caio/clang-p2996
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity
Files
d661b4b57504b965a37dc30b79bdd5ac36fca9ec
clang-p2996/llvm/test/CodeGen/AMDGPU/GlobalISel/extractelement.i8.ll
Jay Foad f2c164c815 [AMDGPU] Do not wait for vscnt on function entry and return 
SIInsertWaitcnts inserts waitcnt instructions to resolve data
dependencies. The GFX10+ vscnt (VMEM store count) counter is never used
in this way. It is only used to resolve memory dependencies, and that is
handled by SIMemoryLegalizer. Hence there is no need to conservatively
wait for vscnt to be 0 on function entry and before returns.

Differential Revision: https://reviews.llvm.org/D153537
2023-07-04 12:22:38 +01:00

2404 lines
90 KiB
LLVM
Raw Blame History

; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
; RUN: llc -global-isel -mtriple=amdgcn-mesa-mesa3d -mcpu=gfx900 -verify-machineinstrs < %s | FileCheck -check-prefixes=GCN,GFX9 %s
; RUN: llc -global-isel -mtriple=amdgcn-mesa-mesa3d -mcpu=fiji -verify-machineinstrs < %s | FileCheck -check-prefixes=GCN,GFX8 %s
; RUN: llc -global-isel -mtriple=amdgcn-mesa-mesa3d -mcpu=hawaii -verify-machineinstrs < %s | FileCheck -check-prefixes=GCN,GFX7 %s
; RUN: llc -global-isel -mtriple=amdgcn-mesa-mesa3d -mcpu=gfx1010 -verify-machineinstrs < %s | FileCheck -check-prefixes=GFX10 %s
; RUN: llc -global-isel -mtriple=amdgcn-mesa-mesa3d -mcpu=gfx1100 -verify-machineinstrs < %s | FileCheck -check-prefixes=GFX11 %s
define amdgpu_ps i8 @extractelement_sgpr_v4i8_sgpr_idx(ptr addrspace(4) inreg %ptr, i32 inreg %idx) {
; GCN-LABEL: extractelement_sgpr_v4i8_sgpr_idx:
; GCN: ; %bb.0:
; GCN-NEXT: s_load_dword s0, s[2:3], 0x0
; GCN-NEXT: s_and_b32 s1, s4, 3
; GCN-NEXT: s_lshl_b32 s1, s1, 3
; GCN-NEXT: s_waitcnt lgkmcnt(0)
; GCN-NEXT: s_lshr_b32 s0, s0, s1
; GCN-NEXT: ; return to shader part epilog
;
; GFX10-LABEL: extractelement_sgpr_v4i8_sgpr_idx:
; GFX10: ; %bb.0:
; GFX10-NEXT: s_load_dword s0, s[2:3], 0x0
; GFX10-NEXT: s_and_b32 s1, s4, 3
; GFX10-NEXT: s_lshl_b32 s1, s1, 3
; GFX10-NEXT: s_waitcnt lgkmcnt(0)
; GFX10-NEXT: s_lshr_b32 s0, s0, s1
; GFX10-NEXT: ; return to shader part epilog
;
; GFX11-LABEL: extractelement_sgpr_v4i8_sgpr_idx:
; GFX11: ; %bb.0:
; GFX11-NEXT: s_load_b32 s0, s[2:3], 0x0
; GFX11-NEXT: s_and_b32 s1, s4, 3
; GFX11-NEXT: s_delay_alu instid0(SALU_CYCLE_1)
; GFX11-NEXT: s_lshl_b32 s1, s1, 3
; GFX11-NEXT: s_waitcnt lgkmcnt(0)
; GFX11-NEXT: s_lshr_b32 s0, s0, s1
; GFX11-NEXT: ; return to shader part epilog
%vector = load <4 x i8>, ptr addrspace(4) %ptr
%element = extractelement <4 x i8> %vector, i32 %idx
ret i8 %element
}
define amdgpu_ps i8 @extractelement_vgpr_v4i8_sgpr_idx(ptr addrspace(1) %ptr, i32 inreg %idx) {
; GFX9-LABEL: extractelement_vgpr_v4i8_sgpr_idx:
; GFX9: ; %bb.0:
; GFX9-NEXT: global_load_dword v0, v[0:1], off
; GFX9-NEXT: s_and_b32 s0, s2, 3
; GFX9-NEXT: s_lshl_b32 s0, s0, 3
; GFX9-NEXT: s_waitcnt vmcnt(0)
; GFX9-NEXT: v_lshrrev_b32_e32 v0, s0, v0
; GFX9-NEXT: v_readfirstlane_b32 s0, v0
; GFX9-NEXT: ; return to shader part epilog
;
; GFX8-LABEL: extractelement_vgpr_v4i8_sgpr_idx:
; GFX8: ; %bb.0:
; GFX8-NEXT: flat_load_dword v0, v[0:1]
; GFX8-NEXT: s_and_b32 s0, s2, 3
; GFX8-NEXT: s_lshl_b32 s0, s0, 3
; GFX8-NEXT: s_waitcnt vmcnt(0)
; GFX8-NEXT: v_lshrrev_b32_e32 v0, s0, v0
; GFX8-NEXT: v_readfirstlane_b32 s0, v0
; GFX8-NEXT: ; return to shader part epilog
;
; GFX7-LABEL: extractelement_vgpr_v4i8_sgpr_idx:
; GFX7: ; %bb.0:
; GFX7-NEXT: s_mov_b32 s6, 0
; GFX7-NEXT: s_mov_b32 s7, 0xf000
; GFX7-NEXT: s_mov_b64 s[4:5], 0
; GFX7-NEXT: buffer_load_dword v0, v[0:1], s[4:7], 0 addr64
; GFX7-NEXT: s_and_b32 s0, s2, 3
; GFX7-NEXT: s_lshl_b32 s0, s0, 3
; GFX7-NEXT: s_waitcnt vmcnt(0)
; GFX7-NEXT: v_lshrrev_b32_e32 v0, s0, v0
; GFX7-NEXT: v_readfirstlane_b32 s0, v0
; GFX7-NEXT: ; return to shader part epilog
;
; GFX10-LABEL: extractelement_vgpr_v4i8_sgpr_idx:
; GFX10: ; %bb.0:
; GFX10-NEXT: global_load_dword v0, v[0:1], off
; GFX10-NEXT: s_and_b32 s0, s2, 3
; GFX10-NEXT: s_lshl_b32 s0, s0, 3
; GFX10-NEXT: s_waitcnt vmcnt(0)
; GFX10-NEXT: v_lshrrev_b32_e32 v0, s0, v0
; GFX10-NEXT: v_readfirstlane_b32 s0, v0
; GFX10-NEXT: ; return to shader part epilog
;
; GFX11-LABEL: extractelement_vgpr_v4i8_sgpr_idx:
; GFX11: ; %bb.0:
; GFX11-NEXT: global_load_b32 v0, v[0:1], off
; GFX11-NEXT: s_and_b32 s0, s2, 3
; GFX11-NEXT: s_delay_alu instid0(SALU_CYCLE_1) | instskip(SKIP_2) | instid1(VALU_DEP_1)
; GFX11-NEXT: s_lshl_b32 s0, s0, 3
; GFX11-NEXT: s_waitcnt vmcnt(0)
; GFX11-NEXT: v_lshrrev_b32_e32 v0, s0, v0
; GFX11-NEXT: v_readfirstlane_b32 s0, v0
; GFX11-NEXT: ; return to shader part epilog
%vector = load <4 x i8>, ptr addrspace(1) %ptr
%element = extractelement <4 x i8> %vector, i32 %idx
ret i8 %element
}
define i8 @extractelement_vgpr_v4i8_vgpr_idx(ptr addrspace(1) %ptr, i32 %idx) {
; GFX9-LABEL: extractelement_vgpr_v4i8_vgpr_idx:
; GFX9: ; %bb.0:
; GFX9-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX9-NEXT: global_load_dword v0, v[0:1], off
; GFX9-NEXT: v_and_b32_e32 v1, 3, v2
; GFX9-NEXT: v_lshlrev_b32_e32 v1, 3, v1
; GFX9-NEXT: s_waitcnt vmcnt(0)
; GFX9-NEXT: v_lshrrev_b32_e32 v0, v1, v0
; GFX9-NEXT: s_setpc_b64 s[30:31]
;
; GFX8-LABEL: extractelement_vgpr_v4i8_vgpr_idx:
; GFX8: ; %bb.0:
; GFX8-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX8-NEXT: flat_load_dword v0, v[0:1]
; GFX8-NEXT: v_and_b32_e32 v1, 3, v2
; GFX8-NEXT: v_lshlrev_b32_e32 v1, 3, v1
; GFX8-NEXT: s_waitcnt vmcnt(0)
; GFX8-NEXT: v_lshrrev_b32_e32 v0, v1, v0
; GFX8-NEXT: s_setpc_b64 s[30:31]
;
; GFX7-LABEL: extractelement_vgpr_v4i8_vgpr_idx:
; GFX7: ; %bb.0:
; GFX7-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX7-NEXT: s_mov_b32 s6, 0
; GFX7-NEXT: s_mov_b32 s7, 0xf000
; GFX7-NEXT: s_mov_b64 s[4:5], 0
; GFX7-NEXT: buffer_load_dword v0, v[0:1], s[4:7], 0 addr64
; GFX7-NEXT: v_and_b32_e32 v1, 3, v2
; GFX7-NEXT: v_lshlrev_b32_e32 v1, 3, v1
; GFX7-NEXT: s_waitcnt vmcnt(0)
; GFX7-NEXT: v_lshrrev_b32_e32 v0, v1, v0
; GFX7-NEXT: s_setpc_b64 s[30:31]
;
; GFX10-LABEL: extractelement_vgpr_v4i8_vgpr_idx:
; GFX10: ; %bb.0:
; GFX10-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX10-NEXT: global_load_dword v0, v[0:1], off
; GFX10-NEXT: v_and_b32_e32 v1, 3, v2
; GFX10-NEXT: v_lshlrev_b32_e32 v1, 3, v1
; GFX10-NEXT: s_waitcnt vmcnt(0)
; GFX10-NEXT: v_lshrrev_b32_e32 v0, v1, v0
; GFX10-NEXT: s_setpc_b64 s[30:31]
;
; GFX11-LABEL: extractelement_vgpr_v4i8_vgpr_idx:
; GFX11: ; %bb.0:
; GFX11-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX11-NEXT: global_load_b32 v0, v[0:1], off
; GFX11-NEXT: v_and_b32_e32 v1, 3, v2
; GFX11-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_1) | instid1(VALU_DEP_1)
; GFX11-NEXT: v_lshlrev_b32_e32 v1, 3, v1
; GFX11-NEXT: s_waitcnt vmcnt(0)
; GFX11-NEXT: v_lshrrev_b32_e32 v0, v1, v0
; GFX11-NEXT: s_setpc_b64 s[30:31]
%vector = load <4 x i8>, ptr addrspace(1) %ptr
%element = extractelement <4 x i8> %vector, i32 %idx
ret i8 %element
}
define amdgpu_ps i8 @extractelement_sgpr_v4i8_vgpr_idx(ptr addrspace(4) inreg %ptr, i32 %idx) {
; GFX9-LABEL: extractelement_sgpr_v4i8_vgpr_idx:
; GFX9: ; %bb.0:
; GFX9-NEXT: s_load_dword s0, s[2:3], 0x0
; GFX9-NEXT: v_and_b32_e32 v0, 3, v0
; GFX9-NEXT: v_lshlrev_b32_e32 v0, 3, v0
; GFX9-NEXT: s_waitcnt lgkmcnt(0)
; GFX9-NEXT: v_lshrrev_b32_e64 v0, v0, s0
; GFX9-NEXT: v_readfirstlane_b32 s0, v0
; GFX9-NEXT: ; return to shader part epilog
;
; GFX8-LABEL: extractelement_sgpr_v4i8_vgpr_idx:
; GFX8: ; %bb.0:
; GFX8-NEXT: s_load_dword s0, s[2:3], 0x0
; GFX8-NEXT: v_and_b32_e32 v0, 3, v0
; GFX8-NEXT: v_lshlrev_b32_e32 v0, 3, v0
; GFX8-NEXT: s_waitcnt lgkmcnt(0)
; GFX8-NEXT: v_lshrrev_b32_e64 v0, v0, s0
; GFX8-NEXT: v_readfirstlane_b32 s0, v0
; GFX8-NEXT: ; return to shader part epilog
;
; GFX7-LABEL: extractelement_sgpr_v4i8_vgpr_idx:
; GFX7: ; %bb.0:
; GFX7-NEXT: s_load_dword s0, s[2:3], 0x0
; GFX7-NEXT: v_and_b32_e32 v0, 3, v0
; GFX7-NEXT: v_lshlrev_b32_e32 v0, 3, v0
; GFX7-NEXT: s_waitcnt lgkmcnt(0)
; GFX7-NEXT: v_lshr_b32_e32 v0, s0, v0
; GFX7-NEXT: v_readfirstlane_b32 s0, v0
; GFX7-NEXT: ; return to shader part epilog
;
; GFX10-LABEL: extractelement_sgpr_v4i8_vgpr_idx:
; GFX10: ; %bb.0:
; GFX10-NEXT: s_load_dword s0, s[2:3], 0x0
; GFX10-NEXT: v_and_b32_e32 v0, 3, v0
; GFX10-NEXT: v_lshlrev_b32_e32 v0, 3, v0
; GFX10-NEXT: s_waitcnt lgkmcnt(0)
; GFX10-NEXT: v_lshrrev_b32_e64 v0, v0, s0
; GFX10-NEXT: v_readfirstlane_b32 s0, v0
; GFX10-NEXT: ; return to shader part epilog
;
; GFX11-LABEL: extractelement_sgpr_v4i8_vgpr_idx:
; GFX11: ; %bb.0:
; GFX11-NEXT: s_load_b32 s0, s[2:3], 0x0
; GFX11-NEXT: v_and_b32_e32 v0, 3, v0
; GFX11-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_1) | instid1(VALU_DEP_1)
; GFX11-NEXT: v_lshlrev_b32_e32 v0, 3, v0
; GFX11-NEXT: s_waitcnt lgkmcnt(0)
; GFX11-NEXT: v_lshrrev_b32_e64 v0, v0, s0
; GFX11-NEXT: s_delay_alu instid0(VALU_DEP_1)
; GFX11-NEXT: v_readfirstlane_b32 s0, v0
; GFX11-NEXT: ; return to shader part epilog
%vector = load <4 x i8>, ptr addrspace(4) %ptr
%element = extractelement <4 x i8> %vector, i32 %idx
ret i8 %element
}
define amdgpu_ps i8 @extractelement_sgpr_v4i8_idx0(ptr addrspace(4) inreg %ptr) {
; GCN-LABEL: extractelement_sgpr_v4i8_idx0:
; GCN: ; %bb.0:
; GCN-NEXT: s_load_dword s0, s[2:3], 0x0
; GCN-NEXT: s_waitcnt lgkmcnt(0)
; GCN-NEXT: ; return to shader part epilog
;
; GFX10-LABEL: extractelement_sgpr_v4i8_idx0:
; GFX10: ; %bb.0:
; GFX10-NEXT: s_load_dword s0, s[2:3], 0x0
; GFX10-NEXT: s_waitcnt lgkmcnt(0)
; GFX10-NEXT: ; return to shader part epilog
;
; GFX11-LABEL: extractelement_sgpr_v4i8_idx0:
; GFX11: ; %bb.0:
; GFX11-NEXT: s_load_b32 s0, s[2:3], 0x0
; GFX11-NEXT: s_waitcnt lgkmcnt(0)
; GFX11-NEXT: ; return to shader part epilog
%vector = load <4 x i8>, ptr addrspace(4) %ptr
%element = extractelement <4 x i8> %vector, i32 0
ret i8 %element
}
define amdgpu_ps i8 @extractelement_sgpr_v4i8_idx1(ptr addrspace(4) inreg %ptr) {
; GCN-LABEL: extractelement_sgpr_v4i8_idx1:
; GCN: ; %bb.0:
; GCN-NEXT: s_load_dword s0, s[2:3], 0x0
; GCN-NEXT: s_waitcnt lgkmcnt(0)
; GCN-NEXT: s_lshr_b32 s0, s0, 8
; GCN-NEXT: ; return to shader part epilog
;
; GFX10-LABEL: extractelement_sgpr_v4i8_idx1:
; GFX10: ; %bb.0:
; GFX10-NEXT: s_load_dword s0, s[2:3], 0x0
; GFX10-NEXT: s_waitcnt lgkmcnt(0)
; GFX10-NEXT: s_lshr_b32 s0, s0, 8
; GFX10-NEXT: ; return to shader part epilog
;
; GFX11-LABEL: extractelement_sgpr_v4i8_idx1:
; GFX11: ; %bb.0:
; GFX11-NEXT: s_load_b32 s0, s[2:3], 0x0
; GFX11-NEXT: s_waitcnt lgkmcnt(0)
; GFX11-NEXT: s_lshr_b32 s0, s0, 8
; GFX11-NEXT: ; return to shader part epilog
%vector = load <4 x i8>, ptr addrspace(4) %ptr
%element = extractelement <4 x i8> %vector, i32 1
ret i8 %element
}
define amdgpu_ps i8 @extractelement_sgpr_v4i8_idx2(ptr addrspace(4) inreg %ptr) {
; GCN-LABEL: extractelement_sgpr_v4i8_idx2:
; GCN: ; %bb.0:
; GCN-NEXT: s_load_dword s0, s[2:3], 0x0
; GCN-NEXT: s_waitcnt lgkmcnt(0)
; GCN-NEXT: s_lshr_b32 s0, s0, 16
; GCN-NEXT: ; return to shader part epilog
;
; GFX10-LABEL: extractelement_sgpr_v4i8_idx2:
; GFX10: ; %bb.0:
; GFX10-NEXT: s_load_dword s0, s[2:3], 0x0
; GFX10-NEXT: s_waitcnt lgkmcnt(0)
; GFX10-NEXT: s_lshr_b32 s0, s0, 16
; GFX10-NEXT: ; return to shader part epilog
;
; GFX11-LABEL: extractelement_sgpr_v4i8_idx2:
; GFX11: ; %bb.0:
; GFX11-NEXT: s_load_b32 s0, s[2:3], 0x0
; GFX11-NEXT: s_waitcnt lgkmcnt(0)
; GFX11-NEXT: s_lshr_b32 s0, s0, 16
; GFX11-NEXT: ; return to shader part epilog
%vector = load <4 x i8>, ptr addrspace(4) %ptr
%element = extractelement <4 x i8> %vector, i32 2
ret i8 %element
}
define amdgpu_ps i8 @extractelement_sgpr_v4i8_idx3(ptr addrspace(4) inreg %ptr) {
; GCN-LABEL: extractelement_sgpr_v4i8_idx3:
; GCN: ; %bb.0:
; GCN-NEXT: s_load_dword s0, s[2:3], 0x0
; GCN-NEXT: s_waitcnt lgkmcnt(0)
; GCN-NEXT: s_lshr_b32 s0, s0, 24
; GCN-NEXT: ; return to shader part epilog
;
; GFX10-LABEL: extractelement_sgpr_v4i8_idx3:
; GFX10: ; %bb.0:
; GFX10-NEXT: s_load_dword s0, s[2:3], 0x0
; GFX10-NEXT: s_waitcnt lgkmcnt(0)
; GFX10-NEXT: s_lshr_b32 s0, s0, 24
; GFX10-NEXT: ; return to shader part epilog
;
; GFX11-LABEL: extractelement_sgpr_v4i8_idx3:
; GFX11: ; %bb.0:
; GFX11-NEXT: s_load_b32 s0, s[2:3], 0x0
; GFX11-NEXT: s_waitcnt lgkmcnt(0)
; GFX11-NEXT: s_lshr_b32 s0, s0, 24
; GFX11-NEXT: ; return to shader part epilog
%vector = load <4 x i8>, ptr addrspace(4) %ptr
%element = extractelement <4 x i8> %vector, i32 3
ret i8 %element
}
define i8 @extractelement_vgpr_v4i8_idx0(ptr addrspace(1) %ptr) {
; GFX9-LABEL: extractelement_vgpr_v4i8_idx0:
; GFX9: ; %bb.0:
; 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: s_setpc_b64 s[30:31]
;
; GFX8-LABEL: extractelement_vgpr_v4i8_idx0:
; GFX8: ; %bb.0:
; GFX8-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX8-NEXT: flat_load_dword v0, v[0:1]
; GFX8-NEXT: s_waitcnt vmcnt(0)
; GFX8-NEXT: s_setpc_b64 s[30:31]
;
; GFX7-LABEL: extractelement_vgpr_v4i8_idx0:
; GFX7: ; %bb.0:
; GFX7-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX7-NEXT: s_mov_b32 s6, 0
; GFX7-NEXT: s_mov_b32 s7, 0xf000
; GFX7-NEXT: s_mov_b64 s[4:5], 0
; GFX7-NEXT: buffer_load_dword v0, v[0:1], s[4:7], 0 addr64
; GFX7-NEXT: s_waitcnt vmcnt(0)
; GFX7-NEXT: s_setpc_b64 s[30:31]
;
; GFX10-LABEL: extractelement_vgpr_v4i8_idx0:
; GFX10: ; %bb.0:
; GFX10-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX10-NEXT: global_load_dword v0, v[0:1], off
; GFX10-NEXT: s_waitcnt vmcnt(0)
; GFX10-NEXT: s_setpc_b64 s[30:31]
;
; GFX11-LABEL: extractelement_vgpr_v4i8_idx0:
; GFX11: ; %bb.0:
; GFX11-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX11-NEXT: global_load_b32 v0, v[0:1], off
; GFX11-NEXT: s_waitcnt vmcnt(0)
; GFX11-NEXT: s_setpc_b64 s[30:31]
%vector = load <4 x i8>, ptr addrspace(1) %ptr
%element = extractelement <4 x i8> %vector, i32 0
ret i8 %element
}
define i8 @extractelement_vgpr_v4i8_idx1(ptr addrspace(1) %ptr) {
; GFX9-LABEL: extractelement_vgpr_v4i8_idx1:
; GFX9: ; %bb.0:
; 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 v0, 8, v0
; GFX9-NEXT: s_setpc_b64 s[30:31]
;
; GFX8-LABEL: extractelement_vgpr_v4i8_idx1:
; GFX8: ; %bb.0:
; GFX8-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX8-NEXT: flat_load_dword v0, v[0:1]
; GFX8-NEXT: s_waitcnt vmcnt(0)
; GFX8-NEXT: v_lshrrev_b32_e32 v0, 8, v0
; GFX8-NEXT: s_setpc_b64 s[30:31]
;
; GFX7-LABEL: extractelement_vgpr_v4i8_idx1:
; GFX7: ; %bb.0:
; GFX7-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX7-NEXT: s_mov_b32 s6, 0
; GFX7-NEXT: s_mov_b32 s7, 0xf000
; GFX7-NEXT: s_mov_b64 s[4:5], 0
; GFX7-NEXT: buffer_load_dword v0, v[0:1], s[4:7], 0 addr64
; GFX7-NEXT: s_waitcnt vmcnt(0)
; GFX7-NEXT: v_lshrrev_b32_e32 v0, 8, v0
; GFX7-NEXT: s_setpc_b64 s[30:31]
;
; GFX10-LABEL: extractelement_vgpr_v4i8_idx1:
; GFX10: ; %bb.0:
; GFX10-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX10-NEXT: global_load_dword v0, v[0:1], off
; GFX10-NEXT: s_waitcnt vmcnt(0)
; GFX10-NEXT: v_lshrrev_b32_e32 v0, 8, v0
; GFX10-NEXT: s_setpc_b64 s[30:31]
;
; GFX11-LABEL: extractelement_vgpr_v4i8_idx1:
; GFX11: ; %bb.0:
; GFX11-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX11-NEXT: global_load_b32 v0, v[0:1], off
; GFX11-NEXT: s_waitcnt vmcnt(0)
; GFX11-NEXT: v_lshrrev_b32_e32 v0, 8, v0
; GFX11-NEXT: s_setpc_b64 s[30:31]
%vector = load <4 x i8>, ptr addrspace(1) %ptr
%element = extractelement <4 x i8> %vector, i32 1
ret i8 %element
}
define i8 @extractelement_vgpr_v4i8_idx2(ptr addrspace(1) %ptr) {
; GFX9-LABEL: extractelement_vgpr_v4i8_idx2:
; GFX9: ; %bb.0:
; 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 v0, 16, v0
; GFX9-NEXT: s_setpc_b64 s[30:31]
;
; GFX8-LABEL: extractelement_vgpr_v4i8_idx2:
; GFX8: ; %bb.0:
; GFX8-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX8-NEXT: flat_load_dword v0, v[0:1]
; GFX8-NEXT: s_waitcnt vmcnt(0)
; GFX8-NEXT: v_lshrrev_b32_e32 v0, 16, v0
; GFX8-NEXT: s_setpc_b64 s[30:31]
;
; GFX7-LABEL: extractelement_vgpr_v4i8_idx2:
; GFX7: ; %bb.0:
; GFX7-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX7-NEXT: s_mov_b32 s6, 0
; GFX7-NEXT: s_mov_b32 s7, 0xf000
; GFX7-NEXT: s_mov_b64 s[4:5], 0
; GFX7-NEXT: buffer_load_dword v0, v[0:1], s[4:7], 0 addr64
; GFX7-NEXT: s_waitcnt vmcnt(0)
; GFX7-NEXT: v_lshrrev_b32_e32 v0, 16, v0
; GFX7-NEXT: s_setpc_b64 s[30:31]
;
; GFX10-LABEL: extractelement_vgpr_v4i8_idx2:
; GFX10: ; %bb.0:
; GFX10-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX10-NEXT: global_load_dword v0, v[0:1], off
; GFX10-NEXT: s_waitcnt vmcnt(0)
; GFX10-NEXT: v_lshrrev_b32_e32 v0, 16, v0
; GFX10-NEXT: s_setpc_b64 s[30:31]
;
; GFX11-LABEL: extractelement_vgpr_v4i8_idx2:
; GFX11: ; %bb.0:
; GFX11-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX11-NEXT: global_load_b32 v0, v[0:1], off
; GFX11-NEXT: s_waitcnt vmcnt(0)
; GFX11-NEXT: v_lshrrev_b32_e32 v0, 16, v0
; GFX11-NEXT: s_setpc_b64 s[30:31]
%vector = load <4 x i8>, ptr addrspace(1) %ptr
%element = extractelement <4 x i8> %vector, i32 2
ret i8 %element
}
define i8 @extractelement_vgpr_v4i8_idx3(ptr addrspace(1) %ptr) {
; GFX9-LABEL: extractelement_vgpr_v4i8_idx3:
; GFX9: ; %bb.0:
; 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 v0, 24, v0
; GFX9-NEXT: s_setpc_b64 s[30:31]
;
; GFX8-LABEL: extractelement_vgpr_v4i8_idx3:
; GFX8: ; %bb.0:
; GFX8-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX8-NEXT: flat_load_dword v0, v[0:1]
; GFX8-NEXT: s_waitcnt vmcnt(0)
; GFX8-NEXT: v_lshrrev_b32_e32 v0, 24, v0
; GFX8-NEXT: s_setpc_b64 s[30:31]
;
; GFX7-LABEL: extractelement_vgpr_v4i8_idx3:
; GFX7: ; %bb.0:
; GFX7-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX7-NEXT: s_mov_b32 s6, 0
; GFX7-NEXT: s_mov_b32 s7, 0xf000
; GFX7-NEXT: s_mov_b64 s[4:5], 0
; GFX7-NEXT: buffer_load_dword v0, v[0:1], s[4:7], 0 addr64
; GFX7-NEXT: s_waitcnt vmcnt(0)
; GFX7-NEXT: v_lshrrev_b32_e32 v0, 24, v0
; GFX7-NEXT: s_setpc_b64 s[30:31]
;
; GFX10-LABEL: extractelement_vgpr_v4i8_idx3:
; GFX10: ; %bb.0:
; GFX10-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX10-NEXT: global_load_dword v0, v[0:1], off
; GFX10-NEXT: s_waitcnt vmcnt(0)
; GFX10-NEXT: v_lshrrev_b32_e32 v0, 24, v0
; GFX10-NEXT: s_setpc_b64 s[30:31]
;
; GFX11-LABEL: extractelement_vgpr_v4i8_idx3:
; GFX11: ; %bb.0:
; GFX11-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX11-NEXT: global_load_b32 v0, v[0:1], off
; GFX11-NEXT: s_waitcnt vmcnt(0)
; GFX11-NEXT: v_lshrrev_b32_e32 v0, 24, v0
; GFX11-NEXT: s_setpc_b64 s[30:31]
%vector = load <4 x i8>, ptr addrspace(1) %ptr
%element = extractelement <4 x i8> %vector, i32 3
ret i8 %element
}
define amdgpu_ps i8 @extractelement_sgpr_v8i8_sgpr_idx(ptr addrspace(4) inreg %ptr, i32 inreg %idx) {
; GCN-LABEL: extractelement_sgpr_v8i8_sgpr_idx:
; GCN: ; %bb.0:
; GCN-NEXT: s_load_dwordx2 s[0:1], s[2:3], 0x0
; GCN-NEXT: s_lshr_b32 s2, s4, 2
; GCN-NEXT: s_cmp_eq_u32 s2, 1
; GCN-NEXT: s_waitcnt lgkmcnt(0)
; GCN-NEXT: s_cselect_b32 s0, s1, s0
; GCN-NEXT: s_and_b32 s1, s4, 3
; GCN-NEXT: s_lshl_b32 s1, s1, 3
; GCN-NEXT: s_lshr_b32 s0, s0, s1
; GCN-NEXT: ; return to shader part epilog
;
; GFX10-LABEL: extractelement_sgpr_v8i8_sgpr_idx:
; GFX10: ; %bb.0:
; GFX10-NEXT: s_load_dwordx2 s[0:1], s[2:3], 0x0
; GFX10-NEXT: s_lshr_b32 s2, s4, 2
; GFX10-NEXT: s_cmp_eq_u32 s2, 1
; GFX10-NEXT: s_waitcnt lgkmcnt(0)
; GFX10-NEXT: s_cselect_b32 s0, s1, s0
; GFX10-NEXT: s_and_b32 s1, s4, 3
; GFX10-NEXT: s_lshl_b32 s1, s1, 3
; GFX10-NEXT: s_lshr_b32 s0, s0, s1
; GFX10-NEXT: ; return to shader part epilog
;
; GFX11-LABEL: extractelement_sgpr_v8i8_sgpr_idx:
; GFX11: ; %bb.0:
; GFX11-NEXT: s_load_b64 s[0:1], s[2:3], 0x0
; GFX11-NEXT: s_lshr_b32 s2, s4, 2
; GFX11-NEXT: s_delay_alu instid0(SALU_CYCLE_1) | instskip(SKIP_3) | instid1(SALU_CYCLE_1)
; GFX11-NEXT: s_cmp_eq_u32 s2, 1
; GFX11-NEXT: s_waitcnt lgkmcnt(0)
; GFX11-NEXT: s_cselect_b32 s0, s1, s0
; GFX11-NEXT: s_and_b32 s1, s4, 3
; GFX11-NEXT: s_lshl_b32 s1, s1, 3
; GFX11-NEXT: s_delay_alu instid0(SALU_CYCLE_1)
; GFX11-NEXT: s_lshr_b32 s0, s0, s1
; GFX11-NEXT: ; return to shader part epilog
%vector = load <8 x i8>, ptr addrspace(4) %ptr
%element = extractelement <8 x i8> %vector, i32 %idx
ret i8 %element
}
define amdgpu_ps i8 @extractelement_vgpr_v8i8_sgpr_idx(ptr addrspace(1) %ptr, i32 inreg %idx) {
; GFX9-LABEL: extractelement_vgpr_v8i8_sgpr_idx:
; GFX9: ; %bb.0:
; GFX9-NEXT: global_load_dwordx2 v[0:1], v[0:1], off
; GFX9-NEXT: s_lshr_b32 s0, s2, 2
; GFX9-NEXT: s_and_b32 s1, s2, 3
; GFX9-NEXT: v_cmp_eq_u32_e64 vcc, s0, 1
; GFX9-NEXT: s_lshl_b32 s0, s1, 3
; GFX9-NEXT: s_waitcnt vmcnt(0)
; GFX9-NEXT: v_cndmask_b32_e32 v0, v0, v1, vcc
; GFX9-NEXT: v_lshrrev_b32_e32 v0, s0, v0
; GFX9-NEXT: v_readfirstlane_b32 s0, v0
; GFX9-NEXT: ; return to shader part epilog
;
; GFX8-LABEL: extractelement_vgpr_v8i8_sgpr_idx:
; GFX8: ; %bb.0:
; GFX8-NEXT: flat_load_dwordx2 v[0:1], v[0:1]
; GFX8-NEXT: s_lshr_b32 s0, s2, 2
; GFX8-NEXT: s_and_b32 s1, s2, 3
; GFX8-NEXT: v_cmp_eq_u32_e64 vcc, s0, 1
; GFX8-NEXT: s_lshl_b32 s0, s1, 3
; GFX8-NEXT: s_waitcnt vmcnt(0)
; GFX8-NEXT: v_cndmask_b32_e32 v0, v0, v1, vcc
; GFX8-NEXT: v_lshrrev_b32_e32 v0, s0, v0
; GFX8-NEXT: v_readfirstlane_b32 s0, v0
; GFX8-NEXT: ; return to shader part epilog
;
; GFX7-LABEL: extractelement_vgpr_v8i8_sgpr_idx:
; GFX7: ; %bb.0:
; GFX7-NEXT: s_mov_b32 s6, 0
; GFX7-NEXT: s_mov_b32 s7, 0xf000
; GFX7-NEXT: s_mov_b64 s[4:5], 0
; GFX7-NEXT: buffer_load_dwordx2 v[0:1], v[0:1], s[4:7], 0 addr64
; GFX7-NEXT: s_lshr_b32 s0, s2, 2
; GFX7-NEXT: s_and_b32 s1, s2, 3
; GFX7-NEXT: v_cmp_eq_u32_e64 vcc, s0, 1
; GFX7-NEXT: s_lshl_b32 s0, s1, 3
; GFX7-NEXT: s_waitcnt vmcnt(0)
; GFX7-NEXT: v_cndmask_b32_e32 v0, v0, v1, vcc
; GFX7-NEXT: v_lshrrev_b32_e32 v0, s0, v0
; GFX7-NEXT: v_readfirstlane_b32 s0, v0
; GFX7-NEXT: ; return to shader part epilog
;
; GFX10-LABEL: extractelement_vgpr_v8i8_sgpr_idx:
; GFX10: ; %bb.0:
; GFX10-NEXT: global_load_dwordx2 v[0:1], v[0:1], off
; GFX10-NEXT: s_lshr_b32 s0, s2, 2
; GFX10-NEXT: v_cmp_eq_u32_e64 vcc_lo, s0, 1
; GFX10-NEXT: s_and_b32 s0, s2, 3
; GFX10-NEXT: s_lshl_b32 s0, s0, 3
; GFX10-NEXT: s_waitcnt vmcnt(0)
; GFX10-NEXT: v_cndmask_b32_e32 v0, v0, v1, vcc_lo
; GFX10-NEXT: v_lshrrev_b32_e32 v0, s0, v0
; GFX10-NEXT: v_readfirstlane_b32 s0, v0
; GFX10-NEXT: ; return to shader part epilog
;
; GFX11-LABEL: extractelement_vgpr_v8i8_sgpr_idx:
; GFX11: ; %bb.0:
; GFX11-NEXT: global_load_b64 v[0:1], v[0:1], off
; GFX11-NEXT: s_lshr_b32 s0, s2, 2
; GFX11-NEXT: s_delay_alu instid0(SALU_CYCLE_1) | instskip(SKIP_1) | instid1(SALU_CYCLE_1)
; GFX11-NEXT: v_cmp_eq_u32_e64 vcc_lo, s0, 1
; GFX11-NEXT: s_and_b32 s0, s2, 3
; GFX11-NEXT: s_lshl_b32 s0, s0, 3
; GFX11-NEXT: s_waitcnt vmcnt(0)
; GFX11-NEXT: v_cndmask_b32_e32 v0, v0, v1, vcc_lo
; GFX11-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
; GFX11-NEXT: v_lshrrev_b32_e32 v0, s0, v0
; GFX11-NEXT: v_readfirstlane_b32 s0, v0
; GFX11-NEXT: ; return to shader part epilog
%vector = load <8 x i8>, ptr addrspace(1) %ptr
%element = extractelement <8 x i8> %vector, i32 %idx
ret i8 %element
}
define i8 @extractelement_vgpr_v8i8_vgpr_idx(ptr addrspace(1) %ptr, i32 %idx) {
; GFX9-LABEL: extractelement_vgpr_v8i8_vgpr_idx:
; 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_lshrrev_b32_e32 v3, 2, v2
; GFX9-NEXT: v_and_b32_e32 v2, 3, v2
; GFX9-NEXT: v_cmp_eq_u32_e32 vcc, 1, v3
; GFX9-NEXT: s_waitcnt vmcnt(0)
; GFX9-NEXT: v_cndmask_b32_e32 v0, v0, v1, vcc
; GFX9-NEXT: v_lshlrev_b32_e32 v1, 3, v2
; GFX9-NEXT: v_lshrrev_b32_e32 v0, v1, v0
; GFX9-NEXT: s_setpc_b64 s[30:31]
;
; GFX8-LABEL: extractelement_vgpr_v8i8_vgpr_idx:
; GFX8: ; %bb.0:
; GFX8-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX8-NEXT: flat_load_dwordx2 v[0:1], v[0:1]
; GFX8-NEXT: v_lshrrev_b32_e32 v3, 2, v2
; GFX8-NEXT: v_and_b32_e32 v2, 3, v2
; GFX8-NEXT: v_cmp_eq_u32_e32 vcc, 1, v3
; GFX8-NEXT: s_waitcnt vmcnt(0)
; GFX8-NEXT: v_cndmask_b32_e32 v0, v0, v1, vcc
; GFX8-NEXT: v_lshlrev_b32_e32 v1, 3, v2
; GFX8-NEXT: v_lshrrev_b32_e32 v0, v1, v0
; GFX8-NEXT: s_setpc_b64 s[30:31]
;
; GFX7-LABEL: extractelement_vgpr_v8i8_vgpr_idx:
; GFX7: ; %bb.0:
; GFX7-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX7-NEXT: s_mov_b32 s6, 0
; GFX7-NEXT: s_mov_b32 s7, 0xf000
; GFX7-NEXT: s_mov_b64 s[4:5], 0
; GFX7-NEXT: buffer_load_dwordx2 v[0:1], v[0:1], s[4:7], 0 addr64
; GFX7-NEXT: v_lshrrev_b32_e32 v3, 2, v2
; GFX7-NEXT: v_and_b32_e32 v2, 3, v2
; GFX7-NEXT: v_cmp_eq_u32_e32 vcc, 1, v3
; GFX7-NEXT: s_waitcnt vmcnt(0)
; GFX7-NEXT: v_cndmask_b32_e32 v0, v0, v1, vcc
; GFX7-NEXT: v_lshlrev_b32_e32 v1, 3, v2
; GFX7-NEXT: v_lshrrev_b32_e32 v0, v1, v0
; GFX7-NEXT: s_setpc_b64 s[30:31]
;
; GFX10-LABEL: extractelement_vgpr_v8i8_vgpr_idx:
; GFX10: ; %bb.0:
; GFX10-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX10-NEXT: global_load_dwordx2 v[0:1], v[0:1], off
; GFX10-NEXT: v_lshrrev_b32_e32 v3, 2, v2
; GFX10-NEXT: v_and_b32_e32 v2, 3, v2
; GFX10-NEXT: v_cmp_eq_u32_e32 vcc_lo, 1, v3
; GFX10-NEXT: s_waitcnt vmcnt(0)
; GFX10-NEXT: v_cndmask_b32_e32 v0, v0, v1, vcc_lo
; GFX10-NEXT: v_lshlrev_b32_e32 v1, 3, v2
; GFX10-NEXT: v_lshrrev_b32_e32 v0, v1, v0
; GFX10-NEXT: s_setpc_b64 s[30:31]
;
; GFX11-LABEL: extractelement_vgpr_v8i8_vgpr_idx:
; GFX11: ; %bb.0:
; GFX11-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX11-NEXT: global_load_b64 v[0:1], v[0:1], off
; GFX11-NEXT: v_lshrrev_b32_e32 v3, 2, v2
; GFX11-NEXT: v_and_b32_e32 v2, 3, v2
; GFX11-NEXT: s_delay_alu instid0(VALU_DEP_2) | instskip(SKIP_1) | instid1(VALU_DEP_2)
; GFX11-NEXT: v_cmp_eq_u32_e32 vcc_lo, 1, v3
; GFX11-NEXT: s_waitcnt vmcnt(0)
; GFX11-NEXT: v_dual_cndmask_b32 v0, v0, v1 :: v_dual_lshlrev_b32 v1, 3, v2
; GFX11-NEXT: s_delay_alu instid0(VALU_DEP_1)
; GFX11-NEXT: v_lshrrev_b32_e32 v0, v1, v0
; GFX11-NEXT: s_setpc_b64 s[30:31]
%vector = load <8 x i8>, ptr addrspace(1) %ptr
%element = extractelement <8 x i8> %vector, i32 %idx
ret i8 %element
}
define amdgpu_ps i8 @extractelement_sgpr_v8i8_vgpr_idx(ptr addrspace(4) inreg %ptr, i32 %idx) {
; GCN-LABEL: extractelement_sgpr_v8i8_vgpr_idx:
; GCN: ; %bb.0:
; GCN-NEXT: s_load_dwordx2 s[0:1], s[2:3], 0x0
; GCN-NEXT: v_lshrrev_b32_e32 v1, 2, v0
; GCN-NEXT: v_cmp_eq_u32_e32 vcc, 1, v1
; GCN-NEXT: v_and_b32_e32 v0, 3, v0
; GCN-NEXT: v_lshlrev_b32_e32 v0, 3, v0
; GCN-NEXT: s_waitcnt lgkmcnt(0)
; GCN-NEXT: v_mov_b32_e32 v2, s0
; GCN-NEXT: v_mov_b32_e32 v3, s1
; GCN-NEXT: v_cndmask_b32_e32 v1, v2, v3, vcc
; GCN-NEXT: v_lshrrev_b32_e32 v0, v0, v1
; GCN-NEXT: v_readfirstlane_b32 s0, v0
; GCN-NEXT: ; return to shader part epilog
;
; GFX10-LABEL: extractelement_sgpr_v8i8_vgpr_idx:
; GFX10: ; %bb.0:
; GFX10-NEXT: s_load_dwordx2 s[0:1], s[2:3], 0x0
; GFX10-NEXT: v_lshrrev_b32_e32 v1, 2, v0
; GFX10-NEXT: v_and_b32_e32 v0, 3, v0
; GFX10-NEXT: v_cmp_eq_u32_e32 vcc_lo, 1, v1
; GFX10-NEXT: v_lshlrev_b32_e32 v0, 3, v0
; GFX10-NEXT: s_waitcnt lgkmcnt(0)
; GFX10-NEXT: v_mov_b32_e32 v2, s1
; GFX10-NEXT: v_cndmask_b32_e32 v1, s0, v2, vcc_lo
; GFX10-NEXT: v_lshrrev_b32_e32 v0, v0, v1
; GFX10-NEXT: v_readfirstlane_b32 s0, v0
; GFX10-NEXT: ; return to shader part epilog
;
; GFX11-LABEL: extractelement_sgpr_v8i8_vgpr_idx:
; GFX11: ; %bb.0:
; GFX11-NEXT: s_load_b64 s[0:1], s[2:3], 0x0
; GFX11-NEXT: v_lshrrev_b32_e32 v1, 2, v0
; GFX11-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_2) | instid1(VALU_DEP_1)
; GFX11-NEXT: v_cmp_eq_u32_e32 vcc_lo, 1, v1
; GFX11-NEXT: s_waitcnt lgkmcnt(0)
; GFX11-NEXT: v_mov_b32_e32 v2, s1
; GFX11-NEXT: v_dual_cndmask_b32 v1, s0, v2 :: v_dual_and_b32 v0, 3, v0
; GFX11-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
; GFX11-NEXT: v_lshlrev_b32_e32 v0, 3, v0
; GFX11-NEXT: v_lshrrev_b32_e32 v0, v0, v1
; GFX11-NEXT: s_delay_alu instid0(VALU_DEP_1)
; GFX11-NEXT: v_readfirstlane_b32 s0, v0
; GFX11-NEXT: ; return to shader part epilog
%vector = load <8 x i8>, ptr addrspace(4) %ptr
%element = extractelement <8 x i8> %vector, i32 %idx
ret i8 %element
}
define amdgpu_ps i8 @extractelement_sgpr_v8i8_idx0(ptr addrspace(4) inreg %ptr) {
; GCN-LABEL: extractelement_sgpr_v8i8_idx0:
; GCN: ; %bb.0:
; GCN-NEXT: s_load_dwordx2 s[0:1], s[2:3], 0x0
; GCN-NEXT: s_waitcnt lgkmcnt(0)
; GCN-NEXT: ; return to shader part epilog
;
; GFX10-LABEL: extractelement_sgpr_v8i8_idx0:
; GFX10: ; %bb.0:
; GFX10-NEXT: s_load_dwordx2 s[0:1], s[2:3], 0x0
; GFX10-NEXT: s_waitcnt lgkmcnt(0)
; GFX10-NEXT: ; return to shader part epilog
;
; GFX11-LABEL: extractelement_sgpr_v8i8_idx0:
; GFX11: ; %bb.0:
; GFX11-NEXT: s_load_b64 s[0:1], s[2:3], 0x0
; GFX11-NEXT: s_waitcnt lgkmcnt(0)
; GFX11-NEXT: ; return to shader part epilog
%vector = load <8 x i8>, ptr addrspace(4) %ptr
%element = extractelement <8 x i8> %vector, i32 0
ret i8 %element
}
define amdgpu_ps i8 @extractelement_sgpr_v8i8_idx1(ptr addrspace(4) inreg %ptr) {
; GCN-LABEL: extractelement_sgpr_v8i8_idx1:
; GCN: ; %bb.0:
; GCN-NEXT: s_load_dwordx2 s[0:1], s[2:3], 0x0
; GCN-NEXT: s_waitcnt lgkmcnt(0)
; GCN-NEXT: s_lshr_b32 s0, s0, 8
; GCN-NEXT: ; return to shader part epilog
;
; GFX10-LABEL: extractelement_sgpr_v8i8_idx1:
; GFX10: ; %bb.0:
; GFX10-NEXT: s_load_dwordx2 s[0:1], s[2:3], 0x0
; GFX10-NEXT: s_waitcnt lgkmcnt(0)
; GFX10-NEXT: s_lshr_b32 s0, s0, 8
; GFX10-NEXT: ; return to shader part epilog
;
; GFX11-LABEL: extractelement_sgpr_v8i8_idx1:
; GFX11: ; %bb.0:
; GFX11-NEXT: s_load_b64 s[0:1], s[2:3], 0x0
; GFX11-NEXT: s_waitcnt lgkmcnt(0)
; GFX11-NEXT: s_lshr_b32 s0, s0, 8
; GFX11-NEXT: ; return to shader part epilog
%vector = load <8 x i8>, ptr addrspace(4) %ptr
%element = extractelement <8 x i8> %vector, i32 1
ret i8 %element
}
define amdgpu_ps i8 @extractelement_sgpr_v8i8_idx2(ptr addrspace(4) inreg %ptr) {
; GCN-LABEL: extractelement_sgpr_v8i8_idx2:
; GCN: ; %bb.0:
; GCN-NEXT: s_load_dwordx2 s[0:1], s[2:3], 0x0
; GCN-NEXT: s_waitcnt lgkmcnt(0)
; GCN-NEXT: s_lshr_b32 s0, s0, 16
; GCN-NEXT: ; return to shader part epilog
;
; GFX10-LABEL: extractelement_sgpr_v8i8_idx2:
; GFX10: ; %bb.0:
; GFX10-NEXT: s_load_dwordx2 s[0:1], s[2:3], 0x0
; GFX10-NEXT: s_waitcnt lgkmcnt(0)
; GFX10-NEXT: s_lshr_b32 s0, s0, 16
; GFX10-NEXT: ; return to shader part epilog
;
; GFX11-LABEL: extractelement_sgpr_v8i8_idx2:
; GFX11: ; %bb.0:
; GFX11-NEXT: s_load_b64 s[0:1], s[2:3], 0x0
; GFX11-NEXT: s_waitcnt lgkmcnt(0)
; GFX11-NEXT: s_lshr_b32 s0, s0, 16
; GFX11-NEXT: ; return to shader part epilog
%vector = load <8 x i8>, ptr addrspace(4) %ptr
%element = extractelement <8 x i8> %vector, i32 2
ret i8 %element
}
define amdgpu_ps i8 @extractelement_sgpr_v8i8_idx3(ptr addrspace(4) inreg %ptr) {
; GCN-LABEL: extractelement_sgpr_v8i8_idx3:
; GCN: ; %bb.0:
; GCN-NEXT: s_load_dwordx2 s[0:1], s[2:3], 0x0
; GCN-NEXT: s_waitcnt lgkmcnt(0)
; GCN-NEXT: s_lshr_b32 s0, s0, 24
; GCN-NEXT: ; return to shader part epilog
;
; GFX10-LABEL: extractelement_sgpr_v8i8_idx3:
; GFX10: ; %bb.0:
; GFX10-NEXT: s_load_dwordx2 s[0:1], s[2:3], 0x0
; GFX10-NEXT: s_waitcnt lgkmcnt(0)
; GFX10-NEXT: s_lshr_b32 s0, s0, 24
; GFX10-NEXT: ; return to shader part epilog
;
; GFX11-LABEL: extractelement_sgpr_v8i8_idx3:
; GFX11: ; %bb.0:
; GFX11-NEXT: s_load_b64 s[0:1], s[2:3], 0x0
; GFX11-NEXT: s_waitcnt lgkmcnt(0)
; GFX11-NEXT: s_lshr_b32 s0, s0, 24
; GFX11-NEXT: ; return to shader part epilog
%vector = load <8 x i8>, ptr addrspace(4) %ptr
%element = extractelement <8 x i8> %vector, i32 3
ret i8 %element
}
define amdgpu_ps i8 @extractelement_sgpr_v8i8_idx4(ptr addrspace(4) inreg %ptr) {
; GCN-LABEL: extractelement_sgpr_v8i8_idx4:
; GCN: ; %bb.0:
; GCN-NEXT: s_load_dwordx2 s[0:1], s[2:3], 0x0
; GCN-NEXT: s_waitcnt lgkmcnt(0)
; GCN-NEXT: s_mov_b32 s0, s1
; GCN-NEXT: ; return to shader part epilog
;
; GFX10-LABEL: extractelement_sgpr_v8i8_idx4:
; GFX10: ; %bb.0:
; GFX10-NEXT: s_load_dwordx2 s[0:1], s[2:3], 0x0
; GFX10-NEXT: s_waitcnt lgkmcnt(0)
; GFX10-NEXT: s_mov_b32 s0, s1
; GFX10-NEXT: ; return to shader part epilog
;
; GFX11-LABEL: extractelement_sgpr_v8i8_idx4:
; GFX11: ; %bb.0:
; GFX11-NEXT: s_load_b64 s[0:1], s[2:3], 0x0
; GFX11-NEXT: s_waitcnt lgkmcnt(0)
; GFX11-NEXT: s_mov_b32 s0, s1
; GFX11-NEXT: ; return to shader part epilog
%vector = load <8 x i8>, ptr addrspace(4) %ptr
%element = extractelement <8 x i8> %vector, i32 4
ret i8 %element
}
define amdgpu_ps i8 @extractelement_sgpr_v8i8_idx5(ptr addrspace(4) inreg %ptr) {
; GCN-LABEL: extractelement_sgpr_v8i8_idx5:
; GCN: ; %bb.0:
; GCN-NEXT: s_load_dwordx2 s[0:1], s[2:3], 0x0
; GCN-NEXT: s_waitcnt lgkmcnt(0)
; GCN-NEXT: s_lshr_b32 s0, s1, 8
; GCN-NEXT: ; return to shader part epilog
;
; GFX10-LABEL: extractelement_sgpr_v8i8_idx5:
; GFX10: ; %bb.0:
; GFX10-NEXT: s_load_dwordx2 s[0:1], s[2:3], 0x0
; GFX10-NEXT: s_waitcnt lgkmcnt(0)
; GFX10-NEXT: s_lshr_b32 s0, s1, 8
; GFX10-NEXT: ; return to shader part epilog
;
; GFX11-LABEL: extractelement_sgpr_v8i8_idx5:
; GFX11: ; %bb.0:
; GFX11-NEXT: s_load_b64 s[0:1], s[2:3], 0x0
; GFX11-NEXT: s_waitcnt lgkmcnt(0)
; GFX11-NEXT: s_lshr_b32 s0, s1, 8
; GFX11-NEXT: ; return to shader part epilog
%vector = load <8 x i8>, ptr addrspace(4) %ptr
%element = extractelement <8 x i8> %vector, i32 5
ret i8 %element
}
define amdgpu_ps i8 @extractelement_sgpr_v8i8_idx6(ptr addrspace(4) inreg %ptr) {
; GCN-LABEL: extractelement_sgpr_v8i8_idx6:
; GCN: ; %bb.0:
; GCN-NEXT: s_load_dwordx2 s[0:1], s[2:3], 0x0
; GCN-NEXT: s_waitcnt lgkmcnt(0)
; GCN-NEXT: s_lshr_b32 s0, s1, 16
; GCN-NEXT: ; return to shader part epilog
;
; GFX10-LABEL: extractelement_sgpr_v8i8_idx6:
; GFX10: ; %bb.0:
; GFX10-NEXT: s_load_dwordx2 s[0:1], s[2:3], 0x0
; GFX10-NEXT: s_waitcnt lgkmcnt(0)
; GFX10-NEXT: s_lshr_b32 s0, s1, 16
; GFX10-NEXT: ; return to shader part epilog
;
; GFX11-LABEL: extractelement_sgpr_v8i8_idx6:
; GFX11: ; %bb.0:
; GFX11-NEXT: s_load_b64 s[0:1], s[2:3], 0x0
; GFX11-NEXT: s_waitcnt lgkmcnt(0)
; GFX11-NEXT: s_lshr_b32 s0, s1, 16
; GFX11-NEXT: ; return to shader part epilog
%vector = load <8 x i8>, ptr addrspace(4) %ptr
%element = extractelement <8 x i8> %vector, i32 6
ret i8 %element
}
define amdgpu_ps i8 @extractelement_sgpr_v8i8_idx7(ptr addrspace(4) inreg %ptr) {
; GCN-LABEL: extractelement_sgpr_v8i8_idx7:
; GCN: ; %bb.0:
; GCN-NEXT: s_load_dwordx2 s[0:1], s[2:3], 0x0
; GCN-NEXT: s_waitcnt lgkmcnt(0)
; GCN-NEXT: s_lshr_b32 s0, s1, 24
; GCN-NEXT: ; return to shader part epilog
;
; GFX10-LABEL: extractelement_sgpr_v8i8_idx7:
; GFX10: ; %bb.0:
; GFX10-NEXT: s_load_dwordx2 s[0:1], s[2:3], 0x0
; GFX10-NEXT: s_waitcnt lgkmcnt(0)
; GFX10-NEXT: s_lshr_b32 s0, s1, 24
; GFX10-NEXT: ; return to shader part epilog
;
; GFX11-LABEL: extractelement_sgpr_v8i8_idx7:
; GFX11: ; %bb.0:
; GFX11-NEXT: s_load_b64 s[0:1], s[2:3], 0x0
; GFX11-NEXT: s_waitcnt lgkmcnt(0)
; GFX11-NEXT: s_lshr_b32 s0, s1, 24
; GFX11-NEXT: ; return to shader part epilog
%vector = load <8 x i8>, ptr addrspace(4) %ptr
%element = extractelement <8 x i8> %vector, i32 7
ret i8 %element
}
define i8 @extractelement_vgpr_v8i8_idx0(ptr addrspace(1) %ptr) {
; GFX9-LABEL: extractelement_vgpr_v8i8_idx0:
; 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: s_setpc_b64 s[30:31]
;
; GFX8-LABEL: extractelement_vgpr_v8i8_idx0:
; GFX8: ; %bb.0:
; GFX8-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX8-NEXT: flat_load_dwordx2 v[0:1], v[0:1]
; GFX8-NEXT: s_waitcnt vmcnt(0)
; GFX8-NEXT: s_setpc_b64 s[30:31]
;
; GFX7-LABEL: extractelement_vgpr_v8i8_idx0:
; GFX7: ; %bb.0:
; GFX7-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX7-NEXT: s_mov_b32 s6, 0
; GFX7-NEXT: s_mov_b32 s7, 0xf000
; GFX7-NEXT: s_mov_b64 s[4:5], 0
; GFX7-NEXT: buffer_load_dwordx2 v[0:1], v[0:1], s[4:7], 0 addr64
; GFX7-NEXT: s_waitcnt vmcnt(0)
; GFX7-NEXT: s_setpc_b64 s[30:31]
;
; GFX10-LABEL: extractelement_vgpr_v8i8_idx0:
; GFX10: ; %bb.0:
; GFX10-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX10-NEXT: global_load_dwordx2 v[0:1], v[0:1], off
; GFX10-NEXT: s_waitcnt vmcnt(0)
; GFX10-NEXT: s_setpc_b64 s[30:31]
;
; GFX11-LABEL: extractelement_vgpr_v8i8_idx0:
; GFX11: ; %bb.0:
; GFX11-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX11-NEXT: global_load_b64 v[0:1], v[0:1], off
; GFX11-NEXT: s_waitcnt vmcnt(0)
; GFX11-NEXT: s_setpc_b64 s[30:31]
%vector = load <8 x i8>, ptr addrspace(1) %ptr
%element = extractelement <8 x i8> %vector, i32 0
ret i8 %element
}
define i8 @extractelement_vgpr_v8i8_idx1(ptr addrspace(1) %ptr) {
; GFX9-LABEL: extractelement_vgpr_v8i8_idx1:
; 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: v_lshrrev_b32_e32 v0, 8, v0
; GFX9-NEXT: s_setpc_b64 s[30:31]
;
; GFX8-LABEL: extractelement_vgpr_v8i8_idx1:
; GFX8: ; %bb.0:
; GFX8-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX8-NEXT: flat_load_dwordx2 v[0:1], v[0:1]
; GFX8-NEXT: s_waitcnt vmcnt(0)
; GFX8-NEXT: v_lshrrev_b32_e32 v0, 8, v0
; GFX8-NEXT: s_setpc_b64 s[30:31]
;
; GFX7-LABEL: extractelement_vgpr_v8i8_idx1:
; GFX7: ; %bb.0:
; GFX7-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX7-NEXT: s_mov_b32 s6, 0
; GFX7-NEXT: s_mov_b32 s7, 0xf000
; GFX7-NEXT: s_mov_b64 s[4:5], 0
; GFX7-NEXT: buffer_load_dwordx2 v[0:1], v[0:1], s[4:7], 0 addr64
; GFX7-NEXT: s_waitcnt vmcnt(0)
; GFX7-NEXT: v_lshrrev_b32_e32 v0, 8, v0
; GFX7-NEXT: s_setpc_b64 s[30:31]
;
; GFX10-LABEL: extractelement_vgpr_v8i8_idx1:
; GFX10: ; %bb.0:
; GFX10-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX10-NEXT: global_load_dwordx2 v[0:1], v[0:1], off
; GFX10-NEXT: s_waitcnt vmcnt(0)
; GFX10-NEXT: v_lshrrev_b32_e32 v0, 8, v0
; GFX10-NEXT: s_setpc_b64 s[30:31]
;
; GFX11-LABEL: extractelement_vgpr_v8i8_idx1:
; GFX11: ; %bb.0:
; GFX11-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX11-NEXT: global_load_b64 v[0:1], v[0:1], off
; GFX11-NEXT: s_waitcnt vmcnt(0)
; GFX11-NEXT: v_lshrrev_b32_e32 v0, 8, v0
; GFX11-NEXT: s_setpc_b64 s[30:31]
%vector = load <8 x i8>, ptr addrspace(1) %ptr
%element = extractelement <8 x i8> %vector, i32 1
ret i8 %element
}
define i8 @extractelement_vgpr_v8i8_idx2(ptr addrspace(1) %ptr) {
; GFX9-LABEL: extractelement_vgpr_v8i8_idx2:
; 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: v_lshrrev_b32_e32 v0, 16, v0
; GFX9-NEXT: s_setpc_b64 s[30:31]
;
; GFX8-LABEL: extractelement_vgpr_v8i8_idx2:
; GFX8: ; %bb.0:
; GFX8-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX8-NEXT: flat_load_dwordx2 v[0:1], v[0:1]
; GFX8-NEXT: s_waitcnt vmcnt(0)
; GFX8-NEXT: v_lshrrev_b32_e32 v0, 16, v0
; GFX8-NEXT: s_setpc_b64 s[30:31]
;
; GFX7-LABEL: extractelement_vgpr_v8i8_idx2:
; GFX7: ; %bb.0:
; GFX7-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX7-NEXT: s_mov_b32 s6, 0
; GFX7-NEXT: s_mov_b32 s7, 0xf000
; GFX7-NEXT: s_mov_b64 s[4:5], 0
; GFX7-NEXT: buffer_load_dwordx2 v[0:1], v[0:1], s[4:7], 0 addr64
; GFX7-NEXT: s_waitcnt vmcnt(0)
; GFX7-NEXT: v_lshrrev_b32_e32 v0, 16, v0
; GFX7-NEXT: s_setpc_b64 s[30:31]
;
; GFX10-LABEL: extractelement_vgpr_v8i8_idx2:
; GFX10: ; %bb.0:
; GFX10-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX10-NEXT: global_load_dwordx2 v[0:1], v[0:1], off
; GFX10-NEXT: s_waitcnt vmcnt(0)
; GFX10-NEXT: v_lshrrev_b32_e32 v0, 16, v0
; GFX10-NEXT: s_setpc_b64 s[30:31]
;
; GFX11-LABEL: extractelement_vgpr_v8i8_idx2:
; GFX11: ; %bb.0:
; GFX11-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX11-NEXT: global_load_b64 v[0:1], v[0:1], off
; GFX11-NEXT: s_waitcnt vmcnt(0)
; GFX11-NEXT: v_lshrrev_b32_e32 v0, 16, v0
; GFX11-NEXT: s_setpc_b64 s[30:31]
%vector = load <8 x i8>, ptr addrspace(1) %ptr
%element = extractelement <8 x i8> %vector, i32 2
ret i8 %element
}
define i8 @extractelement_vgpr_v8i8_idx3(ptr addrspace(1) %ptr) {
; GFX9-LABEL: extractelement_vgpr_v8i8_idx3:
; 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: v_lshrrev_b32_e32 v0, 24, v0
; GFX9-NEXT: s_setpc_b64 s[30:31]
;
; GFX8-LABEL: extractelement_vgpr_v8i8_idx3:
; GFX8: ; %bb.0:
; GFX8-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX8-NEXT: flat_load_dwordx2 v[0:1], v[0:1]
; GFX8-NEXT: s_waitcnt vmcnt(0)
; GFX8-NEXT: v_lshrrev_b32_e32 v0, 24, v0
; GFX8-NEXT: s_setpc_b64 s[30:31]
;
; GFX7-LABEL: extractelement_vgpr_v8i8_idx3:
; GFX7: ; %bb.0:
; GFX7-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX7-NEXT: s_mov_b32 s6, 0
; GFX7-NEXT: s_mov_b32 s7, 0xf000
; GFX7-NEXT: s_mov_b64 s[4:5], 0
; GFX7-NEXT: buffer_load_dwordx2 v[0:1], v[0:1], s[4:7], 0 addr64
; GFX7-NEXT: s_waitcnt vmcnt(0)
; GFX7-NEXT: v_lshrrev_b32_e32 v0, 24, v0
; GFX7-NEXT: s_setpc_b64 s[30:31]
;
; GFX10-LABEL: extractelement_vgpr_v8i8_idx3:
; GFX10: ; %bb.0:
; GFX10-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX10-NEXT: global_load_dwordx2 v[0:1], v[0:1], off
; GFX10-NEXT: s_waitcnt vmcnt(0)
; GFX10-NEXT: v_lshrrev_b32_e32 v0, 24, v0
; GFX10-NEXT: s_setpc_b64 s[30:31]
;
; GFX11-LABEL: extractelement_vgpr_v8i8_idx3:
; GFX11: ; %bb.0:
; GFX11-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX11-NEXT: global_load_b64 v[0:1], v[0:1], off
; GFX11-NEXT: s_waitcnt vmcnt(0)
; GFX11-NEXT: v_lshrrev_b32_e32 v0, 24, v0
; GFX11-NEXT: s_setpc_b64 s[30:31]
%vector = load <8 x i8>, ptr addrspace(1) %ptr
%element = extractelement <8 x i8> %vector, i32 3
ret i8 %element
}
define i8 @extractelement_vgpr_v8i8_idx4(ptr addrspace(1) %ptr) {
; GFX9-LABEL: extractelement_vgpr_v8i8_idx4:
; 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: v_mov_b32_e32 v0, v1
; GFX9-NEXT: s_setpc_b64 s[30:31]
;
; GFX8-LABEL: extractelement_vgpr_v8i8_idx4:
; GFX8: ; %bb.0:
; GFX8-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX8-NEXT: flat_load_dwordx2 v[0:1], v[0:1]
; GFX8-NEXT: s_waitcnt vmcnt(0)
; GFX8-NEXT: v_mov_b32_e32 v0, v1
; GFX8-NEXT: s_setpc_b64 s[30:31]
;
; GFX7-LABEL: extractelement_vgpr_v8i8_idx4:
; GFX7: ; %bb.0:
; GFX7-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX7-NEXT: s_mov_b32 s6, 0
; GFX7-NEXT: s_mov_b32 s7, 0xf000
; GFX7-NEXT: s_mov_b64 s[4:5], 0
; GFX7-NEXT: buffer_load_dwordx2 v[0:1], v[0:1], s[4:7], 0 addr64
; GFX7-NEXT: s_waitcnt vmcnt(0)
; GFX7-NEXT: v_mov_b32_e32 v0, v1
; GFX7-NEXT: s_setpc_b64 s[30:31]
;
; GFX10-LABEL: extractelement_vgpr_v8i8_idx4:
; GFX10: ; %bb.0:
; GFX10-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX10-NEXT: global_load_dwordx2 v[0:1], v[0:1], off
; GFX10-NEXT: s_waitcnt vmcnt(0)
; GFX10-NEXT: v_mov_b32_e32 v0, v1
; GFX10-NEXT: s_setpc_b64 s[30:31]
;
; GFX11-LABEL: extractelement_vgpr_v8i8_idx4:
; GFX11: ; %bb.0:
; GFX11-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX11-NEXT: global_load_b64 v[0:1], v[0:1], off
; GFX11-NEXT: s_waitcnt vmcnt(0)
; GFX11-NEXT: v_mov_b32_e32 v0, v1
; GFX11-NEXT: s_setpc_b64 s[30:31]
%vector = load <8 x i8>, ptr addrspace(1) %ptr
%element = extractelement <8 x i8> %vector, i32 4
ret i8 %element
}
define i8 @extractelement_vgpr_v8i8_idx5(ptr addrspace(1) %ptr) {
; GFX9-LABEL: extractelement_vgpr_v8i8_idx5:
; 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: v_lshrrev_b32_e32 v0, 8, v1
; GFX9-NEXT: s_setpc_b64 s[30:31]
;
; GFX8-LABEL: extractelement_vgpr_v8i8_idx5:
; GFX8: ; %bb.0:
; GFX8-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX8-NEXT: flat_load_dwordx2 v[0:1], v[0:1]
; GFX8-NEXT: s_waitcnt vmcnt(0)
; GFX8-NEXT: v_lshrrev_b32_e32 v0, 8, v1
; GFX8-NEXT: s_setpc_b64 s[30:31]
;
; GFX7-LABEL: extractelement_vgpr_v8i8_idx5:
; GFX7: ; %bb.0:
; GFX7-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX7-NEXT: s_mov_b32 s6, 0
; GFX7-NEXT: s_mov_b32 s7, 0xf000
; GFX7-NEXT: s_mov_b64 s[4:5], 0
; GFX7-NEXT: buffer_load_dwordx2 v[0:1], v[0:1], s[4:7], 0 addr64
; GFX7-NEXT: s_waitcnt vmcnt(0)
; GFX7-NEXT: v_lshrrev_b32_e32 v0, 8, v1
; GFX7-NEXT: s_setpc_b64 s[30:31]
;
; GFX10-LABEL: extractelement_vgpr_v8i8_idx5:
; GFX10: ; %bb.0:
; GFX10-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX10-NEXT: global_load_dwordx2 v[0:1], v[0:1], off
; GFX10-NEXT: s_waitcnt vmcnt(0)
; GFX10-NEXT: v_lshrrev_b32_e32 v0, 8, v1
; GFX10-NEXT: s_setpc_b64 s[30:31]
;
; GFX11-LABEL: extractelement_vgpr_v8i8_idx5:
; GFX11: ; %bb.0:
; GFX11-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX11-NEXT: global_load_b64 v[0:1], v[0:1], off
; GFX11-NEXT: s_waitcnt vmcnt(0)
; GFX11-NEXT: v_lshrrev_b32_e32 v0, 8, v1
; GFX11-NEXT: s_setpc_b64 s[30:31]
%vector = load <8 x i8>, ptr addrspace(1) %ptr
%element = extractelement <8 x i8> %vector, i32 5
ret i8 %element
}
define i8 @extractelement_vgpr_v8i8_idx6(ptr addrspace(1) %ptr) {
; GFX9-LABEL: extractelement_vgpr_v8i8_idx6:
; 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: v_lshrrev_b32_e32 v0, 16, v1
; GFX9-NEXT: s_setpc_b64 s[30:31]
;
; GFX8-LABEL: extractelement_vgpr_v8i8_idx6:
; GFX8: ; %bb.0:
; GFX8-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX8-NEXT: flat_load_dwordx2 v[0:1], v[0:1]
; GFX8-NEXT: s_waitcnt vmcnt(0)
; GFX8-NEXT: v_lshrrev_b32_e32 v0, 16, v1
; GFX8-NEXT: s_setpc_b64 s[30:31]
;
; GFX7-LABEL: extractelement_vgpr_v8i8_idx6:
; GFX7: ; %bb.0:
; GFX7-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX7-NEXT: s_mov_b32 s6, 0
; GFX7-NEXT: s_mov_b32 s7, 0xf000
; GFX7-NEXT: s_mov_b64 s[4:5], 0
; GFX7-NEXT: buffer_load_dwordx2 v[0:1], v[0:1], s[4:7], 0 addr64
; GFX7-NEXT: s_waitcnt vmcnt(0)
; GFX7-NEXT: v_lshrrev_b32_e32 v0, 16, v1
; GFX7-NEXT: s_setpc_b64 s[30:31]
;
; GFX10-LABEL: extractelement_vgpr_v8i8_idx6:
; GFX10: ; %bb.0:
; GFX10-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX10-NEXT: global_load_dwordx2 v[0:1], v[0:1], off
; GFX10-NEXT: s_waitcnt vmcnt(0)
; GFX10-NEXT: v_lshrrev_b32_e32 v0, 16, v1
; GFX10-NEXT: s_setpc_b64 s[30:31]
;
; GFX11-LABEL: extractelement_vgpr_v8i8_idx6:
; GFX11: ; %bb.0:
; GFX11-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX11-NEXT: global_load_b64 v[0:1], v[0:1], off
; GFX11-NEXT: s_waitcnt vmcnt(0)
; GFX11-NEXT: v_lshrrev_b32_e32 v0, 16, v1
; GFX11-NEXT: s_setpc_b64 s[30:31]
%vector = load <8 x i8>, ptr addrspace(1) %ptr
%element = extractelement <8 x i8> %vector, i32 6
ret i8 %element
}
define i8 @extractelement_vgpr_v8i8_idx7(ptr addrspace(1) %ptr) {
; GFX9-LABEL: extractelement_vgpr_v8i8_idx7:
; 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: v_lshrrev_b32_e32 v0, 24, v1
; GFX9-NEXT: s_setpc_b64 s[30:31]
;
; GFX8-LABEL: extractelement_vgpr_v8i8_idx7:
; GFX8: ; %bb.0:
; GFX8-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX8-NEXT: flat_load_dwordx2 v[0:1], v[0:1]
; GFX8-NEXT: s_waitcnt vmcnt(0)
; GFX8-NEXT: v_lshrrev_b32_e32 v0, 24, v1
; GFX8-NEXT: s_setpc_b64 s[30:31]
;
; GFX7-LABEL: extractelement_vgpr_v8i8_idx7:
; GFX7: ; %bb.0:
; GFX7-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX7-NEXT: s_mov_b32 s6, 0
; GFX7-NEXT: s_mov_b32 s7, 0xf000
; GFX7-NEXT: s_mov_b64 s[4:5], 0
; GFX7-NEXT: buffer_load_dwordx2 v[0:1], v[0:1], s[4:7], 0 addr64
; GFX7-NEXT: s_waitcnt vmcnt(0)
; GFX7-NEXT: v_lshrrev_b32_e32 v0, 24, v1
; GFX7-NEXT: s_setpc_b64 s[30:31]
;
; GFX10-LABEL: extractelement_vgpr_v8i8_idx7:
; GFX10: ; %bb.0:
; GFX10-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX10-NEXT: global_load_dwordx2 v[0:1], v[0:1], off
; GFX10-NEXT: s_waitcnt vmcnt(0)
; GFX10-NEXT: v_lshrrev_b32_e32 v0, 24, v1
; GFX10-NEXT: s_setpc_b64 s[30:31]
;
; GFX11-LABEL: extractelement_vgpr_v8i8_idx7:
; GFX11: ; %bb.0:
; GFX11-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX11-NEXT: global_load_b64 v[0:1], v[0:1], off
; GFX11-NEXT: s_waitcnt vmcnt(0)
; GFX11-NEXT: v_lshrrev_b32_e32 v0, 24, v1
; GFX11-NEXT: s_setpc_b64 s[30:31]
%vector = load <8 x i8>, ptr addrspace(1) %ptr
%element = extractelement <8 x i8> %vector, i32 7
ret i8 %element
}
define amdgpu_ps i8 @extractelement_sgpr_v16i8_sgpr_idx(ptr addrspace(4) inreg %ptr, i32 inreg %idx) {
; GCN-LABEL: extractelement_sgpr_v16i8_sgpr_idx:
; GCN: ; %bb.0:
; GCN-NEXT: s_load_dwordx4 s[0:3], s[2:3], 0x0
; GCN-NEXT: s_lshr_b32 s5, s4, 2
; GCN-NEXT: s_cmp_eq_u32 s5, 1
; GCN-NEXT: s_waitcnt lgkmcnt(0)
; GCN-NEXT: s_cselect_b32 s0, s1, s0
; GCN-NEXT: s_cmp_eq_u32 s5, 2
; GCN-NEXT: s_cselect_b32 s0, s2, s0
; GCN-NEXT: s_cmp_eq_u32 s5, 3
; GCN-NEXT: s_cselect_b32 s0, s3, s0
; GCN-NEXT: s_and_b32 s1, s4, 3
; GCN-NEXT: s_lshl_b32 s1, s1, 3
; GCN-NEXT: s_lshr_b32 s0, s0, s1
; GCN-NEXT: ; return to shader part epilog
;
; GFX10-LABEL: extractelement_sgpr_v16i8_sgpr_idx:
; GFX10: ; %bb.0:
; GFX10-NEXT: s_load_dwordx4 s[0:3], s[2:3], 0x0
; GFX10-NEXT: s_lshr_b32 s5, s4, 2
; GFX10-NEXT: s_cmp_eq_u32 s5, 1
; GFX10-NEXT: s_waitcnt lgkmcnt(0)
; GFX10-NEXT: s_cselect_b32 s0, s1, s0
; GFX10-NEXT: s_cmp_eq_u32 s5, 2
; GFX10-NEXT: s_cselect_b32 s0, s2, s0
; GFX10-NEXT: s_cmp_eq_u32 s5, 3
; GFX10-NEXT: s_cselect_b32 s0, s3, s0
; GFX10-NEXT: s_and_b32 s1, s4, 3
; GFX10-NEXT: s_lshl_b32 s1, s1, 3
; GFX10-NEXT: s_lshr_b32 s0, s0, s1
; GFX10-NEXT: ; return to shader part epilog
;
; GFX11-LABEL: extractelement_sgpr_v16i8_sgpr_idx:
; GFX11: ; %bb.0:
; GFX11-NEXT: s_load_b128 s[0:3], s[2:3], 0x0
; GFX11-NEXT: s_lshr_b32 s5, s4, 2
; GFX11-NEXT: s_delay_alu instid0(SALU_CYCLE_1)
; GFX11-NEXT: s_cmp_eq_u32 s5, 1
; GFX11-NEXT: s_waitcnt lgkmcnt(0)
; GFX11-NEXT: s_cselect_b32 s0, s1, s0
; GFX11-NEXT: s_cmp_eq_u32 s5, 2
; GFX11-NEXT: s_cselect_b32 s0, s2, s0
; GFX11-NEXT: s_cmp_eq_u32 s5, 3
; GFX11-NEXT: s_cselect_b32 s0, s3, s0
; GFX11-NEXT: s_and_b32 s1, s4, 3
; GFX11-NEXT: s_delay_alu instid0(SALU_CYCLE_1) | instskip(NEXT) | instid1(SALU_CYCLE_1)
; GFX11-NEXT: s_lshl_b32 s1, s1, 3
; GFX11-NEXT: s_lshr_b32 s0, s0, s1
; GFX11-NEXT: ; return to shader part epilog
%vector = load <16 x i8>, ptr addrspace(4) %ptr
%element = extractelement <16 x i8> %vector, i32 %idx
ret i8 %element
}
define amdgpu_ps i8 @extractelement_vgpr_v16i8_sgpr_idx(ptr addrspace(1) %ptr, i32 inreg %idx) {
; GFX9-LABEL: extractelement_vgpr_v16i8_sgpr_idx:
; GFX9: ; %bb.0:
; GFX9-NEXT: global_load_dwordx4 v[0:3], v[0:1], off
; GFX9-NEXT: s_lshr_b32 s0, s2, 2
; GFX9-NEXT: v_cmp_eq_u32_e64 vcc, s0, 1
; GFX9-NEXT: s_and_b32 s1, s2, 3
; GFX9-NEXT: s_waitcnt vmcnt(0)
; GFX9-NEXT: v_cndmask_b32_e32 v0, v0, v1, vcc
; GFX9-NEXT: v_cmp_eq_u32_e64 vcc, s0, 2
; GFX9-NEXT: v_cndmask_b32_e32 v0, v0, v2, vcc
; GFX9-NEXT: v_cmp_eq_u32_e64 vcc, s0, 3
; GFX9-NEXT: v_cndmask_b32_e32 v0, v0, v3, vcc
; GFX9-NEXT: s_lshl_b32 s0, s1, 3
; GFX9-NEXT: v_lshrrev_b32_e32 v0, s0, v0
; GFX9-NEXT: v_readfirstlane_b32 s0, v0
; GFX9-NEXT: ; return to shader part epilog
;
; GFX8-LABEL: extractelement_vgpr_v16i8_sgpr_idx:
; GFX8: ; %bb.0:
; GFX8-NEXT: flat_load_dwordx4 v[0:3], v[0:1]
; GFX8-NEXT: s_lshr_b32 s0, s2, 2
; GFX8-NEXT: v_cmp_eq_u32_e64 vcc, s0, 1
; GFX8-NEXT: s_and_b32 s1, s2, 3
; GFX8-NEXT: s_waitcnt vmcnt(0)
; GFX8-NEXT: v_cndmask_b32_e32 v0, v0, v1, vcc
; GFX8-NEXT: v_cmp_eq_u32_e64 vcc, s0, 2
; GFX8-NEXT: v_cndmask_b32_e32 v0, v0, v2, vcc
; GFX8-NEXT: v_cmp_eq_u32_e64 vcc, s0, 3
; GFX8-NEXT: v_cndmask_b32_e32 v0, v0, v3, vcc
; GFX8-NEXT: s_lshl_b32 s0, s1, 3
; GFX8-NEXT: v_lshrrev_b32_e32 v0, s0, v0
; GFX8-NEXT: v_readfirstlane_b32 s0, v0
; GFX8-NEXT: ; return to shader part epilog
;
; GFX7-LABEL: extractelement_vgpr_v16i8_sgpr_idx:
; GFX7: ; %bb.0:
; GFX7-NEXT: s_mov_b32 s6, 0
; GFX7-NEXT: s_mov_b32 s7, 0xf000
; GFX7-NEXT: s_mov_b64 s[4:5], 0
; GFX7-NEXT: buffer_load_dwordx4 v[0:3], v[0:1], s[4:7], 0 addr64
; GFX7-NEXT: s_lshr_b32 s0, s2, 2
; GFX7-NEXT: v_cmp_eq_u32_e64 vcc, s0, 1
; GFX7-NEXT: s_and_b32 s1, s2, 3
; GFX7-NEXT: s_waitcnt vmcnt(0)
; GFX7-NEXT: v_cndmask_b32_e32 v0, v0, v1, vcc
; GFX7-NEXT: v_cmp_eq_u32_e64 vcc, s0, 2
; GFX7-NEXT: v_cndmask_b32_e32 v0, v0, v2, vcc
; GFX7-NEXT: v_cmp_eq_u32_e64 vcc, s0, 3
; GFX7-NEXT: v_cndmask_b32_e32 v0, v0, v3, vcc
; GFX7-NEXT: s_lshl_b32 s0, s1, 3
; GFX7-NEXT: v_lshrrev_b32_e32 v0, s0, v0
; GFX7-NEXT: v_readfirstlane_b32 s0, v0
; GFX7-NEXT: ; return to shader part epilog
;
; GFX10-LABEL: extractelement_vgpr_v16i8_sgpr_idx:
; GFX10: ; %bb.0:
; GFX10-NEXT: global_load_dwordx4 v[0:3], v[0:1], off
; GFX10-NEXT: s_lshr_b32 s0, s2, 2
; GFX10-NEXT: v_cmp_eq_u32_e64 vcc_lo, s0, 1
; GFX10-NEXT: s_waitcnt vmcnt(0)
; GFX10-NEXT: v_cndmask_b32_e32 v0, v0, v1, vcc_lo
; GFX10-NEXT: v_cmp_eq_u32_e64 vcc_lo, s0, 2
; GFX10-NEXT: v_cndmask_b32_e32 v0, v0, v2, vcc_lo
; GFX10-NEXT: v_cmp_eq_u32_e64 vcc_lo, s0, 3
; GFX10-NEXT: s_and_b32 s0, s2, 3
; GFX10-NEXT: s_lshl_b32 s0, s0, 3
; GFX10-NEXT: v_cndmask_b32_e32 v0, v0, v3, vcc_lo
; GFX10-NEXT: v_lshrrev_b32_e32 v0, s0, v0
; GFX10-NEXT: v_readfirstlane_b32 s0, v0
; GFX10-NEXT: ; return to shader part epilog
;
; GFX11-LABEL: extractelement_vgpr_v16i8_sgpr_idx:
; GFX11: ; %bb.0:
; GFX11-NEXT: global_load_b128 v[0:3], v[0:1], off
; GFX11-NEXT: s_lshr_b32 s0, s2, 2
; GFX11-NEXT: s_delay_alu instid0(SALU_CYCLE_1) | instskip(SKIP_3) | instid1(VALU_DEP_2)
; GFX11-NEXT: v_cmp_eq_u32_e64 vcc_lo, s0, 1
; GFX11-NEXT: s_waitcnt vmcnt(0)
; GFX11-NEXT: v_cndmask_b32_e32 v0, v0, v1, vcc_lo
; GFX11-NEXT: v_cmp_eq_u32_e64 vcc_lo, s0, 2
; GFX11-NEXT: v_cndmask_b32_e32 v0, v0, v2, vcc_lo
; GFX11-NEXT: v_cmp_eq_u32_e64 vcc_lo, s0, 3
; GFX11-NEXT: s_and_b32 s0, s2, 3
; GFX11-NEXT: s_delay_alu instid0(SALU_CYCLE_1) | instskip(NEXT) | instid1(VALU_DEP_2)
; GFX11-NEXT: s_lshl_b32 s0, s0, 3
; GFX11-NEXT: v_cndmask_b32_e32 v0, v0, v3, vcc_lo
; GFX11-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
; GFX11-NEXT: v_lshrrev_b32_e32 v0, s0, v0
; GFX11-NEXT: v_readfirstlane_b32 s0, v0
; GFX11-NEXT: ; return to shader part epilog
%vector = load <16 x i8>, ptr addrspace(1) %ptr
%element = extractelement <16 x i8> %vector, i32 %idx
ret i8 %element
}
define i8 @extractelement_vgpr_v16i8_vgpr_idx(ptr addrspace(1) %ptr, i32 %idx) {
; GFX9-LABEL: extractelement_vgpr_v16i8_vgpr_idx:
; GFX9: ; %bb.0:
; GFX9-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX9-NEXT: global_load_dwordx4 v[3:6], v[0:1], off
; GFX9-NEXT: v_lshrrev_b32_e32 v0, 2, v2
; GFX9-NEXT: v_cmp_eq_u32_e32 vcc, 1, v0
; GFX9-NEXT: v_and_b32_e32 v1, 3, v2
; GFX9-NEXT: v_lshlrev_b32_e32 v1, 3, v1
; GFX9-NEXT: s_waitcnt vmcnt(0)
; GFX9-NEXT: v_cndmask_b32_e32 v2, v3, v4, vcc
; GFX9-NEXT: v_cmp_eq_u32_e32 vcc, 2, v0
; GFX9-NEXT: v_cndmask_b32_e32 v2, v2, v5, vcc
; GFX9-NEXT: v_cmp_eq_u32_e32 vcc, 3, v0
; GFX9-NEXT: v_cndmask_b32_e32 v0, v2, v6, vcc
; GFX9-NEXT: v_lshrrev_b32_e32 v0, v1, v0
; GFX9-NEXT: s_setpc_b64 s[30:31]
;
; GFX8-LABEL: extractelement_vgpr_v16i8_vgpr_idx:
; GFX8: ; %bb.0:
; GFX8-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX8-NEXT: flat_load_dwordx4 v[3:6], v[0:1]
; GFX8-NEXT: v_lshrrev_b32_e32 v0, 2, v2
; GFX8-NEXT: v_cmp_eq_u32_e32 vcc, 1, v0
; GFX8-NEXT: v_and_b32_e32 v1, 3, v2
; GFX8-NEXT: v_lshlrev_b32_e32 v1, 3, v1
; GFX8-NEXT: s_waitcnt vmcnt(0)
; GFX8-NEXT: v_cndmask_b32_e32 v2, v3, v4, vcc
; GFX8-NEXT: v_cmp_eq_u32_e32 vcc, 2, v0
; GFX8-NEXT: v_cndmask_b32_e32 v2, v2, v5, vcc
; GFX8-NEXT: v_cmp_eq_u32_e32 vcc, 3, v0
; GFX8-NEXT: v_cndmask_b32_e32 v0, v2, v6, vcc
; GFX8-NEXT: v_lshrrev_b32_e32 v0, v1, v0
; GFX8-NEXT: s_setpc_b64 s[30:31]
;
; GFX7-LABEL: extractelement_vgpr_v16i8_vgpr_idx:
; GFX7: ; %bb.0:
; GFX7-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX7-NEXT: s_mov_b32 s6, 0
; GFX7-NEXT: s_mov_b32 s7, 0xf000
; GFX7-NEXT: s_mov_b64 s[4:5], 0
; GFX7-NEXT: buffer_load_dwordx4 v[3:6], v[0:1], s[4:7], 0 addr64
; GFX7-NEXT: v_lshrrev_b32_e32 v0, 2, v2
; GFX7-NEXT: v_cmp_eq_u32_e32 vcc, 1, v0
; GFX7-NEXT: v_and_b32_e32 v1, 3, v2
; GFX7-NEXT: v_lshlrev_b32_e32 v1, 3, v1
; GFX7-NEXT: s_waitcnt vmcnt(0)
; GFX7-NEXT: v_cndmask_b32_e32 v2, v3, v4, vcc
; GFX7-NEXT: v_cmp_eq_u32_e32 vcc, 2, v0
; GFX7-NEXT: v_cndmask_b32_e32 v2, v2, v5, vcc
; GFX7-NEXT: v_cmp_eq_u32_e32 vcc, 3, v0
; GFX7-NEXT: v_cndmask_b32_e32 v0, v2, v6, vcc
; GFX7-NEXT: v_lshrrev_b32_e32 v0, v1, v0
; GFX7-NEXT: s_setpc_b64 s[30:31]
;
; GFX10-LABEL: extractelement_vgpr_v16i8_vgpr_idx:
; GFX10: ; %bb.0:
; GFX10-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX10-NEXT: global_load_dwordx4 v[3:6], v[0:1], off
; GFX10-NEXT: v_lshrrev_b32_e32 v0, 2, v2
; GFX10-NEXT: v_and_b32_e32 v2, 3, v2
; GFX10-NEXT: v_cmp_eq_u32_e32 vcc_lo, 1, v0
; GFX10-NEXT: s_waitcnt vmcnt(0)
; GFX10-NEXT: v_cndmask_b32_e32 v1, v3, v4, vcc_lo
; GFX10-NEXT: v_cmp_eq_u32_e32 vcc_lo, 2, v0
; GFX10-NEXT: v_cndmask_b32_e32 v1, v1, v5, vcc_lo
; GFX10-NEXT: v_cmp_eq_u32_e32 vcc_lo, 3, v0
; GFX10-NEXT: v_cndmask_b32_e32 v0, v1, v6, vcc_lo
; GFX10-NEXT: v_lshlrev_b32_e32 v1, 3, v2
; GFX10-NEXT: v_lshrrev_b32_e32 v0, v1, v0
; GFX10-NEXT: s_setpc_b64 s[30:31]
;
; GFX11-LABEL: extractelement_vgpr_v16i8_vgpr_idx:
; GFX11: ; %bb.0:
; GFX11-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX11-NEXT: global_load_b128 v[3:6], v[0:1], off
; GFX11-NEXT: v_lshrrev_b32_e32 v0, 2, v2
; GFX11-NEXT: v_and_b32_e32 v2, 3, v2
; GFX11-NEXT: s_delay_alu instid0(VALU_DEP_2) | instskip(SKIP_3) | instid1(VALU_DEP_2)
; GFX11-NEXT: v_cmp_eq_u32_e32 vcc_lo, 1, v0
; GFX11-NEXT: s_waitcnt vmcnt(0)
; GFX11-NEXT: v_cndmask_b32_e32 v1, v3, v4, vcc_lo
; GFX11-NEXT: v_cmp_eq_u32_e32 vcc_lo, 2, v0
; GFX11-NEXT: v_cndmask_b32_e32 v1, v1, v5, vcc_lo
; GFX11-NEXT: v_cmp_eq_u32_e32 vcc_lo, 3, v0
; GFX11-NEXT: s_delay_alu instid0(VALU_DEP_2) | instskip(SKIP_1) | instid1(VALU_DEP_1)
; GFX11-NEXT: v_cndmask_b32_e32 v0, v1, v6, vcc_lo
; GFX11-NEXT: v_lshlrev_b32_e32 v1, 3, v2
; GFX11-NEXT: v_lshrrev_b32_e32 v0, v1, v0
; GFX11-NEXT: s_setpc_b64 s[30:31]
%vector = load <16 x i8>, ptr addrspace(1) %ptr
%element = extractelement <16 x i8> %vector, i32 %idx
ret i8 %element
}
define amdgpu_ps i8 @extractelement_sgpr_v16i8_vgpr_idx(ptr addrspace(4) inreg %ptr, i32 %idx) {
; GCN-LABEL: extractelement_sgpr_v16i8_vgpr_idx:
; GCN: ; %bb.0:
; GCN-NEXT: s_load_dwordx4 s[0:3], s[2:3], 0x0
; GCN-NEXT: v_lshrrev_b32_e32 v1, 2, v0
; GCN-NEXT: v_cmp_eq_u32_e32 vcc, 1, v1
; GCN-NEXT: v_and_b32_e32 v0, 3, v0
; GCN-NEXT: v_lshlrev_b32_e32 v0, 3, v0
; GCN-NEXT: s_waitcnt lgkmcnt(0)
; GCN-NEXT: v_mov_b32_e32 v2, s0
; GCN-NEXT: v_mov_b32_e32 v3, s1
; GCN-NEXT: v_mov_b32_e32 v4, s2
; GCN-NEXT: v_cndmask_b32_e32 v2, v2, v3, vcc
; GCN-NEXT: v_cmp_eq_u32_e32 vcc, 2, v1
; GCN-NEXT: v_mov_b32_e32 v5, s3
; GCN-NEXT: v_cndmask_b32_e32 v2, v2, v4, vcc
; GCN-NEXT: v_cmp_eq_u32_e32 vcc, 3, v1
; GCN-NEXT: v_cndmask_b32_e32 v1, v2, v5, vcc
; GCN-NEXT: v_lshrrev_b32_e32 v0, v0, v1
; GCN-NEXT: v_readfirstlane_b32 s0, v0
; GCN-NEXT: ; return to shader part epilog
;
; GFX10-LABEL: extractelement_sgpr_v16i8_vgpr_idx:
; GFX10: ; %bb.0:
; GFX10-NEXT: s_load_dwordx4 s[0:3], s[2:3], 0x0
; GFX10-NEXT: v_lshrrev_b32_e32 v1, 2, v0
; GFX10-NEXT: v_and_b32_e32 v0, 3, v0
; GFX10-NEXT: v_cmp_eq_u32_e32 vcc_lo, 1, v1
; GFX10-NEXT: v_lshlrev_b32_e32 v0, 3, v0
; GFX10-NEXT: s_waitcnt lgkmcnt(0)
; GFX10-NEXT: v_mov_b32_e32 v2, s1
; GFX10-NEXT: v_cndmask_b32_e32 v2, s0, v2, vcc_lo
; GFX10-NEXT: v_cmp_eq_u32_e32 vcc_lo, 2, v1
; GFX10-NEXT: v_cndmask_b32_e64 v2, v2, s2, vcc_lo
; GFX10-NEXT: v_cmp_eq_u32_e32 vcc_lo, 3, v1
; GFX10-NEXT: v_cndmask_b32_e64 v1, v2, s3, vcc_lo
; GFX10-NEXT: v_lshrrev_b32_e32 v0, v0, v1
; GFX10-NEXT: v_readfirstlane_b32 s0, v0
; GFX10-NEXT: ; return to shader part epilog
;
; GFX11-LABEL: extractelement_sgpr_v16i8_vgpr_idx:
; GFX11: ; %bb.0:
; GFX11-NEXT: s_load_b128 s[0:3], s[2:3], 0x0
; GFX11-NEXT: v_lshrrev_b32_e32 v1, 2, v0
; GFX11-NEXT: v_and_b32_e32 v0, 3, v0
; GFX11-NEXT: s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_2)
; GFX11-NEXT: v_cmp_eq_u32_e32 vcc_lo, 1, v1
; GFX11-NEXT: v_lshlrev_b32_e32 v0, 3, v0
; GFX11-NEXT: s_waitcnt lgkmcnt(0)
; GFX11-NEXT: v_mov_b32_e32 v2, s1
; GFX11-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_1) | instid1(VALU_DEP_2)
; GFX11-NEXT: v_cndmask_b32_e32 v2, s0, v2, vcc_lo
; GFX11-NEXT: v_cmp_eq_u32_e32 vcc_lo, 2, v1
; GFX11-NEXT: v_cndmask_b32_e64 v2, v2, s2, vcc_lo
; GFX11-NEXT: v_cmp_eq_u32_e32 vcc_lo, 3, v1
; GFX11-NEXT: s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_1)
; GFX11-NEXT: v_cndmask_b32_e64 v1, v2, s3, vcc_lo
; GFX11-NEXT: v_lshrrev_b32_e32 v0, v0, v1
; GFX11-NEXT: s_delay_alu instid0(VALU_DEP_1)
; GFX11-NEXT: v_readfirstlane_b32 s0, v0
; GFX11-NEXT: ; return to shader part epilog
%vector = load <16 x i8>, ptr addrspace(4) %ptr
%element = extractelement <16 x i8> %vector, i32 %idx
ret i8 %element
}
define i8 @extractelement_vgpr_v16i8_idx0(ptr addrspace(1) %ptr) {
; GFX9-LABEL: extractelement_vgpr_v16i8_idx0:
; GFX9: ; %bb.0:
; GFX9-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX9-NEXT: global_load_dwordx4 v[0:3], v[0:1], off
; GFX9-NEXT: s_waitcnt vmcnt(0)
; GFX9-NEXT: s_setpc_b64 s[30:31]
;
; GFX8-LABEL: extractelement_vgpr_v16i8_idx0:
; GFX8: ; %bb.0:
; GFX8-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX8-NEXT: flat_load_dwordx4 v[0:3], v[0:1]
; GFX8-NEXT: s_waitcnt vmcnt(0)
; GFX8-NEXT: s_setpc_b64 s[30:31]
;
; GFX7-LABEL: extractelement_vgpr_v16i8_idx0:
; GFX7: ; %bb.0:
; GFX7-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX7-NEXT: s_mov_b32 s6, 0
; GFX7-NEXT: s_mov_b32 s7, 0xf000
; GFX7-NEXT: s_mov_b64 s[4:5], 0
; GFX7-NEXT: buffer_load_dwordx4 v[0:3], v[0:1], s[4:7], 0 addr64
; GFX7-NEXT: s_waitcnt vmcnt(0)
; GFX7-NEXT: s_setpc_b64 s[30:31]
;
; GFX10-LABEL: extractelement_vgpr_v16i8_idx0:
; GFX10: ; %bb.0:
; GFX10-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX10-NEXT: global_load_dwordx4 v[0:3], v[0:1], off
; GFX10-NEXT: s_waitcnt vmcnt(0)
; GFX10-NEXT: s_setpc_b64 s[30:31]
;
; GFX11-LABEL: extractelement_vgpr_v16i8_idx0:
; GFX11: ; %bb.0:
; GFX11-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX11-NEXT: global_load_b128 v[0:3], v[0:1], off
; GFX11-NEXT: s_waitcnt vmcnt(0)
; GFX11-NEXT: s_setpc_b64 s[30:31]
%vector = load <16 x i8>, ptr addrspace(1) %ptr
%element = extractelement <16 x i8> %vector, i32 0
ret i8 %element
}
define i8 @extractelement_vgpr_v16i8_idx1(ptr addrspace(1) %ptr) {
; GFX9-LABEL: extractelement_vgpr_v16i8_idx1:
; GFX9: ; %bb.0:
; GFX9-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX9-NEXT: global_load_dwordx4 v[0:3], v[0:1], off
; GFX9-NEXT: s_waitcnt vmcnt(0)
; GFX9-NEXT: v_lshrrev_b32_e32 v0, 8, v0
; GFX9-NEXT: s_setpc_b64 s[30:31]
;
; GFX8-LABEL: extractelement_vgpr_v16i8_idx1:
; GFX8: ; %bb.0:
; GFX8-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX8-NEXT: flat_load_dwordx4 v[0:3], v[0:1]
; GFX8-NEXT: s_waitcnt vmcnt(0)
; GFX8-NEXT: v_lshrrev_b32_e32 v0, 8, v0
; GFX8-NEXT: s_setpc_b64 s[30:31]
;
; GFX7-LABEL: extractelement_vgpr_v16i8_idx1:
; GFX7: ; %bb.0:
; GFX7-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX7-NEXT: s_mov_b32 s6, 0
; GFX7-NEXT: s_mov_b32 s7, 0xf000
; GFX7-NEXT: s_mov_b64 s[4:5], 0
; GFX7-NEXT: buffer_load_dwordx4 v[0:3], v[0:1], s[4:7], 0 addr64
; GFX7-NEXT: s_waitcnt vmcnt(0)
; GFX7-NEXT: v_lshrrev_b32_e32 v0, 8, v0
; GFX7-NEXT: s_setpc_b64 s[30:31]
;
; GFX10-LABEL: extractelement_vgpr_v16i8_idx1:
; GFX10: ; %bb.0:
; GFX10-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX10-NEXT: global_load_dwordx4 v[0:3], v[0:1], off
; GFX10-NEXT: s_waitcnt vmcnt(0)
; GFX10-NEXT: v_lshrrev_b32_e32 v0, 8, v0
; GFX10-NEXT: s_setpc_b64 s[30:31]
;
; GFX11-LABEL: extractelement_vgpr_v16i8_idx1:
; GFX11: ; %bb.0:
; GFX11-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX11-NEXT: global_load_b128 v[0:3], v[0:1], off
; GFX11-NEXT: s_waitcnt vmcnt(0)
; GFX11-NEXT: v_lshrrev_b32_e32 v0, 8, v0
; GFX11-NEXT: s_setpc_b64 s[30:31]
%vector = load <16 x i8>, ptr addrspace(1) %ptr
%element = extractelement <16 x i8> %vector, i32 1
ret i8 %element
}
define i8 @extractelement_vgpr_v16i8_idx2(ptr addrspace(1) %ptr) {
; GFX9-LABEL: extractelement_vgpr_v16i8_idx2:
; GFX9: ; %bb.0:
; GFX9-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX9-NEXT: global_load_dwordx4 v[0:3], v[0:1], off
; GFX9-NEXT: s_waitcnt vmcnt(0)
; GFX9-NEXT: v_lshrrev_b32_e32 v0, 16, v0
; GFX9-NEXT: s_setpc_b64 s[30:31]
;
; GFX8-LABEL: extractelement_vgpr_v16i8_idx2:
; GFX8: ; %bb.0:
; GFX8-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX8-NEXT: flat_load_dwordx4 v[0:3], v[0:1]
; GFX8-NEXT: s_waitcnt vmcnt(0)
; GFX8-NEXT: v_lshrrev_b32_e32 v0, 16, v0
; GFX8-NEXT: s_setpc_b64 s[30:31]
;
; GFX7-LABEL: extractelement_vgpr_v16i8_idx2:
; GFX7: ; %bb.0:
; GFX7-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX7-NEXT: s_mov_b32 s6, 0
; GFX7-NEXT: s_mov_b32 s7, 0xf000
; GFX7-NEXT: s_mov_b64 s[4:5], 0
; GFX7-NEXT: buffer_load_dwordx4 v[0:3], v[0:1], s[4:7], 0 addr64
; GFX7-NEXT: s_waitcnt vmcnt(0)
; GFX7-NEXT: v_lshrrev_b32_e32 v0, 16, v0
; GFX7-NEXT: s_setpc_b64 s[30:31]
;
; GFX10-LABEL: extractelement_vgpr_v16i8_idx2:
; GFX10: ; %bb.0:
; GFX10-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX10-NEXT: global_load_dwordx4 v[0:3], v[0:1], off
; GFX10-NEXT: s_waitcnt vmcnt(0)
; GFX10-NEXT: v_lshrrev_b32_e32 v0, 16, v0
; GFX10-NEXT: s_setpc_b64 s[30:31]
;
; GFX11-LABEL: extractelement_vgpr_v16i8_idx2:
; GFX11: ; %bb.0:
; GFX11-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX11-NEXT: global_load_b128 v[0:3], v[0:1], off
; GFX11-NEXT: s_waitcnt vmcnt(0)
; GFX11-NEXT: v_lshrrev_b32_e32 v0, 16, v0
; GFX11-NEXT: s_setpc_b64 s[30:31]
%vector = load <16 x i8>, ptr addrspace(1) %ptr
%element = extractelement <16 x i8> %vector, i32 2
ret i8 %element
}
define i8 @extractelement_vgpr_v16i8_idx3(ptr addrspace(1) %ptr) {
; GFX9-LABEL: extractelement_vgpr_v16i8_idx3:
; GFX9: ; %bb.0:
; GFX9-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX9-NEXT: global_load_dwordx4 v[0:3], v[0:1], off
; GFX9-NEXT: s_waitcnt vmcnt(0)
; GFX9-NEXT: v_lshrrev_b32_e32 v0, 24, v0
; GFX9-NEXT: s_setpc_b64 s[30:31]
;
; GFX8-LABEL: extractelement_vgpr_v16i8_idx3:
; GFX8: ; %bb.0:
; GFX8-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX8-NEXT: flat_load_dwordx4 v[0:3], v[0:1]
; GFX8-NEXT: s_waitcnt vmcnt(0)
; GFX8-NEXT: v_lshrrev_b32_e32 v0, 24, v0
; GFX8-NEXT: s_setpc_b64 s[30:31]
;
; GFX7-LABEL: extractelement_vgpr_v16i8_idx3:
; GFX7: ; %bb.0:
; GFX7-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX7-NEXT: s_mov_b32 s6, 0
; GFX7-NEXT: s_mov_b32 s7, 0xf000
; GFX7-NEXT: s_mov_b64 s[4:5], 0
; GFX7-NEXT: buffer_load_dwordx4 v[0:3], v[0:1], s[4:7], 0 addr64
; GFX7-NEXT: s_waitcnt vmcnt(0)
; GFX7-NEXT: v_lshrrev_b32_e32 v0, 24, v0
; GFX7-NEXT: s_setpc_b64 s[30:31]
;
; GFX10-LABEL: extractelement_vgpr_v16i8_idx3:
; GFX10: ; %bb.0:
; GFX10-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX10-NEXT: global_load_dwordx4 v[0:3], v[0:1], off
; GFX10-NEXT: s_waitcnt vmcnt(0)
; GFX10-NEXT: v_lshrrev_b32_e32 v0, 24, v0
; GFX10-NEXT: s_setpc_b64 s[30:31]
;
; GFX11-LABEL: extractelement_vgpr_v16i8_idx3:
; GFX11: ; %bb.0:
; GFX11-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX11-NEXT: global_load_b128 v[0:3], v[0:1], off
; GFX11-NEXT: s_waitcnt vmcnt(0)
; GFX11-NEXT: v_lshrrev_b32_e32 v0, 24, v0
; GFX11-NEXT: s_setpc_b64 s[30:31]
%vector = load <16 x i8>, ptr addrspace(1) %ptr
%element = extractelement <16 x i8> %vector, i32 3
ret i8 %element
}
define i8 @extractelement_vgpr_v16i8_idx4(ptr addrspace(1) %ptr) {
; GFX9-LABEL: extractelement_vgpr_v16i8_idx4:
; GFX9: ; %bb.0:
; GFX9-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX9-NEXT: global_load_dwordx4 v[0:3], v[0:1], off
; GFX9-NEXT: s_waitcnt vmcnt(0)
; GFX9-NEXT: v_mov_b32_e32 v0, v1
; GFX9-NEXT: s_setpc_b64 s[30:31]
;
; GFX8-LABEL: extractelement_vgpr_v16i8_idx4:
; GFX8: ; %bb.0:
; GFX8-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX8-NEXT: flat_load_dwordx4 v[0:3], v[0:1]
; GFX8-NEXT: s_waitcnt vmcnt(0)
; GFX8-NEXT: v_mov_b32_e32 v0, v1
; GFX8-NEXT: s_setpc_b64 s[30:31]
;
; GFX7-LABEL: extractelement_vgpr_v16i8_idx4:
; GFX7: ; %bb.0:
; GFX7-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX7-NEXT: s_mov_b32 s6, 0
; GFX7-NEXT: s_mov_b32 s7, 0xf000
; GFX7-NEXT: s_mov_b64 s[4:5], 0
; GFX7-NEXT: buffer_load_dwordx4 v[0:3], v[0:1], s[4:7], 0 addr64
; GFX7-NEXT: s_waitcnt vmcnt(0)
; GFX7-NEXT: v_mov_b32_e32 v0, v1
; GFX7-NEXT: s_setpc_b64 s[30:31]
;
; GFX10-LABEL: extractelement_vgpr_v16i8_idx4:
; GFX10: ; %bb.0:
; GFX10-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX10-NEXT: global_load_dwordx4 v[0:3], v[0:1], off
; GFX10-NEXT: s_waitcnt vmcnt(0)
; GFX10-NEXT: v_mov_b32_e32 v0, v1
; GFX10-NEXT: s_setpc_b64 s[30:31]
;
; GFX11-LABEL: extractelement_vgpr_v16i8_idx4:
; GFX11: ; %bb.0:
; GFX11-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX11-NEXT: global_load_b128 v[0:3], v[0:1], off
; GFX11-NEXT: s_waitcnt vmcnt(0)
; GFX11-NEXT: v_mov_b32_e32 v0, v1
; GFX11-NEXT: s_setpc_b64 s[30:31]
%vector = load <16 x i8>, ptr addrspace(1) %ptr
%element = extractelement <16 x i8> %vector, i32 4
ret i8 %element
}
define i8 @extractelement_vgpr_v16i8_idx5(ptr addrspace(1) %ptr) {
; GFX9-LABEL: extractelement_vgpr_v16i8_idx5:
; GFX9: ; %bb.0:
; GFX9-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX9-NEXT: global_load_dwordx4 v[0:3], v[0:1], off
; GFX9-NEXT: s_waitcnt vmcnt(0)
; GFX9-NEXT: v_lshrrev_b32_e32 v0, 8, v1
; GFX9-NEXT: s_setpc_b64 s[30:31]
;
; GFX8-LABEL: extractelement_vgpr_v16i8_idx5:
; GFX8: ; %bb.0:
; GFX8-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX8-NEXT: flat_load_dwordx4 v[0:3], v[0:1]
; GFX8-NEXT: s_waitcnt vmcnt(0)
; GFX8-NEXT: v_lshrrev_b32_e32 v0, 8, v1
; GFX8-NEXT: s_setpc_b64 s[30:31]
;
; GFX7-LABEL: extractelement_vgpr_v16i8_idx5:
; GFX7: ; %bb.0:
; GFX7-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX7-NEXT: s_mov_b32 s6, 0
; GFX7-NEXT: s_mov_b32 s7, 0xf000
; GFX7-NEXT: s_mov_b64 s[4:5], 0
; GFX7-NEXT: buffer_load_dwordx4 v[0:3], v[0:1], s[4:7], 0 addr64
; GFX7-NEXT: s_waitcnt vmcnt(0)
; GFX7-NEXT: v_lshrrev_b32_e32 v0, 8, v1
; GFX7-NEXT: s_setpc_b64 s[30:31]
;
; GFX10-LABEL: extractelement_vgpr_v16i8_idx5:
; GFX10: ; %bb.0:
; GFX10-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX10-NEXT: global_load_dwordx4 v[0:3], v[0:1], off
; GFX10-NEXT: s_waitcnt vmcnt(0)
; GFX10-NEXT: v_lshrrev_b32_e32 v0, 8, v1
; GFX10-NEXT: s_setpc_b64 s[30:31]
;
; GFX11-LABEL: extractelement_vgpr_v16i8_idx5:
; GFX11: ; %bb.0:
; GFX11-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX11-NEXT: global_load_b128 v[0:3], v[0:1], off
; GFX11-NEXT: s_waitcnt vmcnt(0)
; GFX11-NEXT: v_lshrrev_b32_e32 v0, 8, v1
; GFX11-NEXT: s_setpc_b64 s[30:31]
%vector = load <16 x i8>, ptr addrspace(1) %ptr
%element = extractelement <16 x i8> %vector, i32 5
ret i8 %element
}
define i8 @extractelement_vgpr_v16i8_idx6(ptr addrspace(1) %ptr) {
; GFX9-LABEL: extractelement_vgpr_v16i8_idx6:
; GFX9: ; %bb.0:
; GFX9-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX9-NEXT: global_load_dwordx4 v[0:3], v[0:1], off
; GFX9-NEXT: s_waitcnt vmcnt(0)
; GFX9-NEXT: v_lshrrev_b32_e32 v0, 16, v1
; GFX9-NEXT: s_setpc_b64 s[30:31]
;
; GFX8-LABEL: extractelement_vgpr_v16i8_idx6:
; GFX8: ; %bb.0:
; GFX8-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX8-NEXT: flat_load_dwordx4 v[0:3], v[0:1]
; GFX8-NEXT: s_waitcnt vmcnt(0)
; GFX8-NEXT: v_lshrrev_b32_e32 v0, 16, v1
; GFX8-NEXT: s_setpc_b64 s[30:31]
;
; GFX7-LABEL: extractelement_vgpr_v16i8_idx6:
; GFX7: ; %bb.0:
; GFX7-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX7-NEXT: s_mov_b32 s6, 0
; GFX7-NEXT: s_mov_b32 s7, 0xf000
; GFX7-NEXT: s_mov_b64 s[4:5], 0
; GFX7-NEXT: buffer_load_dwordx4 v[0:3], v[0:1], s[4:7], 0 addr64
; GFX7-NEXT: s_waitcnt vmcnt(0)
; GFX7-NEXT: v_lshrrev_b32_e32 v0, 16, v1
; GFX7-NEXT: s_setpc_b64 s[30:31]
;
; GFX10-LABEL: extractelement_vgpr_v16i8_idx6:
; GFX10: ; %bb.0:
; GFX10-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX10-NEXT: global_load_dwordx4 v[0:3], v[0:1], off
; GFX10-NEXT: s_waitcnt vmcnt(0)
; GFX10-NEXT: v_lshrrev_b32_e32 v0, 16, v1
; GFX10-NEXT: s_setpc_b64 s[30:31]
;
; GFX11-LABEL: extractelement_vgpr_v16i8_idx6:
; GFX11: ; %bb.0:
; GFX11-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX11-NEXT: global_load_b128 v[0:3], v[0:1], off
; GFX11-NEXT: s_waitcnt vmcnt(0)
; GFX11-NEXT: v_lshrrev_b32_e32 v0, 16, v1
; GFX11-NEXT: s_setpc_b64 s[30:31]
%vector = load <16 x i8>, ptr addrspace(1) %ptr
%element = extractelement <16 x i8> %vector, i32 6
ret i8 %element
}
define i8 @extractelement_vgpr_v16i8_idx7(ptr addrspace(1) %ptr) {
; GFX9-LABEL: extractelement_vgpr_v16i8_idx7:
; GFX9: ; %bb.0:
; GFX9-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX9-NEXT: global_load_dwordx4 v[0:3], v[0:1], off
; GFX9-NEXT: s_waitcnt vmcnt(0)
; GFX9-NEXT: v_lshrrev_b32_e32 v0, 24, v1
; GFX9-NEXT: s_setpc_b64 s[30:31]
;
; GFX8-LABEL: extractelement_vgpr_v16i8_idx7:
; GFX8: ; %bb.0:
; GFX8-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX8-NEXT: flat_load_dwordx4 v[0:3], v[0:1]
; GFX8-NEXT: s_waitcnt vmcnt(0)
; GFX8-NEXT: v_lshrrev_b32_e32 v0, 24, v1
; GFX8-NEXT: s_setpc_b64 s[30:31]
;
; GFX7-LABEL: extractelement_vgpr_v16i8_idx7:
; GFX7: ; %bb.0:
; GFX7-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX7-NEXT: s_mov_b32 s6, 0
; GFX7-NEXT: s_mov_b32 s7, 0xf000
; GFX7-NEXT: s_mov_b64 s[4:5], 0
; GFX7-NEXT: buffer_load_dwordx4 v[0:3], v[0:1], s[4:7], 0 addr64
; GFX7-NEXT: s_waitcnt vmcnt(0)
; GFX7-NEXT: v_lshrrev_b32_e32 v0, 24, v1
; GFX7-NEXT: s_setpc_b64 s[30:31]
;
; GFX10-LABEL: extractelement_vgpr_v16i8_idx7:
; GFX10: ; %bb.0:
; GFX10-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX10-NEXT: global_load_dwordx4 v[0:3], v[0:1], off
; GFX10-NEXT: s_waitcnt vmcnt(0)
; GFX10-NEXT: v_lshrrev_b32_e32 v0, 24, v1
; GFX10-NEXT: s_setpc_b64 s[30:31]
;
; GFX11-LABEL: extractelement_vgpr_v16i8_idx7:
; GFX11: ; %bb.0:
; GFX11-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX11-NEXT: global_load_b128 v[0:3], v[0:1], off
; GFX11-NEXT: s_waitcnt vmcnt(0)
; GFX11-NEXT: v_lshrrev_b32_e32 v0, 24, v1
; GFX11-NEXT: s_setpc_b64 s[30:31]
%vector = load <16 x i8>, ptr addrspace(1) %ptr
%element = extractelement <16 x i8> %vector, i32 7
ret i8 %element
}
define i8 @extractelement_vgpr_v16i8_idx8(ptr addrspace(1) %ptr) {
; GFX9-LABEL: extractelement_vgpr_v16i8_idx8:
; GFX9: ; %bb.0:
; GFX9-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX9-NEXT: global_load_dwordx4 v[0:3], v[0:1], off
; GFX9-NEXT: s_waitcnt vmcnt(0)
; GFX9-NEXT: v_mov_b32_e32 v0, v2
; GFX9-NEXT: s_setpc_b64 s[30:31]
;
; GFX8-LABEL: extractelement_vgpr_v16i8_idx8:
; GFX8: ; %bb.0:
; GFX8-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX8-NEXT: flat_load_dwordx4 v[0:3], v[0:1]
; GFX8-NEXT: s_waitcnt vmcnt(0)
; GFX8-NEXT: v_mov_b32_e32 v0, v2
; GFX8-NEXT: s_setpc_b64 s[30:31]
;
; GFX7-LABEL: extractelement_vgpr_v16i8_idx8:
; GFX7: ; %bb.0:
; GFX7-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX7-NEXT: s_mov_b32 s6, 0
; GFX7-NEXT: s_mov_b32 s7, 0xf000
; GFX7-NEXT: s_mov_b64 s[4:5], 0
; GFX7-NEXT: buffer_load_dwordx4 v[0:3], v[0:1], s[4:7], 0 addr64
; GFX7-NEXT: s_waitcnt vmcnt(0)
; GFX7-NEXT: v_mov_b32_e32 v0, v2
; GFX7-NEXT: s_setpc_b64 s[30:31]
;
; GFX10-LABEL: extractelement_vgpr_v16i8_idx8:
; GFX10: ; %bb.0:
; GFX10-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX10-NEXT: global_load_dwordx4 v[0:3], v[0:1], off
; GFX10-NEXT: s_waitcnt vmcnt(0)
; GFX10-NEXT: v_mov_b32_e32 v0, v2
; GFX10-NEXT: s_setpc_b64 s[30:31]
;
; GFX11-LABEL: extractelement_vgpr_v16i8_idx8:
; GFX11: ; %bb.0:
; GFX11-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX11-NEXT: global_load_b128 v[0:3], v[0:1], off
; GFX11-NEXT: s_waitcnt vmcnt(0)
; GFX11-NEXT: v_mov_b32_e32 v0, v2
; GFX11-NEXT: s_setpc_b64 s[30:31]
%vector = load <16 x i8>, ptr addrspace(1) %ptr
%element = extractelement <16 x i8> %vector, i32 8
ret i8 %element
}
define i8 @extractelement_vgpr_v16i8_idx9(ptr addrspace(1) %ptr) {
; GFX9-LABEL: extractelement_vgpr_v16i8_idx9:
; GFX9: ; %bb.0:
; GFX9-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX9-NEXT: global_load_dwordx4 v[0:3], v[0:1], off
; GFX9-NEXT: s_waitcnt vmcnt(0)
; GFX9-NEXT: v_lshrrev_b32_e32 v0, 8, v2
; GFX9-NEXT: s_setpc_b64 s[30:31]
;
; GFX8-LABEL: extractelement_vgpr_v16i8_idx9:
; GFX8: ; %bb.0:
; GFX8-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX8-NEXT: flat_load_dwordx4 v[0:3], v[0:1]
; GFX8-NEXT: s_waitcnt vmcnt(0)
; GFX8-NEXT: v_lshrrev_b32_e32 v0, 8, v2
; GFX8-NEXT: s_setpc_b64 s[30:31]
;
; GFX7-LABEL: extractelement_vgpr_v16i8_idx9:
; GFX7: ; %bb.0:
; GFX7-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX7-NEXT: s_mov_b32 s6, 0
; GFX7-NEXT: s_mov_b32 s7, 0xf000
; GFX7-NEXT: s_mov_b64 s[4:5], 0
; GFX7-NEXT: buffer_load_dwordx4 v[0:3], v[0:1], s[4:7], 0 addr64
; GFX7-NEXT: s_waitcnt vmcnt(0)
; GFX7-NEXT: v_lshrrev_b32_e32 v0, 8, v2
; GFX7-NEXT: s_setpc_b64 s[30:31]
;
; GFX10-LABEL: extractelement_vgpr_v16i8_idx9:
; GFX10: ; %bb.0:
; GFX10-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX10-NEXT: global_load_dwordx4 v[0:3], v[0:1], off
; GFX10-NEXT: s_waitcnt vmcnt(0)
; GFX10-NEXT: v_lshrrev_b32_e32 v0, 8, v2
; GFX10-NEXT: s_setpc_b64 s[30:31]
;
; GFX11-LABEL: extractelement_vgpr_v16i8_idx9:
; GFX11: ; %bb.0:
; GFX11-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX11-NEXT: global_load_b128 v[0:3], v[0:1], off
; GFX11-NEXT: s_waitcnt vmcnt(0)
; GFX11-NEXT: v_lshrrev_b32_e32 v0, 8, v2
; GFX11-NEXT: s_setpc_b64 s[30:31]
%vector = load <16 x i8>, ptr addrspace(1) %ptr
%element = extractelement <16 x i8> %vector, i32 9
ret i8 %element
}
define i8 @extractelement_vgpr_v16i8_idx10(ptr addrspace(1) %ptr) {
; GFX9-LABEL: extractelement_vgpr_v16i8_idx10:
; GFX9: ; %bb.0:
; GFX9-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX9-NEXT: global_load_dwordx4 v[0:3], v[0:1], off
; GFX9-NEXT: s_waitcnt vmcnt(0)
; GFX9-NEXT: v_lshrrev_b32_e32 v0, 16, v2
; GFX9-NEXT: s_setpc_b64 s[30:31]
;
; GFX8-LABEL: extractelement_vgpr_v16i8_idx10:
; GFX8: ; %bb.0:
; GFX8-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX8-NEXT: flat_load_dwordx4 v[0:3], v[0:1]
; GFX8-NEXT: s_waitcnt vmcnt(0)
; GFX8-NEXT: v_lshrrev_b32_e32 v0, 16, v2
; GFX8-NEXT: s_setpc_b64 s[30:31]
;
; GFX7-LABEL: extractelement_vgpr_v16i8_idx10:
; GFX7: ; %bb.0:
; GFX7-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX7-NEXT: s_mov_b32 s6, 0
; GFX7-NEXT: s_mov_b32 s7, 0xf000
; GFX7-NEXT: s_mov_b64 s[4:5], 0
; GFX7-NEXT: buffer_load_dwordx4 v[0:3], v[0:1], s[4:7], 0 addr64
; GFX7-NEXT: s_waitcnt vmcnt(0)
; GFX7-NEXT: v_lshrrev_b32_e32 v0, 16, v2
; GFX7-NEXT: s_setpc_b64 s[30:31]
;
; GFX10-LABEL: extractelement_vgpr_v16i8_idx10:
; GFX10: ; %bb.0:
; GFX10-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX10-NEXT: global_load_dwordx4 v[0:3], v[0:1], off
; GFX10-NEXT: s_waitcnt vmcnt(0)
; GFX10-NEXT: v_lshrrev_b32_e32 v0, 16, v2
; GFX10-NEXT: s_setpc_b64 s[30:31]
;
; GFX11-LABEL: extractelement_vgpr_v16i8_idx10:
; GFX11: ; %bb.0:
; GFX11-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX11-NEXT: global_load_b128 v[0:3], v[0:1], off
; GFX11-NEXT: s_waitcnt vmcnt(0)
; GFX11-NEXT: v_lshrrev_b32_e32 v0, 16, v2
; GFX11-NEXT: s_setpc_b64 s[30:31]
%vector = load <16 x i8>, ptr addrspace(1) %ptr
%element = extractelement <16 x i8> %vector, i32 10
ret i8 %element
}
define i8 @extractelement_vgpr_v16i8_idx11(ptr addrspace(1) %ptr) {
; GFX9-LABEL: extractelement_vgpr_v16i8_idx11:
; GFX9: ; %bb.0:
; GFX9-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX9-NEXT: global_load_dwordx4 v[0:3], v[0:1], off
; GFX9-NEXT: s_waitcnt vmcnt(0)
; GFX9-NEXT: v_lshrrev_b32_e32 v0, 24, v2
; GFX9-NEXT: s_setpc_b64 s[30:31]
;
; GFX8-LABEL: extractelement_vgpr_v16i8_idx11:
; GFX8: ; %bb.0:
; GFX8-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX8-NEXT: flat_load_dwordx4 v[0:3], v[0:1]
; GFX8-NEXT: s_waitcnt vmcnt(0)
; GFX8-NEXT: v_lshrrev_b32_e32 v0, 24, v2
; GFX8-NEXT: s_setpc_b64 s[30:31]
;
; GFX7-LABEL: extractelement_vgpr_v16i8_idx11:
; GFX7: ; %bb.0:
; GFX7-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX7-NEXT: s_mov_b32 s6, 0
; GFX7-NEXT: s_mov_b32 s7, 0xf000
; GFX7-NEXT: s_mov_b64 s[4:5], 0
; GFX7-NEXT: buffer_load_dwordx4 v[0:3], v[0:1], s[4:7], 0 addr64
; GFX7-NEXT: s_waitcnt vmcnt(0)
; GFX7-NEXT: v_lshrrev_b32_e32 v0, 24, v2
; GFX7-NEXT: s_setpc_b64 s[30:31]
;
; GFX10-LABEL: extractelement_vgpr_v16i8_idx11:
; GFX10: ; %bb.0:
; GFX10-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX10-NEXT: global_load_dwordx4 v[0:3], v[0:1], off
; GFX10-NEXT: s_waitcnt vmcnt(0)
; GFX10-NEXT: v_lshrrev_b32_e32 v0, 24, v2
; GFX10-NEXT: s_setpc_b64 s[30:31]
;
; GFX11-LABEL: extractelement_vgpr_v16i8_idx11:
; GFX11: ; %bb.0:
; GFX11-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX11-NEXT: global_load_b128 v[0:3], v[0:1], off
; GFX11-NEXT: s_waitcnt vmcnt(0)
; GFX11-NEXT: v_lshrrev_b32_e32 v0, 24, v2
; GFX11-NEXT: s_setpc_b64 s[30:31]
%vector = load <16 x i8>, ptr addrspace(1) %ptr
%element = extractelement <16 x i8> %vector, i32 11
ret i8 %element
}
define i8 @extractelement_vgpr_v16i8_idx12(ptr addrspace(1) %ptr) {
; GFX9-LABEL: extractelement_vgpr_v16i8_idx12:
; GFX9: ; %bb.0:
; GFX9-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX9-NEXT: global_load_dwordx4 v[0:3], v[0:1], off
; GFX9-NEXT: s_waitcnt vmcnt(0)
; GFX9-NEXT: v_mov_b32_e32 v0, v3
; GFX9-NEXT: s_setpc_b64 s[30:31]
;
; GFX8-LABEL: extractelement_vgpr_v16i8_idx12:
; GFX8: ; %bb.0:
; GFX8-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX8-NEXT: flat_load_dwordx4 v[0:3], v[0:1]
; GFX8-NEXT: s_waitcnt vmcnt(0)
; GFX8-NEXT: v_mov_b32_e32 v0, v3
; GFX8-NEXT: s_setpc_b64 s[30:31]
;
; GFX7-LABEL: extractelement_vgpr_v16i8_idx12:
; GFX7: ; %bb.0:
; GFX7-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX7-NEXT: s_mov_b32 s6, 0
; GFX7-NEXT: s_mov_b32 s7, 0xf000
; GFX7-NEXT: s_mov_b64 s[4:5], 0
; GFX7-NEXT: buffer_load_dwordx4 v[0:3], v[0:1], s[4:7], 0 addr64
; GFX7-NEXT: s_waitcnt vmcnt(0)
; GFX7-NEXT: v_mov_b32_e32 v0, v3
; GFX7-NEXT: s_setpc_b64 s[30:31]
;
; GFX10-LABEL: extractelement_vgpr_v16i8_idx12:
; GFX10: ; %bb.0:
; GFX10-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX10-NEXT: global_load_dwordx4 v[0:3], v[0:1], off
; GFX10-NEXT: s_waitcnt vmcnt(0)
; GFX10-NEXT: v_mov_b32_e32 v0, v3
; GFX10-NEXT: s_setpc_b64 s[30:31]
;
; GFX11-LABEL: extractelement_vgpr_v16i8_idx12:
; GFX11: ; %bb.0:
; GFX11-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX11-NEXT: global_load_b128 v[0:3], v[0:1], off
; GFX11-NEXT: s_waitcnt vmcnt(0)
; GFX11-NEXT: v_mov_b32_e32 v0, v3
; GFX11-NEXT: s_setpc_b64 s[30:31]
%vector = load <16 x i8>, ptr addrspace(1) %ptr
%element = extractelement <16 x i8> %vector, i32 12
ret i8 %element
}
define i8 @extractelement_vgpr_v16i8_idx13(ptr addrspace(1) %ptr) {
; GFX9-LABEL: extractelement_vgpr_v16i8_idx13:
; GFX9: ; %bb.0:
; GFX9-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX9-NEXT: global_load_dwordx4 v[0:3], v[0:1], off
; GFX9-NEXT: s_waitcnt vmcnt(0)
; GFX9-NEXT: v_lshrrev_b32_e32 v0, 8, v3
; GFX9-NEXT: s_setpc_b64 s[30:31]
;
; GFX8-LABEL: extractelement_vgpr_v16i8_idx13:
; GFX8: ; %bb.0:
; GFX8-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX8-NEXT: flat_load_dwordx4 v[0:3], v[0:1]
; GFX8-NEXT: s_waitcnt vmcnt(0)
; GFX8-NEXT: v_lshrrev_b32_e32 v0, 8, v3
; GFX8-NEXT: s_setpc_b64 s[30:31]
;
; GFX7-LABEL: extractelement_vgpr_v16i8_idx13:
; GFX7: ; %bb.0:
; GFX7-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX7-NEXT: s_mov_b32 s6, 0
; GFX7-NEXT: s_mov_b32 s7, 0xf000
; GFX7-NEXT: s_mov_b64 s[4:5], 0
; GFX7-NEXT: buffer_load_dwordx4 v[0:3], v[0:1], s[4:7], 0 addr64
; GFX7-NEXT: s_waitcnt vmcnt(0)
; GFX7-NEXT: v_lshrrev_b32_e32 v0, 8, v3
; GFX7-NEXT: s_setpc_b64 s[30:31]
;
; GFX10-LABEL: extractelement_vgpr_v16i8_idx13:
; GFX10: ; %bb.0:
; GFX10-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX10-NEXT: global_load_dwordx4 v[0:3], v[0:1], off
; GFX10-NEXT: s_waitcnt vmcnt(0)
; GFX10-NEXT: v_lshrrev_b32_e32 v0, 8, v3
; GFX10-NEXT: s_setpc_b64 s[30:31]
;
; GFX11-LABEL: extractelement_vgpr_v16i8_idx13:
; GFX11: ; %bb.0:
; GFX11-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX11-NEXT: global_load_b128 v[0:3], v[0:1], off
; GFX11-NEXT: s_waitcnt vmcnt(0)
; GFX11-NEXT: v_lshrrev_b32_e32 v0, 8, v3
; GFX11-NEXT: s_setpc_b64 s[30:31]
%vector = load <16 x i8>, ptr addrspace(1) %ptr
%element = extractelement <16 x i8> %vector, i32 13
ret i8 %element
}
define i8 @extractelement_vgpr_v16i8_idx14(ptr addrspace(1) %ptr) {
; GFX9-LABEL: extractelement_vgpr_v16i8_idx14:
; GFX9: ; %bb.0:
; GFX9-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX9-NEXT: global_load_dwordx4 v[0:3], v[0:1], off
; GFX9-NEXT: s_waitcnt vmcnt(0)
; GFX9-NEXT: v_lshrrev_b32_e32 v0, 16, v3
; GFX9-NEXT: s_setpc_b64 s[30:31]
;
; GFX8-LABEL: extractelement_vgpr_v16i8_idx14:
; GFX8: ; %bb.0:
; GFX8-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX8-NEXT: flat_load_dwordx4 v[0:3], v[0:1]
; GFX8-NEXT: s_waitcnt vmcnt(0)
; GFX8-NEXT: v_lshrrev_b32_e32 v0, 16, v3
; GFX8-NEXT: s_setpc_b64 s[30:31]
;
; GFX7-LABEL: extractelement_vgpr_v16i8_idx14:
; GFX7: ; %bb.0:
; GFX7-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX7-NEXT: s_mov_b32 s6, 0
; GFX7-NEXT: s_mov_b32 s7, 0xf000
; GFX7-NEXT: s_mov_b64 s[4:5], 0
; GFX7-NEXT: buffer_load_dwordx4 v[0:3], v[0:1], s[4:7], 0 addr64
; GFX7-NEXT: s_waitcnt vmcnt(0)
; GFX7-NEXT: v_lshrrev_b32_e32 v0, 16, v3
; GFX7-NEXT: s_setpc_b64 s[30:31]
;
; GFX10-LABEL: extractelement_vgpr_v16i8_idx14:
; GFX10: ; %bb.0:
; GFX10-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX10-NEXT: global_load_dwordx4 v[0:3], v[0:1], off
; GFX10-NEXT: s_waitcnt vmcnt(0)
; GFX10-NEXT: v_lshrrev_b32_e32 v0, 16, v3
; GFX10-NEXT: s_setpc_b64 s[30:31]
;
; GFX11-LABEL: extractelement_vgpr_v16i8_idx14:
; GFX11: ; %bb.0:
; GFX11-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX11-NEXT: global_load_b128 v[0:3], v[0:1], off
; GFX11-NEXT: s_waitcnt vmcnt(0)
; GFX11-NEXT: v_lshrrev_b32_e32 v0, 16, v3
; GFX11-NEXT: s_setpc_b64 s[30:31]
%vector = load <16 x i8>, ptr addrspace(1) %ptr
%element = extractelement <16 x i8> %vector, i32 14
ret i8 %element
}
define i8 @extractelement_vgpr_v16i8_idx15(ptr addrspace(1) %ptr) {
; GFX9-LABEL: extractelement_vgpr_v16i8_idx15:
; GFX9: ; %bb.0:
; GFX9-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX9-NEXT: global_load_dwordx4 v[0:3], v[0:1], off
; GFX9-NEXT: s_waitcnt vmcnt(0)
; GFX9-NEXT: v_lshrrev_b32_e32 v0, 24, v3
; GFX9-NEXT: s_setpc_b64 s[30:31]
;
; GFX8-LABEL: extractelement_vgpr_v16i8_idx15:
; GFX8: ; %bb.0:
; GFX8-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX8-NEXT: flat_load_dwordx4 v[0:3], v[0:1]
; GFX8-NEXT: s_waitcnt vmcnt(0)
; GFX8-NEXT: v_lshrrev_b32_e32 v0, 24, v3
; GFX8-NEXT: s_setpc_b64 s[30:31]
;
; GFX7-LABEL: extractelement_vgpr_v16i8_idx15:
; GFX7: ; %bb.0:
; GFX7-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX7-NEXT: s_mov_b32 s6, 0
; GFX7-NEXT: s_mov_b32 s7, 0xf000
; GFX7-NEXT: s_mov_b64 s[4:5], 0
; GFX7-NEXT: buffer_load_dwordx4 v[0:3], v[0:1], s[4:7], 0 addr64
; GFX7-NEXT: s_waitcnt vmcnt(0)
; GFX7-NEXT: v_lshrrev_b32_e32 v0, 24, v3
; GFX7-NEXT: s_setpc_b64 s[30:31]
;
; GFX10-LABEL: extractelement_vgpr_v16i8_idx15:
; GFX10: ; %bb.0:
; GFX10-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX10-NEXT: global_load_dwordx4 v[0:3], v[0:1], off
; GFX10-NEXT: s_waitcnt vmcnt(0)
; GFX10-NEXT: v_lshrrev_b32_e32 v0, 24, v3
; GFX10-NEXT: s_setpc_b64 s[30:31]
;
; GFX11-LABEL: extractelement_vgpr_v16i8_idx15:
; GFX11: ; %bb.0:
; GFX11-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX11-NEXT: global_load_b128 v[0:3], v[0:1], off
; GFX11-NEXT: s_waitcnt vmcnt(0)
; GFX11-NEXT: v_lshrrev_b32_e32 v0, 24, v3
; GFX11-NEXT: s_setpc_b64 s[30:31]
%vector = load <16 x i8>, ptr addrspace(1) %ptr
%element = extractelement <16 x i8> %vector, i32 15
ret i8 %element
}
Reference in New Issue View Git Blame Copy Permalink