Merging from mainline

llvm-svn: 31623

2 files changed, 83 insertions, 7 deletions

diff --git a/llvm/lib/Target/X86/X86ISelLowering.cpp b/llvm/lib/Target/X86/X86ISelLowering.cpp
index ac7327f9e953..d5b7c1c90e6a 100644
--- a/llvm/lib/Target/X86/X86ISelLowering.cpp
+++ b/llvm/lib/Target/X86/X86ISelLowering.cpp
@@ -516,10 +516,9 @@ SDOperand X86TargetLowering::LowerCCCArguments(SDOperand Op, SelectionDAG &DAG)
   BytesToPopOnReturn = 0;         // Callee pops nothing.
   BytesCallerReserves = ArgOffset;
 
-  // If this is a struct return on Darwin/X86, the callee pops the hidden struct
-  // pointer.
-  if (MF.getFunction()->getCallingConv() == CallingConv::CSRet &&
-      Subtarget->isTargetDarwin())
+  // If this is a struct return on, the callee pops the hidden struct
+  // pointer. This is common for Darwin/X86, Linux & Mingw32 targets.
+  if (MF.getFunction()->getCallingConv() == CallingConv::CSRet)
     BytesToPopOnReturn = 4;
 
   // Return the new list of results.
@@ -680,9 +679,10 @@ SDOperand X86TargetLowering::LowerCCCCallTo(SDOperand Op, SelectionDAG &DAG) {
   // Create the CALLSEQ_END node.
   unsigned NumBytesForCalleeToPush = 0;
 
-  // If this is is a call to a struct-return function on Darwin/X86, the callee
+  // If this is is a call to a struct-return function, the callee
   // pops the hidden struct pointer, so we have to push it back.
-  if (CallingConv == CallingConv::CSRet && Subtarget->isTargetDarwin())
+  // This is common for Darwin/X86, Linux & Mingw32 targets.
+  if (CallingConv == CallingConv::CSRet)
     NumBytesForCalleeToPush = 4;
   
   NodeTys.clear();
@@ -2582,6 +2582,22 @@ bool X86::isMOVHLPSMask(SDNode *N) {
          isUndefOrEqual(N->getOperand(3), 3);
 }
 
+/// isMOVHLPS_v_undef_Mask - Special case of isMOVHLPSMask for canonical form
+/// of vector_shuffle v, v, <2, 3, 2, 3>, i.e. vector_shuffle v, undef,
+/// <2, 3, 2, 3>
+bool X86::isMOVHLPS_v_undef_Mask(SDNode *N) {
+  assert(N->getOpcode() == ISD::BUILD_VECTOR);
+
+  if (N->getNumOperands() != 4)
+    return false;
+
+  // Expect bit0 == 2, bit1 == 3, bit2 == 2, bit3 == 3
+  return isUndefOrEqual(N->getOperand(0), 2) &&
+         isUndefOrEqual(N->getOperand(1), 3) &&
+         isUndefOrEqual(N->getOperand(2), 2) &&
+         isUndefOrEqual(N->getOperand(3), 3);
+}
+
 /// isMOVLPMask - Return true if the specified VECTOR_SHUFFLE operand
 /// specifies a shuffle of elements that is suitable for input to MOVLP{S|D}.
 bool X86::isMOVLPMask(SDNode *N) {
@@ -3724,7 +3740,7 @@ X86TargetLowering::LowerEXTRACT_VECTOR_ELT(SDOperand Op, SelectionDAG &DAG) {
     SDOperand Mask = DAG.getNode(ISD::BUILD_VECTOR, MaskVT,
                                  &IdxVec[0], IdxVec.size());
     Vec = DAG.getNode(ISD::VECTOR_SHUFFLE, Vec.getValueType(),
-                      Vec, Vec, Mask);
+                      Vec, DAG.getNode(ISD::UNDEF, Vec.getValueType()), Mask);
     return DAG.getNode(ISD::EXTRACT_VECTOR_ELT, VT, Vec,
                        DAG.getConstant(0, getPointerTy()));
   } else if (MVT::getSizeInBits(VT) == 64) {
diff --git a/llvm/test/Regression/CodeGen/X86/2006-11-12-CSRetCC.ll b/llvm/test/Regression/CodeGen/X86/2006-11-12-CSRetCC.ll
new file mode 100644
index 000000000000..03be0aca702e
--- /dev/null
+++ b/llvm/test/Regression/CodeGen/X86/2006-11-12-CSRetCC.ll
@@ -0,0 +1,60 @@
+; RUN: llvm-as < %s | llc -march=x86 | grep 'subl $4, %esp'
+
+%str = internal constant [9 x sbyte] c"%f+%f*i\0A\00"		; <[9 x sbyte]*> [#uses=1]
+
+implementation   ; Functions:
+
+int %main() {
+entry:
+	%retval = alloca int, align 4		; <int*> [#uses=1]
+	%tmp = alloca { double, double }, align 16		; <{ double, double }*> [#uses=4]
+	%tmp1 = alloca { double, double }, align 16		; <{ double, double }*> [#uses=4]
+	%tmp2 = alloca { double, double }, align 16		; <{ double, double }*> [#uses=3]
+	%pi = alloca double, align 8		; <double*> [#uses=2]
+	%z = alloca { double, double }, align 16		; <{ double, double }*> [#uses=4]
+	"alloca point" = cast int 0 to int		; <int> [#uses=0]
+	store double 0x400921FB54442D18, double* %pi
+	%tmp = load double* %pi		; <double> [#uses=1]
+	%real = getelementptr { double, double }* %tmp1, uint 0, uint 0		; <double*> [#uses=1]
+	store double 0.000000e+00, double* %real
+	%real3 = getelementptr { double, double }* %tmp1, uint 0, uint 1		; <double*> [#uses=1]
+	store double %tmp, double* %real3
+	%tmp = getelementptr { double, double }* %tmp, uint 0, uint 0		; <double*> [#uses=1]
+	%tmp4 = getelementptr { double, double }* %tmp1, uint 0, uint 0		; <double*> [#uses=1]
+	%tmp5 = load double* %tmp4		; <double> [#uses=1]
+	store double %tmp5, double* %tmp
+	%tmp6 = getelementptr { double, double }* %tmp, uint 0, uint 1		; <double*> [#uses=1]
+	%tmp7 = getelementptr { double, double }* %tmp1, uint 0, uint 1		; <double*> [#uses=1]
+	%tmp8 = load double* %tmp7		; <double> [#uses=1]
+	store double %tmp8, double* %tmp6
+	%tmp = cast { double, double }* %tmp to { long, long }*		; <{ long, long }*> [#uses=1]
+	%tmp = getelementptr { long, long }* %tmp, uint 0, uint 0		; <long*> [#uses=1]
+	%tmp = load long* %tmp		; <long> [#uses=1]
+	%tmp9 = cast { double, double }* %tmp to { long, long }*		; <{ long, long }*> [#uses=1]
+	%tmp10 = getelementptr { long, long }* %tmp9, uint 0, uint 1		; <long*> [#uses=1]
+	%tmp11 = load long* %tmp10		; <long> [#uses=1]
+	call csretcc void %cexp( { double, double }* %tmp2, long %tmp, long %tmp11 )
+	%tmp12 = getelementptr { double, double }* %z, uint 0, uint 0		; <double*> [#uses=1]
+	%tmp13 = getelementptr { double, double }* %tmp2, uint 0, uint 0		; <double*> [#uses=1]
+	%tmp14 = load double* %tmp13		; <double> [#uses=1]
+	store double %tmp14, double* %tmp12
+	%tmp15 = getelementptr { double, double }* %z, uint 0, uint 1		; <double*> [#uses=1]
+	%tmp16 = getelementptr { double, double }* %tmp2, uint 0, uint 1		; <double*> [#uses=1]
+	%tmp17 = load double* %tmp16		; <double> [#uses=1]
+	store double %tmp17, double* %tmp15
+	%tmp18 = getelementptr { double, double }* %z, uint 0, uint 1		; <double*> [#uses=1]
+	%tmp19 = load double* %tmp18		; <double> [#uses=1]
+	%tmp20 = getelementptr { double, double }* %z, uint 0, uint 0		; <double*> [#uses=1]
+	%tmp21 = load double* %tmp20		; <double> [#uses=1]
+	%tmp = getelementptr [9 x sbyte]* %str, int 0, uint 0		; <sbyte*> [#uses=1]
+	%tmp = call int (sbyte*, ...)* %printf( sbyte* %tmp, double %tmp21, double %tmp19 )		; <int> [#uses=0]
+	br label %return
+
+return:		; preds = %entry
+	%retval = load int* %retval		; <int> [#uses=1]
+	ret int %retval
+}
+
+declare csretcc void %cexp({ double, double }*, long, long)
+
+declare int %printf(sbyte*, ...)