Debian Codegraph @

gcc/testsuite/gcc.target/aarch64/sve2/acle/asm/sub_bf16.c

raw
1 /* { dg-do assemble { target aarch64_asm_sve-b16b16_ok } } */
2 /* { dg-do compile { target { ! aarch64_asm_sve-b16b16_ok } } } */
3 /* { dg-final { check-function-bodies "**" "" "-DCHECK_ASM" } } */
4
5 #include "test_sve_acle.h"
6
7 #pragma GCC target "+sve-b16b16"
8 #ifdef STREAMING_COMPATIBLE
9 #pragma GCC target "+sme2"
10 #endif
11
12 /*
13 ** sub_bf16_m_tied1:
14 ** bfsub z0\.h, p0/m, z0\.h, z1\.h
15 ** ret
16 */
17 TEST_UNIFORM_Z (sub_bf16_m_tied1, svbfloat16_t,
18 z0 = svsub_bf16_m (p0, z0, z1),
19 z0 = svsub_m (p0, z0, z1))
20
21 /*
22 ** sub_bf16_m_tied2:
23 ** mov (z[0-9]+)\.d, z0\.d
24 ** movprfx z0, z1
25 ** bfsub z0\.h, p0/m, z0\.h, \1\.h
26 ** ret
27 */
28 TEST_UNIFORM_Z (sub_bf16_m_tied2, svbfloat16_t,
29 z0 = svsub_bf16_m (p0, z1, z0),
30 z0 = svsub_m (p0, z1, z0))
31
32 /*
33 ** sub_bf16_m_untied:
34 ** movprfx z0, z1
35 ** bfsub z0\.h, p0/m, z0\.h, z2\.h
36 ** ret
37 */
38 TEST_UNIFORM_Z (sub_bf16_m_untied, svbfloat16_t,
39 z0 = svsub_bf16_m (p0, z1, z2),
40 z0 = svsub_m (p0, z1, z2))
41
42 /*
43 ** sub_h4_bf16_m_tied1:
44 ** mov (z[0-9]+\.h), h4
45 ** bfsub z0\.h, p0/m, z0\.h, \1
46 ** ret
47 */
48 TEST_UNIFORM_ZD (sub_h4_bf16_m_tied1, svbfloat16_t, __bf16,
49 z0 = svsub_n_bf16_m (p0, z0, d4),
50 z0 = svsub_m (p0, z0, d4))
51
52 /*
53 ** sub_h4_bf16_m_untied:
54 ** mov (z[0-9]+\.h), h4
55 ** movprfx z0, z1
56 ** bfsub z0\.h, p0/m, z0\.h, \1
57 ** ret
58 */
59 TEST_UNIFORM_ZD (sub_h4_bf16_m_untied, svbfloat16_t, __bf16,
60 z0 = svsub_n_bf16_m (p0, z1, d4),
61 z0 = svsub_m (p0, z1, d4))
62
63 /*
64 ** sub_1_bf16_m:
65 ** fmov (z[0-9]+\.h), #-1\.875(?:e\+0)?
66 ** bfadd z0\.h, p0/m, z0\.h, \1
67 ** ret
68 */
69 TEST_UNIFORM_Z (sub_1_bf16_m, svbfloat16_t,
70 z0 = svsub_n_bf16_m (p0, z0, 1),
71 z0 = svsub_m (p0, z0, 1))
72
73 /*
74 ** sub_0p5_bf16_m:
75 ** fmov (z[0-9]+\.h), #-1\.75(?:e\+0)?
76 ** bfadd z0\.h, p0/m, z0\.h, \1
77 ** ret
78 */
79 TEST_UNIFORM_Z (sub_0p5_bf16_m, svbfloat16_t,
80 z0 = svsub_n_bf16_m (p0, z0, (bfloat16_t) (0.5)),
81 z0 = svsub_m (p0, z0, (bfloat16_t) (0.5)))
82
83 /*
84 ** sub_m1_bf16_m:
85 ** fmov (z[0-9]+\.h), #1\.875(?:e\+0)?
86 ** bfadd z0\.h, p0/m, z0\.h, \1
87 ** ret
88 */
89 TEST_UNIFORM_Z (sub_m1_bf16_m, svbfloat16_t,
90 z0 = svsub_n_bf16_m (p0, z0, -1),
91 z0 = svsub_m (p0, z0, -1))
92
93 /*
94 ** sub_m0p5_bf16_m:
95 ** fmov (z[0-9]+\.h), #1\.75(?:e\+0)?
96 ** bfadd z0\.h, p0/m, z0\.h, \1
97 ** ret
98 */
99 TEST_UNIFORM_Z (sub_m0p5_bf16_m, svbfloat16_t,
100 z0 = svsub_n_bf16_m (p0, z0, (bfloat16_t) (-0.5)),
101 z0 = svsub_m (p0, z0, (bfloat16_t) (-0.5)))
102
103 /*
104 ** sub_m2_bf16_m_tied1:
105 ** fmov (z[0-9]+\.h), #2\.0(?:e\+0)?
106 ** bfadd z0\.h, p0/m, z0\.h, \1
107 ** ret
108 */
109 TEST_UNIFORM_Z (sub_m2_bf16_m_tied1, svbfloat16_t,
110 z0 = svsub_n_bf16_m (p0, z0, -2),
111 z0 = svsub_m (p0, z0, -2))
112
113 /*
114 ** sub_m2_bf16_m_untied:
115 ** fmov (z[0-9]+\.h), #2\.0(?:e\+0)?
116 ** movprfx z0, z1
117 ** bfadd z0\.h, p0/m, z0\.h, \1
118 ** ret
119 */
120 TEST_UNIFORM_Z (sub_m2_bf16_m_untied, svbfloat16_t,
121 z0 = svsub_n_bf16_m (p0, z1, -2),
122 z0 = svsub_m (p0, z1, -2))
123
124 /*
125 ** sub_bf16_z_tied1:
126 ** movprfx z0\.h, p0/z, z0\.h
127 ** bfsub z0\.h, p0/m, z0\.h, z1\.h
128 ** ret
129 */
130 TEST_UNIFORM_Z (sub_bf16_z_tied1, svbfloat16_t,
131 z0 = svsub_bf16_z (p0, z0, z1),
132 z0 = svsub_z (p0, z0, z1))
133
134 /*
135 ** sub_bf16_z_tied2:
136 ** mov (z[0-9]+)\.d, z0\.d
137 ** movprfx z0\.h, p0/z, z1\.h
138 ** bfsub z0\.h, p0/m, z0\.h, \1\.h
139 ** ret
140 */
141 TEST_UNIFORM_Z (sub_bf16_z_tied2, svbfloat16_t,
142 z0 = svsub_bf16_z (p0, z1, z0),
143 z0 = svsub_z (p0, z1, z0))
144
145 /*
146 ** sub_bf16_z_untied:
147 ** movprfx z0\.h, p0/z, z1\.h
148 ** bfsub z0\.h, p0/m, z0\.h, z2\.h
149 ** ret
150 */
151 TEST_UNIFORM_Z (sub_bf16_z_untied, svbfloat16_t,
152 z0 = svsub_bf16_z (p0, z1, z2),
153 z0 = svsub_z (p0, z1, z2))
154
155 /*
156 ** sub_h4_bf16_z_tied1:
157 ** mov (z[0-9]+\.h), h4
158 ** movprfx z0\.h, p0/z, z0\.h
159 ** bfsub z0\.h, p0/m, z0\.h, \1
160 ** ret
161 */
162 TEST_UNIFORM_ZD (sub_h4_bf16_z_tied1, svbfloat16_t, __bf16,
163 z0 = svsub_n_bf16_z (p0, z0, d4),
164 z0 = svsub_z (p0, z0, d4))
165
166 /*
167 ** sub_h4_bf16_z_untied:
168 ** mov (z[0-9]+\.h), h4
169 ** movprfx z0\.h, p0/z, z1\.h
170 ** bfsub z0\.h, p0/m, z0\.h, \1
171 ** ret
172 */
173 TEST_UNIFORM_ZD (sub_h4_bf16_z_untied, svbfloat16_t, __bf16,
174 z0 = svsub_n_bf16_z (p0, z1, d4),
175 z0 = svsub_z (p0, z1, d4))
176
177 /*
178 ** sub_1_bf16_z:
179 ** fmov (z[0-9]+\.h), #-1\.875(?:e\+0)?
180 ** movprfx z0\.h, p0/z, z0\.h
181 ** bfadd z0\.h, p0/m, z0\.h, \1
182 ** ret
183 */
184 TEST_UNIFORM_Z (sub_1_bf16_z, svbfloat16_t,
185 z0 = svsub_n_bf16_z (p0, z0, 1),
186 z0 = svsub_z (p0, z0, 1))
187
188 /*
189 ** sub_bf16_x_tied1:
190 ** bfsub z0\.h, p0/m, z0\.h, z1\.h
191 ** ret
192 */
193 TEST_UNIFORM_Z (sub_bf16_x_tied1, svbfloat16_t,
194 z0 = svsub_bf16_x (p0, z0, z1),
195 z0 = svsub_x (p0, z0, z1))
196
197 /*
198 ** sub_bf16_x_tied2:
199 ** mov (z[0-9]+)\.d, z0\.d
200 ** movprfx z0, z1
201 ** bfsub z0\.h, p0/m, z0\.h, \1\.h
202 ** ret
203 */
204 TEST_UNIFORM_Z (sub_bf16_x_tied2, svbfloat16_t,
205 z0 = svsub_bf16_x (p0, z1, z0),
206 z0 = svsub_x (p0, z1, z0))
207
208 /*
209 ** sub_bf16_x_untied:
210 ** movprfx z0, z1
211 ** bfsub z0\.h, p0/m, z0\.h, z2\.h
212 ** ret
213 */
214 TEST_UNIFORM_Z (sub_bf16_x_untied, svbfloat16_t,
215 z0 = svsub_bf16_x (p0, z1, z2),
216 z0 = svsub_x (p0, z1, z2))
217
218 /*
219 ** sub_h4_bf16_x_tied1:
220 ** mov (z[0-9]+\.h), h4
221 ** bfsub z0\.h, p0/m, z0\.h, \1
222 ** ret
223 */
224 TEST_UNIFORM_ZD (sub_h4_bf16_x_tied1, svbfloat16_t, __bf16,
225 z0 = svsub_n_bf16_x (p0, z0, d4),
226 z0 = svsub_x (p0, z0, d4))
227
228 /*
229 ** sub_h4_bf16_x_untied:
230 ** mov (z[0-9]+\.h), h4
231 ** movprfx z0, z1
232 ** bfsub z0\.h, p0/m, z0\.h, \1
233 ** ret
234 */
235 TEST_UNIFORM_ZD (sub_h4_bf16_x_untied, svbfloat16_t, __bf16,
236 z0 = svsub_n_bf16_x (p0, z1, d4),
237 z0 = svsub_x (p0, z1, d4))
238
239 /*
240 ** sub_1_bf16_x_tied1:
241 ** fmov (z[0-9]+\.h), #-1\.875(?:e\+0)?
242 ** bfadd z0\.h, p0/m, z0\.h, \1
243 ** ret
244 */
245 TEST_UNIFORM_Z (sub_1_bf16_x_tied1, svbfloat16_t,
246 z0 = svsub_n_bf16_x (p0, z0, 1),
247 z0 = svsub_x (p0, z0, 1))
248
249 /*
250 ** sub_1_bf16_x_untied:
251 ** fmov z0\.h, #-1\.875(?:e\+0)?
252 ** bfadd z0\.h, p0/m, z0\.h, z1\.h
253 ** ret
254 */
255 TEST_UNIFORM_Z (sub_1_bf16_x_untied, svbfloat16_t,
256 z0 = svsub_n_bf16_x (p0, z1, 1),
257 z0 = svsub_x (p0, z1, 1))
258
259 /*
260 ** ptrue_sub_bf16_x_tied1:
261 ** bfsub z0\.h, z0\.h, z1\.h
262 ** ret
263 */
264 TEST_UNIFORM_Z (ptrue_sub_bf16_x_tied1, svbfloat16_t,
265 z0 = svsub_bf16_x (svptrue_b16 (), z0, z1),
266 z0 = svsub_x (svptrue_b16 (), z0, z1))
267
268 /*
269 ** ptrue_sub_bf16_x_tied2:
270 ** bfsub z0\.h, z1\.h, z0\.h
271 ** ret
272 */
273 TEST_UNIFORM_Z (ptrue_sub_bf16_x_tied2, svbfloat16_t,
274 z0 = svsub_bf16_x (svptrue_b16 (), z1, z0),
275 z0 = svsub_x (svptrue_b16 (), z1, z0))
276
277 /*
278 ** ptrue_sub_bf16_x_untied:
279 ** bfsub z0\.h, z1\.h, z2\.h
280 ** ret
281 */
282 TEST_UNIFORM_Z (ptrue_sub_bf16_x_untied, svbfloat16_t,
283 z0 = svsub_bf16_x (svptrue_b16 (), z1, z2),
284 z0 = svsub_x (svptrue_b16 (), z1, z2))
285
286 /*
287 ** ptrue_sub_1_bf16_x_tied1:
288 ** fmov (z[0-9]+\.h), #-1\.875(?:e\+0)?
289 ** bfadd z0\.h, (z0\.h, \1|\1, z0\.h)
290 ** ret
291 */
292 TEST_UNIFORM_Z (ptrue_sub_1_bf16_x_tied1, svbfloat16_t,
293 z0 = svsub_n_bf16_x (svptrue_b16 (), z0, 1),
294 z0 = svsub_x (svptrue_b16 (), z0, 1))
295
296 /*
297 ** ptrue_sub_1_bf16_x_untied:
298 ** fmov (z[0-9]+\.h), #-1\.875(?:e\+0)?
299 ** bfadd z0\.h, (z1\.h, \1|\1, z1\.h)
300 ** ret
301 */
302 TEST_UNIFORM_Z (ptrue_sub_1_bf16_x_untied, svbfloat16_t,
303 z0 = svsub_n_bf16_x (svptrue_b16 (), z1, 1),
304 z0 = svsub_x (svptrue_b16 (), z1, 1))