llvm-svn: 35226

Avoid recursion, Use iterative algorithm.

llvm-svn: 35225

llvm-svn: 35224

Get pos/neg infinity the correct way.

llvm-svn: 35223

LoopSimplify::FindPHIToPartitionLoops()
Use ETForest instead of DominatorSet.

llvm-svn: 35222

Use ETForest instead of DominatorSet.

llvm-svn: 35221

llvm-svn: 35220

llvm-svn: 35219

    mov lr, pc
    bx lr
So, the function was not called.

llvm-svn: 35218

llvm-svn: 35217

llvm-svn: 35216

llvm-svn: 35215

llvm-svn: 35214

llvm-svn: 35213

llvm-svn: 35212

llvm-svn: 35211

llvm-svn: 35210

llvm-svn: 35209

llvm-svn: 35208

Make two piece constant generation as a single instruction. It's re-materialized as a load from constantpool.

llvm-svn: 35207

llvm-svn: 35206

llvm-svn: 35205

1) codegen a shift of a register as a shift, not an LEA.
2) teach the RA to convert a shift to an LEA instruction if it wants something
   in three-address form.

This gives us asm diffs like:

-       leal (,%eax,4), %eax
+       shll $2, %eax

which is faster on some processors and smaller on all of them.

and, more interestingly:

-       movl 24(%esi), %eax
-       leal (,%eax,4), %edi
+       movl 24(%esi), %edi
+       shll $2, %edi

Without #2, #1 was a significant pessimization in some cases.

This implements CodeGen/X86/shift-codegen.ll

llvm-svn: 35204

llvm-svn: 35203

llvm-svn: 35202

llvm-svn: 35201

llvm-svn: 35200

llvm-svn: 35199

llvm-svn: 35198

llvm-svn: 35197

llvm-svn: 35196

llvm-svn: 35195

llvm-svn: 35194

llvm-svn: 35193

llvm-svn: 35192

llvm-svn: 35191

llvm-svn: 35190

llvm-svn: 35189

Patch by Sheng Zhou.

llvm-svn: 35188

Patch by Zhou Sheng.

llvm-svn: 35187