This implements Transforms/ScalarRepl/union-packed.ll

llvm-svn: 32422

Regression/Transforms/IndVarsSimplify/2006-12-10-BitCast.ll

llvm-svn: 32420

 * Add support for promoting unions with fp values in them.  This produces
   our new int<->fp bitcast instructions, implementing
   Transforms/ScalarRepl/union-fp-int.ll

As an example, this allows us to compile this:

union intfloat { int i; float f; };
float invsqrt(const float arg_x) {
    union intfloat x = { .f = arg_x };
    const float xhalf = arg_x * 0.5f;
    x.i = 0x5f3759df - (x.i >> 1);
    return x.f * (1.5f - xhalf * x.f * x.f);
}

into:

_invsqrt:
        movss 4(%esp), %xmm0
        movd %xmm0, %eax
        sarl %eax
        movl $1597463007, %ecx
        subl %eax, %ecx
        movd %ecx, %xmm1
        mulss LCPI1_0, %xmm0
        mulss %xmm1, %xmm0
        movss LCPI1_1, %xmm2
        mulss %xmm1, %xmm0
        subss %xmm0, %xmm2
        movl 8(%esp), %eax
        mulss %xmm2, %xmm1
        movss %xmm1, (%eax)
        ret

instead of:

_invsqrt:
        subl $4, %esp
        movss 8(%esp), %xmm0
        movss %xmm0, (%esp)
        movl (%esp), %eax
        movl $1597463007, %ecx
        sarl %eax
        subl %eax, %ecx
        movl %ecx, (%esp)
        mulss LCPI1_0, %xmm0
        movss (%esp), %xmm1
        mulss %xmm1, %xmm0
        mulss %xmm1, %xmm0
        movss LCPI1_1, %xmm2
        subss %xmm0, %xmm2
        mulss %xmm2, %xmm1
        movl 12(%esp), %eax
        movss %xmm1, (%eax)
        addl $4, %esp
        ret

llvm-svn: 32418

code (both asm & cbe) for Mingw32 target.
Removed autoconf checks for underscored versions of setjmp/longjmp.

llvm-svn: 32415

llvm-svn: 32413

llvm-svn: 32412

static dtors.

llvm-svn: 32411

.comm is in bytes
:-(

llvm-svn: 32408

if a variable has no initialization, I->getInitializer() will fail

llvm-svn: 32407

llvm-svn: 32394

llvm-svn: 32393

llvm-svn: 32389

removed now.

llvm-svn: 32388

llvm-svn: 32387

llvm-svn: 32386

pass manager queuer.

llvm-svn: 32385

AnalysisResolver.

llvm-svn: 32383

top level manager. Use this info to findAnalysisPass.

llvm-svn: 32382

llvm-svn: 32381

Use one provided by toplevel manager.

llvm-svn: 32380

llvm-svn: 32379

llvm-svn: 32378

PMDataManager::findAnalysisPass()

llvm-svn: 32377

llvm-svn: 32376

MarkAliveBlocks.

llvm-svn: 32375

llvm-svn: 32373

Instead, the stat info is printed when llvm_shutdown() is called.
These also don't need static ctors, but getting rid of them is uglier:
still investigating.  This reduces the number of static dtors in llvm from
~1400 to ~750.

llvm-svn: 32372

llvm-svn: 32371

llvm-svn: 32370

llvm-svn: 32368

llvm-svn: 32366

llvm-svn: 32364

llvm-svn: 32362

llvm-svn: 32361

tied to another oeprand, 2) whether is is being tied to by another operand. So
the destination operand of a two-address MI can be correctly identified.

llvm-svn: 32354

llvm-svn: 32353

llvm-svn: 32352

llvm-svn: 32351

which allows the code to be above the 2G marker.  We still need to JIT emit
dyld stubs to support external, weak, common, etc globals, but that will
happen tomorrow.

llvm-svn: 32348

llvm-svn: 32347