[Xtensa] Lower ATOMIC_FENCE. Add Atomic Expand pass.

Author: andreisfr

llvm/lib/Target/Xtensa/XtensaISelLowering.cpp

@@ -304,6 +304,10 @@ XtensaTargetLowering::XtensaTargetLowering(const TargetMachine &tm,
 
   setOperationAction(ISD::TRAP, MVT::Other, Legal);
 
+  // to have the best chance and doing something good with fences custom lower
+  // them
+  setOperationAction(ISD::ATOMIC_FENCE, MVT::Other, Custom);
+
   // Compute derived properties from the register classes
   computeRegisterProperties(STI.getRegisterInfo());
 
@@ -1649,6 +1653,13 @@ SDValue XtensaTargetLowering::LowerShiftRightParts(SDValue Op,
   return DAG.getMergeValues(Ops, DL);
 }
 
+SDValue XtensaTargetLowering::LowerATOMIC_FENCE(SDValue Op,
+                                                SelectionDAG &DAG) const {
+  SDLoc DL(Op);
+  SDValue Chain = Op.getOperand(0);
+  return DAG.getNode(XtensaISD::MEMW, DL, MVT::Other, Chain);
+}
+
 SDValue XtensaTargetLowering::LowerOperation(SDValue Op,
                                              SelectionDAG &DAG) const {
   switch (Op.getOpcode()) {
@@ -1688,6 +1699,8 @@ SDValue XtensaTargetLowering::LowerOperation(SDValue Op,
     return LowerVASTART(Op, DAG);
   case ISD::VACOPY:
     return LowerVACOPY(Op, DAG);
+  case ISD::ATOMIC_FENCE:
+    return LowerATOMIC_FENCE(Op, DAG);
   case ISD::SHL_PARTS:
     return LowerShiftLeftParts(Op, DAG);
   case ISD::SRA_PARTS:
@@ -1725,6 +1738,7 @@ const char *XtensaTargetLowering::getTargetNodeName(unsigned Opcode) const {
     OPCODE(MADD);
     OPCODE(MSUB);
     OPCODE(MOVS);
+    OPCODE(MEMW);
     OPCODE(MOVSP);
     OPCODE(RUR);
     OPCODE(SHL);

llvm/lib/Target/Xtensa/XtensaISelLowering.h

@@ -52,6 +52,8 @@ enum {
   // FP move
   MOVS,
 
+  MEMW,
+
   MOVSP,
 
   // Wraps a TargetGlobalAddress that should be loaded using PC-relative
@@ -157,6 +159,10 @@ class XtensaTargetLowering : public TargetLowering {
                       const SmallVectorImpl<SDValue> &OutVals, const SDLoc &DL,
                       SelectionDAG &DAG) const override;
 
+  bool shouldInsertFencesForAtomic(const Instruction *I) const override {
+    return true;
+  }
+
   MachineBasicBlock *
   EmitInstrWithCustomInserter(MachineInstr &MI,
                               MachineBasicBlock *BB) const override;
@@ -190,6 +196,8 @@ class XtensaTargetLowering : public TargetLowering {
   SDValue LowerShiftLeftParts(SDValue Op, SelectionDAG &DAG) const;
   SDValue LowerShiftRightParts(SDValue Op, SelectionDAG &DAG, bool IsSRA) const;
 
+  SDValue LowerATOMIC_FENCE(SDValue Op, SelectionDAG &DAG) const;
+
   SDValue getAddrPCRel(SDValue Op, SelectionDAG &DAG) const;
 
   CCAssignFn *CCAssignFnForCall(CallingConv::ID CC, bool IsVarArg) const;

llvm/lib/Target/Xtensa/XtensaInstrInfo.td

@@ -613,6 +613,8 @@ def EXTW : RRR_Inst<0x00, 0x00, 0x00, (outs), (ins),
   let t = 0xd;
 }
 
+def : Pat<(Xtensa_mem_barrier), (MEMW)>;
+
 //===----------------------------------------------------------------------===//
 // Processor control instructions
 //===----------------------------------------------------------------------===//
@@ -1484,6 +1486,18 @@ def WITLB: RRR_Inst<0x00, 0x00, 0x05, (outs AR:$t), (ins AR:$s),
   let r = 0x6;
 }
 
+//===----------------------------------------------------------------------===//
+// Atomic patterns
+//===----------------------------------------------------------------------===//
+
+def : Pat<(i32 (atomic_load_8  addr_ish1:$addr)), (L8UI addr_ish1:$addr)>;
+def : Pat<(i32 (atomic_load_16 addr_ish2:$addr)), (L16UI addr_ish2:$addr)>;
+def : Pat<(i32 (atomic_load_32 addr_ish4:$addr)), (L32I addr_ish4:$addr)>;
+
+def : Pat<(atomic_store_8  addr_ish1:$addr, AR:$t), (S8I AR:$t, addr_ish1:$addr)>;
+def : Pat<(atomic_store_16 addr_ish2:$addr, AR:$t), (S16I AR:$t, addr_ish2:$addr)>;
+def : Pat<(atomic_store_32 addr_ish4:$addr, AR:$t), (S32I AR:$t, addr_ish4:$addr)>;
+
 //===----------------------------------------------------------------------===//
 // DSP Instructions
 //===----------------------------------------------------------------------===//

llvm/lib/Target/Xtensa/XtensaOperators.td

@@ -40,6 +40,7 @@ def SDT_XtensaSRC                 : SDTypeProfile<1, 2, [SDTCisVT<0, i32>, SDTCi
                                                          SDTCisVT<2, i32>]>;
 def SDT_XtensaSSL                 : SDTypeProfile<0, 1, [SDTCisVT<0, i32>]>;
 def SDT_XtensaSSR                 : SDTypeProfile<0, 1, [SDTCisVT<0, i32>]>;
+def SDT_XtensaMEMBARRIER         : SDTypeProfile<0, 0, []>;
 def SDT_XtensaRUR                : SDTypeProfile<1, 1, [SDTCisVT<0, i32>, SDTCisVT<1, i32>]>;
 
 //===----------------------------------------------------------------------===//
@@ -98,6 +99,8 @@ def Xtensa_ssr: SDNode<"XtensaISD::SSR", SDT_XtensaSSR, [SDNPOutGlue]>;
 def Xtensa_brjt: SDNode<"XtensaISD::BR_JT", SDT_XtensaBrJT, [SDNPHasChain]>;
 def Xtensa_callw: SDNode<"XtensaISD::CALLW", SDT_XtensaCall,
                         [SDNPHasChain, SDNPOutGlue, SDNPOptInGlue, SDNPVariadic]>;
+def Xtensa_mem_barrier: SDNode<"XtensaISD::MEMW", SDT_XtensaMEMBARRIER,
+                              [SDNPHasChain, SDNPSideEffect]>;
 
 def Xtensa_rur: SDNode<"XtensaISD::RUR", SDT_XtensaRUR,
-                            [SDNPInGlue]>;
+                      [SDNPInGlue]>;

llvm/lib/Target/Xtensa/XtensaTargetMachine.cpp

@@ -82,6 +82,7 @@ class XtensaPassConfig : public TargetPassConfig {
     return getTM<XtensaTargetMachine>();
   }
 
+  void addIRPasses() override;
   bool addInstSelector() override;
   void addPreEmitPass() override;
 };
@@ -92,6 +93,8 @@ bool XtensaPassConfig::addInstSelector() {
   return false;
 }
 
+void XtensaPassConfig::addIRPasses() { addPass(createAtomicExpandPass()); }
+
 void XtensaPassConfig::addPreEmitPass() {
   addPass(createXtensaSizeReductionPass());
   addPass(&BranchRelaxationPassID);