[AMDGPU] Enable AAAddressSpace in AMDGPUAttributor (#101593)

shiltian · web-flow · commit a0afcbfb5dd1 · 2024-08-06T15:27:18.000-04:00
diff --git a/llvm/lib/Target/AMDGPU/AMDGPUAttributor.cpp b/llvm/lib/Target/AMDGPU/AMDGPUAttributor.cpp
@@ -1038,7 +1038,7 @@ static bool runImpl(Module &M, AnalysisGetter &AG, TargetMachine &TM) {
        &AAPotentialValues::ID, &AAAMDFlatWorkGroupSize::ID,
        &AAAMDWavesPerEU::ID, &AAAMDGPUNoAGPR::ID, &AACallEdges::ID,
        &AAPointerInfo::ID, &AAPotentialConstantValues::ID,
-       &AAUnderlyingObjects::ID});
+       &AAUnderlyingObjects::ID, &AAAddressSpace::ID});
 
   AttributorConfig AC(CGUpdater);
   AC.Allowed = &Allowed;
@@ -1064,6 +1064,17 @@ static bool runImpl(Module &M, AnalysisGetter &AG, TargetMachine &TM) {
     } else if (CC == CallingConv::AMDGPU_KERNEL) {
       addPreloadKernArgHint(F, TM);
     }
+
+    for (auto &I : instructions(F)) {
+      if (auto *LI = dyn_cast<LoadInst>(&I)) {
+        A.getOrCreateAAFor<AAAddressSpace>(
+            IRPosition::value(*LI->getPointerOperand()));
+      }
+      if (auto *SI = dyn_cast<StoreInst>(&I)) {
+        A.getOrCreateAAFor<AAAddressSpace>(
+            IRPosition::value(*SI->getPointerOperand()));
+      }
+    }
   }
 
   ChangeStatus Change = A.run();
diff --git a/llvm/test/CodeGen/AMDGPU/GlobalISel/flat-scratch-init.ll b/llvm/test/CodeGen/AMDGPU/GlobalISel/flat-scratch-init.ll
@@ -7,10 +7,11 @@ target triple = "amdgcn-amd-amdhsa"
 ; Make sure flat_scratch_init is set
 
 ; GCN-LABEL: {{^}}stack_object_addrspacecast_in_kernel_no_calls:
-; RW-FLAT:     s_add_u32 flat_scratch_lo, s4, s7
-; RW-FLAT:     s_addc_u32 flat_scratch_hi, s5, 0
+; RW-FLAT:     s_add_u32 s0, s0, s7
+; RW-FLAT:     s_addc_u32 s1, s1, 0
 ; RO-FLAT-NOT: flat_scratch
-; GCN:         flat_store_dword
+; RW-FLAT:     buffer_store_dword
+; RO-FLAT:     scratch_store_dword
 ; RO-FLAT-NOT: .amdhsa_user_sgpr_private_segment_buffer
 ; RW-FLAT:     .amdhsa_user_sgpr_flat_scratch_init 1
 ; RO-FLAT-NOT: .amdhsa_user_sgpr_flat_scratch_init
diff --git a/llvm/test/CodeGen/AMDGPU/addrspacecast.ll b/llvm/test/CodeGen/AMDGPU/addrspacecast.ll
@@ -5,22 +5,11 @@ target triple = "amdgcn-amd-amdhsa"
 
 ; HSA-LABEL: {{^}}use_group_to_flat_addrspacecast:
 
-; CI-DAG: s_load_dword [[PTR:s[0-9]+]], s[6:7], 0x0{{$}}
-; CI-DAG: s_load_dword [[APERTURE:s[0-9]+]], s[4:5], 0x10{{$}}
-; CI-DAG: s_cmp_lg_u32 [[PTR]], -1
-; CI-DAG: s_cselect_b32 s[[HI:[0-9]+]], [[APERTURE]], 0
-; CI-DAG: s_cselect_b32 s[[LO:[0-9]+]], [[PTR]], 0
-
-; GFX9-DAG: s_mov_b64 s[{{[0-9]+}}:[[HIBASE:[0-9]+]]], src_shared_base
-
+; CI-DAG: s_load_dword [[APERTURE:s[0-9]+]], s[6:7], 0x0{{$}}
+; GFX9-DAG: s_load_dword [[APERTURE:s[0-9]+]], s[4:5], 0x0{{$}}
+; HSA-DAG: v_mov_b32_e32 [[PTR:v[0-9]+]], [[APERTURE]]
 ; HSA-DAG: v_mov_b32_e32 [[K:v[0-9]+]], 7
-; GFX9-DAG: s_load_dword [[PTR:s[0-9]+]], s[4:5], 0x0{{$}}
-
-; GFX9: s_cmp_lg_u32 [[PTR]], -1
-; GFX9-DAG: s_cselect_b32 s[[LO:[0-9]+]], s[[HIBASE]], 0
-; GFX9-DAG: s_cselect_b32 s[[HI:[0-9]+]], [[PTR]], 0
-
-; HSA: flat_store_dword v[[[LO]]:[[HI]]], [[K]]
+; HSA-DAG: ds_write_b32 [[PTR]], [[K]]
 
 ; HSA:  .amdhsa_user_sgpr_private_segment_buffer 1
 ; HSA:  .amdhsa_user_sgpr_dispatch_ptr 0
@@ -39,22 +28,8 @@ define amdgpu_kernel void @use_group_to_flat_addrspacecast(ptr addrspace(3) %ptr
 
 ; Test handling inside a non-kernel
 ; HSA-LABEL: {{^}}use_group_to_flat_addrspacecast_func:
-; CI-DAG: s_load_dword [[APERTURE:s[0-9]+]], s[6:7], 0x10{{$}}
-; CI-DAG: v_mov_b32_e32 [[VAPERTURE:v[0-9]+]], [[APERTURE]]
-; CI-DAG: v_cmp_ne_u32_e32 vcc, -1, v0
-; CI-DAG: v_cndmask_b32_e32 v[[HI:[0-9]+]], 0, [[VAPERTURE]], vcc
-; CI-DAG: v_cndmask_b32_e32 v[[LO:[0-9]+]], 0, v0
-
-; GFX9-DAG: s_mov_b64 s[{{[0-9]+}}:[[HIBASE:[0-9]+]]], src_shared_base
-
 ; HSA-DAG: v_mov_b32_e32 [[K:v[0-9]+]], 7
-
-; GFX9-DAG: v_mov_b32_e32 v[[VREG_HIBASE:[0-9]+]], s[[HIBASE]]
-; GFX9-DAG: v_cmp_ne_u32_e32 vcc, -1, v0
-; GFX9-DAG: v_cndmask_b32_e32 v[[LO:[0-9]+]], 0, v0, vcc
-; GFX9-DAG: v_cndmask_b32_e32 v[[HI:[0-9]+]], 0, v[[VREG_HIBASE]], vcc
-
-; HSA: flat_store_dword v[[[LO]]:[[HI]]], [[K]]
+; HSA-DAG: ds_write_b32 v0, [[K]]
 define void @use_group_to_flat_addrspacecast_func(ptr addrspace(3) %ptr) #0 {
   %stof = addrspacecast ptr addrspace(3) %ptr to ptr
   store volatile i32 7, ptr %stof
@@ -63,23 +38,16 @@ define void @use_group_to_flat_addrspacecast_func(ptr addrspace(3) %ptr) #0 {
 
 ; HSA-LABEL: {{^}}use_private_to_flat_addrspacecast:
 
-; CI-DAG: s_load_dword [[PTR:s[0-9]+]], s[6:7], 0x0{{$}}
-; CI-DAG: s_load_dword [[APERTURE:s[0-9]+]], s[4:5], 0x11{{$}}
-
-; CI-DAG: v_mov_b32_e32 [[K:v[0-9]+]], 7
-; CI-DAG: s_cmp_lg_u32 [[PTR]], -1
-; CI-DAG: s_cselect_b32 s[[HI:[0-9]+]], [[APERTURE]], 0
-; CI-DAG: s_cselect_b32 s[[LO:[0-9]+]], [[PTR]], 0
-
-; GFX9-DAG: s_load_dword [[PTR:s[0-9]+]], s[4:5], 0x0{{$}}
-; GFX9-DAG: s_mov_b64 s[{{[0-9]+}}:[[HIBASE:[0-9]+]]], src_private_base
-
-; GFX9-DAG: v_mov_b32_e32 [[K:v[0-9]+]], 7
-; GFX9: s_cmp_lg_u32 [[PTR]], -1
-; GFX9: s_cselect_b32 s[[LO:[0-9]+]], s[[HIBASE]], 0
-; GFX9: s_cselect_b32 s[[HI:[0-9]+]], [[PTR]], 0
-
-; HSA: flat_store_dword v[[[LO]]:[[HI]]], [[K]]
+; CI-DAG: s_load_dword [[APERTURE:s[0-9]+]], s[6:7], 0x0{{$}}
+; GFX9-DAG: s_load_dword [[APERTURE:s[0-9]+]], s[4:5], 0x0{{$}}
+; HSA-DAG: v_mov_b32_e32 [[PTR:v[0-9]+]], [[APERTURE]]
+; HSA-DAG: s_mov_b64 s[{{[0-9]+}}:[[RSRCHI:[0-9]+]]], s[2:3]
+; HSA-DAG: s_mov_b64 s[[[BASELO:[0-9]+]]:[[BASEHI:[0-9]+]]], s[0:1]
+; SI-DAG: s_add_u32 s[[BASELO]], s[[BASELO]], s9
+; GFX9-DAG: s_add_u32 s[[BASELO]], s[[BASELO]], s7
+; HSA-DAG: s_addc_u32 s[[BASEHI]], s[[BASEHI]], 0
+; HSA-DAG: v_mov_b32_e32 [[K:v[0-9]+]], 7
+; HSA: buffer_store_dword [[K]], [[PTR]], s[[[BASELO]]:[[RSRCHI]]], 0 offen
 
 ; HSA:  .amdhsa_user_sgpr_private_segment_buffer 1
 ; HSA:  .amdhsa_user_sgpr_dispatch_ptr 0
@@ -97,10 +65,12 @@ define amdgpu_kernel void @use_private_to_flat_addrspacecast(ptr addrspace(5) %p
 ; HSA-LABEL: {{^}}use_global_to_flat_addrspacecast:
 
 ; HSA: s_load_dwordx2 s[[[PTRLO:[0-9]+]]:[[PTRHI:[0-9]+]]]
-; HSA-DAG: v_mov_b32_e32 v[[VPTRLO:[0-9]+]], s[[PTRLO]]
-; HSA-DAG: v_mov_b32_e32 v[[VPTRHI:[0-9]+]], s[[PTRHI]]
+; CI-DAG: v_mov_b32_e32 v[[VPTRLO:[0-9]+]], s[[PTRLO]]
+; CI-DAG: v_mov_b32_e32 v[[VPTRHI:[0-9]+]], s[[PTRHI]]
 ; HSA-DAG: v_mov_b32_e32 [[K:v[0-9]+]], 7
-; HSA: flat_store_dword v[[[VPTRLO]]:[[VPTRHI]]], [[K]]
+; CI: flat_store_dword v[[[VPTRLO]]:[[VPTRHI]]], [[K]]
+; GFX9-DAG: v_mov_b32_e32 [[ADDR:v[0-9]+]], 0
+; GFX9: global_store_dword [[ADDR]], [[K]], s[[[PTRLO]]:[[PTRHI]]]
 
 ; HSA:  .amdhsa_user_sgpr_queue_ptr 0
 define amdgpu_kernel void @use_global_to_flat_addrspacecast(ptr addrspace(1) %ptr) #0 {
@@ -112,9 +82,7 @@ define amdgpu_kernel void @use_global_to_flat_addrspacecast(ptr addrspace(1) %pt
 ; no-op
 ; HSA-LABEL: {{^}}use_constant_to_flat_addrspacecast:
 ; HSA: s_load_dwordx2 s[[[PTRLO:[0-9]+]]:[[PTRHI:[0-9]+]]]
-; HSA-DAG: v_mov_b32_e32 v[[VPTRLO:[0-9]+]], s[[PTRLO]]
-; HSA-DAG: v_mov_b32_e32 v[[VPTRHI:[0-9]+]], s[[PTRHI]]
-; HSA: flat_load_dword v{{[0-9]+}}, v[[[VPTRLO]]:[[VPTRHI]]]
+; HSA-DAG: s_load_dword s0, s[[[PTRLO]]:[[PTRHI]]], 0x0
 define amdgpu_kernel void @use_constant_to_flat_addrspacecast(ptr addrspace(4) %ptr) #0 {
   %stof = addrspacecast ptr addrspace(4) %ptr to ptr
   %ld = load volatile i32, ptr %stof
@@ -215,14 +183,9 @@ define amdgpu_kernel void @use_flat_to_constant_addrspacecast(ptr %ptr) #0 {
 }
 
 ; HSA-LABEL: {{^}}cast_0_group_to_flat_addrspacecast:
-; CI: s_load_dword [[APERTURE:s[0-9]+]], s[4:5], 0x10
-; CI-DAG: v_mov_b32_e32 v[[HI:[0-9]+]], [[APERTURE]]
-
-; GFX9-DAG: s_mov_b64 s[{{[0-9]+}}:[[HI:[0-9]+]]], src_shared_base
-
 ; HSA-DAG: v_mov_b32_e32 v[[LO:[0-9]+]], 0{{$}}
 ; HSA-DAG: v_mov_b32_e32 v[[K:[0-9]+]], 7{{$}}
-; HSA: {{flat|global}}_store_dword v[[[LO]]:[[HI]]], v[[K]]
+; HSA: ds_write_b32 v[[LO]], v[[K]]
 define amdgpu_kernel void @cast_0_group_to_flat_addrspacecast() #0 {
   %cast = addrspacecast ptr addrspace(3) null to ptr
   store volatile i32 7, ptr %cast
@@ -240,10 +203,9 @@ define amdgpu_kernel void @cast_0_flat_to_group_addrspacecast() #0 {
 }
 
 ; HSA-LABEL: {{^}}cast_neg1_group_to_flat_addrspacecast:
-; HSA: v_mov_b32_e32 v[[LO:[0-9]+]], 0{{$}}
 ; HSA-DAG: v_mov_b32_e32 v[[K:[0-9]+]], 7{{$}}
-; HSA-DAG: v_mov_b32_e32 v[[HI:[0-9]+]], 0{{$}}
-; HSA: {{flat|global}}_store_dword v[[[LO]]:[[HI]]], v[[K]]
+; HSA-DAG: v_mov_b32_e32 v[[LO:[0-9]+]], -1
+; HSA: ds_write_b32 v[[LO]], v[[K]]
 define amdgpu_kernel void @cast_neg1_group_to_flat_addrspacecast() #0 {
   %cast = addrspacecast ptr addrspace(3) inttoptr (i32 -1 to ptr addrspace(3)) to ptr
   store volatile i32 7, ptr %cast
@@ -262,14 +224,13 @@ define amdgpu_kernel void @cast_neg1_flat_to_group_addrspacecast() #0 {
 
 ; FIXME: Shouldn't need to enable queue ptr
 ; HSA-LABEL: {{^}}cast_0_private_to_flat_addrspacecast:
-; CI: s_load_dword [[APERTURE:s[0-9]+]], s[4:5], 0x11
-; CI-DAG: v_mov_b32_e32 v[[HI:[0-9]+]], [[APERTURE]]
-
-; GFX9-DAG: s_mov_b64 s[{{[0-9]+}}:[[HI:[0-9]+]]], src_private_base
-
-; HSA-DAG: v_mov_b32_e32 v[[LO:[0-9]+]], 0{{$}}
+; HSA-DAG: s_mov_b64 s[{{[0-9]+}}:[[RSRCHI:[0-9]+]]], s[2:3]
+; HSA-DAG: s_mov_b64 s[[[BASELO:[0-9]+]]:[[BASEHI:[0-9]+]]], s[0:1]
+; CI-DAG: s_add_u32 s[[BASELO]], s[[BASELO]], s7
+; GFX9-DAG: s_add_u32 s[[BASELO]], s[[BASELO]], s5
+; HSA-DAG: s_addc_u32 s[[BASEHI]], s[[BASEHI]], 0
 ; HSA-DAG: v_mov_b32_e32 v[[K:[0-9]+]], 7{{$}}
-; HSA: {{flat|global}}_store_dword v[[[LO]]:[[HI]]], v[[K]]
+; HSA: buffer_store_dword v[[K]], off, s[[[BASELO]]:[[RSRCHI]]], 0
 define amdgpu_kernel void @cast_0_private_to_flat_addrspacecast() #0 {
   %cast = addrspacecast ptr addrspace(5) null to ptr
   store volatile i32 7, ptr %cast
@@ -286,13 +247,16 @@ define amdgpu_kernel void @cast_0_flat_to_private_addrspacecast() #0 {
   ret void
 }
 
-
 ; HSA-LABEL: {{^}}cast_neg1_private_to_flat_addrspacecast:
 
-; HSA: v_mov_b32_e32 v[[LO:[0-9]+]], 0{{$}}
+; HSA-DAG: s_mov_b64 s[{{[0-9]+}}:[[RSRCHI:[0-9]+]]], s[2:3]
+; HSA-DAG: s_mov_b64 s[[[BASELO:[0-9]+]]:[[BASEHI:[0-9]+]]], s[0:1]
+; CI-DAG: s_add_u32 s[[BASELO]], s[[BASELO]], s7
+; GFX9-DAG: s_add_u32 s[[BASELO]], s[[BASELO]], s5
+; HSA-DAG: s_addc_u32 s[[BASEHI]], s[[BASEHI]], 0
+; HSA-DAG: v_mov_b32_e32 [[PTR:v[0-9]+]], -1{{$}}
 ; HSA-DAG: v_mov_b32_e32 v[[K:[0-9]+]], 7{{$}}
-; HSA-DAG: v_mov_b32_e32 v[[HI:[0-9]+]], 0{{$}}
-; HSA: {{flat|global}}_store_dword v[[[LO]]:[[HI]]], v[[K]]
+; HSA: buffer_store_dword v[[K]], [[PTR]], s[[[BASELO]]:[[RSRCHI]]], 0 offen
 
 ; CI:  .amdhsa_user_sgpr_queue_ptr 1
 ; GFX9:  .amdhsa_user_sgpr_queue_ptr 0
@@ -342,16 +306,18 @@ end:
 
 ; Check for prologue initializing special SGPRs pointing to scratch.
 ; HSA-LABEL: {{^}}store_flat_scratch:
-; CI-DAG: s_mov_b32 flat_scratch_lo, s9
 ; CI-DAG: s_add_i32 [[ADD:s[0-9]+]], s8, s11
 ; CI-DAG: s_lshr_b32 flat_scratch_hi, [[ADD]], 8
-
-; GFX9: s_add_u32 flat_scratch_lo, s6, s9
-; GFX9: s_addc_u32 flat_scratch_hi, s7, 0
-
-; HSA: {{flat|global}}_store_dword
-; HSA: s_barrier
-; HSA: {{flat|global}}_load_dword
+; HSA: buffer_store_dword
+; HSA:  s_barrier
+; HSA: buffer_load_dword [[K:v[0-9]+]], v{{[0-9]+}}, s[0:3], 0 offen glc
+; HSA-DAG: s_load_dwordx2
+; CI-DAG: s_mov_b32 flat_scratch_lo, s9
+; CI-DAG: v_mov_b32_e32 v[[LO:[0-9]+]], s4
+; CI-DAG: v_mov_b32_e32 v[[HI:[0-9]+]], s5
+; GFX9-DAG: v_mov_b32_e32 [[PTR:v[0-9]+]], 0
+; CI: flat_store_dword v[[[LO]]:[[HI]]], [[K]]
+; GFX9: global_store_dword [[PTR]], [[K]]
 define amdgpu_kernel void @store_flat_scratch(ptr addrspace(1) noalias %out, i32) #0 {
   %alloca = alloca i32, i32 9, align 4, addrspace(5)
   %x = call i32 @llvm.amdgcn.workitem.id.x() #2
diff --git a/llvm/test/CodeGen/AMDGPU/annotate-kernel-features-hsa.ll b/llvm/test/CodeGen/AMDGPU/annotate-kernel-features-hsa.ll
@@ -425,8 +425,7 @@ define amdgpu_kernel void @use_group_to_flat_addrspacecast(ptr addrspace(3) %ptr
 ;
 ; ATTRIBUTOR_HSA-LABEL: define {{[^@]+}}@use_group_to_flat_addrspacecast
 ; ATTRIBUTOR_HSA-SAME: (ptr addrspace(3) [[PTR:%.*]]) #[[ATTR12:[0-9]+]] {
-; ATTRIBUTOR_HSA-NEXT:    [[STOF:%.*]] = addrspacecast ptr addrspace(3) [[PTR]] to ptr
-; ATTRIBUTOR_HSA-NEXT:    store volatile i32 0, ptr [[STOF]], align 4
+; ATTRIBUTOR_HSA-NEXT:    store volatile i32 0, ptr addrspace(3) [[PTR]], align 4
 ; ATTRIBUTOR_HSA-NEXT:    ret void
 ;
   %stof = addrspacecast ptr addrspace(3) %ptr to ptr
@@ -443,8 +442,7 @@ define amdgpu_kernel void @use_private_to_flat_addrspacecast(ptr addrspace(5) %p
 ;
 ; ATTRIBUTOR_HSA-LABEL: define {{[^@]+}}@use_private_to_flat_addrspacecast
 ; ATTRIBUTOR_HSA-SAME: (ptr addrspace(5) [[PTR:%.*]]) #[[ATTR12]] {
-; ATTRIBUTOR_HSA-NEXT:    [[STOF:%.*]] = addrspacecast ptr addrspace(5) [[PTR]] to ptr
-; ATTRIBUTOR_HSA-NEXT:    store volatile i32 0, ptr [[STOF]], align 4
+; ATTRIBUTOR_HSA-NEXT:    store volatile i32 0, ptr addrspace(5) [[PTR]], align 4
 ; ATTRIBUTOR_HSA-NEXT:    ret void
 ;
   %stof = addrspacecast ptr addrspace(5) %ptr to ptr
@@ -478,23 +476,33 @@ define amdgpu_kernel void @use_flat_to_private_addrspacecast(ptr %ptr) #1 {
 
 ; No-op addrspacecast should not use queue ptr
 define amdgpu_kernel void @use_global_to_flat_addrspacecast(ptr addrspace(1) %ptr) #1 {
-; HSA-LABEL: define {{[^@]+}}@use_global_to_flat_addrspacecast
-; HSA-SAME: (ptr addrspace(1) [[PTR:%.*]]) #[[ATTR1]] {
-; HSA-NEXT:    [[STOF:%.*]] = addrspacecast ptr addrspace(1) [[PTR]] to ptr
-; HSA-NEXT:    store volatile i32 0, ptr [[STOF]], align 4
-; HSA-NEXT:    ret void
+; AKF_HSA-LABEL: define {{[^@]+}}@use_global_to_flat_addrspacecast
+; AKF_HSA-SAME: (ptr addrspace(1) [[PTR:%.*]]) #[[ATTR1]] {
+; AKF_HSA-NEXT:    [[STOF:%.*]] = addrspacecast ptr addrspace(1) [[PTR]] to ptr
+; AKF_HSA-NEXT:    store volatile i32 0, ptr [[STOF]], align 4
+; AKF_HSA-NEXT:    ret void
+;
+; ATTRIBUTOR_HSA-LABEL: define {{[^@]+}}@use_global_to_flat_addrspacecast
+; ATTRIBUTOR_HSA-SAME: (ptr addrspace(1) [[PTR:%.*]]) #[[ATTR1]] {
+; ATTRIBUTOR_HSA-NEXT:    store volatile i32 0, ptr addrspace(1) [[PTR]], align 4
+; ATTRIBUTOR_HSA-NEXT:    ret void
 ;
   %stof = addrspacecast ptr addrspace(1) %ptr to ptr
   store volatile i32 0, ptr %stof
   ret void
 }
 
 define amdgpu_kernel void @use_constant_to_flat_addrspacecast(ptr addrspace(4) %ptr) #1 {
-; HSA-LABEL: define {{[^@]+}}@use_constant_to_flat_addrspacecast
-; HSA-SAME: (ptr addrspace(4) [[PTR:%.*]]) #[[ATTR1]] {
-; HSA-NEXT:    [[STOF:%.*]] = addrspacecast ptr addrspace(4) [[PTR]] to ptr
-; HSA-NEXT:    [[LD:%.*]] = load volatile i32, ptr [[STOF]], align 4
-; HSA-NEXT:    ret void
+; AKF_HSA-LABEL: define {{[^@]+}}@use_constant_to_flat_addrspacecast
+; AKF_HSA-SAME: (ptr addrspace(4) [[PTR:%.*]]) #[[ATTR1]] {
+; AKF_HSA-NEXT:    [[STOF:%.*]] = addrspacecast ptr addrspace(4) [[PTR]] to ptr
+; AKF_HSA-NEXT:    [[LD:%.*]] = load volatile i32, ptr [[STOF]], align 4
+; AKF_HSA-NEXT:    ret void
+;
+; ATTRIBUTOR_HSA-LABEL: define {{[^@]+}}@use_constant_to_flat_addrspacecast
+; ATTRIBUTOR_HSA-SAME: (ptr addrspace(4) [[PTR:%.*]]) #[[ATTR1]] {
+; ATTRIBUTOR_HSA-NEXT:    [[LD:%.*]] = load volatile i32, ptr addrspace(4) [[PTR]], align 4
+; ATTRIBUTOR_HSA-NEXT:    ret void
 ;
   %stof = addrspacecast ptr addrspace(4) %ptr to ptr
   %ld = load volatile i32, ptr %stof
diff --git a/llvm/test/CodeGen/AMDGPU/callee-special-input-sgprs-fixed-abi.ll b/llvm/test/CodeGen/AMDGPU/callee-special-input-sgprs-fixed-abi.ll
@@ -38,15 +38,9 @@ define amdgpu_kernel void @kern_indirect_use_queue_ptr(i32) #1 {
 }
 
 ; GCN-LABEL: {{^}}use_queue_ptr_addrspacecast:
-; CIVI: s_load_dword [[APERTURE_LOAD:s[0-9]+]], s[4:5], 0x0
-; CIVI: v_mov_b32_e32 v[[LO:[0-9]+]], 16
-; CIVI-DAG: v_mov_b32_e32 v[[HI:[0-9]+]], [[APERTURE_LOAD]]
+; GCN: v_mov_b32_e32 v[[LO:[0-9]+]], 0
+; GCN-DAG: ds_write_b32 v[[LO]], v[[LO]] offset:16
 
-; GFX9: s_mov_b64 s[{{[0-9]+}}:[[HI:[0-9]+]]], src_shared_base
-; GFX9-DAG: v_mov_b32_e32 v[[VGPR_HI:[0-9]+]], s[[HI]]
-; GFX9: {{flat|global}}_store_dword v{{\[[0-9]+}}:[[VGPR_HI]]]
-
-; CIVI: {{flat|global}}_store_dword v[[[LO]]:[[HI]]]
 define hidden void @use_queue_ptr_addrspacecast() #1 {
   %asc = addrspacecast ptr addrspace(3) inttoptr (i32 16 to ptr addrspace(3)) to ptr
   store volatile i32 0, ptr %asc
diff --git a/llvm/test/CodeGen/AMDGPU/simple-indirect-call.ll b/llvm/test/CodeGen/AMDGPU/simple-indirect-call.ll
@@ -34,9 +34,8 @@ define amdgpu_kernel void @test_simple_indirect_call() {
 ; ATTRIBUTOR_GCN-LABEL: define {{[^@]+}}@test_simple_indirect_call
 ; ATTRIBUTOR_GCN-SAME: () #[[ATTR1:[0-9]+]] {
 ; ATTRIBUTOR_GCN-NEXT:    [[FPTR:%.*]] = alloca ptr, align 8, addrspace(5)
-; ATTRIBUTOR_GCN-NEXT:    [[FPTR_CAST:%.*]] = addrspacecast ptr addrspace(5) [[FPTR]] to ptr
-; ATTRIBUTOR_GCN-NEXT:    store ptr @indirect, ptr [[FPTR_CAST]], align 8
-; ATTRIBUTOR_GCN-NEXT:    [[FP:%.*]] = load ptr, ptr [[FPTR_CAST]], align 8
+; ATTRIBUTOR_GCN-NEXT:    store ptr @indirect, ptr addrspace(5) [[FPTR]], align 8
+; ATTRIBUTOR_GCN-NEXT:    [[FP:%.*]] = load ptr, ptr addrspace(5) [[FPTR]], align 8
 ; ATTRIBUTOR_GCN-NEXT:    call void [[FP]]()
 ; ATTRIBUTOR_GCN-NEXT:    ret void
 ;
@@ -75,12 +74,16 @@ define amdgpu_kernel void @test_simple_indirect_call() {
   ret void
 }
 
+
+!llvm.module.flags = !{!0}
+!0 = !{i32 1, !"amdhsa_code_object_version", i32 500}
 ;.
 ; AKF_GCN: attributes #[[ATTR0]] = { "amdgpu-calls" "amdgpu-stack-objects" }
 ;.
 ; ATTRIBUTOR_GCN: attributes #[[ATTR0]] = { "amdgpu-no-agpr" "amdgpu-no-completion-action" "amdgpu-no-default-queue" "amdgpu-no-dispatch-id" "amdgpu-no-dispatch-ptr" "amdgpu-no-heap-ptr" "amdgpu-no-hostcall-ptr" "amdgpu-no-implicitarg-ptr" "amdgpu-no-lds-kernel-id" "amdgpu-no-multigrid-sync-arg" "amdgpu-no-queue-ptr" "amdgpu-no-workgroup-id-x" "amdgpu-no-workgroup-id-y" "amdgpu-no-workgroup-id-z" "amdgpu-no-workitem-id-x" "amdgpu-no-workitem-id-y" "amdgpu-no-workitem-id-z" "amdgpu-waves-per-eu"="4,10" "uniform-work-group-size"="false" }
 ; ATTRIBUTOR_GCN: attributes #[[ATTR1]] = { "uniform-work-group-size"="false" }
 ;.
-
-!llvm.module.flags = !{!0}
-!0 = !{i32 1, !"amdhsa_code_object_version", i32 500}
+; AKF_GCN: [[META0:![0-9]+]] = !{i32 1, !"amdhsa_code_object_version", i32 500}
+;.
+; ATTRIBUTOR_GCN: [[META0:![0-9]+]] = !{i32 1, !"amdhsa_code_object_version", i32 500}
+;.
