• Home
  • Line#
  • Scopes#
  • Navigate#
  • Raw
  • Download
1; RUN: llc -march=amdgcn -mcpu=verde -verify-machineinstrs < %s | FileCheck -check-prefix=GCN -check-prefix=FUNC %s
2; RUN: llc -march=amdgcn -mcpu=tonga -verify-machineinstrs < %s | FileCheck -check-prefix=GCN -check-prefix=FUNC %s
3
4; FUNC-LABEL: {{^}}s_abs_i32:
5; GCN: s_abs_i32
6; GCN: s_add_i32
7define void @s_abs_i32(i32 addrspace(1)* %out, i32 %val) nounwind {
8  %neg = sub i32 0, %val
9  %cond = icmp sgt i32 %val, %neg
10  %res = select i1 %cond, i32 %val, i32 %neg
11  %res2 = add i32 %res, 2
12  store i32 %res2, i32 addrspace(1)* %out, align 4
13  ret void
14}
15
16; FUNC-LABEL: {{^}}v_abs_i32:
17; GCN: v_sub_i32_e32 [[NEG:v[0-9]+]], vcc, 0, [[SRC:v[0-9]+]]
18; GCN: v_max_i32_e32 {{v[0-9]+}}, [[NEG]], [[SRC]]
19; GCN: v_add_i32
20define void @v_abs_i32(i32 addrspace(1)* %out, i32 addrspace(1)* %src) nounwind {
21  %val = load i32, i32 addrspace(1)* %src, align 4
22  %neg = sub i32 0, %val
23  %cond = icmp sgt i32 %val, %neg
24  %res = select i1 %cond, i32 %val, i32 %neg
25  %res2 = add i32 %res, 2
26  store i32 %res2, i32 addrspace(1)* %out, align 4
27  ret void
28}
29
30; FUNC-LABEL: {{^}}s_abs_v2i32:
31; GCN: s_abs_i32
32; GCN: s_abs_i32
33; GCN: s_add_i32
34; GCN: s_add_i32
35define void @s_abs_v2i32(<2 x i32> addrspace(1)* %out, <2 x i32> %val) nounwind {
36  %z0 = insertelement <2 x i32> undef, i32 0, i32 0
37  %z1 = insertelement <2 x i32> %z0, i32 0, i32 1
38  %t0 = insertelement <2 x i32> undef, i32 2, i32 0
39  %t1 = insertelement <2 x i32> %t0, i32 2, i32 1
40  %neg = sub <2 x i32> %z1, %val
41  %cond = icmp sgt <2 x i32> %val, %neg
42  %res = select <2 x i1> %cond, <2 x i32> %val, <2 x i32> %neg
43  %res2 = add <2 x i32> %res, %t1
44  store <2 x i32> %res2, <2 x i32> addrspace(1)* %out, align 4
45  ret void
46}
47
48; FUNC-LABEL: {{^}}v_abs_v2i32:
49; GCN: v_sub_i32_e32 [[NEG0:v[0-9]+]], vcc, 0, [[SRC0:v[0-9]+]]
50; GCN: v_sub_i32_e32 [[NEG1:v[0-9]+]], vcc, 0, [[SRC1:v[0-9]+]]
51
52; GCN: v_max_i32_e32 {{v[0-9]+}}, [[NEG0]], [[SRC0]]
53; GCN: v_max_i32_e32 {{v[0-9]+}}, [[NEG1]], [[SRC1]]
54
55; GCN: v_add_i32
56; GCN: v_add_i32
57define void @v_abs_v2i32(<2 x i32> addrspace(1)* %out, <2 x i32> addrspace(1)* %src) nounwind {
58  %z0 = insertelement <2 x i32> undef, i32 0, i32 0
59  %z1 = insertelement <2 x i32> %z0, i32 0, i32 1
60  %t0 = insertelement <2 x i32> undef, i32 2, i32 0
61  %t1 = insertelement <2 x i32> %t0, i32 2, i32 1
62  %val = load <2 x i32>, <2 x i32> addrspace(1)* %src, align 4
63  %neg = sub <2 x i32> %z1, %val
64  %cond = icmp sgt <2 x i32> %val, %neg
65  %res = select <2 x i1> %cond, <2 x i32> %val, <2 x i32> %neg
66  %res2 = add <2 x i32> %res, %t1
67  store <2 x i32> %res2, <2 x i32> addrspace(1)* %out, align 4
68  ret void
69}
70
71; FUNC-LABEL: {{^}}s_abs_v4i32:
72; TODO: this should use s_abs_i32
73; GCN: s_abs_i32
74; GCN: s_abs_i32
75; GCN: s_abs_i32
76; GCN: s_abs_i32
77
78; GCN: s_add_i32
79; GCN: s_add_i32
80; GCN: s_add_i32
81; GCN: s_add_i32
82define void @s_abs_v4i32(<4 x i32> addrspace(1)* %out, <4 x i32> %val) nounwind {
83  %z0 = insertelement <4 x i32> undef, i32 0, i32 0
84  %z1 = insertelement <4 x i32> %z0, i32 0, i32 1
85  %z2 = insertelement <4 x i32> %z1, i32 0, i32 2
86  %z3 = insertelement <4 x i32> %z2, i32 0, i32 3
87  %t0 = insertelement <4 x i32> undef, i32 2, i32 0
88  %t1 = insertelement <4 x i32> %t0, i32 2, i32 1
89  %t2 = insertelement <4 x i32> %t1, i32 2, i32 2
90  %t3 = insertelement <4 x i32> %t2, i32 2, i32 3
91  %neg = sub <4 x i32> %z3, %val
92  %cond = icmp sgt <4 x i32> %val, %neg
93  %res = select <4 x i1> %cond, <4 x i32> %val, <4 x i32> %neg
94  %res2 = add <4 x i32> %res, %t3
95  store <4 x i32> %res2, <4 x i32> addrspace(1)* %out, align 4
96  ret void
97}
98
99; FUNC-LABEL: {{^}}v_abs_v4i32:
100; GCN: v_sub_i32_e32 [[NEG0:v[0-9]+]], vcc, 0, [[SRC0:v[0-9]+]]
101; GCN: v_sub_i32_e32 [[NEG1:v[0-9]+]], vcc, 0, [[SRC1:v[0-9]+]]
102; GCN: v_sub_i32_e32 [[NEG2:v[0-9]+]], vcc, 0, [[SRC2:v[0-9]+]]
103; GCN: v_sub_i32_e32 [[NEG3:v[0-9]+]], vcc, 0, [[SRC3:v[0-9]+]]
104
105; GCN: v_max_i32_e32 {{v[0-9]+}}, [[NEG0]], [[SRC0]]
106; GCN: v_max_i32_e32 {{v[0-9]+}}, [[NEG1]], [[SRC1]]
107; GCN: v_max_i32_e32 {{v[0-9]+}}, [[NEG2]], [[SRC2]]
108; GCN: v_max_i32_e32 {{v[0-9]+}}, [[NEG3]], [[SRC3]]
109
110; GCN: v_add_i32
111; GCN: v_add_i32
112; GCN: v_add_i32
113; GCN: v_add_i32
114define void @v_abs_v4i32(<4 x i32> addrspace(1)* %out, <4 x i32> addrspace(1)* %src) nounwind {
115  %z0 = insertelement <4 x i32> undef, i32 0, i32 0
116  %z1 = insertelement <4 x i32> %z0, i32 0, i32 1
117  %z2 = insertelement <4 x i32> %z1, i32 0, i32 2
118  %z3 = insertelement <4 x i32> %z2, i32 0, i32 3
119  %t0 = insertelement <4 x i32> undef, i32 2, i32 0
120  %t1 = insertelement <4 x i32> %t0, i32 2, i32 1
121  %t2 = insertelement <4 x i32> %t1, i32 2, i32 2
122  %t3 = insertelement <4 x i32> %t2, i32 2, i32 3
123  %val = load <4 x i32>, <4 x i32> addrspace(1)* %src, align 4
124  %neg = sub <4 x i32> %z3, %val
125  %cond = icmp sgt <4 x i32> %val, %neg
126  %res = select <4 x i1> %cond, <4 x i32> %val, <4 x i32> %neg
127  %res2 = add <4 x i32> %res, %t3
128  store <4 x i32> %res2, <4 x i32> addrspace(1)* %out, align 4
129  ret void
130}
131