Split LLVMBasedICFG (#726)

* Out-source globals model

* Mode call-graph construction

* small refactor

* Cleanup resolver headers

* buildCallGraph speedup for non-OTF resolvers

* Move isVirtualCall into Resolver + some refactoring

* Move EntryFunctionUtils into phasar_llvm_controlflow to avoid circular library dependencies between phasar_llvm_db and phasar_llvm_utils

* Add more constness

Author: fabianbs96

include/phasar/PhasarLLVM/ControlFlow.h

@@ -10,6 +10,7 @@
 #ifndef PHASAR_PHASARLLVM_CONTROLFLOW_H
 #define PHASAR_PHASARLLVM_CONTROLFLOW_H
 
+#include "phasar/PhasarLLVM/ControlFlow/EntryFunctionUtils.h"
 #include "phasar/PhasarLLVM/ControlFlow/LLVMBasedBackwardCFG.h"
 #include "phasar/PhasarLLVM/ControlFlow/LLVMBasedBackwardICFG.h"
 #include "phasar/PhasarLLVM/ControlFlow/LLVMBasedCFG.h"

include/phasar/PhasarLLVM/ControlFlow/EntryFunctionUtils.h

@@ -0,0 +1,31 @@
+/******************************************************************************
+ * Copyright (c) 2024 Fabian Schiebel.
+ * All rights reserved. This program and the accompanying materials are made
+ * available under the terms of LICENSE.txt.
+ *
+ * Contributors:
+ *     Fabian Schiebel and others
+ *****************************************************************************/
+
+#ifndef PHASAR_PHASARLLVM_UTILS_ENTRYFUNCTIONUTILS_H
+#define PHASAR_PHASARLLVM_UTILS_ENTRYFUNCTIONUTILS_H
+
+#include "llvm/ADT/ArrayRef.h"
+#include "llvm/IR/Function.h"
+
+#include <string>
+#include <vector>
+
+namespace psr {
+class LLVMProjectIRDB;
+
+[[nodiscard]] std::vector<const llvm::Function *>
+getEntryFunctions(const LLVMProjectIRDB &IRDB,
+                  llvm::ArrayRef<std::string> EntryPoints);
+
+[[nodiscard]] std::vector<llvm::Function *>
+getEntryFunctionsMut(LLVMProjectIRDB &IRDB,
+                     llvm::ArrayRef<std::string> EntryPoints);
+} // namespace psr
+
+#endif // PHASAR_PHASARLLVM_UTILS_ENTRYFUNCTIONUTILS_H

include/phasar/PhasarLLVM/ControlFlow/GlobalCtorsDtorsModel.h

@@ -0,0 +1,45 @@
+/******************************************************************************
+ * Copyright (c) 2024 Fabian Schiebel.
+ * All rights reserved. This program and the accompanying materials are made
+ * available under the terms of LICENSE.txt.
+ *
+ * Contributors:
+ *     Fabian Schiebel and others
+ *****************************************************************************/
+
+#ifndef PHASAR_PHASARLLVM_CONTROLFLOW_GLOBALCTORSDTORSMODEL_H
+#define PHASAR_PHASARLLVM_CONTROLFLOW_GLOBALCTORSDTORSMODEL_H
+
+#include "llvm/ADT/ArrayRef.h"
+#include "llvm/IR/Function.h"
+
+namespace psr {
+class LLVMProjectIRDB;
+
+class GlobalCtorsDtorsModel {
+public:
+  static constexpr llvm::StringLiteral ModelName =
+      "__psrCRuntimeGlobalCtorsModel";
+
+  static constexpr llvm::StringLiteral DtorModelName =
+      "__psrCRuntimeGlobalDtorsModel";
+
+  static constexpr llvm::StringLiteral DtorsCallerName =
+      "__psrGlobalDtorsCaller";
+
+  static constexpr llvm::StringLiteral UserEntrySelectorName =
+      "__psrCRuntimeUserEntrySelector";
+
+  static llvm::Function *
+  buildModel(LLVMProjectIRDB &IRDB,
+             llvm::ArrayRef<llvm::Function *> UserEntryPoints);
+  static llvm::Function *
+  buildModel(LLVMProjectIRDB &IRDB,
+             llvm::ArrayRef<std::string> UserEntryPoints);
+
+  /// Returns true, if a function was generated by phasar.
+  [[nodiscard]] static bool isPhasarGenerated(const llvm::Function &F) noexcept;
+};
+} // namespace psr
+
+#endif // PHASAR_PHASARLLVM_CONTROLFLOW_GLOBALCTORSDTORSMODEL_H

include/phasar/PhasarLLVM/ControlFlow/LLVMBasedCallGraph.h

@@ -0,0 +1,25 @@
+/******************************************************************************
+ * Copyright (c) 2024 Fabian Schiebel.
+ * All rights reserved. This program and the accompanying materials are made
+ * available under the terms of LICENSE.txt.
+ *
+ * Contributors:
+ *     Fabian Schiebel and others
+ *****************************************************************************/
+
+#ifndef PHASAR_PHASARLLVM_CONTROLFLOW_LLVMBASEDCALLGRAPH_H
+#define PHASAR_PHASARLLVM_CONTROLFLOW_LLVMBASEDCALLGRAPH_H
+
+#include "phasar/ControlFlow/CallGraph.h"
+
+namespace llvm {
+class Instruction;
+class Function;
+} // namespace llvm
+
+namespace psr {
+using LLVMBasedCallGraph =
+    CallGraph<const llvm::Instruction *, const llvm::Function *>;
+} // namespace psr
+
+#endif // PHASAR_PHASARLLVM_CONTROLFLOW_LLVMBASEDCALLGRAPH_H

include/phasar/PhasarLLVM/ControlFlow/LLVMBasedCallGraphBuilder.h

@@ -0,0 +1,49 @@
+/******************************************************************************
+ * Copyright (c) 2024 Fabian Schiebel.
+ * All rights reserved. This program and the accompanying materials are made
+ * available under the terms of LICENSE.txt.
+ *
+ * Contributors:
+ *     Fabian Schiebel and others
+ *****************************************************************************/
+
+#ifndef PHASAR_PHASARLLVM_CONTROLFLOW_LLVMBASEDCALLGRAPHBUILDER_H
+#define PHASAR_PHASARLLVM_CONTROLFLOW_LLVMBASEDCALLGRAPHBUILDER_H
+
+#include "phasar/PhasarLLVM/ControlFlow/LLVMBasedCallGraph.h"
+#include "phasar/PhasarLLVM/Pointer/LLVMAliasInfo.h"
+#include "phasar/Utils/Soundness.h"
+
+namespace psr {
+class LLVMProjectIRDB;
+enum class CallGraphAnalysisType;
+class LLVMTypeHierarchy;
+class LLVMVFTableProvider;
+class Resolver;
+
+[[nodiscard]] LLVMBasedCallGraph
+buildLLVMBasedCallGraph(LLVMProjectIRDB &IRDB, CallGraphAnalysisType CGType,
+                        llvm::ArrayRef<const llvm::Function *> EntryPoints,
+                        LLVMTypeHierarchy &TH, LLVMVFTableProvider &VTP,
+                        LLVMAliasInfoRef PT = nullptr,
+                        Soundness S = Soundness::Soundy);
+
+[[nodiscard]] LLVMBasedCallGraph
+buildLLVMBasedCallGraph(const LLVMProjectIRDB &IRDB, Resolver &CGResolver,
+                        llvm::ArrayRef<const llvm::Function *> EntryPoints,
+                        Soundness S = Soundness::Soundy);
+
+[[nodiscard]] LLVMBasedCallGraph
+buildLLVMBasedCallGraph(LLVMProjectIRDB &IRDB, CallGraphAnalysisType CGType,
+                        llvm::ArrayRef<std::string> EntryPoints,
+                        LLVMTypeHierarchy &TH, LLVMVFTableProvider &VTP,
+                        LLVMAliasInfoRef PT = nullptr,
+                        Soundness S = Soundness::Soundy);
+
+[[nodiscard]] LLVMBasedCallGraph
+buildLLVMBasedCallGraph(const LLVMProjectIRDB &IRDB, Resolver &CGResolver,
+                        llvm::ArrayRef<std::string> EntryPoints,
+                        Soundness S = Soundness::Soundy);
+} // namespace psr
+
+#endif // PHASAR_PHASARLLVM_CONTROLFLOW_LLVMBASEDCALLGRAPHBUILDER_H

include/phasar/PhasarLLVM/ControlFlow/LLVMBasedICFG.h

@@ -20,10 +20,13 @@
 #include "phasar/ControlFlow/CallGraph.h"
 #include "phasar/ControlFlow/CallGraphAnalysisType.h"
 #include "phasar/ControlFlow/ICFGBase.h"
+#include "phasar/PhasarLLVM/ControlFlow/GlobalCtorsDtorsModel.h"
 #include "phasar/PhasarLLVM/ControlFlow/LLVMBasedCFG.h"
+#include "phasar/PhasarLLVM/ControlFlow/LLVMBasedCallGraph.h"
 #include "phasar/PhasarLLVM/ControlFlow/LLVMVFTableProvider.h"
 #include "phasar/PhasarLLVM/Pointer/LLVMAliasInfo.h"
 #include "phasar/PhasarLLVM/Utils/LLVMBasedContainerConfig.h"
+#include "phasar/Utils/MaybeUniquePtr.h"
 #include "phasar/Utils/Soundness.h"
 
 #include "llvm/ADT/ArrayRef.h"
@@ -46,20 +49,10 @@ template <> struct CFGTraits<LLVMBasedICFG> : CFGTraits<LLVMBasedCFG> {};
 class LLVMBasedICFG : public LLVMBasedCFG, public ICFGBase<LLVMBasedICFG> {
   friend ICFGBase;
 
-  struct Builder;
-
 public:
+  // For backward compatibility
   static constexpr llvm::StringLiteral GlobalCRuntimeModelName =
-      "__psrCRuntimeGlobalCtorsModel";
-
-  static constexpr llvm::StringLiteral GlobalCRuntimeDtorModelName =
-      "__psrCRuntimeGlobalDtorsModel";
-
-  static constexpr llvm::StringLiteral GlobalCRuntimeDtorsCallerName =
-      "__psrGlobalDtorsCaller";
-
-  static constexpr llvm::StringLiteral GlobalCRuntimeUserEntrySelectorName =
-      "__psrCRuntimeUserEntrySelector";
+      GlobalCtorsDtorsModel::ModelName;
 
   /// Constructs the ICFG based on the given IRDB and the entry-points using a
   /// fixpoint iteration. This may take a long time.
@@ -95,9 +88,9 @@ class LLVMBasedICFG : public LLVMBasedCFG, public ICFGBase<LLVMBasedICFG> {
                          bool IncludeGlobals = true);
 
   /// Creates an ICFG with an already given call-graph
-  explicit LLVMBasedICFG(CallGraph<n_t, f_t> CG, LLVMProjectIRDB *IRDB);
+  explicit LLVMBasedICFG(CallGraph<n_t, f_t> CG, const LLVMProjectIRDB *IRDB);
 
-  explicit LLVMBasedICFG(LLVMProjectIRDB *IRDB,
+  explicit LLVMBasedICFG(const LLVMProjectIRDB *IRDB,
                          const CallGraphData &SerializedCG);
 
   // Deleter of LLVMTypeHierarchy may be unknown here...
@@ -131,10 +124,13 @@ class LLVMBasedICFG : public LLVMBasedCFG, public ICFGBase<LLVMBasedICFG> {
   }
 
   /// Gets the underlying IRDB
-  [[nodiscard]] LLVMProjectIRDB *getIRDB() const noexcept { return IRDB; }
+  [[nodiscard]] const LLVMProjectIRDB *getIRDB() const noexcept { return IRDB; }
 
   /// Returns true, if a function was generated by phasar.
-  [[nodiscard]] static bool isPhasarGenerated(const llvm::Function &) noexcept;
+  [[nodiscard]] static bool
+  isPhasarGenerated(const llvm::Function &F) noexcept {
+    return GlobalCtorsDtorsModel::isPhasarGenerated(F);
+  }
 
   using CFGBase::print;
   using ICFGBase::print;
@@ -157,22 +153,21 @@ class LLVMBasedICFG : public LLVMBasedCFG, public ICFGBase<LLVMBasedICFG> {
   void printImpl(llvm::raw_ostream &OS) const;
   void printAsJsonImpl(llvm::raw_ostream &OS) const;
   [[nodiscard, deprecated]] nlohmann::json getAsJsonImpl() const;
-  [[nodiscard]] const CallGraph<n_t, f_t> &getCallGraphImpl() const noexcept {
+  [[nodiscard]] const LLVMBasedCallGraph &getCallGraphImpl() const noexcept {
     return CG;
   }
 
   [[nodiscard]] llvm::Function *buildCRuntimeGlobalCtorsDtorsModel(
       llvm::Module &M, llvm::ArrayRef<llvm::Function *> UserEntryPoints);
 
   void initialize(LLVMProjectIRDB *IRDB, Resolver &CGResolver,
-                  llvm::ArrayRef<std::string> EntryPoints,
-                  const LLVMVFTableProvider &VTP, Soundness S,
+                  llvm::ArrayRef<std::string> EntryPoints, Soundness S,
                   bool IncludeGlobals);
 
   // ---
 
-  CallGraph<const llvm::Instruction *, const llvm::Function *> CG;
-  LLVMProjectIRDB *IRDB = nullptr;
+  LLVMBasedCallGraph CG;
+  const LLVMProjectIRDB *IRDB = nullptr;
   LLVMVFTableProvider VTP;
 };
 

include/phasar/PhasarLLVM/ControlFlow/Resolver/CHAResolver.h

@@ -22,7 +22,6 @@
 
 namespace llvm {
 class CallBase;
-class Function;
 } // namespace llvm
 
 namespace psr {
@@ -38,6 +37,11 @@ class CHAResolver : public Resolver {
 
   [[nodiscard]] std::string str() const override;
 
+  [[nodiscard]] bool
+  mutatesHelperAnalysisInformation() const noexcept override {
+    return false;
+  }
+
 protected:
   MaybeUniquePtr<const LLVMTypeHierarchy, true> TH;
 };

include/phasar/PhasarLLVM/ControlFlow/Resolver/DTAResolver.h

@@ -37,6 +37,22 @@ class DTAResolver : public CHAResolver {
 public:
   using TypeGraph_t = CachedTypeGraph;
 
+  DTAResolver(const LLVMProjectIRDB *IRDB, const LLVMVFTableProvider *VTP,
+              const LLVMTypeHierarchy *TH);
+
+  ~DTAResolver() override = default;
+
+  FunctionSetTy resolveVirtualCall(const llvm::CallBase *CallSite) override;
+
+  void otherInst(const llvm::Instruction *Inst) override;
+
+  [[nodiscard]] std::string str() const override;
+
+  [[nodiscard]] bool
+  mutatesHelperAnalysisInformation() const noexcept override {
+    return false;
+  }
+
 protected:
   TypeGraph_t TypeGraph;
 
@@ -53,18 +69,6 @@ class DTAResolver : public CHAResolver {
    * of vtable)
    */
   bool heuristicAntiConstructorVtablePos(const llvm::BitCastInst *BitCast);
-
-public:
-  DTAResolver(const LLVMProjectIRDB *IRDB, const LLVMVFTableProvider *VTP,
-              const LLVMTypeHierarchy *TH);
-
-  ~DTAResolver() override = default;
-
-  FunctionSetTy resolveVirtualCall(const llvm::CallBase *CallSite) override;
-
-  void otherInst(const llvm::Instruction *Inst) override;
-
-  [[nodiscard]] std::string str() const override;
 };
 } // namespace psr
 

include/phasar/PhasarLLVM/ControlFlow/Resolver/NOResolver.h

@@ -13,39 +13,27 @@
 #include "phasar/PhasarLLVM/ControlFlow/Resolver/Resolver.h"
 
 namespace llvm {
-class Instruction;
 class CallBase;
-class Function;
-class StructType;
 } // namespace llvm
 
 namespace psr {
 
 class NOResolver final : public Resolver {
-protected:
-  const llvm::Function *
-  getNonPureVirtualVFTEntry(const llvm::StructType *T, unsigned Idx,
-                            const llvm::CallBase *CallSite);
-
 public:
-  NOResolver(const LLVMProjectIRDB *IRDB);
+  NOResolver(const LLVMProjectIRDB *IRDB, const LLVMVFTableProvider *VTP);
 
   ~NOResolver() override = default;
 
-  void preCall(const llvm::Instruction *Inst) override;
-
-  void handlePossibleTargets(const llvm::CallBase *CallSite,
-                             FunctionSetTy &PossibleTargets) override;
-
-  void postCall(const llvm::Instruction *Inst) override;
-
   FunctionSetTy resolveVirtualCall(const llvm::CallBase *CallSite) override;
 
   FunctionSetTy resolveFunctionPointer(const llvm::CallBase *CallSite) override;
 
-  void otherInst(const llvm::Instruction *Inst) override;
-
   [[nodiscard]] std::string str() const override;
+
+  [[nodiscard]] bool
+  mutatesHelperAnalysisInformation() const noexcept override {
+    return false;
+  }
 };
 } // namespace psr