[SystemZ] Optimize selects between 0 and -1

Since z has no setcc instruction as such, the choice of setBooleanContents
is a bit arbitrary.  Currently it's set to ZeroOrOneBooleanContent,
so we produced a branch-free form when selecting between 0 and 1,
but not when selecting between 0 and -1.  This patch handles the latter
case too.

At some point I'd like to measure whether it's better to use conditional
moves for constant selects on z196, but that's future work.


git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@196578 91177308-0d34-0410-b5e6-96231b3b80d8

1 files changed, 44 insertions, 14 deletions

diff --git a/lib/Target/SystemZ/SystemZISelLowering.cpp b/lib/Target/SystemZ/SystemZISelLowering.cpp
index f6e18530f4..0340a876ee 100644
--- a/lib/Target/SystemZ/SystemZISelLowering.cpp
+++ b/lib/Target/SystemZ/SystemZISelLowering.cpp
@@ -1486,16 +1486,11 @@ static void lowerGR128Binary(SelectionDAG &DAG, SDLoc DL, EVT VT,
   Odd = DAG.getTargetExtractSubreg(SystemZ::odd128(Is32Bit), DL, VT, Result);
 }
 
-SDValue SystemZTargetLowering::lowerSETCC(SDValue Op,
-                                          SelectionDAG &DAG) const {
-  SDValue CmpOp0   = Op.getOperand(0);
-  SDValue CmpOp1   = Op.getOperand(1);
-  ISD::CondCode CC = cast<CondCodeSDNode>(Op.getOperand(2))->get();
-  SDLoc DL(Op);
-
-  unsigned CCValid, CCMask;
-  SDValue Glue = emitCmp(TM, DAG, DL, CmpOp0, CmpOp1, CC, CCValid, CCMask);
-
+// Return an i32 value that is 1 if the CC value produced by Glue is
+// in the mask CCMask and 0 otherwise.  CC is known to have a value
+// in CCValid, so other values can be ignored.
+static SDValue emitSETCC(SelectionDAG &DAG, SDLoc DL, SDValue Glue,
+                         unsigned CCValid, unsigned CCMask) {
   IPMConversion Conversion = getIPMConversion(CCValid, CCMask);
   SDValue Result = DAG.getNode(SystemZISD::IPM, DL, MVT::i32, Glue);
 
@@ -1516,6 +1511,18 @@ SDValue SystemZTargetLowering::lowerSETCC(SDValue Op,
   return Result;
 }
 
+SDValue SystemZTargetLowering::lowerSETCC(SDValue Op,
+                                          SelectionDAG &DAG) const {
+  SDValue CmpOp0   = Op.getOperand(0);
+  SDValue CmpOp1   = Op.getOperand(1);
+  ISD::CondCode CC = cast<CondCodeSDNode>(Op.getOperand(2))->get();
+  SDLoc DL(Op);
+
+  unsigned CCValid, CCMask;
+  SDValue Glue = emitCmp(TM, DAG, DL, CmpOp0, CmpOp1, CC, CCValid, CCMask);
+  return emitSETCC(DAG, DL, Glue, CCValid, CCMask);
+}
+
 SDValue SystemZTargetLowering::lowerBR_CC(SDValue Op, SelectionDAG &DAG) const {
   SDValue Chain    = Op.getOperand(0);
   ISD::CondCode CC = cast<CondCodeSDNode>(Op.getOperand(1))->get();
@@ -1525,10 +1532,10 @@ SDValue SystemZTargetLowering::lowerBR_CC(SDValue Op, SelectionDAG &DAG) const {
   SDLoc DL(Op);
 
   unsigned CCValid, CCMask;
-  SDValue Flags = emitCmp(TM, DAG, DL, CmpOp0, CmpOp1, CC, CCValid, CCMask);
+  SDValue Glue = emitCmp(TM, DAG, DL, CmpOp0, CmpOp1, CC, CCValid, CCMask);
   return DAG.getNode(SystemZISD::BR_CCMASK, DL, Op.getValueType(),
                      Chain, DAG.getConstant(CCValid, MVT::i32),
-                     DAG.getConstant(CCMask, MVT::i32), Dest, Flags);
+                     DAG.getConstant(CCMask, MVT::i32), Dest, Glue);
 }
 
 SDValue SystemZTargetLowering::lowerSELECT_CC(SDValue Op,
@@ -1541,14 +1548,37 @@ SDValue SystemZTargetLowering::lowerSELECT_CC(SDValue Op,
   SDLoc DL(Op);
 
   unsigned CCValid, CCMask;
-  SDValue Flags = emitCmp(TM, DAG, DL, CmpOp0, CmpOp1, CC, CCValid, CCMask);
+  SDValue Glue = emitCmp(TM, DAG, DL, CmpOp0, CmpOp1, CC, CCValid, CCMask);
+
+  // Special case for handling -1/0 results.  The shifts we use here
+  // should get optimized with the IPM conversion sequence.
+  ConstantSDNode *TrueC = dyn_cast<ConstantSDNode>(TrueOp);
+  ConstantSDNode *FalseC = dyn_cast<ConstantSDNode>(FalseOp);
+  if (TrueC && FalseC) {
+    int64_t TrueVal = TrueC->getSExtValue();
+    int64_t FalseVal = FalseC->getSExtValue();
+    if ((TrueVal == -1 && FalseVal == 0) || (TrueVal == 0 && FalseVal == -1)) {
+      // Invert the condition if we want -1 on false.
+      if (TrueVal == 0)
+        CCMask ^= CCValid;
+      SDValue Result = emitSETCC(DAG, DL, Glue, CCValid, CCMask);
+      EVT VT = Op.getValueType();
+      // Extend the result to VT.  Upper bits are ignored.
+      if (!is32Bit(VT))
+        Result = DAG.getNode(ISD::ANY_EXTEND, DL, VT, Result);
+      // Sign-extend from the low bit.
+      SDValue ShAmt = DAG.getConstant(VT.getSizeInBits() - 1, MVT::i32);
+      SDValue Shl = DAG.getNode(ISD::SHL, DL, VT, Result, ShAmt);
+      return DAG.getNode(ISD::SRA, DL, VT, Shl, ShAmt);
+    }
+  }
 
   SmallVector<SDValue, 5> Ops;
   Ops.push_back(TrueOp);
   Ops.push_back(FalseOp);
   Ops.push_back(DAG.getConstant(CCValid, MVT::i32));
   Ops.push_back(DAG.getConstant(CCMask, MVT::i32));
-  Ops.push_back(Flags);
+  Ops.push_back(Glue);
 
   SDVTList VTs = DAG.getVTList(Op.getValueType(), MVT::Glue);
   return DAG.getNode(SystemZISD::SELECT_CCMASK, DL, VTs, &Ops[0], Ops.size());