A = shl x, 42
...
B = lshr ..., 38

which can be transformed into:
A = shl x, 4
...

iff we can prove that the would-be-shifted-in bits
are already zero.  This eliminates two shifts in the testcase
and allows eliminate of the whole i128 chain in the real example.

llvm-svn: 112314

Added checking of (x == x) and (x != x) to IdempotentOperationChecker and updated test cases flagged by it.

llvm-svn: 112313

Enabled relaxed LiveVariables analysis in the path-sensitive engine to increase the coverage of bugs. Primarily affects IdempotentOperationChecker.
- Migrated a temporarily separated test back to its original file (bug has been fixed, null-deref-ps-temp.c -> null-deref-ps.c)
- Changed SymbolManager to use relaxed LiveVariables
- Updated several test cases that the IdempotentOperationChecker class now flags
- Added test case to test relaxed LiveVariables use by the IdempotentOperationChecker

llvm-svn: 112312

llvm-svn: 112311

llvm-svn: 112310

llvm-svn: 112309

o Fixed a crasher when getting it via SBTarget.GetExecutable().

>>> filespec = target.GetExecutable()
Segmentation fault

o And renamed SBFileSpec::GetFileName() to GetFilename() to be consistent with FileSpec::GetFilename().

llvm-svn: 112308

llvm-svn: 112307

Add alternate version of LiveVariables analysis that does not kill liveness at assignments. This 'relaxed' liveness is useful in path sensitive analysis for situations where the resulting extended liveness allows us to find some bugs.
- Added killAtAssign flag to LiveVariables
- Added relaxed LiveVariables to AnalysisContext with an accessor

llvm-svn: 112306

llvm-svn: 112305

framework, which is good at ripping through bitfield
operations.  This generalize a bunch of the existing
xforms that instcombine does, such as 
  (x << c) >> c -> and
to handle intermediate logical nodes.  This is useful for
ripping up the "promote to large integer" code produced by
SRoA.

llvm-svn: 112304

llvm-svn: 112303

llvm-svn: 112302

swaps on the variable list, value object list, and disassembly. This avoids
us having to try and update frame indexes and other things that were getting
out of sync.

llvm-svn: 112301

llvm-svn: 112300

adjacent operands at once, instead of just two at a time.

llvm-svn: 112299

Improve wording of diagnostic complaining about a non-void* pointer as the first parameter of operator delete

llvm-svn: 112298

Patch by Cameron Esfahani!

llvm-svn: 112297

Implement CXCursor support for walking C++ base specifiers.  This includes adding the API hooks clang_isVirtualBase() and clang_getCXXAccessSpecifier() to query properties of the base specifier.

Implements <rdar://problem/8274883>.

llvm-svn: 112296

llvm-svn: 112295

member function you're typing in overrides another virtual function,
this fills in a (qualified!) call to that virtual function to make
such delegation easy.

llvm-svn: 112294

llvm-svn: 112293

llvm-svn: 112292

more general simplify demanded bits logic.

llvm-svn: 112291

transformation collect all the addrecs with the same loop
add combine them at once rather than starting everything over
at the first chance.

llvm-svn: 112290

llvm-svn: 112289

llvm-svn: 112288

llvm-svn: 112287

llvm-svn: 112286

computation can be truncated if it is fed by a sext/zext that doesn't
have to be exactly equal to the truncation result type.

llvm-svn: 112285

llvm-svn: 112284

llvm-svn: 112283

we don't see a DeclStmt (failure to instantiate which generally causes
panic).

llvm-svn: 112282

to DenseMap.

llvm-svn: 112281

llvm-svn: 112280

by the SRoA "promote to large integer" code, eliminating
some type conversions like this:

   %94 = zext i16 %93 to i32                       ; <i32> [#uses=2]
   %96 = lshr i32 %94, 8                           ; <i32> [#uses=1]
   %101 = trunc i32 %96 to i8                      ; <i8> [#uses=1]

This also unblocks other xforms from happening, now clang is able to compile:

struct S { float A, B, C, D; };
float foo(struct S A) { return A.A + A.B+A.C+A.D; }

into:

_foo:                                   ## @foo
## BB#0:                                ## %entry
	pshufd	$1, %xmm0, %xmm2
	addss	%xmm0, %xmm2
	movdqa	%xmm1, %xmm3
	addss	%xmm2, %xmm3
	pshufd	$1, %xmm1, %xmm0
	addss	%xmm3, %xmm0
	ret

on x86-64, instead of:

_foo:                                   ## @foo
## BB#0:                                ## %entry
	movd	%xmm0, %rax
	shrq	$32, %rax
	movd	%eax, %xmm2
	addss	%xmm0, %xmm2
	movapd	%xmm1, %xmm3
	addss	%xmm2, %xmm3
	movd	%xmm1, %rax
	shrq	$32, %rax
	movd	%eax, %xmm0
	addss	%xmm3, %xmm0
	ret

This seems pretty close to optimal to me, at least without
using horizontal adds.  This also triggers in lots of other
code, including SPEC.

llvm-svn: 112278

instead of trying to maintain the real frame list (unwind frames) and an
inline frame list. The information is cheap to produce when we already have
looked up a block and was making stack frame uniquing difficult when trying
to use the previous stack when making the current stack.

We now maintain the previous value object lists for common frames between
a previous and current frames so we will be able to tell when variable values
change.

llvm-svn: 112277

and process.  Added comment within the file about issues of using LaunchProcess
of SBTarget to launch a process (rdar://problem/8364687).

llvm-svn: 112276

Patch originally by Alexander Herz.

llvm-svn: 112275

declarator, the very definition of "low-hanging fruit".

llvm-svn: 112274