; GFX9-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX9-NEXT: global_load_dwordx2 v[0:1], v[0:1], off
; GFX9-NEXT: s_waitcnt vmcnt(0)
-; GFX9-NEXT: v_lshrrev_b32_e32 v0, 16, v1
-; GFX9-NEXT: v_and_b32_e32 v1, 0xffff, v1
-; GFX9-NEXT: v_lshl_or_b32 v0, v0, 16, v1
+; GFX9-NEXT: v_mov_b32_e32 v0, v1
; GFX9-NEXT: s_setpc_b64 s[30:31]
%val0 = load <4 x half>, <4 x half> addrspace(1)* %arg0
%val1 = load <4 x half>, <4 x half> addrspace(1)* %arg1
; GFX9-NEXT: global_load_dwordx2 v[4:5], v[0:1], off
; GFX9-NEXT: global_load_dwordx2 v[1:2], v[2:3], off
; GFX9-NEXT: s_waitcnt vmcnt(1)
-; GFX9-NEXT: v_lshrrev_b32_e32 v0, 16, v5
+; GFX9-NEXT: v_mov_b32_e32 v0, v5
; GFX9-NEXT: s_waitcnt vmcnt(0)
-; GFX9-NEXT: v_and_b32_e32 v2, 0xffff, v5
-; GFX9-NEXT: v_lshl_or_b32 v0, v0, 16, v2
; GFX9-NEXT: s_setpc_b64 s[30:31]
%val0 = load <4 x half>, <4 x half> addrspace(1)* %arg0
%val1 = load <4 x half>, <4 x half> addrspace(1)* %arg1
; GFX9-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX9-NEXT: global_load_dwordx2 v[0:1], v[0:1], off
; GFX9-NEXT: s_waitcnt vmcnt(0)
-; GFX9-NEXT: v_lshrrev_b32_e32 v1, 16, v0
-; GFX9-NEXT: v_and_b32_e32 v0, 0xffff, v0
-; GFX9-NEXT: v_lshl_or_b32 v0, v1, 16, v0
; GFX9-NEXT: v_mov_b32_e32 v1, v0
; GFX9-NEXT: s_setpc_b64 s[30:31]
%val0 = load <4 x half>, <4 x half> addrspace(1)* %arg0
; GFX9-NEXT: s_waitcnt vmcnt(0)
; GFX9-NEXT: global_load_dwordx2 v[1:2], v[2:3], off
; GFX9-NEXT: s_waitcnt vmcnt(0)
-; GFX9-NEXT: v_mov_b32_e32 v2, 0xffff
-; GFX9-NEXT: v_lshrrev_b32_e32 v3, 16, v0
-; GFX9-NEXT: v_and_b32_e32 v0, v2, v0
-; GFX9-NEXT: v_lshrrev_b32_e32 v4, 16, v1
-; GFX9-NEXT: v_and_b32_e32 v1, v2, v1
-; GFX9-NEXT: v_lshl_or_b32 v0, v3, 16, v0
-; GFX9-NEXT: v_lshl_or_b32 v1, v4, 16, v1
; GFX9-NEXT: s_setpc_b64 s[30:31]
%val0 = load <4 x half>, <4 x half> addrspace(1)* %arg0
%val1 = load <4 x half>, <4 x half> addrspace(1)* %arg1
; GFX9-NEXT: s_waitcnt vmcnt(0)
; GFX9-NEXT: global_load_dwordx2 v[1:2], v[2:3], off
; GFX9-NEXT: s_waitcnt vmcnt(0)
-; GFX9-NEXT: v_mov_b32_e32 v1, 0xffff
-; GFX9-NEXT: v_lshrrev_b32_e32 v3, 16, v0
-; GFX9-NEXT: v_and_b32_e32 v0, v1, v0
-; GFX9-NEXT: v_lshrrev_b32_e32 v4, 16, v2
-; GFX9-NEXT: v_and_b32_e32 v1, v1, v2
-; GFX9-NEXT: v_lshl_or_b32 v0, v3, 16, v0
-; GFX9-NEXT: v_lshl_or_b32 v1, v4, 16, v1
+; GFX9-NEXT: v_mov_b32_e32 v1, v2
; GFX9-NEXT: s_setpc_b64 s[30:31]
%val0 = load <4 x half>, <4 x half> addrspace(1)* %arg0
%val1 = load <4 x half>, <4 x half> addrspace(1)* %arg1
; GFX9-LABEL: shuffle_v4f16_2301:
; GFX9: ; %bb.0:
; GFX9-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
-; GFX9-NEXT: global_load_dwordx2 v[0:1], v[0:1], off
-; GFX9-NEXT: v_mov_b32_e32 v2, 0xffff
+; GFX9-NEXT: global_load_dwordx2 v[1:2], v[0:1], off
; GFX9-NEXT: s_waitcnt vmcnt(0)
-; GFX9-NEXT: v_lshrrev_b32_e32 v3, 16, v1
-; GFX9-NEXT: v_and_b32_e32 v1, v2, v1
-; GFX9-NEXT: v_lshrrev_b32_e32 v4, 16, v0
-; GFX9-NEXT: v_and_b32_e32 v2, v2, v0
-; GFX9-NEXT: v_lshl_or_b32 v0, v3, 16, v1
-; GFX9-NEXT: v_lshl_or_b32 v1, v4, 16, v2
+; GFX9-NEXT: v_mov_b32_e32 v0, v2
; GFX9-NEXT: s_setpc_b64 s[30:31]
%val0 = load <4 x half>, <4 x half> addrspace(1)* %arg0
%val1 = load <4 x half>, <4 x half> addrspace(1)* %arg1
; GFX9-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX9-NEXT: global_load_dwordx2 v[0:1], v[0:1], off
; GFX9-NEXT: s_waitcnt vmcnt(0)
-; GFX9-NEXT: v_lshrrev_b32_e32 v0, 16, v1
-; GFX9-NEXT: v_and_b32_e32 v1, 0xffff, v1
-; GFX9-NEXT: v_lshl_or_b32 v0, v0, 16, v1
-; GFX9-NEXT: v_mov_b32_e32 v1, v0
+; GFX9-NEXT: v_mov_b32_e32 v0, v1
; GFX9-NEXT: s_setpc_b64 s[30:31]
%val0 = load <4 x half>, <4 x half> addrspace(1)* %arg0
%val1 = load <4 x half>, <4 x half> addrspace(1)* %arg1
; GFX9-LABEL: shuffle_v4f16_2345:
; GFX9: ; %bb.0:
; GFX9-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
-; GFX9-NEXT: global_load_dwordx2 v[0:1], v[0:1], off
-; GFX9-NEXT: global_load_dwordx2 v[2:3], v[2:3], off
+; GFX9-NEXT: global_load_dwordx2 v[4:5], v[0:1], off
+; GFX9-NEXT: global_load_dwordx2 v[1:2], v[2:3], off
; GFX9-NEXT: s_waitcnt vmcnt(1)
-; GFX9-NEXT: v_mov_b32_e32 v0, 0xffff
+; GFX9-NEXT: v_mov_b32_e32 v0, v5
; GFX9-NEXT: s_waitcnt vmcnt(0)
-; GFX9-NEXT: v_lshrrev_b32_e32 v3, 16, v1
-; GFX9-NEXT: v_lshrrev_b32_e32 v4, 16, v2
-; GFX9-NEXT: v_and_b32_e32 v1, v0, v1
-; GFX9-NEXT: v_and_b32_e32 v2, v0, v2
-; GFX9-NEXT: v_lshl_or_b32 v0, v3, 16, v1
-; GFX9-NEXT: v_lshl_or_b32 v1, v4, 16, v2
; GFX9-NEXT: s_setpc_b64 s[30:31]
%val0 = load <4 x half>, <4 x half> addrspace(1)* %arg0
%val1 = load <4 x half>, <4 x half> addrspace(1)* %arg1
; GFX9-LABEL: shuffle_v4f16_2367:
; GFX9: ; %bb.0:
; GFX9-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
-; GFX9-NEXT: global_load_dwordx2 v[0:1], v[0:1], off
-; GFX9-NEXT: global_load_dwordx2 v[2:3], v[2:3], off
-; GFX9-NEXT: s_waitcnt vmcnt(1)
-; GFX9-NEXT: v_mov_b32_e32 v0, 0xffff
+; GFX9-NEXT: global_load_dwordx2 v[4:5], v[0:1], off
+; GFX9-NEXT: global_load_dwordx2 v[0:1], v[2:3], off
; GFX9-NEXT: s_waitcnt vmcnt(0)
-; GFX9-NEXT: v_lshrrev_b32_e32 v2, 16, v1
-; GFX9-NEXT: v_lshrrev_b32_e32 v4, 16, v3
-; GFX9-NEXT: v_and_b32_e32 v1, v0, v1
-; GFX9-NEXT: v_and_b32_e32 v3, v0, v3
-; GFX9-NEXT: v_lshl_or_b32 v0, v2, 16, v1
-; GFX9-NEXT: v_lshl_or_b32 v1, v4, 16, v3
+; GFX9-NEXT: v_mov_b32_e32 v0, v5
; GFX9-NEXT: s_setpc_b64 s[30:31]
%val0 = load <4 x half>, <4 x half> addrspace(1)* %arg0
%val1 = load <4 x half>, <4 x half> addrspace(1)* %arg1
; GFX9-LABEL: shuffle_v4f16_4501:
; GFX9: ; %bb.0:
; GFX9-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
-; GFX9-NEXT: global_load_dwordx2 v[0:1], v[0:1], off
-; GFX9-NEXT: s_waitcnt vmcnt(0)
-; GFX9-NEXT: global_load_dwordx2 v[1:2], v[2:3], off
+; GFX9-NEXT: global_load_dwordx2 v[4:5], v[0:1], off
+; GFX9-NEXT: global_load_dwordx2 v[0:1], v[2:3], off
; GFX9-NEXT: s_waitcnt vmcnt(0)
-; GFX9-NEXT: v_mov_b32_e32 v2, 0xffff
-; GFX9-NEXT: v_lshrrev_b32_e32 v3, 16, v0
-; GFX9-NEXT: v_lshrrev_b32_e32 v4, 16, v1
-; GFX9-NEXT: v_and_b32_e32 v1, v2, v1
-; GFX9-NEXT: v_and_b32_e32 v2, v2, v0
-; GFX9-NEXT: v_lshl_or_b32 v0, v4, 16, v1
-; GFX9-NEXT: v_lshl_or_b32 v1, v3, 16, v2
+; GFX9-NEXT: v_mov_b32_e32 v1, v4
; GFX9-NEXT: s_setpc_b64 s[30:31]
%val0 = load <4 x half>, <4 x half> addrspace(1)* %arg0
%val1 = load <4 x half>, <4 x half> addrspace(1)* %arg1
; GFX9-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX9-NEXT: global_load_dwordx2 v[0:1], v[0:1], off
; GFX9-NEXT: global_load_dwordx2 v[2:3], v[2:3], off
-; GFX9-NEXT: s_waitcnt vmcnt(1)
-; GFX9-NEXT: v_mov_b32_e32 v0, 0xffff
; GFX9-NEXT: s_waitcnt vmcnt(0)
-; GFX9-NEXT: v_lshrrev_b32_e32 v3, 16, v1
-; GFX9-NEXT: v_and_b32_e32 v1, v0, v1
-; GFX9-NEXT: v_lshrrev_b32_e32 v4, 16, v2
-; GFX9-NEXT: v_and_b32_e32 v2, v0, v2
-; GFX9-NEXT: v_lshl_or_b32 v0, v4, 16, v2
-; GFX9-NEXT: v_lshl_or_b32 v1, v3, 16, v1
+; GFX9-NEXT: v_mov_b32_e32 v0, v2
; GFX9-NEXT: s_setpc_b64 s[30:31]
%val0 = load <4 x half>, <4 x half> addrspace(1)* %arg0
%val1 = load <4 x half>, <4 x half> addrspace(1)* %arg1
; GFX9-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX9-NEXT: global_load_dwordx2 v[0:1], v[2:3], off
; GFX9-NEXT: s_waitcnt vmcnt(0)
-; GFX9-NEXT: v_lshrrev_b32_e32 v1, 16, v0
-; GFX9-NEXT: v_and_b32_e32 v0, 0xffff, v0
-; GFX9-NEXT: v_lshl_or_b32 v0, v1, 16, v0
; GFX9-NEXT: v_mov_b32_e32 v1, v0
; GFX9-NEXT: s_setpc_b64 s[30:31]
%val0 = load <4 x half>, <4 x half> addrspace(1)* %arg0
; GFX9-LABEL: shuffle_v4f16_6701:
; GFX9: ; %bb.0:
; GFX9-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
-; GFX9-NEXT: global_load_dwordx2 v[0:1], v[0:1], off
-; GFX9-NEXT: s_waitcnt vmcnt(0)
-; GFX9-NEXT: global_load_dwordx2 v[1:2], v[2:3], off
+; GFX9-NEXT: global_load_dwordx2 v[4:5], v[0:1], off
+; GFX9-NEXT: global_load_dwordx2 v[0:1], v[2:3], off
; GFX9-NEXT: s_waitcnt vmcnt(0)
-; GFX9-NEXT: v_mov_b32_e32 v1, 0xffff
-; GFX9-NEXT: v_lshrrev_b32_e32 v3, 16, v0
-; GFX9-NEXT: v_lshrrev_b32_e32 v4, 16, v2
-; GFX9-NEXT: v_and_b32_e32 v2, v1, v2
-; GFX9-NEXT: v_and_b32_e32 v1, v1, v0
-; GFX9-NEXT: v_lshl_or_b32 v0, v4, 16, v2
-; GFX9-NEXT: v_lshl_or_b32 v1, v3, 16, v1
+; GFX9-NEXT: v_mov_b32_e32 v0, v1
+; GFX9-NEXT: v_mov_b32_e32 v1, v4
; GFX9-NEXT: s_setpc_b64 s[30:31]
%val0 = load <4 x half>, <4 x half> addrspace(1)* %arg0
%val1 = load <4 x half>, <4 x half> addrspace(1)* %arg1
; GFX9-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX9-NEXT: global_load_dwordx2 v[0:1], v[0:1], off
; GFX9-NEXT: global_load_dwordx2 v[2:3], v[2:3], off
-; GFX9-NEXT: s_waitcnt vmcnt(1)
-; GFX9-NEXT: v_mov_b32_e32 v0, 0xffff
; GFX9-NEXT: s_waitcnt vmcnt(0)
-; GFX9-NEXT: v_lshrrev_b32_e32 v2, 16, v1
-; GFX9-NEXT: v_and_b32_e32 v1, v0, v1
-; GFX9-NEXT: v_lshrrev_b32_e32 v4, 16, v3
-; GFX9-NEXT: v_and_b32_e32 v3, v0, v3
-; GFX9-NEXT: v_lshl_or_b32 v0, v4, 16, v3
-; GFX9-NEXT: v_lshl_or_b32 v1, v2, 16, v1
+; GFX9-NEXT: v_mov_b32_e32 v0, v3
; GFX9-NEXT: s_setpc_b64 s[30:31]
%val0 = load <4 x half>, <4 x half> addrspace(1)* %arg0
%val1 = load <4 x half>, <4 x half> addrspace(1)* %arg1
; GFX9-LABEL: shuffle_v4f16_6745:
; GFX9: ; %bb.0:
; GFX9-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
-; GFX9-NEXT: global_load_dwordx2 v[0:1], v[2:3], off
-; GFX9-NEXT: v_mov_b32_e32 v2, 0xffff
+; GFX9-NEXT: global_load_dwordx2 v[1:2], v[2:3], off
; GFX9-NEXT: s_waitcnt vmcnt(0)
-; GFX9-NEXT: v_lshrrev_b32_e32 v3, 16, v1
-; GFX9-NEXT: v_and_b32_e32 v1, v2, v1
-; GFX9-NEXT: v_lshrrev_b32_e32 v4, 16, v0
-; GFX9-NEXT: v_and_b32_e32 v2, v2, v0
-; GFX9-NEXT: v_lshl_or_b32 v0, v3, 16, v1
-; GFX9-NEXT: v_lshl_or_b32 v1, v4, 16, v2
+; GFX9-NEXT: v_mov_b32_e32 v0, v2
; GFX9-NEXT: s_setpc_b64 s[30:31]
%val0 = load <4 x half>, <4 x half> addrspace(1)* %arg0
%val1 = load <4 x half>, <4 x half> addrspace(1)* %arg1
; GFX9-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX9-NEXT: global_load_dwordx2 v[0:1], v[2:3], off
; GFX9-NEXT: s_waitcnt vmcnt(0)
-; GFX9-NEXT: v_lshrrev_b32_e32 v0, 16, v1
-; GFX9-NEXT: v_and_b32_e32 v1, 0xffff, v1
-; GFX9-NEXT: v_lshl_or_b32 v0, v0, 16, v1
-; GFX9-NEXT: v_mov_b32_e32 v1, v0
+; GFX9-NEXT: v_mov_b32_e32 v0, v1
; GFX9-NEXT: s_setpc_b64 s[30:31]
%val0 = load <4 x half>, <4 x half> addrspace(1)* %arg0
%val1 = load <4 x half>, <4 x half> addrspace(1)* %arg1
; GFX9-LABEL: shuffle_v4f16_2356:
; GFX9: ; %bb.0:
; GFX9-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
-; GFX9-NEXT: global_load_dwordx2 v[0:1], v[0:1], off
-; GFX9-NEXT: global_load_dwordx2 v[2:3], v[2:3], off
-; GFX9-NEXT: s_waitcnt vmcnt(1)
-; GFX9-NEXT: v_mov_b32_e32 v0, 0xffff
-; GFX9-NEXT: v_lshrrev_b32_e32 v4, 16, v1
-; GFX9-NEXT: v_and_b32_e32 v1, v0, v1
+; GFX9-NEXT: global_load_dwordx2 v[4:5], v[0:1], off
+; GFX9-NEXT: global_load_dwordx2 v[0:1], v[2:3], off
+; GFX9-NEXT: v_mov_b32_e32 v2, 0xffff
; GFX9-NEXT: s_waitcnt vmcnt(0)
-; GFX9-NEXT: v_and_b32_sdwa v2, v0, v2 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:WORD_1
-; GFX9-NEXT: v_lshl_or_b32 v0, v4, 16, v1
-; GFX9-NEXT: v_lshl_or_b32 v1, v3, 16, v2
+; GFX9-NEXT: v_and_b32_sdwa v0, v2, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:WORD_1
+; GFX9-NEXT: v_lshl_or_b32 v1, v1, 16, v0
+; GFX9-NEXT: v_mov_b32_e32 v0, v5
; GFX9-NEXT: s_setpc_b64 s[30:31]
%val0 = load <4 x half>, <4 x half> addrspace(1)* %arg0
%val1 = load <4 x half>, <4 x half> addrspace(1)* %arg1
; GFX9-NEXT: global_load_dwordx2 v[2:3], v[2:3], off
; GFX9-NEXT: s_waitcnt vmcnt(1)
; GFX9-NEXT: v_mov_b32_e32 v0, 0xffff
-; GFX9-NEXT: v_lshrrev_b32_e32 v4, 16, v1
-; GFX9-NEXT: v_and_b32_e32 v1, v0, v1
; GFX9-NEXT: s_waitcnt vmcnt(0)
-; GFX9-NEXT: v_and_b32_sdwa v2, v0, v2 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:WORD_1
-; GFX9-NEXT: v_lshl_or_b32 v0, v3, 16, v2
-; GFX9-NEXT: v_lshl_or_b32 v1, v4, 16, v1
+; GFX9-NEXT: v_and_b32_sdwa v0, v0, v2 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:WORD_1
+; GFX9-NEXT: v_lshl_or_b32 v0, v3, 16, v0
; GFX9-NEXT: s_setpc_b64 s[30:31]
%val0 = load <4 x half>, <4 x half> addrspace(1)* %arg0
%val1 = load <4 x half>, <4 x half> addrspace(1)* %arg1
; GFX9-LABEL: shuffle_v4i16_2356:
; GFX9: ; %bb.0:
; GFX9-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
-; GFX9-NEXT: global_load_dwordx2 v[0:1], v[0:1], off
-; GFX9-NEXT: global_load_dwordx2 v[2:3], v[2:3], off
-; GFX9-NEXT: v_mov_b32_e32 v4, 0xffff
-; GFX9-NEXT: s_waitcnt vmcnt(1)
-; GFX9-NEXT: v_bfi_b32 v0, v4, v1, v1
+; GFX9-NEXT: global_load_dwordx2 v[4:5], v[0:1], off
+; GFX9-NEXT: global_load_dwordx2 v[0:1], v[2:3], off
+; GFX9-NEXT: v_mov_b32_e32 v2, 0xffff
; GFX9-NEXT: s_waitcnt vmcnt(0)
-; GFX9-NEXT: v_and_b32_sdwa v1, v4, v2 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:WORD_1
-; GFX9-NEXT: v_lshl_or_b32 v1, v3, 16, v1
+; GFX9-NEXT: v_and_b32_sdwa v0, v2, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:WORD_1
+; GFX9-NEXT: v_lshl_or_b32 v1, v1, 16, v0
+; GFX9-NEXT: v_mov_b32_e32 v0, v5
; GFX9-NEXT: s_setpc_b64 s[30:31]
%val0 = load <4 x i16>, <4 x i16> addrspace(1)* %arg0
%val1 = load <4 x i16>, <4 x i16> addrspace(1)* %arg1
; GFX9-NEXT: s_waitcnt vmcnt(0)
; GFX9-NEXT: global_load_dwordx2 v[1:2], v[2:3], off
; GFX9-NEXT: s_waitcnt vmcnt(0)
-; GFX9-NEXT: v_mov_b32_e32 v1, 0xffff
-; GFX9-NEXT: v_bfi_b32 v0, v1, v0, v0
-; GFX9-NEXT: v_bfi_b32 v1, v1, v2, v2
+; GFX9-NEXT: v_mov_b32_e32 v1, v2
; GFX9-NEXT: s_setpc_b64 s[30:31]
%val0 = load <4 x i16>, <4 x i16> addrspace(1)* %arg0
%val1 = load <4 x i16>, <4 x i16> addrspace(1)* %arg1
; GFX9-LABEL: shuffle_v4f16_2333:
; GFX9: ; %bb.0:
; GFX9-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
-; GFX9-NEXT: global_load_dwordx2 v[0:1], v[0:1], off
-; GFX9-NEXT: v_mov_b32_e32 v2, 0xffff
+; GFX9-NEXT: global_load_dwordx2 v[1:2], v[0:1], off
; GFX9-NEXT: s_waitcnt vmcnt(0)
-; GFX9-NEXT: v_lshrrev_b32_e32 v3, 16, v1
-; GFX9-NEXT: v_and_b32_e32 v0, v2, v1
-; GFX9-NEXT: v_and_b32_e32 v1, v2, v3
-; GFX9-NEXT: v_lshl_or_b32 v0, v3, 16, v0
-; GFX9-NEXT: v_lshl_or_b32 v1, v3, 16, v1
+; GFX9-NEXT: v_lshrrev_b32_e32 v0, 16, v2
+; GFX9-NEXT: v_and_b32_e32 v1, 0xffff, v0
+; GFX9-NEXT: v_lshl_or_b32 v1, v0, 16, v1
+; GFX9-NEXT: v_mov_b32_e32 v0, v2
; GFX9-NEXT: s_setpc_b64 s[30:31]
%val0 = load <4 x half>, <4 x half> addrspace(1)* %arg0
%val1 = load <4 x half>, <4 x half> addrspace(1)* %arg1
; GFX9-LABEL: shuffle_v4f16_6667:
; GFX9: ; %bb.0:
; GFX9-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
-; GFX9-NEXT: global_load_dwordx2 v[0:1], v[0:1], off
-; GFX9-NEXT: v_mov_b32_e32 v2, 0xffff
+; GFX9-NEXT: global_load_dwordx2 v[1:2], v[0:1], off
; GFX9-NEXT: s_waitcnt vmcnt(0)
-; GFX9-NEXT: v_lshrrev_b32_e32 v3, 16, v1
-; GFX9-NEXT: v_and_b32_e32 v0, v2, v1
-; GFX9-NEXT: v_and_b32_e32 v1, v2, v3
-; GFX9-NEXT: v_lshl_or_b32 v0, v3, 16, v0
-; GFX9-NEXT: v_lshl_or_b32 v1, v3, 16, v1
+; GFX9-NEXT: v_lshrrev_b32_e32 v0, 16, v2
+; GFX9-NEXT: v_and_b32_e32 v1, 0xffff, v0
+; GFX9-NEXT: v_lshl_or_b32 v1, v0, 16, v1
+; GFX9-NEXT: v_mov_b32_e32 v0, v2
; GFX9-NEXT: s_setpc_b64 s[30:31]
%val0 = load <4 x half>, <4 x half> addrspace(1)* %arg0
%val1 = load <4 x half>, <4 x half> addrspace(1)* %arg1
; GFX9-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX9-NEXT: global_load_dword v0, v[0:1], off
; GFX9-NEXT: s_waitcnt vmcnt(0)
-; GFX9-NEXT: v_lshrrev_b32_e32 v1, 16, v0
-; GFX9-NEXT: v_and_b32_e32 v0, 0xffff, v0
-; GFX9-NEXT: v_lshl_or_b32 v0, v1, 16, v0
; GFX9-NEXT: v_mov_b32_e32 v1, v0
; GFX9-NEXT: s_setpc_b64 s[30:31]
%val0 = load <8 x half>, <8 x half> addrspace(1)* %arg0
; GFX9-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX9-NEXT: global_load_dword v0, v[0:1], off offset:8
; GFX9-NEXT: global_load_dword v1, v[2:3], off
-; GFX9-NEXT: v_mov_b32_e32 v2, 0xffff
-; GFX9-NEXT: s_waitcnt vmcnt(1)
-; GFX9-NEXT: v_lshrrev_b32_e32 v3, 16, v0
-; GFX9-NEXT: v_and_b32_e32 v0, v2, v0
; GFX9-NEXT: s_waitcnt vmcnt(0)
-; GFX9-NEXT: v_lshrrev_b32_e32 v4, 16, v1
-; GFX9-NEXT: v_and_b32_e32 v1, v2, v1
-; GFX9-NEXT: v_lshl_or_b32 v0, v3, 16, v0
-; GFX9-NEXT: v_lshl_or_b32 v1, v4, 16, v1
; GFX9-NEXT: s_setpc_b64 s[30:31]
%val0 = load <8 x half>, <8 x half> addrspace(1)* %arg0
%val1 = load <8 x half>, <8 x half> addrspace(1)* %arg1
; GFX9-LABEL: shuffle_v8f16_10_11_2_3:
; GFX9: ; %bb.0:
; GFX9-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
-; GFX9-NEXT: global_load_dword v0, v[0:1], off offset:4
-; GFX9-NEXT: global_load_dword v1, v[2:3], off offset:4
-; GFX9-NEXT: v_mov_b32_e32 v2, 0xffff
-; GFX9-NEXT: s_waitcnt vmcnt(1)
-; GFX9-NEXT: v_lshrrev_b32_e32 v3, 16, v0
+; GFX9-NEXT: global_load_dword v1, v[0:1], off offset:4
+; GFX9-NEXT: global_load_dword v0, v[2:3], off offset:4
; GFX9-NEXT: s_waitcnt vmcnt(0)
-; GFX9-NEXT: v_lshrrev_b32_e32 v4, 16, v1
-; GFX9-NEXT: v_and_b32_e32 v1, v2, v1
-; GFX9-NEXT: v_and_b32_e32 v2, v2, v0
-; GFX9-NEXT: v_lshl_or_b32 v0, v4, 16, v1
-; GFX9-NEXT: v_lshl_or_b32 v1, v3, 16, v2
; GFX9-NEXT: s_setpc_b64 s[30:31]
%val0 = load <8 x half>, <8 x half> addrspace(1)* %arg0
%val1 = load <8 x half>, <8 x half> addrspace(1)* %arg1
; GFX9-LABEL: shuffle_v8f16_13_14_2_3:
; GFX9: ; %bb.0:
; GFX9-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
-; GFX9-NEXT: global_load_dword v4, v[0:1], off offset:4
-; GFX9-NEXT: global_load_dwordx4 v[0:3], v[2:3], off
-; GFX9-NEXT: s_waitcnt vmcnt(0)
+; GFX9-NEXT: global_load_dword v1, v[0:1], off offset:4
+; GFX9-NEXT: global_load_dwordx4 v[2:5], v[2:3], off
; GFX9-NEXT: v_mov_b32_e32 v0, 0xffff
-; GFX9-NEXT: v_lshrrev_b32_e32 v1, 16, v4
-; GFX9-NEXT: v_and_b32_e32 v4, v0, v4
-; GFX9-NEXT: v_and_b32_sdwa v0, v0, v2 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:WORD_1
-; GFX9-NEXT: v_lshl_or_b32 v1, v1, 16, v4
-; GFX9-NEXT: v_lshl_or_b32 v0, v3, 16, v0
+; GFX9-NEXT: s_waitcnt vmcnt(0)
+; GFX9-NEXT: v_and_b32_sdwa v0, v0, v4 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:WORD_1
+; GFX9-NEXT: v_lshl_or_b32 v0, v5, 16, v0
; GFX9-NEXT: s_setpc_b64 s[30:31]
%val0 = load <8 x half>, <8 x half> addrspace(1)* %arg0
%val1 = load <8 x half>, <8 x half> addrspace(1)* %arg1
; GFX9: ; %bb.0:
; GFX9-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX9-NEXT: global_load_dwordx2 v[0:1], v[0:1], off
-; GFX9-NEXT: v_mov_b32_e32 v2, 0xffff
; GFX9-NEXT: s_waitcnt vmcnt(0)
-; GFX9-NEXT: v_lshrrev_b32_e32 v3, 16, v0
-; GFX9-NEXT: v_and_b32_e32 v0, v2, v0
-; GFX9-NEXT: v_and_b32_e32 v2, v2, v1
-; GFX9-NEXT: v_lshl_or_b32 v0, v3, 16, v0
+; GFX9-NEXT: v_and_b32_e32 v2, 0xffff, v1
; GFX9-NEXT: v_lshl_or_b32 v1, v1, 16, v2
; GFX9-NEXT: s_setpc_b64 s[30:31]
%val0 = load <3 x half>, <3 x half> addrspace(1)* %arg0
; GFX9-LABEL: shuffle_v6f16_452367:
; GFX9: ; %bb.0:
; GFX9-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
-; GFX9-NEXT: global_load_dwordx3 v[4:6], v[0:1], off
-; GFX9-NEXT: global_load_dword v2, v[2:3], off
-; GFX9-NEXT: v_mov_b32_e32 v0, 0xffff
+; GFX9-NEXT: v_mov_b32_e32 v4, v3
+; GFX9-NEXT: v_mov_b32_e32 v3, v2
+; GFX9-NEXT: global_load_dwordx3 v[0:2], v[0:1], off
+; GFX9-NEXT: global_load_dword v3, v[3:4], off
; GFX9-NEXT: s_waitcnt vmcnt(1)
-; GFX9-NEXT: v_lshrrev_b32_e32 v4, 16, v5
-; GFX9-NEXT: v_lshrrev_b32_e32 v1, 16, v6
-; GFX9-NEXT: v_and_b32_e32 v3, v0, v6
-; GFX9-NEXT: v_and_b32_e32 v5, v0, v5
-; GFX9-NEXT: v_lshl_or_b32 v0, v1, 16, v3
-; GFX9-NEXT: v_lshl_or_b32 v1, v4, 16, v5
+; GFX9-NEXT: v_mov_b32_e32 v0, v2
; GFX9-NEXT: s_waitcnt vmcnt(0)
+; GFX9-NEXT: v_mov_b32_e32 v2, v3
; GFX9-NEXT: s_setpc_b64 s[30:31]
%val0 = load <6 x half>, <6 x half> addrspace(1)* %arg0
%val1 = load <6 x half>, <6 x half> addrspace(1)* %arg1
; GFX9-NEXT: v_pk_fma_f16 v2, v1, v2, v7 op_sel_hi:[0,1,1]
; GFX9-NEXT: v_pk_fma_f16 v0, v0, v3, v6 op_sel:[1,0,0]
; GFX9-NEXT: v_pk_fma_f16 v1, v1, v3, v2 op_sel:[1,0,0]
-; GFX9-NEXT: v_lshrrev_b32_e32 v2, 16, v0
-; GFX9-NEXT: v_and_b32_e32 v0, 0xffff, v0
-; GFX9-NEXT: v_lshl_or_b32 v0, v2, 16, v0
; GFX9-NEXT: global_store_dwordx2 v[4:5], v[0:1], off
; GFX9-NEXT: s_endpgm
entry:
ret void
}
+define <4 x half> @shuffle_v4f16_0456(<4 x half> addrspace(1)* %arg0, <4 x half> addrspace(1)* %arg1) {
+; GFX9-LABEL: shuffle_v4f16_0456:
+; GFX9: ; %bb.0:
+; GFX9-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX9-NEXT: global_load_dwordx2 v[0:1], v[0:1], off
+; GFX9-NEXT: s_waitcnt vmcnt(0)
+; GFX9-NEXT: global_load_dwordx2 v[1:2], v[2:3], off
+; GFX9-NEXT: v_mov_b32_e32 v3, 0xffff
+; GFX9-NEXT: v_and_b32_e32 v0, v3, v0
+; GFX9-NEXT: s_waitcnt vmcnt(0)
+; GFX9-NEXT: v_and_b32_sdwa v3, v3, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:WORD_1
+; GFX9-NEXT: v_lshl_or_b32 v0, v1, 16, v0
+; GFX9-NEXT: v_lshl_or_b32 v1, v2, 16, v3
+; GFX9-NEXT: s_setpc_b64 s[30:31]
+ %val0 = load <4 x half>, <4 x half> addrspace(1)* %arg0
+ %val1 = load <4 x half>, <4 x half> addrspace(1)* %arg1
+ %shuffle = shufflevector <4 x half> %val0, <4 x half> %val1, <4 x i32> <i32 0, i32 4, i32 5, i32 6>
+ ret <4 x half> %shuffle
+}
+
declare <2 x half> @llvm.fma.v2f16(<2 x half>, <2 x half>, <2 x half>) #0
declare i32 @llvm.amdgcn.workitem.id.x() #0