Merge upstream branch release_60

CMakeLists.txt

@@ -24,7 +24,7 @@ if(NOT DEFINED LLVM_VERSION_MINOR)
   set(LLVM_VERSION_MINOR 0)
 endif()
 if(NOT DEFINED LLVM_VERSION_PATCH)
-  set(LLVM_VERSION_PATCH 0)
+  set(LLVM_VERSION_PATCH 1)
 endif()
 if(NOT DEFINED LLVM_VERSION_SUFFIX)
   set(LLVM_VERSION_SUFFIX "")

include/llvm/CodeGen/TargetInstrInfo.h

@@ -421,7 +421,8 @@ class TargetInstrInfo : public MCInstrInfo {
   /// Build the equivalent inputs of a REG_SEQUENCE for the given \p MI
   /// and \p DefIdx.
   /// \p [out] InputRegs of the equivalent REG_SEQUENCE. Each element of
-  /// the list is modeled as <Reg:SubReg, SubIdx>.
+  /// the list is modeled as <Reg:SubReg, SubIdx>. Operands with the undef
+  /// flag are not added to this list.
   /// E.g., REG_SEQUENCE %1:sub1, sub0, %2, sub1 would produce
   /// two elements:
   /// - %1:sub1, sub0
@@ -446,7 +447,8 @@ class TargetInstrInfo : public MCInstrInfo {
   /// - %1:sub1, sub0
   ///
   /// \returns true if it is possible to build such an input sequence
-  /// with the pair \p MI, \p DefIdx. False otherwise.
+  /// with the pair \p MI, \p DefIdx and the operand has no undef flag set.
+  /// False otherwise.
   ///
   /// \pre MI.isExtractSubreg() or MI.isExtractSubregLike().
   ///
@@ -465,7 +467,8 @@ class TargetInstrInfo : public MCInstrInfo {
   /// - InsertedReg: %1:sub1, sub3
   ///
   /// \returns true if it is possible to build such an input sequence
-  /// with the pair \p MI, \p DefIdx. False otherwise.
+  /// with the pair \p MI, \p DefIdx and the operand has no undef flag set.
+  /// False otherwise.
   ///
   /// \pre MI.isInsertSubreg() or MI.isInsertSubregLike().
   ///

lib/Analysis/GlobalsModRef.cpp

@@ -502,6 +502,8 @@ void GlobalsAAResult::AnalyzeCallGraph(CallGraph &CG, Module &M) {
     }
 
     FunctionInfo &FI = FunctionInfos[F];
+    Handles.emplace_front(*this, F);
+    Handles.front().I = Handles.begin();
     bool KnowNothing = false;
 
     // Collect the mod/ref properties due to called functions.  We only compute

lib/Analysis/MemorySSA.cpp

@@ -153,9 +153,14 @@ class MemoryLocOrCall {
     if (IsCall != Other.IsCall)
       return false;
 
-    if (IsCall)
-      return CS.getCalledValue() == Other.CS.getCalledValue();
-    return Loc == Other.Loc;
+    if (!IsCall)
+      return Loc == Other.Loc;
+
+    if (CS.getCalledValue() != Other.CS.getCalledValue())
+      return false;
+
+    return CS.arg_size() == Other.CS.arg_size() &&
+           std::equal(CS.arg_begin(), CS.arg_end(), Other.CS.arg_begin());
   }
 
 private:
@@ -179,12 +184,18 @@ template <> struct DenseMapInfo<MemoryLocOrCall> {
   }
 
   static unsigned getHashValue(const MemoryLocOrCall &MLOC) {
-    if (MLOC.IsCall)
-      return hash_combine(MLOC.IsCall,
-                          DenseMapInfo<const Value *>::getHashValue(
-                              MLOC.getCS().getCalledValue()));
-    return hash_combine(
-        MLOC.IsCall, DenseMapInfo<MemoryLocation>::getHashValue(MLOC.getLoc()));
+    if (!MLOC.IsCall)
+      return hash_combine(
+          MLOC.IsCall,
+          DenseMapInfo<MemoryLocation>::getHashValue(MLOC.getLoc()));
+
+    hash_code hash =
+        hash_combine(MLOC.IsCall, DenseMapInfo<const Value *>::getHashValue(
+                                      MLOC.getCS().getCalledValue()));
+
+    for (const Value *Arg : MLOC.getCS().args())
+      hash = hash_combine(hash, DenseMapInfo<const Value *>::getHashValue(Arg));
+    return hash;
   }
 
   static bool isEqual(const MemoryLocOrCall &LHS, const MemoryLocOrCall &RHS) {

lib/CodeGen/LiveDebugVariables.cpp

@@ -557,8 +557,11 @@ bool LDVImpl::handleDebugValue(MachineInstr &MI, SlotIndex Idx) {
       getUserValue(Var, Expr, MI.getDebugLoc());
   if (!Discard)
     UV->addDef(Idx, MI.getOperand(0), IsIndirect);
-  else
-    UV->addDef(Idx, MachineOperand::CreateReg(0U, RegState::Debug), false);
+  else {
+    MachineOperand MO = MachineOperand::CreateReg(0U, false);
+    MO.setIsDebug();
+    UV->addDef(Idx, MO, false);
+  }
   return true;
 }
 

lib/CodeGen/PeepholeOptimizer.cpp

@@ -1882,6 +1882,8 @@ ValueTrackerResult ValueTracker::getNextSourceFromCopy() {
     return ValueTrackerResult();
   // Otherwise, we want the whole source.
   const MachineOperand &Src = Def->getOperand(1);
+  if (Src.isUndef())
+    return ValueTrackerResult();
   return ValueTrackerResult(Src.getReg(), Src.getSubReg());
 }
 
@@ -1925,6 +1927,8 @@ ValueTrackerResult ValueTracker::getNextSourceFromBitcast() {
   }
 
   const MachineOperand &Src = Def->getOperand(SrcIdx);
+  if (Src.isUndef())
+    return ValueTrackerResult();
   return ValueTrackerResult(Src.getReg(), Src.getSubReg());
 }
 
@@ -2093,6 +2097,10 @@ ValueTrackerResult ValueTracker::getNextSourceFromPHI() {
   for (unsigned i = 1, e = Def->getNumOperands(); i < e; i += 2) {
     auto &MO = Def->getOperand(i);
     assert(MO.isReg() && "Invalid PHI instruction");
+    // We have no code to deal with undef operands. They shouldn't happen in
+    // normal programs anyway.
+    if (MO.isUndef())
+      return ValueTrackerResult();
     Res.addSource(MO.getReg(), MO.getSubReg());
   }
 
@@ -2149,9 +2157,14 @@ ValueTrackerResult ValueTracker::getNextSource() {
     // If we can still move up in the use-def chain, move to the next
     // definition.
     if (!TargetRegisterInfo::isPhysicalRegister(Reg) && OneRegSrc) {
-      Def = MRI.getVRegDef(Reg);
-      DefIdx = MRI.def_begin(Reg).getOperandNo();
-      DefSubReg = Res.getSrcSubReg(0);
+      MachineRegisterInfo::def_iterator DI = MRI.def_begin(Reg);
+      if (DI != MRI.def_end()) {
+        Def = DI->getParent();
+        DefIdx = DI.getOperandNo();
+        DefSubReg = Res.getSrcSubReg(0);
+      } else {
+        Def = nullptr;
+      }
       return Res;
     }
   }

lib/CodeGen/TargetInstrInfo.cpp

@@ -1151,6 +1151,8 @@ bool TargetInstrInfo::getRegSequenceInputs(
   for (unsigned OpIdx = 1, EndOpIdx = MI.getNumOperands(); OpIdx != EndOpIdx;
        OpIdx += 2) {
     const MachineOperand &MOReg = MI.getOperand(OpIdx);
+    if (MOReg.isUndef())
+      continue;
     const MachineOperand &MOSubIdx = MI.getOperand(OpIdx + 1);
     assert(MOSubIdx.isImm() &&
            "One of the subindex of the reg_sequence is not an immediate");
@@ -1174,6 +1176,8 @@ bool TargetInstrInfo::getExtractSubregInputs(
   // Def = EXTRACT_SUBREG v0.sub1, sub0.
   assert(DefIdx == 0 && "EXTRACT_SUBREG only has one def");
   const MachineOperand &MOReg = MI.getOperand(1);
+  if (MOReg.isUndef())
+    return false;
   const MachineOperand &MOSubIdx = MI.getOperand(2);
   assert(MOSubIdx.isImm() &&
          "The subindex of the extract_subreg is not an immediate");
@@ -1198,6 +1202,8 @@ bool TargetInstrInfo::getInsertSubregInputs(
   assert(DefIdx == 0 && "INSERT_SUBREG only has one def");
   const MachineOperand &MOBaseReg = MI.getOperand(1);
   const MachineOperand &MOInsertedReg = MI.getOperand(2);
+  if (MOInsertedReg.isUndef())
+    return false;
   const MachineOperand &MOSubIdx = MI.getOperand(3);
   assert(MOSubIdx.isImm() &&
          "One of the subindex of the reg_sequence is not an immediate");

lib/Support/Host.cpp

@@ -1009,7 +1009,7 @@ StringRef sys::getHostCPUName() {
 #include "llvm/Support/X86TargetParser.def"
 
   // Now check types.
-#define X86_CPU_SUBTYPE(ARCHNAME, ENUM) \
+#define X86_CPU_TYPE(ARCHNAME, ENUM) \
   if (Type == X86::ENUM) \
     return ARCHNAME;
 #include "llvm/Support/X86TargetParser.def"

lib/Target/AArch64/AArch64FalkorHWPFFix.cpp

@@ -46,6 +46,7 @@
 #include "llvm/Pass.h"
 #include "llvm/Support/Casting.h"
 #include "llvm/Support/Debug.h"
+#include "llvm/Support/DebugCounter.h"
 #include "llvm/Support/raw_ostream.h"
 #include <cassert>
 #include <iterator>
@@ -60,6 +61,8 @@ STATISTIC(NumCollisionsAvoided,
           "Number of HW prefetch tag collisions avoided");
 STATISTIC(NumCollisionsNotAvoided,
           "Number of HW prefetch tag collisions not avoided due to lack of regsiters");
+DEBUG_COUNTER(FixCounter, "falkor-hwpf",
+              "Controls which tag collisions are avoided");
 
 namespace {
 
@@ -729,6 +732,21 @@ void FalkorHWPFFix::runOnLoop(MachineLoop &L, MachineFunction &Fn) {
       bool Fixed = false;
       DEBUG(dbgs() << "Attempting to fix tag collision: " << MI);
 
+      if (!DebugCounter::shouldExecute(FixCounter)) {
+        DEBUG(dbgs() << "Skipping fix due to debug counter:\n  " << MI);
+        continue;
+      }
+
+      // Add the non-base registers of MI as live so we don't use them as
+      // scratch registers.
+      for (unsigned OpI = 0, OpE = MI.getNumOperands(); OpI < OpE; ++OpI) {
+        if (OpI == static_cast<unsigned>(LdI.BaseRegIdx))
+          continue;
+        MachineOperand &MO = MI.getOperand(OpI);
+        if (MO.isReg() && MO.readsReg())
+          LR.addReg(MO.getReg());
+      }
+
       for (unsigned ScratchReg : AArch64::GPR64RegClass) {
         if (!LR.available(ScratchReg) || MRI.isReserved(ScratchReg))
           continue;

lib/Target/AMDGPU/AMDGPUTargetMachine.cpp

@@ -147,6 +147,7 @@ extern "C" void LLVMInitializeAMDGPUTarget() {
   initializeR600PacketizerPass(*PR);
   initializeR600ExpandSpecialInstrsPassPass(*PR);
   initializeR600VectorRegMergerPass(*PR);
+  initializeGlobalISel(*PR);
   initializeAMDGPUDAGToDAGISelPass(*PR);
   initializeSILowerI1CopiesPass(*PR);
   initializeSIFixSGPRCopiesPass(*PR);

lib/Target/AMDGPU/SIISelLowering.cpp

@@ -358,6 +358,7 @@ SITargetLowering::SITargetLowering(const TargetMachine &TM,
     setOperationAction(ISD::CTTZ_ZERO_UNDEF, MVT::i16, Promote);
     setOperationAction(ISD::CTLZ, MVT::i16, Promote);
     setOperationAction(ISD::CTLZ_ZERO_UNDEF, MVT::i16, Promote);
+    setOperationAction(ISD::CTPOP, MVT::i16, Promote);
 
     setOperationAction(ISD::SELECT_CC, MVT::i16, Expand);
 

lib/Target/AMDGPU/SIInstructions.td

@@ -726,6 +726,10 @@ def : GCNPat <
   (i32 (add (i32 (ctpop i32:$popcnt)), i32:$val)),
   (V_BCNT_U32_B32_e64 $popcnt, $val)
 >;
+def : GCNPat <
+  (i16 (add (i16 (trunc (ctpop i32:$popcnt))), i16:$val)),
+  (V_BCNT_U32_B32_e64 $popcnt, $val)
+>;
 
 /********** ============================================ **********/
 /********** Extraction, Insertion, Building and Casting  **********/

lib/Target/ARM/ARMBaseInstrInfo.cpp

@@ -4864,12 +4864,14 @@ bool ARMBaseInstrInfo::getRegSequenceLikeInputs(
     // Populate the InputRegs accordingly.
     // rY
     const MachineOperand *MOReg = &MI.getOperand(1);
-    InputRegs.push_back(
-        RegSubRegPairAndIdx(MOReg->getReg(), MOReg->getSubReg(), ARM::ssub_0));
+    if (!MOReg->isUndef())
+      InputRegs.push_back(RegSubRegPairAndIdx(MOReg->getReg(),
+                                              MOReg->getSubReg(), ARM::ssub_0));
     // rZ
     MOReg = &MI.getOperand(2);
-    InputRegs.push_back(
-        RegSubRegPairAndIdx(MOReg->getReg(), MOReg->getSubReg(), ARM::ssub_1));
+    if (!MOReg->isUndef())
+      InputRegs.push_back(RegSubRegPairAndIdx(MOReg->getReg(),
+                                              MOReg->getSubReg(), ARM::ssub_1));
     return true;
   }
   llvm_unreachable("Target dependent opcode missing");
@@ -4888,6 +4890,8 @@ bool ARMBaseInstrInfo::getExtractSubregLikeInputs(
     // rX = EXTRACT_SUBREG dZ, ssub_0
     // rY = EXTRACT_SUBREG dZ, ssub_1
     const MachineOperand &MOReg = MI.getOperand(2);
+    if (MOReg.isUndef())
+      return false;
     InputReg.Reg = MOReg.getReg();
     InputReg.SubReg = MOReg.getSubReg();
     InputReg.SubIdx = DefIdx == 0 ? ARM::ssub_0 : ARM::ssub_1;
@@ -4907,6 +4911,8 @@ bool ARMBaseInstrInfo::getInsertSubregLikeInputs(
     // dX = VSETLNi32 dY, rZ, imm
     const MachineOperand &MOBaseReg = MI.getOperand(1);
     const MachineOperand &MOInsertedReg = MI.getOperand(2);
+    if (MOInsertedReg.isUndef())
+      return false;
     const MachineOperand &MOIndex = MI.getOperand(3);
     BaseReg.Reg = MOBaseReg.getReg();
     BaseReg.SubReg = MOBaseReg.getSubReg();

lib/Target/ARM/ARMComputeBlockSize.cpp

@@ -35,6 +35,7 @@ mayOptimizeThumb2Instruction(const MachineInstr *MI) {
     case ARM::tBcc:
     // optimizeThumb2JumpTables.
     case ARM::t2BR_JT:
+    case ARM::tBR_JTr:
       return true;
   }
   return false;

lib/Target/Mips/AsmParser/MipsAsmParser.cpp

@@ -5136,6 +5136,7 @@ unsigned MipsAsmParser::checkTargetMatchPredicate(MCInst &Inst) {
   // It also applies for registers Rt and Rs of microMIPSr6 jalrc.hb instruction
   // and registers Rd and Base for microMIPS lwp instruction
   case Mips::JALR_HB:
+  case Mips::JALR_HB64:
   case Mips::JALRC_HB_MMR6:
   case Mips::JALRC_MMR6:
     if (Inst.getOperand(0).getReg() == Inst.getOperand(1).getReg())

lib/Target/Mips/MicroMips32r6InstrInfo.td

@@ -1886,6 +1886,12 @@ let AddedComplexity = 41 in {
 
 def TAILCALL_MMR6 : TailCall<BC_MMR6, brtarget26_mm>, ISA_MICROMIPS32R6;
 
+def TAILCALLREG_MMR6  : TailCallReg<JRC16_MM, GPR32Opnd>, ISA_MICROMIPS32R6;
+
+def PseudoIndirectBranch_MMR6 : PseudoIndirectBranchBase<JRC16_MMR6,
+                                                         GPR32Opnd>,
+                                ISA_MICROMIPS32R6;
+
 def : MipsPat<(MipsTailCall (iPTR tglobaladdr:$dst)),
               (TAILCALL_MMR6 tglobaladdr:$dst)>, ISA_MICROMIPS32R6;
 

lib/Target/Mips/MicroMipsInstrInfo.td

@@ -1003,6 +1003,12 @@ let DecoderNamespace = "MicroMips", Predicates = [InMicroMips] in {
 
 def TAILCALL_MM : TailCall<J_MM, jmptarget_mm>, ISA_MIPS1_NOT_32R6_64R6;
 
+def TAILCALLREG_MM  : TailCallReg<JRC16_MM, GPR32Opnd>,
+                      ISA_MICROMIPS32_NOT_MIPS32R6;
+
+def PseudoIndirectBranch_MM : PseudoIndirectBranchBase<JR_MM, GPR32Opnd>,
+                              ISA_MICROMIPS32_NOT_MIPS32R6;
+
 let DecoderNamespace = "MicroMips" in {
   def RDHWR_MM : MMRel, R6MMR6Rel, ReadHardware<GPR32Opnd, HWRegsOpnd>,
                  RDHWR_FM_MM, ISA_MICROMIPS32_NOT_MIPS32R6;

lib/Target/Mips/Mips.td

@@ -193,6 +193,10 @@ def FeatureMT : SubtargetFeature<"mt", "HasMT", "true", "Mips MT ASE">;
 def FeatureLongCalls : SubtargetFeature<"long-calls", "UseLongCalls", "true",
                                         "Disable use of the jal instruction">;
 
+def FeatureUseIndirectJumpsHazard : SubtargetFeature<"use-indirect-jump-hazard",
+                                                    "UseIndirectJumpsHazard",
+                                                    "true", "Use indirect jump"
+                        " guards to prevent certain speculation based attacks">;
 //===----------------------------------------------------------------------===//
 // Mips processors supported.
 //===----------------------------------------------------------------------===//

lib/Target/Mips/Mips32r6InstrInfo.td

@@ -1036,3 +1036,42 @@ def : MipsPat<(select i32:$cond, immz, i32:$f),
               (SELEQZ i32:$f, i32:$cond)>,
               ISA_MIPS32R6;
 }
+
+// Pseudo instructions
+let isCall = 1, isTerminator = 1, isReturn = 1, isBarrier = 1, hasDelaySlot = 1,
+    hasExtraSrcRegAllocReq = 1, isCTI = 1, Defs = [AT] in {
+  class TailCallRegR6<Instruction JumpInst, Register RT, RegisterOperand RO> :
+    PseudoSE<(outs), (ins RO:$rs), [(MipsTailCall RO:$rs)], II_JR>,
+    PseudoInstExpansion<(JumpInst RT:$rt, RO:$rs)>;
+}
+
+class PseudoIndirectBranchBaseR6<Instruction JumpInst, Register RT,
+                                 RegisterOperand RO> :
+    MipsPseudo<(outs), (ins RO:$rs), [(brind RO:$rs)],
+               II_IndirectBranchPseudo>,
+    PseudoInstExpansion<(JumpInst RT:$rt, RO:$rs)> {
+  let isTerminator=1;
+  let isBarrier=1;
+  let hasDelaySlot = 1;
+  let isBranch = 1;
+  let isIndirectBranch = 1;
+  bit isCTI = 1;
+}
+
+
+let AdditionalPredicates = [NotInMips16Mode, NotInMicroMips,
+                            NoIndirectJumpGuards] in {
+  def TAILCALLR6REG : TailCallRegR6<JALR, ZERO, GPR32Opnd>, ISA_MIPS32R6;
+  def PseudoIndirectBranchR6 : PseudoIndirectBranchBaseR6<JALR, ZERO,
+                                                          GPR32Opnd>,
+                               ISA_MIPS32R6;
+}
+
+let AdditionalPredicates = [NotInMips16Mode, NotInMicroMips,
+                            UseIndirectJumpsHazard] in {
+  def TAILCALLHBR6REG : TailCallReg<JR_HB_R6, GPR32Opnd>, ISA_MIPS32R6;
+  def PseudoIndrectHazardBranchR6 : PseudoIndirectBranchBase<JR_HB_R6,
+                                                             GPR32Opnd>,
+                                    ISA_MIPS32R6;
+}
+