[SROA] Fix the loop exit placement to be prior to indexing the splits

[oota-llvm.git] / lib / Transforms / Scalar / MemCpyOptimizer.cpp

diff --git a/lib/Transforms/Scalar/MemCpyOptimizer.cpp b/lib/Transforms/Scalar/MemCpyOptimizer.cpp
index b6bc79228824839eec03c17bf516d6d822daaaae..be524be0082af01c6aae8c492f8f5b0065f84f05 100644 (file)
--- a/lib/Transforms/Scalar/MemCpyOptimizer.cpp
+++ b/lib/Transforms/Scalar/MemCpyOptimizer.cpp
@@ -16,6 +16,7 @@
 #include "llvm/ADT/SmallVector.h"
 #include "llvm/ADT/Statistic.h"
 #include "llvm/Analysis/AliasAnalysis.h"
+#include "llvm/Analysis/AssumptionTracker.h"
 #include "llvm/Analysis/MemoryDependenceAnalysis.h"
 #include "llvm/Analysis/ValueTracking.h"
 #include "llvm/IR/DataLayout.h"
@@ -329,6 +330,7 @@ namespace {
     // This transformation requires dominator postdominator info
     void getAnalysisUsage(AnalysisUsage &AU) const override {
       AU.setPreservesCFG();
+      AU.addRequired<AssumptionTracker>();
       AU.addRequired<DominatorTreeWrapperPass>();
       AU.addRequired<MemoryDependenceAnalysis>();
       AU.addRequired<AliasAnalysis>();
@@ -361,6 +363,7 @@ FunctionPass *llvm::createMemCpyOptPass() { return new MemCpyOpt(); }
 
 INITIALIZE_PASS_BEGIN(MemCpyOpt, "memcpyopt", "MemCpy Optimization",
                       false, false)
+INITIALIZE_PASS_DEPENDENCY(AssumptionTracker)
 INITIALIZE_PASS_DEPENDENCY(DominatorTreeWrapperPass)
 INITIALIZE_PASS_DEPENDENCY(MemoryDependenceAnalysis)
 INITIALIZE_PASS_DEPENDENCY(TargetLibraryInfo)
@@ -631,22 +634,24 @@ bool MemCpyOpt::performCallSlotOptzn(Instruction *cpy,
     if (destSize < srcSize)
       return false;
   } else if (Argument *A = dyn_cast<Argument>(cpyDest)) {
-    // If the destination is an sret parameter then only accesses that are
-    // outside of the returned struct type can trap.
-    if (!A->hasStructRetAttr())
-      return false;
+    if (A->getDereferenceableBytes() < srcSize) {
+      // If the destination is an sret parameter then only accesses that are
+      // outside of the returned struct type can trap.
+      if (!A->hasStructRetAttr())
+        return false;
 
-    Type *StructTy = cast<PointerType>(A->getType())->getElementType();
-    if (!StructTy->isSized()) {
-      // The call may never return and hence the copy-instruction may never
-      // be executed, and therefore it's not safe to say "the destination
-      // has at least <cpyLen> bytes, as implied by the copy-instruction",
-      return false;
-    }
+      Type *StructTy = cast<PointerType>(A->getType())->getElementType();
+      if (!StructTy->isSized()) {
+        // The call may never return and hence the copy-instruction may never
+        // be executed, and therefore it's not safe to say "the destination
+        // has at least <cpyLen> bytes, as implied by the copy-instruction",
+        return false;
+      }
 
-    uint64_t destSize = DL->getTypeAllocSize(StructTy);
-    if (destSize < srcSize)
-      return false;
+      uint64_t destSize = DL->getTypeAllocSize(StructTy);
+      if (destSize < srcSize)
+        return false;
+    }
   } else {
     return false;
   }
@@ -673,17 +678,31 @@ bool MemCpyOpt::performCallSlotOptzn(Instruction *cpy,
     if (isa<BitCastInst>(U) || isa<AddrSpaceCastInst>(U)) {
       for (User *UU : U->users())
         srcUseList.push_back(UU);
-    } else if (GetElementPtrInst *G = dyn_cast<GetElementPtrInst>(U)) {
-      if (G->hasAllZeroIndices())
-        for (User *UU : U->users())
-          srcUseList.push_back(UU);
-      else
+      continue;
+    }
+    if (GetElementPtrInst *G = dyn_cast<GetElementPtrInst>(U)) {
+      if (!G->hasAllZeroIndices())
         return false;
-    } else if (U != C && U != cpy) {
-      return false;
+
+      for (User *UU : U->users())
+        srcUseList.push_back(UU);
+      continue;
     }
+    if (const IntrinsicInst *IT = dyn_cast<IntrinsicInst>(U))
+      if (IT->getIntrinsicID() == Intrinsic::lifetime_start ||
+          IT->getIntrinsicID() == Intrinsic::lifetime_end)
+        continue;
+
+    if (U != C && U != cpy)
+      return false;
   }
 
+  // Check that src isn't captured by the called function since the
+  // transformation can cause aliasing issues in that case.
+  for (unsigned i = 0, e = CS.arg_size(); i != e; ++i)
+    if (CS.getArgument(i) == cpySrc && !CS.doesNotCapture(i))
+      return false;
+
   // Since we're changing the parameter to the callsite, we need to make sure
   // that what would be the new parameter dominates the callsite.
   DominatorTree &DT = getAnalysis<DominatorTreeWrapperPass>().getDomTree();
@@ -963,8 +982,11 @@ bool MemCpyOpt::processByValArgument(CallSite CS, unsigned ArgNo) {
 
   // If it is greater than the memcpy, then we check to see if we can force the
   // source of the memcpy to the alignment we need.  If we fail, we bail out.
+  AssumptionTracker *AT = &getAnalysis<AssumptionTracker>();
+  DominatorTree &DT = getAnalysis<DominatorTreeWrapperPass>().getDomTree();
   if (MDep->getAlignment() < ByValAlign &&
-      getOrEnforceKnownAlignment(MDep->getSource(),ByValAlign, DL) < ByValAlign)
+      getOrEnforceKnownAlignment(MDep->getSource(),ByValAlign,
+                                 DL, AT, CS.getInstruction(), &DT) < ByValAlign)
     return false;
 
   // Verify that the copied-from memory doesn't change in between the memcpy and