diff --git a/llvm/test/CodeGen/AMDGPU/branch-folding-implicit-def-subreg.ll b/llvm/test/CodeGen/AMDGPU/branch-folding-implicit-def-subreg.ll
index d89b39348ad9a..0310b7e788ddf 100644
--- a/llvm/test/CodeGen/AMDGPU/branch-folding-implicit-def-subreg.ll
+++ b/llvm/test/CodeGen/AMDGPU/branch-folding-implicit-def-subreg.ll
@@ -62,7 +62,7 @@ define amdgpu_kernel void @f1(ptr addrspace(1) %arg, ptr addrspace(1) %arg1, i64
   ; GFX90A-NEXT:   successors: %bb.35(0x40000000), %bb.5(0x40000000)
   ; GFX90A-NEXT:   liveins: $sgpr14, $sgpr15, $sgpr16, $sgpr17, $vgpr30, $vgpr31, $sgpr4_sgpr5, $sgpr6_sgpr7, $sgpr8_sgpr9:0x000000000000000F, $sgpr10_sgpr11, $sgpr12_sgpr13, $sgpr28_sgpr29, $sgpr30_sgpr31, $sgpr36_sgpr37, $sgpr56_sgpr57:0x000000000000000F, $sgpr20_sgpr21_sgpr22_sgpr23:0x000000000000003F, $sgpr24_sgpr25_sgpr26_sgpr27:0x00000000000000FF, $vgpr4_vgpr5:0x000000000000000F, $sgpr0_sgpr1_sgpr2_sgpr3, $sgpr40_sgpr41
   ; GFX90A-NEXT: {{  $}}
-  ; GFX90A-NEXT:   renamable $vgpr0_vgpr1 = V_LSHLREV_B64_e64 2, $vgpr4_vgpr5, implicit $exec
+  ; GFX90A-NEXT:   renamable $vgpr0_vgpr1 = nsw V_LSHLREV_B64_e64 2, $vgpr4_vgpr5, implicit $exec
   ; GFX90A-NEXT:   renamable $vgpr2 = COPY renamable $sgpr25, implicit $exec
   ; GFX90A-NEXT:   renamable $vgpr46, renamable $vcc = V_ADD_CO_U32_e64 $sgpr24, $vgpr0, 0, implicit $exec
   ; GFX90A-NEXT:   renamable $vgpr47, dead renamable $vcc = V_ADDC_U32_e64 killed $vgpr2, killed $vgpr1, killed $vcc, 0, implicit $exec
@@ -959,7 +959,7 @@ define amdgpu_kernel void @f1(ptr addrspace(1) %arg, ptr addrspace(1) %arg1, i64
   ; GFX90A-NEXT:   successors: %bb.71(0x80000000)
   ; GFX90A-NEXT:   liveins: $sgpr14, $sgpr15, $sgpr16, $vgpr12, $vgpr17, $vgpr30, $vgpr31, $agpr0_agpr1:0x0000000000000003, $sgpr4_sgpr5, $sgpr6_sgpr7, $sgpr8_sgpr9:0x000000000000000F, $sgpr10_sgpr11, $sgpr18_sgpr19, $sgpr34_sgpr35, $sgpr38_sgpr39, $sgpr40_sgpr41, $sgpr42_sgpr43, $sgpr44_sgpr45, $sgpr46_sgpr47, $sgpr48_sgpr49, $sgpr50_sgpr51, $sgpr52_sgpr53, $sgpr54_sgpr55, $sgpr64_sgpr65, $sgpr66_sgpr67, $sgpr20_sgpr21_sgpr22_sgpr23:0x000000000000003C, $sgpr24_sgpr25_sgpr26_sgpr27:0x00000000000000F0, $vgpr0_vgpr1:0x000000000000000F, $vgpr2_vgpr3:0x000000000000000C, $vgpr4_vgpr5:0x000000000000000F, $vgpr6_vgpr7:0x000000000000000F, $vgpr8_vgpr9:0x000000000000000F, $vgpr10_vgpr11:0x000000000000000F, $vgpr14_vgpr15:0x000000000000000C, $vgpr16_vgpr17:0x0000000000000003, $vgpr18_vgpr19:0x000000000000000C, $vgpr24_vgpr25:0x000000000000000C, $vgpr26_vgpr27:0x000000000000000C, $vgpr28_vgpr29:0x000000000000000C, $vgpr32_vgpr33:0x0000000000000003, $vgpr34_vgpr35:0x0000000000000003, $vgpr36_vgpr37:0x0000000000000003, $vgpr38_vgpr39:0x0000000000000003, $vgpr40_vgpr41:0x000000000000000F, $vgpr42_vgpr43:0x000000000000000F, $vgpr44_vgpr45:0x000000000000000F, $vgpr46_vgpr47:0x000000000000000F, $vgpr48_vgpr49:0x0000000000000003, $vgpr50_vgpr51:0x0000000000000003, $vgpr52_vgpr53:0x0000000000000003, $vgpr54_vgpr55:0x0000000000000003, $vgpr56_vgpr57:0x000000000000000F, $vgpr58_vgpr59:0x000000000000000F, $vgpr60_vgpr61:0x000000000000000F, $vgpr62_vgpr63:0x000000000000000F, $sgpr0_sgpr1_sgpr2_sgpr3
   ; GFX90A-NEXT: {{  $}}
-  ; GFX90A-NEXT:   renamable $vgpr4_vgpr5 = V_LSHLREV_B64_e64 3, killed $vgpr4_vgpr5, implicit $exec
+  ; GFX90A-NEXT:   renamable $vgpr4_vgpr5 = nsw V_LSHLREV_B64_e64 3, killed $vgpr4_vgpr5, implicit $exec
   ; GFX90A-NEXT:   renamable $vgpr2 = COPY renamable $sgpr27, implicit $exec
   ; GFX90A-NEXT:   renamable $vgpr4, renamable $vcc = V_ADD_CO_U32_e64 killed $sgpr26, $vgpr4, 0, implicit $exec
   ; GFX90A-NEXT:   renamable $vgpr2, dead renamable $vcc = V_ADDC_U32_e64 killed $vgpr2, killed $vgpr5, killed $vcc, 0, implicit $exec
@@ -1007,12 +1007,12 @@ bb:
   %i11 = icmp eq i32 %i, 0
   %i12 = load i32, ptr addrspace(3) null, align 8
   %i13 = zext i32 %i12 to i64
-  %i14 = getelementptr i32, ptr addrspace(1) %arg, i64 %i13
+  %i14 = getelementptr inbounds i32, ptr addrspace(1) %arg, i64 %i13
   br i1 %arg3, label %bb15, label %bb103
 
 bb15:
   %i16 = zext i32 %i to i64
-  %i17 = getelementptr i32, ptr addrspace(1) %i14, i64 %i16
+  %i17 = getelementptr inbounds i32, ptr addrspace(1) %i14, i64 %i16
   %i18 = ptrtoint ptr addrspace(1) %i17 to i64
   br i1 %arg4, label %bb19, label %bb20
 
@@ -1021,7 +1021,7 @@ bb19:
   unreachable
 
 bb20:
-  %i21 = getelementptr i32, ptr addrspace(1) %i17, i64 256
+  %i21 = getelementptr inbounds i32, ptr addrspace(1) %i17, i64 256
   %i22 = ptrtoint ptr addrspace(1) %i21 to i64
   %i23 = inttoptr i64 %i22 to ptr
   %i24 = load i8, ptr %i23, align 1
@@ -1033,7 +1033,7 @@ bb26:
   unreachable
 
 bb27:
-  %i28 = getelementptr i32, ptr addrspace(1) %i17, i64 512
+  %i28 = getelementptr inbounds i32, ptr addrspace(1) %i17, i64 512
   %i29 = ptrtoint ptr addrspace(1) %i28 to i64
   %i30 = inttoptr i64 %i29 to ptr
   %i31 = load i8, ptr %i30, align 1
@@ -1045,7 +1045,7 @@ bb33:
   unreachable
 
 bb34:
-  %i35 = getelementptr i32, ptr addrspace(1) %i17, i64 768
+  %i35 = getelementptr inbounds i32, ptr addrspace(1) %i17, i64 768
   %i36 = ptrtoint ptr addrspace(1) %i35 to i64
   %i37 = inttoptr i64 %i36 to ptr
   %i38 = load i8, ptr %i37, align 1
@@ -1057,7 +1057,7 @@ bb40:
   unreachable
 
 bb41:
-  %i42 = getelementptr i32, ptr addrspace(1) %i17, i64 1024
+  %i42 = getelementptr inbounds i32, ptr addrspace(1) %i17, i64 1024
   %i43 = ptrtoint ptr addrspace(1) %i42 to i64
   %i44 = inttoptr i64 %i43 to ptr
   %i45 = load i8, ptr %i44, align 1
@@ -1069,7 +1069,7 @@ bb47:
   unreachable
 
 bb48:
-  %i49 = getelementptr i32, ptr addrspace(1) %i17, i64 1280
+  %i49 = getelementptr inbounds i32, ptr addrspace(1) %i17, i64 1280
   %i50 = ptrtoint ptr addrspace(1) %i49 to i64
   %i51 = inttoptr i64 %i50 to ptr
   %i52 = load i8, ptr %i51, align 1
@@ -1081,7 +1081,7 @@ bb54:
   unreachable
 
 bb55:
-  %i56 = getelementptr i32, ptr addrspace(1) %i17, i64 1536
+  %i56 = getelementptr inbounds i32, ptr addrspace(1) %i17, i64 1536
   %i57 = ptrtoint ptr addrspace(1) %i56 to i64
   %i58 = or i64 %i57, 1
   %i59 = inttoptr i64 %i58 to ptr
@@ -1113,7 +1113,7 @@ bb67:
 
 bb68:
   %i69 = zext i1 %arg5 to i8
-  %i70 = getelementptr [2 x i32], ptr addrspace(1) null, i64 %i16
+  %i70 = getelementptr inbounds [2 x i32], ptr addrspace(1) null, i64 %i16
   %i71 = ptrtoint ptr addrspace(1) %i70 to i64
   br i1 %arg5, label %bb72, label %bb73
 
@@ -1122,7 +1122,7 @@ bb72:
   unreachable
 
 bb73:
-  %i74 = getelementptr [2 x i32], ptr addrspace(1) %i70, i64 256
+  %i74 = getelementptr inbounds [2 x i32], ptr addrspace(1) %i70, i64 256
   %i75 = ptrtoint ptr addrspace(1) %i74 to i64
   %i76 = inttoptr i64 %i75 to ptr
   %i77 = load i8, ptr %i76, align 1
@@ -1134,7 +1134,7 @@ bb79:
   unreachable
 
 bb80:
-  %i81 = getelementptr [2 x i32], ptr addrspace(1) %i70, i64 512
+  %i81 = getelementptr inbounds [2 x i32], ptr addrspace(1) %i70, i64 512
   %i82 = ptrtoint ptr addrspace(1) %i81 to i64
   %i83 = or i64 %i82, 1
   br i1 %arg6, label %bb84, label %bb85
@@ -1269,7 +1269,7 @@ bb174:
   %i182 = select i1 %arg3, i32 %i181, i32 0
   %i183 = or i32 %i182, %i154
   %i184 = or i32 %i183, %i156
-  %i185 = getelementptr [2 x i32], ptr addrspace(1) %arg1, i64 %i13
+  %i185 = getelementptr inbounds [2 x i32], ptr addrspace(1) %arg1, i64 %i13
   br i1 %arg3, label %bb186, label %bb196
 
 bb186:
diff --git a/llvm/test/CodeGen/AMDGPU/cgp-addressing-modes-flat.ll b/llvm/test/CodeGen/AMDGPU/cgp-addressing-modes-flat.ll
index 890f4f77ed107..e509d7b2b9b1b 100644
--- a/llvm/test/CodeGen/AMDGPU/cgp-addressing-modes-flat.ll
+++ b/llvm/test/CodeGen/AMDGPU/cgp-addressing-modes-flat.ll
@@ -12,8 +12,8 @@
 define void @test_sinkable_flat_small_offset_i32(ptr %out, ptr %in, i32 %cond) {
 ; OPT-GFX7-LABEL: @test_sinkable_flat_small_offset_i32(
 ; OPT-GFX7-NEXT:  entry:
-; OPT-GFX7-NEXT:    [[OUT_GEP:%.*]] = getelementptr i32, ptr [[OUT:%.*]], i64 999999
-; OPT-GFX7-NEXT:    [[IN_GEP:%.*]] = getelementptr i32, ptr [[IN:%.*]], i64 7
+; OPT-GFX7-NEXT:    [[OUT_GEP:%.*]] = getelementptr inbounds i32, ptr [[OUT:%.*]], i64 999999
+; OPT-GFX7-NEXT:    [[IN_GEP:%.*]] = getelementptr inbounds i32, ptr [[IN:%.*]], i64 7
 ; OPT-GFX7-NEXT:    [[CMP0:%.*]] = icmp eq i32 [[COND:%.*]], 0
 ; OPT-GFX7-NEXT:    br i1 [[CMP0]], label [[ENDIF:%.*]], label [[IF:%.*]]
 ; OPT-GFX7:       if:
@@ -28,8 +28,8 @@ define void @test_sinkable_flat_small_offset_i32(ptr %out, ptr %in, i32 %cond) {
 ;
 ; OPT-GFX8-LABEL: @test_sinkable_flat_small_offset_i32(
 ; OPT-GFX8-NEXT:  entry:
-; OPT-GFX8-NEXT:    [[OUT_GEP:%.*]] = getelementptr i32, ptr [[OUT:%.*]], i64 999999
-; OPT-GFX8-NEXT:    [[IN_GEP:%.*]] = getelementptr i32, ptr [[IN:%.*]], i64 7
+; OPT-GFX8-NEXT:    [[OUT_GEP:%.*]] = getelementptr inbounds i32, ptr [[OUT:%.*]], i64 999999
+; OPT-GFX8-NEXT:    [[IN_GEP:%.*]] = getelementptr inbounds i32, ptr [[IN:%.*]], i64 7
 ; OPT-GFX8-NEXT:    [[CMP0:%.*]] = icmp eq i32 [[COND:%.*]], 0
 ; OPT-GFX8-NEXT:    br i1 [[CMP0]], label [[ENDIF:%.*]], label [[IF:%.*]]
 ; OPT-GFX8:       if:
@@ -44,11 +44,11 @@ define void @test_sinkable_flat_small_offset_i32(ptr %out, ptr %in, i32 %cond) {
 ;
 ; OPT-GFX9-LABEL: @test_sinkable_flat_small_offset_i32(
 ; OPT-GFX9-NEXT:  entry:
-; OPT-GFX9-NEXT:    [[OUT_GEP:%.*]] = getelementptr i32, ptr [[OUT:%.*]], i64 999999
+; OPT-GFX9-NEXT:    [[OUT_GEP:%.*]] = getelementptr inbounds i32, ptr [[OUT:%.*]], i64 999999
 ; OPT-GFX9-NEXT:    [[CMP0:%.*]] = icmp eq i32 [[COND:%.*]], 0
 ; OPT-GFX9-NEXT:    br i1 [[CMP0]], label [[ENDIF:%.*]], label [[IF:%.*]]
 ; OPT-GFX9:       if:
-; OPT-GFX9-NEXT:    [[SUNKADDR:%.*]] = getelementptr i8, ptr [[IN:%.*]], i64 28
+; OPT-GFX9-NEXT:    [[SUNKADDR:%.*]] = getelementptr inbounds i8, ptr [[IN:%.*]], i64 28
 ; OPT-GFX9-NEXT:    [[LOAD:%.*]] = load i32, ptr [[SUNKADDR]], align 4
 ; OPT-GFX9-NEXT:    br label [[ENDIF]]
 ; OPT-GFX9:       endif:
@@ -58,11 +58,11 @@ define void @test_sinkable_flat_small_offset_i32(ptr %out, ptr %in, i32 %cond) {
 ;
 ; OPT-GFX10-LABEL: @test_sinkable_flat_small_offset_i32(
 ; OPT-GFX10-NEXT:  entry:
-; OPT-GFX10-NEXT:    [[OUT_GEP:%.*]] = getelementptr i32, ptr [[OUT:%.*]], i64 999999
+; OPT-GFX10-NEXT:    [[OUT_GEP:%.*]] = getelementptr inbounds i32, ptr [[OUT:%.*]], i64 999999
 ; OPT-GFX10-NEXT:    [[CMP0:%.*]] = icmp eq i32 [[COND:%.*]], 0
 ; OPT-GFX10-NEXT:    br i1 [[CMP0]], label [[ENDIF:%.*]], label [[IF:%.*]]
 ; OPT-GFX10:       if:
-; OPT-GFX10-NEXT:    [[SUNKADDR:%.*]] = getelementptr i8, ptr [[IN:%.*]], i64 28
+; OPT-GFX10-NEXT:    [[SUNKADDR:%.*]] = getelementptr inbounds i8, ptr [[IN:%.*]], i64 28
 ; OPT-GFX10-NEXT:    [[LOAD:%.*]] = load i32, ptr [[SUNKADDR]], align 4
 ; OPT-GFX10-NEXT:    br label [[ENDIF]]
 ; OPT-GFX10:       endif:
@@ -146,8 +146,8 @@ define void @test_sinkable_flat_small_offset_i32(ptr %out, ptr %in, i32 %cond) {
 ; GFX10-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX10-NEXT:    s_setpc_b64 s[30:31]
 entry:
-  %out.gep = getelementptr i32, ptr %out, i64 999999
-  %in.gep = getelementptr i32, ptr %in, i64 7
+  %out.gep = getelementptr inbounds i32, ptr %out, i64 999999
+  %in.gep = getelementptr inbounds i32, ptr %in, i64 7
   %cmp0 = icmp eq i32 %cond, 0
   br i1 %cmp0, label %endif, label %if
 
@@ -167,12 +167,12 @@ done:
 define void @test_sink_noop_addrspacecast_flat_to_global_i32(ptr %out, ptr %in, i32 %cond) {
 ; OPT-GFX7-LABEL: @test_sink_noop_addrspacecast_flat_to_global_i32(
 ; OPT-GFX7-NEXT:  entry:
-; OPT-GFX7-NEXT:    [[OUT_GEP:%.*]] = getelementptr i32, ptr [[OUT:%.*]], i64 999999
+; OPT-GFX7-NEXT:    [[OUT_GEP:%.*]] = getelementptr inbounds i32, ptr [[OUT:%.*]], i64 999999
 ; OPT-GFX7-NEXT:    [[CMP0:%.*]] = icmp eq i32 [[COND:%.*]], 0
 ; OPT-GFX7-NEXT:    br i1 [[CMP0]], label [[ENDIF:%.*]], label [[IF:%.*]]
 ; OPT-GFX7:       if:
 ; OPT-GFX7-NEXT:    [[TMP0:%.*]] = addrspacecast ptr [[IN:%.*]] to ptr addrspace(1)
-; OPT-GFX7-NEXT:    [[SUNKADDR:%.*]] = getelementptr i8, ptr addrspace(1) [[TMP0]], i64 28
+; OPT-GFX7-NEXT:    [[SUNKADDR:%.*]] = getelementptr inbounds i8, ptr addrspace(1) [[TMP0]], i64 28
 ; OPT-GFX7-NEXT:    [[LOAD:%.*]] = load i32, ptr addrspace(1) [[SUNKADDR]], align 4
 ; OPT-GFX7-NEXT:    br label [[ENDIF]]
 ; OPT-GFX7:       endif:
@@ -182,8 +182,8 @@ define void @test_sink_noop_addrspacecast_flat_to_global_i32(ptr %out, ptr %in,
 ;
 ; OPT-GFX8-LABEL: @test_sink_noop_addrspacecast_flat_to_global_i32(
 ; OPT-GFX8-NEXT:  entry:
-; OPT-GFX8-NEXT:    [[OUT_GEP:%.*]] = getelementptr i32, ptr [[OUT:%.*]], i64 999999
-; OPT-GFX8-NEXT:    [[IN_GEP:%.*]] = getelementptr i32, ptr [[IN:%.*]], i64 7
+; OPT-GFX8-NEXT:    [[OUT_GEP:%.*]] = getelementptr inbounds i32, ptr [[OUT:%.*]], i64 999999
+; OPT-GFX8-NEXT:    [[IN_GEP:%.*]] = getelementptr inbounds i32, ptr [[IN:%.*]], i64 7
 ; OPT-GFX8-NEXT:    [[CMP0:%.*]] = icmp eq i32 [[COND:%.*]], 0
 ; OPT-GFX8-NEXT:    br i1 [[CMP0]], label [[ENDIF:%.*]], label [[IF:%.*]]
 ; OPT-GFX8:       if:
@@ -197,12 +197,12 @@ define void @test_sink_noop_addrspacecast_flat_to_global_i32(ptr %out, ptr %in,
 ;
 ; OPT-GFX9-LABEL: @test_sink_noop_addrspacecast_flat_to_global_i32(
 ; OPT-GFX9-NEXT:  entry:
-; OPT-GFX9-NEXT:    [[OUT_GEP:%.*]] = getelementptr i32, ptr [[OUT:%.*]], i64 999999
+; OPT-GFX9-NEXT:    [[OUT_GEP:%.*]] = getelementptr inbounds i32, ptr [[OUT:%.*]], i64 999999
 ; OPT-GFX9-NEXT:    [[CMP0:%.*]] = icmp eq i32 [[COND:%.*]], 0
 ; OPT-GFX9-NEXT:    br i1 [[CMP0]], label [[ENDIF:%.*]], label [[IF:%.*]]
 ; OPT-GFX9:       if:
 ; OPT-GFX9-NEXT:    [[TMP0:%.*]] = addrspacecast ptr [[IN:%.*]] to ptr addrspace(1)
-; OPT-GFX9-NEXT:    [[SUNKADDR:%.*]] = getelementptr i8, ptr addrspace(1) [[TMP0]], i64 28
+; OPT-GFX9-NEXT:    [[SUNKADDR:%.*]] = getelementptr inbounds i8, ptr addrspace(1) [[TMP0]], i64 28
 ; OPT-GFX9-NEXT:    [[LOAD:%.*]] = load i32, ptr addrspace(1) [[SUNKADDR]], align 4
 ; OPT-GFX9-NEXT:    br label [[ENDIF]]
 ; OPT-GFX9:       endif:
@@ -212,12 +212,12 @@ define void @test_sink_noop_addrspacecast_flat_to_global_i32(ptr %out, ptr %in,
 ;
 ; OPT-GFX10-LABEL: @test_sink_noop_addrspacecast_flat_to_global_i32(
 ; OPT-GFX10-NEXT:  entry:
-; OPT-GFX10-NEXT:    [[OUT_GEP:%.*]] = getelementptr i32, ptr [[OUT:%.*]], i64 999999
+; OPT-GFX10-NEXT:    [[OUT_GEP:%.*]] = getelementptr inbounds i32, ptr [[OUT:%.*]], i64 999999
 ; OPT-GFX10-NEXT:    [[CMP0:%.*]] = icmp eq i32 [[COND:%.*]], 0
 ; OPT-GFX10-NEXT:    br i1 [[CMP0]], label [[ENDIF:%.*]], label [[IF:%.*]]
 ; OPT-GFX10:       if:
 ; OPT-GFX10-NEXT:    [[TMP0:%.*]] = addrspacecast ptr [[IN:%.*]] to ptr addrspace(1)
-; OPT-GFX10-NEXT:    [[SUNKADDR:%.*]] = getelementptr i8, ptr addrspace(1) [[TMP0]], i64 28
+; OPT-GFX10-NEXT:    [[SUNKADDR:%.*]] = getelementptr inbounds i8, ptr addrspace(1) [[TMP0]], i64 28
 ; OPT-GFX10-NEXT:    [[LOAD:%.*]] = load i32, ptr addrspace(1) [[SUNKADDR]], align 4
 ; OPT-GFX10-NEXT:    br label [[ENDIF]]
 ; OPT-GFX10:       endif:
@@ -303,8 +303,8 @@ define void @test_sink_noop_addrspacecast_flat_to_global_i32(ptr %out, ptr %in,
 ; GFX10-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX10-NEXT:    s_setpc_b64 s[30:31]
 entry:
-  %out.gep = getelementptr i32, ptr %out, i64 999999
-  %in.gep = getelementptr i32, ptr %in, i64 7
+  %out.gep = getelementptr inbounds i32, ptr %out, i64 999999
+  %in.gep = getelementptr inbounds i32, ptr %in, i64 7
   %cast = addrspacecast ptr %in.gep to ptr addrspace(1)
   %cmp0 = icmp eq i32 %cond, 0
   br i1 %cmp0, label %endif, label %if
@@ -325,12 +325,12 @@ done:
 define void @test_sink_noop_addrspacecast_flat_to_constant_i32(ptr %out, ptr %in, i32 %cond) {
 ; OPT-LABEL: @test_sink_noop_addrspacecast_flat_to_constant_i32(
 ; OPT-NEXT:  entry:
-; OPT-NEXT:    [[OUT_GEP:%.*]] = getelementptr i32, ptr [[OUT:%.*]], i64 999999
+; OPT-NEXT:    [[OUT_GEP:%.*]] = getelementptr inbounds i32, ptr [[OUT:%.*]], i64 999999
 ; OPT-NEXT:    [[CMP0:%.*]] = icmp eq i32 [[COND:%.*]], 0
 ; OPT-NEXT:    br i1 [[CMP0]], label [[ENDIF:%.*]], label [[IF:%.*]]
 ; OPT:       if:
 ; OPT-NEXT:    [[TMP0:%.*]] = addrspacecast ptr [[IN:%.*]] to ptr addrspace(4)
-; OPT-NEXT:    [[SUNKADDR:%.*]] = getelementptr i8, ptr addrspace(4) [[TMP0]], i64 28
+; OPT-NEXT:    [[SUNKADDR:%.*]] = getelementptr inbounds i8, ptr addrspace(4) [[TMP0]], i64 28
 ; OPT-NEXT:    [[LOAD:%.*]] = load i32, ptr addrspace(4) [[SUNKADDR]], align 4
 ; OPT-NEXT:    br label [[ENDIF]]
 ; OPT:       endif:
@@ -416,8 +416,8 @@ define void @test_sink_noop_addrspacecast_flat_to_constant_i32(ptr %out, ptr %in
 ; GFX10-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX10-NEXT:    s_setpc_b64 s[30:31]
 entry:
-  %out.gep = getelementptr i32, ptr %out, i64 999999
-  %in.gep = getelementptr i32, ptr %in, i64 7
+  %out.gep = getelementptr inbounds i32, ptr %out, i64 999999
+  %in.gep = getelementptr inbounds i32, ptr %in, i64 7
   %cast = addrspacecast ptr %in.gep to ptr addrspace(4)
   %cmp0 = icmp eq i32 %cond, 0
   br i1 %cmp0, label %endif, label %if
@@ -438,8 +438,8 @@ done:
 define void @test_sink_flat_small_max_flat_offset(ptr %out, ptr %in) #1 {
 ; OPT-GFX7-LABEL: @test_sink_flat_small_max_flat_offset(
 ; OPT-GFX7-NEXT:  entry:
-; OPT-GFX7-NEXT:    [[OUT_GEP:%.*]] = getelementptr i32, ptr [[OUT:%.*]], i32 1024
-; OPT-GFX7-NEXT:    [[IN_GEP:%.*]] = getelementptr i8, ptr [[IN:%.*]], i64 4095
+; OPT-GFX7-NEXT:    [[OUT_GEP:%.*]] = getelementptr inbounds i32, ptr [[OUT:%.*]], i32 1024
+; OPT-GFX7-NEXT:    [[IN_GEP:%.*]] = getelementptr inbounds i8, ptr [[IN:%.*]], i64 4095
 ; OPT-GFX7-NEXT:    [[TID:%.*]] = call i32 @llvm.amdgcn.mbcnt.lo(i32 -1, i32 0) #[[ATTR3:[0-9]+]]
 ; OPT-GFX7-NEXT:    [[CMP0:%.*]] = icmp eq i32 [[TID]], 0
 ; OPT-GFX7-NEXT:    br i1 [[CMP0]], label [[ENDIF:%.*]], label [[IF:%.*]]
@@ -456,8 +456,8 @@ define void @test_sink_flat_small_max_flat_offset(ptr %out, ptr %in) #1 {
 ;
 ; OPT-GFX8-LABEL: @test_sink_flat_small_max_flat_offset(
 ; OPT-GFX8-NEXT:  entry:
-; OPT-GFX8-NEXT:    [[OUT_GEP:%.*]] = getelementptr i32, ptr [[OUT:%.*]], i32 1024
-; OPT-GFX8-NEXT:    [[IN_GEP:%.*]] = getelementptr i8, ptr [[IN:%.*]], i64 4095
+; OPT-GFX8-NEXT:    [[OUT_GEP:%.*]] = getelementptr inbounds i32, ptr [[OUT:%.*]], i32 1024
+; OPT-GFX8-NEXT:    [[IN_GEP:%.*]] = getelementptr inbounds i8, ptr [[IN:%.*]], i64 4095
 ; OPT-GFX8-NEXT:    [[TID:%.*]] = call i32 @llvm.amdgcn.mbcnt.lo(i32 -1, i32 0) #[[ATTR3:[0-9]+]]
 ; OPT-GFX8-NEXT:    [[CMP0:%.*]] = icmp eq i32 [[TID]], 0
 ; OPT-GFX8-NEXT:    br i1 [[CMP0]], label [[ENDIF:%.*]], label [[IF:%.*]]
@@ -474,12 +474,12 @@ define void @test_sink_flat_small_max_flat_offset(ptr %out, ptr %in) #1 {
 ;
 ; OPT-GFX9-LABEL: @test_sink_flat_small_max_flat_offset(
 ; OPT-GFX9-NEXT:  entry:
-; OPT-GFX9-NEXT:    [[OUT_GEP:%.*]] = getelementptr i32, ptr [[OUT:%.*]], i32 1024
+; OPT-GFX9-NEXT:    [[OUT_GEP:%.*]] = getelementptr inbounds i32, ptr [[OUT:%.*]], i32 1024
 ; OPT-GFX9-NEXT:    [[TID:%.*]] = call i32 @llvm.amdgcn.mbcnt.lo(i32 -1, i32 0) #[[ATTR3:[0-9]+]]
 ; OPT-GFX9-NEXT:    [[CMP0:%.*]] = icmp eq i32 [[TID]], 0
 ; OPT-GFX9-NEXT:    br i1 [[CMP0]], label [[ENDIF:%.*]], label [[IF:%.*]]
 ; OPT-GFX9:       if:
-; OPT-GFX9-NEXT:    [[SUNKADDR:%.*]] = getelementptr i8, ptr [[IN:%.*]], i64 4095
+; OPT-GFX9-NEXT:    [[SUNKADDR:%.*]] = getelementptr inbounds i8, ptr [[IN:%.*]], i64 4095
 ; OPT-GFX9-NEXT:    [[LOAD:%.*]] = load i8, ptr [[SUNKADDR]], align 1
 ; OPT-GFX9-NEXT:    [[CAST:%.*]] = sext i8 [[LOAD]] to i32
 ; OPT-GFX9-NEXT:    br label [[ENDIF]]
@@ -490,8 +490,8 @@ define void @test_sink_flat_small_max_flat_offset(ptr %out, ptr %in) #1 {
 ;
 ; OPT-GFX10-LABEL: @test_sink_flat_small_max_flat_offset(
 ; OPT-GFX10-NEXT:  entry:
-; OPT-GFX10-NEXT:    [[OUT_GEP:%.*]] = getelementptr i32, ptr [[OUT:%.*]], i32 1024
-; OPT-GFX10-NEXT:    [[IN_GEP:%.*]] = getelementptr i8, ptr [[IN:%.*]], i64 4095
+; OPT-GFX10-NEXT:    [[OUT_GEP:%.*]] = getelementptr inbounds i32, ptr [[OUT:%.*]], i32 1024
+; OPT-GFX10-NEXT:    [[IN_GEP:%.*]] = getelementptr inbounds i8, ptr [[IN:%.*]], i64 4095
 ; OPT-GFX10-NEXT:    [[TID:%.*]] = call i32 @llvm.amdgcn.mbcnt.lo(i32 -1, i32 0) #[[ATTR3:[0-9]+]]
 ; OPT-GFX10-NEXT:    [[CMP0:%.*]] = icmp eq i32 [[TID]], 0
 ; OPT-GFX10-NEXT:    br i1 [[CMP0]], label [[ENDIF:%.*]], label [[IF:%.*]]
@@ -588,8 +588,8 @@ define void @test_sink_flat_small_max_flat_offset(ptr %out, ptr %in) #1 {
 ; GFX10-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX10-NEXT:    s_setpc_b64 s[30:31]
 entry:
-  %out.gep = getelementptr i32, ptr %out, i32 1024
-  %in.gep = getelementptr i8, ptr %in, i64 4095
+  %out.gep = getelementptr inbounds i32, ptr %out, i32 1024
+  %in.gep = getelementptr inbounds i8, ptr %in, i64 4095
   %tid = call i32 @llvm.amdgcn.mbcnt.lo(i32 -1, i32 0) #0
   %cmp0 = icmp eq i32 %tid, 0
   br i1 %cmp0, label %endif, label %if
@@ -611,8 +611,8 @@ done:
 define void @test_sink_flat_small_max_plus_1_flat_offset(ptr %out, ptr %in) #1 {
 ; OPT-LABEL: @test_sink_flat_small_max_plus_1_flat_offset(
 ; OPT-NEXT:  entry:
-; OPT-NEXT:    [[OUT_GEP:%.*]] = getelementptr i32, ptr [[OUT:%.*]], i64 99999
-; OPT-NEXT:    [[IN_GEP:%.*]] = getelementptr i8, ptr [[IN:%.*]], i64 4096
+; OPT-NEXT:    [[OUT_GEP:%.*]] = getelementptr inbounds i32, ptr [[OUT:%.*]], i64 99999
+; OPT-NEXT:    [[IN_GEP:%.*]] = getelementptr inbounds i8, ptr [[IN:%.*]], i64 4096
 ; OPT-NEXT:    [[TID:%.*]] = call i32 @llvm.amdgcn.mbcnt.lo(i32 -1, i32 0) #[[ATTR3:[0-9]+]]
 ; OPT-NEXT:    [[CMP0:%.*]] = icmp eq i32 [[TID]], 0
 ; OPT-NEXT:    br i1 [[CMP0]], label [[ENDIF:%.*]], label [[IF:%.*]]
@@ -711,8 +711,8 @@ define void @test_sink_flat_small_max_plus_1_flat_offset(ptr %out, ptr %in) #1 {
 ; GFX10-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX10-NEXT:    s_setpc_b64 s[30:31]
 entry:
-  %out.gep = getelementptr i32, ptr %out, i64 99999
-  %in.gep = getelementptr i8, ptr %in, i64 4096
+  %out.gep = getelementptr inbounds i32, ptr %out, i64 99999
+  %in.gep = getelementptr inbounds i8, ptr %in, i64 4096
   %tid = call i32 @llvm.amdgcn.mbcnt.lo(i32 -1, i32 0) #0
   %cmp0 = icmp eq i32 %tid, 0
   br i1 %cmp0, label %endif, label %if
@@ -734,8 +734,8 @@ done:
 define void @test_sinkable_flat_reg_offset(ptr %out, ptr %in, i64 %reg) #1 {
 ; OPT-LABEL: @test_sinkable_flat_reg_offset(
 ; OPT-NEXT:  entry:
-; OPT-NEXT:    [[OUT_GEP:%.*]] = getelementptr i32, ptr [[OUT:%.*]], i32 1024
-; OPT-NEXT:    [[IN_GEP:%.*]] = getelementptr i8, ptr [[IN:%.*]], i64 [[REG:%.*]]
+; OPT-NEXT:    [[OUT_GEP:%.*]] = getelementptr inbounds i32, ptr [[OUT:%.*]], i32 1024
+; OPT-NEXT:    [[IN_GEP:%.*]] = getelementptr inbounds i8, ptr [[IN:%.*]], i64 [[REG:%.*]]
 ; OPT-NEXT:    [[TID:%.*]] = call i32 @llvm.amdgcn.mbcnt.lo(i32 -1, i32 0) #[[ATTR3]]
 ; OPT-NEXT:    [[CMP0:%.*]] = icmp eq i32 [[TID]], 0
 ; OPT-NEXT:    br i1 [[CMP0]], label [[ENDIF:%.*]], label [[IF:%.*]]
@@ -834,8 +834,8 @@ define void @test_sinkable_flat_reg_offset(ptr %out, ptr %in, i64 %reg) #1 {
 ; GFX10-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX10-NEXT:    s_setpc_b64 s[30:31]
 entry:
-  %out.gep = getelementptr i32, ptr %out, i32 1024
-  %in.gep = getelementptr i8, ptr %in, i64 %reg
+  %out.gep = getelementptr inbounds i32, ptr %out, i32 1024
+  %in.gep = getelementptr inbounds i8, ptr %in, i64 %reg
   %tid = call i32 @llvm.amdgcn.mbcnt.lo(i32 -1, i32 0) #0
   %cmp0 = icmp eq i32 %tid, 0
   br i1 %cmp0, label %endif, label %if
diff --git a/llvm/test/CodeGen/AMDGPU/flat-atomicrmw-fadd.ll b/llvm/test/CodeGen/AMDGPU/flat-atomicrmw-fadd.ll
index 1a4a54b81c78f..119056a72f2c6 100644
--- a/llvm/test/CodeGen/AMDGPU/flat-atomicrmw-fadd.ll
+++ b/llvm/test/CodeGen/AMDGPU/flat-atomicrmw-fadd.ll
@@ -373,7 +373,7 @@ define float @flat_agent_atomic_fadd_ret_f32__offset12b_pos__amdgpu_no_fine_grai
 ; GFX7-NEXT:  ; %bb.2: ; %atomicrmw.end
 ; GFX7-NEXT:    s_or_b64 exec, exec, s[4:5]
 ; GFX7-NEXT:    s_setpc_b64 s[30:31]
-  %gep = getelementptr float, ptr %ptr, i64 511
+  %gep = getelementptr inbounds float, ptr %ptr, i64 511
   %result = atomicrmw fadd ptr %gep, float %val syncscope("agent") seq_cst, !amdgpu.no.fine.grained.memory !0, !amdgpu.ignore.denormal.mode !0
   ret float %result
 }
@@ -570,7 +570,7 @@ define float @flat_agent_atomic_fadd_ret_f32__offset12b_neg__amdgpu_no_fine_grai
 ; GFX7-NEXT:  ; %bb.2: ; %atomicrmw.end
 ; GFX7-NEXT:    s_or_b64 exec, exec, s[4:5]
 ; GFX7-NEXT:    s_setpc_b64 s[30:31]
-  %gep = getelementptr float, ptr %ptr, i64 -512
+  %gep = getelementptr inbounds float, ptr %ptr, i64 -512
   %result = atomicrmw fadd ptr %gep, float %val syncscope("agent") seq_cst, !amdgpu.no.fine.grained.memory !0, !amdgpu.ignore.denormal.mode !0
   ret float %result
 }
@@ -995,7 +995,7 @@ define void @flat_agent_atomic_fadd_noret_f32__offset12b_pos__amdgpu_no_fine_gra
 ; GFX7-NEXT:  ; %bb.2: ; %atomicrmw.end
 ; GFX7-NEXT:    s_or_b64 exec, exec, s[4:5]
 ; GFX7-NEXT:    s_setpc_b64 s[30:31]
-  %gep = getelementptr float, ptr %ptr, i64 511
+  %gep = getelementptr inbounds float, ptr %ptr, i64 511
   %unused = atomicrmw fadd ptr %gep, float %val syncscope("agent") seq_cst, !amdgpu.no.fine.grained.memory !0, !amdgpu.ignore.denormal.mode !0
   ret void
 }
@@ -1219,7 +1219,7 @@ define void @flat_agent_atomic_fadd_noret_f32__offset12b_neg__amdgpu_no_fine_gra
 ; GFX7-NEXT:  ; %bb.2: ; %atomicrmw.end
 ; GFX7-NEXT:    s_or_b64 exec, exec, s[4:5]
 ; GFX7-NEXT:    s_setpc_b64 s[30:31]
-  %gep = getelementptr float, ptr %ptr, i64 -512
+  %gep = getelementptr inbounds float, ptr %ptr, i64 -512
   %unused = atomicrmw fadd ptr %gep, float %val syncscope("agent") seq_cst, !amdgpu.no.fine.grained.memory !0, !amdgpu.ignore.denormal.mode !0
   ret void
 }
@@ -1409,7 +1409,7 @@ define float @flat_system_atomic_fadd_ret_f32__offset12b_pos__amdgpu_no_fine_gra
 ; GFX7-NEXT:  ; %bb.2: ; %atomicrmw.end
 ; GFX7-NEXT:    s_or_b64 exec, exec, s[4:5]
 ; GFX7-NEXT:    s_setpc_b64 s[30:31]
-  %gep = getelementptr float, ptr %ptr, i64 511
+  %gep = getelementptr inbounds float, ptr %ptr, i64 511
   %result = atomicrmw fadd ptr %gep, float %val seq_cst, !amdgpu.no.fine.grained.memory !0, !amdgpu.ignore.denormal.mode !0
   ret float %result
 }
@@ -1630,7 +1630,7 @@ define void @flat_system_atomic_fadd_noret_f32__offset12b_pos__amdgpu_no_fine_gr
 ; GFX7-NEXT:  ; %bb.2: ; %atomicrmw.end
 ; GFX7-NEXT:    s_or_b64 exec, exec, s[4:5]
 ; GFX7-NEXT:    s_setpc_b64 s[30:31]
-  %gep = getelementptr float, ptr %ptr, i64 511
+  %gep = getelementptr inbounds float, ptr %ptr, i64 511
   %unused = atomicrmw fadd ptr %gep, float %val seq_cst, !amdgpu.no.fine.grained.memory !0, !amdgpu.ignore.denormal.mode !0
   ret void
 }
@@ -1795,7 +1795,7 @@ define void @flat_agent_atomic_fadd_noret_f32_maybe_remote(ptr %ptr, float %val)
 ; GFX7-NEXT:  ; %bb.2: ; %atomicrmw.end
 ; GFX7-NEXT:    s_or_b64 exec, exec, s[4:5]
 ; GFX7-NEXT:    s_setpc_b64 s[30:31]
-  %gep = getelementptr float, ptr %ptr, i64 511
+  %gep = getelementptr inbounds float, ptr %ptr, i64 511
   %unused = atomicrmw fadd ptr %gep, float %val syncscope("agent") seq_cst
   ret void
 }
@@ -1947,7 +1947,7 @@ define void @flat_agent_atomic_fadd_noret_f32___amdgpu_no_fine_grained_memory(pt
 ; GFX7-NEXT:  ; %bb.2: ; %atomicrmw.end
 ; GFX7-NEXT:    s_or_b64 exec, exec, s[4:5]
 ; GFX7-NEXT:    s_setpc_b64 s[30:31]
-  %gep = getelementptr float, ptr %ptr, i64 511
+  %gep = getelementptr inbounds float, ptr %ptr, i64 511
   %unused = atomicrmw fadd ptr %gep, float %val syncscope("agent") seq_cst, !amdgpu.no.fine.grained.memory !0
   ret void
 }
@@ -2165,7 +2165,7 @@ define void @flat_agent_atomic_fadd_noret_f32___amdgpu_no_fine_grained_memory__a
 ; GFX7-NEXT:  ; %bb.2: ; %atomicrmw.end
 ; GFX7-NEXT:    s_or_b64 exec, exec, s[4:5]
 ; GFX7-NEXT:    s_setpc_b64 s[30:31]
-  %gep = getelementptr float, ptr %ptr, i64 511
+  %gep = getelementptr inbounds float, ptr %ptr, i64 511
   %unused = atomicrmw fadd ptr %gep, float %val syncscope("agent") seq_cst, !amdgpu.no.fine.grained.memory !0, !amdgpu.ignore.denormal.mode !0
   ret void
 }
@@ -2330,7 +2330,7 @@ define void @flat_agent_atomic_fadd_noret_f32_amdgpu_ignore_denormal_mode(ptr %p
 ; GFX7-NEXT:  ; %bb.2: ; %atomicrmw.end
 ; GFX7-NEXT:    s_or_b64 exec, exec, s[4:5]
 ; GFX7-NEXT:    s_setpc_b64 s[30:31]
-  %gep = getelementptr float, ptr %ptr, i64 511
+  %gep = getelementptr inbounds float, ptr %ptr, i64 511
   %unused = atomicrmw fadd ptr %gep, float %val syncscope("agent") seq_cst, !amdgpu.ignore.denormal.mode !0
   ret void
 }
@@ -2698,7 +2698,7 @@ define float @flat_agent_atomic_fadd_ret_f32__offset12b_pos__ftz__amdgpu_no_fine
 ; GFX7-NEXT:  ; %bb.2: ; %atomicrmw.end
 ; GFX7-NEXT:    s_or_b64 exec, exec, s[4:5]
 ; GFX7-NEXT:    s_setpc_b64 s[30:31]
-  %gep = getelementptr float, ptr %ptr, i64 511
+  %gep = getelementptr inbounds float, ptr %ptr, i64 511
   %result = atomicrmw fadd ptr %gep, float %val syncscope("agent") seq_cst, !amdgpu.no.fine.grained.memory !0
   ret float %result
 }
@@ -2895,7 +2895,7 @@ define float @flat_agent_atomic_fadd_ret_f32__offset12b_neg__ftz__amdgpu_no_fine
 ; GFX7-NEXT:  ; %bb.2: ; %atomicrmw.end
 ; GFX7-NEXT:    s_or_b64 exec, exec, s[4:5]
 ; GFX7-NEXT:    s_setpc_b64 s[30:31]
-  %gep = getelementptr float, ptr %ptr, i64 -512
+  %gep = getelementptr inbounds float, ptr %ptr, i64 -512
   %result = atomicrmw fadd ptr %gep, float %val syncscope("agent") seq_cst, !amdgpu.no.fine.grained.memory !0
   ret float %result
 }
@@ -3320,7 +3320,7 @@ define void @flat_agent_atomic_fadd_noret_f32__offset12b_pos__ftz__amdgpu_no_fin
 ; GFX7-NEXT:  ; %bb.2: ; %atomicrmw.end
 ; GFX7-NEXT:    s_or_b64 exec, exec, s[4:5]
 ; GFX7-NEXT:    s_setpc_b64 s[30:31]
-  %gep = getelementptr float, ptr %ptr, i64 511
+  %gep = getelementptr inbounds float, ptr %ptr, i64 511
   %unused = atomicrmw fadd ptr %gep, float %val syncscope("agent") seq_cst, !amdgpu.no.fine.grained.memory !0
   ret void
 }
@@ -3544,7 +3544,7 @@ define void @flat_agent_atomic_fadd_noret_f32__offset12b_neg__ftz__amdgpu_no_fin
 ; GFX7-NEXT:  ; %bb.2: ; %atomicrmw.end
 ; GFX7-NEXT:    s_or_b64 exec, exec, s[4:5]
 ; GFX7-NEXT:    s_setpc_b64 s[30:31]
-  %gep = getelementptr float, ptr %ptr, i64 -512
+  %gep = getelementptr inbounds float, ptr %ptr, i64 -512
   %unused = atomicrmw fadd ptr %gep, float %val syncscope("agent") seq_cst, !amdgpu.no.fine.grained.memory !0
   ret void
 }
@@ -3734,7 +3734,7 @@ define float @flat_system_atomic_fadd_ret_f32__offset12b_pos__ftz__amdgpu_no_fin
 ; GFX7-NEXT:  ; %bb.2: ; %atomicrmw.end
 ; GFX7-NEXT:    s_or_b64 exec, exec, s[4:5]
 ; GFX7-NEXT:    s_setpc_b64 s[30:31]
-  %gep = getelementptr float, ptr %ptr, i64 511
+  %gep = getelementptr inbounds float, ptr %ptr, i64 511
   %result = atomicrmw fadd ptr %gep, float %val seq_cst, !amdgpu.no.fine.grained.memory !0
   ret float %result
 }
@@ -3955,7 +3955,7 @@ define void @flat_system_atomic_fadd_noret_f32__offset12b_pos__ftz__amdgpu_no_fi
 ; GFX7-NEXT:  ; %bb.2: ; %atomicrmw.end
 ; GFX7-NEXT:    s_or_b64 exec, exec, s[4:5]
 ; GFX7-NEXT:    s_setpc_b64 s[30:31]
-  %gep = getelementptr float, ptr %ptr, i64 511
+  %gep = getelementptr inbounds float, ptr %ptr, i64 511
   %unused = atomicrmw fadd ptr %gep, float %val seq_cst, !amdgpu.no.fine.grained.memory !0
   ret void
 }
@@ -4145,7 +4145,7 @@ define float @flat_agent_atomic_fadd_ret_f32__ieee__amdgpu_no_fine_grained_memor
 ; GFX7-NEXT:  ; %bb.2: ; %atomicrmw.end
 ; GFX7-NEXT:    s_or_b64 exec, exec, s[4:5]
 ; GFX7-NEXT:    s_setpc_b64 s[30:31]
-  %gep = getelementptr float, ptr %ptr, i64 511
+  %gep = getelementptr inbounds float, ptr %ptr, i64 511
   %result = atomicrmw fadd ptr %gep, float %val seq_cst, !amdgpu.no.fine.grained.memory !0, !amdgpu.ignore.denormal.mode !0
   ret float %result
 }
@@ -4366,7 +4366,7 @@ define void @flat_agent_atomic_fadd_noret_f32__ieee__amdgpu_no_fine_grained_memo
 ; GFX7-NEXT:  ; %bb.2: ; %atomicrmw.end
 ; GFX7-NEXT:    s_or_b64 exec, exec, s[4:5]
 ; GFX7-NEXT:    s_setpc_b64 s[30:31]
-  %gep = getelementptr float, ptr %ptr, i64 511
+  %gep = getelementptr inbounds float, ptr %ptr, i64 511
   %unused = atomicrmw fadd ptr %gep, float %val seq_cst, !amdgpu.no.fine.grained.memory !0, !amdgpu.ignore.denormal.mode !0
   ret void
 }
@@ -6590,7 +6590,7 @@ define double @flat_agent_atomic_fadd_ret_f64__offset12b_pos__amdgpu_no_fine_gra
 ; GFX7-NEXT:    s_or_b64 exec, exec, s[4:5]
 ; GFX7-NEXT:    s_waitcnt vmcnt(0)
 ; GFX7-NEXT:    s_setpc_b64 s[30:31]
-  %gep = getelementptr double, ptr %ptr, i64 255
+  %gep = getelementptr inbounds double, ptr %ptr, i64 255
   %result = atomicrmw fadd ptr %gep, double %val syncscope("agent") seq_cst, !amdgpu.no.fine.grained.memory !0
   ret double %result
 }
@@ -7052,7 +7052,7 @@ define double @flat_agent_atomic_fadd_ret_f64__offset12b_neg__amdgpu_no_fine_gra
 ; GFX7-NEXT:    s_or_b64 exec, exec, s[4:5]
 ; GFX7-NEXT:    s_waitcnt vmcnt(0)
 ; GFX7-NEXT:    s_setpc_b64 s[30:31]
-  %gep = getelementptr double, ptr %ptr, i64 -256
+  %gep = getelementptr inbounds double, ptr %ptr, i64 -256
   %result = atomicrmw fadd ptr %gep, double %val syncscope("agent") seq_cst, !amdgpu.no.fine.grained.memory !0
   ret double %result
 }
@@ -7931,7 +7931,7 @@ define void @flat_agent_atomic_fadd_noret_f64__offset12b_pos__amdgpu_no_fine_gra
 ; GFX7-NEXT:    s_or_b64 exec, exec, s[4:5]
 ; GFX7-NEXT:    s_waitcnt vmcnt(0)
 ; GFX7-NEXT:    s_setpc_b64 s[30:31]
-  %gep = getelementptr double, ptr %ptr, i64 255
+  %gep = getelementptr inbounds double, ptr %ptr, i64 255
   %unused = atomicrmw fadd ptr %gep, double %val syncscope("agent") seq_cst, !amdgpu.no.fine.grained.memory !0
   ret void
 }
@@ -8381,7 +8381,7 @@ define void @flat_agent_atomic_fadd_noret_f64__offset12b_neg__amdgpu_no_fine_gra
 ; GFX7-NEXT:    s_or_b64 exec, exec, s[4:5]
 ; GFX7-NEXT:    s_waitcnt vmcnt(0)
 ; GFX7-NEXT:    s_setpc_b64 s[30:31]
-  %gep = getelementptr double, ptr %ptr, i64 -256
+  %gep = getelementptr inbounds double, ptr %ptr, i64 -256
   %unused = atomicrmw fadd ptr %gep, double %val syncscope("agent") seq_cst, !amdgpu.no.fine.grained.memory !0
   ret void
 }
@@ -9141,7 +9141,7 @@ define half @flat_agent_atomic_fadd_ret_f16__offset12b_pos__amdgpu_no_fine_grain
 ; GFX7-NEXT:    v_lshrrev_b32_e32 v0, v2, v5
 ; GFX7-NEXT:    v_cvt_f32_f16_e32 v0, v0
 ; GFX7-NEXT:    s_setpc_b64 s[30:31]
-  %gep = getelementptr half, ptr %ptr, i64 1023
+  %gep = getelementptr inbounds half, ptr %ptr, i64 1023
   %result = atomicrmw fadd ptr %gep, half %val syncscope("agent") seq_cst, !amdgpu.no.fine.grained.memory !0
   ret half %result
 }
@@ -9526,7 +9526,7 @@ define half @flat_agent_atomic_fadd_ret_f16__offset12b_neg__amdgpu_no_fine_grain
 ; GFX7-NEXT:    v_lshrrev_b32_e32 v0, v2, v5
 ; GFX7-NEXT:    v_cvt_f32_f16_e32 v0, v0
 ; GFX7-NEXT:    s_setpc_b64 s[30:31]
-  %gep = getelementptr half, ptr %ptr, i64 -1024
+  %gep = getelementptr inbounds half, ptr %ptr, i64 -1024
   %result = atomicrmw fadd ptr %gep, half %val syncscope("agent") seq_cst, !amdgpu.no.fine.grained.memory !0
   ret half %result
  }
@@ -10256,7 +10256,7 @@ define void @flat_agent_atomic_fadd_noret_f16__offset12b_pos__amdgpu_no_fine_gra
 ; GFX7-NEXT:  ; %bb.2: ; %atomicrmw.end
 ; GFX7-NEXT:    s_or_b64 exec, exec, s[4:5]
 ; GFX7-NEXT:    s_setpc_b64 s[30:31]
-  %gep = getelementptr half, ptr %ptr, i64 1023
+  %gep = getelementptr inbounds half, ptr %ptr, i64 1023
   %unused = atomicrmw fadd ptr %gep, half %val syncscope("agent") seq_cst, !amdgpu.no.fine.grained.memory !0
   ret void
 }
@@ -10628,7 +10628,7 @@ define void @flat_agent_atomic_fadd_noret_f16__offset12b_neg__amdgpu_no_fine_gra
 ; GFX7-NEXT:  ; %bb.2: ; %atomicrmw.end
 ; GFX7-NEXT:    s_or_b64 exec, exec, s[4:5]
 ; GFX7-NEXT:    s_setpc_b64 s[30:31]
-  %gep = getelementptr half, ptr %ptr, i64 -1024
+  %gep = getelementptr inbounds half, ptr %ptr, i64 -1024
   %unused = atomicrmw fadd ptr %gep, half %val syncscope("agent") seq_cst, !amdgpu.no.fine.grained.memory !0
   ret void
 }
@@ -10902,7 +10902,7 @@ define void @flat_agent_atomic_fadd_noret_f16__offset12b__align4_pos__amdgpu_no_
 ; GFX7-NEXT:  ; %bb.2: ; %atomicrmw.end
 ; GFX7-NEXT:    s_or_b64 exec, exec, s[4:5]
 ; GFX7-NEXT:    s_setpc_b64 s[30:31]
-  %gep = getelementptr half, ptr %ptr, i64 1023
+  %gep = getelementptr inbounds half, ptr %ptr, i64 1023
   %unused = atomicrmw fadd ptr %gep, half %val syncscope("agent") seq_cst, align 4, !amdgpu.no.fine.grained.memory !0
   ret void
 }
@@ -11186,7 +11186,7 @@ define half @flat_agent_atomic_fadd_ret_f16__offset12b_pos__align4__amdgpu_no_fi
 ; GFX7-NEXT:    s_or_b64 exec, exec, s[4:5]
 ; GFX7-NEXT:    v_cvt_f32_f16_e32 v0, v3
 ; GFX7-NEXT:    s_setpc_b64 s[30:31]
-  %gep = getelementptr half, ptr %ptr, i64 1023
+  %gep = getelementptr inbounds half, ptr %ptr, i64 1023
   %result = atomicrmw fadd ptr %gep, half %val syncscope("agent") seq_cst, align 4, !amdgpu.no.fine.grained.memory !0
   ret half %result
 }
@@ -11574,7 +11574,7 @@ define half @flat_system_atomic_fadd_ret_f16__offset12b_pos__amdgpu_no_fine_grai
 ; GFX7-NEXT:    v_lshrrev_b32_e32 v0, v2, v5
 ; GFX7-NEXT:    v_cvt_f32_f16_e32 v0, v0
 ; GFX7-NEXT:    s_setpc_b64 s[30:31]
-  %gep = getelementptr half, ptr %ptr, i64 1023
+  %gep = getelementptr inbounds half, ptr %ptr, i64 1023
   %result = atomicrmw fadd ptr %gep, half %val seq_cst, !amdgpu.no.fine.grained.memory !0
   ret half %result
 }
@@ -11949,7 +11949,7 @@ define void @flat_system_atomic_fadd_noret_f16__offset12b_pos__amdgpu_no_fine_gr
 ; GFX7-NEXT:  ; %bb.2: ; %atomicrmw.end
 ; GFX7-NEXT:    s_or_b64 exec, exec, s[4:5]
 ; GFX7-NEXT:    s_setpc_b64 s[30:31]
-  %gep = getelementptr half, ptr %ptr, i64 1023
+  %gep = getelementptr inbounds half, ptr %ptr, i64 1023
   %unused = atomicrmw fadd ptr %gep, half %val seq_cst, !amdgpu.no.fine.grained.memory !0
   ret void
 }
@@ -12872,7 +12872,7 @@ define bfloat @flat_agent_atomic_fadd_ret_bf16__offset12b_pos__amdgpu_no_fine_gr
 ; GFX7-NEXT:    v_lshrrev_b32_e32 v0, v3, v5
 ; GFX7-NEXT:    v_lshlrev_b32_e32 v0, 16, v0
 ; GFX7-NEXT:    s_setpc_b64 s[30:31]
-  %gep = getelementptr bfloat, ptr %ptr, i64 1023
+  %gep = getelementptr inbounds bfloat, ptr %ptr, i64 1023
   %result = atomicrmw fadd ptr %gep, bfloat %val syncscope("agent") seq_cst, !amdgpu.no.fine.grained.memory !0
   ret bfloat %result
 }
@@ -13340,7 +13340,7 @@ define bfloat @flat_agent_atomic_fadd_ret_bf16__offset12b_neg__amdgpu_no_fine_gr
 ; GFX7-NEXT:    v_lshrrev_b32_e32 v0, v3, v5
 ; GFX7-NEXT:    v_lshlrev_b32_e32 v0, 16, v0
 ; GFX7-NEXT:    s_setpc_b64 s[30:31]
-  %gep = getelementptr bfloat, ptr %ptr, i64 -1024
+  %gep = getelementptr inbounds bfloat, ptr %ptr, i64 -1024
   %result = atomicrmw fadd ptr %gep, bfloat %val syncscope("agent") seq_cst, !amdgpu.no.fine.grained.memory !0
   ret bfloat %result
  }
@@ -13792,7 +13792,7 @@ define void @flat_agent_atomic_fadd_noret_bf16__offset12b_pos__amdgpu_no_fine_gr
 ; GFX7-NEXT:  ; %bb.2: ; %atomicrmw.end
 ; GFX7-NEXT:    s_or_b64 exec, exec, s[4:5]
 ; GFX7-NEXT:    s_setpc_b64 s[30:31]
-  %gep = getelementptr bfloat, ptr %ptr, i64 1023
+  %gep = getelementptr inbounds bfloat, ptr %ptr, i64 1023
   %unused = atomicrmw fadd ptr %gep, bfloat %val syncscope("agent") seq_cst, !amdgpu.no.fine.grained.memory !0
   ret void
 }
@@ -14245,7 +14245,7 @@ define void @flat_agent_atomic_fadd_noret_bf16__offset12b_neg__amdgpu_no_fine_gr
 ; GFX7-NEXT:  ; %bb.2: ; %atomicrmw.end
 ; GFX7-NEXT:    s_or_b64 exec, exec, s[4:5]
 ; GFX7-NEXT:    s_setpc_b64 s[30:31]
-  %gep = getelementptr bfloat, ptr %ptr, i64 -1024
+  %gep = getelementptr inbounds bfloat, ptr %ptr, i64 -1024
   %unused = atomicrmw fadd ptr %gep, bfloat %val syncscope("agent") seq_cst, !amdgpu.no.fine.grained.memory !0
   ret void
 }
@@ -14618,7 +14618,7 @@ define bfloat @flat_agent_atomic_fadd_ret_bf16__offset12b_pos__align4__amdgpu_no
 ; GFX7-NEXT:    s_or_b64 exec, exec, s[4:5]
 ; GFX7-NEXT:    v_lshlrev_b32_e32 v0, 16, v3
 ; GFX7-NEXT:    s_setpc_b64 s[30:31]
-  %gep = getelementptr bfloat, ptr %ptr, i64 1023
+  %gep = getelementptr inbounds bfloat, ptr %ptr, i64 1023
   %result = atomicrmw fadd ptr %gep, bfloat %val syncscope("agent") seq_cst, align 4, !amdgpu.no.fine.grained.memory !0
   ret bfloat %result
 }
@@ -14979,7 +14979,7 @@ define void @flat_agent_atomic_fadd_noret_bf16__offset12b__align4_pos__amdgpu_no
 ; GFX7-NEXT:  ; %bb.2: ; %atomicrmw.end
 ; GFX7-NEXT:    s_or_b64 exec, exec, s[4:5]
 ; GFX7-NEXT:    s_setpc_b64 s[30:31]
-  %gep = getelementptr bfloat, ptr %ptr, i64 1023
+  %gep = getelementptr inbounds bfloat, ptr %ptr, i64 1023
   %unused = atomicrmw fadd ptr %gep, bfloat %val syncscope("agent") seq_cst, align 4, !amdgpu.no.fine.grained.memory !0
   ret void
 }
@@ -15887,7 +15887,7 @@ define bfloat @flat_system_atomic_fadd_ret_bf16__offset12b_pos__amdgpu_no_fine_g
 ; GFX7-NEXT:    v_lshrrev_b32_e32 v0, v3, v5
 ; GFX7-NEXT:    v_lshlrev_b32_e32 v0, 16, v0
 ; GFX7-NEXT:    s_setpc_b64 s[30:31]
-  %gep = getelementptr bfloat, ptr %ptr, i64 1023
+  %gep = getelementptr inbounds bfloat, ptr %ptr, i64 1023
   %result = atomicrmw fadd ptr %gep, bfloat %val seq_cst, !amdgpu.no.fine.grained.memory !0
   ret bfloat %result
 }
@@ -16343,7 +16343,7 @@ define void @flat_system_atomic_fadd_noret_bf16__offset12b_pos__amdgpu_no_fine_g
 ; GFX7-NEXT:  ; %bb.2: ; %atomicrmw.end
 ; GFX7-NEXT:    s_or_b64 exec, exec, s[4:5]
 ; GFX7-NEXT:    s_setpc_b64 s[30:31]
-  %gep = getelementptr bfloat, ptr %ptr, i64 1023
+  %gep = getelementptr inbounds bfloat, ptr %ptr, i64 1023
   %unused = atomicrmw fadd ptr %gep, bfloat %val seq_cst, !amdgpu.no.fine.grained.memory !0
   ret void
 }
@@ -16726,7 +16726,7 @@ define <2 x half> @flat_agent_atomic_fadd_ret_v2f16__offset12b_pos__amdgpu_no_fi
 ; GFX7-NEXT:  ; %bb.2: ; %atomicrmw.end
 ; GFX7-NEXT:    s_or_b64 exec, exec, s[4:5]
 ; GFX7-NEXT:    s_setpc_b64 s[30:31]
-  %gep = getelementptr <2 x half>, ptr %ptr, i64 511
+  %gep = getelementptr inbounds <2 x half>, ptr %ptr, i64 511
   %result = atomicrmw fadd ptr %gep, <2 x half> %val syncscope("agent") seq_cst, !amdgpu.no.fine.grained.memory !0
   ret <2 x half> %result
 }
@@ -16930,7 +16930,7 @@ define <2 x half> @flat_agent_atomic_fadd_ret_v2f16__offset12b_neg__amdgpu_no_fi
 ; GFX7-NEXT:  ; %bb.2: ; %atomicrmw.end
 ; GFX7-NEXT:    s_or_b64 exec, exec, s[4:5]
 ; GFX7-NEXT:    s_setpc_b64 s[30:31]
-  %gep = getelementptr <2 x half>, ptr %ptr, i64 -512
+  %gep = getelementptr inbounds <2 x half>, ptr %ptr, i64 -512
   %result = atomicrmw fadd ptr %gep, <2 x half> %val syncscope("agent") seq_cst, !amdgpu.no.fine.grained.memory !0
   ret <2 x half> %result
 }
@@ -17297,7 +17297,7 @@ define void @flat_agent_atomic_fadd_noret_v2f16__offset12b_pos__amdgpu_no_fine_g
 ; GFX7-NEXT:  ; %bb.2: ; %atomicrmw.end
 ; GFX7-NEXT:    s_or_b64 exec, exec, s[4:5]
 ; GFX7-NEXT:    s_setpc_b64 s[30:31]
-  %gep = getelementptr <2 x half>, ptr %ptr, i64 511
+  %gep = getelementptr inbounds <2 x half>, ptr %ptr, i64 511
   %unused = atomicrmw fadd ptr %gep, <2 x half> %val syncscope("agent") seq_cst, !amdgpu.no.fine.grained.memory !0
   ret void
 }
@@ -17500,7 +17500,7 @@ define void @flat_agent_atomic_fadd_noret_v2f16__offset12b_neg__amdgpu_no_fine_g
 ; GFX7-NEXT:  ; %bb.2: ; %atomicrmw.end
 ; GFX7-NEXT:    s_or_b64 exec, exec, s[4:5]
 ; GFX7-NEXT:    s_setpc_b64 s[30:31]
-  %gep = getelementptr <2 x half>, ptr %ptr, i64 -512
+  %gep = getelementptr inbounds <2 x half>, ptr %ptr, i64 -512
   %unused = atomicrmw fadd ptr %gep, <2 x half> %val syncscope("agent") seq_cst, !amdgpu.no.fine.grained.memory !0
   ret void
 }
@@ -17694,7 +17694,7 @@ define <2 x half> @flat_system_atomic_fadd_ret_v2f16__offset12b_pos__amdgpu_no_f
 ; GFX7-NEXT:  ; %bb.2: ; %atomicrmw.end
 ; GFX7-NEXT:    s_or_b64 exec, exec, s[4:5]
 ; GFX7-NEXT:    s_setpc_b64 s[30:31]
-  %gep = getelementptr <2 x half>, ptr %ptr, i64 511
+  %gep = getelementptr inbounds <2 x half>, ptr %ptr, i64 511
   %result = atomicrmw fadd ptr %gep, <2 x half> %val seq_cst, !amdgpu.no.fine.grained.memory !0
   ret <2 x half> %result
 }
@@ -17884,7 +17884,7 @@ define void @flat_system_atomic_fadd_noret_v2f16__offset12b_pos__amdgpu_no_fine_
 ; GFX7-NEXT:  ; %bb.2: ; %atomicrmw.end
 ; GFX7-NEXT:    s_or_b64 exec, exec, s[4:5]
 ; GFX7-NEXT:    s_setpc_b64 s[30:31]
-  %gep = getelementptr <2 x half>, ptr %ptr, i64 511
+  %gep = getelementptr inbounds <2 x half>, ptr %ptr, i64 511
   %unused = atomicrmw fadd ptr %gep, <2 x half> %val seq_cst, !amdgpu.no.fine.grained.memory !0
   ret void
 }
@@ -19273,7 +19273,7 @@ define <2 x bfloat> @flat_agent_atomic_fadd_ret_v2bf16__offset12b_pos__amdgpu_no
 ; GFX7-NEXT:  ; %bb.2: ; %atomicrmw.end
 ; GFX7-NEXT:    s_or_b64 exec, exec, s[4:5]
 ; GFX7-NEXT:    s_setpc_b64 s[30:31]
-  %gep = getelementptr <2 x bfloat>, ptr %ptr, i64 511
+  %gep = getelementptr inbounds <2 x bfloat>, ptr %ptr, i64 511
   %result = atomicrmw fadd ptr %gep, <2 x bfloat> %val syncscope("agent") seq_cst, !amdgpu.no.fine.grained.memory !0
   ret <2 x bfloat> %result
 }
@@ -19613,7 +19613,7 @@ define <2 x bfloat> @flat_agent_atomic_fadd_ret_v2bf16__offset12b_neg__amdgpu_no
 ; GFX7-NEXT:  ; %bb.2: ; %atomicrmw.end
 ; GFX7-NEXT:    s_or_b64 exec, exec, s[4:5]
 ; GFX7-NEXT:    s_setpc_b64 s[30:31]
-  %gep = getelementptr <2 x bfloat>, ptr %ptr, i64 -512
+  %gep = getelementptr inbounds <2 x bfloat>, ptr %ptr, i64 -512
   %result = atomicrmw fadd ptr %gep, <2 x bfloat> %val syncscope("agent") seq_cst, !amdgpu.no.fine.grained.memory !0
   ret <2 x bfloat> %result
 }
@@ -20246,7 +20246,7 @@ define void @flat_agent_atomic_fadd_noret_v2bf16__offset12b_pos__amdgpu_no_fine_
 ; GFX7-NEXT:  ; %bb.2: ; %atomicrmw.end
 ; GFX7-NEXT:    s_or_b64 exec, exec, s[4:5]
 ; GFX7-NEXT:    s_setpc_b64 s[30:31]
-  %gep = getelementptr <2 x bfloat>, ptr %ptr, i64 511
+  %gep = getelementptr inbounds <2 x bfloat>, ptr %ptr, i64 511
   %unused = atomicrmw fadd ptr %gep, <2 x bfloat> %val syncscope("agent") seq_cst, !amdgpu.no.fine.grained.memory !0
   ret void
 }
@@ -20585,7 +20585,7 @@ define void @flat_agent_atomic_fadd_noret_v2bf16__offset12b_neg__amdgpu_no_fine_
 ; GFX7-NEXT:  ; %bb.2: ; %atomicrmw.end
 ; GFX7-NEXT:    s_or_b64 exec, exec, s[4:5]
 ; GFX7-NEXT:    s_setpc_b64 s[30:31]
-  %gep = getelementptr <2 x bfloat>, ptr %ptr, i64 -512
+  %gep = getelementptr inbounds <2 x bfloat>, ptr %ptr, i64 -512
   %unused = atomicrmw fadd ptr %gep, <2 x bfloat> %val syncscope("agent") seq_cst, !amdgpu.no.fine.grained.memory !0
   ret void
 }
@@ -20914,7 +20914,7 @@ define <2 x bfloat> @flat_system_atomic_fadd_ret_v2bf16__offset12b_pos__amdgpu_n
 ; GFX7-NEXT:  ; %bb.2: ; %atomicrmw.end
 ; GFX7-NEXT:    s_or_b64 exec, exec, s[4:5]
 ; GFX7-NEXT:    s_setpc_b64 s[30:31]
-  %gep = getelementptr <2 x bfloat>, ptr %ptr, i64 511
+  %gep = getelementptr inbounds <2 x bfloat>, ptr %ptr, i64 511
   %result = atomicrmw fadd ptr %gep, <2 x bfloat> %val seq_cst, !amdgpu.no.fine.grained.memory !0
   ret <2 x bfloat> %result
 }
@@ -21237,7 +21237,7 @@ define void @flat_system_atomic_fadd_noret_v2bf16__offset12b_pos__amdgpu_no_fine
 ; GFX7-NEXT:  ; %bb.2: ; %atomicrmw.end
 ; GFX7-NEXT:    s_or_b64 exec, exec, s[4:5]
 ; GFX7-NEXT:    s_setpc_b64 s[30:31]
-  %gep = getelementptr <2 x bfloat>, ptr %ptr, i64 511
+  %gep = getelementptr inbounds <2 x bfloat>, ptr %ptr, i64 511
   %unused = atomicrmw fadd ptr %gep, <2 x bfloat> %val seq_cst, !amdgpu.no.fine.grained.memory !0
   ret void
 }
diff --git a/llvm/test/CodeGen/AMDGPU/flat-atomicrmw-fmax.ll b/llvm/test/CodeGen/AMDGPU/flat-atomicrmw-fmax.ll
index 59b0537b817d2..0c592a2097896 100644
--- a/llvm/test/CodeGen/AMDGPU/flat-atomicrmw-fmax.ll
+++ b/llvm/test/CodeGen/AMDGPU/flat-atomicrmw-fmax.ll
@@ -301,7 +301,7 @@ define float @flat_agent_atomic_fmax_ret_f32__offset12b_pos__amdgpu_no_fine_grai
 ; GFX7-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX7-NEXT:    buffer_wbinvl1
 ; GFX7-NEXT:    s_setpc_b64 s[30:31]
-  %gep = getelementptr float, ptr %ptr, i64 511
+  %gep = getelementptr inbounds float, ptr %ptr, i64 511
   %result = atomicrmw fmax ptr %gep, float %val syncscope("agent") seq_cst, !amdgpu.no.fine.grained.memory !0
   ret float %result
 }
@@ -468,7 +468,7 @@ define float @flat_agent_atomic_fmax_ret_f32__offset12b_neg__amdgpu_no_fine_grai
 ; GFX7-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX7-NEXT:    buffer_wbinvl1
 ; GFX7-NEXT:    s_setpc_b64 s[30:31]
-  %gep = getelementptr float, ptr %ptr, i64 -512
+  %gep = getelementptr inbounds float, ptr %ptr, i64 -512
   %result = atomicrmw fmax ptr %gep, float %val syncscope("agent") seq_cst, !amdgpu.no.fine.grained.memory !0
   ret float %result
 }
@@ -757,7 +757,7 @@ define void @flat_agent_atomic_fmax_noret_f32__offset12b_pos__amdgpu_no_fine_gra
 ; GFX7-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX7-NEXT:    buffer_wbinvl1
 ; GFX7-NEXT:    s_setpc_b64 s[30:31]
-  %gep = getelementptr float, ptr %ptr, i64 511
+  %gep = getelementptr inbounds float, ptr %ptr, i64 511
   %unused = atomicrmw fmax ptr %gep, float %val syncscope("agent") seq_cst, !amdgpu.no.fine.grained.memory !0
   ret void
 }
@@ -924,7 +924,7 @@ define void @flat_agent_atomic_fmax_noret_f32__offset12b_neg__amdgpu_no_fine_gra
 ; GFX7-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX7-NEXT:    buffer_wbinvl1
 ; GFX7-NEXT:    s_setpc_b64 s[30:31]
-  %gep = getelementptr float, ptr %ptr, i64 -512
+  %gep = getelementptr inbounds float, ptr %ptr, i64 -512
   %unused = atomicrmw fmax ptr %gep, float %val syncscope("agent") seq_cst, !amdgpu.no.fine.grained.memory !0
   ret void
 }
@@ -1076,7 +1076,7 @@ define float @flat_system_atomic_fmax_ret_f32__offset12b_pos__amdgpu_no_fine_gra
 ; GFX7-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX7-NEXT:    buffer_wbinvl1
 ; GFX7-NEXT:    s_setpc_b64 s[30:31]
-  %gep = getelementptr float, ptr %ptr, i64 511
+  %gep = getelementptr inbounds float, ptr %ptr, i64 511
   %result = atomicrmw fmax ptr %gep, float %val seq_cst, !amdgpu.no.fine.grained.memory !0
   ret float %result
 }
@@ -1227,7 +1227,7 @@ define void @flat_system_atomic_fmax_noret_f32__offset12b_pos__amdgpu_no_fine_gr
 ; GFX7-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX7-NEXT:    buffer_wbinvl1
 ; GFX7-NEXT:    s_setpc_b64 s[30:31]
-  %gep = getelementptr float, ptr %ptr, i64 511
+  %gep = getelementptr inbounds float, ptr %ptr, i64 511
   %unused = atomicrmw fmax ptr %gep, float %val seq_cst, !amdgpu.no.fine.grained.memory !0
   ret void
 }
@@ -1859,7 +1859,7 @@ define float @flat_agent_atomic_fmax_ret_f32__offset12b_pos__ftz__amdgpu_no_fine
 ; GFX7-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX7-NEXT:    buffer_wbinvl1
 ; GFX7-NEXT:    s_setpc_b64 s[30:31]
-  %gep = getelementptr float, ptr %ptr, i64 511
+  %gep = getelementptr inbounds float, ptr %ptr, i64 511
   %result = atomicrmw fmax ptr %gep, float %val syncscope("agent") seq_cst, !amdgpu.no.fine.grained.memory !0
   ret float %result
 }
@@ -2026,7 +2026,7 @@ define float @flat_agent_atomic_fmax_ret_f32__offset12b_neg__ftz__amdgpu_no_fine
 ; GFX7-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX7-NEXT:    buffer_wbinvl1
 ; GFX7-NEXT:    s_setpc_b64 s[30:31]
-  %gep = getelementptr float, ptr %ptr, i64 -512
+  %gep = getelementptr inbounds float, ptr %ptr, i64 -512
   %result = atomicrmw fmax ptr %gep, float %val syncscope("agent") seq_cst, !amdgpu.no.fine.grained.memory !0
   ret float %result
 }
@@ -2315,7 +2315,7 @@ define void @flat_agent_atomic_fmax_noret_f32__offset12b_pos__ftz__amdgpu_no_fin
 ; GFX7-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX7-NEXT:    buffer_wbinvl1
 ; GFX7-NEXT:    s_setpc_b64 s[30:31]
-  %gep = getelementptr float, ptr %ptr, i64 511
+  %gep = getelementptr inbounds float, ptr %ptr, i64 511
   %unused = atomicrmw fmax ptr %gep, float %val syncscope("agent") seq_cst, !amdgpu.no.fine.grained.memory !0
   ret void
 }
@@ -2482,7 +2482,7 @@ define void @flat_agent_atomic_fmax_noret_f32__offset12b_neg__ftz__amdgpu_no_fin
 ; GFX7-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX7-NEXT:    buffer_wbinvl1
 ; GFX7-NEXT:    s_setpc_b64 s[30:31]
-  %gep = getelementptr float, ptr %ptr, i64 -512
+  %gep = getelementptr inbounds float, ptr %ptr, i64 -512
   %unused = atomicrmw fmax ptr %gep, float %val syncscope("agent") seq_cst, !amdgpu.no.fine.grained.memory !0
   ret void
 }
@@ -2634,7 +2634,7 @@ define float @flat_system_atomic_fmax_ret_f32__offset12b_pos__ftz__amdgpu_no_fin
 ; GFX7-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX7-NEXT:    buffer_wbinvl1
 ; GFX7-NEXT:    s_setpc_b64 s[30:31]
-  %gep = getelementptr float, ptr %ptr, i64 511
+  %gep = getelementptr inbounds float, ptr %ptr, i64 511
   %result = atomicrmw fmax ptr %gep, float %val seq_cst, !amdgpu.no.fine.grained.memory !0
   ret float %result
 }
@@ -2785,7 +2785,7 @@ define void @flat_system_atomic_fmax_noret_f32__offset12b_pos__ftz__amdgpu_no_fi
 ; GFX7-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX7-NEXT:    buffer_wbinvl1
 ; GFX7-NEXT:    s_setpc_b64 s[30:31]
-  %gep = getelementptr float, ptr %ptr, i64 511
+  %gep = getelementptr inbounds float, ptr %ptr, i64 511
   %unused = atomicrmw fmax ptr %gep, float %val seq_cst, !amdgpu.no.fine.grained.memory !0
   ret void
 }
@@ -3591,7 +3591,7 @@ define double @flat_agent_atomic_fmax_ret_f64__offset12b_pos__amdgpu_no_fine_gra
 ; GFX7-NEXT:    s_or_b64 exec, exec, s[4:5]
 ; GFX7-NEXT:    s_waitcnt vmcnt(0)
 ; GFX7-NEXT:    s_setpc_b64 s[30:31]
-  %gep = getelementptr double, ptr %ptr, i64 255
+  %gep = getelementptr inbounds double, ptr %ptr, i64 255
   %result = atomicrmw fmax ptr %gep, double %val syncscope("agent") seq_cst, !amdgpu.no.fine.grained.memory !0
   ret double %result
 }
@@ -4004,7 +4004,7 @@ define double @flat_agent_atomic_fmax_ret_f64__offset12b_neg__amdgpu_no_fine_gra
 ; GFX7-NEXT:    s_or_b64 exec, exec, s[4:5]
 ; GFX7-NEXT:    s_waitcnt vmcnt(0)
 ; GFX7-NEXT:    s_setpc_b64 s[30:31]
-  %gep = getelementptr double, ptr %ptr, i64 -256
+  %gep = getelementptr inbounds double, ptr %ptr, i64 -256
   %result = atomicrmw fmax ptr %gep, double %val syncscope("agent") seq_cst, !amdgpu.no.fine.grained.memory !0
   ret double %result
 }
@@ -4795,7 +4795,7 @@ define void @flat_agent_atomic_fmax_noret_f64__offset12b_pos__amdgpu_no_fine_gra
 ; GFX7-NEXT:    s_or_b64 exec, exec, s[4:5]
 ; GFX7-NEXT:    s_waitcnt vmcnt(0)
 ; GFX7-NEXT:    s_setpc_b64 s[30:31]
-  %gep = getelementptr double, ptr %ptr, i64 255
+  %gep = getelementptr inbounds double, ptr %ptr, i64 255
   %unused = atomicrmw fmax ptr %gep, double %val syncscope("agent") seq_cst, !amdgpu.no.fine.grained.memory !0
   ret void
 }
@@ -5201,7 +5201,7 @@ define void @flat_agent_atomic_fmax_noret_f64__offset12b_neg__amdgpu_no_fine_gra
 ; GFX7-NEXT:    s_or_b64 exec, exec, s[4:5]
 ; GFX7-NEXT:    s_waitcnt vmcnt(0)
 ; GFX7-NEXT:    s_setpc_b64 s[30:31]
-  %gep = getelementptr double, ptr %ptr, i64 -256
+  %gep = getelementptr inbounds double, ptr %ptr, i64 -256
   %unused = atomicrmw fmax ptr %gep, double %val syncscope("agent") seq_cst, !amdgpu.no.fine.grained.memory !0
   ret void
 }
@@ -6824,7 +6824,7 @@ define half @flat_agent_atomic_fmax_ret_f16__offset12b_pos__amdgpu_no_fine_grain
 ; GFX7-NEXT:    v_lshrrev_b32_e32 v0, v2, v5
 ; GFX7-NEXT:    v_cvt_f32_f16_e32 v0, v0
 ; GFX7-NEXT:    s_setpc_b64 s[30:31]
-  %gep = getelementptr half, ptr %ptr, i64 1023
+  %gep = getelementptr inbounds half, ptr %ptr, i64 1023
   %result = atomicrmw fmax ptr %gep, half %val syncscope("agent") seq_cst, !amdgpu.no.fine.grained.memory !0
   ret half %result
 }
@@ -7234,7 +7234,7 @@ define half @flat_agent_atomic_fmax_ret_f16__offset12b_neg__amdgpu_no_fine_grain
 ; GFX7-NEXT:    v_lshrrev_b32_e32 v0, v2, v5
 ; GFX7-NEXT:    v_cvt_f32_f16_e32 v0, v0
 ; GFX7-NEXT:    s_setpc_b64 s[30:31]
-  %gep = getelementptr half, ptr %ptr, i64 -1024
+  %gep = getelementptr inbounds half, ptr %ptr, i64 -1024
   %result = atomicrmw fmax ptr %gep, half %val syncscope("agent") seq_cst, !amdgpu.no.fine.grained.memory !0
   ret half %result
  }
@@ -8009,7 +8009,7 @@ define void @flat_agent_atomic_fmax_noret_f16__offset12b_pos__amdgpu_no_fine_gra
 ; GFX7-NEXT:  ; %bb.2: ; %atomicrmw.end
 ; GFX7-NEXT:    s_or_b64 exec, exec, s[4:5]
 ; GFX7-NEXT:    s_setpc_b64 s[30:31]
-  %gep = getelementptr half, ptr %ptr, i64 1023
+  %gep = getelementptr inbounds half, ptr %ptr, i64 1023
   %unused = atomicrmw fmax ptr %gep, half %val syncscope("agent") seq_cst, !amdgpu.no.fine.grained.memory !0
   ret void
 }
@@ -8406,7 +8406,7 @@ define void @flat_agent_atomic_fmax_noret_f16__offset12b_neg__amdgpu_no_fine_gra
 ; GFX7-NEXT:  ; %bb.2: ; %atomicrmw.end
 ; GFX7-NEXT:    s_or_b64 exec, exec, s[4:5]
 ; GFX7-NEXT:    s_setpc_b64 s[30:31]
-  %gep = getelementptr half, ptr %ptr, i64 -1024
+  %gep = getelementptr inbounds half, ptr %ptr, i64 -1024
   %unused = atomicrmw fmax ptr %gep, half %val syncscope("agent") seq_cst, !amdgpu.no.fine.grained.memory !0
   ret void
 }
@@ -8710,7 +8710,7 @@ define half @flat_agent_atomic_fmax_ret_f16__offset12b_pos__align4__amdgpu_no_fi
 ; GFX7-NEXT:    s_or_b64 exec, exec, s[4:5]
 ; GFX7-NEXT:    v_cvt_f32_f16_e32 v0, v3
 ; GFX7-NEXT:    s_setpc_b64 s[30:31]
-  %gep = getelementptr half, ptr %ptr, i64 1023
+  %gep = getelementptr inbounds half, ptr %ptr, i64 1023
   %result = atomicrmw fmax ptr %gep, half %val syncscope("agent") seq_cst, align 4, !amdgpu.no.fine.grained.memory !0
   ret half %result
 }
@@ -9004,7 +9004,7 @@ define void @flat_agent_atomic_fmax_noret_f16__offset12b__align4_pos__amdgpu_no_
 ; GFX7-NEXT:  ; %bb.2: ; %atomicrmw.end
 ; GFX7-NEXT:    s_or_b64 exec, exec, s[4:5]
 ; GFX7-NEXT:    s_setpc_b64 s[30:31]
-  %gep = getelementptr half, ptr %ptr, i64 1023
+  %gep = getelementptr inbounds half, ptr %ptr, i64 1023
   %unused = atomicrmw fmax ptr %gep, half %val syncscope("agent") seq_cst, align 4, !amdgpu.no.fine.grained.memory !0
   ret void
 }
@@ -9417,7 +9417,7 @@ define half @flat_system_atomic_fmax_ret_f16__offset12b_pos__amdgpu_no_fine_grai
 ; GFX7-NEXT:    v_lshrrev_b32_e32 v0, v2, v5
 ; GFX7-NEXT:    v_cvt_f32_f16_e32 v0, v0
 ; GFX7-NEXT:    s_setpc_b64 s[30:31]
-  %gep = getelementptr half, ptr %ptr, i64 1023
+  %gep = getelementptr inbounds half, ptr %ptr, i64 1023
   %result = atomicrmw fmax ptr %gep, half %val seq_cst, !amdgpu.no.fine.grained.memory !0
   ret half %result
 }
@@ -9817,7 +9817,7 @@ define void @flat_system_atomic_fmax_noret_f16__offset12b_pos__amdgpu_no_fine_gr
 ; GFX7-NEXT:  ; %bb.2: ; %atomicrmw.end
 ; GFX7-NEXT:    s_or_b64 exec, exec, s[4:5]
 ; GFX7-NEXT:    s_setpc_b64 s[30:31]
-  %gep = getelementptr half, ptr %ptr, i64 1023
+  %gep = getelementptr inbounds half, ptr %ptr, i64 1023
   %unused = atomicrmw fmax ptr %gep, half %val seq_cst, !amdgpu.no.fine.grained.memory !0
   ret void
 }
@@ -10742,7 +10742,7 @@ define bfloat @flat_agent_atomic_fmax_ret_bf16__offset12b_pos__amdgpu_no_fine_gr
 ; GFX7-NEXT:    v_lshrrev_b32_e32 v0, v3, v5
 ; GFX7-NEXT:    v_lshlrev_b32_e32 v0, 16, v0
 ; GFX7-NEXT:    s_setpc_b64 s[30:31]
-  %gep = getelementptr bfloat, ptr %ptr, i64 1023
+  %gep = getelementptr inbounds bfloat, ptr %ptr, i64 1023
   %result = atomicrmw fmax ptr %gep, bfloat %val syncscope("agent") seq_cst, !amdgpu.no.fine.grained.memory !0
   ret bfloat %result
 }
@@ -11211,7 +11211,7 @@ define bfloat @flat_agent_atomic_fmax_ret_bf16__offset12b_neg__amdgpu_no_fine_gr
 ; GFX7-NEXT:    v_lshrrev_b32_e32 v0, v3, v5
 ; GFX7-NEXT:    v_lshlrev_b32_e32 v0, 16, v0
 ; GFX7-NEXT:    s_setpc_b64 s[30:31]
-  %gep = getelementptr bfloat, ptr %ptr, i64 -1024
+  %gep = getelementptr inbounds bfloat, ptr %ptr, i64 -1024
   %result = atomicrmw fmax ptr %gep, bfloat %val syncscope("agent") seq_cst, !amdgpu.no.fine.grained.memory !0
   ret bfloat %result
  }
@@ -12102,7 +12102,7 @@ define void @flat_agent_atomic_fmax_noret_bf16__offset12b_pos__amdgpu_no_fine_gr
 ; GFX7-NEXT:  ; %bb.2: ; %atomicrmw.end
 ; GFX7-NEXT:    s_or_b64 exec, exec, s[4:5]
 ; GFX7-NEXT:    s_setpc_b64 s[30:31]
-  %gep = getelementptr bfloat, ptr %ptr, i64 1023
+  %gep = getelementptr inbounds bfloat, ptr %ptr, i64 1023
   %unused = atomicrmw fmax ptr %gep, bfloat %val syncscope("agent") seq_cst, !amdgpu.no.fine.grained.memory !0
   ret void
 }
@@ -12556,7 +12556,7 @@ define void @flat_agent_atomic_fmax_noret_bf16__offset12b_neg__amdgpu_no_fine_gr
 ; GFX7-NEXT:  ; %bb.2: ; %atomicrmw.end
 ; GFX7-NEXT:    s_or_b64 exec, exec, s[4:5]
 ; GFX7-NEXT:    s_setpc_b64 s[30:31]
-  %gep = getelementptr bfloat, ptr %ptr, i64 -1024
+  %gep = getelementptr inbounds bfloat, ptr %ptr, i64 -1024
   %unused = atomicrmw fmax ptr %gep, bfloat %val syncscope("agent") seq_cst, !amdgpu.no.fine.grained.memory !0
   ret void
 }
@@ -12930,7 +12930,7 @@ define bfloat @flat_agent_atomic_fmax_ret_bf16__offset12b_pos__align4__amdgpu_no
 ; GFX7-NEXT:    s_or_b64 exec, exec, s[4:5]
 ; GFX7-NEXT:    v_lshlrev_b32_e32 v0, 16, v3
 ; GFX7-NEXT:    s_setpc_b64 s[30:31]
-  %gep = getelementptr bfloat, ptr %ptr, i64 1023
+  %gep = getelementptr inbounds bfloat, ptr %ptr, i64 1023
   %result = atomicrmw fmax ptr %gep, bfloat %val syncscope("agent") seq_cst, align 4, !amdgpu.no.fine.grained.memory !0
   ret bfloat %result
 }
@@ -13292,7 +13292,7 @@ define void @flat_agent_atomic_fmax_noret_bf16__offset12b__align4_pos__amdgpu_no
 ; GFX7-NEXT:  ; %bb.2: ; %atomicrmw.end
 ; GFX7-NEXT:    s_or_b64 exec, exec, s[4:5]
 ; GFX7-NEXT:    s_setpc_b64 s[30:31]
-  %gep = getelementptr bfloat, ptr %ptr, i64 1023
+  %gep = getelementptr inbounds bfloat, ptr %ptr, i64 1023
   %unused = atomicrmw fmax ptr %gep, bfloat %val syncscope("agent") seq_cst, align 4, !amdgpu.no.fine.grained.memory !0
   ret void
 }
@@ -13764,7 +13764,7 @@ define bfloat @flat_system_atomic_fmax_ret_bf16__offset12b_pos__amdgpu_no_fine_g
 ; GFX7-NEXT:    v_lshrrev_b32_e32 v0, v3, v5
 ; GFX7-NEXT:    v_lshlrev_b32_e32 v0, 16, v0
 ; GFX7-NEXT:    s_setpc_b64 s[30:31]
-  %gep = getelementptr bfloat, ptr %ptr, i64 1023
+  %gep = getelementptr inbounds bfloat, ptr %ptr, i64 1023
   %result = atomicrmw fmax ptr %gep, bfloat %val seq_cst, !amdgpu.no.fine.grained.memory !0
   ret bfloat %result
 }
@@ -14221,7 +14221,7 @@ define void @flat_system_atomic_fmax_noret_bf16__offset12b_pos__amdgpu_no_fine_g
 ; GFX7-NEXT:  ; %bb.2: ; %atomicrmw.end
 ; GFX7-NEXT:    s_or_b64 exec, exec, s[4:5]
 ; GFX7-NEXT:    s_setpc_b64 s[30:31]
-  %gep = getelementptr bfloat, ptr %ptr, i64 1023
+  %gep = getelementptr inbounds bfloat, ptr %ptr, i64 1023
   %unused = atomicrmw fmax ptr %gep, bfloat %val seq_cst, !amdgpu.no.fine.grained.memory !0
   ret void
 }
@@ -14700,7 +14700,7 @@ define <2 x half> @flat_agent_atomic_fmax_ret_v2f16__offset12b_pos__amdgpu_no_fi
 ; GFX7-NEXT:  ; %bb.2: ; %atomicrmw.end
 ; GFX7-NEXT:    s_or_b64 exec, exec, s[4:5]
 ; GFX7-NEXT:    s_setpc_b64 s[30:31]
-  %gep = getelementptr <2 x half>, ptr %ptr, i64 511
+  %gep = getelementptr inbounds <2 x half>, ptr %ptr, i64 511
   %result = atomicrmw fmax ptr %gep, <2 x half> %val syncscope("agent") seq_cst, !amdgpu.no.fine.grained.memory !0
   ret <2 x half> %result
 }
@@ -14956,7 +14956,7 @@ define <2 x half> @flat_agent_atomic_fmax_ret_v2f16__offset12b_neg__amdgpu_no_fi
 ; GFX7-NEXT:  ; %bb.2: ; %atomicrmw.end
 ; GFX7-NEXT:    s_or_b64 exec, exec, s[4:5]
 ; GFX7-NEXT:    s_setpc_b64 s[30:31]
-  %gep = getelementptr <2 x half>, ptr %ptr, i64 -512
+  %gep = getelementptr inbounds <2 x half>, ptr %ptr, i64 -512
   %result = atomicrmw fmax ptr %gep, <2 x half> %val syncscope("agent") seq_cst, !amdgpu.no.fine.grained.memory !0
   ret <2 x half> %result
 }
@@ -15417,7 +15417,7 @@ define void @flat_agent_atomic_fmax_noret_v2f16__offset12b_pos__amdgpu_no_fine_g
 ; GFX7-NEXT:  ; %bb.2: ; %atomicrmw.end
 ; GFX7-NEXT:    s_or_b64 exec, exec, s[4:5]
 ; GFX7-NEXT:    s_setpc_b64 s[30:31]
-  %gep = getelementptr <2 x half>, ptr %ptr, i64 511
+  %gep = getelementptr inbounds <2 x half>, ptr %ptr, i64 511
   %unused = atomicrmw fmax ptr %gep, <2 x half> %val syncscope("agent") seq_cst, !amdgpu.no.fine.grained.memory !0
   ret void
 }
@@ -15670,7 +15670,7 @@ define void @flat_agent_atomic_fmax_noret_v2f16__offset12b_neg__amdgpu_no_fine_g
 ; GFX7-NEXT:  ; %bb.2: ; %atomicrmw.end
 ; GFX7-NEXT:    s_or_b64 exec, exec, s[4:5]
 ; GFX7-NEXT:    s_setpc_b64 s[30:31]
-  %gep = getelementptr <2 x half>, ptr %ptr, i64 -512
+  %gep = getelementptr inbounds <2 x half>, ptr %ptr, i64 -512
   %unused = atomicrmw fmax ptr %gep, <2 x half> %val syncscope("agent") seq_cst, !amdgpu.no.fine.grained.memory !0
   ret void
 }
@@ -15912,7 +15912,7 @@ define <2 x half> @flat_system_atomic_fmax_ret_v2f16__offset12b_pos__amdgpu_no_f
 ; GFX7-NEXT:  ; %bb.2: ; %atomicrmw.end
 ; GFX7-NEXT:    s_or_b64 exec, exec, s[4:5]
 ; GFX7-NEXT:    s_setpc_b64 s[30:31]
-  %gep = getelementptr <2 x half>, ptr %ptr, i64 511
+  %gep = getelementptr inbounds <2 x half>, ptr %ptr, i64 511
   %result = atomicrmw fmax ptr %gep, <2 x half> %val seq_cst, !amdgpu.no.fine.grained.memory !0
   ret <2 x half> %result
 }
@@ -16149,7 +16149,7 @@ define void @flat_system_atomic_fmax_noret_v2f16__offset12b_pos__amdgpu_no_fine_
 ; GFX7-NEXT:  ; %bb.2: ; %atomicrmw.end
 ; GFX7-NEXT:    s_or_b64 exec, exec, s[4:5]
 ; GFX7-NEXT:    s_setpc_b64 s[30:31]
-  %gep = getelementptr <2 x half>, ptr %ptr, i64 511
+  %gep = getelementptr inbounds <2 x half>, ptr %ptr, i64 511
   %unused = atomicrmw fmax ptr %gep, <2 x half> %val seq_cst, !amdgpu.no.fine.grained.memory !0
   ret void
 }
@@ -17050,7 +17050,7 @@ define <2 x bfloat> @flat_agent_atomic_fmax_ret_v2bf16__offset12b_pos__amdgpu_no
 ; GFX7-NEXT:  ; %bb.2: ; %atomicrmw.end
 ; GFX7-NEXT:    s_or_b64 exec, exec, s[4:5]
 ; GFX7-NEXT:    s_setpc_b64 s[30:31]
-  %gep = getelementptr <2 x bfloat>, ptr %ptr, i64 511
+  %gep = getelementptr inbounds <2 x bfloat>, ptr %ptr, i64 511
   %result = atomicrmw fmax ptr %gep, <2 x bfloat> %val syncscope("agent") seq_cst, !amdgpu.no.fine.grained.memory !0
   ret <2 x bfloat> %result
 }
@@ -17518,7 +17518,7 @@ define <2 x bfloat> @flat_agent_atomic_fmax_ret_v2bf16__offset12b_neg__amdgpu_no
 ; GFX7-NEXT:  ; %bb.2: ; %atomicrmw.end
 ; GFX7-NEXT:    s_or_b64 exec, exec, s[4:5]
 ; GFX7-NEXT:    s_setpc_b64 s[30:31]
-  %gep = getelementptr <2 x bfloat>, ptr %ptr, i64 -512
+  %gep = getelementptr inbounds <2 x bfloat>, ptr %ptr, i64 -512
   %result = atomicrmw fmax ptr %gep, <2 x bfloat> %val syncscope("agent") seq_cst, !amdgpu.no.fine.grained.memory !0
   ret <2 x bfloat> %result
 }
@@ -18389,7 +18389,7 @@ define void @flat_agent_atomic_fmax_noret_v2bf16__offset12b_pos__amdgpu_no_fine_
 ; GFX7-NEXT:  ; %bb.2: ; %atomicrmw.end
 ; GFX7-NEXT:    s_or_b64 exec, exec, s[4:5]
 ; GFX7-NEXT:    s_setpc_b64 s[30:31]
-  %gep = getelementptr <2 x bfloat>, ptr %ptr, i64 511
+  %gep = getelementptr inbounds <2 x bfloat>, ptr %ptr, i64 511
   %unused = atomicrmw fmax ptr %gep, <2 x bfloat> %val syncscope("agent") seq_cst, !amdgpu.no.fine.grained.memory !0
   ret void
 }
@@ -18850,7 +18850,7 @@ define void @flat_agent_atomic_fmax_noret_v2bf16__offset12b_neg__amdgpu_no_fine_
 ; GFX7-NEXT:  ; %bb.2: ; %atomicrmw.end
 ; GFX7-NEXT:    s_or_b64 exec, exec, s[4:5]
 ; GFX7-NEXT:    s_setpc_b64 s[30:31]
-  %gep = getelementptr <2 x bfloat>, ptr %ptr, i64 -512
+  %gep = getelementptr inbounds <2 x bfloat>, ptr %ptr, i64 -512
   %unused = atomicrmw fmax ptr %gep, <2 x bfloat> %val syncscope("agent") seq_cst, !amdgpu.no.fine.grained.memory !0
   ret void
 }
@@ -19304,7 +19304,7 @@ define <2 x bfloat> @flat_system_atomic_fmax_ret_v2bf16__offset12b_pos__amdgpu_n
 ; GFX7-NEXT:  ; %bb.2: ; %atomicrmw.end
 ; GFX7-NEXT:    s_or_b64 exec, exec, s[4:5]
 ; GFX7-NEXT:    s_setpc_b64 s[30:31]
-  %gep = getelementptr <2 x bfloat>, ptr %ptr, i64 511
+  %gep = getelementptr inbounds <2 x bfloat>, ptr %ptr, i64 511
   %result = atomicrmw fmax ptr %gep, <2 x bfloat> %val seq_cst, !amdgpu.no.fine.grained.memory !0
   ret <2 x bfloat> %result
 }
@@ -19747,7 +19747,7 @@ define void @flat_system_atomic_fmax_noret_v2bf16__offset12b_pos__amdgpu_no_fine
 ; GFX7-NEXT:  ; %bb.2: ; %atomicrmw.end
 ; GFX7-NEXT:    s_or_b64 exec, exec, s[4:5]
 ; GFX7-NEXT:    s_setpc_b64 s[30:31]
-  %gep = getelementptr <2 x bfloat>, ptr %ptr, i64 511
+  %gep = getelementptr inbounds <2 x bfloat>, ptr %ptr, i64 511
   %unused = atomicrmw fmax ptr %gep, <2 x bfloat> %val seq_cst, !amdgpu.no.fine.grained.memory !0
   ret void
 }
diff --git a/llvm/test/CodeGen/AMDGPU/flat-atomicrmw-fmin.ll b/llvm/test/CodeGen/AMDGPU/flat-atomicrmw-fmin.ll
index c9c9f332fe391..d08fdc9809e19 100644
--- a/llvm/test/CodeGen/AMDGPU/flat-atomicrmw-fmin.ll
+++ b/llvm/test/CodeGen/AMDGPU/flat-atomicrmw-fmin.ll
@@ -301,7 +301,7 @@ define float @flat_agent_atomic_fmin_ret_f32__offset12b_pos__amdgpu_no_fine_grai
 ; GFX7-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX7-NEXT:    buffer_wbinvl1
 ; GFX7-NEXT:    s_setpc_b64 s[30:31]
-  %gep = getelementptr float, ptr %ptr, i64 511
+  %gep = getelementptr inbounds float, ptr %ptr, i64 511
   %result = atomicrmw fmin ptr %gep, float %val syncscope("agent") seq_cst, !amdgpu.no.fine.grained.memory !0
   ret float %result
 }
@@ -468,7 +468,7 @@ define float @flat_agent_atomic_fmin_ret_f32__offset12b_neg__amdgpu_no_fine_grai
 ; GFX7-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX7-NEXT:    buffer_wbinvl1
 ; GFX7-NEXT:    s_setpc_b64 s[30:31]
-  %gep = getelementptr float, ptr %ptr, i64 -512
+  %gep = getelementptr inbounds float, ptr %ptr, i64 -512
   %result = atomicrmw fmin ptr %gep, float %val syncscope("agent") seq_cst, !amdgpu.no.fine.grained.memory !0
   ret float %result
 }
@@ -757,7 +757,7 @@ define void @flat_agent_atomic_fmin_noret_f32__offset12b_pos__amdgpu_no_fine_gra
 ; GFX7-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX7-NEXT:    buffer_wbinvl1
 ; GFX7-NEXT:    s_setpc_b64 s[30:31]
-  %gep = getelementptr float, ptr %ptr, i64 511
+  %gep = getelementptr inbounds float, ptr %ptr, i64 511
   %unused = atomicrmw fmin ptr %gep, float %val syncscope("agent") seq_cst, !amdgpu.no.fine.grained.memory !0
   ret void
 }
@@ -924,7 +924,7 @@ define void @flat_agent_atomic_fmin_noret_f32__offset12b_neg__amdgpu_no_fine_gra
 ; GFX7-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX7-NEXT:    buffer_wbinvl1
 ; GFX7-NEXT:    s_setpc_b64 s[30:31]
-  %gep = getelementptr float, ptr %ptr, i64 -512
+  %gep = getelementptr inbounds float, ptr %ptr, i64 -512
   %unused = atomicrmw fmin ptr %gep, float %val syncscope("agent") seq_cst, !amdgpu.no.fine.grained.memory !0
   ret void
 }
@@ -1076,7 +1076,7 @@ define float @flat_system_atomic_fmin_ret_f32__offset12b_pos__amdgpu_no_fine_gra
 ; GFX7-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX7-NEXT:    buffer_wbinvl1
 ; GFX7-NEXT:    s_setpc_b64 s[30:31]
-  %gep = getelementptr float, ptr %ptr, i64 511
+  %gep = getelementptr inbounds float, ptr %ptr, i64 511
   %result = atomicrmw fmin ptr %gep, float %val seq_cst, !amdgpu.no.fine.grained.memory !0
   ret float %result
 }
@@ -1227,7 +1227,7 @@ define void @flat_system_atomic_fmin_noret_f32__offset12b_pos__amdgpu_no_fine_gr
 ; GFX7-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX7-NEXT:    buffer_wbinvl1
 ; GFX7-NEXT:    s_setpc_b64 s[30:31]
-  %gep = getelementptr float, ptr %ptr, i64 511
+  %gep = getelementptr inbounds float, ptr %ptr, i64 511
   %unused = atomicrmw fmin ptr %gep, float %val seq_cst, !amdgpu.no.fine.grained.memory !0
   ret void
 }
@@ -1859,7 +1859,7 @@ define float @flat_agent_atomic_fmin_ret_f32__offset12b_pos__ftz__amdgpu_no_fine
 ; GFX7-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX7-NEXT:    buffer_wbinvl1
 ; GFX7-NEXT:    s_setpc_b64 s[30:31]
-  %gep = getelementptr float, ptr %ptr, i64 511
+  %gep = getelementptr inbounds float, ptr %ptr, i64 511
   %result = atomicrmw fmin ptr %gep, float %val syncscope("agent") seq_cst, !amdgpu.no.fine.grained.memory !0
   ret float %result
 }
@@ -2026,7 +2026,7 @@ define float @flat_agent_atomic_fmin_ret_f32__offset12b_neg__ftz__amdgpu_no_fine
 ; GFX7-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX7-NEXT:    buffer_wbinvl1
 ; GFX7-NEXT:    s_setpc_b64 s[30:31]
-  %gep = getelementptr float, ptr %ptr, i64 -512
+  %gep = getelementptr inbounds float, ptr %ptr, i64 -512
   %result = atomicrmw fmin ptr %gep, float %val syncscope("agent") seq_cst, !amdgpu.no.fine.grained.memory !0
   ret float %result
 }
@@ -2315,7 +2315,7 @@ define void @flat_agent_atomic_fmin_noret_f32__offset12b_pos__ftz__amdgpu_no_fin
 ; GFX7-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX7-NEXT:    buffer_wbinvl1
 ; GFX7-NEXT:    s_setpc_b64 s[30:31]
-  %gep = getelementptr float, ptr %ptr, i64 511
+  %gep = getelementptr inbounds float, ptr %ptr, i64 511
   %unused = atomicrmw fmin ptr %gep, float %val syncscope("agent") seq_cst, !amdgpu.no.fine.grained.memory !0
   ret void
 }
@@ -2482,7 +2482,7 @@ define void @flat_agent_atomic_fmin_noret_f32__offset12b_neg__ftz__amdgpu_no_fin
 ; GFX7-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX7-NEXT:    buffer_wbinvl1
 ; GFX7-NEXT:    s_setpc_b64 s[30:31]
-  %gep = getelementptr float, ptr %ptr, i64 -512
+  %gep = getelementptr inbounds float, ptr %ptr, i64 -512
   %unused = atomicrmw fmin ptr %gep, float %val syncscope("agent") seq_cst, !amdgpu.no.fine.grained.memory !0
   ret void
 }
@@ -2634,7 +2634,7 @@ define float @flat_system_atomic_fmin_ret_f32__offset12b_pos__ftz__amdgpu_no_fin
 ; GFX7-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX7-NEXT:    buffer_wbinvl1
 ; GFX7-NEXT:    s_setpc_b64 s[30:31]
-  %gep = getelementptr float, ptr %ptr, i64 511
+  %gep = getelementptr inbounds float, ptr %ptr, i64 511
   %result = atomicrmw fmin ptr %gep, float %val seq_cst, !amdgpu.no.fine.grained.memory !0
   ret float %result
 }
@@ -2785,7 +2785,7 @@ define void @flat_system_atomic_fmin_noret_f32__offset12b_pos__ftz__amdgpu_no_fi
 ; GFX7-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX7-NEXT:    buffer_wbinvl1
 ; GFX7-NEXT:    s_setpc_b64 s[30:31]
-  %gep = getelementptr float, ptr %ptr, i64 511
+  %gep = getelementptr inbounds float, ptr %ptr, i64 511
   %unused = atomicrmw fmin ptr %gep, float %val seq_cst, !amdgpu.no.fine.grained.memory !0
   ret void
 }
@@ -3591,7 +3591,7 @@ define double @flat_agent_atomic_fmin_ret_f64__offset12b_pos__amdgpu_no_fine_gra
 ; GFX7-NEXT:    s_or_b64 exec, exec, s[4:5]
 ; GFX7-NEXT:    s_waitcnt vmcnt(0)
 ; GFX7-NEXT:    s_setpc_b64 s[30:31]
-  %gep = getelementptr double, ptr %ptr, i64 255
+  %gep = getelementptr inbounds double, ptr %ptr, i64 255
   %result = atomicrmw fmin ptr %gep, double %val syncscope("agent") seq_cst, !amdgpu.no.fine.grained.memory !0
   ret double %result
 }
@@ -4004,7 +4004,7 @@ define double @flat_agent_atomic_fmin_ret_f64__offset12b_neg__amdgpu_no_fine_gra
 ; GFX7-NEXT:    s_or_b64 exec, exec, s[4:5]
 ; GFX7-NEXT:    s_waitcnt vmcnt(0)
 ; GFX7-NEXT:    s_setpc_b64 s[30:31]
-  %gep = getelementptr double, ptr %ptr, i64 -256
+  %gep = getelementptr inbounds double, ptr %ptr, i64 -256
   %result = atomicrmw fmin ptr %gep, double %val syncscope("agent") seq_cst, !amdgpu.no.fine.grained.memory !0
   ret double %result
 }
@@ -4795,7 +4795,7 @@ define void @flat_agent_atomic_fmin_noret_f64__offset12b_pos__amdgpu_no_fine_gra
 ; GFX7-NEXT:    s_or_b64 exec, exec, s[4:5]
 ; GFX7-NEXT:    s_waitcnt vmcnt(0)
 ; GFX7-NEXT:    s_setpc_b64 s[30:31]
-  %gep = getelementptr double, ptr %ptr, i64 255
+  %gep = getelementptr inbounds double, ptr %ptr, i64 255
   %unused = atomicrmw fmin ptr %gep, double %val syncscope("agent") seq_cst, !amdgpu.no.fine.grained.memory !0
   ret void
 }
@@ -5201,7 +5201,7 @@ define void @flat_agent_atomic_fmin_noret_f64__offset12b_neg__amdgpu_no_fine_gra
 ; GFX7-NEXT:    s_or_b64 exec, exec, s[4:5]
 ; GFX7-NEXT:    s_waitcnt vmcnt(0)
 ; GFX7-NEXT:    s_setpc_b64 s[30:31]
-  %gep = getelementptr double, ptr %ptr, i64 -256
+  %gep = getelementptr inbounds double, ptr %ptr, i64 -256
   %unused = atomicrmw fmin ptr %gep, double %val syncscope("agent") seq_cst, !amdgpu.no.fine.grained.memory !0
   ret void
 }
@@ -6824,7 +6824,7 @@ define half @flat_agent_atomic_fmin_ret_f16__offset12b_pos__amdgpu_no_fine_grain
 ; GFX7-NEXT:    v_lshrrev_b32_e32 v0, v2, v5
 ; GFX7-NEXT:    v_cvt_f32_f16_e32 v0, v0
 ; GFX7-NEXT:    s_setpc_b64 s[30:31]
-  %gep = getelementptr half, ptr %ptr, i64 1023
+  %gep = getelementptr inbounds half, ptr %ptr, i64 1023
   %result = atomicrmw fmin ptr %gep, half %val syncscope("agent") seq_cst, !amdgpu.no.fine.grained.memory !0
   ret half %result
 }
@@ -7234,7 +7234,7 @@ define half @flat_agent_atomic_fmin_ret_f16__offset12b_neg__amdgpu_no_fine_grain
 ; GFX7-NEXT:    v_lshrrev_b32_e32 v0, v2, v5
 ; GFX7-NEXT:    v_cvt_f32_f16_e32 v0, v0
 ; GFX7-NEXT:    s_setpc_b64 s[30:31]
-  %gep = getelementptr half, ptr %ptr, i64 -1024
+  %gep = getelementptr inbounds half, ptr %ptr, i64 -1024
   %result = atomicrmw fmin ptr %gep, half %val syncscope("agent") seq_cst, !amdgpu.no.fine.grained.memory !0
   ret half %result
  }
@@ -8009,7 +8009,7 @@ define void @flat_agent_atomic_fmin_noret_f16__offset12b_pos__amdgpu_no_fine_gra
 ; GFX7-NEXT:  ; %bb.2: ; %atomicrmw.end
 ; GFX7-NEXT:    s_or_b64 exec, exec, s[4:5]
 ; GFX7-NEXT:    s_setpc_b64 s[30:31]
-  %gep = getelementptr half, ptr %ptr, i64 1023
+  %gep = getelementptr inbounds half, ptr %ptr, i64 1023
   %unused = atomicrmw fmin ptr %gep, half %val syncscope("agent") seq_cst, !amdgpu.no.fine.grained.memory !0
   ret void
 }
@@ -8406,7 +8406,7 @@ define void @flat_agent_atomic_fmin_noret_f16__offset12b_neg__amdgpu_no_fine_gra
 ; GFX7-NEXT:  ; %bb.2: ; %atomicrmw.end
 ; GFX7-NEXT:    s_or_b64 exec, exec, s[4:5]
 ; GFX7-NEXT:    s_setpc_b64 s[30:31]
-  %gep = getelementptr half, ptr %ptr, i64 -1024
+  %gep = getelementptr inbounds half, ptr %ptr, i64 -1024
   %unused = atomicrmw fmin ptr %gep, half %val syncscope("agent") seq_cst, !amdgpu.no.fine.grained.memory !0
   ret void
 }
@@ -8710,7 +8710,7 @@ define half @flat_agent_atomic_fmin_ret_f16__offset12b_pos__align4__amdgpu_no_fi
 ; GFX7-NEXT:    s_or_b64 exec, exec, s[4:5]
 ; GFX7-NEXT:    v_cvt_f32_f16_e32 v0, v3
 ; GFX7-NEXT:    s_setpc_b64 s[30:31]
-  %gep = getelementptr half, ptr %ptr, i64 1023
+  %gep = getelementptr inbounds half, ptr %ptr, i64 1023
   %result = atomicrmw fmin ptr %gep, half %val syncscope("agent") seq_cst, align 4, !amdgpu.no.fine.grained.memory !0
   ret half %result
 }
@@ -9004,7 +9004,7 @@ define void @flat_agent_atomic_fmin_noret_f16__offset12b__align4_pos__amdgpu_no_
 ; GFX7-NEXT:  ; %bb.2: ; %atomicrmw.end
 ; GFX7-NEXT:    s_or_b64 exec, exec, s[4:5]
 ; GFX7-NEXT:    s_setpc_b64 s[30:31]
-  %gep = getelementptr half, ptr %ptr, i64 1023
+  %gep = getelementptr inbounds half, ptr %ptr, i64 1023
   %unused = atomicrmw fmin ptr %gep, half %val syncscope("agent") seq_cst, align 4, !amdgpu.no.fine.grained.memory !0
   ret void
 }
@@ -9417,7 +9417,7 @@ define half @flat_system_atomic_fmin_ret_f16__offset12b_pos__amdgpu_no_fine_grai
 ; GFX7-NEXT:    v_lshrrev_b32_e32 v0, v2, v5
 ; GFX7-NEXT:    v_cvt_f32_f16_e32 v0, v0
 ; GFX7-NEXT:    s_setpc_b64 s[30:31]
-  %gep = getelementptr half, ptr %ptr, i64 1023
+  %gep = getelementptr inbounds half, ptr %ptr, i64 1023
   %result = atomicrmw fmin ptr %gep, half %val seq_cst, !amdgpu.no.fine.grained.memory !0
   ret half %result
 }
@@ -9817,7 +9817,7 @@ define void @flat_system_atomic_fmin_noret_f16__offset12b_pos__amdgpu_no_fine_gr
 ; GFX7-NEXT:  ; %bb.2: ; %atomicrmw.end
 ; GFX7-NEXT:    s_or_b64 exec, exec, s[4:5]
 ; GFX7-NEXT:    s_setpc_b64 s[30:31]
-  %gep = getelementptr half, ptr %ptr, i64 1023
+  %gep = getelementptr inbounds half, ptr %ptr, i64 1023
   %unused = atomicrmw fmin ptr %gep, half %val seq_cst, !amdgpu.no.fine.grained.memory !0
   ret void
 }
@@ -10742,7 +10742,7 @@ define bfloat @flat_agent_atomic_fmin_ret_bf16__offset12b_pos__amdgpu_no_fine_gr
 ; GFX7-NEXT:    v_lshrrev_b32_e32 v0, v3, v5
 ; GFX7-NEXT:    v_lshlrev_b32_e32 v0, 16, v0
 ; GFX7-NEXT:    s_setpc_b64 s[30:31]
-  %gep = getelementptr bfloat, ptr %ptr, i64 1023
+  %gep = getelementptr inbounds bfloat, ptr %ptr, i64 1023
   %result = atomicrmw fmin ptr %gep, bfloat %val syncscope("agent") seq_cst, !amdgpu.no.fine.grained.memory !0
   ret bfloat %result
 }
@@ -11211,7 +11211,7 @@ define bfloat @flat_agent_atomic_fmin_ret_bf16__offset12b_neg__amdgpu_no_fine_gr
 ; GFX7-NEXT:    v_lshrrev_b32_e32 v0, v3, v5
 ; GFX7-NEXT:    v_lshlrev_b32_e32 v0, 16, v0
 ; GFX7-NEXT:    s_setpc_b64 s[30:31]
-  %gep = getelementptr bfloat, ptr %ptr, i64 -1024
+  %gep = getelementptr inbounds bfloat, ptr %ptr, i64 -1024
   %result = atomicrmw fmin ptr %gep, bfloat %val syncscope("agent") seq_cst, !amdgpu.no.fine.grained.memory !0
   ret bfloat %result
  }
@@ -12102,7 +12102,7 @@ define void @flat_agent_atomic_fmin_noret_bf16__offset12b_pos__amdgpu_no_fine_gr
 ; GFX7-NEXT:  ; %bb.2: ; %atomicrmw.end
 ; GFX7-NEXT:    s_or_b64 exec, exec, s[4:5]
 ; GFX7-NEXT:    s_setpc_b64 s[30:31]
-  %gep = getelementptr bfloat, ptr %ptr, i64 1023
+  %gep = getelementptr inbounds bfloat, ptr %ptr, i64 1023
   %unused = atomicrmw fmin ptr %gep, bfloat %val syncscope("agent") seq_cst, !amdgpu.no.fine.grained.memory !0
   ret void
 }
@@ -12556,7 +12556,7 @@ define void @flat_agent_atomic_fmin_noret_bf16__offset12b_neg__amdgpu_no_fine_gr
 ; GFX7-NEXT:  ; %bb.2: ; %atomicrmw.end
 ; GFX7-NEXT:    s_or_b64 exec, exec, s[4:5]
 ; GFX7-NEXT:    s_setpc_b64 s[30:31]
-  %gep = getelementptr bfloat, ptr %ptr, i64 -1024
+  %gep = getelementptr inbounds bfloat, ptr %ptr, i64 -1024
   %unused = atomicrmw fmin ptr %gep, bfloat %val syncscope("agent") seq_cst, !amdgpu.no.fine.grained.memory !0
   ret void
 }
@@ -12930,7 +12930,7 @@ define bfloat @flat_agent_atomic_fmin_ret_bf16__offset12b_pos__align4__amdgpu_no
 ; GFX7-NEXT:    s_or_b64 exec, exec, s[4:5]
 ; GFX7-NEXT:    v_lshlrev_b32_e32 v0, 16, v3
 ; GFX7-NEXT:    s_setpc_b64 s[30:31]
-  %gep = getelementptr bfloat, ptr %ptr, i64 1023
+  %gep = getelementptr inbounds bfloat, ptr %ptr, i64 1023
   %result = atomicrmw fmin ptr %gep, bfloat %val syncscope("agent") seq_cst, align 4, !amdgpu.no.fine.grained.memory !0
   ret bfloat %result
 }
@@ -13292,7 +13292,7 @@ define void @flat_agent_atomic_fmin_noret_bf16__offset12b__align4_pos__amdgpu_no
 ; GFX7-NEXT:  ; %bb.2: ; %atomicrmw.end
 ; GFX7-NEXT:    s_or_b64 exec, exec, s[4:5]
 ; GFX7-NEXT:    s_setpc_b64 s[30:31]
-  %gep = getelementptr bfloat, ptr %ptr, i64 1023
+  %gep = getelementptr inbounds bfloat, ptr %ptr, i64 1023
   %unused = atomicrmw fmin ptr %gep, bfloat %val syncscope("agent") seq_cst, align 4, !amdgpu.no.fine.grained.memory !0
   ret void
 }
@@ -13764,7 +13764,7 @@ define bfloat @flat_system_atomic_fmin_ret_bf16__offset12b_pos__amdgpu_no_fine_g
 ; GFX7-NEXT:    v_lshrrev_b32_e32 v0, v3, v5
 ; GFX7-NEXT:    v_lshlrev_b32_e32 v0, 16, v0
 ; GFX7-NEXT:    s_setpc_b64 s[30:31]
-  %gep = getelementptr bfloat, ptr %ptr, i64 1023
+  %gep = getelementptr inbounds bfloat, ptr %ptr, i64 1023
   %result = atomicrmw fmin ptr %gep, bfloat %val seq_cst, !amdgpu.no.fine.grained.memory !0
   ret bfloat %result
 }
@@ -14221,7 +14221,7 @@ define void @flat_system_atomic_fmin_noret_bf16__offset12b_pos__amdgpu_no_fine_g
 ; GFX7-NEXT:  ; %bb.2: ; %atomicrmw.end
 ; GFX7-NEXT:    s_or_b64 exec, exec, s[4:5]
 ; GFX7-NEXT:    s_setpc_b64 s[30:31]
-  %gep = getelementptr bfloat, ptr %ptr, i64 1023
+  %gep = getelementptr inbounds bfloat, ptr %ptr, i64 1023
   %unused = atomicrmw fmin ptr %gep, bfloat %val seq_cst, !amdgpu.no.fine.grained.memory !0
   ret void
 }
@@ -14700,7 +14700,7 @@ define <2 x half> @flat_agent_atomic_fmin_ret_v2f16__offset12b_pos__amdgpu_no_fi
 ; GFX7-NEXT:  ; %bb.2: ; %atomicrmw.end
 ; GFX7-NEXT:    s_or_b64 exec, exec, s[4:5]
 ; GFX7-NEXT:    s_setpc_b64 s[30:31]
-  %gep = getelementptr <2 x half>, ptr %ptr, i64 511
+  %gep = getelementptr inbounds <2 x half>, ptr %ptr, i64 511
   %result = atomicrmw fmin ptr %gep, <2 x half> %val syncscope("agent") seq_cst, !amdgpu.no.fine.grained.memory !0
   ret <2 x half> %result
 }
@@ -14956,7 +14956,7 @@ define <2 x half> @flat_agent_atomic_fmin_ret_v2f16__offset12b_neg__amdgpu_no_fi
 ; GFX7-NEXT:  ; %bb.2: ; %atomicrmw.end
 ; GFX7-NEXT:    s_or_b64 exec, exec, s[4:5]
 ; GFX7-NEXT:    s_setpc_b64 s[30:31]
-  %gep = getelementptr <2 x half>, ptr %ptr, i64 -512
+  %gep = getelementptr inbounds <2 x half>, ptr %ptr, i64 -512
   %result = atomicrmw fmin ptr %gep, <2 x half> %val syncscope("agent") seq_cst, !amdgpu.no.fine.grained.memory !0
   ret <2 x half> %result
 }
@@ -15417,7 +15417,7 @@ define void @flat_agent_atomic_fmin_noret_v2f16__offset12b_pos__amdgpu_no_fine_g
 ; GFX7-NEXT:  ; %bb.2: ; %atomicrmw.end
 ; GFX7-NEXT:    s_or_b64 exec, exec, s[4:5]
 ; GFX7-NEXT:    s_setpc_b64 s[30:31]
-  %gep = getelementptr <2 x half>, ptr %ptr, i64 511
+  %gep = getelementptr inbounds <2 x half>, ptr %ptr, i64 511
   %unused = atomicrmw fmin ptr %gep, <2 x half> %val syncscope("agent") seq_cst, !amdgpu.no.fine.grained.memory !0
   ret void
 }
@@ -15670,7 +15670,7 @@ define void @flat_agent_atomic_fmin_noret_v2f16__offset12b_neg__amdgpu_no_fine_g
 ; GFX7-NEXT:  ; %bb.2: ; %atomicrmw.end
 ; GFX7-NEXT:    s_or_b64 exec, exec, s[4:5]
 ; GFX7-NEXT:    s_setpc_b64 s[30:31]
-  %gep = getelementptr <2 x half>, ptr %ptr, i64 -512
+  %gep = getelementptr inbounds <2 x half>, ptr %ptr, i64 -512
   %unused = atomicrmw fmin ptr %gep, <2 x half> %val syncscope("agent") seq_cst, !amdgpu.no.fine.grained.memory !0
   ret void
 }
@@ -15912,7 +15912,7 @@ define <2 x half> @flat_system_atomic_fmin_ret_v2f16__offset12b_pos__amdgpu_no_f
 ; GFX7-NEXT:  ; %bb.2: ; %atomicrmw.end
 ; GFX7-NEXT:    s_or_b64 exec, exec, s[4:5]
 ; GFX7-NEXT:    s_setpc_b64 s[30:31]
-  %gep = getelementptr <2 x half>, ptr %ptr, i64 511
+  %gep = getelementptr inbounds <2 x half>, ptr %ptr, i64 511
   %result = atomicrmw fmin ptr %gep, <2 x half> %val seq_cst, !amdgpu.no.fine.grained.memory !0
   ret <2 x half> %result
 }
@@ -16149,7 +16149,7 @@ define void @flat_system_atomic_fmin_noret_v2f16__offset12b_pos__amdgpu_no_fine_
 ; GFX7-NEXT:  ; %bb.2: ; %atomicrmw.end
 ; GFX7-NEXT:    s_or_b64 exec, exec, s[4:5]
 ; GFX7-NEXT:    s_setpc_b64 s[30:31]
-  %gep = getelementptr <2 x half>, ptr %ptr, i64 511
+  %gep = getelementptr inbounds <2 x half>, ptr %ptr, i64 511
   %unused = atomicrmw fmin ptr %gep, <2 x half> %val seq_cst, !amdgpu.no.fine.grained.memory !0
   ret void
 }
@@ -17050,7 +17050,7 @@ define <2 x bfloat> @flat_agent_atomic_fmin_ret_v2bf16__offset12b_pos__amdgpu_no
 ; GFX7-NEXT:  ; %bb.2: ; %atomicrmw.end
 ; GFX7-NEXT:    s_or_b64 exec, exec, s[4:5]
 ; GFX7-NEXT:    s_setpc_b64 s[30:31]
-  %gep = getelementptr <2 x bfloat>, ptr %ptr, i64 511
+  %gep = getelementptr inbounds <2 x bfloat>, ptr %ptr, i64 511
   %result = atomicrmw fmin ptr %gep, <2 x bfloat> %val syncscope("agent") seq_cst, !amdgpu.no.fine.grained.memory !0
   ret <2 x bfloat> %result
 }
@@ -17518,7 +17518,7 @@ define <2 x bfloat> @flat_agent_atomic_fmin_ret_v2bf16__offset12b_neg__amdgpu_no
 ; GFX7-NEXT:  ; %bb.2: ; %atomicrmw.end
 ; GFX7-NEXT:    s_or_b64 exec, exec, s[4:5]
 ; GFX7-NEXT:    s_setpc_b64 s[30:31]
-  %gep = getelementptr <2 x bfloat>, ptr %ptr, i64 -512
+  %gep = getelementptr inbounds <2 x bfloat>, ptr %ptr, i64 -512
   %result = atomicrmw fmin ptr %gep, <2 x bfloat> %val syncscope("agent") seq_cst, !amdgpu.no.fine.grained.memory !0
   ret <2 x bfloat> %result
 }
@@ -18389,7 +18389,7 @@ define void @flat_agent_atomic_fmin_noret_v2bf16__offset12b_pos__amdgpu_no_fine_
 ; GFX7-NEXT:  ; %bb.2: ; %atomicrmw.end
 ; GFX7-NEXT:    s_or_b64 exec, exec, s[4:5]
 ; GFX7-NEXT:    s_setpc_b64 s[30:31]
-  %gep = getelementptr <2 x bfloat>, ptr %ptr, i64 511
+  %gep = getelementptr inbounds <2 x bfloat>, ptr %ptr, i64 511
   %unused = atomicrmw fmin ptr %gep, <2 x bfloat> %val syncscope("agent") seq_cst, !amdgpu.no.fine.grained.memory !0
   ret void
 }
@@ -18850,7 +18850,7 @@ define void @flat_agent_atomic_fmin_noret_v2bf16__offset12b_neg__amdgpu_no_fine_
 ; GFX7-NEXT:  ; %bb.2: ; %atomicrmw.end
 ; GFX7-NEXT:    s_or_b64 exec, exec, s[4:5]
 ; GFX7-NEXT:    s_setpc_b64 s[30:31]
-  %gep = getelementptr <2 x bfloat>, ptr %ptr, i64 -512
+  %gep = getelementptr inbounds <2 x bfloat>, ptr %ptr, i64 -512
   %unused = atomicrmw fmin ptr %gep, <2 x bfloat> %val syncscope("agent") seq_cst, !amdgpu.no.fine.grained.memory !0
   ret void
 }
@@ -19304,7 +19304,7 @@ define <2 x bfloat> @flat_system_atomic_fmin_ret_v2bf16__offset12b_pos__amdgpu_n
 ; GFX7-NEXT:  ; %bb.2: ; %atomicrmw.end
 ; GFX7-NEXT:    s_or_b64 exec, exec, s[4:5]
 ; GFX7-NEXT:    s_setpc_b64 s[30:31]
-  %gep = getelementptr <2 x bfloat>, ptr %ptr, i64 511
+  %gep = getelementptr inbounds <2 x bfloat>, ptr %ptr, i64 511
   %result = atomicrmw fmin ptr %gep, <2 x bfloat> %val seq_cst, !amdgpu.no.fine.grained.memory !0
   ret <2 x bfloat> %result
 }
@@ -19747,7 +19747,7 @@ define void @flat_system_atomic_fmin_noret_v2bf16__offset12b_pos__amdgpu_no_fine
 ; GFX7-NEXT:  ; %bb.2: ; %atomicrmw.end
 ; GFX7-NEXT:    s_or_b64 exec, exec, s[4:5]
 ; GFX7-NEXT:    s_setpc_b64 s[30:31]
-  %gep = getelementptr <2 x bfloat>, ptr %ptr, i64 511
+  %gep = getelementptr inbounds <2 x bfloat>, ptr %ptr, i64 511
   %unused = atomicrmw fmin ptr %gep, <2 x bfloat> %val seq_cst, !amdgpu.no.fine.grained.memory !0
   ret void
 }
diff --git a/llvm/test/CodeGen/AMDGPU/flat-atomicrmw-fsub.ll b/llvm/test/CodeGen/AMDGPU/flat-atomicrmw-fsub.ll
index 587c2ea885077..e5c967666c9d7 100644
--- a/llvm/test/CodeGen/AMDGPU/flat-atomicrmw-fsub.ll
+++ b/llvm/test/CodeGen/AMDGPU/flat-atomicrmw-fsub.ll
@@ -405,7 +405,7 @@ define float @flat_agent_atomic_fsub_ret_f32__offset12b_pos(ptr %ptr, float %val
 ; GFX7-NEXT:  ; %bb.2: ; %atomicrmw.end
 ; GFX7-NEXT:    s_or_b64 exec, exec, s[4:5]
 ; GFX7-NEXT:    s_setpc_b64 s[30:31]
-  %gep = getelementptr float, ptr %ptr, i64 511
+  %gep = getelementptr inbounds float, ptr %ptr, i64 511
   %result = atomicrmw fsub ptr %gep, float %val syncscope("agent") seq_cst
   ret float %result
 }
@@ -622,7 +622,7 @@ define float @flat_agent_atomic_fsub_ret_f32__offset12b_neg(ptr %ptr, float %val
 ; GFX7-NEXT:  ; %bb.2: ; %atomicrmw.end
 ; GFX7-NEXT:    s_or_b64 exec, exec, s[4:5]
 ; GFX7-NEXT:    s_setpc_b64 s[30:31]
-  %gep = getelementptr float, ptr %ptr, i64 -512
+  %gep = getelementptr inbounds float, ptr %ptr, i64 -512
   %result = atomicrmw fsub ptr %gep, float %val syncscope("agent") seq_cst
   ret float %result
 }
@@ -1001,7 +1001,7 @@ define void @flat_agent_atomic_fsub_noret_f32__offset12b_pos(ptr %ptr, float %va
 ; GFX7-NEXT:  ; %bb.2: ; %atomicrmw.end
 ; GFX7-NEXT:    s_or_b64 exec, exec, s[4:5]
 ; GFX7-NEXT:    s_setpc_b64 s[30:31]
-  %gep = getelementptr float, ptr %ptr, i64 511
+  %gep = getelementptr inbounds float, ptr %ptr, i64 511
   %unused = atomicrmw fsub ptr %gep, float %val syncscope("agent") seq_cst
   ret void
 }
@@ -1213,7 +1213,7 @@ define void @flat_agent_atomic_fsub_noret_f32__offset12b_neg(ptr %ptr, float %va
 ; GFX7-NEXT:  ; %bb.2: ; %atomicrmw.end
 ; GFX7-NEXT:    s_or_b64 exec, exec, s[4:5]
 ; GFX7-NEXT:    s_setpc_b64 s[30:31]
-  %gep = getelementptr float, ptr %ptr, i64 -512
+  %gep = getelementptr inbounds float, ptr %ptr, i64 -512
   %unused = atomicrmw fsub ptr %gep, float %val syncscope("agent") seq_cst
   ret void
 }
@@ -1416,7 +1416,7 @@ define float @flat_system_atomic_fsub_ret_f32__offset12b_pos(ptr %ptr, float %va
 ; GFX7-NEXT:  ; %bb.2: ; %atomicrmw.end
 ; GFX7-NEXT:    s_or_b64 exec, exec, s[4:5]
 ; GFX7-NEXT:    s_setpc_b64 s[30:31]
-  %gep = getelementptr float, ptr %ptr, i64 511
+  %gep = getelementptr inbounds float, ptr %ptr, i64 511
   %result = atomicrmw fsub ptr %gep, float %val seq_cst
   ret float %result
 }
@@ -1612,7 +1612,7 @@ define void @flat_system_atomic_fsub_noret_f32__offset12b_pos(ptr %ptr, float %v
 ; GFX7-NEXT:  ; %bb.2: ; %atomicrmw.end
 ; GFX7-NEXT:    s_or_b64 exec, exec, s[4:5]
 ; GFX7-NEXT:    s_setpc_b64 s[30:31]
-  %gep = getelementptr float, ptr %ptr, i64 511
+  %gep = getelementptr inbounds float, ptr %ptr, i64 511
   %unused = atomicrmw fsub ptr %gep, float %val seq_cst
   ret void
 }
@@ -2012,7 +2012,7 @@ define float @flat_agent_atomic_fsub_ret_f32__offset12b_pos__ftz(ptr %ptr, float
 ; GFX7-NEXT:  ; %bb.2: ; %atomicrmw.end
 ; GFX7-NEXT:    s_or_b64 exec, exec, s[4:5]
 ; GFX7-NEXT:    s_setpc_b64 s[30:31]
-  %gep = getelementptr float, ptr %ptr, i64 511
+  %gep = getelementptr inbounds float, ptr %ptr, i64 511
   %result = atomicrmw fsub ptr %gep, float %val syncscope("agent") seq_cst
   ret float %result
 }
@@ -2229,7 +2229,7 @@ define float @flat_agent_atomic_fsub_ret_f32__offset12b_neg__ftz(ptr %ptr, float
 ; GFX7-NEXT:  ; %bb.2: ; %atomicrmw.end
 ; GFX7-NEXT:    s_or_b64 exec, exec, s[4:5]
 ; GFX7-NEXT:    s_setpc_b64 s[30:31]
-  %gep = getelementptr float, ptr %ptr, i64 -512
+  %gep = getelementptr inbounds float, ptr %ptr, i64 -512
   %result = atomicrmw fsub ptr %gep, float %val syncscope("agent") seq_cst
   ret float %result
 }
@@ -2608,7 +2608,7 @@ define void @flat_agent_atomic_fsub_noret_f32__offset12b_pos__ftz(ptr %ptr, floa
 ; GFX7-NEXT:  ; %bb.2: ; %atomicrmw.end
 ; GFX7-NEXT:    s_or_b64 exec, exec, s[4:5]
 ; GFX7-NEXT:    s_setpc_b64 s[30:31]
-  %gep = getelementptr float, ptr %ptr, i64 511
+  %gep = getelementptr inbounds float, ptr %ptr, i64 511
   %unused = atomicrmw fsub ptr %gep, float %val syncscope("agent") seq_cst
   ret void
 }
@@ -2820,7 +2820,7 @@ define void @flat_agent_atomic_fsub_noret_f32__offset12b_neg__ftz(ptr %ptr, floa
 ; GFX7-NEXT:  ; %bb.2: ; %atomicrmw.end
 ; GFX7-NEXT:    s_or_b64 exec, exec, s[4:5]
 ; GFX7-NEXT:    s_setpc_b64 s[30:31]
-  %gep = getelementptr float, ptr %ptr, i64 -512
+  %gep = getelementptr inbounds float, ptr %ptr, i64 -512
   %unused = atomicrmw fsub ptr %gep, float %val syncscope("agent") seq_cst
   ret void
 }
@@ -3023,7 +3023,7 @@ define float @flat_system_atomic_fsub_ret_f32__offset12b_pos__ftz(ptr %ptr, floa
 ; GFX7-NEXT:  ; %bb.2: ; %atomicrmw.end
 ; GFX7-NEXT:    s_or_b64 exec, exec, s[4:5]
 ; GFX7-NEXT:    s_setpc_b64 s[30:31]
-  %gep = getelementptr float, ptr %ptr, i64 511
+  %gep = getelementptr inbounds float, ptr %ptr, i64 511
   %result = atomicrmw fsub ptr %gep, float %val seq_cst
   ret float %result
 }
@@ -3219,7 +3219,7 @@ define void @flat_system_atomic_fsub_noret_f32__offset12b_pos__ftz(ptr %ptr, flo
 ; GFX7-NEXT:  ; %bb.2: ; %atomicrmw.end
 ; GFX7-NEXT:    s_or_b64 exec, exec, s[4:5]
 ; GFX7-NEXT:    s_setpc_b64 s[30:31]
-  %gep = getelementptr float, ptr %ptr, i64 511
+  %gep = getelementptr inbounds float, ptr %ptr, i64 511
   %unused = atomicrmw fsub ptr %gep, float %val seq_cst
   ret void
 }
@@ -4085,7 +4085,7 @@ define double @flat_agent_atomic_fsub_ret_f64__offset12b_pos(ptr %ptr, double %v
 ; GFX7-NEXT:    s_or_b64 exec, exec, s[4:5]
 ; GFX7-NEXT:    s_waitcnt vmcnt(0)
 ; GFX7-NEXT:    s_setpc_b64 s[30:31]
-  %gep = getelementptr double, ptr %ptr, i64 255
+  %gep = getelementptr inbounds double, ptr %ptr, i64 255
   %result = atomicrmw fsub ptr %gep, double %val syncscope("agent") seq_cst
   ret double %result
 }
@@ -4532,7 +4532,7 @@ define double @flat_agent_atomic_fsub_ret_f64__offset12b_neg(ptr %ptr, double %v
 ; GFX7-NEXT:    s_or_b64 exec, exec, s[4:5]
 ; GFX7-NEXT:    s_waitcnt vmcnt(0)
 ; GFX7-NEXT:    s_setpc_b64 s[30:31]
-  %gep = getelementptr double, ptr %ptr, i64 -256
+  %gep = getelementptr inbounds double, ptr %ptr, i64 -256
   %result = atomicrmw fsub ptr %gep, double %val syncscope("agent") seq_cst
   ret double %result
 }
@@ -5385,7 +5385,7 @@ define void @flat_agent_atomic_fsub_noret_f64__offset12b_pos(ptr %ptr, double %v
 ; GFX7-NEXT:    s_or_b64 exec, exec, s[4:5]
 ; GFX7-NEXT:    s_waitcnt vmcnt(0)
 ; GFX7-NEXT:    s_setpc_b64 s[30:31]
-  %gep = getelementptr double, ptr %ptr, i64 255
+  %gep = getelementptr inbounds double, ptr %ptr, i64 255
   %unused = atomicrmw fsub ptr %gep, double %val syncscope("agent") seq_cst
   ret void
 }
@@ -5822,7 +5822,7 @@ define void @flat_agent_atomic_fsub_noret_f64__offset12b_neg(ptr %ptr, double %v
 ; GFX7-NEXT:    s_or_b64 exec, exec, s[4:5]
 ; GFX7-NEXT:    s_waitcnt vmcnt(0)
 ; GFX7-NEXT:    s_setpc_b64 s[30:31]
-  %gep = getelementptr double, ptr %ptr, i64 -256
+  %gep = getelementptr inbounds double, ptr %ptr, i64 -256
   %unused = atomicrmw fsub ptr %gep, double %val syncscope("agent") seq_cst
   ret void
 }
@@ -6582,7 +6582,7 @@ define half @flat_agent_atomic_fsub_ret_f16__offset12b_pos(ptr %ptr, half %val)
 ; GFX7-NEXT:    v_lshrrev_b32_e32 v0, v2, v5
 ; GFX7-NEXT:    v_cvt_f32_f16_e32 v0, v0
 ; GFX7-NEXT:    s_setpc_b64 s[30:31]
-  %gep = getelementptr half, ptr %ptr, i64 1023
+  %gep = getelementptr inbounds half, ptr %ptr, i64 1023
   %result = atomicrmw fsub ptr %gep, half %val syncscope("agent") seq_cst
   ret half %result
 }
@@ -6967,7 +6967,7 @@ define half @flat_agent_atomic_fsub_ret_f16__offset12b_neg(ptr %ptr, half %val)
 ; GFX7-NEXT:    v_lshrrev_b32_e32 v0, v2, v5
 ; GFX7-NEXT:    v_cvt_f32_f16_e32 v0, v0
 ; GFX7-NEXT:    s_setpc_b64 s[30:31]
-  %gep = getelementptr half, ptr %ptr, i64 -1024
+  %gep = getelementptr inbounds half, ptr %ptr, i64 -1024
   %result = atomicrmw fsub ptr %gep, half %val syncscope("agent") seq_cst
   ret half %result
  }
@@ -7697,7 +7697,7 @@ define void @flat_agent_atomic_fsub_noret_f16__offset12b_pos(ptr %ptr, half %val
 ; GFX7-NEXT:  ; %bb.2: ; %atomicrmw.end
 ; GFX7-NEXT:    s_or_b64 exec, exec, s[4:5]
 ; GFX7-NEXT:    s_setpc_b64 s[30:31]
-  %gep = getelementptr half, ptr %ptr, i64 1023
+  %gep = getelementptr inbounds half, ptr %ptr, i64 1023
   %unused = atomicrmw fsub ptr %gep, half %val syncscope("agent") seq_cst
   ret void
 }
@@ -8069,7 +8069,7 @@ define void @flat_agent_atomic_fsub_noret_f16__offset12b_neg(ptr %ptr, half %val
 ; GFX7-NEXT:  ; %bb.2: ; %atomicrmw.end
 ; GFX7-NEXT:    s_or_b64 exec, exec, s[4:5]
 ; GFX7-NEXT:    s_setpc_b64 s[30:31]
-  %gep = getelementptr half, ptr %ptr, i64 -1024
+  %gep = getelementptr inbounds half, ptr %ptr, i64 -1024
   %unused = atomicrmw fsub ptr %gep, half %val syncscope("agent") seq_cst
   ret void
 }
@@ -8353,7 +8353,7 @@ define half @flat_agent_atomic_fsub_ret_f16__offset12b_pos__align4(ptr %ptr, hal
 ; GFX7-NEXT:    s_or_b64 exec, exec, s[4:5]
 ; GFX7-NEXT:    v_cvt_f32_f16_e32 v0, v3
 ; GFX7-NEXT:    s_setpc_b64 s[30:31]
-  %gep = getelementptr half, ptr %ptr, i64 1023
+  %gep = getelementptr inbounds half, ptr %ptr, i64 1023
   %result = atomicrmw fsub ptr %gep, half %val syncscope("agent") seq_cst, align 4
   ret half %result
 }
@@ -8627,7 +8627,7 @@ define void @flat_agent_atomic_fsub_noret_f16__offset12b__align4_pos(ptr %ptr, h
 ; GFX7-NEXT:  ; %bb.2: ; %atomicrmw.end
 ; GFX7-NEXT:    s_or_b64 exec, exec, s[4:5]
 ; GFX7-NEXT:    s_setpc_b64 s[30:31]
-  %gep = getelementptr half, ptr %ptr, i64 1023
+  %gep = getelementptr inbounds half, ptr %ptr, i64 1023
   %unused = atomicrmw fsub ptr %gep, half %val syncscope("agent") seq_cst, align 4
   ret void
 }
@@ -9015,7 +9015,7 @@ define half @flat_system_atomic_fsub_ret_f16__offset12b_pos(ptr %ptr, half %val)
 ; GFX7-NEXT:    v_lshrrev_b32_e32 v0, v2, v5
 ; GFX7-NEXT:    v_cvt_f32_f16_e32 v0, v0
 ; GFX7-NEXT:    s_setpc_b64 s[30:31]
-  %gep = getelementptr half, ptr %ptr, i64 1023
+  %gep = getelementptr inbounds half, ptr %ptr, i64 1023
   %result = atomicrmw fsub ptr %gep, half %val seq_cst
   ret half %result
 }
@@ -9390,7 +9390,7 @@ define void @flat_system_atomic_fsub_noret_f16__offset12b_pos(ptr %ptr, half %va
 ; GFX7-NEXT:  ; %bb.2: ; %atomicrmw.end
 ; GFX7-NEXT:    s_or_b64 exec, exec, s[4:5]
 ; GFX7-NEXT:    s_setpc_b64 s[30:31]
-  %gep = getelementptr half, ptr %ptr, i64 1023
+  %gep = getelementptr inbounds half, ptr %ptr, i64 1023
   %unused = atomicrmw fsub ptr %gep, half %val seq_cst
   ret void
 }
@@ -10313,7 +10313,7 @@ define bfloat @flat_agent_atomic_fsub_ret_bf16__offset12b_pos(ptr %ptr, bfloat %
 ; GFX7-NEXT:    v_lshrrev_b32_e32 v0, v3, v5
 ; GFX7-NEXT:    v_lshlrev_b32_e32 v0, 16, v0
 ; GFX7-NEXT:    s_setpc_b64 s[30:31]
-  %gep = getelementptr bfloat, ptr %ptr, i64 1023
+  %gep = getelementptr inbounds bfloat, ptr %ptr, i64 1023
   %result = atomicrmw fsub ptr %gep, bfloat %val syncscope("agent") seq_cst
   ret bfloat %result
 }
@@ -10781,7 +10781,7 @@ define bfloat @flat_agent_atomic_fsub_ret_bf16__offset12b_neg(ptr %ptr, bfloat %
 ; GFX7-NEXT:    v_lshrrev_b32_e32 v0, v3, v5
 ; GFX7-NEXT:    v_lshlrev_b32_e32 v0, 16, v0
 ; GFX7-NEXT:    s_setpc_b64 s[30:31]
-  %gep = getelementptr bfloat, ptr %ptr, i64 -1024
+  %gep = getelementptr inbounds bfloat, ptr %ptr, i64 -1024
   %result = atomicrmw fsub ptr %gep, bfloat %val syncscope("agent") seq_cst
   ret bfloat %result
  }
@@ -11670,7 +11670,7 @@ define void @flat_agent_atomic_fsub_noret_bf16__offset12b_pos(ptr %ptr, bfloat %
 ; GFX7-NEXT:  ; %bb.2: ; %atomicrmw.end
 ; GFX7-NEXT:    s_or_b64 exec, exec, s[4:5]
 ; GFX7-NEXT:    s_setpc_b64 s[30:31]
-  %gep = getelementptr bfloat, ptr %ptr, i64 1023
+  %gep = getelementptr inbounds bfloat, ptr %ptr, i64 1023
   %unused = atomicrmw fsub ptr %gep, bfloat %val syncscope("agent") seq_cst
   ret void
 }
@@ -12123,7 +12123,7 @@ define void @flat_agent_atomic_fsub_noret_bf16__offset12b_neg(ptr %ptr, bfloat %
 ; GFX7-NEXT:  ; %bb.2: ; %atomicrmw.end
 ; GFX7-NEXT:    s_or_b64 exec, exec, s[4:5]
 ; GFX7-NEXT:    s_setpc_b64 s[30:31]
-  %gep = getelementptr bfloat, ptr %ptr, i64 -1024
+  %gep = getelementptr inbounds bfloat, ptr %ptr, i64 -1024
   %unused = atomicrmw fsub ptr %gep, bfloat %val syncscope("agent") seq_cst
   ret void
 }
@@ -12496,7 +12496,7 @@ define bfloat @flat_agent_atomic_fsub_ret_bf16__offset12b_pos__align4(ptr %ptr,
 ; GFX7-NEXT:    s_or_b64 exec, exec, s[4:5]
 ; GFX7-NEXT:    v_lshlrev_b32_e32 v0, 16, v3
 ; GFX7-NEXT:    s_setpc_b64 s[30:31]
-  %gep = getelementptr bfloat, ptr %ptr, i64 1023
+  %gep = getelementptr inbounds bfloat, ptr %ptr, i64 1023
   %result = atomicrmw fsub ptr %gep, bfloat %val syncscope("agent") seq_cst, align 4
   ret bfloat %result
 }
@@ -12857,7 +12857,7 @@ define void @flat_agent_atomic_fsub_noret_bf16__offset12b__align4_pos(ptr %ptr,
 ; GFX7-NEXT:  ; %bb.2: ; %atomicrmw.end
 ; GFX7-NEXT:    s_or_b64 exec, exec, s[4:5]
 ; GFX7-NEXT:    s_setpc_b64 s[30:31]
-  %gep = getelementptr bfloat, ptr %ptr, i64 1023
+  %gep = getelementptr inbounds bfloat, ptr %ptr, i64 1023
   %unused = atomicrmw fsub ptr %gep, bfloat %val syncscope("agent") seq_cst, align 4
   ret void
 }
@@ -13328,7 +13328,7 @@ define bfloat @flat_system_atomic_fsub_ret_bf16__offset12b_pos(ptr %ptr, bfloat
 ; GFX7-NEXT:    v_lshrrev_b32_e32 v0, v3, v5
 ; GFX7-NEXT:    v_lshlrev_b32_e32 v0, 16, v0
 ; GFX7-NEXT:    s_setpc_b64 s[30:31]
-  %gep = getelementptr bfloat, ptr %ptr, i64 1023
+  %gep = getelementptr inbounds bfloat, ptr %ptr, i64 1023
   %result = atomicrmw fsub ptr %gep, bfloat %val seq_cst
   ret bfloat %result
 }
@@ -13784,7 +13784,7 @@ define void @flat_system_atomic_fsub_noret_bf16__offset12b_pos(ptr %ptr, bfloat
 ; GFX7-NEXT:  ; %bb.2: ; %atomicrmw.end
 ; GFX7-NEXT:    s_or_b64 exec, exec, s[4:5]
 ; GFX7-NEXT:    s_setpc_b64 s[30:31]
-  %gep = getelementptr bfloat, ptr %ptr, i64 1023
+  %gep = getelementptr inbounds bfloat, ptr %ptr, i64 1023
   %unused = atomicrmw fsub ptr %gep, bfloat %val seq_cst
   ret void
 }
@@ -14229,7 +14229,7 @@ define <2 x half> @flat_agent_atomic_fsub_ret_v2f16__offset12b_pos(ptr %ptr, <2
 ; GFX7-NEXT:  ; %bb.2: ; %atomicrmw.end
 ; GFX7-NEXT:    s_or_b64 exec, exec, s[4:5]
 ; GFX7-NEXT:    s_setpc_b64 s[30:31]
-  %gep = getelementptr <2 x half>, ptr %ptr, i64 511
+  %gep = getelementptr inbounds <2 x half>, ptr %ptr, i64 511
   %result = atomicrmw fsub ptr %gep, <2 x half> %val syncscope("agent") seq_cst
   ret <2 x half> %result
 }
@@ -14468,7 +14468,7 @@ define <2 x half> @flat_agent_atomic_fsub_ret_v2f16__offset12b_neg(ptr %ptr, <2
 ; GFX7-NEXT:  ; %bb.2: ; %atomicrmw.end
 ; GFX7-NEXT:    s_or_b64 exec, exec, s[4:5]
 ; GFX7-NEXT:    s_setpc_b64 s[30:31]
-  %gep = getelementptr <2 x half>, ptr %ptr, i64 -512
+  %gep = getelementptr inbounds <2 x half>, ptr %ptr, i64 -512
   %result = atomicrmw fsub ptr %gep, <2 x half> %val syncscope("agent") seq_cst
   ret <2 x half> %result
 }
@@ -14891,7 +14891,7 @@ define void @flat_agent_atomic_fsub_noret_v2f16__offset12b_pos(ptr %ptr, <2 x ha
 ; GFX7-NEXT:  ; %bb.2: ; %atomicrmw.end
 ; GFX7-NEXT:    s_or_b64 exec, exec, s[4:5]
 ; GFX7-NEXT:    s_setpc_b64 s[30:31]
-  %gep = getelementptr <2 x half>, ptr %ptr, i64 511
+  %gep = getelementptr inbounds <2 x half>, ptr %ptr, i64 511
   %unused = atomicrmw fsub ptr %gep, <2 x half> %val syncscope("agent") seq_cst
   ret void
 }
@@ -15125,7 +15125,7 @@ define void @flat_agent_atomic_fsub_noret_v2f16__offset12b_neg(ptr %ptr, <2 x ha
 ; GFX7-NEXT:  ; %bb.2: ; %atomicrmw.end
 ; GFX7-NEXT:    s_or_b64 exec, exec, s[4:5]
 ; GFX7-NEXT:    s_setpc_b64 s[30:31]
-  %gep = getelementptr <2 x half>, ptr %ptr, i64 -512
+  %gep = getelementptr inbounds <2 x half>, ptr %ptr, i64 -512
   %unused = atomicrmw fsub ptr %gep, <2 x half> %val syncscope("agent") seq_cst
   ret void
 }
@@ -15350,7 +15350,7 @@ define <2 x half> @flat_system_atomic_fsub_ret_v2f16__offset12b_pos(ptr %ptr, <2
 ; GFX7-NEXT:  ; %bb.2: ; %atomicrmw.end
 ; GFX7-NEXT:    s_or_b64 exec, exec, s[4:5]
 ; GFX7-NEXT:    s_setpc_b64 s[30:31]
-  %gep = getelementptr <2 x half>, ptr %ptr, i64 511
+  %gep = getelementptr inbounds <2 x half>, ptr %ptr, i64 511
   %result = atomicrmw fsub ptr %gep, <2 x half> %val seq_cst
   ret <2 x half> %result
 }
@@ -15568,7 +15568,7 @@ define void @flat_system_atomic_fsub_noret_v2f16__offset12b_pos(ptr %ptr, <2 x h
 ; GFX7-NEXT:  ; %bb.2: ; %atomicrmw.end
 ; GFX7-NEXT:    s_or_b64 exec, exec, s[4:5]
 ; GFX7-NEXT:    s_setpc_b64 s[30:31]
-  %gep = getelementptr <2 x half>, ptr %ptr, i64 511
+  %gep = getelementptr inbounds <2 x half>, ptr %ptr, i64 511
   %unused = atomicrmw fsub ptr %gep, <2 x half> %val seq_cst
   ret void
 }
@@ -16469,7 +16469,7 @@ define <2 x bfloat> @flat_agent_atomic_fsub_ret_v2bf16__offset12b_pos(ptr %ptr,
 ; GFX7-NEXT:  ; %bb.2: ; %atomicrmw.end
 ; GFX7-NEXT:    s_or_b64 exec, exec, s[4:5]
 ; GFX7-NEXT:    s_setpc_b64 s[30:31]
-  %gep = getelementptr <2 x bfloat>, ptr %ptr, i64 511
+  %gep = getelementptr inbounds <2 x bfloat>, ptr %ptr, i64 511
   %result = atomicrmw fsub ptr %gep, <2 x bfloat> %val syncscope("agent") seq_cst
   ret <2 x bfloat> %result
 }
@@ -16937,7 +16937,7 @@ define <2 x bfloat> @flat_agent_atomic_fsub_ret_v2bf16__offset12b_neg(ptr %ptr,
 ; GFX7-NEXT:  ; %bb.2: ; %atomicrmw.end
 ; GFX7-NEXT:    s_or_b64 exec, exec, s[4:5]
 ; GFX7-NEXT:    s_setpc_b64 s[30:31]
-  %gep = getelementptr <2 x bfloat>, ptr %ptr, i64 -512
+  %gep = getelementptr inbounds <2 x bfloat>, ptr %ptr, i64 -512
   %result = atomicrmw fsub ptr %gep, <2 x bfloat> %val syncscope("agent") seq_cst
   ret <2 x bfloat> %result
 }
@@ -17808,7 +17808,7 @@ define void @flat_agent_atomic_fsub_noret_v2bf16__offset12b_pos(ptr %ptr, <2 x b
 ; GFX7-NEXT:  ; %bb.2: ; %atomicrmw.end
 ; GFX7-NEXT:    s_or_b64 exec, exec, s[4:5]
 ; GFX7-NEXT:    s_setpc_b64 s[30:31]
-  %gep = getelementptr <2 x bfloat>, ptr %ptr, i64 511
+  %gep = getelementptr inbounds <2 x bfloat>, ptr %ptr, i64 511
   %unused = atomicrmw fsub ptr %gep, <2 x bfloat> %val syncscope("agent") seq_cst
   ret void
 }
@@ -18269,7 +18269,7 @@ define void @flat_agent_atomic_fsub_noret_v2bf16__offset12b_neg(ptr %ptr, <2 x b
 ; GFX7-NEXT:  ; %bb.2: ; %atomicrmw.end
 ; GFX7-NEXT:    s_or_b64 exec, exec, s[4:5]
 ; GFX7-NEXT:    s_setpc_b64 s[30:31]
-  %gep = getelementptr <2 x bfloat>, ptr %ptr, i64 -512
+  %gep = getelementptr inbounds <2 x bfloat>, ptr %ptr, i64 -512
   %unused = atomicrmw fsub ptr %gep, <2 x bfloat> %val syncscope("agent") seq_cst
   ret void
 }
@@ -18723,7 +18723,7 @@ define <2 x bfloat> @flat_system_atomic_fsub_ret_v2bf16__offset12b_pos(ptr %ptr,
 ; GFX7-NEXT:  ; %bb.2: ; %atomicrmw.end
 ; GFX7-NEXT:    s_or_b64 exec, exec, s[4:5]
 ; GFX7-NEXT:    s_setpc_b64 s[30:31]
-  %gep = getelementptr <2 x bfloat>, ptr %ptr, i64 511
+  %gep = getelementptr inbounds <2 x bfloat>, ptr %ptr, i64 511
   %result = atomicrmw fsub ptr %gep, <2 x bfloat> %val seq_cst
   ret <2 x bfloat> %result
 }
@@ -19166,7 +19166,7 @@ define void @flat_system_atomic_fsub_noret_v2bf16__offset12b_pos(ptr %ptr, <2 x
 ; GFX7-NEXT:  ; %bb.2: ; %atomicrmw.end
 ; GFX7-NEXT:    s_or_b64 exec, exec, s[4:5]
 ; GFX7-NEXT:    s_setpc_b64 s[30:31]
-  %gep = getelementptr <2 x bfloat>, ptr %ptr, i64 511
+  %gep = getelementptr inbounds <2 x bfloat>, ptr %ptr, i64 511
   %unused = atomicrmw fsub ptr %gep, <2 x bfloat> %val seq_cst
   ret void
 }
diff --git a/llvm/test/CodeGen/AMDGPU/flat_atomics.ll b/llvm/test/CodeGen/AMDGPU/flat_atomics.ll
index 57be2907da4a0..b35f07002a48a 100644
--- a/llvm/test/CodeGen/AMDGPU/flat_atomics.ll
+++ b/llvm/test/CodeGen/AMDGPU/flat_atomics.ll
@@ -64,7 +64,7 @@ define amdgpu_kernel void @atomic_add_i32_offset(ptr %out, i32 %in) {
 ; GFX11-NEXT:    buffer_gl0_inv
 ; GFX11-NEXT:    s_endpgm
 entry:
-  %gep = getelementptr i32, ptr %out, i32 4
+  %gep = getelementptr inbounds i32, ptr %out, i32 4
   %val = atomicrmw add ptr %gep, i32 %in syncscope("agent") seq_cst
   ret void
 }
@@ -128,7 +128,7 @@ define amdgpu_kernel void @atomic_add_i32_max_offset(ptr %out, i32 %in) {
 ; GFX11-NEXT:    buffer_gl0_inv
 ; GFX11-NEXT:    s_endpgm
 entry:
-  %gep = getelementptr i32, ptr %out, i32 1023
+  %gep = getelementptr inbounds i32, ptr %out, i32 1023
   %val = atomicrmw volatile add ptr %gep, i32 %in syncscope("agent") seq_cst
   ret void
 }
@@ -196,7 +196,7 @@ define amdgpu_kernel void @atomic_add_i32_max_offset_p1(ptr %out, i32 %in) {
 ; GFX11-NEXT:    buffer_gl0_inv
 ; GFX11-NEXT:    s_endpgm
 entry:
-  %gep = getelementptr i32, ptr %out, i32 1024
+  %gep = getelementptr inbounds i32, ptr %out, i32 1024
   %val = atomicrmw volatile add ptr %gep, i32 %in syncscope("agent") seq_cst
   ret void
 }
@@ -270,7 +270,7 @@ define amdgpu_kernel void @atomic_add_i32_ret_offset(ptr %out, ptr %out2, i32 %i
 ; GFX11-NEXT:    flat_store_b32 v[0:1], v2
 ; GFX11-NEXT:    s_endpgm
 entry:
-  %gep = getelementptr i32, ptr %out, i32 4
+  %gep = getelementptr inbounds i32, ptr %out, i32 4
   %val = atomicrmw volatile add ptr %gep, i32 %in syncscope("agent") seq_cst
   store i32 %val, ptr %out2
   ret void
@@ -352,8 +352,8 @@ define amdgpu_kernel void @atomic_add_i32_addr64_offset(ptr %out, i32 %in, i64 %
 ; GFX11-NEXT:    buffer_gl0_inv
 ; GFX11-NEXT:    s_endpgm
 entry:
-  %ptr = getelementptr i32, ptr %out, i64 %index
-  %gep = getelementptr i32, ptr %ptr, i32 4
+  %ptr = getelementptr inbounds i32, ptr %out, i64 %index
+  %gep = getelementptr inbounds i32, ptr %ptr, i32 4
   %val = atomicrmw volatile add ptr %gep, i32 %in syncscope("agent") seq_cst
   ret void
 }
@@ -444,8 +444,8 @@ define amdgpu_kernel void @atomic_add_i32_ret_addr64_offset(ptr %out, ptr %out2,
 ; GFX11-NEXT:    flat_store_b32 v[0:1], v2
 ; GFX11-NEXT:    s_endpgm
 entry:
-  %ptr = getelementptr i32, ptr %out, i64 %index
-  %gep = getelementptr i32, ptr %ptr, i32 4
+  %ptr = getelementptr inbounds i32, ptr %out, i64 %index
+  %gep = getelementptr inbounds i32, ptr %ptr, i32 4
   %val = atomicrmw volatile add ptr %gep, i32 %in syncscope("agent") seq_cst
   store i32 %val, ptr %out2
   ret void
@@ -652,7 +652,7 @@ define amdgpu_kernel void @atomic_add_i32_addr64(ptr %out, i32 %in, i64 %index)
 ; GFX11-NEXT:    buffer_gl0_inv
 ; GFX11-NEXT:    s_endpgm
 entry:
-  %ptr = getelementptr i32, ptr %out, i64 %index
+  %ptr = getelementptr inbounds i32, ptr %out, i64 %index
   %val = atomicrmw volatile add ptr %ptr, i32 %in syncscope("agent") seq_cst
   ret void
 }
@@ -739,7 +739,7 @@ define amdgpu_kernel void @atomic_add_i32_ret_addr64(ptr %out, ptr %out2, i32 %i
 ; GFX11-NEXT:    flat_store_b32 v[0:1], v2
 ; GFX11-NEXT:    s_endpgm
 entry:
-  %ptr = getelementptr i32, ptr %out, i64 %index
+  %ptr = getelementptr inbounds i32, ptr %out, i64 %index
   %val = atomicrmw volatile add ptr %ptr, i32 %in syncscope("agent") seq_cst
   store i32 %val, ptr %out2
   ret void
@@ -804,7 +804,7 @@ define amdgpu_kernel void @atomic_and_i32_offset(ptr %out, i32 %in) {
 ; GFX11-NEXT:    buffer_gl0_inv
 ; GFX11-NEXT:    s_endpgm
 entry:
-  %gep = getelementptr i32, ptr %out, i32 4
+  %gep = getelementptr inbounds i32, ptr %out, i32 4
   %val = atomicrmw volatile and ptr %gep, i32 %in syncscope("agent") seq_cst, !amdgpu.no.fine.grained.memory !0
   ret void
 }
@@ -878,7 +878,7 @@ define amdgpu_kernel void @atomic_and_i32_ret_offset(ptr %out, ptr %out2, i32 %i
 ; GFX11-NEXT:    flat_store_b32 v[0:1], v2
 ; GFX11-NEXT:    s_endpgm
 entry:
-  %gep = getelementptr i32, ptr %out, i32 4
+  %gep = getelementptr inbounds i32, ptr %out, i32 4
   %val = atomicrmw volatile and ptr %gep, i32 %in syncscope("agent") seq_cst, !amdgpu.no.fine.grained.memory !0
   store i32 %val, ptr %out2
   ret void
@@ -960,8 +960,8 @@ define amdgpu_kernel void @atomic_and_i32_addr64_offset(ptr %out, i32 %in, i64 %
 ; GFX11-NEXT:    buffer_gl0_inv
 ; GFX11-NEXT:    s_endpgm
 entry:
-  %ptr = getelementptr i32, ptr %out, i64 %index
-  %gep = getelementptr i32, ptr %ptr, i32 4
+  %ptr = getelementptr inbounds i32, ptr %out, i64 %index
+  %gep = getelementptr inbounds i32, ptr %ptr, i32 4
   %val = atomicrmw volatile and ptr %gep, i32 %in syncscope("agent") seq_cst, !amdgpu.no.fine.grained.memory !0
   ret void
 }
@@ -1052,8 +1052,8 @@ define amdgpu_kernel void @atomic_and_i32_ret_addr64_offset(ptr %out, ptr %out2,
 ; GFX11-NEXT:    flat_store_b32 v[0:1], v2
 ; GFX11-NEXT:    s_endpgm
 entry:
-  %ptr = getelementptr i32, ptr %out, i64 %index
-  %gep = getelementptr i32, ptr %ptr, i32 4
+  %ptr = getelementptr inbounds i32, ptr %out, i64 %index
+  %gep = getelementptr inbounds i32, ptr %ptr, i32 4
   %val = atomicrmw volatile and ptr %gep, i32 %in syncscope("agent") seq_cst, !amdgpu.no.fine.grained.memory !0
   store i32 %val, ptr %out2
   ret void
@@ -1260,7 +1260,7 @@ define amdgpu_kernel void @atomic_and_i32_addr64(ptr %out, i32 %in, i64 %index)
 ; GFX11-NEXT:    buffer_gl0_inv
 ; GFX11-NEXT:    s_endpgm
 entry:
-  %ptr = getelementptr i32, ptr %out, i64 %index
+  %ptr = getelementptr inbounds i32, ptr %out, i64 %index
   %val = atomicrmw volatile and ptr %ptr, i32 %in syncscope("agent") seq_cst, !amdgpu.no.fine.grained.memory !0
   ret void
 }
@@ -1347,7 +1347,7 @@ define amdgpu_kernel void @atomic_and_i32_ret_addr64(ptr %out, ptr %out2, i32 %i
 ; GFX11-NEXT:    flat_store_b32 v[0:1], v2
 ; GFX11-NEXT:    s_endpgm
 entry:
-  %ptr = getelementptr i32, ptr %out, i64 %index
+  %ptr = getelementptr inbounds i32, ptr %out, i64 %index
   %val = atomicrmw volatile and ptr %ptr, i32 %in syncscope("agent") seq_cst, !amdgpu.no.fine.grained.memory !0
   store i32 %val, ptr %out2
   ret void
@@ -1412,7 +1412,7 @@ define amdgpu_kernel void @atomic_sub_i32_offset(ptr %out, i32 %in) {
 ; GFX11-NEXT:    buffer_gl0_inv
 ; GFX11-NEXT:    s_endpgm
 entry:
-  %gep = getelementptr i32, ptr %out, i32 4
+  %gep = getelementptr inbounds i32, ptr %out, i32 4
   %val = atomicrmw volatile sub ptr %gep, i32 %in syncscope("agent") seq_cst, !amdgpu.no.fine.grained.memory !0
   ret void
 }
@@ -1486,7 +1486,7 @@ define amdgpu_kernel void @atomic_sub_i32_ret_offset(ptr %out, ptr %out2, i32 %i
 ; GFX11-NEXT:    flat_store_b32 v[0:1], v2
 ; GFX11-NEXT:    s_endpgm
 entry:
-  %gep = getelementptr i32, ptr %out, i32 4
+  %gep = getelementptr inbounds i32, ptr %out, i32 4
   %val = atomicrmw volatile sub ptr %gep, i32 %in syncscope("agent") seq_cst, !amdgpu.no.fine.grained.memory !0
   store i32 %val, ptr %out2
   ret void
@@ -1568,8 +1568,8 @@ define amdgpu_kernel void @atomic_sub_i32_addr64_offset(ptr %out, i32 %in, i64 %
 ; GFX11-NEXT:    buffer_gl0_inv
 ; GFX11-NEXT:    s_endpgm
 entry:
-  %ptr = getelementptr i32, ptr %out, i64 %index
-  %gep = getelementptr i32, ptr %ptr, i32 4
+  %ptr = getelementptr inbounds i32, ptr %out, i64 %index
+  %gep = getelementptr inbounds i32, ptr %ptr, i32 4
   %val = atomicrmw volatile sub ptr %gep, i32 %in syncscope("agent") seq_cst, !amdgpu.no.fine.grained.memory !0
   ret void
 }
@@ -1660,8 +1660,8 @@ define amdgpu_kernel void @atomic_sub_i32_ret_addr64_offset(ptr %out, ptr %out2,
 ; GFX11-NEXT:    flat_store_b32 v[0:1], v2
 ; GFX11-NEXT:    s_endpgm
 entry:
-  %ptr = getelementptr i32, ptr %out, i64 %index
-  %gep = getelementptr i32, ptr %ptr, i32 4
+  %ptr = getelementptr inbounds i32, ptr %out, i64 %index
+  %gep = getelementptr inbounds i32, ptr %ptr, i32 4
   %val = atomicrmw volatile sub ptr %gep, i32 %in syncscope("agent") seq_cst, !amdgpu.no.fine.grained.memory !0
   store i32 %val, ptr %out2
   ret void
@@ -1868,7 +1868,7 @@ define amdgpu_kernel void @atomic_sub_i32_addr64(ptr %out, i32 %in, i64 %index)
 ; GFX11-NEXT:    buffer_gl0_inv
 ; GFX11-NEXT:    s_endpgm
 entry:
-  %ptr = getelementptr i32, ptr %out, i64 %index
+  %ptr = getelementptr inbounds i32, ptr %out, i64 %index
   %val = atomicrmw volatile sub ptr %ptr, i32 %in syncscope("agent") seq_cst, !amdgpu.no.fine.grained.memory !0
   ret void
 }
@@ -1955,7 +1955,7 @@ define amdgpu_kernel void @atomic_sub_i32_ret_addr64(ptr %out, ptr %out2, i32 %i
 ; GFX11-NEXT:    flat_store_b32 v[0:1], v2
 ; GFX11-NEXT:    s_endpgm
 entry:
-  %ptr = getelementptr i32, ptr %out, i64 %index
+  %ptr = getelementptr inbounds i32, ptr %out, i64 %index
   %val = atomicrmw volatile sub ptr %ptr, i32 %in syncscope("agent") seq_cst, !amdgpu.no.fine.grained.memory !0
   store i32 %val, ptr %out2
   ret void
@@ -2016,7 +2016,7 @@ define amdgpu_kernel void @atomic_max_i32_offset(ptr %out, i32 %in) {
 ; GFX11-NEXT:    buffer_gl0_inv
 ; GFX11-NEXT:    s_endpgm
 entry:
-  %gep = getelementptr i32, ptr %out, i32 4
+  %gep = getelementptr inbounds i32, ptr %out, i32 4
   %val = atomicrmw volatile max ptr %gep, i32 %in syncscope("workgroup") seq_cst, !amdgpu.no.fine.grained.memory !0
   ret void
 }
@@ -2089,7 +2089,7 @@ define amdgpu_kernel void @atomic_max_i32_ret_offset(ptr %out, ptr %out2, i32 %i
 ; GFX11-NEXT:    flat_store_b32 v[0:1], v2
 ; GFX11-NEXT:    s_endpgm
 entry:
-  %gep = getelementptr i32, ptr %out, i32 4
+  %gep = getelementptr inbounds i32, ptr %out, i32 4
   %val = atomicrmw volatile max ptr %gep, i32 %in syncscope("workgroup") seq_cst, !amdgpu.no.fine.grained.memory !0
   store i32 %val, ptr %out2
   ret void
@@ -2167,8 +2167,8 @@ define amdgpu_kernel void @atomic_max_i32_addr64_offset(ptr %out, i32 %in, i64 %
 ; GFX11-NEXT:    buffer_gl0_inv
 ; GFX11-NEXT:    s_endpgm
 entry:
-  %ptr = getelementptr i32, ptr %out, i64 %index
-  %gep = getelementptr i32, ptr %ptr, i32 4
+  %ptr = getelementptr inbounds i32, ptr %out, i64 %index
+  %gep = getelementptr inbounds i32, ptr %ptr, i32 4
   %val = atomicrmw volatile max ptr %gep, i32 %in syncscope("workgroup") seq_cst, !amdgpu.no.fine.grained.memory !0
   ret void
 }
@@ -2258,8 +2258,8 @@ define amdgpu_kernel void @atomic_max_i32_ret_addr64_offset(ptr %out, ptr %out2,
 ; GFX11-NEXT:    flat_store_b32 v[0:1], v2
 ; GFX11-NEXT:    s_endpgm
 entry:
-  %ptr = getelementptr i32, ptr %out, i64 %index
-  %gep = getelementptr i32, ptr %ptr, i32 4
+  %ptr = getelementptr inbounds i32, ptr %out, i64 %index
+  %gep = getelementptr inbounds i32, ptr %ptr, i32 4
   %val = atomicrmw volatile max ptr %gep, i32 %in syncscope("workgroup") seq_cst, !amdgpu.no.fine.grained.memory !0
   store i32 %val, ptr %out2
   ret void
@@ -2457,7 +2457,7 @@ define amdgpu_kernel void @atomic_max_i32_addr64(ptr %out, i32 %in, i64 %index)
 ; GFX11-NEXT:    buffer_gl0_inv
 ; GFX11-NEXT:    s_endpgm
 entry:
-  %ptr = getelementptr i32, ptr %out, i64 %index
+  %ptr = getelementptr inbounds i32, ptr %out, i64 %index
   %val = atomicrmw volatile max ptr %ptr, i32 %in syncscope("workgroup") seq_cst, !amdgpu.no.fine.grained.memory !0
   ret void
 }
@@ -2543,7 +2543,7 @@ define amdgpu_kernel void @atomic_max_i32_ret_addr64(ptr %out, ptr %out2, i32 %i
 ; GFX11-NEXT:    flat_store_b32 v[0:1], v2
 ; GFX11-NEXT:    s_endpgm
 entry:
-  %ptr = getelementptr i32, ptr %out, i64 %index
+  %ptr = getelementptr inbounds i32, ptr %out, i64 %index
   %val = atomicrmw volatile max ptr %ptr, i32 %in syncscope("workgroup") seq_cst, !amdgpu.no.fine.grained.memory !0
   store i32 %val, ptr %out2
   ret void
@@ -2604,7 +2604,7 @@ define amdgpu_kernel void @atomic_umax_i32_offset(ptr %out, i32 %in) {
 ; GFX11-NEXT:    buffer_gl0_inv
 ; GFX11-NEXT:    s_endpgm
 entry:
-  %gep = getelementptr i32, ptr %out, i32 4
+  %gep = getelementptr inbounds i32, ptr %out, i32 4
   %val = atomicrmw volatile umax ptr %gep, i32 %in syncscope("workgroup") seq_cst, !amdgpu.no.fine.grained.memory !0
   ret void
 }
@@ -2677,7 +2677,7 @@ define amdgpu_kernel void @atomic_umax_i32_ret_offset(ptr %out, ptr %out2, i32 %
 ; GFX11-NEXT:    flat_store_b32 v[0:1], v2
 ; GFX11-NEXT:    s_endpgm
 entry:
-  %gep = getelementptr i32, ptr %out, i32 4
+  %gep = getelementptr inbounds i32, ptr %out, i32 4
   %val = atomicrmw volatile umax ptr %gep, i32 %in syncscope("workgroup") seq_cst, !amdgpu.no.fine.grained.memory !0
   store i32 %val, ptr %out2
   ret void
@@ -2755,8 +2755,8 @@ define amdgpu_kernel void @atomic_umax_i32_addr64_offset(ptr %out, i32 %in, i64
 ; GFX11-NEXT:    buffer_gl0_inv
 ; GFX11-NEXT:    s_endpgm
 entry:
-  %ptr = getelementptr i32, ptr %out, i64 %index
-  %gep = getelementptr i32, ptr %ptr, i32 4
+  %ptr = getelementptr inbounds i32, ptr %out, i64 %index
+  %gep = getelementptr inbounds i32, ptr %ptr, i32 4
   %val = atomicrmw volatile umax ptr %gep, i32 %in syncscope("workgroup") seq_cst, !amdgpu.no.fine.grained.memory !0
   ret void
 }
@@ -2846,8 +2846,8 @@ define amdgpu_kernel void @atomic_umax_i32_ret_addr64_offset(ptr %out, ptr %out2
 ; GFX11-NEXT:    flat_store_b32 v[0:1], v2
 ; GFX11-NEXT:    s_endpgm
 entry:
-  %ptr = getelementptr i32, ptr %out, i64 %index
-  %gep = getelementptr i32, ptr %ptr, i32 4
+  %ptr = getelementptr inbounds i32, ptr %out, i64 %index
+  %gep = getelementptr inbounds i32, ptr %ptr, i32 4
   %val = atomicrmw volatile umax ptr %gep, i32 %in syncscope("workgroup") seq_cst, !amdgpu.no.fine.grained.memory !0
   store i32 %val, ptr %out2
   ret void
@@ -3045,7 +3045,7 @@ define amdgpu_kernel void @atomic_umax_i32_addr64(ptr %out, i32 %in, i64 %index)
 ; GFX11-NEXT:    buffer_gl0_inv
 ; GFX11-NEXT:    s_endpgm
 entry:
-  %ptr = getelementptr i32, ptr %out, i64 %index
+  %ptr = getelementptr inbounds i32, ptr %out, i64 %index
   %val = atomicrmw volatile umax ptr %ptr, i32 %in syncscope("workgroup") seq_cst, !amdgpu.no.fine.grained.memory !0
   ret void
 }
@@ -3131,7 +3131,7 @@ define amdgpu_kernel void @atomic_umax_i32_ret_addr64(ptr %out, ptr %out2, i32 %
 ; GFX11-NEXT:    flat_store_b32 v[0:1], v2
 ; GFX11-NEXT:    s_endpgm
 entry:
-  %ptr = getelementptr i32, ptr %out, i64 %index
+  %ptr = getelementptr inbounds i32, ptr %out, i64 %index
   %val = atomicrmw volatile umax ptr %ptr, i32 %in syncscope("workgroup") seq_cst, !amdgpu.no.fine.grained.memory !0
   store i32 %val, ptr %out2
   ret void
@@ -3192,7 +3192,7 @@ define amdgpu_kernel void @atomic_min_i32_offset(ptr %out, i32 %in) {
 ; GFX11-NEXT:    buffer_gl0_inv
 ; GFX11-NEXT:    s_endpgm
 entry:
-  %gep = getelementptr i32, ptr %out, i32 4
+  %gep = getelementptr inbounds i32, ptr %out, i32 4
   %val = atomicrmw volatile min ptr %gep, i32 %in syncscope("workgroup") seq_cst, !amdgpu.no.fine.grained.memory !0
   ret void
 }
@@ -3265,7 +3265,7 @@ define amdgpu_kernel void @atomic_min_i32_ret_offset(ptr %out, ptr %out2, i32 %i
 ; GFX11-NEXT:    flat_store_b32 v[0:1], v2
 ; GFX11-NEXT:    s_endpgm
 entry:
-  %gep = getelementptr i32, ptr %out, i32 4
+  %gep = getelementptr inbounds i32, ptr %out, i32 4
   %val = atomicrmw volatile min ptr %gep, i32 %in syncscope("workgroup") seq_cst, !amdgpu.no.fine.grained.memory !0
   store i32 %val, ptr %out2
   ret void
@@ -3343,8 +3343,8 @@ define amdgpu_kernel void @atomic_min_i32_addr64_offset(ptr %out, i32 %in, i64 %
 ; GFX11-NEXT:    buffer_gl0_inv
 ; GFX11-NEXT:    s_endpgm
 entry:
-  %ptr = getelementptr i32, ptr %out, i64 %index
-  %gep = getelementptr i32, ptr %ptr, i32 4
+  %ptr = getelementptr inbounds i32, ptr %out, i64 %index
+  %gep = getelementptr inbounds i32, ptr %ptr, i32 4
   %val = atomicrmw volatile min ptr %gep, i32 %in syncscope("workgroup") seq_cst, !amdgpu.no.fine.grained.memory !0
   ret void
 }
@@ -3434,8 +3434,8 @@ define amdgpu_kernel void @atomic_min_i32_ret_addr64_offset(ptr %out, ptr %out2,
 ; GFX11-NEXT:    flat_store_b32 v[0:1], v2
 ; GFX11-NEXT:    s_endpgm
 entry:
-  %ptr = getelementptr i32, ptr %out, i64 %index
-  %gep = getelementptr i32, ptr %ptr, i32 4
+  %ptr = getelementptr inbounds i32, ptr %out, i64 %index
+  %gep = getelementptr inbounds i32, ptr %ptr, i32 4
   %val = atomicrmw volatile min ptr %gep, i32 %in syncscope("workgroup") seq_cst, !amdgpu.no.fine.grained.memory !0
   store i32 %val, ptr %out2
   ret void
@@ -3633,7 +3633,7 @@ define amdgpu_kernel void @atomic_min_i32_addr64(ptr %out, i32 %in, i64 %index)
 ; GFX11-NEXT:    buffer_gl0_inv
 ; GFX11-NEXT:    s_endpgm
 entry:
-  %ptr = getelementptr i32, ptr %out, i64 %index
+  %ptr = getelementptr inbounds i32, ptr %out, i64 %index
   %val = atomicrmw volatile min ptr %ptr, i32 %in syncscope("workgroup") seq_cst, !amdgpu.no.fine.grained.memory !0
   ret void
 }
@@ -3719,7 +3719,7 @@ define amdgpu_kernel void @atomic_min_i32_ret_addr64(ptr %out, ptr %out2, i32 %i
 ; GFX11-NEXT:    flat_store_b32 v[0:1], v2
 ; GFX11-NEXT:    s_endpgm
 entry:
-  %ptr = getelementptr i32, ptr %out, i64 %index
+  %ptr = getelementptr inbounds i32, ptr %out, i64 %index
   %val = atomicrmw volatile min ptr %ptr, i32 %in syncscope("workgroup") seq_cst, !amdgpu.no.fine.grained.memory !0
   store i32 %val, ptr %out2
   ret void
@@ -3780,7 +3780,7 @@ define amdgpu_kernel void @atomic_umin_i32_offset(ptr %out, i32 %in) {
 ; GFX11-NEXT:    buffer_gl0_inv
 ; GFX11-NEXT:    s_endpgm
 entry:
-  %gep = getelementptr i32, ptr %out, i32 4
+  %gep = getelementptr inbounds i32, ptr %out, i32 4
   %val = atomicrmw volatile umin ptr %gep, i32 %in syncscope("workgroup") seq_cst, !amdgpu.no.fine.grained.memory !0
   ret void
 }
@@ -3853,7 +3853,7 @@ define amdgpu_kernel void @atomic_umin_i32_ret_offset(ptr %out, ptr %out2, i32 %
 ; GFX11-NEXT:    flat_store_b32 v[0:1], v2
 ; GFX11-NEXT:    s_endpgm
 entry:
-  %gep = getelementptr i32, ptr %out, i32 4
+  %gep = getelementptr inbounds i32, ptr %out, i32 4
   %val = atomicrmw volatile umin ptr %gep, i32 %in syncscope("workgroup") seq_cst, !amdgpu.no.fine.grained.memory !0
   store i32 %val, ptr %out2
   ret void
@@ -3931,8 +3931,8 @@ define amdgpu_kernel void @atomic_umin_i32_addr64_offset(ptr %out, i32 %in, i64
 ; GFX11-NEXT:    buffer_gl0_inv
 ; GFX11-NEXT:    s_endpgm
 entry:
-  %ptr = getelementptr i32, ptr %out, i64 %index
-  %gep = getelementptr i32, ptr %ptr, i32 4
+  %ptr = getelementptr inbounds i32, ptr %out, i64 %index
+  %gep = getelementptr inbounds i32, ptr %ptr, i32 4
   %val = atomicrmw volatile umin ptr %gep, i32 %in syncscope("workgroup") seq_cst, !amdgpu.no.fine.grained.memory !0
   ret void
 }
@@ -4022,8 +4022,8 @@ define amdgpu_kernel void @atomic_umin_i32_ret_addr64_offset(ptr %out, ptr %out2
 ; GFX11-NEXT:    flat_store_b32 v[0:1], v2
 ; GFX11-NEXT:    s_endpgm
 entry:
-  %ptr = getelementptr i32, ptr %out, i64 %index
-  %gep = getelementptr i32, ptr %ptr, i32 4
+  %ptr = getelementptr inbounds i32, ptr %out, i64 %index
+  %gep = getelementptr inbounds i32, ptr %ptr, i32 4
   %val = atomicrmw volatile umin ptr %gep, i32 %in syncscope("workgroup") seq_cst, !amdgpu.no.fine.grained.memory !0
   store i32 %val, ptr %out2
   ret void
@@ -4221,7 +4221,7 @@ define amdgpu_kernel void @atomic_umin_i32_addr64(ptr %out, i32 %in, i64 %index)
 ; GFX11-NEXT:    buffer_gl0_inv
 ; GFX11-NEXT:    s_endpgm
 entry:
-  %ptr = getelementptr i32, ptr %out, i64 %index
+  %ptr = getelementptr inbounds i32, ptr %out, i64 %index
   %val = atomicrmw volatile umin ptr %ptr, i32 %in syncscope("workgroup") seq_cst, !amdgpu.no.fine.grained.memory !0
   ret void
 }
@@ -4307,7 +4307,7 @@ define amdgpu_kernel void @atomic_umin_i32_ret_addr64(ptr %out, ptr %out2, i32 %
 ; GFX11-NEXT:    flat_store_b32 v[0:1], v2
 ; GFX11-NEXT:    s_endpgm
 entry:
-  %ptr = getelementptr i32, ptr %out, i64 %index
+  %ptr = getelementptr inbounds i32, ptr %out, i64 %index
   %val = atomicrmw volatile umin ptr %ptr, i32 %in syncscope("workgroup") seq_cst, !amdgpu.no.fine.grained.memory !0
   store i32 %val, ptr %out2
   ret void
@@ -4372,7 +4372,7 @@ define amdgpu_kernel void @atomic_or_i32_offset(ptr %out, i32 %in) {
 ; GFX11-NEXT:    buffer_gl0_inv
 ; GFX11-NEXT:    s_endpgm
 entry:
-  %gep = getelementptr i32, ptr %out, i32 4
+  %gep = getelementptr inbounds i32, ptr %out, i32 4
   %val = atomicrmw volatile or ptr %gep, i32 %in syncscope("agent") seq_cst, !amdgpu.no.fine.grained.memory !0
   ret void
 }
@@ -4446,7 +4446,7 @@ define amdgpu_kernel void @atomic_or_i32_ret_offset(ptr %out, ptr %out2, i32 %in
 ; GFX11-NEXT:    flat_store_b32 v[0:1], v2
 ; GFX11-NEXT:    s_endpgm
 entry:
-  %gep = getelementptr i32, ptr %out, i32 4
+  %gep = getelementptr inbounds i32, ptr %out, i32 4
   %val = atomicrmw volatile or ptr %gep, i32 %in syncscope("agent") seq_cst, !amdgpu.no.fine.grained.memory !0
   store i32 %val, ptr %out2
   ret void
@@ -4528,8 +4528,8 @@ define amdgpu_kernel void @atomic_or_i32_addr64_offset(ptr %out, i32 %in, i64 %i
 ; GFX11-NEXT:    buffer_gl0_inv
 ; GFX11-NEXT:    s_endpgm
 entry:
-  %ptr = getelementptr i32, ptr %out, i64 %index
-  %gep = getelementptr i32, ptr %ptr, i32 4
+  %ptr = getelementptr inbounds i32, ptr %out, i64 %index
+  %gep = getelementptr inbounds i32, ptr %ptr, i32 4
   %val = atomicrmw volatile or ptr %gep, i32 %in syncscope("agent") seq_cst, !amdgpu.no.fine.grained.memory !0
   ret void
 }
@@ -4620,8 +4620,8 @@ define amdgpu_kernel void @atomic_or_i32_ret_addr64_offset(ptr %out, ptr %out2,
 ; GFX11-NEXT:    flat_store_b32 v[0:1], v2
 ; GFX11-NEXT:    s_endpgm
 entry:
-  %ptr = getelementptr i32, ptr %out, i64 %index
-  %gep = getelementptr i32, ptr %ptr, i32 4
+  %ptr = getelementptr inbounds i32, ptr %out, i64 %index
+  %gep = getelementptr inbounds i32, ptr %ptr, i32 4
   %val = atomicrmw volatile or ptr %gep, i32 %in syncscope("agent") seq_cst, !amdgpu.no.fine.grained.memory !0
   store i32 %val, ptr %out2
   ret void
@@ -4828,7 +4828,7 @@ define amdgpu_kernel void @atomic_or_i32_addr64(ptr %out, i32 %in, i64 %index) {
 ; GFX11-NEXT:    buffer_gl0_inv
 ; GFX11-NEXT:    s_endpgm
 entry:
-  %ptr = getelementptr i32, ptr %out, i64 %index
+  %ptr = getelementptr inbounds i32, ptr %out, i64 %index
   %val = atomicrmw volatile or ptr %ptr, i32 %in syncscope("agent") seq_cst, !amdgpu.no.fine.grained.memory !0
   ret void
 }
@@ -4915,7 +4915,7 @@ define amdgpu_kernel void @atomic_or_i32_ret_addr64(ptr %out, ptr %out2, i32 %in
 ; GFX11-NEXT:    flat_store_b32 v[0:1], v2
 ; GFX11-NEXT:    s_endpgm
 entry:
-  %ptr = getelementptr i32, ptr %out, i64 %index
+  %ptr = getelementptr inbounds i32, ptr %out, i64 %index
   %val = atomicrmw volatile or ptr %ptr, i32 %in syncscope("agent") seq_cst, !amdgpu.no.fine.grained.memory !0
   store i32 %val, ptr %out2
   ret void
@@ -4980,7 +4980,7 @@ define amdgpu_kernel void @atomic_xchg_i32_offset(ptr %out, i32 %in) {
 ; GFX11-NEXT:    buffer_gl0_inv
 ; GFX11-NEXT:    s_endpgm
 entry:
-  %gep = getelementptr i32, ptr %out, i32 4
+  %gep = getelementptr inbounds i32, ptr %out, i32 4
   %val = atomicrmw volatile xchg ptr %gep, i32 %in syncscope("agent") seq_cst
   ret void
 }
@@ -5044,7 +5044,7 @@ define amdgpu_kernel void @atomic_xchg_f32_offset(ptr %out, float %in) {
 ; GFX11-NEXT:    buffer_gl0_inv
 ; GFX11-NEXT:    s_endpgm
 entry:
-  %gep = getelementptr float, ptr %out, i32 4
+  %gep = getelementptr inbounds float, ptr %out, i32 4
   %val = atomicrmw volatile xchg ptr %gep, float %in syncscope("agent") seq_cst
   ret void
 }
@@ -5118,7 +5118,7 @@ define amdgpu_kernel void @atomic_xchg_i32_ret_offset(ptr %out, ptr %out2, i32 %
 ; GFX11-NEXT:    flat_store_b32 v[0:1], v2
 ; GFX11-NEXT:    s_endpgm
 entry:
-  %gep = getelementptr i32, ptr %out, i32 4
+  %gep = getelementptr inbounds i32, ptr %out, i32 4
   %val = atomicrmw volatile xchg ptr %gep, i32 %in syncscope("agent") seq_cst
   store i32 %val, ptr %out2
   ret void
@@ -5200,8 +5200,8 @@ define amdgpu_kernel void @atomic_xchg_i32_addr64_offset(ptr %out, i32 %in, i64
 ; GFX11-NEXT:    buffer_gl0_inv
 ; GFX11-NEXT:    s_endpgm
 entry:
-  %ptr = getelementptr i32, ptr %out, i64 %index
-  %gep = getelementptr i32, ptr %ptr, i32 4
+  %ptr = getelementptr inbounds i32, ptr %out, i64 %index
+  %gep = getelementptr inbounds i32, ptr %ptr, i32 4
   %val = atomicrmw volatile xchg ptr %gep, i32 %in syncscope("agent") seq_cst
   ret void
 }
@@ -5292,8 +5292,8 @@ define amdgpu_kernel void @atomic_xchg_i32_ret_addr64_offset(ptr %out, ptr %out2
 ; GFX11-NEXT:    flat_store_b32 v[0:1], v2
 ; GFX11-NEXT:    s_endpgm
 entry:
-  %ptr = getelementptr i32, ptr %out, i64 %index
-  %gep = getelementptr i32, ptr %ptr, i32 4
+  %ptr = getelementptr inbounds i32, ptr %out, i64 %index
+  %gep = getelementptr inbounds i32, ptr %ptr, i32 4
   %val = atomicrmw volatile xchg ptr %gep, i32 %in syncscope("agent") seq_cst
   store i32 %val, ptr %out2
   ret void
@@ -5500,7 +5500,7 @@ define amdgpu_kernel void @atomic_xchg_i32_addr64(ptr %out, i32 %in, i64 %index)
 ; GFX11-NEXT:    buffer_gl0_inv
 ; GFX11-NEXT:    s_endpgm
 entry:
-  %ptr = getelementptr i32, ptr %out, i64 %index
+  %ptr = getelementptr inbounds i32, ptr %out, i64 %index
   %val = atomicrmw volatile xchg ptr %ptr, i32 %in syncscope("agent") seq_cst
   ret void
 }
@@ -5587,7 +5587,7 @@ define amdgpu_kernel void @atomic_xchg_i32_ret_addr64(ptr %out, ptr %out2, i32 %
 ; GFX11-NEXT:    flat_store_b32 v[0:1], v2
 ; GFX11-NEXT:    s_endpgm
 entry:
-  %ptr = getelementptr i32, ptr %out, i64 %index
+  %ptr = getelementptr inbounds i32, ptr %out, i64 %index
   %val = atomicrmw volatile xchg ptr %ptr, i32 %in syncscope("agent") seq_cst
   store i32 %val, ptr %out2
   ret void
@@ -5652,7 +5652,7 @@ define amdgpu_kernel void @atomic_cmpxchg_i32_offset(ptr %out, i32 %in, i32 %old
 ; GFX11-NEXT:    buffer_gl0_inv
 ; GFX11-NEXT:    s_endpgm
 entry:
-  %gep = getelementptr i32, ptr %out, i32 4
+  %gep = getelementptr inbounds i32, ptr %out, i32 4
   %val = cmpxchg volatile ptr %gep, i32 %old, i32 %in syncscope("agent") seq_cst seq_cst, !amdgpu.no.fine.grained.memory !0
   ret void
 }
@@ -5729,7 +5729,7 @@ define amdgpu_kernel void @atomic_cmpxchg_i32_ret_offset(ptr %out, ptr %out2, i3
 ; GFX11-NEXT:    flat_store_b32 v[0:1], v2
 ; GFX11-NEXT:    s_endpgm
 entry:
-  %gep = getelementptr i32, ptr %out, i32 4
+  %gep = getelementptr inbounds i32, ptr %out, i32 4
   %val = cmpxchg volatile ptr %gep, i32 %old, i32 %in syncscope("agent") seq_cst seq_cst
   %flag = extractvalue { i32, i1 } %val, 0
   store i32 %flag, ptr %out2
@@ -5819,8 +5819,8 @@ define amdgpu_kernel void @atomic_cmpxchg_i32_addr64_offset(ptr %out, i32 %in, i
 ; GFX11-NEXT:    buffer_gl0_inv
 ; GFX11-NEXT:    s_endpgm
 entry:
-  %ptr = getelementptr i32, ptr %out, i64 %index
-  %gep = getelementptr i32, ptr %ptr, i32 4
+  %ptr = getelementptr inbounds i32, ptr %out, i64 %index
+  %gep = getelementptr inbounds i32, ptr %ptr, i32 4
   %val  = cmpxchg volatile ptr %gep, i32 %old, i32 %in syncscope("agent") seq_cst seq_cst
   ret void
 }
@@ -5918,8 +5918,8 @@ define amdgpu_kernel void @atomic_cmpxchg_i32_ret_addr64_offset(ptr %out, ptr %o
 ; GFX11-NEXT:    flat_store_b32 v[0:1], v2
 ; GFX11-NEXT:    s_endpgm
 entry:
-  %ptr = getelementptr i32, ptr %out, i64 %index
-  %gep = getelementptr i32, ptr %ptr, i32 4
+  %ptr = getelementptr inbounds i32, ptr %out, i64 %index
+  %gep = getelementptr inbounds i32, ptr %ptr, i32 4
   %val  = cmpxchg volatile ptr %gep, i32 %old, i32 %in syncscope("agent") seq_cst seq_cst
   %flag = extractvalue { i32, i1 } %val, 0
   store i32 %flag, ptr %out2
@@ -6136,7 +6136,7 @@ define amdgpu_kernel void @atomic_cmpxchg_i32_addr64(ptr %out, i32 %in, i64 %ind
 ; GFX11-NEXT:    buffer_gl0_inv
 ; GFX11-NEXT:    s_endpgm
 entry:
-  %ptr = getelementptr i32, ptr %out, i64 %index
+  %ptr = getelementptr inbounds i32, ptr %out, i64 %index
   %val = cmpxchg volatile ptr %ptr, i32 %old, i32 %in syncscope("agent") seq_cst seq_cst
   ret void
 }
@@ -6230,7 +6230,7 @@ define amdgpu_kernel void @atomic_cmpxchg_i32_ret_addr64(ptr %out, ptr %out2, i3
 ; GFX11-NEXT:    flat_store_b32 v[0:1], v2
 ; GFX11-NEXT:    s_endpgm
 entry:
-  %ptr = getelementptr i32, ptr %out, i64 %index
+  %ptr = getelementptr inbounds i32, ptr %out, i64 %index
   %val = cmpxchg volatile ptr %ptr, i32 %old, i32 %in syncscope("agent") seq_cst seq_cst
   %flag = extractvalue { i32, i1 } %val, 0
   store i32 %flag, ptr %out2
@@ -6296,7 +6296,7 @@ define amdgpu_kernel void @atomic_xor_i32_offset(ptr %out, i32 %in) {
 ; GFX11-NEXT:    buffer_gl0_inv
 ; GFX11-NEXT:    s_endpgm
 entry:
-  %gep = getelementptr i32, ptr %out, i32 4
+  %gep = getelementptr inbounds i32, ptr %out, i32 4
   %val = atomicrmw volatile xor ptr %gep, i32 %in syncscope("agent") seq_cst, !amdgpu.no.fine.grained.memory !0
   ret void
 }
@@ -6370,7 +6370,7 @@ define amdgpu_kernel void @atomic_xor_i32_ret_offset(ptr %out, ptr %out2, i32 %i
 ; GFX11-NEXT:    flat_store_b32 v[0:1], v2
 ; GFX11-NEXT:    s_endpgm
 entry:
-  %gep = getelementptr i32, ptr %out, i32 4
+  %gep = getelementptr inbounds i32, ptr %out, i32 4
   %val = atomicrmw volatile xor ptr %gep, i32 %in syncscope("agent") seq_cst, !amdgpu.no.fine.grained.memory !0
   store i32 %val, ptr %out2
   ret void
@@ -6452,8 +6452,8 @@ define amdgpu_kernel void @atomic_xor_i32_addr64_offset(ptr %out, i32 %in, i64 %
 ; GFX11-NEXT:    buffer_gl0_inv
 ; GFX11-NEXT:    s_endpgm
 entry:
-  %ptr = getelementptr i32, ptr %out, i64 %index
-  %gep = getelementptr i32, ptr %ptr, i32 4
+  %ptr = getelementptr inbounds i32, ptr %out, i64 %index
+  %gep = getelementptr inbounds i32, ptr %ptr, i32 4
   %val = atomicrmw volatile xor ptr %gep, i32 %in syncscope("agent") seq_cst, !amdgpu.no.fine.grained.memory !0
   ret void
 }
@@ -6544,8 +6544,8 @@ define amdgpu_kernel void @atomic_xor_i32_ret_addr64_offset(ptr %out, ptr %out2,
 ; GFX11-NEXT:    flat_store_b32 v[0:1], v2
 ; GFX11-NEXT:    s_endpgm
 entry:
-  %ptr = getelementptr i32, ptr %out, i64 %index
-  %gep = getelementptr i32, ptr %ptr, i32 4
+  %ptr = getelementptr inbounds i32, ptr %out, i64 %index
+  %gep = getelementptr inbounds i32, ptr %ptr, i32 4
   %val = atomicrmw volatile xor ptr %gep, i32 %in syncscope("agent") seq_cst, !amdgpu.no.fine.grained.memory !0
   store i32 %val, ptr %out2
   ret void
@@ -6752,7 +6752,7 @@ define amdgpu_kernel void @atomic_xor_i32_addr64(ptr %out, i32 %in, i64 %index)
 ; GFX11-NEXT:    buffer_gl0_inv
 ; GFX11-NEXT:    s_endpgm
 entry:
-  %ptr = getelementptr i32, ptr %out, i64 %index
+  %ptr = getelementptr inbounds i32, ptr %out, i64 %index
   %val = atomicrmw volatile xor ptr %ptr, i32 %in syncscope("agent") seq_cst, !amdgpu.no.fine.grained.memory !0
   ret void
 }
@@ -6839,7 +6839,7 @@ define amdgpu_kernel void @atomic_xor_i32_ret_addr64(ptr %out, ptr %out2, i32 %i
 ; GFX11-NEXT:    flat_store_b32 v[0:1], v2
 ; GFX11-NEXT:    s_endpgm
 entry:
-  %ptr = getelementptr i32, ptr %out, i64 %index
+  %ptr = getelementptr inbounds i32, ptr %out, i64 %index
   %val = atomicrmw volatile xor ptr %ptr, i32 %in syncscope("agent") seq_cst, !amdgpu.no.fine.grained.memory !0
   store i32 %val, ptr %out2
   ret void
@@ -6905,7 +6905,7 @@ define amdgpu_kernel void @atomic_load_i32_offset(ptr %in, ptr %out) {
 ; GFX11-NEXT:    flat_store_b32 v[0:1], v2
 ; GFX11-NEXT:    s_endpgm
 entry:
-  %gep = getelementptr i32, ptr %in, i32 4
+  %gep = getelementptr inbounds i32, ptr %in, i32 4
   %val = load atomic i32, ptr %gep  seq_cst, align 4
   store i32 %val, ptr %out
   ret void
@@ -7050,8 +7050,8 @@ define amdgpu_kernel void @atomic_load_i32_addr64_offset(ptr %in, ptr %out, i64
 ; GFX11-NEXT:    flat_store_b32 v[0:1], v2
 ; GFX11-NEXT:    s_endpgm
 entry:
-  %ptr = getelementptr i32, ptr %in, i64 %index
-  %gep = getelementptr i32, ptr %ptr, i32 4
+  %ptr = getelementptr inbounds i32, ptr %in, i64 %index
+  %gep = getelementptr inbounds i32, ptr %ptr, i32 4
   %val = load atomic i32, ptr %gep seq_cst, align 4
   store i32 %val, ptr %out
   ret void
@@ -7131,7 +7131,7 @@ define amdgpu_kernel void @atomic_load_i32_addr64(ptr %in, ptr %out, i64 %index)
 ; GFX11-NEXT:    flat_store_b32 v[0:1], v2
 ; GFX11-NEXT:    s_endpgm
 entry:
-  %ptr = getelementptr i32, ptr %in, i64 %index
+  %ptr = getelementptr inbounds i32, ptr %in, i64 %index
   %val = load atomic i32, ptr %ptr seq_cst, align 4
   store i32 %val, ptr %out
   ret void
@@ -7186,7 +7186,7 @@ define amdgpu_kernel void @atomic_store_i32_offset(i32 %in, ptr %out) {
 ; GFX11-NEXT:    flat_store_b32 v[0:1], v2 offset:16
 ; GFX11-NEXT:    s_endpgm
 entry:
-  %gep = getelementptr i32, ptr %out, i32 4
+  %gep = getelementptr inbounds i32, ptr %out, i32 4
   store atomic i32 %in, ptr %gep  seq_cst, align 4
   ret void
 }
@@ -7302,8 +7302,8 @@ define amdgpu_kernel void @atomic_store_i32_addr64_offset(i32 %in, ptr %out, i64
 ; GFX11-NEXT:    flat_store_b32 v[0:1], v2 offset:16
 ; GFX11-NEXT:    s_endpgm
 entry:
-  %ptr = getelementptr i32, ptr %out, i64 %index
-  %gep = getelementptr i32, ptr %ptr, i32 4
+  %ptr = getelementptr inbounds i32, ptr %out, i64 %index
+  %gep = getelementptr inbounds i32, ptr %ptr, i32 4
   store atomic i32 %in, ptr %gep seq_cst, align 4
   ret void
 }
@@ -7366,7 +7366,7 @@ define amdgpu_kernel void @atomic_store_i32_addr64(i32 %in, ptr %out, i64 %index
 ; GFX11-NEXT:    flat_store_b32 v[0:1], v2
 ; GFX11-NEXT:    s_endpgm
 entry:
-  %ptr = getelementptr i32, ptr %out, i64 %index
+  %ptr = getelementptr inbounds i32, ptr %out, i64 %index
   store atomic i32 %in, ptr %ptr seq_cst, align 4
   ret void
 }
@@ -7431,7 +7431,7 @@ define amdgpu_kernel void @atomic_load_f32_offset(ptr %in, ptr %out) {
 ; GFX11-NEXT:    flat_store_b32 v[0:1], v2
 ; GFX11-NEXT:    s_endpgm
 entry:
-  %gep = getelementptr float, ptr %in, i32 4
+  %gep = getelementptr inbounds float, ptr %in, i32 4
   %val = load atomic float, ptr %gep  seq_cst, align 4
   store float %val, ptr %out
   ret void
@@ -7576,8 +7576,8 @@ define amdgpu_kernel void @atomic_load_f32_addr64_offset(ptr %in, ptr %out, i64
 ; GFX11-NEXT:    flat_store_b32 v[0:1], v2
 ; GFX11-NEXT:    s_endpgm
 entry:
-  %ptr = getelementptr float, ptr %in, i64 %index
-  %gep = getelementptr float, ptr %ptr, i32 4
+  %ptr = getelementptr inbounds float, ptr %in, i64 %index
+  %gep = getelementptr inbounds float, ptr %ptr, i32 4
   %val = load atomic float, ptr %gep seq_cst, align 4
   store float %val, ptr %out
   ret void
@@ -7657,7 +7657,7 @@ define amdgpu_kernel void @atomic_load_f32_addr64(ptr %in, ptr %out, i64 %index)
 ; GFX11-NEXT:    flat_store_b32 v[0:1], v2
 ; GFX11-NEXT:    s_endpgm
 entry:
-  %ptr = getelementptr float, ptr %in, i64 %index
+  %ptr = getelementptr inbounds float, ptr %in, i64 %index
   %val = load atomic float, ptr %ptr seq_cst, align 4
   store float %val, ptr %out
   ret void
@@ -7712,7 +7712,7 @@ define amdgpu_kernel void @atomic_store_f32_offset(float %in, ptr %out) {
 ; GFX11-NEXT:    flat_store_b32 v[0:1], v2 offset:16
 ; GFX11-NEXT:    s_endpgm
 entry:
-  %gep = getelementptr float, ptr %out, i32 4
+  %gep = getelementptr inbounds float, ptr %out, i32 4
   store atomic float %in, ptr %gep  seq_cst, align 4
   ret void
 }
@@ -7828,8 +7828,8 @@ define amdgpu_kernel void @atomic_store_f32_addr64_offset(float %in, ptr %out, i
 ; GFX11-NEXT:    flat_store_b32 v[0:1], v2 offset:16
 ; GFX11-NEXT:    s_endpgm
 entry:
-  %ptr = getelementptr float, ptr %out, i64 %index
-  %gep = getelementptr float, ptr %ptr, i32 4
+  %ptr = getelementptr inbounds float, ptr %out, i64 %index
+  %gep = getelementptr inbounds float, ptr %ptr, i32 4
   store atomic float %in, ptr %gep seq_cst, align 4
   ret void
 }
@@ -7892,7 +7892,7 @@ define amdgpu_kernel void @atomic_store_f32_addr64(float %in, ptr %out, i64 %ind
 ; GFX11-NEXT:    flat_store_b32 v[0:1], v2
 ; GFX11-NEXT:    s_endpgm
 entry:
-  %ptr = getelementptr float, ptr %out, i64 %index
+  %ptr = getelementptr inbounds float, ptr %out, i64 %index
   store atomic float %in, ptr %ptr seq_cst, align 4
   ret void
 }
@@ -7971,7 +7971,7 @@ define amdgpu_kernel void @atomic_load_i8_offset(ptr %in, ptr %out) {
 ; GFX11-FAKE16-NEXT:    flat_store_b8 v[0:1], v2
 ; GFX11-FAKE16-NEXT:    s_endpgm
 entry:
-  %gep = getelementptr i8, ptr %in, i64 16
+  %gep = getelementptr inbounds i8, ptr %in, i64 16
   %val = load atomic i8, ptr %gep  seq_cst, align 1
   store i8 %val, ptr %out
   ret void
@@ -8145,8 +8145,8 @@ define amdgpu_kernel void @atomic_load_i8_addr64_offset(ptr %in, ptr %out, i64 %
 ; GFX11-FAKE16-NEXT:    flat_store_b8 v[0:1], v2
 ; GFX11-FAKE16-NEXT:    s_endpgm
 entry:
-  %ptr = getelementptr i8, ptr %in, i64 %index
-  %gep = getelementptr i8, ptr %ptr, i64 16
+  %ptr = getelementptr inbounds i8, ptr %in, i64 %index
+  %gep = getelementptr inbounds i8, ptr %ptr, i64 16
   %val = load atomic i8, ptr %gep seq_cst, align 1
   store i8 %val, ptr %out
   ret void
@@ -8212,7 +8212,7 @@ define amdgpu_kernel void @atomic_store_i8_offset(i8 %in, ptr %out) {
 ; GFX11-FAKE16-NEXT:    flat_store_b8 v[0:1], v2 offset:16
 ; GFX11-FAKE16-NEXT:    s_endpgm
 entry:
-  %gep = getelementptr i8, ptr %out, i64 16
+  %gep = getelementptr inbounds i8, ptr %out, i64 16
   store atomic i8 %in, ptr %gep  seq_cst, align 1
   ret void
 }
@@ -8348,8 +8348,8 @@ define amdgpu_kernel void @atomic_store_i8_addr64_offset(i8 %in, ptr %out, i64 %
 ; GFX11-FAKE16-NEXT:    flat_store_b8 v[0:1], v2 offset:16
 ; GFX11-FAKE16-NEXT:    s_endpgm
 entry:
-  %ptr = getelementptr i8, ptr %out, i64 %index
-  %gep = getelementptr i8, ptr %ptr, i64 16
+  %ptr = getelementptr inbounds i8, ptr %out, i64 %index
+  %gep = getelementptr inbounds i8, ptr %ptr, i64 16
   store atomic i8 %in, ptr %gep seq_cst, align 1
   ret void
 }
@@ -8428,7 +8428,7 @@ define amdgpu_kernel void @atomic_load_i16_offset(ptr %in, ptr %out) {
 ; GFX11-FAKE16-NEXT:    flat_store_b16 v[0:1], v2
 ; GFX11-FAKE16-NEXT:    s_endpgm
 entry:
-  %gep = getelementptr i16, ptr %in, i64 8
+  %gep = getelementptr inbounds i16, ptr %in, i64 8
   %val = load atomic i16, ptr %gep  seq_cst, align 2
   store i16 %val, ptr %out
   ret void
@@ -8607,8 +8607,8 @@ define amdgpu_kernel void @atomic_load_i16_addr64_offset(ptr %in, ptr %out, i64
 ; GFX11-FAKE16-NEXT:    flat_store_b16 v[0:1], v2
 ; GFX11-FAKE16-NEXT:    s_endpgm
 entry:
-  %ptr = getelementptr i16, ptr %in, i64 %index
-  %gep = getelementptr i16, ptr %ptr, i64 8
+  %ptr = getelementptr inbounds i16, ptr %in, i64 %index
+  %gep = getelementptr inbounds i16, ptr %ptr, i64 8
   %val = load atomic i16, ptr %gep seq_cst, align 2
   store i16 %val, ptr %out
   ret void
@@ -8674,7 +8674,7 @@ define amdgpu_kernel void @atomic_store_i16_offset(i16 %in, ptr %out) {
 ; GFX11-FAKE16-NEXT:    flat_store_b16 v[0:1], v2 offset:16
 ; GFX11-FAKE16-NEXT:    s_endpgm
 entry:
-  %gep = getelementptr i16, ptr %out, i64 8
+  %gep = getelementptr inbounds i16, ptr %out, i64 8
   store atomic i16 %in, ptr %gep  seq_cst, align 2
   ret void
 }
@@ -8816,8 +8816,8 @@ define amdgpu_kernel void @atomic_store_i16_addr64_offset(i16 %in, ptr %out, i64
 ; GFX11-FAKE16-NEXT:    flat_store_b16 v[0:1], v2 offset:16
 ; GFX11-FAKE16-NEXT:    s_endpgm
 entry:
-  %ptr = getelementptr i16, ptr %out, i64 %index
-  %gep = getelementptr i16, ptr %ptr, i64 8
+  %ptr = getelementptr inbounds i16, ptr %out, i64 %index
+  %gep = getelementptr inbounds i16, ptr %ptr, i64 8
   store atomic i16 %in, ptr %gep seq_cst, align 2
   ret void
 }
@@ -8882,7 +8882,7 @@ define amdgpu_kernel void @atomic_store_f16_offset(half %in, ptr %out) {
 ; GFX11-FAKE16-NEXT:    flat_store_b16 v[0:1], v2 offset:16
 ; GFX11-FAKE16-NEXT:    s_endpgm
 entry:
-  %gep = getelementptr half, ptr %out, i64 8
+  %gep = getelementptr inbounds half, ptr %out, i64 8
   store atomic half %in, ptr %gep  seq_cst, align 2
   ret void
 }
@@ -9002,7 +9002,7 @@ define amdgpu_kernel void @atomic_store_bf16_offset(bfloat %in, ptr %out) {
 ; GFX11-FAKE16-NEXT:    v_mov_b32_e32 v2, s2
 ; GFX11-FAKE16-NEXT:    flat_store_b16 v[0:1], v2
 ; GFX11-FAKE16-NEXT:    s_endpgm
-  %gep = getelementptr bfloat, ptr %out, i64 8
+  %gep = getelementptr inbounds bfloat, ptr %out, i64 8
   store atomic bfloat %in, ptr %out seq_cst, align 2
   ret void
 }
@@ -9125,7 +9125,7 @@ define amdgpu_kernel void @atomic_inc_i32_offset(ptr %out, i32 %in) {
 ; GFX11-NEXT:    buffer_gl0_inv
 ; GFX11-NEXT:    s_endpgm
 entry:
-  %gep = getelementptr i32, ptr %out, i32 4
+  %gep = getelementptr inbounds i32, ptr %out, i32 4
   %val = atomicrmw volatile uinc_wrap ptr %gep, i32 %in syncscope("agent") seq_cst, !amdgpu.no.fine.grained.memory !0
   ret void
 }
@@ -9189,7 +9189,7 @@ define amdgpu_kernel void @atomic_inc_i32_max_offset(ptr %out, i32 %in) {
 ; GFX11-NEXT:    buffer_gl0_inv
 ; GFX11-NEXT:    s_endpgm
 entry:
-  %gep = getelementptr i32, ptr %out, i32 1023
+  %gep = getelementptr inbounds i32, ptr %out, i32 1023
   %val = atomicrmw volatile uinc_wrap ptr %gep, i32 %in syncscope("agent") seq_cst, !amdgpu.no.fine.grained.memory !0
   ret void
 }
@@ -9257,7 +9257,7 @@ define amdgpu_kernel void @atomic_inc_i32_max_offset_p1(ptr %out, i32 %in) {
 ; GFX11-NEXT:    buffer_gl0_inv
 ; GFX11-NEXT:    s_endpgm
 entry:
-  %gep = getelementptr i32, ptr %out, i32 1024
+  %gep = getelementptr inbounds i32, ptr %out, i32 1024
   %val = atomicrmw volatile uinc_wrap ptr %gep, i32 %in syncscope("agent") seq_cst, !amdgpu.no.fine.grained.memory !0
   ret void
 }
@@ -9331,7 +9331,7 @@ define amdgpu_kernel void @atomic_inc_i32_ret_offset(ptr %out, ptr %out2, i32 %i
 ; GFX11-NEXT:    flat_store_b32 v[0:1], v2
 ; GFX11-NEXT:    s_endpgm
 entry:
-  %gep = getelementptr i32, ptr %out, i32 4
+  %gep = getelementptr inbounds i32, ptr %out, i32 4
   %val = atomicrmw volatile uinc_wrap ptr %gep, i32 %in syncscope("agent") seq_cst, !amdgpu.no.fine.grained.memory !0
   store i32 %val, ptr %out2
   ret void
@@ -9413,8 +9413,8 @@ define amdgpu_kernel void @atomic_inc_i32_incr64_offset(ptr %out, i32 %in, i64 %
 ; GFX11-NEXT:    buffer_gl0_inv
 ; GFX11-NEXT:    s_endpgm
 entry:
-  %ptr = getelementptr i32, ptr %out, i64 %index
-  %gep = getelementptr i32, ptr %ptr, i32 4
+  %ptr = getelementptr inbounds i32, ptr %out, i64 %index
+  %gep = getelementptr inbounds i32, ptr %ptr, i32 4
   %val = atomicrmw volatile uinc_wrap ptr %gep, i32 %in syncscope("agent") seq_cst, !amdgpu.no.fine.grained.memory !0
   ret void
 }
@@ -9505,8 +9505,8 @@ define amdgpu_kernel void @atomic_inc_i32_ret_incr64_offset(ptr %out, ptr %out2,
 ; GFX11-NEXT:    flat_store_b32 v[0:1], v2
 ; GFX11-NEXT:    s_endpgm
 entry:
-  %ptr = getelementptr i32, ptr %out, i64 %index
-  %gep = getelementptr i32, ptr %ptr, i32 4
+  %ptr = getelementptr inbounds i32, ptr %out, i64 %index
+  %gep = getelementptr inbounds i32, ptr %ptr, i32 4
   %val = atomicrmw volatile uinc_wrap ptr %gep, i32 %in syncscope("agent") seq_cst, !amdgpu.no.fine.grained.memory !0
   store i32 %val, ptr %out2
   ret void
@@ -9713,7 +9713,7 @@ define amdgpu_kernel void @atomic_inc_i32_incr64(ptr %out, i32 %in, i64 %index)
 ; GFX11-NEXT:    buffer_gl0_inv
 ; GFX11-NEXT:    s_endpgm
 entry:
-  %ptr = getelementptr i32, ptr %out, i64 %index
+  %ptr = getelementptr inbounds i32, ptr %out, i64 %index
   %val = atomicrmw volatile uinc_wrap ptr %ptr, i32 %in syncscope("agent") seq_cst, !amdgpu.no.fine.grained.memory !0
   ret void
 }
@@ -9800,7 +9800,7 @@ define amdgpu_kernel void @atomic_inc_i32_ret_incr64(ptr %out, ptr %out2, i32 %i
 ; GFX11-NEXT:    flat_store_b32 v[0:1], v2
 ; GFX11-NEXT:    s_endpgm
 entry:
-  %ptr = getelementptr i32, ptr %out, i64 %index
+  %ptr = getelementptr inbounds i32, ptr %out, i64 %index
   %val = atomicrmw volatile uinc_wrap ptr %ptr, i32 %in syncscope("agent") seq_cst, !amdgpu.no.fine.grained.memory !0
   store i32 %val, ptr %out2
   ret void
@@ -9865,7 +9865,7 @@ define amdgpu_kernel void @atomic_dec_i32_offset(ptr %out, i32 %in) {
 ; GFX11-NEXT:    buffer_gl0_inv
 ; GFX11-NEXT:    s_endpgm
 entry:
-  %gep = getelementptr i32, ptr %out, i32 4
+  %gep = getelementptr inbounds i32, ptr %out, i32 4
   %val = atomicrmw volatile udec_wrap ptr %gep, i32 %in syncscope("agent") seq_cst, !amdgpu.no.fine.grained.memory !0
   ret void
 }
@@ -9929,7 +9929,7 @@ define amdgpu_kernel void @atomic_dec_i32_max_offset(ptr %out, i32 %in) {
 ; GFX11-NEXT:    buffer_gl0_inv
 ; GFX11-NEXT:    s_endpgm
 entry:
-  %gep = getelementptr i32, ptr %out, i32 1023
+  %gep = getelementptr inbounds i32, ptr %out, i32 1023
   %val = atomicrmw volatile udec_wrap ptr %gep, i32 %in syncscope("agent") seq_cst, !amdgpu.no.fine.grained.memory !0
   ret void
 }
@@ -9997,7 +9997,7 @@ define amdgpu_kernel void @atomic_dec_i32_max_offset_p1(ptr %out, i32 %in) {
 ; GFX11-NEXT:    buffer_gl0_inv
 ; GFX11-NEXT:    s_endpgm
 entry:
-  %gep = getelementptr i32, ptr %out, i32 1024
+  %gep = getelementptr inbounds i32, ptr %out, i32 1024
   %val = atomicrmw volatile udec_wrap ptr %gep, i32 %in syncscope("agent") seq_cst, !amdgpu.no.fine.grained.memory !0
   ret void
 }
@@ -10071,7 +10071,7 @@ define amdgpu_kernel void @atomic_dec_i32_ret_offset(ptr %out, ptr %out2, i32 %i
 ; GFX11-NEXT:    flat_store_b32 v[0:1], v2
 ; GFX11-NEXT:    s_endpgm
 entry:
-  %gep = getelementptr i32, ptr %out, i32 4
+  %gep = getelementptr inbounds i32, ptr %out, i32 4
   %val = atomicrmw volatile udec_wrap ptr %gep, i32 %in syncscope("agent") seq_cst, !amdgpu.no.fine.grained.memory !0
   store i32 %val, ptr %out2
   ret void
@@ -10153,8 +10153,8 @@ define amdgpu_kernel void @atomic_dec_i32_decr64_offset(ptr %out, i32 %in, i64 %
 ; GFX11-NEXT:    buffer_gl0_inv
 ; GFX11-NEXT:    s_endpgm
 entry:
-  %ptr = getelementptr i32, ptr %out, i64 %index
-  %gep = getelementptr i32, ptr %ptr, i32 4
+  %ptr = getelementptr inbounds i32, ptr %out, i64 %index
+  %gep = getelementptr inbounds i32, ptr %ptr, i32 4
   %val = atomicrmw volatile udec_wrap ptr %gep, i32 %in syncscope("agent") seq_cst, !amdgpu.no.fine.grained.memory !0
   ret void
 }
@@ -10245,8 +10245,8 @@ define amdgpu_kernel void @atomic_dec_i32_ret_decr64_offset(ptr %out, ptr %out2,
 ; GFX11-NEXT:    flat_store_b32 v[0:1], v2
 ; GFX11-NEXT:    s_endpgm
 entry:
-  %ptr = getelementptr i32, ptr %out, i64 %index
-  %gep = getelementptr i32, ptr %ptr, i32 4
+  %ptr = getelementptr inbounds i32, ptr %out, i64 %index
+  %gep = getelementptr inbounds i32, ptr %ptr, i32 4
   %val = atomicrmw volatile udec_wrap ptr %gep, i32 %in syncscope("agent") seq_cst, !amdgpu.no.fine.grained.memory !0
   store i32 %val, ptr %out2
   ret void
@@ -10453,7 +10453,7 @@ define amdgpu_kernel void @atomic_dec_i32_decr64(ptr %out, i32 %in, i64 %index)
 ; GFX11-NEXT:    buffer_gl0_inv
 ; GFX11-NEXT:    s_endpgm
 entry:
-  %ptr = getelementptr i32, ptr %out, i64 %index
+  %ptr = getelementptr inbounds i32, ptr %out, i64 %index
   %val = atomicrmw volatile udec_wrap ptr %ptr, i32 %in syncscope("agent") seq_cst, !amdgpu.no.fine.grained.memory !0
   ret void
 }
@@ -10540,7 +10540,7 @@ define amdgpu_kernel void @atomic_dec_i32_ret_decr64(ptr %out, ptr %out2, i32 %i
 ; GFX11-NEXT:    flat_store_b32 v[0:1], v2
 ; GFX11-NEXT:    s_endpgm
 entry:
-  %ptr = getelementptr i32, ptr %out, i64 %index
+  %ptr = getelementptr inbounds i32, ptr %out, i64 %index
   %val = atomicrmw volatile udec_wrap ptr %ptr, i32 %in syncscope("agent") seq_cst, !amdgpu.no.fine.grained.memory !0
   store i32 %val, ptr %out2
   ret void
@@ -10619,7 +10619,7 @@ define amdgpu_kernel void @atomic_load_f16_offset(ptr %in, ptr %out) {
 ; GFX11-FAKE16-NEXT:    v_dual_mov_b32 v0, s2 :: v_dual_mov_b32 v1, s3
 ; GFX11-FAKE16-NEXT:    flat_store_b16 v[0:1], v2
 ; GFX11-FAKE16-NEXT:    s_endpgm
-  %gep = getelementptr half, ptr %in, i64 8
+  %gep = getelementptr inbounds half, ptr %in, i64 8
   %val = load atomic half, ptr %gep  seq_cst, align 2
   store half %val, ptr %out
   ret void
@@ -10772,7 +10772,7 @@ define amdgpu_kernel void @atomic_load_bf16_offset(ptr %in, ptr %out) {
 ; GFX11-FAKE16-NEXT:    v_dual_mov_b32 v0, s2 :: v_dual_mov_b32 v1, s3
 ; GFX11-FAKE16-NEXT:    flat_store_b16 v[0:1], v2
 ; GFX11-FAKE16-NEXT:    s_endpgm
-  %gep = getelementptr bfloat, ptr %in, i64 8
+  %gep = getelementptr inbounds bfloat, ptr %in, i64 8
   %val = load atomic bfloat, ptr %gep  seq_cst, align 2
   store bfloat %val, ptr %out
   ret void
diff --git a/llvm/test/CodeGen/AMDGPU/flat_atomics_i32_system.ll b/llvm/test/CodeGen/AMDGPU/flat_atomics_i32_system.ll
index 47161954cc332..45f9d9e774079 100644
--- a/llvm/test/CodeGen/AMDGPU/flat_atomics_i32_system.ll
+++ b/llvm/test/CodeGen/AMDGPU/flat_atomics_i32_system.ll
@@ -63,7 +63,7 @@ define void @flat_atomic_xchg_i32_noret_offset(ptr %out, i32 %in) {
 ; GCN3-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GCN3-NEXT:    buffer_wbinvl1_vol
 ; GCN3-NEXT:    s_setpc_b64 s[30:31]
-  %gep = getelementptr i32, ptr %out, i32 4
+  %gep = getelementptr inbounds i32, ptr %out, i32 4
   %tmp0 = atomicrmw xchg ptr %gep, i32 %in seq_cst
   ret void
 }
@@ -124,7 +124,7 @@ define i32 @flat_atomic_xchg_i32_ret_offset(ptr %out, i32 %in) {
 ; GCN3-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GCN3-NEXT:    buffer_wbinvl1_vol
 ; GCN3-NEXT:    s_setpc_b64 s[30:31]
-  %gep = getelementptr i32, ptr %out, i32 4
+  %gep = getelementptr inbounds i32, ptr %out, i32 4
   %result = atomicrmw xchg ptr %gep, i32 %in seq_cst
   ret i32 %result
 }
@@ -203,7 +203,7 @@ define amdgpu_gfx void @flat_atomic_xchg_i32_noret_offset_scalar(ptr inreg %out,
 ; GCN3-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GCN3-NEXT:    buffer_wbinvl1_vol
 ; GCN3-NEXT:    s_setpc_b64 s[30:31]
-  %gep = getelementptr i32, ptr %out, i32 4
+  %gep = getelementptr inbounds i32, ptr %out, i32 4
   %tmp0 = atomicrmw xchg ptr %gep, i32 %in seq_cst
   ret void
 }
@@ -282,7 +282,7 @@ define amdgpu_gfx i32 @flat_atomic_xchg_i32_ret_offset_scalar(ptr inreg %out, i3
 ; GCN3-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GCN3-NEXT:    buffer_wbinvl1_vol
 ; GCN3-NEXT:    s_setpc_b64 s[30:31]
-  %gep = getelementptr i32, ptr %out, i32 4
+  %gep = getelementptr inbounds i32, ptr %out, i32 4
   %result = atomicrmw xchg ptr %gep, i32 %in seq_cst
   ret i32 %result
 }
@@ -315,7 +315,7 @@ define void @flat_atomic_xchg_i32_noret_offset__amdgpu_no_remote_memory(ptr %out
 ; GCN3-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GCN3-NEXT:    buffer_wbinvl1_vol
 ; GCN3-NEXT:    s_setpc_b64 s[30:31]
-  %gep = getelementptr i32, ptr %out, i64 4
+  %gep = getelementptr inbounds i32, ptr %out, i64 4
   %tmp0 = atomicrmw xchg ptr %gep, i32 %in seq_cst, !amdgpu.no.remote.memory !0
   ret void
 }
@@ -348,7 +348,7 @@ define i32 @flat_atomic_xchg_i32_ret_offset__amdgpu_no_remote_memory(ptr %out, i
 ; GCN3-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GCN3-NEXT:    buffer_wbinvl1_vol
 ; GCN3-NEXT:    s_setpc_b64 s[30:31]
-  %gep = getelementptr i32, ptr %out, i64 4
+  %gep = getelementptr inbounds i32, ptr %out, i64 4
   %result = atomicrmw xchg ptr %gep, i32 %in seq_cst, !amdgpu.no.remote.memory !0
   ret i32 %result
 }
@@ -413,7 +413,7 @@ define void @flat_atomic_xchg_f32_noret_offset(ptr %out, float %in) {
 ; GCN3-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GCN3-NEXT:    buffer_wbinvl1_vol
 ; GCN3-NEXT:    s_setpc_b64 s[30:31]
-  %gep = getelementptr float, ptr %out, i32 4
+  %gep = getelementptr inbounds float, ptr %out, i32 4
   %tmp0 = atomicrmw xchg ptr %gep, float %in seq_cst
   ret void
 }
@@ -474,7 +474,7 @@ define float @flat_atomic_xchg_f32_ret_offset(ptr %out, float %in) {
 ; GCN3-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GCN3-NEXT:    buffer_wbinvl1_vol
 ; GCN3-NEXT:    s_setpc_b64 s[30:31]
-  %gep = getelementptr float, ptr %out, i32 4
+  %gep = getelementptr inbounds float, ptr %out, i32 4
   %result = atomicrmw xchg ptr %gep, float %in seq_cst
   ret float %result
 }
@@ -553,7 +553,7 @@ define amdgpu_gfx void @flat_atomic_xchg_f32_noret_offset_scalar(ptr inreg %out,
 ; GCN3-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GCN3-NEXT:    buffer_wbinvl1_vol
 ; GCN3-NEXT:    s_setpc_b64 s[30:31]
-  %gep = getelementptr float, ptr %out, i32 4
+  %gep = getelementptr inbounds float, ptr %out, i32 4
   %tmp0 = atomicrmw xchg ptr %gep, float %in seq_cst
   ret void
 }
@@ -632,7 +632,7 @@ define amdgpu_gfx float @flat_atomic_xchg_f32_ret_offset_scalar(ptr inreg %out,
 ; GCN3-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GCN3-NEXT:    buffer_wbinvl1_vol
 ; GCN3-NEXT:    s_setpc_b64 s[30:31]
-  %gep = getelementptr float, ptr %out, i32 4
+  %gep = getelementptr inbounds float, ptr %out, i32 4
   %result = atomicrmw xchg ptr %gep, float %in seq_cst
   ret float %result
 }
@@ -665,7 +665,7 @@ define void @flat_atomic_xchg_f32_noret_offset__amdgpu_no_remote_memory(ptr %out
 ; GCN3-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GCN3-NEXT:    buffer_wbinvl1_vol
 ; GCN3-NEXT:    s_setpc_b64 s[30:31]
-  %gep = getelementptr float, ptr %out, i64 4
+  %gep = getelementptr inbounds float, ptr %out, i64 4
   %tmp0 = atomicrmw xchg ptr %gep, float %in seq_cst, !amdgpu.no.remote.memory !0
   ret void
 }
@@ -698,7 +698,7 @@ define float @flat_atomic_xchg_f32_ret_offset__amdgpu_no_remote_memory(ptr %out,
 ; GCN3-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GCN3-NEXT:    buffer_wbinvl1_vol
 ; GCN3-NEXT:    s_setpc_b64 s[30:31]
-  %gep = getelementptr float, ptr %out, i64 4
+  %gep = getelementptr inbounds float, ptr %out, i64 4
   %result = atomicrmw xchg ptr %gep, float %in seq_cst, !amdgpu.no.remote.memory !0
   ret float %result
 }
@@ -763,7 +763,7 @@ define void @flat_atomic_add_i32_noret_offset(ptr %out, i32 %in) {
 ; GCN3-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GCN3-NEXT:    buffer_wbinvl1_vol
 ; GCN3-NEXT:    s_setpc_b64 s[30:31]
-  %gep = getelementptr i32, ptr %out, i32 4
+  %gep = getelementptr inbounds i32, ptr %out, i32 4
   %tmp0 = atomicrmw add ptr %gep, i32 %in seq_cst
   ret void
 }
@@ -824,7 +824,7 @@ define i32 @flat_atomic_add_i32_ret_offset(ptr %out, i32 %in) {
 ; GCN3-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GCN3-NEXT:    buffer_wbinvl1_vol
 ; GCN3-NEXT:    s_setpc_b64 s[30:31]
-  %gep = getelementptr i32, ptr %out, i32 4
+  %gep = getelementptr inbounds i32, ptr %out, i32 4
   %result = atomicrmw add ptr %gep, i32 %in seq_cst
   ret i32 %result
 }
@@ -903,7 +903,7 @@ define amdgpu_gfx void @flat_atomic_add_i32_noret_offset_scalar(ptr inreg %out,
 ; GCN3-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GCN3-NEXT:    buffer_wbinvl1_vol
 ; GCN3-NEXT:    s_setpc_b64 s[30:31]
-  %gep = getelementptr i32, ptr %out, i32 4
+  %gep = getelementptr inbounds i32, ptr %out, i32 4
   %tmp0 = atomicrmw add ptr %gep, i32 %in seq_cst
   ret void
 }
@@ -982,7 +982,7 @@ define amdgpu_gfx i32 @flat_atomic_add_i32_ret_offset_scalar(ptr inreg %out, i32
 ; GCN3-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GCN3-NEXT:    buffer_wbinvl1_vol
 ; GCN3-NEXT:    s_setpc_b64 s[30:31]
-  %gep = getelementptr i32, ptr %out, i32 4
+  %gep = getelementptr inbounds i32, ptr %out, i32 4
   %result = atomicrmw add ptr %gep, i32 %in seq_cst
   ret i32 %result
 }
@@ -1015,7 +1015,7 @@ define void @flat_atomic_add_i32_noret_offset__amdgpu_no_remote_memory(ptr %out,
 ; GCN3-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GCN3-NEXT:    buffer_wbinvl1_vol
 ; GCN3-NEXT:    s_setpc_b64 s[30:31]
-  %gep = getelementptr i32, ptr %out, i64 4
+  %gep = getelementptr inbounds i32, ptr %out, i64 4
   %tmp0 = atomicrmw add ptr %gep, i32 %in seq_cst, !amdgpu.no.remote.memory !0
   ret void
 }
@@ -1048,7 +1048,7 @@ define i32 @flat_atomic_add_i32_ret_offset__amdgpu_no_remote_memory(ptr %out, i3
 ; GCN3-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GCN3-NEXT:    buffer_wbinvl1_vol
 ; GCN3-NEXT:    s_setpc_b64 s[30:31]
-  %gep = getelementptr i32, ptr %out, i64 4
+  %gep = getelementptr inbounds i32, ptr %out, i64 4
   %result = atomicrmw add ptr %gep, i32 %in seq_cst, !amdgpu.no.remote.memory !0
   ret i32 %result
 }
@@ -1191,7 +1191,7 @@ define void @flat_atomic_sub_i32_noret_offset(ptr %out, i32 %in) {
 ; GCN3-NEXT:  ; %bb.2: ; %atomicrmw.end
 ; GCN3-NEXT:    s_or_b64 exec, exec, s[4:5]
 ; GCN3-NEXT:    s_setpc_b64 s[30:31]
-  %gep = getelementptr i32, ptr %out, i32 4
+  %gep = getelementptr inbounds i32, ptr %out, i32 4
   %tmp0 = atomicrmw sub ptr %gep, i32 %in seq_cst
   ret void
 }
@@ -1334,7 +1334,7 @@ define i32 @flat_atomic_sub_i32_ret_offset(ptr %out, i32 %in) {
 ; GCN3-NEXT:    s_or_b64 exec, exec, s[4:5]
 ; GCN3-NEXT:    v_mov_b32_e32 v0, v3
 ; GCN3-NEXT:    s_setpc_b64 s[30:31]
-  %gep = getelementptr i32, ptr %out, i32 4
+  %gep = getelementptr inbounds i32, ptr %out, i32 4
   %result = atomicrmw sub ptr %gep, i32 %in seq_cst
   ret i32 %result
 }
@@ -1485,7 +1485,7 @@ define amdgpu_gfx void @flat_atomic_sub_i32_noret_offset_scalar(ptr inreg %out,
 ; GCN3-NEXT:  ; %bb.2: ; %atomicrmw.end
 ; GCN3-NEXT:    s_or_b64 exec, exec, s[34:35]
 ; GCN3-NEXT:    s_setpc_b64 s[30:31]
-  %gep = getelementptr i32, ptr %out, i32 4
+  %gep = getelementptr inbounds i32, ptr %out, i32 4
   %tmp0 = atomicrmw sub ptr %gep, i32 %in seq_cst
   ret void
 }
@@ -1644,7 +1644,7 @@ define amdgpu_gfx i32 @flat_atomic_sub_i32_ret_offset_scalar(ptr inreg %out, i32
 ; GCN3-NEXT:  ; %bb.2: ; %atomicrmw.end
 ; GCN3-NEXT:    s_or_b64 exec, exec, s[34:35]
 ; GCN3-NEXT:    s_setpc_b64 s[30:31]
-  %gep = getelementptr i32, ptr %out, i32 4
+  %gep = getelementptr inbounds i32, ptr %out, i32 4
   %result = atomicrmw sub ptr %gep, i32 %in seq_cst
   ret i32 %result
 }
@@ -1677,7 +1677,7 @@ define void @flat_atomic_sub_i32_noret_offset__amdgpu_no_remote_memory(ptr %out,
 ; GCN3-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GCN3-NEXT:    buffer_wbinvl1_vol
 ; GCN3-NEXT:    s_setpc_b64 s[30:31]
-  %gep = getelementptr i32, ptr %out, i64 4
+  %gep = getelementptr inbounds i32, ptr %out, i64 4
   %tmp0 = atomicrmw sub ptr %gep, i32 %in seq_cst, !amdgpu.no.remote.memory !0
   ret void
 }
@@ -1710,7 +1710,7 @@ define i32 @flat_atomic_sub_i32_ret_offset__amdgpu_no_remote_memory(ptr %out, i3
 ; GCN3-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GCN3-NEXT:    buffer_wbinvl1_vol
 ; GCN3-NEXT:    s_setpc_b64 s[30:31]
-  %gep = getelementptr i32, ptr %out, i64 4
+  %gep = getelementptr inbounds i32, ptr %out, i64 4
   %result = atomicrmw sub ptr %gep, i32 %in seq_cst, !amdgpu.no.remote.memory !0
   ret i32 %result
 }
@@ -1853,7 +1853,7 @@ define void @flat_atomic_and_i32_noret_offset(ptr %out, i32 %in) {
 ; GCN3-NEXT:  ; %bb.2: ; %atomicrmw.end
 ; GCN3-NEXT:    s_or_b64 exec, exec, s[4:5]
 ; GCN3-NEXT:    s_setpc_b64 s[30:31]
-  %gep = getelementptr i32, ptr %out, i32 4
+  %gep = getelementptr inbounds i32, ptr %out, i32 4
   %tmp0 = atomicrmw and ptr %gep, i32 %in seq_cst
   ret void
 }
@@ -1996,7 +1996,7 @@ define i32 @flat_atomic_and_i32_ret_offset(ptr %out, i32 %in) {
 ; GCN3-NEXT:    s_or_b64 exec, exec, s[4:5]
 ; GCN3-NEXT:    v_mov_b32_e32 v0, v3
 ; GCN3-NEXT:    s_setpc_b64 s[30:31]
-  %gep = getelementptr i32, ptr %out, i32 4
+  %gep = getelementptr inbounds i32, ptr %out, i32 4
   %result = atomicrmw and ptr %gep, i32 %in seq_cst
   ret i32 %result
 }
@@ -2147,7 +2147,7 @@ define amdgpu_gfx void @flat_atomic_and_i32_noret_offset_scalar(ptr inreg %out,
 ; GCN3-NEXT:  ; %bb.2: ; %atomicrmw.end
 ; GCN3-NEXT:    s_or_b64 exec, exec, s[34:35]
 ; GCN3-NEXT:    s_setpc_b64 s[30:31]
-  %gep = getelementptr i32, ptr %out, i32 4
+  %gep = getelementptr inbounds i32, ptr %out, i32 4
   %tmp0 = atomicrmw and ptr %gep, i32 %in seq_cst
   ret void
 }
@@ -2306,7 +2306,7 @@ define amdgpu_gfx i32 @flat_atomic_and_i32_ret_offset_scalar(ptr inreg %out, i32
 ; GCN3-NEXT:  ; %bb.2: ; %atomicrmw.end
 ; GCN3-NEXT:    s_or_b64 exec, exec, s[34:35]
 ; GCN3-NEXT:    s_setpc_b64 s[30:31]
-  %gep = getelementptr i32, ptr %out, i32 4
+  %gep = getelementptr inbounds i32, ptr %out, i32 4
   %result = atomicrmw and ptr %gep, i32 %in seq_cst
   ret i32 %result
 }
@@ -2339,7 +2339,7 @@ define void @flat_atomic_and_i32_noret_offset__amdgpu_no_remote_memory(ptr %out,
 ; GCN3-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GCN3-NEXT:    buffer_wbinvl1_vol
 ; GCN3-NEXT:    s_setpc_b64 s[30:31]
-  %gep = getelementptr i32, ptr %out, i64 4
+  %gep = getelementptr inbounds i32, ptr %out, i64 4
   %tmp0 = atomicrmw and ptr %gep, i32 %in seq_cst, !amdgpu.no.remote.memory !0
   ret void
 }
@@ -2372,7 +2372,7 @@ define i32 @flat_atomic_and_i32_ret_offset__amdgpu_no_remote_memory(ptr %out, i3
 ; GCN3-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GCN3-NEXT:    buffer_wbinvl1_vol
 ; GCN3-NEXT:    s_setpc_b64 s[30:31]
-  %gep = getelementptr i32, ptr %out, i64 4
+  %gep = getelementptr inbounds i32, ptr %out, i64 4
   %result = atomicrmw and ptr %gep, i32 %in seq_cst, !amdgpu.no.remote.memory !0
   ret i32 %result
 }
@@ -2521,7 +2521,7 @@ define void @flat_atomic_nand_i32_noret_offset(ptr %out, i32 %in) {
 ; GCN3-NEXT:  ; %bb.2: ; %atomicrmw.end
 ; GCN3-NEXT:    s_or_b64 exec, exec, s[4:5]
 ; GCN3-NEXT:    s_setpc_b64 s[30:31]
-  %gep = getelementptr i32, ptr %out, i32 4
+  %gep = getelementptr inbounds i32, ptr %out, i32 4
   %tmp0 = atomicrmw nand ptr %gep, i32 %in seq_cst
   ret void
 }
@@ -2670,7 +2670,7 @@ define i32 @flat_atomic_nand_i32_ret_offset(ptr %out, i32 %in) {
 ; GCN3-NEXT:    s_or_b64 exec, exec, s[4:5]
 ; GCN3-NEXT:    v_mov_b32_e32 v0, v3
 ; GCN3-NEXT:    s_setpc_b64 s[30:31]
-  %gep = getelementptr i32, ptr %out, i32 4
+  %gep = getelementptr inbounds i32, ptr %out, i32 4
   %result = atomicrmw nand ptr %gep, i32 %in seq_cst
   ret i32 %result
 }
@@ -2827,7 +2827,7 @@ define amdgpu_gfx void @flat_atomic_nand_i32_noret_offset_scalar(ptr inreg %out,
 ; GCN3-NEXT:  ; %bb.2: ; %atomicrmw.end
 ; GCN3-NEXT:    s_or_b64 exec, exec, s[34:35]
 ; GCN3-NEXT:    s_setpc_b64 s[30:31]
-  %gep = getelementptr i32, ptr %out, i32 4
+  %gep = getelementptr inbounds i32, ptr %out, i32 4
   %tmp0 = atomicrmw nand ptr %gep, i32 %in seq_cst
   ret void
 }
@@ -2992,7 +2992,7 @@ define amdgpu_gfx i32 @flat_atomic_nand_i32_ret_offset_scalar(ptr inreg %out, i3
 ; GCN3-NEXT:  ; %bb.2: ; %atomicrmw.end
 ; GCN3-NEXT:    s_or_b64 exec, exec, s[34:35]
 ; GCN3-NEXT:    s_setpc_b64 s[30:31]
-  %gep = getelementptr i32, ptr %out, i32 4
+  %gep = getelementptr inbounds i32, ptr %out, i32 4
   %result = atomicrmw nand ptr %gep, i32 %in seq_cst
   ret i32 %result
 }
@@ -3067,7 +3067,7 @@ define void @flat_atomic_nand_i32_noret_offset__amdgpu_no_remote_memory(ptr %out
 ; GCN3-NEXT:  ; %bb.2: ; %atomicrmw.end
 ; GCN3-NEXT:    s_or_b64 exec, exec, s[4:5]
 ; GCN3-NEXT:    s_setpc_b64 s[30:31]
-  %gep = getelementptr i32, ptr %out, i64 4
+  %gep = getelementptr inbounds i32, ptr %out, i64 4
   %tmp0 = atomicrmw nand ptr %gep, i32 %in seq_cst, !amdgpu.no.remote.memory !0
   ret void
 }
@@ -3143,7 +3143,7 @@ define i32 @flat_atomic_nand_i32_ret_offset__amdgpu_no_remote_memory(ptr %out, i
 ; GCN3-NEXT:    s_or_b64 exec, exec, s[4:5]
 ; GCN3-NEXT:    v_mov_b32_e32 v0, v3
 ; GCN3-NEXT:    s_setpc_b64 s[30:31]
-  %gep = getelementptr i32, ptr %out, i64 4
+  %gep = getelementptr inbounds i32, ptr %out, i64 4
   %result = atomicrmw nand ptr %gep, i32 %in seq_cst, !amdgpu.no.remote.memory !0
   ret i32 %result
 }
@@ -3286,7 +3286,7 @@ define void @flat_atomic_or_i32_noret_offset(ptr %out, i32 %in) {
 ; GCN3-NEXT:  ; %bb.2: ; %atomicrmw.end
 ; GCN3-NEXT:    s_or_b64 exec, exec, s[4:5]
 ; GCN3-NEXT:    s_setpc_b64 s[30:31]
-  %gep = getelementptr i32, ptr %out, i32 4
+  %gep = getelementptr inbounds i32, ptr %out, i32 4
   %tmp0 = atomicrmw or ptr %gep, i32 %in seq_cst
   ret void
 }
@@ -3429,7 +3429,7 @@ define i32 @flat_atomic_or_i32_ret_offset(ptr %out, i32 %in) {
 ; GCN3-NEXT:    s_or_b64 exec, exec, s[4:5]
 ; GCN3-NEXT:    v_mov_b32_e32 v0, v3
 ; GCN3-NEXT:    s_setpc_b64 s[30:31]
-  %gep = getelementptr i32, ptr %out, i32 4
+  %gep = getelementptr inbounds i32, ptr %out, i32 4
   %result = atomicrmw or ptr %gep, i32 %in seq_cst
   ret i32 %result
 }
@@ -3580,7 +3580,7 @@ define amdgpu_gfx void @flat_atomic_or_i32_noret_offset_scalar(ptr inreg %out, i
 ; GCN3-NEXT:  ; %bb.2: ; %atomicrmw.end
 ; GCN3-NEXT:    s_or_b64 exec, exec, s[34:35]
 ; GCN3-NEXT:    s_setpc_b64 s[30:31]
-  %gep = getelementptr i32, ptr %out, i32 4
+  %gep = getelementptr inbounds i32, ptr %out, i32 4
   %tmp0 = atomicrmw or ptr %gep, i32 %in seq_cst
   ret void
 }
@@ -3739,7 +3739,7 @@ define amdgpu_gfx i32 @flat_atomic_or_i32_ret_offset_scalar(ptr inreg %out, i32
 ; GCN3-NEXT:  ; %bb.2: ; %atomicrmw.end
 ; GCN3-NEXT:    s_or_b64 exec, exec, s[34:35]
 ; GCN3-NEXT:    s_setpc_b64 s[30:31]
-  %gep = getelementptr i32, ptr %out, i32 4
+  %gep = getelementptr inbounds i32, ptr %out, i32 4
   %result = atomicrmw or ptr %gep, i32 %in seq_cst
   ret i32 %result
 }
@@ -3772,7 +3772,7 @@ define void @flat_atomic_or_i32_noret_offset__amdgpu_no_remote_memory(ptr %out,
 ; GCN3-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GCN3-NEXT:    buffer_wbinvl1_vol
 ; GCN3-NEXT:    s_setpc_b64 s[30:31]
-  %gep = getelementptr i32, ptr %out, i64 4
+  %gep = getelementptr inbounds i32, ptr %out, i64 4
   %tmp0 = atomicrmw or ptr %gep, i32 %in seq_cst, !amdgpu.no.remote.memory !0
   ret void
 }
@@ -3805,7 +3805,7 @@ define i32 @flat_atomic_or_i32_ret_offset__amdgpu_no_remote_memory(ptr %out, i32
 ; GCN3-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GCN3-NEXT:    buffer_wbinvl1_vol
 ; GCN3-NEXT:    s_setpc_b64 s[30:31]
-  %gep = getelementptr i32, ptr %out, i64 4
+  %gep = getelementptr inbounds i32, ptr %out, i64 4
   %result = atomicrmw or ptr %gep, i32 %in seq_cst, !amdgpu.no.remote.memory !0
   ret i32 %result
 }
@@ -3948,7 +3948,7 @@ define void @flat_atomic_xor_i32_noret_offset(ptr %out, i32 %in) {
 ; GCN3-NEXT:  ; %bb.2: ; %atomicrmw.end
 ; GCN3-NEXT:    s_or_b64 exec, exec, s[4:5]
 ; GCN3-NEXT:    s_setpc_b64 s[30:31]
-  %gep = getelementptr i32, ptr %out, i32 4
+  %gep = getelementptr inbounds i32, ptr %out, i32 4
   %tmp0 = atomicrmw xor ptr %gep, i32 %in seq_cst
   ret void
 }
@@ -4091,7 +4091,7 @@ define i32 @flat_atomic_xor_i32_ret_offset(ptr %out, i32 %in) {
 ; GCN3-NEXT:    s_or_b64 exec, exec, s[4:5]
 ; GCN3-NEXT:    v_mov_b32_e32 v0, v3
 ; GCN3-NEXT:    s_setpc_b64 s[30:31]
-  %gep = getelementptr i32, ptr %out, i32 4
+  %gep = getelementptr inbounds i32, ptr %out, i32 4
   %result = atomicrmw xor ptr %gep, i32 %in seq_cst
   ret i32 %result
 }
@@ -4242,7 +4242,7 @@ define amdgpu_gfx void @flat_atomic_xor_i32_noret_offset_scalar(ptr inreg %out,
 ; GCN3-NEXT:  ; %bb.2: ; %atomicrmw.end
 ; GCN3-NEXT:    s_or_b64 exec, exec, s[34:35]
 ; GCN3-NEXT:    s_setpc_b64 s[30:31]
-  %gep = getelementptr i32, ptr %out, i32 4
+  %gep = getelementptr inbounds i32, ptr %out, i32 4
   %tmp0 = atomicrmw xor ptr %gep, i32 %in seq_cst
   ret void
 }
@@ -4401,7 +4401,7 @@ define amdgpu_gfx i32 @flat_atomic_xor_i32_ret_offset_scalar(ptr inreg %out, i32
 ; GCN3-NEXT:  ; %bb.2: ; %atomicrmw.end
 ; GCN3-NEXT:    s_or_b64 exec, exec, s[34:35]
 ; GCN3-NEXT:    s_setpc_b64 s[30:31]
-  %gep = getelementptr i32, ptr %out, i32 4
+  %gep = getelementptr inbounds i32, ptr %out, i32 4
   %result = atomicrmw xor ptr %gep, i32 %in seq_cst
   ret i32 %result
 }
@@ -4434,7 +4434,7 @@ define void @flat_xor_i32_noret_offset__amdgpu_no_remote_memory(ptr %out, i32 %i
 ; GCN3-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GCN3-NEXT:    buffer_wbinvl1_vol
 ; GCN3-NEXT:    s_setpc_b64 s[30:31]
-  %gep = getelementptr i32, ptr %out, i64 4
+  %gep = getelementptr inbounds i32, ptr %out, i64 4
   %tmp0 = atomicrmw xor ptr %gep, i32 %in seq_cst, !amdgpu.no.remote.memory !0
   ret void
 }
@@ -4467,7 +4467,7 @@ define i32 @flat_atomic_xor_i32_ret_offset__amdgpu_no_remote_memory(ptr %out, i3
 ; GCN3-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GCN3-NEXT:    buffer_wbinvl1_vol
 ; GCN3-NEXT:    s_setpc_b64 s[30:31]
-  %gep = getelementptr i32, ptr %out, i64 4
+  %gep = getelementptr inbounds i32, ptr %out, i64 4
   %result = atomicrmw xor ptr %gep, i32 %in seq_cst, !amdgpu.no.remote.memory !0
   ret i32 %result
 }
@@ -4610,7 +4610,7 @@ define void @flat_atomic_max_i32_noret_offset(ptr %out, i32 %in) {
 ; GCN3-NEXT:  ; %bb.2: ; %atomicrmw.end
 ; GCN3-NEXT:    s_or_b64 exec, exec, s[4:5]
 ; GCN3-NEXT:    s_setpc_b64 s[30:31]
-  %gep = getelementptr i32, ptr %out, i32 4
+  %gep = getelementptr inbounds i32, ptr %out, i32 4
   %tmp0 = atomicrmw max ptr %gep, i32 %in seq_cst
   ret void
 }
@@ -4753,7 +4753,7 @@ define i32 @flat_atomic_max_i32_ret_offset(ptr %out, i32 %in) {
 ; GCN3-NEXT:    s_or_b64 exec, exec, s[4:5]
 ; GCN3-NEXT:    v_mov_b32_e32 v0, v3
 ; GCN3-NEXT:    s_setpc_b64 s[30:31]
-  %gep = getelementptr i32, ptr %out, i32 4
+  %gep = getelementptr inbounds i32, ptr %out, i32 4
   %result = atomicrmw max ptr %gep, i32 %in seq_cst
   ret i32 %result
 }
@@ -4904,7 +4904,7 @@ define amdgpu_gfx void @flat_atomic_max_i32_noret_offset_scalar(ptr inreg %out,
 ; GCN3-NEXT:  ; %bb.2: ; %atomicrmw.end
 ; GCN3-NEXT:    s_or_b64 exec, exec, s[34:35]
 ; GCN3-NEXT:    s_setpc_b64 s[30:31]
-  %gep = getelementptr i32, ptr %out, i32 4
+  %gep = getelementptr inbounds i32, ptr %out, i32 4
   %tmp0 = atomicrmw max ptr %gep, i32 %in seq_cst
   ret void
 }
@@ -5063,7 +5063,7 @@ define amdgpu_gfx i32 @flat_atomic_max_i32_ret_offset_scalar(ptr inreg %out, i32
 ; GCN3-NEXT:  ; %bb.2: ; %atomicrmw.end
 ; GCN3-NEXT:    s_or_b64 exec, exec, s[34:35]
 ; GCN3-NEXT:    s_setpc_b64 s[30:31]
-  %gep = getelementptr i32, ptr %out, i32 4
+  %gep = getelementptr inbounds i32, ptr %out, i32 4
   %result = atomicrmw max ptr %gep, i32 %in seq_cst
   ret i32 %result
 }
@@ -5157,8 +5157,8 @@ define amdgpu_kernel void @atomic_max_i32_addr64_offset(ptr %out, i32 %in, i32 %
 ; GCN3-NEXT:  ; %bb.2: ; %atomicrmw.end
 ; GCN3-NEXT:    s_endpgm
 entry:
-  %ptr = getelementptr i32, ptr %out, i32 %index
-  %gep = getelementptr i32, ptr %ptr, i32 4
+  %ptr = getelementptr inbounds i32, ptr %out, i32 %index
+  %gep = getelementptr inbounds i32, ptr %ptr, i32 4
   %tmp0 = atomicrmw max ptr %gep, i32 %in seq_cst
   ret void
 }
@@ -5267,8 +5267,8 @@ define amdgpu_kernel void @atomic_max_i32_ret_addr64_offset(ptr %out, ptr %out2,
 ; GCN3-NEXT:    flat_store_dword v[0:1], v2
 ; GCN3-NEXT:    s_endpgm
 entry:
-  %ptr = getelementptr i32, ptr %out, i32 %index
-  %gep = getelementptr i32, ptr %ptr, i32 4
+  %ptr = getelementptr inbounds i32, ptr %out, i32 %index
+  %gep = getelementptr inbounds i32, ptr %ptr, i32 4
   %tmp0 = atomicrmw max ptr %gep, i32 %in seq_cst
   store i32 %tmp0, ptr %out2
   ret void
@@ -5359,7 +5359,7 @@ define amdgpu_kernel void @atomic_max_i32_addr64(ptr %out, i32 %in, i32 %index)
 ; GCN3-NEXT:  ; %bb.2: ; %atomicrmw.end
 ; GCN3-NEXT:    s_endpgm
 entry:
-  %ptr = getelementptr i32, ptr %out, i32 %index
+  %ptr = getelementptr inbounds i32, ptr %out, i32 %index
   %tmp0 = atomicrmw max ptr %ptr, i32 %in seq_cst
   ret void
 }
@@ -5464,7 +5464,7 @@ define amdgpu_kernel void @atomic_max_i32_ret_addr64(ptr %out, ptr %out2, i32 %i
 ; GCN3-NEXT:    flat_store_dword v[0:1], v2
 ; GCN3-NEXT:    s_endpgm
 entry:
-  %ptr = getelementptr i32, ptr %out, i32 %index
+  %ptr = getelementptr inbounds i32, ptr %out, i32 %index
   %tmp0 = atomicrmw max ptr %ptr, i32 %in seq_cst
   store i32 %tmp0, ptr %out2
   ret void
@@ -5498,7 +5498,7 @@ define void @flat_max_i32_noret_offset__amdgpu_no_remote_memory(ptr %out, i32 %i
 ; GCN3-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GCN3-NEXT:    buffer_wbinvl1_vol
 ; GCN3-NEXT:    s_setpc_b64 s[30:31]
-  %gep = getelementptr i32, ptr %out, i64 4
+  %gep = getelementptr inbounds i32, ptr %out, i64 4
   %tmp0 = atomicrmw max ptr %gep, i32 %in seq_cst, !amdgpu.no.remote.memory !0
   ret void
 }
@@ -5531,7 +5531,7 @@ define i32 @flat_atomic_max_i32_ret_offset__amdgpu_no_remote_memory(ptr %out, i3
 ; GCN3-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GCN3-NEXT:    buffer_wbinvl1_vol
 ; GCN3-NEXT:    s_setpc_b64 s[30:31]
-  %gep = getelementptr i32, ptr %out, i64 4
+  %gep = getelementptr inbounds i32, ptr %out, i64 4
   %result = atomicrmw max ptr %gep, i32 %in seq_cst, !amdgpu.no.remote.memory !0
   ret i32 %result
 }
@@ -5674,7 +5674,7 @@ define void @flat_atomic_umax_i32_noret_offset(ptr %out, i32 %in) {
 ; GCN3-NEXT:  ; %bb.2: ; %atomicrmw.end
 ; GCN3-NEXT:    s_or_b64 exec, exec, s[4:5]
 ; GCN3-NEXT:    s_setpc_b64 s[30:31]
-  %gep = getelementptr i32, ptr %out, i32 4
+  %gep = getelementptr inbounds i32, ptr %out, i32 4
   %tmp0 = atomicrmw umax ptr %gep, i32 %in seq_cst
   ret void
 }
@@ -5817,7 +5817,7 @@ define i32 @flat_atomic_umax_i32_ret_offset(ptr %out, i32 %in) {
 ; GCN3-NEXT:    s_or_b64 exec, exec, s[4:5]
 ; GCN3-NEXT:    v_mov_b32_e32 v0, v3
 ; GCN3-NEXT:    s_setpc_b64 s[30:31]
-  %gep = getelementptr i32, ptr %out, i32 4
+  %gep = getelementptr inbounds i32, ptr %out, i32 4
   %result = atomicrmw umax ptr %gep, i32 %in seq_cst
   ret i32 %result
 }
@@ -5968,7 +5968,7 @@ define amdgpu_gfx void @flat_atomic_umax_i32_noret_offset_scalar(ptr inreg %out,
 ; GCN3-NEXT:  ; %bb.2: ; %atomicrmw.end
 ; GCN3-NEXT:    s_or_b64 exec, exec, s[34:35]
 ; GCN3-NEXT:    s_setpc_b64 s[30:31]
-  %gep = getelementptr i32, ptr %out, i32 4
+  %gep = getelementptr inbounds i32, ptr %out, i32 4
   %tmp0 = atomicrmw umax ptr %gep, i32 %in seq_cst
   ret void
 }
@@ -6127,7 +6127,7 @@ define amdgpu_gfx i32 @flat_atomic_umax_i32_ret_offset_scalar(ptr inreg %out, i3
 ; GCN3-NEXT:  ; %bb.2: ; %atomicrmw.end
 ; GCN3-NEXT:    s_or_b64 exec, exec, s[34:35]
 ; GCN3-NEXT:    s_setpc_b64 s[30:31]
-  %gep = getelementptr i32, ptr %out, i32 4
+  %gep = getelementptr inbounds i32, ptr %out, i32 4
   %result = atomicrmw umax ptr %gep, i32 %in seq_cst
   ret i32 %result
 }
@@ -6221,8 +6221,8 @@ define amdgpu_kernel void @atomic_umax_i32_addr64_offset(ptr %out, i32 %in, i32
 ; GCN3-NEXT:  ; %bb.2: ; %atomicrmw.end
 ; GCN3-NEXT:    s_endpgm
 entry:
-  %ptr = getelementptr i32, ptr %out, i32 %index
-  %gep = getelementptr i32, ptr %ptr, i32 4
+  %ptr = getelementptr inbounds i32, ptr %out, i32 %index
+  %gep = getelementptr inbounds i32, ptr %ptr, i32 4
   %tmp0 = atomicrmw umax ptr %gep, i32 %in seq_cst
   ret void
 }
@@ -6331,8 +6331,8 @@ define amdgpu_kernel void @atomic_umax_i32_ret_addr64_offset(ptr %out, ptr %out2
 ; GCN3-NEXT:    flat_store_dword v[0:1], v2
 ; GCN3-NEXT:    s_endpgm
 entry:
-  %ptr = getelementptr i32, ptr %out, i32 %index
-  %gep = getelementptr i32, ptr %ptr, i32 4
+  %ptr = getelementptr inbounds i32, ptr %out, i32 %index
+  %gep = getelementptr inbounds i32, ptr %ptr, i32 4
   %tmp0 = atomicrmw umax ptr %gep, i32 %in seq_cst
   store i32 %tmp0, ptr %out2
   ret void
@@ -6438,7 +6438,7 @@ define amdgpu_kernel void @atomic_umax_i32_ret_addr64(ptr %out, ptr %out2, i32 %
 ; GCN3-NEXT:    flat_store_dword v[0:1], v2
 ; GCN3-NEXT:    s_endpgm
 entry:
-  %ptr = getelementptr i32, ptr %out, i32 %index
+  %ptr = getelementptr inbounds i32, ptr %out, i32 %index
   %tmp0 = atomicrmw umax ptr %ptr, i32 %in seq_cst
   store i32 %tmp0, ptr %out2
   ret void
@@ -6472,7 +6472,7 @@ define void @flat_umax_i32_noret_offset__amdgpu_no_remote_memory(ptr %out, i32 %
 ; GCN3-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GCN3-NEXT:    buffer_wbinvl1_vol
 ; GCN3-NEXT:    s_setpc_b64 s[30:31]
-  %gep = getelementptr i32, ptr %out, i64 4
+  %gep = getelementptr inbounds i32, ptr %out, i64 4
   %tmp0 = atomicrmw umax ptr %gep, i32 %in seq_cst, !amdgpu.no.remote.memory !0
   ret void
 }
@@ -6505,7 +6505,7 @@ define i32 @flat_atomic_umax_i32_ret_offset__amdgpu_no_remote_memory(ptr %out, i
 ; GCN3-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GCN3-NEXT:    buffer_wbinvl1_vol
 ; GCN3-NEXT:    s_setpc_b64 s[30:31]
-  %gep = getelementptr i32, ptr %out, i64 4
+  %gep = getelementptr inbounds i32, ptr %out, i64 4
   %result = atomicrmw umax ptr %gep, i32 %in seq_cst, !amdgpu.no.remote.memory !0
   ret i32 %result
 }
@@ -6648,7 +6648,7 @@ define void @flat_atomic_umin_i32_noret_offset(ptr %out, i32 %in) {
 ; GCN3-NEXT:  ; %bb.2: ; %atomicrmw.end
 ; GCN3-NEXT:    s_or_b64 exec, exec, s[4:5]
 ; GCN3-NEXT:    s_setpc_b64 s[30:31]
-  %gep = getelementptr i32, ptr %out, i32 4
+  %gep = getelementptr inbounds i32, ptr %out, i32 4
   %tmp0 = atomicrmw umin ptr %gep, i32 %in seq_cst
   ret void
 }
@@ -6791,7 +6791,7 @@ define i32 @flat_atomic_umin_i32_ret_offset(ptr %out, i32 %in) {
 ; GCN3-NEXT:    s_or_b64 exec, exec, s[4:5]
 ; GCN3-NEXT:    v_mov_b32_e32 v0, v3
 ; GCN3-NEXT:    s_setpc_b64 s[30:31]
-  %gep = getelementptr i32, ptr %out, i32 4
+  %gep = getelementptr inbounds i32, ptr %out, i32 4
   %result = atomicrmw umin ptr %gep, i32 %in seq_cst
   ret i32 %result
 }
@@ -6942,7 +6942,7 @@ define amdgpu_gfx void @flat_atomic_umin_i32_noret_offset_scalar(ptr inreg %out,
 ; GCN3-NEXT:  ; %bb.2: ; %atomicrmw.end
 ; GCN3-NEXT:    s_or_b64 exec, exec, s[34:35]
 ; GCN3-NEXT:    s_setpc_b64 s[30:31]
-  %gep = getelementptr i32, ptr %out, i32 4
+  %gep = getelementptr inbounds i32, ptr %out, i32 4
   %tmp0 = atomicrmw umin ptr %gep, i32 %in seq_cst
   ret void
 }
@@ -7101,7 +7101,7 @@ define amdgpu_gfx i32 @flat_atomic_umin_i32_ret_offset_scalar(ptr inreg %out, i3
 ; GCN3-NEXT:  ; %bb.2: ; %atomicrmw.end
 ; GCN3-NEXT:    s_or_b64 exec, exec, s[34:35]
 ; GCN3-NEXT:    s_setpc_b64 s[30:31]
-  %gep = getelementptr i32, ptr %out, i32 4
+  %gep = getelementptr inbounds i32, ptr %out, i32 4
   %result = atomicrmw umin ptr %gep, i32 %in seq_cst
   ret i32 %result
 }
@@ -7134,7 +7134,7 @@ define void @flat_umin_i32_noret_offset__amdgpu_no_remote_memory(ptr %out, i32 %
 ; GCN3-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GCN3-NEXT:    buffer_wbinvl1_vol
 ; GCN3-NEXT:    s_setpc_b64 s[30:31]
-  %gep = getelementptr i32, ptr %out, i64 4
+  %gep = getelementptr inbounds i32, ptr %out, i64 4
   %tmp0 = atomicrmw umin ptr %gep, i32 %in seq_cst, !amdgpu.no.remote.memory !0
   ret void
 }
@@ -7167,7 +7167,7 @@ define i32 @flat_atomic_umin_i32_ret_offset__amdgpu_no_remote_memory(ptr %out, i
 ; GCN3-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GCN3-NEXT:    buffer_wbinvl1_vol
 ; GCN3-NEXT:    s_setpc_b64 s[30:31]
-  %gep = getelementptr i32, ptr %out, i64 4
+  %gep = getelementptr inbounds i32, ptr %out, i64 4
   %result = atomicrmw umin ptr %gep, i32 %in seq_cst, !amdgpu.no.remote.memory !0
   ret i32 %result
 }
@@ -7310,7 +7310,7 @@ define void @flat_atomic_min_i32_noret_offset(ptr %out, i32 %in) {
 ; GCN3-NEXT:  ; %bb.2: ; %atomicrmw.end
 ; GCN3-NEXT:    s_or_b64 exec, exec, s[4:5]
 ; GCN3-NEXT:    s_setpc_b64 s[30:31]
-  %gep = getelementptr i32, ptr %out, i32 4
+  %gep = getelementptr inbounds i32, ptr %out, i32 4
   %tmp0 = atomicrmw min ptr %gep, i32 %in seq_cst
   ret void
 }
@@ -7453,7 +7453,7 @@ define i32 @flat_atomic_min_i32_ret_offset(ptr %out, i32 %in) {
 ; GCN3-NEXT:    s_or_b64 exec, exec, s[4:5]
 ; GCN3-NEXT:    v_mov_b32_e32 v0, v3
 ; GCN3-NEXT:    s_setpc_b64 s[30:31]
-  %gep = getelementptr i32, ptr %out, i32 4
+  %gep = getelementptr inbounds i32, ptr %out, i32 4
   %result = atomicrmw min ptr %gep, i32 %in seq_cst
   ret i32 %result
 }
@@ -7604,7 +7604,7 @@ define amdgpu_gfx void @flat_atomic_min_i32_noret_offset_scalar(ptr inreg %out,
 ; GCN3-NEXT:  ; %bb.2: ; %atomicrmw.end
 ; GCN3-NEXT:    s_or_b64 exec, exec, s[34:35]
 ; GCN3-NEXT:    s_setpc_b64 s[30:31]
-  %gep = getelementptr i32, ptr %out, i32 4
+  %gep = getelementptr inbounds i32, ptr %out, i32 4
   %tmp0 = atomicrmw min ptr %gep, i32 %in seq_cst
   ret void
 }
@@ -7763,7 +7763,7 @@ define amdgpu_gfx i32 @flat_atomic_min_i32_ret_offset_scalar(ptr inreg %out, i32
 ; GCN3-NEXT:  ; %bb.2: ; %atomicrmw.end
 ; GCN3-NEXT:    s_or_b64 exec, exec, s[34:35]
 ; GCN3-NEXT:    s_setpc_b64 s[30:31]
-  %gep = getelementptr i32, ptr %out, i32 4
+  %gep = getelementptr inbounds i32, ptr %out, i32 4
   %result = atomicrmw min ptr %gep, i32 %in seq_cst
   ret i32 %result
 }
@@ -7857,8 +7857,8 @@ define amdgpu_kernel void @atomic_min_i32_addr64_offset(ptr %out, i32 %in, i32 %
 ; GCN3-NEXT:  ; %bb.2: ; %atomicrmw.end
 ; GCN3-NEXT:    s_endpgm
 entry:
-  %ptr = getelementptr i32, ptr %out, i32 %index
-  %gep = getelementptr i32, ptr %ptr, i32 4
+  %ptr = getelementptr inbounds i32, ptr %out, i32 %index
+  %gep = getelementptr inbounds i32, ptr %ptr, i32 4
   %tmp0 = atomicrmw min ptr %gep, i32 %in seq_cst
   ret void
 }
@@ -7967,8 +7967,8 @@ define amdgpu_kernel void @atomic_min_i32_ret_addr64_offset(ptr %out, ptr %out2,
 ; GCN3-NEXT:    flat_store_dword v[0:1], v2
 ; GCN3-NEXT:    s_endpgm
 entry:
-  %ptr = getelementptr i32, ptr %out, i32 %index
-  %gep = getelementptr i32, ptr %ptr, i32 4
+  %ptr = getelementptr inbounds i32, ptr %out, i32 %index
+  %gep = getelementptr inbounds i32, ptr %ptr, i32 4
   %tmp0 = atomicrmw min ptr %gep, i32 %in seq_cst
   store i32 %tmp0, ptr %out2
   ret void
@@ -8151,7 +8151,7 @@ define amdgpu_kernel void @atomic_min_i32_ret_addr64(ptr %out, ptr %out2, i32 %i
 ; GCN3-NEXT:    flat_store_dword v[0:1], v2
 ; GCN3-NEXT:    s_endpgm
 entry:
-  %ptr = getelementptr i32, ptr %out, i32 %index
+  %ptr = getelementptr inbounds i32, ptr %out, i32 %index
   %tmp0 = atomicrmw min ptr %ptr, i32 %in seq_cst
   store i32 %tmp0, ptr %out2
   ret void
@@ -8185,7 +8185,7 @@ define void @flat_min_i32_noret_offset__amdgpu_no_remote_memory(ptr %out, i32 %i
 ; GCN3-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GCN3-NEXT:    buffer_wbinvl1_vol
 ; GCN3-NEXT:    s_setpc_b64 s[30:31]
-  %gep = getelementptr i32, ptr %out, i64 4
+  %gep = getelementptr inbounds i32, ptr %out, i64 4
   %tmp0 = atomicrmw min ptr %gep, i32 %in seq_cst, !amdgpu.no.remote.memory !0
   ret void
 }
@@ -8218,7 +8218,7 @@ define i32 @flat_atomic_min_i32_ret_offset__amdgpu_no_remote_memory(ptr %out, i3
 ; GCN3-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GCN3-NEXT:    buffer_wbinvl1_vol
 ; GCN3-NEXT:    s_setpc_b64 s[30:31]
-  %gep = getelementptr i32, ptr %out, i64 4
+  %gep = getelementptr inbounds i32, ptr %out, i64 4
   %result = atomicrmw min ptr %gep, i32 %in seq_cst, !amdgpu.no.remote.memory !0
   ret i32 %result
 }
@@ -8373,7 +8373,7 @@ define void @flat_atomic_uinc_wrap_i32_noret_offset(ptr %out, i32 %in) {
 ; GCN3-NEXT:  ; %bb.2: ; %atomicrmw.end
 ; GCN3-NEXT:    s_or_b64 exec, exec, s[4:5]
 ; GCN3-NEXT:    s_setpc_b64 s[30:31]
-  %gep = getelementptr i32, ptr %out, i32 4
+  %gep = getelementptr inbounds i32, ptr %out, i32 4
   %tmp0 = atomicrmw uinc_wrap ptr %gep, i32 %in seq_cst
   ret void
 }
@@ -8528,7 +8528,7 @@ define i32 @flat_atomic_uinc_wrap_i32_ret_offset(ptr %out, i32 %in) {
 ; GCN3-NEXT:    s_or_b64 exec, exec, s[4:5]
 ; GCN3-NEXT:    v_mov_b32_e32 v0, v3
 ; GCN3-NEXT:    s_setpc_b64 s[30:31]
-  %gep = getelementptr i32, ptr %out, i32 4
+  %gep = getelementptr inbounds i32, ptr %out, i32 4
   %result = atomicrmw uinc_wrap ptr %gep, i32 %in seq_cst
   ret i32 %result
 }
@@ -8691,7 +8691,7 @@ define amdgpu_gfx void @flat_atomic_uinc_wrap_i32_noret_offset_scalar(ptr inreg
 ; GCN3-NEXT:  ; %bb.2: ; %atomicrmw.end
 ; GCN3-NEXT:    s_or_b64 exec, exec, s[34:35]
 ; GCN3-NEXT:    s_setpc_b64 s[30:31]
-  %gep = getelementptr i32, ptr %out, i32 4
+  %gep = getelementptr inbounds i32, ptr %out, i32 4
   %tmp0 = atomicrmw uinc_wrap ptr %gep, i32 %in seq_cst
   ret void
 }
@@ -8862,7 +8862,7 @@ define amdgpu_gfx i32 @flat_atomic_uinc_wrap_i32_ret_offset_scalar(ptr inreg %ou
 ; GCN3-NEXT:  ; %bb.2: ; %atomicrmw.end
 ; GCN3-NEXT:    s_or_b64 exec, exec, s[34:35]
 ; GCN3-NEXT:    s_setpc_b64 s[30:31]
-  %gep = getelementptr i32, ptr %out, i32 4
+  %gep = getelementptr inbounds i32, ptr %out, i32 4
   %result = atomicrmw uinc_wrap ptr %gep, i32 %in seq_cst
   ret i32 %result
 }
@@ -8895,7 +8895,7 @@ define void @flat_uinc_wrap_i32_noret_offset__amdgpu_no_remote_memory(ptr %out,
 ; GCN3-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GCN3-NEXT:    buffer_wbinvl1_vol
 ; GCN3-NEXT:    s_setpc_b64 s[30:31]
-  %gep = getelementptr i32, ptr %out, i64 4
+  %gep = getelementptr inbounds i32, ptr %out, i64 4
   %tmp0 = atomicrmw uinc_wrap ptr %gep, i32 %in seq_cst, !amdgpu.no.remote.memory !0
   ret void
 }
@@ -8928,7 +8928,7 @@ define i32 @flat_atomic_uinc_wrap_i32_ret_offset__amdgpu_no_remote_memory(ptr %o
 ; GCN3-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GCN3-NEXT:    buffer_wbinvl1_vol
 ; GCN3-NEXT:    s_setpc_b64 s[30:31]
-  %gep = getelementptr i32, ptr %out, i64 4
+  %gep = getelementptr inbounds i32, ptr %out, i64 4
   %result = atomicrmw uinc_wrap ptr %gep, i32 %in seq_cst, !amdgpu.no.remote.memory !0
   ret i32 %result
 }
@@ -9089,7 +9089,7 @@ define void @flat_atomic_udec_wrap_i32_noret_offset(ptr %out, i32 %in) {
 ; GCN3-NEXT:  ; %bb.2: ; %atomicrmw.end
 ; GCN3-NEXT:    s_or_b64 exec, exec, s[6:7]
 ; GCN3-NEXT:    s_setpc_b64 s[30:31]
-  %gep = getelementptr i32, ptr %out, i32 4
+  %gep = getelementptr inbounds i32, ptr %out, i32 4
   %tmp0 = atomicrmw udec_wrap ptr %gep, i32 %in seq_cst
   ret void
 }
@@ -9250,7 +9250,7 @@ define i32 @flat_atomic_udec_wrap_i32_ret_offset(ptr %out, i32 %in) {
 ; GCN3-NEXT:    s_or_b64 exec, exec, s[6:7]
 ; GCN3-NEXT:    v_mov_b32_e32 v0, v3
 ; GCN3-NEXT:    s_setpc_b64 s[30:31]
-  %gep = getelementptr i32, ptr %out, i32 4
+  %gep = getelementptr inbounds i32, ptr %out, i32 4
   %result = atomicrmw udec_wrap ptr %gep, i32 %in seq_cst
   ret i32 %result
 }
@@ -9425,7 +9425,7 @@ define amdgpu_gfx void @flat_atomic_udec_wrap_i32_noret_offset_scalar(ptr inreg
 ; GCN3-NEXT:  ; %bb.2: ; %atomicrmw.end
 ; GCN3-NEXT:    s_or_b64 exec, exec, s[36:37]
 ; GCN3-NEXT:    s_setpc_b64 s[30:31]
-  %gep = getelementptr i32, ptr %out, i32 4
+  %gep = getelementptr inbounds i32, ptr %out, i32 4
   %tmp0 = atomicrmw udec_wrap ptr %gep, i32 %in seq_cst
   ret void
 }
@@ -9608,7 +9608,7 @@ define amdgpu_gfx i32 @flat_atomic_udec_wrap_i32_ret_offset_scalar(ptr inreg %ou
 ; GCN3-NEXT:  ; %bb.2: ; %atomicrmw.end
 ; GCN3-NEXT:    s_or_b64 exec, exec, s[36:37]
 ; GCN3-NEXT:    s_setpc_b64 s[30:31]
-  %gep = getelementptr i32, ptr %out, i32 4
+  %gep = getelementptr inbounds i32, ptr %out, i32 4
   %result = atomicrmw udec_wrap ptr %gep, i32 %in seq_cst
   ret i32 %result
 }
@@ -9641,7 +9641,7 @@ define void @flat_udec_wrap_i32_noret_offset__amdgpu_no_remote_memory(ptr %out,
 ; GCN3-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GCN3-NEXT:    buffer_wbinvl1_vol
 ; GCN3-NEXT:    s_setpc_b64 s[30:31]
-  %gep = getelementptr i32, ptr %out, i64 4
+  %gep = getelementptr inbounds i32, ptr %out, i64 4
   %tmp0 = atomicrmw udec_wrap ptr %gep, i32 %in seq_cst, !amdgpu.no.remote.memory !0
   ret void
 }
@@ -9674,7 +9674,7 @@ define i32 @flat_atomic_udec_wrap_i32_ret_offset__amdgpu_no_remote_memory(ptr %o
 ; GCN3-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GCN3-NEXT:    buffer_wbinvl1_vol
 ; GCN3-NEXT:    s_setpc_b64 s[30:31]
-  %gep = getelementptr i32, ptr %out, i64 4
+  %gep = getelementptr inbounds i32, ptr %out, i64 4
   %result = atomicrmw udec_wrap ptr %gep, i32 %in seq_cst, !amdgpu.no.remote.memory !0
   ret i32 %result
 }
diff --git a/llvm/test/CodeGen/AMDGPU/flat_atomics_i64.ll b/llvm/test/CodeGen/AMDGPU/flat_atomics_i64.ll
index 1f105e8dd8ba5..6dfe4594a248c 100644
--- a/llvm/test/CodeGen/AMDGPU/flat_atomics_i64.ll
+++ b/llvm/test/CodeGen/AMDGPU/flat_atomics_i64.ll
@@ -138,7 +138,7 @@ define amdgpu_kernel void @atomic_add_i64_offset(ptr %out, i64 %in) {
 ; GFX12-NEXT:    scratch_store_b64 off, v[0:1], s0
 ; GFX12-NEXT:    s_endpgm
 entry:
-  %gep = getelementptr i64, ptr %out, i64 4
+  %gep = getelementptr inbounds i64, ptr %out, i64 4
   %tmp0 = atomicrmw volatile add ptr %gep, i64 %in syncscope("agent") seq_cst
   ret void
 }
@@ -284,7 +284,7 @@ define amdgpu_kernel void @atomic_add_i64_ret_offset(ptr %out, ptr %out2, i64 %i
 ; GFX12-NEXT:    flat_store_b64 v[2:3], v[0:1]
 ; GFX12-NEXT:    s_endpgm
 entry:
-  %gep = getelementptr i64, ptr %out, i64 4
+  %gep = getelementptr inbounds i64, ptr %out, i64 4
   %tmp0 = atomicrmw volatile add ptr %gep, i64 %in syncscope("agent") seq_cst
   store i64 %tmp0, ptr %out2
   ret void
@@ -438,8 +438,8 @@ define amdgpu_kernel void @atomic_add_i64_addr64_offset(ptr %out, i64 %in, i64 %
 ; GFX12-NEXT:    scratch_store_b64 off, v[0:1], s0
 ; GFX12-NEXT:    s_endpgm
 entry:
-  %ptr = getelementptr i64, ptr %out, i64 %index
-  %gep = getelementptr i64, ptr %ptr, i64 4
+  %ptr = getelementptr inbounds i64, ptr %out, i64 %index
+  %gep = getelementptr inbounds i64, ptr %ptr, i64 4
   %tmp0 = atomicrmw volatile add ptr %gep, i64 %in syncscope("agent") seq_cst
   ret void
 }
@@ -590,8 +590,8 @@ define amdgpu_kernel void @atomic_add_i64_ret_addr64_offset(ptr %out, ptr %out2,
 ; GFX12-NEXT:    flat_store_b64 v[2:3], v[0:1]
 ; GFX12-NEXT:    s_endpgm
 entry:
-  %ptr = getelementptr i64, ptr %out, i64 %index
-  %gep = getelementptr i64, ptr %ptr, i64 4
+  %ptr = getelementptr inbounds i64, ptr %out, i64 %index
+  %gep = getelementptr inbounds i64, ptr %ptr, i64 4
   %tmp0 = atomicrmw volatile add ptr %gep, i64 %in syncscope("agent") seq_cst
   store i64 %tmp0, ptr %out2
   ret void
@@ -1015,7 +1015,7 @@ define amdgpu_kernel void @atomic_add_i64_addr64(ptr %out, i64 %in, i64 %index)
 ; GFX12-NEXT:    scratch_store_b64 off, v[0:1], s0
 ; GFX12-NEXT:    s_endpgm
 entry:
-  %ptr = getelementptr i64, ptr %out, i64 %index
+  %ptr = getelementptr inbounds i64, ptr %out, i64 %index
   %tmp0 = atomicrmw volatile add ptr %ptr, i64 %in syncscope("agent") seq_cst
   ret void
 }
@@ -1161,7 +1161,7 @@ define amdgpu_kernel void @atomic_add_i64_ret_addr64(ptr %out, ptr %out2, i64 %i
 ; GFX12-NEXT:    flat_store_b64 v[2:3], v[0:1]
 ; GFX12-NEXT:    s_endpgm
 entry:
-  %ptr = getelementptr i64, ptr %out, i64 %index
+  %ptr = getelementptr inbounds i64, ptr %out, i64 %index
   %tmp0 = atomicrmw volatile add ptr %ptr, i64 %in syncscope("agent") seq_cst
   store i64 %tmp0, ptr %out2
   ret void
@@ -1299,7 +1299,7 @@ define amdgpu_kernel void @atomic_and_i64_offset(ptr %out, i64 %in) {
 ; GFX12-NEXT:    scratch_store_b64 off, v[0:1], s0
 ; GFX12-NEXT:    s_endpgm
 entry:
-  %gep = getelementptr i64, ptr %out, i64 4
+  %gep = getelementptr inbounds i64, ptr %out, i64 4
   %tmp0 = atomicrmw volatile and ptr %gep, i64 %in syncscope("agent") seq_cst, !amdgpu.no.fine.grained.memory !0
   ret void
 }
@@ -1442,7 +1442,7 @@ define amdgpu_kernel void @atomic_and_i64_ret_offset(ptr %out, ptr %out2, i64 %i
 ; GFX12-NEXT:    flat_store_b64 v[2:3], v[0:1]
 ; GFX12-NEXT:    s_endpgm
 entry:
-  %gep = getelementptr i64, ptr %out, i64 4
+  %gep = getelementptr inbounds i64, ptr %out, i64 4
   %tmp0 = atomicrmw volatile and ptr %gep, i64 %in syncscope("agent") seq_cst, !amdgpu.no.fine.grained.memory !0
   store i64 %tmp0, ptr %out2
   ret void
@@ -1593,8 +1593,8 @@ define amdgpu_kernel void @atomic_and_i64_addr64_offset(ptr %out, i64 %in, i64 %
 ; GFX12-NEXT:    scratch_store_b64 off, v[0:1], s0
 ; GFX12-NEXT:    s_endpgm
 entry:
-  %ptr = getelementptr i64, ptr %out, i64 %index
-  %gep = getelementptr i64, ptr %ptr, i64 4
+  %ptr = getelementptr inbounds i64, ptr %out, i64 %index
+  %gep = getelementptr inbounds i64, ptr %ptr, i64 4
   %tmp0 = atomicrmw volatile and ptr %gep, i64 %in syncscope("agent") seq_cst, !amdgpu.no.fine.grained.memory !0
   ret void
 }
@@ -1742,8 +1742,8 @@ define amdgpu_kernel void @atomic_and_i64_ret_addr64_offset(ptr %out, ptr %out2,
 ; GFX12-NEXT:    flat_store_b64 v[2:3], v[0:1]
 ; GFX12-NEXT:    s_endpgm
 entry:
-  %ptr = getelementptr i64, ptr %out, i64 %index
-  %gep = getelementptr i64, ptr %ptr, i64 4
+  %ptr = getelementptr inbounds i64, ptr %out, i64 %index
+  %gep = getelementptr inbounds i64, ptr %ptr, i64 4
   %tmp0 = atomicrmw volatile and ptr %gep, i64 %in syncscope("agent") seq_cst, !amdgpu.no.fine.grained.memory !0
   store i64 %tmp0, ptr %out2
   ret void
@@ -2158,7 +2158,7 @@ define amdgpu_kernel void @atomic_and_i64_addr64(ptr %out, i64 %in, i64 %index)
 ; GFX12-NEXT:    scratch_store_b64 off, v[0:1], s0
 ; GFX12-NEXT:    s_endpgm
 entry:
-  %ptr = getelementptr i64, ptr %out, i64 %index
+  %ptr = getelementptr inbounds i64, ptr %out, i64 %index
   %tmp0 = atomicrmw volatile and ptr %ptr, i64 %in syncscope("agent") seq_cst, !amdgpu.no.fine.grained.memory !0
   ret void
 }
@@ -2301,7 +2301,7 @@ define amdgpu_kernel void @atomic_and_i64_ret_addr64(ptr %out, ptr %out2, i64 %i
 ; GFX12-NEXT:    flat_store_b64 v[2:3], v[0:1]
 ; GFX12-NEXT:    s_endpgm
 entry:
-  %ptr = getelementptr i64, ptr %out, i64 %index
+  %ptr = getelementptr inbounds i64, ptr %out, i64 %index
   %tmp0 = atomicrmw volatile and ptr %ptr, i64 %in syncscope("agent") seq_cst, !amdgpu.no.fine.grained.memory !0
   store i64 %tmp0, ptr %out2
   ret void
@@ -2442,7 +2442,7 @@ define amdgpu_kernel void @atomic_sub_i64_offset(ptr %out, i64 %in) {
 ; GFX12-NEXT:    scratch_store_b64 off, v[0:1], s0
 ; GFX12-NEXT:    s_endpgm
 entry:
-  %gep = getelementptr i64, ptr %out, i64 4
+  %gep = getelementptr inbounds i64, ptr %out, i64 4
   %tmp0 = atomicrmw volatile sub ptr %gep, i64 %in syncscope("agent") seq_cst, !amdgpu.no.fine.grained.memory !0
   ret void
 }
@@ -2588,7 +2588,7 @@ define amdgpu_kernel void @atomic_sub_i64_ret_offset(ptr %out, ptr %out2, i64 %i
 ; GFX12-NEXT:    flat_store_b64 v[2:3], v[0:1]
 ; GFX12-NEXT:    s_endpgm
 entry:
-  %gep = getelementptr i64, ptr %out, i64 4
+  %gep = getelementptr inbounds i64, ptr %out, i64 4
   %tmp0 = atomicrmw volatile sub ptr %gep, i64 %in syncscope("agent") seq_cst, !amdgpu.no.fine.grained.memory !0
   store i64 %tmp0, ptr %out2
   ret void
@@ -2742,8 +2742,8 @@ define amdgpu_kernel void @atomic_sub_i64_addr64_offset(ptr %out, i64 %in, i64 %
 ; GFX12-NEXT:    scratch_store_b64 off, v[0:1], s0
 ; GFX12-NEXT:    s_endpgm
 entry:
-  %ptr = getelementptr i64, ptr %out, i64 %index
-  %gep = getelementptr i64, ptr %ptr, i64 4
+  %ptr = getelementptr inbounds i64, ptr %out, i64 %index
+  %gep = getelementptr inbounds i64, ptr %ptr, i64 4
   %tmp0 = atomicrmw volatile sub ptr %gep, i64 %in syncscope("agent") seq_cst, !amdgpu.no.fine.grained.memory !0
   ret void
 }
@@ -2894,8 +2894,8 @@ define amdgpu_kernel void @atomic_sub_i64_ret_addr64_offset(ptr %out, ptr %out2,
 ; GFX12-NEXT:    flat_store_b64 v[2:3], v[0:1]
 ; GFX12-NEXT:    s_endpgm
 entry:
-  %ptr = getelementptr i64, ptr %out, i64 %index
-  %gep = getelementptr i64, ptr %ptr, i64 4
+  %ptr = getelementptr inbounds i64, ptr %out, i64 %index
+  %gep = getelementptr inbounds i64, ptr %ptr, i64 4
   %tmp0 = atomicrmw volatile sub ptr %gep, i64 %in syncscope("agent") seq_cst, !amdgpu.no.fine.grained.memory !0
   store i64 %tmp0, ptr %out2
   ret void
@@ -3319,7 +3319,7 @@ define amdgpu_kernel void @atomic_sub_i64_addr64(ptr %out, i64 %in, i64 %index)
 ; GFX12-NEXT:    scratch_store_b64 off, v[0:1], s0
 ; GFX12-NEXT:    s_endpgm
 entry:
-  %ptr = getelementptr i64, ptr %out, i64 %index
+  %ptr = getelementptr inbounds i64, ptr %out, i64 %index
   %tmp0 = atomicrmw volatile sub ptr %ptr, i64 %in syncscope("agent") seq_cst, !amdgpu.no.fine.grained.memory !0
   ret void
 }
@@ -3465,7 +3465,7 @@ define amdgpu_kernel void @atomic_sub_i64_ret_addr64(ptr %out, ptr %out2, i64 %i
 ; GFX12-NEXT:    flat_store_b64 v[2:3], v[0:1]
 ; GFX12-NEXT:    s_endpgm
 entry:
-  %ptr = getelementptr i64, ptr %out, i64 %index
+  %ptr = getelementptr inbounds i64, ptr %out, i64 %index
   %tmp0 = atomicrmw volatile sub ptr %ptr, i64 %in syncscope("agent") seq_cst, !amdgpu.no.fine.grained.memory !0
   store i64 %tmp0, ptr %out2
   ret void
@@ -3606,7 +3606,7 @@ define amdgpu_kernel void @atomic_max_i64_offset(ptr %out, i64 %in) {
 ; GFX12-NEXT:    scratch_store_b64 off, v[0:1], s0
 ; GFX12-NEXT:    s_endpgm
 entry:
-  %gep = getelementptr i64, ptr %out, i64 4
+  %gep = getelementptr inbounds i64, ptr %out, i64 4
   %tmp0 = atomicrmw volatile max ptr %gep, i64 %in syncscope("workgroup") seq_cst, !amdgpu.no.fine.grained.memory !0
   ret void
 }
@@ -3754,7 +3754,7 @@ define amdgpu_kernel void @atomic_max_i64_ret_offset(ptr %out, ptr %out2, i64 %i
 ; GFX12-NEXT:    flat_store_b64 v[2:3], v[0:1]
 ; GFX12-NEXT:    s_endpgm
 entry:
-  %gep = getelementptr i64, ptr %out, i64 4
+  %gep = getelementptr inbounds i64, ptr %out, i64 4
   %tmp0 = atomicrmw volatile max ptr %gep, i64 %in syncscope("workgroup") seq_cst, !amdgpu.no.fine.grained.memory !0
   store i64 %tmp0, ptr %out2
   ret void
@@ -3908,8 +3908,8 @@ define amdgpu_kernel void @atomic_max_i64_addr64_offset(ptr %out, i64 %in, i64 %
 ; GFX12-NEXT:    scratch_store_b64 off, v[0:1], s0
 ; GFX12-NEXT:    s_endpgm
 entry:
-  %ptr = getelementptr i64, ptr %out, i64 %index
-  %gep = getelementptr i64, ptr %ptr, i64 4
+  %ptr = getelementptr inbounds i64, ptr %out, i64 %index
+  %gep = getelementptr inbounds i64, ptr %ptr, i64 4
   %tmp0 = atomicrmw volatile max ptr %gep, i64 %in syncscope("workgroup") seq_cst, !amdgpu.no.fine.grained.memory !0
   ret void
 }
@@ -4062,8 +4062,8 @@ define amdgpu_kernel void @atomic_max_i64_ret_addr64_offset(ptr %out, ptr %out2,
 ; GFX12-NEXT:    flat_store_b64 v[2:3], v[0:1]
 ; GFX12-NEXT:    s_endpgm
 entry:
-  %ptr = getelementptr i64, ptr %out, i64 %index
-  %gep = getelementptr i64, ptr %ptr, i64 4
+  %ptr = getelementptr inbounds i64, ptr %out, i64 %index
+  %gep = getelementptr inbounds i64, ptr %ptr, i64 4
   %tmp0 = atomicrmw volatile max ptr %gep, i64 %in syncscope("workgroup") seq_cst, !amdgpu.no.fine.grained.memory !0
   store i64 %tmp0, ptr %out2
   ret void
@@ -4489,7 +4489,7 @@ define amdgpu_kernel void @atomic_max_i64_addr64(ptr %out, i64 %in, i64 %index)
 ; GFX12-NEXT:    scratch_store_b64 off, v[0:1], s0
 ; GFX12-NEXT:    s_endpgm
 entry:
-  %ptr = getelementptr i64, ptr %out, i64 %index
+  %ptr = getelementptr inbounds i64, ptr %out, i64 %index
   %tmp0 = atomicrmw volatile max ptr %ptr, i64 %in syncscope("workgroup") seq_cst, !amdgpu.no.fine.grained.memory !0
   ret void
 }
@@ -4637,7 +4637,7 @@ define amdgpu_kernel void @atomic_max_i64_ret_addr64(ptr %out, ptr %out2, i64 %i
 ; GFX12-NEXT:    flat_store_b64 v[2:3], v[0:1]
 ; GFX12-NEXT:    s_endpgm
 entry:
-  %ptr = getelementptr i64, ptr %out, i64 %index
+  %ptr = getelementptr inbounds i64, ptr %out, i64 %index
   %tmp0 = atomicrmw volatile max ptr %ptr, i64 %in syncscope("workgroup") seq_cst, !amdgpu.no.fine.grained.memory !0
   store i64 %tmp0, ptr %out2
   ret void
@@ -4778,7 +4778,7 @@ define amdgpu_kernel void @atomic_umax_i64_offset(ptr %out, i64 %in) {
 ; GFX12-NEXT:    scratch_store_b64 off, v[0:1], s0
 ; GFX12-NEXT:    s_endpgm
 entry:
-  %gep = getelementptr i64, ptr %out, i64 4
+  %gep = getelementptr inbounds i64, ptr %out, i64 4
   %tmp0 = atomicrmw volatile umax ptr %gep, i64 %in syncscope("workgroup") seq_cst, !amdgpu.no.fine.grained.memory !0
   ret void
 }
@@ -4926,7 +4926,7 @@ define amdgpu_kernel void @atomic_umax_i64_ret_offset(ptr %out, ptr %out2, i64 %
 ; GFX12-NEXT:    flat_store_b64 v[2:3], v[0:1]
 ; GFX12-NEXT:    s_endpgm
 entry:
-  %gep = getelementptr i64, ptr %out, i64 4
+  %gep = getelementptr inbounds i64, ptr %out, i64 4
   %tmp0 = atomicrmw volatile umax ptr %gep, i64 %in syncscope("workgroup") seq_cst, !amdgpu.no.fine.grained.memory !0
   store i64 %tmp0, ptr %out2
   ret void
@@ -5080,8 +5080,8 @@ define amdgpu_kernel void @atomic_umax_i64_addr64_offset(ptr %out, i64 %in, i64
 ; GFX12-NEXT:    scratch_store_b64 off, v[0:1], s0
 ; GFX12-NEXT:    s_endpgm
 entry:
-  %ptr = getelementptr i64, ptr %out, i64 %index
-  %gep = getelementptr i64, ptr %ptr, i64 4
+  %ptr = getelementptr inbounds i64, ptr %out, i64 %index
+  %gep = getelementptr inbounds i64, ptr %ptr, i64 4
   %tmp0 = atomicrmw volatile umax ptr %gep, i64 %in syncscope("workgroup") seq_cst, !amdgpu.no.fine.grained.memory !0
   ret void
 }
@@ -5234,8 +5234,8 @@ define amdgpu_kernel void @atomic_umax_i64_ret_addr64_offset(ptr %out, ptr %out2
 ; GFX12-NEXT:    flat_store_b64 v[2:3], v[0:1]
 ; GFX12-NEXT:    s_endpgm
 entry:
-  %ptr = getelementptr i64, ptr %out, i64 %index
-  %gep = getelementptr i64, ptr %ptr, i64 4
+  %ptr = getelementptr inbounds i64, ptr %out, i64 %index
+  %gep = getelementptr inbounds i64, ptr %ptr, i64 4
   %tmp0 = atomicrmw volatile umax ptr %gep, i64 %in syncscope("workgroup") seq_cst, !amdgpu.no.fine.grained.memory !0
   store i64 %tmp0, ptr %out2
   ret void
@@ -5661,7 +5661,7 @@ define amdgpu_kernel void @atomic_umax_i64_addr64(ptr %out, i64 %in, i64 %index)
 ; GFX12-NEXT:    scratch_store_b64 off, v[0:1], s0
 ; GFX12-NEXT:    s_endpgm
 entry:
-  %ptr = getelementptr i64, ptr %out, i64 %index
+  %ptr = getelementptr inbounds i64, ptr %out, i64 %index
   %tmp0 = atomicrmw volatile umax ptr %ptr, i64 %in syncscope("workgroup") seq_cst, !amdgpu.no.fine.grained.memory !0
   ret void
 }
@@ -5809,7 +5809,7 @@ define amdgpu_kernel void @atomic_umax_i64_ret_addr64(ptr %out, ptr %out2, i64 %
 ; GFX12-NEXT:    flat_store_b64 v[2:3], v[0:1]
 ; GFX12-NEXT:    s_endpgm
 entry:
-  %ptr = getelementptr i64, ptr %out, i64 %index
+  %ptr = getelementptr inbounds i64, ptr %out, i64 %index
   %tmp0 = atomicrmw volatile umax ptr %ptr, i64 %in syncscope("workgroup") seq_cst, !amdgpu.no.fine.grained.memory !0
   store i64 %tmp0, ptr %out2
   ret void
@@ -5950,7 +5950,7 @@ define amdgpu_kernel void @atomic_min_i64_offset(ptr %out, i64 %in) {
 ; GFX12-NEXT:    scratch_store_b64 off, v[0:1], s0
 ; GFX12-NEXT:    s_endpgm
 entry:
-  %gep = getelementptr i64, ptr %out, i64 4
+  %gep = getelementptr inbounds i64, ptr %out, i64 4
   %tmp0 = atomicrmw volatile min ptr %gep, i64 %in syncscope("workgroup") seq_cst, !amdgpu.no.fine.grained.memory !0
   ret void
 }
@@ -6098,7 +6098,7 @@ define amdgpu_kernel void @atomic_min_i64_ret_offset(ptr %out, ptr %out2, i64 %i
 ; GFX12-NEXT:    flat_store_b64 v[2:3], v[0:1]
 ; GFX12-NEXT:    s_endpgm
 entry:
-  %gep = getelementptr i64, ptr %out, i64 4
+  %gep = getelementptr inbounds i64, ptr %out, i64 4
   %tmp0 = atomicrmw volatile min ptr %gep, i64 %in syncscope("workgroup") seq_cst, !amdgpu.no.fine.grained.memory !0
   store i64 %tmp0, ptr %out2
   ret void
@@ -6252,8 +6252,8 @@ define amdgpu_kernel void @atomic_min_i64_addr64_offset(ptr %out, i64 %in, i64 %
 ; GFX12-NEXT:    scratch_store_b64 off, v[0:1], s0
 ; GFX12-NEXT:    s_endpgm
 entry:
-  %ptr = getelementptr i64, ptr %out, i64 %index
-  %gep = getelementptr i64, ptr %ptr, i64 4
+  %ptr = getelementptr inbounds i64, ptr %out, i64 %index
+  %gep = getelementptr inbounds i64, ptr %ptr, i64 4
   %tmp0 = atomicrmw volatile min ptr %gep, i64 %in syncscope("workgroup") seq_cst, !amdgpu.no.fine.grained.memory !0
   ret void
 }
@@ -6406,8 +6406,8 @@ define amdgpu_kernel void @atomic_min_i64_ret_addr64_offset(ptr %out, ptr %out2,
 ; GFX12-NEXT:    flat_store_b64 v[2:3], v[0:1]
 ; GFX12-NEXT:    s_endpgm
 entry:
-  %ptr = getelementptr i64, ptr %out, i64 %index
-  %gep = getelementptr i64, ptr %ptr, i64 4
+  %ptr = getelementptr inbounds i64, ptr %out, i64 %index
+  %gep = getelementptr inbounds i64, ptr %ptr, i64 4
   %tmp0 = atomicrmw volatile min ptr %gep, i64 %in syncscope("workgroup") seq_cst, !amdgpu.no.fine.grained.memory !0
   store i64 %tmp0, ptr %out2
   ret void
@@ -6833,7 +6833,7 @@ define amdgpu_kernel void @atomic_min_i64_addr64(ptr %out, i64 %in, i64 %index)
 ; GFX12-NEXT:    scratch_store_b64 off, v[0:1], s0
 ; GFX12-NEXT:    s_endpgm
 entry:
-  %ptr = getelementptr i64, ptr %out, i64 %index
+  %ptr = getelementptr inbounds i64, ptr %out, i64 %index
   %tmp0 = atomicrmw volatile min ptr %ptr, i64 %in syncscope("workgroup") seq_cst, !amdgpu.no.fine.grained.memory !0
   ret void
 }
@@ -6981,7 +6981,7 @@ define amdgpu_kernel void @atomic_min_i64_ret_addr64(ptr %out, ptr %out2, i64 %i
 ; GFX12-NEXT:    flat_store_b64 v[2:3], v[0:1]
 ; GFX12-NEXT:    s_endpgm
 entry:
-  %ptr = getelementptr i64, ptr %out, i64 %index
+  %ptr = getelementptr inbounds i64, ptr %out, i64 %index
   %tmp0 = atomicrmw volatile min ptr %ptr, i64 %in syncscope("workgroup") seq_cst, !amdgpu.no.fine.grained.memory !0
   store i64 %tmp0, ptr %out2
   ret void
@@ -7122,7 +7122,7 @@ define amdgpu_kernel void @atomic_umin_i64_offset(ptr %out, i64 %in) {
 ; GFX12-NEXT:    scratch_store_b64 off, v[0:1], s0
 ; GFX12-NEXT:    s_endpgm
 entry:
-  %gep = getelementptr i64, ptr %out, i64 4
+  %gep = getelementptr inbounds i64, ptr %out, i64 4
   %tmp0 = atomicrmw volatile umin ptr %gep, i64 %in syncscope("workgroup") seq_cst, !amdgpu.no.fine.grained.memory !0
   ret void
 }
@@ -7270,7 +7270,7 @@ define amdgpu_kernel void @atomic_umin_i64_ret_offset(ptr %out, ptr %out2, i64 %
 ; GFX12-NEXT:    flat_store_b64 v[2:3], v[0:1]
 ; GFX12-NEXT:    s_endpgm
 entry:
-  %gep = getelementptr i64, ptr %out, i64 4
+  %gep = getelementptr inbounds i64, ptr %out, i64 4
   %tmp0 = atomicrmw volatile umin ptr %gep, i64 %in syncscope("workgroup") seq_cst, !amdgpu.no.fine.grained.memory !0
   store i64 %tmp0, ptr %out2
   ret void
@@ -7424,8 +7424,8 @@ define amdgpu_kernel void @atomic_umin_i64_addr64_offset(ptr %out, i64 %in, i64
 ; GFX12-NEXT:    scratch_store_b64 off, v[0:1], s0
 ; GFX12-NEXT:    s_endpgm
 entry:
-  %ptr = getelementptr i64, ptr %out, i64 %index
-  %gep = getelementptr i64, ptr %ptr, i64 4
+  %ptr = getelementptr inbounds i64, ptr %out, i64 %index
+  %gep = getelementptr inbounds i64, ptr %ptr, i64 4
   %tmp0 = atomicrmw volatile umin ptr %gep, i64 %in syncscope("workgroup") seq_cst, !amdgpu.no.fine.grained.memory !0
   ret void
 }
@@ -7578,8 +7578,8 @@ define amdgpu_kernel void @atomic_umin_i64_ret_addr64_offset(ptr %out, ptr %out2
 ; GFX12-NEXT:    flat_store_b64 v[2:3], v[0:1]
 ; GFX12-NEXT:    s_endpgm
 entry:
-  %ptr = getelementptr i64, ptr %out, i64 %index
-  %gep = getelementptr i64, ptr %ptr, i64 4
+  %ptr = getelementptr inbounds i64, ptr %out, i64 %index
+  %gep = getelementptr inbounds i64, ptr %ptr, i64 4
   %tmp0 = atomicrmw volatile umin ptr %gep, i64 %in syncscope("workgroup") seq_cst, !amdgpu.no.fine.grained.memory !0
   store i64 %tmp0, ptr %out2
   ret void
@@ -8005,7 +8005,7 @@ define amdgpu_kernel void @atomic_umin_i64_addr64(ptr %out, i64 %in, i64 %index)
 ; GFX12-NEXT:    scratch_store_b64 off, v[0:1], s0
 ; GFX12-NEXT:    s_endpgm
 entry:
-  %ptr = getelementptr i64, ptr %out, i64 %index
+  %ptr = getelementptr inbounds i64, ptr %out, i64 %index
   %tmp0 = atomicrmw volatile umin ptr %ptr, i64 %in syncscope("workgroup") seq_cst, !amdgpu.no.fine.grained.memory !0
   ret void
 }
@@ -8153,7 +8153,7 @@ define amdgpu_kernel void @atomic_umin_i64_ret_addr64(ptr %out, ptr %out2, i64 %
 ; GFX12-NEXT:    flat_store_b64 v[2:3], v[0:1]
 ; GFX12-NEXT:    s_endpgm
 entry:
-  %ptr = getelementptr i64, ptr %out, i64 %index
+  %ptr = getelementptr inbounds i64, ptr %out, i64 %index
   %tmp0 = atomicrmw volatile umin ptr %ptr, i64 %in syncscope("workgroup") seq_cst, !amdgpu.no.fine.grained.memory !0
   store i64 %tmp0, ptr %out2
   ret void
@@ -8291,7 +8291,7 @@ define amdgpu_kernel void @atomic_or_i64_offset(ptr %out, i64 %in) {
 ; GFX12-NEXT:    scratch_store_b64 off, v[0:1], s0
 ; GFX12-NEXT:    s_endpgm
 entry:
-  %gep = getelementptr i64, ptr %out, i64 4
+  %gep = getelementptr inbounds i64, ptr %out, i64 4
   %tmp0 = atomicrmw volatile or ptr %gep, i64 %in syncscope("agent") seq_cst, !amdgpu.no.fine.grained.memory !0
   ret void
 }
@@ -8434,7 +8434,7 @@ define amdgpu_kernel void @atomic_or_i64_ret_offset(ptr %out, ptr %out2, i64 %in
 ; GFX12-NEXT:    flat_store_b64 v[2:3], v[0:1]
 ; GFX12-NEXT:    s_endpgm
 entry:
-  %gep = getelementptr i64, ptr %out, i64 4
+  %gep = getelementptr inbounds i64, ptr %out, i64 4
   %tmp0 = atomicrmw volatile or ptr %gep, i64 %in syncscope("agent") seq_cst, !amdgpu.no.fine.grained.memory !0
   store i64 %tmp0, ptr %out2
   ret void
@@ -8585,8 +8585,8 @@ define amdgpu_kernel void @atomic_or_i64_addr64_offset(ptr %out, i64 %in, i64 %i
 ; GFX12-NEXT:    scratch_store_b64 off, v[0:1], s0
 ; GFX12-NEXT:    s_endpgm
 entry:
-  %ptr = getelementptr i64, ptr %out, i64 %index
-  %gep = getelementptr i64, ptr %ptr, i64 4
+  %ptr = getelementptr inbounds i64, ptr %out, i64 %index
+  %gep = getelementptr inbounds i64, ptr %ptr, i64 4
   %tmp0 = atomicrmw volatile or ptr %gep, i64 %in syncscope("agent") seq_cst, !amdgpu.no.fine.grained.memory !0
   ret void
 }
@@ -8734,8 +8734,8 @@ define amdgpu_kernel void @atomic_or_i64_ret_addr64_offset(ptr %out, ptr %out2,
 ; GFX12-NEXT:    flat_store_b64 v[2:3], v[0:1]
 ; GFX12-NEXT:    s_endpgm
 entry:
-  %ptr = getelementptr i64, ptr %out, i64 %index
-  %gep = getelementptr i64, ptr %ptr, i64 4
+  %ptr = getelementptr inbounds i64, ptr %out, i64 %index
+  %gep = getelementptr inbounds i64, ptr %ptr, i64 4
   %tmp0 = atomicrmw volatile or ptr %gep, i64 %in syncscope("agent") seq_cst, !amdgpu.no.fine.grained.memory !0
   store i64 %tmp0, ptr %out2
   ret void
@@ -9150,7 +9150,7 @@ define amdgpu_kernel void @atomic_or_i64_addr64(ptr %out, i64 %in, i64 %index) {
 ; GFX12-NEXT:    scratch_store_b64 off, v[0:1], s0
 ; GFX12-NEXT:    s_endpgm
 entry:
-  %ptr = getelementptr i64, ptr %out, i64 %index
+  %ptr = getelementptr inbounds i64, ptr %out, i64 %index
   %tmp0 = atomicrmw volatile or ptr %ptr, i64 %in syncscope("agent") seq_cst, !amdgpu.no.fine.grained.memory !0
   ret void
 }
@@ -9293,7 +9293,7 @@ define amdgpu_kernel void @atomic_or_i64_ret_addr64(ptr %out, ptr %out2, i64 %in
 ; GFX12-NEXT:    flat_store_b64 v[2:3], v[0:1]
 ; GFX12-NEXT:    s_endpgm
 entry:
-  %ptr = getelementptr i64, ptr %out, i64 %index
+  %ptr = getelementptr inbounds i64, ptr %out, i64 %index
   %tmp0 = atomicrmw volatile or ptr %ptr, i64 %in syncscope("agent") seq_cst, !amdgpu.no.fine.grained.memory !0
   store i64 %tmp0, ptr %out2
   ret void
@@ -9420,7 +9420,7 @@ define amdgpu_kernel void @atomic_xchg_i64_offset(ptr %out, i64 %in) {
 ; GFX12-NEXT:    scratch_store_b64 off, v[0:1], s0
 ; GFX12-NEXT:    s_endpgm
 entry:
-  %gep = getelementptr i64, ptr %out, i64 4
+  %gep = getelementptr inbounds i64, ptr %out, i64 4
   %tmp0 = atomicrmw volatile xchg ptr %gep, i64 %in syncscope("agent") seq_cst
   ret void
 }
@@ -9546,7 +9546,7 @@ define amdgpu_kernel void @atomic_xchg_f64_offset(ptr %out, double %in) {
 ; GFX12-NEXT:    scratch_store_b64 off, v[0:1], s0
 ; GFX12-NEXT:    s_endpgm
 entry:
-  %gep = getelementptr double, ptr %out, i64 4
+  %gep = getelementptr inbounds double, ptr %out, i64 4
   %tmp0 = atomicrmw volatile xchg ptr %gep, double %in syncscope("agent") seq_cst
   ret void
 }
@@ -9672,7 +9672,7 @@ define amdgpu_kernel void @atomic_xchg_pointer_offset(ptr %out, ptr %in) {
 ; GFX12-NEXT:    scratch_store_b64 off, v[0:1], s0
 ; GFX12-NEXT:    s_endpgm
 entry:
-  %gep = getelementptr ptr, ptr %out, i32 4
+  %gep = getelementptr inbounds ptr, ptr %out, i32 4
   %val = atomicrmw volatile xchg ptr %gep, ptr %in syncscope("agent") seq_cst
   ret void
 }
@@ -9812,7 +9812,7 @@ define amdgpu_kernel void @atomic_xchg_i64_ret_offset(ptr %out, ptr %out2, i64 %
 ; GFX12-NEXT:    flat_store_b64 v[2:3], v[0:1]
 ; GFX12-NEXT:    s_endpgm
 entry:
-  %gep = getelementptr i64, ptr %out, i64 4
+  %gep = getelementptr inbounds i64, ptr %out, i64 4
   %tmp0 = atomicrmw volatile xchg ptr %gep, i64 %in syncscope("agent") seq_cst
   store i64 %tmp0, ptr %out2
   ret void
@@ -9952,8 +9952,8 @@ define amdgpu_kernel void @atomic_xchg_i64_addr64_offset(ptr %out, i64 %in, i64
 ; GFX12-NEXT:    scratch_store_b64 off, v[0:1], s0
 ; GFX12-NEXT:    s_endpgm
 entry:
-  %ptr = getelementptr i64, ptr %out, i64 %index
-  %gep = getelementptr i64, ptr %ptr, i64 4
+  %ptr = getelementptr inbounds i64, ptr %out, i64 %index
+  %gep = getelementptr inbounds i64, ptr %ptr, i64 4
   %tmp0 = atomicrmw volatile xchg ptr %gep, i64 %in syncscope("agent") seq_cst
   ret void
 }
@@ -10098,8 +10098,8 @@ define amdgpu_kernel void @atomic_xchg_i64_ret_addr64_offset(ptr %out, ptr %out2
 ; GFX12-NEXT:    flat_store_b64 v[2:3], v[0:1]
 ; GFX12-NEXT:    s_endpgm
 entry:
-  %ptr = getelementptr i64, ptr %out, i64 %index
-  %gep = getelementptr i64, ptr %ptr, i64 4
+  %ptr = getelementptr inbounds i64, ptr %out, i64 %index
+  %gep = getelementptr inbounds i64, ptr %ptr, i64 4
   %tmp0 = atomicrmw volatile xchg ptr %gep, i64 %in syncscope("agent") seq_cst
   store i64 %tmp0, ptr %out2
   ret void
@@ -10489,7 +10489,7 @@ define amdgpu_kernel void @atomic_xchg_i64_addr64(ptr %out, i64 %in, i64 %index)
 ; GFX12-NEXT:    scratch_store_b64 off, v[0:1], s0
 ; GFX12-NEXT:    s_endpgm
 entry:
-  %ptr = getelementptr i64, ptr %out, i64 %index
+  %ptr = getelementptr inbounds i64, ptr %out, i64 %index
   %tmp0 = atomicrmw volatile xchg ptr %ptr, i64 %in syncscope("agent") seq_cst
   ret void
 }
@@ -10629,7 +10629,7 @@ define amdgpu_kernel void @atomic_xchg_i64_ret_addr64(ptr %out, ptr %out2, i64 %
 ; GFX12-NEXT:    flat_store_b64 v[2:3], v[0:1]
 ; GFX12-NEXT:    s_endpgm
 entry:
-  %ptr = getelementptr i64, ptr %out, i64 %index
+  %ptr = getelementptr inbounds i64, ptr %out, i64 %index
   %tmp0 = atomicrmw volatile xchg ptr %ptr, i64 %in syncscope("agent") seq_cst
   store i64 %tmp0, ptr %out2
   ret void
@@ -10767,7 +10767,7 @@ define amdgpu_kernel void @atomic_xor_i64_offset(ptr %out, i64 %in) {
 ; GFX12-NEXT:    scratch_store_b64 off, v[0:1], s0
 ; GFX12-NEXT:    s_endpgm
 entry:
-  %gep = getelementptr i64, ptr %out, i64 4
+  %gep = getelementptr inbounds i64, ptr %out, i64 4
   %tmp0 = atomicrmw volatile xor ptr %gep, i64 %in syncscope("agent") seq_cst, !amdgpu.no.fine.grained.memory !0
   ret void
 }
@@ -10910,7 +10910,7 @@ define amdgpu_kernel void @atomic_xor_i64_ret_offset(ptr %out, ptr %out2, i64 %i
 ; GFX12-NEXT:    flat_store_b64 v[2:3], v[0:1]
 ; GFX12-NEXT:    s_endpgm
 entry:
-  %gep = getelementptr i64, ptr %out, i64 4
+  %gep = getelementptr inbounds i64, ptr %out, i64 4
   %tmp0 = atomicrmw volatile xor ptr %gep, i64 %in syncscope("agent") seq_cst, !amdgpu.no.fine.grained.memory !0
   store i64 %tmp0, ptr %out2
   ret void
@@ -11061,8 +11061,8 @@ define amdgpu_kernel void @atomic_xor_i64_addr64_offset(ptr %out, i64 %in, i64 %
 ; GFX12-NEXT:    scratch_store_b64 off, v[0:1], s0
 ; GFX12-NEXT:    s_endpgm
 entry:
-  %ptr = getelementptr i64, ptr %out, i64 %index
-  %gep = getelementptr i64, ptr %ptr, i64 4
+  %ptr = getelementptr inbounds i64, ptr %out, i64 %index
+  %gep = getelementptr inbounds i64, ptr %ptr, i64 4
   %tmp0 = atomicrmw volatile xor ptr %gep, i64 %in syncscope("agent") seq_cst, !amdgpu.no.fine.grained.memory !0
   ret void
 }
@@ -11210,8 +11210,8 @@ define amdgpu_kernel void @atomic_xor_i64_ret_addr64_offset(ptr %out, ptr %out2,
 ; GFX12-NEXT:    flat_store_b64 v[2:3], v[0:1]
 ; GFX12-NEXT:    s_endpgm
 entry:
-  %ptr = getelementptr i64, ptr %out, i64 %index
-  %gep = getelementptr i64, ptr %ptr, i64 4
+  %ptr = getelementptr inbounds i64, ptr %out, i64 %index
+  %gep = getelementptr inbounds i64, ptr %ptr, i64 4
   %tmp0 = atomicrmw volatile xor ptr %gep, i64 %in syncscope("agent") seq_cst, !amdgpu.no.fine.grained.memory !0
   store i64 %tmp0, ptr %out2
   ret void
@@ -11626,7 +11626,7 @@ define amdgpu_kernel void @atomic_xor_i64_addr64(ptr %out, i64 %in, i64 %index)
 ; GFX12-NEXT:    scratch_store_b64 off, v[0:1], s0
 ; GFX12-NEXT:    s_endpgm
 entry:
-  %ptr = getelementptr i64, ptr %out, i64 %index
+  %ptr = getelementptr inbounds i64, ptr %out, i64 %index
   %tmp0 = atomicrmw volatile xor ptr %ptr, i64 %in syncscope("agent") seq_cst, !amdgpu.no.fine.grained.memory !0
   ret void
 }
@@ -11769,7 +11769,7 @@ define amdgpu_kernel void @atomic_xor_i64_ret_addr64(ptr %out, ptr %out2, i64 %i
 ; GFX12-NEXT:    flat_store_b64 v[2:3], v[0:1]
 ; GFX12-NEXT:    s_endpgm
 entry:
-  %ptr = getelementptr i64, ptr %out, i64 %index
+  %ptr = getelementptr inbounds i64, ptr %out, i64 %index
   %tmp0 = atomicrmw volatile xor ptr %ptr, i64 %in syncscope("agent") seq_cst, !amdgpu.no.fine.grained.memory !0
   store i64 %tmp0, ptr %out2
   ret void
@@ -11820,7 +11820,7 @@ define amdgpu_kernel void @atomic_load_i64_offset(ptr %in, ptr %out) {
 ; GFX12-NEXT:    flat_store_b64 v[2:3], v[0:1]
 ; GFX12-NEXT:    s_endpgm
 entry:
-  %gep = getelementptr i64, ptr %in, i64 4
+  %gep = getelementptr inbounds i64, ptr %in, i64 4
   %val = load atomic i64, ptr %gep  seq_cst, align 8
   store i64 %val, ptr %out
   ret void
@@ -11930,8 +11930,8 @@ define amdgpu_kernel void @atomic_load_i64_addr64_offset(ptr %in, ptr %out, i64
 ; GFX12-NEXT:    flat_store_b64 v[2:3], v[0:1]
 ; GFX12-NEXT:    s_endpgm
 entry:
-  %ptr = getelementptr i64, ptr %in, i64 %index
-  %gep = getelementptr i64, ptr %ptr, i64 4
+  %ptr = getelementptr inbounds i64, ptr %in, i64 %index
+  %gep = getelementptr inbounds i64, ptr %ptr, i64 4
   %val = load atomic i64, ptr %gep seq_cst, align 8
   store i64 %val, ptr %out
   ret void
@@ -11991,7 +11991,7 @@ define amdgpu_kernel void @atomic_load_i64_addr64(ptr %in, ptr %out, i64 %index)
 ; GFX12-NEXT:    flat_store_b64 v[2:3], v[0:1]
 ; GFX12-NEXT:    s_endpgm
 entry:
-  %ptr = getelementptr i64, ptr %in, i64 %index
+  %ptr = getelementptr inbounds i64, ptr %in, i64 %index
   %val = load atomic i64, ptr %ptr seq_cst, align 8
   store i64 %val, ptr %out
   ret void
@@ -12035,7 +12035,7 @@ define amdgpu_kernel void @atomic_store_i64_offset(i64 %in, ptr %out) {
 ; GFX12-NEXT:    flat_store_b64 v[2:3], v[0:1] offset:32 scope:SCOPE_SYS
 ; GFX12-NEXT:    s_endpgm
 entry:
-  %gep = getelementptr i64, ptr %out, i64 4
+  %gep = getelementptr inbounds i64, ptr %out, i64 4
   store atomic i64 %in, ptr %gep  seq_cst, align 8
   ret void
 }
@@ -12129,8 +12129,8 @@ define amdgpu_kernel void @atomic_store_i64_addr64_offset(i64 %in, ptr %out, i64
 ; GFX12-NEXT:    flat_store_b64 v[2:3], v[0:1] offset:32 scope:SCOPE_SYS
 ; GFX12-NEXT:    s_endpgm
 entry:
-  %ptr = getelementptr i64, ptr %out, i64 %index
-  %gep = getelementptr i64, ptr %ptr, i64 4
+  %ptr = getelementptr inbounds i64, ptr %out, i64 %index
+  %gep = getelementptr inbounds i64, ptr %ptr, i64 4
   store atomic i64 %in, ptr %gep seq_cst, align 8
   ret void
 }
@@ -12182,7 +12182,7 @@ define amdgpu_kernel void @atomic_store_i64_addr64(i64 %in, ptr %out, i64 %index
 ; GFX12-NEXT:    flat_store_b64 v[2:3], v[0:1] scope:SCOPE_SYS
 ; GFX12-NEXT:    s_endpgm
 entry:
-  %ptr = getelementptr i64, ptr %out, i64 %index
+  %ptr = getelementptr inbounds i64, ptr %out, i64 %index
   store atomic i64 %in, ptr %ptr seq_cst, align 8
   ret void
 }
@@ -12333,7 +12333,7 @@ define amdgpu_kernel void @atomic_cmpxchg_i64_offset(ptr %out, i64 %in, i64 %old
 ; GFX12-NEXT:    scratch_store_b64 off, v[0:1], s0
 ; GFX12-NEXT:    s_endpgm
 entry:
-  %gep = getelementptr i64, ptr %out, i64 4
+  %gep = getelementptr inbounds i64, ptr %out, i64 4
   %val = cmpxchg volatile ptr %gep, i64 %old, i64 %in syncscope("agent") seq_cst seq_cst
   ret void
 }
@@ -12484,7 +12484,7 @@ define amdgpu_kernel void @atomic_cmpxchg_i64_soffset(ptr %out, i64 %in, i64 %ol
 ; GFX12-NEXT:    scratch_store_b64 off, v[0:1], s0
 ; GFX12-NEXT:    s_endpgm
 entry:
-  %gep = getelementptr i64, ptr %out, i64 9000
+  %gep = getelementptr inbounds i64, ptr %out, i64 9000
   %val = cmpxchg volatile ptr %gep, i64 %old, i64 %in syncscope("agent") seq_cst seq_cst
   ret void
 }
@@ -12633,7 +12633,7 @@ define amdgpu_kernel void @atomic_cmpxchg_i64_ret_offset(ptr %out, ptr %out2, i6
 ; GFX12-NEXT:    flat_store_b64 v[2:3], v[0:1]
 ; GFX12-NEXT:    s_endpgm
 entry:
-  %gep = getelementptr i64, ptr %out, i64 4
+  %gep = getelementptr inbounds i64, ptr %out, i64 4
   %val = cmpxchg volatile ptr %gep, i64 %old, i64 %in syncscope("agent") seq_cst seq_cst
   %extract0 = extractvalue { i64, i1 } %val, 0
   store i64 %extract0, ptr %out2
@@ -12791,8 +12791,8 @@ define amdgpu_kernel void @atomic_cmpxchg_i64_addr64_offset(ptr %out, i64 %in, i
 ; GFX12-NEXT:    scratch_store_b64 off, v[0:1], s0
 ; GFX12-NEXT:    s_endpgm
 entry:
-  %ptr = getelementptr i64, ptr %out, i64 %index
-  %gep = getelementptr i64, ptr %ptr, i64 4
+  %ptr = getelementptr inbounds i64, ptr %out, i64 %index
+  %gep = getelementptr inbounds i64, ptr %ptr, i64 4
   %val = cmpxchg volatile ptr %gep, i64 %old, i64 %in syncscope("agent") seq_cst seq_cst
   ret void
 }
@@ -12954,8 +12954,8 @@ define amdgpu_kernel void @atomic_cmpxchg_i64_ret_addr64_offset(ptr %out, ptr %o
 ; GFX12-NEXT:    flat_store_b64 v[2:3], v[0:1]
 ; GFX12-NEXT:    s_endpgm
 entry:
-  %ptr = getelementptr i64, ptr %out, i64 %index
-  %gep = getelementptr i64, ptr %ptr, i64 4
+  %ptr = getelementptr inbounds i64, ptr %out, i64 %index
+  %gep = getelementptr inbounds i64, ptr %ptr, i64 4
   %val = cmpxchg volatile ptr %gep, i64 %old, i64 %in syncscope("agent") seq_cst seq_cst
   %extract0 = extractvalue { i64, i1 } %val, 0
   store i64 %extract0, ptr %out2
@@ -13398,7 +13398,7 @@ define amdgpu_kernel void @atomic_cmpxchg_i64_addr64(ptr %out, i64 %in, i64 %ind
 ; GFX12-NEXT:    scratch_store_b64 off, v[0:1], s0
 ; GFX12-NEXT:    s_endpgm
 entry:
-  %ptr = getelementptr i64, ptr %out, i64 %index
+  %ptr = getelementptr inbounds i64, ptr %out, i64 %index
   %val = cmpxchg volatile ptr %ptr, i64 %old, i64 %in syncscope("agent") seq_cst seq_cst
   ret void
 }
@@ -13555,7 +13555,7 @@ define amdgpu_kernel void @atomic_cmpxchg_i64_ret_addr64(ptr %out, ptr %out2, i6
 ; GFX12-NEXT:    flat_store_b64 v[2:3], v[0:1]
 ; GFX12-NEXT:    s_endpgm
 entry:
-  %ptr = getelementptr i64, ptr %out, i64 %index
+  %ptr = getelementptr inbounds i64, ptr %out, i64 %index
   %val = cmpxchg volatile ptr %ptr, i64 %old, i64 %in syncscope("agent") seq_cst seq_cst
   %extract0 = extractvalue { i64, i1 } %val, 0
   store i64 %extract0, ptr %out2
@@ -13607,7 +13607,7 @@ define amdgpu_kernel void @atomic_load_f64_offset(ptr %in, ptr %out) {
 ; GFX12-NEXT:    flat_store_b64 v[2:3], v[0:1]
 ; GFX12-NEXT:    s_endpgm
 entry:
-  %gep = getelementptr double, ptr %in, i64 4
+  %gep = getelementptr inbounds double, ptr %in, i64 4
   %val = load atomic double, ptr %gep  seq_cst, align 8
   store double %val, ptr %out
   ret void
@@ -13717,8 +13717,8 @@ define amdgpu_kernel void @atomic_load_f64_addr64_offset(ptr %in, ptr %out, i64
 ; GFX12-NEXT:    flat_store_b64 v[2:3], v[0:1]
 ; GFX12-NEXT:    s_endpgm
 entry:
-  %ptr = getelementptr double, ptr %in, i64 %index
-  %gep = getelementptr double, ptr %ptr, i64 4
+  %ptr = getelementptr inbounds double, ptr %in, i64 %index
+  %gep = getelementptr inbounds double, ptr %ptr, i64 4
   %val = load atomic double, ptr %gep seq_cst, align 8
   store double %val, ptr %out
   ret void
@@ -13778,7 +13778,7 @@ define amdgpu_kernel void @atomic_load_f64_addr64(ptr %in, ptr %out, i64 %index)
 ; GFX12-NEXT:    flat_store_b64 v[2:3], v[0:1]
 ; GFX12-NEXT:    s_endpgm
 entry:
-  %ptr = getelementptr double, ptr %in, i64 %index
+  %ptr = getelementptr inbounds double, ptr %in, i64 %index
   %val = load atomic double, ptr %ptr seq_cst, align 8
   store double %val, ptr %out
   ret void
@@ -13822,7 +13822,7 @@ define amdgpu_kernel void @atomic_store_f64_offset(double %in, ptr %out) {
 ; GFX12-NEXT:    flat_store_b64 v[2:3], v[0:1] offset:32 scope:SCOPE_SYS
 ; GFX12-NEXT:    s_endpgm
 entry:
-  %gep = getelementptr double, ptr %out, i64 4
+  %gep = getelementptr inbounds double, ptr %out, i64 4
   store atomic double %in, ptr %gep  seq_cst, align 8
   ret void
 }
@@ -13916,8 +13916,8 @@ define amdgpu_kernel void @atomic_store_f64_addr64_offset(double %in, ptr %out,
 ; GFX12-NEXT:    flat_store_b64 v[2:3], v[0:1] offset:32 scope:SCOPE_SYS
 ; GFX12-NEXT:    s_endpgm
 entry:
-  %ptr = getelementptr double, ptr %out, i64 %index
-  %gep = getelementptr double, ptr %ptr, i64 4
+  %ptr = getelementptr inbounds double, ptr %out, i64 %index
+  %gep = getelementptr inbounds double, ptr %ptr, i64 4
   store atomic double %in, ptr %gep seq_cst, align 8
   ret void
 }
@@ -13969,7 +13969,7 @@ define amdgpu_kernel void @atomic_store_f64_addr64(double %in, ptr %out, i64 %in
 ; GFX12-NEXT:    flat_store_b64 v[2:3], v[0:1] scope:SCOPE_SYS
 ; GFX12-NEXT:    s_endpgm
 entry:
-  %ptr = getelementptr double, ptr %out, i64 %index
+  %ptr = getelementptr inbounds double, ptr %out, i64 %index
   store atomic double %in, ptr %ptr seq_cst, align 8
   ret void
 }
@@ -14116,7 +14116,7 @@ define amdgpu_kernel void @atomic_inc_i64_offset(ptr %out, i64 %in) {
 ; GFX12-NEXT:    scratch_store_b64 off, v[0:1], s0
 ; GFX12-NEXT:    s_endpgm
 entry:
-  %gep = getelementptr i64, ptr %out, i64 4
+  %gep = getelementptr inbounds i64, ptr %out, i64 4
   %tmp0 = atomicrmw volatile uinc_wrap ptr %gep, i64 %in syncscope("agent") seq_cst, !amdgpu.no.fine.grained.memory !0
   ret void
 }
@@ -14269,7 +14269,7 @@ define amdgpu_kernel void @atomic_inc_i64_ret_offset(ptr %out, ptr %out2, i64 %i
 ; GFX12-NEXT:    flat_store_b64 v[2:3], v[0:1]
 ; GFX12-NEXT:    s_endpgm
 entry:
-  %gep = getelementptr i64, ptr %out, i64 4
+  %gep = getelementptr inbounds i64, ptr %out, i64 4
   %tmp0 = atomicrmw volatile uinc_wrap ptr %gep, i64 %in syncscope("agent") seq_cst, !amdgpu.no.fine.grained.memory !0
   store i64 %tmp0, ptr %out2
   ret void
@@ -14430,8 +14430,8 @@ define amdgpu_kernel void @atomic_inc_i64_incr64_offset(ptr %out, i64 %in, i64 %
 ; GFX12-NEXT:    scratch_store_b64 off, v[0:1], s0
 ; GFX12-NEXT:    s_endpgm
 entry:
-  %ptr = getelementptr i64, ptr %out, i64 %index
-  %gep = getelementptr i64, ptr %ptr, i64 4
+  %ptr = getelementptr inbounds i64, ptr %out, i64 %index
+  %gep = getelementptr inbounds i64, ptr %ptr, i64 4
   %tmp0 = atomicrmw volatile uinc_wrap ptr %gep, i64 %in syncscope("agent") seq_cst, !amdgpu.no.fine.grained.memory !0
   ret void
 }
@@ -14589,8 +14589,8 @@ define amdgpu_kernel void @atomic_inc_i64_ret_incr64_offset(ptr %out, ptr %out2,
 ; GFX12-NEXT:    flat_store_b64 v[2:3], v[0:1]
 ; GFX12-NEXT:    s_endpgm
 entry:
-  %ptr = getelementptr i64, ptr %out, i64 %index
-  %gep = getelementptr i64, ptr %ptr, i64 4
+  %ptr = getelementptr inbounds i64, ptr %out, i64 %index
+  %gep = getelementptr inbounds i64, ptr %ptr, i64 4
   %tmp0 = atomicrmw volatile uinc_wrap ptr %gep, i64 %in syncscope("agent") seq_cst, !amdgpu.no.fine.grained.memory !0
   store i64 %tmp0, ptr %out2
   ret void
@@ -15035,7 +15035,7 @@ define amdgpu_kernel void @atomic_inc_i64_incr64(ptr %out, i64 %in, i64 %index)
 ; GFX12-NEXT:    scratch_store_b64 off, v[0:1], s0
 ; GFX12-NEXT:    s_endpgm
 entry:
-  %ptr = getelementptr i64, ptr %out, i64 %index
+  %ptr = getelementptr inbounds i64, ptr %out, i64 %index
   %tmp0 = atomicrmw volatile uinc_wrap ptr %ptr, i64 %in syncscope("agent") seq_cst, !amdgpu.no.fine.grained.memory !0
   ret void
 }
@@ -15188,7 +15188,7 @@ define amdgpu_kernel void @atomic_inc_i64_ret_incr64(ptr %out, ptr %out2, i64 %i
 ; GFX12-NEXT:    flat_store_b64 v[2:3], v[0:1]
 ; GFX12-NEXT:    s_endpgm
 entry:
-  %ptr = getelementptr i64, ptr %out, i64 %index
+  %ptr = getelementptr inbounds i64, ptr %out, i64 %index
   %tmp0 = atomicrmw volatile uinc_wrap ptr %ptr, i64 %in syncscope("agent") seq_cst, !amdgpu.no.fine.grained.memory !0
   store i64 %tmp0, ptr %out2
   ret void
@@ -15345,7 +15345,7 @@ define amdgpu_kernel void @atomic_dec_i64_offset(ptr %out, i64 %in) {
 ; GFX12-NEXT:    scratch_store_b64 off, v[0:1], s4
 ; GFX12-NEXT:    s_endpgm
 entry:
-  %gep = getelementptr i64, ptr %out, i64 4
+  %gep = getelementptr inbounds i64, ptr %out, i64 4
   %tmp0 = atomicrmw volatile udec_wrap ptr %gep, i64 %in syncscope("agent") seq_cst, !amdgpu.no.fine.grained.memory !0
   ret void
 }
@@ -15509,7 +15509,7 @@ define amdgpu_kernel void @atomic_dec_i64_ret_offset(ptr %out, ptr %out2, i64 %i
 ; GFX12-NEXT:    flat_store_b64 v[2:3], v[0:1]
 ; GFX12-NEXT:    s_endpgm
 entry:
-  %gep = getelementptr i64, ptr %out, i64 4
+  %gep = getelementptr inbounds i64, ptr %out, i64 4
   %tmp0 = atomicrmw volatile udec_wrap ptr %gep, i64 %in syncscope("agent") seq_cst, !amdgpu.no.fine.grained.memory !0
   store i64 %tmp0, ptr %out2
   ret void
@@ -15679,8 +15679,8 @@ define amdgpu_kernel void @atomic_dec_i64_decr64_offset(ptr %out, i64 %in, i64 %
 ; GFX12-NEXT:    scratch_store_b64 off, v[0:1], s4
 ; GFX12-NEXT:    s_endpgm
 entry:
-  %ptr = getelementptr i64, ptr %out, i64 %index
-  %gep = getelementptr i64, ptr %ptr, i64 4
+  %ptr = getelementptr inbounds i64, ptr %out, i64 %index
+  %gep = getelementptr inbounds i64, ptr %ptr, i64 4
   %tmp0 = atomicrmw volatile udec_wrap ptr %gep, i64 %in syncscope("agent") seq_cst, !amdgpu.no.fine.grained.memory !0
   ret void
 }
@@ -15849,8 +15849,8 @@ define amdgpu_kernel void @atomic_dec_i64_ret_decr64_offset(ptr %out, ptr %out2,
 ; GFX12-NEXT:    flat_store_b64 v[2:3], v[0:1]
 ; GFX12-NEXT:    s_endpgm
 entry:
-  %ptr = getelementptr i64, ptr %out, i64 %index
-  %gep = getelementptr i64, ptr %ptr, i64 4
+  %ptr = getelementptr inbounds i64, ptr %out, i64 %index
+  %gep = getelementptr inbounds i64, ptr %ptr, i64 4
   %tmp0 = atomicrmw volatile udec_wrap ptr %gep, i64 %in syncscope("agent") seq_cst, !amdgpu.no.fine.grained.memory !0
   store i64 %tmp0, ptr %out2
   ret void
@@ -16324,7 +16324,7 @@ define amdgpu_kernel void @atomic_dec_i64_decr64(ptr %out, i64 %in, i64 %index)
 ; GFX12-NEXT:    scratch_store_b64 off, v[0:1], s4
 ; GFX12-NEXT:    s_endpgm
 entry:
-  %ptr = getelementptr i64, ptr %out, i64 %index
+  %ptr = getelementptr inbounds i64, ptr %out, i64 %index
   %tmp0 = atomicrmw volatile udec_wrap ptr %ptr, i64 %in syncscope("agent") seq_cst, !amdgpu.no.fine.grained.memory !0
   ret void
 }
@@ -16488,7 +16488,7 @@ define amdgpu_kernel void @atomic_dec_i64_ret_decr64(ptr %out, ptr %out2, i64 %i
 ; GFX12-NEXT:    flat_store_b64 v[2:3], v[0:1]
 ; GFX12-NEXT:    s_endpgm
 entry:
-  %ptr = getelementptr i64, ptr %out, i64 %index
+  %ptr = getelementptr inbounds i64, ptr %out, i64 %index
   %tmp0 = atomicrmw volatile udec_wrap ptr %ptr, i64 %in syncscope("agent") seq_cst, !amdgpu.no.fine.grained.memory !0
   store i64 %tmp0, ptr %out2
   ret void
diff --git a/llvm/test/CodeGen/AMDGPU/flat_atomics_i64_noprivate.ll b/llvm/test/CodeGen/AMDGPU/flat_atomics_i64_noprivate.ll
index 757649ca592b3..e5187a811a230 100644
--- a/llvm/test/CodeGen/AMDGPU/flat_atomics_i64_noprivate.ll
+++ b/llvm/test/CodeGen/AMDGPU/flat_atomics_i64_noprivate.ll
@@ -45,7 +45,7 @@ define amdgpu_kernel void @atomic_add_i64_offset(ptr %out, i64 %in) {
 ; GFX12-NEXT:    global_inv scope:SCOPE_DEV
 ; GFX12-NEXT:    s_endpgm
 entry:
-  %gep = getelementptr i64, ptr %out, i64 4
+  %gep = getelementptr inbounds i64, ptr %out, i64 4
   %tmp0 = atomicrmw volatile add ptr %gep, i64 %in syncscope("agent") seq_cst, !noalias.addrspace !0
   ret void
 }
@@ -104,7 +104,7 @@ define amdgpu_kernel void @atomic_add_i64_ret_offset(ptr %out, ptr %out2, i64 %i
 ; GFX12-NEXT:    flat_store_b64 v[2:3], v[0:1]
 ; GFX12-NEXT:    s_endpgm
 entry:
-  %gep = getelementptr i64, ptr %out, i64 4
+  %gep = getelementptr inbounds i64, ptr %out, i64 4
   %tmp0 = atomicrmw volatile add ptr %gep, i64 %in syncscope("agent") seq_cst, !noalias.addrspace !0
   store i64 %tmp0, ptr %out2
   ret void
@@ -165,8 +165,8 @@ define amdgpu_kernel void @atomic_add_i64_addr64_offset(ptr %out, i64 %in, i64 %
 ; GFX12-NEXT:    global_inv scope:SCOPE_DEV
 ; GFX12-NEXT:    s_endpgm
 entry:
-  %ptr = getelementptr i64, ptr %out, i64 %index
-  %gep = getelementptr i64, ptr %ptr, i64 4
+  %ptr = getelementptr inbounds i64, ptr %out, i64 %index
+  %gep = getelementptr inbounds i64, ptr %ptr, i64 4
   %tmp0 = atomicrmw volatile add ptr %gep, i64 %in syncscope("agent") seq_cst, !noalias.addrspace !0
   ret void
 }
@@ -230,8 +230,8 @@ define amdgpu_kernel void @atomic_add_i64_ret_addr64_offset(ptr %out, ptr %out2,
 ; GFX12-NEXT:    flat_store_b64 v[2:3], v[0:1]
 ; GFX12-NEXT:    s_endpgm
 entry:
-  %ptr = getelementptr i64, ptr %out, i64 %index
-  %gep = getelementptr i64, ptr %ptr, i64 4
+  %ptr = getelementptr inbounds i64, ptr %out, i64 %index
+  %gep = getelementptr inbounds i64, ptr %ptr, i64 4
   %tmp0 = atomicrmw volatile add ptr %gep, i64 %in syncscope("agent") seq_cst, !noalias.addrspace !0
   store i64 %tmp0, ptr %out2
   ret void
@@ -385,7 +385,7 @@ define amdgpu_kernel void @atomic_add_i64_addr64(ptr %out, i64 %in, i64 %index)
 ; GFX12-NEXT:    global_inv scope:SCOPE_DEV
 ; GFX12-NEXT:    s_endpgm
 entry:
-  %ptr = getelementptr i64, ptr %out, i64 %index
+  %ptr = getelementptr inbounds i64, ptr %out, i64 %index
   %tmp0 = atomicrmw volatile add ptr %ptr, i64 %in syncscope("agent") seq_cst, !noalias.addrspace !0
   ret void
 }
@@ -445,7 +445,7 @@ define amdgpu_kernel void @atomic_add_i64_ret_addr64(ptr %out, ptr %out2, i64 %i
 ; GFX12-NEXT:    flat_store_b64 v[2:3], v[0:1]
 ; GFX12-NEXT:    s_endpgm
 entry:
-  %ptr = getelementptr i64, ptr %out, i64 %index
+  %ptr = getelementptr inbounds i64, ptr %out, i64 %index
   %tmp0 = atomicrmw volatile add ptr %ptr, i64 %in syncscope("agent") seq_cst, !noalias.addrspace !0
   store i64 %tmp0, ptr %out2
   ret void
@@ -517,7 +517,7 @@ define amdgpu_kernel void @atomic_and_i64_offset(ptr %out, i64 %in) {
 ; GFX12-NEXT:    global_inv scope:SCOPE_DEV
 ; GFX12-NEXT:    s_endpgm
 entry:
-  %gep = getelementptr i64, ptr %out, i64 4
+  %gep = getelementptr inbounds i64, ptr %out, i64 4
   %tmp0 = atomicrmw volatile and ptr %gep, i64 %in syncscope("agent") seq_cst, !noalias.addrspace !0
   ret void
 }
@@ -602,7 +602,7 @@ define amdgpu_kernel void @atomic_and_i64_ret_offset(ptr %out, ptr %out2, i64 %i
 ; GFX12-NEXT:    flat_store_b64 v[2:3], v[0:1]
 ; GFX12-NEXT:    s_endpgm
 entry:
-  %gep = getelementptr i64, ptr %out, i64 4
+  %gep = getelementptr inbounds i64, ptr %out, i64 4
   %tmp0 = atomicrmw volatile and ptr %gep, i64 %in syncscope("agent") seq_cst, !noalias.addrspace !0
   store i64 %tmp0, ptr %out2
   ret void
@@ -687,8 +687,8 @@ define amdgpu_kernel void @atomic_and_i64_addr64_offset(ptr %out, i64 %in, i64 %
 ; GFX12-NEXT:    global_inv scope:SCOPE_DEV
 ; GFX12-NEXT:    s_endpgm
 entry:
-  %ptr = getelementptr i64, ptr %out, i64 %index
-  %gep = getelementptr i64, ptr %ptr, i64 4
+  %ptr = getelementptr inbounds i64, ptr %out, i64 %index
+  %gep = getelementptr inbounds i64, ptr %ptr, i64 4
   %tmp0 = atomicrmw volatile and ptr %gep, i64 %in syncscope("agent") seq_cst, !noalias.addrspace !0
   ret void
 }
@@ -778,8 +778,8 @@ define amdgpu_kernel void @atomic_and_i64_ret_addr64_offset(ptr %out, ptr %out2,
 ; GFX12-NEXT:    flat_store_b64 v[2:3], v[0:1]
 ; GFX12-NEXT:    s_endpgm
 entry:
-  %ptr = getelementptr i64, ptr %out, i64 %index
-  %gep = getelementptr i64, ptr %ptr, i64 4
+  %ptr = getelementptr inbounds i64, ptr %out, i64 %index
+  %gep = getelementptr inbounds i64, ptr %ptr, i64 4
   %tmp0 = atomicrmw volatile and ptr %gep, i64 %in syncscope("agent") seq_cst, !noalias.addrspace !0
   store i64 %tmp0, ptr %out2
   ret void
@@ -1015,7 +1015,7 @@ define amdgpu_kernel void @atomic_and_i64_addr64(ptr %out, i64 %in, i64 %index)
 ; GFX12-NEXT:    global_inv scope:SCOPE_DEV
 ; GFX12-NEXT:    s_endpgm
 entry:
-  %ptr = getelementptr i64, ptr %out, i64 %index
+  %ptr = getelementptr inbounds i64, ptr %out, i64 %index
   %tmp0 = atomicrmw volatile and ptr %ptr, i64 %in syncscope("agent") seq_cst, !noalias.addrspace !0
   ret void
 }
@@ -1101,7 +1101,7 @@ define amdgpu_kernel void @atomic_and_i64_ret_addr64(ptr %out, ptr %out2, i64 %i
 ; GFX12-NEXT:    flat_store_b64 v[2:3], v[0:1]
 ; GFX12-NEXT:    s_endpgm
 entry:
-  %ptr = getelementptr i64, ptr %out, i64 %index
+  %ptr = getelementptr inbounds i64, ptr %out, i64 %index
   %tmp0 = atomicrmw volatile and ptr %ptr, i64 %in syncscope("agent") seq_cst, !noalias.addrspace !0
   store i64 %tmp0, ptr %out2
   ret void
@@ -1175,7 +1175,7 @@ define amdgpu_kernel void @atomic_sub_i64_offset(ptr %out, i64 %in) {
 ; GFX12-NEXT:    global_inv scope:SCOPE_DEV
 ; GFX12-NEXT:    s_endpgm
 entry:
-  %gep = getelementptr i64, ptr %out, i64 4
+  %gep = getelementptr inbounds i64, ptr %out, i64 4
   %tmp0 = atomicrmw volatile sub ptr %gep, i64 %in syncscope("agent") seq_cst, !noalias.addrspace !0
   ret void
 }
@@ -1262,7 +1262,7 @@ define amdgpu_kernel void @atomic_sub_i64_ret_offset(ptr %out, ptr %out2, i64 %i
 ; GFX12-NEXT:    flat_store_b64 v[2:3], v[0:1]
 ; GFX12-NEXT:    s_endpgm
 entry:
-  %gep = getelementptr i64, ptr %out, i64 4
+  %gep = getelementptr inbounds i64, ptr %out, i64 4
   %tmp0 = atomicrmw volatile sub ptr %gep, i64 %in syncscope("agent") seq_cst, !noalias.addrspace !0
   store i64 %tmp0, ptr %out2
   ret void
@@ -1349,8 +1349,8 @@ define amdgpu_kernel void @atomic_sub_i64_addr64_offset(ptr %out, i64 %in, i64 %
 ; GFX12-NEXT:    global_inv scope:SCOPE_DEV
 ; GFX12-NEXT:    s_endpgm
 entry:
-  %ptr = getelementptr i64, ptr %out, i64 %index
-  %gep = getelementptr i64, ptr %ptr, i64 4
+  %ptr = getelementptr inbounds i64, ptr %out, i64 %index
+  %gep = getelementptr inbounds i64, ptr %ptr, i64 4
   %tmp0 = atomicrmw volatile sub ptr %gep, i64 %in syncscope("agent") seq_cst, !noalias.addrspace !0
   ret void
 }
@@ -1442,8 +1442,8 @@ define amdgpu_kernel void @atomic_sub_i64_ret_addr64_offset(ptr %out, ptr %out2,
 ; GFX12-NEXT:    flat_store_b64 v[2:3], v[0:1]
 ; GFX12-NEXT:    s_endpgm
 entry:
-  %ptr = getelementptr i64, ptr %out, i64 %index
-  %gep = getelementptr i64, ptr %ptr, i64 4
+  %ptr = getelementptr inbounds i64, ptr %out, i64 %index
+  %gep = getelementptr inbounds i64, ptr %ptr, i64 4
   %tmp0 = atomicrmw volatile sub ptr %gep, i64 %in syncscope("agent") seq_cst, !noalias.addrspace !0
   store i64 %tmp0, ptr %out2
   ret void
@@ -1685,7 +1685,7 @@ define amdgpu_kernel void @atomic_sub_i64_addr64(ptr %out, i64 %in, i64 %index)
 ; GFX12-NEXT:    global_inv scope:SCOPE_DEV
 ; GFX12-NEXT:    s_endpgm
 entry:
-  %ptr = getelementptr i64, ptr %out, i64 %index
+  %ptr = getelementptr inbounds i64, ptr %out, i64 %index
   %tmp0 = atomicrmw volatile sub ptr %ptr, i64 %in syncscope("agent") seq_cst, !noalias.addrspace !0
   ret void
 }
@@ -1773,7 +1773,7 @@ define amdgpu_kernel void @atomic_sub_i64_ret_addr64(ptr %out, ptr %out2, i64 %i
 ; GFX12-NEXT:    flat_store_b64 v[2:3], v[0:1]
 ; GFX12-NEXT:    s_endpgm
 entry:
-  %ptr = getelementptr i64, ptr %out, i64 %index
+  %ptr = getelementptr inbounds i64, ptr %out, i64 %index
   %tmp0 = atomicrmw volatile sub ptr %ptr, i64 %in syncscope("agent") seq_cst, !noalias.addrspace !0
   store i64 %tmp0, ptr %out2
   ret void
@@ -1849,7 +1849,7 @@ define amdgpu_kernel void @atomic_max_i64_offset(ptr %out, i64 %in) {
 ; GFX12-NEXT:    global_inv scope:SCOPE_SE
 ; GFX12-NEXT:    s_endpgm
 entry:
-  %gep = getelementptr i64, ptr %out, i64 4
+  %gep = getelementptr inbounds i64, ptr %out, i64 4
   %tmp0 = atomicrmw volatile max ptr %gep, i64 %in syncscope("workgroup") seq_cst, !noalias.addrspace !0
   ret void
 }
@@ -1938,7 +1938,7 @@ define amdgpu_kernel void @atomic_max_i64_ret_offset(ptr %out, ptr %out2, i64 %i
 ; GFX12-NEXT:    flat_store_b64 v[2:3], v[0:1]
 ; GFX12-NEXT:    s_endpgm
 entry:
-  %gep = getelementptr i64, ptr %out, i64 4
+  %gep = getelementptr inbounds i64, ptr %out, i64 4
   %tmp0 = atomicrmw volatile max ptr %gep, i64 %in syncscope("workgroup") seq_cst, !noalias.addrspace !0
   store i64 %tmp0, ptr %out2
   ret void
@@ -2027,8 +2027,8 @@ define amdgpu_kernel void @atomic_max_i64_addr64_offset(ptr %out, i64 %in, i64 %
 ; GFX12-NEXT:    global_inv scope:SCOPE_SE
 ; GFX12-NEXT:    s_endpgm
 entry:
-  %ptr = getelementptr i64, ptr %out, i64 %index
-  %gep = getelementptr i64, ptr %ptr, i64 4
+  %ptr = getelementptr inbounds i64, ptr %out, i64 %index
+  %gep = getelementptr inbounds i64, ptr %ptr, i64 4
   %tmp0 = atomicrmw volatile max ptr %gep, i64 %in syncscope("workgroup") seq_cst, !noalias.addrspace !0
   ret void
 }
@@ -2122,8 +2122,8 @@ define amdgpu_kernel void @atomic_max_i64_ret_addr64_offset(ptr %out, ptr %out2,
 ; GFX12-NEXT:    flat_store_b64 v[2:3], v[0:1]
 ; GFX12-NEXT:    s_endpgm
 entry:
-  %ptr = getelementptr i64, ptr %out, i64 %index
-  %gep = getelementptr i64, ptr %ptr, i64 4
+  %ptr = getelementptr inbounds i64, ptr %out, i64 %index
+  %gep = getelementptr inbounds i64, ptr %ptr, i64 4
   %tmp0 = atomicrmw volatile max ptr %gep, i64 %in syncscope("workgroup") seq_cst, !noalias.addrspace !0
   store i64 %tmp0, ptr %out2
   ret void
@@ -2371,7 +2371,7 @@ define amdgpu_kernel void @atomic_max_i64_addr64(ptr %out, i64 %in, i64 %index)
 ; GFX12-NEXT:    global_inv scope:SCOPE_SE
 ; GFX12-NEXT:    s_endpgm
 entry:
-  %ptr = getelementptr i64, ptr %out, i64 %index
+  %ptr = getelementptr inbounds i64, ptr %out, i64 %index
   %tmp0 = atomicrmw volatile max ptr %ptr, i64 %in syncscope("workgroup") seq_cst, !noalias.addrspace !0
   ret void
 }
@@ -2461,7 +2461,7 @@ define amdgpu_kernel void @atomic_max_i64_ret_addr64(ptr %out, ptr %out2, i64 %i
 ; GFX12-NEXT:    flat_store_b64 v[2:3], v[0:1]
 ; GFX12-NEXT:    s_endpgm
 entry:
-  %ptr = getelementptr i64, ptr %out, i64 %index
+  %ptr = getelementptr inbounds i64, ptr %out, i64 %index
   %tmp0 = atomicrmw volatile max ptr %ptr, i64 %in syncscope("workgroup") seq_cst, !noalias.addrspace !0
   store i64 %tmp0, ptr %out2
   ret void
@@ -2537,7 +2537,7 @@ define amdgpu_kernel void @atomic_umax_i64_offset(ptr %out, i64 %in) {
 ; GFX12-NEXT:    global_inv scope:SCOPE_SE
 ; GFX12-NEXT:    s_endpgm
 entry:
-  %gep = getelementptr i64, ptr %out, i64 4
+  %gep = getelementptr inbounds i64, ptr %out, i64 4
   %tmp0 = atomicrmw volatile umax ptr %gep, i64 %in syncscope("workgroup") seq_cst, !noalias.addrspace !0
   ret void
 }
@@ -2626,7 +2626,7 @@ define amdgpu_kernel void @atomic_umax_i64_ret_offset(ptr %out, ptr %out2, i64 %
 ; GFX12-NEXT:    flat_store_b64 v[2:3], v[0:1]
 ; GFX12-NEXT:    s_endpgm
 entry:
-  %gep = getelementptr i64, ptr %out, i64 4
+  %gep = getelementptr inbounds i64, ptr %out, i64 4
   %tmp0 = atomicrmw volatile umax ptr %gep, i64 %in syncscope("workgroup") seq_cst, !noalias.addrspace !0
   store i64 %tmp0, ptr %out2
   ret void
@@ -2715,8 +2715,8 @@ define amdgpu_kernel void @atomic_umax_i64_addr64_offset(ptr %out, i64 %in, i64
 ; GFX12-NEXT:    global_inv scope:SCOPE_SE
 ; GFX12-NEXT:    s_endpgm
 entry:
-  %ptr = getelementptr i64, ptr %out, i64 %index
-  %gep = getelementptr i64, ptr %ptr, i64 4
+  %ptr = getelementptr inbounds i64, ptr %out, i64 %index
+  %gep = getelementptr inbounds i64, ptr %ptr, i64 4
   %tmp0 = atomicrmw volatile umax ptr %gep, i64 %in syncscope("workgroup") seq_cst, !noalias.addrspace !0
   ret void
 }
@@ -2810,8 +2810,8 @@ define amdgpu_kernel void @atomic_umax_i64_ret_addr64_offset(ptr %out, ptr %out2
 ; GFX12-NEXT:    flat_store_b64 v[2:3], v[0:1]
 ; GFX12-NEXT:    s_endpgm
 entry:
-  %ptr = getelementptr i64, ptr %out, i64 %index
-  %gep = getelementptr i64, ptr %ptr, i64 4
+  %ptr = getelementptr inbounds i64, ptr %out, i64 %index
+  %gep = getelementptr inbounds i64, ptr %ptr, i64 4
   %tmp0 = atomicrmw volatile umax ptr %gep, i64 %in syncscope("workgroup") seq_cst, !noalias.addrspace !0
   store i64 %tmp0, ptr %out2
   ret void
@@ -3059,7 +3059,7 @@ define amdgpu_kernel void @atomic_umax_i64_addr64(ptr %out, i64 %in, i64 %index)
 ; GFX12-NEXT:    global_inv scope:SCOPE_SE
 ; GFX12-NEXT:    s_endpgm
 entry:
-  %ptr = getelementptr i64, ptr %out, i64 %index
+  %ptr = getelementptr inbounds i64, ptr %out, i64 %index
   %tmp0 = atomicrmw volatile umax ptr %ptr, i64 %in syncscope("workgroup") seq_cst, !noalias.addrspace !0
   ret void
 }
@@ -3149,7 +3149,7 @@ define amdgpu_kernel void @atomic_umax_i64_ret_addr64(ptr %out, ptr %out2, i64 %
 ; GFX12-NEXT:    flat_store_b64 v[2:3], v[0:1]
 ; GFX12-NEXT:    s_endpgm
 entry:
-  %ptr = getelementptr i64, ptr %out, i64 %index
+  %ptr = getelementptr inbounds i64, ptr %out, i64 %index
   %tmp0 = atomicrmw volatile umax ptr %ptr, i64 %in syncscope("workgroup") seq_cst, !noalias.addrspace !0
   store i64 %tmp0, ptr %out2
   ret void
@@ -3225,7 +3225,7 @@ define amdgpu_kernel void @atomic_min_i64_offset(ptr %out, i64 %in) {
 ; GFX12-NEXT:    global_inv scope:SCOPE_SE
 ; GFX12-NEXT:    s_endpgm
 entry:
-  %gep = getelementptr i64, ptr %out, i64 4
+  %gep = getelementptr inbounds i64, ptr %out, i64 4
   %tmp0 = atomicrmw volatile min ptr %gep, i64 %in syncscope("workgroup") seq_cst, !noalias.addrspace !0
   ret void
 }
@@ -3314,7 +3314,7 @@ define amdgpu_kernel void @atomic_min_i64_ret_offset(ptr %out, ptr %out2, i64 %i
 ; GFX12-NEXT:    flat_store_b64 v[2:3], v[0:1]
 ; GFX12-NEXT:    s_endpgm
 entry:
-  %gep = getelementptr i64, ptr %out, i64 4
+  %gep = getelementptr inbounds i64, ptr %out, i64 4
   %tmp0 = atomicrmw volatile min ptr %gep, i64 %in syncscope("workgroup") seq_cst, !noalias.addrspace !0
   store i64 %tmp0, ptr %out2
   ret void
@@ -3403,8 +3403,8 @@ define amdgpu_kernel void @atomic_min_i64_addr64_offset(ptr %out, i64 %in, i64 %
 ; GFX12-NEXT:    global_inv scope:SCOPE_SE
 ; GFX12-NEXT:    s_endpgm
 entry:
-  %ptr = getelementptr i64, ptr %out, i64 %index
-  %gep = getelementptr i64, ptr %ptr, i64 4
+  %ptr = getelementptr inbounds i64, ptr %out, i64 %index
+  %gep = getelementptr inbounds i64, ptr %ptr, i64 4
   %tmp0 = atomicrmw volatile min ptr %gep, i64 %in syncscope("workgroup") seq_cst, !noalias.addrspace !0
   ret void
 }
@@ -3498,8 +3498,8 @@ define amdgpu_kernel void @atomic_min_i64_ret_addr64_offset(ptr %out, ptr %out2,
 ; GFX12-NEXT:    flat_store_b64 v[2:3], v[0:1]
 ; GFX12-NEXT:    s_endpgm
 entry:
-  %ptr = getelementptr i64, ptr %out, i64 %index
-  %gep = getelementptr i64, ptr %ptr, i64 4
+  %ptr = getelementptr inbounds i64, ptr %out, i64 %index
+  %gep = getelementptr inbounds i64, ptr %ptr, i64 4
   %tmp0 = atomicrmw volatile min ptr %gep, i64 %in syncscope("workgroup") seq_cst, !noalias.addrspace !0
   store i64 %tmp0, ptr %out2
   ret void
@@ -3747,7 +3747,7 @@ define amdgpu_kernel void @atomic_min_i64_addr64(ptr %out, i64 %in, i64 %index)
 ; GFX12-NEXT:    global_inv scope:SCOPE_SE
 ; GFX12-NEXT:    s_endpgm
 entry:
-  %ptr = getelementptr i64, ptr %out, i64 %index
+  %ptr = getelementptr inbounds i64, ptr %out, i64 %index
   %tmp0 = atomicrmw volatile min ptr %ptr, i64 %in syncscope("workgroup") seq_cst, !noalias.addrspace !0
   ret void
 }
@@ -3837,7 +3837,7 @@ define amdgpu_kernel void @atomic_min_i64_ret_addr64(ptr %out, ptr %out2, i64 %i
 ; GFX12-NEXT:    flat_store_b64 v[2:3], v[0:1]
 ; GFX12-NEXT:    s_endpgm
 entry:
-  %ptr = getelementptr i64, ptr %out, i64 %index
+  %ptr = getelementptr inbounds i64, ptr %out, i64 %index
   %tmp0 = atomicrmw volatile min ptr %ptr, i64 %in syncscope("workgroup") seq_cst, !noalias.addrspace !0
   store i64 %tmp0, ptr %out2
   ret void
@@ -3913,7 +3913,7 @@ define amdgpu_kernel void @atomic_umin_i64_offset(ptr %out, i64 %in) {
 ; GFX12-NEXT:    global_inv scope:SCOPE_SE
 ; GFX12-NEXT:    s_endpgm
 entry:
-  %gep = getelementptr i64, ptr %out, i64 4
+  %gep = getelementptr inbounds i64, ptr %out, i64 4
   %tmp0 = atomicrmw volatile umin ptr %gep, i64 %in syncscope("workgroup") seq_cst, !noalias.addrspace !0
   ret void
 }
@@ -4002,7 +4002,7 @@ define amdgpu_kernel void @atomic_umin_i64_ret_offset(ptr %out, ptr %out2, i64 %
 ; GFX12-NEXT:    flat_store_b64 v[2:3], v[0:1]
 ; GFX12-NEXT:    s_endpgm
 entry:
-  %gep = getelementptr i64, ptr %out, i64 4
+  %gep = getelementptr inbounds i64, ptr %out, i64 4
   %tmp0 = atomicrmw volatile umin ptr %gep, i64 %in syncscope("workgroup") seq_cst, !noalias.addrspace !0
   store i64 %tmp0, ptr %out2
   ret void
@@ -4091,8 +4091,8 @@ define amdgpu_kernel void @atomic_umin_i64_addr64_offset(ptr %out, i64 %in, i64
 ; GFX12-NEXT:    global_inv scope:SCOPE_SE
 ; GFX12-NEXT:    s_endpgm
 entry:
-  %ptr = getelementptr i64, ptr %out, i64 %index
-  %gep = getelementptr i64, ptr %ptr, i64 4
+  %ptr = getelementptr inbounds i64, ptr %out, i64 %index
+  %gep = getelementptr inbounds i64, ptr %ptr, i64 4
   %tmp0 = atomicrmw volatile umin ptr %gep, i64 %in syncscope("workgroup") seq_cst, !noalias.addrspace !0
   ret void
 }
@@ -4186,8 +4186,8 @@ define amdgpu_kernel void @atomic_umin_i64_ret_addr64_offset(ptr %out, ptr %out2
 ; GFX12-NEXT:    flat_store_b64 v[2:3], v[0:1]
 ; GFX12-NEXT:    s_endpgm
 entry:
-  %ptr = getelementptr i64, ptr %out, i64 %index
-  %gep = getelementptr i64, ptr %ptr, i64 4
+  %ptr = getelementptr inbounds i64, ptr %out, i64 %index
+  %gep = getelementptr inbounds i64, ptr %ptr, i64 4
   %tmp0 = atomicrmw volatile umin ptr %gep, i64 %in syncscope("workgroup") seq_cst, !noalias.addrspace !0
   store i64 %tmp0, ptr %out2
   ret void
@@ -4435,7 +4435,7 @@ define amdgpu_kernel void @atomic_umin_i64_addr64(ptr %out, i64 %in, i64 %index)
 ; GFX12-NEXT:    global_inv scope:SCOPE_SE
 ; GFX12-NEXT:    s_endpgm
 entry:
-  %ptr = getelementptr i64, ptr %out, i64 %index
+  %ptr = getelementptr inbounds i64, ptr %out, i64 %index
   %tmp0 = atomicrmw volatile umin ptr %ptr, i64 %in syncscope("workgroup") seq_cst, !noalias.addrspace !0
   ret void
 }
@@ -4525,7 +4525,7 @@ define amdgpu_kernel void @atomic_umin_i64_ret_addr64(ptr %out, ptr %out2, i64 %
 ; GFX12-NEXT:    flat_store_b64 v[2:3], v[0:1]
 ; GFX12-NEXT:    s_endpgm
 entry:
-  %ptr = getelementptr i64, ptr %out, i64 %index
+  %ptr = getelementptr inbounds i64, ptr %out, i64 %index
   %tmp0 = atomicrmw volatile umin ptr %ptr, i64 %in syncscope("workgroup") seq_cst, !noalias.addrspace !0
   store i64 %tmp0, ptr %out2
   ret void
@@ -4597,7 +4597,7 @@ define amdgpu_kernel void @atomic_or_i64_offset(ptr %out, i64 %in) {
 ; GFX12-NEXT:    global_inv scope:SCOPE_DEV
 ; GFX12-NEXT:    s_endpgm
 entry:
-  %gep = getelementptr i64, ptr %out, i64 4
+  %gep = getelementptr inbounds i64, ptr %out, i64 4
   %tmp0 = atomicrmw volatile or ptr %gep, i64 %in syncscope("agent") seq_cst, !noalias.addrspace !0
   ret void
 }
@@ -4682,7 +4682,7 @@ define amdgpu_kernel void @atomic_or_i64_ret_offset(ptr %out, ptr %out2, i64 %in
 ; GFX12-NEXT:    flat_store_b64 v[2:3], v[0:1]
 ; GFX12-NEXT:    s_endpgm
 entry:
-  %gep = getelementptr i64, ptr %out, i64 4
+  %gep = getelementptr inbounds i64, ptr %out, i64 4
   %tmp0 = atomicrmw volatile or ptr %gep, i64 %in syncscope("agent") seq_cst, !noalias.addrspace !0
   store i64 %tmp0, ptr %out2
   ret void
@@ -4767,8 +4767,8 @@ define amdgpu_kernel void @atomic_or_i64_addr64_offset(ptr %out, i64 %in, i64 %i
 ; GFX12-NEXT:    global_inv scope:SCOPE_DEV
 ; GFX12-NEXT:    s_endpgm
 entry:
-  %ptr = getelementptr i64, ptr %out, i64 %index
-  %gep = getelementptr i64, ptr %ptr, i64 4
+  %ptr = getelementptr inbounds i64, ptr %out, i64 %index
+  %gep = getelementptr inbounds i64, ptr %ptr, i64 4
   %tmp0 = atomicrmw volatile or ptr %gep, i64 %in syncscope("agent") seq_cst, !noalias.addrspace !0
   ret void
 }
@@ -4858,8 +4858,8 @@ define amdgpu_kernel void @atomic_or_i64_ret_addr64_offset(ptr %out, ptr %out2,
 ; GFX12-NEXT:    flat_store_b64 v[2:3], v[0:1]
 ; GFX12-NEXT:    s_endpgm
 entry:
-  %ptr = getelementptr i64, ptr %out, i64 %index
-  %gep = getelementptr i64, ptr %ptr, i64 4
+  %ptr = getelementptr inbounds i64, ptr %out, i64 %index
+  %gep = getelementptr inbounds i64, ptr %ptr, i64 4
   %tmp0 = atomicrmw volatile or ptr %gep, i64 %in syncscope("agent") seq_cst, !noalias.addrspace !0
   store i64 %tmp0, ptr %out2
   ret void
@@ -5095,7 +5095,7 @@ define amdgpu_kernel void @atomic_or_i64_addr64(ptr %out, i64 %in, i64 %index) {
 ; GFX12-NEXT:    global_inv scope:SCOPE_DEV
 ; GFX12-NEXT:    s_endpgm
 entry:
-  %ptr = getelementptr i64, ptr %out, i64 %index
+  %ptr = getelementptr inbounds i64, ptr %out, i64 %index
   %tmp0 = atomicrmw volatile or ptr %ptr, i64 %in syncscope("agent") seq_cst, !noalias.addrspace !0
   ret void
 }
@@ -5181,7 +5181,7 @@ define amdgpu_kernel void @atomic_or_i64_ret_addr64(ptr %out, ptr %out2, i64 %in
 ; GFX12-NEXT:    flat_store_b64 v[2:3], v[0:1]
 ; GFX12-NEXT:    s_endpgm
 entry:
-  %ptr = getelementptr i64, ptr %out, i64 %index
+  %ptr = getelementptr inbounds i64, ptr %out, i64 %index
   %tmp0 = atomicrmw volatile or ptr %ptr, i64 %in syncscope("agent") seq_cst, !noalias.addrspace !0
   store i64 %tmp0, ptr %out2
   ret void
@@ -5229,7 +5229,7 @@ define amdgpu_kernel void @atomic_xchg_i64_offset(ptr %out, i64 %in) {
 ; GFX12-NEXT:    global_inv scope:SCOPE_DEV
 ; GFX12-NEXT:    s_endpgm
 entry:
-  %gep = getelementptr i64, ptr %out, i64 4
+  %gep = getelementptr inbounds i64, ptr %out, i64 4
   %tmp0 = atomicrmw volatile xchg ptr %gep, i64 %in syncscope("agent") seq_cst, !noalias.addrspace !0
   ret void
 }
@@ -5276,7 +5276,7 @@ define amdgpu_kernel void @atomic_xchg_f64_offset(ptr %out, double %in) {
 ; GFX12-NEXT:    global_inv scope:SCOPE_DEV
 ; GFX12-NEXT:    s_endpgm
 entry:
-  %gep = getelementptr double, ptr %out, i64 4
+  %gep = getelementptr inbounds double, ptr %out, i64 4
   %tmp0 = atomicrmw volatile xchg ptr %gep, double %in syncscope("agent") seq_cst, !noalias.addrspace !0
   ret void
 }
@@ -5323,7 +5323,7 @@ define amdgpu_kernel void @atomic_xchg_pointer_offset(ptr %out, ptr %in) {
 ; GFX12-NEXT:    global_inv scope:SCOPE_DEV
 ; GFX12-NEXT:    s_endpgm
 entry:
-  %gep = getelementptr ptr, ptr %out, i32 4
+  %gep = getelementptr inbounds ptr, ptr %out, i32 4
   %val = atomicrmw volatile xchg ptr %gep, ptr %in syncscope("agent") seq_cst, !noalias.addrspace !0
   ret void
 }
@@ -5382,7 +5382,7 @@ define amdgpu_kernel void @atomic_xchg_i64_ret_offset(ptr %out, ptr %out2, i64 %
 ; GFX12-NEXT:    flat_store_b64 v[2:3], v[0:1]
 ; GFX12-NEXT:    s_endpgm
 entry:
-  %gep = getelementptr i64, ptr %out, i64 4
+  %gep = getelementptr inbounds i64, ptr %out, i64 4
   %tmp0 = atomicrmw volatile xchg ptr %gep, i64 %in syncscope("agent") seq_cst, !noalias.addrspace !0
   store i64 %tmp0, ptr %out2
   ret void
@@ -5443,8 +5443,8 @@ define amdgpu_kernel void @atomic_xchg_i64_addr64_offset(ptr %out, i64 %in, i64
 ; GFX12-NEXT:    global_inv scope:SCOPE_DEV
 ; GFX12-NEXT:    s_endpgm
 entry:
-  %ptr = getelementptr i64, ptr %out, i64 %index
-  %gep = getelementptr i64, ptr %ptr, i64 4
+  %ptr = getelementptr inbounds i64, ptr %out, i64 %index
+  %gep = getelementptr inbounds i64, ptr %ptr, i64 4
   %tmp0 = atomicrmw volatile xchg ptr %gep, i64 %in syncscope("agent") seq_cst, !noalias.addrspace !0
   ret void
 }
@@ -5508,8 +5508,8 @@ define amdgpu_kernel void @atomic_xchg_i64_ret_addr64_offset(ptr %out, ptr %out2
 ; GFX12-NEXT:    flat_store_b64 v[2:3], v[0:1]
 ; GFX12-NEXT:    s_endpgm
 entry:
-  %ptr = getelementptr i64, ptr %out, i64 %index
-  %gep = getelementptr i64, ptr %ptr, i64 4
+  %ptr = getelementptr inbounds i64, ptr %out, i64 %index
+  %gep = getelementptr inbounds i64, ptr %ptr, i64 4
   %tmp0 = atomicrmw volatile xchg ptr %gep, i64 %in syncscope("agent") seq_cst, !noalias.addrspace !0
   store i64 %tmp0, ptr %out2
   ret void
@@ -5663,7 +5663,7 @@ define amdgpu_kernel void @atomic_xchg_i64_addr64(ptr %out, i64 %in, i64 %index)
 ; GFX12-NEXT:    global_inv scope:SCOPE_DEV
 ; GFX12-NEXT:    s_endpgm
 entry:
-  %ptr = getelementptr i64, ptr %out, i64 %index
+  %ptr = getelementptr inbounds i64, ptr %out, i64 %index
   %tmp0 = atomicrmw volatile xchg ptr %ptr, i64 %in syncscope("agent") seq_cst, !noalias.addrspace !0
   ret void
 }
@@ -5723,7 +5723,7 @@ define amdgpu_kernel void @atomic_xchg_i64_ret_addr64(ptr %out, ptr %out2, i64 %
 ; GFX12-NEXT:    flat_store_b64 v[2:3], v[0:1]
 ; GFX12-NEXT:    s_endpgm
 entry:
-  %ptr = getelementptr i64, ptr %out, i64 %index
+  %ptr = getelementptr inbounds i64, ptr %out, i64 %index
   %tmp0 = atomicrmw volatile xchg ptr %ptr, i64 %in syncscope("agent") seq_cst, !noalias.addrspace !0
   store i64 %tmp0, ptr %out2
   ret void
@@ -5795,7 +5795,7 @@ define amdgpu_kernel void @atomic_xor_i64_offset(ptr %out, i64 %in) {
 ; GFX12-NEXT:    global_inv scope:SCOPE_DEV
 ; GFX12-NEXT:    s_endpgm
 entry:
-  %gep = getelementptr i64, ptr %out, i64 4
+  %gep = getelementptr inbounds i64, ptr %out, i64 4
   %tmp0 = atomicrmw volatile xor ptr %gep, i64 %in syncscope("agent") seq_cst, !noalias.addrspace !0
   ret void
 }
@@ -5880,7 +5880,7 @@ define amdgpu_kernel void @atomic_xor_i64_ret_offset(ptr %out, ptr %out2, i64 %i
 ; GFX12-NEXT:    flat_store_b64 v[2:3], v[0:1]
 ; GFX12-NEXT:    s_endpgm
 entry:
-  %gep = getelementptr i64, ptr %out, i64 4
+  %gep = getelementptr inbounds i64, ptr %out, i64 4
   %tmp0 = atomicrmw volatile xor ptr %gep, i64 %in syncscope("agent") seq_cst, !noalias.addrspace !0
   store i64 %tmp0, ptr %out2
   ret void
@@ -5965,8 +5965,8 @@ define amdgpu_kernel void @atomic_xor_i64_addr64_offset(ptr %out, i64 %in, i64 %
 ; GFX12-NEXT:    global_inv scope:SCOPE_DEV
 ; GFX12-NEXT:    s_endpgm
 entry:
-  %ptr = getelementptr i64, ptr %out, i64 %index
-  %gep = getelementptr i64, ptr %ptr, i64 4
+  %ptr = getelementptr inbounds i64, ptr %out, i64 %index
+  %gep = getelementptr inbounds i64, ptr %ptr, i64 4
   %tmp0 = atomicrmw volatile xor ptr %gep, i64 %in syncscope("agent") seq_cst, !noalias.addrspace !0
   ret void
 }
@@ -6056,8 +6056,8 @@ define amdgpu_kernel void @atomic_xor_i64_ret_addr64_offset(ptr %out, ptr %out2,
 ; GFX12-NEXT:    flat_store_b64 v[2:3], v[0:1]
 ; GFX12-NEXT:    s_endpgm
 entry:
-  %ptr = getelementptr i64, ptr %out, i64 %index
-  %gep = getelementptr i64, ptr %ptr, i64 4
+  %ptr = getelementptr inbounds i64, ptr %out, i64 %index
+  %gep = getelementptr inbounds i64, ptr %ptr, i64 4
   %tmp0 = atomicrmw volatile xor ptr %gep, i64 %in syncscope("agent") seq_cst, !noalias.addrspace !0
   store i64 %tmp0, ptr %out2
   ret void
@@ -6293,7 +6293,7 @@ define amdgpu_kernel void @atomic_xor_i64_addr64(ptr %out, i64 %in, i64 %index)
 ; GFX12-NEXT:    global_inv scope:SCOPE_DEV
 ; GFX12-NEXT:    s_endpgm
 entry:
-  %ptr = getelementptr i64, ptr %out, i64 %index
+  %ptr = getelementptr inbounds i64, ptr %out, i64 %index
   %tmp0 = atomicrmw volatile xor ptr %ptr, i64 %in syncscope("agent") seq_cst, !noalias.addrspace !0
   ret void
 }
@@ -6379,7 +6379,7 @@ define amdgpu_kernel void @atomic_xor_i64_ret_addr64(ptr %out, ptr %out2, i64 %i
 ; GFX12-NEXT:    flat_store_b64 v[2:3], v[0:1]
 ; GFX12-NEXT:    s_endpgm
 entry:
-  %ptr = getelementptr i64, ptr %out, i64 %index
+  %ptr = getelementptr inbounds i64, ptr %out, i64 %index
   %tmp0 = atomicrmw volatile xor ptr %ptr, i64 %in syncscope("agent") seq_cst, !noalias.addrspace !0
   store i64 %tmp0, ptr %out2
   ret void
@@ -6430,7 +6430,7 @@ define amdgpu_kernel void @atomic_load_i64_offset(ptr %in, ptr %out) {
 ; GFX12-NEXT:    flat_store_b64 v[2:3], v[0:1]
 ; GFX12-NEXT:    s_endpgm
 entry:
-  %gep = getelementptr i64, ptr %in, i64 4
+  %gep = getelementptr inbounds i64, ptr %in, i64 4
   %val = load atomic i64, ptr %gep  seq_cst, align 8
   store i64 %val, ptr %out
   ret void
@@ -6540,8 +6540,8 @@ define amdgpu_kernel void @atomic_load_i64_addr64_offset(ptr %in, ptr %out, i64
 ; GFX12-NEXT:    flat_store_b64 v[2:3], v[0:1]
 ; GFX12-NEXT:    s_endpgm
 entry:
-  %ptr = getelementptr i64, ptr %in, i64 %index
-  %gep = getelementptr i64, ptr %ptr, i64 4
+  %ptr = getelementptr inbounds i64, ptr %in, i64 %index
+  %gep = getelementptr inbounds i64, ptr %ptr, i64 4
   %val = load atomic i64, ptr %gep seq_cst, align 8
   store i64 %val, ptr %out
   ret void
@@ -6601,7 +6601,7 @@ define amdgpu_kernel void @atomic_load_i64_addr64(ptr %in, ptr %out, i64 %index)
 ; GFX12-NEXT:    flat_store_b64 v[2:3], v[0:1]
 ; GFX12-NEXT:    s_endpgm
 entry:
-  %ptr = getelementptr i64, ptr %in, i64 %index
+  %ptr = getelementptr inbounds i64, ptr %in, i64 %index
   %val = load atomic i64, ptr %ptr seq_cst, align 8
   store i64 %val, ptr %out
   ret void
@@ -6645,7 +6645,7 @@ define amdgpu_kernel void @atomic_store_i64_offset(i64 %in, ptr %out) {
 ; GFX12-NEXT:    flat_store_b64 v[2:3], v[0:1] offset:32 scope:SCOPE_SYS
 ; GFX12-NEXT:    s_endpgm
 entry:
-  %gep = getelementptr i64, ptr %out, i64 4
+  %gep = getelementptr inbounds i64, ptr %out, i64 4
   store atomic i64 %in, ptr %gep  seq_cst, align 8
   ret void
 }
@@ -6739,8 +6739,8 @@ define amdgpu_kernel void @atomic_store_i64_addr64_offset(i64 %in, ptr %out, i64
 ; GFX12-NEXT:    flat_store_b64 v[2:3], v[0:1] offset:32 scope:SCOPE_SYS
 ; GFX12-NEXT:    s_endpgm
 entry:
-  %ptr = getelementptr i64, ptr %out, i64 %index
-  %gep = getelementptr i64, ptr %ptr, i64 4
+  %ptr = getelementptr inbounds i64, ptr %out, i64 %index
+  %gep = getelementptr inbounds i64, ptr %ptr, i64 4
   store atomic i64 %in, ptr %gep seq_cst, align 8
   ret void
 }
@@ -6792,7 +6792,7 @@ define amdgpu_kernel void @atomic_store_i64_addr64(i64 %in, ptr %out, i64 %index
 ; GFX12-NEXT:    flat_store_b64 v[2:3], v[0:1] scope:SCOPE_SYS
 ; GFX12-NEXT:    s_endpgm
 entry:
-  %ptr = getelementptr i64, ptr %out, i64 %index
+  %ptr = getelementptr inbounds i64, ptr %out, i64 %index
   store atomic i64 %in, ptr %ptr seq_cst, align 8
   ret void
 }
@@ -6848,7 +6848,7 @@ define amdgpu_kernel void @atomic_cmpxchg_i64_offset(ptr %out, i64 %in, i64 %old
 ; GFX12-NEXT:    global_inv scope:SCOPE_DEV
 ; GFX12-NEXT:    s_endpgm
 entry:
-  %gep = getelementptr i64, ptr %out, i64 4
+  %gep = getelementptr inbounds i64, ptr %out, i64 4
   %val = cmpxchg volatile ptr %gep, i64 %old, i64 %in syncscope("agent") seq_cst seq_cst, !noalias.addrspace !0
   ret void
 }
@@ -6904,7 +6904,7 @@ define amdgpu_kernel void @atomic_cmpxchg_i64_soffset(ptr %out, i64 %in, i64 %ol
 ; GFX12-NEXT:    global_inv scope:SCOPE_DEV
 ; GFX12-NEXT:    s_endpgm
 entry:
-  %gep = getelementptr i64, ptr %out, i64 9000
+  %gep = getelementptr inbounds i64, ptr %out, i64 9000
   %val = cmpxchg volatile ptr %gep, i64 %old, i64 %in syncscope("agent") seq_cst seq_cst, !noalias.addrspace !0
   ret void
 }
@@ -6964,7 +6964,7 @@ define amdgpu_kernel void @atomic_cmpxchg_i64_ret_offset(ptr %out, ptr %out2, i6
 ; GFX12-NEXT:    flat_store_b64 v[2:3], v[0:1]
 ; GFX12-NEXT:    s_endpgm
 entry:
-  %gep = getelementptr i64, ptr %out, i64 4
+  %gep = getelementptr inbounds i64, ptr %out, i64 4
   %val = cmpxchg volatile ptr %gep, i64 %old, i64 %in syncscope("agent") seq_cst seq_cst, !noalias.addrspace !0
   %extract0 = extractvalue { i64, i1 } %val, 0
   store i64 %extract0, ptr %out2
@@ -7026,8 +7026,8 @@ define amdgpu_kernel void @atomic_cmpxchg_i64_addr64_offset(ptr %out, i64 %in, i
 ; GFX12-NEXT:    global_inv scope:SCOPE_DEV
 ; GFX12-NEXT:    s_endpgm
 entry:
-  %ptr = getelementptr i64, ptr %out, i64 %index
-  %gep = getelementptr i64, ptr %ptr, i64 4
+  %ptr = getelementptr inbounds i64, ptr %out, i64 %index
+  %gep = getelementptr inbounds i64, ptr %ptr, i64 4
   %val = cmpxchg volatile ptr %gep, i64 %old, i64 %in syncscope("agent") seq_cst seq_cst, !noalias.addrspace !0
   ret void
 }
@@ -7099,8 +7099,8 @@ define amdgpu_kernel void @atomic_cmpxchg_i64_ret_addr64_offset(ptr %out, ptr %o
 ; GFX12-NEXT:    flat_store_b64 v[2:3], v[0:1]
 ; GFX12-NEXT:    s_endpgm
 entry:
-  %ptr = getelementptr i64, ptr %out, i64 %index
-  %gep = getelementptr i64, ptr %ptr, i64 4
+  %ptr = getelementptr inbounds i64, ptr %out, i64 %index
+  %gep = getelementptr inbounds i64, ptr %ptr, i64 4
   %val = cmpxchg volatile ptr %gep, i64 %old, i64 %in syncscope("agent") seq_cst seq_cst, !noalias.addrspace !0
   %extract0 = extractvalue { i64, i1 } %val, 0
   store i64 %extract0, ptr %out2
@@ -7266,7 +7266,7 @@ define amdgpu_kernel void @atomic_cmpxchg_i64_addr64(ptr %out, i64 %in, i64 %ind
 ; GFX12-NEXT:    global_inv scope:SCOPE_DEV
 ; GFX12-NEXT:    s_endpgm
 entry:
-  %ptr = getelementptr i64, ptr %out, i64 %index
+  %ptr = getelementptr inbounds i64, ptr %out, i64 %index
   %val = cmpxchg volatile ptr %ptr, i64 %old, i64 %in syncscope("agent") seq_cst seq_cst, !noalias.addrspace !0
   ret void
 }
@@ -7334,7 +7334,7 @@ define amdgpu_kernel void @atomic_cmpxchg_i64_ret_addr64(ptr %out, ptr %out2, i6
 ; GFX12-NEXT:    flat_store_b64 v[2:3], v[0:1]
 ; GFX12-NEXT:    s_endpgm
 entry:
-  %ptr = getelementptr i64, ptr %out, i64 %index
+  %ptr = getelementptr inbounds i64, ptr %out, i64 %index
   %val = cmpxchg volatile ptr %ptr, i64 %old, i64 %in syncscope("agent") seq_cst seq_cst, !noalias.addrspace !0
   %extract0 = extractvalue { i64, i1 } %val, 0
   store i64 %extract0, ptr %out2
@@ -7386,7 +7386,7 @@ define amdgpu_kernel void @atomic_load_f64_offset(ptr %in, ptr %out) {
 ; GFX12-NEXT:    flat_store_b64 v[2:3], v[0:1]
 ; GFX12-NEXT:    s_endpgm
 entry:
-  %gep = getelementptr double, ptr %in, i64 4
+  %gep = getelementptr inbounds double, ptr %in, i64 4
   %val = load atomic double, ptr %gep  seq_cst, align 8, !noalias.addrspace !0
   store double %val, ptr %out
   ret void
@@ -7496,8 +7496,8 @@ define amdgpu_kernel void @atomic_load_f64_addr64_offset(ptr %in, ptr %out, i64
 ; GFX12-NEXT:    flat_store_b64 v[2:3], v[0:1]
 ; GFX12-NEXT:    s_endpgm
 entry:
-  %ptr = getelementptr double, ptr %in, i64 %index
-  %gep = getelementptr double, ptr %ptr, i64 4
+  %ptr = getelementptr inbounds double, ptr %in, i64 %index
+  %gep = getelementptr inbounds double, ptr %ptr, i64 4
   %val = load atomic double, ptr %gep seq_cst, align 8, !noalias.addrspace !0
   store double %val, ptr %out
   ret void
@@ -7557,7 +7557,7 @@ define amdgpu_kernel void @atomic_load_f64_addr64(ptr %in, ptr %out, i64 %index)
 ; GFX12-NEXT:    flat_store_b64 v[2:3], v[0:1]
 ; GFX12-NEXT:    s_endpgm
 entry:
-  %ptr = getelementptr double, ptr %in, i64 %index
+  %ptr = getelementptr inbounds double, ptr %in, i64 %index
   %val = load atomic double, ptr %ptr seq_cst, align 8, !noalias.addrspace !0
   store double %val, ptr %out
   ret void
@@ -7601,7 +7601,7 @@ define amdgpu_kernel void @atomic_store_f64_offset(double %in, ptr %out) {
 ; GFX12-NEXT:    flat_store_b64 v[2:3], v[0:1] offset:32 scope:SCOPE_SYS
 ; GFX12-NEXT:    s_endpgm
 entry:
-  %gep = getelementptr double, ptr %out, i64 4
+  %gep = getelementptr inbounds double, ptr %out, i64 4
   store atomic double %in, ptr %gep  seq_cst, align 8, !noalias.addrspace !0
   ret void
 }
@@ -7695,8 +7695,8 @@ define amdgpu_kernel void @atomic_store_f64_addr64_offset(double %in, ptr %out,
 ; GFX12-NEXT:    flat_store_b64 v[2:3], v[0:1] offset:32 scope:SCOPE_SYS
 ; GFX12-NEXT:    s_endpgm
 entry:
-  %ptr = getelementptr double, ptr %out, i64 %index
-  %gep = getelementptr double, ptr %ptr, i64 4
+  %ptr = getelementptr inbounds double, ptr %out, i64 %index
+  %gep = getelementptr inbounds double, ptr %ptr, i64 4
   store atomic double %in, ptr %gep seq_cst, align 8, !noalias.addrspace !0
   ret void
 }
@@ -7748,7 +7748,7 @@ define amdgpu_kernel void @atomic_store_f64_addr64(double %in, ptr %out, i64 %in
 ; GFX12-NEXT:    flat_store_b64 v[2:3], v[0:1] scope:SCOPE_SYS
 ; GFX12-NEXT:    s_endpgm
 entry:
-  %ptr = getelementptr double, ptr %out, i64 %index
+  %ptr = getelementptr inbounds double, ptr %out, i64 %index
   store atomic double %in, ptr %ptr seq_cst, align 8, !noalias.addrspace !0
   ret void
 }
@@ -7825,7 +7825,7 @@ define amdgpu_kernel void @atomic_inc_i64_offset(ptr %out, i64 %in) {
 ; GFX12-NEXT:    global_inv scope:SCOPE_DEV
 ; GFX12-NEXT:    s_endpgm
 entry:
-  %gep = getelementptr i64, ptr %out, i64 4
+  %gep = getelementptr inbounds i64, ptr %out, i64 4
   %tmp0 = atomicrmw volatile uinc_wrap ptr %gep, i64 %in syncscope("agent") seq_cst, !noalias.addrspace !0
   ret void
 }
@@ -7916,7 +7916,7 @@ define amdgpu_kernel void @atomic_inc_i64_ret_offset(ptr %out, ptr %out2, i64 %i
 ; GFX12-NEXT:    flat_store_b64 v[2:3], v[0:1]
 ; GFX12-NEXT:    s_endpgm
 entry:
-  %gep = getelementptr i64, ptr %out, i64 4
+  %gep = getelementptr inbounds i64, ptr %out, i64 4
   %tmp0 = atomicrmw volatile uinc_wrap ptr %gep, i64 %in syncscope("agent") seq_cst, !noalias.addrspace !0
   store i64 %tmp0, ptr %out2
   ret void
@@ -8007,8 +8007,8 @@ define amdgpu_kernel void @atomic_inc_i64_incr64_offset(ptr %out, i64 %in, i64 %
 ; GFX12-NEXT:    global_inv scope:SCOPE_DEV
 ; GFX12-NEXT:    s_endpgm
 entry:
-  %ptr = getelementptr i64, ptr %out, i64 %index
-  %gep = getelementptr i64, ptr %ptr, i64 4
+  %ptr = getelementptr inbounds i64, ptr %out, i64 %index
+  %gep = getelementptr inbounds i64, ptr %ptr, i64 4
   %tmp0 = atomicrmw volatile uinc_wrap ptr %gep, i64 %in syncscope("agent") seq_cst, !noalias.addrspace !0
   ret void
 }
@@ -8104,8 +8104,8 @@ define amdgpu_kernel void @atomic_inc_i64_ret_incr64_offset(ptr %out, ptr %out2,
 ; GFX12-NEXT:    flat_store_b64 v[2:3], v[0:1]
 ; GFX12-NEXT:    s_endpgm
 entry:
-  %ptr = getelementptr i64, ptr %out, i64 %index
-  %gep = getelementptr i64, ptr %ptr, i64 4
+  %ptr = getelementptr inbounds i64, ptr %out, i64 %index
+  %gep = getelementptr inbounds i64, ptr %ptr, i64 4
   %tmp0 = atomicrmw volatile uinc_wrap ptr %gep, i64 %in syncscope("agent") seq_cst, !noalias.addrspace !0
   store i64 %tmp0, ptr %out2
   ret void
@@ -8355,7 +8355,7 @@ define amdgpu_kernel void @atomic_inc_i64_incr64(ptr %out, i64 %in, i64 %index)
 ; GFX12-NEXT:    global_inv scope:SCOPE_DEV
 ; GFX12-NEXT:    s_endpgm
 entry:
-  %ptr = getelementptr i64, ptr %out, i64 %index
+  %ptr = getelementptr inbounds i64, ptr %out, i64 %index
   %tmp0 = atomicrmw volatile uinc_wrap ptr %ptr, i64 %in syncscope("agent") seq_cst, !noalias.addrspace !0
   ret void
 }
@@ -8447,7 +8447,7 @@ define amdgpu_kernel void @atomic_inc_i64_ret_incr64(ptr %out, ptr %out2, i64 %i
 ; GFX12-NEXT:    flat_store_b64 v[2:3], v[0:1]
 ; GFX12-NEXT:    s_endpgm
 entry:
-  %ptr = getelementptr i64, ptr %out, i64 %index
+  %ptr = getelementptr inbounds i64, ptr %out, i64 %index
   %tmp0 = atomicrmw volatile uinc_wrap ptr %ptr, i64 %in syncscope("agent") seq_cst, !noalias.addrspace !0
   store i64 %tmp0, ptr %out2
   ret void
@@ -8533,7 +8533,7 @@ define amdgpu_kernel void @atomic_dec_i64_offset(ptr %out, i64 %in) {
 ; GFX12-NEXT:    global_inv scope:SCOPE_DEV
 ; GFX12-NEXT:    s_endpgm
 entry:
-  %gep = getelementptr i64, ptr %out, i64 4
+  %gep = getelementptr inbounds i64, ptr %out, i64 4
   %tmp0 = atomicrmw volatile udec_wrap ptr %gep, i64 %in syncscope("agent") seq_cst, !noalias.addrspace !0
   ret void
 }
@@ -8632,7 +8632,7 @@ define amdgpu_kernel void @atomic_dec_i64_ret_offset(ptr %out, ptr %out2, i64 %i
 ; GFX12-NEXT:    flat_store_b64 v[2:3], v[0:1]
 ; GFX12-NEXT:    s_endpgm
 entry:
-  %gep = getelementptr i64, ptr %out, i64 4
+  %gep = getelementptr inbounds i64, ptr %out, i64 4
   %tmp0 = atomicrmw volatile udec_wrap ptr %gep, i64 %in syncscope("agent") seq_cst, !noalias.addrspace !0
   store i64 %tmp0, ptr %out2
   ret void
@@ -8731,8 +8731,8 @@ define amdgpu_kernel void @atomic_dec_i64_decr64_offset(ptr %out, i64 %in, i64 %
 ; GFX12-NEXT:    global_inv scope:SCOPE_DEV
 ; GFX12-NEXT:    s_endpgm
 entry:
-  %ptr = getelementptr i64, ptr %out, i64 %index
-  %gep = getelementptr i64, ptr %ptr, i64 4
+  %ptr = getelementptr inbounds i64, ptr %out, i64 %index
+  %gep = getelementptr inbounds i64, ptr %ptr, i64 4
   %tmp0 = atomicrmw volatile udec_wrap ptr %gep, i64 %in syncscope("agent") seq_cst, !noalias.addrspace !0
   ret void
 }
@@ -8836,8 +8836,8 @@ define amdgpu_kernel void @atomic_dec_i64_ret_decr64_offset(ptr %out, ptr %out2,
 ; GFX12-NEXT:    flat_store_b64 v[2:3], v[0:1]
 ; GFX12-NEXT:    s_endpgm
 entry:
-  %ptr = getelementptr i64, ptr %out, i64 %index
-  %gep = getelementptr i64, ptr %ptr, i64 4
+  %ptr = getelementptr inbounds i64, ptr %out, i64 %index
+  %gep = getelementptr inbounds i64, ptr %ptr, i64 4
   %tmp0 = atomicrmw volatile udec_wrap ptr %gep, i64 %in syncscope("agent") seq_cst, !noalias.addrspace !0
   store i64 %tmp0, ptr %out2
   ret void
@@ -9111,7 +9111,7 @@ define amdgpu_kernel void @atomic_dec_i64_decr64(ptr %out, i64 %in, i64 %index)
 ; GFX12-NEXT:    global_inv scope:SCOPE_DEV
 ; GFX12-NEXT:    s_endpgm
 entry:
-  %ptr = getelementptr i64, ptr %out, i64 %index
+  %ptr = getelementptr inbounds i64, ptr %out, i64 %index
   %tmp0 = atomicrmw volatile udec_wrap ptr %ptr, i64 %in syncscope("agent") seq_cst, !noalias.addrspace !0
   ret void
 }
@@ -9211,7 +9211,7 @@ define amdgpu_kernel void @atomic_dec_i64_ret_decr64(ptr %out, ptr %out2, i64 %i
 ; GFX12-NEXT:    flat_store_b64 v[2:3], v[0:1]
 ; GFX12-NEXT:    s_endpgm
 entry:
-  %ptr = getelementptr i64, ptr %out, i64 %index
+  %ptr = getelementptr inbounds i64, ptr %out, i64 %index
   %tmp0 = atomicrmw volatile udec_wrap ptr %ptr, i64 %in syncscope("agent") seq_cst, !noalias.addrspace !0
   store i64 %tmp0, ptr %out2
   ret void
diff --git a/llvm/test/CodeGen/AMDGPU/flat_atomics_i64_system_noprivate.ll b/llvm/test/CodeGen/AMDGPU/flat_atomics_i64_system_noprivate.ll
index 4dea4495b36fb..f655d4761fa31 100644
--- a/llvm/test/CodeGen/AMDGPU/flat_atomics_i64_system_noprivate.ll
+++ b/llvm/test/CodeGen/AMDGPU/flat_atomics_i64_system_noprivate.ll
@@ -63,7 +63,7 @@ define void @flat_atomic_xchg_i64_noret_offset(ptr %out, i64 %in) {
 ; GFX9-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX9-NEXT:    buffer_wbinvl1_vol
 ; GFX9-NEXT:    s_setpc_b64 s[30:31]
-  %gep = getelementptr i64, ptr %out, i64 4
+  %gep = getelementptr inbounds i64, ptr %out, i64 4
   %tmp0 = atomicrmw xchg ptr %gep, i64 %in seq_cst, !noalias.addrspace !1
   ret void
 }
@@ -124,7 +124,7 @@ define i64 @flat_atomic_xchg_i64_ret_offset(ptr %out, i64 %in) {
 ; GFX9-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX9-NEXT:    buffer_wbinvl1_vol
 ; GFX9-NEXT:    s_setpc_b64 s[30:31]
-  %gep = getelementptr i64, ptr %out, i64 4
+  %gep = getelementptr inbounds i64, ptr %out, i64 4
   %result = atomicrmw xchg ptr %gep, i64 %in seq_cst, !noalias.addrspace !1
   ret i64 %result
 }
@@ -209,7 +209,7 @@ define amdgpu_gfx void @flat_atomic_xchg_i64_noret_offset_scalar(ptr inreg %out,
 ; GFX9-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX9-NEXT:    buffer_wbinvl1_vol
 ; GFX9-NEXT:    s_setpc_b64 s[30:31]
-  %gep = getelementptr i64, ptr %out, i64 4
+  %gep = getelementptr inbounds i64, ptr %out, i64 4
   %tmp0 = atomicrmw xchg ptr %gep, i64 %in seq_cst, !noalias.addrspace !1
   ret void
 }
@@ -294,7 +294,7 @@ define amdgpu_gfx i64 @flat_atomic_xchg_i64_ret_offset_scalar(ptr inreg %out, i6
 ; GFX9-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX9-NEXT:    buffer_wbinvl1_vol
 ; GFX9-NEXT:    s_setpc_b64 s[30:31]
-  %gep = getelementptr i64, ptr %out, i64 4
+  %gep = getelementptr inbounds i64, ptr %out, i64 4
   %result = atomicrmw xchg ptr %gep, i64 %in seq_cst, !noalias.addrspace !1
   ret i64 %result
 }
@@ -327,7 +327,7 @@ define void @flat_atomic_xchg_i64_noret_offset__amdgpu_no_remote_memory(ptr %out
 ; GFX9-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX9-NEXT:    buffer_wbinvl1_vol
 ; GFX9-NEXT:    s_setpc_b64 s[30:31]
-  %gep = getelementptr i64, ptr %out, i64 4
+  %gep = getelementptr inbounds i64, ptr %out, i64 4
   %tmp0 = atomicrmw xchg ptr %gep, i64 %in seq_cst, !amdgpu.no.remote.memory !0, !noalias.addrspace !1
   ret void
 }
@@ -360,7 +360,7 @@ define i64 @flat_atomic_xchg_i64_ret_offset__amdgpu_no_remote_memory(ptr %out, i
 ; GFX9-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX9-NEXT:    buffer_wbinvl1_vol
 ; GFX9-NEXT:    s_setpc_b64 s[30:31]
-  %gep = getelementptr i64, ptr %out, i64 4
+  %gep = getelementptr inbounds i64, ptr %out, i64 4
   %result = atomicrmw xchg ptr %gep, i64 %in seq_cst, !amdgpu.no.remote.memory !0, !noalias.addrspace !1
   ret i64 %result
 }
@@ -425,7 +425,7 @@ define void @flat_atomic_xchg_f64_noret_offset(ptr %out, double %in) {
 ; GFX9-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX9-NEXT:    buffer_wbinvl1_vol
 ; GFX9-NEXT:    s_setpc_b64 s[30:31]
-  %gep = getelementptr double, ptr %out, i32 4
+  %gep = getelementptr inbounds double, ptr %out, i32 4
   %tmp0 = atomicrmw xchg ptr %gep, double %in seq_cst, !noalias.addrspace !1
   ret void
 }
@@ -486,7 +486,7 @@ define double @flat_atomic_xchg_f64_ret_offset(ptr %out, double %in) {
 ; GFX9-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX9-NEXT:    buffer_wbinvl1_vol
 ; GFX9-NEXT:    s_setpc_b64 s[30:31]
-  %gep = getelementptr double, ptr %out, i32 4
+  %gep = getelementptr inbounds double, ptr %out, i32 4
   %result = atomicrmw xchg ptr %gep, double %in seq_cst, !noalias.addrspace !1
   ret double %result
 }
@@ -571,7 +571,7 @@ define amdgpu_gfx void @flat_atomic_xchg_f64_noret_offset_scalar(ptr inreg %out,
 ; GFX9-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX9-NEXT:    buffer_wbinvl1_vol
 ; GFX9-NEXT:    s_setpc_b64 s[30:31]
-  %gep = getelementptr double, ptr %out, i32 4
+  %gep = getelementptr inbounds double, ptr %out, i32 4
   %tmp0 = atomicrmw xchg ptr %gep, double %in seq_cst, !noalias.addrspace !1
   ret void
 }
@@ -656,7 +656,7 @@ define amdgpu_gfx double @flat_atomic_xchg_f64_ret_offset_scalar(ptr inreg %out,
 ; GFX9-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX9-NEXT:    buffer_wbinvl1_vol
 ; GFX9-NEXT:    s_setpc_b64 s[30:31]
-  %gep = getelementptr double, ptr %out, i32 4
+  %gep = getelementptr inbounds double, ptr %out, i32 4
   %result = atomicrmw xchg ptr %gep, double %in seq_cst, !noalias.addrspace !1
   ret double %result
 }
@@ -689,7 +689,7 @@ define void @flat_atomic_xchg_f64_noret_offset__amdgpu_no_remote_memory(ptr %out
 ; GFX9-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX9-NEXT:    buffer_wbinvl1_vol
 ; GFX9-NEXT:    s_setpc_b64 s[30:31]
-  %gep = getelementptr double, ptr %out, i64 4
+  %gep = getelementptr inbounds double, ptr %out, i64 4
   %tmp0 = atomicrmw xchg ptr %gep, double %in seq_cst, !amdgpu.no.remote.memory !0, !noalias.addrspace !1
   ret void
 }
@@ -722,7 +722,7 @@ define double @flat_atomic_xchg_f64_ret_offset__amdgpu_no_remote_memory(ptr %out
 ; GFX9-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX9-NEXT:    buffer_wbinvl1_vol
 ; GFX9-NEXT:    s_setpc_b64 s[30:31]
-  %gep = getelementptr double, ptr %out, i64 4
+  %gep = getelementptr inbounds double, ptr %out, i64 4
   %result = atomicrmw xchg ptr %gep, double %in seq_cst, !amdgpu.no.remote.memory !0, !noalias.addrspace !1
   ret double %result
 }
@@ -787,7 +787,7 @@ define void @flat_atomic_add_i64_noret_offset(ptr %out, i64 %in) {
 ; GFX9-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX9-NEXT:    buffer_wbinvl1_vol
 ; GFX9-NEXT:    s_setpc_b64 s[30:31]
-  %gep = getelementptr i64, ptr %out, i64 4
+  %gep = getelementptr inbounds i64, ptr %out, i64 4
   %tmp0 = atomicrmw add ptr %gep, i64 %in seq_cst, !noalias.addrspace !1
   ret void
 }
@@ -848,7 +848,7 @@ define i64 @flat_atomic_add_i64_ret_offset(ptr %out, i64 %in) {
 ; GFX9-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX9-NEXT:    buffer_wbinvl1_vol
 ; GFX9-NEXT:    s_setpc_b64 s[30:31]
-  %gep = getelementptr i64, ptr %out, i64 4
+  %gep = getelementptr inbounds i64, ptr %out, i64 4
   %result = atomicrmw add ptr %gep, i64 %in seq_cst, !noalias.addrspace !1
   ret i64 %result
 }
@@ -933,7 +933,7 @@ define amdgpu_gfx void @flat_atomic_add_i64_noret_offset_scalar(ptr inreg %out,
 ; GFX9-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX9-NEXT:    buffer_wbinvl1_vol
 ; GFX9-NEXT:    s_setpc_b64 s[30:31]
-  %gep = getelementptr i64, ptr %out, i64 4
+  %gep = getelementptr inbounds i64, ptr %out, i64 4
   %tmp0 = atomicrmw add ptr %gep, i64 %in seq_cst, !noalias.addrspace !1
   ret void
 }
@@ -1018,7 +1018,7 @@ define amdgpu_gfx i64 @flat_atomic_add_i64_ret_offset_scalar(ptr inreg %out, i64
 ; GFX9-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX9-NEXT:    buffer_wbinvl1_vol
 ; GFX9-NEXT:    s_setpc_b64 s[30:31]
-  %gep = getelementptr i64, ptr %out, i64 4
+  %gep = getelementptr inbounds i64, ptr %out, i64 4
   %result = atomicrmw add ptr %gep, i64 %in seq_cst, !noalias.addrspace !1
   ret i64 %result
 }
@@ -1051,7 +1051,7 @@ define void @flat_atomic_add_i64_noret_offset__amdgpu_no_remote_memory(ptr %out,
 ; GFX9-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX9-NEXT:    buffer_wbinvl1_vol
 ; GFX9-NEXT:    s_setpc_b64 s[30:31]
-  %gep = getelementptr i64, ptr %out, i64 4
+  %gep = getelementptr inbounds i64, ptr %out, i64 4
   %tmp0 = atomicrmw add ptr %gep, i64 %in seq_cst, !amdgpu.no.remote.memory !0, !noalias.addrspace !1
   ret void
 }
@@ -1084,7 +1084,7 @@ define i64 @flat_atomic_add_i64_ret_offset__amdgpu_no_remote_memory(ptr %out, i6
 ; GFX9-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX9-NEXT:    buffer_wbinvl1_vol
 ; GFX9-NEXT:    s_setpc_b64 s[30:31]
-  %gep = getelementptr i64, ptr %out, i64 4
+  %gep = getelementptr inbounds i64, ptr %out, i64 4
   %result = atomicrmw add ptr %gep, i64 %in seq_cst, !amdgpu.no.remote.memory !0, !noalias.addrspace !1
   ret i64 %result
 }
@@ -1251,7 +1251,7 @@ define void @flat_atomic_sub_i64_noret_offset(ptr %out, i64 %in) {
 ; GFX9-NEXT:  ; %bb.2: ; %atomicrmw.end
 ; GFX9-NEXT:    s_or_b64 exec, exec, s[4:5]
 ; GFX9-NEXT:    s_setpc_b64 s[30:31]
-  %gep = getelementptr i64, ptr %out, i64 4
+  %gep = getelementptr inbounds i64, ptr %out, i64 4
   %tmp0 = atomicrmw sub ptr %gep, i64 %in seq_cst, !noalias.addrspace !1
   ret void
 }
@@ -1422,7 +1422,7 @@ define i64 @flat_atomic_sub_i64_ret_offset(ptr %out, i64 %in) {
 ; GFX9-NEXT:    v_mov_b32_e32 v0, v4
 ; GFX9-NEXT:    v_mov_b32_e32 v1, v5
 ; GFX9-NEXT:    s_setpc_b64 s[30:31]
-  %gep = getelementptr i64, ptr %out, i64 4
+  %gep = getelementptr inbounds i64, ptr %out, i64 4
   %result = atomicrmw sub ptr %gep, i64 %in seq_cst, !noalias.addrspace !1
   ret i64 %result
 }
@@ -1619,7 +1619,7 @@ define amdgpu_gfx void @flat_atomic_sub_i64_noret_offset_scalar(ptr inreg %out,
 ; GFX9-NEXT:  ; %bb.2: ; %atomicrmw.end
 ; GFX9-NEXT:    s_or_b64 exec, exec, s[34:35]
 ; GFX9-NEXT:    s_setpc_b64 s[30:31]
-  %gep = getelementptr i64, ptr %out, i64 4
+  %gep = getelementptr inbounds i64, ptr %out, i64 4
   %tmp0 = atomicrmw sub ptr %gep, i64 %in seq_cst, !noalias.addrspace !1
   ret void
 }
@@ -1816,7 +1816,7 @@ define amdgpu_gfx i64 @flat_atomic_sub_i64_ret_offset_scalar(ptr inreg %out, i64
 ; GFX9-NEXT:  ; %bb.2: ; %atomicrmw.end
 ; GFX9-NEXT:    s_or_b64 exec, exec, s[34:35]
 ; GFX9-NEXT:    s_setpc_b64 s[30:31]
-  %gep = getelementptr i64, ptr %out, i64 4
+  %gep = getelementptr inbounds i64, ptr %out, i64 4
   %result = atomicrmw sub ptr %gep, i64 %in seq_cst, !noalias.addrspace !1
   ret i64 %result
 }
@@ -1849,7 +1849,7 @@ define void @flat_atomic_sub_i64_noret_offset__amdgpu_no_remote_memory(ptr %out,
 ; GFX9-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX9-NEXT:    buffer_wbinvl1_vol
 ; GFX9-NEXT:    s_setpc_b64 s[30:31]
-  %gep = getelementptr i64, ptr %out, i64 4
+  %gep = getelementptr inbounds i64, ptr %out, i64 4
   %tmp0 = atomicrmw sub ptr %gep, i64 %in seq_cst, !amdgpu.no.remote.memory !0, !noalias.addrspace !1
   ret void
 }
@@ -1882,7 +1882,7 @@ define i64 @flat_atomic_sub_i64_ret_offset__amdgpu_no_remote_memory(ptr %out, i6
 ; GFX9-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX9-NEXT:    buffer_wbinvl1_vol
 ; GFX9-NEXT:    s_setpc_b64 s[30:31]
-  %gep = getelementptr i64, ptr %out, i64 4
+  %gep = getelementptr inbounds i64, ptr %out, i64 4
   %result = atomicrmw sub ptr %gep, i64 %in seq_cst, !amdgpu.no.remote.memory !0, !noalias.addrspace !1
   ret i64 %result
 }
@@ -2049,7 +2049,7 @@ define void @flat_atomic_and_i64_noret_offset(ptr %out, i64 %in) {
 ; GFX9-NEXT:  ; %bb.2: ; %atomicrmw.end
 ; GFX9-NEXT:    s_or_b64 exec, exec, s[4:5]
 ; GFX9-NEXT:    s_setpc_b64 s[30:31]
-  %gep = getelementptr i64, ptr %out, i64 4
+  %gep = getelementptr inbounds i64, ptr %out, i64 4
   %tmp0 = atomicrmw and ptr %gep, i64 %in seq_cst, !noalias.addrspace !1
   ret void
 }
@@ -2220,7 +2220,7 @@ define i64 @flat_atomic_and_i64_ret_offset(ptr %out, i64 %in) {
 ; GFX9-NEXT:    v_mov_b32_e32 v0, v4
 ; GFX9-NEXT:    v_mov_b32_e32 v1, v5
 ; GFX9-NEXT:    s_setpc_b64 s[30:31]
-  %gep = getelementptr i64, ptr %out, i64 4
+  %gep = getelementptr inbounds i64, ptr %out, i64 4
   %result = atomicrmw and ptr %gep, i64 %in seq_cst, !noalias.addrspace !1
   ret i64 %result
 }
@@ -2411,7 +2411,7 @@ define amdgpu_gfx void @flat_atomic_and_i64_noret_offset_scalar(ptr inreg %out,
 ; GFX9-NEXT:  ; %bb.2: ; %atomicrmw.end
 ; GFX9-NEXT:    s_or_b64 exec, exec, s[34:35]
 ; GFX9-NEXT:    s_setpc_b64 s[30:31]
-  %gep = getelementptr i64, ptr %out, i64 4
+  %gep = getelementptr inbounds i64, ptr %out, i64 4
   %tmp0 = atomicrmw and ptr %gep, i64 %in seq_cst, !noalias.addrspace !1
   ret void
 }
@@ -2602,7 +2602,7 @@ define amdgpu_gfx i64 @flat_atomic_and_i64_ret_offset_scalar(ptr inreg %out, i64
 ; GFX9-NEXT:  ; %bb.2: ; %atomicrmw.end
 ; GFX9-NEXT:    s_or_b64 exec, exec, s[34:35]
 ; GFX9-NEXT:    s_setpc_b64 s[30:31]
-  %gep = getelementptr i64, ptr %out, i64 4
+  %gep = getelementptr inbounds i64, ptr %out, i64 4
   %result = atomicrmw and ptr %gep, i64 %in seq_cst, !noalias.addrspace !1
   ret i64 %result
 }
@@ -2635,7 +2635,7 @@ define void @flat_atomic_and_i64_noret_offset__amdgpu_no_remote_memory(ptr %out,
 ; GFX9-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX9-NEXT:    buffer_wbinvl1_vol
 ; GFX9-NEXT:    s_setpc_b64 s[30:31]
-  %gep = getelementptr i64, ptr %out, i64 4
+  %gep = getelementptr inbounds i64, ptr %out, i64 4
   %tmp0 = atomicrmw and ptr %gep, i64 %in seq_cst, !amdgpu.no.remote.memory !0, !noalias.addrspace !1
   ret void
 }
@@ -2668,7 +2668,7 @@ define i64 @flat_atomic_and_i64_ret_offset__amdgpu_no_remote_memory(ptr %out, i6
 ; GFX9-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX9-NEXT:    buffer_wbinvl1_vol
 ; GFX9-NEXT:    s_setpc_b64 s[30:31]
-  %gep = getelementptr i64, ptr %out, i64 4
+  %gep = getelementptr inbounds i64, ptr %out, i64 4
   %result = atomicrmw and ptr %gep, i64 %in seq_cst, !amdgpu.no.remote.memory !0, !noalias.addrspace !1
   ret i64 %result
 }
@@ -2847,7 +2847,7 @@ define void @flat_atomic_nand_i64_noret_offset(ptr %out, i64 %in) {
 ; GFX9-NEXT:  ; %bb.2: ; %atomicrmw.end
 ; GFX9-NEXT:    s_or_b64 exec, exec, s[4:5]
 ; GFX9-NEXT:    s_setpc_b64 s[30:31]
-  %gep = getelementptr i64, ptr %out, i64 4
+  %gep = getelementptr inbounds i64, ptr %out, i64 4
   %tmp0 = atomicrmw nand ptr %gep, i64 %in seq_cst, !noalias.addrspace !1
   ret void
 }
@@ -3030,7 +3030,7 @@ define i64 @flat_atomic_nand_i64_ret_offset(ptr %out, i64 %in) {
 ; GFX9-NEXT:    v_mov_b32_e32 v0, v4
 ; GFX9-NEXT:    v_mov_b32_e32 v1, v5
 ; GFX9-NEXT:    s_setpc_b64 s[30:31]
-  %gep = getelementptr i64, ptr %out, i64 4
+  %gep = getelementptr inbounds i64, ptr %out, i64 4
   %result = atomicrmw nand ptr %gep, i64 %in seq_cst, !noalias.addrspace !1
   ret i64 %result
 }
@@ -3233,7 +3233,7 @@ define amdgpu_gfx void @flat_atomic_nand_i64_noret_offset_scalar(ptr inreg %out,
 ; GFX9-NEXT:  ; %bb.2: ; %atomicrmw.end
 ; GFX9-NEXT:    s_or_b64 exec, exec, s[34:35]
 ; GFX9-NEXT:    s_setpc_b64 s[30:31]
-  %gep = getelementptr i64, ptr %out, i64 4
+  %gep = getelementptr inbounds i64, ptr %out, i64 4
   %tmp0 = atomicrmw nand ptr %gep, i64 %in seq_cst, !noalias.addrspace !1
   ret void
 }
@@ -3436,7 +3436,7 @@ define amdgpu_gfx i64 @flat_atomic_nand_i64_ret_offset_scalar(ptr inreg %out, i6
 ; GFX9-NEXT:  ; %bb.2: ; %atomicrmw.end
 ; GFX9-NEXT:    s_or_b64 exec, exec, s[34:35]
 ; GFX9-NEXT:    s_setpc_b64 s[30:31]
-  %gep = getelementptr i64, ptr %out, i64 4
+  %gep = getelementptr inbounds i64, ptr %out, i64 4
   %result = atomicrmw nand ptr %gep, i64 %in seq_cst, !noalias.addrspace !1
   ret i64 %result
 }
@@ -3526,7 +3526,7 @@ define void @flat_atomic_nand_i64_noret_offset__amdgpu_no_remote_memory(ptr %out
 ; GFX9-NEXT:  ; %bb.2: ; %atomicrmw.end
 ; GFX9-NEXT:    s_or_b64 exec, exec, s[4:5]
 ; GFX9-NEXT:    s_setpc_b64 s[30:31]
-  %gep = getelementptr i64, ptr %out, i64 4
+  %gep = getelementptr inbounds i64, ptr %out, i64 4
   %tmp0 = atomicrmw nand ptr %gep, i64 %in seq_cst, !amdgpu.no.remote.memory !0, !noalias.addrspace !1
   ret void
 }
@@ -3618,7 +3618,7 @@ define i64 @flat_atomic_nand_i64_ret_offset__amdgpu_no_remote_memory(ptr %out, i
 ; GFX9-NEXT:    v_mov_b32_e32 v0, v4
 ; GFX9-NEXT:    v_mov_b32_e32 v1, v5
 ; GFX9-NEXT:    s_setpc_b64 s[30:31]
-  %gep = getelementptr i64, ptr %out, i64 4
+  %gep = getelementptr inbounds i64, ptr %out, i64 4
   %result = atomicrmw nand ptr %gep, i64 %in seq_cst, !amdgpu.no.remote.memory !0, !noalias.addrspace !1
   ret i64 %result
 }
@@ -3785,7 +3785,7 @@ define void @flat_atomic_or_i64_noret_offset(ptr %out, i64 %in) {
 ; GFX9-NEXT:  ; %bb.2: ; %atomicrmw.end
 ; GFX9-NEXT:    s_or_b64 exec, exec, s[4:5]
 ; GFX9-NEXT:    s_setpc_b64 s[30:31]
-  %gep = getelementptr i64, ptr %out, i64 4
+  %gep = getelementptr inbounds i64, ptr %out, i64 4
   %tmp0 = atomicrmw or ptr %gep, i64 %in seq_cst, !noalias.addrspace !1
   ret void
 }
@@ -3956,7 +3956,7 @@ define i64 @flat_atomic_or_i64_ret_offset(ptr %out, i64 %in) {
 ; GFX9-NEXT:    v_mov_b32_e32 v0, v4
 ; GFX9-NEXT:    v_mov_b32_e32 v1, v5
 ; GFX9-NEXT:    s_setpc_b64 s[30:31]
-  %gep = getelementptr i64, ptr %out, i64 4
+  %gep = getelementptr inbounds i64, ptr %out, i64 4
   %result = atomicrmw or ptr %gep, i64 %in seq_cst, !noalias.addrspace !1
   ret i64 %result
 }
@@ -4147,7 +4147,7 @@ define amdgpu_gfx void @flat_atomic_or_i64_noret_offset_scalar(ptr inreg %out, i
 ; GFX9-NEXT:  ; %bb.2: ; %atomicrmw.end
 ; GFX9-NEXT:    s_or_b64 exec, exec, s[34:35]
 ; GFX9-NEXT:    s_setpc_b64 s[30:31]
-  %gep = getelementptr i64, ptr %out, i64 4
+  %gep = getelementptr inbounds i64, ptr %out, i64 4
   %tmp0 = atomicrmw or ptr %gep, i64 %in seq_cst, !noalias.addrspace !1
   ret void
 }
@@ -4338,7 +4338,7 @@ define amdgpu_gfx i64 @flat_atomic_or_i64_ret_offset_scalar(ptr inreg %out, i64
 ; GFX9-NEXT:  ; %bb.2: ; %atomicrmw.end
 ; GFX9-NEXT:    s_or_b64 exec, exec, s[34:35]
 ; GFX9-NEXT:    s_setpc_b64 s[30:31]
-  %gep = getelementptr i64, ptr %out, i64 4
+  %gep = getelementptr inbounds i64, ptr %out, i64 4
   %result = atomicrmw or ptr %gep, i64 %in seq_cst, !noalias.addrspace !1
   ret i64 %result
 }
@@ -4371,7 +4371,7 @@ define void @flat_atomic_or_i64_noret_offset__amdgpu_no_remote_memory(ptr %out,
 ; GFX9-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX9-NEXT:    buffer_wbinvl1_vol
 ; GFX9-NEXT:    s_setpc_b64 s[30:31]
-  %gep = getelementptr i64, ptr %out, i64 4
+  %gep = getelementptr inbounds i64, ptr %out, i64 4
   %tmp0 = atomicrmw or ptr %gep, i64 %in seq_cst, !amdgpu.no.remote.memory !0, !noalias.addrspace !1
   ret void
 }
@@ -4404,7 +4404,7 @@ define i64 @flat_atomic_or_i64_ret_offset__amdgpu_no_remote_memory(ptr %out, i64
 ; GFX9-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX9-NEXT:    buffer_wbinvl1_vol
 ; GFX9-NEXT:    s_setpc_b64 s[30:31]
-  %gep = getelementptr i64, ptr %out, i64 4
+  %gep = getelementptr inbounds i64, ptr %out, i64 4
   %result = atomicrmw or ptr %gep, i64 %in seq_cst, !amdgpu.no.remote.memory !0, !noalias.addrspace !1
   ret i64 %result
 }
@@ -4571,7 +4571,7 @@ define void @flat_atomic_xor_i64_noret_offset(ptr %out, i64 %in) {
 ; GFX9-NEXT:  ; %bb.2: ; %atomicrmw.end
 ; GFX9-NEXT:    s_or_b64 exec, exec, s[4:5]
 ; GFX9-NEXT:    s_setpc_b64 s[30:31]
-  %gep = getelementptr i64, ptr %out, i64 4
+  %gep = getelementptr inbounds i64, ptr %out, i64 4
   %tmp0 = atomicrmw xor ptr %gep, i64 %in seq_cst, !noalias.addrspace !1
   ret void
 }
@@ -4742,7 +4742,7 @@ define i64 @flat_atomic_xor_i64_ret_offset(ptr %out, i64 %in) {
 ; GFX9-NEXT:    v_mov_b32_e32 v0, v4
 ; GFX9-NEXT:    v_mov_b32_e32 v1, v5
 ; GFX9-NEXT:    s_setpc_b64 s[30:31]
-  %gep = getelementptr i64, ptr %out, i64 4
+  %gep = getelementptr inbounds i64, ptr %out, i64 4
   %result = atomicrmw xor ptr %gep, i64 %in seq_cst, !noalias.addrspace !1
   ret i64 %result
 }
@@ -4933,7 +4933,7 @@ define amdgpu_gfx void @flat_atomic_xor_i64_noret_offset_scalar(ptr inreg %out,
 ; GFX9-NEXT:  ; %bb.2: ; %atomicrmw.end
 ; GFX9-NEXT:    s_or_b64 exec, exec, s[34:35]
 ; GFX9-NEXT:    s_setpc_b64 s[30:31]
-  %gep = getelementptr i64, ptr %out, i64 4
+  %gep = getelementptr inbounds i64, ptr %out, i64 4
   %tmp0 = atomicrmw xor ptr %gep, i64 %in seq_cst, !noalias.addrspace !1
   ret void
 }
@@ -5124,7 +5124,7 @@ define amdgpu_gfx i64 @flat_atomic_xor_i64_ret_offset_scalar(ptr inreg %out, i64
 ; GFX9-NEXT:  ; %bb.2: ; %atomicrmw.end
 ; GFX9-NEXT:    s_or_b64 exec, exec, s[34:35]
 ; GFX9-NEXT:    s_setpc_b64 s[30:31]
-  %gep = getelementptr i64, ptr %out, i64 4
+  %gep = getelementptr inbounds i64, ptr %out, i64 4
   %result = atomicrmw xor ptr %gep, i64 %in seq_cst, !noalias.addrspace !1
   ret i64 %result
 }
@@ -5157,7 +5157,7 @@ define void @flat_atomic_xor_i64_noret_offset__amdgpu_no_remote_memory(ptr %out,
 ; GFX9-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX9-NEXT:    buffer_wbinvl1_vol
 ; GFX9-NEXT:    s_setpc_b64 s[30:31]
-  %gep = getelementptr i64, ptr %out, i64 4
+  %gep = getelementptr inbounds i64, ptr %out, i64 4
   %tmp0 = atomicrmw xor ptr %gep, i64 %in seq_cst, !amdgpu.no.remote.memory !0, !noalias.addrspace !1
   ret void
 }
@@ -5190,7 +5190,7 @@ define i64 @flat_atomic_xor_i64_ret_offset__amdgpu_no_remote_memory(ptr %out, i6
 ; GFX9-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX9-NEXT:    buffer_wbinvl1_vol
 ; GFX9-NEXT:    s_setpc_b64 s[30:31]
-  %gep = getelementptr i64, ptr %out, i64 4
+  %gep = getelementptr inbounds i64, ptr %out, i64 4
   %result = atomicrmw xor ptr %gep, i64 %in seq_cst, !amdgpu.no.remote.memory !0, !noalias.addrspace !1
   ret i64 %result
 }
@@ -5363,7 +5363,7 @@ define void @flat_atomic_max_i64_noret_offset(ptr %out, i64 %in) {
 ; GFX9-NEXT:  ; %bb.2: ; %atomicrmw.end
 ; GFX9-NEXT:    s_or_b64 exec, exec, s[4:5]
 ; GFX9-NEXT:    s_setpc_b64 s[30:31]
-  %gep = getelementptr i64, ptr %out, i64 4
+  %gep = getelementptr inbounds i64, ptr %out, i64 4
   %tmp0 = atomicrmw max ptr %gep, i64 %in seq_cst, !noalias.addrspace !1
   ret void
 }
@@ -5540,7 +5540,7 @@ define i64 @flat_atomic_max_i64_ret_offset(ptr %out, i64 %in) {
 ; GFX9-NEXT:    v_mov_b32_e32 v0, v4
 ; GFX9-NEXT:    v_mov_b32_e32 v1, v5
 ; GFX9-NEXT:    s_setpc_b64 s[30:31]
-  %gep = getelementptr i64, ptr %out, i64 4
+  %gep = getelementptr inbounds i64, ptr %out, i64 4
   %result = atomicrmw max ptr %gep, i64 %in seq_cst, !noalias.addrspace !1
   ret i64 %result
 }
@@ -5749,7 +5749,7 @@ define amdgpu_gfx void @flat_atomic_max_i64_noret_offset_scalar(ptr inreg %out,
 ; GFX9-NEXT:  ; %bb.2: ; %atomicrmw.end
 ; GFX9-NEXT:    s_or_b64 exec, exec, s[34:35]
 ; GFX9-NEXT:    s_setpc_b64 s[30:31]
-  %gep = getelementptr i64, ptr %out, i64 4
+  %gep = getelementptr inbounds i64, ptr %out, i64 4
   %tmp0 = atomicrmw max ptr %gep, i64 %in seq_cst, !noalias.addrspace !1
   ret void
 }
@@ -5958,7 +5958,7 @@ define amdgpu_gfx i64 @flat_atomic_max_i64_ret_offset_scalar(ptr inreg %out, i64
 ; GFX9-NEXT:  ; %bb.2: ; %atomicrmw.end
 ; GFX9-NEXT:    s_or_b64 exec, exec, s[34:35]
 ; GFX9-NEXT:    s_setpc_b64 s[30:31]
-  %gep = getelementptr i64, ptr %out, i64 4
+  %gep = getelementptr inbounds i64, ptr %out, i64 4
   %result = atomicrmw max ptr %gep, i64 %in seq_cst, !noalias.addrspace !1
   ret i64 %result
 }
@@ -6064,8 +6064,8 @@ define amdgpu_kernel void @atomic_max_i64_addr64_offset(ptr %out, i64 %in, i64 %
 ; GFX9-NEXT:  ; %bb.2: ; %atomicrmw.end
 ; GFX9-NEXT:    s_endpgm
 entry:
-  %ptr = getelementptr i64, ptr %out, i64 %index
-  %gep = getelementptr i64, ptr %ptr, i64 4
+  %ptr = getelementptr inbounds i64, ptr %out, i64 %index
+  %gep = getelementptr inbounds i64, ptr %ptr, i64 4
   %tmp0 = atomicrmw max ptr %gep, i64 %in seq_cst, !noalias.addrspace !1
   ret void
 }
@@ -6180,8 +6180,8 @@ define amdgpu_kernel void @atomic_max_i64_ret_addr64_offset(ptr %out, ptr %out2,
 ; GFX9-NEXT:    flat_store_dwordx2 v[0:1], v[2:3]
 ; GFX9-NEXT:    s_endpgm
 entry:
-  %ptr = getelementptr i64, ptr %out, i64 %index
-  %gep = getelementptr i64, ptr %ptr, i64 4
+  %ptr = getelementptr inbounds i64, ptr %out, i64 %index
+  %gep = getelementptr inbounds i64, ptr %ptr, i64 4
   %tmp0 = atomicrmw max ptr %gep, i64 %in seq_cst, !noalias.addrspace !1
   store i64 %tmp0, ptr %out2
   ret void
@@ -6284,7 +6284,7 @@ define amdgpu_kernel void @atomic_max_i64_addr64(ptr %out, i64 %in, i64 %index)
 ; GFX9-NEXT:  ; %bb.2: ; %atomicrmw.end
 ; GFX9-NEXT:    s_endpgm
 entry:
-  %ptr = getelementptr i64, ptr %out, i64 %index
+  %ptr = getelementptr inbounds i64, ptr %out, i64 %index
   %tmp0 = atomicrmw max ptr %ptr, i64 %in seq_cst, !noalias.addrspace !1
   ret void
 }
@@ -6395,7 +6395,7 @@ define amdgpu_kernel void @atomic_max_i64_ret_addr64(ptr %out, ptr %out2, i64 %i
 ; GFX9-NEXT:    flat_store_dwordx2 v[0:1], v[2:3]
 ; GFX9-NEXT:    s_endpgm
 entry:
-  %ptr = getelementptr i64, ptr %out, i64 %index
+  %ptr = getelementptr inbounds i64, ptr %out, i64 %index
   %tmp0 = atomicrmw max ptr %ptr, i64 %in seq_cst, !noalias.addrspace !1
   store i64 %tmp0, ptr %out2
   ret void
@@ -6429,7 +6429,7 @@ define void @flat_atomic_max_i64_noret_offset__amdgpu_no_remote_memory(ptr %out,
 ; GFX9-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX9-NEXT:    buffer_wbinvl1_vol
 ; GFX9-NEXT:    s_setpc_b64 s[30:31]
-  %gep = getelementptr i64, ptr %out, i64 4
+  %gep = getelementptr inbounds i64, ptr %out, i64 4
   %tmp0 = atomicrmw max ptr %gep, i64 %in seq_cst, !amdgpu.no.remote.memory !0, !noalias.addrspace !1
   ret void
 }
@@ -6462,7 +6462,7 @@ define i64 @flat_atomic_max_i64_ret_offset__amdgpu_no_remote_memory(ptr %out, i6
 ; GFX9-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX9-NEXT:    buffer_wbinvl1_vol
 ; GFX9-NEXT:    s_setpc_b64 s[30:31]
-  %gep = getelementptr i64, ptr %out, i64 4
+  %gep = getelementptr inbounds i64, ptr %out, i64 4
   %result = atomicrmw max ptr %gep, i64 %in seq_cst, !amdgpu.no.remote.memory !0, !noalias.addrspace !1
   ret i64 %result
 }
@@ -6635,7 +6635,7 @@ define void @flat_atomic_umax_i64_noret_offset(ptr %out, i64 %in) {
 ; GFX9-NEXT:  ; %bb.2: ; %atomicrmw.end
 ; GFX9-NEXT:    s_or_b64 exec, exec, s[4:5]
 ; GFX9-NEXT:    s_setpc_b64 s[30:31]
-  %gep = getelementptr i64, ptr %out, i64 4
+  %gep = getelementptr inbounds i64, ptr %out, i64 4
   %tmp0 = atomicrmw umax ptr %gep, i64 %in seq_cst, !noalias.addrspace !1
   ret void
 }
@@ -6812,7 +6812,7 @@ define i64 @flat_atomic_umax_i64_ret_offset(ptr %out, i64 %in) {
 ; GFX9-NEXT:    v_mov_b32_e32 v0, v4
 ; GFX9-NEXT:    v_mov_b32_e32 v1, v5
 ; GFX9-NEXT:    s_setpc_b64 s[30:31]
-  %gep = getelementptr i64, ptr %out, i64 4
+  %gep = getelementptr inbounds i64, ptr %out, i64 4
   %result = atomicrmw umax ptr %gep, i64 %in seq_cst, !noalias.addrspace !1
   ret i64 %result
 }
@@ -7021,7 +7021,7 @@ define amdgpu_gfx void @flat_atomic_umax_i64_noret_offset_scalar(ptr inreg %out,
 ; GFX9-NEXT:  ; %bb.2: ; %atomicrmw.end
 ; GFX9-NEXT:    s_or_b64 exec, exec, s[34:35]
 ; GFX9-NEXT:    s_setpc_b64 s[30:31]
-  %gep = getelementptr i64, ptr %out, i64 4
+  %gep = getelementptr inbounds i64, ptr %out, i64 4
   %tmp0 = atomicrmw umax ptr %gep, i64 %in seq_cst, !noalias.addrspace !1
   ret void
 }
@@ -7230,7 +7230,7 @@ define amdgpu_gfx i64 @flat_atomic_umax_i64_ret_offset_scalar(ptr inreg %out, i6
 ; GFX9-NEXT:  ; %bb.2: ; %atomicrmw.end
 ; GFX9-NEXT:    s_or_b64 exec, exec, s[34:35]
 ; GFX9-NEXT:    s_setpc_b64 s[30:31]
-  %gep = getelementptr i64, ptr %out, i64 4
+  %gep = getelementptr inbounds i64, ptr %out, i64 4
   %result = atomicrmw umax ptr %gep, i64 %in seq_cst, !noalias.addrspace !1
   ret i64 %result
 }
@@ -7336,8 +7336,8 @@ define amdgpu_kernel void @atomic_umax_i64_addr64_offset(ptr %out, i64 %in, i64
 ; GFX9-NEXT:  ; %bb.2: ; %atomicrmw.end
 ; GFX9-NEXT:    s_endpgm
 entry:
-  %ptr = getelementptr i64, ptr %out, i64 %index
-  %gep = getelementptr i64, ptr %ptr, i64 4
+  %ptr = getelementptr inbounds i64, ptr %out, i64 %index
+  %gep = getelementptr inbounds i64, ptr %ptr, i64 4
   %tmp0 = atomicrmw umax ptr %gep, i64 %in seq_cst, !noalias.addrspace !1
   ret void
 }
@@ -7452,8 +7452,8 @@ define amdgpu_kernel void @atomic_umax_i64_ret_addr64_offset(ptr %out, ptr %out2
 ; GFX9-NEXT:    flat_store_dwordx2 v[0:1], v[2:3]
 ; GFX9-NEXT:    s_endpgm
 entry:
-  %ptr = getelementptr i64, ptr %out, i64 %index
-  %gep = getelementptr i64, ptr %ptr, i64 4
+  %ptr = getelementptr inbounds i64, ptr %out, i64 %index
+  %gep = getelementptr inbounds i64, ptr %ptr, i64 4
   %tmp0 = atomicrmw umax ptr %gep, i64 %in seq_cst, !noalias.addrspace !1
   store i64 %tmp0, ptr %out2
   ret void
@@ -7565,7 +7565,7 @@ define amdgpu_kernel void @atomic_umax_i64_ret_addr64(ptr %out, ptr %out2, i64 %
 ; GFX9-NEXT:    flat_store_dwordx2 v[0:1], v[2:3]
 ; GFX9-NEXT:    s_endpgm
 entry:
-  %ptr = getelementptr i64, ptr %out, i64 %index
+  %ptr = getelementptr inbounds i64, ptr %out, i64 %index
   %tmp0 = atomicrmw umax ptr %ptr, i64 %in seq_cst, !noalias.addrspace !1
   store i64 %tmp0, ptr %out2
   ret void
@@ -7599,7 +7599,7 @@ define void @flat_atomic_umax_i64_noret_offset__amdgpu_no_remote_memory(ptr %out
 ; GFX9-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX9-NEXT:    buffer_wbinvl1_vol
 ; GFX9-NEXT:    s_setpc_b64 s[30:31]
-  %gep = getelementptr i64, ptr %out, i64 4
+  %gep = getelementptr inbounds i64, ptr %out, i64 4
   %tmp0 = atomicrmw umax ptr %gep, i64 %in seq_cst, !amdgpu.no.remote.memory !0, !noalias.addrspace !1
   ret void
 }
@@ -7632,7 +7632,7 @@ define i64 @flat_atomic_umax_i64_ret_offset__amdgpu_no_remote_memory(ptr %out, i
 ; GFX9-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX9-NEXT:    buffer_wbinvl1_vol
 ; GFX9-NEXT:    s_setpc_b64 s[30:31]
-  %gep = getelementptr i64, ptr %out, i64 4
+  %gep = getelementptr inbounds i64, ptr %out, i64 4
   %result = atomicrmw umax ptr %gep, i64 %in seq_cst, !amdgpu.no.remote.memory !0, !noalias.addrspace !1
   ret i64 %result
 }
@@ -7805,7 +7805,7 @@ define void @flat_atomic_umin_i64_noret_offset(ptr %out, i64 %in) {
 ; GFX9-NEXT:  ; %bb.2: ; %atomicrmw.end
 ; GFX9-NEXT:    s_or_b64 exec, exec, s[4:5]
 ; GFX9-NEXT:    s_setpc_b64 s[30:31]
-  %gep = getelementptr i64, ptr %out, i64 4
+  %gep = getelementptr inbounds i64, ptr %out, i64 4
   %tmp0 = atomicrmw umin ptr %gep, i64 %in seq_cst, !noalias.addrspace !1
   ret void
 }
@@ -7982,7 +7982,7 @@ define i64 @flat_atomic_umin_i64_ret_offset(ptr %out, i64 %in) {
 ; GFX9-NEXT:    v_mov_b32_e32 v0, v4
 ; GFX9-NEXT:    v_mov_b32_e32 v1, v5
 ; GFX9-NEXT:    s_setpc_b64 s[30:31]
-  %gep = getelementptr i64, ptr %out, i64 4
+  %gep = getelementptr inbounds i64, ptr %out, i64 4
   %result = atomicrmw umin ptr %gep, i64 %in seq_cst, !noalias.addrspace !1
   ret i64 %result
 }
@@ -8191,7 +8191,7 @@ define amdgpu_gfx void @flat_atomic_umin_i64_noret_offset_scalar(ptr inreg %out,
 ; GFX9-NEXT:  ; %bb.2: ; %atomicrmw.end
 ; GFX9-NEXT:    s_or_b64 exec, exec, s[34:35]
 ; GFX9-NEXT:    s_setpc_b64 s[30:31]
-  %gep = getelementptr i64, ptr %out, i64 4
+  %gep = getelementptr inbounds i64, ptr %out, i64 4
   %tmp0 = atomicrmw umin ptr %gep, i64 %in seq_cst, !noalias.addrspace !1
   ret void
 }
@@ -8400,7 +8400,7 @@ define amdgpu_gfx i64 @flat_atomic_umin_i64_ret_offset_scalar(ptr inreg %out, i6
 ; GFX9-NEXT:  ; %bb.2: ; %atomicrmw.end
 ; GFX9-NEXT:    s_or_b64 exec, exec, s[34:35]
 ; GFX9-NEXT:    s_setpc_b64 s[30:31]
-  %gep = getelementptr i64, ptr %out, i64 4
+  %gep = getelementptr inbounds i64, ptr %out, i64 4
   %result = atomicrmw umin ptr %gep, i64 %in seq_cst, !noalias.addrspace !1
   ret i64 %result
 }
@@ -8433,7 +8433,7 @@ define void @flat_atomic_umin_i64_noret_offset__amdgpu_no_remote_memory(ptr %out
 ; GFX9-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX9-NEXT:    buffer_wbinvl1_vol
 ; GFX9-NEXT:    s_setpc_b64 s[30:31]
-  %gep = getelementptr i64, ptr %out, i64 4
+  %gep = getelementptr inbounds i64, ptr %out, i64 4
   %tmp0 = atomicrmw umin ptr %gep, i64 %in seq_cst, !amdgpu.no.remote.memory !0, !noalias.addrspace !1
   ret void
 }
@@ -8466,7 +8466,7 @@ define i64 @flat_atomic_umin_i64_ret_offset__amdgpu_no_remote_memory(ptr %out, i
 ; GFX9-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX9-NEXT:    buffer_wbinvl1_vol
 ; GFX9-NEXT:    s_setpc_b64 s[30:31]
-  %gep = getelementptr i64, ptr %out, i64 4
+  %gep = getelementptr inbounds i64, ptr %out, i64 4
   %result = atomicrmw umin ptr %gep, i64 %in seq_cst, !amdgpu.no.remote.memory !0, !noalias.addrspace !1
   ret i64 %result
 }
@@ -8639,7 +8639,7 @@ define void @flat_atomic_min_i64_noret_offset(ptr %out, i64 %in) {
 ; GFX9-NEXT:  ; %bb.2: ; %atomicrmw.end
 ; GFX9-NEXT:    s_or_b64 exec, exec, s[4:5]
 ; GFX9-NEXT:    s_setpc_b64 s[30:31]
-  %gep = getelementptr i64, ptr %out, i64 4
+  %gep = getelementptr inbounds i64, ptr %out, i64 4
   %tmp0 = atomicrmw min ptr %gep, i64 %in seq_cst, !noalias.addrspace !1
   ret void
 }
@@ -8816,7 +8816,7 @@ define i64 @flat_atomic_min_i64_ret_offset(ptr %out, i64 %in) {
 ; GFX9-NEXT:    v_mov_b32_e32 v0, v4
 ; GFX9-NEXT:    v_mov_b32_e32 v1, v5
 ; GFX9-NEXT:    s_setpc_b64 s[30:31]
-  %gep = getelementptr i64, ptr %out, i64 4
+  %gep = getelementptr inbounds i64, ptr %out, i64 4
   %result = atomicrmw min ptr %gep, i64 %in seq_cst, !noalias.addrspace !1
   ret i64 %result
 }
@@ -9025,7 +9025,7 @@ define amdgpu_gfx void @flat_atomic_min_i64_noret_offset_scalar(ptr inreg %out,
 ; GFX9-NEXT:  ; %bb.2: ; %atomicrmw.end
 ; GFX9-NEXT:    s_or_b64 exec, exec, s[34:35]
 ; GFX9-NEXT:    s_setpc_b64 s[30:31]
-  %gep = getelementptr i64, ptr %out, i64 4
+  %gep = getelementptr inbounds i64, ptr %out, i64 4
   %tmp0 = atomicrmw min ptr %gep, i64 %in seq_cst, !noalias.addrspace !1
   ret void
 }
@@ -9234,7 +9234,7 @@ define amdgpu_gfx i64 @flat_atomic_min_i64_ret_offset_scalar(ptr inreg %out, i64
 ; GFX9-NEXT:  ; %bb.2: ; %atomicrmw.end
 ; GFX9-NEXT:    s_or_b64 exec, exec, s[34:35]
 ; GFX9-NEXT:    s_setpc_b64 s[30:31]
-  %gep = getelementptr i64, ptr %out, i64 4
+  %gep = getelementptr inbounds i64, ptr %out, i64 4
   %result = atomicrmw min ptr %gep, i64 %in seq_cst, !noalias.addrspace !1
   ret i64 %result
 }
@@ -9340,8 +9340,8 @@ define amdgpu_kernel void @atomic_min_i64_addr64_offset(ptr %out, i64 %in, i64 %
 ; GFX9-NEXT:  ; %bb.2: ; %atomicrmw.end
 ; GFX9-NEXT:    s_endpgm
 entry:
-  %ptr = getelementptr i64, ptr %out, i64 %index
-  %gep = getelementptr i64, ptr %ptr, i64 4
+  %ptr = getelementptr inbounds i64, ptr %out, i64 %index
+  %gep = getelementptr inbounds i64, ptr %ptr, i64 4
   %tmp0 = atomicrmw min ptr %gep, i64 %in seq_cst, !noalias.addrspace !1
   ret void
 }
@@ -9456,8 +9456,8 @@ define amdgpu_kernel void @atomic_min_i64_ret_addr64_offset(ptr %out, ptr %out2,
 ; GFX9-NEXT:    flat_store_dwordx2 v[0:1], v[2:3]
 ; GFX9-NEXT:    s_endpgm
 entry:
-  %ptr = getelementptr i64, ptr %out, i64 %index
-  %gep = getelementptr i64, ptr %ptr, i64 4
+  %ptr = getelementptr inbounds i64, ptr %out, i64 %index
+  %gep = getelementptr inbounds i64, ptr %ptr, i64 4
   %tmp0 = atomicrmw min ptr %gep, i64 %in seq_cst, !noalias.addrspace !1
   store i64 %tmp0, ptr %out2
   ret void
@@ -9664,7 +9664,7 @@ define amdgpu_kernel void @atomic_min_i64_ret_addr64(ptr %out, ptr %out2, i64 %i
 ; GFX9-NEXT:    flat_store_dwordx2 v[0:1], v[2:3]
 ; GFX9-NEXT:    s_endpgm
 entry:
-  %ptr = getelementptr i64, ptr %out, i64 %index
+  %ptr = getelementptr inbounds i64, ptr %out, i64 %index
   %tmp0 = atomicrmw min ptr %ptr, i64 %in seq_cst, !noalias.addrspace !1, !noalias.addrspace !1
   store i64 %tmp0, ptr %out2
   ret void
@@ -9698,7 +9698,7 @@ define void @flat_atomic_min_i64_noret_offset__amdgpu_no_remote_memory(ptr %out,
 ; GFX9-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX9-NEXT:    buffer_wbinvl1_vol
 ; GFX9-NEXT:    s_setpc_b64 s[30:31]
-  %gep = getelementptr i64, ptr %out, i64 4
+  %gep = getelementptr inbounds i64, ptr %out, i64 4
   %tmp0 = atomicrmw min ptr %gep, i64 %in seq_cst, !amdgpu.no.remote.memory !0, !noalias.addrspace !1
   ret void
 }
@@ -9731,7 +9731,7 @@ define i64 @flat_atomic_min_i64_ret_offset__amdgpu_no_remote_memory(ptr %out, i6
 ; GFX9-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX9-NEXT:    buffer_wbinvl1_vol
 ; GFX9-NEXT:    s_setpc_b64 s[30:31]
-  %gep = getelementptr i64, ptr %out, i64 4
+  %gep = getelementptr inbounds i64, ptr %out, i64 4
   %result = atomicrmw min ptr %gep, i64 %in seq_cst, !amdgpu.no.remote.memory !0, !noalias.addrspace !1
   ret i64 %result
 }
@@ -9916,7 +9916,7 @@ define void @flat_atomic_uinc_wrap_i64_noret_offset(ptr %out, i64 %in) {
 ; GFX9-NEXT:  ; %bb.2: ; %atomicrmw.end
 ; GFX9-NEXT:    s_or_b64 exec, exec, s[4:5]
 ; GFX9-NEXT:    s_setpc_b64 s[30:31]
-  %gep = getelementptr i64, ptr %out, i64 4
+  %gep = getelementptr inbounds i64, ptr %out, i64 4
   %tmp0 = atomicrmw uinc_wrap ptr %gep, i64 %in seq_cst, !noalias.addrspace !1
   ret void
 }
@@ -10105,7 +10105,7 @@ define i64 @flat_atomic_uinc_wrap_i64_ret_offset(ptr %out, i64 %in) {
 ; GFX9-NEXT:    v_mov_b32_e32 v0, v4
 ; GFX9-NEXT:    v_mov_b32_e32 v1, v5
 ; GFX9-NEXT:    s_setpc_b64 s[30:31]
-  %gep = getelementptr i64, ptr %out, i64 4
+  %gep = getelementptr inbounds i64, ptr %out, i64 4
   %result = atomicrmw uinc_wrap ptr %gep, i64 %in seq_cst, !noalias.addrspace !1
   ret i64 %result
 }
@@ -10314,7 +10314,7 @@ define amdgpu_gfx void @flat_atomic_uinc_wrap_i64_noret_offset_scalar(ptr inreg
 ; GFX9-NEXT:  ; %bb.2: ; %atomicrmw.end
 ; GFX9-NEXT:    s_or_b64 exec, exec, s[34:35]
 ; GFX9-NEXT:    s_setpc_b64 s[30:31]
-  %gep = getelementptr i64, ptr %out, i64 4
+  %gep = getelementptr inbounds i64, ptr %out, i64 4
   %tmp0 = atomicrmw uinc_wrap ptr %gep, i64 %in seq_cst, !noalias.addrspace !1
   ret void
 }
@@ -10523,7 +10523,7 @@ define amdgpu_gfx i64 @flat_atomic_uinc_wrap_i64_ret_offset_scalar(ptr inreg %ou
 ; GFX9-NEXT:  ; %bb.2: ; %atomicrmw.end
 ; GFX9-NEXT:    s_or_b64 exec, exec, s[34:35]
 ; GFX9-NEXT:    s_setpc_b64 s[30:31]
-  %gep = getelementptr i64, ptr %out, i64 4
+  %gep = getelementptr inbounds i64, ptr %out, i64 4
   %result = atomicrmw uinc_wrap ptr %gep, i64 %in seq_cst, !noalias.addrspace !1
   ret i64 %result
 }
@@ -10556,7 +10556,7 @@ define void @flat_atomic_uinc_wrap_i64_noret_offset__amdgpu_no_remote_memory(ptr
 ; GFX9-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX9-NEXT:    buffer_wbinvl1_vol
 ; GFX9-NEXT:    s_setpc_b64 s[30:31]
-  %gep = getelementptr i64, ptr %out, i64 4
+  %gep = getelementptr inbounds i64, ptr %out, i64 4
   %tmp0 = atomicrmw uinc_wrap ptr %gep, i64 %in seq_cst, !amdgpu.no.remote.memory !0, !noalias.addrspace !1
   ret void
 }
@@ -10589,7 +10589,7 @@ define i64 @flat_atomic_uinc_wrap_i64_ret_offset__amdgpu_no_remote_memory(ptr %o
 ; GFX9-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX9-NEXT:    buffer_wbinvl1_vol
 ; GFX9-NEXT:    s_setpc_b64 s[30:31]
-  %gep = getelementptr i64, ptr %out, i64 4
+  %gep = getelementptr inbounds i64, ptr %out, i64 4
   %result = atomicrmw uinc_wrap ptr %gep, i64 %in seq_cst, !amdgpu.no.remote.memory !0, !noalias.addrspace !1
   ret i64 %result
 }
@@ -10786,7 +10786,7 @@ define void @flat_atomic_udec_wrap_i64_noret_offset(ptr %out, i64 %in) {
 ; GFX9-NEXT:  ; %bb.2: ; %atomicrmw.end
 ; GFX9-NEXT:    s_or_b64 exec, exec, s[8:9]
 ; GFX9-NEXT:    s_setpc_b64 s[30:31]
-  %gep = getelementptr i64, ptr %out, i64 4
+  %gep = getelementptr inbounds i64, ptr %out, i64 4
   %tmp0 = atomicrmw udec_wrap ptr %gep, i64 %in seq_cst, !noalias.addrspace !1
   ret void
 }
@@ -10987,7 +10987,7 @@ define i64 @flat_atomic_udec_wrap_i64_ret_offset(ptr %out, i64 %in) {
 ; GFX9-NEXT:    v_mov_b32_e32 v0, v4
 ; GFX9-NEXT:    v_mov_b32_e32 v1, v5
 ; GFX9-NEXT:    s_setpc_b64 s[30:31]
-  %gep = getelementptr i64, ptr %out, i64 4
+  %gep = getelementptr inbounds i64, ptr %out, i64 4
   %result = atomicrmw udec_wrap ptr %gep, i64 %in seq_cst, !noalias.addrspace !1
   ret i64 %result
 }
@@ -11220,7 +11220,7 @@ define amdgpu_gfx void @flat_atomic_udec_wrap_i64_noret_offset_scalar(ptr inreg
 ; GFX9-NEXT:  ; %bb.2: ; %atomicrmw.end
 ; GFX9-NEXT:    s_or_b64 exec, exec, s[38:39]
 ; GFX9-NEXT:    s_setpc_b64 s[30:31]
-  %gep = getelementptr i64, ptr %out, i64 4
+  %gep = getelementptr inbounds i64, ptr %out, i64 4
   %tmp0 = atomicrmw udec_wrap ptr %gep, i64 %in seq_cst, !noalias.addrspace !1
   ret void
 }
@@ -11453,7 +11453,7 @@ define amdgpu_gfx i64 @flat_atomic_udec_wrap_i64_ret_offset_scalar(ptr inreg %ou
 ; GFX9-NEXT:  ; %bb.2: ; %atomicrmw.end
 ; GFX9-NEXT:    s_or_b64 exec, exec, s[38:39]
 ; GFX9-NEXT:    s_setpc_b64 s[30:31]
-  %gep = getelementptr i64, ptr %out, i64 4
+  %gep = getelementptr inbounds i64, ptr %out, i64 4
   %result = atomicrmw udec_wrap ptr %gep, i64 %in seq_cst, !noalias.addrspace !1
   ret i64 %result
 }
@@ -11486,7 +11486,7 @@ define void @flat_atomic_udec_wrap_i64_noret_offset__amdgpu_no_remote_memory(ptr
 ; GFX9-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX9-NEXT:    buffer_wbinvl1_vol
 ; GFX9-NEXT:    s_setpc_b64 s[30:31]
-  %gep = getelementptr i64, ptr %out, i64 4
+  %gep = getelementptr inbounds i64, ptr %out, i64 4
   %tmp0 = atomicrmw udec_wrap ptr %gep, i64 %in seq_cst, !amdgpu.no.remote.memory !0, !noalias.addrspace !1
   ret void
 }
@@ -11519,7 +11519,7 @@ define i64 @flat_atomic_udec_wrap_i64_ret_offset__amdgpu_no_remote_memory(ptr %o
 ; GFX9-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX9-NEXT:    buffer_wbinvl1_vol
 ; GFX9-NEXT:    s_setpc_b64 s[30:31]
-  %gep = getelementptr i64, ptr %out, i64 4
+  %gep = getelementptr inbounds i64, ptr %out, i64 4
   %result = atomicrmw udec_wrap ptr %gep, i64 %in seq_cst, !amdgpu.no.remote.memory !0, !noalias.addrspace !1
   ret i64 %result
 }
diff --git a/llvm/test/CodeGen/AMDGPU/llvm.amdgcn.atomic.dec.ll b/llvm/test/CodeGen/AMDGPU/llvm.amdgcn.atomic.dec.ll
index 37c57ef57570e..cfa03402ef048 100644
--- a/llvm/test/CodeGen/AMDGPU/llvm.amdgcn.atomic.dec.ll
+++ b/llvm/test/CodeGen/AMDGPU/llvm.amdgcn.atomic.dec.ll
@@ -579,7 +579,7 @@ define amdgpu_kernel void @flat_atomic_dec_ret_i32_offset(ptr %out, ptr %ptr) #0
 ; GFX9-NEXT:    v_mov_b32_e32 v1, s1
 ; GFX9-NEXT:    flat_store_dword v[0:1], v2
 ; GFX9-NEXT:    s_endpgm
-  %gep = getelementptr i32, ptr %ptr, i32 4
+  %gep = getelementptr inbounds i32, ptr %ptr, i32 4
   %result = call i32 @llvm.amdgcn.atomic.dec.i32.p0(ptr %gep, i32 42, i32 0, i32 0, i1 false)
   store i32 %result, ptr %out
   ret void
@@ -665,7 +665,7 @@ define amdgpu_kernel void @flat_atomic_dec_noret_i32_offset(ptr %ptr) nounwind {
 ; GFX9-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX9-NEXT:    buffer_wbinvl1_vol
 ; GFX9-NEXT:    s_endpgm
-  %gep = getelementptr i32, ptr %ptr, i32 4
+  %gep = getelementptr inbounds i32, ptr %ptr, i32 4
   %result = call i32 @llvm.amdgcn.atomic.dec.i32.p0(ptr %gep, i32 42, i32 0, i32 0, i1 false)
   ret void
 }
@@ -729,9 +729,9 @@ define amdgpu_kernel void @flat_atomic_dec_ret_i32_offset_addr64(ptr %out, ptr %
 ; GFX9-NEXT:    flat_store_dword v[0:1], v3
 ; GFX9-NEXT:    s_endpgm
   %id = call i32 @llvm.amdgcn.workitem.id.x()
-  %gep.tid = getelementptr i32, ptr %ptr, i32 %id
-  %out.gep = getelementptr i32, ptr %out, i32 %id
-  %gep = getelementptr i32, ptr %gep.tid, i32 5
+  %gep.tid = getelementptr inbounds i32, ptr %ptr, i32 %id
+  %out.gep = getelementptr inbounds i32, ptr %out, i32 %id
+  %gep = getelementptr inbounds i32, ptr %gep.tid, i32 5
   %result = call i32 @llvm.amdgcn.atomic.dec.i32.p0(ptr %gep, i32 42, i32 0, i32 0, i1 false)
   store i32 %result, ptr %out.gep
   ret void
@@ -784,8 +784,8 @@ define amdgpu_kernel void @flat_atomic_dec_noret_i32_offset_addr64(ptr %ptr) #0
 ; GFX9-NEXT:    buffer_wbinvl1_vol
 ; GFX9-NEXT:    s_endpgm
   %id = call i32 @llvm.amdgcn.workitem.id.x()
-  %gep.tid = getelementptr i32, ptr %ptr, i32 %id
-  %gep = getelementptr i32, ptr %gep.tid, i32 5
+  %gep.tid = getelementptr inbounds i32, ptr %ptr, i32 %id
+  %gep = getelementptr inbounds i32, ptr %gep.tid, i32 5
   %result = call i32 @llvm.amdgcn.atomic.dec.i32.p0(ptr %gep, i32 42, i32 0, i32 0, i1 false)
   ret void
 }
@@ -895,7 +895,7 @@ define amdgpu_kernel void @flat_atomic_dec_ret_i64_offset(ptr %out, ptr %ptr) #0
 ; GFX9-NEXT:    v_mov_b32_e32 v3, s1
 ; GFX9-NEXT:    flat_store_dwordx2 v[2:3], v[0:1]
 ; GFX9-NEXT:    s_endpgm
-  %gep = getelementptr i64, ptr %ptr, i32 4
+  %gep = getelementptr inbounds i64, ptr %ptr, i32 4
   %result = call i64 @llvm.amdgcn.atomic.dec.i64.p0(ptr %gep, i64 42, i32 0, i32 0, i1 false)
   store i64 %result, ptr %out
   ret void
@@ -987,7 +987,7 @@ define amdgpu_kernel void @flat_atomic_dec_noret_i64_offset(ptr %ptr) nounwind {
 ; GFX9-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX9-NEXT:    buffer_wbinvl1_vol
 ; GFX9-NEXT:    s_endpgm
-  %gep = getelementptr i64, ptr %ptr, i32 4
+  %gep = getelementptr inbounds i64, ptr %ptr, i32 4
   %result = call i64 @llvm.amdgcn.atomic.dec.i64.p0(ptr %gep, i64 42, i32 0, i32 0, i1 false)
   ret void
 }
@@ -1054,9 +1054,9 @@ define amdgpu_kernel void @flat_atomic_dec_ret_i64_offset_addr64(ptr %out, ptr %
 ; GFX9-NEXT:    flat_store_dwordx2 v[2:3], v[0:1]
 ; GFX9-NEXT:    s_endpgm
   %id = call i32 @llvm.amdgcn.workitem.id.x()
-  %gep.tid = getelementptr i64, ptr %ptr, i32 %id
-  %out.gep = getelementptr i64, ptr %out, i32 %id
-  %gep = getelementptr i64, ptr %gep.tid, i32 5
+  %gep.tid = getelementptr inbounds i64, ptr %ptr, i32 %id
+  %out.gep = getelementptr inbounds i64, ptr %out, i32 %id
+  %gep = getelementptr inbounds i64, ptr %gep.tid, i32 5
   %result = call i64 @llvm.amdgcn.atomic.dec.i64.p0(ptr %gep, i64 42, i32 0, i32 0, i1 false)
   store i64 %result, ptr %out.gep
   ret void
@@ -1112,8 +1112,8 @@ define amdgpu_kernel void @flat_atomic_dec_noret_i64_offset_addr64(ptr %ptr) #0
 ; GFX9-NEXT:    buffer_wbinvl1_vol
 ; GFX9-NEXT:    s_endpgm
   %id = call i32 @llvm.amdgcn.workitem.id.x()
-  %gep.tid = getelementptr i64, ptr %ptr, i32 %id
-  %gep = getelementptr i64, ptr %gep.tid, i32 5
+  %gep.tid = getelementptr inbounds i64, ptr %ptr, i32 %id
+  %gep = getelementptr inbounds i64, ptr %gep.tid, i32 5
   %result = call i64 @llvm.amdgcn.atomic.dec.i64.p0(ptr %gep, i64 42, i32 0, i32 0, i1 false)
   ret void
 }
diff --git a/llvm/test/CodeGen/AMDGPU/llvm.amdgcn.atomic.inc.ll b/llvm/test/CodeGen/AMDGPU/llvm.amdgcn.atomic.inc.ll
index 2f4ecb8b0de92..cccd2449c3f01 100644
--- a/llvm/test/CodeGen/AMDGPU/llvm.amdgcn.atomic.inc.ll
+++ b/llvm/test/CodeGen/AMDGPU/llvm.amdgcn.atomic.inc.ll
@@ -1145,7 +1145,7 @@ define amdgpu_kernel void @flat_atomic_inc_ret_i32_offset(ptr %out, ptr %ptr) #0
 ; GFX9-NEXT:    v_mov_b32_e32 v1, s1
 ; GFX9-NEXT:    flat_store_dword v[0:1], v2
 ; GFX9-NEXT:    s_endpgm
-  %gep = getelementptr i32, ptr %ptr, i32 4
+  %gep = getelementptr inbounds i32, ptr %ptr, i32 4
   %result = call i32 @llvm.amdgcn.atomic.inc.i32.p0(ptr %gep, i32 42, i32 0, i32 0, i1 false)
   store i32 %result, ptr %out
   ret void
@@ -1231,7 +1231,7 @@ define amdgpu_kernel void @flat_atomic_inc_noret_i32_offset(ptr %ptr) nounwind {
 ; GFX9-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX9-NEXT:    buffer_wbinvl1_vol
 ; GFX9-NEXT:    s_endpgm
-  %gep = getelementptr i32, ptr %ptr, i32 4
+  %gep = getelementptr inbounds i32, ptr %ptr, i32 4
   %result = call i32 @llvm.amdgcn.atomic.inc.i32.p0(ptr %gep, i32 42, i32 0, i32 0, i1 false)
   ret void
 }
@@ -1295,9 +1295,9 @@ define amdgpu_kernel void @flat_atomic_inc_ret_i32_offset_addr64(ptr %out, ptr %
 ; GFX9-NEXT:    flat_store_dword v[0:1], v3
 ; GFX9-NEXT:    s_endpgm
   %id = call i32 @llvm.amdgcn.workitem.id.x()
-  %gep.tid = getelementptr i32, ptr %ptr, i32 %id
-  %out.gep = getelementptr i32, ptr %out, i32 %id
-  %gep = getelementptr i32, ptr %gep.tid, i32 5
+  %gep.tid = getelementptr inbounds i32, ptr %ptr, i32 %id
+  %out.gep = getelementptr inbounds i32, ptr %out, i32 %id
+  %gep = getelementptr inbounds i32, ptr %gep.tid, i32 5
   %result = call i32 @llvm.amdgcn.atomic.inc.i32.p0(ptr %gep, i32 42, i32 0, i32 0, i1 false)
   store i32 %result, ptr %out.gep
   ret void
@@ -1350,8 +1350,8 @@ define amdgpu_kernel void @flat_atomic_inc_noret_i32_offset_addr64(ptr %ptr) #0
 ; GFX9-NEXT:    buffer_wbinvl1_vol
 ; GFX9-NEXT:    s_endpgm
   %id = call i32 @llvm.amdgcn.workitem.id.x()
-  %gep.tid = getelementptr i32, ptr %ptr, i32 %id
-  %gep = getelementptr i32, ptr %gep.tid, i32 5
+  %gep.tid = getelementptr inbounds i32, ptr %ptr, i32 %id
+  %gep = getelementptr inbounds i32, ptr %gep.tid, i32 5
   %result = call i32 @llvm.amdgcn.atomic.inc.i32.p0(ptr %gep, i32 42, i32 0, i32 0, i1 false)
   ret void
 }
@@ -1533,7 +1533,7 @@ define amdgpu_kernel void @flat_atomic_inc_ret_i64_offset(ptr %out, ptr %ptr) #0
 ; GFX9-NEXT:    v_mov_b32_e32 v3, s1
 ; GFX9-NEXT:    flat_store_dwordx2 v[2:3], v[0:1]
 ; GFX9-NEXT:    s_endpgm
-  %gep = getelementptr i64, ptr %ptr, i32 4
+  %gep = getelementptr inbounds i64, ptr %ptr, i32 4
   %result = call i64 @llvm.amdgcn.atomic.inc.i64.p0(ptr %gep, i64 42, i32 0, i32 0, i1 false)
   store i64 %result, ptr %out
   ret void
@@ -1625,7 +1625,7 @@ define amdgpu_kernel void @flat_atomic_inc_noret_i64_offset(ptr %ptr) nounwind {
 ; GFX9-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX9-NEXT:    buffer_wbinvl1_vol
 ; GFX9-NEXT:    s_endpgm
-  %gep = getelementptr i64, ptr %ptr, i32 4
+  %gep = getelementptr inbounds i64, ptr %ptr, i32 4
   %result = call i64 @llvm.amdgcn.atomic.inc.i64.p0(ptr %gep, i64 42, i32 0, i32 0, i1 false)
   ret void
 }
@@ -1692,9 +1692,9 @@ define amdgpu_kernel void @flat_atomic_inc_ret_i64_offset_addr64(ptr %out, ptr %
 ; GFX9-NEXT:    flat_store_dwordx2 v[2:3], v[0:1]
 ; GFX9-NEXT:    s_endpgm
   %id = call i32 @llvm.amdgcn.workitem.id.x()
-  %gep.tid = getelementptr i64, ptr %ptr, i32 %id
-  %out.gep = getelementptr i64, ptr %out, i32 %id
-  %gep = getelementptr i64, ptr %gep.tid, i32 5
+  %gep.tid = getelementptr inbounds i64, ptr %ptr, i32 %id
+  %out.gep = getelementptr inbounds i64, ptr %out, i32 %id
+  %gep = getelementptr inbounds i64, ptr %gep.tid, i32 5
   %result = call i64 @llvm.amdgcn.atomic.inc.i64.p0(ptr %gep, i64 42, i32 0, i32 0, i1 false)
   store i64 %result, ptr %out.gep
   ret void
@@ -1750,8 +1750,8 @@ define amdgpu_kernel void @flat_atomic_inc_noret_i64_offset_addr64(ptr %ptr) #0
 ; GFX9-NEXT:    buffer_wbinvl1_vol
 ; GFX9-NEXT:    s_endpgm
   %id = call i32 @llvm.amdgcn.workitem.id.x()
-  %gep.tid = getelementptr i64, ptr %ptr, i32 %id
-  %gep = getelementptr i64, ptr %gep.tid, i32 5
+  %gep.tid = getelementptr inbounds i64, ptr %ptr, i32 %id
+  %gep = getelementptr inbounds i64, ptr %gep.tid, i32 5
   %result = call i64 @llvm.amdgcn.atomic.inc.i64.p0(ptr %gep, i64 42, i32 0, i32 0, i1 false)
   ret void
 }
diff --git a/llvm/test/CodeGen/AMDGPU/memory-legalizer-flat-agent.ll b/llvm/test/CodeGen/AMDGPU/memory-legalizer-flat-agent.ll
index 9e2906cf85432..bf516f8b91c91 100644
--- a/llvm/test/CodeGen/AMDGPU/memory-legalizer-flat-agent.ll
+++ b/llvm/test/CodeGen/AMDGPU/memory-legalizer-flat-agent.ll
@@ -3486,7 +3486,7 @@ define amdgpu_kernel void @flat_agent_monotonic_monotonic_cmpxchg(
 ; GFX1250-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
-  %gep = getelementptr i32, ptr %out, i32 4
+  %gep = getelementptr inbounds i32, ptr %out, i32 4
   %val = cmpxchg volatile ptr %gep, i32 %old, i32 %in syncscope("agent") monotonic monotonic
   ret void
 }
@@ -3772,7 +3772,7 @@ define amdgpu_kernel void @flat_agent_acquire_monotonic_cmpxchg(
 ; GFX1250-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
-  %gep = getelementptr i32, ptr %out, i32 4
+  %gep = getelementptr inbounds i32, ptr %out, i32 4
   %val = cmpxchg volatile ptr %gep, i32 %old, i32 %in syncscope("agent") acquire monotonic
   ret void
 }
@@ -4052,7 +4052,7 @@ define amdgpu_kernel void @flat_agent_release_monotonic_cmpxchg(
 ; GFX1250-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
-  %gep = getelementptr i32, ptr %out, i32 4
+  %gep = getelementptr inbounds i32, ptr %out, i32 4
   %val = cmpxchg volatile ptr %gep, i32 %old, i32 %in syncscope("agent") release monotonic
   ret void
 }
@@ -4365,7 +4365,7 @@ define amdgpu_kernel void @flat_agent_acq_rel_monotonic_cmpxchg(
 ; GFX1250-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
-  %gep = getelementptr i32, ptr %out, i32 4
+  %gep = getelementptr inbounds i32, ptr %out, i32 4
   %val = cmpxchg volatile ptr %gep, i32 %old, i32 %in syncscope("agent") acq_rel monotonic
   ret void
 }
@@ -4678,7 +4678,7 @@ define amdgpu_kernel void @flat_agent_seq_cst_monotonic_cmpxchg(
 ; GFX1250-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
-  %gep = getelementptr i32, ptr %out, i32 4
+  %gep = getelementptr inbounds i32, ptr %out, i32 4
   %val = cmpxchg volatile ptr %gep, i32 %old, i32 %in syncscope("agent") seq_cst monotonic
   ret void
 }
@@ -4964,7 +4964,7 @@ define amdgpu_kernel void @flat_agent_monotonic_acquire_cmpxchg(
 ; GFX1250-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
-  %gep = getelementptr i32, ptr %out, i32 4
+  %gep = getelementptr inbounds i32, ptr %out, i32 4
   %val = cmpxchg volatile ptr %gep, i32 %old, i32 %in syncscope("agent") monotonic acquire
   ret void
 }
@@ -5250,7 +5250,7 @@ define amdgpu_kernel void @flat_agent_acquire_acquire_cmpxchg(
 ; GFX1250-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
-  %gep = getelementptr i32, ptr %out, i32 4
+  %gep = getelementptr inbounds i32, ptr %out, i32 4
   %val = cmpxchg volatile ptr %gep, i32 %old, i32 %in syncscope("agent") acquire acquire
   ret void
 }
@@ -5563,7 +5563,7 @@ define amdgpu_kernel void @flat_agent_release_acquire_cmpxchg(
 ; GFX1250-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
-  %gep = getelementptr i32, ptr %out, i32 4
+  %gep = getelementptr inbounds i32, ptr %out, i32 4
   %val = cmpxchg volatile ptr %gep, i32 %old, i32 %in syncscope("agent") release acquire
   ret void
 }
@@ -5876,7 +5876,7 @@ define amdgpu_kernel void @flat_agent_acq_rel_acquire_cmpxchg(
 ; GFX1250-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
-  %gep = getelementptr i32, ptr %out, i32 4
+  %gep = getelementptr inbounds i32, ptr %out, i32 4
   %val = cmpxchg volatile ptr %gep, i32 %old, i32 %in syncscope("agent") acq_rel acquire
   ret void
 }
@@ -6189,7 +6189,7 @@ define amdgpu_kernel void @flat_agent_seq_cst_acquire_cmpxchg(
 ; GFX1250-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
-  %gep = getelementptr i32, ptr %out, i32 4
+  %gep = getelementptr inbounds i32, ptr %out, i32 4
   %val = cmpxchg volatile ptr %gep, i32 %old, i32 %in syncscope("agent") seq_cst acquire
   ret void
 }
@@ -6502,7 +6502,7 @@ define amdgpu_kernel void @flat_agent_monotonic_seq_cst_cmpxchg(
 ; GFX1250-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
-  %gep = getelementptr i32, ptr %out, i32 4
+  %gep = getelementptr inbounds i32, ptr %out, i32 4
   %val = cmpxchg volatile ptr %gep, i32 %old, i32 %in syncscope("agent") monotonic seq_cst
   ret void
 }
@@ -6815,7 +6815,7 @@ define amdgpu_kernel void @flat_agent_acquire_seq_cst_cmpxchg(
 ; GFX1250-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
-  %gep = getelementptr i32, ptr %out, i32 4
+  %gep = getelementptr inbounds i32, ptr %out, i32 4
   %val = cmpxchg volatile ptr %gep, i32 %old, i32 %in syncscope("agent") acquire seq_cst
   ret void
 }
@@ -7128,7 +7128,7 @@ define amdgpu_kernel void @flat_agent_release_seq_cst_cmpxchg(
 ; GFX1250-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
-  %gep = getelementptr i32, ptr %out, i32 4
+  %gep = getelementptr inbounds i32, ptr %out, i32 4
   %val = cmpxchg volatile ptr %gep, i32 %old, i32 %in syncscope("agent") release seq_cst
   ret void
 }
@@ -7441,7 +7441,7 @@ define amdgpu_kernel void @flat_agent_acq_rel_seq_cst_cmpxchg(
 ; GFX1250-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
-  %gep = getelementptr i32, ptr %out, i32 4
+  %gep = getelementptr inbounds i32, ptr %out, i32 4
   %val = cmpxchg volatile ptr %gep, i32 %old, i32 %in syncscope("agent") acq_rel seq_cst
   ret void
 }
@@ -7754,7 +7754,7 @@ define amdgpu_kernel void @flat_agent_seq_cst_seq_cst_cmpxchg(
 ; GFX1250-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
-  %gep = getelementptr i32, ptr %out, i32 4
+  %gep = getelementptr inbounds i32, ptr %out, i32 4
   %val = cmpxchg volatile ptr %gep, i32 %old, i32 %in syncscope("agent") seq_cst seq_cst
   ret void
 }
@@ -8053,7 +8053,7 @@ define amdgpu_kernel void @flat_agent_monotonic_monotonic_ret_cmpxchg(
 ; GFX1250-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
-  %gep = getelementptr i32, ptr %out, i32 4
+  %gep = getelementptr inbounds i32, ptr %out, i32 4
   %val = cmpxchg volatile ptr %gep, i32 %old, i32 %in syncscope("agent") monotonic monotonic
   %val0 = extractvalue { i32, i1 } %val, 0
   store i32 %val0, ptr %out, align 4
@@ -8370,7 +8370,7 @@ define amdgpu_kernel void @flat_agent_acquire_monotonic_ret_cmpxchg(
 ; GFX1250-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
-  %gep = getelementptr i32, ptr %out, i32 4
+  %gep = getelementptr inbounds i32, ptr %out, i32 4
   %val = cmpxchg volatile ptr %gep, i32 %old, i32 %in syncscope("agent") acquire monotonic
   %val0 = extractvalue { i32, i1 } %val, 0
   store i32 %val0, ptr %out, align 4
@@ -8698,7 +8698,7 @@ define amdgpu_kernel void @flat_agent_release_monotonic_ret_cmpxchg(
 ; GFX1250-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
-  %gep = getelementptr i32, ptr %out, i32 4
+  %gep = getelementptr inbounds i32, ptr %out, i32 4
   %val = cmpxchg volatile ptr %gep, i32 %old, i32 %in syncscope("agent") release monotonic
   %val0 = extractvalue { i32, i1 } %val, 0
   store i32 %val0, ptr %out, align 4
@@ -9042,7 +9042,7 @@ define amdgpu_kernel void @flat_agent_acq_rel_monotonic_ret_cmpxchg(
 ; GFX1250-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
-  %gep = getelementptr i32, ptr %out, i32 4
+  %gep = getelementptr inbounds i32, ptr %out, i32 4
   %val = cmpxchg volatile ptr %gep, i32 %old, i32 %in syncscope("agent") acq_rel monotonic
   %val0 = extractvalue { i32, i1 } %val, 0
   store i32 %val0, ptr %out, align 4
@@ -9386,7 +9386,7 @@ define amdgpu_kernel void @flat_agent_seq_cst_monotonic_ret_cmpxchg(
 ; GFX1250-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
-  %gep = getelementptr i32, ptr %out, i32 4
+  %gep = getelementptr inbounds i32, ptr %out, i32 4
   %val = cmpxchg volatile ptr %gep, i32 %old, i32 %in syncscope("agent") seq_cst monotonic
   %val0 = extractvalue { i32, i1 } %val, 0
   store i32 %val0, ptr %out, align 4
@@ -9703,7 +9703,7 @@ define amdgpu_kernel void @flat_agent_monotonic_acquire_ret_cmpxchg(
 ; GFX1250-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
-  %gep = getelementptr i32, ptr %out, i32 4
+  %gep = getelementptr inbounds i32, ptr %out, i32 4
   %val = cmpxchg volatile ptr %gep, i32 %old, i32 %in syncscope("agent") monotonic acquire
   %val0 = extractvalue { i32, i1 } %val, 0
   store i32 %val0, ptr %out, align 4
@@ -10020,7 +10020,7 @@ define amdgpu_kernel void @flat_agent_acquire_acquire_ret_cmpxchg(
 ; GFX1250-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
-  %gep = getelementptr i32, ptr %out, i32 4
+  %gep = getelementptr inbounds i32, ptr %out, i32 4
   %val = cmpxchg volatile ptr %gep, i32 %old, i32 %in syncscope("agent") acquire acquire
   %val0 = extractvalue { i32, i1 } %val, 0
   store i32 %val0, ptr %out, align 4
@@ -10364,7 +10364,7 @@ define amdgpu_kernel void @flat_agent_release_acquire_ret_cmpxchg(
 ; GFX1250-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
-  %gep = getelementptr i32, ptr %out, i32 4
+  %gep = getelementptr inbounds i32, ptr %out, i32 4
   %val = cmpxchg volatile ptr %gep, i32 %old, i32 %in syncscope("agent") release acquire
   %val0 = extractvalue { i32, i1 } %val, 0
   store i32 %val0, ptr %out, align 4
@@ -10708,7 +10708,7 @@ define amdgpu_kernel void @flat_agent_acq_rel_acquire_ret_cmpxchg(
 ; GFX1250-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
-  %gep = getelementptr i32, ptr %out, i32 4
+  %gep = getelementptr inbounds i32, ptr %out, i32 4
   %val = cmpxchg volatile ptr %gep, i32 %old, i32 %in syncscope("agent") acq_rel acquire
   %val0 = extractvalue { i32, i1 } %val, 0
   store i32 %val0, ptr %out, align 4
@@ -11052,7 +11052,7 @@ define amdgpu_kernel void @flat_agent_seq_cst_acquire_ret_cmpxchg(
 ; GFX1250-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
-  %gep = getelementptr i32, ptr %out, i32 4
+  %gep = getelementptr inbounds i32, ptr %out, i32 4
   %val = cmpxchg volatile ptr %gep, i32 %old, i32 %in syncscope("agent") seq_cst acquire
   %val0 = extractvalue { i32, i1 } %val, 0
   store i32 %val0, ptr %out, align 4
@@ -11396,7 +11396,7 @@ define amdgpu_kernel void @flat_agent_monotonic_seq_cst_ret_cmpxchg(
 ; GFX1250-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
-  %gep = getelementptr i32, ptr %out, i32 4
+  %gep = getelementptr inbounds i32, ptr %out, i32 4
   %val = cmpxchg volatile ptr %gep, i32 %old, i32 %in syncscope("agent") monotonic seq_cst
   %val0 = extractvalue { i32, i1 } %val, 0
   store i32 %val0, ptr %out, align 4
@@ -11740,7 +11740,7 @@ define amdgpu_kernel void @flat_agent_acquire_seq_cst_ret_cmpxchg(
 ; GFX1250-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
-  %gep = getelementptr i32, ptr %out, i32 4
+  %gep = getelementptr inbounds i32, ptr %out, i32 4
   %val = cmpxchg volatile ptr %gep, i32 %old, i32 %in syncscope("agent") acquire seq_cst
   %val0 = extractvalue { i32, i1 } %val, 0
   store i32 %val0, ptr %out, align 4
@@ -12084,7 +12084,7 @@ define amdgpu_kernel void @flat_agent_release_seq_cst_ret_cmpxchg(
 ; GFX1250-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
-  %gep = getelementptr i32, ptr %out, i32 4
+  %gep = getelementptr inbounds i32, ptr %out, i32 4
   %val = cmpxchg volatile ptr %gep, i32 %old, i32 %in syncscope("agent") release seq_cst
   %val0 = extractvalue { i32, i1 } %val, 0
   store i32 %val0, ptr %out, align 4
@@ -12428,7 +12428,7 @@ define amdgpu_kernel void @flat_agent_acq_rel_seq_cst_ret_cmpxchg(
 ; GFX1250-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
-  %gep = getelementptr i32, ptr %out, i32 4
+  %gep = getelementptr inbounds i32, ptr %out, i32 4
   %val = cmpxchg volatile ptr %gep, i32 %old, i32 %in syncscope("agent") acq_rel seq_cst
   %val0 = extractvalue { i32, i1 } %val, 0
   store i32 %val0, ptr %out, align 4
@@ -12772,7 +12772,7 @@ define amdgpu_kernel void @flat_agent_seq_cst_seq_cst_ret_cmpxchg(
 ; GFX1250-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
-  %gep = getelementptr i32, ptr %out, i32 4
+  %gep = getelementptr inbounds i32, ptr %out, i32 4
   %val = cmpxchg volatile ptr %gep, i32 %old, i32 %in syncscope("agent") seq_cst seq_cst
   %val0 = extractvalue { i32, i1 } %val, 0
   store i32 %val0, ptr %out, align 4
@@ -16295,7 +16295,7 @@ define amdgpu_kernel void @flat_agent_one_as_monotonic_monotonic_cmpxchg(
 ; GFX1250-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
-  %gep = getelementptr i32, ptr %out, i32 4
+  %gep = getelementptr inbounds i32, ptr %out, i32 4
   %val = cmpxchg volatile ptr %gep, i32 %old, i32 %in syncscope("agent-one-as") monotonic monotonic
   ret void
 }
@@ -16577,7 +16577,7 @@ define amdgpu_kernel void @flat_agent_one_as_acquire_monotonic_cmpxchg(
 ; GFX1250-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
-  %gep = getelementptr i32, ptr %out, i32 4
+  %gep = getelementptr inbounds i32, ptr %out, i32 4
   %val = cmpxchg volatile ptr %gep, i32 %old, i32 %in syncscope("agent-one-as") acquire monotonic
   ret void
 }
@@ -16857,7 +16857,7 @@ define amdgpu_kernel void @flat_agent_one_as_release_monotonic_cmpxchg(
 ; GFX1250-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
-  %gep = getelementptr i32, ptr %out, i32 4
+  %gep = getelementptr inbounds i32, ptr %out, i32 4
   %val = cmpxchg volatile ptr %gep, i32 %old, i32 %in syncscope("agent-one-as") release monotonic
   ret void
 }
@@ -17166,7 +17166,7 @@ define amdgpu_kernel void @flat_agent_one_as_acq_rel_monotonic_cmpxchg(
 ; GFX1250-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
-  %gep = getelementptr i32, ptr %out, i32 4
+  %gep = getelementptr inbounds i32, ptr %out, i32 4
   %val = cmpxchg volatile ptr %gep, i32 %old, i32 %in syncscope("agent-one-as") acq_rel monotonic
   ret void
 }
@@ -17475,7 +17475,7 @@ define amdgpu_kernel void @flat_agent_one_as_seq_cst_monotonic_cmpxchg(
 ; GFX1250-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
-  %gep = getelementptr i32, ptr %out, i32 4
+  %gep = getelementptr inbounds i32, ptr %out, i32 4
   %val = cmpxchg volatile ptr %gep, i32 %old, i32 %in syncscope("agent-one-as") seq_cst monotonic
   ret void
 }
@@ -17757,7 +17757,7 @@ define amdgpu_kernel void @flat_agent_one_as_monotonic_acquire_cmpxchg(
 ; GFX1250-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
-  %gep = getelementptr i32, ptr %out, i32 4
+  %gep = getelementptr inbounds i32, ptr %out, i32 4
   %val = cmpxchg volatile ptr %gep, i32 %old, i32 %in syncscope("agent-one-as") monotonic acquire
   ret void
 }
@@ -18039,7 +18039,7 @@ define amdgpu_kernel void @flat_agent_one_as_acquire_acquire_cmpxchg(
 ; GFX1250-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
-  %gep = getelementptr i32, ptr %out, i32 4
+  %gep = getelementptr inbounds i32, ptr %out, i32 4
   %val = cmpxchg volatile ptr %gep, i32 %old, i32 %in syncscope("agent-one-as") acquire acquire
   ret void
 }
@@ -18348,7 +18348,7 @@ define amdgpu_kernel void @flat_agent_one_as_release_acquire_cmpxchg(
 ; GFX1250-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
-  %gep = getelementptr i32, ptr %out, i32 4
+  %gep = getelementptr inbounds i32, ptr %out, i32 4
   %val = cmpxchg volatile ptr %gep, i32 %old, i32 %in syncscope("agent-one-as") release acquire
   ret void
 }
@@ -18657,7 +18657,7 @@ define amdgpu_kernel void @flat_agent_one_as_acq_rel_acquire_cmpxchg(
 ; GFX1250-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
-  %gep = getelementptr i32, ptr %out, i32 4
+  %gep = getelementptr inbounds i32, ptr %out, i32 4
   %val = cmpxchg volatile ptr %gep, i32 %old, i32 %in syncscope("agent-one-as") acq_rel acquire
   ret void
 }
@@ -18966,7 +18966,7 @@ define amdgpu_kernel void @flat_agent_one_as_seq_cst_acquire_cmpxchg(
 ; GFX1250-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
-  %gep = getelementptr i32, ptr %out, i32 4
+  %gep = getelementptr inbounds i32, ptr %out, i32 4
   %val = cmpxchg volatile ptr %gep, i32 %old, i32 %in syncscope("agent-one-as") seq_cst acquire
   ret void
 }
@@ -19275,7 +19275,7 @@ define amdgpu_kernel void @flat_agent_one_as_monotonic_seq_cst_cmpxchg(
 ; GFX1250-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
-  %gep = getelementptr i32, ptr %out, i32 4
+  %gep = getelementptr inbounds i32, ptr %out, i32 4
   %val = cmpxchg volatile ptr %gep, i32 %old, i32 %in syncscope("agent-one-as") monotonic seq_cst
   ret void
 }
@@ -19584,7 +19584,7 @@ define amdgpu_kernel void @flat_agent_one_as_acquire_seq_cst_cmpxchg(
 ; GFX1250-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
-  %gep = getelementptr i32, ptr %out, i32 4
+  %gep = getelementptr inbounds i32, ptr %out, i32 4
   %val = cmpxchg volatile ptr %gep, i32 %old, i32 %in syncscope("agent-one-as") acquire seq_cst
   ret void
 }
@@ -19893,7 +19893,7 @@ define amdgpu_kernel void @flat_agent_one_as_release_seq_cst_cmpxchg(
 ; GFX1250-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
-  %gep = getelementptr i32, ptr %out, i32 4
+  %gep = getelementptr inbounds i32, ptr %out, i32 4
   %val = cmpxchg volatile ptr %gep, i32 %old, i32 %in syncscope("agent-one-as") release seq_cst
   ret void
 }
@@ -20202,7 +20202,7 @@ define amdgpu_kernel void @flat_agent_one_as_acq_rel_seq_cst_cmpxchg(
 ; GFX1250-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
-  %gep = getelementptr i32, ptr %out, i32 4
+  %gep = getelementptr inbounds i32, ptr %out, i32 4
   %val = cmpxchg volatile ptr %gep, i32 %old, i32 %in syncscope("agent-one-as") acq_rel seq_cst
   ret void
 }
@@ -20511,7 +20511,7 @@ define amdgpu_kernel void @flat_agent_one_as_seq_cst_seq_cst_cmpxchg(
 ; GFX1250-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
-  %gep = getelementptr i32, ptr %out, i32 4
+  %gep = getelementptr inbounds i32, ptr %out, i32 4
   %val = cmpxchg volatile ptr %gep, i32 %old, i32 %in syncscope("agent-one-as") seq_cst seq_cst
   ret void
 }
@@ -20810,7 +20810,7 @@ define amdgpu_kernel void @flat_agent_one_as_monotonic_monotonic_ret_cmpxchg(
 ; GFX1250-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
-  %gep = getelementptr i32, ptr %out, i32 4
+  %gep = getelementptr inbounds i32, ptr %out, i32 4
   %val = cmpxchg volatile ptr %gep, i32 %old, i32 %in syncscope("agent-one-as") monotonic monotonic
   %val0 = extractvalue { i32, i1 } %val, 0
   store i32 %val0, ptr %out, align 4
@@ -21138,7 +21138,7 @@ define amdgpu_kernel void @flat_agent_one_as_acquire_monotonic_ret_cmpxchg(
 ; GFX1250-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
-  %gep = getelementptr i32, ptr %out, i32 4
+  %gep = getelementptr inbounds i32, ptr %out, i32 4
   %val = cmpxchg volatile ptr %gep, i32 %old, i32 %in syncscope("agent-one-as") acquire monotonic
   %val0 = extractvalue { i32, i1 } %val, 0
   store i32 %val0, ptr %out, align 4
@@ -21466,7 +21466,7 @@ define amdgpu_kernel void @flat_agent_one_as_release_monotonic_ret_cmpxchg(
 ; GFX1250-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
-  %gep = getelementptr i32, ptr %out, i32 4
+  %gep = getelementptr inbounds i32, ptr %out, i32 4
   %val = cmpxchg volatile ptr %gep, i32 %old, i32 %in syncscope("agent-one-as") release monotonic
   %val0 = extractvalue { i32, i1 } %val, 0
   store i32 %val0, ptr %out, align 4
@@ -21821,7 +21821,7 @@ define amdgpu_kernel void @flat_agent_one_as_acq_rel_monotonic_ret_cmpxchg(
 ; GFX1250-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
-  %gep = getelementptr i32, ptr %out, i32 4
+  %gep = getelementptr inbounds i32, ptr %out, i32 4
   %val = cmpxchg volatile ptr %gep, i32 %old, i32 %in syncscope("agent-one-as") acq_rel monotonic
   %val0 = extractvalue { i32, i1 } %val, 0
   store i32 %val0, ptr %out, align 4
@@ -22176,7 +22176,7 @@ define amdgpu_kernel void @flat_agent_one_as_seq_cst_monotonic_ret_cmpxchg(
 ; GFX1250-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
-  %gep = getelementptr i32, ptr %out, i32 4
+  %gep = getelementptr inbounds i32, ptr %out, i32 4
   %val = cmpxchg volatile ptr %gep, i32 %old, i32 %in syncscope("agent-one-as") seq_cst monotonic
   %val0 = extractvalue { i32, i1 } %val, 0
   store i32 %val0, ptr %out, align 4
@@ -22504,7 +22504,7 @@ define amdgpu_kernel void @flat_agent_one_as_monotonic_acquire_ret_cmpxchg(
 ; GFX1250-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
-  %gep = getelementptr i32, ptr %out, i32 4
+  %gep = getelementptr inbounds i32, ptr %out, i32 4
   %val = cmpxchg volatile ptr %gep, i32 %old, i32 %in syncscope("agent-one-as") monotonic acquire
   %val0 = extractvalue { i32, i1 } %val, 0
   store i32 %val0, ptr %out, align 4
@@ -22832,7 +22832,7 @@ define amdgpu_kernel void @flat_agent_one_as_acquire_acquire_ret_cmpxchg(
 ; GFX1250-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
-  %gep = getelementptr i32, ptr %out, i32 4
+  %gep = getelementptr inbounds i32, ptr %out, i32 4
   %val = cmpxchg volatile ptr %gep, i32 %old, i32 %in syncscope("agent-one-as") acquire acquire
   %val0 = extractvalue { i32, i1 } %val, 0
   store i32 %val0, ptr %out, align 4
@@ -23187,7 +23187,7 @@ define amdgpu_kernel void @flat_agent_one_as_release_acquire_ret_cmpxchg(
 ; GFX1250-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
-  %gep = getelementptr i32, ptr %out, i32 4
+  %gep = getelementptr inbounds i32, ptr %out, i32 4
   %val = cmpxchg volatile ptr %gep, i32 %old, i32 %in syncscope("agent-one-as") release acquire
   %val0 = extractvalue { i32, i1 } %val, 0
   store i32 %val0, ptr %out, align 4
@@ -23542,7 +23542,7 @@ define amdgpu_kernel void @flat_agent_one_as_acq_rel_acquire_ret_cmpxchg(
 ; GFX1250-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
-  %gep = getelementptr i32, ptr %out, i32 4
+  %gep = getelementptr inbounds i32, ptr %out, i32 4
   %val = cmpxchg volatile ptr %gep, i32 %old, i32 %in syncscope("agent-one-as") acq_rel acquire
   %val0 = extractvalue { i32, i1 } %val, 0
   store i32 %val0, ptr %out, align 4
@@ -23897,7 +23897,7 @@ define amdgpu_kernel void @flat_agent_one_as_seq_cst_acquire_ret_cmpxchg(
 ; GFX1250-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
-  %gep = getelementptr i32, ptr %out, i32 4
+  %gep = getelementptr inbounds i32, ptr %out, i32 4
   %val = cmpxchg volatile ptr %gep, i32 %old, i32 %in syncscope("agent-one-as") seq_cst acquire
   %val0 = extractvalue { i32, i1 } %val, 0
   store i32 %val0, ptr %out, align 4
@@ -24252,7 +24252,7 @@ define amdgpu_kernel void @flat_agent_one_as_monotonic_seq_cst_ret_cmpxchg(
 ; GFX1250-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
-  %gep = getelementptr i32, ptr %out, i32 4
+  %gep = getelementptr inbounds i32, ptr %out, i32 4
   %val = cmpxchg volatile ptr %gep, i32 %old, i32 %in syncscope("agent-one-as") monotonic seq_cst
   %val0 = extractvalue { i32, i1 } %val, 0
   store i32 %val0, ptr %out, align 4
@@ -24607,7 +24607,7 @@ define amdgpu_kernel void @flat_agent_one_as_acquire_seq_cst_ret_cmpxchg(
 ; GFX1250-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
-  %gep = getelementptr i32, ptr %out, i32 4
+  %gep = getelementptr inbounds i32, ptr %out, i32 4
   %val = cmpxchg volatile ptr %gep, i32 %old, i32 %in syncscope("agent-one-as") acquire seq_cst
   %val0 = extractvalue { i32, i1 } %val, 0
   store i32 %val0, ptr %out, align 4
@@ -24962,7 +24962,7 @@ define amdgpu_kernel void @flat_agent_one_as_release_seq_cst_ret_cmpxchg(
 ; GFX1250-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
-  %gep = getelementptr i32, ptr %out, i32 4
+  %gep = getelementptr inbounds i32, ptr %out, i32 4
   %val = cmpxchg volatile ptr %gep, i32 %old, i32 %in syncscope("agent-one-as") release seq_cst
   %val0 = extractvalue { i32, i1 } %val, 0
   store i32 %val0, ptr %out, align 4
@@ -25317,7 +25317,7 @@ define amdgpu_kernel void @flat_agent_one_as_acq_rel_seq_cst_ret_cmpxchg(
 ; GFX1250-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
-  %gep = getelementptr i32, ptr %out, i32 4
+  %gep = getelementptr inbounds i32, ptr %out, i32 4
   %val = cmpxchg volatile ptr %gep, i32 %old, i32 %in syncscope("agent-one-as") acq_rel seq_cst
   %val0 = extractvalue { i32, i1 } %val, 0
   store i32 %val0, ptr %out, align 4
@@ -25672,7 +25672,7 @@ define amdgpu_kernel void @flat_agent_one_as_seq_cst_seq_cst_ret_cmpxchg(
 ; GFX1250-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
-  %gep = getelementptr i32, ptr %out, i32 4
+  %gep = getelementptr inbounds i32, ptr %out, i32 4
   %val = cmpxchg volatile ptr %gep, i32 %old, i32 %in syncscope("agent-one-as") seq_cst seq_cst
   %val0 = extractvalue { i32, i1 } %val, 0
   store i32 %val0, ptr %out, align 4
diff --git a/llvm/test/CodeGen/AMDGPU/memory-legalizer-flat-cluster.ll b/llvm/test/CodeGen/AMDGPU/memory-legalizer-flat-cluster.ll
index 27283beb4b877..b2b71c246c97b 100644
--- a/llvm/test/CodeGen/AMDGPU/memory-legalizer-flat-cluster.ll
+++ b/llvm/test/CodeGen/AMDGPU/memory-legalizer-flat-cluster.ll
@@ -3479,7 +3479,7 @@ define amdgpu_kernel void @flat_cluster_monotonic_monotonic_cmpxchg(
 ; GFX1250-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
-  %gep = getelementptr i32, ptr %out, i32 4
+  %gep = getelementptr inbounds i32, ptr %out, i32 4
   %val = cmpxchg volatile ptr %gep, i32 %old, i32 %in syncscope("cluster") monotonic monotonic
   ret void
 }
@@ -3765,7 +3765,7 @@ define amdgpu_kernel void @flat_cluster_acquire_monotonic_cmpxchg(
 ; GFX1250-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
-  %gep = getelementptr i32, ptr %out, i32 4
+  %gep = getelementptr inbounds i32, ptr %out, i32 4
   %val = cmpxchg volatile ptr %gep, i32 %old, i32 %in syncscope("cluster") acquire monotonic
   ret void
 }
@@ -4044,7 +4044,7 @@ define amdgpu_kernel void @flat_cluster_release_monotonic_cmpxchg(
 ; GFX1250-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
-  %gep = getelementptr i32, ptr %out, i32 4
+  %gep = getelementptr inbounds i32, ptr %out, i32 4
   %val = cmpxchg volatile ptr %gep, i32 %old, i32 %in syncscope("cluster") release monotonic
   ret void
 }
@@ -4356,7 +4356,7 @@ define amdgpu_kernel void @flat_cluster_acq_rel_monotonic_cmpxchg(
 ; GFX1250-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
-  %gep = getelementptr i32, ptr %out, i32 4
+  %gep = getelementptr inbounds i32, ptr %out, i32 4
   %val = cmpxchg volatile ptr %gep, i32 %old, i32 %in syncscope("cluster") acq_rel monotonic
   ret void
 }
@@ -4668,7 +4668,7 @@ define amdgpu_kernel void @flat_cluster_seq_cst_monotonic_cmpxchg(
 ; GFX1250-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
-  %gep = getelementptr i32, ptr %out, i32 4
+  %gep = getelementptr inbounds i32, ptr %out, i32 4
   %val = cmpxchg volatile ptr %gep, i32 %old, i32 %in syncscope("cluster") seq_cst monotonic
   ret void
 }
@@ -4954,7 +4954,7 @@ define amdgpu_kernel void @flat_cluster_monotonic_acquire_cmpxchg(
 ; GFX1250-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
-  %gep = getelementptr i32, ptr %out, i32 4
+  %gep = getelementptr inbounds i32, ptr %out, i32 4
   %val = cmpxchg volatile ptr %gep, i32 %old, i32 %in syncscope("cluster") monotonic acquire
   ret void
 }
@@ -5240,7 +5240,7 @@ define amdgpu_kernel void @flat_cluster_acquire_acquire_cmpxchg(
 ; GFX1250-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
-  %gep = getelementptr i32, ptr %out, i32 4
+  %gep = getelementptr inbounds i32, ptr %out, i32 4
   %val = cmpxchg volatile ptr %gep, i32 %old, i32 %in syncscope("cluster") acquire acquire
   ret void
 }
@@ -5552,7 +5552,7 @@ define amdgpu_kernel void @flat_cluster_release_acquire_cmpxchg(
 ; GFX1250-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
-  %gep = getelementptr i32, ptr %out, i32 4
+  %gep = getelementptr inbounds i32, ptr %out, i32 4
   %val = cmpxchg volatile ptr %gep, i32 %old, i32 %in syncscope("cluster") release acquire
   ret void
 }
@@ -5864,7 +5864,7 @@ define amdgpu_kernel void @flat_cluster_acq_rel_acquire_cmpxchg(
 ; GFX1250-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
-  %gep = getelementptr i32, ptr %out, i32 4
+  %gep = getelementptr inbounds i32, ptr %out, i32 4
   %val = cmpxchg volatile ptr %gep, i32 %old, i32 %in syncscope("cluster") acq_rel acquire
   ret void
 }
@@ -6176,7 +6176,7 @@ define amdgpu_kernel void @flat_cluster_seq_cst_acquire_cmpxchg(
 ; GFX1250-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
-  %gep = getelementptr i32, ptr %out, i32 4
+  %gep = getelementptr inbounds i32, ptr %out, i32 4
   %val = cmpxchg volatile ptr %gep, i32 %old, i32 %in syncscope("cluster") seq_cst acquire
   ret void
 }
@@ -6488,7 +6488,7 @@ define amdgpu_kernel void @flat_cluster_monotonic_seq_cst_cmpxchg(
 ; GFX1250-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
-  %gep = getelementptr i32, ptr %out, i32 4
+  %gep = getelementptr inbounds i32, ptr %out, i32 4
   %val = cmpxchg volatile ptr %gep, i32 %old, i32 %in syncscope("cluster") monotonic seq_cst
   ret void
 }
@@ -6800,7 +6800,7 @@ define amdgpu_kernel void @flat_cluster_acquire_seq_cst_cmpxchg(
 ; GFX1250-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
-  %gep = getelementptr i32, ptr %out, i32 4
+  %gep = getelementptr inbounds i32, ptr %out, i32 4
   %val = cmpxchg volatile ptr %gep, i32 %old, i32 %in syncscope("cluster") acquire seq_cst
   ret void
 }
@@ -7112,7 +7112,7 @@ define amdgpu_kernel void @flat_cluster_release_seq_cst_cmpxchg(
 ; GFX1250-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
-  %gep = getelementptr i32, ptr %out, i32 4
+  %gep = getelementptr inbounds i32, ptr %out, i32 4
   %val = cmpxchg volatile ptr %gep, i32 %old, i32 %in syncscope("cluster") release seq_cst
   ret void
 }
@@ -7424,7 +7424,7 @@ define amdgpu_kernel void @flat_cluster_acq_rel_seq_cst_cmpxchg(
 ; GFX1250-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
-  %gep = getelementptr i32, ptr %out, i32 4
+  %gep = getelementptr inbounds i32, ptr %out, i32 4
   %val = cmpxchg volatile ptr %gep, i32 %old, i32 %in syncscope("cluster") acq_rel seq_cst
   ret void
 }
@@ -7736,7 +7736,7 @@ define amdgpu_kernel void @flat_cluster_seq_cst_seq_cst_cmpxchg(
 ; GFX1250-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
-  %gep = getelementptr i32, ptr %out, i32 4
+  %gep = getelementptr inbounds i32, ptr %out, i32 4
   %val = cmpxchg volatile ptr %gep, i32 %old, i32 %in syncscope("cluster") seq_cst seq_cst
   ret void
 }
@@ -8035,7 +8035,7 @@ define amdgpu_kernel void @flat_cluster_monotonic_monotonic_ret_cmpxchg(
 ; GFX1250-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
-  %gep = getelementptr i32, ptr %out, i32 4
+  %gep = getelementptr inbounds i32, ptr %out, i32 4
   %val = cmpxchg volatile ptr %gep, i32 %old, i32 %in syncscope("cluster") monotonic monotonic
   %val0 = extractvalue { i32, i1 } %val, 0
   store i32 %val0, ptr %out, align 4
@@ -8352,7 +8352,7 @@ define amdgpu_kernel void @flat_cluster_acquire_monotonic_ret_cmpxchg(
 ; GFX1250-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
-  %gep = getelementptr i32, ptr %out, i32 4
+  %gep = getelementptr inbounds i32, ptr %out, i32 4
   %val = cmpxchg volatile ptr %gep, i32 %old, i32 %in syncscope("cluster") acquire monotonic
   %val0 = extractvalue { i32, i1 } %val, 0
   store i32 %val0, ptr %out, align 4
@@ -8679,7 +8679,7 @@ define amdgpu_kernel void @flat_cluster_release_monotonic_ret_cmpxchg(
 ; GFX1250-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
-  %gep = getelementptr i32, ptr %out, i32 4
+  %gep = getelementptr inbounds i32, ptr %out, i32 4
   %val = cmpxchg volatile ptr %gep, i32 %old, i32 %in syncscope("cluster") release monotonic
   %val0 = extractvalue { i32, i1 } %val, 0
   store i32 %val0, ptr %out, align 4
@@ -9022,7 +9022,7 @@ define amdgpu_kernel void @flat_cluster_acq_rel_monotonic_ret_cmpxchg(
 ; GFX1250-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
-  %gep = getelementptr i32, ptr %out, i32 4
+  %gep = getelementptr inbounds i32, ptr %out, i32 4
   %val = cmpxchg volatile ptr %gep, i32 %old, i32 %in syncscope("cluster") acq_rel monotonic
   %val0 = extractvalue { i32, i1 } %val, 0
   store i32 %val0, ptr %out, align 4
@@ -9365,7 +9365,7 @@ define amdgpu_kernel void @flat_cluster_seq_cst_monotonic_ret_cmpxchg(
 ; GFX1250-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
-  %gep = getelementptr i32, ptr %out, i32 4
+  %gep = getelementptr inbounds i32, ptr %out, i32 4
   %val = cmpxchg volatile ptr %gep, i32 %old, i32 %in syncscope("cluster") seq_cst monotonic
   %val0 = extractvalue { i32, i1 } %val, 0
   store i32 %val0, ptr %out, align 4
@@ -9682,7 +9682,7 @@ define amdgpu_kernel void @flat_cluster_monotonic_acquire_ret_cmpxchg(
 ; GFX1250-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
-  %gep = getelementptr i32, ptr %out, i32 4
+  %gep = getelementptr inbounds i32, ptr %out, i32 4
   %val = cmpxchg volatile ptr %gep, i32 %old, i32 %in syncscope("cluster") monotonic acquire
   %val0 = extractvalue { i32, i1 } %val, 0
   store i32 %val0, ptr %out, align 4
@@ -9999,7 +9999,7 @@ define amdgpu_kernel void @flat_cluster_acquire_acquire_ret_cmpxchg(
 ; GFX1250-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
-  %gep = getelementptr i32, ptr %out, i32 4
+  %gep = getelementptr inbounds i32, ptr %out, i32 4
   %val = cmpxchg volatile ptr %gep, i32 %old, i32 %in syncscope("cluster") acquire acquire
   %val0 = extractvalue { i32, i1 } %val, 0
   store i32 %val0, ptr %out, align 4
@@ -10342,7 +10342,7 @@ define amdgpu_kernel void @flat_cluster_release_acquire_ret_cmpxchg(
 ; GFX1250-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
-  %gep = getelementptr i32, ptr %out, i32 4
+  %gep = getelementptr inbounds i32, ptr %out, i32 4
   %val = cmpxchg volatile ptr %gep, i32 %old, i32 %in syncscope("cluster") release acquire
   %val0 = extractvalue { i32, i1 } %val, 0
   store i32 %val0, ptr %out, align 4
@@ -10685,7 +10685,7 @@ define amdgpu_kernel void @flat_cluster_acq_rel_acquire_ret_cmpxchg(
 ; GFX1250-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
-  %gep = getelementptr i32, ptr %out, i32 4
+  %gep = getelementptr inbounds i32, ptr %out, i32 4
   %val = cmpxchg volatile ptr %gep, i32 %old, i32 %in syncscope("cluster") acq_rel acquire
   %val0 = extractvalue { i32, i1 } %val, 0
   store i32 %val0, ptr %out, align 4
@@ -11028,7 +11028,7 @@ define amdgpu_kernel void @flat_cluster_seq_cst_acquire_ret_cmpxchg(
 ; GFX1250-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
-  %gep = getelementptr i32, ptr %out, i32 4
+  %gep = getelementptr inbounds i32, ptr %out, i32 4
   %val = cmpxchg volatile ptr %gep, i32 %old, i32 %in syncscope("cluster") seq_cst acquire
   %val0 = extractvalue { i32, i1 } %val, 0
   store i32 %val0, ptr %out, align 4
@@ -11371,7 +11371,7 @@ define amdgpu_kernel void @flat_cluster_monotonic_seq_cst_ret_cmpxchg(
 ; GFX1250-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
-  %gep = getelementptr i32, ptr %out, i32 4
+  %gep = getelementptr inbounds i32, ptr %out, i32 4
   %val = cmpxchg volatile ptr %gep, i32 %old, i32 %in syncscope("cluster") monotonic seq_cst
   %val0 = extractvalue { i32, i1 } %val, 0
   store i32 %val0, ptr %out, align 4
@@ -11714,7 +11714,7 @@ define amdgpu_kernel void @flat_cluster_acquire_seq_cst_ret_cmpxchg(
 ; GFX1250-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
-  %gep = getelementptr i32, ptr %out, i32 4
+  %gep = getelementptr inbounds i32, ptr %out, i32 4
   %val = cmpxchg volatile ptr %gep, i32 %old, i32 %in syncscope("cluster") acquire seq_cst
   %val0 = extractvalue { i32, i1 } %val, 0
   store i32 %val0, ptr %out, align 4
@@ -12057,7 +12057,7 @@ define amdgpu_kernel void @flat_cluster_release_seq_cst_ret_cmpxchg(
 ; GFX1250-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
-  %gep = getelementptr i32, ptr %out, i32 4
+  %gep = getelementptr inbounds i32, ptr %out, i32 4
   %val = cmpxchg volatile ptr %gep, i32 %old, i32 %in syncscope("cluster") release seq_cst
   %val0 = extractvalue { i32, i1 } %val, 0
   store i32 %val0, ptr %out, align 4
@@ -12400,7 +12400,7 @@ define amdgpu_kernel void @flat_cluster_acq_rel_seq_cst_ret_cmpxchg(
 ; GFX1250-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
-  %gep = getelementptr i32, ptr %out, i32 4
+  %gep = getelementptr inbounds i32, ptr %out, i32 4
   %val = cmpxchg volatile ptr %gep, i32 %old, i32 %in syncscope("cluster") acq_rel seq_cst
   %val0 = extractvalue { i32, i1 } %val, 0
   store i32 %val0, ptr %out, align 4
@@ -12743,7 +12743,7 @@ define amdgpu_kernel void @flat_cluster_seq_cst_seq_cst_ret_cmpxchg(
 ; GFX1250-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
-  %gep = getelementptr i32, ptr %out, i32 4
+  %gep = getelementptr inbounds i32, ptr %out, i32 4
   %val = cmpxchg volatile ptr %gep, i32 %old, i32 %in syncscope("cluster") seq_cst seq_cst
   %val0 = extractvalue { i32, i1 } %val, 0
   store i32 %val0, ptr %out, align 4
@@ -16259,7 +16259,7 @@ define amdgpu_kernel void @flat_cluster_one_as_monotonic_monotonic_cmpxchg(
 ; GFX1250-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
-  %gep = getelementptr i32, ptr %out, i32 4
+  %gep = getelementptr inbounds i32, ptr %out, i32 4
   %val = cmpxchg volatile ptr %gep, i32 %old, i32 %in syncscope("cluster-one-as") monotonic monotonic
   ret void
 }
@@ -16541,7 +16541,7 @@ define amdgpu_kernel void @flat_cluster_one_as_acquire_monotonic_cmpxchg(
 ; GFX1250-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
-  %gep = getelementptr i32, ptr %out, i32 4
+  %gep = getelementptr inbounds i32, ptr %out, i32 4
   %val = cmpxchg volatile ptr %gep, i32 %old, i32 %in syncscope("cluster-one-as") acquire monotonic
   ret void
 }
@@ -16820,7 +16820,7 @@ define amdgpu_kernel void @flat_cluster_one_as_release_monotonic_cmpxchg(
 ; GFX1250-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
-  %gep = getelementptr i32, ptr %out, i32 4
+  %gep = getelementptr inbounds i32, ptr %out, i32 4
   %val = cmpxchg volatile ptr %gep, i32 %old, i32 %in syncscope("cluster-one-as") release monotonic
   ret void
 }
@@ -17128,7 +17128,7 @@ define amdgpu_kernel void @flat_cluster_one_as_acq_rel_monotonic_cmpxchg(
 ; GFX1250-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
-  %gep = getelementptr i32, ptr %out, i32 4
+  %gep = getelementptr inbounds i32, ptr %out, i32 4
   %val = cmpxchg volatile ptr %gep, i32 %old, i32 %in syncscope("cluster-one-as") acq_rel monotonic
   ret void
 }
@@ -17436,7 +17436,7 @@ define amdgpu_kernel void @flat_cluster_one_as_seq_cst_monotonic_cmpxchg(
 ; GFX1250-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
-  %gep = getelementptr i32, ptr %out, i32 4
+  %gep = getelementptr inbounds i32, ptr %out, i32 4
   %val = cmpxchg volatile ptr %gep, i32 %old, i32 %in syncscope("cluster-one-as") seq_cst monotonic
   ret void
 }
@@ -17718,7 +17718,7 @@ define amdgpu_kernel void @flat_cluster_one_as_monotonic_acquire_cmpxchg(
 ; GFX1250-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
-  %gep = getelementptr i32, ptr %out, i32 4
+  %gep = getelementptr inbounds i32, ptr %out, i32 4
   %val = cmpxchg volatile ptr %gep, i32 %old, i32 %in syncscope("cluster-one-as") monotonic acquire
   ret void
 }
@@ -18000,7 +18000,7 @@ define amdgpu_kernel void @flat_cluster_one_as_acquire_acquire_cmpxchg(
 ; GFX1250-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
-  %gep = getelementptr i32, ptr %out, i32 4
+  %gep = getelementptr inbounds i32, ptr %out, i32 4
   %val = cmpxchg volatile ptr %gep, i32 %old, i32 %in syncscope("cluster-one-as") acquire acquire
   ret void
 }
@@ -18308,7 +18308,7 @@ define amdgpu_kernel void @flat_cluster_one_as_release_acquire_cmpxchg(
 ; GFX1250-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
-  %gep = getelementptr i32, ptr %out, i32 4
+  %gep = getelementptr inbounds i32, ptr %out, i32 4
   %val = cmpxchg volatile ptr %gep, i32 %old, i32 %in syncscope("cluster-one-as") release acquire
   ret void
 }
@@ -18616,7 +18616,7 @@ define amdgpu_kernel void @flat_cluster_one_as_acq_rel_acquire_cmpxchg(
 ; GFX1250-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
-  %gep = getelementptr i32, ptr %out, i32 4
+  %gep = getelementptr inbounds i32, ptr %out, i32 4
   %val = cmpxchg volatile ptr %gep, i32 %old, i32 %in syncscope("cluster-one-as") acq_rel acquire
   ret void
 }
@@ -18924,7 +18924,7 @@ define amdgpu_kernel void @flat_cluster_one_as_seq_cst_acquire_cmpxchg(
 ; GFX1250-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
-  %gep = getelementptr i32, ptr %out, i32 4
+  %gep = getelementptr inbounds i32, ptr %out, i32 4
   %val = cmpxchg volatile ptr %gep, i32 %old, i32 %in syncscope("cluster-one-as") seq_cst acquire
   ret void
 }
@@ -19232,7 +19232,7 @@ define amdgpu_kernel void @flat_cluster_one_as_monotonic_seq_cst_cmpxchg(
 ; GFX1250-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
-  %gep = getelementptr i32, ptr %out, i32 4
+  %gep = getelementptr inbounds i32, ptr %out, i32 4
   %val = cmpxchg volatile ptr %gep, i32 %old, i32 %in syncscope("cluster-one-as") monotonic seq_cst
   ret void
 }
@@ -19540,7 +19540,7 @@ define amdgpu_kernel void @flat_cluster_one_as_acquire_seq_cst_cmpxchg(
 ; GFX1250-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
-  %gep = getelementptr i32, ptr %out, i32 4
+  %gep = getelementptr inbounds i32, ptr %out, i32 4
   %val = cmpxchg volatile ptr %gep, i32 %old, i32 %in syncscope("cluster-one-as") acquire seq_cst
   ret void
 }
@@ -19848,7 +19848,7 @@ define amdgpu_kernel void @flat_cluster_one_as_release_seq_cst_cmpxchg(
 ; GFX1250-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
-  %gep = getelementptr i32, ptr %out, i32 4
+  %gep = getelementptr inbounds i32, ptr %out, i32 4
   %val = cmpxchg volatile ptr %gep, i32 %old, i32 %in syncscope("cluster-one-as") release seq_cst
   ret void
 }
@@ -20156,7 +20156,7 @@ define amdgpu_kernel void @flat_cluster_one_as_acq_rel_seq_cst_cmpxchg(
 ; GFX1250-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
-  %gep = getelementptr i32, ptr %out, i32 4
+  %gep = getelementptr inbounds i32, ptr %out, i32 4
   %val = cmpxchg volatile ptr %gep, i32 %old, i32 %in syncscope("cluster-one-as") acq_rel seq_cst
   ret void
 }
@@ -20464,7 +20464,7 @@ define amdgpu_kernel void @flat_cluster_one_as_seq_cst_seq_cst_cmpxchg(
 ; GFX1250-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
-  %gep = getelementptr i32, ptr %out, i32 4
+  %gep = getelementptr inbounds i32, ptr %out, i32 4
   %val = cmpxchg volatile ptr %gep, i32 %old, i32 %in syncscope("cluster-one-as") seq_cst seq_cst
   ret void
 }
@@ -20763,7 +20763,7 @@ define amdgpu_kernel void @flat_cluster_one_as_monotonic_monotonic_ret_cmpxchg(
 ; GFX1250-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
-  %gep = getelementptr i32, ptr %out, i32 4
+  %gep = getelementptr inbounds i32, ptr %out, i32 4
   %val = cmpxchg volatile ptr %gep, i32 %old, i32 %in syncscope("cluster-one-as") monotonic monotonic
   %val0 = extractvalue { i32, i1 } %val, 0
   store i32 %val0, ptr %out, align 4
@@ -21091,7 +21091,7 @@ define amdgpu_kernel void @flat_cluster_one_as_acquire_monotonic_ret_cmpxchg(
 ; GFX1250-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
-  %gep = getelementptr i32, ptr %out, i32 4
+  %gep = getelementptr inbounds i32, ptr %out, i32 4
   %val = cmpxchg volatile ptr %gep, i32 %old, i32 %in syncscope("cluster-one-as") acquire monotonic
   %val0 = extractvalue { i32, i1 } %val, 0
   store i32 %val0, ptr %out, align 4
@@ -21418,7 +21418,7 @@ define amdgpu_kernel void @flat_cluster_one_as_release_monotonic_ret_cmpxchg(
 ; GFX1250-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
-  %gep = getelementptr i32, ptr %out, i32 4
+  %gep = getelementptr inbounds i32, ptr %out, i32 4
   %val = cmpxchg volatile ptr %gep, i32 %old, i32 %in syncscope("cluster-one-as") release monotonic
   %val0 = extractvalue { i32, i1 } %val, 0
   store i32 %val0, ptr %out, align 4
@@ -21772,7 +21772,7 @@ define amdgpu_kernel void @flat_cluster_one_as_acq_rel_monotonic_ret_cmpxchg(
 ; GFX1250-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
-  %gep = getelementptr i32, ptr %out, i32 4
+  %gep = getelementptr inbounds i32, ptr %out, i32 4
   %val = cmpxchg volatile ptr %gep, i32 %old, i32 %in syncscope("cluster-one-as") acq_rel monotonic
   %val0 = extractvalue { i32, i1 } %val, 0
   store i32 %val0, ptr %out, align 4
@@ -22126,7 +22126,7 @@ define amdgpu_kernel void @flat_cluster_one_as_seq_cst_monotonic_ret_cmpxchg(
 ; GFX1250-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
-  %gep = getelementptr i32, ptr %out, i32 4
+  %gep = getelementptr inbounds i32, ptr %out, i32 4
   %val = cmpxchg volatile ptr %gep, i32 %old, i32 %in syncscope("cluster-one-as") seq_cst monotonic
   %val0 = extractvalue { i32, i1 } %val, 0
   store i32 %val0, ptr %out, align 4
@@ -22454,7 +22454,7 @@ define amdgpu_kernel void @flat_cluster_one_as_monotonic_acquire_ret_cmpxchg(
 ; GFX1250-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
-  %gep = getelementptr i32, ptr %out, i32 4
+  %gep = getelementptr inbounds i32, ptr %out, i32 4
   %val = cmpxchg volatile ptr %gep, i32 %old, i32 %in syncscope("cluster-one-as") monotonic acquire
   %val0 = extractvalue { i32, i1 } %val, 0
   store i32 %val0, ptr %out, align 4
@@ -22782,7 +22782,7 @@ define amdgpu_kernel void @flat_cluster_one_as_acquire_acquire_ret_cmpxchg(
 ; GFX1250-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
-  %gep = getelementptr i32, ptr %out, i32 4
+  %gep = getelementptr inbounds i32, ptr %out, i32 4
   %val = cmpxchg volatile ptr %gep, i32 %old, i32 %in syncscope("cluster-one-as") acquire acquire
   %val0 = extractvalue { i32, i1 } %val, 0
   store i32 %val0, ptr %out, align 4
@@ -23136,7 +23136,7 @@ define amdgpu_kernel void @flat_cluster_one_as_release_acquire_ret_cmpxchg(
 ; GFX1250-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
-  %gep = getelementptr i32, ptr %out, i32 4
+  %gep = getelementptr inbounds i32, ptr %out, i32 4
   %val = cmpxchg volatile ptr %gep, i32 %old, i32 %in syncscope("cluster-one-as") release acquire
   %val0 = extractvalue { i32, i1 } %val, 0
   store i32 %val0, ptr %out, align 4
@@ -23490,7 +23490,7 @@ define amdgpu_kernel void @flat_cluster_one_as_acq_rel_acquire_ret_cmpxchg(
 ; GFX1250-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
-  %gep = getelementptr i32, ptr %out, i32 4
+  %gep = getelementptr inbounds i32, ptr %out, i32 4
   %val = cmpxchg volatile ptr %gep, i32 %old, i32 %in syncscope("cluster-one-as") acq_rel acquire
   %val0 = extractvalue { i32, i1 } %val, 0
   store i32 %val0, ptr %out, align 4
@@ -23844,7 +23844,7 @@ define amdgpu_kernel void @flat_cluster_one_as_seq_cst_acquire_ret_cmpxchg(
 ; GFX1250-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
-  %gep = getelementptr i32, ptr %out, i32 4
+  %gep = getelementptr inbounds i32, ptr %out, i32 4
   %val = cmpxchg volatile ptr %gep, i32 %old, i32 %in syncscope("cluster-one-as") seq_cst acquire
   %val0 = extractvalue { i32, i1 } %val, 0
   store i32 %val0, ptr %out, align 4
@@ -24198,7 +24198,7 @@ define amdgpu_kernel void @flat_cluster_one_as_monotonic_seq_cst_ret_cmpxchg(
 ; GFX1250-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
-  %gep = getelementptr i32, ptr %out, i32 4
+  %gep = getelementptr inbounds i32, ptr %out, i32 4
   %val = cmpxchg volatile ptr %gep, i32 %old, i32 %in syncscope("cluster-one-as") monotonic seq_cst
   %val0 = extractvalue { i32, i1 } %val, 0
   store i32 %val0, ptr %out, align 4
@@ -24552,7 +24552,7 @@ define amdgpu_kernel void @flat_cluster_one_as_acquire_seq_cst_ret_cmpxchg(
 ; GFX1250-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
-  %gep = getelementptr i32, ptr %out, i32 4
+  %gep = getelementptr inbounds i32, ptr %out, i32 4
   %val = cmpxchg volatile ptr %gep, i32 %old, i32 %in syncscope("cluster-one-as") acquire seq_cst
   %val0 = extractvalue { i32, i1 } %val, 0
   store i32 %val0, ptr %out, align 4
@@ -24906,7 +24906,7 @@ define amdgpu_kernel void @flat_cluster_one_as_release_seq_cst_ret_cmpxchg(
 ; GFX1250-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
-  %gep = getelementptr i32, ptr %out, i32 4
+  %gep = getelementptr inbounds i32, ptr %out, i32 4
   %val = cmpxchg volatile ptr %gep, i32 %old, i32 %in syncscope("cluster-one-as") release seq_cst
   %val0 = extractvalue { i32, i1 } %val, 0
   store i32 %val0, ptr %out, align 4
@@ -25260,7 +25260,7 @@ define amdgpu_kernel void @flat_cluster_one_as_acq_rel_seq_cst_ret_cmpxchg(
 ; GFX1250-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
-  %gep = getelementptr i32, ptr %out, i32 4
+  %gep = getelementptr inbounds i32, ptr %out, i32 4
   %val = cmpxchg volatile ptr %gep, i32 %old, i32 %in syncscope("cluster-one-as") acq_rel seq_cst
   %val0 = extractvalue { i32, i1 } %val, 0
   store i32 %val0, ptr %out, align 4
@@ -25614,7 +25614,7 @@ define amdgpu_kernel void @flat_cluster_one_as_seq_cst_seq_cst_ret_cmpxchg(
 ; GFX1250-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
-  %gep = getelementptr i32, ptr %out, i32 4
+  %gep = getelementptr inbounds i32, ptr %out, i32 4
   %val = cmpxchg volatile ptr %gep, i32 %old, i32 %in syncscope("cluster-one-as") seq_cst seq_cst
   %val0 = extractvalue { i32, i1 } %val, 0
   store i32 %val0, ptr %out, align 4
diff --git a/llvm/test/CodeGen/AMDGPU/memory-legalizer-flat-singlethread.ll b/llvm/test/CodeGen/AMDGPU/memory-legalizer-flat-singlethread.ll
index a05f4c718c351..7d357922ac307 100644
--- a/llvm/test/CodeGen/AMDGPU/memory-legalizer-flat-singlethread.ll
+++ b/llvm/test/CodeGen/AMDGPU/memory-legalizer-flat-singlethread.ll
@@ -3094,7 +3094,7 @@ define amdgpu_kernel void @flat_singlethread_monotonic_monotonic_cmpxchg(
 ; GFX1250-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
-  %gep = getelementptr i32, ptr %out, i32 4
+  %gep = getelementptr inbounds i32, ptr %out, i32 4
   %val = cmpxchg volatile ptr %gep, i32 %old, i32 %in syncscope("singlethread") monotonic monotonic
   ret void
 }
@@ -3347,7 +3347,7 @@ define amdgpu_kernel void @flat_singlethread_acquire_monotonic_cmpxchg(
 ; GFX1250-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
-  %gep = getelementptr i32, ptr %out, i32 4
+  %gep = getelementptr inbounds i32, ptr %out, i32 4
   %val = cmpxchg volatile ptr %gep, i32 %old, i32 %in syncscope("singlethread") acquire monotonic
   ret void
 }
@@ -3600,7 +3600,7 @@ define amdgpu_kernel void @flat_singlethread_release_monotonic_cmpxchg(
 ; GFX1250-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
-  %gep = getelementptr i32, ptr %out, i32 4
+  %gep = getelementptr inbounds i32, ptr %out, i32 4
   %val = cmpxchg volatile ptr %gep, i32 %old, i32 %in syncscope("singlethread") release monotonic
   ret void
 }
@@ -3853,7 +3853,7 @@ define amdgpu_kernel void @flat_singlethread_acq_rel_monotonic_cmpxchg(
 ; GFX1250-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
-  %gep = getelementptr i32, ptr %out, i32 4
+  %gep = getelementptr inbounds i32, ptr %out, i32 4
   %val = cmpxchg volatile ptr %gep, i32 %old, i32 %in syncscope("singlethread") acq_rel monotonic
   ret void
 }
@@ -4106,7 +4106,7 @@ define amdgpu_kernel void @flat_singlethread_seq_cst_monotonic_cmpxchg(
 ; GFX1250-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
-  %gep = getelementptr i32, ptr %out, i32 4
+  %gep = getelementptr inbounds i32, ptr %out, i32 4
   %val = cmpxchg volatile ptr %gep, i32 %old, i32 %in syncscope("singlethread") seq_cst monotonic
   ret void
 }
@@ -4359,7 +4359,7 @@ define amdgpu_kernel void @flat_singlethread_monotonic_acquire_cmpxchg(
 ; GFX1250-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
-  %gep = getelementptr i32, ptr %out, i32 4
+  %gep = getelementptr inbounds i32, ptr %out, i32 4
   %val = cmpxchg volatile ptr %gep, i32 %old, i32 %in syncscope("singlethread") monotonic acquire
   ret void
 }
@@ -4612,7 +4612,7 @@ define amdgpu_kernel void @flat_singlethread_acquire_acquire_cmpxchg(
 ; GFX1250-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
-  %gep = getelementptr i32, ptr %out, i32 4
+  %gep = getelementptr inbounds i32, ptr %out, i32 4
   %val = cmpxchg volatile ptr %gep, i32 %old, i32 %in syncscope("singlethread") acquire acquire
   ret void
 }
@@ -4865,7 +4865,7 @@ define amdgpu_kernel void @flat_singlethread_release_acquire_cmpxchg(
 ; GFX1250-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
-  %gep = getelementptr i32, ptr %out, i32 4
+  %gep = getelementptr inbounds i32, ptr %out, i32 4
   %val = cmpxchg volatile ptr %gep, i32 %old, i32 %in syncscope("singlethread") release acquire
   ret void
 }
@@ -5118,7 +5118,7 @@ define amdgpu_kernel void @flat_singlethread_acq_rel_acquire_cmpxchg(
 ; GFX1250-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
-  %gep = getelementptr i32, ptr %out, i32 4
+  %gep = getelementptr inbounds i32, ptr %out, i32 4
   %val = cmpxchg volatile ptr %gep, i32 %old, i32 %in syncscope("singlethread") acq_rel acquire
   ret void
 }
@@ -5371,7 +5371,7 @@ define amdgpu_kernel void @flat_singlethread_seq_cst_acquire_cmpxchg(
 ; GFX1250-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
-  %gep = getelementptr i32, ptr %out, i32 4
+  %gep = getelementptr inbounds i32, ptr %out, i32 4
   %val = cmpxchg volatile ptr %gep, i32 %old, i32 %in syncscope("singlethread") seq_cst acquire
   ret void
 }
@@ -5624,7 +5624,7 @@ define amdgpu_kernel void @flat_singlethread_monotonic_seq_cst_cmpxchg(
 ; GFX1250-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
-  %gep = getelementptr i32, ptr %out, i32 4
+  %gep = getelementptr inbounds i32, ptr %out, i32 4
   %val = cmpxchg volatile ptr %gep, i32 %old, i32 %in syncscope("singlethread") monotonic seq_cst
   ret void
 }
@@ -5877,7 +5877,7 @@ define amdgpu_kernel void @flat_singlethread_acquire_seq_cst_cmpxchg(
 ; GFX1250-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
-  %gep = getelementptr i32, ptr %out, i32 4
+  %gep = getelementptr inbounds i32, ptr %out, i32 4
   %val = cmpxchg volatile ptr %gep, i32 %old, i32 %in syncscope("singlethread") acquire seq_cst
   ret void
 }
@@ -6130,7 +6130,7 @@ define amdgpu_kernel void @flat_singlethread_release_seq_cst_cmpxchg(
 ; GFX1250-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
-  %gep = getelementptr i32, ptr %out, i32 4
+  %gep = getelementptr inbounds i32, ptr %out, i32 4
   %val = cmpxchg volatile ptr %gep, i32 %old, i32 %in syncscope("singlethread") release seq_cst
   ret void
 }
@@ -6383,7 +6383,7 @@ define amdgpu_kernel void @flat_singlethread_acq_rel_seq_cst_cmpxchg(
 ; GFX1250-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
-  %gep = getelementptr i32, ptr %out, i32 4
+  %gep = getelementptr inbounds i32, ptr %out, i32 4
   %val = cmpxchg volatile ptr %gep, i32 %old, i32 %in syncscope("singlethread") acq_rel seq_cst
   ret void
 }
@@ -6636,7 +6636,7 @@ define amdgpu_kernel void @flat_singlethread_seq_cst_seq_cst_cmpxchg(
 ; GFX1250-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
-  %gep = getelementptr i32, ptr %out, i32 4
+  %gep = getelementptr inbounds i32, ptr %out, i32 4
   %val = cmpxchg volatile ptr %gep, i32 %old, i32 %in syncscope("singlethread") seq_cst seq_cst
   ret void
 }
@@ -6935,7 +6935,7 @@ define amdgpu_kernel void @flat_singlethread_monotonic_monotonic_ret_cmpxchg(
 ; GFX1250-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
-  %gep = getelementptr i32, ptr %out, i32 4
+  %gep = getelementptr inbounds i32, ptr %out, i32 4
   %val = cmpxchg volatile ptr %gep, i32 %old, i32 %in syncscope("singlethread") monotonic monotonic
   %val0 = extractvalue { i32, i1 } %val, 0
   store i32 %val0, ptr %out, align 4
@@ -7236,7 +7236,7 @@ define amdgpu_kernel void @flat_singlethread_acquire_monotonic_ret_cmpxchg(
 ; GFX1250-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
-  %gep = getelementptr i32, ptr %out, i32 4
+  %gep = getelementptr inbounds i32, ptr %out, i32 4
   %val = cmpxchg volatile ptr %gep, i32 %old, i32 %in syncscope("singlethread") acquire monotonic
   %val0 = extractvalue { i32, i1 } %val, 0
   store i32 %val0, ptr %out, align 4
@@ -7537,7 +7537,7 @@ define amdgpu_kernel void @flat_singlethread_release_monotonic_ret_cmpxchg(
 ; GFX1250-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
-  %gep = getelementptr i32, ptr %out, i32 4
+  %gep = getelementptr inbounds i32, ptr %out, i32 4
   %val = cmpxchg volatile ptr %gep, i32 %old, i32 %in syncscope("singlethread") release monotonic
   %val0 = extractvalue { i32, i1 } %val, 0
   store i32 %val0, ptr %out, align 4
@@ -7838,7 +7838,7 @@ define amdgpu_kernel void @flat_singlethread_acq_rel_monotonic_ret_cmpxchg(
 ; GFX1250-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
-  %gep = getelementptr i32, ptr %out, i32 4
+  %gep = getelementptr inbounds i32, ptr %out, i32 4
   %val = cmpxchg volatile ptr %gep, i32 %old, i32 %in syncscope("singlethread") acq_rel monotonic
   %val0 = extractvalue { i32, i1 } %val, 0
   store i32 %val0, ptr %out, align 4
@@ -8139,7 +8139,7 @@ define amdgpu_kernel void @flat_singlethread_seq_cst_monotonic_ret_cmpxchg(
 ; GFX1250-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
-  %gep = getelementptr i32, ptr %out, i32 4
+  %gep = getelementptr inbounds i32, ptr %out, i32 4
   %val = cmpxchg volatile ptr %gep, i32 %old, i32 %in syncscope("singlethread") seq_cst monotonic
   %val0 = extractvalue { i32, i1 } %val, 0
   store i32 %val0, ptr %out, align 4
@@ -8440,7 +8440,7 @@ define amdgpu_kernel void @flat_singlethread_monotonic_acquire_ret_cmpxchg(
 ; GFX1250-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
-  %gep = getelementptr i32, ptr %out, i32 4
+  %gep = getelementptr inbounds i32, ptr %out, i32 4
   %val = cmpxchg volatile ptr %gep, i32 %old, i32 %in syncscope("singlethread") monotonic acquire
   %val0 = extractvalue { i32, i1 } %val, 0
   store i32 %val0, ptr %out, align 4
@@ -8741,7 +8741,7 @@ define amdgpu_kernel void @flat_singlethread_acquire_acquire_ret_cmpxchg(
 ; GFX1250-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
-  %gep = getelementptr i32, ptr %out, i32 4
+  %gep = getelementptr inbounds i32, ptr %out, i32 4
   %val = cmpxchg volatile ptr %gep, i32 %old, i32 %in syncscope("singlethread") acquire acquire
   %val0 = extractvalue { i32, i1 } %val, 0
   store i32 %val0, ptr %out, align 4
@@ -9042,7 +9042,7 @@ define amdgpu_kernel void @flat_singlethread_release_acquire_ret_cmpxchg(
 ; GFX1250-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
-  %gep = getelementptr i32, ptr %out, i32 4
+  %gep = getelementptr inbounds i32, ptr %out, i32 4
   %val = cmpxchg volatile ptr %gep, i32 %old, i32 %in syncscope("singlethread") release acquire
   %val0 = extractvalue { i32, i1 } %val, 0
   store i32 %val0, ptr %out, align 4
@@ -9343,7 +9343,7 @@ define amdgpu_kernel void @flat_singlethread_acq_rel_acquire_ret_cmpxchg(
 ; GFX1250-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
-  %gep = getelementptr i32, ptr %out, i32 4
+  %gep = getelementptr inbounds i32, ptr %out, i32 4
   %val = cmpxchg volatile ptr %gep, i32 %old, i32 %in syncscope("singlethread") acq_rel acquire
   %val0 = extractvalue { i32, i1 } %val, 0
   store i32 %val0, ptr %out, align 4
@@ -9644,7 +9644,7 @@ define amdgpu_kernel void @flat_singlethread_seq_cst_acquire_ret_cmpxchg(
 ; GFX1250-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
-  %gep = getelementptr i32, ptr %out, i32 4
+  %gep = getelementptr inbounds i32, ptr %out, i32 4
   %val = cmpxchg volatile ptr %gep, i32 %old, i32 %in syncscope("singlethread") seq_cst acquire
   %val0 = extractvalue { i32, i1 } %val, 0
   store i32 %val0, ptr %out, align 4
@@ -9945,7 +9945,7 @@ define amdgpu_kernel void @flat_singlethread_monotonic_seq_cst_ret_cmpxchg(
 ; GFX1250-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
-  %gep = getelementptr i32, ptr %out, i32 4
+  %gep = getelementptr inbounds i32, ptr %out, i32 4
   %val = cmpxchg volatile ptr %gep, i32 %old, i32 %in syncscope("singlethread") monotonic seq_cst
   %val0 = extractvalue { i32, i1 } %val, 0
   store i32 %val0, ptr %out, align 4
@@ -10246,7 +10246,7 @@ define amdgpu_kernel void @flat_singlethread_acquire_seq_cst_ret_cmpxchg(
 ; GFX1250-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
-  %gep = getelementptr i32, ptr %out, i32 4
+  %gep = getelementptr inbounds i32, ptr %out, i32 4
   %val = cmpxchg volatile ptr %gep, i32 %old, i32 %in syncscope("singlethread") acquire seq_cst
   %val0 = extractvalue { i32, i1 } %val, 0
   store i32 %val0, ptr %out, align 4
@@ -10547,7 +10547,7 @@ define amdgpu_kernel void @flat_singlethread_release_seq_cst_ret_cmpxchg(
 ; GFX1250-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
-  %gep = getelementptr i32, ptr %out, i32 4
+  %gep = getelementptr inbounds i32, ptr %out, i32 4
   %val = cmpxchg volatile ptr %gep, i32 %old, i32 %in syncscope("singlethread") release seq_cst
   %val0 = extractvalue { i32, i1 } %val, 0
   store i32 %val0, ptr %out, align 4
@@ -10848,7 +10848,7 @@ define amdgpu_kernel void @flat_singlethread_acq_rel_seq_cst_ret_cmpxchg(
 ; GFX1250-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
-  %gep = getelementptr i32, ptr %out, i32 4
+  %gep = getelementptr inbounds i32, ptr %out, i32 4
   %val = cmpxchg volatile ptr %gep, i32 %old, i32 %in syncscope("singlethread") acq_rel seq_cst
   %val0 = extractvalue { i32, i1 } %val, 0
   store i32 %val0, ptr %out, align 4
@@ -11149,7 +11149,7 @@ define amdgpu_kernel void @flat_singlethread_seq_cst_seq_cst_ret_cmpxchg(
 ; GFX1250-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
-  %gep = getelementptr i32, ptr %out, i32 4
+  %gep = getelementptr inbounds i32, ptr %out, i32 4
   %val = cmpxchg volatile ptr %gep, i32 %old, i32 %in syncscope("singlethread") seq_cst seq_cst
   %val0 = extractvalue { i32, i1 } %val, 0
   store i32 %val0, ptr %out, align 4
@@ -14237,7 +14237,7 @@ define amdgpu_kernel void @flat_singlethread_one_as_monotonic_monotonic_cmpxchg(
 ; GFX1250-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
-  %gep = getelementptr i32, ptr %out, i32 4
+  %gep = getelementptr inbounds i32, ptr %out, i32 4
   %val = cmpxchg volatile ptr %gep, i32 %old, i32 %in syncscope("singlethread-one-as") monotonic monotonic
   ret void
 }
@@ -14490,7 +14490,7 @@ define amdgpu_kernel void @flat_singlethread_one_as_acquire_monotonic_cmpxchg(
 ; GFX1250-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
-  %gep = getelementptr i32, ptr %out, i32 4
+  %gep = getelementptr inbounds i32, ptr %out, i32 4
   %val = cmpxchg volatile ptr %gep, i32 %old, i32 %in syncscope("singlethread-one-as") acquire monotonic
   ret void
 }
@@ -14743,7 +14743,7 @@ define amdgpu_kernel void @flat_singlethread_one_as_release_monotonic_cmpxchg(
 ; GFX1250-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
-  %gep = getelementptr i32, ptr %out, i32 4
+  %gep = getelementptr inbounds i32, ptr %out, i32 4
   %val = cmpxchg volatile ptr %gep, i32 %old, i32 %in syncscope("singlethread-one-as") release monotonic
   ret void
 }
@@ -14996,7 +14996,7 @@ define amdgpu_kernel void @flat_singlethread_one_as_acq_rel_monotonic_cmpxchg(
 ; GFX1250-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
-  %gep = getelementptr i32, ptr %out, i32 4
+  %gep = getelementptr inbounds i32, ptr %out, i32 4
   %val = cmpxchg volatile ptr %gep, i32 %old, i32 %in syncscope("singlethread-one-as") acq_rel monotonic
   ret void
 }
@@ -15249,7 +15249,7 @@ define amdgpu_kernel void @flat_singlethread_one_as_seq_cst_monotonic_cmpxchg(
 ; GFX1250-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
-  %gep = getelementptr i32, ptr %out, i32 4
+  %gep = getelementptr inbounds i32, ptr %out, i32 4
   %val = cmpxchg volatile ptr %gep, i32 %old, i32 %in syncscope("singlethread-one-as") seq_cst monotonic
   ret void
 }
@@ -15502,7 +15502,7 @@ define amdgpu_kernel void @flat_singlethread_one_as_monotonic_acquire_cmpxchg(
 ; GFX1250-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
-  %gep = getelementptr i32, ptr %out, i32 4
+  %gep = getelementptr inbounds i32, ptr %out, i32 4
   %val = cmpxchg volatile ptr %gep, i32 %old, i32 %in syncscope("singlethread-one-as") monotonic acquire
   ret void
 }
@@ -15755,7 +15755,7 @@ define amdgpu_kernel void @flat_singlethread_one_as_acquire_acquire_cmpxchg(
 ; GFX1250-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
-  %gep = getelementptr i32, ptr %out, i32 4
+  %gep = getelementptr inbounds i32, ptr %out, i32 4
   %val = cmpxchg volatile ptr %gep, i32 %old, i32 %in syncscope("singlethread-one-as") acquire acquire
   ret void
 }
@@ -16008,7 +16008,7 @@ define amdgpu_kernel void @flat_singlethread_one_as_release_acquire_cmpxchg(
 ; GFX1250-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
-  %gep = getelementptr i32, ptr %out, i32 4
+  %gep = getelementptr inbounds i32, ptr %out, i32 4
   %val = cmpxchg volatile ptr %gep, i32 %old, i32 %in syncscope("singlethread-one-as") release acquire
   ret void
 }
@@ -16261,7 +16261,7 @@ define amdgpu_kernel void @flat_singlethread_one_as_acq_rel_acquire_cmpxchg(
 ; GFX1250-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
-  %gep = getelementptr i32, ptr %out, i32 4
+  %gep = getelementptr inbounds i32, ptr %out, i32 4
   %val = cmpxchg volatile ptr %gep, i32 %old, i32 %in syncscope("singlethread-one-as") acq_rel acquire
   ret void
 }
@@ -16514,7 +16514,7 @@ define amdgpu_kernel void @flat_singlethread_one_as_seq_cst_acquire_cmpxchg(
 ; GFX1250-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
-  %gep = getelementptr i32, ptr %out, i32 4
+  %gep = getelementptr inbounds i32, ptr %out, i32 4
   %val = cmpxchg volatile ptr %gep, i32 %old, i32 %in syncscope("singlethread-one-as") seq_cst acquire
   ret void
 }
@@ -16767,7 +16767,7 @@ define amdgpu_kernel void @flat_singlethread_one_as_monotonic_seq_cst_cmpxchg(
 ; GFX1250-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
-  %gep = getelementptr i32, ptr %out, i32 4
+  %gep = getelementptr inbounds i32, ptr %out, i32 4
   %val = cmpxchg volatile ptr %gep, i32 %old, i32 %in syncscope("singlethread-one-as") monotonic seq_cst
   ret void
 }
@@ -17020,7 +17020,7 @@ define amdgpu_kernel void @flat_singlethread_one_as_acquire_seq_cst_cmpxchg(
 ; GFX1250-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
-  %gep = getelementptr i32, ptr %out, i32 4
+  %gep = getelementptr inbounds i32, ptr %out, i32 4
   %val = cmpxchg volatile ptr %gep, i32 %old, i32 %in syncscope("singlethread-one-as") acquire seq_cst
   ret void
 }
@@ -17273,7 +17273,7 @@ define amdgpu_kernel void @flat_singlethread_one_as_release_seq_cst_cmpxchg(
 ; GFX1250-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
-  %gep = getelementptr i32, ptr %out, i32 4
+  %gep = getelementptr inbounds i32, ptr %out, i32 4
   %val = cmpxchg volatile ptr %gep, i32 %old, i32 %in syncscope("singlethread-one-as") release seq_cst
   ret void
 }
@@ -17526,7 +17526,7 @@ define amdgpu_kernel void @flat_singlethread_one_as_acq_rel_seq_cst_cmpxchg(
 ; GFX1250-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
-  %gep = getelementptr i32, ptr %out, i32 4
+  %gep = getelementptr inbounds i32, ptr %out, i32 4
   %val = cmpxchg volatile ptr %gep, i32 %old, i32 %in syncscope("singlethread-one-as") acq_rel seq_cst
   ret void
 }
@@ -17779,7 +17779,7 @@ define amdgpu_kernel void @flat_singlethread_one_as_seq_cst_seq_cst_cmpxchg(
 ; GFX1250-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
-  %gep = getelementptr i32, ptr %out, i32 4
+  %gep = getelementptr inbounds i32, ptr %out, i32 4
   %val = cmpxchg volatile ptr %gep, i32 %old, i32 %in syncscope("singlethread-one-as") seq_cst seq_cst
   ret void
 }
@@ -18078,7 +18078,7 @@ define amdgpu_kernel void @flat_singlethread_one_as_monotonic_monotonic_ret_cmpx
 ; GFX1250-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
-  %gep = getelementptr i32, ptr %out, i32 4
+  %gep = getelementptr inbounds i32, ptr %out, i32 4
   %val = cmpxchg volatile ptr %gep, i32 %old, i32 %in syncscope("singlethread-one-as") monotonic monotonic
   %val0 = extractvalue { i32, i1 } %val, 0
   store i32 %val0, ptr %out, align 4
@@ -18379,7 +18379,7 @@ define amdgpu_kernel void @flat_singlethread_one_as_acquire_monotonic_ret_cmpxch
 ; GFX1250-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
-  %gep = getelementptr i32, ptr %out, i32 4
+  %gep = getelementptr inbounds i32, ptr %out, i32 4
   %val = cmpxchg volatile ptr %gep, i32 %old, i32 %in syncscope("singlethread-one-as") acquire monotonic
   %val0 = extractvalue { i32, i1 } %val, 0
   store i32 %val0, ptr %out, align 4
@@ -18680,7 +18680,7 @@ define amdgpu_kernel void @flat_singlethread_one_as_release_monotonic_ret_cmpxch
 ; GFX1250-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
-  %gep = getelementptr i32, ptr %out, i32 4
+  %gep = getelementptr inbounds i32, ptr %out, i32 4
   %val = cmpxchg volatile ptr %gep, i32 %old, i32 %in syncscope("singlethread-one-as") release monotonic
   %val0 = extractvalue { i32, i1 } %val, 0
   store i32 %val0, ptr %out, align 4
@@ -18981,7 +18981,7 @@ define amdgpu_kernel void @flat_singlethread_one_as_acq_rel_monotonic_ret_cmpxch
 ; GFX1250-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
-  %gep = getelementptr i32, ptr %out, i32 4
+  %gep = getelementptr inbounds i32, ptr %out, i32 4
   %val = cmpxchg volatile ptr %gep, i32 %old, i32 %in syncscope("singlethread-one-as") acq_rel monotonic
   %val0 = extractvalue { i32, i1 } %val, 0
   store i32 %val0, ptr %out, align 4
@@ -19282,7 +19282,7 @@ define amdgpu_kernel void @flat_singlethread_one_as_seq_cst_monotonic_ret_cmpxch
 ; GFX1250-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
-  %gep = getelementptr i32, ptr %out, i32 4
+  %gep = getelementptr inbounds i32, ptr %out, i32 4
   %val = cmpxchg volatile ptr %gep, i32 %old, i32 %in syncscope("singlethread-one-as") seq_cst monotonic
   %val0 = extractvalue { i32, i1 } %val, 0
   store i32 %val0, ptr %out, align 4
@@ -19583,7 +19583,7 @@ define amdgpu_kernel void @flat_singlethread_one_as_monotonic_acquire_ret_cmpxch
 ; GFX1250-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
-  %gep = getelementptr i32, ptr %out, i32 4
+  %gep = getelementptr inbounds i32, ptr %out, i32 4
   %val = cmpxchg volatile ptr %gep, i32 %old, i32 %in syncscope("singlethread-one-as") monotonic acquire
   %val0 = extractvalue { i32, i1 } %val, 0
   store i32 %val0, ptr %out, align 4
@@ -19884,7 +19884,7 @@ define amdgpu_kernel void @flat_singlethread_one_as_acquire_acquire_ret_cmpxchg(
 ; GFX1250-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
-  %gep = getelementptr i32, ptr %out, i32 4
+  %gep = getelementptr inbounds i32, ptr %out, i32 4
   %val = cmpxchg volatile ptr %gep, i32 %old, i32 %in syncscope("singlethread-one-as") acquire acquire
   %val0 = extractvalue { i32, i1 } %val, 0
   store i32 %val0, ptr %out, align 4
@@ -20185,7 +20185,7 @@ define amdgpu_kernel void @flat_singlethread_one_as_release_acquire_ret_cmpxchg(
 ; GFX1250-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
-  %gep = getelementptr i32, ptr %out, i32 4
+  %gep = getelementptr inbounds i32, ptr %out, i32 4
   %val = cmpxchg volatile ptr %gep, i32 %old, i32 %in syncscope("singlethread-one-as") release acquire
   %val0 = extractvalue { i32, i1 } %val, 0
   store i32 %val0, ptr %out, align 4
@@ -20486,7 +20486,7 @@ define amdgpu_kernel void @flat_singlethread_one_as_acq_rel_acquire_ret_cmpxchg(
 ; GFX1250-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
-  %gep = getelementptr i32, ptr %out, i32 4
+  %gep = getelementptr inbounds i32, ptr %out, i32 4
   %val = cmpxchg volatile ptr %gep, i32 %old, i32 %in syncscope("singlethread-one-as") acq_rel acquire
   %val0 = extractvalue { i32, i1 } %val, 0
   store i32 %val0, ptr %out, align 4
@@ -20787,7 +20787,7 @@ define amdgpu_kernel void @flat_singlethread_one_as_seq_cst_acquire_ret_cmpxchg(
 ; GFX1250-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
-  %gep = getelementptr i32, ptr %out, i32 4
+  %gep = getelementptr inbounds i32, ptr %out, i32 4
   %val = cmpxchg volatile ptr %gep, i32 %old, i32 %in syncscope("singlethread-one-as") seq_cst acquire
   %val0 = extractvalue { i32, i1 } %val, 0
   store i32 %val0, ptr %out, align 4
@@ -21088,7 +21088,7 @@ define amdgpu_kernel void @flat_singlethread_one_as_monotonic_seq_cst_ret_cmpxch
 ; GFX1250-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
-  %gep = getelementptr i32, ptr %out, i32 4
+  %gep = getelementptr inbounds i32, ptr %out, i32 4
   %val = cmpxchg volatile ptr %gep, i32 %old, i32 %in syncscope("singlethread-one-as") monotonic seq_cst
   %val0 = extractvalue { i32, i1 } %val, 0
   store i32 %val0, ptr %out, align 4
@@ -21389,7 +21389,7 @@ define amdgpu_kernel void @flat_singlethread_one_as_acquire_seq_cst_ret_cmpxchg(
 ; GFX1250-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
-  %gep = getelementptr i32, ptr %out, i32 4
+  %gep = getelementptr inbounds i32, ptr %out, i32 4
   %val = cmpxchg volatile ptr %gep, i32 %old, i32 %in syncscope("singlethread-one-as") acquire seq_cst
   %val0 = extractvalue { i32, i1 } %val, 0
   store i32 %val0, ptr %out, align 4
@@ -21690,7 +21690,7 @@ define amdgpu_kernel void @flat_singlethread_one_as_release_seq_cst_ret_cmpxchg(
 ; GFX1250-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
-  %gep = getelementptr i32, ptr %out, i32 4
+  %gep = getelementptr inbounds i32, ptr %out, i32 4
   %val = cmpxchg volatile ptr %gep, i32 %old, i32 %in syncscope("singlethread-one-as") release seq_cst
   %val0 = extractvalue { i32, i1 } %val, 0
   store i32 %val0, ptr %out, align 4
@@ -21991,7 +21991,7 @@ define amdgpu_kernel void @flat_singlethread_one_as_acq_rel_seq_cst_ret_cmpxchg(
 ; GFX1250-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
-  %gep = getelementptr i32, ptr %out, i32 4
+  %gep = getelementptr inbounds i32, ptr %out, i32 4
   %val = cmpxchg volatile ptr %gep, i32 %old, i32 %in syncscope("singlethread-one-as") acq_rel seq_cst
   %val0 = extractvalue { i32, i1 } %val, 0
   store i32 %val0, ptr %out, align 4
@@ -22292,7 +22292,7 @@ define amdgpu_kernel void @flat_singlethread_one_as_seq_cst_seq_cst_ret_cmpxchg(
 ; GFX1250-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
-  %gep = getelementptr i32, ptr %out, i32 4
+  %gep = getelementptr inbounds i32, ptr %out, i32 4
   %val = cmpxchg volatile ptr %gep, i32 %old, i32 %in syncscope("singlethread-one-as") seq_cst seq_cst
   %val0 = extractvalue { i32, i1 } %val, 0
   store i32 %val0, ptr %out, align 4
diff --git a/llvm/test/CodeGen/AMDGPU/memory-legalizer-flat-system.ll b/llvm/test/CodeGen/AMDGPU/memory-legalizer-flat-system.ll
index 74065146fd385..d5b37650ae9cc 100644
--- a/llvm/test/CodeGen/AMDGPU/memory-legalizer-flat-system.ll
+++ b/llvm/test/CodeGen/AMDGPU/memory-legalizer-flat-system.ll
@@ -3530,7 +3530,7 @@ define amdgpu_kernel void @flat_system_monotonic_monotonic_cmpxchg(
 ; GFX1250-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
-  %gep = getelementptr i32, ptr %out, i32 4
+  %gep = getelementptr inbounds i32, ptr %out, i32 4
   %val = cmpxchg volatile ptr %gep, i32 %old, i32 %in monotonic monotonic
   ret void
 }
@@ -3818,7 +3818,7 @@ define amdgpu_kernel void @flat_system_acquire_monotonic_cmpxchg(
 ; GFX1250-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
-  %gep = getelementptr i32, ptr %out, i32 4
+  %gep = getelementptr inbounds i32, ptr %out, i32 4
   %val = cmpxchg volatile ptr %gep, i32 %old, i32 %in acquire monotonic
   ret void
 }
@@ -4102,7 +4102,7 @@ define amdgpu_kernel void @flat_system_release_monotonic_cmpxchg(
 ; GFX1250-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
-  %gep = getelementptr i32, ptr %out, i32 4
+  %gep = getelementptr inbounds i32, ptr %out, i32 4
   %val = cmpxchg volatile ptr %gep, i32 %old, i32 %in release monotonic
   ret void
 }
@@ -4421,7 +4421,7 @@ define amdgpu_kernel void @flat_system_acq_rel_monotonic_cmpxchg(
 ; GFX1250-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
-  %gep = getelementptr i32, ptr %out, i32 4
+  %gep = getelementptr inbounds i32, ptr %out, i32 4
   %val = cmpxchg volatile ptr %gep, i32 %old, i32 %in acq_rel monotonic
   ret void
 }
@@ -4740,7 +4740,7 @@ define amdgpu_kernel void @flat_system_seq_cst_monotonic_cmpxchg(
 ; GFX1250-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
-  %gep = getelementptr i32, ptr %out, i32 4
+  %gep = getelementptr inbounds i32, ptr %out, i32 4
   %val = cmpxchg volatile ptr %gep, i32 %old, i32 %in seq_cst monotonic
   ret void
 }
@@ -5028,7 +5028,7 @@ define amdgpu_kernel void @flat_system_monotonic_acquire_cmpxchg(
 ; GFX1250-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
-  %gep = getelementptr i32, ptr %out, i32 4
+  %gep = getelementptr inbounds i32, ptr %out, i32 4
   %val = cmpxchg volatile ptr %gep, i32 %old, i32 %in monotonic acquire
   ret void
 }
@@ -5316,7 +5316,7 @@ define amdgpu_kernel void @flat_system_acquire_acquire_cmpxchg(
 ; GFX1250-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
-  %gep = getelementptr i32, ptr %out, i32 4
+  %gep = getelementptr inbounds i32, ptr %out, i32 4
   %val = cmpxchg volatile ptr %gep, i32 %old, i32 %in acquire acquire
   ret void
 }
@@ -5635,7 +5635,7 @@ define amdgpu_kernel void @flat_system_release_acquire_cmpxchg(
 ; GFX1250-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
-  %gep = getelementptr i32, ptr %out, i32 4
+  %gep = getelementptr inbounds i32, ptr %out, i32 4
   %val = cmpxchg volatile ptr %gep, i32 %old, i32 %in release acquire
   ret void
 }
@@ -5954,7 +5954,7 @@ define amdgpu_kernel void @flat_system_acq_rel_acquire_cmpxchg(
 ; GFX1250-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
-  %gep = getelementptr i32, ptr %out, i32 4
+  %gep = getelementptr inbounds i32, ptr %out, i32 4
   %val = cmpxchg volatile ptr %gep, i32 %old, i32 %in acq_rel acquire
   ret void
 }
@@ -6273,7 +6273,7 @@ define amdgpu_kernel void @flat_system_seq_cst_acquire_cmpxchg(
 ; GFX1250-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
-  %gep = getelementptr i32, ptr %out, i32 4
+  %gep = getelementptr inbounds i32, ptr %out, i32 4
   %val = cmpxchg volatile ptr %gep, i32 %old, i32 %in seq_cst acquire
   ret void
 }
@@ -6592,7 +6592,7 @@ define amdgpu_kernel void @flat_system_monotonic_seq_cst_cmpxchg(
 ; GFX1250-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
-  %gep = getelementptr i32, ptr %out, i32 4
+  %gep = getelementptr inbounds i32, ptr %out, i32 4
   %val = cmpxchg volatile ptr %gep, i32 %old, i32 %in monotonic seq_cst
   ret void
 }
@@ -6911,7 +6911,7 @@ define amdgpu_kernel void @flat_system_acquire_seq_cst_cmpxchg(
 ; GFX1250-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
-  %gep = getelementptr i32, ptr %out, i32 4
+  %gep = getelementptr inbounds i32, ptr %out, i32 4
   %val = cmpxchg volatile ptr %gep, i32 %old, i32 %in acquire seq_cst
   ret void
 }
@@ -7230,7 +7230,7 @@ define amdgpu_kernel void @flat_system_release_seq_cst_cmpxchg(
 ; GFX1250-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
-  %gep = getelementptr i32, ptr %out, i32 4
+  %gep = getelementptr inbounds i32, ptr %out, i32 4
   %val = cmpxchg volatile ptr %gep, i32 %old, i32 %in release seq_cst
   ret void
 }
@@ -7549,7 +7549,7 @@ define amdgpu_kernel void @flat_system_acq_rel_seq_cst_cmpxchg(
 ; GFX1250-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
-  %gep = getelementptr i32, ptr %out, i32 4
+  %gep = getelementptr inbounds i32, ptr %out, i32 4
   %val = cmpxchg volatile ptr %gep, i32 %old, i32 %in acq_rel seq_cst
   ret void
 }
@@ -7868,7 +7868,7 @@ define amdgpu_kernel void @flat_system_seq_cst_seq_cst_cmpxchg(
 ; GFX1250-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
-  %gep = getelementptr i32, ptr %out, i32 4
+  %gep = getelementptr inbounds i32, ptr %out, i32 4
   %val = cmpxchg volatile ptr %gep, i32 %old, i32 %in seq_cst seq_cst
   ret void
 }
@@ -8167,7 +8167,7 @@ define amdgpu_kernel void @flat_system_monotonic_monotonic_ret_cmpxchg(
 ; GFX1250-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
-  %gep = getelementptr i32, ptr %out, i32 4
+  %gep = getelementptr inbounds i32, ptr %out, i32 4
   %val = cmpxchg volatile ptr %gep, i32 %old, i32 %in monotonic monotonic
   %val0 = extractvalue { i32, i1 } %val, 0
   store i32 %val0, ptr %out, align 4
@@ -8486,7 +8486,7 @@ define amdgpu_kernel void @flat_system_acquire_monotonic_ret_cmpxchg(
 ; GFX1250-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
-  %gep = getelementptr i32, ptr %out, i32 4
+  %gep = getelementptr inbounds i32, ptr %out, i32 4
   %val = cmpxchg volatile ptr %gep, i32 %old, i32 %in acquire monotonic
   %val0 = extractvalue { i32, i1 } %val, 0
   store i32 %val0, ptr %out, align 4
@@ -8818,7 +8818,7 @@ define amdgpu_kernel void @flat_system_release_monotonic_ret_cmpxchg(
 ; GFX1250-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
-  %gep = getelementptr i32, ptr %out, i32 4
+  %gep = getelementptr inbounds i32, ptr %out, i32 4
   %val = cmpxchg volatile ptr %gep, i32 %old, i32 %in release monotonic
   %val0 = extractvalue { i32, i1 } %val, 0
   store i32 %val0, ptr %out, align 4
@@ -9168,7 +9168,7 @@ define amdgpu_kernel void @flat_system_acq_rel_monotonic_ret_cmpxchg(
 ; GFX1250-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
-  %gep = getelementptr i32, ptr %out, i32 4
+  %gep = getelementptr inbounds i32, ptr %out, i32 4
   %val = cmpxchg volatile ptr %gep, i32 %old, i32 %in acq_rel monotonic
   %val0 = extractvalue { i32, i1 } %val, 0
   store i32 %val0, ptr %out, align 4
@@ -9518,7 +9518,7 @@ define amdgpu_kernel void @flat_system_seq_cst_monotonic_ret_cmpxchg(
 ; GFX1250-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
-  %gep = getelementptr i32, ptr %out, i32 4
+  %gep = getelementptr inbounds i32, ptr %out, i32 4
   %val = cmpxchg volatile ptr %gep, i32 %old, i32 %in seq_cst monotonic
   %val0 = extractvalue { i32, i1 } %val, 0
   store i32 %val0, ptr %out, align 4
@@ -9837,7 +9837,7 @@ define amdgpu_kernel void @flat_system_monotonic_acquire_ret_cmpxchg(
 ; GFX1250-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
-  %gep = getelementptr i32, ptr %out, i32 4
+  %gep = getelementptr inbounds i32, ptr %out, i32 4
   %val = cmpxchg volatile ptr %gep, i32 %old, i32 %in monotonic acquire
   %val0 = extractvalue { i32, i1 } %val, 0
   store i32 %val0, ptr %out, align 4
@@ -10156,7 +10156,7 @@ define amdgpu_kernel void @flat_system_acquire_acquire_ret_cmpxchg(
 ; GFX1250-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
-  %gep = getelementptr i32, ptr %out, i32 4
+  %gep = getelementptr inbounds i32, ptr %out, i32 4
   %val = cmpxchg volatile ptr %gep, i32 %old, i32 %in acquire acquire
   %val0 = extractvalue { i32, i1 } %val, 0
   store i32 %val0, ptr %out, align 4
@@ -10506,7 +10506,7 @@ define amdgpu_kernel void @flat_system_release_acquire_ret_cmpxchg(
 ; GFX1250-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
-  %gep = getelementptr i32, ptr %out, i32 4
+  %gep = getelementptr inbounds i32, ptr %out, i32 4
   %val = cmpxchg volatile ptr %gep, i32 %old, i32 %in release acquire
   %val0 = extractvalue { i32, i1 } %val, 0
   store i32 %val0, ptr %out, align 4
@@ -10856,7 +10856,7 @@ define amdgpu_kernel void @flat_system_acq_rel_acquire_ret_cmpxchg(
 ; GFX1250-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
-  %gep = getelementptr i32, ptr %out, i32 4
+  %gep = getelementptr inbounds i32, ptr %out, i32 4
   %val = cmpxchg volatile ptr %gep, i32 %old, i32 %in acq_rel acquire
   %val0 = extractvalue { i32, i1 } %val, 0
   store i32 %val0, ptr %out, align 4
@@ -11206,7 +11206,7 @@ define amdgpu_kernel void @flat_system_seq_cst_acquire_ret_cmpxchg(
 ; GFX1250-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
-  %gep = getelementptr i32, ptr %out, i32 4
+  %gep = getelementptr inbounds i32, ptr %out, i32 4
   %val = cmpxchg volatile ptr %gep, i32 %old, i32 %in seq_cst acquire
   %val0 = extractvalue { i32, i1 } %val, 0
   store i32 %val0, ptr %out, align 4
@@ -11556,7 +11556,7 @@ define amdgpu_kernel void @flat_system_monotonic_seq_cst_ret_cmpxchg(
 ; GFX1250-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
-  %gep = getelementptr i32, ptr %out, i32 4
+  %gep = getelementptr inbounds i32, ptr %out, i32 4
   %val = cmpxchg volatile ptr %gep, i32 %old, i32 %in monotonic seq_cst
   %val0 = extractvalue { i32, i1 } %val, 0
   store i32 %val0, ptr %out, align 4
@@ -11906,7 +11906,7 @@ define amdgpu_kernel void @flat_system_acquire_seq_cst_ret_cmpxchg(
 ; GFX1250-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
-  %gep = getelementptr i32, ptr %out, i32 4
+  %gep = getelementptr inbounds i32, ptr %out, i32 4
   %val = cmpxchg volatile ptr %gep, i32 %old, i32 %in acquire seq_cst
   %val0 = extractvalue { i32, i1 } %val, 0
   store i32 %val0, ptr %out, align 4
@@ -12256,7 +12256,7 @@ define amdgpu_kernel void @flat_system_release_seq_cst_ret_cmpxchg(
 ; GFX1250-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
-  %gep = getelementptr i32, ptr %out, i32 4
+  %gep = getelementptr inbounds i32, ptr %out, i32 4
   %val = cmpxchg volatile ptr %gep, i32 %old, i32 %in release seq_cst
   %val0 = extractvalue { i32, i1 } %val, 0
   store i32 %val0, ptr %out, align 4
@@ -12606,7 +12606,7 @@ define amdgpu_kernel void @flat_system_acq_rel_seq_cst_ret_cmpxchg(
 ; GFX1250-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
-  %gep = getelementptr i32, ptr %out, i32 4
+  %gep = getelementptr inbounds i32, ptr %out, i32 4
   %val = cmpxchg volatile ptr %gep, i32 %old, i32 %in acq_rel seq_cst
   %val0 = extractvalue { i32, i1 } %val, 0
   store i32 %val0, ptr %out, align 4
@@ -12956,7 +12956,7 @@ define amdgpu_kernel void @flat_system_seq_cst_seq_cst_ret_cmpxchg(
 ; GFX1250-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
-  %gep = getelementptr i32, ptr %out, i32 4
+  %gep = getelementptr inbounds i32, ptr %out, i32 4
   %val = cmpxchg volatile ptr %gep, i32 %old, i32 %in seq_cst seq_cst
   %val0 = extractvalue { i32, i1 } %val, 0
   store i32 %val0, ptr %out, align 4
@@ -16523,7 +16523,7 @@ define amdgpu_kernel void @flat_system_one_as_monotonic_monotonic_cmpxchg(
 ; GFX1250-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
-  %gep = getelementptr i32, ptr %out, i32 4
+  %gep = getelementptr inbounds i32, ptr %out, i32 4
   %val = cmpxchg volatile ptr %gep, i32 %old, i32 %in syncscope("one-as") monotonic monotonic
   ret void
 }
@@ -16807,7 +16807,7 @@ define amdgpu_kernel void @flat_system_one_as_acquire_monotonic_cmpxchg(
 ; GFX1250-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
-  %gep = getelementptr i32, ptr %out, i32 4
+  %gep = getelementptr inbounds i32, ptr %out, i32 4
   %val = cmpxchg volatile ptr %gep, i32 %old, i32 %in syncscope("one-as") acquire monotonic
   ret void
 }
@@ -17091,7 +17091,7 @@ define amdgpu_kernel void @flat_system_one_as_release_monotonic_cmpxchg(
 ; GFX1250-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
-  %gep = getelementptr i32, ptr %out, i32 4
+  %gep = getelementptr inbounds i32, ptr %out, i32 4
   %val = cmpxchg volatile ptr %gep, i32 %old, i32 %in syncscope("one-as") release monotonic
   ret void
 }
@@ -17406,7 +17406,7 @@ define amdgpu_kernel void @flat_system_one_as_acq_rel_monotonic_cmpxchg(
 ; GFX1250-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
-  %gep = getelementptr i32, ptr %out, i32 4
+  %gep = getelementptr inbounds i32, ptr %out, i32 4
   %val = cmpxchg volatile ptr %gep, i32 %old, i32 %in syncscope("one-as") acq_rel monotonic
   ret void
 }
@@ -17721,7 +17721,7 @@ define amdgpu_kernel void @flat_system_one_as_seq_cst_monotonic_cmpxchg(
 ; GFX1250-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
-  %gep = getelementptr i32, ptr %out, i32 4
+  %gep = getelementptr inbounds i32, ptr %out, i32 4
   %val = cmpxchg volatile ptr %gep, i32 %old, i32 %in syncscope("one-as") seq_cst monotonic
   ret void
 }
@@ -18005,7 +18005,7 @@ define amdgpu_kernel void @flat_system_one_as_monotonic_acquire_cmpxchg(
 ; GFX1250-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
-  %gep = getelementptr i32, ptr %out, i32 4
+  %gep = getelementptr inbounds i32, ptr %out, i32 4
   %val = cmpxchg volatile ptr %gep, i32 %old, i32 %in syncscope("one-as") monotonic acquire
   ret void
 }
@@ -18289,7 +18289,7 @@ define amdgpu_kernel void @flat_system_one_as_acquire_acquire_cmpxchg(
 ; GFX1250-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
-  %gep = getelementptr i32, ptr %out, i32 4
+  %gep = getelementptr inbounds i32, ptr %out, i32 4
   %val = cmpxchg volatile ptr %gep, i32 %old, i32 %in syncscope("one-as") acquire acquire
   ret void
 }
@@ -18604,7 +18604,7 @@ define amdgpu_kernel void @flat_system_one_as_release_acquire_cmpxchg(
 ; GFX1250-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
-  %gep = getelementptr i32, ptr %out, i32 4
+  %gep = getelementptr inbounds i32, ptr %out, i32 4
   %val = cmpxchg volatile ptr %gep, i32 %old, i32 %in syncscope("one-as") release acquire
   ret void
 }
@@ -18919,7 +18919,7 @@ define amdgpu_kernel void @flat_system_one_as_acq_rel_acquire_cmpxchg(
 ; GFX1250-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
-  %gep = getelementptr i32, ptr %out, i32 4
+  %gep = getelementptr inbounds i32, ptr %out, i32 4
   %val = cmpxchg volatile ptr %gep, i32 %old, i32 %in syncscope("one-as") acq_rel acquire
   ret void
 }
@@ -19234,7 +19234,7 @@ define amdgpu_kernel void @flat_system_one_as_seq_cst_acquire_cmpxchg(
 ; GFX1250-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
-  %gep = getelementptr i32, ptr %out, i32 4
+  %gep = getelementptr inbounds i32, ptr %out, i32 4
   %val = cmpxchg volatile ptr %gep, i32 %old, i32 %in syncscope("one-as") seq_cst acquire
   ret void
 }
@@ -19549,7 +19549,7 @@ define amdgpu_kernel void @flat_system_one_as_monotonic_seq_cst_cmpxchg(
 ; GFX1250-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
-  %gep = getelementptr i32, ptr %out, i32 4
+  %gep = getelementptr inbounds i32, ptr %out, i32 4
   %val = cmpxchg volatile ptr %gep, i32 %old, i32 %in syncscope("one-as") monotonic seq_cst
   ret void
 }
@@ -19864,7 +19864,7 @@ define amdgpu_kernel void @flat_system_one_as_acquire_seq_cst_cmpxchg(
 ; GFX1250-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
-  %gep = getelementptr i32, ptr %out, i32 4
+  %gep = getelementptr inbounds i32, ptr %out, i32 4
   %val = cmpxchg volatile ptr %gep, i32 %old, i32 %in syncscope("one-as") acquire seq_cst
   ret void
 }
@@ -20179,7 +20179,7 @@ define amdgpu_kernel void @flat_system_one_as_release_seq_cst_cmpxchg(
 ; GFX1250-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
-  %gep = getelementptr i32, ptr %out, i32 4
+  %gep = getelementptr inbounds i32, ptr %out, i32 4
   %val = cmpxchg volatile ptr %gep, i32 %old, i32 %in syncscope("one-as") release seq_cst
   ret void
 }
@@ -20494,7 +20494,7 @@ define amdgpu_kernel void @flat_system_one_as_acq_rel_seq_cst_cmpxchg(
 ; GFX1250-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
-  %gep = getelementptr i32, ptr %out, i32 4
+  %gep = getelementptr inbounds i32, ptr %out, i32 4
   %val = cmpxchg volatile ptr %gep, i32 %old, i32 %in syncscope("one-as") acq_rel seq_cst
   ret void
 }
@@ -20809,7 +20809,7 @@ define amdgpu_kernel void @flat_system_one_as_seq_cst_seq_cst_cmpxchg(
 ; GFX1250-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
-  %gep = getelementptr i32, ptr %out, i32 4
+  %gep = getelementptr inbounds i32, ptr %out, i32 4
   %val = cmpxchg volatile ptr %gep, i32 %old, i32 %in syncscope("one-as") seq_cst seq_cst
   ret void
 }
@@ -21108,7 +21108,7 @@ define amdgpu_kernel void @flat_system_one_as_monotonic_monotonic_ret_cmpxchg(
 ; GFX1250-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
-  %gep = getelementptr i32, ptr %out, i32 4
+  %gep = getelementptr inbounds i32, ptr %out, i32 4
   %val = cmpxchg volatile ptr %gep, i32 %old, i32 %in syncscope("one-as") monotonic monotonic
   %val0 = extractvalue { i32, i1 } %val, 0
   store i32 %val0, ptr %out, align 4
@@ -21438,7 +21438,7 @@ define amdgpu_kernel void @flat_system_one_as_acquire_monotonic_ret_cmpxchg(
 ; GFX1250-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
-  %gep = getelementptr i32, ptr %out, i32 4
+  %gep = getelementptr inbounds i32, ptr %out, i32 4
   %val = cmpxchg volatile ptr %gep, i32 %old, i32 %in syncscope("one-as") acquire monotonic
   %val0 = extractvalue { i32, i1 } %val, 0
   store i32 %val0, ptr %out, align 4
@@ -21770,7 +21770,7 @@ define amdgpu_kernel void @flat_system_one_as_release_monotonic_ret_cmpxchg(
 ; GFX1250-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
-  %gep = getelementptr i32, ptr %out, i32 4
+  %gep = getelementptr inbounds i32, ptr %out, i32 4
   %val = cmpxchg volatile ptr %gep, i32 %old, i32 %in syncscope("one-as") release monotonic
   %val0 = extractvalue { i32, i1 } %val, 0
   store i32 %val0, ptr %out, align 4
@@ -22131,7 +22131,7 @@ define amdgpu_kernel void @flat_system_one_as_acq_rel_monotonic_ret_cmpxchg(
 ; GFX1250-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
-  %gep = getelementptr i32, ptr %out, i32 4
+  %gep = getelementptr inbounds i32, ptr %out, i32 4
   %val = cmpxchg volatile ptr %gep, i32 %old, i32 %in syncscope("one-as") acq_rel monotonic
   %val0 = extractvalue { i32, i1 } %val, 0
   store i32 %val0, ptr %out, align 4
@@ -22492,7 +22492,7 @@ define amdgpu_kernel void @flat_system_one_as_seq_cst_monotonic_ret_cmpxchg(
 ; GFX1250-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
-  %gep = getelementptr i32, ptr %out, i32 4
+  %gep = getelementptr inbounds i32, ptr %out, i32 4
   %val = cmpxchg volatile ptr %gep, i32 %old, i32 %in syncscope("one-as") seq_cst monotonic
   %val0 = extractvalue { i32, i1 } %val, 0
   store i32 %val0, ptr %out, align 4
@@ -22822,7 +22822,7 @@ define amdgpu_kernel void @flat_system_one_as_monotonic_acquire_ret_cmpxchg(
 ; GFX1250-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
-  %gep = getelementptr i32, ptr %out, i32 4
+  %gep = getelementptr inbounds i32, ptr %out, i32 4
   %val = cmpxchg volatile ptr %gep, i32 %old, i32 %in syncscope("one-as") monotonic acquire
   %val0 = extractvalue { i32, i1 } %val, 0
   store i32 %val0, ptr %out, align 4
@@ -23152,7 +23152,7 @@ define amdgpu_kernel void @flat_system_one_as_acquire_acquire_ret_cmpxchg(
 ; GFX1250-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
-  %gep = getelementptr i32, ptr %out, i32 4
+  %gep = getelementptr inbounds i32, ptr %out, i32 4
   %val = cmpxchg volatile ptr %gep, i32 %old, i32 %in syncscope("one-as") acquire acquire
   %val0 = extractvalue { i32, i1 } %val, 0
   store i32 %val0, ptr %out, align 4
@@ -23513,7 +23513,7 @@ define amdgpu_kernel void @flat_system_one_as_release_acquire_ret_cmpxchg(
 ; GFX1250-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
-  %gep = getelementptr i32, ptr %out, i32 4
+  %gep = getelementptr inbounds i32, ptr %out, i32 4
   %val = cmpxchg volatile ptr %gep, i32 %old, i32 %in syncscope("one-as") release acquire
   %val0 = extractvalue { i32, i1 } %val, 0
   store i32 %val0, ptr %out, align 4
@@ -23874,7 +23874,7 @@ define amdgpu_kernel void @flat_system_one_as_acq_rel_acquire_ret_cmpxchg(
 ; GFX1250-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
-  %gep = getelementptr i32, ptr %out, i32 4
+  %gep = getelementptr inbounds i32, ptr %out, i32 4
   %val = cmpxchg volatile ptr %gep, i32 %old, i32 %in syncscope("one-as") acq_rel acquire
   %val0 = extractvalue { i32, i1 } %val, 0
   store i32 %val0, ptr %out, align 4
@@ -24235,7 +24235,7 @@ define amdgpu_kernel void @flat_system_one_as_seq_cst_acquire_ret_cmpxchg(
 ; GFX1250-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
-  %gep = getelementptr i32, ptr %out, i32 4
+  %gep = getelementptr inbounds i32, ptr %out, i32 4
   %val = cmpxchg volatile ptr %gep, i32 %old, i32 %in syncscope("one-as") seq_cst acquire
   %val0 = extractvalue { i32, i1 } %val, 0
   store i32 %val0, ptr %out, align 4
@@ -24596,7 +24596,7 @@ define amdgpu_kernel void @flat_system_one_as_monotonic_seq_cst_ret_cmpxchg(
 ; GFX1250-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
-  %gep = getelementptr i32, ptr %out, i32 4
+  %gep = getelementptr inbounds i32, ptr %out, i32 4
   %val = cmpxchg volatile ptr %gep, i32 %old, i32 %in syncscope("one-as") monotonic seq_cst
   %val0 = extractvalue { i32, i1 } %val, 0
   store i32 %val0, ptr %out, align 4
@@ -24957,7 +24957,7 @@ define amdgpu_kernel void @flat_system_one_as_acquire_seq_cst_ret_cmpxchg(
 ; GFX1250-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
-  %gep = getelementptr i32, ptr %out, i32 4
+  %gep = getelementptr inbounds i32, ptr %out, i32 4
   %val = cmpxchg volatile ptr %gep, i32 %old, i32 %in syncscope("one-as") acquire seq_cst
   %val0 = extractvalue { i32, i1 } %val, 0
   store i32 %val0, ptr %out, align 4
@@ -25318,7 +25318,7 @@ define amdgpu_kernel void @flat_system_one_as_release_seq_cst_ret_cmpxchg(
 ; GFX1250-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
-  %gep = getelementptr i32, ptr %out, i32 4
+  %gep = getelementptr inbounds i32, ptr %out, i32 4
   %val = cmpxchg volatile ptr %gep, i32 %old, i32 %in syncscope("one-as") release seq_cst
   %val0 = extractvalue { i32, i1 } %val, 0
   store i32 %val0, ptr %out, align 4
@@ -25679,7 +25679,7 @@ define amdgpu_kernel void @flat_system_one_as_acq_rel_seq_cst_ret_cmpxchg(
 ; GFX1250-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
-  %gep = getelementptr i32, ptr %out, i32 4
+  %gep = getelementptr inbounds i32, ptr %out, i32 4
   %val = cmpxchg volatile ptr %gep, i32 %old, i32 %in syncscope("one-as") acq_rel seq_cst
   %val0 = extractvalue { i32, i1 } %val, 0
   store i32 %val0, ptr %out, align 4
@@ -26040,7 +26040,7 @@ define amdgpu_kernel void @flat_system_one_as_seq_cst_seq_cst_ret_cmpxchg(
 ; GFX1250-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
-  %gep = getelementptr i32, ptr %out, i32 4
+  %gep = getelementptr inbounds i32, ptr %out, i32 4
   %val = cmpxchg volatile ptr %gep, i32 %old, i32 %in syncscope("one-as") seq_cst seq_cst
   %val0 = extractvalue { i32, i1 } %val, 0
   store i32 %val0, ptr %out, align 4
diff --git a/llvm/test/CodeGen/AMDGPU/memory-legalizer-flat-wavefront.ll b/llvm/test/CodeGen/AMDGPU/memory-legalizer-flat-wavefront.ll
index 8734e7152e281..b8e324ff5f458 100644
--- a/llvm/test/CodeGen/AMDGPU/memory-legalizer-flat-wavefront.ll
+++ b/llvm/test/CodeGen/AMDGPU/memory-legalizer-flat-wavefront.ll
@@ -3094,7 +3094,7 @@ define amdgpu_kernel void @flat_wavefront_monotonic_monotonic_cmpxchg(
 ; GFX1250-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
-  %gep = getelementptr i32, ptr %out, i32 4
+  %gep = getelementptr inbounds i32, ptr %out, i32 4
   %val = cmpxchg volatile ptr %gep, i32 %old, i32 %in syncscope("wavefront") monotonic monotonic
   ret void
 }
@@ -3347,7 +3347,7 @@ define amdgpu_kernel void @flat_wavefront_acquire_monotonic_cmpxchg(
 ; GFX1250-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
-  %gep = getelementptr i32, ptr %out, i32 4
+  %gep = getelementptr inbounds i32, ptr %out, i32 4
   %val = cmpxchg volatile ptr %gep, i32 %old, i32 %in syncscope("wavefront") acquire monotonic
   ret void
 }
@@ -3600,7 +3600,7 @@ define amdgpu_kernel void @flat_wavefront_release_monotonic_cmpxchg(
 ; GFX1250-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
-  %gep = getelementptr i32, ptr %out, i32 4
+  %gep = getelementptr inbounds i32, ptr %out, i32 4
   %val = cmpxchg volatile ptr %gep, i32 %old, i32 %in syncscope("wavefront") release monotonic
   ret void
 }
@@ -3853,7 +3853,7 @@ define amdgpu_kernel void @flat_wavefront_acq_rel_monotonic_cmpxchg(
 ; GFX1250-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
-  %gep = getelementptr i32, ptr %out, i32 4
+  %gep = getelementptr inbounds i32, ptr %out, i32 4
   %val = cmpxchg volatile ptr %gep, i32 %old, i32 %in syncscope("wavefront") acq_rel monotonic
   ret void
 }
@@ -4106,7 +4106,7 @@ define amdgpu_kernel void @flat_wavefront_seq_cst_monotonic_cmpxchg(
 ; GFX1250-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
-  %gep = getelementptr i32, ptr %out, i32 4
+  %gep = getelementptr inbounds i32, ptr %out, i32 4
   %val = cmpxchg volatile ptr %gep, i32 %old, i32 %in syncscope("wavefront") seq_cst monotonic
   ret void
 }
@@ -4359,7 +4359,7 @@ define amdgpu_kernel void @flat_wavefront_monotonic_acquire_cmpxchg(
 ; GFX1250-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
-  %gep = getelementptr i32, ptr %out, i32 4
+  %gep = getelementptr inbounds i32, ptr %out, i32 4
   %val = cmpxchg volatile ptr %gep, i32 %old, i32 %in syncscope("wavefront") monotonic acquire
   ret void
 }
@@ -4612,7 +4612,7 @@ define amdgpu_kernel void @flat_wavefront_acquire_acquire_cmpxchg(
 ; GFX1250-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
-  %gep = getelementptr i32, ptr %out, i32 4
+  %gep = getelementptr inbounds i32, ptr %out, i32 4
   %val = cmpxchg volatile ptr %gep, i32 %old, i32 %in syncscope("wavefront") acquire acquire
   ret void
 }
@@ -4865,7 +4865,7 @@ define amdgpu_kernel void @flat_wavefront_release_acquire_cmpxchg(
 ; GFX1250-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
-  %gep = getelementptr i32, ptr %out, i32 4
+  %gep = getelementptr inbounds i32, ptr %out, i32 4
   %val = cmpxchg volatile ptr %gep, i32 %old, i32 %in syncscope("wavefront") release acquire
   ret void
 }
@@ -5118,7 +5118,7 @@ define amdgpu_kernel void @flat_wavefront_acq_rel_acquire_cmpxchg(
 ; GFX1250-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
-  %gep = getelementptr i32, ptr %out, i32 4
+  %gep = getelementptr inbounds i32, ptr %out, i32 4
   %val = cmpxchg volatile ptr %gep, i32 %old, i32 %in syncscope("wavefront") acq_rel acquire
   ret void
 }
@@ -5371,7 +5371,7 @@ define amdgpu_kernel void @flat_wavefront_seq_cst_acquire_cmpxchg(
 ; GFX1250-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
-  %gep = getelementptr i32, ptr %out, i32 4
+  %gep = getelementptr inbounds i32, ptr %out, i32 4
   %val = cmpxchg volatile ptr %gep, i32 %old, i32 %in syncscope("wavefront") seq_cst acquire
   ret void
 }
@@ -5624,7 +5624,7 @@ define amdgpu_kernel void @flat_wavefront_monotonic_seq_cst_cmpxchg(
 ; GFX1250-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
-  %gep = getelementptr i32, ptr %out, i32 4
+  %gep = getelementptr inbounds i32, ptr %out, i32 4
   %val = cmpxchg volatile ptr %gep, i32 %old, i32 %in syncscope("wavefront") monotonic seq_cst
   ret void
 }
@@ -5877,7 +5877,7 @@ define amdgpu_kernel void @flat_wavefront_acquire_seq_cst_cmpxchg(
 ; GFX1250-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
-  %gep = getelementptr i32, ptr %out, i32 4
+  %gep = getelementptr inbounds i32, ptr %out, i32 4
   %val = cmpxchg volatile ptr %gep, i32 %old, i32 %in syncscope("wavefront") acquire seq_cst
   ret void
 }
@@ -6130,7 +6130,7 @@ define amdgpu_kernel void @flat_wavefront_release_seq_cst_cmpxchg(
 ; GFX1250-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
-  %gep = getelementptr i32, ptr %out, i32 4
+  %gep = getelementptr inbounds i32, ptr %out, i32 4
   %val = cmpxchg volatile ptr %gep, i32 %old, i32 %in syncscope("wavefront") release seq_cst
   ret void
 }
@@ -6383,7 +6383,7 @@ define amdgpu_kernel void @flat_wavefront_acq_rel_seq_cst_cmpxchg(
 ; GFX1250-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
-  %gep = getelementptr i32, ptr %out, i32 4
+  %gep = getelementptr inbounds i32, ptr %out, i32 4
   %val = cmpxchg volatile ptr %gep, i32 %old, i32 %in syncscope("wavefront") acq_rel seq_cst
   ret void
 }
@@ -6636,7 +6636,7 @@ define amdgpu_kernel void @flat_wavefront_seq_cst_seq_cst_cmpxchg(
 ; GFX1250-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
-  %gep = getelementptr i32, ptr %out, i32 4
+  %gep = getelementptr inbounds i32, ptr %out, i32 4
   %val = cmpxchg volatile ptr %gep, i32 %old, i32 %in syncscope("wavefront") seq_cst seq_cst
   ret void
 }
@@ -6935,7 +6935,7 @@ define amdgpu_kernel void @flat_wavefront_monotonic_monotonic_ret_cmpxchg(
 ; GFX1250-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
-  %gep = getelementptr i32, ptr %out, i32 4
+  %gep = getelementptr inbounds i32, ptr %out, i32 4
   %val = cmpxchg volatile ptr %gep, i32 %old, i32 %in syncscope("wavefront") monotonic monotonic
   %val0 = extractvalue { i32, i1 } %val, 0
   store i32 %val0, ptr %out, align 4
@@ -7236,7 +7236,7 @@ define amdgpu_kernel void @flat_wavefront_acquire_monotonic_ret_cmpxchg(
 ; GFX1250-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
-  %gep = getelementptr i32, ptr %out, i32 4
+  %gep = getelementptr inbounds i32, ptr %out, i32 4
   %val = cmpxchg volatile ptr %gep, i32 %old, i32 %in syncscope("wavefront") acquire monotonic
   %val0 = extractvalue { i32, i1 } %val, 0
   store i32 %val0, ptr %out, align 4
@@ -7537,7 +7537,7 @@ define amdgpu_kernel void @flat_wavefront_release_monotonic_ret_cmpxchg(
 ; GFX1250-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
-  %gep = getelementptr i32, ptr %out, i32 4
+  %gep = getelementptr inbounds i32, ptr %out, i32 4
   %val = cmpxchg volatile ptr %gep, i32 %old, i32 %in syncscope("wavefront") release monotonic
   %val0 = extractvalue { i32, i1 } %val, 0
   store i32 %val0, ptr %out, align 4
@@ -7838,7 +7838,7 @@ define amdgpu_kernel void @flat_wavefront_acq_rel_monotonic_ret_cmpxchg(
 ; GFX1250-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
-  %gep = getelementptr i32, ptr %out, i32 4
+  %gep = getelementptr inbounds i32, ptr %out, i32 4
   %val = cmpxchg volatile ptr %gep, i32 %old, i32 %in syncscope("wavefront") acq_rel monotonic
   %val0 = extractvalue { i32, i1 } %val, 0
   store i32 %val0, ptr %out, align 4
@@ -8139,7 +8139,7 @@ define amdgpu_kernel void @flat_wavefront_seq_cst_monotonic_ret_cmpxchg(
 ; GFX1250-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
-  %gep = getelementptr i32, ptr %out, i32 4
+  %gep = getelementptr inbounds i32, ptr %out, i32 4
   %val = cmpxchg volatile ptr %gep, i32 %old, i32 %in syncscope("wavefront") seq_cst monotonic
   %val0 = extractvalue { i32, i1 } %val, 0
   store i32 %val0, ptr %out, align 4
@@ -8440,7 +8440,7 @@ define amdgpu_kernel void @flat_wavefront_monotonic_acquire_ret_cmpxchg(
 ; GFX1250-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
-  %gep = getelementptr i32, ptr %out, i32 4
+  %gep = getelementptr inbounds i32, ptr %out, i32 4
   %val = cmpxchg volatile ptr %gep, i32 %old, i32 %in syncscope("wavefront") monotonic acquire
   %val0 = extractvalue { i32, i1 } %val, 0
   store i32 %val0, ptr %out, align 4
@@ -8741,7 +8741,7 @@ define amdgpu_kernel void @flat_wavefront_acquire_acquire_ret_cmpxchg(
 ; GFX1250-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
-  %gep = getelementptr i32, ptr %out, i32 4
+  %gep = getelementptr inbounds i32, ptr %out, i32 4
   %val = cmpxchg volatile ptr %gep, i32 %old, i32 %in syncscope("wavefront") acquire acquire
   %val0 = extractvalue { i32, i1 } %val, 0
   store i32 %val0, ptr %out, align 4
@@ -9042,7 +9042,7 @@ define amdgpu_kernel void @flat_wavefront_release_acquire_ret_cmpxchg(
 ; GFX1250-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
-  %gep = getelementptr i32, ptr %out, i32 4
+  %gep = getelementptr inbounds i32, ptr %out, i32 4
   %val = cmpxchg volatile ptr %gep, i32 %old, i32 %in syncscope("wavefront") release acquire
   %val0 = extractvalue { i32, i1 } %val, 0
   store i32 %val0, ptr %out, align 4
@@ -9343,7 +9343,7 @@ define amdgpu_kernel void @flat_wavefront_acq_rel_acquire_ret_cmpxchg(
 ; GFX1250-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
-  %gep = getelementptr i32, ptr %out, i32 4
+  %gep = getelementptr inbounds i32, ptr %out, i32 4
   %val = cmpxchg volatile ptr %gep, i32 %old, i32 %in syncscope("wavefront") acq_rel acquire
   %val0 = extractvalue { i32, i1 } %val, 0
   store i32 %val0, ptr %out, align 4
@@ -9644,7 +9644,7 @@ define amdgpu_kernel void @flat_wavefront_seq_cst_acquire_ret_cmpxchg(
 ; GFX1250-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
-  %gep = getelementptr i32, ptr %out, i32 4
+  %gep = getelementptr inbounds i32, ptr %out, i32 4
   %val = cmpxchg volatile ptr %gep, i32 %old, i32 %in syncscope("wavefront") seq_cst acquire
   %val0 = extractvalue { i32, i1 } %val, 0
   store i32 %val0, ptr %out, align 4
@@ -9945,7 +9945,7 @@ define amdgpu_kernel void @flat_wavefront_monotonic_seq_cst_ret_cmpxchg(
 ; GFX1250-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
-  %gep = getelementptr i32, ptr %out, i32 4
+  %gep = getelementptr inbounds i32, ptr %out, i32 4
   %val = cmpxchg volatile ptr %gep, i32 %old, i32 %in syncscope("wavefront") monotonic seq_cst
   %val0 = extractvalue { i32, i1 } %val, 0
   store i32 %val0, ptr %out, align 4
@@ -10246,7 +10246,7 @@ define amdgpu_kernel void @flat_wavefront_acquire_seq_cst_ret_cmpxchg(
 ; GFX1250-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
-  %gep = getelementptr i32, ptr %out, i32 4
+  %gep = getelementptr inbounds i32, ptr %out, i32 4
   %val = cmpxchg volatile ptr %gep, i32 %old, i32 %in syncscope("wavefront") acquire seq_cst
   %val0 = extractvalue { i32, i1 } %val, 0
   store i32 %val0, ptr %out, align 4
@@ -10547,7 +10547,7 @@ define amdgpu_kernel void @flat_wavefront_release_seq_cst_ret_cmpxchg(
 ; GFX1250-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
-  %gep = getelementptr i32, ptr %out, i32 4
+  %gep = getelementptr inbounds i32, ptr %out, i32 4
   %val = cmpxchg volatile ptr %gep, i32 %old, i32 %in syncscope("wavefront") release seq_cst
   %val0 = extractvalue { i32, i1 } %val, 0
   store i32 %val0, ptr %out, align 4
@@ -10848,7 +10848,7 @@ define amdgpu_kernel void @flat_wavefront_acq_rel_seq_cst_ret_cmpxchg(
 ; GFX1250-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
-  %gep = getelementptr i32, ptr %out, i32 4
+  %gep = getelementptr inbounds i32, ptr %out, i32 4
   %val = cmpxchg volatile ptr %gep, i32 %old, i32 %in syncscope("wavefront") acq_rel seq_cst
   %val0 = extractvalue { i32, i1 } %val, 0
   store i32 %val0, ptr %out, align 4
@@ -11149,7 +11149,7 @@ define amdgpu_kernel void @flat_wavefront_seq_cst_seq_cst_ret_cmpxchg(
 ; GFX1250-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
-  %gep = getelementptr i32, ptr %out, i32 4
+  %gep = getelementptr inbounds i32, ptr %out, i32 4
   %val = cmpxchg volatile ptr %gep, i32 %old, i32 %in syncscope("wavefront") seq_cst seq_cst
   %val0 = extractvalue { i32, i1 } %val, 0
   store i32 %val0, ptr %out, align 4
@@ -14237,7 +14237,7 @@ define amdgpu_kernel void @flat_wavefront_one_as_monotonic_monotonic_cmpxchg(
 ; GFX1250-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
-  %gep = getelementptr i32, ptr %out, i32 4
+  %gep = getelementptr inbounds i32, ptr %out, i32 4
   %val = cmpxchg volatile ptr %gep, i32 %old, i32 %in syncscope("wavefront-one-as") monotonic monotonic
   ret void
 }
@@ -14490,7 +14490,7 @@ define amdgpu_kernel void @flat_wavefront_one_as_acquire_monotonic_cmpxchg(
 ; GFX1250-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
-  %gep = getelementptr i32, ptr %out, i32 4
+  %gep = getelementptr inbounds i32, ptr %out, i32 4
   %val = cmpxchg volatile ptr %gep, i32 %old, i32 %in syncscope("wavefront-one-as") acquire monotonic
   ret void
 }
@@ -14743,7 +14743,7 @@ define amdgpu_kernel void @flat_wavefront_one_as_release_monotonic_cmpxchg(
 ; GFX1250-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
-  %gep = getelementptr i32, ptr %out, i32 4
+  %gep = getelementptr inbounds i32, ptr %out, i32 4
   %val = cmpxchg volatile ptr %gep, i32 %old, i32 %in syncscope("wavefront-one-as") release monotonic
   ret void
 }
@@ -14996,7 +14996,7 @@ define amdgpu_kernel void @flat_wavefront_one_as_acq_rel_monotonic_cmpxchg(
 ; GFX1250-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
-  %gep = getelementptr i32, ptr %out, i32 4
+  %gep = getelementptr inbounds i32, ptr %out, i32 4
   %val = cmpxchg volatile ptr %gep, i32 %old, i32 %in syncscope("wavefront-one-as") acq_rel monotonic
   ret void
 }
@@ -15249,7 +15249,7 @@ define amdgpu_kernel void @flat_wavefront_one_as_seq_cst_monotonic_cmpxchg(
 ; GFX1250-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
-  %gep = getelementptr i32, ptr %out, i32 4
+  %gep = getelementptr inbounds i32, ptr %out, i32 4
   %val = cmpxchg volatile ptr %gep, i32 %old, i32 %in syncscope("wavefront-one-as") seq_cst monotonic
   ret void
 }
@@ -15502,7 +15502,7 @@ define amdgpu_kernel void @flat_wavefront_one_as_monotonic_acquire_cmpxchg(
 ; GFX1250-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
-  %gep = getelementptr i32, ptr %out, i32 4
+  %gep = getelementptr inbounds i32, ptr %out, i32 4
   %val = cmpxchg volatile ptr %gep, i32 %old, i32 %in syncscope("wavefront-one-as") monotonic acquire
   ret void
 }
@@ -15755,7 +15755,7 @@ define amdgpu_kernel void @flat_wavefront_one_as_acquire_acquire_cmpxchg(
 ; GFX1250-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
-  %gep = getelementptr i32, ptr %out, i32 4
+  %gep = getelementptr inbounds i32, ptr %out, i32 4
   %val = cmpxchg volatile ptr %gep, i32 %old, i32 %in syncscope("wavefront-one-as") acquire acquire
   ret void
 }
@@ -16008,7 +16008,7 @@ define amdgpu_kernel void @flat_wavefront_one_as_release_acquire_cmpxchg(
 ; GFX1250-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
-  %gep = getelementptr i32, ptr %out, i32 4
+  %gep = getelementptr inbounds i32, ptr %out, i32 4
   %val = cmpxchg volatile ptr %gep, i32 %old, i32 %in syncscope("wavefront-one-as") release acquire
   ret void
 }
@@ -16261,7 +16261,7 @@ define amdgpu_kernel void @flat_wavefront_one_as_acq_rel_acquire_cmpxchg(
 ; GFX1250-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
-  %gep = getelementptr i32, ptr %out, i32 4
+  %gep = getelementptr inbounds i32, ptr %out, i32 4
   %val = cmpxchg volatile ptr %gep, i32 %old, i32 %in syncscope("wavefront-one-as") acq_rel acquire
   ret void
 }
@@ -16514,7 +16514,7 @@ define amdgpu_kernel void @flat_wavefront_one_as_seq_cst_acquire_cmpxchg(
 ; GFX1250-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
-  %gep = getelementptr i32, ptr %out, i32 4
+  %gep = getelementptr inbounds i32, ptr %out, i32 4
   %val = cmpxchg volatile ptr %gep, i32 %old, i32 %in syncscope("wavefront-one-as") seq_cst acquire
   ret void
 }
@@ -16767,7 +16767,7 @@ define amdgpu_kernel void @flat_wavefront_one_as_monotonic_seq_cst_cmpxchg(
 ; GFX1250-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
-  %gep = getelementptr i32, ptr %out, i32 4
+  %gep = getelementptr inbounds i32, ptr %out, i32 4
   %val = cmpxchg volatile ptr %gep, i32 %old, i32 %in syncscope("wavefront-one-as") monotonic seq_cst
   ret void
 }
@@ -17020,7 +17020,7 @@ define amdgpu_kernel void @flat_wavefront_one_as_acquire_seq_cst_cmpxchg(
 ; GFX1250-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
-  %gep = getelementptr i32, ptr %out, i32 4
+  %gep = getelementptr inbounds i32, ptr %out, i32 4
   %val = cmpxchg volatile ptr %gep, i32 %old, i32 %in syncscope("wavefront-one-as") acquire seq_cst
   ret void
 }
@@ -17273,7 +17273,7 @@ define amdgpu_kernel void @flat_wavefront_one_as_release_seq_cst_cmpxchg(
 ; GFX1250-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
-  %gep = getelementptr i32, ptr %out, i32 4
+  %gep = getelementptr inbounds i32, ptr %out, i32 4
   %val = cmpxchg volatile ptr %gep, i32 %old, i32 %in syncscope("wavefront-one-as") release seq_cst
   ret void
 }
@@ -17526,7 +17526,7 @@ define amdgpu_kernel void @flat_wavefront_one_as_acq_rel_seq_cst_cmpxchg(
 ; GFX1250-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
-  %gep = getelementptr i32, ptr %out, i32 4
+  %gep = getelementptr inbounds i32, ptr %out, i32 4
   %val = cmpxchg volatile ptr %gep, i32 %old, i32 %in syncscope("wavefront-one-as") acq_rel seq_cst
   ret void
 }
@@ -17779,7 +17779,7 @@ define amdgpu_kernel void @flat_wavefront_one_as_seq_cst_seq_cst_cmpxchg(
 ; GFX1250-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
-  %gep = getelementptr i32, ptr %out, i32 4
+  %gep = getelementptr inbounds i32, ptr %out, i32 4
   %val = cmpxchg volatile ptr %gep, i32 %old, i32 %in syncscope("wavefront-one-as") seq_cst seq_cst
   ret void
 }
@@ -18078,7 +18078,7 @@ define amdgpu_kernel void @flat_wavefront_one_as_monotonic_monotonic_ret_cmpxchg
 ; GFX1250-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
-  %gep = getelementptr i32, ptr %out, i32 4
+  %gep = getelementptr inbounds i32, ptr %out, i32 4
   %val = cmpxchg volatile ptr %gep, i32 %old, i32 %in syncscope("wavefront-one-as") monotonic monotonic
   %val0 = extractvalue { i32, i1 } %val, 0
   store i32 %val0, ptr %out, align 4
@@ -18379,7 +18379,7 @@ define amdgpu_kernel void @flat_wavefront_one_as_acquire_monotonic_ret_cmpxchg(
 ; GFX1250-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
-  %gep = getelementptr i32, ptr %out, i32 4
+  %gep = getelementptr inbounds i32, ptr %out, i32 4
   %val = cmpxchg volatile ptr %gep, i32 %old, i32 %in syncscope("wavefront-one-as") acquire monotonic
   %val0 = extractvalue { i32, i1 } %val, 0
   store i32 %val0, ptr %out, align 4
@@ -18680,7 +18680,7 @@ define amdgpu_kernel void @flat_wavefront_one_as_acq_rel_monotonic_ret_cmpxchg(
 ; GFX1250-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
-  %gep = getelementptr i32, ptr %out, i32 4
+  %gep = getelementptr inbounds i32, ptr %out, i32 4
   %val = cmpxchg volatile ptr %gep, i32 %old, i32 %in syncscope("wavefront-one-as") acq_rel monotonic
   %val0 = extractvalue { i32, i1 } %val, 0
   store i32 %val0, ptr %out, align 4
@@ -18981,7 +18981,7 @@ define amdgpu_kernel void @flat_wavefront_one_as_seq_cst_monotonic_ret_cmpxchg(
 ; GFX1250-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
-  %gep = getelementptr i32, ptr %out, i32 4
+  %gep = getelementptr inbounds i32, ptr %out, i32 4
   %val = cmpxchg volatile ptr %gep, i32 %old, i32 %in syncscope("wavefront-one-as") seq_cst monotonic
   %val0 = extractvalue { i32, i1 } %val, 0
   store i32 %val0, ptr %out, align 4
@@ -19282,7 +19282,7 @@ define amdgpu_kernel void @flat_wavefront_one_as_monotonic_acquire_ret_cmpxchg(
 ; GFX1250-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
-  %gep = getelementptr i32, ptr %out, i32 4
+  %gep = getelementptr inbounds i32, ptr %out, i32 4
   %val = cmpxchg volatile ptr %gep, i32 %old, i32 %in syncscope("wavefront-one-as") monotonic acquire
   %val0 = extractvalue { i32, i1 } %val, 0
   store i32 %val0, ptr %out, align 4
@@ -19583,7 +19583,7 @@ define amdgpu_kernel void @flat_wavefront_one_as_acquire_acquire_ret_cmpxchg(
 ; GFX1250-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
-  %gep = getelementptr i32, ptr %out, i32 4
+  %gep = getelementptr inbounds i32, ptr %out, i32 4
   %val = cmpxchg volatile ptr %gep, i32 %old, i32 %in syncscope("wavefront-one-as") acquire acquire
   %val0 = extractvalue { i32, i1 } %val, 0
   store i32 %val0, ptr %out, align 4
@@ -19884,7 +19884,7 @@ define amdgpu_kernel void @flat_wavefront_one_as_release_acquire_ret_cmpxchg(
 ; GFX1250-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
-  %gep = getelementptr i32, ptr %out, i32 4
+  %gep = getelementptr inbounds i32, ptr %out, i32 4
   %val = cmpxchg volatile ptr %gep, i32 %old, i32 %in syncscope("wavefront-one-as") release acquire
   %val0 = extractvalue { i32, i1 } %val, 0
   store i32 %val0, ptr %out, align 4
@@ -20185,7 +20185,7 @@ define amdgpu_kernel void @flat_wavefront_one_as_acq_rel_acquire_ret_cmpxchg(
 ; GFX1250-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
-  %gep = getelementptr i32, ptr %out, i32 4
+  %gep = getelementptr inbounds i32, ptr %out, i32 4
   %val = cmpxchg volatile ptr %gep, i32 %old, i32 %in syncscope("wavefront-one-as") acq_rel acquire
   %val0 = extractvalue { i32, i1 } %val, 0
   store i32 %val0, ptr %out, align 4
@@ -20486,7 +20486,7 @@ define amdgpu_kernel void @flat_wavefront_one_as_seq_cst_acquire_ret_cmpxchg(
 ; GFX1250-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
-  %gep = getelementptr i32, ptr %out, i32 4
+  %gep = getelementptr inbounds i32, ptr %out, i32 4
   %val = cmpxchg volatile ptr %gep, i32 %old, i32 %in syncscope("wavefront-one-as") seq_cst acquire
   %val0 = extractvalue { i32, i1 } %val, 0
   store i32 %val0, ptr %out, align 4
@@ -20787,7 +20787,7 @@ define amdgpu_kernel void @flat_wavefront_one_as_monotonic_seq_cst_ret_cmpxchg(
 ; GFX1250-NEXT:    s_endpgm
    ptr %out, i32 %in, i32 %old) {
 entry:
-  %gep = getelementptr i32, ptr %out, i32 4
+  %gep = getelementptr inbounds i32, ptr %out, i32 4
   %val = cmpxchg volatile ptr %gep, i32 %old, i32 %in syncscope("wavefront-one-as") monotonic seq_cst
   %val0 = extractvalue { i32, i1 } %val, 0
   store i32 %val0, ptr %out, align 4
@@ -21088,7 +21088,7 @@ define amdgpu_kernel void @flat_wavefront_one_as_acquire_seq_cst_ret_cmpxchg(
 ; GFX1250-NEXT:    s_endpgm
    ptr %out, i32 %in, i32 %old) {
 entry:
-  %gep = getelementptr i32, ptr %out, i32 4
+  %gep = getelementptr inbounds i32, ptr %out, i32 4
   %val = cmpxchg volatile ptr %gep, i32 %old, i32 %in syncscope("wavefront-one-as") acquire seq_cst
   %val0 = extractvalue { i32, i1 } %val, 0
   store i32 %val0, ptr %out, align 4
@@ -21389,7 +21389,7 @@ define amdgpu_kernel void @flat_wavefront_one_as_release_seq_cst_ret_cmpxchg(
 ; GFX1250-NEXT:    s_endpgm
    ptr %out, i32 %in, i32 %old) {
 entry:
-  %gep = getelementptr i32, ptr %out, i32 4
+  %gep = getelementptr inbounds i32, ptr %out, i32 4
   %val = cmpxchg volatile ptr %gep, i32 %old, i32 %in syncscope("wavefront-one-as") release seq_cst
   %val0 = extractvalue { i32, i1 } %val, 0
   store i32 %val0, ptr %out, align 4
@@ -21690,7 +21690,7 @@ define amdgpu_kernel void @flat_wavefront_one_as_acq_relc_seq_cst_ret_cmpxchg(
 ; GFX1250-NEXT:    s_endpgm
    ptr %out, i32 %in, i32 %old) {
 entry:
-  %gep = getelementptr i32, ptr %out, i32 4
+  %gep = getelementptr inbounds i32, ptr %out, i32 4
   %val = cmpxchg volatile ptr %gep, i32 %old, i32 %in syncscope("wavefront-one-as") acq_rel seq_cst
   %val0 = extractvalue { i32, i1 } %val, 0
   store i32 %val0, ptr %out, align 4
@@ -21991,7 +21991,7 @@ define amdgpu_kernel void @flat_wavefront_one_as_seq_cst_seq_cst_ret_cmpxchg(
 ; GFX1250-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
-  %gep = getelementptr i32, ptr %out, i32 4
+  %gep = getelementptr inbounds i32, ptr %out, i32 4
   %val = cmpxchg volatile ptr %gep, i32 %old, i32 %in syncscope("wavefront-one-as") seq_cst seq_cst
   %val0 = extractvalue { i32, i1 } %val, 0
   store i32 %val0, ptr %out, align 4
diff --git a/llvm/test/CodeGen/AMDGPU/memory-legalizer-flat-workgroup.ll b/llvm/test/CodeGen/AMDGPU/memory-legalizer-flat-workgroup.ll
index d384aec2a2b19..d44e7fff2359f 100644
--- a/llvm/test/CodeGen/AMDGPU/memory-legalizer-flat-workgroup.ll
+++ b/llvm/test/CodeGen/AMDGPU/memory-legalizer-flat-workgroup.ll
@@ -3401,7 +3401,7 @@ define amdgpu_kernel void @flat_workgroup_monotonic_monotonic_cmpxchg(
 ; GFX1250-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
-  %gep = getelementptr i32, ptr %out, i32 4
+  %gep = getelementptr inbounds i32, ptr %out, i32 4
   %val = cmpxchg volatile ptr %gep, i32 %old, i32 %in syncscope("workgroup") monotonic monotonic
   ret void
 }
@@ -3674,7 +3674,7 @@ define amdgpu_kernel void @flat_workgroup_acquire_monotonic_cmpxchg(
 ; GFX1250-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
-  %gep = getelementptr i32, ptr %out, i32 4
+  %gep = getelementptr inbounds i32, ptr %out, i32 4
   %val = cmpxchg volatile ptr %gep, i32 %old, i32 %in syncscope("workgroup") acquire monotonic
   ret void
 }
@@ -3951,7 +3951,7 @@ define amdgpu_kernel void @flat_workgroup_release_monotonic_cmpxchg(
 ; GFX1250-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
-  %gep = getelementptr i32, ptr %out, i32 4
+  %gep = getelementptr inbounds i32, ptr %out, i32 4
   %val = cmpxchg volatile ptr %gep, i32 %old, i32 %in syncscope("workgroup") release monotonic
   ret void
 }
@@ -4250,7 +4250,7 @@ define amdgpu_kernel void @flat_workgroup_acq_rel_monotonic_cmpxchg(
 ; GFX1250-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
-  %gep = getelementptr i32, ptr %out, i32 4
+  %gep = getelementptr inbounds i32, ptr %out, i32 4
   %val = cmpxchg volatile ptr %gep, i32 %old, i32 %in syncscope("workgroup") acq_rel monotonic
   ret void
 }
@@ -4549,7 +4549,7 @@ define amdgpu_kernel void @flat_workgroup_seq_cst_monotonic_cmpxchg(
 ; GFX1250-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
-  %gep = getelementptr i32, ptr %out, i32 4
+  %gep = getelementptr inbounds i32, ptr %out, i32 4
   %val = cmpxchg volatile ptr %gep, i32 %old, i32 %in syncscope("workgroup") seq_cst monotonic
   ret void
 }
@@ -4822,7 +4822,7 @@ define amdgpu_kernel void @flat_workgroup_monotonic_acquire_cmpxchg(
 ; GFX1250-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
-  %gep = getelementptr i32, ptr %out, i32 4
+  %gep = getelementptr inbounds i32, ptr %out, i32 4
   %val = cmpxchg volatile ptr %gep, i32 %old, i32 %in syncscope("workgroup") monotonic acquire
   ret void
 }
@@ -5095,7 +5095,7 @@ define amdgpu_kernel void @flat_workgroup_acquire_acquire_cmpxchg(
 ; GFX1250-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
-  %gep = getelementptr i32, ptr %out, i32 4
+  %gep = getelementptr inbounds i32, ptr %out, i32 4
   %val = cmpxchg volatile ptr %gep, i32 %old, i32 %in syncscope("workgroup") acquire acquire
   ret void
 }
@@ -5394,7 +5394,7 @@ define amdgpu_kernel void @flat_workgroup_release_acquire_cmpxchg(
 ; GFX1250-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
-  %gep = getelementptr i32, ptr %out, i32 4
+  %gep = getelementptr inbounds i32, ptr %out, i32 4
   %val = cmpxchg volatile ptr %gep, i32 %old, i32 %in syncscope("workgroup") release acquire
   ret void
 }
@@ -5693,7 +5693,7 @@ define amdgpu_kernel void @flat_workgroup_acq_rel_acquire_cmpxchg(
 ; GFX1250-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
-  %gep = getelementptr i32, ptr %out, i32 4
+  %gep = getelementptr inbounds i32, ptr %out, i32 4
   %val = cmpxchg volatile ptr %gep, i32 %old, i32 %in syncscope("workgroup") acq_rel acquire
   ret void
 }
@@ -5992,7 +5992,7 @@ define amdgpu_kernel void @flat_workgroup_seq_cst_acquire_cmpxchg(
 ; GFX1250-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
-  %gep = getelementptr i32, ptr %out, i32 4
+  %gep = getelementptr inbounds i32, ptr %out, i32 4
   %val = cmpxchg volatile ptr %gep, i32 %old, i32 %in syncscope("workgroup") seq_cst acquire
   ret void
 }
@@ -6291,7 +6291,7 @@ define amdgpu_kernel void @flat_workgroup_seq_cst_seq_cst_cmpxchg(
 ; GFX1250-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
-  %gep = getelementptr i32, ptr %out, i32 4
+  %gep = getelementptr inbounds i32, ptr %out, i32 4
   %val = cmpxchg volatile ptr %gep, i32 %old, i32 %in syncscope("workgroup") seq_cst seq_cst
   ret void
 }
@@ -6590,7 +6590,7 @@ define amdgpu_kernel void @flat_workgroup_monotonic_monotonic_ret_cmpxchg(
 ; GFX1250-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
-  %gep = getelementptr i32, ptr %out, i32 4
+  %gep = getelementptr inbounds i32, ptr %out, i32 4
   %val = cmpxchg volatile ptr %gep, i32 %old, i32 %in syncscope("workgroup") monotonic monotonic
   %val0 = extractvalue { i32, i1 } %val, 0
   store i32 %val0, ptr %out, align 4
@@ -6903,7 +6903,7 @@ define amdgpu_kernel void @flat_workgroup_acquire_monotonic_ret_cmpxchg(
 ; GFX1250-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
-  %gep = getelementptr i32, ptr %out, i32 4
+  %gep = getelementptr inbounds i32, ptr %out, i32 4
   %val = cmpxchg volatile ptr %gep, i32 %old, i32 %in syncscope("workgroup") acquire monotonic
   %val0 = extractvalue { i32, i1 } %val, 0
   store i32 %val0, ptr %out, align 4
@@ -7228,7 +7228,7 @@ define amdgpu_kernel void @flat_workgroup_release_monotonic_ret_cmpxchg(
 ; GFX1250-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
-  %gep = getelementptr i32, ptr %out, i32 4
+  %gep = getelementptr inbounds i32, ptr %out, i32 4
   %val = cmpxchg volatile ptr %gep, i32 %old, i32 %in syncscope("workgroup") release monotonic
   %val0 = extractvalue { i32, i1 } %val, 0
   store i32 %val0, ptr %out, align 4
@@ -7562,7 +7562,7 @@ define amdgpu_kernel void @flat_workgroup_acq_rel_monotonic_ret_cmpxchg(
 ; GFX1250-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
-  %gep = getelementptr i32, ptr %out, i32 4
+  %gep = getelementptr inbounds i32, ptr %out, i32 4
   %val = cmpxchg volatile ptr %gep, i32 %old, i32 %in syncscope("workgroup") acq_rel monotonic
   %val0 = extractvalue { i32, i1 } %val, 0
   store i32 %val0, ptr %out, align 4
@@ -7896,7 +7896,7 @@ define amdgpu_kernel void @flat_workgroup_seq_cst_monotonic_ret_cmpxchg(
 ; GFX1250-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
-  %gep = getelementptr i32, ptr %out, i32 4
+  %gep = getelementptr inbounds i32, ptr %out, i32 4
   %val = cmpxchg volatile ptr %gep, i32 %old, i32 %in syncscope("workgroup") seq_cst monotonic
   %val0 = extractvalue { i32, i1 } %val, 0
   store i32 %val0, ptr %out, align 4
@@ -8209,7 +8209,7 @@ define amdgpu_kernel void @flat_workgroup_monotonic_acquire_ret_cmpxchg(
 ; GFX1250-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
-  %gep = getelementptr i32, ptr %out, i32 4
+  %gep = getelementptr inbounds i32, ptr %out, i32 4
   %val = cmpxchg volatile ptr %gep, i32 %old, i32 %in syncscope("workgroup") monotonic acquire
   %val0 = extractvalue { i32, i1 } %val, 0
   store i32 %val0, ptr %out, align 4
@@ -8522,7 +8522,7 @@ define amdgpu_kernel void @flat_workgroup_acquire_acquire_ret_cmpxchg(
 ; GFX1250-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
-  %gep = getelementptr i32, ptr %out, i32 4
+  %gep = getelementptr inbounds i32, ptr %out, i32 4
   %val = cmpxchg volatile ptr %gep, i32 %old, i32 %in syncscope("workgroup") acquire acquire
   %val0 = extractvalue { i32, i1 } %val, 0
   store i32 %val0, ptr %out, align 4
@@ -8856,7 +8856,7 @@ define amdgpu_kernel void @flat_workgroup_release_acquire_ret_cmpxchg(
 ; GFX1250-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
-  %gep = getelementptr i32, ptr %out, i32 4
+  %gep = getelementptr inbounds i32, ptr %out, i32 4
   %val = cmpxchg volatile ptr %gep, i32 %old, i32 %in syncscope("workgroup") release acquire
   %val0 = extractvalue { i32, i1 } %val, 0
   store i32 %val0, ptr %out, align 4
@@ -9190,7 +9190,7 @@ define amdgpu_kernel void @flat_workgroup_acq_rel_acquire_ret_cmpxchg(
 ; GFX1250-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
-  %gep = getelementptr i32, ptr %out, i32 4
+  %gep = getelementptr inbounds i32, ptr %out, i32 4
   %val = cmpxchg volatile ptr %gep, i32 %old, i32 %in syncscope("workgroup") acq_rel acquire
   %val0 = extractvalue { i32, i1 } %val, 0
   store i32 %val0, ptr %out, align 4
@@ -9524,7 +9524,7 @@ define amdgpu_kernel void @flat_workgroup_seq_cst_acquire_ret_cmpxchg(
 ; GFX1250-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
-  %gep = getelementptr i32, ptr %out, i32 4
+  %gep = getelementptr inbounds i32, ptr %out, i32 4
   %val = cmpxchg volatile ptr %gep, i32 %old, i32 %in syncscope("workgroup") seq_cst acquire
   %val0 = extractvalue { i32, i1 } %val, 0
   store i32 %val0, ptr %out, align 4
@@ -9861,7 +9861,7 @@ define amdgpu_kernel void @flat_workgroup_monotonic_seq_cst_ret_cmpxchg(
 ; GFX1250-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
-  %gep = getelementptr i32, ptr %out, i32 4
+  %gep = getelementptr inbounds i32, ptr %out, i32 4
   %val = cmpxchg volatile ptr %gep, i32 %old, i32 %in syncscope("workgroup") monotonic seq_cst
   %val0 = extractvalue { i32, i1 } %val, 0
   store i32 %val0, ptr %out, align 4
@@ -10198,7 +10198,7 @@ define amdgpu_kernel void @flat_workgroup_acquire_seq_cst_ret_cmpxchg(
 ; GFX1250-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
-  %gep = getelementptr i32, ptr %out, i32 4
+  %gep = getelementptr inbounds i32, ptr %out, i32 4
   %val = cmpxchg volatile ptr %gep, i32 %old, i32 %in syncscope("workgroup") acquire seq_cst
   %val0 = extractvalue { i32, i1 } %val, 0
   store i32 %val0, ptr %out, align 4
@@ -10532,7 +10532,7 @@ define amdgpu_kernel void @flat_workgroup_release_seq_cst_ret_cmpxchg(
 ; GFX1250-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
-  %gep = getelementptr i32, ptr %out, i32 4
+  %gep = getelementptr inbounds i32, ptr %out, i32 4
   %val = cmpxchg volatile ptr %gep, i32 %old, i32 %in syncscope("workgroup") release seq_cst
   %val0 = extractvalue { i32, i1 } %val, 0
   store i32 %val0, ptr %out, align 4
@@ -10866,7 +10866,7 @@ define amdgpu_kernel void @flat_workgroup_acq_rel_seq_cst_ret_cmpxchg(
 ; GFX1250-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
-  %gep = getelementptr i32, ptr %out, i32 4
+  %gep = getelementptr inbounds i32, ptr %out, i32 4
   %val = cmpxchg volatile ptr %gep, i32 %old, i32 %in syncscope("workgroup") acq_rel seq_cst
   %val0 = extractvalue { i32, i1 } %val, 0
   store i32 %val0, ptr %out, align 4
@@ -11200,7 +11200,7 @@ define amdgpu_kernel void @flat_workgroup_seq_cst_seq_cst_ret_cmpxchg(
 ; GFX1250-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
-  %gep = getelementptr i32, ptr %out, i32 4
+  %gep = getelementptr inbounds i32, ptr %out, i32 4
   %val = cmpxchg volatile ptr %gep, i32 %old, i32 %in syncscope("workgroup") seq_cst seq_cst
   %val0 = extractvalue { i32, i1 } %val, 0
   store i32 %val0, ptr %out, align 4
@@ -14536,7 +14536,7 @@ define amdgpu_kernel void @flat_workgroup_one_as_monotonic_monotonic_cmpxchg(
 ; GFX1250-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
-  %gep = getelementptr i32, ptr %out, i32 4
+  %gep = getelementptr inbounds i32, ptr %out, i32 4
   %val = cmpxchg volatile ptr %gep, i32 %old, i32 %in syncscope("workgroup-one-as") monotonic monotonic
   ret void
 }
@@ -14800,7 +14800,7 @@ define amdgpu_kernel void @flat_workgroup_one_as_acquire_monotonic_cmpxchg(
 ; GFX1250-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
-  %gep = getelementptr i32, ptr %out, i32 4
+  %gep = getelementptr inbounds i32, ptr %out, i32 4
   %val = cmpxchg volatile ptr %gep, i32 %old, i32 %in syncscope("workgroup-one-as") acquire monotonic
   ret void
 }
@@ -15073,7 +15073,7 @@ define amdgpu_kernel void @flat_workgroup_one_as_release_monotonic_cmpxchg(
 ; GFX1250-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
-  %gep = getelementptr i32, ptr %out, i32 4
+  %gep = getelementptr inbounds i32, ptr %out, i32 4
   %val = cmpxchg volatile ptr %gep, i32 %old, i32 %in syncscope("workgroup-one-as") release monotonic
   ret void
 }
@@ -15360,7 +15360,7 @@ define amdgpu_kernel void @flat_workgroup_one_as_acq_rel_monotonic_cmpxchg(
 ; GFX1250-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
-  %gep = getelementptr i32, ptr %out, i32 4
+  %gep = getelementptr inbounds i32, ptr %out, i32 4
   %val = cmpxchg volatile ptr %gep, i32 %old, i32 %in syncscope("workgroup-one-as") acq_rel monotonic
   ret void
 }
@@ -15647,7 +15647,7 @@ define amdgpu_kernel void @flat_workgroup_one_as_seq_cst_monotonic_cmpxchg(
 ; GFX1250-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
-  %gep = getelementptr i32, ptr %out, i32 4
+  %gep = getelementptr inbounds i32, ptr %out, i32 4
   %val = cmpxchg volatile ptr %gep, i32 %old, i32 %in syncscope("workgroup-one-as") seq_cst monotonic
   ret void
 }
@@ -15911,7 +15911,7 @@ define amdgpu_kernel void @flat_workgroup_one_as_monotonic_acquire_cmpxchg(
 ; GFX1250-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
-  %gep = getelementptr i32, ptr %out, i32 4
+  %gep = getelementptr inbounds i32, ptr %out, i32 4
   %val = cmpxchg volatile ptr %gep, i32 %old, i32 %in syncscope("workgroup-one-as") monotonic acquire
   ret void
 }
@@ -16175,7 +16175,7 @@ define amdgpu_kernel void @flat_workgroup_one_as_acquire_acquire_cmpxchg(
 ; GFX1250-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
-  %gep = getelementptr i32, ptr %out, i32 4
+  %gep = getelementptr inbounds i32, ptr %out, i32 4
   %val = cmpxchg volatile ptr %gep, i32 %old, i32 %in syncscope("workgroup-one-as") acquire acquire
   ret void
 }
@@ -16462,7 +16462,7 @@ define amdgpu_kernel void @flat_workgroup_one_as_release_acquire_cmpxchg(
 ; GFX1250-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
-  %gep = getelementptr i32, ptr %out, i32 4
+  %gep = getelementptr inbounds i32, ptr %out, i32 4
   %val = cmpxchg volatile ptr %gep, i32 %old, i32 %in syncscope("workgroup-one-as") release acquire
   ret void
 }
@@ -16749,7 +16749,7 @@ define amdgpu_kernel void @flat_workgroup_one_as_acq_rel_acquire_cmpxchg(
 ; GFX1250-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
-  %gep = getelementptr i32, ptr %out, i32 4
+  %gep = getelementptr inbounds i32, ptr %out, i32 4
   %val = cmpxchg volatile ptr %gep, i32 %old, i32 %in syncscope("workgroup-one-as") acq_rel acquire
   ret void
 }
@@ -17036,7 +17036,7 @@ define amdgpu_kernel void @flat_workgroup_one_as_seq_cst_acquire_cmpxchg(
 ; GFX1250-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
-  %gep = getelementptr i32, ptr %out, i32 4
+  %gep = getelementptr inbounds i32, ptr %out, i32 4
   %val = cmpxchg volatile ptr %gep, i32 %old, i32 %in syncscope("workgroup-one-as") seq_cst acquire
   ret void
 }
@@ -17320,7 +17320,7 @@ define amdgpu_kernel void @flat_workgroup_one_as_monotonic_seq_cst_cmpxchg(
 ; GFX1250-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
-  %gep = getelementptr i32, ptr %out, i32 4
+  %gep = getelementptr inbounds i32, ptr %out, i32 4
   %val = cmpxchg volatile ptr %gep, i32 %old, i32 %in syncscope("workgroup-one-as") monotonic seq_cst
   ret void
 }
@@ -17604,7 +17604,7 @@ define amdgpu_kernel void @flat_workgroup_one_as_acquire_seq_cst_cmpxchg(
 ; GFX1250-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
-  %gep = getelementptr i32, ptr %out, i32 4
+  %gep = getelementptr inbounds i32, ptr %out, i32 4
   %val = cmpxchg volatile ptr %gep, i32 %old, i32 %in syncscope("workgroup-one-as") acquire seq_cst
   ret void
 }
@@ -17891,7 +17891,7 @@ define amdgpu_kernel void @flat_workgroup_one_as_release_seq_cst_cmpxchg(
 ; GFX1250-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
-  %gep = getelementptr i32, ptr %out, i32 4
+  %gep = getelementptr inbounds i32, ptr %out, i32 4
   %val = cmpxchg volatile ptr %gep, i32 %old, i32 %in syncscope("workgroup-one-as") release seq_cst
   ret void
 }
@@ -18178,7 +18178,7 @@ define amdgpu_kernel void @flat_workgroup_one_as_acq_rel_seq_cst_cmpxchg(
 ; GFX1250-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
-  %gep = getelementptr i32, ptr %out, i32 4
+  %gep = getelementptr inbounds i32, ptr %out, i32 4
   %val = cmpxchg volatile ptr %gep, i32 %old, i32 %in syncscope("workgroup-one-as") acq_rel seq_cst
   ret void
 }
@@ -18465,7 +18465,7 @@ define amdgpu_kernel void @flat_workgroup_one_as_seq_cst_seq_cst_cmpxchg(
 ; GFX1250-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
-  %gep = getelementptr i32, ptr %out, i32 4
+  %gep = getelementptr inbounds i32, ptr %out, i32 4
   %val = cmpxchg volatile ptr %gep, i32 %old, i32 %in syncscope("workgroup-one-as") seq_cst seq_cst
   ret void
 }
@@ -18764,7 +18764,7 @@ define amdgpu_kernel void @flat_workgroup_one_as_monotonicmonotonic_ret_cmpxchg(
 ; GFX1250-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
-  %gep = getelementptr i32, ptr %out, i32 4
+  %gep = getelementptr inbounds i32, ptr %out, i32 4
   %val = cmpxchg volatile ptr %gep, i32 %old, i32 %in syncscope("workgroup-one-as") monotonic monotonic
   %val0 = extractvalue { i32, i1 } %val, 0
   store i32 %val0, ptr %out, align 4
@@ -19073,7 +19073,7 @@ define amdgpu_kernel void @flat_workgroup_one_as_acquire_monotonic_ret_cmpxchg(
 ; GFX1250-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
-  %gep = getelementptr i32, ptr %out, i32 4
+  %gep = getelementptr inbounds i32, ptr %out, i32 4
   %val = cmpxchg volatile ptr %gep, i32 %old, i32 %in syncscope("workgroup-one-as") acquire monotonic
   %val0 = extractvalue { i32, i1 } %val, 0
   store i32 %val0, ptr %out, align 4
@@ -19394,7 +19394,7 @@ define amdgpu_kernel void @flat_workgroup_one_as_release_monotonic_ret_cmpxchg(
 ; GFX1250-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
-  %gep = getelementptr i32, ptr %out, i32 4
+  %gep = getelementptr inbounds i32, ptr %out, i32 4
   %val = cmpxchg volatile ptr %gep, i32 %old, i32 %in syncscope("workgroup-one-as") release monotonic
   %val0 = extractvalue { i32, i1 } %val, 0
   store i32 %val0, ptr %out, align 4
@@ -19726,7 +19726,7 @@ define amdgpu_kernel void @flat_workgroup_one_as_acq_rel_monotonic_ret_cmpxchg(
 ; GFX1250-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
-  %gep = getelementptr i32, ptr %out, i32 4
+  %gep = getelementptr inbounds i32, ptr %out, i32 4
   %val = cmpxchg volatile ptr %gep, i32 %old, i32 %in syncscope("workgroup-one-as") acq_rel monotonic
   %val0 = extractvalue { i32, i1 } %val, 0
   store i32 %val0, ptr %out, align 4
@@ -20058,7 +20058,7 @@ define amdgpu_kernel void @flat_workgroup_one_as_seq_cst_monotonic_ret_cmpxchg(
 ; GFX1250-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
-  %gep = getelementptr i32, ptr %out, i32 4
+  %gep = getelementptr inbounds i32, ptr %out, i32 4
   %val = cmpxchg volatile ptr %gep, i32 %old, i32 %in syncscope("workgroup-one-as") seq_cst monotonic
   %val0 = extractvalue { i32, i1 } %val, 0
   store i32 %val0, ptr %out, align 4
@@ -20367,7 +20367,7 @@ define amdgpu_kernel void @flat_workgroup_one_as_monotonic_acquire_ret_cmpxchg(
 ; GFX1250-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
-  %gep = getelementptr i32, ptr %out, i32 4
+  %gep = getelementptr inbounds i32, ptr %out, i32 4
   %val = cmpxchg volatile ptr %gep, i32 %old, i32 %in syncscope("workgroup-one-as") monotonic acquire
   %val0 = extractvalue { i32, i1 } %val, 0
   store i32 %val0, ptr %out, align 4
@@ -20676,7 +20676,7 @@ define amdgpu_kernel void @flat_workgroup_one_as_acquire_acquire_ret_cmpxchg(
 ; GFX1250-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
-  %gep = getelementptr i32, ptr %out, i32 4
+  %gep = getelementptr inbounds i32, ptr %out, i32 4
   %val = cmpxchg volatile ptr %gep, i32 %old, i32 %in syncscope("workgroup-one-as") acquire acquire
   %val0 = extractvalue { i32, i1 } %val, 0
   store i32 %val0, ptr %out, align 4
@@ -21008,7 +21008,7 @@ define amdgpu_kernel void @flat_workgroup_one_as_release_acquire_ret_cmpxchg(
 ; GFX1250-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
-  %gep = getelementptr i32, ptr %out, i32 4
+  %gep = getelementptr inbounds i32, ptr %out, i32 4
   %val = cmpxchg volatile ptr %gep, i32 %old, i32 %in syncscope("workgroup-one-as") release acquire
   %val0 = extractvalue { i32, i1 } %val, 0
   store i32 %val0, ptr %out, align 4
@@ -21340,7 +21340,7 @@ define amdgpu_kernel void @flat_workgroup_one_as_acq_rel_acquire_ret_cmpxchg(
 ; GFX1250-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
-  %gep = getelementptr i32, ptr %out, i32 4
+  %gep = getelementptr inbounds i32, ptr %out, i32 4
   %val = cmpxchg volatile ptr %gep, i32 %old, i32 %in syncscope("workgroup-one-as") acq_rel acquire
   %val0 = extractvalue { i32, i1 } %val, 0
   store i32 %val0, ptr %out, align 4
@@ -21672,7 +21672,7 @@ define amdgpu_kernel void @flat_workgroup_one_as_seq_cst_acquire_ret_cmpxchg(
 ; GFX1250-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
-  %gep = getelementptr i32, ptr %out, i32 4
+  %gep = getelementptr inbounds i32, ptr %out, i32 4
   %val = cmpxchg volatile ptr %gep, i32 %old, i32 %in syncscope("workgroup-one-as") seq_cst acquire
   %val0 = extractvalue { i32, i1 } %val, 0
   store i32 %val0, ptr %out, align 4
@@ -22001,7 +22001,7 @@ define amdgpu_kernel void @flat_workgroup_one_as_monotonic_seq_cst_ret_cmpxchg(
 ; GFX1250-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
-  %gep = getelementptr i32, ptr %out, i32 4
+  %gep = getelementptr inbounds i32, ptr %out, i32 4
   %val = cmpxchg volatile ptr %gep, i32 %old, i32 %in syncscope("workgroup-one-as") monotonic seq_cst
   %val0 = extractvalue { i32, i1 } %val, 0
   store i32 %val0, ptr %out, align 4
@@ -22330,7 +22330,7 @@ define amdgpu_kernel void @flat_workgroup_one_as_acquire_seq_cst_ret_cmpxchg(
 ; GFX1250-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
-  %gep = getelementptr i32, ptr %out, i32 4
+  %gep = getelementptr inbounds i32, ptr %out, i32 4
   %val = cmpxchg volatile ptr %gep, i32 %old, i32 %in syncscope("workgroup-one-as") acquire seq_cst
   %val0 = extractvalue { i32, i1 } %val, 0
   store i32 %val0, ptr %out, align 4
@@ -22662,7 +22662,7 @@ define amdgpu_kernel void @flat_workgroup_one_as_release_seq_cst_ret_cmpxchg(
 ; GFX1250-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
-  %gep = getelementptr i32, ptr %out, i32 4
+  %gep = getelementptr inbounds i32, ptr %out, i32 4
   %val = cmpxchg volatile ptr %gep, i32 %old, i32 %in syncscope("workgroup-one-as") release seq_cst
   %val0 = extractvalue { i32, i1 } %val, 0
   store i32 %val0, ptr %out, align 4
@@ -22994,7 +22994,7 @@ define amdgpu_kernel void @flat_workgroup_one_as_acq_rel_seq_cst_ret_cmpxchg(
 ; GFX1250-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
-  %gep = getelementptr i32, ptr %out, i32 4
+  %gep = getelementptr inbounds i32, ptr %out, i32 4
   %val = cmpxchg volatile ptr %gep, i32 %old, i32 %in syncscope("workgroup-one-as") acq_rel seq_cst
   %val0 = extractvalue { i32, i1 } %val, 0
   store i32 %val0, ptr %out, align 4
@@ -23326,7 +23326,7 @@ define amdgpu_kernel void @flat_workgroup_one_as_seq_cst_seq_cst_ret_cmpxchg(
 ; GFX1250-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
-  %gep = getelementptr i32, ptr %out, i32 4
+  %gep = getelementptr inbounds i32, ptr %out, i32 4
   %val = cmpxchg volatile ptr %gep, i32 %old, i32 %in syncscope("workgroup-one-as") seq_cst seq_cst
   %val0 = extractvalue { i32, i1 } %val, 0
   store i32 %val0, ptr %out, align 4
diff --git a/llvm/test/CodeGen/AMDGPU/offset-split-flat.ll b/llvm/test/CodeGen/AMDGPU/offset-split-flat.ll
index e50ed3ee95140..4d189d7a9673d 100644
--- a/llvm/test/CodeGen/AMDGPU/offset-split-flat.ll
+++ b/llvm/test/CodeGen/AMDGPU/offset-split-flat.ll
@@ -1492,7 +1492,7 @@ define i8 @flat_inst_valu_offset_2x_neg_24bit_max(ptr %p) {
 ; GFX12-GISEL-NEXT:    flat_load_u8 v0, v[0:1]
 ; GFX12-GISEL-NEXT:    s_wait_loadcnt_dscnt 0x0
 ; GFX12-GISEL-NEXT:    s_setpc_b64 s[30:31]
-  %gep = getelementptr i8, ptr %p, i64 -16777215
+  %gep = getelementptr inbounds i8, ptr %p, i64 -16777215
   %load = load i8, ptr %gep, align 4
   ret i8 %load
 }
@@ -1597,7 +1597,7 @@ define i8 @flat_inst_valu_offset_64bit_11bit_split0(ptr %p) {
 ; GFX12-GISEL-NEXT:    flat_load_u8 v0, v[0:1]
 ; GFX12-GISEL-NEXT:    s_wait_loadcnt_dscnt 0x0
 ; GFX12-GISEL-NEXT:    s_setpc_b64 s[30:31]
-  %gep = getelementptr i8, ptr %p, i64 8589936639
+  %gep = getelementptr inbounds i8, ptr %p, i64 8589936639
   %load = load i8, ptr %gep, align 4
   ret i8 %load
 }
@@ -1702,7 +1702,7 @@ define i8 @flat_inst_valu_offset_64bit_11bit_split1(ptr %p) {
 ; GFX12-GISEL-NEXT:    flat_load_u8 v0, v[0:1]
 ; GFX12-GISEL-NEXT:    s_wait_loadcnt_dscnt 0x0
 ; GFX12-GISEL-NEXT:    s_setpc_b64 s[30:31]
-  %gep = getelementptr i8, ptr %p, i64 8589936640
+  %gep = getelementptr inbounds i8, ptr %p, i64 8589936640
   %load = load i8, ptr %gep, align 4
   ret i8 %load
 }
@@ -1807,7 +1807,7 @@ define i8 @flat_inst_valu_offset_64bit_12bit_split0(ptr %p) {
 ; GFX12-GISEL-NEXT:    flat_load_u8 v0, v[0:1]
 ; GFX12-GISEL-NEXT:    s_wait_loadcnt_dscnt 0x0
 ; GFX12-GISEL-NEXT:    s_setpc_b64 s[30:31]
-  %gep = getelementptr i8, ptr %p, i64 8589938687
+  %gep = getelementptr inbounds i8, ptr %p, i64 8589938687
   %load = load i8, ptr %gep, align 4
   ret i8 %load
 }
@@ -1903,7 +1903,7 @@ define i8 @flat_inst_valu_offset_64bit_12bit_split1(ptr %p) {
 ; GFX12-GISEL-NEXT:    flat_load_u8 v0, v[0:1]
 ; GFX12-GISEL-NEXT:    s_wait_loadcnt_dscnt 0x0
 ; GFX12-GISEL-NEXT:    s_setpc_b64 s[30:31]
-  %gep = getelementptr i8, ptr %p, i64 8589938688
+  %gep = getelementptr inbounds i8, ptr %p, i64 8589938688
   %load = load i8, ptr %gep, align 4
   ret i8 %load
 }
@@ -2008,7 +2008,7 @@ define i8 @flat_inst_valu_offset_64bit_13bit_split0(ptr %p) {
 ; GFX12-GISEL-NEXT:    flat_load_u8 v0, v[0:1]
 ; GFX12-GISEL-NEXT:    s_wait_loadcnt_dscnt 0x0
 ; GFX12-GISEL-NEXT:    s_setpc_b64 s[30:31]
-  %gep = getelementptr i8, ptr %p, i64 8589942783
+  %gep = getelementptr inbounds i8, ptr %p, i64 8589942783
   %load = load i8, ptr %gep, align 4
   ret i8 %load
 }
@@ -2104,7 +2104,7 @@ define i8 @flat_inst_valu_offset_64bit_13bit_split1(ptr %p) {
 ; GFX12-GISEL-NEXT:    flat_load_u8 v0, v[0:1]
 ; GFX12-GISEL-NEXT:    s_wait_loadcnt_dscnt 0x0
 ; GFX12-GISEL-NEXT:    s_setpc_b64 s[30:31]
-  %gep = getelementptr i8, ptr %p, i64 8589942784
+  %gep = getelementptr inbounds i8, ptr %p, i64 8589942784
   %load = load i8, ptr %gep, align 4
   ret i8 %load
 }
@@ -2211,7 +2211,7 @@ define i8 @flat_inst_valu_offset_64bit_11bit_neg_high_split0(ptr %p) {
 ; GFX12-GISEL-NEXT:    flat_load_u8 v0, v[0:1]
 ; GFX12-GISEL-NEXT:    s_wait_loadcnt_dscnt 0x0
 ; GFX12-GISEL-NEXT:    s_setpc_b64 s[30:31]
-  %gep = getelementptr i8, ptr %p, i64 -9223372036854773761
+  %gep = getelementptr inbounds i8, ptr %p, i64 -9223372036854773761
   %load = load i8, ptr %gep, align 4
   ret i8 %load
 }
@@ -2318,7 +2318,7 @@ define i8 @flat_inst_valu_offset_64bit_11bit_neg_high_split1(ptr %p) {
 ; GFX12-GISEL-NEXT:    flat_load_u8 v0, v[0:1]
 ; GFX12-GISEL-NEXT:    s_wait_loadcnt_dscnt 0x0
 ; GFX12-GISEL-NEXT:    s_setpc_b64 s[30:31]
-  %gep = getelementptr i8, ptr %p, i64 -9223372036854773760
+  %gep = getelementptr inbounds i8, ptr %p, i64 -9223372036854773760
   %load = load i8, ptr %gep, align 4
   ret i8 %load
 }
@@ -2425,7 +2425,7 @@ define i8 @flat_inst_valu_offset_64bit_12bit_neg_high_split0(ptr %p) {
 ; GFX12-GISEL-NEXT:    flat_load_u8 v0, v[0:1]
 ; GFX12-GISEL-NEXT:    s_wait_loadcnt_dscnt 0x0
 ; GFX12-GISEL-NEXT:    s_setpc_b64 s[30:31]
-  %gep = getelementptr i8, ptr %p, i64 -9223372036854771713
+  %gep = getelementptr inbounds i8, ptr %p, i64 -9223372036854771713
   %load = load i8, ptr %gep, align 4
   ret i8 %load
 }
@@ -2532,7 +2532,7 @@ define i8 @flat_inst_valu_offset_64bit_12bit_neg_high_split1(ptr %p) {
 ; GFX12-GISEL-NEXT:    flat_load_u8 v0, v[0:1]
 ; GFX12-GISEL-NEXT:    s_wait_loadcnt_dscnt 0x0
 ; GFX12-GISEL-NEXT:    s_setpc_b64 s[30:31]
-  %gep = getelementptr i8, ptr %p, i64 -9223372036854771712
+  %gep = getelementptr inbounds i8, ptr %p, i64 -9223372036854771712
   %load = load i8, ptr %gep, align 4
   ret i8 %load
 }
@@ -2639,7 +2639,7 @@ define i8 @flat_inst_valu_offset_64bit_13bit_neg_high_split0(ptr %p) {
 ; GFX12-GISEL-NEXT:    flat_load_u8 v0, v[0:1]
 ; GFX12-GISEL-NEXT:    s_wait_loadcnt_dscnt 0x0
 ; GFX12-GISEL-NEXT:    s_setpc_b64 s[30:31]
-  %gep = getelementptr i8, ptr %p, i64 -9223372036854767617
+  %gep = getelementptr inbounds i8, ptr %p, i64 -9223372036854767617
   %load = load i8, ptr %gep, align 4
   ret i8 %load
 }
@@ -2746,7 +2746,7 @@ define i8 @flat_inst_valu_offset_64bit_13bit_neg_high_split1(ptr %p) {
 ; GFX12-GISEL-NEXT:    flat_load_u8 v0, v[0:1]
 ; GFX12-GISEL-NEXT:    s_wait_loadcnt_dscnt 0x0
 ; GFX12-GISEL-NEXT:    s_setpc_b64 s[30:31]
-  %gep = getelementptr i8, ptr %p, i64 -9223372036854767616
+  %gep = getelementptr inbounds i8, ptr %p, i64 -9223372036854767616
   %load = load i8, ptr %gep, align 4
   ret i8 %load
 }
@@ -4232,7 +4232,7 @@ define amdgpu_kernel void @flat_inst_salu_offset_64bit_11bit_split0(ptr %p) {
 ; GFX12-GISEL-NEXT:    s_wait_loadcnt_dscnt 0x0
 ; GFX12-GISEL-NEXT:    flat_store_b8 v[0:1], v0
 ; GFX12-GISEL-NEXT:    s_endpgm
-  %gep = getelementptr i8, ptr %p, i64 8589936639
+  %gep = getelementptr inbounds i8, ptr %p, i64 8589936639
   %load = load volatile i8, ptr %gep, align 1
   store i8 %load, ptr poison
   ret void
@@ -4351,7 +4351,7 @@ define amdgpu_kernel void @flat_inst_salu_offset_64bit_11bit_split1(ptr %p) {
 ; GFX12-GISEL-NEXT:    s_wait_loadcnt_dscnt 0x0
 ; GFX12-GISEL-NEXT:    flat_store_b8 v[0:1], v0
 ; GFX12-GISEL-NEXT:    s_endpgm
-  %gep = getelementptr i8, ptr %p, i64 8589936640
+  %gep = getelementptr inbounds i8, ptr %p, i64 8589936640
   %load = load volatile i8, ptr %gep, align 1
   store i8 %load, ptr poison
   ret void
@@ -4470,7 +4470,7 @@ define amdgpu_kernel void @flat_inst_salu_offset_64bit_12bit_split0(ptr %p) {
 ; GFX12-GISEL-NEXT:    s_wait_loadcnt_dscnt 0x0
 ; GFX12-GISEL-NEXT:    flat_store_b8 v[0:1], v0
 ; GFX12-GISEL-NEXT:    s_endpgm
-  %gep = getelementptr i8, ptr %p, i64 8589938687
+  %gep = getelementptr inbounds i8, ptr %p, i64 8589938687
   %load = load volatile i8, ptr %gep, align 1
   store i8 %load, ptr poison
   ret void
@@ -4590,7 +4590,7 @@ define amdgpu_kernel void @flat_inst_salu_offset_64bit_12bit_split1(ptr %p) {
 ; GFX12-GISEL-NEXT:    s_wait_loadcnt_dscnt 0x0
 ; GFX12-GISEL-NEXT:    flat_store_b8 v[0:1], v0
 ; GFX12-GISEL-NEXT:    s_endpgm
-  %gep = getelementptr i8, ptr %p, i64 8589938688
+  %gep = getelementptr inbounds i8, ptr %p, i64 8589938688
   %load = load volatile i8, ptr %gep, align 1
   store i8 %load, ptr poison
   ret void
@@ -4710,7 +4710,7 @@ define amdgpu_kernel void @flat_inst_salu_offset_64bit_13bit_split0(ptr %p) {
 ; GFX12-GISEL-NEXT:    s_wait_loadcnt_dscnt 0x0
 ; GFX12-GISEL-NEXT:    flat_store_b8 v[0:1], v0
 ; GFX12-GISEL-NEXT:    s_endpgm
-  %gep = getelementptr i8, ptr %p, i64 8589942783
+  %gep = getelementptr inbounds i8, ptr %p, i64 8589942783
   %load = load volatile i8, ptr %gep, align 1
   store i8 %load, ptr poison
   ret void
@@ -4830,7 +4830,7 @@ define amdgpu_kernel void @flat_inst_salu_offset_64bit_13bit_split1(ptr %p) {
 ; GFX12-GISEL-NEXT:    s_wait_loadcnt_dscnt 0x0
 ; GFX12-GISEL-NEXT:    flat_store_b8 v[0:1], v0
 ; GFX12-GISEL-NEXT:    s_endpgm
-  %gep = getelementptr i8, ptr %p, i64 8589942784
+  %gep = getelementptr inbounds i8, ptr %p, i64 8589942784
   %load = load volatile i8, ptr %gep, align 1
   store i8 %load, ptr poison
   ret void
@@ -4955,7 +4955,7 @@ define amdgpu_kernel void @flat_inst_salu_offset_64bit_11bit_neg_high_split0(ptr
 ; GFX12-GISEL-NEXT:    s_wait_loadcnt_dscnt 0x0
 ; GFX12-GISEL-NEXT:    flat_store_b8 v[0:1], v0
 ; GFX12-GISEL-NEXT:    s_endpgm
-  %gep = getelementptr i8, ptr %p, i64 -9223372036854773761
+  %gep = getelementptr inbounds i8, ptr %p, i64 -9223372036854773761
   %load = load volatile i8, ptr %gep, align 1
   store i8 %load, ptr poison
   ret void
@@ -5080,7 +5080,7 @@ define amdgpu_kernel void @flat_inst_salu_offset_64bit_11bit_neg_high_split1(ptr
 ; GFX12-GISEL-NEXT:    s_wait_loadcnt_dscnt 0x0
 ; GFX12-GISEL-NEXT:    flat_store_b8 v[0:1], v0
 ; GFX12-GISEL-NEXT:    s_endpgm
-  %gep = getelementptr i8, ptr %p, i64 -9223372036854773760
+  %gep = getelementptr inbounds i8, ptr %p, i64 -9223372036854773760
   %load = load volatile i8, ptr %gep, align 1
   store i8 %load, ptr poison
   ret void
@@ -5205,7 +5205,7 @@ define amdgpu_kernel void @flat_inst_salu_offset_64bit_12bit_neg_high_split0(ptr
 ; GFX12-GISEL-NEXT:    s_wait_loadcnt_dscnt 0x0
 ; GFX12-GISEL-NEXT:    flat_store_b8 v[0:1], v0
 ; GFX12-GISEL-NEXT:    s_endpgm
-  %gep = getelementptr i8, ptr %p, i64 -9223372036854771713
+  %gep = getelementptr inbounds i8, ptr %p, i64 -9223372036854771713
   %load = load volatile i8, ptr %gep, align 1
   store i8 %load, ptr poison
   ret void
@@ -5330,7 +5330,7 @@ define amdgpu_kernel void @flat_inst_salu_offset_64bit_12bit_neg_high_split1(ptr
 ; GFX12-GISEL-NEXT:    s_wait_loadcnt_dscnt 0x0
 ; GFX12-GISEL-NEXT:    flat_store_b8 v[0:1], v0
 ; GFX12-GISEL-NEXT:    s_endpgm
-  %gep = getelementptr i8, ptr %p, i64 -9223372036854771712
+  %gep = getelementptr inbounds i8, ptr %p, i64 -9223372036854771712
   %load = load volatile i8, ptr %gep, align 1
   store i8 %load, ptr poison
   ret void
@@ -5455,7 +5455,7 @@ define amdgpu_kernel void @flat_inst_salu_offset_64bit_13bit_neg_high_split0(ptr
 ; GFX12-GISEL-NEXT:    s_wait_loadcnt_dscnt 0x0
 ; GFX12-GISEL-NEXT:    flat_store_b8 v[0:1], v0
 ; GFX12-GISEL-NEXT:    s_endpgm
-  %gep = getelementptr i8, ptr %p, i64 -9223372036854767617
+  %gep = getelementptr inbounds i8, ptr %p, i64 -9223372036854767617
   %load = load volatile i8, ptr %gep, align 1
   store i8 %load, ptr poison
   ret void
@@ -5580,7 +5580,7 @@ define amdgpu_kernel void @flat_inst_salu_offset_64bit_13bit_neg_high_split1(ptr
 ; GFX12-GISEL-NEXT:    s_wait_loadcnt_dscnt 0x0
 ; GFX12-GISEL-NEXT:    flat_store_b8 v[0:1], v0
 ; GFX12-GISEL-NEXT:    s_endpgm
-  %gep = getelementptr i8, ptr %p, i64 -9223372036854767616
+  %gep = getelementptr inbounds i8, ptr %p, i64 -9223372036854767616
   %load = load volatile i8, ptr %gep, align 1
   store i8 %load, ptr poison
   ret void
diff --git a/llvm/test/CodeGen/AMDGPU/promote-constOffset-to-imm.ll b/llvm/test/CodeGen/AMDGPU/promote-constOffset-to-imm.ll
index 85a9aba1a0e51..57eb25700c76a 100644
--- a/llvm/test/CodeGen/AMDGPU/promote-constOffset-to-imm.ll
+++ b/llvm/test/CodeGen/AMDGPU/promote-constOffset-to-imm.ll
@@ -2658,7 +2658,7 @@ define amdgpu_kernel void @negativeoffsetnullptr(ptr %buffer) {
 ; GFX11-FAKE16-NEXT:    s_endpgm
 entry:
   %null = select i1 false, ptr %buffer, ptr addrspacecast (ptr addrspace(5) null to ptr)
-  %gep = getelementptr i8, ptr %null, i64 -1
+  %gep = getelementptr inbounds i8, ptr %null, i64 -1
   %ld = load i8, ptr %gep
   %cmp = icmp eq i8 %ld, 0
   br label %branch