1 /* { dg-do compile } */
2 /* { dg-options "-O2 -ftree-vectorize" } */
3
4 #include <stdint.h>
5
6 #define N 100
7
8 #define eq(A, B) ((A) == (B))
9 #define ne(A, B) ((A) != (B))
10 #define lt(A, B) ((A) < (B))
11 #define le(A, B) ((A) <= (B))
12 #define ge(A, B) ((A) >= (B))
13 #define gt(A, B) ((A) > (B))
14 #define unordered(A, B) (__builtin_isunordered (A, B))
15
16 #define DEF_CONST_LOOP(NAME, SUFFIX, TYPE, CONST) \
17 void __attribute__ ((noipa)) \
18 NAME##_##SUFFIX##_##TYPE (TYPE *restrict dst, TYPE *restrict src) \
19 { \
20 for (int i = 0; i < N; ++i) \
21 if (NAME (src[i], CONST)) \
22 dst[i] = 1; \
23 }
24
25 #define DEF_LOOP(NAME, TYPE, CONST1, CONST2) \
26 void __attribute__ ((noipa)) \
27 NAME##_var_##TYPE (TYPE *restrict dst, TYPE *restrict src, TYPE x) \
28 { \
29 for (int i = 0; i < N; ++i) \
30 if (NAME (src[i], x)) \
31 dst[i] = x; \
32 } \
33 DEF_CONST_LOOP (NAME, const1, TYPE, CONST1) \
34 DEF_CONST_LOOP (NAME, const2, TYPE, CONST2)
35
36 #define FOR_EACH_INT_OPERATOR(T, TYPE, CONST1, CONST2) \
37 T (eq, TYPE, CONST1, CONST2) \
38 T (ne, TYPE, CONST1, CONST2) \
39 T (le, TYPE, CONST1, CONST2) \
40 T (lt, TYPE, CONST1, CONST2) \
41 T (gt, TYPE, CONST1, CONST2) \
42 T (ge, TYPE, CONST1, CONST2)
43
44 #define FOR_EACH_FLOAT_OPERATOR(T, TYPE, CONST1, CONST2) \
45 FOR_EACH_INT_OPERATOR(T, TYPE, CONST1, CONST2) \
46 T (unordered, TYPE, CONST1, CONST2)
47
48 #define FOR_EACH_TYPE(T) \
49 FOR_EACH_INT_OPERATOR (T, int8_t, 2, 100) \
50 FOR_EACH_INT_OPERATOR (T, int16_t, 3, 1000) \
51 FOR_EACH_INT_OPERATOR (T, int32_t, 4, 2000) \
52 FOR_EACH_INT_OPERATOR (T, int64_t, 5, 3000) \
53 FOR_EACH_INT_OPERATOR (T, uint8_t, 2, 160) \
54 FOR_EACH_INT_OPERATOR (T, uint16_t, 3, 500) \
55 FOR_EACH_INT_OPERATOR (T, uint32_t, 4, 1500) \
56 FOR_EACH_INT_OPERATOR (T, uint64_t, 5, 2500) \
57 FOR_EACH_FLOAT_OPERATOR (T, _Float16, 0, 1) \
58 FOR_EACH_FLOAT_OPERATOR (T, float, 0, 1) \
59 FOR_EACH_FLOAT_OPERATOR (T, double, 0, 1)
60
61 FOR_EACH_TYPE (DEF_LOOP)
62
63 /* { dg-final { scan-assembler-not {\tand\t} } } */
64
65 /* { dg-final { scan-assembler-times {\tcmpeq\tp[0-7]\.b, p[0-7]/z, z[0-9]+\.b, z[0-9]+\.b\n} 4 } } */
66 /* { dg-final { scan-assembler-times {\tcmpeq\tp[0-7]\.h, p[0-7]/z, z[0-9]+\.h, z[0-9]+\.h\n} 4 } } */
67 /* { dg-final { scan-assembler-times {\tcmpeq\tp[0-7]\.s, p[0-7]/z, z[0-9]+\.s, z[0-9]+\.s\n} 4 } } */
68 /* { dg-final { scan-assembler-times {\tcmpeq\tp[0-7]\.d, p[0-7]/z, z[0-9]+\.d, z[0-9]+\.d\n} 4 } } */
69
70 /* { dg-final { scan-assembler-times {\tcmpeq\tp[0-7]\.b, p[0-7]/z, z[0-9]+\.b, #2\n} 2 } } */
71 /* { dg-final { scan-assembler-times {\tcmpeq\tp[0-7]\.h, p[0-7]/z, z[0-9]+\.h, #3\n} 2 } } */
72 /* { dg-final { scan-assembler-times {\tcmpeq\tp[0-7]\.s, p[0-7]/z, z[0-9]+\.s, #4\n} 2 } } */
73 /* { dg-final { scan-assembler-times {\tcmpeq\tp[0-7]\.d, p[0-7]/z, z[0-9]+\.d, #5\n} 2 } } */
74
75 /* { dg-final { scan-assembler-times {\tcmpne\tp[0-7]\.b, p[0-7]/z, z[0-9]+\.b, z[0-9]+\.b\n} 4 } } */
76 /* { dg-final { scan-assembler-times {\tcmpne\tp[0-7]\.h, p[0-7]/z, z[0-9]+\.h, z[0-9]+\.h\n} 4 } } */
77 /* { dg-final { scan-assembler-times {\tcmpne\tp[0-7]\.s, p[0-7]/z, z[0-9]+\.s, z[0-9]+\.s\n} 4 } } */
78 /* { dg-final { scan-assembler-times {\tcmpne\tp[0-7]\.d, p[0-7]/z, z[0-9]+\.d, z[0-9]+\.d\n} 4 } } */
79
80 /* { dg-final { scan-assembler-times {\tcmpne\tp[0-7]\.b, p[0-7]/z, z[0-9]+\.b, #2\n} 2 } } */
81 /* { dg-final { scan-assembler-times {\tcmpne\tp[0-7]\.h, p[0-7]/z, z[0-9]+\.h, #3\n} 2 } } */
82 /* { dg-final { scan-assembler-times {\tcmpne\tp[0-7]\.s, p[0-7]/z, z[0-9]+\.s, #4\n} 2 } } */
83 /* { dg-final { scan-assembler-times {\tcmpne\tp[0-7]\.d, p[0-7]/z, z[0-9]+\.d, #5\n} 2 } } */
84
85 /* { dg-final { scan-assembler-times {\tcmplt\tp[0-7]\.b, p[0-7]/z, z[0-9]+\.b, z[0-9]+\.b\n} 1 } } */
86 /* { dg-final { scan-assembler-times {\tcmplt\tp[0-7]\.h, p[0-7]/z, z[0-9]+\.h, z[0-9]+\.h\n} 1 } } */
87 /* { dg-final { scan-assembler-times {\tcmplt\tp[0-7]\.s, p[0-7]/z, z[0-9]+\.s, z[0-9]+\.s\n} 1 } } */
88 /* { dg-final { scan-assembler-times {\tcmplt\tp[0-7]\.d, p[0-7]/z, z[0-9]+\.d, z[0-9]+\.d\n} 1 } } */
89
90 /* { dg-final { scan-assembler-times {\tcmple\tp[0-7]\.b, p[0-7]/z, z[0-9]+\.b, #1\n} 1 } } */
91 /* { dg-final { scan-assembler-times {\tcmple\tp[0-7]\.h, p[0-7]/z, z[0-9]+\.h, #2\n} 1 } } */
92 /* { dg-final { scan-assembler-times {\tcmple\tp[0-7]\.s, p[0-7]/z, z[0-9]+\.s, #3\n} 1 } } */
93 /* { dg-final { scan-assembler-times {\tcmple\tp[0-7]\.d, p[0-7]/z, z[0-9]+\.d, #4\n} 1 } } */
94
95 /* { dg-final { scan-assembler-times {\tcmple\tp[0-7]\.b, p[0-7]/z, z[0-9]+\.b, z[0-9]+\.b\n} 3 } } */
96 /* { dg-final { scan-assembler-times {\tcmple\tp[0-7]\.h, p[0-7]/z, z[0-9]+\.h, z[0-9]+\.h\n} 3 } } */
97 /* { dg-final { scan-assembler-times {\tcmple\tp[0-7]\.s, p[0-7]/z, z[0-9]+\.s, z[0-9]+\.s\n} 3 } } */
98 /* { dg-final { scan-assembler-times {\tcmple\tp[0-7]\.d, p[0-7]/z, z[0-9]+\.d, z[0-9]+\.d\n} 3 } } */
99
100 /* { dg-final { scan-assembler-times {\tcmple\tp[0-7]\.b, p[0-7]/z, z[0-9]+\.b, #2\n} 1 } } */
101 /* { dg-final { scan-assembler-times {\tcmple\tp[0-7]\.h, p[0-7]/z, z[0-9]+\.h, #3\n} 1 } } */
102 /* { dg-final { scan-assembler-times {\tcmple\tp[0-7]\.s, p[0-7]/z, z[0-9]+\.s, #4\n} 1 } } */
103 /* { dg-final { scan-assembler-times {\tcmple\tp[0-7]\.d, p[0-7]/z, z[0-9]+\.d, #5\n} 1 } } */
104
105 /* { dg-final { scan-assembler-times {\tcmpgt\tp[0-7]\.b, p[0-7]/z, z[0-9]+\.b, z[0-9]+\.b\n} 3 } } */
106 /* { dg-final { scan-assembler-times {\tcmpgt\tp[0-7]\.h, p[0-7]/z, z[0-9]+\.h, z[0-9]+\.h\n} 3 } } */
107 /* { dg-final { scan-assembler-times {\tcmpgt\tp[0-7]\.s, p[0-7]/z, z[0-9]+\.s, z[0-9]+\.s\n} 3 } } */
108 /* { dg-final { scan-assembler-times {\tcmpgt\tp[0-7]\.d, p[0-7]/z, z[0-9]+\.d, z[0-9]+\.d\n} 3 } } */
109
110 /* { dg-final { scan-assembler-times {\tcmpgt\tp[0-7]\.b, p[0-7]/z, z[0-9]+\.b, #2\n} 1 } } */
111 /* { dg-final { scan-assembler-times {\tcmpgt\tp[0-7]\.h, p[0-7]/z, z[0-9]+\.h, #3\n} 1 } } */
112 /* { dg-final { scan-assembler-times {\tcmpgt\tp[0-7]\.s, p[0-7]/z, z[0-9]+\.s, #4\n} 1 } } */
113 /* { dg-final { scan-assembler-times {\tcmpgt\tp[0-7]\.d, p[0-7]/z, z[0-9]+\.d, #5\n} 1 } } */
114
115 /* { dg-final { scan-assembler-times {\tcmpge\tp[0-7]\.b, p[0-7]/z, z[0-9]+\.b, z[0-9]+\.b\n} 1 } } */
116 /* { dg-final { scan-assembler-times {\tcmpge\tp[0-7]\.h, p[0-7]/z, z[0-9]+\.h, z[0-9]+\.h\n} 1 } } */
117 /* { dg-final { scan-assembler-times {\tcmpge\tp[0-7]\.s, p[0-7]/z, z[0-9]+\.s, z[0-9]+\.s\n} 1 } } */
118 /* { dg-final { scan-assembler-times {\tcmpge\tp[0-7]\.d, p[0-7]/z, z[0-9]+\.d, z[0-9]+\.d\n} 1 } } */
119
120 /* { dg-final { scan-assembler-times {\tcmpgt\tp[0-7]\.b, p[0-7]/z, z[0-9]+\.b, #1\n} 1 } } */
121 /* { dg-final { scan-assembler-times {\tcmpgt\tp[0-7]\.h, p[0-7]/z, z[0-9]+\.h, #2\n} 1 } } */
122 /* { dg-final { scan-assembler-times {\tcmpgt\tp[0-7]\.s, p[0-7]/z, z[0-9]+\.s, #3\n} 1 } } */
123 /* { dg-final { scan-assembler-times {\tcmpgt\tp[0-7]\.d, p[0-7]/z, z[0-9]+\.d, #4\n} 1 } } */
124
125 /* { dg-final { scan-assembler-times {\tcmplo\tp[0-7]\.b, p[0-7]/z, z[0-9]+\.b, z[0-9]+\.b\n} 1 } } */
126 /* { dg-final { scan-assembler-times {\tcmplo\tp[0-7]\.h, p[0-7]/z, z[0-9]+\.h, z[0-9]+\.h\n} 1 } } */
127 /* { dg-final { scan-assembler-times {\tcmplo\tp[0-7]\.s, p[0-7]/z, z[0-9]+\.s, z[0-9]+\.s\n} 1 } } */
128 /* { dg-final { scan-assembler-times {\tcmplo\tp[0-7]\.d, p[0-7]/z, z[0-9]+\.d, z[0-9]+\.d\n} 1 } } */
129
130 /* { dg-final { scan-assembler-times {\tcmpls\tp[0-7]\.b, p[0-7]/z, z[0-9]+\.b, #1\n} 1 } } */
131 /* { dg-final { scan-assembler-times {\tcmpls\tp[0-7]\.h, p[0-7]/z, z[0-9]+\.h, #2\n} 1 } } */
132 /* { dg-final { scan-assembler-times {\tcmpls\tp[0-7]\.s, p[0-7]/z, z[0-9]+\.s, #3\n} 1 } } */
133 /* { dg-final { scan-assembler-times {\tcmpls\tp[0-7]\.d, p[0-7]/z, z[0-9]+\.d, #4\n} 1 } } */
134
135 /* { dg-final { scan-assembler-times {\tcmpls\tp[0-7]\.b, p[0-7]/z, z[0-9]+\.b, z[0-9]+\.b\n} 3 } } */
136 /* { dg-final { scan-assembler-times {\tcmpls\tp[0-7]\.h, p[0-7]/z, z[0-9]+\.h, z[0-9]+\.h\n} 3 } } */
137 /* { dg-final { scan-assembler-times {\tcmpls\tp[0-7]\.s, p[0-7]/z, z[0-9]+\.s, z[0-9]+\.s\n} 3 } } */
138 /* { dg-final { scan-assembler-times {\tcmpls\tp[0-7]\.d, p[0-7]/z, z[0-9]+\.d, z[0-9]+\.d\n} 3 } } */
139
140 /* { dg-final { scan-assembler-times {\tcmpls\tp[0-7]\.b, p[0-7]/z, z[0-9]+\.b, #2\n} 1 } } */
141 /* { dg-final { scan-assembler-times {\tcmpls\tp[0-7]\.h, p[0-7]/z, z[0-9]+\.h, #3\n} 1 } } */
142 /* { dg-final { scan-assembler-times {\tcmpls\tp[0-7]\.s, p[0-7]/z, z[0-9]+\.s, #4\n} 1 } } */
143 /* { dg-final { scan-assembler-times {\tcmpls\tp[0-7]\.d, p[0-7]/z, z[0-9]+\.d, #5\n} 1 } } */
144
145 /* { dg-final { scan-assembler-times {\tcmphi\tp[0-7]\.b, p[0-7]/z, z[0-9]+\.b, z[0-9]+\.b\n} 3 } } */
146 /* { dg-final { scan-assembler-times {\tcmphi\tp[0-7]\.h, p[0-7]/z, z[0-9]+\.h, z[0-9]+\.h\n} 3 } } */
147 /* { dg-final { scan-assembler-times {\tcmphi\tp[0-7]\.s, p[0-7]/z, z[0-9]+\.s, z[0-9]+\.s\n} 3 } } */
148 /* { dg-final { scan-assembler-times {\tcmphi\tp[0-7]\.d, p[0-7]/z, z[0-9]+\.d, z[0-9]+\.d\n} 3 } } */
149
150 /* { dg-final { scan-assembler-times {\tcmphi\tp[0-7]\.b, p[0-7]/z, z[0-9]+\.b, #2\n} 1 } } */
151 /* { dg-final { scan-assembler-times {\tcmphi\tp[0-7]\.h, p[0-7]/z, z[0-9]+\.h, #3\n} 1 } } */
152 /* { dg-final { scan-assembler-times {\tcmphi\tp[0-7]\.s, p[0-7]/z, z[0-9]+\.s, #4\n} 1 } } */
153 /* { dg-final { scan-assembler-times {\tcmphi\tp[0-7]\.d, p[0-7]/z, z[0-9]+\.d, #5\n} 1 } } */
154
155 /* { dg-final { scan-assembler-times {\tcmphs\tp[0-7]\.b, p[0-7]/z, z[0-9]+\.b, z[0-9]+\.b\n} 1 } } */
156 /* { dg-final { scan-assembler-times {\tcmphs\tp[0-7]\.h, p[0-7]/z, z[0-9]+\.h, z[0-9]+\.h\n} 1 } } */
157 /* { dg-final { scan-assembler-times {\tcmphs\tp[0-7]\.s, p[0-7]/z, z[0-9]+\.s, z[0-9]+\.s\n} 1 } } */
158 /* { dg-final { scan-assembler-times {\tcmphs\tp[0-7]\.d, p[0-7]/z, z[0-9]+\.d, z[0-9]+\.d\n} 1 } } */
159
160 /* { dg-final { scan-assembler-times {\tcmphi\tp[0-7]\.b, p[0-7]/z, z[0-9]+\.b, #1\n} 1 } } */
161 /* { dg-final { scan-assembler-times {\tcmphi\tp[0-7]\.h, p[0-7]/z, z[0-9]+\.h, #2\n} 1 } } */
162 /* { dg-final { scan-assembler-times {\tcmphi\tp[0-7]\.s, p[0-7]/z, z[0-9]+\.s, #3\n} 1 } } */
163 /* { dg-final { scan-assembler-times {\tcmphi\tp[0-7]\.d, p[0-7]/z, z[0-9]+\.d, #4\n} 1 } } */
164
165
166 /* { dg-final { scan-assembler-times {\tfcmeq\tp[0-7]\.h, p[0-7]/z, z[0-9]+\.h, z[0-9]+\.h\n} 2 } } */
167 /* { dg-final { scan-assembler-times {\tfcmeq\tp[0-7]\.s, p[0-7]/z, z[0-9]+\.s, z[0-9]+\.s\n} 2 } } */
168 /* { dg-final { scan-assembler-times {\tfcmeq\tp[0-7]\.d, p[0-7]/z, z[0-9]+\.d, z[0-9]+\.d\n} 2 } } */
169
170 /* { dg-final { scan-assembler-times {\tfcmeq\tp[0-7]\.h, p[0-7]/z, z[0-9]+\.h, #0\.0\n} 1 } } */
171 /* { dg-final { scan-assembler-times {\tfcmeq\tp[0-7]\.s, p[0-7]/z, z[0-9]+\.s, #0\.0\n} 1 } } */
172 /* { dg-final { scan-assembler-times {\tfcmeq\tp[0-7]\.d, p[0-7]/z, z[0-9]+\.d, #0\.0\n} 1 } } */
173
174 /* { dg-final { scan-assembler-times {\tfcmne\tp[0-7]\.h, p[0-7]/z, z[0-9]+\.h, z[0-9]+\.h\n} 2 } } */
175 /* { dg-final { scan-assembler-times {\tfcmne\tp[0-7]\.s, p[0-7]/z, z[0-9]+\.s, z[0-9]+\.s\n} 2 } } */
176 /* { dg-final { scan-assembler-times {\tfcmne\tp[0-7]\.d, p[0-7]/z, z[0-9]+\.d, z[0-9]+\.d\n} 2 } } */
177
178 /* { dg-final { scan-assembler-times {\tfcmne\tp[0-7]\.h, p[0-7]/z, z[0-9]+\.h, #0\.0\n} 1 } } */
179 /* { dg-final { scan-assembler-times {\tfcmne\tp[0-7]\.s, p[0-7]/z, z[0-9]+\.s, #0\.0\n} 1 } } */
180 /* { dg-final { scan-assembler-times {\tfcmne\tp[0-7]\.d, p[0-7]/z, z[0-9]+\.d, #0\.0\n} 1 } } */
181
182 /* { dg-final { scan-assembler-times {\tfcmle\tp[0-7]\.h, p[0-7]/z, z[0-9]+\.h, z[0-9]+\.h\n} 2 } } */
183 /* { dg-final { scan-assembler-times {\tfcmle\tp[0-7]\.s, p[0-7]/z, z[0-9]+\.s, z[0-9]+\.s\n} 2 } } */
184 /* { dg-final { scan-assembler-times {\tfcmle\tp[0-7]\.d, p[0-7]/z, z[0-9]+\.d, z[0-9]+\.d\n} 2 } } */
185
186 /* { dg-final { scan-assembler-times {\tfcmle\tp[0-7]\.h, p[0-7]/z, z[0-9]+\.h, #0\.0\n} 1 } } */
187 /* { dg-final { scan-assembler-times {\tfcmle\tp[0-7]\.s, p[0-7]/z, z[0-9]+\.s, #0\.0\n} 1 } } */
188 /* { dg-final { scan-assembler-times {\tfcmle\tp[0-7]\.d, p[0-7]/z, z[0-9]+\.d, #0\.0\n} 1 } } */
189
190 /* { dg-final { scan-assembler-times {\tfcmlt\tp[0-7]\.h, p[0-7]/z, z[0-9]+\.h, z[0-9]+\.h\n} 2 } } */
191 /* { dg-final { scan-assembler-times {\tfcmlt\tp[0-7]\.s, p[0-7]/z, z[0-9]+\.s, z[0-9]+\.s\n} 2 } } */
192 /* { dg-final { scan-assembler-times {\tfcmlt\tp[0-7]\.d, p[0-7]/z, z[0-9]+\.d, z[0-9]+\.d\n} 2 } } */
193
194 /* { dg-final { scan-assembler-times {\tfcmlt\tp[0-7]\.h, p[0-7]/z, z[0-9]+\.h, #0\.0\n} 1 } } */
195 /* { dg-final { scan-assembler-times {\tfcmlt\tp[0-7]\.s, p[0-7]/z, z[0-9]+\.s, #0\.0\n} 1 } } */
196 /* { dg-final { scan-assembler-times {\tfcmlt\tp[0-7]\.d, p[0-7]/z, z[0-9]+\.d, #0\.0\n} 1 } } */
197
198 /* { dg-final { scan-assembler-times {\tfcmge\tp[0-7]\.h, p[0-7]/z, z[0-9]+\.h, z[0-9]+\.h\n} 2 } } */
199 /* { dg-final { scan-assembler-times {\tfcmge\tp[0-7]\.s, p[0-7]/z, z[0-9]+\.s, z[0-9]+\.s\n} 2 } } */
200 /* { dg-final { scan-assembler-times {\tfcmge\tp[0-7]\.d, p[0-7]/z, z[0-9]+\.d, z[0-9]+\.d\n} 2 } } */
201
202 /* { dg-final { scan-assembler-times {\tfcmge\tp[0-7]\.h, p[0-7]/z, z[0-9]+\.h, #0\.0\n} 1 } } */
203 /* { dg-final { scan-assembler-times {\tfcmge\tp[0-7]\.s, p[0-7]/z, z[0-9]+\.s, #0\.0\n} 1 } } */
204 /* { dg-final { scan-assembler-times {\tfcmge\tp[0-7]\.d, p[0-7]/z, z[0-9]+\.d, #0\.0\n} 1 } } */
205
206 /* { dg-final { scan-assembler-times {\tfcmgt\tp[0-7]\.h, p[0-7]/z, z[0-9]+\.h, z[0-9]+\.h\n} 2 } } */
207 /* { dg-final { scan-assembler-times {\tfcmgt\tp[0-7]\.s, p[0-7]/z, z[0-9]+\.s, z[0-9]+\.s\n} 2 } } */
208 /* { dg-final { scan-assembler-times {\tfcmgt\tp[0-7]\.d, p[0-7]/z, z[0-9]+\.d, z[0-9]+\.d\n} 2 } } */
209
210 /* { dg-final { scan-assembler-times {\tfcmgt\tp[0-7]\.h, p[0-7]/z, z[0-9]+\.h, #0\.0\n} 1 } } */
211 /* { dg-final { scan-assembler-times {\tfcmgt\tp[0-7]\.s, p[0-7]/z, z[0-9]+\.s, #0\.0\n} 1 } } */
212 /* { dg-final { scan-assembler-times {\tfcmgt\tp[0-7]\.d, p[0-7]/z, z[0-9]+\.d, #0\.0\n} 1 } } */
213
214 /* { dg-final { scan-assembler-times {\tfcmuo\tp[0-7]\.h, p[0-7]/z, z[0-9]+\.h, z[0-9]+\.h\n} 3 } } */
215 /* { dg-final { scan-assembler-times {\tfcmuo\tp[0-7]\.s, p[0-7]/z, z[0-9]+\.s, z[0-9]+\.s\n} 3 } } */
216 /* { dg-final { scan-assembler-times {\tfcmuo\tp[0-7]\.d, p[0-7]/z, z[0-9]+\.d, z[0-9]+\.d\n} 3 } } */