llvm-svn: 47598

llvm-svn: 47597

llvm-svn: 47596

llvm-svn: 47595

GOT-style position independent code. Before only tail calls to
protected/hidden functions within the same module were optimized.
Now all function calls are tail call optimized.

llvm-svn: 47594

calls. Before arguments that could overwrite each other were
explicitly lowered to a stack slot, not giving the register allocator
a chance to optimize. Now a sequence of copyto/copyfrom virtual
registers ensures that arguments are loaded in (virtual) registers
before they are lowered to the stack slot (and might overwrite each
other). Also parameter stack slots are marked mutable for
(potentially) tail calling functions.

llvm-svn: 47593

vr1 = extract_subreg vr2, 3
...
vr3 = extract_subreg vr1, 2
The end result is vr3 is equal to vr2 with subidx 2.

llvm-svn: 47592

after legalize.  Just because a constant is legal (e.g. 0.0 in SSE) 
doesn't mean that its negated value is legal (-0.0).  We could make
this stronger by checking to see if the negated constant is actually
legal post negation, but it doesn't seem like a big deal.

llvm-svn: 47591

llvm-svn: 47590

operands into inline asm block.

llvm-svn: 47589

llvm-svn: 47588

llvm-svn: 47587

Added recording of divide-by-zero and divide-by-uninitialized nodes.

llvm-svn: 47586

llvm-svn: 47585

llvm-svn: 47584

Take 2.

llvm-svn: 47583

Patch by Mike Stump!

llvm-svn: 47582

sign of its type in the early exit case.  No testcase, because this
doesn't manifest as a failure.

llvm-svn: 47581

llvm-svn: 47580

This caused the whole body to get dropped from the CFG.

llvm-svn: 47579

llvm-svn: 47578

llvm-svn: 47577

llvm-svn: 47576

pointed out that this isn't correct at -O0.

llvm-svn: 47575

llvm-svn: 47574

llvm-svn: 47573

that aren't VarDecls.

llvm-svn: 47572

and StoreSDNode into LSBaseSDNode.

llvm-svn: 47570

{S,U}MUL_LOHI with an unused high value.

llvm-svn: 47569

failures on OS X in some cases.  (Thank you valgrind.)

llvm-svn: 47568

result into a MUL late in the X86 codegen process. ISD::MUL is
once again Legal on X86, so this is no longer needed. And, the
hack was suboptimal; see PR1874 for details.

llvm-svn: 47567

llvm-svn: 47566

bits VC++ right now.

llvm-svn: 47565

llvm-svn: 47564

llvm-svn: 47563

llvm-svn: 47562

a SignBitIsZero function to simplify a common use case.

llvm-svn: 47561

llvm-svn: 47560

llvm-svn: 47558

All remat'ed loads cannot be folded into two-address code. Not just argument loads. This change doesn't really have any impact on codegen.

llvm-svn: 47557