[AA] Hoist the logic to reformulate various AA queries in terms of other

parts of the AA interface out of the base class of every single AA
result object.

Because this logic reformulates the query in terms of some other aspect
of the API, it would easily cause O(n^2) query patterns in alias
analysis. These could in turn be magnified further based on the number
of call arguments, and then further based on the number of AA queries
made for a particular call. This ended up causing problems for Rust that
were actually noticable enough to get a bug (PR26564) and probably other
places as well.

When originally re-working the AA infrastructure, the desire was to
regularize the pattern of refinement without losing any generality.
While I think it was successful, that is clearly proving to be too
costly. And the cost is needless: we gain no actual improvement for this
generality of making a direct query to tbaa actually be able to
re-use some other alias analysis's refinement logic for one of the other
APIs, or some such. In short, this is entirely wasted work.

To the extent possible, delegation to other API surfaces should be done
at the aggregation layer so that we can avoid re-walking the
aggregation. In fact, this significantly simplifies the logic as we no
longer need to smuggle the aggregation layer into each alias analysis
(or the TargetLibraryInfo into each alias analysis just so we can form
argument memory locations!).

However, we also have some delegation logic inside of BasicAA and some
of it even makes sense. When the delegation logic is baking in specific
knowledge of aliasing properties of the LLVM IR, as opposed to simply
reformulating the query to utilize a different alias analysis interface
entry point, it makes a lot of sense to restrict that logic to
a different layer such as BasicAA. So one aspect of the delegation that
was in every AA base class is that when we don't have operand bundles,
we re-use function AA results as a fallback for callsite alias results.
This relies on the IR properties of calls and functions w.r.t. aliasing,
and so seems a better fit to BasicAA. I've lifted the logic up to that
point where it seems to be a natural fit. This still does a bit of
redundant work (we query function attributes twice, once via the
callsite and once via the function AA query) but it is *exactly* twice
here, no more.

The end result is that all of the delegation logic is hoisted out of the
base class and into either the aggregation layer when it is a pure
retargeting to a different API surface, or into BasicAA when it relies
on the IR's aliasing properties. This should fix the quadratic query
pattern reported in PR26564, although I don't have a stand-alone test
case to reproduce it.

It also seems general goodness. Now the numerous AAs that don't need
target library info don't carry it around and depend on it. I think
I can even rip out the general access to the aggregation layer and only
expose that in BasicAA as it is the only place where we re-query in that
manner.

However, this is a non-trivial change to the AA infrastructure so I want
to get some additional eyes on this before it lands. Sadly, it can't
wait long because we should really cherry pick this into 3.8 if we're
going to go this route.

Differential Revision: http://reviews.llvm.org/D17329

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@262490 91177308-0d34-0410-b5e6-96231b3b80d8

Author: chandlerc

include/llvm/Analysis/AliasAnalysis.h

@@ -43,14 +43,14 @@
 #include "llvm/IR/Metadata.h"
 #include "llvm/IR/PassManager.h"
 #include "llvm/Analysis/MemoryLocation.h"
+#include "llvm/Analysis/TargetLibraryInfo.h"
 
 namespace llvm {
 class BasicAAResult;
 class LoadInst;
 class StoreInst;
 class VAArgInst;
 class DataLayout;
-class TargetLibraryInfo;
 class Pass;
 class AnalysisUsage;
 class MemTransferInst;
@@ -161,9 +161,8 @@ class AAResults {
 public:
   // Make these results default constructable and movable. We have to spell
   // these out because MSVC won't synthesize them.
-  AAResults() {}
+  AAResults(const TargetLibraryInfo &TLI) : TLI(TLI) {}
   AAResults(AAResults &&Arg);
-  AAResults &operator=(AAResults &&Arg);
   ~AAResults();
 
   /// Register a specific AA result.
@@ -557,6 +556,8 @@ class AAResults {
 
   template <typename T> friend class AAResultBase;
 
+  const TargetLibraryInfo &TLI;
+
   std::vector<std::unique_ptr<Concept>> AAs;
 };
 
@@ -753,20 +754,23 @@ template <typename DerivedT> class AAResultBase {
     }
   };
 
-  const TargetLibraryInfo &TLI;
-
-  explicit AAResultBase(const TargetLibraryInfo &TLI) : TLI(TLI) {}
+  explicit AAResultBase() {}
 
   // Provide all the copy and move constructors so that derived types aren't
   // constrained.
-  AAResultBase(const AAResultBase &Arg) : TLI(Arg.TLI) {}
-  AAResultBase(AAResultBase &&Arg) : TLI(Arg.TLI) {}
+  AAResultBase(const AAResultBase &Arg) {}
+  AAResultBase(AAResultBase &&Arg) {}
 
   /// Get a proxy for the best AA result set to query at this time.
   ///
   /// When this result is part of a larger aggregation, this will proxy to that
   /// aggregation. When this result is used in isolation, it will just delegate
   /// back to the derived class's implementation.
+  ///
+  /// Note that callers of this need to take considerable care to not cause
+  /// performance problems when they use this routine, in the case of a large
+  /// number of alias analyses being aggregated, it can be expensive to walk
+  /// back across the chain.
   AAResultsProxy getBestAAResults() { return AAResultsProxy(AAR, derived()); }
 
 public:
@@ -783,173 +787,22 @@ template <typename DerivedT> class AAResultBase {
   }
 
   FunctionModRefBehavior getModRefBehavior(ImmutableCallSite CS) {
-    if (!CS.hasOperandBundles())
-      // If CS has operand bundles then aliasing attributes from the function it
-      // calls do not directly apply to the CallSite.  This can be made more
-      // precise in the future.
-      if (const Function *F = CS.getCalledFunction())
-        return getBestAAResults().getModRefBehavior(F);
-
     return FMRB_UnknownModRefBehavior;
   }
 
   FunctionModRefBehavior getModRefBehavior(const Function *F) {
     return FMRB_UnknownModRefBehavior;
   }
 
-  ModRefInfo getModRefInfo(ImmutableCallSite CS, const MemoryLocation &Loc);
-
-  ModRefInfo getModRefInfo(ImmutableCallSite CS1, ImmutableCallSite CS2);
-};
-
-/// Synthesize \c ModRefInfo for a call site and memory location by examining
-/// the general behavior of the call site and any specific information for its
-/// arguments.
-///
-/// This essentially, delegates across the alias analysis interface to collect
-/// information which may be enough to (conservatively) fulfill the query.
-template <typename DerivedT>
-ModRefInfo AAResultBase<DerivedT>::getModRefInfo(ImmutableCallSite CS,
-                                                 const MemoryLocation &Loc) {
-  auto MRB = getBestAAResults().getModRefBehavior(CS);
-  if (MRB == FMRB_DoesNotAccessMemory)
-    return MRI_NoModRef;
-
-  ModRefInfo Mask = MRI_ModRef;
-  if (AAResults::onlyReadsMemory(MRB))
-    Mask = MRI_Ref;
-
-  if (AAResults::onlyAccessesArgPointees(MRB)) {
-    bool DoesAlias = false;
-    ModRefInfo AllArgsMask = MRI_NoModRef;
-    if (AAResults::doesAccessArgPointees(MRB)) {
-      for (ImmutableCallSite::arg_iterator AI = CS.arg_begin(),
-                                           AE = CS.arg_end();
-           AI != AE; ++AI) {
-        const Value *Arg = *AI;
-        if (!Arg->getType()->isPointerTy())
-          continue;
-        unsigned ArgIdx = std::distance(CS.arg_begin(), AI);
-        MemoryLocation ArgLoc = MemoryLocation::getForArgument(CS, ArgIdx, TLI);
-        AliasResult ArgAlias = getBestAAResults().alias(ArgLoc, Loc);
-        if (ArgAlias != NoAlias) {
-          ModRefInfo ArgMask = getBestAAResults().getArgModRefInfo(CS, ArgIdx);
-          DoesAlias = true;
-          AllArgsMask = ModRefInfo(AllArgsMask | ArgMask);
-        }
-      }
-    }
-    if (!DoesAlias)
-      return MRI_NoModRef;
-    Mask = ModRefInfo(Mask & AllArgsMask);
-  }
-
-  // If Loc is a constant memory location, the call definitely could not
-  // modify the memory location.
-  if ((Mask & MRI_Mod) &&
-      getBestAAResults().pointsToConstantMemory(Loc, /*OrLocal*/ false))
-    Mask = ModRefInfo(Mask & ~MRI_Mod);
-
-  return Mask;
-}
-
-/// Synthesize \c ModRefInfo for two call sites by examining the general
-/// behavior of the call site and any specific information for its arguments.
-///
-/// This essentially, delegates across the alias analysis interface to collect
-/// information which may be enough to (conservatively) fulfill the query.
-template <typename DerivedT>
-ModRefInfo AAResultBase<DerivedT>::getModRefInfo(ImmutableCallSite CS1,
-                                                 ImmutableCallSite CS2) {
-  // If CS1 or CS2 are readnone, they don't interact.
-  auto CS1B = getBestAAResults().getModRefBehavior(CS1);
-  if (CS1B == FMRB_DoesNotAccessMemory)
-    return MRI_NoModRef;
-
-  auto CS2B = getBestAAResults().getModRefBehavior(CS2);
-  if (CS2B == FMRB_DoesNotAccessMemory)
-    return MRI_NoModRef;
-
-  // If they both only read from memory, there is no dependence.
-  if (AAResults::onlyReadsMemory(CS1B) && AAResults::onlyReadsMemory(CS2B))
-    return MRI_NoModRef;
-
-  ModRefInfo Mask = MRI_ModRef;
-
-  // If CS1 only reads memory, the only dependence on CS2 can be
-  // from CS1 reading memory written by CS2.
-  if (AAResults::onlyReadsMemory(CS1B))
-    Mask = ModRefInfo(Mask & MRI_Ref);
-
-  // If CS2 only access memory through arguments, accumulate the mod/ref
-  // information from CS1's references to the memory referenced by
-  // CS2's arguments.
-  if (AAResults::onlyAccessesArgPointees(CS2B)) {
-    ModRefInfo R = MRI_NoModRef;
-    if (AAResults::doesAccessArgPointees(CS2B)) {
-      for (ImmutableCallSite::arg_iterator I = CS2.arg_begin(),
-                                           E = CS2.arg_end();
-           I != E; ++I) {
-        const Value *Arg = *I;
-        if (!Arg->getType()->isPointerTy())
-          continue;
-        unsigned CS2ArgIdx = std::distance(CS2.arg_begin(), I);
-        auto CS2ArgLoc = MemoryLocation::getForArgument(CS2, CS2ArgIdx, TLI);
-
-        // ArgMask indicates what CS2 might do to CS2ArgLoc, and the dependence
-        // of CS1 on that location is the inverse.
-        ModRefInfo ArgMask =
-            getBestAAResults().getArgModRefInfo(CS2, CS2ArgIdx);
-        if (ArgMask == MRI_Mod)
-          ArgMask = MRI_ModRef;
-        else if (ArgMask == MRI_Ref)
-          ArgMask = MRI_Mod;
-
-        ArgMask = ModRefInfo(ArgMask &
-                             getBestAAResults().getModRefInfo(CS1, CS2ArgLoc));
-
-        R = ModRefInfo((R | ArgMask) & Mask);
-        if (R == Mask)
-          break;
-      }
-    }
-    return R;
+  ModRefInfo getModRefInfo(ImmutableCallSite CS, const MemoryLocation &Loc) {
+    return MRI_ModRef;
   }
 
-  // If CS1 only accesses memory through arguments, check if CS2 references
-  // any of the memory referenced by CS1's arguments. If not, return NoModRef.
-  if (AAResults::onlyAccessesArgPointees(CS1B)) {
-    ModRefInfo R = MRI_NoModRef;
-    if (AAResults::doesAccessArgPointees(CS1B)) {
-      for (ImmutableCallSite::arg_iterator I = CS1.arg_begin(),
-                                           E = CS1.arg_end();
-           I != E; ++I) {
-        const Value *Arg = *I;
-        if (!Arg->getType()->isPointerTy())
-          continue;
-        unsigned CS1ArgIdx = std::distance(CS1.arg_begin(), I);
-        auto CS1ArgLoc = MemoryLocation::getForArgument(CS1, CS1ArgIdx, TLI);
-
-        // ArgMask indicates what CS1 might do to CS1ArgLoc; if CS1 might Mod
-        // CS1ArgLoc, then we care about either a Mod or a Ref by CS2. If CS1
-        // might Ref, then we care only about a Mod by CS2.
-        ModRefInfo ArgMask = getBestAAResults().getArgModRefInfo(CS1, CS1ArgIdx);
-        ModRefInfo ArgR = getBestAAResults().getModRefInfo(CS2, CS1ArgLoc);
-        if (((ArgMask & MRI_Mod) != MRI_NoModRef &&
-             (ArgR & MRI_ModRef) != MRI_NoModRef) ||
-            ((ArgMask & MRI_Ref) != MRI_NoModRef &&
-             (ArgR & MRI_Mod) != MRI_NoModRef))
-          R = ModRefInfo((R | ArgMask) & Mask);
-
-        if (R == Mask)
-          break;
-      }
-    }
-    return R;
+  ModRefInfo getModRefInfo(ImmutableCallSite CS1, ImmutableCallSite CS2) {
+    return MRI_ModRef;
   }
+};
 
-  return Mask;
-}
 
 /// isNoAliasCall - Return true if this pointer is returned by a noalias
 /// function.
@@ -1019,7 +872,7 @@ class AAManager {
   }
 
   Result run(Function &F, AnalysisManager<Function> *AM) {
-    Result R;
+    Result R(AM->getResult<TargetLibraryAnalysis>(F));
     for (auto &Getter : FunctionResultGetters)
       (*Getter)(F, *AM, R);
     return R;
@@ -1075,13 +928,13 @@ ImmutablePass *createExternalAAWrapperPass(
 /// inside of a \c ModulePass or a \c CallGraphSCCPass.
 ///
 /// If a \c ModulePass or a \c CallGraphSCCPass calls \p
-/// createLegacyPMAAResults, it also needs to call \p addUsedAAAnalyses in \p
-/// getAnalysisUsage.
+/// createLegacyPMAAResults, it also needs to call \p getAAResultsAnalysisUsage
+/// in \p / getAnalysisUsage.
 AAResults createLegacyPMAAResults(Pass &P, Function &F, BasicAAResult &BAR);
 
 /// A helper for the legacy pass manager to populate \p AU to add uses to make
 /// sure the analyses required by \p createLegacyPMAAResults are available.
-void addUsedAAAnalyses(AnalysisUsage &AU);
+void getAAResultsAnalysisUsage(AnalysisUsage &AU);
 
 } // End llvm namespace
 

include/llvm/Analysis/BasicAliasAnalysis.h

@@ -40,6 +40,7 @@ class BasicAAResult : public AAResultBase<BasicAAResult> {
   friend AAResultBase<BasicAAResult>;
 
   const DataLayout &DL;
+  const TargetLibraryInfo &TLI;
   AssumptionCache &AC;
   DominatorTree *DT;
   LoopInfo *LI;
@@ -48,13 +49,14 @@ class BasicAAResult : public AAResultBase<BasicAAResult> {
   BasicAAResult(const DataLayout &DL, const TargetLibraryInfo &TLI,
                 AssumptionCache &AC, DominatorTree *DT = nullptr,
                 LoopInfo *LI = nullptr)
-      : AAResultBase(TLI), DL(DL), AC(AC), DT(DT), LI(LI) {}
+      : AAResultBase(), DL(DL), TLI(TLI), AC(AC), DT(DT), LI(LI) {}
 
   BasicAAResult(const BasicAAResult &Arg)
-      : AAResultBase(Arg), DL(Arg.DL), AC(Arg.AC), DT(Arg.DT), LI(Arg.LI) {}
-  BasicAAResult(BasicAAResult &&Arg)
-      : AAResultBase(std::move(Arg)), DL(Arg.DL), AC(Arg.AC), DT(Arg.DT),
+      : AAResultBase(Arg), DL(Arg.DL), TLI(Arg.TLI), AC(Arg.AC), DT(Arg.DT),
         LI(Arg.LI) {}
+  BasicAAResult(BasicAAResult &&Arg)
+      : AAResultBase(std::move(Arg)), DL(Arg.DL), TLI(Arg.TLI), AC(Arg.AC),
+        DT(Arg.DT), LI(Arg.LI) {}
 
   /// Handle invalidation events from the new pass manager.
   ///

include/llvm/Analysis/CFLAliasAnalysis.h

@@ -32,7 +32,7 @@ class CFLAAResult : public AAResultBase<CFLAAResult> {
   struct FunctionInfo;
 
 public:
-  explicit CFLAAResult(const TargetLibraryInfo &TLI);
+  explicit CFLAAResult();
   CFLAAResult(CFLAAResult &&Arg);
 
   /// Handle invalidation events from the new pass manager.

include/llvm/Analysis/GlobalsModRef.h

@@ -35,6 +35,7 @@ class GlobalsAAResult : public AAResultBase<GlobalsAAResult> {
   class FunctionInfo;
 
   const DataLayout &DL;
+  const TargetLibraryInfo &TLI;
 
   /// The globals that do not have their addresses taken.
   SmallPtrSet<const GlobalValue *, 8> NonAddressTakenGlobals;

include/llvm/Analysis/ObjCARCAliasAnalysis.h

@@ -24,7 +24,6 @@
 #define LLVM_ANALYSIS_OBJCARCALIASANALYSIS_H
 
 #include "llvm/Analysis/AliasAnalysis.h"
-#include "llvm/Analysis/TargetLibraryInfo.h"
 #include "llvm/Pass.h"
 
 namespace llvm {
@@ -42,8 +41,7 @@ class ObjCARCAAResult : public AAResultBase<ObjCARCAAResult> {
   const DataLayout &DL;
 
 public:
-  explicit ObjCARCAAResult(const DataLayout &DL, const TargetLibraryInfo &TLI)
-      : AAResultBase(TLI), DL(DL) {}
+  explicit ObjCARCAAResult(const DataLayout &DL) : AAResultBase(), DL(DL) {}
   ObjCARCAAResult(ObjCARCAAResult &&Arg)
       : AAResultBase(std::move(Arg)), DL(Arg.DL) {}
 

include/llvm/Analysis/ScalarEvolutionAliasAnalysis.h

@@ -28,8 +28,7 @@ class SCEVAAResult : public AAResultBase<SCEVAAResult> {
   ScalarEvolution &SE;
 
 public:
-  explicit SCEVAAResult(const TargetLibraryInfo &TLI, ScalarEvolution &SE)
-      : AAResultBase(TLI), SE(SE) {}
+  explicit SCEVAAResult(ScalarEvolution &SE) : AAResultBase(), SE(SE) {}
   SCEVAAResult(SCEVAAResult &&Arg) : AAResultBase(std::move(Arg)), SE(Arg.SE) {}
 
   AliasResult alias(const MemoryLocation &LocA, const MemoryLocation &LocB);

include/llvm/Analysis/ScopedNoAliasAA.h

@@ -27,8 +27,7 @@ class ScopedNoAliasAAResult : public AAResultBase<ScopedNoAliasAAResult> {
   friend AAResultBase<ScopedNoAliasAAResult>;
 
 public:
-  explicit ScopedNoAliasAAResult(const TargetLibraryInfo &TLI)
-      : AAResultBase(TLI) {}
+  explicit ScopedNoAliasAAResult() : AAResultBase() {}
   ScopedNoAliasAAResult(ScopedNoAliasAAResult &&Arg)
       : AAResultBase(std::move(Arg)) {}
 

include/llvm/Analysis/TypeBasedAliasAnalysis.h

@@ -27,8 +27,7 @@ class TypeBasedAAResult : public AAResultBase<TypeBasedAAResult> {
   friend AAResultBase<TypeBasedAAResult>;
 
 public:
-  explicit TypeBasedAAResult(const TargetLibraryInfo &TLI)
-      : AAResultBase(TLI) {}
+  explicit TypeBasedAAResult() {}
   TypeBasedAAResult(TypeBasedAAResult &&Arg) : AAResultBase(std::move(Arg)) {}
 
   /// Handle invalidation events from the new pass manager.

include/llvm/LinkAllPasses.h

@@ -31,6 +31,7 @@
 #include "llvm/Analysis/ScalarEvolution.h"
 #include "llvm/Analysis/ScalarEvolutionAliasAnalysis.h"
 #include "llvm/Analysis/ScopedNoAliasAA.h"
+#include "llvm/Analysis/TargetLibraryInfo.h"
 #include "llvm/Analysis/TypeBasedAliasAnalysis.h"
 #include "llvm/CodeGen/Passes.h"
 #include "llvm/IR/Function.h"
@@ -190,7 +191,9 @@ namespace {
       (void)new llvm::ScalarEvolutionWrapperPass();
       llvm::Function::Create(nullptr, llvm::GlobalValue::ExternalLinkage)->viewCFGOnly();
       llvm::RGPassManager RGM;
-      llvm::AliasAnalysis AA;
+      llvm::TargetLibraryInfoImpl TLII;
+      llvm::TargetLibraryInfo TLI(TLII);
+      llvm::AliasAnalysis AA(TLI);
       llvm::AliasSetTracker X(AA);
       X.add(nullptr, 0, llvm::AAMDNodes()); // for -print-alias-sets
       (void) llvm::AreStatisticsEnabled();