llvm-svn: 29153

This implements PR828 and CodeGen/X86/2006-07-12-InlineAsmQConstraint.ll

llvm-svn: 29118

llvm-svn: 29110

CodeGen/X86/2006-07-10-InlineAsmAConstraint.ll

llvm-svn: 29101

llvm-svn: 29091

Use that information when doing the transformation to merge multiple loads
into a 128-bit load.

llvm-svn: 29090

llvm-svn: 29085

(load x+8), (load x+12), <0, 1, 2, 3> to a single 128-bit load (aligned and
unaligned).

e.g.

__m128 test(float a, float b, float c, float d) {
  return _mm_set_ps(d, c, b, a);
}

_test:
        movups 4(%esp), %xmm0
        ret

llvm-svn: 29042

llvm-svn: 28999

llvm-svn: 28992

llvm-svn: 28984

Let X86CompilationCallback pass previous frame and return address to X86CompilationCallback2. Remove alloca hack.

llvm-svn: 28982

llvm-svn: 28980

llvm-svn: 28979

llvm-svn: 28978

llvm-svn: 28976

llvm-svn: 28963

llvm-svn: 28958

llvm-svn: 28957

llvm-svn: 28938

Simplify X86CompilationCallback: always align to 16-byte boundary; don't save EAX/EDX if unnecessary.

llvm-svn: 28910

code in sections other than ".text", including weak sections like ctors and
dtors.

llvm-svn: 28909

llvm-svn: 28907

llvm-svn: 28906

llvm-svn: 28864

llvm-svn: 28862

llvm-svn: 28860

llvm-svn: 28848

llvm-svn: 28843

Remove explicit casts to std::string now that there is no overload resolution
issues in the TargetData ctors.

llvm-svn: 28830

llvm-svn: 28808

llvm-svn: 28796

vector types are passed via XMM registers.

llvm-svn: 28789

llvm-svn: 28787

llvm-svn: 28783

functions.

llvm-svn: 28781

llvm-svn: 28764

llvm-svn: 28737

llvm-svn: 28736

llvm-svn: 28728