1 ; RUN: llc -march=r600 -mcpu=redwood < %s | FileCheck --check-prefix=R600 -check-prefix=FUNC %s
2 ; RUN: llc -march=amdgcn -verify-machineinstrs < %s | FileCheck -check-prefix=SI -check-prefix=FUNC %s
3 ; RUN: llc -march=amdgcn -mcpu=tonga -verify-machineinstrs < %s | FileCheck -check-prefix=SI -check-prefix=FUNC %s
5 ; FUNC-LABEL: {{^}}rotr_i32:
9 define amdgpu_kernel void @rotr_i32(i32 addrspace(1)* %in, i32 %x, i32 %y) {
11 %tmp0 = sub i32 32, %y
12 %tmp1 = shl i32 %x, %tmp0
13 %tmp2 = lshr i32 %x, %y
14 %tmp3 = or i32 %tmp1, %tmp2
15 store i32 %tmp3, i32 addrspace(1)* %in
19 ; FUNC-LABEL: {{^}}rotr_v2i32:
25 define amdgpu_kernel void @rotr_v2i32(<2 x i32> addrspace(1)* %in, <2 x i32> %x, <2 x i32> %y) {
27 %tmp0 = sub <2 x i32> <i32 32, i32 32>, %y
28 %tmp1 = shl <2 x i32> %x, %tmp0
29 %tmp2 = lshr <2 x i32> %x, %y
30 %tmp3 = or <2 x i32> %tmp1, %tmp2
31 store <2 x i32> %tmp3, <2 x i32> addrspace(1)* %in
35 ; FUNC-LABEL: {{^}}rotr_v4i32:
45 define amdgpu_kernel void @rotr_v4i32(<4 x i32> addrspace(1)* %in, <4 x i32> %x, <4 x i32> %y) {
47 %tmp0 = sub <4 x i32> <i32 32, i32 32, i32 32, i32 32>, %y
48 %tmp1 = shl <4 x i32> %x, %tmp0
49 %tmp2 = lshr <4 x i32> %x, %y
50 %tmp3 = or <4 x i32> %tmp1, %tmp2
51 store <4 x i32> %tmp3, <4 x i32> addrspace(1)* %in
55 ; GCN-LABEL: @test_rotr_i16
56 ; GCN: global_load_ushort [[X:v[0-9]+]]
57 ; GCN: global_load_ushort [[D:v[0-9]+]]
58 ; GCN: v_sub_nc_u16_e64 [[NX:v[0-9]+]], 0, [[X]]
59 ; GCN: v_and_b32_e32 [[XAND:v[0-9]+]], 15, [[X]]
60 ; GCN: v_and_b32_e32 [[NXAND:v[0-9]+]], 15, [[NX]]
61 ; GCN: v_lshrrev_b16_e64 [[LO:v[0-9]+]], [[XAND]], [[D]]
62 ; GCN: v_lshlrev_b16_e64 [[HI:v[0-9]+]], [[NXAND]], [[D]]
63 ; GCN: v_or_b32_e32 [[RES:v[0-9]+]], [[LO]], [[HI]]
64 ; GCN: global_store_short v{{\[[0-9]+:[0-9]+\]}}, [[RES]]
66 declare i16 @llvm.fshr.i16(i16, i16, i16)
68 define void @test_rotr_i16(i16 addrspace(1)* nocapture readonly %sourceA, i16 addrspace(1)* nocapture readonly %sourceB, i16 addrspace(1)* nocapture %destValues) {
70 %arrayidx = getelementptr inbounds i16, i16 addrspace(1)* %sourceA, i64 16
71 %a = load i16, i16 addrspace(1)* %arrayidx
72 %arrayidx2 = getelementptr inbounds i16, i16 addrspace(1)* %sourceB, i64 24
73 %b = load i16, i16 addrspace(1)* %arrayidx2
74 %c = tail call i16 @llvm.fshr.i16(i16 %a, i16 %a, i16 %b)
75 %arrayidx5 = getelementptr inbounds i16, i16 addrspace(1)* %destValues, i64 4
76 store i16 %c, i16 addrspace(1)* %arrayidx5