[SCEV][LAA] Re-commit r260085 and r260086, this time with a fix for the memory

sanitizer issue. The PredicatedScalarEvolution's copy constructor
wasn't copying the Generation value, and was leaving it un-initialized.

Original commit message:

[SCEV][LAA] Add no wrap SCEV predicates and use use them to improve strided pointer detection

Summary:
This change adds no wrap SCEV predicates with:
  - support for runtime checking
  - support for expression rewriting:
      (sext ({x,+,y}) -> {sext(x),+,sext(y)}
      (zext ({x,+,y}) -> {zext(x),+,sext(y)}

Note that we are sign extending the increment of the SCEV, even for
the zext case. This is needed to cover the fairly common case where y would
be a (small) negative integer. In order to do this, this change adds two new
flags: nusw and nssw that are applicable to AddRecExprs and permit the
transformations above.

We also change isStridedPtr in LAA to be able to make use of
these predicates. With this feature we should now always be able to
work around overflow issues in the dependence analysis.

Reviewers: mzolotukhin, sanjoy, anemet

Subscribers: mzolotukhin, sanjoy, llvm-commits, rengolin, jmolloy, hfinkel

Differential Revision: http://reviews.llvm.org/D15412

llvm-svn: 260112

1 files changed, 43 insertions, 18 deletions

diff --git a/llvm/lib/Analysis/LoopAccessAnalysis.cpp b/llvm/lib/Analysis/LoopAccessAnalysis.cpp
index a2ab231a62d..f371feb8142 100644
--- a/llvm/lib/Analysis/LoopAccessAnalysis.cpp
+++ b/llvm/lib/Analysis/LoopAccessAnalysis.cpp
@@ -773,7 +773,7 @@ static bool isInBoundsGep(Value *Ptr) {
 /// \brief Return true if an AddRec pointer \p Ptr is unsigned non-wrapping,
 /// i.e. monotonically increasing/decreasing.
 static bool isNoWrapAddRec(Value *Ptr, const SCEVAddRecExpr *AR,
-                           ScalarEvolution *SE, const Loop *L) {
+                           PredicatedScalarEvolution &PSE, const Loop *L) {
   // FIXME: This should probably only return true for NUW.
   if (AR->getNoWrapFlags(SCEV::NoWrapMask))
     return true;
@@ -809,7 +809,7 @@ static bool isNoWrapAddRec(Value *Ptr, const SCEVAddRecExpr *AR,
         // Assume constant for other the operand so that the AddRec can be
         // easily found.
         isa<ConstantInt>(OBO->getOperand(1))) {
-      auto *OpScev = SE->getSCEV(OBO->getOperand(0));
+      auto *OpScev = PSE.getSCEV(OBO->getOperand(0));
 
       if (auto *OpAR = dyn_cast<SCEVAddRecExpr>(OpScev))
         return OpAR->getLoop() == L && OpAR->getNoWrapFlags(SCEV::FlagNSW);
@@ -820,31 +820,35 @@ static bool isNoWrapAddRec(Value *Ptr, const SCEVAddRecExpr *AR,
 
 /// \brief Check whether the access through \p Ptr has a constant stride.
 int llvm::isStridedPtr(PredicatedScalarEvolution &PSE, Value *Ptr,
-                       const Loop *Lp, const ValueToValueMap &StridesMap) {
+                       const Loop *Lp, const ValueToValueMap &StridesMap,
+                       bool Assume) {
   Type *Ty = Ptr->getType();
   assert(Ty->isPointerTy() && "Unexpected non-ptr");
 
   // Make sure that the pointer does not point to aggregate types.
   auto *PtrTy = cast<PointerType>(Ty);
   if (PtrTy->getElementType()->isAggregateType()) {
-    DEBUG(dbgs() << "LAA: Bad stride - Not a pointer to a scalar type"
-          << *Ptr << "\n");
+    DEBUG(dbgs() << "LAA: Bad stride - Not a pointer to a scalar type" << *Ptr
+                 << "\n");
     return 0;
   }
 
   const SCEV *PtrScev = replaceSymbolicStrideSCEV(PSE, StridesMap, Ptr);
 
   const SCEVAddRecExpr *AR = dyn_cast<SCEVAddRecExpr>(PtrScev);
+  if (Assume && !AR)
+    AR = dyn_cast<SCEVAddRecExpr>(PSE.getAsAddRec(Ptr));
+
   if (!AR) {
-    DEBUG(dbgs() << "LAA: Bad stride - Not an AddRecExpr pointer "
-          << *Ptr << " SCEV: " << *PtrScev << "\n");
+    DEBUG(dbgs() << "LAA: Bad stride - Not an AddRecExpr pointer " << *Ptr
+                 << " SCEV: " << *PtrScev << "\n");
     return 0;
   }
 
   // The accesss function must stride over the innermost loop.
   if (Lp != AR->getLoop()) {
     DEBUG(dbgs() << "LAA: Bad stride - Not striding over innermost loop " <<
-          *Ptr << " SCEV: " << *PtrScev << "\n");
+          *Ptr << " SCEV: " << *AR << "\n");
     return 0;
   }
 
@@ -856,12 +860,23 @@ int llvm::isStridedPtr(PredicatedScalarEvolution &PSE, Value *Ptr,
   // to access the pointer value "0" which is undefined behavior in address
   // space 0, therefore we can also vectorize this case.
   bool IsInBoundsGEP = isInBoundsGep(Ptr);
-  bool IsNoWrapAddRec = isNoWrapAddRec(Ptr, AR, PSE.getSE(), Lp);
+  bool IsNoWrapAddRec =
+      PSE.hasNoOverflow(Ptr, SCEVWrapPredicate::IncrementNUSW) ||
+      isNoWrapAddRec(Ptr, AR, PSE, Lp);
   bool IsInAddressSpaceZero = PtrTy->getAddressSpace() == 0;
   if (!IsNoWrapAddRec && !IsInBoundsGEP && !IsInAddressSpaceZero) {
-    DEBUG(dbgs() << "LAA: Bad stride - Pointer may wrap in the address space "
-                 << *Ptr << " SCEV: " << *PtrScev << "\n");
-    return 0;
+    if (Assume) {
+      PSE.setNoOverflow(Ptr, SCEVWrapPredicate::IncrementNUSW);
+      IsNoWrapAddRec = true;
+      DEBUG(dbgs() << "LAA: Pointer may wrap in the address space:\n"
+                   << "LAA:   Pointer: " << *Ptr << "\n"
+                   << "LAA:   SCEV: " << *AR << "\n"
+                   << "LAA:   Added an overflow assumption\n");
+    } else {
+      DEBUG(dbgs() << "LAA: Bad stride - Pointer may wrap in the address space "
+                   << *Ptr << " SCEV: " << *AR << "\n");
+      return 0;
+    }
   }
 
   // Check the step is constant.
@@ -871,7 +886,7 @@ int llvm::isStridedPtr(PredicatedScalarEvolution &PSE, Value *Ptr,
   const SCEVConstant *C = dyn_cast<SCEVConstant>(Step);
   if (!C) {
     DEBUG(dbgs() << "LAA: Bad stride - Not a constant strided " << *Ptr <<
-          " SCEV: " << *PtrScev << "\n");
+          " SCEV: " << *AR << "\n");
     return 0;
   }
 
@@ -895,8 +910,18 @@ int llvm::isStridedPtr(PredicatedScalarEvolution &PSE, Value *Ptr,
   // know we can't "wrap around the address space". In case of address space
   // zero we know that this won't happen without triggering undefined behavior.
   if (!IsNoWrapAddRec && (IsInBoundsGEP || IsInAddressSpaceZero) &&
-      Stride != 1 && Stride != -1)
-    return 0;
+      Stride != 1 && Stride != -1) {
+    if (Assume) {
+      // We can avoid this case by adding a run-time check.
+      DEBUG(dbgs() << "LAA: Non unit strided pointer which is not either "
+                   << "inbouds or in address space 0 may wrap:\n"
+                   << "LAA:   Pointer: " << *Ptr << "\n"
+                   << "LAA:   SCEV: " << *AR << "\n"
+                   << "LAA:   Added an overflow assumption\n");
+      PSE.setNoOverflow(Ptr, SCEVWrapPredicate::IncrementNUSW);
+    } else
+      return 0;
+  }
 
   return Stride;
 }
@@ -1123,8 +1148,8 @@ MemoryDepChecker::isDependent(const MemAccessInfo &A, unsigned AIdx,
   const SCEV *AScev = replaceSymbolicStrideSCEV(PSE, Strides, APtr);
   const SCEV *BScev = replaceSymbolicStrideSCEV(PSE, Strides, BPtr);
 
-  int StrideAPtr = isStridedPtr(PSE, APtr, InnermostLoop, Strides);
-  int StrideBPtr = isStridedPtr(PSE, BPtr, InnermostLoop, Strides);
+  int StrideAPtr = isStridedPtr(PSE, APtr, InnermostLoop, Strides, true);
+  int StrideBPtr = isStridedPtr(PSE, BPtr, InnermostLoop, Strides, true);
 
   const SCEV *Src = AScev;
   const SCEV *Sink = BScev;
@@ -1824,7 +1849,7 @@ LoopAccessInfo::LoopAccessInfo(Loop *L, ScalarEvolution *SE,
                                const TargetLibraryInfo *TLI, AliasAnalysis *AA,
                                DominatorTree *DT, LoopInfo *LI,
                                const ValueToValueMap &Strides)
-    : PSE(*SE), PtrRtChecking(SE), DepChecker(PSE, L), TheLoop(L), DL(DL),
+    : PSE(*SE, *L), PtrRtChecking(SE), DepChecker(PSE, L), TheLoop(L), DL(DL),
       TLI(TLI), AA(AA), DT(DT), LI(LI), NumLoads(0), NumStores(0),
       MaxSafeDepDistBytes(-1U), CanVecMem(false),
       StoreToLoopInvariantAddress(false) {