bug where spill instructions were added to the next basic block
instead of the end of the current one if the instruction that required
the spill was the last in the block.

llvm-svn: 10272

llvm-svn: 10270

llvm-svn: 10268

If this doesn't work Misha, feel free to revert it.

llvm-svn: 10267

allow unaligned loads, that is probably the problem I've been seeing in numerous
SPARC test cases failing. X86, on the other hand, just slows down unaligned
accesses, since it must make 2 aligned accesses for each unaligned one.

llvm-svn: 10266

Some gep generalization changes

llvm-svn: 10252

llvm-svn: 10236

llvm-svn: 10232

llvm-svn: 10231

llvm-svn: 10229

llvm-svn: 10228

llvm-svn: 10226

llvm-svn: 10225

llvm-svn: 10224

llvm-svn: 10221

llvm-svn: 10219

llvm-svn: 10218

llvm-svn: 10217

llvm-svn: 10210

llvm-svn: 10204

llvm-svn: 10197

llvm-svn: 10193

the other way around, instead of failing a large, tumor-like assertion.

llvm-svn: 10171

llvm-svn: 10170

llvm-svn: 10168

llvm-svn: 10167

llvm-svn: 10164

llvm-svn: 10163

llvm-svn: 10161

The module stripping pass should not strip symbols on external globals

llvm-svn: 10157

* Restore registers *after* everything else to avoid any possible side effects

This fixes McCat-imp.

llvm-svn: 10147

llvm-svn: 10146

would have caught this.  *sigh*

llvm-svn: 10142

llvm-svn: 10138

llvm-svn: 10137

* Make the cost metric for passing constants in as arguments to functions MUCH
  more accurate, by actually estimating the amount of code that will be constant
  propagated away.

llvm-svn: 10136

Print out the costs for functions that AREN'T inlined as well

llvm-svn: 10135

llvm-svn: 10133

llvm-svn: 10132

llvm-svn: 10131