[AMDGPU] Regenerate v2i16 insertelement tests.

To help show the diffs from an upcoming SimplifyDemandedBits patch.

llvm-svn: 367213

1 files changed, 1557 insertions, 337 deletions

diff --git a/llvm/test/CodeGen/AMDGPU/insert_vector_elt.v2i16.ll b/llvm/test/CodeGen/AMDGPU/insert_vector_elt.v2i16.ll
index a8a298045ff..f31c1b5c87f 100644
--- a/llvm/test/CodeGen/AMDGPU/insert_vector_elt.v2i16.ll
+++ b/llvm/test/CodeGen/AMDGPU/insert_vector_elt.v2i16.ll
@@ -1,64 +1,152 @@
-; RUN: llc -verify-machineinstrs -mtriple=amdgcn-amd-amdhsa -mcpu=gfx900 -enable-amdgpu-aa=0 -mattr=+flat-for-global,-fp64-fp16-denormals < %s | FileCheck -enable-var-scope -check-prefixes=GCN,GFX9,GFX89 %s
-; RUN: llc -verify-machineinstrs -mtriple=amdgcn-amd-amdhsa -mcpu=fiji -enable-amdgpu-aa=0 -mattr=+flat-for-global < %s | FileCheck -enable-var-scope -check-prefixes=GCN,CIVI,VI,GFX89 %s
+; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
+; RUN: llc -verify-machineinstrs -mtriple=amdgcn-amd-amdhsa -mcpu=gfx900 -enable-amdgpu-aa=0 -mattr=+flat-for-global,-fp64-fp16-denormals < %s | FileCheck -enable-var-scope -check-prefixes=GCN,GFX89,GFX9 %s
+; RUN: llc -verify-machineinstrs -mtriple=amdgcn-amd-amdhsa -mcpu=fiji -enable-amdgpu-aa=0 -mattr=+flat-for-global < %s | FileCheck -enable-var-scope -check-prefixes=GCN,GFX89,CIVI,VI %s
 ; RUN: llc -verify-machineinstrs -mtriple=amdgcn-amd-amdhsa -mcpu=hawaii -enable-amdgpu-aa=0 -mattr=+flat-for-global < %s | FileCheck -enable-var-scope -check-prefixes=GCN,CIVI,CI %s
 
-; GCN-LABEL: {{^}}s_insertelement_v2i16_0:
-; GCN: s_load_dword [[VEC:s[0-9]+]], s{{\[[0-9]+:[0-9]+\]}}, 0x0
-
-; CIVI: s_and_b32 [[ELT1:s[0-9]+]], [[VEC]], 0xffff0000{{$}}
-; CIVI: s_or_b32 s{{[0-9]+}}, [[ELT1]], 0x3e7{{$}}
-
-; GFX9-NOT: lshr
-; GFX9: s_pack_lh_b32_b16 s{{[0-9]+}}, 0x3e7, [[VEC]]
 define amdgpu_kernel void @s_insertelement_v2i16_0(<2 x i16> addrspace(1)* %out, <2 x i16> addrspace(4)* %vec.ptr) #0 {
+; GFX9-LABEL: s_insertelement_v2i16_0:
+; GFX9:       ; %bb.0:
+; GFX9-NEXT:    s_load_dwordx4 s[0:3], s[4:5], 0x0
+; GFX9-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX9-NEXT:    s_load_dword s2, s[2:3], 0x0
+; GFX9-NEXT:    v_mov_b32_e32 v0, s0
+; GFX9-NEXT:    v_mov_b32_e32 v1, s1
+; GFX9-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX9-NEXT:    s_pack_lh_b32_b16 s0, 0x3e7, s2
+; GFX9-NEXT:    v_mov_b32_e32 v2, s0
+; GFX9-NEXT:    global_store_dword v[0:1], v2, off
+; GFX9-NEXT:    s_endpgm
+;
+; CIVI-LABEL: s_insertelement_v2i16_0:
+; CIVI:       ; %bb.0:
+; CIVI-NEXT:    s_load_dwordx4 s[0:3], s[4:5], 0x0
+; CIVI-NEXT:    s_waitcnt lgkmcnt(0)
+; CIVI-NEXT:    s_load_dword s2, s[2:3], 0x0
+; CIVI-NEXT:    v_mov_b32_e32 v0, s0
+; CIVI-NEXT:    v_mov_b32_e32 v1, s1
+; CIVI-NEXT:    s_waitcnt lgkmcnt(0)
+; CIVI-NEXT:    s_and_b32 s0, s2, 0xffff0000
+; CIVI-NEXT:    s_or_b32 s0, s0, 0x3e7
+; CIVI-NEXT:    v_mov_b32_e32 v2, s0
+; CIVI-NEXT:    flat_store_dword v[0:1], v2
+; CIVI-NEXT:    s_endpgm
   %vec = load <2 x i16>, <2 x i16> addrspace(4)* %vec.ptr
   %vecins = insertelement <2 x i16> %vec, i16 999, i32 0
   store <2 x i16> %vecins, <2 x i16> addrspace(1)* %out
   ret void
 }
 
-; GCN-LABEL: {{^}}s_insertelement_v2i16_0_reg:
-; GCN-DAG: s_load_dword [[ELT_LOAD:s[0-9]+]], s[4:5],
-; GCN-DAG: s_load_dword [[VEC:s[0-9]+]], s{{\[[0-9]+:[0-9]+\]}}, 0x0
-
-; CIVI-DAG: s_and_b32 [[ELT0:s[0-9]+]], [[ELT_LOAD]], 0xffff{{$}}
-; CIVI-DAG: s_and_b32 [[ELT1:s[0-9]+]], [[VEC]], 0xffff0000{{$}}
-; CIVI: s_or_b32 s{{[0-9]+}}, [[ELT0]], [[ELT1]]
 
-; GFX9-NOT: [[ELT0]]
-; GFX9-NOT: [[VEC]]
-; GFX9: s_pack_lh_b32_b16 s{{[0-9]+}}, [[ELT_LOAD]], [[VEC]]
 define amdgpu_kernel void @s_insertelement_v2i16_0_reg(<2 x i16> addrspace(1)* %out, <2 x i16> addrspace(4)* %vec.ptr, [8 x i32], i16 %elt) #0 {
+; GFX9-LABEL: s_insertelement_v2i16_0_reg:
+; GFX9:       ; %bb.0:
+; GFX9-NEXT:    s_load_dwordx4 s[0:3], s[4:5], 0x0
+; GFX9-NEXT:    s_load_dword s4, s[4:5], 0x30
+; GFX9-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX9-NEXT:    s_load_dword s2, s[2:3], 0x0
+; GFX9-NEXT:    v_mov_b32_e32 v0, s0
+; GFX9-NEXT:    v_mov_b32_e32 v1, s1
+; GFX9-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX9-NEXT:    s_pack_lh_b32_b16 s0, s4, s2
+; GFX9-NEXT:    v_mov_b32_e32 v2, s0
+; GFX9-NEXT:    global_store_dword v[0:1], v2, off
+; GFX9-NEXT:    s_endpgm
+;
+; VI-LABEL: s_insertelement_v2i16_0_reg:
+; VI:       ; %bb.0:
+; VI-NEXT:    s_load_dwordx4 s[0:3], s[4:5], 0x0
+; VI-NEXT:    s_load_dword s4, s[4:5], 0x30
+; VI-NEXT:    s_waitcnt lgkmcnt(0)
+; VI-NEXT:    s_load_dword s2, s[2:3], 0x0
+; VI-NEXT:    v_mov_b32_e32 v0, s0
+; VI-NEXT:    v_mov_b32_e32 v1, s1
+; VI-NEXT:    s_and_b32 s0, s4, 0xffff
+; VI-NEXT:    s_waitcnt lgkmcnt(0)
+; VI-NEXT:    s_and_b32 s1, s2, 0xffff0000
+; VI-NEXT:    s_or_b32 s0, s0, s1
+; VI-NEXT:    v_mov_b32_e32 v2, s0
+; VI-NEXT:    flat_store_dword v[0:1], v2
+; VI-NEXT:    s_endpgm
+;
+; CI-LABEL: s_insertelement_v2i16_0_reg:
+; CI:       ; %bb.0:
+; CI-NEXT:    s_load_dwordx4 s[0:3], s[4:5], 0x0
+; CI-NEXT:    s_load_dword s4, s[4:5], 0xc
+; CI-NEXT:    s_waitcnt lgkmcnt(0)
+; CI-NEXT:    s_load_dword s2, s[2:3], 0x0
+; CI-NEXT:    v_mov_b32_e32 v0, s0
+; CI-NEXT:    v_mov_b32_e32 v1, s1
+; CI-NEXT:    s_and_b32 s1, s4, 0xffff
+; CI-NEXT:    s_waitcnt lgkmcnt(0)
+; CI-NEXT:    s_and_b32 s0, s2, 0xffff0000
+; CI-NEXT:    s_or_b32 s0, s1, s0
+; CI-NEXT:    v_mov_b32_e32 v2, s0
+; CI-NEXT:    flat_store_dword v[0:1], v2
+; CI-NEXT:    s_endpgm
   %vec = load <2 x i16>, <2 x i16> addrspace(4)* %vec.ptr
   %vecins = insertelement <2 x i16> %vec, i16 %elt, i32 0
   store <2 x i16> %vecins, <2 x i16> addrspace(1)* %out
   ret void
 }
 
-; GCN-LABEL: {{^}}s_insertelement_v2i16_0_multi_use_hi_reg:
-; GCN-DAG: s_load_dword [[ELT_LOAD:s[0-9]+]], s[4:5],
-; GCN-DAG: s_load_dword [[VEC:s[0-9]+]], s{{\[[0-9]+:[0-9]+\]}}, 0x0
-
-; CI-DAG: s_and_b32 [[ELT0_MASKED:s[0-9]+]], [[ELT_LOAD]], 0xffff{{$}}
-; CI: s_lshr_b32 [[SHR:s[0-9]+]], [[VEC]], 16
-; CI: s_lshl_b32 [[ELT1:s[0-9]+]], [[SHR]], 16
-; CI-DAG: s_or_b32 s{{[0-9]+}}, [[ELT0_MASKED]], [[ELT1]]
-; CI-DAG: ; use [[SHR]]
-
-
-; FIXME: Should be able to void mask of upper bits
-; VI-DAG: s_and_b32 [[ELT_MASKED:s[0-9]+]], [[ELT_LOAD]], 0xffff{{$}}
-; VI-DAG: s_and_b32 [[VEC_HIMASK:s[0-9]+]], [[VEC]], 0xffff0000{{$}}
-; VI-DAG: s_or_b32 [[OR:s[0-9]+]], [[ELT_MASKED]], [[VEC_HIMASK]]
-; VI-DAG: s_lshr_b32 [[SHR:s[0-9]+]], [[VEC]], 16
-
-; VI-DAG: ; use [[SHR]]
-
-
-; GFX9: s_lshr_b32 [[ELT1:s[0-9]+]], [[VEC]], 16
-; GFX9-DAG: s_pack_ll_b32_b16 s{{[0-9]+}}, [[ELT_LOAD]], [[ELT1]]
-; GFX9-DAG: ; use [[ELT1]]
 define amdgpu_kernel void @s_insertelement_v2i16_0_multi_use_hi_reg(<2 x i16> addrspace(1)* %out, <2 x i16> addrspace(4)* %vec.ptr, [8 x i32], i16 %elt) #0 {
+; GFX9-LABEL: s_insertelement_v2i16_0_multi_use_hi_reg:
+; GFX9:       ; %bb.0:
+; GFX9-NEXT:    s_load_dwordx4 s[0:3], s[4:5], 0x0
+; GFX9-NEXT:    s_load_dword s4, s[4:5], 0x30
+; GFX9-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX9-NEXT:    s_load_dword s2, s[2:3], 0x0
+; GFX9-NEXT:    v_mov_b32_e32 v0, s0
+; GFX9-NEXT:    v_mov_b32_e32 v1, s1
+; GFX9-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX9-NEXT:    s_lshr_b32 s0, s2, 16
+; GFX9-NEXT:    s_pack_ll_b32_b16 s1, s4, s0
+; GFX9-NEXT:    v_mov_b32_e32 v2, s1
+; GFX9-NEXT:    global_store_dword v[0:1], v2, off
+; GFX9-NEXT:    ;;#ASMSTART
+; GFX9-NEXT:    ; use s0
+; GFX9-NEXT:    ;;#ASMEND
+; GFX9-NEXT:    s_endpgm
+;
+; VI-LABEL: s_insertelement_v2i16_0_multi_use_hi_reg:
+; VI:       ; %bb.0:
+; VI-NEXT:    s_load_dwordx4 s[0:3], s[4:5], 0x0
+; VI-NEXT:    s_load_dword s4, s[4:5], 0x30
+; VI-NEXT:    s_waitcnt lgkmcnt(0)
+; VI-NEXT:    s_load_dword s2, s[2:3], 0x0
+; VI-NEXT:    v_mov_b32_e32 v0, s0
+; VI-NEXT:    v_mov_b32_e32 v1, s1
+; VI-NEXT:    s_and_b32 s0, s4, 0xffff
+; VI-NEXT:    s_waitcnt lgkmcnt(0)
+; VI-NEXT:    s_lshr_b32 s1, s2, 16
+; VI-NEXT:    s_and_b32 s2, s2, 0xffff0000
+; VI-NEXT:    s_or_b32 s0, s0, s2
+; VI-NEXT:    v_mov_b32_e32 v2, s0
+; VI-NEXT:    flat_store_dword v[0:1], v2
+; VI-NEXT:    ;;#ASMSTART
+; VI-NEXT:    ; use s1
+; VI-NEXT:    ;;#ASMEND
+; VI-NEXT:    s_endpgm
+;
+; CI-LABEL: s_insertelement_v2i16_0_multi_use_hi_reg:
+; CI:       ; %bb.0:
+; CI-NEXT:    s_load_dwordx4 s[0:3], s[4:5], 0x0
+; CI-NEXT:    s_load_dword s4, s[4:5], 0xc
+; CI-NEXT:    s_waitcnt lgkmcnt(0)
+; CI-NEXT:    s_load_dword s2, s[2:3], 0x0
+; CI-NEXT:    v_mov_b32_e32 v1, s1
+; CI-NEXT:    v_mov_b32_e32 v0, s0
+; CI-NEXT:    s_and_b32 s0, s4, 0xffff
+; CI-NEXT:    s_waitcnt lgkmcnt(0)
+; CI-NEXT:    s_lshr_b32 s1, s2, 16
+; CI-NEXT:    s_lshl_b32 s2, s1, 16
+; CI-NEXT:    s_or_b32 s0, s0, s2
+; CI-NEXT:    v_mov_b32_e32 v2, s0
+; CI-NEXT:    flat_store_dword v[0:1], v2
+; CI-NEXT:    ;;#ASMSTART
+; CI-NEXT:    ; use s1
+; CI-NEXT:    ;;#ASMEND
+; CI-NEXT:    s_endpgm
   %vec = load <2 x i16>, <2 x i16> addrspace(4)* %vec.ptr
   %elt1 = extractelement <2 x i16> %vec, i32 1
   %vecins = insertelement <2 x i16> %vec, i16 %elt, i32 0
@@ -68,18 +156,52 @@ define amdgpu_kernel void @s_insertelement_v2i16_0_multi_use_hi_reg(<2 x i16> ad
   ret void
 }
 
-; GCN-LABEL: {{^}}s_insertelement_v2i16_0_reghi:
-; GCN-DAG: s_load_dword [[ELT_ARG:s[0-9]+]], s[4:5],
-; GCN-DAG: s_load_dword [[VEC:s[0-9]+]], s{{\[[0-9]+:[0-9]+\]}}, 0x0
-
-; CIVI: s_lshr_b32 [[ELT_HI:s[0-9]+]], [[ELT_ARG]], 16
-; CIVI-DAG: s_and_b32 [[ELT1:s[0-9]+]], [[VEC]], 0xffff0000{{$}}
-; CIVI: s_or_b32 s{{[0-9]+}}, [[ELT_HI]], [[ELT1]]
-
-; GFX9-NOT: [[ELT0]]
-; GFX9-NOT: [[VEC]]
-; GFX9: s_pack_hh_b32_b16 s{{[0-9]+}}, [[ELT_ARG]], [[VEC]]
 define amdgpu_kernel void @s_insertelement_v2i16_0_reghi(<2 x i16> addrspace(1)* %out, <2 x i16> addrspace(4)* %vec.ptr, [8 x i32], i32 %elt.arg) #0 {
+; GFX9-LABEL: s_insertelement_v2i16_0_reghi:
+; GFX9:       ; %bb.0:
+; GFX9-NEXT:    s_load_dwordx4 s[0:3], s[4:5], 0x0
+; GFX9-NEXT:    s_load_dword s4, s[4:5], 0x30
+; GFX9-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX9-NEXT:    s_load_dword s2, s[2:3], 0x0
+; GFX9-NEXT:    v_mov_b32_e32 v0, s0
+; GFX9-NEXT:    v_mov_b32_e32 v1, s1
+; GFX9-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX9-NEXT:    s_pack_hh_b32_b16 s0, s4, s2
+; GFX9-NEXT:    v_mov_b32_e32 v2, s0
+; GFX9-NEXT:    global_store_dword v[0:1], v2, off
+; GFX9-NEXT:    s_endpgm
+;
+; VI-LABEL: s_insertelement_v2i16_0_reghi:
+; VI:       ; %bb.0:
+; VI-NEXT:    s_load_dwordx4 s[0:3], s[4:5], 0x0
+; VI-NEXT:    s_load_dword s4, s[4:5], 0x30
+; VI-NEXT:    s_waitcnt lgkmcnt(0)
+; VI-NEXT:    s_load_dword s2, s[2:3], 0x0
+; VI-NEXT:    v_mov_b32_e32 v0, s0
+; VI-NEXT:    v_mov_b32_e32 v1, s1
+; VI-NEXT:    s_lshr_b32 s0, s4, 16
+; VI-NEXT:    s_waitcnt lgkmcnt(0)
+; VI-NEXT:    s_and_b32 s1, s2, 0xffff0000
+; VI-NEXT:    s_or_b32 s0, s0, s1
+; VI-NEXT:    v_mov_b32_e32 v2, s0
+; VI-NEXT:    flat_store_dword v[0:1], v2
+; VI-NEXT:    s_endpgm
+;
+; CI-LABEL: s_insertelement_v2i16_0_reghi:
+; CI:       ; %bb.0:
+; CI-NEXT:    s_load_dwordx4 s[0:3], s[4:5], 0x0
+; CI-NEXT:    s_load_dword s4, s[4:5], 0xc
+; CI-NEXT:    s_waitcnt lgkmcnt(0)
+; CI-NEXT:    s_load_dword s2, s[2:3], 0x0
+; CI-NEXT:    v_mov_b32_e32 v0, s0
+; CI-NEXT:    v_mov_b32_e32 v1, s1
+; CI-NEXT:    s_lshr_b32 s1, s4, 16
+; CI-NEXT:    s_waitcnt lgkmcnt(0)
+; CI-NEXT:    s_and_b32 s0, s2, 0xffff0000
+; CI-NEXT:    s_or_b32 s0, s1, s0
+; CI-NEXT:    v_mov_b32_e32 v2, s0
+; CI-NEXT:    flat_store_dword v[0:1], v2
+; CI-NEXT:    s_endpgm
   %vec = load <2 x i16>, <2 x i16> addrspace(4)* %vec.ptr
   %elt.hi = lshr i32 %elt.arg, 16
   %elt = trunc i32 %elt.hi to i16
@@ -88,18 +210,62 @@ define amdgpu_kernel void @s_insertelement_v2i16_0_reghi(<2 x i16> addrspace(1)*
   ret void
 }
 
-; GCN-LABEL: {{^}}s_insertelement_v2i16_0_reghi_multi_use_1:
-; GCN: s_load_dword [[ELT_ARG:s[0-9]+]],
-; GCN: s_load_dword [[VEC:s[0-9]+]],
-
-; CIVI-DAG: s_lshr_b32 [[ELT1:s[0-9]+]], [[ELT_ARG]], 16
-; CIVI-DAG: s_and_b32 [[ELT0:s[0-9]+]], [[VEC]], 0xffff0000{{$}}
-; CIVI: s_or_b32 s{{[0-9]+}}, [[ELT1]], [[ELT0]]
-
-; GFX9: s_lshr_b32 [[ELT1:s[0-9]+]], [[ELT_ARG]], 16
-; GFX9: s_pack_lh_b32_b16 s{{[0-9]+}}, [[ELT1]], [[VEC]]
-; GFX9: ; use [[ELT1]]
 define amdgpu_kernel void @s_insertelement_v2i16_0_reghi_multi_use_1(<2 x i16> addrspace(1)* %out, <2 x i16> addrspace(4)* %vec.ptr, i32 %elt.arg) #0 {
+; GFX9-LABEL: s_insertelement_v2i16_0_reghi_multi_use_1:
+; GFX9:       ; %bb.0:
+; GFX9-NEXT:    s_load_dwordx4 s[0:3], s[4:5], 0x0
+; GFX9-NEXT:    s_load_dword s4, s[4:5], 0x10
+; GFX9-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX9-NEXT:    s_load_dword s2, s[2:3], 0x0
+; GFX9-NEXT:    v_mov_b32_e32 v0, s0
+; GFX9-NEXT:    s_lshr_b32 s0, s4, 16
+; GFX9-NEXT:    v_mov_b32_e32 v1, s1
+; GFX9-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX9-NEXT:    s_pack_lh_b32_b16 s1, s0, s2
+; GFX9-NEXT:    v_mov_b32_e32 v2, s1
+; GFX9-NEXT:    global_store_dword v[0:1], v2, off
+; GFX9-NEXT:    ;;#ASMSTART
+; GFX9-NEXT:    ; use s0
+; GFX9-NEXT:    ;;#ASMEND
+; GFX9-NEXT:    s_endpgm
+;
+; VI-LABEL: s_insertelement_v2i16_0_reghi_multi_use_1:
+; VI:       ; %bb.0:
+; VI-NEXT:    s_load_dwordx4 s[0:3], s[4:5], 0x0
+; VI-NEXT:    s_load_dword s4, s[4:5], 0x10
+; VI-NEXT:    s_waitcnt lgkmcnt(0)
+; VI-NEXT:    s_load_dword s2, s[2:3], 0x0
+; VI-NEXT:    v_mov_b32_e32 v0, s0
+; VI-NEXT:    v_mov_b32_e32 v1, s1
+; VI-NEXT:    s_lshr_b32 s0, s4, 16
+; VI-NEXT:    s_waitcnt lgkmcnt(0)
+; VI-NEXT:    s_and_b32 s1, s2, 0xffff0000
+; VI-NEXT:    s_or_b32 s1, s0, s1
+; VI-NEXT:    v_mov_b32_e32 v2, s1
+; VI-NEXT:    flat_store_dword v[0:1], v2
+; VI-NEXT:    ;;#ASMSTART
+; VI-NEXT:    ; use s0
+; VI-NEXT:    ;;#ASMEND
+; VI-NEXT:    s_endpgm
+;
+; CI-LABEL: s_insertelement_v2i16_0_reghi_multi_use_1:
+; CI:       ; %bb.0:
+; CI-NEXT:    s_load_dwordx4 s[0:3], s[4:5], 0x0
+; CI-NEXT:    s_load_dword s4, s[4:5], 0x4
+; CI-NEXT:    s_waitcnt lgkmcnt(0)
+; CI-NEXT:    s_load_dword s2, s[2:3], 0x0
+; CI-NEXT:    v_mov_b32_e32 v0, s0
+; CI-NEXT:    v_mov_b32_e32 v1, s1
+; CI-NEXT:    s_lshr_b32 s0, s4, 16
+; CI-NEXT:    s_waitcnt lgkmcnt(0)
+; CI-NEXT:    s_and_b32 s1, s2, 0xffff0000
+; CI-NEXT:    s_or_b32 s1, s0, s1
+; CI-NEXT:    v_mov_b32_e32 v2, s1
+; CI-NEXT:    flat_store_dword v[0:1], v2
+; CI-NEXT:    ;;#ASMSTART
+; CI-NEXT:    ; use s0
+; CI-NEXT:    ;;#ASMEND
+; CI-NEXT:    s_endpgm
   %vec = load <2 x i16>, <2 x i16> addrspace(4)* %vec.ptr
   %elt.hi = lshr i32 %elt.arg, 16
   %elt = trunc i32 %elt.hi to i16
@@ -110,27 +276,74 @@ define amdgpu_kernel void @s_insertelement_v2i16_0_reghi_multi_use_1(<2 x i16> a
   ret void
 }
 
-; GCN-LABEL: {{^}}s_insertelement_v2i16_0_reghi_both_multi_use_1:
-; GCN: s_load_dword [[ELT_ARG:s[0-9]+]],
-; GCN: s_load_dword [[VEC:s[0-9]+]],
-
-; CI-DAG: s_lshr_b32 [[ELT_HI:s[0-9]+]], [[ELT_ARG]], 16
-; CI-DAG: s_lshr_b32 [[SHR:s[0-9]+]], [[VEC]], 16
-; CI-DAG: s_lshl_b32 [[VEC_HI:s[0-9]+]], [[SHR]], 16
-; CI: s_or_b32 s{{[0-9]+}}, [[ELT_HI]], [[VEC_HI]]
-
-
-; VI-DAG: s_lshr_b32 [[ELT_HI:s[0-9]+]], [[ELT_ARG]], 16
-; VI-DAG: s_lshr_b32 [[VEC_HI:s[0-9]+]], [[VEC]], 16
-; VI: s_and_b32 [[MASK_HI:s[0-9]+]], [[VEC]], 0xffff0000
-; VI: s_or_b32 s{{[0-9]+}}, [[ELT_HI]], [[MASK_HI]]
-
-; GFX9-DAG: s_lshr_b32 [[ELT_HI:s[0-9]+]], [[ELT_ARG]], 16
-; GFX9-DAG: s_lshr_b32 [[VEC_HI:s[0-9]+]], [[VEC]], 16
-; GFX9: s_pack_ll_b32_b16 s{{[0-9]+}}, [[ELT_HI]], [[VEC_HI]]
-; GFX9: ; use [[ELT_HI]]
-; GFX9: ; use [[VEC_HI]]
 define amdgpu_kernel void @s_insertelement_v2i16_0_reghi_both_multi_use_1(<2 x i16> addrspace(1)* %out, <2 x i16> addrspace(4)* %vec.ptr, i32 %elt.arg) #0 {
+; GFX9-LABEL: s_insertelement_v2i16_0_reghi_both_multi_use_1:
+; GFX9:       ; %bb.0:
+; GFX9-NEXT:    s_load_dwordx4 s[0:3], s[4:5], 0x0
+; GFX9-NEXT:    s_load_dword s4, s[4:5], 0x10
+; GFX9-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX9-NEXT:    s_load_dword s2, s[2:3], 0x0
+; GFX9-NEXT:    v_mov_b32_e32 v0, s0
+; GFX9-NEXT:    v_mov_b32_e32 v1, s1
+; GFX9-NEXT:    s_lshr_b32 s0, s4, 16
+; GFX9-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX9-NEXT:    s_lshr_b32 s1, s2, 16
+; GFX9-NEXT:    s_pack_ll_b32_b16 s2, s0, s1
+; GFX9-NEXT:    v_mov_b32_e32 v2, s2
+; GFX9-NEXT:    global_store_dword v[0:1], v2, off
+; GFX9-NEXT:    ;;#ASMSTART
+; GFX9-NEXT:    ; use s0
+; GFX9-NEXT:    ;;#ASMEND
+; GFX9-NEXT:    ;;#ASMSTART
+; GFX9-NEXT:    ; use s1
+; GFX9-NEXT:    ;;#ASMEND
+; GFX9-NEXT:    s_endpgm
+;
+; VI-LABEL: s_insertelement_v2i16_0_reghi_both_multi_use_1:
+; VI:       ; %bb.0:
+; VI-NEXT:    s_load_dwordx4 s[0:3], s[4:5], 0x0
+; VI-NEXT:    s_load_dword s4, s[4:5], 0x10
+; VI-NEXT:    s_waitcnt lgkmcnt(0)
+; VI-NEXT:    s_load_dword s2, s[2:3], 0x0
+; VI-NEXT:    v_mov_b32_e32 v0, s0
+; VI-NEXT:    v_mov_b32_e32 v1, s1
+; VI-NEXT:    s_lshr_b32 s0, s4, 16
+; VI-NEXT:    s_waitcnt lgkmcnt(0)
+; VI-NEXT:    s_lshr_b32 s1, s2, 16
+; VI-NEXT:    s_and_b32 s2, s2, 0xffff0000
+; VI-NEXT:    s_or_b32 s2, s0, s2
+; VI-NEXT:    v_mov_b32_e32 v2, s2
+; VI-NEXT:    flat_store_dword v[0:1], v2
+; VI-NEXT:    ;;#ASMSTART
+; VI-NEXT:    ; use s0
+; VI-NEXT:    ;;#ASMEND
+; VI-NEXT:    ;;#ASMSTART
+; VI-NEXT:    ; use s1
+; VI-NEXT:    ;;#ASMEND
+; VI-NEXT:    s_endpgm
+;
+; CI-LABEL: s_insertelement_v2i16_0_reghi_both_multi_use_1:
+; CI:       ; %bb.0:
+; CI-NEXT:    s_load_dwordx4 s[0:3], s[4:5], 0x0
+; CI-NEXT:    s_load_dword s4, s[4:5], 0x4
+; CI-NEXT:    s_waitcnt lgkmcnt(0)
+; CI-NEXT:    s_load_dword s2, s[2:3], 0x0
+; CI-NEXT:    v_mov_b32_e32 v1, s1
+; CI-NEXT:    v_mov_b32_e32 v0, s0
+; CI-NEXT:    s_lshr_b32 s0, s4, 16
+; CI-NEXT:    s_waitcnt lgkmcnt(0)
+; CI-NEXT:    s_lshr_b32 s1, s2, 16
+; CI-NEXT:    s_lshl_b32 s2, s1, 16
+; CI-NEXT:    s_or_b32 s2, s0, s2
+; CI-NEXT:    v_mov_b32_e32 v2, s2
+; CI-NEXT:    flat_store_dword v[0:1], v2
+; CI-NEXT:    ;;#ASMSTART
+; CI-NEXT:    ; use s0
+; CI-NEXT:    ;;#ASMEND
+; CI-NEXT:    ;;#ASMSTART
+; CI-NEXT:    ; use s1
+; CI-NEXT:    ;;#ASMEND
+; CI-NEXT:    s_endpgm
   %vec = load <2 x i16>, <2 x i16> addrspace(4)* %vec.ptr
   %elt.hi = lshr i32 %elt.arg, 16
   %elt = trunc i32 %elt.hi to i16
@@ -145,78 +358,213 @@ define amdgpu_kernel void @s_insertelement_v2i16_0_reghi_both_multi_use_1(<2 x i
   ret void
 }
 
-; GCN-LABEL: {{^}}s_insertelement_v2i16_1:
-; GCN: s_load_dword [[VEC:s[0-9]+]]
-
-; GCN-NOT: s_lshr
-
-; CIVI: s_and_b32 [[ELT0:s[0-9]+]], [[VEC]], 0xffff{{$}}
-; CIVI: s_or_b32 [[INS:s[0-9]+]], [[ELT0]], 0x3e70000
-
-; GFX9: s_pack_ll_b32_b16 s{{[0-9]+}}, [[VEC]], 0x3e7
 define amdgpu_kernel void @s_insertelement_v2i16_1(<2 x i16> addrspace(1)* %out, <2 x i16> addrspace(4)* %vec.ptr) #0 {
+; GFX9-LABEL: s_insertelement_v2i16_1:
+; GFX9:       ; %bb.0:
+; GFX9-NEXT:    s_load_dwordx4 s[0:3], s[4:5], 0x0
+; GFX9-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX9-NEXT:    s_load_dword s2, s[2:3], 0x0
+; GFX9-NEXT:    v_mov_b32_e32 v0, s0
+; GFX9-NEXT:    v_mov_b32_e32 v1, s1
+; GFX9-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX9-NEXT:    s_pack_ll_b32_b16 s0, s2, 0x3e7
+; GFX9-NEXT:    v_mov_b32_e32 v2, s0
+; GFX9-NEXT:    global_store_dword v[0:1], v2, off
+; GFX9-NEXT:    s_endpgm
+;
+; CIVI-LABEL: s_insertelement_v2i16_1:
+; CIVI:       ; %bb.0:
+; CIVI-NEXT:    s_load_dwordx4 s[0:3], s[4:5], 0x0
+; CIVI-NEXT:    s_waitcnt lgkmcnt(0)
+; CIVI-NEXT:    s_load_dword s2, s[2:3], 0x0
+; CIVI-NEXT:    v_mov_b32_e32 v0, s0
+; CIVI-NEXT:    v_mov_b32_e32 v1, s1
+; CIVI-NEXT:    s_waitcnt lgkmcnt(0)
+; CIVI-NEXT:    s_and_b32 s0, s2, 0xffff
+; CIVI-NEXT:    s_or_b32 s0, s0, 0x3e70000
+; CIVI-NEXT:    v_mov_b32_e32 v2, s0
+; CIVI-NEXT:    flat_store_dword v[0:1], v2
+; CIVI-NEXT:    s_endpgm
   %vec = load <2 x i16>, <2 x i16> addrspace(4)* %vec.ptr
   %vecins = insertelement <2 x i16> %vec, i16 999, i32 1
   store <2 x i16> %vecins, <2 x i16> addrspace(1)* %out
   ret void
 }
 
-; GCN-LABEL: {{^}}s_insertelement_v2i16_1_reg:
-; GCN-DAG: s_load_dword [[ELT1_LOAD:s[0-9]+]], s[4:5],
-; GCN-DAG: s_load_dword [[VEC:s[0-9]+]], s{{\[[0-9]+:[0-9]+\]}}, 0x0
-
-; CIVI: s_lshl_b32 [[ELT1:s[0-9]+]], [[ELT1_LOAD]], 16
-; CIVI: s_and_b32 [[ELT0:s[0-9]+]], [[VEC]], 0xffff{{$}}
-; CIVI: s_or_b32 s{{[0-9]+}}, [[ELT0]], [[ELT1]]
-
-; GCN-NOT: shlr
-; GFX9: s_pack_ll_b32_b16 s{{[0-9]+}}, [[VEC]], [[ELT1_LOAD]]
 define amdgpu_kernel void @s_insertelement_v2i16_1_reg(<2 x i16> addrspace(1)* %out, <2 x i16> addrspace(4)* %vec.ptr, [8 x i32], i16 %elt) #0 {
+; GFX9-LABEL: s_insertelement_v2i16_1_reg:
+; GFX9:       ; %bb.0:
+; GFX9-NEXT:    s_load_dwordx4 s[0:3], s[4:5], 0x0
+; GFX9-NEXT:    s_load_dword s4, s[4:5], 0x30
+; GFX9-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX9-NEXT:    s_load_dword s2, s[2:3], 0x0
+; GFX9-NEXT:    v_mov_b32_e32 v0, s0
+; GFX9-NEXT:    v_mov_b32_e32 v1, s1
+; GFX9-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX9-NEXT:    s_pack_ll_b32_b16 s0, s2, s4
+; GFX9-NEXT:    v_mov_b32_e32 v2, s0
+; GFX9-NEXT:    global_store_dword v[0:1], v2, off
+; GFX9-NEXT:    s_endpgm
+;
+; VI-LABEL: s_insertelement_v2i16_1_reg:
+; VI:       ; %bb.0:
+; VI-NEXT:    s_load_dwordx4 s[0:3], s[4:5], 0x0
+; VI-NEXT:    s_load_dword s4, s[4:5], 0x30
+; VI-NEXT:    s_waitcnt lgkmcnt(0)
+; VI-NEXT:    s_load_dword s2, s[2:3], 0x0
+; VI-NEXT:    v_mov_b32_e32 v0, s0
+; VI-NEXT:    v_mov_b32_e32 v1, s1
+; VI-NEXT:    s_lshl_b32 s0, s4, 16
+; VI-NEXT:    s_waitcnt lgkmcnt(0)
+; VI-NEXT:    s_and_b32 s1, s2, 0xffff
+; VI-NEXT:    s_or_b32 s0, s1, s0
+; VI-NEXT:    v_mov_b32_e32 v2, s0
+; VI-NEXT:    flat_store_dword v[0:1], v2
+; VI-NEXT:    s_endpgm
+;
+; CI-LABEL: s_insertelement_v2i16_1_reg:
+; CI:       ; %bb.0:
+; CI-NEXT:    s_load_dwordx4 s[0:3], s[4:5], 0x0
+; CI-NEXT:    s_load_dword s4, s[4:5], 0xc
+; CI-NEXT:    s_waitcnt lgkmcnt(0)
+; CI-NEXT:    s_load_dword s2, s[2:3], 0x0
+; CI-NEXT:    v_mov_b32_e32 v0, s0
+; CI-NEXT:    v_mov_b32_e32 v1, s1
+; CI-NEXT:    s_lshl_b32 s1, s4, 16
+; CI-NEXT:    s_waitcnt lgkmcnt(0)
+; CI-NEXT:    s_and_b32 s0, s2, 0xffff
+; CI-NEXT:    s_or_b32 s0, s0, s1
+; CI-NEXT:    v_mov_b32_e32 v2, s0
+; CI-NEXT:    flat_store_dword v[0:1], v2
+; CI-NEXT:    s_endpgm
   %vec = load <2 x i16>, <2 x i16> addrspace(4)* %vec.ptr
   %vecins = insertelement <2 x i16> %vec, i16 %elt, i32 1
   store <2 x i16> %vecins, <2 x i16> addrspace(1)* %out
   ret void
 }
 
-; GCN-LABEL: {{^}}s_insertelement_v2f16_0:
-; GCN: s_load_dword [[VEC:s[0-9]+]]
-; CIVI: s_and_b32 [[ELT1:s[0-9]+]], [[VEC:s[0-9]+]], 0xffff0000
-; CIVI: s_or_b32 s{{[0-9]+}}, [[ELT1]], 0x4500
-
-; GFX9: s_lshr_b32 [[ELT1:s[0-9]+]], [[VEC]], 16
-; GFX9: s_pack_ll_b32_b16 s{{[0-9]+}}, 0x4500, [[ELT1]]
 define amdgpu_kernel void @s_insertelement_v2f16_0(<2 x half> addrspace(1)* %out, <2 x half> addrspace(4)* %vec.ptr) #0 {
+; GFX9-LABEL: s_insertelement_v2f16_0:
+; GFX9:       ; %bb.0:
+; GFX9-NEXT:    s_load_dwordx4 s[0:3], s[4:5], 0x0
+; GFX9-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX9-NEXT:    s_load_dword s2, s[2:3], 0x0
+; GFX9-NEXT:    v_mov_b32_e32 v0, s0
+; GFX9-NEXT:    v_mov_b32_e32 v1, s1
+; GFX9-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX9-NEXT:    s_lshr_b32 s0, s2, 16
+; GFX9-NEXT:    s_pack_ll_b32_b16 s0, 0x4500, s0
+; GFX9-NEXT:    v_mov_b32_e32 v2, s0
+; GFX9-NEXT:    global_store_dword v[0:1], v2, off
+; GFX9-NEXT:    s_endpgm
+;
+; CIVI-LABEL: s_insertelement_v2f16_0:
+; CIVI:       ; %bb.0:
+; CIVI-NEXT:    s_load_dwordx4 s[0:3], s[4:5], 0x0
+; CIVI-NEXT:    s_waitcnt lgkmcnt(0)
+; CIVI-NEXT:    s_load_dword s2, s[2:3], 0x0
+; CIVI-NEXT:    v_mov_b32_e32 v0, s0
+; CIVI-NEXT:    v_mov_b32_e32 v1, s1
+; CIVI-NEXT:    s_waitcnt lgkmcnt(0)
+; CIVI-NEXT:    s_and_b32 s0, s2, 0xffff0000
+; CIVI-NEXT:    s_or_b32 s0, s0, 0x4500
+; CIVI-NEXT:    v_mov_b32_e32 v2, s0
+; CIVI-NEXT:    flat_store_dword v[0:1], v2
+; CIVI-NEXT:    s_endpgm
   %vec = load <2 x half>, <2 x half> addrspace(4)* %vec.ptr
   %vecins = insertelement <2 x half> %vec, half 5.000000e+00, i32 0
   store <2 x half> %vecins, <2 x half> addrspace(1)* %out
   ret void
 }
 
-; GCN-LABEL: {{^}}s_insertelement_v2f16_1:
-; GCN: s_load_dword [[VEC:s[0-9]+]]
-; GCN-NOT: s_lshr
-
-; CIVI: s_and_b32 [[ELT0:s[0-9]+]], [[VEC]], 0xffff{{$}}
-; CIVI: s_or_b32 [[INS:s[0-9]+]], [[ELT0]], 0x45000000
-
-; GFX9: s_pack_ll_b32_b16 s{{[0-9]+}}, [[VEC]], 0x4500
 define amdgpu_kernel void @s_insertelement_v2f16_1(<2 x half> addrspace(1)* %out, <2 x half> addrspace(4)* %vec.ptr) #0 {
+; GFX9-LABEL: s_insertelement_v2f16_1:
+; GFX9:       ; %bb.0:
+; GFX9-NEXT:    s_load_dwordx4 s[0:3], s[4:5], 0x0
+; GFX9-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX9-NEXT:    s_load_dword s2, s[2:3], 0x0
+; GFX9-NEXT:    v_mov_b32_e32 v0, s0
+; GFX9-NEXT:    v_mov_b32_e32 v1, s1
+; GFX9-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX9-NEXT:    s_pack_ll_b32_b16 s0, s2, 0x4500
+; GFX9-NEXT:    v_mov_b32_e32 v2, s0
+; GFX9-NEXT:    global_store_dword v[0:1], v2, off
+; GFX9-NEXT:    s_endpgm
+;
+; CIVI-LABEL: s_insertelement_v2f16_1:
+; CIVI:       ; %bb.0:
+; CIVI-NEXT:    s_load_dwordx4 s[0:3], s[4:5], 0x0
+; CIVI-NEXT:    s_waitcnt lgkmcnt(0)
+; CIVI-NEXT:    s_load_dword s2, s[2:3], 0x0
+; CIVI-NEXT:    v_mov_b32_e32 v0, s0
+; CIVI-NEXT:    v_mov_b32_e32 v1, s1
+; CIVI-NEXT:    s_waitcnt lgkmcnt(0)
+; CIVI-NEXT:    s_and_b32 s0, s2, 0xffff
+; CIVI-NEXT:    s_or_b32 s0, s0, 0x45000000
+; CIVI-NEXT:    v_mov_b32_e32 v2, s0
+; CIVI-NEXT:    flat_store_dword v[0:1], v2
+; CIVI-NEXT:    s_endpgm
   %vec = load <2 x half>, <2 x half> addrspace(4)* %vec.ptr
   %vecins = insertelement <2 x half> %vec, half 5.000000e+00, i32 1
   store <2 x half> %vecins, <2 x half> addrspace(1)* %out
   ret void
 }
 
-; GCN-LABEL: {{^}}v_insertelement_v2i16_0:
-; GCN-DAG: {{flat|global}}_load_dword [[VEC:v[0-9]+]]
-; CIVI: v_and_b32_e32 [[ELT1:v[0-9]+]], 0xffff0000, [[VEC]]
-; CIVI: v_or_b32_e32 [[RES:v[0-9]+]], 0x3e7, [[ELT1]]
-
-; GFX9-DAG: s_movk_i32 [[ELT0:s[0-9]+]], 0x3e7{{$}}
-; GFX9-DAG: v_mov_b32_e32 [[MASK:v[0-9]+]], 0xffff{{$}}
-; GFX9: v_bfi_b32 [[RES:v[0-9]+]], [[MASK]], [[ELT0]], [[VEC]]
-; GCN: {{flat|global}}_store_dword v{{\[[0-9]+:[0-9]+\]}}, [[RES]]
 define amdgpu_kernel void @v_insertelement_v2i16_0(<2 x i16> addrspace(1)* %out, <2 x i16> addrspace(1)* %in) #0 {
+; GFX9-LABEL: v_insertelement_v2i16_0:
+; GFX9:       ; %bb.0:
+; GFX9-NEXT:    s_load_dwordx4 s[0:3], s[4:5], 0x0
+; GFX9-NEXT:    v_lshlrev_b32_e32 v2, 2, v0
+; GFX9-NEXT:    s_movk_i32 s4, 0x3e7
+; GFX9-NEXT:    v_mov_b32_e32 v3, 0xffff
+; GFX9-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX9-NEXT:    v_mov_b32_e32 v1, s3
+; GFX9-NEXT:    v_add_co_u32_e32 v0, vcc, s2, v2
+; GFX9-NEXT:    v_addc_co_u32_e32 v1, vcc, 0, v1, vcc
+; GFX9-NEXT:    global_load_dword v4, v[0:1], off
+; GFX9-NEXT:    v_add_co_u32_e32 v0, vcc, s0, v2
+; GFX9-NEXT:    v_mov_b32_e32 v1, s1
+; GFX9-NEXT:    v_addc_co_u32_e32 v1, vcc, 0, v1, vcc
+; GFX9-NEXT:    s_waitcnt vmcnt(0)
+; GFX9-NEXT:    v_bfi_b32 v2, v3, s4, v4
+; GFX9-NEXT:    global_store_dword v[0:1], v2, off
+; GFX9-NEXT:    s_endpgm
+;
+; VI-LABEL: v_insertelement_v2i16_0:
+; VI:       ; %bb.0:
+; VI-NEXT:    s_load_dwordx4 s[0:3], s[4:5], 0x0
+; VI-NEXT:    v_lshlrev_b32_e32 v2, 2, v0
+; VI-NEXT:    s_waitcnt lgkmcnt(0)
+; VI-NEXT:    v_mov_b32_e32 v1, s3
+; VI-NEXT:    v_add_u32_e32 v0, vcc, s2, v2
+; VI-NEXT:    v_addc_u32_e32 v1, vcc, 0, v1, vcc
+; VI-NEXT:    flat_load_dword v3, v[0:1]
+; VI-NEXT:    v_add_u32_e32 v0, vcc, s0, v2
+; VI-NEXT:    v_mov_b32_e32 v1, s1
+; VI-NEXT:    v_addc_u32_e32 v1, vcc, 0, v1, vcc
+; VI-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; VI-NEXT:    v_and_b32_e32 v2, 0xffff0000, v3
+; VI-NEXT:    v_or_b32_e32 v2, 0x3e7, v2
+; VI-NEXT:    flat_store_dword v[0:1], v2
+; VI-NEXT:    s_endpgm
+;
+; CI-LABEL: v_insertelement_v2i16_0:
+; CI:       ; %bb.0:
+; CI-NEXT:    s_load_dwordx4 s[0:3], s[4:5], 0x0
+; CI-NEXT:    v_lshlrev_b32_e32 v2, 2, v0
+; CI-NEXT:    s_waitcnt lgkmcnt(0)
+; CI-NEXT:    v_mov_b32_e32 v1, s3
+; CI-NEXT:    v_add_i32_e32 v0, vcc, s2, v2
+; CI-NEXT:    v_addc_u32_e32 v1, vcc, 0, v1, vcc
+; CI-NEXT:    flat_load_dword v3, v[0:1]
+; CI-NEXT:    v_add_i32_e32 v0, vcc, s0, v2
+; CI-NEXT:    v_mov_b32_e32 v1, s1
+; CI-NEXT:    v_addc_u32_e32 v1, vcc, 0, v1, vcc
+; CI-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; CI-NEXT:    v_and_b32_e32 v2, 0xffff0000, v3
+; CI-NEXT:    v_or_b32_e32 v2, 0x3e7, v2
+; CI-NEXT:    flat_store_dword v[0:1], v2
+; CI-NEXT:    s_endpgm
   %tid = call i32 @llvm.amdgcn.workitem.id.x() #1
   %tid.ext = sext i32 %tid to i64
   %in.gep = getelementptr inbounds <2 x i16>, <2 x i16> addrspace(1)* %in, i64 %tid.ext
@@ -227,20 +575,66 @@ define amdgpu_kernel void @v_insertelement_v2i16_0(<2 x i16> addrspace(1)* %out,
   ret void
 }
 
-; GCN-LABEL: {{^}}v_insertelement_v2i16_0_reghi:
-; GCN-DAG: {{flat|global}}_load_dword [[VEC:v[0-9]+]]
-; GCN-DAG: s_load_dword [[ELT0:s[0-9]+]]
-
-; CIVI-DAG: s_lshr_b32 [[ELT0_SHIFT:s[0-9]+]], [[ELT0]], 16
-; CIVI-DAG: v_and_b32_e32 [[ELT1:v[0-9]+]], 0xffff0000, [[VEC]]
-; CIVI: v_or_b32_e32 [[RES:v[0-9]+]], [[ELT0_SHIFT]], [[ELT1]]
-
-; GFX9-DAG: v_mov_b32_e32 [[MASK:v[0-9]+]], 0xffff0000{{$}}
-; GFX9-DAG: v_lshrrev_b32_e64 [[ELT0_SHIFT:v[0-9]+]], 16, [[ELT0]]
-; GFX9: v_and_or_b32 [[RES:v[0-9]+]], [[VEC]], [[MASK]], [[ELT0_SHIFT]]
-
-; GCN: {{flat|global}}_store_dword v{{\[[0-9]+:[0-9]+\]}}, [[RES]]
 define amdgpu_kernel void @v_insertelement_v2i16_0_reghi(<2 x i16> addrspace(1)* %out, <2 x i16> addrspace(1)* %in, i32 %elt.arg) #0 {
+; GFX9-LABEL: v_insertelement_v2i16_0_reghi:
+; GFX9:       ; %bb.0:
+; GFX9-NEXT:    s_load_dwordx4 s[0:3], s[4:5], 0x0
+; GFX9-NEXT:    v_lshlrev_b32_e32 v2, 2, v0
+; GFX9-NEXT:    s_load_dword s4, s[4:5], 0x10
+; GFX9-NEXT:    v_mov_b32_e32 v3, 0xffff0000
+; GFX9-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX9-NEXT:    v_mov_b32_e32 v1, s3
+; GFX9-NEXT:    v_add_co_u32_e32 v0, vcc, s2, v2
+; GFX9-NEXT:    v_addc_co_u32_e32 v1, vcc, 0, v1, vcc
+; GFX9-NEXT:    global_load_dword v4, v[0:1], off
+; GFX9-NEXT:    v_add_co_u32_e32 v0, vcc, s0, v2
+; GFX9-NEXT:    v_mov_b32_e32 v1, s1
+; GFX9-NEXT:    v_lshrrev_b32_e64 v2, 16, s4
+; GFX9-NEXT:    v_addc_co_u32_e32 v1, vcc, 0, v1, vcc
+; GFX9-NEXT:    s_waitcnt vmcnt(0)
+; GFX9-NEXT:    v_and_or_b32 v2, v4, v3, v2
+; GFX9-NEXT:    global_store_dword v[0:1], v2, off
+; GFX9-NEXT:    s_endpgm
+;
+; VI-LABEL: v_insertelement_v2i16_0_reghi:
+; VI:       ; %bb.0:
+; VI-NEXT:    s_load_dwordx4 s[0:3], s[4:5], 0x0
+; VI-NEXT:    v_lshlrev_b32_e32 v2, 2, v0
+; VI-NEXT:    s_load_dword s4, s[4:5], 0x10
+; VI-NEXT:    s_waitcnt lgkmcnt(0)
+; VI-NEXT:    v_mov_b32_e32 v1, s3
+; VI-NEXT:    v_add_u32_e32 v0, vcc, s2, v2
+; VI-NEXT:    v_addc_u32_e32 v1, vcc, 0, v1, vcc
+; VI-NEXT:    flat_load_dword v3, v[0:1]
+; VI-NEXT:    v_mov_b32_e32 v1, s1
+; VI-NEXT:    v_add_u32_e32 v0, vcc, s0, v2
+; VI-NEXT:    s_lshr_b32 s1, s4, 16
+; VI-NEXT:    v_addc_u32_e32 v1, vcc, 0, v1, vcc
+; VI-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; VI-NEXT:    v_and_b32_e32 v2, 0xffff0000, v3
+; VI-NEXT:    v_or_b32_e32 v2, s1, v2
+; VI-NEXT:    flat_store_dword v[0:1], v2
+; VI-NEXT:    s_endpgm
+;
+; CI-LABEL: v_insertelement_v2i16_0_reghi:
+; CI:       ; %bb.0:
+; CI-NEXT:    s_load_dwordx4 s[0:3], s[4:5], 0x0
+; CI-NEXT:    v_lshlrev_b32_e32 v2, 2, v0
+; CI-NEXT:    s_load_dword s4, s[4:5], 0x4
+; CI-NEXT:    s_waitcnt lgkmcnt(0)
+; CI-NEXT:    v_mov_b32_e32 v1, s3
+; CI-NEXT:    v_add_i32_e32 v0, vcc, s2, v2
+; CI-NEXT:    v_addc_u32_e32 v1, vcc, 0, v1, vcc
+; CI-NEXT:    flat_load_dword v3, v[0:1]
+; CI-NEXT:    v_mov_b32_e32 v1, s1
+; CI-NEXT:    v_add_i32_e32 v0, vcc, s0, v2
+; CI-NEXT:    s_lshr_b32 s1, s4, 16
+; CI-NEXT:    v_addc_u32_e32 v1, vcc, 0, v1, vcc
+; CI-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; CI-NEXT:    v_and_b32_e32 v2, 0xffff0000, v3
+; CI-NEXT:    v_or_b32_e32 v2, s1, v2
+; CI-NEXT:    flat_store_dword v[0:1], v2
+; CI-NEXT:    s_endpgm
   %tid = call i32 @llvm.amdgcn.workitem.id.x() #1
   %tid.ext = sext i32 %tid to i64
   %in.gep = getelementptr inbounds <2 x i16>, <2 x i16> addrspace(1)* %in, i64 %tid.ext
@@ -253,17 +647,60 @@ define amdgpu_kernel void @v_insertelement_v2i16_0_reghi(<2 x i16> addrspace(1)*
   ret void
 }
 
-; GCN-LABEL: {{^}}v_insertelement_v2i16_0_inlineimm:
-; GCN-DAG: {{flat|global}}_load_dword [[VEC:v[0-9]+]]
-
-; CIVI: v_and_b32_e32 [[ELT1:v[0-9]+]], 0xffff0000, [[VEC]]
-; CIVI: v_or_b32_e32 [[RES:v[0-9]+]], 53, [[ELT1]]
-
-; GFX9-DAG: v_mov_b32_e32 [[MASK:v[0-9]+]], 0xffff{{$}}
-; GFX9: v_bfi_b32 [[RES:v[0-9]+]], [[MASK]], 53, [[VEC]]
-
-; GCN: {{flat|global}}_store_dword v{{\[[0-9]+:[0-9]+\]}}, [[RES]]
 define amdgpu_kernel void @v_insertelement_v2i16_0_inlineimm(<2 x i16> addrspace(1)* %out, <2 x i16> addrspace(1)* %in) #0 {
+; GFX9-LABEL: v_insertelement_v2i16_0_inlineimm:
+; GFX9:       ; %bb.0:
+; GFX9-NEXT:    s_load_dwordx4 s[0:3], s[4:5], 0x0
+; GFX9-NEXT:    v_lshlrev_b32_e32 v2, 2, v0
+; GFX9-NEXT:    v_mov_b32_e32 v3, 0xffff
+; GFX9-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX9-NEXT:    v_mov_b32_e32 v1, s3
+; GFX9-NEXT:    v_add_co_u32_e32 v0, vcc, s2, v2
+; GFX9-NEXT:    v_addc_co_u32_e32 v1, vcc, 0, v1, vcc
+; GFX9-NEXT:    global_load_dword v4, v[0:1], off
+; GFX9-NEXT:    v_add_co_u32_e32 v0, vcc, s0, v2
+; GFX9-NEXT:    v_mov_b32_e32 v1, s1
+; GFX9-NEXT:    v_addc_co_u32_e32 v1, vcc, 0, v1, vcc
+; GFX9-NEXT:    s_waitcnt vmcnt(0)
+; GFX9-NEXT:    v_bfi_b32 v2, v3, 53, v4
+; GFX9-NEXT:    global_store_dword v[0:1], v2, off
+; GFX9-NEXT:    s_endpgm
+;
+; VI-LABEL: v_insertelement_v2i16_0_inlineimm:
+; VI:       ; %bb.0:
+; VI-NEXT:    s_load_dwordx4 s[0:3], s[4:5], 0x0
+; VI-NEXT:    v_lshlrev_b32_e32 v2, 2, v0
+; VI-NEXT:    s_waitcnt lgkmcnt(0)
+; VI-NEXT:    v_mov_b32_e32 v1, s3
+; VI-NEXT:    v_add_u32_e32 v0, vcc, s2, v2
+; VI-NEXT:    v_addc_u32_e32 v1, vcc, 0, v1, vcc
+; VI-NEXT:    flat_load_dword v3, v[0:1]
+; VI-NEXT:    v_add_u32_e32 v0, vcc, s0, v2
+; VI-NEXT:    v_mov_b32_e32 v1, s1
+; VI-NEXT:    v_addc_u32_e32 v1, vcc, 0, v1, vcc
+; VI-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; VI-NEXT:    v_and_b32_e32 v2, 0xffff0000, v3
+; VI-NEXT:    v_or_b32_e32 v2, 53, v2
+; VI-NEXT:    flat_store_dword v[0:1], v2
+; VI-NEXT:    s_endpgm
+;
+; CI-LABEL: v_insertelement_v2i16_0_inlineimm:
+; CI:       ; %bb.0:
+; CI-NEXT:    s_load_dwordx4 s[0:3], s[4:5], 0x0
+; CI-NEXT:    v_lshlrev_b32_e32 v2, 2, v0
+; CI-NEXT:    s_waitcnt lgkmcnt(0)
+; CI-NEXT:    v_mov_b32_e32 v1, s3
+; CI-NEXT:    v_add_i32_e32 v0, vcc, s2, v2
+; CI-NEXT:    v_addc_u32_e32 v1, vcc, 0, v1, vcc
+; CI-NEXT:    flat_load_dword v3, v[0:1]
+; CI-NEXT:    v_add_i32_e32 v0, vcc, s0, v2
+; CI-NEXT:    v_mov_b32_e32 v1, s1
+; CI-NEXT:    v_addc_u32_e32 v1, vcc, 0, v1, vcc
+; CI-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; CI-NEXT:    v_and_b32_e32 v2, 0xffff0000, v3
+; CI-NEXT:    v_or_b32_e32 v2, 53, v2
+; CI-NEXT:    flat_store_dword v[0:1], v2
+; CI-NEXT:    s_endpgm
   %tid = call i32 @llvm.amdgcn.workitem.id.x() #1
   %tid.ext = sext i32 %tid to i64
   %in.gep = getelementptr inbounds <2 x i16>, <2 x i16> addrspace(1)* %in, i64 %tid.ext
@@ -275,21 +712,61 @@ define amdgpu_kernel void @v_insertelement_v2i16_0_inlineimm(<2 x i16> addrspace
 }
 
 ; FIXME: fold lshl_or c0, c1, v0 -> or (c0 << c1), v0
-
-; GCN-LABEL: {{^}}v_insertelement_v2i16_1:
-; VI: v_mov_b32_e32 [[K:v[0-9]+]], 0x3e70000
-; GCN-DAG: {{flat|global}}_load_dword [[VEC:v[0-9]+]]
-
-; GFX9-DAG: s_movk_i32 [[K:s[0-9]+]], 0x3e7
-; GFX9-DAG: v_and_b32_e32 [[ELT0:v[0-9]+]], 0xffff, [[VEC]]
-; GFX9: v_lshl_or_b32 [[RES:v[0-9]+]], [[K]], 16, [[ELT0]]
-
-; CI: v_and_b32_e32 [[AND:v[0-9]+]], 0xffff, [[VEC]]
-; CI: v_or_b32_e32 [[RES:v[0-9]+]], 0x3e70000, [[AND]]
-; VI: v_or_b32_sdwa [[RES:v[0-9]+]], [[K]], [[VEC]] dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:WORD_0
-
-; GCN: {{flat|global}}_store_dword v{{\[[0-9]+:[0-9]+\]}}, [[RES]]
 define amdgpu_kernel void @v_insertelement_v2i16_1(<2 x i16> addrspace(1)* %out, <2 x i16> addrspace(1)* %in) #0 {
+; GFX9-LABEL: v_insertelement_v2i16_1:
+; GFX9:       ; %bb.0:
+; GFX9-NEXT:    s_load_dwordx4 s[0:3], s[4:5], 0x0
+; GFX9-NEXT:    v_lshlrev_b32_e32 v2, 2, v0
+; GFX9-NEXT:    s_movk_i32 s4, 0x3e7
+; GFX9-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX9-NEXT:    v_mov_b32_e32 v1, s3
+; GFX9-NEXT:    v_add_co_u32_e32 v0, vcc, s2, v2
+; GFX9-NEXT:    v_addc_co_u32_e32 v1, vcc, 0, v1, vcc
+; GFX9-NEXT:    global_load_dword v3, v[0:1], off
+; GFX9-NEXT:    v_add_co_u32_e32 v0, vcc, s0, v2
+; GFX9-NEXT:    v_mov_b32_e32 v1, s1
+; GFX9-NEXT:    v_addc_co_u32_e32 v1, vcc, 0, v1, vcc
+; GFX9-NEXT:    s_waitcnt vmcnt(0)
+; GFX9-NEXT:    v_and_b32_e32 v2, 0xffff, v3
+; GFX9-NEXT:    v_lshl_or_b32 v2, s4, 16, v2
+; GFX9-NEXT:    global_store_dword v[0:1], v2, off
+; GFX9-NEXT:    s_endpgm
+;
+; VI-LABEL: v_insertelement_v2i16_1:
+; VI:       ; %bb.0:
+; VI-NEXT:    s_load_dwordx4 s[0:3], s[4:5], 0x0
+; VI-NEXT:    v_lshlrev_b32_e32 v2, 2, v0
+; VI-NEXT:    v_mov_b32_e32 v3, 0x3e70000
+; VI-NEXT:    s_waitcnt lgkmcnt(0)
+; VI-NEXT:    v_mov_b32_e32 v1, s3
+; VI-NEXT:    v_add_u32_e32 v0, vcc, s2, v2
+; VI-NEXT:    v_addc_u32_e32 v1, vcc, 0, v1, vcc
+; VI-NEXT:    flat_load_dword v4, v[0:1]
+; VI-NEXT:    v_add_u32_e32 v0, vcc, s0, v2
+; VI-NEXT:    v_mov_b32_e32 v1, s1
+; VI-NEXT:    v_addc_u32_e32 v1, vcc, 0, v1, vcc
+; VI-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; VI-NEXT:    v_or_b32_sdwa v2, v3, v4 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:WORD_0
+; VI-NEXT:    flat_store_dword v[0:1], v2
+; VI-NEXT:    s_endpgm
+;
+; CI-LABEL: v_insertelement_v2i16_1:
+; CI:       ; %bb.0:
+; CI-NEXT:    s_load_dwordx4 s[0:3], s[4:5], 0x0
+; CI-NEXT:    v_lshlrev_b32_e32 v2, 2, v0
+; CI-NEXT:    s_waitcnt lgkmcnt(0)
+; CI-NEXT:    v_mov_b32_e32 v1, s3
+; CI-NEXT:    v_add_i32_e32 v0, vcc, s2, v2
+; CI-NEXT:    v_addc_u32_e32 v1, vcc, 0, v1, vcc
+; CI-NEXT:    flat_load_dword v3, v[0:1]
+; CI-NEXT:    v_add_i32_e32 v0, vcc, s0, v2
+; CI-NEXT:    v_mov_b32_e32 v1, s1
+; CI-NEXT:    v_addc_u32_e32 v1, vcc, 0, v1, vcc
+; CI-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; CI-NEXT:    v_and_b32_e32 v2, 0xffff, v3
+; CI-NEXT:    v_or_b32_e32 v2, 0x3e70000, v2
+; CI-NEXT:    flat_store_dword v[0:1], v2
+; CI-NEXT:    s_endpgm
   %tid = call i32 @llvm.amdgcn.workitem.id.x() #1
   %tid.ext = sext i32 %tid to i64
   %in.gep = getelementptr inbounds <2 x i16>, <2 x i16> addrspace(1)* %in, i64 %tid.ext
@@ -300,16 +777,60 @@ define amdgpu_kernel void @v_insertelement_v2i16_1(<2 x i16> addrspace(1)* %out,
   ret void
 }
 
-; GCN-LABEL: {{^}}v_insertelement_v2i16_1_inlineimm:
-; VI: v_mov_b32_e32 [[K:v[0-9]+]], 0xfff10000
-; GCN: {{flat|global}}_load_dword [[VEC:v[0-9]+]]
-; CI:   v_and_b32_e32 [[ELT0:v[0-9]+]], 0xffff, [[VEC]]
-; GFX9: v_and_b32_e32 [[ELT0:v[0-9]+]], 0xffff, [[VEC]]
-; CI: v_or_b32_e32 [[RES:v[0-9]+]], 0xfff10000, [[ELT0]]
-; VI: v_or_b32_sdwa [[RES:v[0-9]+]], [[K]], [[VEC]] dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:WORD_0
-; GFX9: v_lshl_or_b32 [[RES:v[0-9]+]], -15, 16, [[ELT0]]
-; GCN: {{flat|global}}_store_dword v{{\[[0-9]+:[0-9]+\]}}, [[RES]]
 define amdgpu_kernel void @v_insertelement_v2i16_1_inlineimm(<2 x i16> addrspace(1)* %out, <2 x i16> addrspace(1)* %in) #0 {
+; GFX9-LABEL: v_insertelement_v2i16_1_inlineimm:
+; GFX9:       ; %bb.0:
+; GFX9-NEXT:    s_load_dwordx4 s[0:3], s[4:5], 0x0
+; GFX9-NEXT:    v_lshlrev_b32_e32 v2, 2, v0
+; GFX9-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX9-NEXT:    v_mov_b32_e32 v1, s3
+; GFX9-NEXT:    v_add_co_u32_e32 v0, vcc, s2, v2
+; GFX9-NEXT:    v_addc_co_u32_e32 v1, vcc, 0, v1, vcc
+; GFX9-NEXT:    global_load_dword v3, v[0:1], off
+; GFX9-NEXT:    v_add_co_u32_e32 v0, vcc, s0, v2
+; GFX9-NEXT:    v_mov_b32_e32 v1, s1
+; GFX9-NEXT:    v_addc_co_u32_e32 v1, vcc, 0, v1, vcc
+; GFX9-NEXT:    s_waitcnt vmcnt(0)
+; GFX9-NEXT:    v_and_b32_e32 v2, 0xffff, v3
+; GFX9-NEXT:    v_lshl_or_b32 v2, -15, 16, v2
+; GFX9-NEXT:    global_store_dword v[0:1], v2, off
+; GFX9-NEXT:    s_endpgm
+;
+; VI-LABEL: v_insertelement_v2i16_1_inlineimm:
+; VI:       ; %bb.0:
+; VI-NEXT:    s_load_dwordx4 s[0:3], s[4:5], 0x0
+; VI-NEXT:    v_lshlrev_b32_e32 v2, 2, v0
+; VI-NEXT:    v_mov_b32_e32 v3, 0xfff10000
+; VI-NEXT:    s_waitcnt lgkmcnt(0)
+; VI-NEXT:    v_mov_b32_e32 v1, s3
+; VI-NEXT:    v_add_u32_e32 v0, vcc, s2, v2
+; VI-NEXT:    v_addc_u32_e32 v1, vcc, 0, v1, vcc
+; VI-NEXT:    flat_load_dword v4, v[0:1]
+; VI-NEXT:    v_add_u32_e32 v0, vcc, s0, v2
+; VI-NEXT:    v_mov_b32_e32 v1, s1
+; VI-NEXT:    v_addc_u32_e32 v1, vcc, 0, v1, vcc
+; VI-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; VI-NEXT:    v_or_b32_sdwa v2, v3, v4 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:WORD_0
+; VI-NEXT:    flat_store_dword v[0:1], v2
+; VI-NEXT:    s_endpgm
+;
+; CI-LABEL: v_insertelement_v2i16_1_inlineimm:
+; CI:       ; %bb.0:
+; CI-NEXT:    s_load_dwordx4 s[0:3], s[4:5], 0x0
+; CI-NEXT:    v_lshlrev_b32_e32 v2, 2, v0
+; CI-NEXT:    s_waitcnt lgkmcnt(0)
+; CI-NEXT:    v_mov_b32_e32 v1, s3
+; CI-NEXT:    v_add_i32_e32 v0, vcc, s2, v2
+; CI-NEXT:    v_addc_u32_e32 v1, vcc, 0, v1, vcc
+; CI-NEXT:    flat_load_dword v3, v[0:1]
+; CI-NEXT:    v_add_i32_e32 v0, vcc, s0, v2
+; CI-NEXT:    v_mov_b32_e32 v1, s1
+; CI-NEXT:    v_addc_u32_e32 v1, vcc, 0, v1, vcc
+; CI-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; CI-NEXT:    v_and_b32_e32 v2, 0xffff, v3
+; CI-NEXT:    v_or_b32_e32 v2, 0xfff10000, v2
+; CI-NEXT:    flat_store_dword v[0:1], v2
+; CI-NEXT:    s_endpgm
   %tid = call i32 @llvm.amdgcn.workitem.id.x() #1
   %tid.ext = sext i32 %tid to i64
   %in.gep = getelementptr inbounds <2 x i16>, <2 x i16> addrspace(1)* %in, i64 %tid.ext
@@ -320,18 +841,61 @@ define amdgpu_kernel void @v_insertelement_v2i16_1_inlineimm(<2 x i16> addrspace
   ret void
 }
 
-; GCN-LABEL: {{^}}v_insertelement_v2f16_0:
-; GCN-DAG: {{flat|global}}_load_dword [[VEC:v[0-9]+]]
-
-; CIVI: v_and_b32_e32 [[ELT1:v[0-9]+]], 0xffff0000, [[VEC]]
-; CIVI: v_or_b32_e32 [[RES:v[0-9]+]], 0x4500, [[ELT1]]
-
-; GFX9-DAG: v_mov_b32_e32 [[ELT0:v[0-9]+]], 0x4500{{$}}
-; GFX9-DAG: v_lshrrev_b32_e32 [[ELT1:v[0-9]+]], 16, [[VEC]]
-; GFX9: v_lshl_or_b32 [[RES:v[0-9]+]], [[ELT1]], 16, [[ELT0]]
-
-; GCN: {{flat|global}}_store_dword v{{\[[0-9]+:[0-9]+\]}}, [[RES]]
 define amdgpu_kernel void @v_insertelement_v2f16_0(<2 x half> addrspace(1)* %out, <2 x half> addrspace(1)* %in) #0 {
+; GFX9-LABEL: v_insertelement_v2f16_0:
+; GFX9:       ; %bb.0:
+; GFX9-NEXT:    s_load_dwordx4 s[0:3], s[4:5], 0x0
+; GFX9-NEXT:    v_lshlrev_b32_e32 v2, 2, v0
+; GFX9-NEXT:    v_mov_b32_e32 v3, 0x4500
+; GFX9-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX9-NEXT:    v_mov_b32_e32 v1, s3
+; GFX9-NEXT:    v_add_co_u32_e32 v0, vcc, s2, v2
+; GFX9-NEXT:    v_addc_co_u32_e32 v1, vcc, 0, v1, vcc
+; GFX9-NEXT:    global_load_dword v4, v[0:1], off
+; GFX9-NEXT:    v_add_co_u32_e32 v0, vcc, s0, v2
+; GFX9-NEXT:    v_mov_b32_e32 v1, s1
+; GFX9-NEXT:    v_addc_co_u32_e32 v1, vcc, 0, v1, vcc
+; GFX9-NEXT:    s_waitcnt vmcnt(0)
+; GFX9-NEXT:    v_lshrrev_b32_e32 v2, 16, v4
+; GFX9-NEXT:    v_lshl_or_b32 v2, v2, 16, v3
+; GFX9-NEXT:    global_store_dword v[0:1], v2, off
+; GFX9-NEXT:    s_endpgm
+;
+; VI-LABEL: v_insertelement_v2f16_0:
+; VI:       ; %bb.0:
+; VI-NEXT:    s_load_dwordx4 s[0:3], s[4:5], 0x0
+; VI-NEXT:    v_lshlrev_b32_e32 v2, 2, v0
+; VI-NEXT:    s_waitcnt lgkmcnt(0)
+; VI-NEXT:    v_mov_b32_e32 v1, s3
+; VI-NEXT:    v_add_u32_e32 v0, vcc, s2, v2
+; VI-NEXT:    v_addc_u32_e32 v1, vcc, 0, v1, vcc
+; VI-NEXT:    flat_load_dword v3, v[0:1]
+; VI-NEXT:    v_add_u32_e32 v0, vcc, s0, v2
+; VI-NEXT:    v_mov_b32_e32 v1, s1
+; VI-NEXT:    v_addc_u32_e32 v1, vcc, 0, v1, vcc
+; VI-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; VI-NEXT:    v_and_b32_e32 v2, 0xffff0000, v3
+; VI-NEXT:    v_or_b32_e32 v2, 0x4500, v2
+; VI-NEXT:    flat_store_dword v[0:1], v2
+; VI-NEXT:    s_endpgm
+;
+; CI-LABEL: v_insertelement_v2f16_0:
+; CI:       ; %bb.0:
+; CI-NEXT:    s_load_dwordx4 s[0:3], s[4:5], 0x0
+; CI-NEXT:    v_lshlrev_b32_e32 v2, 2, v0
+; CI-NEXT:    s_waitcnt lgkmcnt(0)
+; CI-NEXT:    v_mov_b32_e32 v1, s3
+; CI-NEXT:    v_add_i32_e32 v0, vcc, s2, v2
+; CI-NEXT:    v_addc_u32_e32 v1, vcc, 0, v1, vcc
+; CI-NEXT:    flat_load_dword v3, v[0:1]
+; CI-NEXT:    v_add_i32_e32 v0, vcc, s0, v2
+; CI-NEXT:    v_mov_b32_e32 v1, s1
+; CI-NEXT:    v_addc_u32_e32 v1, vcc, 0, v1, vcc
+; CI-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; CI-NEXT:    v_and_b32_e32 v2, 0xffff0000, v3
+; CI-NEXT:    v_or_b32_e32 v2, 0x4500, v2
+; CI-NEXT:    flat_store_dword v[0:1], v2
+; CI-NEXT:    s_endpgm
   %tid = call i32 @llvm.amdgcn.workitem.id.x() #1
   %tid.ext = sext i32 %tid to i64
   %in.gep = getelementptr inbounds <2 x half>, <2 x half> addrspace(1)* %in, i64 %tid.ext
@@ -342,16 +906,60 @@ define amdgpu_kernel void @v_insertelement_v2f16_0(<2 x half> addrspace(1)* %out
   ret void
 }
 
-; GCN-LABEL: {{^}}v_insertelement_v2f16_0_inlineimm:
-; GCN: {{flat|global}}_load_dword [[VEC:v[0-9]+]]
-
-; CIVI: v_and_b32_e32 [[ELT1:v[0-9]+]], 0xffff0000, [[VEC]]
-; CIVI: v_or_b32_e32 [[RES:v[0-9]+]], 53, [[ELT1]]
-
-; GFX9: v_lshrrev_b32_e32 [[ELT1:v[0-9]+]], 16, [[VEC]]
-; GFX9: v_lshl_or_b32 [[RES:v[0-9]+]], [[ELT1]], 16, 53
-; GCN: {{flat|global}}_store_dword v{{\[[0-9]+:[0-9]+\]}}, [[RES]]
 define amdgpu_kernel void @v_insertelement_v2f16_0_inlineimm(<2 x half> addrspace(1)* %out, <2 x half> addrspace(1)* %in) #0 {
+; GFX9-LABEL: v_insertelement_v2f16_0_inlineimm:
+; GFX9:       ; %bb.0:
+; GFX9-NEXT:    s_load_dwordx4 s[0:3], s[4:5], 0x0
+; GFX9-NEXT:    v_lshlrev_b32_e32 v2, 2, v0
+; GFX9-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX9-NEXT:    v_mov_b32_e32 v1, s3
+; GFX9-NEXT:    v_add_co_u32_e32 v0, vcc, s2, v2
+; GFX9-NEXT:    v_addc_co_u32_e32 v1, vcc, 0, v1, vcc
+; GFX9-NEXT:    global_load_dword v3, v[0:1], off
+; GFX9-NEXT:    v_add_co_u32_e32 v0, vcc, s0, v2
+; GFX9-NEXT:    v_mov_b32_e32 v1, s1
+; GFX9-NEXT:    v_addc_co_u32_e32 v1, vcc, 0, v1, vcc
+; GFX9-NEXT:    s_waitcnt vmcnt(0)
+; GFX9-NEXT:    v_lshrrev_b32_e32 v2, 16, v3
+; GFX9-NEXT:    v_lshl_or_b32 v2, v2, 16, 53
+; GFX9-NEXT:    global_store_dword v[0:1], v2, off
+; GFX9-NEXT:    s_endpgm
+;
+; VI-LABEL: v_insertelement_v2f16_0_inlineimm:
+; VI:       ; %bb.0:
+; VI-NEXT:    s_load_dwordx4 s[0:3], s[4:5], 0x0
+; VI-NEXT:    v_lshlrev_b32_e32 v2, 2, v0
+; VI-NEXT:    s_waitcnt lgkmcnt(0)
+; VI-NEXT:    v_mov_b32_e32 v1, s3
+; VI-NEXT:    v_add_u32_e32 v0, vcc, s2, v2
+; VI-NEXT:    v_addc_u32_e32 v1, vcc, 0, v1, vcc
+; VI-NEXT:    flat_load_dword v3, v[0:1]
+; VI-NEXT:    v_add_u32_e32 v0, vcc, s0, v2
+; VI-NEXT:    v_mov_b32_e32 v1, s1
+; VI-NEXT:    v_addc_u32_e32 v1, vcc, 0, v1, vcc
+; VI-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; VI-NEXT:    v_and_b32_e32 v2, 0xffff0000, v3
+; VI-NEXT:    v_or_b32_e32 v2, 53, v2
+; VI-NEXT:    flat_store_dword v[0:1], v2
+; VI-NEXT:    s_endpgm
+;
+; CI-LABEL: v_insertelement_v2f16_0_inlineimm:
+; CI:       ; %bb.0:
+; CI-NEXT:    s_load_dwordx4 s[0:3], s[4:5], 0x0
+; CI-NEXT:    v_lshlrev_b32_e32 v2, 2, v0
+; CI-NEXT:    s_waitcnt lgkmcnt(0)
+; CI-NEXT:    v_mov_b32_e32 v1, s3
+; CI-NEXT:    v_add_i32_e32 v0, vcc, s2, v2
+; CI-NEXT:    v_addc_u32_e32 v1, vcc, 0, v1, vcc
+; CI-NEXT:    flat_load_dword v3, v[0:1]
+; CI-NEXT:    v_add_i32_e32 v0, vcc, s0, v2
+; CI-NEXT:    v_mov_b32_e32 v1, s1
+; CI-NEXT:    v_addc_u32_e32 v1, vcc, 0, v1, vcc
+; CI-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; CI-NEXT:    v_and_b32_e32 v2, 0xffff0000, v3
+; CI-NEXT:    v_or_b32_e32 v2, 53, v2
+; CI-NEXT:    flat_store_dword v[0:1], v2
+; CI-NEXT:    s_endpgm
   %tid = call i32 @llvm.amdgcn.workitem.id.x() #1
   %tid.ext = sext i32 %tid to i64
   %in.gep = getelementptr inbounds <2 x half>, <2 x half> addrspace(1)* %in, i64 %tid.ext
@@ -362,21 +970,61 @@ define amdgpu_kernel void @v_insertelement_v2f16_0_inlineimm(<2 x half> addrspac
   ret void
 }
 
-; GCN-LABEL: {{^}}v_insertelement_v2f16_1:
-; VI: v_mov_b32_e32 [[K:v[0-9]+]], 0x45000000
-; GCN-DAG: {{flat|global}}_load_dword [[VEC:v[0-9]+]]
-
-; GFX9-DAG: s_movk_i32 [[K:s[0-9]+]], 0x4500
-; GFX9-DAG: v_and_b32_e32 [[ELT0:v[0-9]+]], 0xffff, [[VEC]]
-; GFX9: v_lshl_or_b32 [[RES:v[0-9]+]], [[K]], 16, [[ELT0]]
-
-; CI: v_and_b32_e32 [[AND:v[0-9]+]], 0xffff, [[VEC]]
-; CI: v_or_b32_e32 [[RES:v[0-9]+]], 0x45000000, [[AND]]
-
-; VI: v_or_b32_sdwa [[RES:v[0-9]+]], [[K]], [[VEC]] dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:WORD_0
-
-; GCN: {{flat|global}}_store_dword v{{\[[0-9]+:[0-9]+\]}}, [[RES]]
 define amdgpu_kernel void @v_insertelement_v2f16_1(<2 x half> addrspace(1)* %out, <2 x half> addrspace(1)* %in) #0 {
+; GFX9-LABEL: v_insertelement_v2f16_1:
+; GFX9:       ; %bb.0:
+; GFX9-NEXT:    s_load_dwordx4 s[0:3], s[4:5], 0x0
+; GFX9-NEXT:    v_lshlrev_b32_e32 v2, 2, v0
+; GFX9-NEXT:    s_movk_i32 s4, 0x4500
+; GFX9-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX9-NEXT:    v_mov_b32_e32 v1, s3
+; GFX9-NEXT:    v_add_co_u32_e32 v0, vcc, s2, v2
+; GFX9-NEXT:    v_addc_co_u32_e32 v1, vcc, 0, v1, vcc
+; GFX9-NEXT:    global_load_dword v3, v[0:1], off
+; GFX9-NEXT:    v_add_co_u32_e32 v0, vcc, s0, v2
+; GFX9-NEXT:    v_mov_b32_e32 v1, s1
+; GFX9-NEXT:    v_addc_co_u32_e32 v1, vcc, 0, v1, vcc
+; GFX9-NEXT:    s_waitcnt vmcnt(0)
+; GFX9-NEXT:    v_and_b32_e32 v2, 0xffff, v3
+; GFX9-NEXT:    v_lshl_or_b32 v2, s4, 16, v2
+; GFX9-NEXT:    global_store_dword v[0:1], v2, off
+; GFX9-NEXT:    s_endpgm
+;
+; VI-LABEL: v_insertelement_v2f16_1:
+; VI:       ; %bb.0:
+; VI-NEXT:    s_load_dwordx4 s[0:3], s[4:5], 0x0
+; VI-NEXT:    v_lshlrev_b32_e32 v2, 2, v0
+; VI-NEXT:    v_mov_b32_e32 v3, 0x45000000
+; VI-NEXT:    s_waitcnt lgkmcnt(0)
+; VI-NEXT:    v_mov_b32_e32 v1, s3
+; VI-NEXT:    v_add_u32_e32 v0, vcc, s2, v2
+; VI-NEXT:    v_addc_u32_e32 v1, vcc, 0, v1, vcc
+; VI-NEXT:    flat_load_dword v4, v[0:1]
+; VI-NEXT:    v_add_u32_e32 v0, vcc, s0, v2
+; VI-NEXT:    v_mov_b32_e32 v1, s1
+; VI-NEXT:    v_addc_u32_e32 v1, vcc, 0, v1, vcc
+; VI-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; VI-NEXT:    v_or_b32_sdwa v2, v3, v4 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:WORD_0
+; VI-NEXT:    flat_store_dword v[0:1], v2
+; VI-NEXT:    s_endpgm
+;
+; CI-LABEL: v_insertelement_v2f16_1:
+; CI:       ; %bb.0:
+; CI-NEXT:    s_load_dwordx4 s[0:3], s[4:5], 0x0
+; CI-NEXT:    v_lshlrev_b32_e32 v2, 2, v0
+; CI-NEXT:    s_waitcnt lgkmcnt(0)
+; CI-NEXT:    v_mov_b32_e32 v1, s3
+; CI-NEXT:    v_add_i32_e32 v0, vcc, s2, v2
+; CI-NEXT:    v_addc_u32_e32 v1, vcc, 0, v1, vcc
+; CI-NEXT:    flat_load_dword v3, v[0:1]
+; CI-NEXT:    v_add_i32_e32 v0, vcc, s0, v2
+; CI-NEXT:    v_mov_b32_e32 v1, s1
+; CI-NEXT:    v_addc_u32_e32 v1, vcc, 0, v1, vcc
+; CI-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; CI-NEXT:    v_and_b32_e32 v2, 0xffff, v3
+; CI-NEXT:    v_or_b32_e32 v2, 0x45000000, v2
+; CI-NEXT:    flat_store_dword v[0:1], v2
+; CI-NEXT:    s_endpgm
   %tid = call i32 @llvm.amdgcn.workitem.id.x() #1
   %tid.ext = sext i32 %tid to i64
   %in.gep = getelementptr inbounds <2 x half>, <2 x half> addrspace(1)* %in, i64 %tid.ext
@@ -387,16 +1035,60 @@ define amdgpu_kernel void @v_insertelement_v2f16_1(<2 x half> addrspace(1)* %out
   ret void
 }
 
-; GCN-LABEL: {{^}}v_insertelement_v2f16_1_inlineimm:
-; VI: v_mov_b32_e32 [[K:v[0-9]+]], 0x230000
-; GCN: {{flat|global}}_load_dword [[VEC:v[0-9]+]]
-; CI: v_and_b32_e32 [[ELT0:v[0-9]+]], 0xffff, [[VEC]]
-; GFX9: v_and_b32_e32 [[ELT0:v[0-9]+]], 0xffff, [[VEC]]
-; CI: v_or_b32_e32 [[RES:v[0-9]+]], 0x230000, [[ELT0]]
-; VI: v_or_b32_sdwa [[RES:v[0-9]+]], [[K]], [[VEC]] dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:WORD_0
-; GFX9: v_lshl_or_b32 [[RES:v[0-9]+]], 35, 16, [[ELT0]]
-; GCN: {{flat|global}}_store_dword v{{\[[0-9]+:[0-9]+\]}}, [[RES]]
 define amdgpu_kernel void @v_insertelement_v2f16_1_inlineimm(<2 x half> addrspace(1)* %out, <2 x half> addrspace(1)* %in) #0 {
+; GFX9-LABEL: v_insertelement_v2f16_1_inlineimm:
+; GFX9:       ; %bb.0:
+; GFX9-NEXT:    s_load_dwordx4 s[0:3], s[4:5], 0x0
+; GFX9-NEXT:    v_lshlrev_b32_e32 v2, 2, v0
+; GFX9-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX9-NEXT:    v_mov_b32_e32 v1, s3
+; GFX9-NEXT:    v_add_co_u32_e32 v0, vcc, s2, v2
+; GFX9-NEXT:    v_addc_co_u32_e32 v1, vcc, 0, v1, vcc
+; GFX9-NEXT:    global_load_dword v3, v[0:1], off
+; GFX9-NEXT:    v_add_co_u32_e32 v0, vcc, s0, v2
+; GFX9-NEXT:    v_mov_b32_e32 v1, s1
+; GFX9-NEXT:    v_addc_co_u32_e32 v1, vcc, 0, v1, vcc
+; GFX9-NEXT:    s_waitcnt vmcnt(0)
+; GFX9-NEXT:    v_and_b32_e32 v2, 0xffff, v3
+; GFX9-NEXT:    v_lshl_or_b32 v2, 35, 16, v2
+; GFX9-NEXT:    global_store_dword v[0:1], v2, off
+; GFX9-NEXT:    s_endpgm
+;
+; VI-LABEL: v_insertelement_v2f16_1_inlineimm:
+; VI:       ; %bb.0:
+; VI-NEXT:    s_load_dwordx4 s[0:3], s[4:5], 0x0
+; VI-NEXT:    v_lshlrev_b32_e32 v2, 2, v0
+; VI-NEXT:    v_mov_b32_e32 v3, 0x230000
+; VI-NEXT:    s_waitcnt lgkmcnt(0)
+; VI-NEXT:    v_mov_b32_e32 v1, s3
+; VI-NEXT:    v_add_u32_e32 v0, vcc, s2, v2
+; VI-NEXT:    v_addc_u32_e32 v1, vcc, 0, v1, vcc
+; VI-NEXT:    flat_load_dword v4, v[0:1]
+; VI-NEXT:    v_add_u32_e32 v0, vcc, s0, v2
+; VI-NEXT:    v_mov_b32_e32 v1, s1
+; VI-NEXT:    v_addc_u32_e32 v1, vcc, 0, v1, vcc
+; VI-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; VI-NEXT:    v_or_b32_sdwa v2, v3, v4 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:WORD_0
+; VI-NEXT:    flat_store_dword v[0:1], v2
+; VI-NEXT:    s_endpgm
+;
+; CI-LABEL: v_insertelement_v2f16_1_inlineimm:
+; CI:       ; %bb.0:
+; CI-NEXT:    s_load_dwordx4 s[0:3], s[4:5], 0x0
+; CI-NEXT:    v_lshlrev_b32_e32 v2, 2, v0
+; CI-NEXT:    s_waitcnt lgkmcnt(0)
+; CI-NEXT:    v_mov_b32_e32 v1, s3
+; CI-NEXT:    v_add_i32_e32 v0, vcc, s2, v2
+; CI-NEXT:    v_addc_u32_e32 v1, vcc, 0, v1, vcc
+; CI-NEXT:    flat_load_dword v3, v[0:1]
+; CI-NEXT:    v_add_i32_e32 v0, vcc, s0, v2
+; CI-NEXT:    v_mov_b32_e32 v1, s1
+; CI-NEXT:    v_addc_u32_e32 v1, vcc, 0, v1, vcc
+; CI-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; CI-NEXT:    v_and_b32_e32 v2, 0xffff, v3
+; CI-NEXT:    v_or_b32_e32 v2, 0x230000, v2
+; CI-NEXT:    flat_store_dword v[0:1], v2
+; CI-NEXT:    s_endpgm
   %tid = call i32 @llvm.amdgcn.workitem.id.x() #1
   %tid.ext = sext i32 %tid to i64
   %in.gep = getelementptr inbounds <2 x half>, <2 x half> addrspace(1)* %in, i64 %tid.ext
@@ -408,16 +1100,60 @@ define amdgpu_kernel void @v_insertelement_v2f16_1_inlineimm(<2 x half> addrspac
 }
 
 ; FIXME: Enable for others when argument load not split
-; GCN-LABEL: {{^}}s_insertelement_v2i16_dynamic:
-; GCN-DAG: v_mov_b32_e32 [[K:v[0-9]+]], 0x3e7
-; GCN: s_load_dword [[IDX:s[0-9]+]]
-; GCN: s_load_dword [[VEC:s[0-9]+]]
-; GCN-DAG: v_mov_b32_e32 [[VVEC:v[0-9]+]], [[VEC]]
-; GCN-DAG: s_lshl_b32 [[SCALED_IDX:s[0-9]+]], [[IDX]], 4
-; GCN-DAG: s_lshl_b32 [[MASK:s[0-9]+]], 0xffff, [[SCALED_IDX]]
-; GCN: v_bfi_b32 [[RESULT:v[0-9]+]], [[MASK]], [[K]], [[VVEC]]
-; GCN: {{flat|global}}_store_dword v{{\[[0-9]+:[0-9]+\]}}, [[RESULT]]
 define amdgpu_kernel void @s_insertelement_v2i16_dynamic(<2 x i16> addrspace(1)* %out, <2 x i16> addrspace(4)* %vec.ptr, i32 addrspace(4)* %idx.ptr) #0 {
+; GFX9-LABEL: s_insertelement_v2i16_dynamic:
+; GFX9:       ; %bb.0:
+; GFX9-NEXT:    s_load_dwordx4 s[0:3], s[4:5], 0x0
+; GFX9-NEXT:    s_load_dwordx4 s[4:7], s[4:5], 0x10
+; GFX9-NEXT:    v_mov_b32_e32 v2, 0x3e703e7
+; GFX9-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX9-NEXT:    v_mov_b32_e32 v0, s0
+; GFX9-NEXT:    v_mov_b32_e32 v1, s1
+; GFX9-NEXT:    s_load_dword s0, s[4:5], 0x0
+; GFX9-NEXT:    s_load_dword s1, s[2:3], 0x0
+; GFX9-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX9-NEXT:    s_lshl_b32 s0, s0, 4
+; GFX9-NEXT:    s_lshl_b32 s0, 0xffff, s0
+; GFX9-NEXT:    v_mov_b32_e32 v3, s1
+; GFX9-NEXT:    v_bfi_b32 v2, s0, v2, v3
+; GFX9-NEXT:    global_store_dword v[0:1], v2, off
+; GFX9-NEXT:    s_endpgm
+;
+; VI-LABEL: s_insertelement_v2i16_dynamic:
+; VI:       ; %bb.0:
+; VI-NEXT:    s_load_dwordx4 s[0:3], s[4:5], 0x0
+; VI-NEXT:    s_load_dwordx4 s[4:7], s[4:5], 0x10
+; VI-NEXT:    v_mov_b32_e32 v2, 0x3e703e7
+; VI-NEXT:    s_waitcnt lgkmcnt(0)
+; VI-NEXT:    v_mov_b32_e32 v0, s0
+; VI-NEXT:    v_mov_b32_e32 v1, s1
+; VI-NEXT:    s_load_dword s0, s[4:5], 0x0
+; VI-NEXT:    s_load_dword s1, s[2:3], 0x0
+; VI-NEXT:    s_waitcnt lgkmcnt(0)
+; VI-NEXT:    s_lshl_b32 s0, s0, 4
+; VI-NEXT:    s_lshl_b32 s0, 0xffff, s0
+; VI-NEXT:    v_mov_b32_e32 v3, s1
+; VI-NEXT:    v_bfi_b32 v2, s0, v2, v3
+; VI-NEXT:    flat_store_dword v[0:1], v2
+; VI-NEXT:    s_endpgm
+;
+; CI-LABEL: s_insertelement_v2i16_dynamic:
+; CI:       ; %bb.0:
+; CI-NEXT:    s_load_dwordx4 s[0:3], s[4:5], 0x0
+; CI-NEXT:    s_load_dwordx4 s[4:7], s[4:5], 0x4
+; CI-NEXT:    v_mov_b32_e32 v2, 0x3e703e7
+; CI-NEXT:    s_waitcnt lgkmcnt(0)
+; CI-NEXT:    v_mov_b32_e32 v0, s0
+; CI-NEXT:    v_mov_b32_e32 v1, s1
+; CI-NEXT:    s_load_dword s0, s[4:5], 0x0
+; CI-NEXT:    s_load_dword s1, s[2:3], 0x0
+; CI-NEXT:    s_waitcnt lgkmcnt(0)
+; CI-NEXT:    s_lshl_b32 s0, s0, 4
+; CI-NEXT:    s_lshl_b32 s0, 0xffff, s0
+; CI-NEXT:    v_mov_b32_e32 v3, s1
+; CI-NEXT:    v_bfi_b32 v2, s0, v2, v3
+; CI-NEXT:    flat_store_dword v[0:1], v2
+; CI-NEXT:    s_endpgm
   %idx = load volatile i32, i32 addrspace(4)* %idx.ptr
   %vec = load <2 x i16>, <2 x i16> addrspace(4)* %vec.ptr
   %vecins = insertelement <2 x i16> %vec, i16 999, i32 %idx
@@ -425,15 +1161,69 @@ define amdgpu_kernel void @s_insertelement_v2i16_dynamic(<2 x i16> addrspace(1)*
   ret void
 }
 
-; GCN-LABEL: {{^}}v_insertelement_v2i16_dynamic_sgpr:
-; GCN-DAG: {{flat|global}}_load_dword [[VEC:v[0-9]+]]
-; GCN-DAG: s_load_dword [[IDX:s[0-9]+]]
-; GCN-DAG: v_mov_b32_e32 [[K:v[0-9]+]], 0x3e7
-; GCN-DAG: s_lshl_b32 [[SCALED_IDX:s[0-9]+]], [[IDX]], 4
-; GCN-DAG: s_lshl_b32 [[MASK:s[0-9]+]], 0xffff, [[SCALED_IDX]]
-; GCN: v_bfi_b32 [[RESULT:v[0-9]+]], [[MASK]], [[K]], [[VEC]]
-; GCN: {{flat|global}}_store_dword v{{\[[0-9]+:[0-9]+\]}}, [[RESULT]]
 define amdgpu_kernel void @v_insertelement_v2i16_dynamic_sgpr(<2 x i16> addrspace(1)* %out, <2 x i16> addrspace(1)* %in, i32 %idx) #0 {
+; GFX9-LABEL: v_insertelement_v2i16_dynamic_sgpr:
+; GFX9:       ; %bb.0:
+; GFX9-NEXT:    s_load_dwordx4 s[0:3], s[4:5], 0x0
+; GFX9-NEXT:    v_lshlrev_b32_e32 v2, 2, v0
+; GFX9-NEXT:    s_load_dword s4, s[4:5], 0x10
+; GFX9-NEXT:    v_mov_b32_e32 v3, 0x3e703e7
+; GFX9-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX9-NEXT:    v_add_co_u32_e32 v0, vcc, s2, v2
+; GFX9-NEXT:    v_mov_b32_e32 v1, s3
+; GFX9-NEXT:    v_addc_co_u32_e32 v1, vcc, 0, v1, vcc
+; GFX9-NEXT:    global_load_dword v4, v[0:1], off
+; GFX9-NEXT:    s_lshl_b32 s2, s4, 4
+; GFX9-NEXT:    v_add_co_u32_e32 v0, vcc, s0, v2
+; GFX9-NEXT:    v_mov_b32_e32 v1, s1
+; GFX9-NEXT:    s_lshl_b32 s0, 0xffff, s2
+; GFX9-NEXT:    v_addc_co_u32_e32 v1, vcc, 0, v1, vcc
+; GFX9-NEXT:    s_waitcnt vmcnt(0)
+; GFX9-NEXT:    v_bfi_b32 v2, s0, v3, v4
+; GFX9-NEXT:    global_store_dword v[0:1], v2, off
+; GFX9-NEXT:    s_endpgm
+;
+; VI-LABEL: v_insertelement_v2i16_dynamic_sgpr:
+; VI:       ; %bb.0:
+; VI-NEXT:    s_load_dwordx4 s[0:3], s[4:5], 0x0
+; VI-NEXT:    v_lshlrev_b32_e32 v2, 2, v0
+; VI-NEXT:    s_load_dword s4, s[4:5], 0x10
+; VI-NEXT:    v_mov_b32_e32 v3, 0x3e703e7
+; VI-NEXT:    s_waitcnt lgkmcnt(0)
+; VI-NEXT:    v_add_u32_e32 v0, vcc, s2, v2
+; VI-NEXT:    v_mov_b32_e32 v1, s3
+; VI-NEXT:    v_addc_u32_e32 v1, vcc, 0, v1, vcc
+; VI-NEXT:    flat_load_dword v4, v[0:1]
+; VI-NEXT:    s_lshl_b32 s2, s4, 4
+; VI-NEXT:    v_add_u32_e32 v0, vcc, s0, v2
+; VI-NEXT:    v_mov_b32_e32 v1, s1
+; VI-NEXT:    s_lshl_b32 s0, 0xffff, s2
+; VI-NEXT:    v_addc_u32_e32 v1, vcc, 0, v1, vcc
+; VI-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; VI-NEXT:    v_bfi_b32 v2, s0, v3, v4
+; VI-NEXT:    flat_store_dword v[0:1], v2
+; VI-NEXT:    s_endpgm
+;
+; CI-LABEL: v_insertelement_v2i16_dynamic_sgpr:
+; CI:       ; %bb.0:
+; CI-NEXT:    s_load_dwordx4 s[0:3], s[4:5], 0x0
+; CI-NEXT:    v_lshlrev_b32_e32 v2, 2, v0
+; CI-NEXT:    s_load_dword s4, s[4:5], 0x4
+; CI-NEXT:    v_mov_b32_e32 v3, 0x3e703e7
+; CI-NEXT:    s_waitcnt lgkmcnt(0)
+; CI-NEXT:    v_add_i32_e32 v0, vcc, s2, v2
+; CI-NEXT:    v_mov_b32_e32 v1, s3
+; CI-NEXT:    v_addc_u32_e32 v1, vcc, 0, v1, vcc
+; CI-NEXT:    flat_load_dword v4, v[0:1]
+; CI-NEXT:    s_lshl_b32 s2, s4, 4
+; CI-NEXT:    v_add_i32_e32 v0, vcc, s0, v2
+; CI-NEXT:    v_mov_b32_e32 v1, s1
+; CI-NEXT:    s_lshl_b32 s0, 0xffff, s2
+; CI-NEXT:    v_addc_u32_e32 v1, vcc, 0, v1, vcc
+; CI-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; CI-NEXT:    v_bfi_b32 v2, s0, v3, v4
+; CI-NEXT:    flat_store_dword v[0:1], v2
+; CI-NEXT:    s_endpgm
   %tid = call i32 @llvm.amdgcn.workitem.id.x() #1
   %tid.ext = sext i32 %tid to i64
   %in.gep = getelementptr inbounds <2 x i16>, <2 x i16> addrspace(1)* %in, i64 %tid.ext
@@ -444,22 +1234,86 @@ define amdgpu_kernel void @v_insertelement_v2i16_dynamic_sgpr(<2 x i16> addrspac
   ret void
 }
 
-; GCN-LABEL: {{^}}v_insertelement_v2f16_dynamic_vgpr:
-; GFX89-DAG: s_mov_b32 [[MASKK:s[0-9]+]], 0xffff{{$}}
-; GCN-DAG: s_mov_b32 [[K:s[0-9]+]], 0x12341234
-
-; GCN-DAG: {{flat|global}}_load_dword [[IDX:v[0-9]+]]
-; GCN-DAG: {{flat|global}}_load_dword [[VEC:v[0-9]+]]
-
-; GFX89-DAG: v_lshlrev_b32_e32 [[SCALED_IDX:v[0-9]+]], 4, [[IDX]]
-; GFX89-DAG: v_lshlrev_b32_e64 [[MASK:v[0-9]+]], [[SCALED_IDX]], [[MASKK]]
-
-; CI-DAG: v_lshlrev_b32_e32 [[SCALED_IDX:v[0-9]+]], 4, [[IDX]]
-; CI-DAG: v_lshl_b32_e32 [[MASK:v[0-9]+]], 0xffff, [[SCALED_IDX]]
-
-; GCN: v_bfi_b32 [[RESULT:v[0-9]+]], [[MASK]], [[K]], [[VEC]]
-; GCN: {{flat|global}}_store_dword v{{\[[0-9]+:[0-9]+\]}}, [[RESULT]]
 define amdgpu_kernel void @v_insertelement_v2f16_dynamic_vgpr(<2 x half> addrspace(1)* %out, <2 x half> addrspace(1)* %in, i32 addrspace(1)* %idx.ptr) #0 {
+; GFX9-LABEL: v_insertelement_v2f16_dynamic_vgpr:
+; GFX9:       ; %bb.0:
+; GFX9-NEXT:    s_load_dwordx4 s[0:3], s[4:5], 0x0
+; GFX9-NEXT:    s_load_dwordx4 s[4:7], s[4:5], 0x10
+; GFX9-NEXT:    v_lshlrev_b32_e32 v2, 2, v0
+; GFX9-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX9-NEXT:    s_mov_b32 s6, 0xffff
+; GFX9-NEXT:    s_mov_b32 s7, 0x12341234
+; GFX9-NEXT:    v_mov_b32_e32 v3, s3
+; GFX9-NEXT:    v_mov_b32_e32 v1, s5
+; GFX9-NEXT:    v_add_co_u32_e32 v0, vcc, s4, v2
+; GFX9-NEXT:    v_addc_co_u32_e32 v1, vcc, 0, v1, vcc
+; GFX9-NEXT:    global_load_dword v4, v[0:1], off
+; GFX9-NEXT:    v_add_co_u32_e32 v0, vcc, s2, v2
+; GFX9-NEXT:    v_addc_co_u32_e32 v1, vcc, 0, v3, vcc
+; GFX9-NEXT:    global_load_dword v3, v[0:1], off
+; GFX9-NEXT:    v_add_co_u32_e32 v0, vcc, s0, v2
+; GFX9-NEXT:    v_mov_b32_e32 v1, s1
+; GFX9-NEXT:    v_addc_co_u32_e32 v1, vcc, 0, v1, vcc
+; GFX9-NEXT:    s_waitcnt vmcnt(1)
+; GFX9-NEXT:    v_lshlrev_b32_e32 v2, 4, v4
+; GFX9-NEXT:    v_lshlrev_b32_e64 v2, v2, s6
+; GFX9-NEXT:    s_waitcnt vmcnt(0)
+; GFX9-NEXT:    v_bfi_b32 v2, v2, s7, v3
+; GFX9-NEXT:    global_store_dword v[0:1], v2, off
+; GFX9-NEXT:    s_endpgm
+;
+; VI-LABEL: v_insertelement_v2f16_dynamic_vgpr:
+; VI:       ; %bb.0:
+; VI-NEXT:    s_load_dwordx4 s[0:3], s[4:5], 0x0
+; VI-NEXT:    s_load_dwordx4 s[4:7], s[4:5], 0x10
+; VI-NEXT:    v_lshlrev_b32_e32 v2, 2, v0
+; VI-NEXT:    s_waitcnt lgkmcnt(0)
+; VI-NEXT:    s_mov_b32 s6, 0xffff
+; VI-NEXT:    s_mov_b32 s7, 0x12341234
+; VI-NEXT:    v_mov_b32_e32 v3, s3
+; VI-NEXT:    v_mov_b32_e32 v1, s5
+; VI-NEXT:    v_add_u32_e32 v0, vcc, s4, v2
+; VI-NEXT:    v_addc_u32_e32 v1, vcc, 0, v1, vcc
+; VI-NEXT:    flat_load_dword v4, v[0:1]
+; VI-NEXT:    v_add_u32_e32 v0, vcc, s2, v2
+; VI-NEXT:    v_addc_u32_e32 v1, vcc, 0, v3, vcc
+; VI-NEXT:    flat_load_dword v3, v[0:1]
+; VI-NEXT:    v_add_u32_e32 v0, vcc, s0, v2
+; VI-NEXT:    v_mov_b32_e32 v1, s1
+; VI-NEXT:    v_addc_u32_e32 v1, vcc, 0, v1, vcc
+; VI-NEXT:    s_waitcnt vmcnt(1) lgkmcnt(1)
+; VI-NEXT:    v_lshlrev_b32_e32 v2, 4, v4
+; VI-NEXT:    v_lshlrev_b32_e64 v2, v2, s6
+; VI-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; VI-NEXT:    v_bfi_b32 v2, v2, s7, v3
+; VI-NEXT:    flat_store_dword v[0:1], v2
+; VI-NEXT:    s_endpgm
+;
+; CI-LABEL: v_insertelement_v2f16_dynamic_vgpr:
+; CI:       ; %bb.0:
+; CI-NEXT:    s_load_dwordx4 s[0:3], s[4:5], 0x0
+; CI-NEXT:    s_load_dwordx4 s[4:7], s[4:5], 0x4
+; CI-NEXT:    v_lshlrev_b32_e32 v2, 2, v0
+; CI-NEXT:    s_waitcnt lgkmcnt(0)
+; CI-NEXT:    s_mov_b32 s6, 0x12341234
+; CI-NEXT:    v_mov_b32_e32 v3, s3
+; CI-NEXT:    v_mov_b32_e32 v1, s5
+; CI-NEXT:    v_add_i32_e32 v0, vcc, s4, v2
+; CI-NEXT:    v_addc_u32_e32 v1, vcc, 0, v1, vcc
+; CI-NEXT:    flat_load_dword v4, v[0:1]
+; CI-NEXT:    v_add_i32_e32 v0, vcc, s2, v2
+; CI-NEXT:    v_addc_u32_e32 v1, vcc, 0, v3, vcc
+; CI-NEXT:    flat_load_dword v3, v[0:1]
+; CI-NEXT:    v_add_i32_e32 v0, vcc, s0, v2
+; CI-NEXT:    v_mov_b32_e32 v1, s1
+; CI-NEXT:    v_addc_u32_e32 v1, vcc, 0, v1, vcc
+; CI-NEXT:    s_waitcnt vmcnt(1) lgkmcnt(1)
+; CI-NEXT:    v_lshlrev_b32_e32 v2, 4, v4
+; CI-NEXT:    v_lshl_b32_e32 v2, 0xffff, v2
+; CI-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; CI-NEXT:    v_bfi_b32 v2, v2, s6, v3
+; CI-NEXT:    flat_store_dword v[0:1], v2
+; CI-NEXT:    s_endpgm
   %tid = call i32 @llvm.amdgcn.workitem.id.x() #1
   %tid.ext = sext i32 %tid to i64
   %in.gep = getelementptr inbounds <2 x half>, <2 x half> addrspace(1)* %in, i64 %tid.ext
@@ -472,19 +1326,65 @@ define amdgpu_kernel void @v_insertelement_v2f16_dynamic_vgpr(<2 x half> addrspa
   ret void
 }
 
-; GCN-LABEL: {{^}}v_insertelement_v4f16_0:
-; GCN-DAG: s_load_dword [[VAL:s[0-9]+]], s[4:5],
-; GCN-DAG: {{flat|global}}_load_dwordx2 v{{\[}}[[LO:[0-9]+]]:[[HI:[0-9]+]]{{\]}}
-
-; GFX9-DAG: v_mov_b32_e32 [[BFI_MASK:v[0-9]+]], 0xffff{{$}}
-; GFX9: v_bfi_b32 v[[INS_LO:[0-9]+]], [[BFI_MASK]], [[VAL]], v[[LO]]
-
-; CIVI: s_and_b32 [[VAL_MASKED:s[0-9]+]], [[VAL]], 0xffff{{$}}
-; CIVI: v_and_b32_e32 [[AND:v[0-9]+]], 0xffff0000, v[[LO]]
-; CIVI: v_or_b32_e32 v[[INS_LO:[0-9]+]], [[VAL_MASKED]], [[AND]]
-
-; GCN: {{flat|global}}_store_dwordx2 v{{\[[0-9]+:[0-9]+\]}}, v{{\[}}[[INS_LO]]:[[HI]]{{\]}}
 define amdgpu_kernel void @v_insertelement_v4f16_0(<4 x half> addrspace(1)* %out, <4 x half> addrspace(1)* %in, [8 x i32], i32 %val) #0 {
+; GFX9-LABEL: v_insertelement_v4f16_0:
+; GFX9:       ; %bb.0:
+; GFX9-NEXT:    s_load_dwordx4 s[0:3], s[4:5], 0x0
+; GFX9-NEXT:    v_lshlrev_b32_e32 v2, 3, v0
+; GFX9-NEXT:    s_load_dword s4, s[4:5], 0x30
+; GFX9-NEXT:    v_mov_b32_e32 v4, 0xffff
+; GFX9-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX9-NEXT:    v_mov_b32_e32 v1, s3
+; GFX9-NEXT:    v_add_co_u32_e32 v0, vcc, s2, v2
+; GFX9-NEXT:    v_addc_co_u32_e32 v1, vcc, 0, v1, vcc
+; GFX9-NEXT:    global_load_dwordx2 v[0:1], v[0:1], off
+; GFX9-NEXT:    v_mov_b32_e32 v3, s1
+; GFX9-NEXT:    v_add_co_u32_e32 v2, vcc, s0, v2
+; GFX9-NEXT:    v_addc_co_u32_e32 v3, vcc, 0, v3, vcc
+; GFX9-NEXT:    s_waitcnt vmcnt(0)
+; GFX9-NEXT:    v_bfi_b32 v0, v4, s4, v0
+; GFX9-NEXT:    global_store_dwordx2 v[2:3], v[0:1], off
+; GFX9-NEXT:    s_endpgm
+;
+; VI-LABEL: v_insertelement_v4f16_0:
+; VI:       ; %bb.0:
+; VI-NEXT:    s_load_dwordx4 s[0:3], s[4:5], 0x0
+; VI-NEXT:    v_lshlrev_b32_e32 v2, 3, v0
+; VI-NEXT:    s_load_dword s4, s[4:5], 0x30
+; VI-NEXT:    s_waitcnt lgkmcnt(0)
+; VI-NEXT:    v_mov_b32_e32 v1, s3
+; VI-NEXT:    v_add_u32_e32 v0, vcc, s2, v2
+; VI-NEXT:    v_addc_u32_e32 v1, vcc, 0, v1, vcc
+; VI-NEXT:    flat_load_dwordx2 v[0:1], v[0:1]
+; VI-NEXT:    v_mov_b32_e32 v3, s1
+; VI-NEXT:    v_add_u32_e32 v2, vcc, s0, v2
+; VI-NEXT:    s_and_b32 s1, s4, 0xffff
+; VI-NEXT:    v_addc_u32_e32 v3, vcc, 0, v3, vcc
+; VI-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; VI-NEXT:    v_and_b32_e32 v0, 0xffff0000, v0
+; VI-NEXT:    v_or_b32_e32 v0, s1, v0
+; VI-NEXT:    flat_store_dwordx2 v[2:3], v[0:1]
+; VI-NEXT:    s_endpgm
+;
+; CI-LABEL: v_insertelement_v4f16_0:
+; CI:       ; %bb.0:
+; CI-NEXT:    s_load_dwordx4 s[0:3], s[4:5], 0x0
+; CI-NEXT:    v_lshlrev_b32_e32 v2, 3, v0
+; CI-NEXT:    s_load_dword s4, s[4:5], 0xc
+; CI-NEXT:    s_waitcnt lgkmcnt(0)
+; CI-NEXT:    v_mov_b32_e32 v1, s3
+; CI-NEXT:    v_add_i32_e32 v0, vcc, s2, v2
+; CI-NEXT:    v_addc_u32_e32 v1, vcc, 0, v1, vcc
+; CI-NEXT:    flat_load_dwordx2 v[0:1], v[0:1]
+; CI-NEXT:    v_mov_b32_e32 v3, s1
+; CI-NEXT:    v_add_i32_e32 v2, vcc, s0, v2
+; CI-NEXT:    s_and_b32 s1, s4, 0xffff
+; CI-NEXT:    v_addc_u32_e32 v3, vcc, 0, v3, vcc
+; CI-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; CI-NEXT:    v_and_b32_e32 v0, 0xffff0000, v0
+; CI-NEXT:    v_or_b32_e32 v0, s1, v0
+; CI-NEXT:    flat_store_dwordx2 v[2:3], v[0:1]
+; CI-NEXT:    s_endpgm
   %tid = call i32 @llvm.amdgcn.workitem.id.x() #1
   %tid.ext = sext i32 %tid to i64
   %in.gep = getelementptr inbounds <4 x half>, <4 x half> addrspace(1)* %in, i64 %tid.ext
@@ -497,23 +1397,65 @@ define amdgpu_kernel void @v_insertelement_v4f16_0(<4 x half> addrspace(1)* %out
   ret void
 }
 
-; GCN-LABEL: {{^}}v_insertelement_v4f16_1:
-; GCN-DAG: s_load_dword [[VAL:s[0-9]+]]
-; GCN-DAG: {{flat|global}}_load_dwordx2 v{{\[}}[[LO:[0-9]+]]:[[HI:[0-9]+]]{{\]}}
-
-; GFX9: v_and_b32_e32 [[AND:v[0-9]+]], 0xffff, v[[LO]]
-; GFX9: v_lshl_or_b32 v[[INS_HALF:[0-9]+]], [[VAL]], 16, [[AND]]
-
-; VI: s_lshl_b32 [[VAL_HI:s[0-9]+]], [[VAL]], 16
-; VI-DAG: v_mov_b32_e32 [[COPY_VAL:v[0-9]+]], [[VAL_HI]]
-; VI: v_or_b32_sdwa v[[INS_HALF:[0-9]+]], [[COPY_VAL]], v[[LO]] dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:WORD_0
-
-; CI: s_lshl_b32 [[VAL_HI:s[0-9]+]], [[VAL]], 16
-; CI: v_and_b32_e32 [[AND:v[0-9]+]], 0xffff, v[[LO]]
-; CI: v_or_b32_e32 v[[INS_HALF:[0-9]+]], [[VAL_HI]], [[AND]]
-
-; GCN: {{flat|global}}_store_dwordx2 v{{\[[0-9]+:[0-9]+\]}}, v{{\[}}[[INS_HALF]]:[[HI]]{{\]}}
 define amdgpu_kernel void @v_insertelement_v4f16_1(<4 x half> addrspace(1)* %out, <4 x half> addrspace(1)* %in, i32 %val) #0 {
+; GFX9-LABEL: v_insertelement_v4f16_1:
+; GFX9:       ; %bb.0:
+; GFX9-NEXT:    s_load_dwordx4 s[0:3], s[4:5], 0x0
+; GFX9-NEXT:    v_lshlrev_b32_e32 v2, 3, v0
+; GFX9-NEXT:    s_load_dword s4, s[4:5], 0x10
+; GFX9-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX9-NEXT:    v_mov_b32_e32 v1, s3
+; GFX9-NEXT:    v_add_co_u32_e32 v0, vcc, s2, v2
+; GFX9-NEXT:    v_addc_co_u32_e32 v1, vcc, 0, v1, vcc
+; GFX9-NEXT:    global_load_dwordx2 v[0:1], v[0:1], off
+; GFX9-NEXT:    v_mov_b32_e32 v3, s1
+; GFX9-NEXT:    v_add_co_u32_e32 v2, vcc, s0, v2
+; GFX9-NEXT:    v_addc_co_u32_e32 v3, vcc, 0, v3, vcc
+; GFX9-NEXT:    s_waitcnt vmcnt(0)
+; GFX9-NEXT:    v_and_b32_e32 v0, 0xffff, v0
+; GFX9-NEXT:    v_lshl_or_b32 v0, s4, 16, v0
+; GFX9-NEXT:    global_store_dwordx2 v[2:3], v[0:1], off
+; GFX9-NEXT:    s_endpgm
+;
+; VI-LABEL: v_insertelement_v4f16_1:
+; VI:       ; %bb.0:
+; VI-NEXT:    s_load_dwordx4 s[0:3], s[4:5], 0x0
+; VI-NEXT:    v_lshlrev_b32_e32 v2, 3, v0
+; VI-NEXT:    s_load_dword s4, s[4:5], 0x10
+; VI-NEXT:    s_waitcnt lgkmcnt(0)
+; VI-NEXT:    v_add_u32_e32 v0, vcc, s2, v2
+; VI-NEXT:    v_mov_b32_e32 v1, s3
+; VI-NEXT:    v_addc_u32_e32 v1, vcc, 0, v1, vcc
+; VI-NEXT:    flat_load_dwordx2 v[0:1], v[0:1]
+; VI-NEXT:    s_lshl_b32 s2, s4, 16
+; VI-NEXT:    v_mov_b32_e32 v4, s2
+; VI-NEXT:    v_mov_b32_e32 v3, s1
+; VI-NEXT:    v_add_u32_e32 v2, vcc, s0, v2
+; VI-NEXT:    v_addc_u32_e32 v3, vcc, 0, v3, vcc
+; VI-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; VI-NEXT:    v_or_b32_sdwa v0, v4, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:WORD_0
+; VI-NEXT:    flat_store_dwordx2 v[2:3], v[0:1]
+; VI-NEXT:    s_endpgm
+;
+; CI-LABEL: v_insertelement_v4f16_1:
+; CI:       ; %bb.0:
+; CI-NEXT:    s_load_dwordx4 s[0:3], s[4:5], 0x0
+; CI-NEXT:    v_lshlrev_b32_e32 v2, 3, v0
+; CI-NEXT:    s_load_dword s4, s[4:5], 0x4
+; CI-NEXT:    s_waitcnt lgkmcnt(0)
+; CI-NEXT:    v_mov_b32_e32 v1, s3
+; CI-NEXT:    v_add_i32_e32 v0, vcc, s2, v2
+; CI-NEXT:    v_addc_u32_e32 v1, vcc, 0, v1, vcc
+; CI-NEXT:    flat_load_dwordx2 v[0:1], v[0:1]
+; CI-NEXT:    v_mov_b32_e32 v3, s1
+; CI-NEXT:    v_add_i32_e32 v2, vcc, s0, v2
+; CI-NEXT:    s_lshl_b32 s1, s4, 16
+; CI-NEXT:    v_addc_u32_e32 v3, vcc, 0, v3, vcc
+; CI-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; CI-NEXT:    v_and_b32_e32 v0, 0xffff, v0
+; CI-NEXT:    v_or_b32_e32 v0, s1, v0
+; CI-NEXT:    flat_store_dwordx2 v[2:3], v[0:1]
+; CI-NEXT:    s_endpgm
   %tid = call i32 @llvm.amdgcn.workitem.id.x() #1
   %tid.ext = sext i32 %tid to i64
   %in.gep = getelementptr inbounds <4 x half>, <4 x half> addrspace(1)* %in, i64 %tid.ext
@@ -526,19 +1468,65 @@ define amdgpu_kernel void @v_insertelement_v4f16_1(<4 x half> addrspace(1)* %out
   ret void
 }
 
-; GCN-LABEL: {{^}}v_insertelement_v4f16_2:
-; GCN-DAG: s_load_dword [[VAL:s[0-9]+]], s[4:5],
-; GCN-DAG: {{flat|global}}_load_dwordx2 v{{\[}}[[LO:[0-9]+]]:[[HI:[0-9]+]]{{\]}}
-
-; GFX9-DAG: v_mov_b32_e32 [[BFI_MASK:v[0-9]+]], 0xffff{{$}}
-; GFX9: v_bfi_b32 v[[INS_HI:[0-9]+]], [[BFI_MASK]], [[VAL]], v[[HI]]
-
-; CIVI: s_and_b32 [[VAL_MASKED:s[0-9]+]], [[VAL]], 0xffff{{$}}
-; CIVI: v_and_b32_e32 [[AND:v[0-9]+]], 0xffff0000, v[[HI]]
-; CIVI: v_or_b32_e32 v[[INS_HI:[0-9]+]], [[VAL_MASKED]], [[AND]]
-
-; GCN: {{flat|global}}_store_dwordx2 v{{\[[0-9]+:[0-9]+\]}}, v{{\[}}[[LO]]:[[INS_HI]]{{\]}}
 define amdgpu_kernel void @v_insertelement_v4f16_2(<4 x half> addrspace(1)* %out, <4 x half> addrspace(1)* %in, [8 x i32], i32 %val) #0 {
+; GFX9-LABEL: v_insertelement_v4f16_2:
+; GFX9:       ; %bb.0:
+; GFX9-NEXT:    s_load_dwordx4 s[0:3], s[4:5], 0x0
+; GFX9-NEXT:    v_lshlrev_b32_e32 v2, 3, v0
+; GFX9-NEXT:    s_load_dword s4, s[4:5], 0x30
+; GFX9-NEXT:    v_mov_b32_e32 v4, 0xffff
+; GFX9-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX9-NEXT:    v_mov_b32_e32 v1, s3
+; GFX9-NEXT:    v_add_co_u32_e32 v0, vcc, s2, v2
+; GFX9-NEXT:    v_addc_co_u32_e32 v1, vcc, 0, v1, vcc
+; GFX9-NEXT:    global_load_dwordx2 v[0:1], v[0:1], off
+; GFX9-NEXT:    v_mov_b32_e32 v3, s1
+; GFX9-NEXT:    v_add_co_u32_e32 v2, vcc, s0, v2
+; GFX9-NEXT:    v_addc_co_u32_e32 v3, vcc, 0, v3, vcc
+; GFX9-NEXT:    s_waitcnt vmcnt(0)
+; GFX9-NEXT:    v_bfi_b32 v1, v4, s4, v1
+; GFX9-NEXT:    global_store_dwordx2 v[2:3], v[0:1], off
+; GFX9-NEXT:    s_endpgm
+;
+; VI-LABEL: v_insertelement_v4f16_2:
+; VI:       ; %bb.0:
+; VI-NEXT:    s_load_dwordx4 s[0:3], s[4:5], 0x0
+; VI-NEXT:    v_lshlrev_b32_e32 v2, 3, v0
+; VI-NEXT:    s_load_dword s4, s[4:5], 0x30
+; VI-NEXT:    s_waitcnt lgkmcnt(0)
+; VI-NEXT:    v_mov_b32_e32 v1, s3
+; VI-NEXT:    v_add_u32_e32 v0, vcc, s2, v2
+; VI-NEXT:    v_addc_u32_e32 v1, vcc, 0, v1, vcc
+; VI-NEXT:    flat_load_dwordx2 v[0:1], v[0:1]
+; VI-NEXT:    v_mov_b32_e32 v3, s1
+; VI-NEXT:    v_add_u32_e32 v2, vcc, s0, v2
+; VI-NEXT:    s_and_b32 s1, s4, 0xffff
+; VI-NEXT:    v_addc_u32_e32 v3, vcc, 0, v3, vcc
+; VI-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; VI-NEXT:    v_and_b32_e32 v1, 0xffff0000, v1
+; VI-NEXT:    v_or_b32_e32 v1, s1, v1
+; VI-NEXT:    flat_store_dwordx2 v[2:3], v[0:1]
+; VI-NEXT:    s_endpgm
+;
+; CI-LABEL: v_insertelement_v4f16_2:
+; CI:       ; %bb.0:
+; CI-NEXT:    s_load_dwordx4 s[0:3], s[4:5], 0x0
+; CI-NEXT:    v_lshlrev_b32_e32 v2, 3, v0
+; CI-NEXT:    s_load_dword s4, s[4:5], 0xc
+; CI-NEXT:    s_waitcnt lgkmcnt(0)
+; CI-NEXT:    v_mov_b32_e32 v1, s3
+; CI-NEXT:    v_add_i32_e32 v0, vcc, s2, v2
+; CI-NEXT:    v_addc_u32_e32 v1, vcc, 0, v1, vcc
+; CI-NEXT:    flat_load_dwordx2 v[0:1], v[0:1]
+; CI-NEXT:    v_mov_b32_e32 v3, s1
+; CI-NEXT:    v_add_i32_e32 v2, vcc, s0, v2
+; CI-NEXT:    s_and_b32 s1, s4, 0xffff
+; CI-NEXT:    v_addc_u32_e32 v3, vcc, 0, v3, vcc
+; CI-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; CI-NEXT:    v_and_b32_e32 v1, 0xffff0000, v1
+; CI-NEXT:    v_or_b32_e32 v1, s1, v1
+; CI-NEXT:    flat_store_dwordx2 v[2:3], v[0:1]
+; CI-NEXT:    s_endpgm
   %tid = call i32 @llvm.amdgcn.workitem.id.x() #1
   %tid.ext = sext i32 %tid to i64
   %in.gep = getelementptr inbounds <4 x half>, <4 x half> addrspace(1)* %in, i64 %tid.ext
@@ -551,23 +1539,65 @@ define amdgpu_kernel void @v_insertelement_v4f16_2(<4 x half> addrspace(1)* %out
   ret void
 }
 
-; GCN-LABEL: {{^}}v_insertelement_v4f16_3:
-; GCN-DAG: s_load_dword [[VAL:s[0-9]+]]
-; GCN-DAG: {{flat|global}}_load_dwordx2 v{{\[}}[[LO:[0-9]+]]:[[HI:[0-9]+]]{{\]}}
-
-; GFX9: v_and_b32_e32 [[AND:v[0-9]+]], 0xffff, v[[HI]]
-; GFX9: v_lshl_or_b32 v[[INS_HI:[0-9]+]], [[VAL]], 16, [[AND]]
-
-; VI: s_lshl_b32 [[VAL_HI:s[0-9]+]], [[VAL]], 16
-; VI-DAG: v_mov_b32_e32 [[COPY_VAL:v[0-9]+]], [[VAL_HI]]
-; VI: v_or_b32_sdwa v[[INS_HI:[0-9]+]], [[COPY_VAL]], v[[HI]] dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:WORD_0
-
-; CI: s_lshl_b32 [[VAL_HI:s[0-9]+]], [[VAL]], 16
-; CI: v_and_b32_e32 [[AND:v[0-9]+]], 0xffff, v[[HI]]
-; CI: v_or_b32_e32 v[[INS_HI:[0-9]+]], [[VAL_HI]], [[AND]]
-
-; GCN: {{flat|global}}_store_dwordx2 v{{\[[0-9]+:[0-9]+\]}}, v{{\[}}[[LO]]:[[INS_HI]]{{\]}}
 define amdgpu_kernel void @v_insertelement_v4f16_3(<4 x half> addrspace(1)* %out, <4 x half> addrspace(1)* %in, i32 %val) #0 {
+; GFX9-LABEL: v_insertelement_v4f16_3:
+; GFX9:       ; %bb.0:
+; GFX9-NEXT:    s_load_dwordx4 s[0:3], s[4:5], 0x0
+; GFX9-NEXT:    v_lshlrev_b32_e32 v2, 3, v0
+; GFX9-NEXT:    s_load_dword s4, s[4:5], 0x10
+; GFX9-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX9-NEXT:    v_mov_b32_e32 v1, s3
+; GFX9-NEXT:    v_add_co_u32_e32 v0, vcc, s2, v2
+; GFX9-NEXT:    v_addc_co_u32_e32 v1, vcc, 0, v1, vcc
+; GFX9-NEXT:    global_load_dwordx2 v[0:1], v[0:1], off
+; GFX9-NEXT:    v_mov_b32_e32 v3, s1
+; GFX9-NEXT:    v_add_co_u32_e32 v2, vcc, s0, v2
+; GFX9-NEXT:    v_addc_co_u32_e32 v3, vcc, 0, v3, vcc
+; GFX9-NEXT:    s_waitcnt vmcnt(0)
+; GFX9-NEXT:    v_and_b32_e32 v1, 0xffff, v1
+; GFX9-NEXT:    v_lshl_or_b32 v1, s4, 16, v1
+; GFX9-NEXT:    global_store_dwordx2 v[2:3], v[0:1], off
+; GFX9-NEXT:    s_endpgm
+;
+; VI-LABEL: v_insertelement_v4f16_3:
+; VI:       ; %bb.0:
+; VI-NEXT:    s_load_dwordx4 s[0:3], s[4:5], 0x0
+; VI-NEXT:    v_lshlrev_b32_e32 v2, 3, v0
+; VI-NEXT:    s_load_dword s4, s[4:5], 0x10
+; VI-NEXT:    s_waitcnt lgkmcnt(0)
+; VI-NEXT:    v_add_u32_e32 v0, vcc, s2, v2
+; VI-NEXT:    v_mov_b32_e32 v1, s3
+; VI-NEXT:    v_addc_u32_e32 v1, vcc, 0, v1, vcc
+; VI-NEXT:    flat_load_dwordx2 v[0:1], v[0:1]
+; VI-NEXT:    s_lshl_b32 s2, s4, 16
+; VI-NEXT:    v_mov_b32_e32 v4, s2
+; VI-NEXT:    v_mov_b32_e32 v3, s1
+; VI-NEXT:    v_add_u32_e32 v2, vcc, s0, v2
+; VI-NEXT:    v_addc_u32_e32 v3, vcc, 0, v3, vcc
+; VI-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; VI-NEXT:    v_or_b32_sdwa v1, v4, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:WORD_0
+; VI-NEXT:    flat_store_dwordx2 v[2:3], v[0:1]
+; VI-NEXT:    s_endpgm
+;
+; CI-LABEL: v_insertelement_v4f16_3:
+; CI:       ; %bb.0:
+; CI-NEXT:    s_load_dwordx4 s[0:3], s[4:5], 0x0
+; CI-NEXT:    v_lshlrev_b32_e32 v2, 3, v0
+; CI-NEXT:    s_load_dword s4, s[4:5], 0x4
+; CI-NEXT:    s_waitcnt lgkmcnt(0)
+; CI-NEXT:    v_mov_b32_e32 v1, s3
+; CI-NEXT:    v_add_i32_e32 v0, vcc, s2, v2
+; CI-NEXT:    v_addc_u32_e32 v1, vcc, 0, v1, vcc
+; CI-NEXT:    flat_load_dwordx2 v[0:1], v[0:1]
+; CI-NEXT:    v_mov_b32_e32 v3, s1
+; CI-NEXT:    v_add_i32_e32 v2, vcc, s0, v2
+; CI-NEXT:    s_lshl_b32 s1, s4, 16
+; CI-NEXT:    v_addc_u32_e32 v3, vcc, 0, v3, vcc
+; CI-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; CI-NEXT:    v_and_b32_e32 v1, 0xffff, v1
+; CI-NEXT:    v_or_b32_e32 v1, s1, v1
+; CI-NEXT:    flat_store_dwordx2 v[2:3], v[0:1]
+; CI-NEXT:    s_endpgm
   %tid = call i32 @llvm.amdgcn.workitem.id.x() #1
   %tid.ext = sext i32 %tid to i64
   %in.gep = getelementptr inbounds <4 x half>, <4 x half> addrspace(1)* %in, i64 %tid.ext
@@ -580,19 +1610,65 @@ define amdgpu_kernel void @v_insertelement_v4f16_3(<4 x half> addrspace(1)* %out
   ret void
 }
 
-; GCN-LABEL: {{^}}v_insertelement_v4i16_2:
-; GCN-DAG: s_load_dword [[VAL:s[0-9]+]]
-; GCN-DAG: {{flat|global}}_load_dwordx2 v{{\[}}[[LO:[0-9]+]]:[[HI:[0-9]+]]{{\]}}
-
-; GFX9-DAG: v_mov_b32_e32 [[BFI_MASK:v[0-9]+]], 0xffff{{$}}
-; GFX9: v_bfi_b32 v[[INS_HI:[0-9]+]], [[BFI_MASK]], [[VAL]], v[[HI]]
-
-; CIVI: s_and_b32 [[VAL_MASKED:s[0-9]+]], [[VAL]], 0xffff{{$}}
-; CIVI: v_and_b32_e32 [[AND:v[0-9]+]], 0xffff0000, v[[HI]]
-; CIVI: v_or_b32_e32 v[[INS_HI:[0-9]+]], [[VAL_MASKED]], [[AND]]
-
-; GCN: {{flat|global}}_store_dwordx2 v{{\[[0-9]+:[0-9]+\]}}, v{{\[}}[[LO]]:[[INS_HI]]{{\]}}
 define amdgpu_kernel void @v_insertelement_v4i16_2(<4 x i16> addrspace(1)* %out, <4 x i16> addrspace(1)* %in, i32 %val) #0 {
+; GFX9-LABEL: v_insertelement_v4i16_2:
+; GFX9:       ; %bb.0:
+; GFX9-NEXT:    s_load_dwordx4 s[0:3], s[4:5], 0x0
+; GFX9-NEXT:    v_lshlrev_b32_e32 v2, 3, v0
+; GFX9-NEXT:    s_load_dword s4, s[4:5], 0x10
+; GFX9-NEXT:    v_mov_b32_e32 v4, 0xffff
+; GFX9-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX9-NEXT:    v_mov_b32_e32 v1, s3
+; GFX9-NEXT:    v_add_co_u32_e32 v0, vcc, s2, v2
+; GFX9-NEXT:    v_addc_co_u32_e32 v1, vcc, 0, v1, vcc
+; GFX9-NEXT:    global_load_dwordx2 v[0:1], v[0:1], off
+; GFX9-NEXT:    v_mov_b32_e32 v3, s1
+; GFX9-NEXT:    v_add_co_u32_e32 v2, vcc, s0, v2
+; GFX9-NEXT:    v_addc_co_u32_e32 v3, vcc, 0, v3, vcc
+; GFX9-NEXT:    s_waitcnt vmcnt(0)
+; GFX9-NEXT:    v_bfi_b32 v1, v4, s4, v1
+; GFX9-NEXT:    global_store_dwordx2 v[2:3], v[0:1], off
+; GFX9-NEXT:    s_endpgm
+;
+; VI-LABEL: v_insertelement_v4i16_2:
+; VI:       ; %bb.0:
+; VI-NEXT:    s_load_dwordx4 s[0:3], s[4:5], 0x0
+; VI-NEXT:    v_lshlrev_b32_e32 v2, 3, v0
+; VI-NEXT:    s_load_dword s4, s[4:5], 0x10
+; VI-NEXT:    s_waitcnt lgkmcnt(0)
+; VI-NEXT:    v_mov_b32_e32 v1, s3
+; VI-NEXT:    v_add_u32_e32 v0, vcc, s2, v2
+; VI-NEXT:    v_addc_u32_e32 v1, vcc, 0, v1, vcc
+; VI-NEXT:    flat_load_dwordx2 v[0:1], v[0:1]
+; VI-NEXT:    v_mov_b32_e32 v3, s1
+; VI-NEXT:    v_add_u32_e32 v2, vcc, s0, v2
+; VI-NEXT:    s_and_b32 s1, s4, 0xffff
+; VI-NEXT:    v_addc_u32_e32 v3, vcc, 0, v3, vcc
+; VI-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; VI-NEXT:    v_and_b32_e32 v1, 0xffff0000, v1
+; VI-NEXT:    v_or_b32_e32 v1, s1, v1
+; VI-NEXT:    flat_store_dwordx2 v[2:3], v[0:1]
+; VI-NEXT:    s_endpgm
+;
+; CI-LABEL: v_insertelement_v4i16_2:
+; CI:       ; %bb.0:
+; CI-NEXT:    s_load_dwordx4 s[0:3], s[4:5], 0x0
+; CI-NEXT:    v_lshlrev_b32_e32 v2, 3, v0
+; CI-NEXT:    s_load_dword s4, s[4:5], 0x4
+; CI-NEXT:    s_waitcnt lgkmcnt(0)
+; CI-NEXT:    v_mov_b32_e32 v1, s3
+; CI-NEXT:    v_add_i32_e32 v0, vcc, s2, v2
+; CI-NEXT:    v_addc_u32_e32 v1, vcc, 0, v1, vcc
+; CI-NEXT:    flat_load_dwordx2 v[0:1], v[0:1]
+; CI-NEXT:    v_mov_b32_e32 v3, s1
+; CI-NEXT:    v_add_i32_e32 v2, vcc, s0, v2
+; CI-NEXT:    s_and_b32 s1, s4, 0xffff
+; CI-NEXT:    v_addc_u32_e32 v3, vcc, 0, v3, vcc
+; CI-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; CI-NEXT:    v_and_b32_e32 v1, 0xffff0000, v1
+; CI-NEXT:    v_or_b32_e32 v1, s1, v1
+; CI-NEXT:    flat_store_dwordx2 v[2:3], v[0:1]
+; CI-NEXT:    s_endpgm
   %tid = call i32 @llvm.amdgcn.workitem.id.x() #1
   %tid.ext = sext i32 %tid to i64
   %in.gep = getelementptr inbounds <4 x i16>, <4 x i16> addrspace(1)* %in, i64 %tid.ext
@@ -606,26 +1682,90 @@ define amdgpu_kernel void @v_insertelement_v4i16_2(<4 x i16> addrspace(1)* %out,
 }
 
 ; FIXME: Better code on CI?
-; GCN-LABEL: {{^}}v_insertelement_v4i16_dynamic_vgpr:
-; GCN-DAG: {{flat|global}}_load_dword [[IDX:v[0-9]+]],
-; GCN-DAG: s_load_dword [[VAL:s[0-9]+]]
-; GCN-DAG: {{flat|global}}_load_dwordx2 v{{\[}}[[LO:[0-9]+]]:[[HI:[0-9]+]]{{\]}}
-
-; GCN-DAG: s_mov_b32 s[[MASK_LO:[0-9]+]], 0xffff
-; GCN-DAG: s_mov_b32 s[[MASK_HI:[0-9]+]], 0
-; CIVI-DAG: s_and_b32 [[MASKED_VAL:s[0-9]+]], [[VAL]], s[[MASK_LO]]
-; VI-DAG: s_lshl_b32 [[SHIFTED_VAL:s[0-9]+]], [[MASKED_VAL]], 16
-; CI-DAG: s_lshl_b32 [[SHIFTED_VAL:s[0-9]+]], [[VAL]], 16
-; CIVI: s_or_b32 [[DUP_VAL:s[0-9]+]], [[MASKED_VAL]], [[SHIFTED_VAL]]
-; GCN-DAG: v_lshlrev_b32_e32 [[SCALED_IDX:v[0-9]+]], 4, [[IDX]]
-; GFX9-DAG: s_pack_ll_b32_b16 [[DUP_VAL:s[0-9]+]], [[VAL]], [[VAL]]
-; GFX89: v_lshlrev_b64 v[{{[0-9:]+}}], [[SCALED_IDX]], s{{\[}}[[MASK_LO]]:[[MASK_HI]]{{\]}}
-; CI: v_lshl_b64 v[{{[0-9:]+}}], s[{{[0-9:]+}}], [[SCALED_IDX]]
-; GCN: v_bfi_b32 v{{[0-9]+}}, v{{[0-9]+}}, [[DUP_VAL]], v{{[0-9]+}}
-; GCN: v_bfi_b32 v{{[0-9]+}}, v{{[0-9]+}}, [[DUP_VAL]], v{{[0-9]+}}
-
-; GCN: {{flat|global}}_store_dwordx2 v{{\[[0-9]+:[0-9]+\]}}, v{{\[[0-9]+:[0-9]+\]}}
 define amdgpu_kernel void @v_insertelement_v4i16_dynamic_vgpr(<4 x i16> addrspace(1)* %out, <4 x i16> addrspace(1)* %in, i32 %val) #0 {
+; GFX9-LABEL: v_insertelement_v4i16_dynamic_vgpr:
+; GFX9:       ; %bb.0:
+; GFX9-NEXT:    s_load_dwordx4 s[0:3], s[4:5], 0x0
+; GFX9-NEXT:    global_load_dword v4, v[0:1], off
+; GFX9-NEXT:    v_lshlrev_b32_e32 v2, 3, v0
+; GFX9-NEXT:    s_load_dword s6, s[4:5], 0x10
+; GFX9-NEXT:    s_mov_b32 s5, 0
+; GFX9-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX9-NEXT:    v_mov_b32_e32 v1, s3
+; GFX9-NEXT:    v_add_co_u32_e32 v0, vcc, s2, v2
+; GFX9-NEXT:    v_addc_co_u32_e32 v1, vcc, 0, v1, vcc
+; GFX9-NEXT:    global_load_dwordx2 v[0:1], v[0:1], off
+; GFX9-NEXT:    s_mov_b32 s4, 0xffff
+; GFX9-NEXT:    v_mov_b32_e32 v3, s1
+; GFX9-NEXT:    v_add_co_u32_e32 v2, vcc, s0, v2
+; GFX9-NEXT:    s_pack_ll_b32_b16 s1, s6, s6
+; GFX9-NEXT:    v_addc_co_u32_e32 v3, vcc, 0, v3, vcc
+; GFX9-NEXT:    s_waitcnt vmcnt(1)
+; GFX9-NEXT:    v_lshlrev_b32_e32 v4, 4, v4
+; GFX9-NEXT:    v_lshlrev_b64 v[4:5], v4, s[4:5]
+; GFX9-NEXT:    s_waitcnt vmcnt(0)
+; GFX9-NEXT:    v_bfi_b32 v1, v5, s1, v1
+; GFX9-NEXT:    v_bfi_b32 v0, v4, s1, v0
+; GFX9-NEXT:    global_store_dwordx2 v[2:3], v[0:1], off
+; GFX9-NEXT:    s_endpgm
+;
+; VI-LABEL: v_insertelement_v4i16_dynamic_vgpr:
+; VI:       ; %bb.0:
+; VI-NEXT:    s_load_dwordx4 s[0:3], s[4:5], 0x0
+; VI-NEXT:    s_waitcnt lgkmcnt(0)
+; VI-NEXT:    flat_load_dword v4, v[0:1]
+; VI-NEXT:    v_lshlrev_b32_e32 v2, 3, v0
+; VI-NEXT:    s_load_dword s6, s[4:5], 0x10
+; VI-NEXT:    s_mov_b32 s4, 0xffff
+; VI-NEXT:    v_mov_b32_e32 v1, s3
+; VI-NEXT:    v_add_u32_e32 v0, vcc, s2, v2
+; VI-NEXT:    v_addc_u32_e32 v1, vcc, 0, v1, vcc
+; VI-NEXT:    flat_load_dwordx2 v[0:1], v[0:1]
+; VI-NEXT:    v_mov_b32_e32 v3, s1
+; VI-NEXT:    s_mov_b32 s5, 0
+; VI-NEXT:    s_waitcnt lgkmcnt(0)
+; VI-NEXT:    s_and_b32 s1, s6, s4
+; VI-NEXT:    v_add_u32_e32 v2, vcc, s0, v2
+; VI-NEXT:    s_lshl_b32 s0, s1, 16
+; VI-NEXT:    s_or_b32 s0, s1, s0
+; VI-NEXT:    v_addc_u32_e32 v3, vcc, 0, v3, vcc
+; VI-NEXT:    s_waitcnt vmcnt(1)
+; VI-NEXT:    v_lshlrev_b32_e32 v4, 4, v4
+; VI-NEXT:    v_lshlrev_b64 v[4:5], v4, s[4:5]
+; VI-NEXT:    s_waitcnt vmcnt(0)
+; VI-NEXT:    v_bfi_b32 v1, v5, s0, v1
+; VI-NEXT:    v_bfi_b32 v0, v4, s0, v0
+; VI-NEXT:    flat_store_dwordx2 v[2:3], v[0:1]
+; VI-NEXT:    s_endpgm
+;
+; CI-LABEL: v_insertelement_v4i16_dynamic_vgpr:
+; CI:       ; %bb.0:
+; CI-NEXT:    s_load_dwordx4 s[0:3], s[4:5], 0x0
+; CI-NEXT:    s_waitcnt lgkmcnt(0)
+; CI-NEXT:    flat_load_dword v4, v[0:1]
+; CI-NEXT:    v_lshlrev_b32_e32 v2, 3, v0
+; CI-NEXT:    s_load_dword s6, s[4:5], 0x4
+; CI-NEXT:    s_mov_b32 s4, 0xffff
+; CI-NEXT:    v_mov_b32_e32 v1, s3
+; CI-NEXT:    v_add_i32_e32 v0, vcc, s2, v2
+; CI-NEXT:    v_addc_u32_e32 v1, vcc, 0, v1, vcc
+; CI-NEXT:    flat_load_dwordx2 v[0:1], v[0:1]
+; CI-NEXT:    s_mov_b32 s5, 0
+; CI-NEXT:    s_waitcnt lgkmcnt(0)
+; CI-NEXT:    s_lshl_b32 s2, s6, 16
+; CI-NEXT:    s_and_b32 s3, s6, s4
+; CI-NEXT:    v_mov_b32_e32 v3, s1
+; CI-NEXT:    v_add_i32_e32 v2, vcc, s0, v2
+; CI-NEXT:    s_or_b32 s1, s3, s2
+; CI-NEXT:    v_addc_u32_e32 v3, vcc, 0, v3, vcc
+; CI-NEXT:    s_waitcnt vmcnt(1)
+; CI-NEXT:    v_lshlrev_b32_e32 v4, 4, v4
+; CI-NEXT:    v_lshl_b64 v[4:5], s[4:5], v4
+; CI-NEXT:    s_waitcnt vmcnt(0)
+; CI-NEXT:    v_bfi_b32 v1, v5, s1, v1
+; CI-NEXT:    v_bfi_b32 v0, v4, s1, v0
+; CI-NEXT:    flat_store_dwordx2 v[2:3], v[0:1]
+; CI-NEXT:    s_endpgm
   %tid = call i32 @llvm.amdgcn.workitem.id.x() #1
   %tid.ext = sext i32 %tid to i64
   %in.gep = getelementptr inbounds <4 x i16>, <4 x i16> addrspace(1)* %in, i64 %tid.ext
@@ -639,8 +1779,88 @@ define amdgpu_kernel void @v_insertelement_v4i16_dynamic_vgpr(<4 x i16> addrspac
   ret void
 }
 
-; GCN-LABEL: {{^}}v_insertelement_v4f16_dynamic_sgpr:
 define amdgpu_kernel void @v_insertelement_v4f16_dynamic_sgpr(<4 x half> addrspace(1)* %out, <4 x half> addrspace(1)* %in, i32 %val, i32 %idxval) #0 {
+; GFX9-LABEL: v_insertelement_v4f16_dynamic_sgpr:
+; GFX9:       ; %bb.0:
+; GFX9-NEXT:    s_load_dwordx4 s[0:3], s[4:5], 0x0
+; GFX9-NEXT:    v_lshlrev_b32_e32 v2, 3, v0
+; GFX9-NEXT:    s_load_dwordx2 s[4:5], s[4:5], 0x10
+; GFX9-NEXT:    s_mov_b32 s7, 0
+; GFX9-NEXT:    s_mov_b32 s6, 0xffff
+; GFX9-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX9-NEXT:    v_mov_b32_e32 v1, s3
+; GFX9-NEXT:    v_add_co_u32_e32 v0, vcc, s2, v2
+; GFX9-NEXT:    v_addc_co_u32_e32 v1, vcc, 0, v1, vcc
+; GFX9-NEXT:    global_load_dwordx2 v[0:1], v[0:1], off
+; GFX9-NEXT:    s_pack_ll_b32_b16 s3, s4, s4
+; GFX9-NEXT:    s_lshl_b32 s2, s5, 4
+; GFX9-NEXT:    v_mov_b32_e32 v3, s1
+; GFX9-NEXT:    v_add_co_u32_e32 v2, vcc, s0, v2
+; GFX9-NEXT:    s_lshl_b64 s[0:1], s[6:7], s2
+; GFX9-NEXT:    v_mov_b32_e32 v4, s3
+; GFX9-NEXT:    v_mov_b32_e32 v5, s3
+; GFX9-NEXT:    v_addc_co_u32_e32 v3, vcc, 0, v3, vcc
+; GFX9-NEXT:    s_waitcnt vmcnt(0)
+; GFX9-NEXT:    v_bfi_b32 v1, s1, v4, v1
+; GFX9-NEXT:    v_bfi_b32 v0, s0, v5, v0
+; GFX9-NEXT:    global_store_dwordx2 v[2:3], v[0:1], off
+; GFX9-NEXT:    s_endpgm
+;
+; VI-LABEL: v_insertelement_v4f16_dynamic_sgpr:
+; VI:       ; %bb.0:
+; VI-NEXT:    s_load_dwordx4 s[0:3], s[4:5], 0x0
+; VI-NEXT:    v_lshlrev_b32_e32 v2, 3, v0
+; VI-NEXT:    s_load_dwordx2 s[4:5], s[4:5], 0x10
+; VI-NEXT:    s_mov_b32 s6, 0xffff
+; VI-NEXT:    s_mov_b32 s7, 0
+; VI-NEXT:    s_waitcnt lgkmcnt(0)
+; VI-NEXT:    v_add_u32_e32 v0, vcc, s2, v2
+; VI-NEXT:    v_mov_b32_e32 v1, s3
+; VI-NEXT:    v_addc_u32_e32 v1, vcc, 0, v1, vcc
+; VI-NEXT:    flat_load_dwordx2 v[0:1], v[0:1]
+; VI-NEXT:    s_and_b32 s2, s4, s6
+; VI-NEXT:    s_lshl_b32 s3, s2, 16
+; VI-NEXT:    s_or_b32 s2, s2, s3
+; VI-NEXT:    s_lshl_b32 s4, s5, 4
+; VI-NEXT:    v_mov_b32_e32 v3, s1
+; VI-NEXT:    v_add_u32_e32 v2, vcc, s0, v2
+; VI-NEXT:    s_lshl_b64 s[0:1], s[6:7], s4
+; VI-NEXT:    v_mov_b32_e32 v4, s2
+; VI-NEXT:    v_mov_b32_e32 v5, s2
+; VI-NEXT:    v_addc_u32_e32 v3, vcc, 0, v3, vcc
+; VI-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; VI-NEXT:    v_bfi_b32 v1, s1, v4, v1
+; VI-NEXT:    v_bfi_b32 v0, s0, v5, v0
+; VI-NEXT:    flat_store_dwordx2 v[2:3], v[0:1]
+; VI-NEXT:    s_endpgm
+;
+; CI-LABEL: v_insertelement_v4f16_dynamic_sgpr:
+; CI:       ; %bb.0:
+; CI-NEXT:    s_load_dwordx4 s[0:3], s[4:5], 0x0
+; CI-NEXT:    v_lshlrev_b32_e32 v2, 3, v0
+; CI-NEXT:    s_load_dwordx2 s[4:5], s[4:5], 0x4
+; CI-NEXT:    s_mov_b32 s6, 0xffff
+; CI-NEXT:    s_mov_b32 s7, 0
+; CI-NEXT:    s_waitcnt lgkmcnt(0)
+; CI-NEXT:    v_mov_b32_e32 v1, s3
+; CI-NEXT:    v_add_i32_e32 v0, vcc, s2, v2
+; CI-NEXT:    v_addc_u32_e32 v1, vcc, 0, v1, vcc
+; CI-NEXT:    flat_load_dwordx2 v[0:1], v[0:1]
+; CI-NEXT:    s_and_b32 s2, s4, s6
+; CI-NEXT:    s_lshl_b32 s3, s4, 16
+; CI-NEXT:    s_or_b32 s2, s2, s3
+; CI-NEXT:    s_lshl_b32 s4, s5, 4
+; CI-NEXT:    v_mov_b32_e32 v3, s1
+; CI-NEXT:    v_add_i32_e32 v2, vcc, s0, v2
+; CI-NEXT:    s_lshl_b64 s[0:1], s[6:7], s4
+; CI-NEXT:    v_mov_b32_e32 v4, s2
+; CI-NEXT:    v_mov_b32_e32 v5, s2
+; CI-NEXT:    v_addc_u32_e32 v3, vcc, 0, v3, vcc
+; CI-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; CI-NEXT:    v_bfi_b32 v1, s1, v4, v1
+; CI-NEXT:    v_bfi_b32 v0, s0, v5, v0
+; CI-NEXT:    flat_store_dwordx2 v[2:3], v[0:1]
+; CI-NEXT:    s_endpgm
   %tid = call i32 @llvm.amdgcn.workitem.id.x() #1
   %tid.ext = sext i32 %tid to i64
   %in.gep = getelementptr inbounds <4 x half>, <4 x half> addrspace(1)* %in, i64 %tid.ext