Index: lib/Target/AMDGPU/AMDGPUPromoteAlloca.cpp
===================================================================
--- lib/Target/AMDGPU/AMDGPUPromoteAlloca.cpp
+++ lib/Target/AMDGPU/AMDGPUPromoteAlloca.cpp
@@ -426,6 +426,13 @@
     if (!User->getType()->isPointerTy())
       continue;
 
+    if (GetElementPtrInst *GEP = dyn_cast<GetElementPtrInst>(UseInst)) {
+      // Be conservative if an address could be computed outside the bounds of
+      // the alloca.
+      if (!GEP->isInBounds())
+        return false;
+    }
+
     WorkList.push_back(User);
     Success = collectUsesWithPtrTypes(User, WorkList);
   }
Index: test/CodeGen/AMDGPU/array-ptr-calc-i32.ll
===================================================================
--- test/CodeGen/AMDGPU/array-ptr-calc-i32.ll
+++ test/CodeGen/AMDGPU/array-ptr-calc-i32.ll
@@ -20,24 +20,24 @@
 ; FIXME: The AMDGPUPromoteAlloca pass should be able to convert this
 ; alloca to a vector.  It currently fails because it does not know how
 ; to interpret:
-; getelementptr [4 x i32], [4 x i32]* %alloca, i32 1, i32 %b
+; getelementptr inbounds [4 x i32], [4 x i32]* %alloca, i32 1, i32 %b
 
 ; SI-PROMOTE: v_add_i32_e32 [[PTRREG:v[0-9]+]], vcc, 16
 ; SI-PROMOTE: ds_write_b32 [[PTRREG]]
 define void @test_private_array_ptr_calc(i32 addrspace(1)* noalias %out, i32 addrspace(1)* noalias %inA, i32 addrspace(1)* noalias %inB) {
   %alloca = alloca [4 x i32], i32 4, align 16
   %tid = call i32 @llvm.SI.tid() readnone
-  %a_ptr = getelementptr i32, i32 addrspace(1)* %inA, i32 %tid
-  %b_ptr = getelementptr i32, i32 addrspace(1)* %inB, i32 %tid
+  %a_ptr = getelementptr inbounds i32, i32 addrspace(1)* %inA, i32 %tid
+  %b_ptr = getelementptr inbounds i32, i32 addrspace(1)* %inB, i32 %tid
   %a = load i32, i32 addrspace(1)* %a_ptr
   %b = load i32, i32 addrspace(1)* %b_ptr
   %result = add i32 %a, %b
-  %alloca_ptr = getelementptr [4 x i32], [4 x i32]* %alloca, i32 1, i32 %b
+  %alloca_ptr = getelementptr inbounds [4 x i32], [4 x i32]* %alloca, i32 1, i32 %b
   store i32 %result, i32* %alloca_ptr, align 4
   ; Dummy call
   call void @llvm.AMDGPU.barrier.local() nounwind convergent
   %reload = load i32, i32* %alloca_ptr, align 4
-  %out_ptr = getelementptr i32, i32 addrspace(1)* %out, i32 %tid
+  %out_ptr = getelementptr inbounds i32, i32 addrspace(1)* %out, i32 %tid
   store i32 %reload, i32 addrspace(1)* %out_ptr, align 4
   ret void
 }
Index: test/CodeGen/AMDGPU/indirect-private-64.ll
===================================================================
--- test/CodeGen/AMDGPU/indirect-private-64.ll
+++ test/CodeGen/AMDGPU/indirect-private-64.ll
@@ -16,7 +16,7 @@
 define void @private_access_f64_alloca(double addrspace(1)* noalias %out, double addrspace(1)* noalias %in, i32 %b) nounwind {
   %val = load double, double addrspace(1)* %in, align 8
   %array = alloca double, i32 16, align 8
-  %ptr = getelementptr double, double* %array, i32 %b
+  %ptr = getelementptr inbounds double, double* %array, i32 %b
   store double %val, double* %ptr, align 8
   call void @llvm.AMDGPU.barrier.local() convergent nounwind
   %result = load double, double* %ptr, align 8
@@ -36,7 +36,7 @@
 define void @private_access_v2f64_alloca(<2 x double> addrspace(1)* noalias %out, <2 x double> addrspace(1)* noalias %in, i32 %b) nounwind {
   %val = load <2 x double>, <2 x double> addrspace(1)* %in, align 16
   %array = alloca <2 x double>, i32 16, align 16
-  %ptr = getelementptr <2 x double>, <2 x double>* %array, i32 %b
+  %ptr = getelementptr inbounds <2 x double>, <2 x double>* %array, i32 %b
   store <2 x double> %val, <2 x double>* %ptr, align 16
   call void @llvm.AMDGPU.barrier.local() convergent nounwind
   %result = load <2 x double>, <2 x double>* %ptr, align 16
@@ -54,7 +54,7 @@
 define void @private_access_i64_alloca(i64 addrspace(1)* noalias %out, i64 addrspace(1)* noalias %in, i32 %b) nounwind {
   %val = load i64, i64 addrspace(1)* %in, align 8
   %array = alloca i64, i32 16, align 8
-  %ptr = getelementptr i64, i64* %array, i32 %b
+  %ptr = getelementptr inbounds i64, i64* %array, i32 %b
   store i64 %val, i64* %ptr, align 8
   call void @llvm.AMDGPU.barrier.local() convergent nounwind
   %result = load i64, i64* %ptr, align 8
@@ -74,7 +74,7 @@
 define void @private_access_v2i64_alloca(<2 x i64> addrspace(1)* noalias %out, <2 x i64> addrspace(1)* noalias %in, i32 %b) nounwind {
   %val = load <2 x i64>, <2 x i64> addrspace(1)* %in, align 16
   %array = alloca <2 x i64>, i32 16, align 16
-  %ptr = getelementptr <2 x i64>, <2 x i64>* %array, i32 %b
+  %ptr = getelementptr inbounds <2 x i64>, <2 x i64>* %array, i32 %b
   store <2 x i64> %val, <2 x i64>* %ptr, align 16
   call void @llvm.AMDGPU.barrier.local() convergent nounwind
   %result = load <2 x i64>, <2 x i64>* %ptr, align 16
Index: test/CodeGen/AMDGPU/private-memory-atomics.ll
===================================================================
--- test/CodeGen/AMDGPU/private-memory-atomics.ll
+++ test/CodeGen/AMDGPU/private-memory-atomics.ll
@@ -7,11 +7,11 @@
 define void @atomicrmw_private(i32 addrspace(1)* %out, i32 %in) nounwind {
 entry:
   %tmp = alloca [2 x i32]
-  %tmp1 = getelementptr [2 x i32], [2 x i32]* %tmp, i32 0, i32 0
-  %tmp2 = getelementptr [2 x i32], [2 x i32]* %tmp, i32 0, i32 1
+  %tmp1 = getelementptr inbounds [2 x i32], [2 x i32]* %tmp, i32 0, i32 0
+  %tmp2 = getelementptr inbounds [2 x i32], [2 x i32]* %tmp, i32 0, i32 1
   store i32 0, i32* %tmp1
   store i32 1, i32* %tmp2
-  %tmp3 = getelementptr [2 x i32], [2 x i32]* %tmp, i32 0, i32 %in
+  %tmp3 = getelementptr inbounds [2 x i32], [2 x i32]* %tmp, i32 0, i32 %in
   %tmp4 = atomicrmw add i32* %tmp3, i32 7 acq_rel
   store i32 %tmp4, i32 addrspace(1)* %out
   ret void
@@ -20,11 +20,11 @@
 define void @cmpxchg_private(i32 addrspace(1)* %out, i32 %in) nounwind {
 entry:
   %tmp = alloca [2 x i32]
-  %tmp1 = getelementptr [2 x i32], [2 x i32]* %tmp, i32 0, i32 0
-  %tmp2 = getelementptr [2 x i32], [2 x i32]* %tmp, i32 0, i32 1
+  %tmp1 = getelementptr inbounds [2 x i32], [2 x i32]* %tmp, i32 0, i32 0
+  %tmp2 = getelementptr inbounds [2 x i32], [2 x i32]* %tmp, i32 0, i32 1
   store i32 0, i32* %tmp1
   store i32 1, i32* %tmp2
-  %tmp3 = getelementptr [2 x i32], [2 x i32]* %tmp, i32 0, i32 %in
+  %tmp3 = getelementptr inbounds [2 x i32], [2 x i32]* %tmp, i32 0, i32 %in
   %tmp4 = cmpxchg i32* %tmp3, i32 0, i32 1 acq_rel monotonic
   %val = extractvalue { i32, i1 } %tmp4, 0
   store i32 %val, i32 addrspace(1)* %out
Index: test/CodeGen/AMDGPU/private-memory-r600.ll
===================================================================
--- test/CodeGen/AMDGPU/private-memory-r600.ll
+++ test/CodeGen/AMDGPU/private-memory-r600.ll
@@ -51,16 +51,16 @@
 entry:
   %a = alloca %struct.point
   %b = alloca %struct.point
-  %a.x.ptr = getelementptr %struct.point, %struct.point* %a, i32 0, i32 0
-  %a.y.ptr = getelementptr %struct.point, %struct.point* %a, i32 0, i32 1
-  %b.x.ptr = getelementptr %struct.point, %struct.point* %b, i32 0, i32 0
-  %b.y.ptr = getelementptr %struct.point, %struct.point* %b, i32 0, i32 1
+  %a.x.ptr = getelementptr inbounds %struct.point, %struct.point* %a, i32 0, i32 0
+  %a.y.ptr = getelementptr inbounds %struct.point, %struct.point* %a, i32 0, i32 1
+  %b.x.ptr = getelementptr inbounds %struct.point, %struct.point* %b, i32 0, i32 0
+  %b.y.ptr = getelementptr inbounds %struct.point, %struct.point* %b, i32 0, i32 1
   store i32 0, i32* %a.x.ptr
   store i32 1, i32* %a.y.ptr
   store i32 2, i32* %b.x.ptr
   store i32 3, i32* %b.y.ptr
-  %a.indirect.ptr = getelementptr %struct.point, %struct.point* %a, i32 0, i32 0
-  %b.indirect.ptr = getelementptr %struct.point, %struct.point* %b, i32 0, i32 0
+  %a.indirect.ptr = getelementptr inbounds %struct.point, %struct.point* %a, i32 0, i32 0
+  %b.indirect.ptr = getelementptr inbounds %struct.point, %struct.point* %b, i32 0, i32 0
   %a.indirect = load i32, i32* %a.indirect.ptr
   %b.indirect = load i32, i32* %b.indirect.ptr
   %0 = add i32 %a.indirect, %b.indirect
@@ -80,19 +80,19 @@
   %prv_array_const = alloca [2 x i32]
   %prv_array = alloca [2 x i32]
   %a = load i32, i32 addrspace(1)* %in
-  %b_src_ptr = getelementptr i32, i32 addrspace(1)* %in, i32 1
+  %b_src_ptr = getelementptr inbounds i32, i32 addrspace(1)* %in, i32 1
   %b = load i32, i32 addrspace(1)* %b_src_ptr
-  %a_dst_ptr = getelementptr [2 x i32], [2 x i32]* %prv_array_const, i32 0, i32 0
+  %a_dst_ptr = getelementptr inbounds [2 x i32], [2 x i32]* %prv_array_const, i32 0, i32 0
   store i32 %a, i32* %a_dst_ptr
-  %b_dst_ptr = getelementptr [2 x i32], [2 x i32]* %prv_array_const, i32 0, i32 1
+  %b_dst_ptr = getelementptr inbounds [2 x i32], [2 x i32]* %prv_array_const, i32 0, i32 1
   store i32 %b, i32* %b_dst_ptr
   br label %for.body
 
 for.body:
   %inc = phi i32 [0, %entry], [%count, %for.body]
-  %x_ptr = getelementptr [2 x i32], [2 x i32]* %prv_array_const, i32 0, i32 0
+  %x_ptr = getelementptr inbounds [2 x i32], [2 x i32]* %prv_array_const, i32 0, i32 0
   %x = load i32, i32* %x_ptr
-  %y_ptr = getelementptr [2 x i32], [2 x i32]* %prv_array, i32 0, i32 0
+  %y_ptr = getelementptr inbounds [2 x i32], [2 x i32]* %prv_array, i32 0, i32 0
   %y = load i32, i32* %y_ptr
   %xy = add i32 %x, %y
   store i32 %xy, i32* %y_ptr
@@ -101,7 +101,7 @@
   br i1 %done, label %for.end, label %for.body
 
 for.end:
-  %value_ptr = getelementptr [2 x i32], [2 x i32]* %prv_array, i32 0, i32 0
+  %value_ptr = getelementptr inbounds [2 x i32], [2 x i32]* %prv_array, i32 0, i32 0
   %value = load i32, i32* %value_ptr
   store i32 %value, i32 addrspace(1)* %out
   ret void
@@ -113,11 +113,11 @@
 define void @short_array(i32 addrspace(1)* %out, i32 %index) {
 entry:
   %0 = alloca [2 x i16]
-  %1 = getelementptr [2 x i16], [2 x i16]* %0, i32 0, i32 0
-  %2 = getelementptr [2 x i16], [2 x i16]* %0, i32 0, i32 1
+  %1 = getelementptr inbounds [2 x i16], [2 x i16]* %0, i32 0, i32 0
+  %2 = getelementptr inbounds [2 x i16], [2 x i16]* %0, i32 0, i32 1
   store i16 0, i16* %1
   store i16 1, i16* %2
-  %3 = getelementptr [2 x i16], [2 x i16]* %0, i32 0, i32 %index
+  %3 = getelementptr inbounds [2 x i16], [2 x i16]* %0, i32 0, i32 %index
   %4 = load i16, i16* %3
   %5 = sext i16 %4 to i32
   store i32 %5, i32 addrspace(1)* %out
@@ -130,11 +130,11 @@
 define void @char_array(i32 addrspace(1)* %out, i32 %index) {
 entry:
   %0 = alloca [2 x i8]
-  %1 = getelementptr [2 x i8], [2 x i8]* %0, i32 0, i32 0
-  %2 = getelementptr [2 x i8], [2 x i8]* %0, i32 0, i32 1
+  %1 = getelementptr inbounds [2 x i8], [2 x i8]* %0, i32 0, i32 0
+  %2 = getelementptr inbounds [2 x i8], [2 x i8]* %0, i32 0, i32 1
   store i8 0, i8* %1
   store i8 1, i8* %2
-  %3 = getelementptr [2 x i8], [2 x i8]* %0, i32 0, i32 %index
+  %3 = getelementptr inbounds [2 x i8], [2 x i8]* %0, i32 0, i32 %index
   %4 = load i8, i8* %3
   %5 = sext i8 %4 to i32
   store i32 %5, i32 addrspace(1)* %out
@@ -151,11 +151,11 @@
 define void @work_item_info(i32 addrspace(1)* %out, i32 %in) {
 entry:
   %0 = alloca [2 x i32]
-  %1 = getelementptr [2 x i32], [2 x i32]* %0, i32 0, i32 0
-  %2 = getelementptr [2 x i32], [2 x i32]* %0, i32 0, i32 1
+  %1 = getelementptr inbounds [2 x i32], [2 x i32]* %0, i32 0, i32 0
+  %2 = getelementptr inbounds [2 x i32], [2 x i32]* %0, i32 0, i32 1
   store i32 0, i32* %1
   store i32 1, i32* %2
-  %3 = getelementptr [2 x i32], [2 x i32]* %0, i32 0, i32 %in
+  %3 = getelementptr inbounds [2 x i32], [2 x i32]* %0, i32 0, i32 %in
   %4 = load i32, i32* %3
   %5 = call i32 @llvm.r600.read.tidig.x()
   %6 = add i32 %4, %5
@@ -173,18 +173,18 @@
 entry:
   %0 = alloca [3 x i8], align 1
   %1 = alloca [2 x i8], align 1
-  %2 = getelementptr [3 x i8], [3 x i8]* %0, i32 0, i32 0
-  %3 = getelementptr [3 x i8], [3 x i8]* %0, i32 0, i32 1
-  %4 = getelementptr [3 x i8], [3 x i8]* %0, i32 0, i32 2
-  %5 = getelementptr [2 x i8], [2 x i8]* %1, i32 0, i32 0
-  %6 = getelementptr [2 x i8], [2 x i8]* %1, i32 0, i32 1
+  %2 = getelementptr inbounds [3 x i8], [3 x i8]* %0, i32 0, i32 0
+  %3 = getelementptr inbounds [3 x i8], [3 x i8]* %0, i32 0, i32 1
+  %4 = getelementptr inbounds [3 x i8], [3 x i8]* %0, i32 0, i32 2
+  %5 = getelementptr inbounds [2 x i8], [2 x i8]* %1, i32 0, i32 0
+  %6 = getelementptr inbounds [2 x i8], [2 x i8]* %1, i32 0, i32 1
   store i8 0, i8* %2
   store i8 1, i8* %3
   store i8 2, i8* %4
   store i8 1, i8* %5
   store i8 0, i8* %6
-  %7 = getelementptr [3 x i8], [3 x i8]* %0, i32 0, i32 %in
-  %8 = getelementptr [2 x i8], [2 x i8]* %1, i32 0, i32 %in
+  %7 = getelementptr inbounds [3 x i8], [3 x i8]* %0, i32 0, i32 %in
+  %8 = getelementptr inbounds [2 x i8], [2 x i8]* %1, i32 0, i32 %in
   %9 = load i8, i8* %7
   %10 = load i8, i8* %8
   %11 = add i8 %9, %10
@@ -196,11 +196,11 @@
 define void @char_array_array(i32 addrspace(1)* %out, i32 %index) {
 entry:
   %alloca = alloca [2 x [2 x i8]]
-  %gep0 = getelementptr [2 x [2 x i8]], [2 x [2 x i8]]* %alloca, i32 0, i32 0, i32 0
-  %gep1 = getelementptr [2 x [2 x i8]], [2 x [2 x i8]]* %alloca, i32 0, i32 0, i32 1
+  %gep0 = getelementptr inbounds [2 x [2 x i8]], [2 x [2 x i8]]* %alloca, i32 0, i32 0, i32 0
+  %gep1 = getelementptr inbounds [2 x [2 x i8]], [2 x [2 x i8]]* %alloca, i32 0, i32 0, i32 1
   store i8 0, i8* %gep0
   store i8 1, i8* %gep1
-  %gep2 = getelementptr [2 x [2 x i8]], [2 x [2 x i8]]* %alloca, i32 0, i32 0, i32 %index
+  %gep2 = getelementptr inbounds [2 x [2 x i8]], [2 x [2 x i8]]* %alloca, i32 0, i32 0, i32 %index
   %load = load i8, i8* %gep2
   %sext = sext i8 %load to i32
   store i32 %sext, i32 addrspace(1)* %out
@@ -210,11 +210,11 @@
 define void @i32_array_array(i32 addrspace(1)* %out, i32 %index) {
 entry:
   %alloca = alloca [2 x [2 x i32]]
-  %gep0 = getelementptr [2 x [2 x i32]], [2 x [2 x i32]]* %alloca, i32 0, i32 0, i32 0
-  %gep1 = getelementptr [2 x [2 x i32]], [2 x [2 x i32]]* %alloca, i32 0, i32 0, i32 1
+  %gep0 = getelementptr inbounds [2 x [2 x i32]], [2 x [2 x i32]]* %alloca, i32 0, i32 0, i32 0
+  %gep1 = getelementptr inbounds [2 x [2 x i32]], [2 x [2 x i32]]* %alloca, i32 0, i32 0, i32 1
   store i32 0, i32* %gep0
   store i32 1, i32* %gep1
-  %gep2 = getelementptr [2 x [2 x i32]], [2 x [2 x i32]]* %alloca, i32 0, i32 0, i32 %index
+  %gep2 = getelementptr inbounds [2 x [2 x i32]], [2 x [2 x i32]]* %alloca, i32 0, i32 0, i32 %index
   %load = load i32, i32* %gep2
   store i32 %load, i32 addrspace(1)* %out
   ret void
@@ -223,11 +223,11 @@
 define void @i64_array_array(i64 addrspace(1)* %out, i32 %index) {
 entry:
   %alloca = alloca [2 x [2 x i64]]
-  %gep0 = getelementptr [2 x [2 x i64]], [2 x [2 x i64]]* %alloca, i32 0, i32 0, i32 0
-  %gep1 = getelementptr [2 x [2 x i64]], [2 x [2 x i64]]* %alloca, i32 0, i32 0, i32 1
+  %gep0 = getelementptr inbounds [2 x [2 x i64]], [2 x [2 x i64]]* %alloca, i32 0, i32 0, i32 0
+  %gep1 = getelementptr inbounds [2 x [2 x i64]], [2 x [2 x i64]]* %alloca, i32 0, i32 0, i32 1
   store i64 0, i64* %gep0
   store i64 1, i64* %gep1
-  %gep2 = getelementptr [2 x [2 x i64]], [2 x [2 x i64]]* %alloca, i32 0, i32 0, i32 %index
+  %gep2 = getelementptr inbounds [2 x [2 x i64]], [2 x [2 x i64]]* %alloca, i32 0, i32 0, i32 %index
   %load = load i64, i64* %gep2
   store i64 %load, i64 addrspace(1)* %out
   ret void
@@ -238,11 +238,11 @@
 define void @struct_array_array(i32 addrspace(1)* %out, i32 %index) {
 entry:
   %alloca = alloca [2 x [2 x %struct.pair32]]
-  %gep0 = getelementptr [2 x [2 x %struct.pair32]], [2 x [2 x %struct.pair32]]* %alloca, i32 0, i32 0, i32 0, i32 1
-  %gep1 = getelementptr [2 x [2 x %struct.pair32]], [2 x [2 x %struct.pair32]]* %alloca, i32 0, i32 0, i32 1, i32 1
+  %gep0 = getelementptr inbounds [2 x [2 x %struct.pair32]], [2 x [2 x %struct.pair32]]* %alloca, i32 0, i32 0, i32 0, i32 1
+  %gep1 = getelementptr inbounds [2 x [2 x %struct.pair32]], [2 x [2 x %struct.pair32]]* %alloca, i32 0, i32 0, i32 1, i32 1
   store i32 0, i32* %gep0
   store i32 1, i32* %gep1
-  %gep2 = getelementptr [2 x [2 x %struct.pair32]], [2 x [2 x %struct.pair32]]* %alloca, i32 0, i32 0, i32 %index, i32 0
+  %gep2 = getelementptr inbounds [2 x [2 x %struct.pair32]], [2 x [2 x %struct.pair32]]* %alloca, i32 0, i32 0, i32 %index, i32 0
   %load = load i32, i32* %gep2
   store i32 %load, i32 addrspace(1)* %out
   ret void
@@ -251,11 +251,11 @@
 define void @struct_pair32_array(i32 addrspace(1)* %out, i32 %index) {
 entry:
   %alloca = alloca [2 x %struct.pair32]
-  %gep0 = getelementptr [2 x %struct.pair32], [2 x %struct.pair32]* %alloca, i32 0, i32 0, i32 1
-  %gep1 = getelementptr [2 x %struct.pair32], [2 x %struct.pair32]* %alloca, i32 0, i32 1, i32 0
+  %gep0 = getelementptr inbounds [2 x %struct.pair32], [2 x %struct.pair32]* %alloca, i32 0, i32 0, i32 1
+  %gep1 = getelementptr inbounds [2 x %struct.pair32], [2 x %struct.pair32]* %alloca, i32 0, i32 1, i32 0
   store i32 0, i32* %gep0
   store i32 1, i32* %gep1
-  %gep2 = getelementptr [2 x %struct.pair32], [2 x %struct.pair32]* %alloca, i32 0, i32 %index, i32 0
+  %gep2 = getelementptr inbounds [2 x %struct.pair32], [2 x %struct.pair32]* %alloca, i32 0, i32 %index, i32 0
   %load = load i32, i32* %gep2
   store i32 %load, i32 addrspace(1)* %out
   ret void
@@ -264,8 +264,8 @@
 define void @select_private(i32 addrspace(1)* %out, i32 %in) nounwind {
 entry:
   %tmp = alloca [2 x i32]
-  %tmp1 = getelementptr [2 x i32], [2 x i32]* %tmp, i32 0, i32 0
-  %tmp2 = getelementptr [2 x i32], [2 x i32]* %tmp, i32 0, i32 1
+  %tmp1 = getelementptr inbounds  [2 x i32], [2 x i32]* %tmp, i32 0, i32 0
+  %tmp2 = getelementptr inbounds [2 x i32], [2 x i32]* %tmp, i32 0, i32 1
   store i32 0, i32* %tmp1
   store i32 1, i32* %tmp2
   %cmp = icmp eq i32 %in, 0
@@ -284,12 +284,12 @@
 ; SI: buffer_load_dword v{{[0-9]+}}, v{{[0-9]+}}, s[{{[0-9]+:[0-9]+}}], s{{[0-9]+}} offen ;
 define void @ptrtoint(i32 addrspace(1)* %out, i32 %a, i32 %b) {
   %alloca = alloca [16 x i32]
-  %tmp0 = getelementptr [16 x i32], [16 x i32]* %alloca, i32 0, i32 %a
+  %tmp0 = getelementptr inbounds [16 x i32], [16 x i32]* %alloca, i32 0, i32 %a
   store i32 5, i32* %tmp0
   %tmp1 = ptrtoint [16 x i32]* %alloca to i32
   %tmp2 = add i32 %tmp1, 5
   %tmp3 = inttoptr i32 %tmp2 to i32*
-  %tmp4 = getelementptr i32, i32* %tmp3, i32 %b
+  %tmp4 = getelementptr inbounds i32, i32* %tmp3, i32 %b
   %tmp5 = load i32, i32* %tmp4
   store i32 %tmp5, i32 addrspace(1)* %out
   ret void
Index: test/CodeGen/AMDGPU/private-memory.ll
===================================================================
--- test/CodeGen/AMDGPU/private-memory.ll
+++ test/CodeGen/AMDGPU/private-memory.ll
@@ -5,12 +5,13 @@
 ; RUN: llc -show-mc-encoding -mattr=+promote-alloca -verify-machineinstrs -march=amdgcn -mcpu=tonga < %s | FileCheck %s -check-prefix=SI-PROMOTE -check-prefix=SI -check-prefix=FUNC
 ; RUN: llc -show-mc-encoding -mattr=-promote-alloca -verify-machineinstrs -march=amdgcn -mcpu=tonga < %s | FileCheck %s -check-prefix=SI-ALLOCA -check-prefix=SI -check-prefix=FUNC
 
-; RUN: opt -S -mtriple=amdgcn-unknown-amdhsa -mcpu=kaveri -amdgpu-promote-alloca < %s | FileCheck -check-prefix=HSAOPT %s
-; RUN: opt -S -mtriple=amdgcn-unknown-unknown -mcpu=kaveri -amdgpu-promote-alloca < %s | FileCheck -check-prefix=NOHSAOPT %s
+; RUN: opt -S -mtriple=amdgcn-unknown-amdhsa -mcpu=kaveri -amdgpu-promote-alloca < %s | FileCheck -check-prefix=HSAOPT -check-prefix=OPT %s
+; RUN: opt -S -mtriple=amdgcn-unknown-unknown -mcpu=kaveri -amdgpu-promote-alloca < %s | FileCheck -check-prefix=NOHSAOPT -check-prefix=OPT %s
 
 declare i32 @llvm.amdgcn.workitem.id.x() nounwind readnone
 
 ; FUNC-LABEL: {{^}}mova_same_clause:
+; OPT-LABEL: @mova_same_clause(
 
 ; R600: LDS_WRITE
 ; R600: LDS_WRITE
@@ -90,6 +91,31 @@
   ret void
 }
 
+; FUNC-LABEL: {{^}}no_replace_inbounds_gep:
+; OPT-LABEL: @no_replace_inbounds_gep(
+; OPT: alloca [5 x i32]
+
+; SI-NOT: ds_write
+define void @no_replace_inbounds_gep(i32 addrspace(1)* nocapture %out, i32 addrspace(1)* nocapture %in) {
+entry:
+  %stack = alloca [5 x i32], align 4
+  %0 = load i32, i32 addrspace(1)* %in, align 4
+  %arrayidx1 = getelementptr [5 x i32], [5 x i32]* %stack, i32 0, i32 %0
+  store i32 4, i32* %arrayidx1, align 4
+  %arrayidx2 = getelementptr inbounds i32, i32 addrspace(1)* %in, i32 1
+  %1 = load i32, i32 addrspace(1)* %arrayidx2, align 4
+  %arrayidx3 = getelementptr inbounds [5 x i32], [5 x i32]* %stack, i32 0, i32 %1
+  store i32 5, i32* %arrayidx3, align 4
+  %arrayidx10 = getelementptr inbounds [5 x i32], [5 x i32]* %stack, i32 0, i32 0
+  %2 = load i32, i32* %arrayidx10, align 4
+  store i32 %2, i32 addrspace(1)* %out, align 4
+  %arrayidx12 = getelementptr inbounds [5 x i32], [5 x i32]* %stack, i32 0, i32 1
+  %3 = load i32, i32* %arrayidx12
+  %arrayidx13 = getelementptr inbounds i32, i32 addrspace(1)* %out, i32 1
+  store i32 %3, i32 addrspace(1)* %arrayidx13
+  ret void
+}
+
 ; This test checks that the stack offset is calculated correctly for structs.
 ; All register loads/stores should be optimized away, so there shouldn't be
 ; any MOVA instructions.
@@ -98,6 +124,8 @@
 ; this.
 
 ; FUNC-LABEL: {{^}}multiple_structs:
+; OPT-LABEL: @multiple_structs(
+
 ; R600-NOT: MOVA_INT
 ; SI-NOT: v_movrel
 ; SI-NOT: v_movrel
@@ -137,19 +165,19 @@
   %prv_array_const = alloca [2 x i32]
   %prv_array = alloca [2 x i32]
   %a = load i32, i32 addrspace(1)* %in
-  %b_src_ptr = getelementptr i32, i32 addrspace(1)* %in, i32 1
+  %b_src_ptr = getelementptr inbounds i32, i32 addrspace(1)* %in, i32 1
   %b = load i32, i32 addrspace(1)* %b_src_ptr
-  %a_dst_ptr = getelementptr [2 x i32], [2 x i32]* %prv_array_const, i32 0, i32 0
+  %a_dst_ptr = getelementptr inbounds [2 x i32], [2 x i32]* %prv_array_const, i32 0, i32 0
   store i32 %a, i32* %a_dst_ptr
-  %b_dst_ptr = getelementptr [2 x i32], [2 x i32]* %prv_array_const, i32 0, i32 1
+  %b_dst_ptr = getelementptr inbounds [2 x i32], [2 x i32]* %prv_array_const, i32 0, i32 1
   store i32 %b, i32* %b_dst_ptr
   br label %for.body
 
 for.body:
   %inc = phi i32 [0, %entry], [%count, %for.body]
-  %x_ptr = getelementptr [2 x i32], [2 x i32]* %prv_array_const, i32 0, i32 0
+  %x_ptr = getelementptr inbounds [2 x i32], [2 x i32]* %prv_array_const, i32 0, i32 0
   %x = load i32, i32* %x_ptr
-  %y_ptr = getelementptr [2 x i32], [2 x i32]* %prv_array, i32 0, i32 0
+  %y_ptr = getelementptr inbounds [2 x i32], [2 x i32]* %prv_array, i32 0, i32 0
   %y = load i32, i32* %y_ptr
   %xy = add i32 %x, %y
   store i32 %xy, i32* %y_ptr
@@ -158,7 +186,7 @@
   br i1 %done, label %for.end, label %for.body
 
 for.end:
-  %value_ptr = getelementptr [2 x i32], [2 x i32]* %prv_array, i32 0, i32 0
+  %value_ptr = getelementptr inbounds [2 x i32], [2 x i32]* %prv_array, i32 0, i32 0
   %value = load i32, i32* %value_ptr
   store i32 %value, i32 addrspace(1)* %out
   ret void
@@ -174,11 +202,11 @@
 define void @short_array(i32 addrspace(1)* %out, i32 %index) {
 entry:
   %0 = alloca [2 x i16]
-  %1 = getelementptr [2 x i16], [2 x i16]* %0, i32 0, i32 0
-  %2 = getelementptr [2 x i16], [2 x i16]* %0, i32 0, i32 1
+  %1 = getelementptr inbounds [2 x i16], [2 x i16]* %0, i32 0, i32 0
+  %2 = getelementptr inbounds [2 x i16], [2 x i16]* %0, i32 0, i32 1
   store i16 0, i16* %1
   store i16 1, i16* %2
-  %3 = getelementptr [2 x i16], [2 x i16]* %0, i32 0, i32 %index
+  %3 = getelementptr inbounds [2 x i16], [2 x i16]* %0, i32 0, i32 %index
   %4 = load i16, i16* %3
   %5 = sext i16 %4 to i32
   store i32 %5, i32 addrspace(1)* %out
@@ -194,11 +222,11 @@
 define void @char_array(i32 addrspace(1)* %out, i32 %index) {
 entry:
   %0 = alloca [2 x i8]
-  %1 = getelementptr [2 x i8], [2 x i8]* %0, i32 0, i32 0
-  %2 = getelementptr [2 x i8], [2 x i8]* %0, i32 0, i32 1
+  %1 = getelementptr inbounds [2 x i8], [2 x i8]* %0, i32 0, i32 0
+  %2 = getelementptr inbounds [2 x i8], [2 x i8]* %0, i32 0, i32 1
   store i8 0, i8* %1
   store i8 1, i8* %2
-  %3 = getelementptr [2 x i8], [2 x i8]* %0, i32 0, i32 %index
+  %3 = getelementptr inbounds [2 x i8], [2 x i8]* %0, i32 0, i32 %index
   %4 = load i8, i8* %3
   %5 = sext i8 %4 to i32
   store i32 %5, i32 addrspace(1)* %out