This was because we were deleting large chunks of functions without an exit block, because the post-dominance
information was not useful.  This broke crafty and twolf.

llvm-svn: 6698

llvm-svn: 6669

llvm-svn: 6667

llvm-svn: 6630

llvm-svn: 6622

llvm-svn: 6620

llvm-svn: 6613

llvm-svn: 6606

Add new combination to turn seteq X, 0 -> not(cast X to bool)

llvm-svn: 6604

llvm-svn: 6599

llvm-svn: 6506

llvm-svn: 6502

llvm-svn: 6486

Count resolutions correctly.

llvm-svn: 6482

llvm-svn: 6480

llvm-svn: 6479

llvm-svn: 6476

broken up into their elements.  Too many programs break because of this.

llvm-svn: 6440

behavior is technically undefined

llvm-svn: 6438

internal phi nodes, and returns a new vector of basic blocks.

llvm-svn: 6431

llvm-svn: 6428

llvm-svn: 6425

llvm-svn: 6420

llvm-svn: 6415

llvm-svn: 6397

* Make the function inliner _significantly_ smarter.  :)

llvm-svn: 6396

llvm-svn: 6352

* Implement SRoA for arrays

llvm-svn: 6349

llvm-svn: 6346

llvm-svn: 6338

llvm-svn: 6306

llvm-svn: 6295

  function and no symbols were explicitly marked to be externalized.
* Add new -internalize-public-api-list option that can be used if the symbol
  list is small, and making a new file is annoying.

llvm-svn: 6289

should not be internalized

llvm-svn: 6288

This hunk:
-    } else if (Src->getNumOperands() == 2 && Src->use_size() == 1) {
+    } else if (Src->getNumOperands() == 2) {

Allows GEP folding to be more aggressive, which reduces the number of instructions
and can dramatically speed up BasicAA in some cases.

llvm-svn: 6286

llvm-svn: 6273

llvm-svn: 6251

llvm-svn: 6248

llvm-svn: 6232

llvm-svn: 6225