platform/upstream/llvm.git
15 years agoit turns out that isWeak() was basically dead anyway. Kill off SectionInfo :-/
Chris Lattner [Sat, 1 Aug 2009 21:46:23 +0000 (21:46 +0000)]
it turns out that isWeak() was basically dead anyway.  Kill off SectionInfo :-/

llvm-svn: 77812

15 years agoAdditional AuroraUX header path.
Eli Friedman [Sat, 1 Aug 2009 21:46:03 +0000 (21:46 +0000)]
Additional AuroraUX header path.

llvm-svn: 77811

15 years agodon't use isWeak anymore.
Chris Lattner [Sat, 1 Aug 2009 21:42:58 +0000 (21:42 +0000)]
don't use isWeak anymore.

llvm-svn: 77810

15 years agofix a layering violation by moving SectionKind out to its own header.
Chris Lattner [Sat, 1 Aug 2009 21:30:49 +0000 (21:30 +0000)]
fix a layering violation by moving SectionKind out to its own header.

llvm-svn: 77808

15 years agoMinor whitespace tidiness.
Dan Gohman [Sat, 1 Aug 2009 21:25:46 +0000 (21:25 +0000)]
Minor whitespace tidiness.

llvm-svn: 77807

15 years agoFix typos in comments.
Dan Gohman [Sat, 1 Aug 2009 21:25:00 +0000 (21:25 +0000)]
Fix typos in comments.

llvm-svn: 77806

15 years agowith the previous refactoring, fixme fixed!
Chris Lattner [Sat, 1 Aug 2009 21:21:43 +0000 (21:21 +0000)]
with the previous refactoring, fixme fixed!

llvm-svn: 77805

15 years agoupdate for API change.
Chris Lattner [Sat, 1 Aug 2009 21:14:30 +0000 (21:14 +0000)]
update for API change.

llvm-svn: 77804

15 years agoChange SectionKind to be a property that is true of a *section*, it
Chris Lattner [Sat, 1 Aug 2009 21:11:14 +0000 (21:11 +0000)]
Change SectionKind to be a property that is true of a *section*, it
should have no state that is specific to particular globals in the
section.  In this case, it means the removal of the "isWeak" and
"ExplicitSection" bits.  MCSection uses the new form of SectionKind.

To handle isWeak, I introduced a new SectionInfo class, which is
SectionKind + isWeak, and it is used by the part of the code generator
that does classification of a specific global.

The ExplicitSection disappears.  It is moved onto MCSection as a new
"IsDirective" bit.  Since the Name of a section is either a section
or directive, it makes sense to keep this bit in MCSection.  Ultimately
the creator of MCSection should canonicalize (e.g.) .text to whatever
the actual section is.

llvm-svn: 77803

15 years agoDon't error on C++ linkage; if we don't support that, there are
Eli Friedman [Sat, 1 Aug 2009 20:48:04 +0000 (20:48 +0000)]
Don't error on C++ linkage; if we don't support that, there are
serious issues :)

llvm-svn: 77800

15 years agoadd a note about the build bots
Chris Lattner [Sat, 1 Aug 2009 19:25:25 +0000 (19:25 +0000)]
add a note about the build bots

llvm-svn: 77796

15 years agoMinor code cleanups.
Dan Gohman [Sat, 1 Aug 2009 19:14:37 +0000 (19:14 +0000)]
Minor code cleanups.

llvm-svn: 77795

15 years agoPrint the target flags as an int instead of a char, as they aren't
Dan Gohman [Sat, 1 Aug 2009 19:13:38 +0000 (19:13 +0000)]
Print the target flags as an int instead of a char, as they aren't
actually characters.

llvm-svn: 77794

15 years agoUse the default copy ctor and copy-assignment operators.
Dan Gohman [Sat, 1 Aug 2009 19:11:31 +0000 (19:11 +0000)]
Use the default copy ctor and copy-assignment operators.

llvm-svn: 77793

15 years agoAdd nounwind to this test.
Dan Gohman [Sat, 1 Aug 2009 19:11:04 +0000 (19:11 +0000)]
Add nounwind to this test.

llvm-svn: 77792

15 years agoAdd newline at end of file to remove gcc warning.
Nick Lewycky [Sat, 1 Aug 2009 19:09:44 +0000 (19:09 +0000)]
Add newline at end of file to remove gcc warning.

llvm-svn: 77791

15 years agoRemove now empty unit test directory.
Benjamin Kramer [Sat, 1 Aug 2009 19:07:29 +0000 (19:07 +0000)]
Remove now empty unit test directory.

llvm-svn: 77790

15 years agodaniel says it's fine to nuke this.
Chris Lattner [Sat, 1 Aug 2009 18:38:21 +0000 (18:38 +0000)]
daniel says it's fine to nuke this.

llvm-svn: 77789

15 years agoupdate for new api
Chris Lattner [Sat, 1 Aug 2009 18:35:49 +0000 (18:35 +0000)]
update for new api

llvm-svn: 77788

15 years agoAll MCSections are now required to have a SectionKind.
Chris Lattner [Sat, 1 Aug 2009 18:25:49 +0000 (18:25 +0000)]
All MCSections are now required to have a SectionKind.

llvm-svn: 77787

15 years agoAdd temp C++ include path for AuroraUX; patch by Edward O'Callaghan.
Eli Friedman [Sat, 1 Aug 2009 17:10:21 +0000 (17:10 +0000)]
Add temp C++ include path for AuroraUX; patch by Edward O'Callaghan.

llvm-svn: 77786

15 years agoHack to make this test work on platforms which aren't Macs. Fixing this
Eli Friedman [Sat, 1 Aug 2009 16:37:18 +0000 (16:37 +0000)]
Hack to make this test work on platforms which aren't Macs.  Fixing this
myself because I'm getting tired of seeing the red buildbots, which have
been red since 5:30PM PDT last night.

Proposed supplement to developer policy: committers should make sure to
be around to watch for buildbot failures after committing.

llvm-svn: 77785

15 years agolit: Add internal script execution.
Daniel Dunbar [Sat, 1 Aug 2009 10:18:01 +0000 (10:18 +0000)]
lit: Add internal script execution.
 - Off by default, you can test it with the --no-sh argument.

 - For me it works for all but 3 tests, but there a number of FIXMEs and QOI
   issues:
     o Redirection isn't completely accurate -- in practice it can't portably
       be, but I would like to error out if someone writes something which isn't
       going to work. This is the source of the 3 test failures.

     o Some pipe configurations have the potential to deadlock.

     o It is significantly slower when multithreaded. I believe this is due to
       locking happening under the hood, there is probably some kind of solution
       but I haven't investigated yet.

     o Log output is ugly.

llvm-svn: 77784

15 years agolit: Two more sh lex/parse bugs (but its so simple!)
Daniel Dunbar [Sat, 1 Aug 2009 09:41:09 +0000 (09:41 +0000)]
lit: Two more sh lex/parse bugs (but its so simple!)
 - Empty arguments weren't handled correctly.
 - Escapes outside quoted strings weren't handled.

llvm-svn: 77783

15 years agoThis is a fairly large patch, which resulted from a cascade of changes
Ted Kremenek [Sat, 1 Aug 2009 06:17:29 +0000 (06:17 +0000)]
This is a fairly large patch, which resulted from a cascade of changes
made to RegionStore (and related classes) in order to handle some
analyzer failures involving casts and manipulation of symbolic memory.

The root of the change is in StoreManager::CastRegion().  Instead of
using ad hoc heuristics to decide when to layer an ElementRegion on a
casted MemRegion, we now always layer an ElementRegion when the cast
type is different than the original type of the region.  This carries
the current cast information associated with a region around without
resorting to the error prone recording of "casted types" in GRState.

Along with this new policy of layering ElementRegions, I added a new
algorithm to strip away existing ElementRegions when they simply
represented casts of a base memory object.  This algorithm computes
the raw "byte offset" that an ElementRegion represents from the base
region, and allows the new ElementRegion to be based off that offset.
The added benefit is that this naturally handles a series of casts of
a MemRegion without building up a set of redundant ElementRegions
(thus canonicalizing the region view).

Other related changes that cascaded from this one (as tests were
failing in RegionStore):

- Revamped RegionStoreManager::InvalidateRegion() to completely remove
  all bindings and default values from a region and all subregions.
  Now invalidated fields are not bound directly to new symbolic
  values; instead the base region has a "default" symbol value from
  which "derived symbols" can be created.  The main advantage of this
  approach is that it allows us to invalidate a region hierarchy and
  then lazily instantiate new values no matter how deep the hierarchy
  went (i.e., regardless of the number of field accesses,
  e.g. x->f->y->z->...).  The previous approach did not do this.

- Slightly reworked RegionStoreManager::RemoveDeadBindings() to also
  incorporate live symbols and live regions that do not have direct
  bindings but also have "default values" used for lazy instantiation.
  The changes to 'InvalidateRegion' revealed that these were necessary
  in order to achieve lazy instantiation of values in the region store
  with those bindings being removed too early.

- The changes to InvalidateRegion() and RemoveDeadBindings() revealed
  a serious bug in 'getSubRegionMap()' where not all region -> subregion
  relationships involved in actually bindings (explicit and implicit)
  were being recorded.  This has been fixed by using a worklist algorithm
  to iteratively fill in the region map.

- Added special support to RegionStoreManager::Bind()/Retrieve() to handle
  OSAtomicCompareAndSwap in light of the new 'CastRegion' changes and the
  layering of ElementRegions.

- Fixed a bug in SymbolReaper::isLive() where derived symbols were not
  being marked live if the symbol they were derived from was also live.
  This fix was critical for getting lazy instantiation in RegionStore
  to work.

- Tidied up the implementation of ValueManager::getXXXSymbolVal() methods
  to use SymbolManager::canSymbolicate() to decide whether or not a
  symbol should be symbolicated.

- 'test/Analysis/misc-ps-xfail.m' now passes; that test case has been
  moved to 'test/Analysis/misc-ps.m'.

- Tweaked some pretty-printing of MemRegions, and implemented
  'ElementRegion::getRawOffset()' for use with the CastRegion changes.

llvm-svn: 77782

15 years agoWorkaround a couple of Darwin assembler bugs.
Evan Cheng [Sat, 1 Aug 2009 06:13:52 +0000 (06:13 +0000)]
Workaround a couple of Darwin assembler bugs.

llvm-svn: 77781

15 years agoOk, ok, I give in. Fix tests for unused result warning.
Daniel Dunbar [Sat, 1 Aug 2009 06:07:15 +0000 (06:07 +0000)]
Ok, ok, I give in. Fix tests for unused result warning.

llvm-svn: 77780

15 years agoTemporarily disable out-of-bounds checking. The current checking logic will not...
Ted Kremenek [Sat, 1 Aug 2009 05:59:39 +0000 (05:59 +0000)]
Temporarily disable out-of-bounds checking.  The current checking logic will not work quite right with the changes I'm about to commit.

llvm-svn: 77779

15 years agoUse %S, not `pwd`, and enable a line that *does* work.
Daniel Dunbar [Sat, 1 Aug 2009 05:57:38 +0000 (05:57 +0000)]
Use %S, not `pwd`, and enable a line that *does* work.
 - Doug, please check.

llvm-svn: 77778

15 years agolit: Fix two sh lexing bugs.
Daniel Dunbar [Sat, 1 Aug 2009 05:52:04 +0000 (05:52 +0000)]
lit: Fix two sh lexing bugs.
 - '\\\\' inside a "..." string becomes '\\'.
 - The '<' token wasn't being recognized.

llvm-svn: 77777

15 years ago'unset' isn't needed in these tests anymore, we always run with a controlled
Daniel Dunbar [Sat, 1 Aug 2009 04:51:30 +0000 (04:51 +0000)]
'unset' isn't needed in these tests anymore, we always run with a controlled
environment.

llvm-svn: 77776

15 years agoGive MachineFunctionAnalysis a destructor so it can verify that
Dan Gohman [Sat, 1 Aug 2009 04:19:43 +0000 (04:19 +0000)]
Give MachineFunctionAnalysis a destructor so it can verify that
that it released its allocated memory.

llvm-svn: 77775

15 years agoDelete a redundant variable.
Dan Gohman [Sat, 1 Aug 2009 04:18:29 +0000 (04:18 +0000)]
Delete a redundant variable.

llvm-svn: 77774

15 years agolit: Fix a name lookup problem, which only occurred on a race condition. This is
Daniel Dunbar [Sat, 1 Aug 2009 04:11:36 +0000 (04:11 +0000)]
lit: Fix a name lookup problem, which only occurred on a race condition. This is
why dynamic binding sucks.

llvm-svn: 77773

15 years agolit: Pull a few more variables into the TestingConfig object.
Daniel Dunbar [Sat, 1 Aug 2009 04:06:02 +0000 (04:06 +0000)]
lit: Pull a few more variables into the TestingConfig object.

llvm-svn: 77772

15 years agoMinor code simplifications.
Dan Gohman [Sat, 1 Aug 2009 03:51:09 +0000 (03:51 +0000)]
Minor code simplifications.

llvm-svn: 77769

15 years agoMinor code simplifications.
Dan Gohman [Sat, 1 Aug 2009 03:42:59 +0000 (03:42 +0000)]
Minor code simplifications.

llvm-svn: 77768

15 years agoAvoid a problem with ulimit on Solaris & friends, patch by Edward O'Callaghan!
Daniel Dunbar [Sat, 1 Aug 2009 03:37:54 +0000 (03:37 +0000)]
Avoid a problem with ulimit on Solaris & friends, patch by Edward O'Callaghan!

llvm-svn: 77767

15 years agolit: Don't use threads when only running one test, or with -j 1.
Daniel Dunbar [Sat, 1 Aug 2009 03:35:40 +0000 (03:35 +0000)]
lit: Don't use threads when only running one test, or with -j 1.

llvm-svn: 77766

15 years agoMultiTestRunner: Add 'sh' parsing to ShUtil.
Daniel Dunbar [Sat, 1 Aug 2009 03:22:27 +0000 (03:22 +0000)]
MultiTestRunner: Add 'sh' parsing to ShUtil.

llvm-svn: 77765

15 years agoSplit t2MOVCCs since some assemblers do not recognize mov shifted register alias...
Evan Cheng [Sat, 1 Aug 2009 01:43:45 +0000 (01:43 +0000)]
Split t2MOVCCs since some assemblers do not recognize mov shifted register alias with predicate.

llvm-svn: 77764

15 years agoMore warnings for unused expressions.
Anders Carlsson [Sat, 1 Aug 2009 01:39:59 +0000 (01:39 +0000)]
More warnings for unused expressions.

llvm-svn: 77763

15 years agoMake default ctor for ImmutableSet::iterator public.
Ted Kremenek [Sat, 1 Aug 2009 01:28:23 +0000 (01:28 +0000)]
Make default ctor for ImmutableSet::iterator public.

llvm-svn: 77762

15 years agollvm-mc: More quoted identifier support.
Daniel Dunbar [Sat, 1 Aug 2009 00:48:30 +0000 (00:48 +0000)]
llvm-mc: More quoted identifier support.

llvm-svn: 77761

15 years agoSelectionDAGISel no longer needs to check hasAvailableExternallyLinkage,
Dan Gohman [Sat, 1 Aug 2009 00:42:23 +0000 (00:42 +0000)]
SelectionDAGISel no longer needs to check hasAvailableExternallyLinkage,
as it is now a MachineFunctionPass, and MachineFunctionPass now handles
this.

llvm-svn: 77760

15 years agoMake UnreachableMachineBlockElim preserve MachineDominatorTree and
Dan Gohman [Sat, 1 Aug 2009 00:34:30 +0000 (00:34 +0000)]
Make UnreachableMachineBlockElim preserve MachineDominatorTree and
MachineLoopInfo.

llc now runs MachineLoopInfo and MachineDominatorTree only twice, instead
of three times.

llvm-svn: 77759

15 years agoThe X86 maximal stack alignment calculator preserves the CFG. Also,
Dan Gohman [Sat, 1 Aug 2009 00:31:02 +0000 (00:31 +0000)]
The X86 maximal stack alignment calculator preserves the CFG. Also,
be more careful about the return value of runOnMachineFunction.

llvm-svn: 77758

15 years agoX86 floating-point passes don't modify the CFG.
Dan Gohman [Sat, 1 Aug 2009 00:26:16 +0000 (00:26 +0000)]
X86 floating-point passes don't modify the CFG.

llvm-svn: 77757

15 years agoFix Thumb2 function call isel. Thumb1 and Thumb2 should share the same
Evan Cheng [Sat, 1 Aug 2009 00:16:10 +0000 (00:16 +0000)]
Fix Thumb2 function call isel. Thumb1 and Thumb2 should share the same
instructions for calls since BL and BLX are always 32-bit long and BX is always
16-bit long.

Also, we should be using BLX to call external function stubs.

llvm-svn: 77756

15 years agoUse setPreservesAll in X86CodeEmitter.
Dan Gohman [Fri, 31 Jul 2009 23:44:16 +0000 (23:44 +0000)]
Use setPreservesAll in X86CodeEmitter.

llvm-svn: 77755

15 years agoUse setPreservesAll and setPreservesCFG in CodeGen passes.
Dan Gohman [Fri, 31 Jul 2009 23:37:33 +0000 (23:37 +0000)]
Use setPreservesAll and setPreservesCFG in CodeGen passes.

llvm-svn: 77754

15 years agoSelectionDAGISel does not "preserve all", since it makes lots of changes
Dan Gohman [Fri, 31 Jul 2009 23:36:22 +0000 (23:36 +0000)]
SelectionDAGISel does not "preserve all", since it makes lots of changes
to the MachineFunction.

llvm-svn: 77753

15 years agoUse a range insert instead of an explicit loop.
Dan Gohman [Fri, 31 Jul 2009 23:36:06 +0000 (23:36 +0000)]
Use a range insert instead of an explicit loop.

llvm-svn: 77752

15 years agoAdd beginnigs of rtti generation, wire up more of -fno-exceptions.
Mike Stump [Fri, 31 Jul 2009 23:15:31 +0000 (23:15 +0000)]
Add beginnigs of rtti generation, wire up more of -fno-exceptions.

llvm-svn: 77751

15 years agollvm-mc: Add -triple, and fix some typos
Daniel Dunbar [Fri, 31 Jul 2009 23:13:12 +0000 (23:13 +0000)]
llvm-mc: Add -triple, and fix some typos

llvm-svn: 77750

15 years agollvm-mc: Fix .s output to quote section & symbol names when necessary.
Daniel Dunbar [Fri, 31 Jul 2009 23:04:32 +0000 (23:04 +0000)]
llvm-mc: Fix .s output to quote section & symbol names when necessary.

llvm-svn: 77749

15 years agoPrivatize all but one of the remaining constant tables.
Owen Anderson [Fri, 31 Jul 2009 22:45:43 +0000 (22:45 +0000)]
Privatize all but one of the remaining constant tables.

llvm-svn: 77748

15 years agoAllow target intrinsics that return multiple values, i.e., struct types,
Bob Wilson [Fri, 31 Jul 2009 22:41:21 +0000 (22:41 +0000)]
Allow target intrinsics that return multiple values, i.e., struct types,
in SelectionDAGLowering::visitTargetIntrinsic.

This removes a bit of special-case code for vector types.  After staring
at it for a while, I managed to convince myself that it is not necessary.
The only case where TLI.getValueType() differs from MVT::getMVT is for iPTR,
so this code could potentially make a difference for a vector of pointers.
But, it looks like that is not supported.  Calling TLI.getValueType() on
a vector of pointers leads to the following sequence of calls:

TargetLowering::getValueType
MVT::getMVT
MVT::getVectorVT(iPTR, num elements)
MVT::getExtendedVectorVT
MVT::getTypeForMVT for iPTR
assertion fails "Type is not extended!"

So, unless I'm really missing something, this bit of code is irrelevant to
the current version of LLVM, which is consistent with the fact that I don't
see this code in other similar places.

llvm-svn: 77747

15 years agoAdd a warning.
Mikhail Glushenkov [Fri, 31 Jul 2009 22:24:20 +0000 (22:24 +0000)]
Add a warning.

llvm-svn: 77746

15 years agollvm-mc: A few more parsing / match tweaks.
Daniel Dunbar [Fri, 31 Jul 2009 22:22:54 +0000 (22:22 +0000)]
llvm-mc: A few more parsing / match tweaks.
 - Operands which are just a label should be parsed as immediates, not memory
   operands (from the assembler perspective).

 - Match a few more flavors of immediates.

 - Distinguish match functions for memory operands which don't take a segment
   register.

 - We match the .s for "hello world" now!

llvm-svn: 77745

15 years agot2BR_JT is mov pc, it's 2 byte long, not 4.
Evan Cheng [Fri, 31 Jul 2009 22:22:22 +0000 (22:22 +0000)]
t2BR_JT is mov pc, it's 2 byte long, not 4.

llvm-svn: 77744

15 years agoThumb2 movcc need .w suffix.
Evan Cheng [Fri, 31 Jul 2009 22:21:55 +0000 (22:21 +0000)]
Thumb2 movcc need .w suffix.

llvm-svn: 77743

15 years agoadd some comments on how this is *supposed* to work. We don't
Chris Lattner [Fri, 31 Jul 2009 22:18:14 +0000 (22:18 +0000)]
add some comments on how this is *supposed* to work.  We don't
need the PreferredEHDataFormat hook, but I have yet-more refactoring to
do before I can zap it.

llvm-svn: 77742

15 years agorearrange a conditional. Even if this weren't #if 0'd out, this would
Chris Lattner [Fri, 31 Jul 2009 22:03:47 +0000 (22:03 +0000)]
rearrange a conditional.  Even if this weren't #if 0'd out, this would
have no functionality change.

llvm-svn: 77741

15 years agoSimplify operand padding by keying off tabs in the asm stream. If
David Greene [Fri, 31 Jul 2009 21:57:10 +0000 (21:57 +0000)]
Simplify operand padding by keying off tabs in the asm stream.  If
padding is disabled, tabs get replaced by spaces except in the case of
the first operand, where the tab is output to line up the operands after
the mnemonics.

Add some better comments and eliminate redundant code.

Fix some testcases to not assume tabs.

llvm-svn: 77740

15 years agollvm-mc: Support quoted identifiers.
Daniel Dunbar [Fri, 31 Jul 2009 21:55:09 +0000 (21:55 +0000)]
llvm-mc: Support quoted identifiers.
 - Uses MCAsmToken::getIdentifier which returns the (sub)string representing the
   meaningfull contents a string or identifier token.

 - Directives aren't done yet.

llvm-svn: 77739

15 years agoRemove the old struct builder code.
Anders Carlsson [Fri, 31 Jul 2009 21:48:56 +0000 (21:48 +0000)]
Remove the old struct builder code.

llvm-svn: 77738

15 years agoAnd now we can generate a simple vtable. Still a work in progress...
Mike Stump [Fri, 31 Jul 2009 21:43:43 +0000 (21:43 +0000)]
And now we can generate a simple vtable.  Still a work in progress...

llvm-svn: 77737

15 years agoPreferredEHDataFormat is always call with data and global, but this whole
Chris Lattner [Fri, 31 Jul 2009 21:39:55 +0000 (21:39 +0000)]
PreferredEHDataFormat is always call with data and global, but this whole
thing is #if0'd out anyway.  Just simplify the code by reducing the interface.
Not deleting this is essential for Bill's continuing happiness.

llvm-svn: 77736

15 years agoMove code from EmitUnion directly into the function that handles cast-to-union.
Anders Carlsson [Fri, 31 Jul 2009 21:38:39 +0000 (21:38 +0000)]
Move code from EmitUnion directly into the function that handles cast-to-union.

llvm-svn: 77735

15 years agoFix unit tests.
Owen Anderson [Fri, 31 Jul 2009 21:38:10 +0000 (21:38 +0000)]
Fix unit tests.

llvm-svn: 77734

15 years agoMove the metadata constructors back to 2.5 syntax.
Owen Anderson [Fri, 31 Jul 2009 21:35:40 +0000 (21:35 +0000)]
Move the metadata constructors back to 2.5 syntax.

llvm-svn: 77733

15 years agoUse the struct builder for unions.
Anders Carlsson [Fri, 31 Jul 2009 21:34:04 +0000 (21:34 +0000)]
Use the struct builder for unions.

llvm-svn: 77732

15 years agoFixup spacing and 80-col violations.
Mike Stump [Fri, 31 Jul 2009 21:31:32 +0000 (21:31 +0000)]
Fixup spacing and 80-col violations.

llvm-svn: 77731

15 years agoUpdate unittest for LLVM API change.
Benjamin Kramer [Fri, 31 Jul 2009 20:56:31 +0000 (20:56 +0000)]
Update unittest for LLVM API change.

llvm-svn: 77730

15 years agoLibCallAliasAnalysis doesn't use TargetData.
Dan Gohman [Fri, 31 Jul 2009 20:56:29 +0000 (20:56 +0000)]
LibCallAliasAnalysis doesn't use TargetData.

llvm-svn: 77729

15 years agollvm-mc/X86: Sketch match functions for immediates and memory operands.
Daniel Dunbar [Fri, 31 Jul 2009 20:53:16 +0000 (20:53 +0000)]
llvm-mc/X86: Sketch match functions for immediates and memory operands.

Also, change scale value to always be 1 when unspecified to machine MachineInst
encoding.

llvm-svn: 77728

15 years agoMake TargetData optional in MemoryDependenceAnalysis.
Dan Gohman [Fri, 31 Jul 2009 20:53:12 +0000 (20:53 +0000)]
Make TargetData optional in MemoryDependenceAnalysis.

llvm-svn: 77727

15 years agomove emitUsedDirectiveFor to TargetLoweringObjectFile and rename it to
Chris Lattner [Fri, 31 Jul 2009 20:52:39 +0000 (20:52 +0000)]
move emitUsedDirectiveFor to TargetLoweringObjectFile and rename it to
indicate that it is a predicate, not an emitter.  This eliminates TAI
dependencies on Mangler and GlobalValue.

llvm-svn: 77726

15 years agoRemove an unnecessary header.
Dan Gohman [Fri, 31 Jul 2009 20:47:45 +0000 (20:47 +0000)]
Remove an unnecessary header.

llvm-svn: 77725

15 years agoremove the PPCLinuxTargetAsmInfo implementation of PreferredEHDataFormat,
Chris Lattner [Fri, 31 Jul 2009 20:43:26 +0000 (20:43 +0000)]
remove the PPCLinuxTargetAsmInfo implementation of PreferredEHDataFormat,
because it just calls the default impl.

Remove the PPCDarwinTargetAsmInfo version of PreferredEHDataFormat because
it just returns DW_EH_PE_absptr unless on 10.6.  However, 10.6 doesn't support
PPC, so the default impl is just fine.

llvm-svn: 77724

15 years agoremove a pointless override.
Chris Lattner [Fri, 31 Jul 2009 20:36:15 +0000 (20:36 +0000)]
remove a pointless override.

llvm-svn: 77723

15 years agoUpdate for LLVM API change.
Owen Anderson [Fri, 31 Jul 2009 20:28:54 +0000 (20:28 +0000)]
Update for LLVM API change.

llvm-svn: 77722

15 years agoMove a few more APIs back to 2.5 forms. The only remaining ones left to change back are
Owen Anderson [Fri, 31 Jul 2009 20:28:14 +0000 (20:28 +0000)]
Move a few more APIs back to 2.5 forms.  The only remaining ones left to change back are
metadata related, which I'm waiting on to avoid conflicting with Devang.

llvm-svn: 77721

15 years agoFix GVN's debug output, now that operator<< on Value* doesn't print
Dan Gohman [Fri, 31 Jul 2009 20:24:18 +0000 (20:24 +0000)]
Fix GVN's debug output, now that operator<< on Value* doesn't print
a trailing newline.

llvm-svn: 77719

15 years agoWhitespace and 80-col cleanup.
Eric Christopher [Fri, 31 Jul 2009 20:07:27 +0000 (20:07 +0000)]
Whitespace and 80-col cleanup.

llvm-svn: 77718

15 years ago- Convert the rest of the DOUTs to DEBUG+errs().
Bill Wendling [Fri, 31 Jul 2009 19:52:24 +0000 (19:52 +0000)]
- Convert the rest of the DOUTs to DEBUG+errs().
- One formatting change.

No intended functionality change.

llvm-svn: 77717

15 years agoPR4662: Fix a crash introduced by the recent LLVMContext changes.
Eli Friedman [Fri, 31 Jul 2009 19:36:47 +0000 (19:36 +0000)]
PR4662: Fix a crash introduced by the recent LLVMContext changes.

llvm-svn: 77716

15 years agoFix a typo in a comment.
Dan Gohman [Fri, 31 Jul 2009 19:26:54 +0000 (19:26 +0000)]
Fix a typo in a comment.

llvm-svn: 77715

15 years agoAdjust unit test for the MCSection changes.
Benjamin Kramer [Fri, 31 Jul 2009 19:12:33 +0000 (19:12 +0000)]
Adjust unit test for the MCSection changes.

llvm-svn: 77714

15 years agoFor a CXXOperatorCallExpr, fix the order that StmtLocResolver uses to check subexpres...
Argyrios Kyrtzidis [Fri, 31 Jul 2009 19:02:11 +0000 (19:02 +0000)]
For a CXXOperatorCallExpr, fix the order that StmtLocResolver uses to check subexpressions.

llvm-svn: 77713

15 years agoDelete spurious semicolons.
Dan Gohman [Fri, 31 Jul 2009 18:59:29 +0000 (18:59 +0000)]
Delete spurious semicolons.

llvm-svn: 77712

15 years agoFix build.
Benjamin Kramer [Fri, 31 Jul 2009 18:58:46 +0000 (18:58 +0000)]
Fix build.

llvm-svn: 77711

15 years agoUpdate CMake files.
Ted Kremenek [Fri, 31 Jul 2009 18:50:22 +0000 (18:50 +0000)]
Update CMake files.

llvm-svn: 77709

15 years agoswitch off of 'Section' onto MCSection. We're not properly using
Chris Lattner [Fri, 31 Jul 2009 18:48:30 +0000 (18:48 +0000)]
switch off of 'Section' onto MCSection.  We're not properly using
MCSection subclasses yet, but this is a step in the right direction.

llvm-svn: 77708

15 years agoMake the recanonicalization-for-an-out-of-line-definition test case a bit trickier
Douglas Gregor [Fri, 31 Jul 2009 18:36:42 +0000 (18:36 +0000)]
Make the recanonicalization-for-an-out-of-line-definition test case a bit trickier

llvm-svn: 77707

15 years agoRemove Annotation.h, which is no longer used in the LLVM tree.
Dan Gohman [Fri, 31 Jul 2009 18:36:25 +0000 (18:36 +0000)]
Remove Annotation.h, which is no longer used in the LLVM tree.

llvm-svn: 77706

15 years agoUntil we have a "ALIGN" pseudo instruction, have asm printer emitted a .align
Evan Cheng [Fri, 31 Jul 2009 18:35:56 +0000 (18:35 +0000)]
Until we have a "ALIGN" pseudo instruction, have asm printer emitted a .align
to ensure the instruction that follows a TBB (when the number of table entries
is odd) is 2-byte aligned.
Patch by Sandeep Patel.

llvm-svn: 77705

15 years agoMachineFunction no longer needs Annotation.
Dan Gohman [Fri, 31 Jul 2009 18:35:51 +0000 (18:35 +0000)]
MachineFunction no longer needs Annotation.

llvm-svn: 77704

15 years agoAllow one to update the type representation of the declaration specifiers. Will be...
Douglas Gregor [Fri, 31 Jul 2009 18:34:30 +0000 (18:34 +0000)]
Allow one to update the type representation of the declaration specifiers. Will be used by an upcoming commit.

llvm-svn: 77703