[X86] Remove CustomInserters for RDPKRU/WRPKRU. Use some custom lowering and new ISD opcodes instead.

These inserters inserted some instructions to zero some registers and copied from virtual registers to physical registers.

This change instead inserts the zeros directly into the DAG at lowering time using new ISD opcodes
that take the extra zeroes as inputs. The zeros will then go through isel on their own to select
the MOV32r0 pseudo. Then we just need to mention the physical registers directly
in the isel patterns and the isel table and InstrEmitter will take care of inserting the necessary
copies to/from physical registers.

llvm-svn: 357659

5 files changed, 39 insertions, 54 deletions

diff --git a/llvm/lib/Target/X86/X86ISelLowering.cpp b/llvm/lib/Target/X86/X86ISelLowering.cpp
index 99cf132..b17d727 100644
--- a/llvm/lib/Target/X86/X86ISelLowering.cpp
+++ b/llvm/lib/Target/X86/X86ISelLowering.cpp
@@ -23035,6 +23035,22 @@ static SDValue LowerINTRINSIC_W_CHAIN(SDValue Op, const X86Subtarget &Subtarget,
       return MarkEHRegistrationNode(Op, DAG);
     case llvm::Intrinsic::x86_seh_ehguard:
       return MarkEHGuard(Op, DAG);
+    case llvm::Intrinsic::x86_rdpkru: {
+      SDLoc dl(Op);
+      SDVTList VTs = DAG.getVTList(MVT::i32, MVT::Other);
+      // Create a RDPKRU node and pass 0 to the ECX parameter.
+      return DAG.getNode(X86ISD::RDPKRU, dl, VTs, Op.getOperand(0),
+                         DAG.getConstant(0, dl, MVT::i32));
+    }
+    case llvm::Intrinsic::x86_wrpkru: {
+      SDLoc dl(Op);
+      // Create a WRPKRU node, pass the input to the EAX parameter,  and pass 0
+      // to the EDX and ECX parameters.
+      return DAG.getNode(X86ISD::WRPKRU, dl, MVT::Other,
+                         Op.getOperand(0), Op.getOperand(2),
+                         DAG.getConstant(0, dl, MVT::i32),
+                         DAG.getConstant(0, dl, MVT::i32));
+    }
     case llvm::Intrinsic::x86_flags_read_u32:
     case llvm::Intrinsic::x86_flags_read_u64:
     case llvm::Intrinsic::x86_flags_write_u32:
@@ -27808,6 +27824,8 @@ const char *X86TargetLowering::getTargetNodeName(unsigned Opcode) const {
   case X86ISD::SAHF:               return "X86ISD::SAHF";
   case X86ISD::RDRAND:             return "X86ISD::RDRAND";
   case X86ISD::RDSEED:             return "X86ISD::RDSEED";
+  case X86ISD::RDPKRU:             return "X86ISD::RDPKRU";
+  case X86ISD::WRPKRU:             return "X86ISD::WRPKRU";
   case X86ISD::VPMADDUBSW:         return "X86ISD::VPMADDUBSW";
   case X86ISD::VPMADDWD:           return "X86ISD::VPMADDWD";
   case X86ISD::VPSHA:              return "X86ISD::VPSHA";
@@ -28268,44 +28286,6 @@ static MachineBasicBlock *emitXBegin(MachineInstr &MI, MachineBasicBlock *MBB,
   return sinkMBB;
 }
 
-static MachineBasicBlock *emitWRPKRU(MachineInstr &MI, MachineBasicBlock *BB,
-                                     const X86Subtarget &Subtarget) {
-  DebugLoc dl = MI.getDebugLoc();
-  const TargetInstrInfo *TII = Subtarget.getInstrInfo();
-
-  // insert input VAL into EAX
-  BuildMI(*BB, MI, dl, TII->get(TargetOpcode::COPY), X86::EAX)
-      .addReg(MI.getOperand(0).getReg());
-  // insert zero to ECX
-  BuildMI(*BB, MI, dl, TII->get(X86::MOV32r0), X86::ECX);
-
-  // insert zero to EDX
-  BuildMI(*BB, MI, dl, TII->get(X86::MOV32r0), X86::EDX);
-
-  // insert WRPKRU instruction
-  BuildMI(*BB, MI, dl, TII->get(X86::WRPKRUr));
-
-  MI.eraseFromParent(); // The pseudo is gone now.
-  return BB;
-}
-
-static MachineBasicBlock *emitRDPKRU(MachineInstr &MI, MachineBasicBlock *BB,
-                                     const X86Subtarget &Subtarget) {
-  DebugLoc dl = MI.getDebugLoc();
-  const TargetInstrInfo *TII = Subtarget.getInstrInfo();
-
-  // insert zero to ECX
-  BuildMI(*BB, MI, dl, TII->get(X86::MOV32r0), X86::ECX);
-
-  // insert RDPKRU instruction
-  BuildMI(*BB, MI, dl, TII->get(X86::RDPKRUr));
-  BuildMI(*BB, MI, dl, TII->get(TargetOpcode::COPY), MI.getOperand(0).getReg())
-      .addReg(X86::EAX);
-
-  MI.eraseFromParent(); // The pseudo is gone now.
-  return BB;
-}
-
 
 
 MachineBasicBlock *
@@ -30418,11 +30398,6 @@ X86TargetLowering::EmitInstrWithCustomInserter(MachineInstr &MI,
     return BB;
   }
 
-  // PKU feature
-  case X86::WRPKRU:
-    return emitWRPKRU(MI, BB, Subtarget);
-  case X86::RDPKRU:
-    return emitRDPKRU(MI, BB, Subtarget);
   // xbegin
   case X86::XBEGIN:
     return emitXBegin(MI, BB, Subtarget.getInstrInfo());
diff --git a/llvm/lib/Target/X86/X86ISelLowering.h b/llvm/lib/Target/X86/X86ISelLowering.h
index 0f7065b..7042a29 100644
--- a/llvm/lib/Target/X86/X86ISelLowering.h
+++ b/llvm/lib/Target/X86/X86ISelLowering.h
@@ -538,6 +538,12 @@ namespace llvm {
       // indicate whether it is valid in CF.
       RDSEED,
 
+      // Protection keys
+      // RDPKRU - Operand 0 is chain. Operand 1 is value for ECX.
+      // WRPKRU - Operand 0 is chain. Operand 1 is value for EDX. Operand 2 is
+      // value for ECX.
+      RDPKRU, WRPKRU,
+
       // SSE42 string comparisons.
       // These nodes produce 3 results, index, mask, and flags. X86ISelDAGToDAG
       // will emit one or two instructions based on which results are used. If
diff --git a/llvm/lib/Target/X86/X86InstrInfo.td b/llvm/lib/Target/X86/X86InstrInfo.td
index e399f8b..e07553f 100644
--- a/llvm/lib/Target/X86/X86InstrInfo.td
+++ b/llvm/lib/Target/X86/X86InstrInfo.td
@@ -63,6 +63,10 @@ def SDTX86sahf : SDTypeProfile<1, 1, [SDTCisVT<0, i32>, SDTCisVT<1, i8>]>;
 
 def SDTX86rdrand : SDTypeProfile<2, 0, [SDTCisInt<0>, SDTCisVT<1, i32>]>;
 
+def SDTX86rdpkru : SDTypeProfile<1, 1, [SDTCisVT<0, i32>, SDTCisVT<1, i32>]>;
+def SDTX86wrpkru : SDTypeProfile<0, 3, [SDTCisVT<0, i32>, SDTCisVT<1, i32>,
+                                        SDTCisVT<2, i32>]>;
+
 def SDTX86cas : SDTypeProfile<0, 3, [SDTCisPtrTy<0>, SDTCisInt<1>,
                                      SDTCisVT<2, i8>]>;
 def SDTX86caspair : SDTypeProfile<0, 1, [SDTCisPtrTy<0>]>;
@@ -151,6 +155,11 @@ def X86rdrand  : SDNode<"X86ISD::RDRAND",   SDTX86rdrand,
 def X86rdseed  : SDNode<"X86ISD::RDSEED",   SDTX86rdrand,
                         [SDNPHasChain, SDNPSideEffect]>;
 
+def X86rdpkru : SDNode<"X86ISD::RDPKRU",    SDTX86rdpkru,
+                       [SDNPHasChain, SDNPSideEffect]>;
+def X86wrpkru : SDNode<"X86ISD::WRPKRU",    SDTX86wrpkru,
+                       [SDNPHasChain, SDNPSideEffect]>;
+
 def X86cas : SDNode<"X86ISD::LCMPXCHG_DAG", SDTX86cas,
                         [SDNPHasChain, SDNPInGlue, SDNPOutGlue, SDNPMayStore,
                          SDNPMayLoad, SDNPMemOperand]>;
diff --git a/llvm/lib/Target/X86/X86InstrSystem.td b/llvm/lib/Target/X86/X86InstrSystem.td
index e08c508..7050e19 100644
--- a/llvm/lib/Target/X86/X86InstrSystem.td
+++ b/llvm/lib/Target/X86/X86InstrSystem.td
@@ -587,18 +587,13 @@ let Defs = [RAX, RDX, RSI], Uses = [RAX, RSI] in
 
 //==-----------------------------------------------------------------------===//
 // PKU  - enable protection key
-let usesCustomInserter = 1, hasNoSchedulingInfo = 1 in {
-  def WRPKRU : PseudoI<(outs), (ins GR32:$src),
-                [(int_x86_wrpkru GR32:$src)]>;
-  def RDPKRU : PseudoI<(outs GR32:$dst), (ins),
-                [(set GR32:$dst, (int_x86_rdpkru))]>;
-}
-
 let SchedRW = [WriteSystem] in {
 let Defs = [EAX, EDX], Uses = [ECX] in
-  def RDPKRUr : I<0x01, MRM_EE, (outs), (ins), "rdpkru", []>, TB;
+  def RDPKRUr : I<0x01, MRM_EE, (outs), (ins), "rdpkru",
+                  [(set EAX, (X86rdpkru ECX)), (implicit EDX)]>, TB;
 let Uses = [EAX, ECX, EDX] in
-  def WRPKRUr : I<0x01, MRM_EF, (outs), (ins), "wrpkru", []>, TB;
+  def WRPKRUr : I<0x01, MRM_EF, (outs), (ins), "wrpkru",
+                  [(X86wrpkru EAX, EDX, ECX)]>, TB;
 } // SchedRW
 
 //===----------------------------------------------------------------------===//
diff --git a/llvm/test/CodeGen/X86/pku.ll b/llvm/test/CodeGen/X86/pku.ll
index e5b47ecc..b6b2f98 100644
--- a/llvm/test/CodeGen/X86/pku.ll
+++ b/llvm/test/CodeGen/X86/pku.ll
@@ -9,16 +9,16 @@ define void @test_x86_wrpkru(i32 %src) {
 ; X86-LABEL: test_x86_wrpkru:
 ; X86:       ## %bb.0:
 ; X86-NEXT:    movl {{[0-9]+}}(%esp), %eax ## encoding: [0x8b,0x44,0x24,0x04]
-; X86-NEXT:    xorl %ecx, %ecx ## encoding: [0x31,0xc9]
 ; X86-NEXT:    xorl %edx, %edx ## encoding: [0x31,0xd2]
+; X86-NEXT:    xorl %ecx, %ecx ## encoding: [0x31,0xc9]
 ; X86-NEXT:    wrpkru ## encoding: [0x0f,0x01,0xef]
 ; X86-NEXT:    retl ## encoding: [0xc3]
 ;
 ; X64-LABEL: test_x86_wrpkru:
 ; X64:       ## %bb.0:
 ; X64-NEXT:    movl %edi, %eax ## encoding: [0x89,0xf8]
-; X64-NEXT:    xorl %ecx, %ecx ## encoding: [0x31,0xc9]
 ; X64-NEXT:    xorl %edx, %edx ## encoding: [0x31,0xd2]
+; X64-NEXT:    xorl %ecx, %ecx ## encoding: [0x31,0xc9]
 ; X64-NEXT:    wrpkru ## encoding: [0x0f,0x01,0xef]
 ; X64-NEXT:    retq ## encoding: [0xc3]
   call void @llvm.x86.wrpkru(i32 %src)