EXTRACT_SUBREG no longer appears as a machine instruction. Use COPY instead.

Add isCopy() checks in many places using isMoveInstr() and isExtractSubreg().
The isMoveInstr hook will be removed later.

llvm-svn: 107879

Fix X86InstrInfo::convertToThreeAddressWithLEA to generate COPY instead of
INSERT_SUBREG.

llvm-svn: 107878

ExtWarn to an Extension. Let the broken code propagate!

llvm-svn: 107875

surprised.

llvm-svn: 107874

For TagType and TemplateSpecializationType, isDependent calculation may be invalid because some decls that the
calculation is using may still be initializing.
Thus, store the isDependent flag to PCH and restore directly to Type.

llvm-svn: 107873

llvm-svn: 107872

llvm-svn: 107871

llvm-svn: 107870

causes some versions of gcc to crash when building LLVM.

llvm-svn: 107869

llvm-svn: 107868

compiled with MSVC 2010 (PR7367).  Instead use a SmallVector.

llvm-svn: 107867

(X >s -1) ? C1 : C2 and (X <s  0) ? C2 : C1
into ((X >>s 31) & (C2 - C1)) + C1, avoiding the conditional.

This optimization could be extended to take non-const C1 and C2 but we better
stay conservative to avoid code size bloat for now.

for
int sel(int n) {
     return n >= 0 ? 60 : 100;
}

we now generate
  sarl  $31, %edi
  andl  $40, %edi
  leal  60(%rdi), %eax

instead of
  testl %edi, %edi
  movl  $60, %ecx
  movl  $100, %eax
  cmovnsl %ecx, %eax

llvm-svn: 107866

list in a diagnostic group so it can be turned on and off. A terrifying amount
of code, including large chunks of open source code, still do this so it's
important to be able to suppress it when necessary.

Doug, is this a reasonable compromise? I'd lean toward making it a normal
extension, but I don't feel strongly as long as we can turn the warnings off.

llvm-svn: 107865

unknown attributes that we discard. Add a diagnostic group for unknown
attribute warnings to allow turning these off when we don't care. Also
consolidates the tests for this case.

llvm-svn: 107864

llvm-svn: 107863

llvm-svn: 107862

correct the testcase for valid assembly.

Needs more tests.

llvm-svn: 107860

as well.

llvm-svn: 107858

suppressing copies of objects with trivial copy constructors.

llvm-svn: 107857

llvm-svn: 107856

This target hook is intended to replace copyRegToReg entirely, but for now it
calls copyRegToReg.

Any remaining calls to copyRegToReg wil be replaced by COPY instructions.

llvm-svn: 107854

Also fixed our build to define NDEBUG; code that
uses LLVM headers without NDEBUG is
binary-incompatible with libraries built with
NDEBUG.

llvm-svn: 107853

Optimize some vfp comparisons to integer ones. This patch implements the simplest case when the following conditions are met:
1. The arguments are f32.
2. The arguments are loads and they have no uses other than the comparison.
3. The comparison code is EQ or NE.

e.g.
        vldr.32 s0, [r1]
        vldr.32 s1, [r0]
        vcmpe.f32       s1, s0
        vmrs    apsr_nzcv, fpscr
	beq     LBB0_2
=>
        ldr     r1, [r1]
        ldr     r0, [r0]
        cmp     r0, r1
        beq     LBB0_2

More complicated cases will be implemented in subsequent patches.

llvm-svn: 107852

Add explicit testcases for tail calls within the same module.
Duplicate some code to humor those who think .w doesn't apply on ARM.
Leave this disabled on Thumb1, and add some comments explaining why it's hard
and won't gain much.

llvm-svn: 107851

Debug info intrinsics win for now.

llvm-svn: 107850

(if there are any) and use the one which remains available for the longest
rather than just using the first one. This should help enable better re-use
of the loaded frame index values. rdar://7318760

llvm-svn: 107847

address spaces when SRoA'ing memcpy's.

llvm-svn: 107846

llvm-svn: 107845

llvm-svn: 107844

was not producing a memcpy with the right address
spaces because of two places in it doing casts of
the arguments to i8, one of which that didn't
preserve the address space.

There is also an optimizer bug here.

llvm-svn: 107842

llvm-svn: 107841

prefer to materialize as local constants. This fixes the clang
bootstrap abort.

llvm-svn: 107840

llvm-svn: 107839

llvm-svn: 107838

breaking bootstrap on Linux.

llvm-svn: 107837

llvm-svn: 107836

Rip out the C++0x-specific handling of destructor names. The specification is still in flux and unclear, and our interim workaround was broken. Fixes PR7467.

llvm-svn: 107835

PR7399.  The asm parser already handles this.  This is of dubious 
utility (see the PR) but the asmprinter was clearly broken here.

llvm-svn: 107834

typedefs won't have the same canonical declaration (since they are
distinct), so we need to check for this case specifically. Fixes
<rdar://problem/8018262>.

llvm-svn: 107833

llvm-svn: 107832