Index: llvm/lib/Transforms/AggressiveInstCombine/AggressiveInstCombine.cpp
===================================================================
--- llvm/lib/Transforms/AggressiveInstCombine/AggressiveInstCombine.cpp
+++ llvm/lib/Transforms/AggressiveInstCombine/AggressiveInstCombine.cpp
@@ -644,6 +644,7 @@
 /// shift amount, zero extend type and loadSize.
 struct LoadOps {
   LoadInst *Root = nullptr;
+  LoadInst *InsertPoint = nullptr;
   bool FoundRoot = false;
   uint64_t LoadSize = 0;
   Value *Shift = nullptr;
@@ -778,7 +779,9 @@
   if (LOps.FoundRoot == false) {
     LOps.FoundRoot = true;
     AATags1 = LI1->getAAMetadata();
-  }
+    LOps.InsertPoint = Start;
+  } else if (LOps.InsertPoint && Start->comesBefore(LOps.InsertPoint))
+    LOps.InsertPoint = Start;
   LOps.LoadSize = LoadSize1 + LoadSize2;
 
   // Concatenate the AATags of the Merged Loads.
@@ -817,7 +820,7 @@
 
   // New load can be generated
   Value *Load1Ptr = LI1->getPointerOperand();
-  Builder.SetInsertPoint(LI1);
+  Builder.SetInsertPoint(LOps.InsertPoint);
   Value *NewPtr = Builder.CreateBitCast(Load1Ptr, WiderType->getPointerTo(AS));
   NewLoad = Builder.CreateAlignedLoad(WiderType, NewPtr, LI1->getAlign(),
                                       LI1->isVolatile(), "");
Index: llvm/test/Transforms/AggressiveInstCombine/AArch64/or-load.ll
===================================================================
--- llvm/test/Transforms/AggressiveInstCombine/AArch64/or-load.ll
+++ llvm/test/Transforms/AggressiveInstCombine/AArch64/or-load.ll
@@ -1772,8 +1772,8 @@
 define i32 @loadCombine_4consecutive_badinsert(ptr %p) {
 ; LE-LABEL: @loadCombine_4consecutive_badinsert(
 ; LE-NEXT:    [[P1:%.*]] = getelementptr i8, ptr [[P:%.*]], i32 1
-; LE-NEXT:    store i8 0, ptr [[P1]], align 1
 ; LE-NEXT:    [[L1:%.*]] = load i32, ptr [[P]], align 1
+; LE-NEXT:    store i8 0, ptr [[P1]], align 1
 ; LE-NEXT:    ret i32 [[L1]]
 ;
 ; BE-LABEL: @loadCombine_4consecutive_badinsert(
Index: llvm/test/Transforms/AggressiveInstCombine/X86/or-load.ll
===================================================================
--- llvm/test/Transforms/AggressiveInstCombine/X86/or-load.ll
+++ llvm/test/Transforms/AggressiveInstCombine/X86/or-load.ll
@@ -1863,8 +1863,8 @@
 define i16 @loadCombine_2consecutive_badinsert(ptr %p) {
 ; LE-LABEL: @loadCombine_2consecutive_badinsert(
 ; LE-NEXT:    [[P1:%.*]] = getelementptr i8, ptr [[P:%.*]], i32 1
-; LE-NEXT:    store i8 0, ptr [[P1]], align 1
 ; LE-NEXT:    [[L1:%.*]] = load i16, ptr [[P]], align 1
+; LE-NEXT:    store i8 0, ptr [[P1]], align 1
 ; LE-NEXT:    ret i16 [[L1]]
 ;
 ; BE-LABEL: @loadCombine_2consecutive_badinsert(
@@ -1892,8 +1892,8 @@
 define i32 @loadCombine_4consecutive_badinsert(ptr %p) {
 ; LE-LABEL: @loadCombine_4consecutive_badinsert(
 ; LE-NEXT:    [[P1:%.*]] = getelementptr i8, ptr [[P:%.*]], i32 1
-; LE-NEXT:    store i8 0, ptr [[P1]], align 1
 ; LE-NEXT:    [[L1:%.*]] = load i32, ptr [[P]], align 1
+; LE-NEXT:    store i8 0, ptr [[P1]], align 1
 ; LE-NEXT:    ret i32 [[L1]]
 ;
 ; BE-LABEL: @loadCombine_4consecutive_badinsert(