diff --git a/llvm/include/llvm/IR/Operator.h b/llvm/include/llvm/IR/Operator.h
--- a/llvm/include/llvm/IR/Operator.h
+++ b/llvm/include/llvm/IR/Operator.h
@@ -545,6 +545,9 @@
       });
   }
 
+  /// 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
diff --git a/llvm/include/llvm/IR/Value.h b/llvm/include/llvm/IR/Value.h
--- a/llvm/include/llvm/IR/Value.h
+++ b/llvm/include/llvm/IR/Value.h
@@ -640,10 +640,12 @@
   ///
   /// Returns the original pointer value.  If this is called on a non-pointer
   /// value, it returns 'this'.
-  const Value *stripInBoundsOffsets() const;
-  Value *stripInBoundsOffsets() {
+  const Value *stripInBoundsOffsets(function_ref<void(const Value *)> Func =
+                                        [](const Value *) {}) const;
+  Value *stripInBoundsOffsets(function_ref<void(const Value *)> Func =
+                                  [](const Value *) {}) {
     return const_cast<Value *>(
-                      static_cast<const Value *>(this)->stripInBoundsOffsets());
+        static_cast<const Value *>(this)->stripInBoundsOffsets(Func));
   }
 
   /// Returns the number of bytes known to be dereferenceable for the
diff --git a/llvm/lib/IR/Operator.cpp b/llvm/lib/IR/Operator.cpp
--- a/llvm/lib/IR/Operator.cpp
+++ b/llvm/lib/IR/Operator.cpp
@@ -31,6 +31,33 @@
   return cast<GetElementPtrConstantExpr>(this)->getResultElementType();
 }
 
+Align GEPOperator::getMaxPreservedAlignment(const DataLayout &DL) const {
+  /// compute the worse possible offset for every level of the GEP et accumulate
+  /// the minimum alignment into Result.
+
+  Align Result = Align(llvm::Value::MaximumAlignment);
+  for (gep_type_iterator GTI = gep_type_begin(this), GTE = gep_type_end(this);
+       GTI != GTE; ++GTI) {
+    int64_t Offset = 1;
+    ConstantInt *OpC = dyn_cast<ConstantInt>(GTI.getOperand());
+
+    if (StructType *STy = GTI.getStructTypeOrNull()) {
+      const StructLayout *SL = DL.getStructLayout(STy);
+      Offset = SL->getElementOffset(OpC->getZExtValue());
+    } else {
+      assert(GTI.isSequential() && "should be sequencial");
+      /// If the index isn't know we take 1 because it is the index that will
+      /// give the worse alignment of the offset.
+      int64_t ElemCount = 1;
+      if (OpC)
+        ElemCount = OpC->getZExtValue();
+      Offset = DL.getTypeAllocSize(GTI.getIndexedType()) * ElemCount;
+    }
+    Result = Align(MinAlign(Offset, Result.value()));
+  }
+  return Result;
+}
+
 bool GEPOperator::accumulateConstantOffset(const DataLayout &DL,
                                            APInt &Offset) const {
   assert(Offset.getBitWidth() ==
diff --git a/llvm/lib/IR/Value.cpp b/llvm/lib/IR/Value.cpp
--- a/llvm/lib/IR/Value.cpp
+++ b/llvm/lib/IR/Value.cpp
@@ -516,7 +516,9 @@
 };
 
 template <PointerStripKind StripKind>
-static const Value *stripPointerCastsAndOffsets(const Value *V) {
+static const Value *stripPointerCastsAndOffsets(
+    const Value *V,
+    function_ref<void(const Value *)> Func = [](const Value *) {}) {
   if (!V->getType()->isPointerTy())
     return V;
 
@@ -526,6 +528,7 @@
 
   Visited.insert(V);
   do {
+    Func(V);
     if (auto *GEP = dyn_cast<GEPOperator>(V)) {
       switch (StripKind) {
       case PSK_ZeroIndices:
@@ -654,8 +657,9 @@
   return V;
 }
 
-const Value *Value::stripInBoundsOffsets() const {
-  return stripPointerCastsAndOffsets<PSK_InBounds>(this);
+const Value *
+Value::stripInBoundsOffsets(function_ref<void(const Value *)> Func) const {
+  return stripPointerCastsAndOffsets<PSK_InBounds>(this, Func);
 }
 
 uint64_t Value::getPointerDereferenceableBytes(const DataLayout &DL,
diff --git a/llvm/lib/Transforms/Utils/AssumeBundleBuilder.cpp b/llvm/lib/Transforms/Utils/AssumeBundleBuilder.cpp
--- a/llvm/lib/Transforms/Utils/AssumeBundleBuilder.cpp
+++ b/llvm/lib/Transforms/Utils/AssumeBundleBuilder.cpp
@@ -48,6 +48,39 @@
   }
 }
 
+/// This function will try to transform the given knowledge into a more
+/// canonical one. the canonical knowledge maybe the given one.
+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 = RK.WasOn->stripInBoundsOffsets([&](const Value *Strip) {
+      if (auto *GEP = dyn_cast<GEPOperator>(Strip))
+        RK.ArgValue =
+            MinAlign(RK.ArgValue,
+                     GEP->getMaxPreservedAlignment(M->getDataLayout()).value());
+    });
+    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;
+  }
+  }
+  return RK;
+}
+
 /// This class contain all knowledge that have been gather while building an
 /// llvm.assume and the function to manipulate it.
 struct AssumeBuilderState {
@@ -64,7 +97,7 @@
       : 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;
@@ -92,8 +125,41 @@
     return HasBeenPreserved;
   }
 
+  bool isKnowledgeWorthPreserving(RetainedKnowledge RK) {
+    if (!RK)
+      return false;
+    if (!RK.WasOn)
+      return true;
+    if (RK.WasOn->getType()->isPointerTy()) {
+      Value *UnderlyingPtr = GetUnderlyingObject(RK.WasOn, M->getDataLayout());
+      if (isa<AllocaInst>(UnderlyingPtr) || isa<GlobalValue>(UnderlyingPtr))
+        return false;
+    }
+    if (auto *Arg = dyn_cast<Argument>(RK.WasOn)) {
+      if (Arg->hasAttribute(RK.AttrKind) &&
+          (!Attribute::doesAttrKindHaveArgument(RK.AttrKind) ||
+           Arg->getAttribute(RK.AttrKind).getValueAsInt() >= RK.ArgValue))
+        return false;
+      return true;
+    }
+    if (auto *Inst = dyn_cast<Instruction>(RK.WasOn))
+      if (wouldInstructionBeTriviallyDead(Inst)) {
+        if (RK.WasOn->use_empty())
+          return false;
+        Use *SingleUse = RK.WasOn->getSingleUndroppableUse();
+        if (SingleUse && SingleUse->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);
diff --git a/llvm/test/Analysis/BasicAA/featuretest.ll b/llvm/test/Analysis/BasicAA/featuretest.ll
--- a/llvm/test/Analysis/BasicAA/featuretest.ll
+++ b/llvm/test/Analysis/BasicAA/featuretest.ll
@@ -15,31 +15,17 @@
 ; operations on another array.  Important for scientific codes.
 ;
 define i32 @different_array_test(i64 %A, i64 %B) {
-; NO_ASSUME-LABEL: @different_array_test(
-; NO_ASSUME-NEXT:    [[ARRAY11:%.*]] = alloca [100 x i32], align 4
-; NO_ASSUME-NEXT:    [[ARRAY22:%.*]] = alloca [200 x i32], align 4
-; NO_ASSUME-NEXT:    [[ARRAY22_SUB:%.*]] = getelementptr inbounds [200 x i32], [200 x i32]* [[ARRAY22]], i64 0, i64 0
-; NO_ASSUME-NEXT:    [[ARRAY11_SUB:%.*]] = getelementptr inbounds [100 x i32], [100 x i32]* [[ARRAY11]], i64 0, i64 0
-; NO_ASSUME-NEXT:    call void @llvm.assume(i1 true) [ "align"(i32* [[ARRAY11_SUB]], i32 4) ]
-; NO_ASSUME-NEXT:    call void @external(i32* nonnull [[ARRAY11_SUB]])
-; NO_ASSUME-NEXT:    call void @external(i32* nonnull [[ARRAY22_SUB]])
-; NO_ASSUME-NEXT:    [[POINTER2:%.*]] = getelementptr [200 x i32], [200 x i32]* [[ARRAY22]], i64 0, i64 [[B:%.*]]
-; NO_ASSUME-NEXT:    store i32 7, i32* [[POINTER2]], align 4
-; NO_ASSUME-NEXT:    ret i32 0
-;
-; USE_ASSUME-LABEL: @different_array_test(
-; USE_ASSUME-NEXT:    [[ARRAY11:%.*]] = alloca [100 x i32], align 4
-; USE_ASSUME-NEXT:    [[ARRAY22:%.*]] = alloca [200 x i32], align 4
-; 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:    ret i32 0
+; CHECK-LABEL: @different_array_test(
+; CHECK-NEXT:    [[ARRAY11:%.*]] = alloca [100 x i32], align 4
+; CHECK-NEXT:    [[ARRAY22:%.*]] = alloca [200 x i32], align 4
+; CHECK-NEXT:    [[ARRAY22_SUB:%.*]] = getelementptr inbounds [200 x i32], [200 x i32]* [[ARRAY22]], i64 0, i64 0
+; CHECK-NEXT:    [[ARRAY11_SUB:%.*]] = getelementptr inbounds [100 x i32], [100 x i32]* [[ARRAY11]], i64 0, i64 0
+; CHECK-NEXT:    call void @llvm.assume(i1 true) [ "align"(i32* [[ARRAY11_SUB]], i32 4) ]
+; CHECK-NEXT:    call void @external(i32* nonnull [[ARRAY11_SUB]])
+; CHECK-NEXT:    call void @external(i32* nonnull [[ARRAY22_SUB]])
+; CHECK-NEXT:    [[POINTER2:%.*]] = getelementptr [200 x i32], [200 x i32]* [[ARRAY22]], i64 0, i64 [[B:%.*]]
+; CHECK-NEXT:    store i32 7, i32* [[POINTER2]], align 4
+; CHECK-NEXT:    ret i32 0
 ;
   %Array1 = alloca i32, i32 100
   %Array2 = alloca i32, i32 200
@@ -63,23 +49,13 @@
 ; interfere with each other.  Again, important for scientific codes.
 ;
 define i32 @constant_array_index_test() {
-; NO_ASSUME-LABEL: @constant_array_index_test(
-; NO_ASSUME-NEXT:    [[ARRAY1:%.*]] = alloca [100 x i32], align 4
-; NO_ASSUME-NEXT:    [[ARRAY1_SUB:%.*]] = getelementptr inbounds [100 x i32], [100 x i32]* [[ARRAY1]], i64 0, i64 0
-; NO_ASSUME-NEXT:    call void @external(i32* nonnull [[ARRAY1_SUB]])
-; NO_ASSUME-NEXT:    [[P2:%.*]] = getelementptr inbounds [100 x i32], [100 x i32]* [[ARRAY1]], i64 0, i64 6
-; NO_ASSUME-NEXT:    store i32 1, i32* [[P2]], align 4
-; NO_ASSUME-NEXT:    ret i32 0
-;
-; 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:    ret i32 0
+; CHECK-LABEL: @constant_array_index_test(
+; CHECK-NEXT:    [[ARRAY1:%.*]] = alloca [100 x i32], align 4
+; CHECK-NEXT:    [[ARRAY1_SUB:%.*]] = getelementptr inbounds [100 x i32], [100 x i32]* [[ARRAY1]], i64 0, i64 0
+; CHECK-NEXT:    call void @external(i32* nonnull [[ARRAY1_SUB]])
+; CHECK-NEXT:    [[P2:%.*]] = getelementptr inbounds [100 x i32], [100 x i32]* [[ARRAY1]], i64 0, i64 6
+; CHECK-NEXT:    store i32 1, i32* [[P2]], align 4
+; CHECK-NEXT:    ret i32 0
 ;
   %Array = alloca i32, i32 100
   call void @external(i32* %Array)
@@ -128,7 +104,7 @@
 ; 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
@@ -167,18 +143,11 @@
 
 ; 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)
diff --git a/llvm/test/Transforms/Attributor/ArgumentPromotion/byval-2-preservation.ll b/llvm/test/Transforms/Attributor/ArgumentPromotion/byval-2-preservation.ll
--- a/llvm/test/Transforms/Attributor/ArgumentPromotion/byval-2-preservation.ll
+++ b/llvm/test/Transforms/Attributor/ArgumentPromotion/byval-2-preservation.ll
@@ -1,5 +1,5 @@
 ; NOTE: Assertions have been autogenerated by utils/update_test_checks.py
-; RUN: opt -S -passes=attributor --enable-knowledge-retention -aa-pipeline='basic-aa' -attributor-max-iterations-verify -attributor-max-iterations=3 < %s | FileCheck %s
+; RUN: opt -S -passes=attributor --enable-knowledge-retention -aa-pipeline='basic-aa' -attributor-max-iterations-verify -attributor-max-iterations=4 < %s | FileCheck %s
 
 %struct.ss = type { i32, i64 }
 
@@ -22,7 +22,7 @@
 ; 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) [ "align"(i32* [[X:%.*]], i64 4), "align"(%struct.ss* [[S]], i64 8), "dereferenceable"(i32* [[X]], i64 4), "dereferenceable"(%struct.ss* [[S]], i64 12), "nonnull"(%struct.ss* [[S]]), "nonnull"(i32* [[X]]) ]
+; CHECK-NEXT:    call void @llvm.assume(i1 true) [ "nonnull"(i32* [[X:%.*]]), "dereferenceable"(i32* [[X]], i64 4) ]
 ; CHECK-NEXT:    ret i32 0
 ;
 entry:
diff --git a/llvm/test/Transforms/Inline/arg-attr-propagation.ll b/llvm/test/Transforms/Inline/arg-attr-propagation.ll
--- a/llvm/test/Transforms/Inline/arg-attr-propagation.ll
+++ b/llvm/test/Transforms/Inline/arg-attr-propagation.ll
@@ -57,16 +57,10 @@
 ; Make sure that we don't propagate a smaller dereferenceable amount.
 
 define i32 @caller3(i32* dereferenceable(33) %t1) {
-; NO_ASSUME-LABEL: define {{[^@]+}}@caller3
-; NO_ASSUME-SAME: (i32* dereferenceable(33) [[T1:%.*]])
-; NO_ASSUME-NEXT:    [[T2_I:%.*]] = load i32, i32* [[T1]]
-; NO_ASSUME-NEXT:    ret i32 [[T2_I]]
-;
-; USE_ASSUME-LABEL: define {{[^@]+}}@caller3
-; USE_ASSUME-SAME: (i32* dereferenceable(33) [[T1:%.*]])
-; USE_ASSUME-NEXT:    call void @llvm.assume(i1 true) [ "dereferenceable"(i32* [[T1]], i64 32) ]
-; USE_ASSUME-NEXT:    [[T2_I:%.*]] = load i32, i32* [[T1]]
-; USE_ASSUME-NEXT:    ret i32 [[T2_I]]
+; CHECK-LABEL: define {{[^@]+}}@caller3
+; CHECK-SAME: (i32* dereferenceable(33) [[T1:%.*]])
+; CHECK-NEXT:    [[T2_I:%.*]] = load i32, i32* [[T1]]
+; CHECK-NEXT:    ret i32 [[T2_I]]
 ;
   %t2 = tail call i32 @callee(i32* dereferenceable(32) %t1)
   ret i32 %t2
diff --git a/llvm/test/Transforms/InstSimplify/load.ll b/llvm/test/Transforms/InstSimplify/load.ll
--- a/llvm/test/Transforms/InstSimplify/load.ll
+++ b/llvm/test/Transforms/InstSimplify/load.ll
@@ -6,24 +6,16 @@
 @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:    ret i32 0
+; CHECK-LABEL: @crash_on_zeroinit(
+; CHECK-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:    ret i32 undef
+; CHECK-LABEL: @crash_on_undef(
+; CHECK-NEXT:    ret i32 undef
 ;
   %load = load i32, i32* bitcast ({}* @undef to i32*)
   ret i32 %load
@@ -32,12 +24,8 @@
 @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:    ret <8 x i32> <i32 0, i32 42, i32 43, i32 44, i32 45, i32 46, i32 47, i32 48>
+; CHECK-LABEL: @partial_load(
+; CHECK-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
@@ -47,12 +35,8 @@
 
 ; 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:    ret <3 x float> undef
+; CHECK-LABEL: @load_vec3(
+; CHECK-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
diff --git a/llvm/test/Transforms/LICM/pr23608.ll b/llvm/test/Transforms/LICM/pr23608.ll
--- a/llvm/test/Transforms/LICM/pr23608.ll
+++ b/llvm/test/Transforms/LICM/pr23608.ll
@@ -51,7 +51,7 @@
 ; 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:%.*]]
diff --git a/llvm/test/Transforms/NewGVN/2007-07-26-PhiErasure.ll b/llvm/test/Transforms/NewGVN/2007-07-26-PhiErasure.ll
--- a/llvm/test/Transforms/NewGVN/2007-07-26-PhiErasure.ll
+++ b/llvm/test/Transforms/NewGVN/2007-07-26-PhiErasure.ll
@@ -10,32 +10,18 @@
 @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]
diff --git a/llvm/test/Transforms/Util/assume-builder.ll b/llvm/test/Transforms/Util/assume-builder.ll
--- a/llvm/test/Transforms/Util/assume-builder.ll
+++ b/llvm/test/Transforms/Util/assume-builder.ll
@@ -19,11 +19,11 @@
 ; 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]])
@@ -33,7 +33,7 @@
 ; 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
 ;
@@ -41,11 +41,11 @@
 ; 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]])
@@ -55,7 +55,7 @@
 ; 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
 ;
@@ -106,9 +106,9 @@
 ; 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]])
@@ -118,7 +118,7 @@
 ; 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
 ;
@@ -144,44 +144,32 @@
 ; 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 4) ]
 ; 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]]
@@ -192,44 +180,32 @@
 ; 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 4) ]
 ; 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]]
@@ -240,19 +216,13 @@
 ; 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]]
@@ -260,19 +230,18 @@
 ; 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 4) ]
 ; 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]]
@@ -283,19 +252,13 @@
 ; 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]]
@@ -303,19 +266,18 @@
 ; 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 4) ]
 ; 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]]
@@ -347,9 +309,9 @@
 ; 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 4), "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]]
@@ -393,45 +355,33 @@
 ; 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 4) ]
+; 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]]
@@ -442,45 +392,33 @@
 ; 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 4) ]
+; 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]]
@@ -491,18 +429,13 @@
 ; 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
@@ -511,19 +444,18 @@
 ; 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 4) ]
+; 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]]
@@ -534,18 +466,13 @@
 ; 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
@@ -554,19 +481,18 @@
 ; 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 4) ]
+; 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]]
@@ -584,7 +510,6 @@
 ; 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
@@ -596,13 +521,13 @@
 ; 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 4), "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]]
@@ -629,7 +554,7 @@
   %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
@@ -649,9 +574,8 @@
 ; 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:
@@ -675,9 +599,8 @@
 ; 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:
@@ -747,7 +670,7 @@
 ; 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]]
@@ -790,13 +713,11 @@
 ; 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:
@@ -812,13 +733,11 @@
 ; 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:
@@ -979,3 +898,389 @@
 }
 
 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"(i16* [[TMP6]], i64 2), "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:    call void @llvm.assume(i1 true) [ "dereferenceable"(i64* [[TMP9]], i64 8) ]
+; 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"(i16* [[TMP6]], i64 2), "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:    call void @llvm.assume(i1 true) [ "dereferenceable"(i64* [[TMP9]], i64 8) ]
+; 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"(i16* [[TMP6]], i64 2), "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:    call void @llvm.assume(i1 true) [ "dereferenceable"(i64* [[TMP9]], i64 8) ]
+; 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"(i16* [[TMP6]], i64 2), "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:    call void @llvm.assume(i1 true) [ "dereferenceable"(i64* [[TMP9]], i64 8) ]
+; 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 [[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:    call void @llvm.assume(i1 true) [ "dereferenceable"(i16* [[TMP6]], i64 2), "dereferenceable"(i64* [[TMP9]], i64 8) ]
+; 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 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) [ "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:    call void @llvm.assume(i1 true) [ "dereferenceable"(i32* [[TMP13]], i64 4) ]
+; 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) [ "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) [ "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:    call void @llvm.assume(i1 true) [ "dereferenceable"(i32* [[TMP13]], i64 4) ]
+; 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) [ "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) [ "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:    call void @llvm.assume(i1 true) [ "dereferenceable"(i32* [[TMP13]], i64 4) ]
+; 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) [ "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) [ "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:    call void @llvm.assume(i1 true) [ "dereferenceable"(i32* [[TMP13]], i64 4) ]
+; 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) [ "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:    [[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:    call void @llvm.assume(i1 true) [ "nonnull"(i32* [[TMP0]]), "align"(i32* [[TMP0]], i64 4), "dereferenceable"(i32* [[TMP13]], i64 4) ]
+; 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) [ "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
+}
+
+%struct.A = type { i8*, i64*, [4 x [4 x %struct.D]], i64 }
+%struct.D = type { i64, i64 }
+
+define i32 @test7(%struct.A* nonnull %0, i32 %1) {
+; BASIC-LABEL: define {{[^@]+}}@test7
+; BASIC-SAME: (%struct.A* nonnull [[TMP0:%.*]], i32 [[TMP1:%.*]])
+; BASIC-NEXT:    [[TMP3:%.*]] = sext i32 [[TMP1]] to i64
+; BASIC-NEXT:    [[TMP4:%.*]] = getelementptr inbounds [[STRUCT_A:%.*]], %struct.A* [[TMP0]], i64 0, i32 3
+; BASIC-NEXT:    call void @llvm.assume(i1 true) [ "dereferenceable"(%struct.A* [[TMP0]], i64 280), "align"(%struct.A* [[TMP0]], i64 16) ]
+; BASIC-NEXT:    [[TMP5:%.*]] = load i64, i64* [[TMP4]], align 8
+; BASIC-NEXT:    [[TMP6:%.*]] = getelementptr inbounds [[STRUCT_A]], %struct.A* [[TMP0]], i64 0, i32 2, i64 [[TMP3]], i64 [[TMP5]], i32 0
+; BASIC-NEXT:    [[TMP7:%.*]] = load i64, i64* [[TMP6]], align 32
+; BASIC-NEXT:    [[TMP8:%.*]] = getelementptr inbounds [[STRUCT_A]], %struct.A* [[TMP0]], i64 0, i32 1
+; BASIC-NEXT:    [[TMP9:%.*]] = load i64*, i64** [[TMP8]], align 8
+; BASIC-NEXT:    [[TMP10:%.*]] = getelementptr inbounds i64, i64* [[TMP9]], i64 [[TMP5]]
+; BASIC-NEXT:    call void @llvm.assume(i1 true) [ "dereferenceable"(i64* [[TMP10]], i64 8), "nonnull"(i64* [[TMP9]]), "align"(i64* [[TMP9]], i64 8) ]
+; BASIC-NEXT:    store i64 [[TMP7]], i64* [[TMP10]], align 8
+; BASIC-NEXT:    store i64 [[TMP7]], i64* [[TMP10]], align 8
+; BASIC-NEXT:    [[TMP11:%.*]] = bitcast %struct.A* [[TMP0]] to i32**
+; BASIC-NEXT:    [[TMP12:%.*]] = load i32*, i32** [[TMP11]], align 8
+; BASIC-NEXT:    [[TMP13:%.*]] = load i32, i32* [[TMP12]], align 4
+; BASIC-NEXT:    ret i32 [[TMP13]]
+;
+; ALL-LABEL: define {{[^@]+}}@test7
+; ALL-SAME: (%struct.A* nonnull [[TMP0:%.*]], i32 [[TMP1:%.*]])
+; ALL-NEXT:    [[TMP3:%.*]] = sext i32 [[TMP1]] to i64
+; ALL-NEXT:    [[TMP4:%.*]] = getelementptr inbounds [[STRUCT_A:%.*]], %struct.A* [[TMP0]], i64 0, i32 3
+; ALL-NEXT:    call void @llvm.assume(i1 true) [ "dereferenceable"(%struct.A* [[TMP0]], i64 280), "align"(%struct.A* [[TMP0]], i64 16) ]
+; ALL-NEXT:    [[TMP5:%.*]] = load i64, i64* [[TMP4]], align 8
+; ALL-NEXT:    [[TMP6:%.*]] = getelementptr inbounds [[STRUCT_A]], %struct.A* [[TMP0]], i64 0, i32 2, i64 [[TMP3]], i64 [[TMP5]], i32 0
+; ALL-NEXT:    [[TMP7:%.*]] = load i64, i64* [[TMP6]], align 32
+; ALL-NEXT:    [[TMP8:%.*]] = getelementptr inbounds [[STRUCT_A]], %struct.A* [[TMP0]], i64 0, i32 1
+; ALL-NEXT:    [[TMP9:%.*]] = load i64*, i64** [[TMP8]], align 8
+; ALL-NEXT:    [[TMP10:%.*]] = getelementptr inbounds i64, i64* [[TMP9]], i64 [[TMP5]]
+; ALL-NEXT:    call void @llvm.assume(i1 true) [ "dereferenceable"(i64* [[TMP10]], i64 8), "nonnull"(i64* [[TMP9]]), "align"(i64* [[TMP9]], i64 8) ]
+; ALL-NEXT:    store i64 [[TMP7]], i64* [[TMP10]], align 8
+; ALL-NEXT:    store i64 [[TMP7]], i64* [[TMP10]], align 8
+; ALL-NEXT:    [[TMP11:%.*]] = bitcast %struct.A* [[TMP0]] to i32**
+; ALL-NEXT:    [[TMP12:%.*]] = load i32*, i32** [[TMP11]], align 8
+; ALL-NEXT:    [[TMP13:%.*]] = load i32, i32* [[TMP12]], align 4
+; ALL-NEXT:    ret i32 [[TMP13]]
+;
+; WITH-AC-LABEL: define {{[^@]+}}@test7
+; WITH-AC-SAME: (%struct.A* nonnull [[TMP0:%.*]], i32 [[TMP1:%.*]])
+; WITH-AC-NEXT:    [[TMP3:%.*]] = sext i32 [[TMP1]] to i64
+; WITH-AC-NEXT:    [[TMP4:%.*]] = getelementptr inbounds [[STRUCT_A:%.*]], %struct.A* [[TMP0]], i64 0, i32 3
+; WITH-AC-NEXT:    call void @llvm.assume(i1 true) [ "dereferenceable"(%struct.A* [[TMP0]], i64 280), "align"(%struct.A* [[TMP0]], i64 16) ]
+; WITH-AC-NEXT:    [[TMP5:%.*]] = load i64, i64* [[TMP4]], align 8
+; WITH-AC-NEXT:    [[TMP6:%.*]] = getelementptr inbounds [[STRUCT_A]], %struct.A* [[TMP0]], i64 0, i32 2, i64 [[TMP3]], i64 [[TMP5]], i32 0
+; WITH-AC-NEXT:    [[TMP7:%.*]] = load i64, i64* [[TMP6]], align 32
+; WITH-AC-NEXT:    [[TMP8:%.*]] = getelementptr inbounds [[STRUCT_A]], %struct.A* [[TMP0]], i64 0, i32 1
+; WITH-AC-NEXT:    [[TMP9:%.*]] = load i64*, i64** [[TMP8]], align 8
+; WITH-AC-NEXT:    [[TMP10:%.*]] = getelementptr inbounds i64, i64* [[TMP9]], i64 [[TMP5]]
+; WITH-AC-NEXT:    call void @llvm.assume(i1 true) [ "dereferenceable"(i64* [[TMP10]], i64 8), "nonnull"(i64* [[TMP9]]), "align"(i64* [[TMP9]], i64 8) ]
+; WITH-AC-NEXT:    store i64 [[TMP7]], i64* [[TMP10]], align 8
+; WITH-AC-NEXT:    store i64 [[TMP7]], i64* [[TMP10]], align 8
+; WITH-AC-NEXT:    [[TMP11:%.*]] = bitcast %struct.A* [[TMP0]] to i32**
+; WITH-AC-NEXT:    [[TMP12:%.*]] = load i32*, i32** [[TMP11]], align 8
+; WITH-AC-NEXT:    [[TMP13:%.*]] = load i32, i32* [[TMP12]], align 4
+; WITH-AC-NEXT:    ret i32 [[TMP13]]
+;
+; CROSS-BLOCK-LABEL: define {{[^@]+}}@test7
+; CROSS-BLOCK-SAME: (%struct.A* nonnull [[TMP0:%.*]], i32 [[TMP1:%.*]])
+; CROSS-BLOCK-NEXT:    [[TMP3:%.*]] = sext i32 [[TMP1]] to i64
+; CROSS-BLOCK-NEXT:    [[TMP4:%.*]] = getelementptr inbounds [[STRUCT_A:%.*]], %struct.A* [[TMP0]], i64 0, i32 3
+; CROSS-BLOCK-NEXT:    call void @llvm.assume(i1 true) [ "dereferenceable"(%struct.A* [[TMP0]], i64 280), "align"(%struct.A* [[TMP0]], i64 16) ]
+; CROSS-BLOCK-NEXT:    [[TMP5:%.*]] = load i64, i64* [[TMP4]], align 8
+; CROSS-BLOCK-NEXT:    [[TMP6:%.*]] = getelementptr inbounds [[STRUCT_A]], %struct.A* [[TMP0]], i64 0, i32 2, i64 [[TMP3]], i64 [[TMP5]], i32 0
+; CROSS-BLOCK-NEXT:    [[TMP7:%.*]] = load i64, i64* [[TMP6]], align 32
+; CROSS-BLOCK-NEXT:    [[TMP8:%.*]] = getelementptr inbounds [[STRUCT_A]], %struct.A* [[TMP0]], i64 0, i32 1
+; CROSS-BLOCK-NEXT:    [[TMP9:%.*]] = load i64*, i64** [[TMP8]], align 8
+; CROSS-BLOCK-NEXT:    [[TMP10:%.*]] = getelementptr inbounds i64, i64* [[TMP9]], i64 [[TMP5]]
+; CROSS-BLOCK-NEXT:    call void @llvm.assume(i1 true) [ "dereferenceable"(i64* [[TMP10]], i64 8), "nonnull"(i64* [[TMP9]]), "align"(i64* [[TMP9]], i64 8) ]
+; CROSS-BLOCK-NEXT:    store i64 [[TMP7]], i64* [[TMP10]], align 8
+; CROSS-BLOCK-NEXT:    store i64 [[TMP7]], i64* [[TMP10]], align 8
+; CROSS-BLOCK-NEXT:    [[TMP11:%.*]] = bitcast %struct.A* [[TMP0]] to i32**
+; CROSS-BLOCK-NEXT:    [[TMP12:%.*]] = load i32*, i32** [[TMP11]], align 8
+; CROSS-BLOCK-NEXT:    [[TMP13:%.*]] = load i32, i32* [[TMP12]], align 4
+; CROSS-BLOCK-NEXT:    ret i32 [[TMP13]]
+;
+; FULL-SIMPLIFY-LABEL: define {{[^@]+}}@test7
+; FULL-SIMPLIFY-SAME: (%struct.A* nonnull align 16 dereferenceable(280) [[TMP0:%.*]], i32 [[TMP1:%.*]])
+; FULL-SIMPLIFY-NEXT:    [[TMP3:%.*]] = sext i32 [[TMP1]] to i64
+; FULL-SIMPLIFY-NEXT:    [[TMP4:%.*]] = getelementptr inbounds [[STRUCT_A:%.*]], %struct.A* [[TMP0]], i64 0, i32 3
+; FULL-SIMPLIFY-NEXT:    [[TMP5:%.*]] = load i64, i64* [[TMP4]], align 8
+; FULL-SIMPLIFY-NEXT:    [[TMP6:%.*]] = getelementptr inbounds [[STRUCT_A]], %struct.A* [[TMP0]], i64 0, i32 2, i64 [[TMP3]], i64 [[TMP5]], i32 0
+; FULL-SIMPLIFY-NEXT:    [[TMP7:%.*]] = load i64, i64* [[TMP6]], align 32
+; FULL-SIMPLIFY-NEXT:    [[TMP8:%.*]] = getelementptr inbounds [[STRUCT_A]], %struct.A* [[TMP0]], i64 0, i32 1
+; FULL-SIMPLIFY-NEXT:    [[TMP9:%.*]] = load i64*, i64** [[TMP8]], align 8
+; FULL-SIMPLIFY-NEXT:    [[TMP10:%.*]] = getelementptr inbounds i64, i64* [[TMP9]], i64 [[TMP5]]
+; FULL-SIMPLIFY-NEXT:    call void @llvm.assume(i1 true) [ "dereferenceable"(i64* [[TMP10]], i64 8), "nonnull"(i64* [[TMP9]]), "align"(i64* [[TMP9]], i64 8) ]
+; FULL-SIMPLIFY-NEXT:    store i64 [[TMP7]], i64* [[TMP10]], align 8
+; FULL-SIMPLIFY-NEXT:    store i64 [[TMP7]], i64* [[TMP10]], align 8
+; FULL-SIMPLIFY-NEXT:    [[TMP11:%.*]] = bitcast %struct.A* [[TMP0]] to i32**
+; FULL-SIMPLIFY-NEXT:    [[TMP12:%.*]] = load i32*, i32** [[TMP11]], align 8
+; FULL-SIMPLIFY-NEXT:    [[TMP13:%.*]] = load i32, i32* [[TMP12]], align 4
+; FULL-SIMPLIFY-NEXT:    ret i32 [[TMP13]]
+;
+  %3 = sext i32 %1 to i64
+  %4 = getelementptr inbounds %struct.A, %struct.A* %0, i64 0, i32 3
+  %5 = load i64, i64* %4, align 8
+  %6 = getelementptr inbounds %struct.A, %struct.A* %0, i64 0, i32 2, i64 %3, i64 %5, i32 0
+  %7 = load i64, i64* %6, align 32
+  %8 = getelementptr inbounds %struct.A, %struct.A* %0, i64 0, i32 1
+  %9 = load i64*, i64** %8, align 8
+  %10 = getelementptr inbounds i64, i64* %9, i64 %5
+  store i64 %7, i64* %10, align 8
+  store i64 %7, i64* %10, align 8
+  %11 = bitcast %struct.A* %0 to i32**
+  %12 = load i32*, i32** %11, align 8
+  %13 = load i32, i32* %12, align 4
+  ret i32 %13
+}
diff --git a/llvm/test/Transforms/Util/assume-simplify.ll b/llvm/test/Transforms/Util/assume-simplify.ll
--- a/llvm/test/Transforms/Util/assume-simplify.ll
+++ b/llvm/test/Transforms/Util/assume-simplify.ll
@@ -76,7 +76,7 @@
 ; 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) [ "dereferenceable"(i32* [[TMP1]], i64 12), "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
@@ -303,7 +303,7 @@
 ; 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"(), "nonnull"(i32* [[P]]) ]
 ; CHECK-NEXT:    ret i32 0
 ;
   %p1 = bitcast i32* %p to i8*