Fix ppcf128 component access on little-endian systems

The PowerPC 128-bit long double data type (ppcf128 in LLVM) is in fact a
pair of two doubles, where one is considered the "high" or
more-significant part, and the other is considered the "low" or
less-significant part.  When a ppcf128 value is stored in memory or a
register pair, the high part always comes first, i.e. at the lower
memory address or in the lower-numbered register, and the low part
always comes second.  This is true both on big-endian and little-endian
PowerPC systems.  (Similar to how with a complex number, the real part
always comes first and the imaginary part second, no matter the byte
order of the system.)

This was implemented incorrectly for little-endian systems in LLVM.
This commit fixes three related issues:

- When printing an immediate ppcf128 constant to assembler output
  in emitGlobalConstantFP, emit the high part first on both big-
  and little-endian systems.

- When lowering a ppcf128 type to a pair of f64 types in SelectionDAG
  (which is used e.g. when generating code to load an argument into a
  register pair), use correct low/high part ordering on little-endian
  systems.

- In a related issue, because lowering ppcf128 into a pair of f64 must
  operate differently from lowering an int128 into a pair of i64,
  bitcasts between ppcf128 and int128 must not be optimized away by the
  DAG combiner on little-endian systems, but must effect a word-swap.

Reviewed by Hal Finkel.

llvm-svn: 212274
diff --git a/llvm/lib/CodeGen/AsmPrinter/AsmPrinter.cpp b/llvm/lib/CodeGen/AsmPrinter/AsmPrinter.cpp
index 5a809aa..f80fdea 100644
--- a/llvm/lib/CodeGen/AsmPrinter/AsmPrinter.cpp
+++ b/llvm/lib/CodeGen/AsmPrinter/AsmPrinter.cpp
@@ -1881,7 +1881,8 @@
 
   // PPC's long double has odd notions of endianness compared to how LLVM
   // handles it: p[0] goes first for *big* endian on PPC.
-  if (AP.TM.getDataLayout()->isBigEndian() != CFP->getType()->isPPC_FP128Ty()) {
+  if (AP.TM.getDataLayout()->isBigEndian() &&
+      !CFP->getType()->isPPC_FP128Ty()) {
     int Chunk = API.getNumWords() - 1;
 
     if (TrailingBytes)
diff --git a/llvm/lib/CodeGen/SelectionDAG/DAGCombiner.cpp b/llvm/lib/CodeGen/SelectionDAG/DAGCombiner.cpp
index 0f50184..7198203 100644
--- a/llvm/lib/CodeGen/SelectionDAG/DAGCombiner.cpp
+++ b/llvm/lib/CodeGen/SelectionDAG/DAGCombiner.cpp
@@ -6210,6 +6210,9 @@
   if (ISD::isNormalLoad(N0.getNode()) && N0.hasOneUse() &&
       // Do not change the width of a volatile load.
       !cast<LoadSDNode>(N0)->isVolatile() &&
+      // Do not remove the cast if the types differ in endian layout.
+      TLI.hasBigEndianPartOrdering(N0.getValueType()) ==
+      TLI.hasBigEndianPartOrdering(VT) &&
       (!LegalOperations || TLI.isOperationLegal(ISD::LOAD, VT)) &&
       TLI.isLoadBitCastBeneficial(N0.getValueType(), VT)) {
     LoadSDNode *LN0 = cast<LoadSDNode>(N0);
diff --git a/llvm/lib/CodeGen/SelectionDAG/LegalizeTypesGeneric.cpp b/llvm/lib/CodeGen/SelectionDAG/LegalizeTypesGeneric.cpp
index f40ed76..7e2f7b6 100644
--- a/llvm/lib/CodeGen/SelectionDAG/LegalizeTypesGeneric.cpp
+++ b/llvm/lib/CodeGen/SelectionDAG/LegalizeTypesGeneric.cpp
@@ -60,12 +60,15 @@
     case TargetLowering::TypeExpandFloat:
       // Convert the expanded pieces of the input.
       GetExpandedOp(InOp, Lo, Hi);
+      if (TLI.hasBigEndianPartOrdering(InVT) !=
+          TLI.hasBigEndianPartOrdering(OutVT))
+        std::swap(Lo, Hi);
       Lo = DAG.getNode(ISD::BITCAST, dl, NOutVT, Lo);
       Hi = DAG.getNode(ISD::BITCAST, dl, NOutVT, Hi);
       return;
     case TargetLowering::TypeSplitVector:
       GetSplitVector(InOp, Lo, Hi);
-      if (TLI.isBigEndian())
+      if (TLI.hasBigEndianPartOrdering(OutVT))
         std::swap(Lo, Hi);
       Lo = DAG.getNode(ISD::BITCAST, dl, NOutVT, Lo);
       Hi = DAG.getNode(ISD::BITCAST, dl, NOutVT, Hi);
@@ -82,7 +85,7 @@
       EVT LoVT, HiVT;
       std::tie(LoVT, HiVT) = DAG.GetSplitDestVTs(InVT);
       std::tie(Lo, Hi) = DAG.SplitVector(InOp, dl, LoVT, HiVT);
-      if (TLI.isBigEndian())
+      if (TLI.hasBigEndianPartOrdering(OutVT))
         std::swap(Lo, Hi);
       Lo = DAG.getNode(ISD::BITCAST, dl, NOutVT, Lo);
       Hi = DAG.getNode(ISD::BITCAST, dl, NOutVT, Hi);
@@ -176,7 +179,7 @@
                    false, false, MinAlign(Alignment, IncrementSize));
 
   // Handle endianness of the load.
-  if (TLI.isBigEndian())
+  if (TLI.hasBigEndianPartOrdering(OutVT))
     std::swap(Lo, Hi);
 }
 
@@ -245,7 +248,8 @@
   SDLoc dl(N);
 
   LoadSDNode *LD = cast<LoadSDNode>(N);
-  EVT NVT = TLI.getTypeToTransformTo(*DAG.getContext(), LD->getValueType(0));
+  EVT ValueVT = LD->getValueType(0);
+  EVT NVT = TLI.getTypeToTransformTo(*DAG.getContext(), ValueVT);
   SDValue Chain = LD->getChain();
   SDValue Ptr = LD->getBasePtr();
   unsigned Alignment = LD->getAlignment();
@@ -275,7 +279,7 @@
                       Hi.getValue(1));
 
   // Handle endianness of the load.
-  if (TLI.isBigEndian())
+  if (TLI.hasBigEndianPartOrdering(ValueVT))
     std::swap(Lo, Hi);
 
   // Modified the chain - switch anything that used the old chain to use
@@ -295,7 +299,7 @@
   Hi = DAG.getVAArg(NVT, dl, Lo.getValue(1), Ptr, N->getOperand(2), 0);
 
   // Handle endianness of the load.
-  if (TLI.isBigEndian())
+  if (TLI.hasBigEndianPartOrdering(OVT))
     std::swap(Lo, Hi);
 
   // Modified the chain - switch anything that used the old chain to use
@@ -459,8 +463,8 @@
   SDLoc dl(N);
 
   StoreSDNode *St = cast<StoreSDNode>(N);
-  EVT NVT = TLI.getTypeToTransformTo(*DAG.getContext(),
-                                     St->getValue().getValueType());
+  EVT ValueVT = St->getValue().getValueType();
+  EVT NVT = TLI.getTypeToTransformTo(*DAG.getContext(), ValueVT);
   SDValue Chain = St->getChain();
   SDValue Ptr = St->getBasePtr();
   unsigned Alignment = St->getAlignment();
@@ -474,7 +478,7 @@
   SDValue Lo, Hi;
   GetExpandedOp(St->getValue(), Lo, Hi);
 
-  if (TLI.isBigEndian())
+  if (TLI.hasBigEndianPartOrdering(ValueVT))
     std::swap(Lo, Hi);
 
   Lo = DAG.getStore(Chain, dl, Lo, Ptr, St->getPointerInfo(),
diff --git a/llvm/lib/CodeGen/SelectionDAG/SelectionDAGBuilder.cpp b/llvm/lib/CodeGen/SelectionDAG/SelectionDAGBuilder.cpp
index 249683f..28d8e98 100644
--- a/llvm/lib/CodeGen/SelectionDAG/SelectionDAGBuilder.cpp
+++ b/llvm/lib/CodeGen/SelectionDAG/SelectionDAGBuilder.cpp
@@ -169,7 +169,7 @@
       SDValue Lo, Hi;
       Lo = DAG.getNode(ISD::BITCAST, DL, EVT(MVT::f64), Parts[0]);
       Hi = DAG.getNode(ISD::BITCAST, DL, EVT(MVT::f64), Parts[1]);
-      if (TLI.isBigEndian())
+      if (TLI.hasBigEndianPartOrdering(ValueVT))
         std::swap(Lo, Hi);
       Val = DAG.getNode(ISD::BUILD_PAIR, DL, ValueVT, Lo, Hi);
     } else {