[PGO] Convert InstrProfRecord based serialization methods to use common C methods

[oota-llvm.git] / lib / Target / AArch64 / AArch64ISelLowering.cpp

diff --git a/lib/Target/AArch64/AArch64ISelLowering.cpp b/lib/Target/AArch64/AArch64ISelLowering.cpp
index 428624cd894017253721af345a0dbb706109510a..26f03ef765db3104ccbb2484af89cfcf103f6d2b 100644 (file)
--- a/lib/Target/AArch64/AArch64ISelLowering.cpp
+++ b/lib/Target/AArch64/AArch64ISelLowering.cpp
@@ -196,11 +196,6 @@ AArch64TargetLowering::AArch64TargetLowering(const TargetMachine &TM,
   setOperationAction(ISD::STACKRESTORE, MVT::Other, Expand);
   setOperationAction(ISD::DYNAMIC_STACKALLOC, MVT::i64, Expand);
 
-  // Exception handling.
-  // FIXME: These are guesses. Has this been defined yet?
-  setExceptionPointerRegister(AArch64::X0);
-  setExceptionSelectorRegister(AArch64::X1);
-
   // Constant pool entries
   setOperationAction(ISD::ConstantPool, MVT::i64, Custom);
 
@@ -494,6 +489,8 @@ AArch64TargetLowering::AArch64TargetLowering(const TargetMachine &TM,
   setTargetDAGCombine(ISD::BITCAST);
   setTargetDAGCombine(ISD::CONCAT_VECTORS);
   setTargetDAGCombine(ISD::STORE);
+  if (Subtarget->supportsAddressTopByteIgnored())
+    setTargetDAGCombine(ISD::LOAD);
 
   setTargetDAGCombine(ISD::MUL);
 
@@ -3890,7 +3887,13 @@ SDValue AArch64TargetLowering::LowerSELECT_CC(ISD::CondCode CC, SDValue LHS,
     }
   }
 
-  // Handle integers first.
+  // Also handle f16, for which we need to do a f32 comparison.
+  if (LHS.getValueType() == MVT::f16) {
+    LHS = DAG.getNode(ISD::FP_EXTEND, dl, MVT::f32, LHS);
+    RHS = DAG.getNode(ISD::FP_EXTEND, dl, MVT::f32, RHS);
+  }
+
+  // Next, handle integers.
   if (LHS.getValueType().isInteger()) {
     assert((LHS.getValueType() == RHS.getValueType()) &&
            (LHS.getValueType() == MVT::i32 || LHS.getValueType() == MVT::i64));
@@ -6400,24 +6403,11 @@ SDValue AArch64TargetLowering::LowerEXTRACT_SUBVECTOR(SDValue Op,
   unsigned Val = Cst->getZExtValue();
 
   unsigned Size = Op.getValueType().getSizeInBits();
-  if (Val == 0) {
-    switch (Size) {
-    case 8:
-      return DAG.getTargetExtractSubreg(AArch64::bsub, dl, Op.getValueType(),
-                                        Op.getOperand(0));
-    case 16:
-      return DAG.getTargetExtractSubreg(AArch64::hsub, dl, Op.getValueType(),
-                                        Op.getOperand(0));
-    case 32:
-      return DAG.getTargetExtractSubreg(AArch64::ssub, dl, Op.getValueType(),
-                                        Op.getOperand(0));
-    case 64:
-      return DAG.getTargetExtractSubreg(AArch64::dsub, dl, Op.getValueType(),
-                                        Op.getOperand(0));
-    default:
-      llvm_unreachable("Unexpected vector type in extract_subvector!");
-    }
-  }
+
+  // This will get lowered to an appropriate EXTRACT_SUBREG in ISel.
+  if (Val == 0)
+    return Op;
+
   // If this is extracting the upper 64-bits of a 128-bit vector, we match
   // that directly.
   if (Size == 64 && Val * VT.getVectorElementType().getSizeInBits() == 64)
@@ -8573,10 +8563,9 @@ static SDValue replaceSplatVectorStore(SelectionDAG &DAG, StoreSDNode *St) {
   return NewST1;
 }
 
-static SDValue performSTORECombine(SDNode *N,
-                                   TargetLowering::DAGCombinerInfo &DCI,
-                                   SelectionDAG &DAG,
-                                   const AArch64Subtarget *Subtarget) {
+static SDValue split16BStores(SDNode *N, TargetLowering::DAGCombinerInfo &DCI,
+                              SelectionDAG &DAG,
+                              const AArch64Subtarget *Subtarget) {
   if (!DCI.isBeforeLegalize())
     return SDValue();
 
@@ -8738,7 +8727,39 @@ static SDValue performPostLD1Combine(SDNode *N,
   return SDValue();
 }
 
-/// This function handles the log2-shuffle pattern produced by the
+/// Simplify \Addr given that the top byte of it is ignored by HW during
+/// address translation.
+static bool performTBISimplification(SDValue Addr,
+                                     TargetLowering::DAGCombinerInfo &DCI,
+                                     SelectionDAG &DAG) {
+  APInt DemandedMask = APInt::getLowBitsSet(64, 56);
+  APInt KnownZero, KnownOne;
+  TargetLowering::TargetLoweringOpt TLO(DAG, DCI.isBeforeLegalize(),
+                                        DCI.isBeforeLegalizeOps());
+  const TargetLowering &TLI = DAG.getTargetLoweringInfo();
+  if (TLI.SimplifyDemandedBits(Addr, DemandedMask, KnownZero, KnownOne, TLO)) {
+    DCI.CommitTargetLoweringOpt(TLO);
+    return true;
+  }
+  return false;
+}
+
+static SDValue performSTORECombine(SDNode *N,
+                                   TargetLowering::DAGCombinerInfo &DCI,
+                                   SelectionDAG &DAG,
+                                   const AArch64Subtarget *Subtarget) {
+  SDValue Split = split16BStores(N, DCI, DAG, Subtarget);
+  if (Split.getNode())
+    return Split;
+
+  if (Subtarget->supportsAddressTopByteIgnored() &&
+      performTBISimplification(N->getOperand(2), DCI, DAG))
+    return SDValue(N, 0);
+
+  return SDValue();
+}
+
+  /// This function handles the log2-shuffle pattern produced by the
 /// LoopVectorizer for the across vector reduction. It consists of
 /// log2(NumVectorElements) steps and, in each step, 2^(s) elements
 /// are reduced, where s is an induction variable from 0 to
@@ -9593,6 +9614,10 @@ SDValue AArch64TargetLowering::PerformDAGCombine(SDNode *N,
   }
   case ISD::VSELECT:
     return performVSelectCombine(N, DCI.DAG);
+  case ISD::LOAD:
+    if (performTBISimplification(N->getOperand(1), DCI, DAG))
+      return SDValue(N, 0);
+    break;
   case ISD::STORE:
     return performSTORECombine(N, DCI, DAG, Subtarget);
   case AArch64ISD::BRCOND: