codesigned debugserver. I did this by adding
a DebugClang configuration to debugserver that's
just a clone of the Debug configuration.
llvm-svn: 168746
NAKAMURA Takumi [Tue, 27 Nov 2012 23:34:28 +0000 (23:34 +0000)]
docs: [CMake] Add Xcode to the list of project formats CMake can generate.
Suggested by Sean McBride, thanks!
llvm-svn: 168745
NAKAMURA Takumi [Tue, 27 Nov 2012 23:34:08 +0000 (23:34 +0000)]
clang/www/get_started.html: s/cmake/CMake/
Suggested by Sean McBride, thanks!
llvm-svn: 168744
Enrico Granata [Tue, 27 Nov 2012 23:28:32 +0000 (23:28 +0000)]
<rdar://problem/
12754509>
Make sure that ValueObjectDynamicValue clears itself when no dynamic type information can be found
This behavior was supposed to be already happening (as per the comment lines)
llvm-svn: 168743
Ted Kremenek [Tue, 27 Nov 2012 23:05:37 +0000 (23:05 +0000)]
Provide stop-gap solution to crash reported in PR 14436.
This was also covered by <rdar://problem/
12753384>. The static analyzer
evaluates a CXXConstructExpr within an initializer expression and
RegionStore doesn't know how to handle the resulting CXXTempObjectRegion
that gets created. We need a better solution than just dropping the
value, but we need to better understand how to implement the right
semantics here.
Thanks to Jordan for his help diagnosing the behavior here.
llvm-svn: 168741
Fariborz Jahanian [Tue, 27 Nov 2012 23:02:53 +0000 (23:02 +0000)]
objective-C arc: load of a __weak object happens via call to
objc_loadWeak. This retains and autorelease the weakly-refereced
object. This hidden autorelease sometimes makes __weak variable alive even
after the weak reference is erased, because the object is still referenced
by an autorelease pool. This patch overcomes this behavior by loading a
weak object via call to objc_loadWeakRetained(), followng it by objc_release
at appropriate place, thereby removing the hidden autorelease. // rdar://
10849570
llvm-svn: 168740
Andrew Kaylor [Tue, 27 Nov 2012 22:53:57 +0000 (22:53 +0000)]
Fix comment formatting in RuntimeDyld.h
llvm-svn: 168739
Argyrios Kyrtzidis [Tue, 27 Nov 2012 22:49:42 +0000 (22:49 +0000)]
[libclang] Remove WorkingDir field from CIndexer, it has been rendered useless.
No functionality change.
llvm-svn: 168738
Eric Christopher [Tue, 27 Nov 2012 22:43:45 +0000 (22:43 +0000)]
Attempt to make the comments for dwarf debug look more like
the coding standard would like.
llvm-svn: 168737
Eric Christopher [Tue, 27 Nov 2012 22:43:42 +0000 (22:43 +0000)]
Reapply section moving, make sure string section is output last.
llvm-svn: 168736
Chad Rosier [Tue, 27 Nov 2012 22:29:43 +0000 (22:29 +0000)]
[arm fast-isel] Appease the machine verifier by using the proper register
classes. The vast majority of the remaining issues are due to uses of
invalid registers, which are defined by getRegForValue(). Those will be
a little more challenging to cleanup.
rdar://
12719844
llvm-svn: 168735
Greg Clayton [Tue, 27 Nov 2012 22:18:23 +0000 (22:18 +0000)]
<rdar://problem/
12636970>
Properly detect the if unnamed bitfields are supported by clang if the major calng version is higher than 425.
llvm-svn: 168734
Chad Rosier [Tue, 27 Nov 2012 22:12:11 +0000 (22:12 +0000)]
[arm fast-isel] Appease the machine verifier by using the proper register
classes.
rdar://
12719844
llvm-svn: 168733
Nico Weber [Tue, 27 Nov 2012 21:57:34 +0000 (21:57 +0000)]
Move PrettyStackTraceParserEntry to ParseAST.cpp
r128056 moved PrettyStackTraceParserEntry construction from Parser.h
to ParseAST.cpp, so there's no need to keep this class in a header.
llvm-svn: 168731
Richard Smith [Tue, 27 Nov 2012 21:51:36 +0000 (21:51 +0000)]
Remove unused internal linkage variable.
llvm-svn: 168729
Chad Rosier [Tue, 27 Nov 2012 21:46:46 +0000 (21:46 +0000)]
[arm fast-isel] Appease the machine verifier by using the proper register
classes. Also a bit of cleanup.
rdar://
12719844
llvm-svn: 168728
Daniel Malea [Tue, 27 Nov 2012 21:33:41 +0000 (21:33 +0000)]
Mark expected failures on Linux (due to bugzilla #14437)
llvm-svn: 168727
Richard Smith [Tue, 27 Nov 2012 21:31:01 +0000 (21:31 +0000)]
Allow an ASTConsumer to selectively skip function bodies while parsing. Patch
by Olivier Goffart!
llvm-svn: 168726
Richard Smith [Tue, 27 Nov 2012 21:20:31 +0000 (21:20 +0000)]
Simplify checking for whether we should implicitly declare special members and
add some assertions. No functionality change.
llvm-svn: 168725
Sebastian Pop [Tue, 27 Nov 2012 21:15:15 +0000 (21:15 +0000)]
do not execute the OpenMP tests when cloog is not found
llvm-svn: 168724
Chad Rosier [Tue, 27 Nov 2012 20:49:56 +0000 (20:49 +0000)]
Add -verify-machineinstrs to these fast-isel test cases.
llvm-svn: 168723
Andrew Kaylor [Tue, 27 Nov 2012 19:49:00 +0000 (19:49 +0000)]
Modifying lli to use the SectionMemoryManager.
The functionality of SectionMemoryManager is equivalent to the LLIMCJITMemoryManager being replaced except that it allocates memory as RW and later changes it to RX or R as needed. The page permissions are set in the call to MCJIT::finalizeObject.
llvm-svn: 168722
Andrew Kaylor [Tue, 27 Nov 2012 19:42:02 +0000 (19:42 +0000)]
Moving SectionMemoryManager into RuntimeDyld and adding unit tests for it.
The SectionMemoryManager now supports (and requires) applying section-specific page permissions. Clients using this memory manager must call either MCJIT::finalizeObject() or SectionMemoryManager::applyPermissions() before executing JITed code.
See r168718 for changes from the previous implementation.
llvm-svn: 168721
Han Ming Ong [Tue, 27 Nov 2012 19:21:03 +0000 (19:21 +0000)]
<rdar://problem/
12759744> Provide physical memory distribution as part of profile data
Make use of unix system calls to provide physical memory usage profile data.
llvm-svn: 168720
Preston Briggs [Tue, 27 Nov 2012 19:12:26 +0000 (19:12 +0000)]
Modified depends() to recognize that when all levels are "=" and
there's no possible loo-independent dependence, then there's no
dependence.
Updated all test result appropriately.
llvm-svn: 168719
Andrew Kaylor [Tue, 27 Nov 2012 19:00:17 +0000 (19:00 +0000)]
Implementing page permission setting in MCJIT unit test SectionMemoryManager.cpp
This commit is primarily here for the revision history. I'm about to move the SectionMemoryManager into the RuntimeDyld library, but I wanted to check the changes in here so people could see the differences in the updated implementation.
llvm-svn: 168718
Manman Ren [Tue, 27 Nov 2012 18:58:41 +0000 (18:58 +0000)]
CSE: allow PerformTrivialCoalescing to check copies across basic block
boundaries.
Given the following case:
BB0
%vreg1<def> = SUBrr %vreg0, %vreg7
%vreg2<def> = COPY %vreg7
BB1
%vreg10<def> = SUBrr %vreg0, %vreg2
We should be able to CSE between SUBrr in BB0 and SUBrr in BB1.
rdar://
12462006
llvm-svn: 168717
Meador Inge [Tue, 27 Nov 2012 18:52:49 +0000 (18:52 +0000)]
instcombine: Don't replace all uses for instructions with no uses
My commit to migrate the printf simplifiers from the simplify-libcalls
in r168604 introduced a regression reported by Duncan [1]. The problem
is that in some cases the library call simplifier can return a new value
that has no uses and the new value's type is different than the old value's
type (which is fine because there are no uses). The specific case that
triggered the bug looked something like:
declare void @printf(i8*, ...)
...
call void (i8*, ...)* @printf(i8* %fmt)
Which we want to optimized into:
call i32 @putchar(i32 104)
However, the code was attempting to replace all uses of the printf with
the putchar and the types differ, hence a crash. This is fixed by *just*
deleting the original instruction when there are no uses. The old
simplify-libcalls pass is already doing something similar.
[1] http://lists.cs.uiuc.edu/pipermail/llvmdev/2012-November/056338.html
llvm-svn: 168716
Howard Hinnant [Tue, 27 Nov 2012 18:52:32 +0000 (18:52 +0000)]
Fix type-o.
llvm-svn: 168715
Sebastian Pop [Tue, 27 Nov 2012 18:50:41 +0000 (18:50 +0000)]
make IslAstInfo::printScop compatible with CloogInfo::printScop
llvm-svn: 168714
Howard Hinnant [Tue, 27 Nov 2012 18:35:09 +0000 (18:35 +0000)]
Remove by-chapter breakdown of what is implemented. The chart is now obsolete.
llvm-svn: 168713
Jakub Staszak [Tue, 27 Nov 2012 18:27:14 +0000 (18:27 +0000)]
Remove duplicated #includes.
llvm-svn: 168712
Benjamin Kramer [Tue, 27 Nov 2012 18:16:32 +0000 (18:16 +0000)]
SCEV: Even if the latch terminator is foldable we can't deduce the result of an unrelated condition with it.
Fixes PR14432.
llvm-svn: 168711
Manman Ren [Tue, 27 Nov 2012 18:09:26 +0000 (18:09 +0000)]
X86: do not fold load instructions such as [V]MOVS[S|D] to other instructions
when the destination register is wider than the memory load.
These load instructions load from m32 or m64 and set the upper bits to zero,
while the folded instructions may accept m128.
rdar://
12721174
llvm-svn: 168710
Pedro Artigas [Tue, 27 Nov 2012 17:39:20 +0000 (17:39 +0000)]
Test commit only modifying comments
llvm-svn: 168709
Bill Schmidt [Tue, 27 Nov 2012 17:35:46 +0000 (17:35 +0000)]
This patch implements medium code model support for 64-bit PowerPC.
The default for 64-bit PowerPC is small code model, in which TOC entries
must be addressable using a 16-bit offset from the TOC pointer. Additionally,
only TOC entries are addressed via the TOC pointer.
With medium code model, TOC entries and data sections can all be addressed
via the TOC pointer using a 32-bit offset. Cooperation with the linker
allows 16-bit offsets to be used when these are sufficient, reducing the
number of extra instructions that need to be executed. Medium code model
also does not generate explicit TOC entries in ".section toc" for variables
that are wholly internal to the compilation unit.
Consider a load of an external 4-byte integer. With small code model, the
compiler generates:
ld 3, .LC1@toc(2)
lwz 4, 0(3)
.section .toc,"aw",@progbits
.LC1:
.tc ei[TC],ei
With medium model, it instead generates:
addis 3, 2, .LC1@toc@ha
ld 3, .LC1@toc@l(3)
lwz 4, 0(3)
.section .toc,"aw",@progbits
.LC1:
.tc ei[TC],ei
Here .LC1@toc@ha is a relocation requesting the upper 16 bits of the
32-bit offset of ei's TOC entry from the TOC base pointer. Similarly,
.LC1@toc@l is a relocation requesting the lower 16 bits. Note that if
the linker determines that ei's TOC entry is within a 16-bit offset of
the TOC base pointer, it will replace the "addis" with a "nop", and
replace the "ld" with the identical "ld" instruction from the small
code model example.
Consider next a load of a function-scope static integer. For small code
model, the compiler generates:
ld 3, .LC1@toc(2)
lwz 4, 0(3)
.section .toc,"aw",@progbits
.LC1:
.tc test_fn_static.si[TC],test_fn_static.si
.type test_fn_static.si,@object
.local test_fn_static.si
.comm test_fn_static.si,4,4
For medium code model, the compiler generates:
addis 3, 2, test_fn_static.si@toc@ha
addi 3, 3, test_fn_static.si@toc@l
lwz 4, 0(3)
.type test_fn_static.si,@object
.local test_fn_static.si
.comm test_fn_static.si,4,4
Again, the linker may replace the "addis" with a "nop", calculating only
a 16-bit offset when this is sufficient.
Note that it would be more efficient for the compiler to generate:
addis 3, 2, test_fn_static.si@toc@ha
lwz 4, test_fn_static.si@toc@l(3)
The current patch does not perform this optimization yet. This will be
addressed as a peephole optimization in a later patch.
For the moment, the default code model for 64-bit PowerPC will remain the
small code model. We plan to eventually change the default to medium code
model, which matches current upstream GCC behavior. Note that the different
code models are ABI-compatible, so code compiled with different models will
be linked and execute correctly.
I've tested the regression suite and the application/benchmark test suite in
two ways: Once with the patch as submitted here, and once with additional
logic to force medium code model as the default. The tests all compile
cleanly, with one exception. The mandel-2 application test fails due to an
unrelated ABI compatibility with passing complex numbers. It just so happens
that small code model was incredibly lucky, in that temporary values in
floating-point registers held the expected values needed by the external
library routine that was called incorrectly. My current thought is to correct
the ABI problems with _Complex before making medium code model the default,
to avoid introducing this "regression."
Here are a few comments on how the patch works, since the selection code
can be difficult to follow:
The existing logic for small code model defines three pseudo-instructions:
LDtoc for most uses, LDtocJTI for jump table addresses, and LDtocCPT for
constant pool addresses. These are expanded by SelectCodeCommon(). The
pseudo-instruction approach doesn't work for medium code model, because
we need to generate two instructions when we match the same pattern.
Instead, new logic in PPCDAGToDAGISel::Select() intercepts the TOC_ENTRY
node for medium code model, and generates an ADDIStocHA followed by either
a LDtocL or an ADDItocL. These new node types correspond naturally to
the sequences described above.
The addis/ld sequence is generated for the following cases:
* Jump table addresses
* Function addresses
* External global variables
* Tentative definitions of global variables (common linkage)
The addis/addi sequence is generated for the following cases:
* Constant pool entries
* File-scope static global variables
* Function-scope static variables
Expanding to the two-instruction sequences at select time exposes the
instructions to subsequent optimization, particularly scheduling.
The rest of the processing occurs at assembly time, in
PPCAsmPrinter::EmitInstruction. Each of the instructions is converted to
a "real" PowerPC instruction. When a TOC entry needs to be created, this
is done here in the same manner as for the existing LDtoc, LDtocJTI, and
LDtocCPT pseudo-instructions (I factored out a new routine to handle this).
I had originally thought that if a TOC entry was needed for LDtocL or
ADDItocL, it would already have been generated for the previous ADDIStocHA.
However, at higher optimization levels, the ADDIStocHA may appear in a
different block, which may be assembled textually following the block
containing the LDtocL or ADDItocL. So it is necessary to include the
possibility of creating a new TOC entry for those two instructions.
Note that for LDtocL, we generate a new form of LD called LDrs. This
allows specifying the @toc@l relocation for the offset field of the LD
instruction (i.e., the offset is replaced by a SymbolLo relocation).
When the peephole optimization described above is added, we will need
to do similar things for all immediate-form load and store operations.
The seven "mcm-n.ll" test cases are kept separate because otherwise the
intermingling of various TOC entries and so forth makes the tests fragile
and hard to understand.
The above assumes use of an external assembler. For use of the
integrated assembler, new relocations are added and used by
PPCELFObjectWriter. Testing is done with "mcm-obj.ll", which tests for
proper generation of the various relocations for the same sequences
tested with the external assembler.
llvm-svn: 168708
Chad Rosier [Tue, 27 Nov 2012 17:31:26 +0000 (17:31 +0000)]
Remove an extra semicolon.
llvm-svn: 168707
Rafael Espindola [Tue, 27 Nov 2012 16:16:02 +0000 (16:16 +0000)]
Fix examples.
llvm-svn: 168705
Ulrich Weigand [Tue, 27 Nov 2012 16:11:16 +0000 (16:11 +0000)]
Never use .lcomm on platforms where it does not accept an alignment
argument. Instead, use a pair of .local and .comm directives.
This avoids spurious differences between binaries built by the
integrated assembler vs. those built by the external assembler,
since the external assembler may impose alignment requirements
on .lcomm symbols where the integrated assembler does not.
llvm-svn: 168704