Debian Codegraph @

gcc/testsuite/gcc.target/aarch64/sme2/acle-asm/mla_za16_bf16_vg1x2.c

raw
1 /* { dg-do assemble { target aarch64_asm_sme-b16b16_ok } } */
2 /* { dg-do compile { target { ! aarch64_asm_sme-b16b16_ok } } } */
3 /* { dg-final { check-function-bodies "**" "" "-DCHECK_ASM" } } */
4
5 #include "test_sme2_acle.h"
6
7 #pragma GCC target "+sme-b16b16"
8
9 /*
10 ** mla_0_z0_z0:
11 ** mov (w8|w9|w10|w11), #?0
12 ** bfmla za\.h\[\1, 0, vgx2\], {z0\.h - z1\.h}, {z0\.h - z1\.h}
13 ** ret
14 */
15 TEST_ZA_XN (mla_0_z0_z0, svbfloat16x2_t,
16 svmla_za16_bf16_vg1x2 (0, z0, z0),
17 svmla_za16_vg1x2 (0, z0, z0))
18
19 /*
20 ** mla_w0_z0_z0:
21 ** mov (w8|w9|w10|w11), w0
22 ** bfmla za\.h\[\1, 0, vgx2\], {z0\.h - z1\.h}, {z0\.h - z1\.h}
23 ** ret
24 */
25 TEST_ZA_XN (mla_w0_z0_z0, svbfloat16x2_t,
26 svmla_za16_bf16_vg1x2 (w0, z0, z0),
27 svmla_za16_vg1x2 (w0, z0, z0))
28
29 /*
30 ** mla_w8_z0_z4:
31 ** bfmla za\.h\[w8, 0, vgx2\], {z0\.h - z1\.h}, {z4\.h - z5\.h}
32 ** ret
33 */
34 TEST_ZA_XN (mla_w8_z0_z4, svbfloat16x2_t,
35 svmla_za16_bf16_vg1x2 (w8, z0, z4),
36 svmla_za16_vg1x2 (w8, z0, z4))
37
38 /*
39 ** mla_w8_z4_z18:
40 ** bfmla za\.h\[w8, 0, vgx2\], {z4\.h - z5\.h}, {z18\.h - z19\.h}
41 ** ret
42 */
43 TEST_ZA_XN (mla_w8_z4_z18, svbfloat16x2_t,
44 svmla_za16_bf16_vg1x2 (w8, z4, z18),
45 svmla_za16_vg1x2 (w8, z4, z18))
46
47 /* Leave the assembler to check for correctness for misaligned registers. */
48
49 /*
50 ** mla_w8_z23_z0:
51 ** ...
52 ** bfmla za\.h\[w8, 0, vgx2\], [^\n]+, {z0\.h - z1\.h}
53 ** ret
54 */
55 TEST_ZA_XN (mla_w8_z23_z0, svbfloat16x2_t,
56 svmla_za16_bf16_vg1x2 (w8, z23, z0),
57 svmla_za16_vg1x2 (w8, z23, z0))
58
59 /*
60 ** mla_w8_z18_z23:
61 ** ...
62 ** bfmla za\.h\[w8, 0, vgx2\], {z18\.h - z19\.h}, [^\n]+
63 ** ret
64 */
65 TEST_ZA_XN (mla_w8_z18_z23, svbfloat16x2_t,
66 svmla_za16_bf16_vg1x2 (w8, z18, z23),
67 svmla_za16_vg1x2 (w8, z18, z23))
68
69 /*
70 ** mla_w8_z4_z28:
71 ** bfmla za\.h\[w8, 0, vgx2\], {z4\.h - z5\.h}, {z28\.h - z29\.h}
72 ** ret
73 */
74 TEST_ZA_XN (mla_w8_z4_z28, svbfloat16x2_t,
75 svmla_za16_bf16_vg1x2 (w8, z4, z28),
76 svmla_za16_vg1x2 (w8, z4, z28))
77
78 /*
79 ** mla_w8p7_z4_z0:
80 ** bfmla za\.h\[w8, 7, vgx2\], {z4\.h - z5\.h}, {z0\.h - z1\.h}
81 ** ret
82 */
83 TEST_ZA_XN (mla_w8p7_z4_z0, svbfloat16x2_t,
84 svmla_za16_bf16_vg1x2 (w8 + 7, z4, z0),
85 svmla_za16_vg1x2 (w8 + 7, z4, z0))
86
87 /*
88 ** mla_w8p8_z4_z4:
89 ** add (w8|w9|w10|w11), w8, #?8
90 ** bfmla za\.h\[\1, 0, vgx2\], {z4\.h - z5\.h}, {z4\.h - z5\.h}
91 ** ret
92 */
93 TEST_ZA_XN (mla_w8p8_z4_z4, svbfloat16x2_t,
94 svmla_za16_bf16_vg1x2 (w8 + 8, z4, z4),
95 svmla_za16_vg1x2 (w8 + 8, z4, z4))
96
97 /*
98 ** mla_w8m1_z4_z0:
99 ** sub (w8|w9|w10|w11), w8, #?1
100 ** bfmla za\.h\[\1, 0, vgx2\], {z4\.h - z5\.h}, {z0\.h - z1\.h}
101 ** ret
102 */
103 TEST_ZA_XN (mla_w8m1_z4_z0, svbfloat16x2_t,
104 svmla_za16_bf16_vg1x2 (w8 - 1, z4, z0),
105 svmla_za16_vg1x2 (w8 - 1, z4, z0))
106
107 /*
108 ** mla_single_0_z1_z0:
109 ** mov (w8|w9|w10|w11), #?0
110 ** bfmla za\.h\[\1, 0, vgx2\], {z1\.h - z2\.h}, z0\.h
111 ** ret
112 */
113 TEST_ZA_SINGLE (mla_single_0_z1_z0, svbfloat16x2_t, svbfloat16_t,
114 svmla_single_za16_bf16_vg1x2 (0, z1, z0),
115 svmla_za16_vg1x2 (0, z1, z0))
116
117 /*
118 ** mla_single_w0_z1_z0:
119 ** mov (w8|w9|w10|w11), w0
120 ** bfmla za\.h\[\1, 0, vgx2\], {z1\.h - z2\.h}, z0\.h
121 ** ret
122 */
123 TEST_ZA_SINGLE (mla_single_w0_z1_z0, svbfloat16x2_t, svbfloat16_t,
124 svmla_single_za16_bf16_vg1x2 (w0, z1, z0),
125 svmla_za16_vg1x2 (w0, z1, z0))
126
127 /*
128 ** mla_single_w8_z1_z0:
129 ** bfmla za\.h\[w8, 0, vgx2\], {z1\.h - z2\.h}, z0\.h
130 ** ret
131 */
132 TEST_ZA_SINGLE (mla_single_w8_z1_z0, svbfloat16x2_t, svbfloat16_t,
133 svmla_single_za16_bf16_vg1x2 (w8, z1, z0),
134 svmla_za16_vg1x2 (w8, z1, z0))
135
136 /*
137 ** mla_single_w8p7_z1_z0:
138 ** bfmla za\.h\[w8, 7, vgx2\], {z1\.h - z2\.h}, z0\.h
139 ** ret
140 */
141 TEST_ZA_SINGLE (mla_single_w8p7_z1_z0, svbfloat16x2_t, svbfloat16_t,
142 svmla_single_za16_bf16_vg1x2 (w8 + 7, z1, z0),
143 svmla_za16_vg1x2 (w8 + 7, z1, z0))
144
145 /*
146 ** mla_single_w8p8_z1_z0:
147 ** add (w8|w9|w10|w11), w8, #?8
148 ** bfmla za\.h\[\1, 0, vgx2\], {z1\.h - z2\.h}, z0\.h
149 ** ret
150 */
151 TEST_ZA_SINGLE (mla_single_w8p8_z1_z0, svbfloat16x2_t, svbfloat16_t,
152 svmla_single_za16_bf16_vg1x2 (w8 + 8, z1, z0),
153 svmla_za16_vg1x2 (w8 + 8, z1, z0))
154
155 /*
156 ** mla_single_w0m1_z1_z0:
157 ** sub (w8|w9|w10|w11), w0, #?1
158 ** bfmla za\.h\[\1, 0, vgx2\], {z1\.h - z2\.h}, z0\.h
159 ** ret
160 */
161 TEST_ZA_SINGLE (mla_single_w0m1_z1_z0, svbfloat16x2_t, svbfloat16_t,
162 svmla_single_za16_bf16_vg1x2 (w0 - 1, z1, z0),
163 svmla_za16_vg1x2 (w0 - 1, z1, z0))
164
165 /*
166 ** mla_single_w8_z0_z15:
167 ** str d15, \[sp, #?-16\]!
168 ** bfmla za\.h\[w8, 0, vgx2\], {z0\.h - z1\.h}, z15\.h
169 ** ldr d15, \[sp\], #?16
170 ** ret
171 */
172 TEST_ZA_SINGLE_Z15 (mla_single_w8_z0_z15, svbfloat16x2_t, svbfloat16_t,
173 svmla_single_za16_bf16_vg1x2 (w8, z0, z15),
174 svmla_za16_vg1x2 (w8, z0, z15))
175
176 /*
177 ** mla_single_w8_z20_z16:
178 ** mov (z[0-7]).d, z16.d
179 ** bfmla za\.h\[w8, 0, vgx2\], {z20\.h - z21\.h}, \1\.h
180 ** ret
181 */
182 TEST_ZA_SINGLE (mla_single_w8_z20_z16, svbfloat16x2_t, svbfloat16_t,
183 svmla_single_za16_bf16_vg1x2 (w8, z20, z16),
184 svmla_za16_vg1x2 (w8, z20, z16))