| blob | f284df4d8a70b1bc1b84a99abf861911b6c1bb24 |
1 ; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
2 ; RUN: llc -march=amdgcn -verify-machineinstrs < %s | FileCheck -check-prefix=GCN %s
4 define amdgpu_kernel void @negated_cond(ptr addrspace(1) %arg1) {
5 ; GCN-LABEL: negated_cond:
6 ; GCN: ; %bb.0: ; %bb
7 ; GCN-NEXT: s_load_dwordx2 s[4:5], s[0:1], 0x9
8 ; GCN-NEXT: s_mov_b32 s7, 0xf000
9 ; GCN-NEXT: s_mov_b32 s10, -1
10 ; GCN-NEXT: s_mov_b32 s6, 0
11 ; GCN-NEXT: s_waitcnt lgkmcnt(0)
12 ; GCN-NEXT: s_mov_b64 s[8:9], s[4:5]
13 ; GCN-NEXT: v_mov_b32_e32 v0, 0
14 ; GCN-NEXT: s_branch .LBB0_2
15 ; GCN-NEXT: .LBB0_1: ; %loop.exit.guard
16 ; GCN-NEXT: ; in Loop: Header=BB0_2 Depth=1
17 ; GCN-NEXT: s_and_b64 vcc, exec, s[14:15]
18 ; GCN-NEXT: s_cbranch_vccnz .LBB0_9
19 ; GCN-NEXT: .LBB0_2: ; %bb1
20 ; GCN-NEXT: ; =>This Loop Header: Depth=1
21 ; GCN-NEXT: ; Child Loop BB0_4 Depth 2
22 ; GCN-NEXT: s_mov_b32 s11, s7
23 ; GCN-NEXT: buffer_load_dword v1, off, s[8:11], 0
24 ; GCN-NEXT: s_waitcnt vmcnt(0)
25 ; GCN-NEXT: v_cmp_ne_u32_e64 s[2:3], 0, v1
26 ; GCN-NEXT: v_cmp_eq_u32_e32 vcc, 0, v1
27 ; GCN-NEXT: v_cndmask_b32_e64 v1, 0, 1, vcc
28 ; GCN-NEXT: v_cmp_ne_u32_e64 s[0:1], 1, v1
29 ; GCN-NEXT: s_mov_b32 s12, s6
30 ; GCN-NEXT: s_branch .LBB0_4
31 ; GCN-NEXT: .LBB0_3: ; %Flow1
32 ; GCN-NEXT: ; in Loop: Header=BB0_4 Depth=2
33 ; GCN-NEXT: s_andn2_b64 vcc, exec, s[16:17]
34 ; GCN-NEXT: s_cbranch_vccz .LBB0_1
35 ; GCN-NEXT: .LBB0_4: ; %bb2
36 ; GCN-NEXT: ; Parent Loop BB0_2 Depth=1
37 ; GCN-NEXT: ; => This Inner Loop Header: Depth=2
38 ; GCN-NEXT: s_and_b64 vcc, exec, s[0:1]
39 ; GCN-NEXT: s_lshl_b32 s12, s12, 5
40 ; GCN-NEXT: s_cbranch_vccz .LBB0_6
41 ; GCN-NEXT: ; %bb.5: ; in Loop: Header=BB0_4 Depth=2
42 ; GCN-NEXT: s_mov_b64 s[14:15], s[2:3]
43 ; GCN-NEXT: s_branch .LBB0_7
44 ; GCN-NEXT: .LBB0_6: ; %bb3
45 ; GCN-NEXT: ; in Loop: Header=BB0_4 Depth=2
46 ; GCN-NEXT: s_add_i32 s12, s12, 1
47 ; GCN-NEXT: s_mov_b64 s[14:15], -1
48 ; GCN-NEXT: .LBB0_7: ; %Flow
49 ; GCN-NEXT: ; in Loop: Header=BB0_4 Depth=2
50 ; GCN-NEXT: s_andn2_b64 vcc, exec, s[14:15]
51 ; GCN-NEXT: s_mov_b64 s[16:17], -1
52 ; GCN-NEXT: s_cbranch_vccnz .LBB0_3
53 ; GCN-NEXT: ; %bb.8: ; %bb4
54 ; GCN-NEXT: ; in Loop: Header=BB0_4 Depth=2
55 ; GCN-NEXT: s_ashr_i32 s13, s12, 31
56 ; GCN-NEXT: s_lshl_b64 s[16:17], s[12:13], 2
57 ; GCN-NEXT: s_mov_b64 s[14:15], 0
58 ; GCN-NEXT: v_mov_b32_e32 v1, s16
59 ; GCN-NEXT: v_mov_b32_e32 v2, s17
60 ; GCN-NEXT: buffer_store_dword v0, v[1:2], s[4:7], 0 addr64
61 ; GCN-NEXT: s_cmp_eq_u32 s12, 32
62 ; GCN-NEXT: s_cselect_b64 s[16:17], -1, 0
63 ; GCN-NEXT: s_branch .LBB0_3
64 ; GCN-NEXT: .LBB0_9: ; %DummyReturnBlock
65 ; GCN-NEXT: s_endpgm
66 bb:
67 br label %bb1
69 bb1:
70 %tmp1 = load i32, ptr addrspace(1) %arg1
71 %tmp2 = icmp eq i32 %tmp1, 0
72 br label %bb2
74 bb2:
75 %tmp3 = phi i32 [ 0, %bb1 ], [ %tmp6, %bb4 ]
76 %tmp4 = shl i32 %tmp3, 5
77 br i1 %tmp2, label %bb3, label %bb4
79 bb3:
80 %tmp5 = add i32 %tmp4, 1
81 br label %bb4
83 bb4:
84 %tmp6 = phi i32 [ %tmp5, %bb3 ], [ %tmp4, %bb2 ]
85 %gep = getelementptr inbounds i32, ptr addrspace(1) %arg1, i32 %tmp6
86 store i32 0, ptr addrspace(1) %gep
87 %tmp7 = icmp eq i32 %tmp6, 32
88 br i1 %tmp7, label %bb1, label %bb2
89 }
91 define amdgpu_kernel void @negated_cond_dominated_blocks(ptr addrspace(1) %arg1) {
92 ; GCN-LABEL: negated_cond_dominated_blocks:
93 ; GCN: ; %bb.0: ; %bb
94 ; GCN-NEXT: s_load_dwordx2 s[4:5], s[0:1], 0x9
95 ; GCN-NEXT: s_waitcnt lgkmcnt(0)
96 ; GCN-NEXT: s_load_dword s0, s[4:5], 0x0
97 ; GCN-NEXT: s_mov_b32 s6, 0
98 ; GCN-NEXT: s_mov_b32 s7, 0xf000
99 ; GCN-NEXT: s_waitcnt lgkmcnt(0)
100 ; GCN-NEXT: s_cmp_lg_u32 s0, 0
101 ; GCN-NEXT: s_cselect_b64 s[0:1], -1, 0
102 ; GCN-NEXT: s_and_b64 s[0:1], exec, s[0:1]
103 ; GCN-NEXT: v_mov_b32_e32 v0, 0
104 ; GCN-NEXT: s_mov_b32 s3, s6
105 ; GCN-NEXT: s_branch .LBB1_2
106 ; GCN-NEXT: .LBB1_1: ; %bb7
107 ; GCN-NEXT: ; in Loop: Header=BB1_2 Depth=1
108 ; GCN-NEXT: s_ashr_i32 s3, s2, 31
109 ; GCN-NEXT: s_lshl_b64 s[8:9], s[2:3], 2
110 ; GCN-NEXT: v_mov_b32_e32 v1, s8
111 ; GCN-NEXT: v_mov_b32_e32 v2, s9
112 ; GCN-NEXT: s_cmp_eq_u32 s2, 32
113 ; GCN-NEXT: buffer_store_dword v0, v[1:2], s[4:7], 0 addr64
114 ; GCN-NEXT: s_mov_b32 s3, s2
115 ; GCN-NEXT: s_cbranch_scc1 .LBB1_6
116 ; GCN-NEXT: .LBB1_2: ; %bb4
117 ; GCN-NEXT: ; =>This Inner Loop Header: Depth=1
118 ; GCN-NEXT: s_mov_b64 vcc, s[0:1]
119 ; GCN-NEXT: s_cbranch_vccz .LBB1_4
120 ; GCN-NEXT: ; %bb.3: ; %bb6
121 ; GCN-NEXT: ; in Loop: Header=BB1_2 Depth=1
122 ; GCN-NEXT: s_add_i32 s2, s3, 1
123 ; GCN-NEXT: s_mov_b64 vcc, exec
124 ; GCN-NEXT: s_cbranch_execnz .LBB1_1
125 ; GCN-NEXT: s_branch .LBB1_5
126 ; GCN-NEXT: .LBB1_4: ; in Loop: Header=BB1_2 Depth=1
127 ; GCN-NEXT: ; implicit-def: $sgpr2
128 ; GCN-NEXT: s_mov_b64 vcc, 0
129 ; GCN-NEXT: .LBB1_5: ; %bb5
130 ; GCN-NEXT: ; in Loop: Header=BB1_2 Depth=1
131 ; GCN-NEXT: s_lshl_b32 s2, s3, 5
132 ; GCN-NEXT: s_or_b32 s2, s2, 1
133 ; GCN-NEXT: s_branch .LBB1_1
134 ; GCN-NEXT: .LBB1_6: ; %bb3
135 ; GCN-NEXT: s_endpgm
136 bb:
137 br label %bb2
139 bb2:
140 %tmp1 = load i32, ptr addrspace(1) %arg1
141 %tmp2 = icmp eq i32 %tmp1, 0
142 br label %bb4
144 bb3:
145 ret void
147 bb4:
148 %tmp3 = phi i32 [ 0, %bb2 ], [ %tmp7, %bb7 ]
149 %tmp4 = shl i32 %tmp3, 5
150 br i1 %tmp2, label %bb5, label %bb6
152 bb5:
153 %tmp5 = add i32 %tmp4, 1
154 br label %bb7
156 bb6:
157 %tmp6 = add i32 %tmp3, 1
158 br label %bb7
160 bb7:
161 %tmp7 = phi i32 [ %tmp5, %bb5 ], [ %tmp6, %bb6 ]
162 %gep = getelementptr inbounds i32, ptr addrspace(1) %arg1, i32 %tmp7
163 store i32 0, ptr addrspace(1) %gep
164 %tmp8 = icmp eq i32 %tmp7, 32
165 br i1 %tmp8, label %bb3, label %bb4
166 }