Teach BasicAA about the aliasing properties of memset_pattern16.

Fixes PR10872 and <rdar://problem/10065079>.


git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@139204 91177308-0d34-0410-b5e6-96231b3b80d8

2 files changed, 61 insertions, 1 deletions

diff --git a/lib/Analysis/BasicAliasAnalysis.cpp b/lib/Analysis/BasicAliasAnalysis.cpp
index 01257a1..4964c8c 100644
--- a/lib/Analysis/BasicAliasAnalysis.cpp
+++ b/lib/Analysis/BasicAliasAnalysis.cpp
@@ -30,6 +30,7 @@
 #include "llvm/Analysis/InstructionSimplify.h"
 #include "llvm/Analysis/ValueTracking.h"
 #include "llvm/Target/TargetData.h"
+#include "llvm/Target/TargetLibraryInfo.h"
 #include "llvm/ADT/SmallPtrSet.h"
 #include "llvm/ADT/SmallVector.h"
 #include "llvm/Support/ErrorHandling.h"
@@ -468,6 +469,7 @@ namespace {
 
     virtual void getAnalysisUsage(AnalysisUsage &AU) const {
       AU.addRequired<AliasAnalysis>();
+      AU.addRequired<TargetLibraryInfo>();
     }
 
     virtual AliasResult alias(const Location &LocA,
@@ -550,10 +552,15 @@ namespace {
 
 // Register this pass...
 char BasicAliasAnalysis::ID = 0;
-INITIALIZE_AG_PASS(BasicAliasAnalysis, AliasAnalysis, "basicaa",
+INITIALIZE_AG_PASS_BEGIN(BasicAliasAnalysis, AliasAnalysis, "basicaa",
+                   "Basic Alias Analysis (stateless AA impl)",
+                   false, true, false)
+INITIALIZE_PASS_DEPENDENCY(TargetLibraryInfo)
+INITIALIZE_AG_PASS_END(BasicAliasAnalysis, AliasAnalysis, "basicaa",
                    "Basic Alias Analysis (stateless AA impl)",
                    false, true, false)
 
+
 ImmutablePass *llvm::createBasicAliasAnalysisPass() {
   return new BasicAliasAnalysis();
 }
@@ -717,6 +724,7 @@ BasicAliasAnalysis::getModRefInfo(ImmutableCallSite CS,
       return NoModRef;
   }
 
+  const TargetLibraryInfo &TLI = getAnalysis<TargetLibraryInfo>();
   ModRefResult Min = ModRef;
 
   // Finally, handle specific knowledge of intrinsics.
@@ -819,6 +827,37 @@ BasicAliasAnalysis::getModRefInfo(ImmutableCallSite CS,
     }
     }
 
+  // We can bound the aliasing properties of memset_pattern16 just as we can
+  // for memcpy/memset.  This is particularly important because the 
+  // LoopIdiomRecognizer likes to turn loops into calls to memset_pattern16
+  // whenever possible.
+  else if (TLI.has(LibFunc::memset_pattern16) &&
+           CS.getCalledFunction() &&
+           CS.getCalledFunction()->getName() == "memset_pattern16") {
+    const Function *MS = CS.getCalledFunction();
+    FunctionType *MemsetType = MS->getFunctionType();
+    if (!MemsetType->isVarArg() && MemsetType->getNumParams() == 3 &&
+        isa<PointerType>(MemsetType->getParamType(0)) &&
+        isa<PointerType>(MemsetType->getParamType(1)) &&
+        isa<IntegerType>(MemsetType->getParamType(2))) {
+      uint64_t Len = UnknownSize;
+      if (const ConstantInt *LenCI = dyn_cast<ConstantInt>(CS.getArgument(2)))
+        Len = LenCI->getZExtValue();
+      const Value *Dest = CS.getArgument(0);
+      const Value *Src = CS.getArgument(1);
+      // If it can't overlap the source dest, then it doesn't modref the loc.
+      if (isNoAlias(Location(Dest, Len), Loc)) {
+        if (isNoAlias(Location(Src, 2), Loc))
+          return NoModRef;
+        // If it can't overlap the dest, then worst case it reads the loc.
+        Min = Ref;
+      } else if (isNoAlias(Location(Src, 2), Loc)) {
+        // If it can't overlap the source, then worst case it mutates the loc.
+        Min = Mod;
+      }
+    }
+  }
+
   // The AliasAnalysis base class has some smarts, lets use them.
   return ModRefResult(AliasAnalysis::getModRefInfo(CS, Loc) & Min);
 }
diff --git a/test/Analysis/BasicAA/memset_pattern.ll b/test/Analysis/BasicAA/memset_pattern.ll
new file mode 100644
index 0000000..590664c
--- /dev/null
+++ b/test/Analysis/BasicAA/memset_pattern.ll
@@ -0,0 +1,21 @@
+; RUN: opt -S -basicaa -gvn < %s | FileCheck %s
+; PR10872
+target datalayout = "e-p:64:64:64-i1:8:8-i8:8:8-i16:16:16-i32:32:32-i64:64:64-f32:32:32-f64:64:64-v64:64:64-v128:128:128-a0:0:64-s0:64:64-f80:128:128-n8:16:32:64"
+target triple = "x86_64-apple-macosx10.7"
+
+@z = internal global i32 0, align 4
+@y = internal global i32 0, align 4
+@x = internal constant i32 0, align 4
+
+; CHECK: @test
+define i32 @test() nounwind uwtable ssp {
+entry:
+  store i32 1, i32* @z
+  tail call void @memset_pattern16(i8* bitcast (i32* @y to i8*), i8* bitcast (i32* @x to i8*), i64 4) nounwind
+; CHECK-NOT: load
+  %l = load i32* @z
+; CHECK: ret i32 1
+  ret i32 %l
+}
+
+declare void @memset_pattern16(i8*, i8*, i64)