[AssumeBundles] add cannonicalisation to the assume builder

[Tyker]

Summary

this reduces significantly the number of assumes generated without aftecting too much
the information that is preserved. this improves the compile-time cost
of enable-knowledge-retention significantly.

cost of enable-knowledge-retention before the patch

Metric: inst_count
Program                                                      baseline      patched       diff
 test-suite :: CTMark/tramp3d-v4/tramp3d-v4.test              118759192395  155068323810 30.6%
 test-suite :: CTMark/Bullet/bullet.test                      132173382943  172092586752 30.2%
 test-suite :: CTMark/7zip/7zip-benchmark.test                193010622254  201710179449  4.5%
 test-suite :: CTMark/kimwitu++/kc.test                       61391711480   64129872325   4.5%
 test-suite :: CTMark/ClamAV/clamscan.test                    80896366241   79491189655  -1.7%
 test-suite :: CTMark/sqlite3/sqlite3.test                    77983944971   77321823801  -0.8%
 test-suite :: CTMark/consumer-typeset/consumer-typeset.test  56565054641   56127618358  -0.8%
 test-suite :: CTMark/mafft/pairlocalalign.test               54476149303   54268277583  -0.4%
 test-suite :: CTMark/SPASS/SPASS.test                        65588484451   65405266465  -0.3%
 Geomean difference                                                                       6.6%

Metric: compile_time
Program                                                      baseline patched diff
 test-suite :: CTMark/tramp3d-v4/tramp3d-v4.test              38.60    49.38  27.9%
 test-suite :: CTMark/Bullet/bullet.test                      43.90    54.25  23.6%
 test-suite :: CTMark/kimwitu++/kc.test                       22.88    24.20   5.8%
 test-suite :: CTMark/7zip/7zip-benchmark.test                66.93    70.27   5.0%
 test-suite :: CTMark/sqlite3/sqlite3.test                    24.70    24.47  -0.9%
 test-suite :: CTMark/ClamAV/clamscan.test                    24.12    23.91  -0.9%
 test-suite :: CTMark/mafft/pairlocalalign.test               15.91    15.84  -0.5%
 test-suite :: CTMark/consumer-typeset/consumer-typeset.test  17.09    17.10   0.1%
 test-suite :: CTMark/SPASS/SPASS.test                        21.81    21.82   0.1%
 Geomean difference                                                            6.2%

cost after this patch

Metric: inst_count
Program                                                      baseline      patched       diff
 test-suite :: CTMark/tramp3d-v4/tramp3d-v4.test              118763840162  138467480074 16.6%
 test-suite :: CTMark/Bullet/bullet.test                      132174006965  136291227509  3.1%
 test-suite :: CTMark/kimwitu++/kc.test                       61393799157   62997635803   2.6%
 test-suite :: CTMark/7zip/7zip-benchmark.test                192990097336  197727495942  2.5%
 test-suite :: CTMark/ClamAV/clamscan.test                    80896861556   79484165212  -1.7%
 test-suite :: CTMark/sqlite3/sqlite3.test                    77990896082   77343953874  -0.8%
 test-suite :: CTMark/consumer-typeset/consumer-typeset.test  56564081236   56118858184  -0.8%
 test-suite :: CTMark/mafft/pairlocalalign.test               54474697759   54263079601  -0.4%
 test-suite :: CTMark/SPASS/SPASS.test                        65590532152   65405772667  -0.3%
 Geomean difference                                                                       2.2%

Metric: compile_time
Program                                                      baseline patched diff
 test-suite :: CTMark/tramp3d-v4/tramp3d-v4.test              38.91    46.75  20.1%
 test-suite :: CTMark/7zip/7zip-benchmark.test                67.93    71.22   4.8%
 test-suite :: CTMark/Bullet/bullet.test                      44.90    46.55   3.7%
 test-suite :: CTMark/kimwitu++/kc.test                       23.58    24.28   3.0%
 test-suite :: CTMark/mafft/pairlocalalign.test               16.09    16.39   1.8%
 test-suite :: CTMark/ClamAV/clamscan.test                    24.63    24.19  -1.8%
 test-suite :: CTMark/sqlite3/sqlite3.test                    25.25    25.37   0.5%
 test-suite :: CTMark/consumer-typeset/consumer-typeset.test  17.28    17.33   0.2%
 test-suite :: CTMark/SPASS/SPASS.test                        22.36    22.39   0.1%
 Geomean difference                                                            3.4%

[jdoerfert]

Nice. I guess we should explicitly look into tramp3d afterwards, 20% probably offers nice opportunities to reduce the cost ;)

Some comments and questions below.

llvm/lib/IR/Operator.cpp

46

The maximum alignment is MaximumAlignment in Value.h, please use that over the shifting stuff.

Can you explain the logic here in a comment at the function entry, please.

llvm/lib/Transforms/Utils/AssumeBundleBuilder.cpp

85

It will return \p RK not an empty one, right?

117

The above functions can be static I guess. I'm a little worried that we add yet another "stripPointer" function here. I guess you cannot reuse the existing ones but maybe we should extend those instead.

Why do you strip this again:
RK.WasOn = V->stripInBoundsOffsets();
That seems wrong. V is the base we can use, stripping dynamic inbounds offsets will invalidate the deref value, right?

168

I'm not super happy about the last part of the check but it is not incorrect. Let's keep it for now.

Btw. We can also weed out (if we don't do it already) the trivial cases:
deref and nonnull for alloca, byval arguments, or certain global values for example.

llvm/test/Analysis/BasicAA/featuretest.ll

41

Can you explain why we drop nonnull but not deref here? I mean, it is good but I don't get how. (As mentioned above, we could also drop deref).

[Tyker]

addressed comments

llvm/lib/Transforms/Utils/AssumeBundleBuilder.cpp

85

yes the comment was outdated.

117

The above functions can be static I guess.

it is in an inline namespace so it is already implicitly static.

I'm a little worried that we add yet another "stripPointer" function here. I guess you cannot reuse the
existing ones but maybe we should extend those instead.

Yes i agree that there are already implementations of this (which i partially ripped of) but they are not general enough for my use case because i need to update the alignment value during the stripping based on the instructions being stripped.

We should probably refactor the various striping and GetPointerBaseWithConstantOffset such that we have one generic implementation and all similar utilities are specialization of that generic implementation.

Why do you strip this again:
RK.WasOn = V->stripInBoundsOffsets();
That seems wrong. V is the base we can use, stripping dynamic inbounds offsets will invalidate the deref value, right?

yeah we can't assume this because dynamic offsets maybe negative.

168

i added a few rules to remove more useless knowledge to cover byval and globals.

I'm not super happy about the last part of the check but it is not incorrect

can you elaborate on why ? the idea is that salvageKnowledge is intended to be called on an instruction about to be removed. and usually when removing an instruction we keep going recursively. this changed makes sure we don't preserve knowledge in a way that is going to prevent from removing instructions recursively.
we will need to deal with droppable uses in RecursivelyDeleteTriviallyDeadInstructions, isInstructionTriviallyDead and others. but there are also a few custom implementations of those.

llvm/test/Analysis/BasicAA/featuretest.ll

41

nonnull and align is dropped because the canonicalizedKnowledge was able to canonicalize it through the inbounds gep of dynamic index and see that the underlying pointer is an alloca so it isn't worth preserving.

however i don't think it is possible to canonicalize a dereferencable through the inbounds gep of dynamic index since the index may be negative.

but i will add a GetUnderlyingObject to the check of whether the pointer is an alloca this way dereferencable will also be removed.

[jdoerfert]

I left two comments. Could you verify the Value.h change does not impact compile time. Other than than, LGTM.

llvm/lib/Transforms/Utils/AssumeBundleBuilder.cpp

168

can you elaborate on why ?

Maybe I am wrong but I thought there is no enforced connection between InsertBeforeInstruction and the knowledge (or the instruction from which we retain knowledge). We use it a certain way, true, but my concerns are with potential other usages. As I said, it is not wrong and what I fear might never happen.

170

If the kind is Deref you can ask the value for the known deref bytes with Value::getPointerDereferenceableBytes. For align it is Value::getPointerAlignment. Unsure if that is needed.

llvm/include/llvm/IR/Operator.h

public:
   }
 
   unsigned countNonConstantIndices() const {
     return count_if(make_range(idx_begin(), idx_end()), [](const Use& use) {
         return !isa<ConstantInt>(*use);
       });
   }
 
+  /// Compute the maximum alignment that this GEP is garranteed to preserve.
+  Align getMaxPreservedAlignment(const DataLayout &DL) const;
+
   /// Accumulate the constant address offset of this GEP if possible.
   ///
   /// This routine accepts an APInt into which it will accumulate the constant
   /// offset of this GEP if the GEP is in fact constant. If the GEP is not
   /// all-constant, it returns false and the value of the offset APInt is
   /// undefined (it is *not* preserved!). The APInt passed into this routine
   /// must be at exactly as wide as the IntPtr type for the address space of the
   /// base GEP pointer.

llvm/lib/IR/Operator.cpp

 }
 
 Type *GEPOperator::getResultElementType() const {
   if (auto *I = dyn_cast<GetElementPtrInst>(this))
     return I->getResultElementType();
   return cast<GetElementPtrConstantExpr>(this)->getResultElementType();
 }
 
+Align GEPOperator::getMaxPreservedAlignment(const DataLayout &DL) const {
+  Align Result =
+      Align(((std::size_t)1) << (sizeof(std::size_t) * CHAR_BIT - 1));
+  for (gep_type_iterator GTI = gep_type_begin(this), GTE = gep_type_end(this);
+       GTI != GTE; ++GTI) {
+    int64_t TypeSize = DL.getTypeStoreSize(GTI.getIndexedType());
+    int64_t ElemCount = 1;
+    if (ConstantInt *OpC = dyn_cast<ConstantInt>(GTI.getOperand()))
+      ElemCount = OpC->getZExtValue();
+    Result = Align(MinAlign(ElemCount * TypeSize, Result.value()));
+  }
+  return Result;
+}
+
 bool GEPOperator::accumulateConstantOffset(const DataLayout &DL,
                                            APInt &Offset) const {
   assert(Offset.getBitWidth() ==
              DL.getIndexSizeInBits(getPointerAddressSpace()) &&
          "The offset bit width does not match DL specification.");
 
   for (gep_type_iterator GTI = gep_type_begin(this), GTE = gep_type_end(this);
        GTI != GTE; ++GTI) {

llvm/lib/Transforms/Utils/AssumeBundleBuilder.cpp

switch (Kind) {
     case Attribute::DereferenceableOrNull:
     case Attribute::Cold:
       return true;
     default:
       return false;
   }
 }
 
+Value *stripGEPandCasts(Value *V, function_ref<bool(Value *)> Func) {
+  if (!V->getType()->isPointerTy())
+    return V;
+
+  // Even though we don't look through PHI nodes, we could be called on an
+  // instruction in an unreachable block, which may be on a cycle.
+  SmallPtrSet<Value *, 4> Visited;
+
+  Visited.insert(V);
+  do {
+    if (!Func(V))
+      return V;
+    if (auto *GEP = dyn_cast<GEPOperator>(V)) {
+      V = GEP->getPointerOperand();
+    } else if (Operator::getOpcode(V) == Instruction::BitCast) {
+      V = cast<Operator>(V)->getOperand(0);
+    } else if (isa<GlobalAlias>(V)) {
+      V = cast<GlobalAlias>(V)->getAliasee();
+    } else {
+      if (const auto *Call = dyn_cast<CallBase>(V)) {
+        if (Value *RV = Call->getReturnedArgOperand()) {
+          V = RV;
+          continue;
+        }
+      }
+      return V;
+    }
+    assert(V->getType()->isPointerTy() && "Unexpected operand type!");
+  } while (Visited.insert(V).second);
+  return V;
+}
+
+/// This function will try to transform the given knowledge into a more
+/// canonical one. If it succeeds it will return that new knowledge otherwise it
+/// will return an emtpy knowledge.
+RetainedKnowledge canonicalizedKnowledge(RetainedKnowledge RK, Module *M) {
+  switch (RK.AttrKind) {
+  default:
+    return RK;
+  case Attribute::NonNull:
+    RK.WasOn = GetUnderlyingObject(RK.WasOn, M->getDataLayout());
+    return RK;
+  case Attribute::Alignment: {
+    Value *V = stripGEPandCasts(RK.WasOn, [&](const Value *Strip) {
+      if (auto *GEP = dyn_cast<GEPOperator>(Strip))
+        RK.ArgValue =
+            MinAlign(RK.ArgValue,
+                     GEP->getMaxPreservedAlignment(M->getDataLayout()).value());
+      return true;
+    });
+    RK.WasOn = V;
+    return RK;
+  }
+  case Attribute::Dereferenceable:
+  case Attribute::DereferenceableOrNull: {
+    int64_t Offset = 0;
+    Value *V = GetPointerBaseWithConstantOffset(
+        RK.WasOn, Offset, M->getDataLayout(), /*AllowNonInBounds*/ false);
+    if (Offset < 0)
+      return RK;
+    RK.ArgValue = RK.ArgValue + Offset;
+    RK.WasOn = V->stripInBoundsOffsets();
+  }
+  }
+  return RK;
+}
+
 /// This class contain all knowledge that have been gather while building an
 /// llvm.assume and the function to manipulate it.
 struct AssumeBuilderState {
   Module *M;
 
   using MapKey = std::pair<Value *, Attribute::AttrKind>;
   SmallMapVector<MapKey, unsigned, 8> AssumedKnowledgeMap;
   Instruction *InsertBeforeInstruction = nullptr;
   AssumptionCache *AC = nullptr;
   DominatorTree *DT = nullptr;
 
   AssumeBuilderState(Module *M, Instruction *I = nullptr,
                      AssumptionCache *AC = nullptr, DominatorTree *DT = nullptr)
       : M(M), InsertBeforeInstruction(I), AC(AC), DT(DT) {}
 
   bool tryToPreserveWithoutAddingAssume(RetainedKnowledge RK) {
-    if (!InsertBeforeInstruction || !AC || !RK.WasOn)
+    if (!InsertBeforeInstruction || !RK.WasOn)
       return false;
     bool HasBeenPreserved = false;
     Use *ToUpdate = nullptr;
     getKnowledgeForValue(
         RK.WasOn, {RK.AttrKind}, AC,
         [&](RetainedKnowledge RKOther, Instruction *Assume,
             const CallInst::BundleOpInfo *Bundle) {
           if (!isValidAssumeForContext(Assume, InsertBeforeInstruction, DT))
getKnowledgeForValue(
           return false;
         });
     if (ToUpdate)
       ToUpdate->set(
           ConstantInt::get(Type::getInt64Ty(M->getContext()), RK.ArgValue));
     return HasBeenPreserved;
   }
 
+  bool isKnowledgeWorthPreserving(RetainedKnowledge RK) {
+    if (!RK)
+      return false;
+    if (RK.WasOn && !isa<Argument>(RK.WasOn) &&
+        (isa<AllocaInst>(RK.WasOn) || RK.WasOn->use_empty() ||
+         (RK.WasOn->hasOneUse() &&
+          RK.WasOn->use_begin()->getUser() == InsertBeforeInstruction)))
+      return false;
+
+    return true;
+  }
+
   void addKnowledge(RetainedKnowledge RK) {
-    if (RK.AttrKind == Attribute::None || tryToPreserveWithoutAddingAssume(RK))
+    RK = canonicalizedKnowledge(RK, M);
+
+    if (!isKnowledgeWorthPreserving(RK))
+      return;
+
+    if (tryToPreserveWithoutAddingAssume(RK))
       return;
     MapKey Key{RK.WasOn, RK.AttrKind};
     auto Lookup = AssumedKnowledgeMap.find(Key);
     if (Lookup == AssumedKnowledgeMap.end()) {
       AssumedKnowledgeMap[Key] = RK.ArgValue;
       return;
     }
     assert(((Lookup->second == 0 && RK.ArgValue == 0) ||

llvm/test/Analysis/BasicAA/featuretest.ll

 ; USE_ASSUME-NEXT:    [[ARRAY22_SUB:%.*]] = getelementptr inbounds [200 x i32], [200 x i32]* [[ARRAY22]], i64 0, i64 0
 ; USE_ASSUME-NEXT:    [[ARRAY11_SUB:%.*]] = getelementptr inbounds [100 x i32], [100 x i32]* [[ARRAY11]], i64 0, i64 0
 ; USE_ASSUME-NEXT:    call void @llvm.assume(i1 true) [ "align"(i32* [[ARRAY11_SUB]], i32 4) ]
 ; USE_ASSUME-NEXT:    call void @external(i32* nonnull [[ARRAY11_SUB]])
 ; USE_ASSUME-NEXT:    call void @external(i32* nonnull [[ARRAY22_SUB]])
 ; USE_ASSUME-NEXT:    [[POINTER:%.*]] = getelementptr [100 x i32], [100 x i32]* [[ARRAY11]], i64 0, i64 [[A:%.*]]
 ; USE_ASSUME-NEXT:    [[POINTER2:%.*]] = getelementptr [200 x i32], [200 x i32]* [[ARRAY22]], i64 0, i64 [[B:%.*]]
 ; USE_ASSUME-NEXT:    store i32 7, i32* [[POINTER2]], align 4
-; USE_ASSUME-NEXT:    call void @llvm.assume(i1 true) [ "dereferenceable"(i32* [[POINTER]], i64 4), "nonnull"(i32* [[POINTER]]) ]
+; USE_ASSUME-NEXT:    call void @llvm.assume(i1 true) [ "dereferenceable"(i32* [[POINTER]], i64 4) ]
 ; USE_ASSUME-NEXT:    ret i32 0
 ;
   %Array1 = alloca i32, i32 100
   %Array2 = alloca i32, i32 200
   call void @llvm.assume(i1 true) ["align"(i32* %Array1, i32 4)]
 
   call void @external(i32* %Array1)
   call void @external(i32* %Array2)

 ;
 ; USE_ASSUME-LABEL: @constant_array_index_test(
 ; USE_ASSUME-NEXT:    [[ARRAY1:%.*]] = alloca [100 x i32], align 4
 ; USE_ASSUME-NEXT:    [[ARRAY1_SUB:%.*]] = getelementptr inbounds [100 x i32], [100 x i32]* [[ARRAY1]], i64 0, i64 0
 ; USE_ASSUME-NEXT:    call void @external(i32* nonnull [[ARRAY1_SUB]])
 ; USE_ASSUME-NEXT:    [[P1:%.*]] = getelementptr inbounds [100 x i32], [100 x i32]* [[ARRAY1]], i64 0, i64 7
 ; USE_ASSUME-NEXT:    [[P2:%.*]] = getelementptr inbounds [100 x i32], [100 x i32]* [[ARRAY1]], i64 0, i64 6
 ; USE_ASSUME-NEXT:    store i32 1, i32* [[P2]], align 4
-; USE_ASSUME-NEXT:    call void @llvm.assume(i1 true) [ "dereferenceable"(i32* [[P1]], i64 4), "nonnull"(i32* [[P1]]) ]
+; USE_ASSUME-NEXT:    call void @llvm.assume(i1 true) [ "dereferenceable"(i32* [[P1]], i64 4) ]
 ; USE_ASSUME-NEXT:    ret i32 0
 ;
   %Array = alloca i32, i32 100
   call void @external(i32* %Array)
 
   %P1 = getelementptr i32, i32* %Array, i64 7
   %P2 = getelementptr i32, i32* %Array, i64 6
 

 ; NO_ASSUME-NEXT:    [[B:%.*]] = getelementptr { i32, i32 }, { i32, i32 }* [[A:%.*]], i64 [[DISTANCE:%.*]], i32 1
 ; NO_ASSUME-NEXT:    store i32 7, i32* [[B]], align 4
 ; NO_ASSUME-NEXT:    ret i32 0
 ;
 ; USE_ASSUME-LABEL: @gep_distance_test2(
 ; USE_ASSUME-NEXT:    [[A1:%.*]] = getelementptr { i32, i32 }, { i32, i32 }* [[A:%.*]], i64 0, i32 0
 ; USE_ASSUME-NEXT:    [[B:%.*]] = getelementptr { i32, i32 }, { i32, i32 }* [[A]], i64 [[DISTANCE:%.*]], i32 1
 ; USE_ASSUME-NEXT:    store i32 7, i32* [[B]], align 4
-; USE_ASSUME-NEXT:    call void @llvm.assume(i1 true) [ "dereferenceable"(i32* [[A1]], i64 4), "nonnull"(i32* [[A1]]) ]
+; USE_ASSUME-NEXT:    call void @llvm.assume(i1 true) [ "dereferenceable"(i32* [[A1]], i64 4), "nonnull"({ i32, i32 }* [[A]]) ]
 ; USE_ASSUME-NEXT:    ret i32 0
 ;
   %A1 = getelementptr {i32,i32}, {i32,i32}* %A, i64 0, i32 0
   %REMOVEu = load i32, i32* %A1
   %B = getelementptr {i32,i32}, {i32,i32}* %A, i64 %distance, i32 1
   store i32 7, i32* %B    ; B cannot alias A, it's at least 4 bytes away
   %REMOVEv = load i32, i32* %A1
   %r = sub i32 %REMOVEu, %REMOVEv
;
   store i8 42, i8* %C
   %Y = load i32, i32* %A
   %R = sub i32 %X, %Y
   ret i32 %R
 }
 
 ; Test that we can disambiguate globals reached through constantexpr geps
 define i32 @constexpr_test() {
-; NO_ASSUME-LABEL: @constexpr_test(
-; NO_ASSUME-NEXT:    [[X:%.*]] = alloca i32, align 4
-; NO_ASSUME-NEXT:    call void @external(i32* nonnull [[X]])
-; NO_ASSUME-NEXT:    store i32 5, i32* getelementptr inbounds ({ i32 }, { i32 }* @Global, i64 0, i32 0), align 4
-; NO_ASSUME-NEXT:    ret i32 0
-;
-; USE_ASSUME-LABEL: @constexpr_test(
-; USE_ASSUME-NEXT:    [[X:%.*]] = alloca i32, align 4
-; USE_ASSUME-NEXT:    call void @external(i32* nonnull [[X]])
-; USE_ASSUME-NEXT:    store i32 5, i32* getelementptr inbounds ({ i32 }, { i32 }* @Global, i64 0, i32 0), align 4
-; USE_ASSUME-NEXT:    call void @llvm.assume(i1 true) [ "dereferenceable"(i32* [[X]], i64 4), "nonnull"(i32* [[X]]) ]
-; USE_ASSUME-NEXT:    ret i32 0
+; CHECK-LABEL: @constexpr_test(
+; CHECK-NEXT:    [[X:%.*]] = alloca i32, align 4
+; CHECK-NEXT:    call void @external(i32* nonnull [[X]])
+; CHECK-NEXT:    store i32 5, i32* getelementptr inbounds ({ i32 }, { i32 }* @Global, i64 0, i32 0), align 4
+; CHECK-NEXT:    ret i32 0
 ;
   %X = alloca i32
   call void @external(i32* %X)
 
   %Y = load i32, i32* %X
   store i32 5, i32* getelementptr ({ i32 }, { i32 }* @Global, i64 0, i32 0)
   %REMOVE = load i32, i32* %X
   %retval = sub i32 %Y, %REMOVE

llvm/test/Transforms/Attributor/ArgumentPromotion/byval-2-preservation.ll

 define i32 @test(i32* %X) {
 ; CHECK-LABEL: @test(
 ; CHECK-NEXT:  entry:
 ; CHECK-NEXT:    [[S:%.*]] = alloca [[STRUCT_SS:%.*]]
 ; CHECK-NEXT:    [[TMP1:%.*]] = getelementptr [[STRUCT_SS]], %struct.ss* [[S]], i32 0, i32 0
 ; CHECK-NEXT:    store i32 1, i32* [[TMP1]], align 8
 ; CHECK-NEXT:    [[TMP4:%.*]] = getelementptr [[STRUCT_SS]], %struct.ss* [[S]], i32 0, i32 1
 ; CHECK-NEXT:    store i64 2, i64* [[TMP4]], align 4
-; CHECK-NEXT:    call void @llvm.assume(i1 true) [ "nonnull"(%struct.ss* [[S]]), "align"(%struct.ss* [[S]], i64 8), "dereferenceable"(%struct.ss* [[S]], i64 12), "align"(i32* [[X:%.*]], i64 4), "nonnull"(i32* [[X]]), "dereferenceable"(i32* [[X]], i64 4) ]
+; CHECK-NEXT:    call void @llvm.assume(i1 true) [ "align"(i32* [[X:%.*]], i64 4), "nonnull"(i32* [[X]]), "dereferenceable"(i32* [[X]], i64 4) ]
 ; CHECK-NEXT:    ret i32 0
 ;
 entry:
   %S = alloca %struct.ss
   %tmp1 = getelementptr %struct.ss, %struct.ss* %S, i32 0, i32 0
   store i32 1, i32* %tmp1, align 8
   %tmp4 = getelementptr %struct.ss, %struct.ss* %S, i32 0, i32 1
   store i64 2, i64* %tmp4, align 4
   call void @f( %struct.ss* byval %S, i32* byval %X)
   ret i32 0
 }

llvm/test/Transforms/InstSimplify/load.ll

 ; NOTE: Assertions have been autogenerated by utils/update_test_checks.py
 ; RUN: opt < %s -instsimplify -S | FileCheck %s  --check-prefixes=CHECK,NO_ASSUME
 ; RUN: opt < %s -instsimplify --enable-knowledge-retention -S | FileCheck %s  --check-prefixes=CHECK,USE_ASSUME
 
 @zeroinit = constant {} zeroinitializer
 @undef = constant {} undef
 
 define i32 @crash_on_zeroinit() {
 ; NO_ASSUME-LABEL: @crash_on_zeroinit(
 ; NO_ASSUME-NEXT:    ret i32 0
 ;
 ; USE_ASSUME-LABEL: @crash_on_zeroinit(
-; USE_ASSUME-NEXT:    call void @llvm.assume(i1 true) [ "dereferenceable"(i32* bitcast ({}* @zeroinit to i32*), i64 4), "nonnull"(i32* bitcast ({}* @zeroinit to i32*)) ]
+; USE_ASSUME-NEXT:    call void @llvm.assume(i1 true) [ "dereferenceable"({}* @zeroinit, i64 4), "nonnull"({}* @zeroinit) ]
 ; USE_ASSUME-NEXT:    ret i32 0
 ;
   %load = load i32, i32* bitcast ({}* @zeroinit to i32*)
   ret i32 %load
 }
 
 define i32 @crash_on_undef() {
 ; NO_ASSUME-LABEL: @crash_on_undef(
 ; NO_ASSUME-NEXT:    ret i32 undef
 ;
 ; USE_ASSUME-LABEL: @crash_on_undef(
-; USE_ASSUME-NEXT:    call void @llvm.assume(i1 true) [ "dereferenceable"(i32* bitcast ({}* @undef to i32*), i64 4), "nonnull"(i32* bitcast ({}* @undef to i32*)) ]
+; USE_ASSUME-NEXT:    call void @llvm.assume(i1 true) [ "dereferenceable"({}* @undef, i64 4), "nonnull"({}* @undef) ]
 ; USE_ASSUME-NEXT:    ret i32 undef
 ;
   %load = load i32, i32* bitcast ({}* @undef to i32*)
   ret i32 %load
 }
 
 @GV = private constant [8 x i32] [i32 42, i32 43, i32 44, i32 45, i32 46, i32 47, i32 48, i32 49]
 
 define <8 x i32> @partial_load() {
 ; NO_ASSUME-LABEL: @partial_load(
 ; NO_ASSUME-NEXT:    ret <8 x i32> <i32 0, i32 42, i32 43, i32 44, i32 45, i32 46, i32 47, i32 48>
 ;
 ; USE_ASSUME-LABEL: @partial_load(
-; USE_ASSUME-NEXT:    call void @llvm.assume(i1 true) [ "dereferenceable"(<8 x i32>* bitcast (i32* getelementptr ([8 x i32], [8 x i32]* @GV, i64 0, i64 -1) to <8 x i32>*), i64 32), "nonnull"(<8 x i32>* bitcast (i32* getelementptr ([8 x i32], [8 x i32]* @GV, i64 0, i64 -1) to <8 x i32>*)) ]
+; USE_ASSUME-NEXT:    call void @llvm.assume(i1 true) [ "dereferenceable"(i32* getelementptr ([8 x i32], [8 x i32]* @GV, i64 0, i64 -1), i64 32), "nonnull"([8 x i32]* @GV) ]
 ; USE_ASSUME-NEXT:    ret <8 x i32> <i32 0, i32 42, i32 43, i32 44, i32 45, i32 46, i32 47, i32 48>
 ;
   %load = load <8 x i32>, <8 x i32>* bitcast (i32* getelementptr ([8 x i32], [8 x i32]* @GV, i64 0, i64 -1) to <8 x i32>*)
   ret <8 x i32> %load
 }
 
 @constvec = internal constant <3 x float> <float 0xBFDA20BC40000000, float 0xBFE6A09EE0000000, float 0x3FE279A760000000>
 
 ; This does an out of bounds load from the global constant
 define <3 x float> @load_vec3() {
 ; NO_ASSUME-LABEL: @load_vec3(
 ; NO_ASSUME-NEXT:    ret <3 x float> undef
 ;
 ; USE_ASSUME-LABEL: @load_vec3(
-; USE_ASSUME-NEXT:    call void @llvm.assume(i1 true) [ "dereferenceable"(<3 x float>* getelementptr inbounds (<3 x float>, <3 x float>* @constvec, i64 1), i64 12), "nonnull"(<3 x float>* getelementptr inbounds (<3 x float>, <3 x float>* @constvec, i64 1)) ]
+; USE_ASSUME-NEXT:    call void @llvm.assume(i1 true) [ "dereferenceable"(<3 x float>* @constvec, i64 28), "nonnull"(<3 x float>* @constvec) ]
 ; USE_ASSUME-NEXT:    ret <3 x float> undef
 ;
   %1 = load <3 x float>, <3 x float>* getelementptr inbounds (<3 x float>, <3 x float>* @constvec, i64 1)
   ret <3 x float> %1
 }

llvm/test/Transforms/LICM/pr23608.ll

 ; NO_ASSUME:       indirectgoto:
 ; NO_ASSUME-NEXT:    indirectbr i8* null, [label [[INDIRECTGOTO]], label %while.cond]
 ;
 ; USE_ASSUME-LABEL: @fn1(
 ; USE_ASSUME-NEXT:  entry:
 ; USE_ASSUME-NEXT:    br label [[INDIRECTGOTO:%.*]]
 ; USE_ASSUME:       while.cond:
 ; USE_ASSUME-NEXT:    [[TMP:%.*]] = load %struct.PyFrameObject*, %struct.PyFrameObject** @a, align 8
-; USE_ASSUME-NEXT:    call void @llvm.assume(i1 true) [ "dereferenceable"(i64* inttoptr (i64 and (i64 ptrtoint (%struct.PyFrameObject** @a to i64), i64 -70368744177665) to i64*), i64 8), "nonnull"(i64* inttoptr (i64 and (i64 ptrtoint (%struct.PyFrameObject** @a to i64), i64 -70368744177665) to i64*)), "align"(i64* inttoptr (i64 and (i64 ptrtoint (%struct.PyFrameObject** @a to i64), i64 -70368744177665) to i64*), i64 8) ]
-; USE_ASSUME-NEXT:    call void @llvm.assume(i1 true) [ "dereferenceable"(i32* inttoptr (i64 add (i64 and (i64 ptrtoint (%struct.PyFrameObject** @a to i64), i64 -70368744177665), i64 35184372088832) to i32*), i64 4), "nonnull"(i32* inttoptr (i64 add (i64 and (i64 ptrtoint (%struct.PyFrameObject** @a to i64), i64 -70368744177665), i64 35184372088832) to i32*)), "align"(i32* inttoptr (i64 add (i64 and (i64 ptrtoint (%struct.PyFrameObject** @a to i64), i64 -70368744177665), i64 35184372088832) to i32*), i64 8) ]
 ; USE_ASSUME-NEXT:    [[F_IBLOCK:%.*]] = getelementptr inbounds [[STRUCT_PYFRAMEOBJECT:%.*]], %struct.PyFrameObject* [[TMP]], i64 0, i32 0
 ; USE_ASSUME-NEXT:    br label [[BB2:%.*]]
 ; USE_ASSUME:       bb:
 ; USE_ASSUME-NEXT:    call void @__msan_warning_noreturn()
 ; USE_ASSUME-NEXT:    unreachable
 ; USE_ASSUME:       bb2:
-; USE_ASSUME-NEXT:    call void @llvm.assume(i1 true) [ "dereferenceable"(i32* [[F_IBLOCK]], i64 4), "nonnull"(i32* [[F_IBLOCK]]), "align"(i32* [[F_IBLOCK]], i64 4) ]
+; USE_ASSUME-NEXT:    call void @llvm.assume(i1 true) [ "dereferenceable"(%struct.PyFrameObject* [[TMP]], i64 4), "nonnull"(%struct.PyFrameObject* [[TMP]]), "align"(%struct.PyFrameObject* [[TMP]], i64 4) ]
 ; USE_ASSUME-NEXT:    [[TMP4:%.*]] = ptrtoint i32* [[F_IBLOCK]] to i64
 ; USE_ASSUME-NEXT:    [[TOBOOL:%.*]] = icmp eq i64 [[TMP4]], 0
 ; USE_ASSUME-NEXT:    br i1 [[TOBOOL]], label [[BB13:%.*]], label [[BB15:%.*]]
 ; USE_ASSUME:       bb13:
 ; USE_ASSUME-NEXT:    [[F_IBLOCK_LCSSA:%.*]] = phi i32* [ [[F_IBLOCK]], [[BB2]] ]
 ; USE_ASSUME-NEXT:    [[TMP4_LE:%.*]] = ptrtoint i32* [[F_IBLOCK_LCSSA]] to i64
 ; USE_ASSUME-NEXT:    [[TMP8_LE:%.*]] = inttoptr i64 [[TMP4_LE]] to i32*
 ; USE_ASSUME-NEXT:    call void @__msan_warning_noreturn()

llvm/test/Transforms/NewGVN/2007-07-26-PhiErasure.ll

 ; NOTE: Assertions have been autogenerated by utils/update_test_checks.py
 ; RUN: opt < %s -newgvn -S | FileCheck %s --check-prefixes=CHECK,NO_ASSUME
 ; RUN: opt < %s -newgvn --enable-knowledge-retention -S | FileCheck %s --check-prefixes=CHECK,USE_ASSUME
 
   %struct..0anon = type { i32 }
   %struct.FILE = type { i8*, i32, i32, i16, i16, %struct.__sbuf, i32, i8*, i32 (i8*)*, i32 (i8*, i8*, i32)*, i64 (i8*, i64, i32)*, i32 (i8*, i8*, i32)*, %struct.__sbuf, %struct.__sFILEX*, i32, [3 x i8], [1 x i8], %struct.__sbuf, i32, i64 }
   %struct.__sFILEX = type opaque
   %struct.__sbuf = type { i8*, i32 }
   %struct.rtx_def = type { i16, i8, i8, [1 x %struct..0anon] }
 @n_spills = external global i32   ; <i32*> [#uses=2]
 
 define i32 @reload(%struct.rtx_def* %first, i32 %global, %struct.FILE* %dumpfile) {
-; NO_ASSUME-LABEL: @reload(
-; NO_ASSUME-NEXT:  cond_next2835.1:
-; NO_ASSUME-NEXT:    br label [[BB2928:%.*]]
-; NO_ASSUME:       bb2928:
-; NO_ASSUME-NEXT:    br i1 false, label [[COND_NEXT2943:%.*]], label [[COND_TRUE2935:%.*]]
-; NO_ASSUME:       cond_true2935:
-; NO_ASSUME-NEXT:    br label [[COND_NEXT2943]]
-; NO_ASSUME:       cond_next2943:
-; NO_ASSUME-NEXT:    br i1 false, label [[BB2982_PREHEADER:%.*]], label [[BB2928]]
-; NO_ASSUME:       bb2982.preheader:
-; NO_ASSUME-NEXT:    store i8 undef, i8* null
-; NO_ASSUME-NEXT:    ret i32 undef
-;
-; USE_ASSUME-LABEL: @reload(
-; USE_ASSUME-NEXT:  cond_next2835.1:
-; USE_ASSUME-NEXT:    br label [[BB2928:%.*]]
-; USE_ASSUME:       bb2928:
-; USE_ASSUME-NEXT:    br i1 false, label [[COND_NEXT2943:%.*]], label [[COND_TRUE2935:%.*]]
-; USE_ASSUME:       cond_true2935:
-; USE_ASSUME-NEXT:    br label [[COND_NEXT2943]]
-; USE_ASSUME:       cond_next2943:
-; USE_ASSUME-NEXT:    br i1 false, label [[BB2982_PREHEADER:%.*]], label [[BB2928]]
-; USE_ASSUME:       bb2982.preheader:
-; USE_ASSUME-NEXT:    call void @llvm.assume(i1 true) [ "dereferenceable"(i32* @n_spills, i64 4), "nonnull"(i32* @n_spills), "align"(i32* @n_spills, i64 4) ]
-; USE_ASSUME-NEXT:    store i8 undef, i8* null
-; USE_ASSUME-NEXT:    ret i32 undef
+; CHECK-LABEL: @reload(
+; CHECK-NEXT:  cond_next2835.1:
+; CHECK-NEXT:    br label [[BB2928:%.*]]
+; CHECK:       bb2928:
+; CHECK-NEXT:    br i1 false, label [[COND_NEXT2943:%.*]], label [[COND_TRUE2935:%.*]]
+; CHECK:       cond_true2935:
+; CHECK-NEXT:    br label [[COND_NEXT2943]]
+; CHECK:       cond_next2943:
+; CHECK-NEXT:    br i1 false, label [[BB2982_PREHEADER:%.*]], label [[BB2928]]
+; CHECK:       bb2982.preheader:
+; CHECK-NEXT:    store i8 undef, i8* null
+; CHECK-NEXT:    ret i32 undef
 ;
 cond_next2835.1:    ; preds = %cond_next2861
   %tmp2922 = load i32, i32* @n_spills, align 4    ; <i32> [#uses=0]
   br label %bb2928
 
 bb2928:   ; preds = %cond_next2835.1, %cond_next2943
   br i1 false, label %cond_next2943, label %cond_true2935
 

llvm/test/Transforms/Util/assume-builder.ll

 declare void @func_argattr(i32* align 8, i32* nonnull) nounwind
 declare void @may_throw()
 
 define void @test(i32* %P, i32* %P1, i32* %P2, i32* %P3) {
 ; BASIC-LABEL: define {{[^@]+}}@test
 ; BASIC-SAME: (i32* [[P:%.*]], i32* [[P1:%.*]], i32* [[P2:%.*]], i32* [[P3:%.*]])
 ; BASIC-NEXT:    call void @llvm.assume(i1 true) [ "nonnull"(i32* [[P]]), "dereferenceable"(i32* [[P]], i64 16) ]
 ; BASIC-NEXT:    call void @func(i32* nonnull dereferenceable(16) [[P]], i32* null)
-; BASIC-NEXT:    call void @llvm.assume(i1 true) [ "dereferenceable"(i32* [[P1]], i64 12), "nonnull"(i32* [[P]]) ]
+; BASIC-NEXT:    call void @llvm.assume(i1 true) [ "dereferenceable"(i32* [[P1]], i64 12) ]
 ; BASIC-NEXT:    call void @func(i32* dereferenceable(12) [[P1]], i32* nonnull [[P]])
-; BASIC-NEXT:    call void @llvm.assume(i1 true) [ "dereferenceable"(i32* [[P1]], i64 12), "cold"() ]
+; BASIC-NEXT:    call void @llvm.assume(i1 true) [ "cold"() ]
 ; BASIC-NEXT:    call void @func_cold(i32* dereferenceable(12) [[P1]]) #6
-; BASIC-NEXT:    call void @llvm.assume(i1 true) [ "dereferenceable"(i32* [[P1]], i64 12), "cold"() ]
+; BASIC-NEXT:    call void @llvm.assume(i1 true) [ "cold"() ]
 ; BASIC-NEXT:    call void @func_cold(i32* dereferenceable(12) [[P1]])
 ; BASIC-NEXT:    call void @func(i32* [[P1]], i32* [[P]])
 ; BASIC-NEXT:    call void @func_strbool(i32* [[P1]])
 ; BASIC-NEXT:    call void @llvm.assume(i1 true) [ "dereferenceable"(i32* [[P]], i64 32) ]
 ; BASIC-NEXT:    call void @func(i32* dereferenceable(32) [[P]], i32* dereferenceable(8) [[P]])
 ; BASIC-NEXT:    call void @llvm.assume(i1 true) [ "align"(i32* [[P1]], i64 8) ]
 ; BASIC-NEXT:    call void @func_many(i32* align 8 [[P1]])
 ; BASIC-NEXT:    call void @llvm.assume(i1 true) [ "align"(i32* [[P2]], i64 8), "nonnull"(i32* [[P3]]) ]
 ; BASIC-NEXT:    call void @func_argattr(i32* [[P2]], i32* [[P3]])
-; BASIC-NEXT:    call void @llvm.assume(i1 true) [ "nonnull"(i32* [[P1]]), "nonnull"(i32* [[P]]) ]
+; BASIC-NEXT:    call void @llvm.assume(i1 true) [ "nonnull"(i32* [[P1]]) ]
 ; BASIC-NEXT:    call void @func(i32* nonnull [[P1]], i32* nonnull [[P]])
 ; BASIC-NEXT:    ret void
 ;
 ; ALL-LABEL: define {{[^@]+}}@test
 ; ALL-SAME: (i32* [[P:%.*]], i32* [[P1:%.*]], i32* [[P2:%.*]], i32* [[P3:%.*]])
 ; ALL-NEXT:    call void @llvm.assume(i1 true) [ "nonnull"(i32* [[P]]), "dereferenceable"(i32* [[P]], i64 16) ]
 ; ALL-NEXT:    call void @func(i32* nonnull dereferenceable(16) [[P]], i32* null)
-; ALL-NEXT:    call void @llvm.assume(i1 true) [ "dereferenceable"(i32* [[P1]], i64 12), "nonnull"(i32* [[P]]) ]
+; ALL-NEXT:    call void @llvm.assume(i1 true) [ "dereferenceable"(i32* [[P1]], i64 12) ]
 ; ALL-NEXT:    call void @func(i32* dereferenceable(12) [[P1]], i32* nonnull [[P]])
-; ALL-NEXT:    call void @llvm.assume(i1 true) [ "dereferenceable"(i32* [[P1]], i64 12), "cold"(), "nounwind"(), "willreturn"() ]
+; ALL-NEXT:    call void @llvm.assume(i1 true) [ "cold"(), "nounwind"(), "willreturn"() ]
 ; ALL-NEXT:    call void @func_cold(i32* dereferenceable(12) [[P1]]) #6
-; ALL-NEXT:    call void @llvm.assume(i1 true) [ "dereferenceable"(i32* [[P1]], i64 12), "cold"(), "nounwind"(), "willreturn"() ]
+; ALL-NEXT:    call void @llvm.assume(i1 true) [ "cold"(), "nounwind"(), "willreturn"() ]
 ; ALL-NEXT:    call void @func_cold(i32* dereferenceable(12) [[P1]])
 ; ALL-NEXT:    call void @func(i32* [[P1]], i32* [[P]])
 ; ALL-NEXT:    call void @func_strbool(i32* [[P1]])
 ; ALL-NEXT:    call void @llvm.assume(i1 true) [ "dereferenceable"(i32* [[P]], i64 32) ]
 ; ALL-NEXT:    call void @func(i32* dereferenceable(32) [[P]], i32* dereferenceable(8) [[P]])
 ; ALL-NEXT:    call void @llvm.assume(i1 true) [ "align"(i32* [[P1]], i64 8), "norecurse"(), "nounwind"(), "willreturn"() ]
 ; ALL-NEXT:    call void @func_many(i32* align 8 [[P1]])
 ; ALL-NEXT:    call void @llvm.assume(i1 true) [ "align"(i32* [[P2]], i64 8), "nonnull"(i32* [[P3]]), "nounwind"() ]
 ; ALL-NEXT:    call void @func_argattr(i32* [[P2]], i32* [[P3]])
-; ALL-NEXT:    call void @llvm.assume(i1 true) [ "nonnull"(i32* [[P1]]), "nonnull"(i32* [[P]]) ]
+; ALL-NEXT:    call void @llvm.assume(i1 true) [ "nonnull"(i32* [[P1]]) ]
 ; ALL-NEXT:    call void @func(i32* nonnull [[P1]], i32* nonnull [[P]])
 ; ALL-NEXT:    ret void
 ;
 ; WITH-AC-LABEL: define {{[^@]+}}@test
 ; WITH-AC-SAME: (i32* [[P:%.*]], i32* [[P1:%.*]], i32* [[P2:%.*]], i32* [[P3:%.*]])
 ; WITH-AC-NEXT:    call void @llvm.assume(i1 true) [ "nonnull"(i32* [[P]]), "dereferenceable"(i32* [[P]], i64 16) ]
 ; WITH-AC-NEXT:    call void @func(i32* nonnull dereferenceable(16) [[P]], i32* null)
 ; WITH-AC-NEXT:    call void @llvm.assume(i1 true) [ "dereferenceable"(i32* [[P1]], i64 12) ]

 ; CROSS-BLOCK-NEXT:    call void @func_argattr(i32* [[P2]], i32* [[P3]])
 ; CROSS-BLOCK-NEXT:    call void @llvm.assume(i1 true) [ "nonnull"(i32* [[P1]]) ]
 ; CROSS-BLOCK-NEXT:    call void @func(i32* nonnull [[P1]], i32* nonnull [[P]])
 ; CROSS-BLOCK-NEXT:    ret void
 ;
 ; FULL-SIMPLIFY-LABEL: define {{[^@]+}}@test
 ; FULL-SIMPLIFY-SAME: (i32* nonnull dereferenceable(16) [[P:%.*]], i32* [[P1:%.*]], i32* [[P2:%.*]], i32* [[P3:%.*]])
 ; FULL-SIMPLIFY-NEXT:    call void @func(i32* nonnull dereferenceable(16) [[P]], i32* null)
-; FULL-SIMPLIFY-NEXT:    call void @llvm.assume(i1 true) [ "dereferenceable"(i32* [[P1]], i64 12), "ignore"(i32* undef) ]
+; FULL-SIMPLIFY-NEXT:    call void @llvm.assume(i1 true) [ "dereferenceable"(i32* [[P1]], i64 12) ]
 ; FULL-SIMPLIFY-NEXT:    call void @func(i32* dereferenceable(12) [[P1]], i32* nonnull [[P]])
-; FULL-SIMPLIFY-NEXT:    call void @llvm.assume(i1 true) [ "ignore"(i32* undef, i64 12), "cold"() ]
+; FULL-SIMPLIFY-NEXT:    call void @llvm.assume(i1 true) [ "cold"() ]
 ; FULL-SIMPLIFY-NEXT:    call void @func_cold(i32* dereferenceable(12) [[P1]]) #6
 ; FULL-SIMPLIFY-NEXT:    call void @func_cold(i32* dereferenceable(12) [[P1]])
 ; FULL-SIMPLIFY-NEXT:    call void @func(i32* [[P1]], i32* [[P]])
 ; FULL-SIMPLIFY-NEXT:    call void @func_strbool(i32* [[P1]])
 ; FULL-SIMPLIFY-NEXT:    call void @llvm.assume(i1 true) [ "dereferenceable"(i32* [[P]], i64 32) ]
 ; FULL-SIMPLIFY-NEXT:    call void @func(i32* dereferenceable(32) [[P]], i32* dereferenceable(8) [[P]])
 ; FULL-SIMPLIFY-NEXT:    call void @llvm.assume(i1 true) [ "align"(i32* [[P1]], i64 8), "align"(i32* [[P2]], i64 8), "nonnull"(i32* [[P3]]) ]
 ; FULL-SIMPLIFY-NEXT:    call void @func_many(i32* align 8 [[P1]])
 ; FULL-SIMPLIFY-NEXT:    call void @func_argattr(i32* [[P2]], i32* [[P3]])
-; FULL-SIMPLIFY-NEXT:    call void @llvm.assume(i1 true) [ "nonnull"(i32* [[P1]]), "ignore"(i32* undef) ]
+; FULL-SIMPLIFY-NEXT:    call void @llvm.assume(i1 true) [ "nonnull"(i32* [[P1]]) ]
 ; FULL-SIMPLIFY-NEXT:    call void @func(i32* nonnull [[P1]], i32* nonnull [[P]])
 ; FULL-SIMPLIFY-NEXT:    ret void
 ;
   call void @func(i32* nonnull dereferenceable(16) %P, i32* null)
   call void @func(i32* dereferenceable(12) %P1, i32* nonnull %P)
   call void @func_cold(i32* dereferenceable(12) %P1) cold
   call void @func_cold(i32* dereferenceable(12) %P1)
   call void @func(i32* %P1, i32* %P)

 
 define i32 @test2(%struct.S* %0, i32* %1, i8* %2) {
 ; BASIC-LABEL: define {{[^@]+}}@test2
 ; BASIC-SAME: (%struct.S* [[TMP0:%.*]], i32* [[TMP1:%.*]], i8* [[TMP2:%.*]])
 ; BASIC-NEXT:    [[TMP4:%.*]] = alloca %struct.S*, align 8
 ; BASIC-NEXT:    [[TMP5:%.*]] = alloca i32*, align 8
 ; BASIC-NEXT:    [[TMP6:%.*]] = alloca i8*, align 8
 ; BASIC-NEXT:    [[TMP7:%.*]] = alloca [[STRUCT_S:%.*]], align 8
-; BASIC-NEXT:    call void @llvm.assume(i1 true) [ "dereferenceable"(%struct.S** [[TMP4]], i64 8), "nonnull"(%struct.S** [[TMP4]]), "align"(%struct.S** [[TMP4]], i64 8) ]
 ; BASIC-NEXT:    store %struct.S* [[TMP0]], %struct.S** [[TMP4]], align 8
-; BASIC-NEXT:    call void @llvm.assume(i1 true) [ "dereferenceable"(i32** [[TMP5]], i64 8), "nonnull"(i32** [[TMP5]]), "align"(i32** [[TMP5]], i64 8) ]
 ; BASIC-NEXT:    store i32* [[TMP1]], i32** [[TMP5]], align 8
-; BASIC-NEXT:    call void @llvm.assume(i1 true) [ "dereferenceable"(i8** [[TMP6]], i64 8), "nonnull"(i8** [[TMP6]]) ]
 ; BASIC-NEXT:    store i8* [[TMP2]], i8** [[TMP6]]
-; BASIC-NEXT:    call void @llvm.assume(i1 true) [ "dereferenceable"(i32** [[TMP5]], i64 8), "nonnull"(i32** [[TMP5]]), "align"(i32** [[TMP5]], i64 8) ]
 ; BASIC-NEXT:    [[TMP8:%.*]] = load i32*, i32** [[TMP5]], align 8
-; BASIC-NEXT:    call void @llvm.assume(i1 true) [ "dereferenceable"(i32* [[TMP8]], i64 4), "nonnull"(i32* [[TMP8]]), "align"(i32* [[TMP8]], i64 4) ]
 ; BASIC-NEXT:    [[TMP9:%.*]] = load i32, i32* [[TMP8]], align 4
 ; BASIC-NEXT:    [[TMP10:%.*]] = trunc i32 [[TMP9]] to i8
-; BASIC-NEXT:    call void @llvm.assume(i1 true) [ "dereferenceable"(i8** [[TMP6]], i64 8), "nonnull"(i8** [[TMP6]]), "align"(i8** [[TMP6]], i64 8) ]
 ; BASIC-NEXT:    [[TMP11:%.*]] = load i8*, i8** [[TMP6]], align 8
-; BASIC-NEXT:    call void @llvm.assume(i1 true) [ "dereferenceable"(i8* [[TMP11]], i64 1), "nonnull"(i8* [[TMP11]]) ]
 ; BASIC-NEXT:    store i8 [[TMP10]], i8* [[TMP11]], align 1
 ; BASIC-NEXT:    [[TMP12:%.*]] = bitcast %struct.S* [[TMP7]] to i8*
-; BASIC-NEXT:    call void @llvm.assume(i1 true) [ "dereferenceable"(%struct.S** [[TMP4]], i64 8), "nonnull"(%struct.S** [[TMP4]]) ]
 ; BASIC-NEXT:    [[TMP13:%.*]] = load %struct.S*, %struct.S** [[TMP4]]
 ; BASIC-NEXT:    [[TMP14:%.*]] = bitcast %struct.S* [[TMP13]] to i8*
 ; BASIC-NEXT:    [[TMP15:%.*]] = bitcast %struct.S* [[TMP7]] to i8*
-; BASIC-NEXT:    call void @llvm.assume(i1 true) [ "dereferenceable"(%struct.S** [[TMP4]], i64 8), "nonnull"(%struct.S** [[TMP4]]), "align"(%struct.S** [[TMP4]], i64 8) ]
 ; BASIC-NEXT:    [[TMP16:%.*]] = load %struct.S*, %struct.S** [[TMP4]], align 8
 ; BASIC-NEXT:    [[TMP17:%.*]] = getelementptr inbounds [[STRUCT_S]], %struct.S* [[TMP16]], i32 0, i32 0
-; BASIC-NEXT:    call void @llvm.assume(i1 true) [ "dereferenceable"(i32* [[TMP17]], i64 4), "nonnull"(i32* [[TMP17]]), "align"(i32* [[TMP17]], i64 8) ]
+; BASIC-NEXT:    call void @llvm.assume(i1 true) [ "dereferenceable"(%struct.S* [[TMP16]], i64 4), "nonnull"(%struct.S* [[TMP16]]), "align"(%struct.S* [[TMP16]], i64 8) ]
 ; BASIC-NEXT:    [[TMP18:%.*]] = load i32, i32* [[TMP17]], align 8
-; BASIC-NEXT:    call void @llvm.assume(i1 true) [ "dereferenceable"(%struct.S** [[TMP4]], i64 8), "nonnull"(%struct.S** [[TMP4]]), "align"(%struct.S** [[TMP4]], i64 8) ]
 ; BASIC-NEXT:    [[TMP19:%.*]] = load %struct.S*, %struct.S** [[TMP4]], align 8
 ; BASIC-NEXT:    [[TMP20:%.*]] = getelementptr inbounds [[STRUCT_S]], %struct.S* [[TMP19]], i32 0, i32 1
-; BASIC-NEXT:    call void @llvm.assume(i1 true) [ "dereferenceable"(i8* [[TMP20]], i64 1), "nonnull"(i8* [[TMP20]]), "align"(i8* [[TMP20]], i64 4) ]
+; BASIC-NEXT:    call void @llvm.assume(i1 true) [ "dereferenceable"(%struct.S* [[TMP19]], i64 5), "nonnull"(%struct.S* [[TMP19]]), "align"(%struct.S* [[TMP19]], i64 1) ]
 ; BASIC-NEXT:    [[TMP21:%.*]] = load i8, i8* [[TMP20]], align 4
 ; BASIC-NEXT:    [[TMP22:%.*]] = sext i8 [[TMP21]] to i32
 ; BASIC-NEXT:    [[TMP23:%.*]] = add nsw i32 [[TMP18]], [[TMP22]]
-; BASIC-NEXT:    call void @llvm.assume(i1 true) [ "dereferenceable"(%struct.S** [[TMP4]], i64 8), "nonnull"(%struct.S** [[TMP4]]), "align"(%struct.S** [[TMP4]], i64 8) ]
 ; BASIC-NEXT:    [[TMP24:%.*]] = load %struct.S*, %struct.S** [[TMP4]], align 8
 ; BASIC-NEXT:    [[TMP25:%.*]] = getelementptr inbounds [[STRUCT_S]], %struct.S* [[TMP24]], i32 0, i32 2
-; BASIC-NEXT:    call void @llvm.assume(i1 true) [ "dereferenceable"(i32** [[TMP25]], i64 8), "nonnull"(i32** [[TMP25]]), "align"(i32** [[TMP25]], i64 8) ]
+; BASIC-NEXT:    call void @llvm.assume(i1 true) [ "dereferenceable"(%struct.S* [[TMP24]], i64 16), "nonnull"(%struct.S* [[TMP24]]), "align"(%struct.S* [[TMP24]], i64 8) ]
 ; BASIC-NEXT:    [[TMP26:%.*]] = load i32*, i32** [[TMP25]], align 8
-; BASIC-NEXT:    call void @llvm.assume(i1 true) [ "dereferenceable"(i32* [[TMP26]], i64 4), "nonnull"(i32* [[TMP26]]), "align"(i32* [[TMP26]], i64 4) ]
 ; BASIC-NEXT:    [[TMP27:%.*]] = load i32, i32* [[TMP26]], align 4
 ; BASIC-NEXT:    [[TMP28:%.*]] = add nsw i32 [[TMP23]], [[TMP27]]
 ; BASIC-NEXT:    ret i32 [[TMP28]]
 ;
 ; ALL-LABEL: define {{[^@]+}}@test2
 ; ALL-SAME: (%struct.S* [[TMP0:%.*]], i32* [[TMP1:%.*]], i8* [[TMP2:%.*]])
 ; ALL-NEXT:    [[TMP4:%.*]] = alloca %struct.S*, align 8
 ; ALL-NEXT:    [[TMP5:%.*]] = alloca i32*, align 8
 ; ALL-NEXT:    [[TMP6:%.*]] = alloca i8*, align 8
 ; ALL-NEXT:    [[TMP7:%.*]] = alloca [[STRUCT_S:%.*]], align 8
-; ALL-NEXT:    call void @llvm.assume(i1 true) [ "dereferenceable"(%struct.S** [[TMP4]], i64 8), "nonnull"(%struct.S** [[TMP4]]), "align"(%struct.S** [[TMP4]], i64 8) ]
 ; ALL-NEXT:    store %struct.S* [[TMP0]], %struct.S** [[TMP4]], align 8
-; ALL-NEXT:    call void @llvm.assume(i1 true) [ "dereferenceable"(i32** [[TMP5]], i64 8), "nonnull"(i32** [[TMP5]]), "align"(i32** [[TMP5]], i64 8) ]
 ; ALL-NEXT:    store i32* [[TMP1]], i32** [[TMP5]], align 8
-; ALL-NEXT:    call void @llvm.assume(i1 true) [ "dereferenceable"(i8** [[TMP6]], i64 8), "nonnull"(i8** [[TMP6]]) ]
 ; ALL-NEXT:    store i8* [[TMP2]], i8** [[TMP6]]
-; ALL-NEXT:    call void @llvm.assume(i1 true) [ "dereferenceable"(i32** [[TMP5]], i64 8), "nonnull"(i32** [[TMP5]]), "align"(i32** [[TMP5]], i64 8) ]
 ; ALL-NEXT:    [[TMP8:%.*]] = load i32*, i32** [[TMP5]], align 8
-; ALL-NEXT:    call void @llvm.assume(i1 true) [ "dereferenceable"(i32* [[TMP8]], i64 4), "nonnull"(i32* [[TMP8]]), "align"(i32* [[TMP8]], i64 4) ]
 ; ALL-NEXT:    [[TMP9:%.*]] = load i32, i32* [[TMP8]], align 4
 ; ALL-NEXT:    [[TMP10:%.*]] = trunc i32 [[TMP9]] to i8
-; ALL-NEXT:    call void @llvm.assume(i1 true) [ "dereferenceable"(i8** [[TMP6]], i64 8), "nonnull"(i8** [[TMP6]]), "align"(i8** [[TMP6]], i64 8) ]
 ; ALL-NEXT:    [[TMP11:%.*]] = load i8*, i8** [[TMP6]], align 8
-; ALL-NEXT:    call void @llvm.assume(i1 true) [ "dereferenceable"(i8* [[TMP11]], i64 1), "nonnull"(i8* [[TMP11]]) ]
 ; ALL-NEXT:    store i8 [[TMP10]], i8* [[TMP11]], align 1
 ; ALL-NEXT:    [[TMP12:%.*]] = bitcast %struct.S* [[TMP7]] to i8*
-; ALL-NEXT:    call void @llvm.assume(i1 true) [ "dereferenceable"(%struct.S** [[TMP4]], i64 8), "nonnull"(%struct.S** [[TMP4]]) ]
 ; ALL-NEXT:    [[TMP13:%.*]] = load %struct.S*, %struct.S** [[TMP4]]
 ; ALL-NEXT:    [[TMP14:%.*]] = bitcast %struct.S* [[TMP13]] to i8*
 ; ALL-NEXT:    [[TMP15:%.*]] = bitcast %struct.S* [[TMP7]] to i8*
-; ALL-NEXT:    call void @llvm.assume(i1 true) [ "dereferenceable"(%struct.S** [[TMP4]], i64 8), "nonnull"(%struct.S** [[TMP4]]), "align"(%struct.S** [[TMP4]], i64 8) ]
 ; ALL-NEXT:    [[TMP16:%.*]] = load %struct.S*, %struct.S** [[TMP4]], align 8
 ; ALL-NEXT:    [[TMP17:%.*]] = getelementptr inbounds [[STRUCT_S]], %struct.S* [[TMP16]], i32 0, i32 0
-; ALL-NEXT:    call void @llvm.assume(i1 true) [ "dereferenceable"(i32* [[TMP17]], i64 4), "nonnull"(i32* [[TMP17]]), "align"(i32* [[TMP17]], i64 8) ]
+; ALL-NEXT:    call void @llvm.assume(i1 true) [ "dereferenceable"(%struct.S* [[TMP16]], i64 4), "nonnull"(%struct.S* [[TMP16]]), "align"(%struct.S* [[TMP16]], i64 8) ]
 ; ALL-NEXT:    [[TMP18:%.*]] = load i32, i32* [[TMP17]], align 8
-; ALL-NEXT:    call void @llvm.assume(i1 true) [ "dereferenceable"(%struct.S** [[TMP4]], i64 8), "nonnull"(%struct.S** [[TMP4]]), "align"(%struct.S** [[TMP4]], i64 8) ]
 ; ALL-NEXT:    [[TMP19:%.*]] = load %struct.S*, %struct.S** [[TMP4]], align 8
 ; ALL-NEXT:    [[TMP20:%.*]] = getelementptr inbounds [[STRUCT_S]], %struct.S* [[TMP19]], i32 0, i32 1
-; ALL-NEXT:    call void @llvm.assume(i1 true) [ "dereferenceable"(i8* [[TMP20]], i64 1), "nonnull"(i8* [[TMP20]]), "align"(i8* [[TMP20]], i64 4) ]
+; ALL-NEXT:    call void @llvm.assume(i1 true) [ "dereferenceable"(%struct.S* [[TMP19]], i64 5), "nonnull"(%struct.S* [[TMP19]]), "align"(%struct.S* [[TMP19]], i64 1) ]
 ; ALL-NEXT:    [[TMP21:%.*]] = load i8, i8* [[TMP20]], align 4
 ; ALL-NEXT:    [[TMP22:%.*]] = sext i8 [[TMP21]] to i32
 ; ALL-NEXT:    [[TMP23:%.*]] = add nsw i32 [[TMP18]], [[TMP22]]
-; ALL-NEXT:    call void @llvm.assume(i1 true) [ "dereferenceable"(%struct.S** [[TMP4]], i64 8), "nonnull"(%struct.S** [[TMP4]]), "align"(%struct.S** [[TMP4]], i64 8) ]
 ; ALL-NEXT:    [[TMP24:%.*]] = load %struct.S*, %struct.S** [[TMP4]], align 8
 ; ALL-NEXT:    [[TMP25:%.*]] = getelementptr inbounds [[STRUCT_S]], %struct.S* [[TMP24]], i32 0, i32 2
-; ALL-NEXT:    call void @llvm.assume(i1 true) [ "dereferenceable"(i32** [[TMP25]], i64 8), "nonnull"(i32** [[TMP25]]), "align"(i32** [[TMP25]], i64 8) ]
+; ALL-NEXT:    call void @llvm.assume(i1 true) [ "dereferenceable"(%struct.S* [[TMP24]], i64 16), "nonnull"(%struct.S* [[TMP24]]), "align"(%struct.S* [[TMP24]], i64 8) ]
 ; ALL-NEXT:    [[TMP26:%.*]] = load i32*, i32** [[TMP25]], align 8
-; ALL-NEXT:    call void @llvm.assume(i1 true) [ "dereferenceable"(i32* [[TMP26]], i64 4), "nonnull"(i32* [[TMP26]]), "align"(i32* [[TMP26]], i64 4) ]
 ; ALL-NEXT:    [[TMP27:%.*]] = load i32, i32* [[TMP26]], align 4
 ; ALL-NEXT:    [[TMP28:%.*]] = add nsw i32 [[TMP23]], [[TMP27]]
 ; ALL-NEXT:    ret i32 [[TMP28]]
 ;
 ; WITH-AC-LABEL: define {{[^@]+}}@test2
 ; WITH-AC-SAME: (%struct.S* [[TMP0:%.*]], i32* [[TMP1:%.*]], i8* [[TMP2:%.*]])
 ; WITH-AC-NEXT:    [[TMP4:%.*]] = alloca %struct.S*, align 8
 ; WITH-AC-NEXT:    [[TMP5:%.*]] = alloca i32*, align 8
 ; WITH-AC-NEXT:    [[TMP6:%.*]] = alloca i8*, align 8
 ; WITH-AC-NEXT:    [[TMP7:%.*]] = alloca [[STRUCT_S:%.*]], align 8
-; WITH-AC-NEXT:    call void @llvm.assume(i1 true) [ "dereferenceable"(%struct.S** [[TMP4]], i64 8), "nonnull"(%struct.S** [[TMP4]]), "align"(%struct.S** [[TMP4]], i64 8) ]
 ; WITH-AC-NEXT:    store %struct.S* [[TMP0]], %struct.S** [[TMP4]], align 8
-; WITH-AC-NEXT:    call void @llvm.assume(i1 true) [ "dereferenceable"(i32** [[TMP5]], i64 8), "nonnull"(i32** [[TMP5]]), "align"(i32** [[TMP5]], i64 8) ]
 ; WITH-AC-NEXT:    store i32* [[TMP1]], i32** [[TMP5]], align 8
-; WITH-AC-NEXT:    call void @llvm.assume(i1 true) [ "dereferenceable"(i8** [[TMP6]], i64 8), "nonnull"(i8** [[TMP6]]) ]
 ; WITH-AC-NEXT:    store i8* [[TMP2]], i8** [[TMP6]]
 ; WITH-AC-NEXT:    [[TMP8:%.*]] = load i32*, i32** [[TMP5]], align 8
-; WITH-AC-NEXT:    call void @llvm.assume(i1 true) [ "dereferenceable"(i32* [[TMP8]], i64 4), "nonnull"(i32* [[TMP8]]), "align"(i32* [[TMP8]], i64 4) ]
 ; WITH-AC-NEXT:    [[TMP9:%.*]] = load i32, i32* [[TMP8]], align 4
 ; WITH-AC-NEXT:    [[TMP10:%.*]] = trunc i32 [[TMP9]] to i8
-; WITH-AC-NEXT:    call void @llvm.assume(i1 true) [ "align"(i8** [[TMP6]], i64 8) ]
 ; WITH-AC-NEXT:    [[TMP11:%.*]] = load i8*, i8** [[TMP6]], align 8
-; WITH-AC-NEXT:    call void @llvm.assume(i1 true) [ "dereferenceable"(i8* [[TMP11]], i64 1), "nonnull"(i8* [[TMP11]]) ]
 ; WITH-AC-NEXT:    store i8 [[TMP10]], i8* [[TMP11]], align 1
 ; WITH-AC-NEXT:    [[TMP12:%.*]] = bitcast %struct.S* [[TMP7]] to i8*
 ; WITH-AC-NEXT:    [[TMP13:%.*]] = load %struct.S*, %struct.S** [[TMP4]]
 ; WITH-AC-NEXT:    [[TMP14:%.*]] = bitcast %struct.S* [[TMP13]] to i8*
 ; WITH-AC-NEXT:    [[TMP15:%.*]] = bitcast %struct.S* [[TMP7]] to i8*
 ; WITH-AC-NEXT:    [[TMP16:%.*]] = load %struct.S*, %struct.S** [[TMP4]], align 8
 ; WITH-AC-NEXT:    [[TMP17:%.*]] = getelementptr inbounds [[STRUCT_S]], %struct.S* [[TMP16]], i32 0, i32 0
-; WITH-AC-NEXT:    call void @llvm.assume(i1 true) [ "dereferenceable"(i32* [[TMP17]], i64 4), "nonnull"(i32* [[TMP17]]), "align"(i32* [[TMP17]], i64 8) ]
+; WITH-AC-NEXT:    call void @llvm.assume(i1 true) [ "dereferenceable"(%struct.S* [[TMP16]], i64 4), "nonnull"(%struct.S* [[TMP16]]), "align"(%struct.S* [[TMP16]], i64 8) ]
 ; WITH-AC-NEXT:    [[TMP18:%.*]] = load i32, i32* [[TMP17]], align 8
 ; WITH-AC-NEXT:    [[TMP19:%.*]] = load %struct.S*, %struct.S** [[TMP4]], align 8
 ; WITH-AC-NEXT:    [[TMP20:%.*]] = getelementptr inbounds [[STRUCT_S]], %struct.S* [[TMP19]], i32 0, i32 1
-; WITH-AC-NEXT:    call void @llvm.assume(i1 true) [ "dereferenceable"(i8* [[TMP20]], i64 1), "nonnull"(i8* [[TMP20]]), "align"(i8* [[TMP20]], i64 4) ]
+; WITH-AC-NEXT:    call void @llvm.assume(i1 true) [ "dereferenceable"(%struct.S* [[TMP19]], i64 5), "nonnull"(%struct.S* [[TMP19]]), "align"(%struct.S* [[TMP19]], i64 1) ]
 ; WITH-AC-NEXT:    [[TMP21:%.*]] = load i8, i8* [[TMP20]], align 4
 ; WITH-AC-NEXT:    [[TMP22:%.*]] = sext i8 [[TMP21]] to i32
 ; WITH-AC-NEXT:    [[TMP23:%.*]] = add nsw i32 [[TMP18]], [[TMP22]]
 ; WITH-AC-NEXT:    [[TMP24:%.*]] = load %struct.S*, %struct.S** [[TMP4]], align 8
 ; WITH-AC-NEXT:    [[TMP25:%.*]] = getelementptr inbounds [[STRUCT_S]], %struct.S* [[TMP24]], i32 0, i32 2
-; WITH-AC-NEXT:    call void @llvm.assume(i1 true) [ "dereferenceable"(i32** [[TMP25]], i64 8), "nonnull"(i32** [[TMP25]]), "align"(i32** [[TMP25]], i64 8) ]
+; WITH-AC-NEXT:    call void @llvm.assume(i1 true) [ "dereferenceable"(%struct.S* [[TMP24]], i64 16), "nonnull"(%struct.S* [[TMP24]]), "align"(%struct.S* [[TMP24]], i64 8) ]
 ; WITH-AC-NEXT:    [[TMP26:%.*]] = load i32*, i32** [[TMP25]], align 8
-; WITH-AC-NEXT:    call void @llvm.assume(i1 true) [ "dereferenceable"(i32* [[TMP26]], i64 4), "nonnull"(i32* [[TMP26]]), "align"(i32* [[TMP26]], i64 4) ]
 ; WITH-AC-NEXT:    [[TMP27:%.*]] = load i32, i32* [[TMP26]], align 4
 ; WITH-AC-NEXT:    [[TMP28:%.*]] = add nsw i32 [[TMP23]], [[TMP27]]
 ; WITH-AC-NEXT:    ret i32 [[TMP28]]
 ;
 ; CROSS-BLOCK-LABEL: define {{[^@]+}}@test2
 ; CROSS-BLOCK-SAME: (%struct.S* [[TMP0:%.*]], i32* [[TMP1:%.*]], i8* [[TMP2:%.*]])
 ; CROSS-BLOCK-NEXT:    [[TMP4:%.*]] = alloca %struct.S*, align 8
 ; CROSS-BLOCK-NEXT:    [[TMP5:%.*]] = alloca i32*, align 8
 ; CROSS-BLOCK-NEXT:    [[TMP6:%.*]] = alloca i8*, align 8
 ; CROSS-BLOCK-NEXT:    [[TMP7:%.*]] = alloca [[STRUCT_S:%.*]], align 8
-; CROSS-BLOCK-NEXT:    call void @llvm.assume(i1 true) [ "dereferenceable"(%struct.S** [[TMP4]], i64 8), "nonnull"(%struct.S** [[TMP4]]), "align"(%struct.S** [[TMP4]], i64 8) ]
 ; CROSS-BLOCK-NEXT:    store %struct.S* [[TMP0]], %struct.S** [[TMP4]], align 8
-; CROSS-BLOCK-NEXT:    call void @llvm.assume(i1 true) [ "dereferenceable"(i32** [[TMP5]], i64 8), "nonnull"(i32** [[TMP5]]), "align"(i32** [[TMP5]], i64 8) ]
 ; CROSS-BLOCK-NEXT:    store i32* [[TMP1]], i32** [[TMP5]], align 8
-; CROSS-BLOCK-NEXT:    call void @llvm.assume(i1 true) [ "dereferenceable"(i8** [[TMP6]], i64 8), "nonnull"(i8** [[TMP6]]) ]
 ; CROSS-BLOCK-NEXT:    store i8* [[TMP2]], i8** [[TMP6]]
 ; CROSS-BLOCK-NEXT:    [[TMP8:%.*]] = load i32*, i32** [[TMP5]], align 8
-; CROSS-BLOCK-NEXT:    call void @llvm.assume(i1 true) [ "dereferenceable"(i32* [[TMP8]], i64 4), "nonnull"(i32* [[TMP8]]), "align"(i32* [[TMP8]], i64 4) ]
 ; CROSS-BLOCK-NEXT:    [[TMP9:%.*]] = load i32, i32* [[TMP8]], align 4
 ; CROSS-BLOCK-NEXT:    [[TMP10:%.*]] = trunc i32 [[TMP9]] to i8
-; CROSS-BLOCK-NEXT:    call void @llvm.assume(i1 true) [ "align"(i8** [[TMP6]], i64 8) ]
 ; CROSS-BLOCK-NEXT:    [[TMP11:%.*]] = load i8*, i8** [[TMP6]], align 8
-; CROSS-BLOCK-NEXT:    call void @llvm.assume(i1 true) [ "dereferenceable"(i8* [[TMP11]], i64 1), "nonnull"(i8* [[TMP11]]) ]
 ; CROSS-BLOCK-NEXT:    store i8 [[TMP10]], i8* [[TMP11]], align 1
 ; CROSS-BLOCK-NEXT:    [[TMP12:%.*]] = bitcast %struct.S* [[TMP7]] to i8*
 ; CROSS-BLOCK-NEXT:    [[TMP13:%.*]] = load %struct.S*, %struct.S** [[TMP4]]
 ; CROSS-BLOCK-NEXT:    [[TMP14:%.*]] = bitcast %struct.S* [[TMP13]] to i8*
 ; CROSS-BLOCK-NEXT:    [[TMP15:%.*]] = bitcast %struct.S* [[TMP7]] to i8*
 ; CROSS-BLOCK-NEXT:    [[TMP16:%.*]] = load %struct.S*, %struct.S** [[TMP4]], align 8
 ; CROSS-BLOCK-NEXT:    [[TMP17:%.*]] = getelementptr inbounds [[STRUCT_S]], %struct.S* [[TMP16]], i32 0, i32 0
-; CROSS-BLOCK-NEXT:    call void @llvm.assume(i1 true) [ "dereferenceable"(i32* [[TMP17]], i64 4), "nonnull"(i32* [[TMP17]]), "align"(i32* [[TMP17]], i64 8) ]
+; CROSS-BLOCK-NEXT:    call void @llvm.assume(i1 true) [ "dereferenceable"(%struct.S* [[TMP16]], i64 4), "nonnull"(%struct.S* [[TMP16]]), "align"(%struct.S* [[TMP16]], i64 8) ]
 ; CROSS-BLOCK-NEXT:    [[TMP18:%.*]] = load i32, i32* [[TMP17]], align 8
 ; CROSS-BLOCK-NEXT:    [[TMP19:%.*]] = load %struct.S*, %struct.S** [[TMP4]], align 8
 ; CROSS-BLOCK-NEXT:    [[TMP20:%.*]] = getelementptr inbounds [[STRUCT_S]], %struct.S* [[TMP19]], i32 0, i32 1
-; CROSS-BLOCK-NEXT:    call void @llvm.assume(i1 true) [ "dereferenceable"(i8* [[TMP20]], i64 1), "nonnull"(i8* [[TMP20]]), "align"(i8* [[TMP20]], i64 4) ]
+; CROSS-BLOCK-NEXT:    call void @llvm.assume(i1 true) [ "dereferenceable"(%struct.S* [[TMP19]], i64 5), "nonnull"(%struct.S* [[TMP19]]), "align"(%struct.S* [[TMP19]], i64 1) ]
 ; CROSS-BLOCK-NEXT:    [[TMP21:%.*]] = load i8, i8* [[TMP20]], align 4
 ; CROSS-BLOCK-NEXT:    [[TMP22:%.*]] = sext i8 [[TMP21]] to i32
 ; CROSS-BLOCK-NEXT:    [[TMP23:%.*]] = add nsw i32 [[TMP18]], [[TMP22]]
 ; CROSS-BLOCK-NEXT:    [[TMP24:%.*]] = load %struct.S*, %struct.S** [[TMP4]], align 8
 ; CROSS-BLOCK-NEXT:    [[TMP25:%.*]] = getelementptr inbounds [[STRUCT_S]], %struct.S* [[TMP24]], i32 0, i32 2
-; CROSS-BLOCK-NEXT:    call void @llvm.assume(i1 true) [ "dereferenceable"(i32** [[TMP25]], i64 8), "nonnull"(i32** [[TMP25]]), "align"(i32** [[TMP25]], i64 8) ]
+; CROSS-BLOCK-NEXT:    call void @llvm.assume(i1 true) [ "dereferenceable"(%struct.S* [[TMP24]], i64 16), "nonnull"(%struct.S* [[TMP24]]), "align"(%struct.S* [[TMP24]], i64 8) ]
 ; CROSS-BLOCK-NEXT:    [[TMP26:%.*]] = load i32*, i32** [[TMP25]], align 8
-; CROSS-BLOCK-NEXT:    call void @llvm.assume(i1 true) [ "dereferenceable"(i32* [[TMP26]], i64 4), "nonnull"(i32* [[TMP26]]), "align"(i32* [[TMP26]], i64 4) ]
 ; CROSS-BLOCK-NEXT:    [[TMP27:%.*]] = load i32, i32* [[TMP26]], align 4
 ; CROSS-BLOCK-NEXT:    [[TMP28:%.*]] = add nsw i32 [[TMP23]], [[TMP27]]
 ; CROSS-BLOCK-NEXT:    ret i32 [[TMP28]]
 ;
 ; FULL-SIMPLIFY-LABEL: define {{[^@]+}}@test2
 ; FULL-SIMPLIFY-SAME: (%struct.S* [[TMP0:%.*]], i32* [[TMP1:%.*]], i8* [[TMP2:%.*]])
 ; FULL-SIMPLIFY-NEXT:    [[TMP4:%.*]] = alloca %struct.S*, align 8
 ; FULL-SIMPLIFY-NEXT:    [[TMP5:%.*]] = alloca i32*, align 8

 ; FULL-SIMPLIFY-NEXT:    [[TMP17:%.*]] = getelementptr inbounds [[STRUCT_S]], %struct.S* [[TMP16]], i32 0, i32 0
 ; FULL-SIMPLIFY-NEXT:    [[TMP18:%.*]] = load i32, i32* [[TMP17]], align 8
 ; FULL-SIMPLIFY-NEXT:    [[TMP19:%.*]] = load %struct.S*, %struct.S** [[TMP4]], align 8
 ; FULL-SIMPLIFY-NEXT:    [[TMP20:%.*]] = getelementptr inbounds [[STRUCT_S]], %struct.S* [[TMP19]], i32 0, i32 1
 ; FULL-SIMPLIFY-NEXT:    [[TMP21:%.*]] = load i8, i8* [[TMP20]], align 4
 ; FULL-SIMPLIFY-NEXT:    [[TMP22:%.*]] = sext i8 [[TMP21]] to i32
 ; FULL-SIMPLIFY-NEXT:    [[TMP23:%.*]] = add nsw i32 [[TMP18]], [[TMP22]]
 ; FULL-SIMPLIFY-NEXT:    [[TMP24:%.*]] = load %struct.S*, %struct.S** [[TMP4]], align 8
+; FULL-SIMPLIFY-NEXT:    call void @llvm.assume(i1 true) [ "dereferenceable"(%struct.S* [[TMP16]], i64 4), "nonnull"(%struct.S* [[TMP16]]), "align"(%struct.S* [[TMP16]], i64 8), "dereferenceable"(%struct.S* [[TMP19]], i64 5), "nonnull"(%struct.S* [[TMP19]]), "align"(%struct.S* [[TMP19]], i64 1), "dereferenceable"(%struct.S* [[TMP24]], i64 16), "nonnull"(%struct.S* [[TMP24]]), "align"(%struct.S* [[TMP24]], i64 8) ]
 ; FULL-SIMPLIFY-NEXT:    [[TMP25:%.*]] = getelementptr inbounds [[STRUCT_S]], %struct.S* [[TMP24]], i32 0, i32 2
 ; FULL-SIMPLIFY-NEXT:    [[TMP26:%.*]] = load i32*, i32** [[TMP25]], align 8
-; FULL-SIMPLIFY-NEXT:    call void @llvm.assume(i1 true) [ "dereferenceable"(%struct.S** [[TMP4]], i64 8), "nonnull"(%struct.S** [[TMP4]]), "align"(%struct.S** [[TMP4]], i64 8), "dereferenceable"(i32** [[TMP5]], i64 8), "nonnull"(i32** [[TMP5]]), "align"(i32** [[TMP5]], i64 8), "dereferenceable"(i8** [[TMP6]], i64 8), "nonnull"(i8** [[TMP6]]), "dereferenceable"(i32* [[TMP8]], i64 4), "nonnull"(i32* [[TMP8]]), "align"(i32* [[TMP8]], i64 4), "align"(i8** [[TMP6]], i64 8), "dereferenceable"(i8* [[TMP11]], i64 1), "nonnull"(i8* [[TMP11]]), "dereferenceable"(i32* [[TMP17]], i64 4), "nonnull"(i32* [[TMP17]]), "align"(i32* [[TMP17]], i64 8), "dereferenceable"(i8* [[TMP20]], i64 1), "nonnull"(i8* [[TMP20]]), "align"(i8* [[TMP20]], i64 4), "dereferenceable"(i32** [[TMP25]], i64 8), "nonnull"(i32** [[TMP25]]), "align"(i32** [[TMP25]], i64 8), "dereferenceable"(i32* [[TMP26]], i64 4), "nonnull"(i32* [[TMP26]]), "align"(i32* [[TMP26]], i64 4) ]
 ; FULL-SIMPLIFY-NEXT:    [[TMP27:%.*]] = load i32, i32* [[TMP26]], align 4
 ; FULL-SIMPLIFY-NEXT:    [[TMP28:%.*]] = add nsw i32 [[TMP23]], [[TMP27]]
 ; FULL-SIMPLIFY-NEXT:    ret i32 [[TMP28]]
 ;
   %4 = alloca %struct.S*, align 8
   %5 = alloca i32*, align 8
   %6 = alloca i8*, align 8
   %7 = alloca %struct.S, align 8

 
 define i32 @test3(%struct.S* %0, i32* %1, i8* %2) "null-pointer-is-valid"="true" {
 ; BASIC-LABEL: define {{[^@]+}}@test3
 ; BASIC-SAME: (%struct.S* [[TMP0:%.*]], i32* [[TMP1:%.*]], i8* [[TMP2:%.*]]) #4
 ; BASIC-NEXT:    [[TMP4:%.*]] = alloca %struct.S*, align 8
 ; BASIC-NEXT:    [[TMP5:%.*]] = alloca i32*, align 8
 ; BASIC-NEXT:    [[TMP6:%.*]] = alloca i8*, align 8
 ; BASIC-NEXT:    [[TMP7:%.*]] = alloca [[STRUCT_S:%.*]], align 8
-; BASIC-NEXT:    call void @llvm.assume(i1 true) [ "dereferenceable"(%struct.S** [[TMP4]], i64 8), "align"(%struct.S** [[TMP4]], i64 8) ]
 ; BASIC-NEXT:    store %struct.S* [[TMP0]], %struct.S** [[TMP4]], align 8
-; BASIC-NEXT:    call void @llvm.assume(i1 true) [ "dereferenceable"(i32** [[TMP5]], i64 8), "align"(i32** [[TMP5]], i64 8) ]
 ; BASIC-NEXT:    store i32* [[TMP1]], i32** [[TMP5]], align 8
-; BASIC-NEXT:    call void @llvm.assume(i1 true) [ "dereferenceable"(i8** [[TMP6]], i64 8), "align"(i8** [[TMP6]], i64 8) ]
 ; BASIC-NEXT:    store i8* [[TMP2]], i8** [[TMP6]], align 8
-; BASIC-NEXT:    call void @llvm.assume(i1 true) [ "dereferenceable"(i32** [[TMP5]], i64 8), "align"(i32** [[TMP5]], i64 8) ]
 ; BASIC-NEXT:    [[TMP8:%.*]] = load i32*, i32** [[TMP5]], align 8
-; BASIC-NEXT:    call void @llvm.assume(i1 true) [ "dereferenceable"(i32* [[TMP8]], i64 4), "align"(i32* [[TMP8]], i64 4) ]
 ; BASIC-NEXT:    [[TMP9:%.*]] = load i32, i32* [[TMP8]], align 4
 ; BASIC-NEXT:    [[TMP10:%.*]] = trunc i32 [[TMP9]] to i8
-; BASIC-NEXT:    call void @llvm.assume(i1 true) [ "dereferenceable"(i8** [[TMP6]], i64 8) ]
 ; BASIC-NEXT:    [[TMP11:%.*]] = load i8*, i8** [[TMP6]]
-; BASIC-NEXT:    call void @llvm.assume(i1 true) [ "dereferenceable"(i8* [[TMP11]], i64 1) ]
 ; BASIC-NEXT:    store i8 [[TMP10]], i8* [[TMP11]], align 1
 ; BASIC-NEXT:    [[TMP12:%.*]] = bitcast %struct.S* [[TMP7]] to i8*
-; BASIC-NEXT:    call void @llvm.assume(i1 true) [ "dereferenceable"(%struct.S** [[TMP4]], i64 8), "align"(%struct.S** [[TMP4]], i64 32) ]
 ; BASIC-NEXT:    [[TMP13:%.*]] = load %struct.S*, %struct.S** [[TMP4]], align 32
 ; BASIC-NEXT:    [[TMP14:%.*]] = bitcast %struct.S* [[TMP13]] to i8*
 ; BASIC-NEXT:    call void @may_throw()
 ; BASIC-NEXT:    [[TMP15:%.*]] = bitcast %struct.S* [[TMP7]] to i8*
-; BASIC-NEXT:    call void @llvm.assume(i1 true) [ "dereferenceable"(%struct.S** [[TMP4]], i64 8), "align"(%struct.S** [[TMP4]], i64 8) ]
 ; BASIC-NEXT:    [[TMP16:%.*]] = load %struct.S*, %struct.S** [[TMP4]], align 8
 ; BASIC-NEXT:    [[TMP17:%.*]] = getelementptr inbounds [[STRUCT_S]], %struct.S* [[TMP16]], i32 0, i32 0
-; BASIC-NEXT:    call void @llvm.assume(i1 true) [ "dereferenceable"(i32* [[TMP17]], i64 4), "align"(i32* [[TMP17]], i64 8) ]
+; BASIC-NEXT:    call void @llvm.assume(i1 true) [ "dereferenceable"(%struct.S* [[TMP16]], i64 4), "align"(%struct.S* [[TMP16]], i64 8) ]
 ; BASIC-NEXT:    [[TMP18:%.*]] = load i32, i32* [[TMP17]], align 8
-; BASIC-NEXT:    call void @llvm.assume(i1 true) [ "dereferenceable"(%struct.S** [[TMP4]], i64 8), "align"(%struct.S** [[TMP4]], i64 8) ]
 ; BASIC-NEXT:    [[TMP19:%.*]] = load %struct.S*, %struct.S** [[TMP4]], align 8
 ; BASIC-NEXT:    [[TMP20:%.*]] = getelementptr inbounds [[STRUCT_S]], %struct.S* [[TMP19]], i32 0, i32 1
-; BASIC-NEXT:    call void @llvm.assume(i1 true) [ "dereferenceable"(i8* [[TMP20]], i64 1), "align"(i8* [[TMP20]], i64 4) ]
-; BASIC-NEXT:    [[TMP21:%.*]] = load i8, i8* [[TMP20]], align 4
+; BASIC-NEXT:    call void @llvm.assume(i1 true) [ "dereferenceable"(%struct.S* [[TMP19]], i64 5), "align"(%struct.S* [[TMP19]], i64 1) ]
+; BASIC-NEXT:    [[TMP21:%.*]] = load i8, i8* [[TMP20]], align 8
 ; BASIC-NEXT:    [[TMP22:%.*]] = sext i8 [[TMP21]] to i32
 ; BASIC-NEXT:    [[TMP23:%.*]] = add nsw i32 [[TMP18]], [[TMP22]]
-; BASIC-NEXT:    call void @llvm.assume(i1 true) [ "dereferenceable"(%struct.S** [[TMP4]], i64 8), "align"(%struct.S** [[TMP4]], i64 8) ]
 ; BASIC-NEXT:    [[TMP24:%.*]] = load %struct.S*, %struct.S** [[TMP4]], align 8
 ; BASIC-NEXT:    [[TMP25:%.*]] = getelementptr inbounds [[STRUCT_S]], %struct.S* [[TMP24]], i32 0, i32 2
-; BASIC-NEXT:    call void @llvm.assume(i1 true) [ "dereferenceable"(i32** [[TMP25]], i64 8) ]
+; BASIC-NEXT:    call void @llvm.assume(i1 true) [ "dereferenceable"(%struct.S* [[TMP24]], i64 16) ]
 ; BASIC-NEXT:    [[TMP26:%.*]] = load i32*, i32** [[TMP25]]
-; BASIC-NEXT:    call void @llvm.assume(i1 true) [ "dereferenceable"(i32* [[TMP26]], i64 4), "align"(i32* [[TMP26]], i64 4) ]
 ; BASIC-NEXT:    [[TMP27:%.*]] = load i32, i32* [[TMP26]], align 4
 ; BASIC-NEXT:    [[TMP28:%.*]] = add nsw i32 [[TMP23]], [[TMP27]]
 ; BASIC-NEXT:    ret i32 [[TMP28]]
 ;
 ; ALL-LABEL: define {{[^@]+}}@test3
 ; ALL-SAME: (%struct.S* [[TMP0:%.*]], i32* [[TMP1:%.*]], i8* [[TMP2:%.*]]) #4
 ; ALL-NEXT:    [[TMP4:%.*]] = alloca %struct.S*, align 8
 ; ALL-NEXT:    [[TMP5:%.*]] = alloca i32*, align 8
 ; ALL-NEXT:    [[TMP6:%.*]] = alloca i8*, align 8
 ; ALL-NEXT:    [[TMP7:%.*]] = alloca [[STRUCT_S:%.*]], align 8
-; ALL-NEXT:    call void @llvm.assume(i1 true) [ "dereferenceable"(%struct.S** [[TMP4]], i64 8), "align"(%struct.S** [[TMP4]], i64 8) ]
 ; ALL-NEXT:    store %struct.S* [[TMP0]], %struct.S** [[TMP4]], align 8
-; ALL-NEXT:    call void @llvm.assume(i1 true) [ "dereferenceable"(i32** [[TMP5]], i64 8), "align"(i32** [[TMP5]], i64 8) ]
 ; ALL-NEXT:    store i32* [[TMP1]], i32** [[TMP5]], align 8
-; ALL-NEXT:    call void @llvm.assume(i1 true) [ "dereferenceable"(i8** [[TMP6]], i64 8), "align"(i8** [[TMP6]], i64 8) ]
 ; ALL-NEXT:    store i8* [[TMP2]], i8** [[TMP6]], align 8
-; ALL-NEXT:    call void @llvm.assume(i1 true) [ "dereferenceable"(i32** [[TMP5]], i64 8), "align"(i32** [[TMP5]], i64 8) ]
 ; ALL-NEXT:    [[TMP8:%.*]] = load i32*, i32** [[TMP5]], align 8
-; ALL-NEXT:    call void @llvm.assume(i1 true) [ "dereferenceable"(i32* [[TMP8]], i64 4), "align"(i32* [[TMP8]], i64 4) ]
 ; ALL-NEXT:    [[TMP9:%.*]] = load i32, i32* [[TMP8]], align 4
 ; ALL-NEXT:    [[TMP10:%.*]] = trunc i32 [[TMP9]] to i8
-; ALL-NEXT:    call void @llvm.assume(i1 true) [ "dereferenceable"(i8** [[TMP6]], i64 8) ]
 ; ALL-NEXT:    [[TMP11:%.*]] = load i8*, i8** [[TMP6]]
-; ALL-NEXT:    call void @llvm.assume(i1 true) [ "dereferenceable"(i8* [[TMP11]], i64 1) ]
 ; ALL-NEXT:    store i8 [[TMP10]], i8* [[TMP11]], align 1
 ; ALL-NEXT:    [[TMP12:%.*]] = bitcast %struct.S* [[TMP7]] to i8*
-; ALL-NEXT:    call void @llvm.assume(i1 true) [ "dereferenceable"(%struct.S** [[TMP4]], i64 8), "align"(%struct.S** [[TMP4]], i64 32) ]
 ; ALL-NEXT:    [[TMP13:%.*]] = load %struct.S*, %struct.S** [[TMP4]], align 32
 ; ALL-NEXT:    [[TMP14:%.*]] = bitcast %struct.S* [[TMP13]] to i8*
 ; ALL-NEXT:    call void @may_throw()
 ; ALL-NEXT:    [[TMP15:%.*]] = bitcast %struct.S* [[TMP7]] to i8*
-; ALL-NEXT:    call void @llvm.assume(i1 true) [ "dereferenceable"(%struct.S** [[TMP4]], i64 8), "align"(%struct.S** [[TMP4]], i64 8) ]
 ; ALL-NEXT:    [[TMP16:%.*]] = load %struct.S*, %struct.S** [[TMP4]], align 8
 ; ALL-NEXT:    [[TMP17:%.*]] = getelementptr inbounds [[STRUCT_S]], %struct.S* [[TMP16]], i32 0, i32 0
-; ALL-NEXT:    call void @llvm.assume(i1 true) [ "dereferenceable"(i32* [[TMP17]], i64 4), "align"(i32* [[TMP17]], i64 8) ]
+; ALL-NEXT:    call void @llvm.assume(i1 true) [ "dereferenceable"(%struct.S* [[TMP16]], i64 4), "align"(%struct.S* [[TMP16]], i64 8) ]
 ; ALL-NEXT:    [[TMP18:%.*]] = load i32, i32* [[TMP17]], align 8
-; ALL-NEXT:    call void @llvm.assume(i1 true) [ "dereferenceable"(%struct.S** [[TMP4]], i64 8), "align"(%struct.S** [[TMP4]], i64 8) ]
 ; ALL-NEXT:    [[TMP19:%.*]] = load %struct.S*, %struct.S** [[TMP4]], align 8
 ; ALL-NEXT:    [[TMP20:%.*]] = getelementptr inbounds [[STRUCT_S]], %struct.S* [[TMP19]], i32 0, i32 1
-; ALL-NEXT:    call void @llvm.assume(i1 true) [ "dereferenceable"(i8* [[TMP20]], i64 1), "align"(i8* [[TMP20]], i64 4) ]
-; ALL-NEXT:    [[TMP21:%.*]] = load i8, i8* [[TMP20]], align 4
+; ALL-NEXT:    call void @llvm.assume(i1 true) [ "dereferenceable"(%struct.S* [[TMP19]], i64 5), "align"(%struct.S* [[TMP19]], i64 1) ]
+; ALL-NEXT:    [[TMP21:%.*]] = load i8, i8* [[TMP20]], align 8
 ; ALL-NEXT:    [[TMP22:%.*]] = sext i8 [[TMP21]] to i32
 ; ALL-NEXT:    [[TMP23:%.*]] = add nsw i32 [[TMP18]], [[TMP22]]
-; ALL-NEXT:    call void @llvm.assume(i1 true) [ "dereferenceable"(%struct.S** [[TMP4]], i64 8), "align"(%struct.S** [[TMP4]], i64 8) ]
 ; ALL-NEXT:    [[TMP24:%.*]] = load %struct.S*, %struct.S** [[TMP4]], align 8
 ; ALL-NEXT:    [[TMP25:%.*]] = getelementptr inbounds [[STRUCT_S]], %struct.S* [[TMP24]], i32 0, i32 2
-; ALL-NEXT:    call void @llvm.assume(i1 true) [ "dereferenceable"(i32** [[TMP25]], i64 8) ]
+; ALL-NEXT:    call void @llvm.assume(i1 true) [ "dereferenceable"(%struct.S* [[TMP24]], i64 16) ]
 ; ALL-NEXT:    [[TMP26:%.*]] = load i32*, i32** [[TMP25]]
-; ALL-NEXT:    call void @llvm.assume(i1 true) [ "dereferenceable"(i32* [[TMP26]], i64 4), "align"(i32* [[TMP26]], i64 4) ]
 ; ALL-NEXT:    [[TMP27:%.*]] = load i32, i32* [[TMP26]], align 4
 ; ALL-NEXT:    [[TMP28:%.*]] = add nsw i32 [[TMP23]], [[TMP27]]
 ; ALL-NEXT:    ret i32 [[TMP28]]
 ;
 ; WITH-AC-LABEL: define {{[^@]+}}@test3
 ; WITH-AC-SAME: (%struct.S* [[TMP0:%.*]], i32* [[TMP1:%.*]], i8* [[TMP2:%.*]]) #4
 ; WITH-AC-NEXT:    [[TMP4:%.*]] = alloca %struct.S*, align 8
 ; WITH-AC-NEXT:    [[TMP5:%.*]] = alloca i32*, align 8
 ; WITH-AC-NEXT:    [[TMP6:%.*]] = alloca i8*, align 8
 ; WITH-AC-NEXT:    [[TMP7:%.*]] = alloca [[STRUCT_S:%.*]], align 8
-; WITH-AC-NEXT:    call void @llvm.assume(i1 true) [ "dereferenceable"(%struct.S** [[TMP4]], i64 8), "align"(%struct.S** [[TMP4]], i64 32) ]
 ; WITH-AC-NEXT:    store %struct.S* [[TMP0]], %struct.S** [[TMP4]], align 8
-; WITH-AC-NEXT:    call void @llvm.assume(i1 true) [ "dereferenceable"(i32** [[TMP5]], i64 8), "align"(i32** [[TMP5]], i64 8) ]
 ; WITH-AC-NEXT:    store i32* [[TMP1]], i32** [[TMP5]], align 8
-; WITH-AC-NEXT:    call void @llvm.assume(i1 true) [ "dereferenceable"(i8** [[TMP6]], i64 8), "align"(i8** [[TMP6]], i64 8) ]
 ; WITH-AC-NEXT:    store i8* [[TMP2]], i8** [[TMP6]], align 8
 ; WITH-AC-NEXT:    [[TMP8:%.*]] = load i32*, i32** [[TMP5]], align 8
-; WITH-AC-NEXT:    call void @llvm.assume(i1 true) [ "dereferenceable"(i32* [[TMP8]], i64 4), "align"(i32* [[TMP8]], i64 4) ]
 ; WITH-AC-NEXT:    [[TMP9:%.*]] = load i32, i32* [[TMP8]], align 4
 ; WITH-AC-NEXT:    [[TMP10:%.*]] = trunc i32 [[TMP9]] to i8
 ; WITH-AC-NEXT:    [[TMP11:%.*]] = load i8*, i8** [[TMP6]]
-; WITH-AC-NEXT:    call void @llvm.assume(i1 true) [ "dereferenceable"(i8* [[TMP11]], i64 1) ]
 ; WITH-AC-NEXT:    store i8 [[TMP10]], i8* [[TMP11]], align 1
 ; WITH-AC-NEXT:    [[TMP12:%.*]] = bitcast %struct.S* [[TMP7]] to i8*
 ; WITH-AC-NEXT:    [[TMP13:%.*]] = load %struct.S*, %struct.S** [[TMP4]], align 32
 ; WITH-AC-NEXT:    [[TMP14:%.*]] = bitcast %struct.S* [[TMP13]] to i8*
 ; WITH-AC-NEXT:    call void @may_throw()
 ; WITH-AC-NEXT:    [[TMP15:%.*]] = bitcast %struct.S* [[TMP7]] to i8*
 ; WITH-AC-NEXT:    [[TMP16:%.*]] = load %struct.S*, %struct.S** [[TMP4]], align 8
 ; WITH-AC-NEXT:    [[TMP17:%.*]] = getelementptr inbounds [[STRUCT_S]], %struct.S* [[TMP16]], i32 0, i32 0
-; WITH-AC-NEXT:    call void @llvm.assume(i1 true) [ "dereferenceable"(i32* [[TMP17]], i64 4), "align"(i32* [[TMP17]], i64 8) ]
+; WITH-AC-NEXT:    call void @llvm.assume(i1 true) [ "dereferenceable"(%struct.S* [[TMP16]], i64 4), "align"(%struct.S* [[TMP16]], i64 8) ]
 ; WITH-AC-NEXT:    [[TMP18:%.*]] = load i32, i32* [[TMP17]], align 8
 ; WITH-AC-NEXT:    [[TMP19:%.*]] = load %struct.S*, %struct.S** [[TMP4]], align 8
 ; WITH-AC-NEXT:    [[TMP20:%.*]] = getelementptr inbounds [[STRUCT_S]], %struct.S* [[TMP19]], i32 0, i32 1
-; WITH-AC-NEXT:    call void @llvm.assume(i1 true) [ "dereferenceable"(i8* [[TMP20]], i64 1), "align"(i8* [[TMP20]], i64 4) ]
-; WITH-AC-NEXT:    [[TMP21:%.*]] = load i8, i8* [[TMP20]], align 4
+; WITH-AC-NEXT:    call void @llvm.assume(i1 true) [ "dereferenceable"(%struct.S* [[TMP19]], i64 5), "align"(%struct.S* [[TMP19]], i64 1) ]
+; WITH-AC-NEXT:    [[TMP21:%.*]] = load i8, i8* [[TMP20]], align 8
 ; WITH-AC-NEXT:    [[TMP22:%.*]] = sext i8 [[TMP21]] to i32
 ; WITH-AC-NEXT:    [[TMP23:%.*]] = add nsw i32 [[TMP18]], [[TMP22]]
 ; WITH-AC-NEXT:    [[TMP24:%.*]] = load %struct.S*, %struct.S** [[TMP4]], align 8
 ; WITH-AC-NEXT:    [[TMP25:%.*]] = getelementptr inbounds [[STRUCT_S]], %struct.S* [[TMP24]], i32 0, i32 2
-; WITH-AC-NEXT:    call void @llvm.assume(i1 true) [ "dereferenceable"(i32** [[TMP25]], i64 8) ]
+; WITH-AC-NEXT:    call void @llvm.assume(i1 true) [ "dereferenceable"(%struct.S* [[TMP24]], i64 16) ]
 ; WITH-AC-NEXT:    [[TMP26:%.*]] = load i32*, i32** [[TMP25]]
-; WITH-AC-NEXT:    call void @llvm.assume(i1 true) [ "dereferenceable"(i32* [[TMP26]], i64 4), "align"(i32* [[TMP26]], i64 4) ]
 ; WITH-AC-NEXT:    [[TMP27:%.*]] = load i32, i32* [[TMP26]], align 4
 ; WITH-AC-NEXT:    [[TMP28:%.*]] = add nsw i32 [[TMP23]], [[TMP27]]
 ; WITH-AC-NEXT:    ret i32 [[TMP28]]
 ;
 ; CROSS-BLOCK-LABEL: define {{[^@]+}}@test3
 ; CROSS-BLOCK-SAME: (%struct.S* [[TMP0:%.*]], i32* [[TMP1:%.*]], i8* [[TMP2:%.*]]) #4
 ; CROSS-BLOCK-NEXT:    [[TMP4:%.*]] = alloca %struct.S*, align 8
 ; CROSS-BLOCK-NEXT:    [[TMP5:%.*]] = alloca i32*, align 8
 ; CROSS-BLOCK-NEXT:    [[TMP6:%.*]] = alloca i8*, align 8
 ; CROSS-BLOCK-NEXT:    [[TMP7:%.*]] = alloca [[STRUCT_S:%.*]], align 8
-; CROSS-BLOCK-NEXT:    call void @llvm.assume(i1 true) [ "dereferenceable"(%struct.S** [[TMP4]], i64 8), "align"(%struct.S** [[TMP4]], i64 32) ]
 ; CROSS-BLOCK-NEXT:    store %struct.S* [[TMP0]], %struct.S** [[TMP4]], align 8
-; CROSS-BLOCK-NEXT:    call void @llvm.assume(i1 true) [ "dereferenceable"(i32** [[TMP5]], i64 8), "align"(i32** [[TMP5]], i64 8) ]
 ; CROSS-BLOCK-NEXT:    store i32* [[TMP1]], i32** [[TMP5]], align 8
-; CROSS-BLOCK-NEXT:    call void @llvm.assume(i1 true) [ "dereferenceable"(i8** [[TMP6]], i64 8), "align"(i8** [[TMP6]], i64 8) ]
 ; CROSS-BLOCK-NEXT:    store i8* [[TMP2]], i8** [[TMP6]], align 8
 ; CROSS-BLOCK-NEXT:    [[TMP8:%.*]] = load i32*, i32** [[TMP5]], align 8
-; CROSS-BLOCK-NEXT:    call void @llvm.assume(i1 true) [ "dereferenceable"(i32* [[TMP8]], i64 4), "align"(i32* [[TMP8]], i64 4) ]
 ; CROSS-BLOCK-NEXT:    [[TMP9:%.*]] = load i32, i32* [[TMP8]], align 4
 ; CROSS-BLOCK-NEXT:    [[TMP10:%.*]] = trunc i32 [[TMP9]] to i8
 ; CROSS-BLOCK-NEXT:    [[TMP11:%.*]] = load i8*, i8** [[TMP6]]
-; CROSS-BLOCK-NEXT:    call void @llvm.assume(i1 true) [ "dereferenceable"(i8* [[TMP11]], i64 1) ]
 ; CROSS-BLOCK-NEXT:    store i8 [[TMP10]], i8* [[TMP11]], align 1
 ; CROSS-BLOCK-NEXT:    [[TMP12:%.*]] = bitcast %struct.S* [[TMP7]] to i8*
 ; CROSS-BLOCK-NEXT:    [[TMP13:%.*]] = load %struct.S*, %struct.S** [[TMP4]], align 32
 ; CROSS-BLOCK-NEXT:    [[TMP14:%.*]] = bitcast %struct.S* [[TMP13]] to i8*
 ; CROSS-BLOCK-NEXT:    call void @may_throw()
 ; CROSS-BLOCK-NEXT:    [[TMP15:%.*]] = bitcast %struct.S* [[TMP7]] to i8*
 ; CROSS-BLOCK-NEXT:    [[TMP16:%.*]] = load %struct.S*, %struct.S** [[TMP4]], align 8
 ; CROSS-BLOCK-NEXT:    [[TMP17:%.*]] = getelementptr inbounds [[STRUCT_S]], %struct.S* [[TMP16]], i32 0, i32 0
-; CROSS-BLOCK-NEXT:    call void @llvm.assume(i1 true) [ "dereferenceable"(i32* [[TMP17]], i64 4), "align"(i32* [[TMP17]], i64 8) ]
+; CROSS-BLOCK-NEXT:    call void @llvm.assume(i1 true) [ "dereferenceable"(%struct.S* [[TMP16]], i64 4), "align"(%struct.S* [[TMP16]], i64 8) ]
 ; CROSS-BLOCK-NEXT:    [[TMP18:%.*]] = load i32, i32* [[TMP17]], align 8
 ; CROSS-BLOCK-NEXT:    [[TMP19:%.*]] = load %struct.S*, %struct.S** [[TMP4]], align 8
 ; CROSS-BLOCK-NEXT:    [[TMP20:%.*]] = getelementptr inbounds [[STRUCT_S]], %struct.S* [[TMP19]], i32 0, i32 1
-; CROSS-BLOCK-NEXT:    call void @llvm.assume(i1 true) [ "dereferenceable"(i8* [[TMP20]], i64 1), "align"(i8* [[TMP20]], i64 4) ]
-; CROSS-BLOCK-NEXT:    [[TMP21:%.*]] = load i8, i8* [[TMP20]], align 4
+; CROSS-BLOCK-NEXT:    call void @llvm.assume(i1 true) [ "dereferenceable"(%struct.S* [[TMP19]], i64 5), "align"(%struct.S* [[TMP19]], i64 1) ]
+; CROSS-BLOCK-NEXT:    [[TMP21:%.*]] = load i8, i8* [[TMP20]], align 8
 ; CROSS-BLOCK-NEXT:    [[TMP22:%.*]] = sext i8 [[TMP21]] to i32
 ; CROSS-BLOCK-NEXT:    [[TMP23:%.*]] = add nsw i32 [[TMP18]], [[TMP22]]
 ; CROSS-BLOCK-NEXT:    [[TMP24:%.*]] = load %struct.S*, %struct.S** [[TMP4]], align 8
 ; CROSS-BLOCK-NEXT:    [[TMP25:%.*]] = getelementptr inbounds [[STRUCT_S]], %struct.S* [[TMP24]], i32 0, i32 2
-; CROSS-BLOCK-NEXT:    call void @llvm.assume(i1 true) [ "dereferenceable"(i32** [[TMP25]], i64 8) ]
+; CROSS-BLOCK-NEXT:    call void @llvm.assume(i1 true) [ "dereferenceable"(%struct.S* [[TMP24]], i64 16) ]
 ; CROSS-BLOCK-NEXT:    [[TMP26:%.*]] = load i32*, i32** [[TMP25]]
-; CROSS-BLOCK-NEXT:    call void @llvm.assume(i1 true) [ "dereferenceable"(i32* [[TMP26]], i64 4), "align"(i32* [[TMP26]], i64 4) ]
 ; CROSS-BLOCK-NEXT:    [[TMP27:%.*]] = load i32, i32* [[TMP26]], align 4
 ; CROSS-BLOCK-NEXT:    [[TMP28:%.*]] = add nsw i32 [[TMP23]], [[TMP27]]
 ; CROSS-BLOCK-NEXT:    ret i32 [[TMP28]]
 ;
 ; FULL-SIMPLIFY-LABEL: define {{[^@]+}}@test3
 ; FULL-SIMPLIFY-SAME: (%struct.S* [[TMP0:%.*]], i32* [[TMP1:%.*]], i8* [[TMP2:%.*]]) #4
 ; FULL-SIMPLIFY-NEXT:    [[TMP4:%.*]] = alloca %struct.S*, align 8
 ; FULL-SIMPLIFY-NEXT:    [[TMP5:%.*]] = alloca i32*, align 8
 ; FULL-SIMPLIFY-NEXT:    [[TMP6:%.*]] = alloca i8*, align 8
 ; FULL-SIMPLIFY-NEXT:    [[TMP7:%.*]] = alloca [[STRUCT_S:%.*]], align 8
 ; FULL-SIMPLIFY-NEXT:    store %struct.S* [[TMP0]], %struct.S** [[TMP4]], align 8
 ; FULL-SIMPLIFY-NEXT:    store i32* [[TMP1]], i32** [[TMP5]], align 8
 ; FULL-SIMPLIFY-NEXT:    store i8* [[TMP2]], i8** [[TMP6]], align 8
 ; FULL-SIMPLIFY-NEXT:    [[TMP8:%.*]] = load i32*, i32** [[TMP5]], align 8
 ; FULL-SIMPLIFY-NEXT:    [[TMP9:%.*]] = load i32, i32* [[TMP8]], align 4
 ; FULL-SIMPLIFY-NEXT:    [[TMP10:%.*]] = trunc i32 [[TMP9]] to i8
 ; FULL-SIMPLIFY-NEXT:    [[TMP11:%.*]] = load i8*, i8** [[TMP6]]
-; FULL-SIMPLIFY-NEXT:    call void @llvm.assume(i1 true) [ "dereferenceable"(%struct.S** [[TMP4]], i64 8), "align"(%struct.S** [[TMP4]], i64 32), "dereferenceable"(i32** [[TMP5]], i64 8), "align"(i32** [[TMP5]], i64 8), "dereferenceable"(i8** [[TMP6]], i64 8), "align"(i8** [[TMP6]], i64 8), "dereferenceable"(i32* [[TMP8]], i64 4), "align"(i32* [[TMP8]], i64 4), "dereferenceable"(i8* [[TMP11]], i64 1) ]
 ; FULL-SIMPLIFY-NEXT:    store i8 [[TMP10]], i8* [[TMP11]], align 1
 ; FULL-SIMPLIFY-NEXT:    [[TMP12:%.*]] = bitcast %struct.S* [[TMP7]] to i8*
 ; FULL-SIMPLIFY-NEXT:    [[TMP13:%.*]] = load %struct.S*, %struct.S** [[TMP4]], align 32
 ; FULL-SIMPLIFY-NEXT:    [[TMP14:%.*]] = bitcast %struct.S* [[TMP13]] to i8*
 ; FULL-SIMPLIFY-NEXT:    call void @may_throw()
 ; FULL-SIMPLIFY-NEXT:    [[TMP15:%.*]] = bitcast %struct.S* [[TMP7]] to i8*
 ; FULL-SIMPLIFY-NEXT:    [[TMP16:%.*]] = load %struct.S*, %struct.S** [[TMP4]], align 8
 ; FULL-SIMPLIFY-NEXT:    [[TMP17:%.*]] = getelementptr inbounds [[STRUCT_S]], %struct.S* [[TMP16]], i32 0, i32 0
 ; FULL-SIMPLIFY-NEXT:    [[TMP18:%.*]] = load i32, i32* [[TMP17]], align 8
 ; FULL-SIMPLIFY-NEXT:    [[TMP19:%.*]] = load %struct.S*, %struct.S** [[TMP4]], align 8
 ; FULL-SIMPLIFY-NEXT:    [[TMP20:%.*]] = getelementptr inbounds [[STRUCT_S]], %struct.S* [[TMP19]], i32 0, i32 1
-; FULL-SIMPLIFY-NEXT:    [[TMP21:%.*]] = load i8, i8* [[TMP20]], align 4
+; FULL-SIMPLIFY-NEXT:    [[TMP21:%.*]] = load i8, i8* [[TMP20]], align 8
 ; FULL-SIMPLIFY-NEXT:    [[TMP22:%.*]] = sext i8 [[TMP21]] to i32
 ; FULL-SIMPLIFY-NEXT:    [[TMP23:%.*]] = add nsw i32 [[TMP18]], [[TMP22]]
 ; FULL-SIMPLIFY-NEXT:    [[TMP24:%.*]] = load %struct.S*, %struct.S** [[TMP4]], align 8
+; FULL-SIMPLIFY-NEXT:    call void @llvm.assume(i1 true) [ "dereferenceable"(%struct.S* [[TMP16]], i64 4), "align"(%struct.S* [[TMP16]], i64 8), "dereferenceable"(%struct.S* [[TMP19]], i64 5), "align"(%struct.S* [[TMP19]], i64 1), "dereferenceable"(%struct.S* [[TMP24]], i64 16) ]
 ; FULL-SIMPLIFY-NEXT:    [[TMP25:%.*]] = getelementptr inbounds [[STRUCT_S]], %struct.S* [[TMP24]], i32 0, i32 2
 ; FULL-SIMPLIFY-NEXT:    [[TMP26:%.*]] = load i32*, i32** [[TMP25]]
-; FULL-SIMPLIFY-NEXT:    call void @llvm.assume(i1 true) [ "dereferenceable"(i32* [[TMP17]], i64 4), "align"(i32* [[TMP17]], i64 8), "dereferenceable"(i8* [[TMP20]], i64 1), "align"(i8* [[TMP20]], i64 4), "dereferenceable"(i32** [[TMP25]], i64 8), "dereferenceable"(i32* [[TMP26]], i64 4), "align"(i32* [[TMP26]], i64 4) ]
 ; FULL-SIMPLIFY-NEXT:    [[TMP27:%.*]] = load i32, i32* [[TMP26]], align 4
 ; FULL-SIMPLIFY-NEXT:    [[TMP28:%.*]] = add nsw i32 [[TMP23]], [[TMP27]]
 ; FULL-SIMPLIFY-NEXT:    ret i32 [[TMP28]]
 ;
   %4 = alloca %struct.S*, align 8
   %5 = alloca i32*, align 8
   %6 = alloca i8*, align 8
   %7 = alloca %struct.S, align 8
;
   %14 = bitcast %struct.S* %13 to i8*
   call void @may_throw()
   %15 = bitcast %struct.S* %7 to i8*
   %16 = load %struct.S*, %struct.S** %4, align 8
   %17 = getelementptr inbounds %struct.S, %struct.S* %16, i32 0, i32 0
   %18 = load i32, i32* %17, align 8
   %19 = load %struct.S*, %struct.S** %4, align 8
   %20 = getelementptr inbounds %struct.S, %struct.S* %19, i32 0, i32 1
-  %21 = load i8, i8* %20, align 4
+  %21 = load i8, i8* %20, align 8
   %22 = sext i8 %21 to i32
   %23 = add nsw i32 %18, %22
   %24 = load %struct.S*, %struct.S** %4, align 8
   %25 = getelementptr inbounds %struct.S, %struct.S* %24, i32 0, i32 2
   %26 = load i32*, i32** %25
   %27 = load i32, i32* %26, align 4
   %28 = add nsw i32 %23, %27
   ret i32 %28
 }
 
 define dso_local i32 @_Z6squarePi(i32* %P, i32* %P1, i1 %cond) {
 ; BASIC-LABEL: define {{[^@]+}}@_Z6squarePi
 ; BASIC-SAME: (i32* [[P:%.*]], i32* [[P1:%.*]], i1 [[COND:%.*]])
 ; BASIC-NEXT:    call void @llvm.assume(i1 true) [ "dereferenceable"(i32* [[P]], i64 4), "nonnull"(i32* [[P]]), "align"(i32* [[P]], i64 4) ]
 ; BASIC-NEXT:    store i32 0, i32* [[P]], align 4
 ; BASIC-NEXT:    call void @llvm.assume(i1 true) [ "dereferenceable"(i32* [[P1]], i64 4), "nonnull"(i32* [[P1]]), "align"(i32* [[P1]], i64 8) ]
 ; BASIC-NEXT:    store i32 0, i32* [[P1]], align 8
 ; BASIC-NEXT:    br i1 [[COND]], label [[A:%.*]], label [[B:%.*]]
 ; BASIC:       A:
-; BASIC-NEXT:    call void @llvm.assume(i1 true) [ "dereferenceable"(i32* [[P]], i64 4), "nonnull"(i32* [[P]]), "align"(i32* [[P]], i64 8) ]
+; BASIC-NEXT:    call void @llvm.assume(i1 true) [ "align"(i32* [[P]], i64 8) ]
 ; BASIC-NEXT:    store i32 0, i32* [[P]], align 8
-; BASIC-NEXT:    call void @llvm.assume(i1 true) [ "dereferenceable"(i32* [[P1]], i64 4), "nonnull"(i32* [[P1]]), "align"(i32* [[P1]], i64 4) ]
 ; BASIC-NEXT:    store i32 0, i32* [[P1]], align 4
 ; BASIC-NEXT:    br i1 [[COND]], label [[C:%.*]], label [[B]]
 ; BASIC:       B:
 ; BASIC-NEXT:    call void @llvm.assume(i1 true) [ "dereferenceable"(i32* [[P]], i64 4), "nonnull"(i32* [[P]]), "align"(i32* [[P]], i64 8) ]
 ; BASIC-NEXT:    store i32 0, i32* [[P]], align 8
 ; BASIC-NEXT:    call void @llvm.assume(i1 true) [ "dereferenceable"(i32* [[P1]], i64 4), "nonnull"(i32* [[P1]]), "align"(i32* [[P1]], i64 8) ]
 ; BASIC-NEXT:    store i32 0, i32* [[P1]], align 8
 ; BASIC-NEXT:    br label [[C]]
 ; BASIC:       C:
 ; BASIC-NEXT:    call void @llvm.assume(i1 true) [ "dereferenceable"(i32* [[P]], i64 4), "nonnull"(i32* [[P]]), "align"(i32* [[P]], i64 32) ]
 ; BASIC-NEXT:    store i32 0, i32* [[P]], align 32
 ; BASIC-NEXT:    call void @llvm.assume(i1 true) [ "dereferenceable"(i32* [[P1]], i64 4), "nonnull"(i32* [[P1]]), "align"(i32* [[P1]], i64 4) ]
 ; BASIC-NEXT:    store i32 0, i32* [[P1]], align 4
 ; BASIC-NEXT:    ret i32 0
 ;
 ; ALL-LABEL: define {{[^@]+}}@_Z6squarePi
 ; ALL-SAME: (i32* [[P:%.*]], i32* [[P1:%.*]], i1 [[COND:%.*]])
 ; ALL-NEXT:    call void @llvm.assume(i1 true) [ "dereferenceable"(i32* [[P]], i64 4), "nonnull"(i32* [[P]]), "align"(i32* [[P]], i64 4) ]
 ; ALL-NEXT:    store i32 0, i32* [[P]], align 4
 ; ALL-NEXT:    call void @llvm.assume(i1 true) [ "dereferenceable"(i32* [[P1]], i64 4), "nonnull"(i32* [[P1]]), "align"(i32* [[P1]], i64 8) ]
 ; ALL-NEXT:    store i32 0, i32* [[P1]], align 8
 ; ALL-NEXT:    br i1 [[COND]], label [[A:%.*]], label [[B:%.*]]
 ; ALL:       A:
-; ALL-NEXT:    call void @llvm.assume(i1 true) [ "dereferenceable"(i32* [[P]], i64 4), "nonnull"(i32* [[P]]), "align"(i32* [[P]], i64 8) ]
+; ALL-NEXT:    call void @llvm.assume(i1 true) [ "align"(i32* [[P]], i64 8) ]
 ; ALL-NEXT:    store i32 0, i32* [[P]], align 8
-; ALL-NEXT:    call void @llvm.assume(i1 true) [ "dereferenceable"(i32* [[P1]], i64 4), "nonnull"(i32* [[P1]]), "align"(i32* [[P1]], i64 4) ]
 ; ALL-NEXT:    store i32 0, i32* [[P1]], align 4
 ; ALL-NEXT:    br i1 [[COND]], label [[C:%.*]], label [[B]]
 ; ALL:       B:
 ; ALL-NEXT:    call void @llvm.assume(i1 true) [ "dereferenceable"(i32* [[P]], i64 4), "nonnull"(i32* [[P]]), "align"(i32* [[P]], i64 8) ]
 ; ALL-NEXT:    store i32 0, i32* [[P]], align 8
 ; ALL-NEXT:    call void @llvm.assume(i1 true) [ "dereferenceable"(i32* [[P1]], i64 4), "nonnull"(i32* [[P1]]), "align"(i32* [[P1]], i64 8) ]
 ; ALL-NEXT:    store i32 0, i32* [[P1]], align 8
 ; ALL-NEXT:    br label [[C]]

 ; CROSS-BLOCK-NEXT:    ret i32 0
 ;
 ; FULL-SIMPLIFY-LABEL: define {{[^@]+}}@_Z6squarePi
 ; FULL-SIMPLIFY-SAME: (i32* nonnull align 4 dereferenceable(4) [[P:%.*]], i32* nonnull align 8 dereferenceable(4) [[P1:%.*]], i1 [[COND:%.*]])
 ; FULL-SIMPLIFY-NEXT:    store i32 0, i32* [[P]], align 4
 ; FULL-SIMPLIFY-NEXT:    store i32 0, i32* [[P1]], align 8
 ; FULL-SIMPLIFY-NEXT:    br i1 [[COND]], label [[A:%.*]], label [[B:%.*]]
 ; FULL-SIMPLIFY:       A:
-; FULL-SIMPLIFY-NEXT:    call void @llvm.assume(i1 true) [ "ignore"(i32* undef, i64 4), "ignore"(i32* undef), "align"(i32* [[P]], i64 8) ]
+; FULL-SIMPLIFY-NEXT:    call void @llvm.assume(i1 true) [ "align"(i32* [[P]], i64 8) ]
 ; FULL-SIMPLIFY-NEXT:    store i32 0, i32* [[P]], align 8
 ; FULL-SIMPLIFY-NEXT:    store i32 0, i32* [[P1]], align 4
 ; FULL-SIMPLIFY-NEXT:    br i1 [[COND]], label [[C:%.*]], label [[B]]
 ; FULL-SIMPLIFY:       B:
 ; FULL-SIMPLIFY-NEXT:    call void @llvm.assume(i1 true) [ "ignore"(i32* undef, i64 4), "ignore"(i32* undef), "align"(i32* [[P]], i64 8) ]
 ; FULL-SIMPLIFY-NEXT:    store i32 0, i32* [[P]], align 8
 ; FULL-SIMPLIFY-NEXT:    store i32 0, i32* [[P1]], align 8
 ; FULL-SIMPLIFY-NEXT:    br label [[C]]

 ; BASIC-NEXT:    [[TMP5:%.*]] = icmp ne i32* [[TMP1]], null
 ; BASIC-NEXT:    br i1 [[TMP5]], label [[TMP6:%.*]], label [[TMP12:%.*]]
 ; BASIC:       6:
 ; BASIC-NEXT:    [[TMP7:%.*]] = add nsw i32 [[TMP3]], [[TMP2]]
 ; BASIC-NEXT:    call void @may_throw()
 ; BASIC-NEXT:    call void @llvm.assume(i1 true) [ "dereferenceable"(i32* [[TMP0]], i64 4), "nonnull"(i32* [[TMP0]]), "align"(i32* [[TMP0]], i64 4) ]
 ; BASIC-NEXT:    [[TMP8:%.*]] = load i32, i32* [[TMP0]], align 4
 ; BASIC-NEXT:    [[TMP9:%.*]] = add nsw i32 [[TMP7]], [[TMP8]]
-; BASIC-NEXT:    call void @llvm.assume(i1 true) [ "dereferenceable"(i32* [[TMP0]], i64 4), "nonnull"(i32* [[TMP0]]), "align"(i32* [[TMP0]], i64 4) ]
 ; BASIC-NEXT:    store i32 0, i32* [[TMP0]], align 4
 ; BASIC-NEXT:    call void @llvm.assume(i1 true) [ "dereferenceable"(i32* [[TMP1]], i64 4), "nonnull"(i32* [[TMP1]]), "align"(i32* [[TMP1]], i64 4) ]
 ; BASIC-NEXT:    [[TMP10:%.*]] = load i32, i32* [[TMP1]], align 4
 ; BASIC-NEXT:    [[TMP11:%.*]] = add nsw i32 [[TMP9]], [[TMP10]]
 ; BASIC-NEXT:    call void @may_throw()
-; BASIC-NEXT:    call void @llvm.assume(i1 true) [ "dereferenceable"(i32* [[TMP1]], i64 4), "nonnull"(i32* [[TMP1]]), "align"(i32* [[TMP1]], i64 4) ]
 ; BASIC-NEXT:    store i32 [[TMP11]], i32* [[TMP1]], align 4
 ; BASIC-NEXT:    br label [[TMP12]]
 ; BASIC:       12:
 ; BASIC-NEXT:    ret i32 0
 ;
 ; ALL-LABEL: define {{[^@]+}}@test4A
 ; ALL-SAME: (i32* [[TMP0:%.*]], i32* [[TMP1:%.*]], i32 [[TMP2:%.*]], i32 [[TMP3:%.*]])
 ; ALL-NEXT:    [[TMP5:%.*]] = icmp ne i32* [[TMP1]], null
 ; ALL-NEXT:    br i1 [[TMP5]], label [[TMP6:%.*]], label [[TMP12:%.*]]
 ; ALL:       6:
 ; ALL-NEXT:    [[TMP7:%.*]] = add nsw i32 [[TMP3]], [[TMP2]]
 ; ALL-NEXT:    call void @may_throw()
 ; ALL-NEXT:    call void @llvm.assume(i1 true) [ "dereferenceable"(i32* [[TMP0]], i64 4), "nonnull"(i32* [[TMP0]]), "align"(i32* [[TMP0]], i64 4) ]
 ; ALL-NEXT:    [[TMP8:%.*]] = load i32, i32* [[TMP0]], align 4
 ; ALL-NEXT:    [[TMP9:%.*]] = add nsw i32 [[TMP7]], [[TMP8]]
-; ALL-NEXT:    call void @llvm.assume(i1 true) [ "dereferenceable"(i32* [[TMP0]], i64 4), "nonnull"(i32* [[TMP0]]), "align"(i32* [[TMP0]], i64 4) ]
 ; ALL-NEXT:    store i32 0, i32* [[TMP0]], align 4
 ; ALL-NEXT:    call void @llvm.assume(i1 true) [ "dereferenceable"(i32* [[TMP1]], i64 4), "nonnull"(i32* [[TMP1]]), "align"(i32* [[TMP1]], i64 4) ]
 ; ALL-NEXT:    [[TMP10:%.*]] = load i32, i32* [[TMP1]], align 4
 ; ALL-NEXT:    [[TMP11:%.*]] = add nsw i32 [[TMP9]], [[TMP10]]
 ; ALL-NEXT:    call void @may_throw()
-; ALL-NEXT:    call void @llvm.assume(i1 true) [ "dereferenceable"(i32* [[TMP1]], i64 4), "nonnull"(i32* [[TMP1]]), "align"(i32* [[TMP1]], i64 4) ]
 ; ALL-NEXT:    store i32 [[TMP11]], i32* [[TMP1]], align 4
 ; ALL-NEXT:    br label [[TMP12]]
 ; ALL:       12:
 ; ALL-NEXT:    ret i32 0
 ;
 ; WITH-AC-LABEL: define {{[^@]+}}@test4A
 ; WITH-AC-SAME: (i32* [[TMP0:%.*]], i32* [[TMP1:%.*]], i32 [[TMP2:%.*]], i32 [[TMP3:%.*]])
 ; WITH-AC-NEXT:    [[TMP5:%.*]] = icmp ne i32* [[TMP1]], null
Catch:                                                ; preds = %0
   br label %Exit
 
 Exit:                                                ; preds = %7, %5
   %.0 = phi i32 [ 1, %0 ], [ 0, %Catch ]
   ret i32 %.0
 }
 
 declare dso_local i32 @__gxx_personality_v0(...)
+
+define dso_local i32 @test5(i8* %0, i32 %1) {
+; BASIC-LABEL: define {{[^@]+}}@test5
+; BASIC-SAME: (i8* [[TMP0:%.*]], i32 [[TMP1:%.*]])
+; BASIC-NEXT:    [[TMP3:%.*]] = bitcast i8* [[TMP0]] to i64*
+; BASIC-NEXT:    [[TMP4:%.*]] = bitcast i8* [[TMP0]] to i16*
+; BASIC-NEXT:    [[TMP5:%.*]] = sext i32 [[TMP1]] to i64
+; BASIC-NEXT:    [[TMP6:%.*]] = getelementptr inbounds i16, i16* [[TMP4]], i64 [[TMP5]]
+; BASIC-NEXT:    call void @llvm.assume(i1 true) [ "dereferenceable"(i8* [[TMP0]], i64 8), "nonnull"(i8* [[TMP0]]), "align"(i8* [[TMP0]], i64 8) ]
+; BASIC-NEXT:    [[TMP7:%.*]] = load i16, i16* [[TMP6]], align 2
+; BASIC-NEXT:    [[A:%.*]] = load i16, i16* [[TMP6]], align 4
+; BASIC-NEXT:    [[TMP8:%.*]] = sext i16 [[TMP7]] to i64
+; BASIC-NEXT:    [[TMP9:%.*]] = getelementptr inbounds i64, i64* [[TMP3]], i64 [[TMP8]]
+; BASIC-NEXT:    [[TMP10:%.*]] = load i64, i64* [[TMP9]], align 16
+; BASIC-NEXT:    [[B:%.*]] = load i64, i64* [[TMP9]], align 32
+; BASIC-NEXT:    [[TMP11:%.*]] = trunc i64 [[TMP10]] to i32
+; BASIC-NEXT:    ret i32 [[TMP11]]
+;
+; ALL-LABEL: define {{[^@]+}}@test5
+; ALL-SAME: (i8* [[TMP0:%.*]], i32 [[TMP1:%.*]])
+; ALL-NEXT:    [[TMP3:%.*]] = bitcast i8* [[TMP0]] to i64*
+; ALL-NEXT:    [[TMP4:%.*]] = bitcast i8* [[TMP0]] to i16*
+; ALL-NEXT:    [[TMP5:%.*]] = sext i32 [[TMP1]] to i64
+; ALL-NEXT:    [[TMP6:%.*]] = getelementptr inbounds i16, i16* [[TMP4]], i64 [[TMP5]]
+; ALL-NEXT:    call void @llvm.assume(i1 true) [ "dereferenceable"(i8* [[TMP0]], i64 8), "nonnull"(i8* [[TMP0]]), "align"(i8* [[TMP0]], i64 8) ]
+; ALL-NEXT:    [[TMP7:%.*]] = load i16, i16* [[TMP6]], align 2
+; ALL-NEXT:    [[A:%.*]] = load i16, i16* [[TMP6]], align 4
+; ALL-NEXT:    [[TMP8:%.*]] = sext i16 [[TMP7]] to i64
+; ALL-NEXT:    [[TMP9:%.*]] = getelementptr inbounds i64, i64* [[TMP3]], i64 [[TMP8]]
+; ALL-NEXT:    [[TMP10:%.*]] = load i64, i64* [[TMP9]], align 16
+; ALL-NEXT:    [[B:%.*]] = load i64, i64* [[TMP9]], align 32
+; ALL-NEXT:    [[TMP11:%.*]] = trunc i64 [[TMP10]] to i32
+; ALL-NEXT:    ret i32 [[TMP11]]
+;
+; WITH-AC-LABEL: define {{[^@]+}}@test5
+; WITH-AC-SAME: (i8* [[TMP0:%.*]], i32 [[TMP1:%.*]])
+; WITH-AC-NEXT:    [[TMP3:%.*]] = bitcast i8* [[TMP0]] to i64*
+; WITH-AC-NEXT:    [[TMP4:%.*]] = bitcast i8* [[TMP0]] to i16*
+; WITH-AC-NEXT:    [[TMP5:%.*]] = sext i32 [[TMP1]] to i64
+; WITH-AC-NEXT:    [[TMP6:%.*]] = getelementptr inbounds i16, i16* [[TMP4]], i64 [[TMP5]]
+; WITH-AC-NEXT:    call void @llvm.assume(i1 true) [ "dereferenceable"(i8* [[TMP0]], i64 8), "nonnull"(i8* [[TMP0]]), "align"(i8* [[TMP0]], i64 8) ]
+; WITH-AC-NEXT:    [[TMP7:%.*]] = load i16, i16* [[TMP6]], align 2
+; WITH-AC-NEXT:    [[A:%.*]] = load i16, i16* [[TMP6]], align 4
+; WITH-AC-NEXT:    [[TMP8:%.*]] = sext i16 [[TMP7]] to i64
+; WITH-AC-NEXT:    [[TMP9:%.*]] = getelementptr inbounds i64, i64* [[TMP3]], i64 [[TMP8]]
+; WITH-AC-NEXT:    [[TMP10:%.*]] = load i64, i64* [[TMP9]], align 16
+; WITH-AC-NEXT:    [[B:%.*]] = load i64, i64* [[TMP9]], align 32
+; WITH-AC-NEXT:    [[TMP11:%.*]] = trunc i64 [[TMP10]] to i32
+; WITH-AC-NEXT:    ret i32 [[TMP11]]
+;
+; CROSS-BLOCK-LABEL: define {{[^@]+}}@test5
+; CROSS-BLOCK-SAME: (i8* [[TMP0:%.*]], i32 [[TMP1:%.*]])
+; CROSS-BLOCK-NEXT:    [[TMP3:%.*]] = bitcast i8* [[TMP0]] to i64*
+; CROSS-BLOCK-NEXT:    [[TMP4:%.*]] = bitcast i8* [[TMP0]] to i16*
+; CROSS-BLOCK-NEXT:    [[TMP5:%.*]] = sext i32 [[TMP1]] to i64
+; CROSS-BLOCK-NEXT:    [[TMP6:%.*]] = getelementptr inbounds i16, i16* [[TMP4]], i64 [[TMP5]]
+; CROSS-BLOCK-NEXT:    call void @llvm.assume(i1 true) [ "dereferenceable"(i8* [[TMP0]], i64 8), "nonnull"(i8* [[TMP0]]), "align"(i8* [[TMP0]], i64 8) ]
+; CROSS-BLOCK-NEXT:    [[TMP7:%.*]] = load i16, i16* [[TMP6]], align 2
+; CROSS-BLOCK-NEXT:    [[A:%.*]] = load i16, i16* [[TMP6]], align 4
+; CROSS-BLOCK-NEXT:    [[TMP8:%.*]] = sext i16 [[TMP7]] to i64
+; CROSS-BLOCK-NEXT:    [[TMP9:%.*]] = getelementptr inbounds i64, i64* [[TMP3]], i64 [[TMP8]]
+; CROSS-BLOCK-NEXT:    [[TMP10:%.*]] = load i64, i64* [[TMP9]], align 16
+; CROSS-BLOCK-NEXT:    [[B:%.*]] = load i64, i64* [[TMP9]], align 32
+; CROSS-BLOCK-NEXT:    [[TMP11:%.*]] = trunc i64 [[TMP10]] to i32
+; CROSS-BLOCK-NEXT:    ret i32 [[TMP11]]
+;
+; FULL-SIMPLIFY-LABEL: define {{[^@]+}}@test5
+; FULL-SIMPLIFY-SAME: (i8* nonnull align 8 dereferenceable(8) [[TMP0:%.*]], i32 [[TMP1:%.*]])
+; FULL-SIMPLIFY-NEXT:    [[TMP3:%.*]] = bitcast i8* [[TMP0]] to i64*
+; FULL-SIMPLIFY-NEXT:    [[TMP4:%.*]] = bitcast i8* [[TMP0]] to i16*
+; FULL-SIMPLIFY-NEXT:    [[TMP5:%.*]] = sext i32 [[TMP1]] to i64
+; FULL-SIMPLIFY-NEXT:    [[TMP6:%.*]] = getelementptr inbounds i16, i16* [[TMP4]], i64 [[TMP5]]
+; FULL-SIMPLIFY-NEXT:    [[TMP7:%.*]] = load i16, i16* [[TMP6]], align 2
+; FULL-SIMPLIFY-NEXT:    [[A:%.*]] = load i16, i16* [[TMP6]], align 4
+; FULL-SIMPLIFY-NEXT:    [[TMP8:%.*]] = sext i16 [[TMP7]] to i64
+; FULL-SIMPLIFY-NEXT:    [[TMP9:%.*]] = getelementptr inbounds i64, i64* [[TMP3]], i64 [[TMP8]]
+; FULL-SIMPLIFY-NEXT:    [[TMP10:%.*]] = load i64, i64* [[TMP9]], align 16
+; FULL-SIMPLIFY-NEXT:    [[B:%.*]] = load i64, i64* [[TMP9]], align 32
+; FULL-SIMPLIFY-NEXT:    [[TMP11:%.*]] = trunc i64 [[TMP10]] to i32
+; FULL-SIMPLIFY-NEXT:    ret i32 [[TMP11]]
+;
+  %3 = bitcast i8* %0 to i64*
+  %4 = bitcast i8* %0 to i16*
+  %5 = sext i32 %1 to i64
+  %6 = getelementptr inbounds i16, i16* %4, i64 %5
+  %7 = load i16, i16* %6, align 2
+  %A = load i16, i16* %6, align 4
+  %8 = sext i16 %7 to i64
+  %9 = getelementptr inbounds i64, i64* %3, i64 %8
+  %10 = load i64, i64* %9, align 16
+  %B = load i64, i64* %9, align 32
+  %11 = trunc i64 %10 to i32
+  ret i32 %11
+}
+
+define dso_local i32 @test6(i32* %0, i32 %1, i32* %2) {
+; BASIC-LABEL: define {{[^@]+}}@test6
+; BASIC-SAME: (i32* [[TMP0:%.*]], i32 [[TMP1:%.*]], i32* [[TMP2:%.*]])
+; BASIC-NEXT:    br label [[TMP4:%.*]]
+; BASIC:       4:
+; BASIC-NEXT:    [[DOT0:%.*]] = phi i32 [ 0, [[TMP3:%.*]] ], [ [[TMP16:%.*]], [[TMP6:%.*]] ]
+; BASIC-NEXT:    [[TMP5:%.*]] = icmp slt i32 [[DOT0]], [[TMP1]]
+; BASIC-NEXT:    br i1 [[TMP5]], label [[TMP6]], label [[TMP17:%.*]]
+; BASIC:       6:
+; BASIC-NEXT:    [[TMP7:%.*]] = add nsw i32 [[TMP1]], [[DOT0]]
+; BASIC-NEXT:    [[TMP8:%.*]] = sext i32 [[TMP7]] to i64
+; BASIC-NEXT:    [[TMP9:%.*]] = getelementptr inbounds i32, i32* [[TMP0]], i64 [[TMP8]]
+; BASIC-NEXT:    call void @llvm.assume(i1 true) [ "dereferenceable"(i32* [[TMP0]], i64 4), "nonnull"(i32* [[TMP0]]), "align"(i32* [[TMP0]], i64 4) ]
+; BASIC-NEXT:    [[TMP10:%.*]] = load i32, i32* [[TMP9]], align 4
+; BASIC-NEXT:    [[TMP11:%.*]] = mul nsw i32 [[DOT0]], [[TMP10]]
+; BASIC-NEXT:    [[TMP12:%.*]] = sext i32 [[DOT0]] to i64
+; BASIC-NEXT:    [[TMP13:%.*]] = getelementptr inbounds i32, i32* [[TMP0]], i64 [[TMP12]]
+; BASIC-NEXT:    [[TMP14:%.*]] = load i32, i32* [[TMP13]], align 4
+; BASIC-NEXT:    [[TMP15:%.*]] = add nsw i32 [[TMP14]], [[TMP11]]
+; BASIC-NEXT:    store i32 [[TMP15]], i32* [[TMP13]], align 4
+; BASIC-NEXT:    [[TMP16]] = add nsw i32 [[DOT0]], 1
+; BASIC-NEXT:    br label [[TMP4]]
+; BASIC:       17:
+; BASIC-NEXT:    [[TMP18:%.*]] = sext i32 [[TMP1]] to i64
+; BASIC-NEXT:    [[TMP19:%.*]] = getelementptr inbounds i32, i32* [[TMP2]], i64 [[TMP18]]
+; BASIC-NEXT:    call void @llvm.assume(i1 true) [ "dereferenceable"(i32* [[TMP2]], i64 4), "nonnull"(i32* [[TMP2]]), "align"(i32* [[TMP2]], i64 4) ]
+; BASIC-NEXT:    [[TMP20:%.*]] = load i32, i32* [[TMP19]], align 4
+; BASIC-NEXT:    ret i32 [[TMP20]]
+;
+; ALL-LABEL: define {{[^@]+}}@test6
+; ALL-SAME: (i32* [[TMP0:%.*]], i32 [[TMP1:%.*]], i32* [[TMP2:%.*]])
+; ALL-NEXT:    br label [[TMP4:%.*]]
+; ALL:       4:
+; ALL-NEXT:    [[DOT0:%.*]] = phi i32 [ 0, [[TMP3:%.*]] ], [ [[TMP16:%.*]], [[TMP6:%.*]] ]
+; ALL-NEXT:    [[TMP5:%.*]] = icmp slt i32 [[DOT0]], [[TMP1]]
+; ALL-NEXT:    br i1 [[TMP5]], label [[TMP6]], label [[TMP17:%.*]]
+; ALL:       6:
+; ALL-NEXT:    [[TMP7:%.*]] = add nsw i32 [[TMP1]], [[DOT0]]
+; ALL-NEXT:    [[TMP8:%.*]] = sext i32 [[TMP7]] to i64
+; ALL-NEXT:    [[TMP9:%.*]] = getelementptr inbounds i32, i32* [[TMP0]], i64 [[TMP8]]
+; ALL-NEXT:    call void @llvm.assume(i1 true) [ "dereferenceable"(i32* [[TMP0]], i64 4), "nonnull"(i32* [[TMP0]]), "align"(i32* [[TMP0]], i64 4) ]
+; ALL-NEXT:    [[TMP10:%.*]] = load i32, i32* [[TMP9]], align 4
+; ALL-NEXT:    [[TMP11:%.*]] = mul nsw i32 [[DOT0]], [[TMP10]]
+; ALL-NEXT:    [[TMP12:%.*]] = sext i32 [[DOT0]] to i64
+; ALL-NEXT:    [[TMP13:%.*]] = getelementptr inbounds i32, i32* [[TMP0]], i64 [[TMP12]]
+; ALL-NEXT:    [[TMP14:%.*]] = load i32, i32* [[TMP13]], align 4
+; ALL-NEXT:    [[TMP15:%.*]] = add nsw i32 [[TMP14]], [[TMP11]]
+; ALL-NEXT:    store i32 [[TMP15]], i32* [[TMP13]], align 4
+; ALL-NEXT:    [[TMP16]] = add nsw i32 [[DOT0]], 1
+; ALL-NEXT:    br label [[TMP4]]
+; ALL:       17:
+; ALL-NEXT:    [[TMP18:%.*]] = sext i32 [[TMP1]] to i64
+; ALL-NEXT:    [[TMP19:%.*]] = getelementptr inbounds i32, i32* [[TMP2]], i64 [[TMP18]]
+; ALL-NEXT:    call void @llvm.assume(i1 true) [ "dereferenceable"(i32* [[TMP2]], i64 4), "nonnull"(i32* [[TMP2]]), "align"(i32* [[TMP2]], i64 4) ]
+; ALL-NEXT:    [[TMP20:%.*]] = load i32, i32* [[TMP19]], align 4
+; ALL-NEXT:    ret i32 [[TMP20]]
+;
+; WITH-AC-LABEL: define {{[^@]+}}@test6
+; WITH-AC-SAME: (i32* [[TMP0:%.*]], i32 [[TMP1:%.*]], i32* [[TMP2:%.*]])
+; WITH-AC-NEXT:    br label [[TMP4:%.*]]
+; WITH-AC:       4:
+; WITH-AC-NEXT:    [[DOT0:%.*]] = phi i32 [ 0, [[TMP3:%.*]] ], [ [[TMP16:%.*]], [[TMP6:%.*]] ]
+; WITH-AC-NEXT:    [[TMP5:%.*]] = icmp slt i32 [[DOT0]], [[TMP1]]
+; WITH-AC-NEXT:    br i1 [[TMP5]], label [[TMP6]], label [[TMP17:%.*]]
+; WITH-AC:       6:
+; WITH-AC-NEXT:    [[TMP7:%.*]] = add nsw i32 [[TMP1]], [[DOT0]]
+; WITH-AC-NEXT:    [[TMP8:%.*]] = sext i32 [[TMP7]] to i64
+; WITH-AC-NEXT:    [[TMP9:%.*]] = getelementptr inbounds i32, i32* [[TMP0]], i64 [[TMP8]]
+; WITH-AC-NEXT:    call void @llvm.assume(i1 true) [ "dereferenceable"(i32* [[TMP0]], i64 4), "nonnull"(i32* [[TMP0]]), "align"(i32* [[TMP0]], i64 4) ]
+; WITH-AC-NEXT:    [[TMP10:%.*]] = load i32, i32* [[TMP9]], align 4
+; WITH-AC-NEXT:    [[TMP11:%.*]] = mul nsw i32 [[DOT0]], [[TMP10]]
+; WITH-AC-NEXT:    [[TMP12:%.*]] = sext i32 [[DOT0]] to i64
+; WITH-AC-NEXT:    [[TMP13:%.*]] = getelementptr inbounds i32, i32* [[TMP0]], i64 [[TMP12]]
+; WITH-AC-NEXT:    [[TMP14:%.*]] = load i32, i32* [[TMP13]], align 4
+; WITH-AC-NEXT:    [[TMP15:%.*]] = add nsw i32 [[TMP14]], [[TMP11]]
+; WITH-AC-NEXT:    store i32 [[TMP15]], i32* [[TMP13]], align 4
+; WITH-AC-NEXT:    [[TMP16]] = add nsw i32 [[DOT0]], 1
+; WITH-AC-NEXT:    br label [[TMP4]]
+; WITH-AC:       17:
+; WITH-AC-NEXT:    [[TMP18:%.*]] = sext i32 [[TMP1]] to i64
+; WITH-AC-NEXT:    [[TMP19:%.*]] = getelementptr inbounds i32, i32* [[TMP2]], i64 [[TMP18]]
+; WITH-AC-NEXT:    call void @llvm.assume(i1 true) [ "dereferenceable"(i32* [[TMP2]], i64 4), "nonnull"(i32* [[TMP2]]), "align"(i32* [[TMP2]], i64 4) ]
+; WITH-AC-NEXT:    [[TMP20:%.*]] = load i32, i32* [[TMP19]], align 4
+; WITH-AC-NEXT:    ret i32 [[TMP20]]
+;
+; CROSS-BLOCK-LABEL: define {{[^@]+}}@test6
+; CROSS-BLOCK-SAME: (i32* [[TMP0:%.*]], i32 [[TMP1:%.*]], i32* [[TMP2:%.*]])
+; CROSS-BLOCK-NEXT:    br label [[TMP4:%.*]]
+; CROSS-BLOCK:       4:
+; CROSS-BLOCK-NEXT:    [[DOT0:%.*]] = phi i32 [ 0, [[TMP3:%.*]] ], [ [[TMP16:%.*]], [[TMP6:%.*]] ]
+; CROSS-BLOCK-NEXT:    [[TMP5:%.*]] = icmp slt i32 [[DOT0]], [[TMP1]]
+; CROSS-BLOCK-NEXT:    br i1 [[TMP5]], label [[TMP6]], label [[TMP17:%.*]]
+; CROSS-BLOCK:       6:
+; CROSS-BLOCK-NEXT:    [[TMP7:%.*]] = add nsw i32 [[TMP1]], [[DOT0]]
+; CROSS-BLOCK-NEXT:    [[TMP8:%.*]] = sext i32 [[TMP7]] to i64
+; CROSS-BLOCK-NEXT:    [[TMP9:%.*]] = getelementptr inbounds i32, i32* [[TMP0]], i64 [[TMP8]]
+; CROSS-BLOCK-NEXT:    call void @llvm.assume(i1 true) [ "dereferenceable"(i32* [[TMP0]], i64 4), "nonnull"(i32* [[TMP0]]), "align"(i32* [[TMP0]], i64 4) ]
+; CROSS-BLOCK-NEXT:    [[TMP10:%.*]] = load i32, i32* [[TMP9]], align 4
+; CROSS-BLOCK-NEXT:    [[TMP11:%.*]] = mul nsw i32 [[DOT0]], [[TMP10]]
+; CROSS-BLOCK-NEXT:    [[TMP12:%.*]] = sext i32 [[DOT0]] to i64
+; CROSS-BLOCK-NEXT:    [[TMP13:%.*]] = getelementptr inbounds i32, i32* [[TMP0]], i64 [[TMP12]]
+; CROSS-BLOCK-NEXT:    [[TMP14:%.*]] = load i32, i32* [[TMP13]], align 4
+; CROSS-BLOCK-NEXT:    [[TMP15:%.*]] = add nsw i32 [[TMP14]], [[TMP11]]
+; CROSS-BLOCK-NEXT:    store i32 [[TMP15]], i32* [[TMP13]], align 4
+; CROSS-BLOCK-NEXT:    [[TMP16]] = add nsw i32 [[DOT0]], 1
+; CROSS-BLOCK-NEXT:    br label [[TMP4]]
+; CROSS-BLOCK:       17:
+; CROSS-BLOCK-NEXT:    [[TMP18:%.*]] = sext i32 [[TMP1]] to i64
+; CROSS-BLOCK-NEXT:    [[TMP19:%.*]] = getelementptr inbounds i32, i32* [[TMP2]], i64 [[TMP18]]
+; CROSS-BLOCK-NEXT:    call void @llvm.assume(i1 true) [ "dereferenceable"(i32* [[TMP2]], i64 4), "nonnull"(i32* [[TMP2]]), "align"(i32* [[TMP2]], i64 4) ]
+; CROSS-BLOCK-NEXT:    [[TMP20:%.*]] = load i32, i32* [[TMP19]], align 4
+; CROSS-BLOCK-NEXT:    ret i32 [[TMP20]]
+;
+; FULL-SIMPLIFY-LABEL: define {{[^@]+}}@test6
+; FULL-SIMPLIFY-SAME: (i32* [[TMP0:%.*]], i32 [[TMP1:%.*]], i32* [[TMP2:%.*]])
+; FULL-SIMPLIFY-NEXT:    br label [[TMP4:%.*]]
+; FULL-SIMPLIFY:       4:
+; FULL-SIMPLIFY-NEXT:    [[DOT0:%.*]] = phi i32 [ 0, [[TMP3:%.*]] ], [ [[TMP16:%.*]], [[TMP6:%.*]] ]
+; FULL-SIMPLIFY-NEXT:    [[TMP5:%.*]] = icmp slt i32 [[DOT0]], [[TMP1]]
+; FULL-SIMPLIFY-NEXT:    br i1 [[TMP5]], label [[TMP6]], label [[TMP17:%.*]]
+; FULL-SIMPLIFY:       6:
+; FULL-SIMPLIFY-NEXT:    [[TMP7:%.*]] = add nsw i32 [[TMP1]], [[DOT0]]
+; FULL-SIMPLIFY-NEXT:    [[TMP8:%.*]] = sext i32 [[TMP7]] to i64
+; FULL-SIMPLIFY-NEXT:    [[TMP9:%.*]] = getelementptr inbounds i32, i32* [[TMP0]], i64 [[TMP8]]
+; FULL-SIMPLIFY-NEXT:    call void @llvm.assume(i1 true) [ "dereferenceable"(i32* [[TMP0]], i64 4), "nonnull"(i32* [[TMP0]]), "align"(i32* [[TMP0]], i64 4) ]
+; FULL-SIMPLIFY-NEXT:    [[TMP10:%.*]] = load i32, i32* [[TMP9]], align 4
+; FULL-SIMPLIFY-NEXT:    [[TMP11:%.*]] = mul nsw i32 [[DOT0]], [[TMP10]]
+; FULL-SIMPLIFY-NEXT:    [[TMP12:%.*]] = sext i32 [[DOT0]] to i64
+; FULL-SIMPLIFY-NEXT:    [[TMP13:%.*]] = getelementptr inbounds i32, i32* [[TMP0]], i64 [[TMP12]]
+; FULL-SIMPLIFY-NEXT:    [[TMP14:%.*]] = load i32, i32* [[TMP13]], align 4
+; FULL-SIMPLIFY-NEXT:    [[TMP15:%.*]] = add nsw i32 [[TMP14]], [[TMP11]]
+; FULL-SIMPLIFY-NEXT:    store i32 [[TMP15]], i32* [[TMP13]], align 4
+; FULL-SIMPLIFY-NEXT:    [[TMP16]] = add nsw i32 [[DOT0]], 1
+; FULL-SIMPLIFY-NEXT:    br label [[TMP4]]
+; FULL-SIMPLIFY:       17:
+; FULL-SIMPLIFY-NEXT:    [[TMP18:%.*]] = sext i32 [[TMP1]] to i64
+; FULL-SIMPLIFY-NEXT:    [[TMP19:%.*]] = getelementptr inbounds i32, i32* [[TMP2]], i64 [[TMP18]]
+; FULL-SIMPLIFY-NEXT:    call void @llvm.assume(i1 true) [ "dereferenceable"(i32* [[TMP2]], i64 4), "nonnull"(i32* [[TMP2]]), "align"(i32* [[TMP2]], i64 4) ]
+; FULL-SIMPLIFY-NEXT:    [[TMP20:%.*]] = load i32, i32* [[TMP19]], align 4
+; FULL-SIMPLIFY-NEXT:    ret i32 [[TMP20]]
+;
+  br label %4
+
+4:                                                ; preds = %6, %3
+  %.0 = phi i32 [ 0, %3 ], [ %16, %6 ]
+  %5 = icmp slt i32 %.0, %1
+  br i1 %5, label %6, label %17
+
+6:                                                ; preds = %4
+  %7 = add nsw i32 %1, %.0
+  %8 = sext i32 %7 to i64
+  %9 = getelementptr inbounds i32, i32* %0, i64 %8
+  %10 = load i32, i32* %9, align 4
+  %11 = mul nsw i32 %.0, %10
+  %12 = sext i32 %.0 to i64
+  %13 = getelementptr inbounds i32, i32* %0, i64 %12
+  %14 = load i32, i32* %13, align 4
+  %15 = add nsw i32 %14, %11
+  store i32 %15, i32* %13, align 4
+  %16 = add nsw i32 %.0, 1
+  br label %4
+
+17:                                               ; preds = %4
+  %18 = sext i32 %1 to i64
+  %19 = getelementptr inbounds i32, i32* %2, i64 %18
+  %20 = load i32, i32* %19, align 4
+  ret i32 %20
+}

llvm/test/Transforms/Util/assume-simplify.ll

 ; CHECK-NEXT:    [[TMP10:%.*]] = load i32, i32* [[TMP8]], align 4
 ; CHECK-NEXT:    [[TMP11:%.*]] = getelementptr inbounds i32, i32* [[TMP1]], i64 2
 ; CHECK-NEXT:    store i32 [[TMP10]], i32* [[TMP11]], align 4
 ; CHECK-NEXT:    call void @may_throw()
 ; CHECK-NEXT:    call void @may_throw()
 ; CHECK-NEXT:    [[TMP12:%.*]] = getelementptr inbounds i32*, i32** [[TMP0]], i64 1
 ; CHECK-NEXT:    [[TMP13:%.*]] = load i32*, i32** [[TMP12]], align 8
 ; CHECK-NEXT:    [[TMP14:%.*]] = getelementptr inbounds i32, i32* [[TMP13]], i64 0
-; CHECK-NEXT:    call void @llvm.assume(i1 true) [ "align"(i32* [[TMP11]], i64 4), "dereferenceable"(i32* [[TMP11]], i64 4), "nonnull"(i32* [[TMP11]]), "align"(i32* [[TMP14]], i64 4), "dereferenceable"(i32* [[TMP14]], i64 4), "nonnull"(i32* [[TMP14]]) ]
+; CHECK-NEXT:    call void @llvm.assume(i1 true) [ "align"(i32* [[TMP1]], i64 4), "dereferenceable"(i32* [[TMP1]], i64 12), "nonnull"(i32* [[TMP1]]), "align"(i32* [[TMP13]], i64 4), "dereferenceable"(i32* [[TMP13]], i64 4), "nonnull"(i32* [[TMP13]]) ]
 ; CHECK-NEXT:    [[TMP15:%.*]] = load i32, i32* [[TMP14]], align 4
 ; CHECK-NEXT:    [[TMP16:%.*]] = getelementptr inbounds i32*, i32** [[TMP0]], i64 1
 ; CHECK-NEXT:    [[TMP17:%.*]] = load i32*, i32** [[TMP16]], align 8
 ; CHECK-NEXT:    [[TMP18:%.*]] = getelementptr inbounds i32, i32* [[TMP17]], i64 2
 ; CHECK-NEXT:    store i32 [[TMP15]], i32* [[TMP18]], align 4
 ; CHECK-NEXT:    call void @may_throw()
 ; CHECK-NEXT:    call void @llvm.assume(i1 true) [ "align"(i32** [[TMP0]], i64 4), "dereferenceable"(i32** [[TMP0]], i64 4), "nonnull"(i32** [[TMP0]]) ]
 ; CHECK-NEXT:    br label [[TMP35:%.*]]
;
   call void @llvm.assume(i1 true) [ "cold"() ]
   ret i32 0
 }
 
 define i32 @test7(i32* %p) {
 ; CHECK-LABEL: define {{[^@]+}}@test7
 ; CHECK-SAME: (i32* align 4 dereferenceable(4) [[P:%.*]])
 ; CHECK-NEXT:    [[P1:%.*]] = bitcast i32* [[P]] to i8*
-; CHECK-NEXT:    call void @llvm.assume(i1 true) [ "cold"(), "align"(i8* [[P1]], i64 4), "nonnull"(i8* [[P1]]) ]
+; CHECK-NEXT:    call void @llvm.assume(i1 true) [ "cold"(), "align"(i32* [[P]], i64 4), "nonnull"(i32* [[P]]) ]
 ; CHECK-NEXT:    ret i32 0
 ;
   %p1 = bitcast i32* %p to i8*
   call void @llvm.assume(i1 true) [ "cold"() ]
   call void @llvm.assume(i1 true) [ "align"(i32* %p, i32 4) ]
   call void @llvm.assume(i1 true) [ "dereferenceable"(i32* %p, i32 4) ]
   call void @llvm.assume(i1 true) [ "align"(i8* %p1, i32 4), "nonnull"(i8* %p1) ]
   ret i32 0
 }