[llvm-exegesis] Generate snippet setup code.

[courbet]

Summary

This ensures that the snippet always sees the same values for registers,
making measurements reproducible.
This will also allow exploring different values.

[courbet]

rebase

[courbet]

rebase

[courbet]

Simplify: VR128X (xmm0-31) includes VR128 (xmm0-8) and FR(32|64|128)X? (low subword of xmm).

[courbet]

Y/ZMM loads require a different opcode.

[gchatelet]

tools/llvm-exegesis/lib/BenchmarkRunner.cpp

142 ↗

(On Diff #152667)

"a mov" -> "an assignment"

unittests/tools/llvm-exegesis/X86/SnippetGeneratorTest.cpp

253 ↗

(On Diff #152667)

remove

unittests/tools/llvm-exegesis/X86/TargetTest.cpp

41 ↗

(On Diff #152667)

SizeIs(Gt(0)) ?

[courbet]

address review comments

[pcc]

llvm/trunk/unittests/tools/llvm-exegesis/X86/TargetTest.cpp

40

Looks like this line has caused a build warning:

utils/unittest/googlemock/include/gmock/gmock-matchers.h:216:60: warning: comparison of integers of different signs: 'const unsigned long' and 'const int' [-Wsign-compare]
  bool operator()(const A& a, const B& b) const { return a > b; }
                                                         ~ ^ ~
[...]
unittests/tools/llvm-exegesis/X86/TargetTest.cpp:40:3: note: in instantiation of function template specialization 'testing::internal::PredicateFormatterFromMatcher<testing::internal::SizeIsMatcher<testing::internal::GtMatcher<int> > >::operator()<std::vector<llvm::MCInst, std::allocator<llvm::MCInst> > >' requested here
  EXPECT_THAT(Insts, SizeIs(Gt(0)));
  ^

[courbet]

llvm/trunk/unittests/tools/llvm-exegesis/X86/TargetTest.cpp

40

Sorry about that. This was fixed in r335547.

llvm/trunk/tools/llvm-exegesis/lib/Assembler.h

 #include "llvm/MC/MCInst.h"
 #include "llvm/Object/Binary.h"
 #include "llvm/Object/ObjectFile.h"
 #include "llvm/Support/raw_ostream.h"
 #include "llvm/Target/TargetMachine.h"
 
 namespace exegesis {
 
+class ExegesisTarget;
+
 // Gather the set of reserved registers (depends on function's calling
 // convention and target machine).
 llvm::BitVector getFunctionReservedRegs(const llvm::TargetMachine &TM);
 
 // Creates a temporary `void foo()` function containing the provided
 // Instructions. Runs a set of llvm Passes to provide correct prologue and
 // epilogue. Once the MachineFunction is ready, it is assembled for TM to
 // AsmStream, the temporary function is eventually discarded.
-void assembleToStream(std::unique_ptr<llvm::LLVMTargetMachine> TM,
+void assembleToStream(const ExegesisTarget *ET,
+                      std::unique_ptr<llvm::LLVMTargetMachine> TM,
+                      llvm::ArrayRef<unsigned> RegsToDef,
                       llvm::ArrayRef<llvm::MCInst> Instructions,
                       llvm::raw_pwrite_stream &AsmStream);
 
 // Creates an ObjectFile in the format understood by the host.
 // Note: the resulting object keeps a copy of Buffer so it can be discarded once
 // this function returns.
 llvm::object::OwningBinary<llvm::object::ObjectFile>
 getObjectFromBuffer(llvm::StringRef Buffer);

llvm/trunk/tools/llvm-exegesis/lib/Assembler.cpp

 #include "llvm/MC/MCInstrInfo.h"
 #include "llvm/Support/MemoryBuffer.h"
 
 namespace exegesis {
 
 static constexpr const char ModuleID[] = "ExegesisInfoTest";
 static constexpr const char FunctionID[] = "foo";
 
+static std::vector<llvm::MCInst>
+generateSnippetSetupCode(const llvm::ArrayRef<unsigned> RegsToDef,
+                         const ExegesisTarget &ET, bool &IsComplete) {
+  IsComplete = true;
+  std::vector<llvm::MCInst> Result;
+  for (const unsigned Reg : RegsToDef) {
+    // Load a constant in the register.
+    const auto Code = ET.setRegToConstant(Reg);
+    if (Code.empty())
+      IsComplete = false;
+    Result.insert(Result.end(), Code.begin(), Code.end());
+  }
+  return Result;
+}
+
 // Small utility function to add named passes.
 static bool addPass(llvm::PassManagerBase &PM, llvm::StringRef PassName,
                     llvm::TargetPassConfig &TPC) {
   const llvm::PassRegistry *PR = llvm::PassRegistry::getPassRegistry();
   const llvm::PassInfo *PI = PR->getPassInfo(PassName);
   if (!PI) {
     llvm::errs() << " run-pass " << PassName << " is not registered.\n";
     return true;
llvm::BitVector getFunctionReservedRegs(const llvm::TargetMachine &TM) {
   std::unique_ptr<llvm::MachineModuleInfo> MMI =
       llvm::make_unique<llvm::MachineModuleInfo>(&TM);
   llvm::MachineFunction &MF =
       createVoidVoidMachineFunction(FunctionID, Module.get(), MMI.get());
   // Saving reserved registers for client.
   return MF.getSubtarget().getRegisterInfo()->getReservedRegs(MF);
 }
 
-void assembleToStream(std::unique_ptr<llvm::LLVMTargetMachine> TM,
+void assembleToStream(const ExegesisTarget *ET,
+                      std::unique_ptr<llvm::LLVMTargetMachine> TM,
+                      llvm::ArrayRef<unsigned> RegsToDef,
                       llvm::ArrayRef<llvm::MCInst> Instructions,
                       llvm::raw_pwrite_stream &AsmStream) {
   std::unique_ptr<llvm::LLVMContext> Context =
       llvm::make_unique<llvm::LLVMContext>();
   std::unique_ptr<llvm::Module> Module =
       createModule(Context, TM->createDataLayout());
   std::unique_ptr<llvm::MachineModuleInfo> MMI =
       llvm::make_unique<llvm::MachineModuleInfo>(TM.get());
   llvm::MachineFunction &MF =
       createVoidVoidMachineFunction(FunctionID, Module.get(), MMI.get());
 
   // We need to instruct the passes that we're done with SSA and virtual
   // registers.
   auto &Properties = MF.getProperties();
   Properties.set(llvm::MachineFunctionProperties::Property::NoVRegs);
   Properties.reset(llvm::MachineFunctionProperties::Property::IsSSA);
-  // FIXME: Remove this when we assign all used registers as config step. This
-  // essentially disables checks that used registers are def'ed somewhere.
+  std::vector<llvm::MCInst> SnippetWithSetup;
+  bool IsSnippetSetupComplete;
+  if (ET) {
+    SnippetWithSetup =
+        generateSnippetSetupCode(RegsToDef, *ET, IsSnippetSetupComplete);
+    SnippetWithSetup.insert(SnippetWithSetup.end(), Instructions.begin(),
+                            Instructions.end());
+    Instructions = SnippetWithSetup;
+  }
+  // If the snippet setup is not complete, we disable liveliness tracking. This
+  // means that we won't know what values are in the registers.
+  if (!IsSnippetSetupComplete)
     Properties.reset(llvm::MachineFunctionProperties::Property::TracksLiveness);
+
   // prologue/epilogue pass needs the reserved registers to be frozen, this
   // is usually done by the SelectionDAGISel pass.
   MF.getRegInfo().freezeReservedRegs(MF);
 
   // Fill the MachineFunction from the instructions.
   fillMachineFunction(MF, Instructions);
 
   // We create the pass manager, run the passes to populate AsmBuffer.
   llvm::MCContext &MCContext = MMI->getContext();
   llvm::legacy::PassManager PM;
 
   llvm::TargetLibraryInfoImpl TLII(llvm::Triple(Module->getTargetTriple()));
   PM.add(new llvm::TargetLibraryInfoWrapperPass(TLII));
 
   llvm::TargetPassConfig *TPC = TM->createPassConfig(PM);
   PM.add(TPC);
   PM.add(MMI.release());
   TPC->printAndVerify("MachineFunctionGenerator::assemble");
   // Add target-specific passes.
-  if (const auto *ET = ExegesisTarget::lookup(TM->getTargetTriple())) {
+  if (ET) {
     ET->addTargetSpecificPasses(PM);
     TPC->printAndVerify("After ExegesisTarget::addTargetSpecificPasses");
   }
   // Adding the following passes:
   // - machineverifier: checks that the MachineFunction is well formed.
   // - prologepilog: saves and restore callee saved registers.
   for (const char *PassName : {"machineverifier", "prologepilog"})
     if (addPass(PM, PassName, *TPC))

llvm/trunk/tools/llvm-exegesis/lib/BenchmarkRunner.h

public:
   BenchmarkFailure(const llvm::Twine &S);
 };
 
 // A collection of instructions that are to be assembled, executed and measured.
 struct BenchmarkConfiguration {
   // This code is run before the Snippet is iterated. Since it is part of the
   // measurement it should be as short as possible. It is usually used to setup
   // the content of the Registers.
-  std::vector<llvm::MCInst> SnippetSetup;
+  struct Setup {
+    std::vector<unsigned> RegsToDef;
+  };
+  Setup SnippetSetup;
 
   // The sequence of instructions that are to be repeated.
   std::vector<llvm::MCInst> Snippet;
 
   // Informations about how this configuration was built.
   std::string Info;
 };
 
public:
   };
 
   virtual ~BenchmarkRunner();
 
   llvm::Expected<std::vector<InstructionBenchmark>>
   run(unsigned Opcode, const InstructionFilter &Filter,
       unsigned NumRepetitions);
 
+  // Given a snippet, computes which registers the setup code needs to define.
+  std::vector<unsigned>
+  computeRegsToDef(const std::vector<InstructionInstance> &Snippet) const;
+
 protected:
   const LLVMState &State;
   const llvm::MCInstrInfo &MCInstrInfo;
   const llvm::MCRegisterInfo &MCRegisterInfo;
   const RegisterAliasingTrackerCache RATC;
 
 private:
   InstructionBenchmark runOne(const BenchmarkConfiguration &Configuration,
private:
   virtual llvm::Expected<SnippetPrototype>
   generatePrototype(unsigned Opcode) const = 0;
 
   virtual std::vector<BenchmarkMeasure>
   runMeasurements(const ExecutableFunction &EF,
                   const unsigned NumRepetitions) const = 0;
 
   llvm::Expected<std::string>
-  writeObjectFile(llvm::ArrayRef<llvm::MCInst> Code) const;
-  llvm::Expected<ExecutableFunction>
-  createExecutableFunction(llvm::ArrayRef<llvm::MCInst> Code) const;
+  writeObjectFile(const BenchmarkConfiguration::Setup &Setup,
+                  llvm::ArrayRef<llvm::MCInst> Code) const;
 };
 
 } // namespace exegesis
 
 #endif // LLVM_TOOLS_LLVM_EXEGESIS_BENCHMARKRUNNER_H

llvm/trunk/tools/llvm-exegesis/lib/BenchmarkRunner.cpp

if (Snippet.empty()) {
     InstrBenchmark.Error = "Empty snippet";
     return InstrBenchmark;
   }
 
   InstrBenchmark.Key.Instructions = Snippet;
 
   // Repeat the snippet until there are at least NumInstructions in the
   // resulting code. The snippet is always repeated at least once.
-  const auto GenerateInstructions = [&Snippet](const int MinInstructions) {
-    std::vector<llvm::MCInst> Code = Snippet;
+  const auto GenerateInstructions = [&Configuration](
+                                        const int MinInstructions) {
+    std::vector<llvm::MCInst> Code = Configuration.Snippet;
     for (int I = 0; I < MinInstructions; ++I)
-      Code.push_back(Snippet[I % Snippet.size()]);
+      Code.push_back(Configuration.Snippet[I % Configuration.Snippet.size()]);
     return Code;
   };
 
   // Assemble at least kMinInstructionsForSnippet instructions by repeating the
   // snippet for debug/analysis. This is so that the user clearly understands
   // that the inside instructions are repeated.
   constexpr const int kMinInstructionsForSnippet = 16;
   {
     auto ObjectFilePath =
-        writeObjectFile(GenerateInstructions(kMinInstructionsForSnippet));
+        writeObjectFile(Configuration.SnippetSetup,
+                        GenerateInstructions(kMinInstructionsForSnippet));
     if (llvm::Error E = ObjectFilePath.takeError()) {
       InstrBenchmark.Error = llvm::toString(std::move(E));
       return InstrBenchmark;
     }
     const ExecutableFunction EF(State.createTargetMachine(),
                                 getObjectFromFile(*ObjectFilePath));
     const auto FnBytes = EF.getFunctionBytes();
     InstrBenchmark.AssembledSnippet.assign(FnBytes.begin(), FnBytes.end());
   }
 
   // Assemble NumRepetitions instructions repetitions of the snippet for
   // measurements.
   auto ObjectFilePath =
-      writeObjectFile(GenerateInstructions(InstrBenchmark.NumRepetitions));
+      writeObjectFile(Configuration.SnippetSetup,
+                      GenerateInstructions(InstrBenchmark.NumRepetitions));
   if (llvm::Error E = ObjectFilePath.takeError()) {
     InstrBenchmark.Error = llvm::toString(std::move(E));
     return InstrBenchmark;
   }
   llvm::outs() << "Check generated assembly with: /usr/bin/objdump -d "
                << *ObjectFilePath << "\n";
   const ExecutableFunction EF(State.createTargetMachine(),
                               getObjectFromFile(*ObjectFilePath));
   InstrBenchmark.Measurements = runMeasurements(EF, NumRepetitions);
 
   return InstrBenchmark;
 }
 
 llvm::Expected<std::vector<BenchmarkConfiguration>>
 BenchmarkRunner::generateConfigurations(unsigned Opcode) const {
   if (auto E = generatePrototype(Opcode)) {
     SnippetPrototype &Prototype = E.get();
     // TODO: Generate as many configurations as needed here.
     BenchmarkConfiguration Configuration;
     Configuration.Info = Prototype.Explanation;
-    for (InstructionInstance &II : Prototype.Snippet)
-      Configuration.Snippet.push_back(II.randomizeUnsetVariablesAndBuild());
+    for (InstructionInstance &II : Prototype.Snippet) {
+      II.randomizeUnsetVariables();
+      Configuration.Snippet.push_back(II.build());
+    }
+    Configuration.SnippetSetup.RegsToDef = computeRegsToDef(Prototype.Snippet);
     return std::vector<BenchmarkConfiguration>{Configuration};
   } else
     return E.takeError();
 }
 
+std::vector<unsigned> BenchmarkRunner::computeRegsToDef(
+    const std::vector<InstructionInstance> &Snippet) const {
+  // Collect all register uses and create an assignment for each of them.
+  // Loop invariant: DefinedRegs[i] is true iif it has been set at least once
+  // before the current instruction.
+  llvm::BitVector DefinedRegs = RATC.emptyRegisters();
+  std::vector<unsigned> RegsToDef;
+  for (const InstructionInstance &II : Snippet) {
+    // Returns the register that this Operand sets or uses, or 0 if this is not
+    // a register.
+    const auto GetOpReg = [&II](const Operand &Op) -> unsigned {
+      if (Op.ImplicitReg) {
+        return *Op.ImplicitReg;
+      } else if (Op.IsExplicit && II.getValueFor(Op).isReg()) {
+        return II.getValueFor(Op).getReg();
+      }
+      return 0;
+    };
+    // Collect used registers that have never been def'ed.
+    for (const Operand &Op : II.Instr.Operands) {
+      if (!Op.IsDef) {
+        const unsigned Reg = GetOpReg(Op);
+        if (Reg > 0 && !DefinedRegs.test(Reg)) {
+          RegsToDef.push_back(Reg);
+          DefinedRegs.set(Reg);
+        }
+      }
+    }
+    // Mark defs as having been def'ed.
+    for (const Operand &Op : II.Instr.Operands) {
+      if (Op.IsDef) {
+        const unsigned Reg = GetOpReg(Op);
+        if (Reg > 0) {
+          DefinedRegs.set(Reg);
+        }
+      }
+    }
+  }
+  return RegsToDef;
+}
+
 llvm::Expected<std::string>
-BenchmarkRunner::writeObjectFile(llvm::ArrayRef<llvm::MCInst> Code) const {
+BenchmarkRunner::writeObjectFile(const BenchmarkConfiguration::Setup &Setup,
+                                 llvm::ArrayRef<llvm::MCInst> Code) const {
   int ResultFD = 0;
   llvm::SmallString<256> ResultPath;
   if (llvm::Error E = llvm::errorCodeToError(llvm::sys::fs::createTemporaryFile(
           "snippet", "o", ResultFD, ResultPath)))
     return std::move(E);
   llvm::raw_fd_ostream OFS(ResultFD, true /*ShouldClose*/);
-  assembleToStream(State.createTargetMachine(), Code, OFS);
+  assembleToStream(State.getExegesisTarget(), State.createTargetMachine(),
+                   Setup.RegsToDef, Code, OFS);
   return ResultPath.str();
 }
 
 } // namespace exegesis

llvm/trunk/tools/llvm-exegesis/lib/LlvmState.h

 /// \file
 /// A class to set up and access common LLVM objects.
 ///
 //===----------------------------------------------------------------------===//
 
 #ifndef LLVM_TOOLS_LLVM_EXEGESIS_LLVMSTATE_H
 #define LLVM_TOOLS_LLVM_EXEGESIS_LLVMSTATE_H
 
+#include "Target.h"
 #include "llvm/MC/MCAsmInfo.h"
 #include "llvm/MC/MCInst.h"
 #include "llvm/MC/MCInstrInfo.h"
 #include "llvm/MC/MCRegisterInfo.h"
 #include "llvm/MC/MCSubtargetInfo.h"
 #include "llvm/Target/TargetMachine.h"
 #include <memory>
 #include <string>

llvm/trunk/tools/llvm-exegesis/lib/LlvmState.cpp

LLVMState::LLVMState(const std::string &Triple, const std::string &CpuName) {
   std::string Error;
   const llvm::Target *const TheTarget =
       llvm::TargetRegistry::lookupTarget(Triple, Error);
   assert(TheTarget && "unknown target for host");
   const llvm::TargetOptions Options;
   TargetMachine.reset(static_cast<llvm::LLVMTargetMachine *>(
       TheTarget->createTargetMachine(Triple, CpuName, /*Features*/ "", Options,
                                      llvm::Reloc::Model::Static)));
+  TheExegesisTarget = ExegesisTarget::lookup(TargetMachine->getTargetTriple());
 }
 
 LLVMState::LLVMState()
     : LLVMState(llvm::sys::getProcessTriple(),
                 llvm::sys::getHostCPUName().str()) {}
 
 std::unique_ptr<llvm::LLVMTargetMachine>
 LLVMState::createTargetMachine() const {

llvm/trunk/tools/llvm-exegesis/lib/MCInstrDescView.h

struct InstructionInstance {
   InstructionInstance &operator=(const InstructionInstance &) = delete;
 
   // Moving is OK.
   InstructionInstance(InstructionInstance &&);
   InstructionInstance &operator=(InstructionInstance &&);
 
   unsigned getOpcode() const;
   llvm::MCOperand &getValueFor(const Variable &Var);
+  const llvm::MCOperand &getValueFor(const Variable &Var) const;
   llvm::MCOperand &getValueFor(const Operand &Op);
+  const llvm::MCOperand &getValueFor(const Operand &Op) const;
   bool hasImmediateVariables() const;
 
-  // Assigns a Random Value to all Variables that are still Invalid and returns
-  // the instance as an llvm::MCInst.
-  llvm::MCInst randomizeUnsetVariablesAndBuild();
+  // Assigns a Random Value to all Variables that are still Invalid.
+  void randomizeUnsetVariables();
+
+  // Returns the instance as an llvm::MCInst. The InstructionInstance must be
+  // fully allocated (no invalid variables).
+  llvm::MCInst build() const;
 
   Instruction Instr;
   llvm::SmallVector<llvm::MCOperand, 4> VariableValues;
 };
 
 // A prototype is a set of InstructionInstances with an explanation of how
 // it's been built. The prototype can then be randomized to exercice several
 // immediate values. It is also used to gather the used registers and define

llvm/trunk/tools/llvm-exegesis/lib/MCInstrDescView.cpp

 unsigned InstructionInstance::getOpcode() const {
   return Instr.Description->getOpcode();
 }
 
 llvm::MCOperand &InstructionInstance::getValueFor(const Variable &Var) {
   return VariableValues[Var.Index];
 }
 
+const llvm::MCOperand &
+InstructionInstance::getValueFor(const Variable &Var) const {
+  return VariableValues[Var.Index];
+}
+
 llvm::MCOperand &InstructionInstance::getValueFor(const Operand &Op) {
   assert(Op.VariableIndex >= 0);
   return getValueFor(Instr.Variables[Op.VariableIndex]);
 }
 
+const llvm::MCOperand &
+InstructionInstance::getValueFor(const Operand &Op) const {
+  assert(Op.VariableIndex >= 0);
+  return getValueFor(Instr.Variables[Op.VariableIndex]);
+}
+
 // forward declaration.
 static void randomize(const Instruction &Instr, const Variable &Var,
                       llvm::MCOperand &AssignedValue);
 
 bool InstructionInstance::hasImmediateVariables() const {
   return llvm::any_of(Instr.Variables, [this](const Variable &Var) {
     assert(!Var.TiedOperands.empty());
     const unsigned OpIndex = Var.TiedOperands[0];
     const Operand &Op = Instr.Operands[OpIndex];
     assert(Op.Info);
     return Op.Info->OperandType == llvm::MCOI::OPERAND_IMMEDIATE;
   });
 }
 
-llvm::MCInst InstructionInstance::randomizeUnsetVariablesAndBuild() {
+void InstructionInstance::randomizeUnsetVariables() {
   for (const Variable &Var : Instr.Variables) {
     llvm::MCOperand &AssignedValue = getValueFor(Var);
     if (!AssignedValue.isValid())
       randomize(Instr, Var, AssignedValue);
   }
+}
+
+llvm::MCInst InstructionInstance::build() const {
   llvm::MCInst Result;
   Result.setOpcode(Instr.Description->Opcode);
   for (const auto &Op : Instr.Operands)
     if (Op.IsExplicit)
       Result.addOperand(getValueFor(Op));
   return Result;
 }
 

llvm/trunk/tools/llvm-exegesis/lib/Target.h

 //===----------------------------------------------------------------------===//
 
 #ifndef LLVM_TOOLS_LLVM_EXEGESIS_TARGET_H
 #define LLVM_TOOLS_LLVM_EXEGESIS_TARGET_H
 
 #include "llvm/ADT/Triple.h"
 #include "llvm/CodeGen/TargetPassConfig.h"
 #include "llvm/IR/LegacyPassManager.h"
+#include "llvm/MC/MCInst.h"
+#include "llvm/MC/MCRegisterInfo.h"
 
 namespace exegesis {
 
 class ExegesisTarget {
 public:
   // Targets can use this to add target-specific passes in assembleToStream();
   virtual void addTargetSpecificPasses(llvm::PassManagerBase &PM) const {}
 
+  // Generates code to move a constant into a the given register.
+  virtual std::vector<llvm::MCInst> setRegToConstant(unsigned Reg) const {
+    return {};
+  }
+
   // Returns the ExegesisTarget for the given triple or nullptr if the target
   // does not exist.
   static const ExegesisTarget *lookup(llvm::Triple TT);
   // Registers a target. Not thread safe.
   static void registerTarget(ExegesisTarget *T);
 
   virtual ~ExegesisTarget();
 
 private:
   virtual bool matchesArch(llvm::Triple::ArchType Arch) const = 0;
   const ExegesisTarget *Next = nullptr;
 };
 
 } // namespace exegesis
 
 #endif // LLVM_TOOLS_LLVM_EXEGESIS_TARGET_H

llvm/trunk/tools/llvm-exegesis/lib/X86/Target.cpp

 //===-- Target.cpp ----------------------------------------------*- C++ -*-===//
 //
 //                     The LLVM Compiler Infrastructure
 //
 // This file is distributed under the University of Illinois Open Source
 // License. See LICENSE.TXT for details.
 //
 //===----------------------------------------------------------------------===//
 #include "../Target.h"
 
+#include "MCTargetDesc/X86MCTargetDesc.h"
 #include "X86.h"
+#include "X86RegisterInfo.h"
+#include "llvm/MC/MCInstBuilder.h"
 
 namespace exegesis {
 
 namespace {
 
 class ExegesisX86Target : public ExegesisTarget {
   void addTargetSpecificPasses(llvm::PassManagerBase &PM) const override {
     // Lowers FP pseudo-instructions, e.g. ABS_Fp32 -> ABS_F.
     // FIXME: Enable when the exegesis assembler no longer does
     // Properties.reset(TracksLiveness);
     // PM.add(llvm::createX86FloatingPointStackifierPass());
   }
 
+  std::vector<llvm::MCInst>
+  setRegToConstant(const unsigned Reg) const override {
+    // FIXME: Handle FP stack:
+    // llvm::X86::RFP32RegClass
+    // llvm::X86::RFP64RegClass
+    // llvm::X86::RFP80RegClass
+    if (llvm::X86::GR8RegClass.contains(Reg)) {
+      return {llvm::MCInstBuilder(llvm::X86::MOV8ri).addReg(Reg).addImm(1)};
+    }
+    if (llvm::X86::GR16RegClass.contains(Reg)) {
+      return {llvm::MCInstBuilder(llvm::X86::MOV16ri).addReg(Reg).addImm(1)};
+    }
+    if (llvm::X86::GR32RegClass.contains(Reg)) {
+      return {llvm::MCInstBuilder(llvm::X86::MOV32ri).addReg(Reg).addImm(1)};
+    }
+    if (llvm::X86::GR64RegClass.contains(Reg)) {
+      return {llvm::MCInstBuilder(llvm::X86::MOV64ri32).addReg(Reg).addImm(1)};
+    }
+    if (llvm::X86::VR128XRegClass.contains(Reg)) {
+      return setVectorRegToConstant(Reg, 16, llvm::X86::VMOVDQUrm);
+    }
+    if (llvm::X86::VR256XRegClass.contains(Reg)) {
+      return setVectorRegToConstant(Reg, 32, llvm::X86::VMOVDQUYrm);
+    }
+    if (llvm::X86::VR512RegClass.contains(Reg)) {
+      return setVectorRegToConstant(Reg, 64, llvm::X86::VMOVDQU64Zrm);
+    }
+    return {};
+  }
+
   bool matchesArch(llvm::Triple::ArchType Arch) const override {
     return Arch == llvm::Triple::x86_64 || Arch == llvm::Triple::x86;
   }
+
+private:
+  // setRegToConstant() specialized for a vector register of size
+  // `RegSizeBytes`. `RMOpcode` is the opcode used to do a memory -> vector
+  // register load.
+  static std::vector<llvm::MCInst>
+  setVectorRegToConstant(const unsigned Reg, const unsigned RegSizeBytes,
+                         const unsigned RMOpcode) {
+    // There is no instruction to directly set XMM, go through memory.
+    // Since vector values can be interpreted as integers of various sizes (8
+    // to 64 bits) as well as floats and double, so we chose an immediate
+    // value that has set bits for all byte values and is a normal float/
+    // double. 0x40404040 is ~32.5 when interpreted as a double and ~3.0f when
+    // interpreted as a float.
+    constexpr const uint64_t kImmValue = 0x40404040ull;
+    std::vector<llvm::MCInst> Result;
+    // Allocate scratch memory on the stack.
+    Result.push_back(llvm::MCInstBuilder(llvm::X86::SUB64ri8)
+                         .addReg(llvm::X86::RSP)
+                         .addReg(llvm::X86::RSP)
+                         .addImm(RegSizeBytes));
+    // Fill scratch memory.
+    for (unsigned Disp = 0; Disp < RegSizeBytes; Disp += 4) {
+      Result.push_back(llvm::MCInstBuilder(llvm::X86::MOV32mi)
+                           // Address = ESP
+                           .addReg(llvm::X86::RSP) // BaseReg
+                           .addImm(1)              // ScaleAmt
+                           .addReg(0)              // IndexReg
+                           .addImm(Disp)           // Disp
+                           .addReg(0)              // Segment
+                           // Immediate.
+                           .addImm(kImmValue));
+    }
+    // Load Reg from scratch memory.
+    Result.push_back(llvm::MCInstBuilder(RMOpcode)
+                         .addReg(Reg)
+                         // Address = ESP
+                         .addReg(llvm::X86::RSP) // BaseReg
+                         .addImm(1)              // ScaleAmt
+                         .addReg(0)              // IndexReg
+                         .addImm(0)              // Disp
+                         .addReg(0));            // Segment
+    // Release scratch memory.
+    Result.push_back(llvm::MCInstBuilder(llvm::X86::ADD64ri8)
+                         .addReg(llvm::X86::RSP)
+                         .addReg(llvm::X86::RSP)
+                         .addImm(RegSizeBytes));
+    return Result;
+  }
 };
 
 } // namespace
 
 static ExegesisTarget *getTheExegesisX86Target() {
   static ExegesisX86Target Target;
   return &Target;
 }
 
 void InitializeX86ExegesisTarget() {
   ExegesisTarget::registerTarget(getTheExegesisX86Target());
 }
 
 } // namespace exegesis

llvm/trunk/unittests/tools/llvm-exegesis/Common/AssemblerUtils.h

MachineFunctionGeneratorBaseTest(const std::string &TT,
     if (!CanExecute) {
       llvm::outs() << "Skipping execution, host:"
                    << llvm::sys::getProcessTriple() << ", target:" << TT
                    << "\n";
     }
   }
 
   template <class... Bs> inline void Check(llvm::MCInst MCInst, Bs... Bytes) {
-    ExecutableFunction Function = (MCInst.getOpcode() == 0)
-                                      ? assembleToFunction({})
-                                      : assembleToFunction({MCInst});
+    CheckWithSetup(nullptr, {}, MCInst, Bytes...);
+  }
+
+  template <class... Bs>
+  inline void CheckWithSetup(const ExegesisTarget *ET,
+                             llvm::ArrayRef<unsigned> RegsToDef,
+                             llvm::MCInst MCInst, Bs... Bytes) {
+    ExecutableFunction Function =
+        (MCInst.getOpcode() == 0) ? assembleToFunction(ET, RegsToDef, {})
+                                  : assembleToFunction(ET, RegsToDef, {MCInst});
     ASSERT_THAT(Function.getFunctionBytes().str(),
                 testing::ElementsAre(Bytes...));
     if (CanExecute)
       Function();
   }
 
 private:
   std::unique_ptr<llvm::LLVMTargetMachine> createTargetMachine() {
     std::string Error;
     const llvm::Target *TheTarget =
         llvm::TargetRegistry::lookupTarget(TT, Error);
     EXPECT_TRUE(TheTarget) << Error << " " << TT;
     const llvm::TargetOptions Options;
     llvm::TargetMachine *TM = TheTarget->createTargetMachine(
         TT, CpuName, "", Options, llvm::Reloc::Model::Static);
     EXPECT_TRUE(TM) << TT << " " << CpuName;
     return std::unique_ptr<llvm::LLVMTargetMachine>(
         static_cast<llvm::LLVMTargetMachine *>(TM));
   }
 
   ExecutableFunction
-  assembleToFunction(llvm::ArrayRef<llvm::MCInst> Instructions) {
+  assembleToFunction(const ExegesisTarget *ET,
+                     llvm::ArrayRef<unsigned> RegsToDef,
+                     llvm::ArrayRef<llvm::MCInst> Instructions) {
     llvm::SmallString<256> Buffer;
     llvm::raw_svector_ostream AsmStream(Buffer);
-    assembleToStream(createTargetMachine(), Instructions, AsmStream);
+    assembleToStream(ET, createTargetMachine(), RegsToDef, Instructions,
+                     AsmStream);
     return ExecutableFunction(createTargetMachine(),
                               getObjectFromBuffer(AsmStream.str()));
   }
 
   const std::string TT;
   const std::string CpuName;
   const bool CanExecute;
 };
 
 } // namespace exegesis

llvm/trunk/unittests/tools/llvm-exegesis/X86/SnippetGeneratorTest.cpp

 namespace exegesis {
 namespace {
 
 using testing::AnyOf;
 using testing::ElementsAre;
 using testing::HasSubstr;
 using testing::Not;
 using testing::SizeIs;
+using testing::UnorderedElementsAre;
 
 MATCHER(IsInvalid, "") { return !arg.isValid(); }
 MATCHER(IsReg, "") { return arg.isReg(); }
 
 class X86SnippetGeneratorTest : public ::testing::Test {
 protected:
   X86SnippetGeneratorTest()
       : State("x86_64-unknown-linux", "haswell"),
TEST_F(UopsSnippetGeneratorTest, NoTiedVariables) {
   ASSERT_THAT(II.VariableValues, SizeIs(4));
   EXPECT_THAT(II.VariableValues[0].getReg(), Not(II.VariableValues[1].getReg()))
       << "Def is different from first Use";
   EXPECT_THAT(II.VariableValues[0].getReg(), Not(II.VariableValues[2].getReg()))
       << "Def is different from second Use";
   EXPECT_THAT(II.VariableValues[3], IsInvalid());
 }
 
+class FakeBenchmarkRunner : public BenchmarkRunner {
+public:
+  using BenchmarkRunner::BenchmarkRunner;
+
+  Instruction createInstruction(unsigned Opcode) {
+    return Instruction(MCInstrInfo.get(Opcode), RATC);
+  }
+
+private:
+  InstructionBenchmark::ModeE getMode() const override {
+    return InstructionBenchmark::Unknown;
+  }
+
+  llvm::Expected<SnippetPrototype>
+  generatePrototype(unsigned Opcode) const override {
+    return llvm::make_error<llvm::StringError>("not implemented",
+                                               llvm::inconvertibleErrorCode());
+  }
+
+  std::vector<BenchmarkMeasure>
+  runMeasurements(const ExecutableFunction &EF,
+                  const unsigned NumRepetitions) const override {
+    return {};
+  }
+};
+
+using FakeSnippetGeneratorTest = SnippetGeneratorTest<FakeBenchmarkRunner>;
+
+TEST_F(FakeSnippetGeneratorTest, ComputeRegsToDefAdd16ri) {
+  // ADD16ri:
+  // explicit def 0       : reg RegClass=GR16
+  // explicit use 1       : reg RegClass=GR16 | TIED_TO:0
+  // explicit use 2       : imm
+  // implicit def         : EFLAGS
+  InstructionInstance II(Runner.createInstruction(llvm::X86::ADD16ri));
+  II.getValueFor(II.Instr.Variables[0]) =
+      llvm::MCOperand::createReg(llvm::X86::AX);
+  std::vector<InstructionInstance> Snippet;
+  Snippet.push_back(std::move(II));
+  const auto RegsToDef = Runner.computeRegsToDef(Snippet);
+  EXPECT_THAT(RegsToDef, UnorderedElementsAre(llvm::X86::AX));
+}
+
+TEST_F(FakeSnippetGeneratorTest, ComputeRegsToDefAdd64rr) {
+  // ADD64rr:
+  //  mov64ri rax, 42
+  //  add64rr rax, rax, rbx
+  // -> only rbx needs defining.
+  std::vector<InstructionInstance> Snippet;
+  {
+    InstructionInstance Mov(Runner.createInstruction(llvm::X86::MOV64ri));
+    Mov.getValueFor(Mov.Instr.Variables[0]) =
+        llvm::MCOperand::createReg(llvm::X86::RAX);
+    Mov.getValueFor(Mov.Instr.Variables[1]) = llvm::MCOperand::createImm(42);
+    Snippet.push_back(std::move(Mov));
+  }
+  {
+    InstructionInstance Add(Runner.createInstruction(llvm::X86::ADD64rr));
+    Add.getValueFor(Add.Instr.Variables[0]) =
+        llvm::MCOperand::createReg(llvm::X86::RAX);
+    Add.getValueFor(Add.Instr.Variables[1]) =
+        llvm::MCOperand::createReg(llvm::X86::RBX);
+    Snippet.push_back(std::move(Add));
+  }
+
+  const auto RegsToDef = Runner.computeRegsToDef(Snippet);
+  EXPECT_THAT(RegsToDef, UnorderedElementsAre(llvm::X86::RBX));
+}
+
 } // namespace
 } // namespace exegesis

llvm/trunk/unittests/tools/llvm-exegesis/X86/TargetTest.cpp

 #include "Target.h"
 
 #include <cassert>
 #include <memory>
 
+#include "MCTargetDesc/X86MCTargetDesc.h"
 #include "gmock/gmock.h"
 #include "gtest/gtest.h"
 
 namespace exegesis {
 
 void InitializeX86ExegesisTarget();
 
 namespace {
 
+using testing::Gt;
 using testing::NotNull;
+using testing::SizeIs;
 
 class X86TargetTest : public ::testing::Test {
 protected:
+  X86TargetTest()
+      : Target_(ExegesisTarget::lookup(llvm::Triple("x86_64-unknown-linux"))) {
+    EXPECT_THAT(Target_, NotNull());
+  }
   static void SetUpTestCase() { InitializeX86ExegesisTarget(); }
+
+  const ExegesisTarget *const Target_;
 };
 
-TEST_F(X86TargetTest, Lookup) {
-  EXPECT_THAT(ExegesisTarget::lookup(llvm::Triple("x86_64-unknown-linux")),
-              NotNull());
+TEST_F(X86TargetTest, SetRegToConstantGPR) {
+  const auto Insts = Target_->setRegToConstant(llvm::X86::EAX);
+  EXPECT_THAT(Insts, SizeIs(1));
+  EXPECT_EQ(Insts[0].getOpcode(), llvm::X86::MOV32ri);
+  EXPECT_EQ(Insts[0].getOperand(0).getReg(), llvm::X86::EAX);
+}
+
+TEST_F(X86TargetTest, SetRegToConstantXMM) {
+  const auto Insts = Target_->setRegToConstant(llvm::X86::XMM1);
+  EXPECT_THAT(Insts, SizeIs(Gt(0)));
 }
 
 } // namespace
 } // namespace exegesis