Index: lib/Transforms/Scalar/SROA.cpp
===================================================================
--- lib/Transforms/Scalar/SROA.cpp
+++ lib/Transforms/Scalar/SROA.cpp
@@ -3627,9 +3627,12 @@
             PLoad->getType()->getPointerTo(SI->getPointerAddressSpace());
 
         StoreInst *PStore = IRB.CreateAlignedStore(
-            PLoad, getAdjustedPtr(IRB, DL, StoreBasePtr,
-                                  APInt(DL.getPointerSizeInBits(), PartOffset),
-                                  PartPtrTy, StoreBasePtr->getName() + "."),
+            PLoad,
+            getAdjustedPtr(
+                IRB, DL, StoreBasePtr,
+                APInt(DL.getPointerSizeInBits(SI->getPointerAddressSpace()),
+                      PartOffset),
+                PartPtrTy, StoreBasePtr->getName() + "."),
             getAdjustedAlignment(SI, PartOffset, DL), /*IsVolatile*/ false);
         PStore->copyMetadata(*LI, LLVMContext::MD_mem_parallel_loop_access);
         DEBUG(dbgs() << "      +" << PartOffset << ":" << *PStore << "\n");
@@ -3698,7 +3701,8 @@
     int Idx = 0, Size = Offsets.Splits.size();
     for (;;) {
       auto *PartTy = Type::getIntNTy(Ty->getContext(), PartSize * 8);
-      auto *PartPtrTy = PartTy->getPointerTo(SI->getPointerAddressSpace());
+      auto *StorePartPtrTy = PartTy->getPointerTo(SI->getPointerAddressSpace());
+      auto *LoadPartPtrTy = PartTy->getPointerTo(LI->getPointerAddressSpace());
 
       // Either lookup a split load or create one.
       LoadInst *PLoad;
@@ -3707,9 +3711,11 @@
       } else {
         IRB.SetInsertPoint(LI);
         PLoad = IRB.CreateAlignedLoad(
-            getAdjustedPtr(IRB, DL, LoadBasePtr,
-                           APInt(DL.getPointerSizeInBits(), PartOffset),
-                           PartPtrTy, LoadBasePtr->getName() + "."),
+            getAdjustedPtr(
+                IRB, DL, LoadBasePtr,
+                APInt(DL.getPointerSizeInBits(LI->getPointerAddressSpace()),
+                      PartOffset),
+                LoadPartPtrTy, LoadBasePtr->getName() + "."),
             getAdjustedAlignment(LI, PartOffset, DL), /*IsVolatile*/ false,
             LI->getName());
       }
@@ -3717,9 +3723,12 @@
       // And store this partition.
       IRB.SetInsertPoint(SI);
       StoreInst *PStore = IRB.CreateAlignedStore(
-          PLoad, getAdjustedPtr(IRB, DL, StoreBasePtr,
-                                APInt(DL.getPointerSizeInBits(), PartOffset),
-                                PartPtrTy, StoreBasePtr->getName() + "."),
+          PLoad,
+          getAdjustedPtr(
+              IRB, DL, StoreBasePtr,
+              APInt(DL.getPointerSizeInBits(SI->getPointerAddressSpace()),
+                    PartOffset),
+              StorePartPtrTy, StoreBasePtr->getName() + "."),
           getAdjustedAlignment(SI, PartOffset, DL), /*IsVolatile*/ false);
 
       // Now build a new slice for the alloca.
Index: test/Transforms/SROA/address-spaces.ll
===================================================================
--- test/Transforms/SROA/address-spaces.ll
+++ test/Transforms/SROA/address-spaces.ll
@@ -83,3 +83,31 @@
   store i32 addrspace(3)* @l, i32 addrspace(3)** %3, align 8
   ret void
 }
+
+; Test load from and store to non-zero address space.
+define void @test_load_store_diff_addr_space([2 x float] addrspace(1)* %complex1, [2 x float] addrspace(1)* %complex2) {
+; CHECK-LABEL: @test_load_store_diff_addr_space
+; CHECK-NOT: alloca
+; CHECK: load i32, i32 addrspace(1)*
+; CHECK: load i32, i32 addrspace(1)*
+; CHECK: store i32 %{{.*}}, i32 addrspace(1)*
+; CHECK: store i32 %{{.*}}, i32 addrspace(1)*
+  %a = alloca i64
+  %a.cast = bitcast i64* %a to [2 x float]*
+  %a.gep1 = getelementptr [2 x float], [2 x float]* %a.cast, i32 0, i32 0
+  %a.gep2 = getelementptr [2 x float], [2 x float]* %a.cast, i32 0, i32 1
+  %complex1.gep = getelementptr [2 x float], [2 x float] addrspace(1)* %complex1, i32 0, i32 0
+  %p1 = bitcast float addrspace(1)* %complex1.gep to i64 addrspace(1)*
+  %v1 = load i64, i64 addrspace(1)* %p1
+  store i64 %v1, i64* %a
+  %f1 = load float, float* %a.gep1
+  %f2 = load float, float* %a.gep2
+  %sum = fadd float %f1, %f2
+  store float %sum, float* %a.gep1
+  store float %sum, float* %a.gep2
+  %v2 = load i64, i64* %a
+  %complex2.gep = getelementptr [2 x float], [2 x float] addrspace(1)* %complex2, i32 0, i32 0
+  %p2 = bitcast float addrspace(1)* %complex2.gep to i64 addrspace(1)*
+  store i64 %v2, i64 addrspace(1)* %p2
+  ret void
+}