When splitting a VAARG, remember its alignment.

This produces terrible but correct code.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@106952 91177308-0d34-0410-b5e6-96231b3b80d8

6 files changed, 51 insertions, 11 deletions

diff --git a/include/llvm/CodeGen/ISDOpcodes.h b/include/llvm/CodeGen/ISDOpcodes.h
index 6203b5e4ad..69de5986dd 100644
--- a/include/llvm/CodeGen/ISDOpcodes.h
+++ b/include/llvm/CodeGen/ISDOpcodes.h
@@ -508,8 +508,9 @@ namespace ISD {
     CALLSEQ_START,  // Beginning of a call sequence
     CALLSEQ_END,    // End of a call sequence
 
-    // VAARG - VAARG has three operands: an input chain, a pointer, and a
-    // SRCVALUE.  It returns a pair of values: the vaarg value and a new chain.
+    // VAARG - VAARG has four operands: an input chain, a pointer, a SRCVALUE,
+    // and the alignment. It returns a pair of values: the vaarg value and a
+    // new chain.
     VAARG,
 
     // VACOPY - VACOPY has five operands: an input chain, a destination pointer,
diff --git a/include/llvm/CodeGen/SelectionDAG.h b/include/llvm/CodeGen/SelectionDAG.h
index 1cd5191487..d13e20e095 100644
--- a/include/llvm/CodeGen/SelectionDAG.h
+++ b/include/llvm/CodeGen/SelectionDAG.h
@@ -582,7 +582,7 @@ public:
   /// getVAArg - VAArg produces a result and token chain, and takes a pointer
   /// and a source value as input.
   SDValue getVAArg(EVT VT, DebugLoc dl, SDValue Chain, SDValue Ptr,
-                   SDValue SV);
+                   SDValue SV, unsigned Align = 0);
 
   /// getAtomic - Gets a node for an atomic op, produces result and chain and
   /// takes 3 operands
diff --git a/lib/CodeGen/SelectionDAG/LegalizeDAG.cpp b/lib/CodeGen/SelectionDAG/LegalizeDAG.cpp
index f360a01737..8fb4654853 100644
--- a/lib/CodeGen/SelectionDAG/LegalizeDAG.cpp
+++ b/lib/CodeGen/SelectionDAG/LegalizeDAG.cpp
@@ -2642,15 +2642,29 @@ void SelectionDAGLegalize::ExpandNode(SDNode *Node,
     EVT VT = Node->getValueType(0);
     Tmp1 = Node->getOperand(0);
     Tmp2 = Node->getOperand(1);
-    SDValue VAList = DAG.getLoad(TLI.getPointerTy(), dl, Tmp1, Tmp2, V, 0,
-                                 false, false, 0);
+    unsigned Align = Node->getConstantOperandVal(3);
+
+    SDValue VAListLoad = DAG.getLoad(TLI.getPointerTy(), dl, Tmp1, Tmp2, V, 0,
+                                     false, false, 0);
+    SDValue VAList = VAListLoad;
+
+    if (Align != 0 ) {
+      VAList = DAG.getNode(ISD::ADD, dl, TLI.getPointerTy(), VAList,
+                           DAG.getConstant(Align - 1,
+                                           TLI.getPointerTy()));
+
+      VAList = DAG.getNode(ISD::AND, dl, TLI.getPointerTy(), VAList,
+                           DAG.getConstant(-Align,
+                                           TLI.getPointerTy()));
+    }
+
     // Increment the pointer, VAList, to the next vaarg
     Tmp3 = DAG.getNode(ISD::ADD, dl, TLI.getPointerTy(), VAList,
                        DAG.getConstant(TLI.getTargetData()->
                           getTypeAllocSize(VT.getTypeForEVT(*DAG.getContext())),
                                        TLI.getPointerTy()));
     // Store the incremented VAList to the legalized pointer
-    Tmp3 = DAG.getStore(VAList.getValue(1), dl, Tmp3, Tmp2, V, 0,
+    Tmp3 = DAG.getStore(VAListLoad.getValue(1), dl, Tmp3, Tmp2, V, 0,
                         false, false, 0);
     // Load the actual argument out of the pointer VAList
     Results.push_back(DAG.getLoad(VT, dl, Tmp3, VAList, NULL, 0,
diff --git a/lib/CodeGen/SelectionDAG/LegalizeTypesGeneric.cpp b/lib/CodeGen/SelectionDAG/LegalizeTypesGeneric.cpp
index 88e1e624ae..71adb33617 100644
--- a/lib/CodeGen/SelectionDAG/LegalizeTypesGeneric.cpp
+++ b/lib/CodeGen/SelectionDAG/LegalizeTypesGeneric.cpp
@@ -238,12 +238,17 @@ void DAGTypeLegalizer::ExpandRes_NormalLoad(SDNode *N, SDValue &Lo,
 }
 
 void DAGTypeLegalizer::ExpandRes_VAARG(SDNode *N, SDValue &Lo, SDValue &Hi) {
-  EVT NVT = TLI.getTypeToTransformTo(*DAG.getContext(), N->getValueType(0));
+  EVT OVT = N->getValueType(0);
+  EVT NVT = TLI.getTypeToTransformTo(*DAG.getContext(), OVT);
   SDValue Chain = N->getOperand(0);
   SDValue Ptr = N->getOperand(1);
   DebugLoc dl = N->getDebugLoc();
+  const unsigned OldAlign = N->getConstantOperandVal(3);
+  const Type *Type = OVT.getTypeForEVT(*DAG.getContext());
+  const unsigned TypeAlign = TLI.getTargetData()->getABITypeAlignment(Type);
+  const unsigned Align = std::max(OldAlign, TypeAlign);
 
-  Lo = DAG.getVAArg(NVT, dl, Chain, Ptr, N->getOperand(2));
+  Lo = DAG.getVAArg(NVT, dl, Chain, Ptr, N->getOperand(2), Align);
   Hi = DAG.getVAArg(NVT, dl, Lo.getValue(1), Ptr, N->getOperand(2));
 
   // Handle endianness of the load.
diff --git a/lib/CodeGen/SelectionDAG/SelectionDAG.cpp b/lib/CodeGen/SelectionDAG/SelectionDAG.cpp
index 5c35df70c5..0a504dbe9f 100644
--- a/lib/CodeGen/SelectionDAG/SelectionDAG.cpp
+++ b/lib/CodeGen/SelectionDAG/SelectionDAG.cpp
@@ -4141,9 +4141,10 @@ SelectionDAG::getIndexedStore(SDValue OrigStore, DebugLoc dl, SDValue Base,
 
 SDValue SelectionDAG::getVAArg(EVT VT, DebugLoc dl,
                                SDValue Chain, SDValue Ptr,
-                               SDValue SV) {
-  SDValue Ops[] = { Chain, Ptr, SV };
-  return getNode(ISD::VAARG, dl, getVTList(VT, MVT::Other), Ops, 3);
+                               SDValue SV,
+                               unsigned Align) {
+  SDValue Ops[] = { Chain, Ptr, SV, getTargetConstant(Align, MVT::i32) };
+  return getNode(ISD::VAARG, dl, getVTList(VT, MVT::Other), Ops, 4);
 }
 
 SDValue SelectionDAG::getNode(unsigned Opcode, DebugLoc DL, EVT VT,
diff --git a/test/CodeGen/ARM/va_arg.ll b/test/CodeGen/ARM/va_arg.ll
new file mode 100644
index 0000000000..a54a499cf1
--- /dev/null
+++ b/test/CodeGen/ARM/va_arg.ll
@@ -0,0 +1,19 @@
+; RUN: llc < %s -mtriple=armv7-none-linux-gnueabi | FileCheck %s
+; Test that we correctly align elements when using va_arg
+
+; CHECK: add	r0, r0, #7
+; CHECK: bfc	r0, #0, #3
+
+define i64 @f8(i32 %i, ...) nounwind optsize {
+entry:
+  %g = alloca i8*, align 4
+  %g1 = bitcast i8** %g to i8*
+  call void @llvm.va_start(i8* %g1)
+  %0 = va_arg i8** %g, i64
+  call void @llvm.va_end(i8* %g1)
+  ret i64 %0
+}
+
+declare void @llvm.va_start(i8*) nounwind
+
+declare void @llvm.va_end(i8*) nounwind