Kostya Serebryany [Thu, 20 Dec 2012 12:26:09 +0000 (12:26 +0000)]
[tsan] fix cmake build
llvm-svn: 170689
Alexander Kornienko [Thu, 20 Dec 2012 12:23:54 +0000 (12:23 +0000)]
Minor coding style issue: NULL -> 0
llvm-svn: 170688
Kostya Serebryany [Thu, 20 Dec 2012 12:11:52 +0000 (12:11 +0000)]
[asan] add memalign tests
llvm-svn: 170687
Evgeniy Stepanov [Thu, 20 Dec 2012 12:03:13 +0000 (12:03 +0000)]
Add __has_feature(memory_sanitizer).
llvm-svn: 170686
Kostya Serebryany [Thu, 20 Dec 2012 11:54:21 +0000 (11:54 +0000)]
[asan] add a flag poison_heap to allow better allocator benchmarking, implemenet malloc_stats() on Linux
llvm-svn: 170685
Alexander Kornienko [Thu, 20 Dec 2012 11:08:38 +0000 (11:08 +0000)]
ASTDumper coding style fixes.
http://llvm-reviews.chandlerc.com/D226
Patch by Philip Craig!
llvm-svn: 170684
NAKAMURA Takumi [Thu, 20 Dec 2012 11:08:13 +0000 (11:08 +0000)]
clang/test/Driver/warning-options.cpp: Appease MSYS bash.
llvm-svn: 170683
NAKAMURA Takumi [Thu, 20 Dec 2012 10:35:18 +0000 (10:35 +0000)]
llvmbuild/main.py: Let LibraryDependencies.inc deterministic.
FYI, llvm and clang can be built deterministically between stage 2 and stage3, among iterative clean rebuilds, with GNU ar;
configure --disable-timestamps
make AR.Flags=crsD RANLIB=echo
llvm-svn: 170682
Dmitry Vyukov [Thu, 20 Dec 2012 10:21:30 +0000 (10:21 +0000)]
tsan: add java interface implementation stub
llvm-svn: 170681
Kostya Serebryany [Thu, 20 Dec 2012 08:53:41 +0000 (08:53 +0000)]
[asan] asan_allocator2: make all remaining tests pass.
llvm-svn: 170680
Dmitry Vyukov [Thu, 20 Dec 2012 07:47:04 +0000 (07:47 +0000)]
tsan: add Java interface
llvm-svn: 170679
Kostya Serebryany [Thu, 20 Dec 2012 07:26:33 +0000 (07:26 +0000)]
[asan] add ASAN_PCRE_DOTALL to two new multi-line regexps
llvm-svn: 170677
Craig Topper [Thu, 20 Dec 2012 07:15:54 +0000 (07:15 +0000)]
Formatting fixes. Remove some unnecessary 'else' after 'return'. No functional change.
llvm-svn: 170676
Craig Topper [Thu, 20 Dec 2012 07:09:41 +0000 (07:09 +0000)]
Removing trailing whitespace
llvm-svn: 170675
Reed Kotler [Thu, 20 Dec 2012 06:59:37 +0000 (06:59 +0000)]
Implement cfi_def_cfa_offset. "Make check" test case for this comming in the
next few days but it's already tested a lot from test-suite and works fine.
This patch completes almost 100% pass of test-suite for mips 16.
llvm-svn: 170674
Reed Kotler [Thu, 20 Dec 2012 06:57:00 +0000 (06:57 +0000)]
There is one more patch to finish large frames. Make sure we assert
on code that has large frames which will not yet compile correctly.
llvm-svn: 170673
Jyotsna Verma [Thu, 20 Dec 2012 06:52:46 +0000 (06:52 +0000)]
Add constant extender support to GP-relative load/store instructions.
llvm-svn: 170672
Jyotsna Verma [Thu, 20 Dec 2012 06:45:39 +0000 (06:45 +0000)]
Add TSFlags to ALU32 type instructions for constant-extender/Relationship maps.
llvm-svn: 170671
Alexey Samsonov [Thu, 20 Dec 2012 06:16:50 +0000 (06:16 +0000)]
[ASan] don't use -fsanitize=address linker flag for unit tests on Android
llvm-svn: 170670
Reed Kotler [Thu, 20 Dec 2012 06:06:35 +0000 (06:06 +0000)]
set register class properly for mips16 here
llvm-svn: 170669
Rafael Espindola [Thu, 20 Dec 2012 05:13:09 +0000 (05:13 +0000)]
Undefine PPC harder.
This was causing a build failure while trying to build on ppc ubuntu 12.10 with
cmake.
llvm-svn: 170668
Reed Kotler [Thu, 20 Dec 2012 05:09:15 +0000 (05:09 +0000)]
This assert is overly restrictive and does not work for mips16.
llvm-svn: 170667
Richard Smith [Thu, 20 Dec 2012 05:00:13 +0000 (05:00 +0000)]
Move C++ name demangling support from ubsan into sanitizer_common.
llvm-svn: 170666
Reed Kotler [Thu, 20 Dec 2012 04:44:58 +0000 (04:44 +0000)]
Turn on register scavenger for Mips 16
We use an unused Mips 32 register for the emergency slot
instead of using the stack.
llvm-svn: 170665
Akira Hatanaka [Thu, 20 Dec 2012 04:27:52 +0000 (04:27 +0000)]
[mips] Refactor SLT (set on less than) instructions. Separate encoding
information from the rest.
llvm-svn: 170664
Akira Hatanaka [Thu, 20 Dec 2012 04:22:39 +0000 (04:22 +0000)]
[mips] Refactor unconditional branch instruction. Separate encoding information
from the rest.
llvm-svn: 170663
Richard Smith [Thu, 20 Dec 2012 04:20:28 +0000 (04:20 +0000)]
Implement std::is_base_of for the case where we don't have a compiler
intrinsic. This relies upon the fact that overload resolution does not check
access and ambiguity for a derived-to-base conversion. This passes all
is_base_of tests in the test suite.
llvm-svn: 170662
Akira Hatanaka [Thu, 20 Dec 2012 04:20:09 +0000 (04:20 +0000)]
[mips] Remove asm string parameter from pseudo instructions. Add InstrItinClass
parameter.
llvm-svn: 170661
Akira Hatanaka [Thu, 20 Dec 2012 04:15:30 +0000 (04:15 +0000)]
[mips] Delete definition of CPRESTORE instruction.
llvm-svn: 170660
Akira Hatanaka [Thu, 20 Dec 2012 04:13:23 +0000 (04:13 +0000)]
[mips] Refactor conditional branch instructions with one register operand.
Separate encoding information from the rest.
llvm-svn: 170659
Richard Smith [Thu, 20 Dec 2012 04:11:02 +0000 (04:11 +0000)]
Don't use isa<CallInst>(this) in the constructor for CallInst's base class.
This has undefined behavior, because the classof implementation attempts to
access parts of the not-yet-constructed derived class. Found by clang
-fsanitize=vptr.
llvm-svn: 170658
Akira Hatanaka [Thu, 20 Dec 2012 04:10:13 +0000 (04:10 +0000)]
[mips] Refactor conditional branch instructions with two register operands.
Separate encoding information from the rest.
llvm-svn: 170657
Reed Kotler [Thu, 20 Dec 2012 04:07:42 +0000 (04:07 +0000)]
fix most of remaining issues with large frames.
these patches are tested a lot by test-suite but
make check tests are forthcoming once the next
few patches that complete this are committed.
with the next few patches the pass rate for mips16 is
near 100%
llvm-svn: 170656
Akira Hatanaka [Thu, 20 Dec 2012 04:06:06 +0000 (04:06 +0000)]
[mips] Use "or $r0, $r1, $zero" instead of "addu $r0, $zero, $r1" to copy
physical register $r1 to $r0.
GNU disassembler recognizes an "or" instruction as a "move", and this change
makes the disassembled code easier to read.
Original patch by Reed Kotler.
llvm-svn: 170655
Richard Smith [Thu, 20 Dec 2012 04:04:17 +0000 (04:04 +0000)]
Fix use-before-construction of X86TargetLowering.
llvm-svn: 170654
Richard Smith [Thu, 20 Dec 2012 04:02:58 +0000 (04:02 +0000)]
Don't use -1 as a value of an unsigned 7-bit enumeration; that has undefined
behavior and violates the !range constraints we put on loads of this enum.
Found by clang -fsanitize=enum.
llvm-svn: 170653
Richard Smith [Thu, 20 Dec 2012 03:59:24 +0000 (03:59 +0000)]
Don't leave IsUnsigned uninitialized in a default-constructed APSInt. Copying
such a structure has undefined behavior. Caught by -fsanitize=bool.
llvm-svn: 170652
Akira Hatanaka [Thu, 20 Dec 2012 03:52:08 +0000 (03:52 +0000)]
[mips] Change the order of template parameters. Move the default parameters to
the end.
llvm-svn: 170651
Akira Hatanaka [Thu, 20 Dec 2012 03:48:24 +0000 (03:48 +0000)]
[mips] Refactor shift instructions with register operands. Separate encoding
information from the rest.
llvm-svn: 170650
Akira Hatanaka [Thu, 20 Dec 2012 03:44:41 +0000 (03:44 +0000)]
[mips] Refactor shift immediate instructions. Separate encoding information
from the rest.
llvm-svn: 170649
Akira Hatanaka [Thu, 20 Dec 2012 03:40:03 +0000 (03:40 +0000)]
[mips] Refactor arithmetic and logic instructions with immediate operands.
Separate encoding information from the rest.
llvm-svn: 170648
Akira Hatanaka [Thu, 20 Dec 2012 03:34:05 +0000 (03:34 +0000)]
[mips] Refactor arithmetic and logic instructions. Separate encoding
information from the rest.
llvm-svn: 170647
Sean Silva [Thu, 20 Dec 2012 03:32:39 +0000 (03:32 +0000)]
docs: Show TOC for GettingStarted.rst.
This is a pretty lengthy document, so put the table of contents in your
face so that it's easier to scope out the content.
This document is a mess currently and needs to be
refactored/revised/split-up.
llvm-svn: 170646
NAKAMURA Takumi [Thu, 20 Dec 2012 03:30:30 +0000 (03:30 +0000)]
clang/test/Misc/ast-dump-decl.c: Add triple. __thread is not supported on all targets.
llvm-svn: 170645
Michael Gottesman [Thu, 20 Dec 2012 03:23:52 +0000 (03:23 +0000)]
Updated link in LanguageExtensions.rst to point to BlockLanguageSpec.rst instead of BlockLanguageSpec.txt.
Thanks to Jean-Daniel Dupas for comments!
llvm-svn: 170644
Michael Gottesman [Thu, 20 Dec 2012 03:23:49 +0000 (03:23 +0000)]
Changed UTF-8 emdashes to --- ala Dmitri Gribenko's comments.
llvm-svn: 170643
Akira Hatanaka [Thu, 20 Dec 2012 03:00:16 +0000 (03:00 +0000)]
[mips] Delete ArithOverflowR and ArithOverflow and use ArithLogicR and
ArithLogicI as the instruction base classes.
llvm-svn: 170642
Jason Molenda [Thu, 20 Dec 2012 02:53:18 +0000 (02:53 +0000)]
Clarify the comments in PlatformDarwin::LocateExecutableScriptingResource
a bit.
llvm-svn: 170641
Richard Smith [Thu, 20 Dec 2012 02:47:01 +0000 (02:47 +0000)]
Fix another uninitialized bool member bug found by -fsanitize=bool. This one
appears to currently be benign (we happen to test the flags in the right
order, so we never depend on the uninitialized value).
llvm-svn: 170640
John McCall [Thu, 20 Dec 2012 02:45:14 +0000 (02:45 +0000)]
Robustify the @encode code and add support for _Atomic.
Fixes PR14628.
llvm-svn: 170639
Sean Silva [Thu, 20 Dec 2012 02:40:45 +0000 (02:40 +0000)]
docs: Clean up adornments.
For whatever reason the usage of '^^^' and '---' adornments were
reversed compared to the "canonical" style of the LLVM docs (which is
currently "the style used in SphinxQuickstartTemplate.rst"). This change
doesn't affect the document structure at all, I'm just doing it for
trivial stylistic consistency (the document content is *much* more
important---thanks Nadav for writing this up!).
Also, trim the adornments to be the same length as the section names.
llvm-svn: 170638
Sean Silva [Thu, 20 Dec 2012 02:23:25 +0000 (02:23 +0000)]
docs: ASCII-fy
llvm-svn: 170637
Richard Smith [Thu, 20 Dec 2012 02:22:15 +0000 (02:22 +0000)]
Fix code that attempted to produce a diagnostic with one DiagnosticEngine, then
produce a note for that diagnostic either with a different DiagnosticEngine or
after calling DiagnosticEngine::Reset(). That didn't make any sense, and did the
wrong thing if the original diagnostic was suppressed.
llvm-svn: 170636
Fariborz Jahanian [Thu, 20 Dec 2012 02:20:09 +0000 (02:20 +0000)]
Provide couple of DeclPrint tests for my last patch.
llvm-svn: 170635
Alexander Kornienko [Thu, 20 Dec 2012 02:09:13 +0000 (02:09 +0000)]
Implement AST dumper for Decls.
http://llvm-reviews.chandlerc.com/D52
Patch by Philip Craig!
llvm-svn: 170634
Enrico Granata [Thu, 20 Dec 2012 02:07:45 +0000 (02:07 +0000)]
<rdar://problem/
12903081>
When looking for the Python script to load for a given module, replace dots with underscores if there are any after stripping the extension
This means that for a module named foo.devel.xyz a file named foo_devel.py will be what we try to load
llvm-svn: 170633
Nadav Rotem [Thu, 20 Dec 2012 02:00:02 +0000 (02:00 +0000)]
Loop Vectorizer: Enable if-conversion.
llvm-svn: 170632
Bill Wendling [Thu, 20 Dec 2012 01:36:59 +0000 (01:36 +0000)]
s/AttributesImpl/AttributeImpl/g This is going to apply to Attribute, not Attributes.
llvm-svn: 170631
Bob Wilson [Thu, 20 Dec 2012 01:36:20 +0000 (01:36 +0000)]
Do not introduce vector operations in functions marked with noimplicitfloat.
<rdar://problem/
12879313>
llvm-svn: 170630
Jim Grosbach [Thu, 20 Dec 2012 01:14:48 +0000 (01:14 +0000)]
Clean up some DOxygen comments.
llvm-svn: 170629
Jim Grosbach [Thu, 20 Dec 2012 01:14:45 +0000 (01:14 +0000)]
Clean up some DOxygen comments.
llvm-svn: 170628
Richard Smith [Thu, 20 Dec 2012 01:05:39 +0000 (01:05 +0000)]
Fix an uninitialized member variable, found by -fsanitize=bool.
llvm-svn: 170627
Nadav Rotem [Thu, 20 Dec 2012 00:49:56 +0000 (00:49 +0000)]
whitespace
llvm-svn: 170626
Anna Zaks [Thu, 20 Dec 2012 00:38:25 +0000 (00:38 +0000)]
[analyzer] Add the pointer escaped callback.
Instead of using several callbacks to identify the pointer escape event,
checkers now can register for the checkPointerEscape.
Converted the Malloc checker to use the new callback.
SimpleStreamChecker will be converted next.
llvm-svn: 170625
Michael J. Spencer [Thu, 20 Dec 2012 00:37:24 +0000 (00:37 +0000)]
[ELF][Reader] Ignore SHF_GROUP when switching on flags.
llvm-svn: 170624
Michael J. Spencer [Thu, 20 Dec 2012 00:37:10 +0000 (00:37 +0000)]
Style fixes.
llvm-svn: 170623
Nadav Rotem [Thu, 20 Dec 2012 00:29:18 +0000 (00:29 +0000)]
doc: resize the image.
llvm-svn: 170622
NAKAMURA Takumi [Thu, 20 Dec 2012 00:22:11 +0000 (00:22 +0000)]
Target/R600: Update MIB according to r170588.
llvm-svn: 170620
Andrew Kaylor [Thu, 20 Dec 2012 00:19:27 +0000 (00:19 +0000)]
Setting TestInferiorChanged.py to expectedFailureLinux because of bugzilla 14662.
Note that the test actually fails before reaching the assertion described in that bug, but the failure is a trivial matter of adding OS-conditional expected stop reason strings (as noted by new comments in this patch). However, with the stop reason strings changed, the test fails for the more substantial reason noted in bugzilla 14662.
llvm-svn: 170619
Nadav Rotem [Thu, 20 Dec 2012 00:03:36 +0000 (00:03 +0000)]
Doc: update the chart.
llvm-svn: 170618
Argyrios Kyrtzidis [Wed, 19 Dec 2012 23:55:44 +0000 (23:55 +0000)]
[preprocessor] When "merging" macro argument tokens into one SLocEntry chunk,
make sure they came from the same kind of FileIDs.
Thanks to Abramo Bagnara for providing the test case.
llvm-svn: 170616
Bill Wendling [Wed, 19 Dec 2012 23:55:43 +0000 (23:55 +0000)]
Add a context so that once we uniquify strings we can access them easily.
llvm-svn: 170615
Chad Rosier [Wed, 19 Dec 2012 23:47:16 +0000 (23:47 +0000)]
Typo.
llvm-svn: 170613
Enrico Granata [Wed, 19 Dec 2012 23:42:07 +0000 (23:42 +0000)]
<rdar://problem/
12446222>
Implement the ability for Python commands to be interrupted by pressing CTRL+C
Also add a new Mutex subclass that attempts to be helpful for debugging by logging actions performed on it
FYI of all interested - there is a separate deadlocking issue related to how LLDB dispatches CTRL+C that might cause LLDB to deadlock upon pressing CTRL+C while in a Python command.
This is not a regression, and was just previously masked by us not even trying to bail out of Python commands, so that it would not be clear from a user perspective whether we were
deadlocked or stuck in an inconsistent state within the Python interpreter.
llvm-svn: 170612
Chad Rosier [Wed, 19 Dec 2012 23:41:50 +0000 (23:41 +0000)]
[driver] Have -isysroot warn on nonexistent paths.
rdar://
12282267
llvm-svn: 170611
Jim Grosbach [Wed, 19 Dec 2012 23:38:53 +0000 (23:38 +0000)]
MC: Add MCInstrDesc::mayAffectControlFlow() method.
MC disassembler clients (LLDB) are interested in querying if an
instruction may affect control flow other than by virtue of being
an explicit branch instruction. For example, instructions which
write directly to the PC on some architectures.
llvm-svn: 170610
Jim Grosbach [Wed, 19 Dec 2012 23:38:49 +0000 (23:38 +0000)]
Add isSubRegisterEq() and isSuperRegisterEq().
isSub and isSuper return false if RegA == RegB. Add variants which also
include the identity function.
llvm-svn: 170609
Jim Grosbach [Wed, 19 Dec 2012 23:38:46 +0000 (23:38 +0000)]
Move isSubRegister() and isSuperRegister to MCRegisterInfo.
These were defined on TargetRegisterInfo, but they don't use any information
that's not available in MCRegisterInfo, so sink them down to be available
at the MC layer.
llvm-svn: 170608
Jim Grosbach [Wed, 19 Dec 2012 23:38:44 +0000 (23:38 +0000)]
Fix doc comment. '///' not '//'.
llvm-svn: 170607
Fariborz Jahanian [Wed, 19 Dec 2012 23:36:00 +0000 (23:36 +0000)]
Audit DeclPrinter with -ast-dump on Cocoa.h and
fix any bad objectiveC syntax coming out of
DeclPrinter. This is on going. Also, introduce a new
PrintPolicy and use it as needed when declaration tag
is to be produced via DeclPrinter.
llvm-svn: 170606
Andrew Kaylor [Wed, 19 Dec 2012 23:31:12 +0000 (23:31 +0000)]
Mark register tests as expectedFailureLinux because of bugzilla reports 14600 and 14661.
llvm-svn: 170605
Daniel Malea [Wed, 19 Dec 2012 23:22:11 +0000 (23:22 +0000)]
Disable confirmation prompt in TestProcessLaunch (to avoid hang on Linux once bug is fixed)
llvm-svn: 170603
Michael Ilseman [Wed, 19 Dec 2012 23:17:20 +0000 (23:17 +0000)]
Refactor isIntrinsic() to be quicker, and change classof() (and thus, isa<IntrinsicInst>()) to use it. This decreases the number of occurrences of the slow-path string matching performed by getIntrinsicID().
llvm-svn: 170602
Sean Callanan [Wed, 19 Dec 2012 23:05:01 +0000 (23:05 +0000)]
This patch removes the SymbolFileSymtab support
for reporting class types from Objective-C runtime
class symbols. Instead, LLDB now queries the
Objective-C runtime for class types.
We have also added a (minimal) Objective-C runtime
type vendor for Objective-C runtime version 1, to
prevent regressions when calling class methods in
the V1 runtime.
Other components of this fix include:
- We search the Objective-C runtime in a few more
places.
- We enable enumeration of all members of
Objective-C classes, which Clang does in certain
circumstances.
- SBTarget::FindFirstType and SBTarget::FindTypes
now query the Objective-C runtime as needed.
- I fixed several test cases.
<rdar://problem/
12885034>
llvm-svn: 170601
Bill Wendling [Wed, 19 Dec 2012 22:42:22 +0000 (22:42 +0000)]
s/AttributeListImpl/AttributeSetImpl/g to match the namechange of AttributeList.
llvm-svn: 170600
Jakob Stoklund Olesen [Wed, 19 Dec 2012 22:35:46 +0000 (22:35 +0000)]
Always use addOperand(MF, MO) from MachineInstrBuilder.
The single-argument MachineInstr::addOperand(MO) will be removed soon.
llvm-svn: 170599
Argyrios Kyrtzidis [Wed, 19 Dec 2012 22:21:18 +0000 (22:21 +0000)]
[PCH/Modules] In ASTReader::completeVisibleDeclsMap, make sure to visit all
modules when getting the decls for a namespace or translation unit.
Otherwise the code-completion results will not be complete.
rdar://
12889089
llvm-svn: 170596
Dmitri Gribenko [Wed, 19 Dec 2012 22:13:01 +0000 (22:13 +0000)]
Add a missing 'else'. Found by grep '} if'
No testcase because it is apparently not so trivial to construct.
llvm-svn: 170595
Tom Stellard [Wed, 19 Dec 2012 22:10:35 +0000 (22:10 +0000)]
R600: Add entry in CODE_OWNERS.TXT
llvm-svn: 170594
Tom Stellard [Wed, 19 Dec 2012 22:10:34 +0000 (22:10 +0000)]
R600: Remove unecessary VREG alignment.
Unlike SGPRs VGPRs doesn't need to be aligned.
Patch by: Christian König
Reviewed-by: Tom Stellard <thomas.stellard@amd.com>
Tested-by: Michel Dänzer <michel.daenzer@amd.com>
Signed-off-by: Christian König <deathsimple@vodafone.de>
llvm-svn: 170593
Tom Stellard [Wed, 19 Dec 2012 22:10:33 +0000 (22:10 +0000)]
R600: control flow optimization
Branch if we have enough instructions so that it makes sense.
Also remove branches if they don't make sense.
Patch by: Christian König
Reviewed-by: Tom Stellard <thomas.stellard@amd.com>
Tested-by: Michel Dänzer <michel.daenzer@amd.com>
Signed-off-by: Christian König <deathsimple@vodafone.de>
llvm-svn: 170592
Tom Stellard [Wed, 19 Dec 2012 22:10:31 +0000 (22:10 +0000)]
R600: New control flow for SI v2
This patch replaces the control flow handling with a new
pass which structurize the graph before transforming it to
machine instruction. This has a couple of different advantages
and currently fixes 20 piglit tests without a single regression.
It is now a general purpose transformation that could be not
only be used for SI/R6xx, but also for other hardware
implementations that use a form of structurized control flow.
v2: further cleanup, fixes and documentation
Patch by: Christian König
Signed-off-by: Christian König <deathsimple@vodafone.de>
Reviewed-by: Tom Stellard <thomas.stellard@amd.com>
Tested-by: Michel Dänzer <michel.daenzer@amd.com>
llvm-svn: 170591
Dmitri Gribenko [Wed, 19 Dec 2012 22:06:59 +0000 (22:06 +0000)]
Documentation: UsersManual.rst: some cleanups after semi-automatic conversion
llvm-svn: 170590
Eric Christopher [Wed, 19 Dec 2012 22:02:53 +0000 (22:02 +0000)]
Split out abbreviations for the skeleton info from the rest of
the abbreviations. Part of implementing split dwarf.
llvm-svn: 170589
Jakob Stoklund Olesen [Wed, 19 Dec 2012 21:31:56 +0000 (21:31 +0000)]
Remove the explicit MachineInstrBuilder(MI) constructor.
Use the version that also takes an MF reference instead.
It would technically be possible to extract an MF reference from the MI
as MI->getParent()->getParent(), but that would not work for MIs that
are not inserted into any basic block.
Given the reasonably small number of places this constructor was used at
all, I preferred the compile time check to a run time assertion.
llvm-svn: 170588
Nadav Rotem [Wed, 19 Dec 2012 20:47:04 +0000 (20:47 +0000)]
Fix a bug that was found by building clang with -fsanitize.
I introduced it in r166785. PR14291.
If TD is unavailable use getScalarSizeInBits, but don't optimize
pointers or vectors of pointers.
llvm-svn: 170586
Meador Inge [Wed, 19 Dec 2012 20:16:40 +0000 (20:16 +0000)]
docs: Fix title underline warnings
Building Vectorizers.rst produces a few warnings of the form:
WARNING: Title underline too short.
Fixed by adding the extra needed dashes under the title.
llvm-svn: 170582
Evan Cheng [Wed, 19 Dec 2012 20:16:09 +0000 (20:16 +0000)]
LLVM sdisel normalize bit extraction of the form:
((x & 0xff00) >> 8) << 2
to
(x >> 6) & 0x3fc
This is general goodness since it folds a left shift into the mask. However,
the trailing zeros in the mask prevents the ARM backend from using the bit
extraction instructions. And worse since the mask materialization may require
an addition instruction. This comes up fairly frequently when the result of
the bit twiddling is used as memory address. e.g.
= ptr[(x & 0xFF0000) >> 16]
We want to generate:
ubfx r3, r1, #16, #8
ldr.w r3, [r0, r3, lsl #2]
vs.
mov.w r9, #1020
and.w r2, r9, r1, lsr #14
ldr r2, [r0, r2]
Add a late ARM specific isel optimization to
ARMDAGToDAGISel::PreprocessISelDAG(). It folds the left shift to the
'base + offset' address computation; change the mask to one which doesn't have
trailing zeros and enable the use of ubfx.
Note the optimization has to be done late since it's target specific and we
don't want to change the DAG normalization. It's also fairly restrictive
as shifter operands are not always free. It's only done for lsh 1 / 2. It's
known to be free on some cpus and they are most common for address
computation.
This is a slight win for blowfish, rijndael, etc.
rdar://
12870177
llvm-svn: 170581
Benjamin Kramer [Wed, 19 Dec 2012 20:11:17 +0000 (20:11 +0000)]
Remove edis remnant.
llvm-svn: 170580
Roman Divacky [Wed, 19 Dec 2012 19:55:47 +0000 (19:55 +0000)]
Remove edis - the enhanced disassembler. Fixes PR14654.
llvm-svn: 170578