Debian Codegraph @

gcc/testsuite/gcc.target/aarch64/sme2/acle-asm/dot_za32_bf16_vg1x4.c

raw
1 /* { dg-final { check-function-bodies "**" "" "-DCHECK_ASM" } } */
2
3 #include "test_sme2_acle.h"
4
5 /*
6 ** dot_0_z0_z0:
7 ** mov (w8|w9|w10|w11), #?0
8 ** bfdot za\.s\[\1, 0, vgx4\], {z0\.h - z3\.h}, {z0\.h - z3\.h}
9 ** ret
10 */
11 TEST_ZA_XN (dot_0_z0_z0, svbfloat16x4_t,
12 svdot_za32_bf16_vg1x4 (0, z0, z0),
13 svdot_za32_vg1x4 (0, z0, z0))
14
15 /*
16 ** dot_w0_z0_z0:
17 ** mov (w8|w9|w10|w11), w0
18 ** bfdot za\.s\[\1, 0, vgx4\], {z0\.h - z3\.h}, {z0\.h - z3\.h}
19 ** ret
20 */
21 TEST_ZA_XN (dot_w0_z0_z0, svbfloat16x4_t,
22 svdot_za32_bf16_vg1x4 (w0, z0, z0),
23 svdot_za32_vg1x4 (w0, z0, z0))
24
25 /*
26 ** dot_w8_z0_z4:
27 ** bfdot za\.s\[w8, 0, vgx4\], {z0\.h - z3\.h}, {z4\.h - z7\.h}
28 ** ret
29 */
30 TEST_ZA_XN (dot_w8_z0_z4, svbfloat16x4_t,
31 svdot_za32_bf16_vg1x4 (w8, z0, z4),
32 svdot_za32_vg1x4 (w8, z0, z4))
33
34 /* Leave the assembler to check for correctness for misaligned registers. */
35
36 /*
37 ** dot_w8_z0_z18:
38 ** ...
39 ** bfdot za\.s\[w8, 0, vgx4\], {z0\.h - z3\.h}, [^\n]+
40 ** ret
41 */
42 TEST_ZA_XN (dot_w8_z0_z18, svbfloat16x4_t,
43 svdot_za32_bf16_vg1x4 (w8, z0, z18),
44 svdot_za32_vg1x4 (w8, z0, z18))
45
46 /*
47 ** dot_w8_z18_z0:
48 ** ...
49 ** bfdot za\.s\[w8, 0, vgx4\], [^\n]+, {z0\.h - z3\.h}
50 ** ret
51 */
52 TEST_ZA_XN (dot_w8_z18_z0, svbfloat16x4_t,
53 svdot_za32_bf16_vg1x4 (w8, z18, z0),
54 svdot_za32_vg1x4 (w8, z18, z0))
55
56 /*
57 ** dot_w8_z0_z23:
58 ** ...
59 ** bfdot za\.s\[w8, 0, vgx4\], {z0\.h - z3\.h}, [^\n]+
60 ** ret
61 */
62 TEST_ZA_XN (dot_w8_z0_z23, svbfloat16x4_t,
63 svdot_za32_bf16_vg1x4 (w8, z0, z23),
64 svdot_za32_vg1x4 (w8, z0, z23))
65
66 /*
67 ** dot_w8_z23_z0:
68 ** ...
69 ** bfdot za\.s\[w8, 0, vgx4\], [^\n]+, {z0\.h - z3\.h}
70 ** ret
71 */
72 TEST_ZA_XN (dot_w8_z23_z0, svbfloat16x4_t,
73 svdot_za32_bf16_vg1x4 (w8, z23, z0),
74 svdot_za32_vg1x4 (w8, z23, z0))
75
76 /*
77 ** dot_w8_z4_z28:
78 ** bfdot za\.s\[w8, 0, vgx4\], {z4\.h - z7\.h}, {z28\.h - z31\.h}
79 ** ret
80 */
81 TEST_ZA_XN (dot_w8_z4_z28, svbfloat16x4_t,
82 svdot_za32_bf16_vg1x4 (w8, z4, z28),
83 svdot_za32_vg1x4 (w8, z4, z28))
84
85 /*
86 ** dot_w8_z28_z0:
87 ** bfdot za\.s\[w8, 0, vgx4\], {z28\.h - z31\.h}, {z0\.h - z3\.h}
88 ** ret
89 */
90 TEST_ZA_XN (dot_w8_z28_z0, svbfloat16x4_t,
91 svdot_za32_bf16_vg1x4 (w8, z28, z0),
92 svdot_za32_vg1x4 (w8, z28, z0))
93
94 /*
95 ** dot_w8p1_z4_z0:
96 ** bfdot za\.s\[w8, 1, vgx4\], {z4\.h - z7\.h}, {z0\.h - z3\.h}
97 ** ret
98 */
99 TEST_ZA_XN (dot_w8p1_z4_z0, svbfloat16x4_t,
100 svdot_za32_bf16_vg1x4 (w8 + 1, z4, z0),
101 svdot_za32_vg1x4 (w8 + 1, z4, z0))
102
103 /*
104 ** dot_w8p2_z4_z0:
105 ** bfdot za\.s\[w8, 2, vgx4\], {z4\.h - z7\.h}, {z0\.h - z3\.h}
106 ** ret
107 */
108 TEST_ZA_XN (dot_w8p2_z4_z0, svbfloat16x4_t,
109 svdot_za32_bf16_vg1x4 (w8 + 2, z4, z0),
110 svdot_za32_vg1x4 (w8 + 2, z4, z0))
111
112 /*
113 ** dot_w11p4_z4_z0:
114 ** bfdot za\.s\[w11, 4, vgx4\], {z4\.h - z7\.h}, {z0\.h - z3\.h}
115 ** ret
116 */
117 TEST_ZA_XN (dot_w11p4_z4_z0, svbfloat16x4_t,
118 svdot_za32_bf16_vg1x4 (w11 + 4, z4, z0),
119 svdot_za32_vg1x4 (w11 + 4, z4, z0))
120
121 /*
122 ** dot_w8p7_z4_z0:
123 ** bfdot za\.s\[w8, 7, vgx4\], {z4\.h - z7\.h}, {z0\.h - z3\.h}
124 ** ret
125 */
126 TEST_ZA_XN (dot_w8p7_z4_z0, svbfloat16x4_t,
127 svdot_za32_bf16_vg1x4 (w8 + 7, z4, z0),
128 svdot_za32_vg1x4 (w8 + 7, z4, z0))
129
130 /*
131 ** dot_w8p8_z4_z4:
132 ** add (w8|w9|w10|w11), w8, #?8
133 ** bfdot za\.s\[\1, 0, vgx4\], {z4\.h - z7\.h}, {z4\.h - z7\.h}
134 ** ret
135 */
136 TEST_ZA_XN (dot_w8p8_z4_z4, svbfloat16x4_t,
137 svdot_za32_bf16_vg1x4 (w8 + 8, z4, z4),
138 svdot_za32_vg1x4 (w8 + 8, z4, z4))
139
140 /*
141 ** dot_w8m1_z4_z0:
142 ** sub (w8|w9|w10|w11), w8, #?1
143 ** bfdot za\.s\[\1, 0, vgx4\], {z4\.h - z7\.h}, {z0\.h - z3\.h}
144 ** ret
145 */
146 TEST_ZA_XN (dot_w8m1_z4_z0, svbfloat16x4_t,
147 svdot_za32_bf16_vg1x4 (w8 - 1, z4, z0),
148 svdot_za32_vg1x4 (w8 - 1, z4, z0))
149
150 /*
151 ** dot_single_0_z1_z0:
152 ** mov (w8|w9|w10|w11), #?0
153 ** bfdot za\.s\[\1, 0, vgx4\], {z1\.h - z4\.h}, z0\.h
154 ** ret
155 */
156 TEST_ZA_SINGLE (dot_single_0_z1_z0, svbfloat16x4_t, svbfloat16_t,
157 svdot_single_za32_bf16_vg1x4 (0, z1, z0),
158 svdot_za32_vg1x4 (0, z1, z0))
159
160 /*
161 ** dot_single_w0_z1_z0:
162 ** mov (w8|w9|w10|w11), w0
163 ** bfdot za\.s\[\1, 0, vgx4\], {z1\.h - z4\.h}, z0\.h
164 ** ret
165 */
166 TEST_ZA_SINGLE (dot_single_w0_z1_z0, svbfloat16x4_t, svbfloat16_t,
167 svdot_single_za32_bf16_vg1x4 (w0, z1, z0),
168 svdot_za32_vg1x4 (w0, z1, z0))
169
170 /*
171 ** dot_single_w8_z1_z0:
172 ** bfdot za\.s\[w8, 0, vgx4\], {z1\.h - z4\.h}, z0\.h
173 ** ret
174 */
175 TEST_ZA_SINGLE (dot_single_w8_z1_z0, svbfloat16x4_t, svbfloat16_t,
176 svdot_single_za32_bf16_vg1x4 (w8, z1, z0),
177 svdot_za32_vg1x4 (w8, z1, z0))
178
179 /*
180 ** dot_single_w8p1_z1_z0:
181 ** bfdot za\.s\[w8, 1, vgx4\], {z1\.h - z4\.h}, z0\.h
182 ** ret
183 */
184 TEST_ZA_SINGLE (dot_single_w8p1_z1_z0, svbfloat16x4_t, svbfloat16_t,
185 svdot_single_za32_bf16_vg1x4 (w8 + 1, z1, z0),
186 svdot_za32_vg1x4 (w8 + 1, z1, z0))
187
188 /*
189 ** dot_single_w8p4_z20_z0:
190 ** bfdot za\.s\[w8, 4, vgx4\], {z20\.h - z23\.h}, z0\.h
191 ** ret
192 */
193 TEST_ZA_SINGLE (dot_single_w8p4_z20_z0, svbfloat16x4_t, svbfloat16_t,
194 svdot_single_za32_bf16_vg1x4 (w8 + 4, z20, z0),
195 svdot_za32_vg1x4 (w8 + 4, z20, z0))
196
197 /*
198 ** dot_single_w8p6_z27_z0:
199 ** bfdot za\.s\[w8, 6, vgx4\], {z27\.h - z30\.h}, z0\.h
200 ** ret
201 */
202 TEST_ZA_SINGLE (dot_single_w8p6_z27_z0, svbfloat16x4_t, svbfloat16_t,
203 svdot_single_za32_bf16_vg1x4 (w8 + 6, z27, z0),
204 svdot_za32_vg1x4 (w8 + 6, z27, z0))
205
206 /*
207 ** dot_single_w8p7_z1_z0:
208 ** bfdot za\.s\[w8, 7, vgx4\], {z1\.h - z4\.h}, z0\.h
209 ** ret
210 */
211 TEST_ZA_SINGLE (dot_single_w8p7_z1_z0, svbfloat16x4_t, svbfloat16_t,
212 svdot_single_za32_bf16_vg1x4 (w8 + 7, z1, z0),
213 svdot_za32_vg1x4 (w8 + 7, z1, z0))
214
215 /*
216 ** dot_single_w8p8_z1_z0:
217 ** add (w8|w9|w10|w11), w8, #?8
218 ** bfdot za\.s\[\1, 0, vgx4\], {z1\.h - z4\.h}, z0\.h
219 ** ret
220 */
221 TEST_ZA_SINGLE (dot_single_w8p8_z1_z0, svbfloat16x4_t, svbfloat16_t,
222 svdot_single_za32_bf16_vg1x4 (w8 + 8, z1, z0),
223 svdot_za32_vg1x4 (w8 + 8, z1, z0))
224
225 /*
226 ** dot_single_w0m1_z1_z0:
227 ** sub (w8|w9|w10|w11), w0, #?1
228 ** bfdot za\.s\[\1, 0, vgx4\], {z1\.h - z4\.h}, z0\.h
229 ** ret
230 */
231 TEST_ZA_SINGLE (dot_single_w0m1_z1_z0, svbfloat16x4_t, svbfloat16_t,
232 svdot_single_za32_bf16_vg1x4 (w0 - 1, z1, z0),
233 svdot_za32_vg1x4 (w0 - 1, z1, z0))
234
235 /*
236 ** dot_single_w8_z0_z15:
237 ** str d15, \[sp, #?-16\]!
238 ** bfdot za\.s\[w8, 0, vgx4\], {z0\.h - z3\.h}, z15\.h
239 ** ldr d15, \[sp\], #?16
240 ** ret
241 */
242 TEST_ZA_SINGLE_Z15 (dot_single_w8_z0_z15, svbfloat16x4_t, svbfloat16_t,
243 svdot_single_za32_bf16_vg1x4 (w8, z0, z15),
244 svdot_za32_vg1x4 (w8, z0, z15))
245
246 /*
247 ** dot_single_w8_z20_z16:
248 ** mov (z[0-7]).d, z16.d
249 ** bfdot za\.s\[w8, 0, vgx4\], {z20\.h - z23\.h}, \1\.h
250 ** ret
251 */
252 TEST_ZA_SINGLE (dot_single_w8_z20_z16, svbfloat16x4_t, svbfloat16_t,
253 svdot_single_za32_bf16_vg1x4 (w8, z20, z16),
254 svdot_za32_vg1x4 (w8, z20, z16))