Sparse IDE

.clang-tidy

@@ -17,6 +17,8 @@ Checks: '-*,
            -readability-convert-member-functions-to-static,
            -readability-isolate-declaration,
            -readability-identifier-length,
+           -readability-redundant-member-init,
+           -readability-use-anyofallof,
          cppcoreguidelines-*,
            -cppcoreguidelines-avoid-non-const-global-variables,
            -cppcoreguidelines-pro-bounds-array-to-pointer-decay,

CMakeLists.txt

@@ -65,7 +65,7 @@ set(RELEASE_CONFIGURATIONS RELWITHDEBINFO RELEASE CACHE INTERNAL "" FORCE)
 # https://reviews.llvm.org/D157613
 
 string(APPEND CMAKE_CXX_FLAGS " -MP -fstack-protector-strong -ffunction-sections -fdata-sections -pipe")
-string(APPEND CMAKE_CXX_FLAGS_DEBUG " -Og -fno-omit-frame-pointer")
+string(APPEND CMAKE_CXX_FLAGS_DEBUG " -fno-omit-frame-pointer")
 string(APPEND CMAKE_CXX_FLAGS_RELWITHDEBINFO " -fno-omit-frame-pointer")
 string(APPEND CMAKE_CXX_FLAGS_RELEASE "")
 

cmake/phasar_macros.cmake

@@ -125,10 +125,21 @@ function(generate_ll_file)
   endif()
 
   if(GEN_LL_MEM2REG)
+    if(CMAKE_CXX_COMPILER_ID MATCHES "Clang")
+      get_filename_component(COMPILER_PATH_STR ${CMAKE_CXX_COMPILER} DIRECTORY)
+      find_program(OPT_TOOL opt HINTS ${COMPILER_PATH_STR})
+    else()
+      find_program(OPT_TOOL opt)
+    endif()
+
+    if(NOT OPT_TOOL)
+      set(OPT_TOOL opt)
+    endif()
+
     add_custom_command(
       OUTPUT ${test_code_ll_file}
       COMMAND ${GEN_CMD} ${test_code_file_path} -o ${test_code_ll_file}
-      COMMAND ${CMAKE_CXX_COMPILER_LAUNCHER} opt -mem2reg -S -opaque-pointers=0 ${test_code_ll_file} -o ${test_code_ll_file}
+      COMMAND ${CMAKE_CXX_COMPILER_LAUNCHER} ${OPT_TOOL} -mem2reg -S -opaque-pointers=0 ${test_code_ll_file} -o ${test_code_ll_file}
       COMMENT ${GEN_CMD_COMMENT}
       DEPENDS ${GEN_LL_FILE}
       VERBATIM

config/double-free-config.json

@@ -0,0 +1,28 @@
+{
+  "name": "double-free",
+  "version": 1.0,
+  "functions": [
+    {
+      "name": "free",
+      "params": {
+        "source": [
+          0
+        ],
+        "sink": [
+          0
+        ]
+      }
+    },
+    {
+      "name": "_ZdlPv",
+      "params": {
+        "source": [
+          0
+        ],
+        "sink": [
+          0
+        ]
+      }
+    }
+  ]
+}

include/phasar/ControlFlow/CFGBase.h

@@ -136,7 +136,7 @@ template <typename Derived> class CFGBase {
     return self().getAsJsonImpl(Fun);
   }
 
-private:
+protected:
   Derived &self() noexcept { return static_cast<Derived &>(*this); }
   const Derived &self() const noexcept {
     return static_cast<const Derived &>(*this);

include/phasar/ControlFlow/SparseCFGBase.h

@@ -0,0 +1,43 @@
+/******************************************************************************
+ * 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_CONTROLFLOW_SPARSECFGBASE_H
+#define PHASAR_CONTROLFLOW_SPARSECFGBASE_H
+
+#include "phasar/ControlFlow/CFGBase.h"
+#include "phasar/Utils/ByRef.h"
+#include "phasar/Utils/Nullable.h"
+
+namespace psr {
+template <typename Derived> class SparseCFGBase : public CFGBase<Derived> {
+public:
+  using typename CFGBase<Derived>::n_t;
+  using typename CFGBase<Derived>::f_t;
+
+  /// Gets the next instruction in control-flow order, starting from
+  /// FromInstruction, that may use or define Val.
+  /// If the next user is ambiguous, returns null.
+  [[nodiscard]] Nullable<n_t>
+  nextUserOrNull(ByConstRef<n_t> FromInstruction) const {
+    return self().nextUserOrNullImpl(FromInstruction);
+  }
+
+protected:
+  using CFGBase<Derived>::self;
+};
+
+template <typename ICF, typename Domain>
+// NOLINTNEXTLINE(readability-identifier-naming)
+constexpr bool is_sparse_cfg_v = is_crtp_base_of_v<SparseCFGBase, ICF>
+    &&std::is_same_v<typename ICF::n_t, typename Domain::n_t>
+        &&std::is_same_v<typename ICF::f_t, typename Domain::f_t>;
+
+} // namespace psr
+
+#endif // PHASAR_CONTROLFLOW_SPARSECFGBASE_H

include/phasar/ControlFlow/SparseCFGProvider.h

@@ -0,0 +1,54 @@
+/******************************************************************************
+ * 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_CONTROLFLOW_SPARSECFGPROVIDER_H
+#define PHASAR_CONTROLFLOW_SPARSECFGPROVIDER_H
+
+#include "phasar/Utils/ByRef.h"
+
+#include <type_traits>
+
+namespace psr {
+template <typename T> T valueOf(T Val) { return Val; }
+
+template <typename Derived, typename F, typename V> class SparseCFGProvider {
+public:
+  using f_t = F;
+  using v_t = V;
+
+  template <typename D>
+  [[nodiscard]] decltype(auto) getSparseCFG(ByConstRef<f_t> Fun,
+                                            const D &Val) const {
+    using psr::valueOf;
+    static_assert(std::is_convertible_v<decltype(valueOf(Val)), v_t>);
+    return self().getSparseCFGImpl(Fun, valueOf(Val));
+  }
+
+private:
+  Derived &self() noexcept { return static_cast<Derived &>(*this); }
+  const Derived &self() const noexcept {
+    return static_cast<const Derived &>(*this);
+  }
+};
+
+template <typename T, typename D, typename = void>
+struct has_getSparseCFG : std::false_type {}; // NOLINT
+template <typename T, typename D>
+struct has_getSparseCFG<
+    T, D,
+    std::void_t<decltype(std::declval<const T>().getSparseCFG(
+        std::declval<typename T::f_t>(), std::declval<D>()))>>
+    : std::true_type {};
+
+template <typename T, typename D>
+// NOLINTNEXTLINE
+static constexpr bool has_getSparseCFG_v = has_getSparseCFG<T, D>::value;
+} // namespace psr
+
+#endif // PHASAR_CONTROLFLOW_SPARSECFGPROVIDER_H

include/phasar/DataFlow/IfdsIde/FlowFunctions.h

@@ -17,6 +17,7 @@
 #ifndef PHASAR_DATAFLOW_IFDSIDE_FLOWFUNCTIONS_H
 #define PHASAR_DATAFLOW_IFDSIDE_FLOWFUNCTIONS_H
 
+#include "phasar/Utils/Macros.h"
 #include "phasar/Utils/TypeTraits.h"
 
 #include "llvm/ADT/ArrayRef.h"
@@ -131,7 +132,7 @@ Container makeContainer(Range &&Rng) {
     Container C;
     reserveIfPossible(C, Rng.size());
     for (auto &&Fact : Rng) {
-      C.insert(std::forward<decltype(Fact)>(Fact));
+      C.insert(PSR_FWD(Fact));
     }
     return C;
   }

include/phasar/DataFlow/IfdsIde/Solver/IDESolver.h

@@ -18,6 +18,7 @@
 #define PHASAR_DATAFLOW_IFDSIDE_SOLVER_IDESOLVER_H
 
 #include "phasar/Config/Configuration.h"
+#include "phasar/ControlFlow/SparseCFGProvider.h"
 #include "phasar/DB/ProjectIRDBBase.h"
 #include "phasar/DataFlow/IfdsIde/EdgeFunction.h"
 #include "phasar/DataFlow/IfdsIde/EdgeFunctionStats.h"
@@ -35,15 +36,19 @@
 #include "phasar/DataFlow/IfdsIde/SolverResults.h"
 #include "phasar/Domain/AnalysisDomain.h"
 #include "phasar/Utils/Average.h"
+#include "phasar/Utils/ByRef.h"
 #include "phasar/Utils/DOTGraph.h"
 #include "phasar/Utils/JoinLattice.h"
 #include "phasar/Utils/Logger.h"
+#include "phasar/Utils/Macros.h"
+#include "phasar/Utils/Nullable.h"
 #include "phasar/Utils/PAMMMacros.h"
 #include "phasar/Utils/Table.h"
 #include "phasar/Utils/Utilities.h"
 
 #include "llvm/ADT/DenseSet.h"
 #include "llvm/ADT/StringRef.h"
+#include "llvm/Support/TypeName.h"
 #include "llvm/Support/raw_ostream.h"
 
 #include "nlohmann/json.hpp"
@@ -81,14 +86,24 @@ class IDESolver
   using t_t = typename AnalysisDomainTy::t_t;
   using v_t = typename AnalysisDomainTy::v_t;
 
+  template <typename I>
   IDESolver(IDETabulationProblem<AnalysisDomainTy, Container> &Problem,
-            const i_t *ICF)
-      : IDEProblem(Problem), ZeroValue(Problem.getZeroValue()), ICF(ICF),
+            const I *ICF)
+      : IDEProblem(Problem), ZeroValue(Problem.getZeroValue()),
+        ICF(&static_cast<const i_t &>(*ICF)), SVFG(ICF),
         SolverConfig(Problem.getIFDSIDESolverConfig()),
         CachedFlowEdgeFunctions(Problem), AllTop(Problem.allTopFunction()),
         JumpFn(std::make_shared<JumpFunctions<AnalysisDomainTy, Container>>()),
         Seeds(Problem.initialSeeds()) {
     assert(ICF != nullptr);
+
+    if constexpr (has_getSparseCFG_v<I, d_t>) {
+      NextUserOrNullCB = [](const void *SVFG, ByConstRef<f_t> Fun,
+                            ByConstRef<d_t> d3, ByConstRef<n_t> n) {
+        auto &&SCFG = static_cast<const I *>(SVFG)->getSparseCFG(Fun, d3);
+        return SCFG.nextUserOrNull(n);
+      };
+    }
   }
 
   IDESolver(IDETabulationProblem<AnalysisDomainTy, Container> *Problem,
@@ -339,6 +354,15 @@ class IDESolver
   }
 
 protected:
+  Nullable<n_t> getNextUserOrNull(ByConstRef<f_t> Fun, ByConstRef<d_t> d3,
+                                  ByConstRef<n_t> n) {
+    if (!NextUserOrNullCB || IDEProblem.isZeroValue(d3)) {
+      return {};
+    }
+
+    return NextUserOrNullCB(SVFG, Fun, d3, n);
+  }
+
   /// Lines 13-20 of the algorithm; processing a call site in the caller's
   /// context.
   ///
@@ -382,6 +406,15 @@ class IDESolver
 
     bool HasNoCalleeInformation = true;
 
+    auto &&Fun = ICF->getFunctionOf(n);
+    auto GetNextUse = [this, &Fun, &n](n_t nPrime, ByConstRef<d_t> d3) {
+      if (auto &&NextUser = getNextUserOrNull(Fun, d3, n)) {
+        return psr::unwrapNullable(PSR_FWD(NextUser));
+      }
+
+      return nPrime;
+    };
+
     // for each possible callee
     for (f_t SCalledProcN : Callees) { // still line 14
       // check if a special summary for the called procedure exists
@@ -409,7 +442,9 @@ class IDESolver
                              "Queried Summary Edge Function: " << SumEdgFnE);
             PHASAR_LOG_LEVEL(DEBUG,
                              "Compose: " << SumEdgFnE << " * " << f << '\n');
-            WorkList.emplace_back(PathEdge(d1, ReturnSiteN, std::move(d3)),
+
+            auto DestN = GetNextUse(ReturnSiteN, d3);
+            WorkList.emplace_back(PathEdge(d1, DestN, std::move(d3)),
                                   IDEProblem.extend(f, SumEdgFnE));
           }
         }
@@ -508,8 +543,10 @@ class IDESolver
                   d_t d5_restoredCtx = restoreContextOnReturnedFact(n, d2, d5);
                   // propagte the effects of the entire call
                   PHASAR_LOG_LEVEL(DEBUG, "Compose: " << fPrime << " * " << f);
+
+                  auto DestN = GetNextUse(RetSiteN, d5_restoredCtx);
                   WorkList.emplace_back(
-                      PathEdge(d1, RetSiteN, std::move(d5_restoredCtx)),
+                      PathEdge(d1, DestN, std::move(d5_restoredCtx)),
                       IDEProblem.extend(f, fPrime));
                 }
               }
@@ -545,7 +582,8 @@ class IDESolver
         auto fPrime = IDEProblem.extend(f, EdgeFnE);
         PHASAR_LOG_LEVEL(DEBUG, "Compose: " << EdgeFnE << " * " << f << " = "
                                             << fPrime);
-        WorkList.emplace_back(PathEdge(d1, ReturnSiteN, std::move(d3)),
+        auto DestN = GetNextUse(ReturnSiteN, d3);
+        WorkList.emplace_back(PathEdge(d1, DestN, std::move(d3)),
                               std::move(fPrime));
       }
     }
@@ -563,6 +601,8 @@ class IDESolver
     EdgeFunction<l_t> f = jumpFunction(Edge);
     auto [d1, n, d2] = Edge.consume();
 
+    const auto &Fun = ICF->getFunctionOf(n);
+
     for (const auto nPrime : ICF->getSuccsOf(n)) {
       FlowFunctionPtrType FlowFunc =
           CachedFlowEdgeFunctions.getNormalFlowFunction(n, nPrime);
@@ -575,14 +615,23 @@ class IDESolver
             CachedFlowEdgeFunctions.getNormalEdgeFunction(n, d2, nPrime, d3);
         PHASAR_LOG_LEVEL(DEBUG, "Queried Normal Edge Function: " << g);
         EdgeFunction<l_t> fPrime = IDEProblem.extend(f, g);
+
+        auto DestN = [&, &n = n] {
+          if (auto &&NextUser = getNextUserOrNull(Fun, d3, n)) {
+            return psr::unwrapNullable(PSR_FWD(NextUser));
+          }
+
+          return nPrime;
+        }();
+
         if (SolverConfig.emitESG()) {
-          IntermediateEdgeFunctions[std::make_tuple(n, d2, nPrime, d3)]
+          IntermediateEdgeFunctions[std::make_tuple(n, d2, DestN, d3)]
               .push_back(g);
         }
         PHASAR_LOG_LEVEL(DEBUG,
                          "Compose: " << g << " * " << f << " = " << fPrime);
         INC_COUNTER("EF Queries", 1, Full);
-        WorkList.emplace_back(PathEdge(d1, nPrime, std::move(d3)),
+        WorkList.emplace_back(PathEdge(d1, DestN, std::move(d3)),
                               std::move(fPrime));
       }
     }
@@ -915,6 +964,7 @@ class IDESolver
     for (const auto &Entry : Inc) {
       // line 22
       n_t c = Entry.first;
+      auto &&Fun = ICF->getFunctionOf(c);
       // for each return site
       for (n_t RetSiteC : ICF->getReturnSitesOfCallAt(c)) {
         // compute return-flow function
@@ -968,9 +1018,19 @@ class IDESolver
                   d_t d3 = ValAndFunc.first;
                   d_t d5_restoredCtx = restoreContextOnReturnedFact(c, d4, d5);
                   PHASAR_LOG_LEVEL(DEBUG, "Compose: " << fPrime << " * " << f3);
-                  WorkList.emplace_back(PathEdge(std::move(d3), RetSiteC,
-                                                 std::move(d5_restoredCtx)),
-                                        IDEProblem.extend(f3, fPrime));
+
+                  auto DestN = [&] {
+                    if (auto &&NextUser =
+                            getNextUserOrNull(Fun, d5_restoredCtx, c)) {
+                      return psr::unwrapNullable(PSR_FWD(NextUser));
+                    }
+
+                    return RetSiteC;
+                  }();
+
+                  WorkList.emplace_back(
+                      PathEdge(std::move(d3), DestN, std::move(d5_restoredCtx)),
+                      IDEProblem.extend(f3, fPrime));
                 }
               }
             }
@@ -1809,7 +1869,10 @@ class IDESolver
   IDETabulationProblem<AnalysisDomainTy, Container> &IDEProblem;
   d_t ZeroValue;
   const i_t *ICF;
+  const void *SVFG;
   IFDSIDESolverConfig &SolverConfig;
+  Nullable<n_t> (*NextUserOrNullCB)(const void *, ByConstRef<f_t>,
+                                    ByConstRef<d_t>, ByConstRef<n_t>) = nullptr;
 
   std::vector<std::pair<PathEdge<n_t, d_t>, EdgeFunction<l_t>>> WorkList;
   std::vector<std::pair<n_t, d_t>> ValuePropWL;