Debian Codegraph @

gcc/testsuite/gcc.target/aarch64/sme2/acle-asm/mla_za32_bf16_vg2x2.c

raw
1 /* { dg-final { check-function-bodies "**" "" "-DCHECK_ASM" } } */
2
3 #include "test_sme2_acle.h"
4
5 /*
6 ** mla_0_z0_z0:
7 ** mov (w8|w9|w10|w11), #?0
8 ** bfmlal za\.s\[\1, 0:1, vgx2\], {z0\.h - z1\.h}, {z0\.h - z1\.h}
9 ** ret
10 */
11 TEST_ZA_XN (mla_0_z0_z0, svbfloat16x2_t,
12 svmla_za32_bf16_vg2x2 (0, z0, z0),
13 svmla_za32_vg2x2 (0, z0, z0))
14
15 /*
16 ** mla_w0_z0_z0:
17 ** mov (w8|w9|w10|w11), w0
18 ** bfmlal za\.s\[\1, 0:1, vgx2\], {z0\.h - z1\.h}, {z0\.h - z1\.h}
19 ** ret
20 */
21 TEST_ZA_XN (mla_w0_z0_z0, svbfloat16x2_t,
22 svmla_za32_bf16_vg2x2 (w0, z0, z0),
23 svmla_za32_vg2x2 (w0, z0, z0))
24
25 /*
26 ** mla_w8_z0_z4:
27 ** bfmlal za\.s\[w8, 0:1, vgx2\], {z0\.h - z1\.h}, {z4\.h - z5\.h}
28 ** ret
29 */
30 TEST_ZA_XN (mla_w8_z0_z4, svbfloat16x2_t,
31 svmla_za32_bf16_vg2x2 (w8, z0, z4),
32 svmla_za32_vg2x2 (w8, z0, z4))
33
34 /*
35 ** mla_w8_z4_z18:
36 ** bfmlal za\.s\[w8, 0:1, vgx2\], {z4\.h - z5\.h}, {z18\.h - z19\.h}
37 ** ret
38 */
39 TEST_ZA_XN (mla_w8_z4_z18, svbfloat16x2_t,
40 svmla_za32_bf16_vg2x2 (w8, z4, z18),
41 svmla_za32_vg2x2 (w8, z4, z18))
42
43 /* Leave the assembler to check for correctness for misaligned registers. */
44
45 /*
46 ** mla_w8_z0_z23:
47 ** ...
48 ** bfmlal za\.s\[w8, 0:1, vgx2\], {z0\.h - z1\.h}, [^\n]+
49 ** ret
50 */
51 TEST_ZA_XN (mla_w8_z0_z23, svbfloat16x2_t,
52 svmla_za32_bf16_vg2x2 (w8, z0, z23),
53 svmla_za32_vg2x2 (w8, z0, z23))
54
55 /*
56 ** mla_w8_z23_z0:
57 ** ...
58 ** bfmlal za\.s\[w8, 0:1, vgx2\], [^\n]+, {z0\.h - z1\.h}
59 ** ret
60 */
61 TEST_ZA_XN (mla_w8_z23_z0, svbfloat16x2_t,
62 svmla_za32_bf16_vg2x2 (w8, z23, z0),
63 svmla_za32_vg2x2 (w8, z23, z0))
64
65 /*
66 ** mla_w8_z18_z28:
67 ** bfmlal za\.s\[w8, 0:1, vgx2\], {z18\.h - z19\.h}, {z28\.h - z29\.h}
68 ** ret
69 */
70 TEST_ZA_XN (mla_w8_z18_z28, svbfloat16x2_t,
71 svmla_za32_bf16_vg2x2 (w8, z18, z28),
72 svmla_za32_vg2x2 (w8, z18, z28))
73
74 /*
75 ** mla_w8_z28_z4:
76 ** bfmlal za\.s\[w8, 0:1, vgx2\], {z28\.h - z29\.h}, {z4\.h - z5\.h}
77 ** ret
78 */
79 TEST_ZA_XN (mla_w8_z28_z4, svbfloat16x2_t,
80 svmla_za32_bf16_vg2x2 (w8, z28, z4),
81 svmla_za32_vg2x2 (w8, z28, z4))
82
83 /*
84 ** mla_w8p1_z4_z0:
85 ** add (w8|w9|w10|w11), w8, #?1
86 ** bfmlal za\.s\[\1, 0:1, vgx2\], {z4\.h - z5\.h}, {z0\.h - z1\.h}
87 ** ret
88 */
89 TEST_ZA_XN (mla_w8p1_z4_z0, svbfloat16x2_t,
90 svmla_za32_bf16_vg2x2 (w8 + 1, z4, z0),
91 svmla_za32_vg2x2 (w8 + 1, z4, z0))
92
93 /*
94 ** mla_w8p2_z4_z0:
95 ** bfmlal za\.s\[w8, 2:3, vgx2\], {z4\.h - z5\.h}, {z0\.h - z1\.h}
96 ** ret
97 */
98 TEST_ZA_XN (mla_w8p2_z4_z0, svbfloat16x2_t,
99 svmla_za32_bf16_vg2x2 (w8 + 2, z4, z0),
100 svmla_za32_vg2x2 (w8 + 2, z4, z0))
101
102 /*
103 ** mla_w8p6_z4_z0:
104 ** bfmlal za\.s\[w8, 6:7, vgx2\], {z4\.h - z5\.h}, {z0\.h - z1\.h}
105 ** ret
106 */
107 TEST_ZA_XN (mla_w8p6_z4_z0, svbfloat16x2_t,
108 svmla_za32_bf16_vg2x2 (w8 + 6, z4, z0),
109 svmla_za32_vg2x2 (w8 + 6, z4, z0))
110
111 /*
112 ** mla_w8p7_z4_z0:
113 ** add (w8|w9|w10|w11), w8, #?7
114 ** bfmlal za\.s\[\1, 0:1, vgx2\], {z4\.h - z5\.h}, {z0\.h - z1\.h}
115 ** ret
116 */
117 TEST_ZA_XN (mla_w8p7_z4_z0, svbfloat16x2_t,
118 svmla_za32_bf16_vg2x2 (w8 + 7, z4, z0),
119 svmla_za32_vg2x2 (w8 + 7, z4, z0))
120
121 /*
122 ** mla_w8p8_z4_z4:
123 ** add (w8|w9|w10|w11), w8, #?8
124 ** bfmlal za\.s\[\1, 0:1, vgx2\], {z4\.h - z5\.h}, {z4\.h - z5\.h}
125 ** ret
126 */
127 TEST_ZA_XN (mla_w8p8_z4_z4, svbfloat16x2_t,
128 svmla_za32_bf16_vg2x2 (w8 + 8, z4, z4),
129 svmla_za32_vg2x2 (w8 + 8, z4, z4))
130
131 /*
132 ** mla_w8m1_z4_z0:
133 ** sub (w8|w9|w10|w11), w8, #?1
134 ** bfmlal za\.s\[\1, 0:1, vgx2\], {z4\.h - z5\.h}, {z0\.h - z1\.h}
135 ** ret
136 */
137 TEST_ZA_XN (mla_w8m1_z4_z0, svbfloat16x2_t,
138 svmla_za32_bf16_vg2x2 (w8 - 1, z4, z0),
139 svmla_za32_vg2x2 (w8 - 1, z4, z0))
140
141 /*
142 ** mla_single_0_z1_z0:
143 ** mov (w8|w9|w10|w11), #?0
144 ** bfmlal za\.s\[\1, 0:1, vgx2\], {z1\.h - z2\.h}, z0\.h
145 ** ret
146 */
147 TEST_ZA_SINGLE (mla_single_0_z1_z0, svbfloat16x2_t, svbfloat16_t,
148 svmla_single_za32_bf16_vg2x2 (0, z1, z0),
149 svmla_za32_vg2x2 (0, z1, z0))
150
151 /*
152 ** mla_single_w0_z1_z0:
153 ** mov (w8|w9|w10|w11), w0
154 ** bfmlal za\.s\[\1, 0:1, vgx2\], {z1\.h - z2\.h}, z0\.h
155 ** ret
156 */
157 TEST_ZA_SINGLE (mla_single_w0_z1_z0, svbfloat16x2_t, svbfloat16_t,
158 svmla_single_za32_bf16_vg2x2 (w0, z1, z0),
159 svmla_za32_vg2x2 (w0, z1, z0))
160
161 /*
162 ** mla_single_w8_z1_z0:
163 ** bfmlal za\.s\[w8, 0:1, vgx2\], {z1\.h - z2\.h}, z0\.h
164 ** ret
165 */
166 TEST_ZA_SINGLE (mla_single_w8_z1_z0, svbfloat16x2_t, svbfloat16_t,
167 svmla_single_za32_bf16_vg2x2 (w8, z1, z0),
168 svmla_za32_vg2x2 (w8, z1, z0))
169
170 /*
171 ** mla_single_w8p1_z1_z0:
172 ** add (w8|w9|w10|w11), w8, #?1
173 ** bfmlal za\.s\[\1, 0:1, vgx2\], {z1\.h - z2\.h}, z0\.h
174 ** ret
175 */
176 TEST_ZA_SINGLE (mla_single_w8p1_z1_z0, svbfloat16x2_t, svbfloat16_t,
177 svmla_single_za32_bf16_vg2x2 (w8 + 1, z1, z0),
178 svmla_za32_vg2x2 (w8 + 1, z1, z0))
179
180 /*
181 ** mla_single_w8p4_z20_z0:
182 ** bfmlal za\.s\[w8, 4:5, vgx2\], {z20\.h - z21\.h}, z0\.h
183 ** ret
184 */
185 TEST_ZA_SINGLE (mla_single_w8p4_z20_z0, svbfloat16x2_t, svbfloat16_t,
186 svmla_single_za32_bf16_vg2x2 (w8 + 4, z20, z0),
187 svmla_za32_vg2x2 (w8 + 4, z20, z0))
188
189 /*
190 ** mla_single_w8p6_z27_z0:
191 ** bfmlal za\.s\[w8, 6:7, vgx2\], {z27\.h - z28\.h}, z0\.h
192 ** ret
193 */
194 TEST_ZA_SINGLE (mla_single_w8p6_z27_z0, svbfloat16x2_t, svbfloat16_t,
195 svmla_single_za32_bf16_vg2x2 (w8 + 6, z27, z0),
196 svmla_za32_vg2x2 (w8 + 6, z27, z0))
197
198 /*
199 ** mla_single_w8p7_z1_z0:
200 ** add (w8|w9|w10|w11), w8, #?7
201 ** bfmlal za\.s\[\1, 0:1, vgx2\], {z1\.h - z2\.h}, z0\.h
202 ** ret
203 */
204 TEST_ZA_SINGLE (mla_single_w8p7_z1_z0, svbfloat16x2_t, svbfloat16_t,
205 svmla_single_za32_bf16_vg2x2 (w8 + 7, z1, z0),
206 svmla_za32_vg2x2 (w8 + 7, z1, z0))
207
208 /*
209 ** mla_single_w8p8_z1_z0:
210 ** add (w8|w9|w10|w11), w8, #?8
211 ** bfmlal za\.s\[\1, 0:1, vgx2\], {z1\.h - z2\.h}, z0\.h
212 ** ret
213 */
214 TEST_ZA_SINGLE (mla_single_w8p8_z1_z0, svbfloat16x2_t, svbfloat16_t,
215 svmla_single_za32_bf16_vg2x2 (w8 + 8, z1, z0),
216 svmla_za32_vg2x2 (w8 + 8, z1, z0))
217
218 /*
219 ** mla_single_w0m1_z1_z0:
220 ** sub (w8|w9|w10|w11), w0, #?1
221 ** bfmlal za\.s\[\1, 0:1, vgx2\], {z1\.h - z2\.h}, z0\.h
222 ** ret
223 */
224 TEST_ZA_SINGLE (mla_single_w0m1_z1_z0, svbfloat16x2_t, svbfloat16_t,
225 svmla_single_za32_bf16_vg2x2 (w0 - 1, z1, z0),
226 svmla_za32_vg2x2 (w0 - 1, z1, z0))
227
228 /*
229 ** mla_single_w8_z0_z15:
230 ** str d15, \[sp, #?-16\]!
231 ** bfmlal za\.s\[w8, 0:1, vgx2\], {z0\.h - z1\.h}, z15\.h
232 ** ldr d15, \[sp\], #?16
233 ** ret
234 */
235 TEST_ZA_SINGLE_Z15 (mla_single_w8_z0_z15, svbfloat16x2_t, svbfloat16_t,
236 svmla_single_za32_bf16_vg2x2 (w8, z0, z15),
237 svmla_za32_vg2x2 (w8, z0, z15))
238
239 /*
240 ** mla_single_w8_z20_z16:
241 ** mov (z[0-7]).d, z16.d
242 ** bfmlal za\.s\[w8, 0:1, vgx2\], {z20\.h - z21\.h}, \1\.h
243 ** ret
244 */
245 TEST_ZA_SINGLE (mla_single_w8_z20_z16, svbfloat16x2_t, svbfloat16_t,
246 svmla_single_za32_bf16_vg2x2 (w8, z20, z16),
247 svmla_za32_vg2x2 (w8, z20, z16))