AMDGPU: Switch some half using-tests to use amdhsa

The default clover ABI weirdly promotes half to float,
which should probably be fixed.

llvm-svn: 333730

1 files changed, 110 insertions, 107 deletions

diff --git a/llvm/test/CodeGen/AMDGPU/half.ll b/llvm/test/CodeGen/AMDGPU/half.ll
index f31b2ab5563..453d8fc387d 100644
--- a/llvm/test/CodeGen/AMDGPU/half.ll
+++ b/llvm/test/CodeGen/AMDGPU/half.ll
@@ -1,13 +1,12 @@
-; RUN: llc -amdgpu-scalarize-global-loads=false -march=amdgcn -mcpu=tahiti -verify-machineinstrs < %s | FileCheck -enable-var-scope -check-prefixes=GCN,SI %s
-; RUN: llc -amdgpu-scalarize-global-loads=false -march=amdgcn -mcpu=tonga -mattr=-flat-for-global -verify-machineinstrs < %s | FileCheck -enable-var-scope -check-prefixes=GCN,VI %s
+; RUN: llc -amdgpu-scalarize-global-loads=false -mtriple=amdgcn-amd-amdhsa -mcpu=kaveri -verify-machineinstrs < %s | FileCheck -enable-var-scope -check-prefixes=GCN,SI %s
+; RUN: llc -amdgpu-scalarize-global-loads=false -mtriple=amdgcn-amd-amdhsa -mcpu=tonga -verify-machineinstrs < %s | FileCheck -enable-var-scope -check-prefixes=GCN,VI %s
 
 ; half args should be promoted to float for SI and lower.
 
 ; GCN-LABEL: {{^}}load_f16_arg:
-; GCN: s_load_dword [[ARG:s[0-9]+]]
-; SI: v_cvt_f16_f32_e32 [[CVT:v[0-9]+]], [[ARG]]
-; VI: v_trunc_f16_e32 [[CVT:v[0-9]+]], [[ARG]]
-; GCN: buffer_store_short [[CVT]]
+; GCN: flat_load_ushort [[ARG:v[0-9]+]]
+; GCN-NOT: [[ARG]]
+; GCN: flat_store_short v{{\[[0-9]+:[0-9]+\]}}, [[ARG]]
 define amdgpu_kernel void @load_f16_arg(half addrspace(1)* %out, half %arg) #0 {
   store half %arg, half addrspace(1)* %out
   ret void
@@ -16,20 +15,20 @@ define amdgpu_kernel void @load_f16_arg(half addrspace(1)* %out, half %arg) #0 {
 ; GCN-LABEL: {{^}}load_v2f16_arg:
 ; GCN: s_load_dword [[ARG:s[0-9]+]]
 ; GCN: v_mov_b32_e32 [[V_ARG:v[0-9]+]], [[ARG]]
-; GCN: buffer_store_dword [[V_ARG]]
+; GCN: flat_store_dword v{{\[[0-9]+:[0-9]+\]}}, [[V_ARG]]
 define amdgpu_kernel void @load_v2f16_arg(<2 x half> addrspace(1)* %out, <2 x half> %arg) #0 {
   store <2 x half> %arg, <2 x half> addrspace(1)* %out
   ret void
 }
 
 ; GCN-LABEL: {{^}}load_v3f16_arg:
-; GCN: buffer_load_ushort
+; GCN: flat_load_ushort
 ; GCN: s_load_dword s
 
-; GCN-NOT: buffer_load
-; GCN-DAG: buffer_store_dword
-; GCN-DAG: buffer_store_short
-; GCN-NOT: buffer_store
+; GCN-NOT: _load
+; GCN-DAG: _store_dword
+; GCN-DAG: _store_short
+; GCN-NOT: _store
 ; GCN: s_endpgm
 define amdgpu_kernel void @load_v3f16_arg(<3 x half> addrspace(1)* %out, <3 x half> %arg) #0 {
   store <3 x half> %arg, <3 x half> addrspace(1)* %out
@@ -39,11 +38,11 @@ define amdgpu_kernel void @load_v3f16_arg(<3 x half> addrspace(1)* %out, <3 x ha
 
 ; FIXME: Why not one load?
 ; GCN-LABEL: {{^}}load_v4f16_arg:
-; GCN-DAG: s_load_dword [[ARG0_LO:s[0-9]+]], s{{\[[0-9]+:[0-9]+\]}}, {{0xb|0x2c}}
-; GCN-DAG: s_load_dword [[ARG0_HI:s[0-9]+]], s{{\[[0-9]+:[0-9]+\]}}, {{0xc|0x30}}
+; GCN-DAG: s_load_dword [[ARG0_LO:s[0-9]+]], s{{\[[0-9]+:[0-9]+\]}}, {{0x2|0x8}}
+; GCN-DAG: s_load_dword [[ARG0_HI:s[0-9]+]], s{{\[[0-9]+:[0-9]+\]}}, {{0x3|0xc}}
 ; GCN-DAG: v_mov_b32_e32 v[[V_ARG0_LO:[0-9]+]], [[ARG0_LO]]
 ; GCN-DAG: v_mov_b32_e32 v[[V_ARG0_HI:[0-9]+]], [[ARG0_HI]]
-; GCN: buffer_store_dwordx2 v{{\[}}[[V_ARG0_LO]]:[[V_ARG0_HI]]{{\]}}
+; GCN: flat_store_dwordx2 v{{\[[0-9]+:[0-9]+\]}}, v{{\[}}[[V_ARG0_LO]]:[[V_ARG0_HI]]{{\]}}
 define amdgpu_kernel void @load_v4f16_arg(<4 x half> addrspace(1)* %out, <4 x half> %arg) #0 {
   store <4 x half> %arg, <4 x half> addrspace(1)* %out
   ret void
@@ -77,16 +76,16 @@ define amdgpu_kernel void @extload_v2f16_to_v2f32_arg(<2 x float> addrspace(1)*
 }
 
 ; GCN-LABEL: {{^}}extload_v3f16_to_v3f32_arg:
-; GCN: buffer_load_ushort
-; GCN: buffer_load_ushort
-; GCN: buffer_load_ushort
-; GCN-NOT: buffer_load
+; GCN: flat_load_ushort
+; GCN: flat_load_ushort
+; GCN: flat_load_ushort
+; GCN-NOT: {{buffer|flat|global}}_load
 ; GCN: v_cvt_f32_f16_e32
 ; GCN: v_cvt_f32_f16_e32
 ; GCN: v_cvt_f32_f16_e32
 ; GCN-NOT: v_cvt_f32_f16
-; GCN-DAG: buffer_store_dword
-; GCN-DAG: buffer_store_dwordx2
+; GCN-DAG: _store_dword
+; GCN-DAG: _store_dwordx2
 ; GCN: s_endpgm
 define amdgpu_kernel void @extload_v3f16_to_v3f32_arg(<3 x float> addrspace(1)* %out, <3 x half> %arg) #0 {
   %ext = fpext <3 x half> %arg to <3 x float>
@@ -102,14 +101,14 @@ define amdgpu_kernel void @extload_v4f16_to_v4f32_arg(<4 x float> addrspace(1)*
 }
 
 ; GCN-LABEL: {{^}}extload_v8f16_to_v8f32_arg:
-; SI: buffer_load_ushort
-; SI: buffer_load_ushort
-; SI: buffer_load_ushort
-; SI: buffer_load_ushort
-; SI: buffer_load_ushort
-; SI: buffer_load_ushort
-; SI: buffer_load_ushort
-; SI: buffer_load_ushort
+; SI: flat_load_ushort
+; SI: flat_load_ushort
+; SI: flat_load_ushort
+; SI: flat_load_ushort
+; SI: flat_load_ushort
+; SI: flat_load_ushort
+; SI: flat_load_ushort
+; SI: flat_load_ushort
 
 
 ; VI: s_load_dword s
@@ -126,8 +125,8 @@ define amdgpu_kernel void @extload_v4f16_to_v4f32_arg(<4 x float> addrspace(1)*
 ; GCN: v_cvt_f32_f16_e32
 ; GCN: v_cvt_f32_f16_e32
 
-; GCN: buffer_store_dwordx4
-; GCN: buffer_store_dwordx4
+; GCN: flat_store_dwordx4
+; GCN: flat_store_dwordx4
 define amdgpu_kernel void @extload_v8f16_to_v8f32_arg(<8 x float> addrspace(1)* %out, <8 x half> %arg) #0 {
   %ext = fpext <8 x half> %arg to <8 x float>
   store <8 x float> %ext, <8 x float> addrspace(1)* %out
@@ -135,13 +134,10 @@ define amdgpu_kernel void @extload_v8f16_to_v8f32_arg(<8 x float> addrspace(1)*
 }
 
 ; GCN-LABEL: {{^}}extload_f16_to_f64_arg:
-; SI: s_load_dword [[ARG:s[0-9]+]], s{{\[[0-9]+:[0-9]+\]}}, 0xb{{$}}
-; SI: v_cvt_f64_f32_e32 [[RESULT:v\[[0-9]+:[0-9]+\]]], [[ARG]]
-; VI: s_load_dword [[ARG:s[0-9]+]], s{{\[[0-9]+:[0-9]+\]}}, 0x2c{{$}}
-; VI: v_trunc_f16_e32 v[[VARG:[0-9]+]], [[ARG]]
-; VI: v_cvt_f32_f16_e32 v[[VARG_F32:[0-9]+]], v[[VARG]]
-; VI: v_cvt_f64_f32_e32 [[RESULT:v\[[0-9]+:[0-9]+\]]], v[[VARG_F32]]
-; GCN: buffer_store_dwordx2 [[RESULT]]
+; GCN: flat_load_ushort [[ARG:v[0-9]+]]
+; GCN: v_cvt_f32_f16_e32 v[[ARG_F32:[0-9]+]], [[ARG]]
+; GCN: v_cvt_f64_f32_e32 [[RESULT:v\[[0-9]+:[0-9]+\]]], v[[ARG_F32]]
+; GCN: flat_store_dwordx2 v{{\[[0-9]+:[0-9]+\]}}, [[RESULT]]
 define amdgpu_kernel void @extload_f16_to_f64_arg(double addrspace(1)* %out, half %arg) #0 {
   %ext = fpext half %arg to double
   store double %ext, double addrspace(1)* %out
@@ -149,8 +145,8 @@ define amdgpu_kernel void @extload_f16_to_f64_arg(double addrspace(1)* %out, hal
 }
 
 ; GCN-LABEL: {{^}}extload_v2f16_to_v2f64_arg:
-; SI-DAG: buffer_load_ushort v
-; SI-DAG: buffer_load_ushort v
+; SI-DAG: flat_load_ushort v
+; SI-DAG: flat_load_ushort v
 
 ; VI-DAG: s_load_dword s
 ; VI: s_lshr_b32
@@ -167,9 +163,9 @@ define amdgpu_kernel void @extload_v2f16_to_v2f64_arg(<2 x double> addrspace(1)*
 }
 
 ; GCN-LABEL: {{^}}extload_v3f16_to_v3f64_arg:
-; GCN-DAG: buffer_load_ushort v
-; GCN-DAG: buffer_load_ushort v
-; GCN-DAG: buffer_load_ushort v
+; GCN-DAG: flat_load_ushort v
+; GCN-DAG: flat_load_ushort v
+; GCN-DAG: flat_load_ushort v
 ; GCN-DAG: v_cvt_f32_f16_e32
 ; GCN-DAG: v_cvt_f32_f16_e32
 ; GCN-DAG: v_cvt_f32_f16_e32
@@ -184,10 +180,10 @@ define amdgpu_kernel void @extload_v3f16_to_v3f64_arg(<3 x double> addrspace(1)*
 }
 
 ; GCN-LABEL: {{^}}extload_v4f16_to_v4f64_arg:
-; SI: buffer_load_ushort v
-; SI: buffer_load_ushort v
-; SI: buffer_load_ushort v
-; SI: buffer_load_ushort v
+; SI: flat_load_ushort v
+; SI: flat_load_ushort v
+; SI: flat_load_ushort v
+; SI: flat_load_ushort v
 
 ; VI: s_load_dword s
 ; VI: s_load_dword s
@@ -208,15 +204,15 @@ define amdgpu_kernel void @extload_v4f16_to_v4f64_arg(<4 x double> addrspace(1)*
 }
 
 ; GCN-LABEL: {{^}}extload_v8f16_to_v8f64_arg:
-; SI: buffer_load_ushort v
-; SI: buffer_load_ushort v
-; SI: buffer_load_ushort v
-; SI: buffer_load_ushort v
+; SI: flat_load_ushort v
+; SI: flat_load_ushort v
+; SI: flat_load_ushort v
+; SI: flat_load_ushort v
 
-; SI: buffer_load_ushort v
-; SI: buffer_load_ushort v
-; SI: buffer_load_ushort v
-; SI: buffer_load_ushort v
+; SI: flat_load_ushort v
+; SI: flat_load_ushort v
+; SI: flat_load_ushort v
+; SI: flat_load_ushort v
 
 
 ; VI: s_load_dword s
@@ -254,8 +250,8 @@ define amdgpu_kernel void @extload_v8f16_to_v8f64_arg(<8 x double> addrspace(1)*
 }
 
 ; GCN-LABEL: {{^}}global_load_store_f16:
-; GCN: buffer_load_ushort [[TMP:v[0-9]+]]
-; GCN: buffer_store_short [[TMP]]
+; GCN: flat_load_ushort [[TMP:v[0-9]+]]
+; GCN: flat_store_short v{{\[[0-9]+:[0-9]+\]}}, [[TMP]]
 define amdgpu_kernel void @global_load_store_f16(half addrspace(1)* %out, half addrspace(1)* %in) #0 {
   %val = load half, half addrspace(1)* %in
   store half %val, half addrspace(1)* %out
@@ -263,8 +259,8 @@ define amdgpu_kernel void @global_load_store_f16(half addrspace(1)* %out, half a
 }
 
 ; GCN-LABEL: {{^}}global_load_store_v2f16:
-; GCN: buffer_load_dword [[TMP:v[0-9]+]]
-; GCN: buffer_store_dword [[TMP]]
+; GCN: flat_load_dword [[TMP:v[0-9]+]]
+; GCN: flat_store_dword v{{\[[0-9]+:[0-9]+\]}}, [[TMP]]
 define amdgpu_kernel void @global_load_store_v2f16(<2 x half> addrspace(1)* %out, <2 x half> addrspace(1)* %in) #0 {
   %val = load <2 x half>, <2 x half> addrspace(1)* %in
   store <2 x half> %val, <2 x half> addrspace(1)* %out
@@ -272,8 +268,8 @@ define amdgpu_kernel void @global_load_store_v2f16(<2 x half> addrspace(1)* %out
 }
 
 ; GCN-LABEL: {{^}}global_load_store_v4f16:
-; GCN: buffer_load_dwordx2 [[TMP:v\[[0-9]+:[0-9]+\]]]
-; GCN: buffer_store_dwordx2 [[TMP]]
+; GCN: flat_load_dwordx2 [[TMP:v\[[0-9]+:[0-9]+\]]]
+; GCN: flat_store_dwordx2 v{{\[[0-9]+:[0-9]+\]}}, [[TMP]]
 define amdgpu_kernel void @global_load_store_v4f16(<4 x half> addrspace(1)* %in, <4 x half> addrspace(1)* %out) #0 {
   %val = load <4 x half>, <4 x half> addrspace(1)* %in
   store <4 x half> %val, <4 x half> addrspace(1)* %out
@@ -281,8 +277,8 @@ define amdgpu_kernel void @global_load_store_v4f16(<4 x half> addrspace(1)* %in,
 }
 
 ; GCN-LABEL: {{^}}global_load_store_v8f16:
-; GCN: buffer_load_dwordx4 [[TMP:v\[[0-9]+:[0-9]+\]]]
-; GCN: buffer_store_dwordx4 [[TMP:v\[[0-9]+:[0-9]+\]]]
+; GCN: flat_load_dwordx4 [[TMP:v\[[0-9]+:[0-9]+\]]]
+; GCN: flat_store_dwordx4 v{{\[[0-9]+:[0-9]+\]}}, [[TMP:v\[[0-9]+:[0-9]+\]]]
 ; GCN: s_endpgm
 define amdgpu_kernel void @global_load_store_v8f16(<8 x half> addrspace(1)* %out, <8 x half> addrspace(1)* %in) #0 {
   %val = load <8 x half>, <8 x half> addrspace(1)* %in
@@ -291,9 +287,9 @@ define amdgpu_kernel void @global_load_store_v8f16(<8 x half> addrspace(1)* %out
 }
 
 ; GCN-LABEL: {{^}}global_extload_f16_to_f32:
-; GCN: buffer_load_ushort [[LOAD:v[0-9]+]]
+; GCN: flat_load_ushort [[LOAD:v[0-9]+]]
 ; GCN: v_cvt_f32_f16_e32 [[CVT:v[0-9]+]], [[LOAD]]
-; GCN: buffer_store_dword [[CVT]]
+; GCN: flat_store_dword v{{\[[0-9]+:[0-9]+\]}}, [[CVT]]
 define amdgpu_kernel void @global_extload_f16_to_f32(float addrspace(1)* %out, half addrspace(1)* %in) #0 {
   %val = load half, half addrspace(1)* %in
   %cvt = fpext half %val to float
@@ -302,12 +298,16 @@ define amdgpu_kernel void @global_extload_f16_to_f32(float addrspace(1)* %out, h
 }
 
 ; GCN-LABEL: {{^}}global_extload_v2f16_to_v2f32:
-; GCN: buffer_load_dword [[LOAD:v[0-9]+]], off, s{{\[[0-9]+:[0-9]+\]}}, 0{{$}}
-; GCN: v_cvt_f32_f16_e32 v[[CVT0:[0-9]+]], [[LOAD]]
+; GCN: flat_load_dword [[LOAD:v[0-9]+]],
+
 ; SI: v_lshrrev_b32_e32 [[HI:v[0-9]+]], 16, [[LOAD]]
+; SI: v_cvt_f32_f16_e32 v[[CVT0:[0-9]+]], [[LOAD]]
 ; SI: v_cvt_f32_f16_e32 v[[CVT1:[0-9]+]], [[HI]]
+
 ; VI: v_cvt_f32_f16_sdwa v[[CVT1:[0-9]+]], [[LOAD]] dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_1
-; GCN: buffer_store_dwordx2 v{{\[}}[[CVT0]]:[[CVT1]]{{\]}}
+; VI: v_cvt_f32_f16_e32 v[[CVT0:[0-9]+]], [[LOAD]]
+
+; GCN: flat_store_dwordx2 v{{\[[0-9]+:[0-9]+\]}}, v{{\[}}[[CVT0]]:[[CVT1]]{{\]}}
 ; GCN: s_endpgm
 define amdgpu_kernel void @global_extload_v2f16_to_v2f32(<2 x float> addrspace(1)* %out, <2 x half> addrspace(1)* %in) #0 {
   %val = load <2 x half>, <2 x half> addrspace(1)* %in
@@ -341,14 +341,17 @@ define amdgpu_kernel void @global_extload_v8f16_to_v8f32(<8 x float> addrspace(1
 }
 
 ; GCN-LABEL: {{^}}global_extload_v16f16_to_v16f32:
-; GCN: buffer_load_dwordx4
-; GCN: buffer_load_dwordx4
+; GCN: flat_load_dwordx4
+; GCN: flat_load_dwordx4
 
 ; SI: v_cvt_f32_f16_e32
 ; SI: v_cvt_f32_f16_e32
 ; SI: v_cvt_f32_f16_e32
 ; SI: v_cvt_f32_f16_e32
 ; SI: v_cvt_f32_f16_e32
+
+; GCN: flat_store_dwordx4
+
 ; SI: v_cvt_f32_f16_e32
 ; SI: v_cvt_f32_f16_e32
 ; SI: v_cvt_f32_f16_e32
@@ -363,12 +366,11 @@ define amdgpu_kernel void @global_extload_v8f16_to_v8f32(<8 x float> addrspace(1
 
 ; VI: v_cvt_f32_f16_e32
 ; VI: v_cvt_f32_f16_sdwa
-; ...
 
-; GCN: buffer_store_dwordx4
-; GCN: buffer_store_dwordx4
-; GCN: buffer_store_dwordx4
-; GCN: buffer_store_dwordx4
+
+; GCN: flat_store_dwordx4
+; GCN: flat_store_dwordx4
+; GCN: flat_store_dwordx4
 
 ; GCN: s_endpgm
 define amdgpu_kernel void @global_extload_v16f16_to_v16f32(<16 x float> addrspace(1)* %out, <16 x half> addrspace(1)* %in) #0 {
@@ -379,10 +381,10 @@ define amdgpu_kernel void @global_extload_v16f16_to_v16f32(<16 x float> addrspac
 }
 
 ; GCN-LABEL: {{^}}global_extload_f16_to_f64:
-; GCN: buffer_load_ushort [[LOAD:v[0-9]+]]
+; GCN: flat_load_ushort [[LOAD:v[0-9]+]]
 ; GCN: v_cvt_f32_f16_e32 [[CVT0:v[0-9]+]], [[LOAD]]
 ; GCN: v_cvt_f64_f32_e32 [[CVT1:v\[[0-9]+:[0-9]+\]]], [[CVT0]]
-; GCN: buffer_store_dwordx2 [[CVT1]]
+; GCN: flat_store_dwordx2 v{{\[[0-9]+:[0-9]+\]}}, [[CVT1]]
 define amdgpu_kernel void @global_extload_f16_to_f64(double addrspace(1)* %out, half addrspace(1)* %in) #0 {
   %val = load half, half addrspace(1)* %in
   %cvt = fpext half %val to double
@@ -391,7 +393,7 @@ define amdgpu_kernel void @global_extload_f16_to_f64(double addrspace(1)* %out,
 }
 
 ; GCN-LABEL: {{^}}global_extload_v2f16_to_v2f64:
-; GCN-DAG: buffer_load_dword [[LOAD:v[0-9]+]], off, s{{\[[0-9]+:[0-9]+\]}}, 0{{$}}
+; GCN-DAG: flat_load_dword [[LOAD:v[0-9]+]],
 
 ; SI-DAG: v_lshrrev_b32_e32 [[HI:v[0-9]+]], 16, [[LOAD]]
 ; SI-DAG: v_cvt_f32_f16_e32 v[[CVT0:[0-9]+]], [[LOAD]]
@@ -404,7 +406,7 @@ define amdgpu_kernel void @global_extload_f16_to_f64(double addrspace(1)* %out,
 ; VI-DAG: v_cvt_f64_f32_e32 v{{\[}}[[CVT3_LO:[0-9]+]]:[[CVT3_HI:[0-9]+]]{{\]}}, v[[CVT0]]
 ; VI-DAG: v_cvt_f64_f32_e32 v{{\[}}[[CVT2_LO:[0-9]+]]:[[CVT2_HI:[0-9]+]]{{\]}}, v[[CVT1]]
 
-; GCN-DAG: buffer_store_dwordx4 v{{\[}}[[CVT2_LO]]:[[CVT3_HI]]{{\]}}
+; GCN-DAG: flat_store_dwordx4 v{{\[[0-9]+:[0-9]+\]}}, v{{\[}}[[CVT2_LO]]:[[CVT3_HI]]{{\]}}
 ; GCN: s_endpgm
 define amdgpu_kernel void @global_extload_v2f16_to_v2f64(<2 x double> addrspace(1)* %out, <2 x half> addrspace(1)* %in) #0 {
   %val = load <2 x half>, <2 x half> addrspace(1)* %in
@@ -415,23 +417,23 @@ define amdgpu_kernel void @global_extload_v2f16_to_v2f64(<2 x double> addrspace(
 
 ; GCN-LABEL: {{^}}global_extload_v3f16_to_v3f64:
 
-; XSI: buffer_load_dwordx2 [[LOAD:v\[[0-9]+:[0-9]+\]]]
+; XSI: flat_load_dwordx2 [[LOAD:v\[[0-9]+:[0-9]+\]]]
 ; XSI: v_cvt_f32_f16_e32
 ; XSI: v_cvt_f32_f16_e32
 ; XSI-DAG: v_lshrrev_b32_e32 {{v[0-9]+}}, 16, {{v[0-9]+}}
 ; XSI: v_cvt_f32_f16_e32
 ; XSI-NOT: v_cvt_f32_f16
 
-; XVI: buffer_load_dwordx2 [[LOAD:v\[[0-9]+:[0-9]+\]]]
+; XVI: flat_load_dwordx2 [[LOAD:v\[[0-9]+:[0-9]+\]]]
 ; XVI: v_cvt_f32_f16_e32
 ; XVI: v_cvt_f32_f16_e32
 ; XVI: v_cvt_f32_f16_sdwa
 ; XVI-NOT: v_cvt_f32_f16
 
-; GCN: buffer_load_dwordx2 v{{\[}}[[IN_LO:[0-9]+]]:[[IN_HI:[0-9]+]]
+; GCN: flat_load_dwordx2 v{{\[}}[[IN_LO:[0-9]+]]:[[IN_HI:[0-9]+]]
 ; GCN-DAG: v_cvt_f32_f16_e32 [[Z32:v[0-9]+]], v[[IN_HI]]
 ; GCN-DAG: v_cvt_f32_f16_e32 [[X32:v[0-9]+]], v[[IN_LO]]
-; SI:      v_lshrrev_b32_e32 [[Y16:v[0-9]+]], 16, v[[IN_LO]]
+; SI-DAG:      v_lshrrev_b32_e32 [[Y16:v[0-9]+]], 16, v[[IN_LO]]
 ; SI-DAG:  v_cvt_f32_f16_e32 [[Y32:v[0-9]+]], [[Y16]]
 ; VI-DAG:  v_cvt_f32_f16_sdwa [[Y32:v[0-9]+]], v[[IN_LO]] dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_1
 
@@ -440,8 +442,8 @@ define amdgpu_kernel void @global_extload_v2f16_to_v2f64(<2 x double> addrspace(
 ; GCN-DAG: v_cvt_f64_f32_e32 v[{{[0-9]+}}:[[YHI:[0-9]+]]{{\]}}, [[Y32]]
 ; GCN-NOT: v_cvt_f64_f32_e32
 
-; GCN-DAG: buffer_store_dwordx4 v{{\[}}[[XLO]]:[[YHI]]{{\]}}, off, s{{\[[0-9]+:[0-9]+\]}}, 0{{$}}
-; GCN-DAG: buffer_store_dwordx2 [[Z]], off, s{{\[[0-9]+:[0-9]+\]}}, 0 offset:16
+; GCN-DAG: flat_store_dwordx4 v{{\[[0-9]+:[0-9]+\]}}, v{{\[}}[[XLO]]:[[YHI]]{{\]}}
+; GCN-DAG: flat_store_dwordx2 v{{\[[0-9]+:[0-9]+\]}}, [[Z]]
 ; GCN: s_endpgm
 define amdgpu_kernel void @global_extload_v3f16_to_v3f64(<3 x double> addrspace(1)* %out, <3 x half> addrspace(1)* %in) #0 {
   %val = load <3 x half>, <3 x half> addrspace(1)* %in
@@ -475,9 +477,9 @@ define amdgpu_kernel void @global_extload_v16f16_to_v16f64(<16 x double> addrspa
 }
 
 ; GCN-LABEL: {{^}}global_truncstore_f32_to_f16:
-; GCN: buffer_load_dword [[LOAD:v[0-9]+]]
+; GCN: flat_load_dword [[LOAD:v[0-9]+]]
 ; GCN: v_cvt_f16_f32_e32 [[CVT:v[0-9]+]], [[LOAD]]
-; GCN: buffer_store_short [[CVT]]
+; GCN: flat_store_short v{{\[[0-9]+:[0-9]+\]}}, [[CVT]]
 define amdgpu_kernel void @global_truncstore_f32_to_f16(half addrspace(1)* %out, float addrspace(1)* %in) #0 {
   %val = load float, float addrspace(1)* %in
   %cvt = fptrunc float %val to half
@@ -486,7 +488,7 @@ define amdgpu_kernel void @global_truncstore_f32_to_f16(half addrspace(1)* %out,
 }
 
 ; GCN-LABEL: {{^}}global_truncstore_v2f32_to_v2f16:
-; GCN: buffer_load_dwordx2 v{{\[}}[[LO:[0-9]+]]:[[HI:[0-9]+]]{{\]}}
+; GCN: flat_load_dwordx2 v{{\[}}[[LO:[0-9]+]]:[[HI:[0-9]+]]{{\]}}
 ; GCN-DAG: v_cvt_f16_f32_e32 [[CVT0:v[0-9]+]], v[[LO]]
 
 ; SI-DAG: v_cvt_f16_f32_e32 [[CVT1:v[0-9]+]], v[[HI]]
@@ -496,7 +498,7 @@ define amdgpu_kernel void @global_truncstore_f32_to_f16(half addrspace(1)* %out,
 ; VI-DAG: v_cvt_f16_f32_sdwa [[CVT1:v[0-9]+]], v[[HI]] dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:DWORD
 ; VI:     v_or_b32_e32 [[PACKED:v[0-9]+]], [[CVT0]], [[CVT1]]
 
-; GCN-DAG: buffer_store_dword [[PACKED]]
+; GCN-DAG: flat_store_dword v{{\[[0-9]+:[0-9]+\]}}, [[PACKED]]
 ; GCN: s_endpgm
 define amdgpu_kernel void @global_truncstore_v2f32_to_v2f16(<2 x half> addrspace(1)* %out, <2 x float> addrspace(1)* %in) #0 {
   %val = load <2 x float>, <2 x float> addrspace(1)* %in
@@ -506,13 +508,13 @@ define amdgpu_kernel void @global_truncstore_v2f32_to_v2f16(<2 x half> addrspace
 }
 
 ; GCN-LABEL: {{^}}global_truncstore_v3f32_to_v3f16:
-; GCN: buffer_load_dwordx4
+; GCN: flat_load_dwordx4
 ; GCN-DAG: v_cvt_f16_f32_e32
 ; SI-DAG:  v_cvt_f16_f32_e32
 ; VI-DAG:  v_cvt_f16_f32_sdwa
 ; GCN-DAG: v_cvt_f16_f32_e32
-; GCN: buffer_store_short
-; GCN: buffer_store_dword
+; GCN: flat_store_short
+; GCN: flat_store_dword
 ; GCN: s_endpgm
 define amdgpu_kernel void @global_truncstore_v3f32_to_v3f16(<3 x half> addrspace(1)* %out, <3 x float> addrspace(1)* %in) #0 {
   %val = load <3 x float>, <3 x float> addrspace(1)* %in
@@ -522,14 +524,14 @@ define amdgpu_kernel void @global_truncstore_v3f32_to_v3f16(<3 x half> addrspace
 }
 
 ; GCN-LABEL: {{^}}global_truncstore_v4f32_to_v4f16:
-; GCN: buffer_load_dwordx4
+; GCN: flat_load_dwordx4
 ; GCN-DAG: v_cvt_f16_f32_e32
 ; SI-DAG:  v_cvt_f16_f32_e32
 ; SI-DAG:  v_cvt_f16_f32_e32
 ; VI-DAG:  v_cvt_f16_f32_sdwa
 ; VI-DAG:  v_cvt_f16_f32_sdwa
 ; GCN-DAG: v_cvt_f16_f32_e32
-; GCN: buffer_store_dwordx2
+; GCN: flat_store_dwordx2
 ; GCN: s_endpgm
 define amdgpu_kernel void @global_truncstore_v4f32_to_v4f16(<4 x half> addrspace(1)* %out, <4 x float> addrspace(1)* %in) #0 {
   %val = load <4 x float>, <4 x float> addrspace(1)* %in
@@ -539,8 +541,8 @@ define amdgpu_kernel void @global_truncstore_v4f32_to_v4f16(<4 x half> addrspace
 }
 
 ; GCN-LABEL: {{^}}global_truncstore_v8f32_to_v8f16:
-; GCN: buffer_load_dwordx4
-; GCN: buffer_load_dwordx4
+; GCN: flat_load_dwordx4
+; GCN: flat_load_dwordx4
 ; SI:  v_cvt_f16_f32_e32
 ; SI:  v_cvt_f16_f32_e32
 ; SI:  v_cvt_f16_f32_e32
@@ -557,7 +559,7 @@ define amdgpu_kernel void @global_truncstore_v4f32_to_v4f16(<4 x half> addrspace
 ; VI-DAG:  v_cvt_f16_f32_sdwa
 ; VI-DAG:  v_cvt_f16_f32_sdwa
 ; VI-DAG:  v_cvt_f16_f32_sdwa
-; GCN: buffer_store_dwordx4
+; GCN: flat_store_dwordx4
 ; GCN: s_endpgm
 define amdgpu_kernel void @global_truncstore_v8f32_to_v8f16(<8 x half> addrspace(1)* %out, <8 x float> addrspace(1)* %in) #0 {
   %val = load <8 x float>, <8 x float> addrspace(1)* %in
@@ -567,10 +569,10 @@ define amdgpu_kernel void @global_truncstore_v8f32_to_v8f16(<8 x half> addrspace
 }
 
 ; GCN-LABEL: {{^}}global_truncstore_v16f32_to_v16f16:
-; GCN: buffer_load_dwordx4
-; GCN: buffer_load_dwordx4
-; GCN: buffer_load_dwordx4
-; GCN: buffer_load_dwordx4
+; GCN: flat_load_dwordx4
+; GCN: flat_load_dwordx4
+; GCN: flat_load_dwordx4
+; GCN: flat_load_dwordx4
 ; GCN-DAG: v_cvt_f16_f32_e32
 ; GCN-DAG: v_cvt_f16_f32_e32
 ; GCN-DAG: v_cvt_f16_f32_e32
@@ -587,8 +589,8 @@ define amdgpu_kernel void @global_truncstore_v8f32_to_v8f16(<8 x half> addrspace
 ; GCN-DAG: v_cvt_f16_f32_e32
 ; GCN-DAG: v_cvt_f16_f32_e32
 ; GCN-DAG: v_cvt_f16_f32_e32
-; GCN-DAG: buffer_store_dwordx4
-; GCN-DAG: buffer_store_dwordx4
+; GCN-DAG: flat_store_dwordx4
+; GCN-DAG: flat_store_dwordx4
 ; GCN: s_endpgm
 define amdgpu_kernel void @global_truncstore_v16f32_to_v16f16(<16 x half> addrspace(1)* %out, <16 x float> addrspace(1)* %in) #0 {
   %val = load <16 x float>, <16 x float> addrspace(1)* %in
@@ -653,8 +655,9 @@ define amdgpu_kernel void @fadd_v8f16(<8 x half> addrspace(1)* %out, <8 x half>
 }
 
 ; GCN-LABEL: {{^}}test_bitcast_from_half:
-; GCN: buffer_load_ushort [[TMP:v[0-9]+]]
-; GCN: buffer_store_short [[TMP]]
+; GCN: flat_load_ushort [[TMP:v[0-9]+]]
+; GCN-NOT: [[TMP]]
+; GCN: flat_store_short v{{\[[0-9]+:[0-9]+\]}}, [[TMP]]
 define amdgpu_kernel void @test_bitcast_from_half(half addrspace(1)* %in, i16 addrspace(1)* %out) #0 {
   %val = load half, half addrspace(1)* %in
   %val_int = bitcast half %val to i16
@@ -663,8 +666,8 @@ define amdgpu_kernel void @test_bitcast_from_half(half addrspace(1)* %in, i16 ad
 }
 
 ; GCN-LABEL: {{^}}test_bitcast_to_half:
-; GCN: buffer_load_ushort [[TMP:v[0-9]+]]
-; GCN: buffer_store_short [[TMP]]
+; GCN: flat_load_ushort [[TMP:v[0-9]+]]
+; GCN: flat_store_short v{{\[[0-9]+:[0-9]+\]}}, [[TMP]]
 define amdgpu_kernel void @test_bitcast_to_half(half addrspace(1)* %out, i16 addrspace(1)* %in) #0 {
   %val = load i16, i16 addrspace(1)* %in
   %val_fp = bitcast i16 %val to half