1 /* { dg-do run } */
2 /* { dg-require-effective-target arm_v8_2a_fp16_neon_hw } */
3 /* { dg-add-options arm_v8_2a_fp16_neon } */
4 /* { dg-skip-if "" { arm*-*-* } } */
5
6 #include <arm_neon.h>
7 #include "arm-neon-ref.h"
8 #include "compute-ref-data.h"
9
10 #define FP16_C(a) ((__fp16) a)
11 #define A FP16_C (13.4)
12 #define B FP16_C (__builtin_inff ())
13 #define C FP16_C (-34.8)
14 #define D FP16_C (-__builtin_inff ())
15 #define E FP16_C (-0.0)
16 #define F FP16_C (19.1)
17 #define G FP16_C (-4.8)
18 #define H FP16_C (0.0)
19
20 float16_t elemE = E;
21 float16_t elemF = F;
22 float16_t elemG = G;
23 float16_t elemH = H;
24
25 #define I FP16_C (0.7)
26 #define J FP16_C (-78)
27 #define K FP16_C (11.23)
28 #define L FP16_C (98)
29 #define M FP16_C (87.1)
30 #define N FP16_C (-8)
31 #define O FP16_C (-1.1)
32 #define P FP16_C (-9.7)
33
34 /* Expected results for vmulx_n. */
35 VECT_VAR_DECL (expected0_static, hfloat, 16, 4) []
36 = { 0x8000 /* A * E. */,
37 0xC000 /* FP16_C (-2.0f). */,
38 0x0000 /* C * E. */,
39 0x4000 /* FP16_C (2.0f). */ };
40
41 VECT_VAR_DECL (expected1_static, hfloat, 16, 4) []
42 = { 0x5BFF /* A * F. */,
43 0x7C00 /* B * F. */,
44 0xE131 /* C * F. */,
45 0xFC00 /* D * F. */ };
46
47 VECT_VAR_DECL (expected2_static, hfloat, 16, 4) []
48 = { 0xD405 /* A * G. */,
49 0xFC00 /* B * G. */,
50 0x5939 /* C * G. */,
51 0x7C00 /* D * G. */ };
52
53 VECT_VAR_DECL (expected3_static, hfloat, 16, 4) []
54 = { 0x0000 /* A * H. */,
55 0x4000 /* FP16_C (2.0f). */,
56 0x8000 /* C * H. */,
57 0xC000 /* FP16_C (-2.0f). */ };
58
59 VECT_VAR_DECL (expected0_static, hfloat, 16, 8) []
60 = { 0x8000 /* A * E. */,
61 0xC000 /* FP16_C (-2.0f). */,
62 0x0000 /* C * E. */,
63 0x4000 /* FP16_C (2.0f). */,
64 0x8000 /* I * E. */,
65 0x0000 /* J * E. */,
66 0x8000 /* K * E. */,
67 0x8000 /* L * E. */ };
68
69 VECT_VAR_DECL (expected1_static, hfloat, 16, 8) []
70 = { 0x5BFF /* A * F. */,
71 0x7C00 /* B * F. */,
72 0xE131 /* C * F. */,
73 0xFC00 /* D * F. */,
74 0x4AAF /* I * F. */,
75 0xE5D1 /* J * F. */,
76 0x5AB3 /* K * F. */,
77 0x674F /* L * F. */ };
78
79 VECT_VAR_DECL (expected2_static, hfloat, 16, 8) []
80 = { 0xD405 /* A * G. */,
81 0xFC00 /* B * G. */,
82 0x5939 /* C * G. */,
83 0x7C00 /* D * G. */,
84 0xC2B9 /* I * G. */,
85 0x5DDA /* J * G. */,
86 0xD2BD /* K * G. */,
87 0xDF5A /* L * G. */ };
88
89 VECT_VAR_DECL (expected3_static, hfloat, 16, 8) []
90 = { 0x0000 /* A * H. */,
91 0x4000 /* FP16_C (2.0f). */,
92 0x8000 /* C * H. */,
93 0xC000 /* FP16_C (-2.0f). */,
94 0x0000 /* I * H. */,
95 0x8000 /* J * H. */,
96 0x0000 /* K * H. */,
97 0x0000 /* L * H. */ };
98
99 void exec_vmulx_n_f16 (void)
100 {
101 #undef TEST_MSG
102 #define TEST_MSG "VMULX_N (FP16)"
103 clean_results ();
104
105 DECL_VARIABLE (vsrc_1, float, 16, 4);
106 VECT_VAR_DECL (buf_src_1, float, 16, 4) [] = {A, B, C, D};
107 VLOAD (vsrc_1, buf_src_1, , float, f, 16, 4);
108 DECL_VARIABLE (vector_res, float, 16, 4)
109 = vmulx_n_f16 (VECT_VAR (vsrc_1, float, 16, 4), elemE);
110 vst1_f16 (VECT_VAR (result, float, 16, 4),
111 VECT_VAR (vector_res, float, 16, 4));
112
113 CHECK_FP (TEST_MSG, float, 16, 4, PRIx16, expected0_static, "");
114
115 VECT_VAR (vector_res, float, 16, 4)
116 = vmulx_n_f16 (VECT_VAR (vsrc_1, float, 16, 4), elemF);
117 vst1_f16 (VECT_VAR (result, float, 16, 4),
118 VECT_VAR (vector_res, float, 16, 4));
119
120 CHECK_FP (TEST_MSG, float, 16, 4, PRIx16, expected1_static, "");
121
122 VECT_VAR (vector_res, float, 16, 4)
123 = vmulx_n_f16 (VECT_VAR (vsrc_1, float, 16, 4), elemG);
124 vst1_f16 (VECT_VAR (result, float, 16, 4),
125 VECT_VAR (vector_res, float, 16, 4));
126
127 CHECK_FP (TEST_MSG, float, 16, 4, PRIx16, expected2_static, "");
128
129 VECT_VAR (vector_res, float, 16, 4)
130 = vmulx_n_f16 (VECT_VAR (vsrc_1, float, 16, 4), elemH);
131 vst1_f16 (VECT_VAR (result, float, 16, 4),
132 VECT_VAR (vector_res, float, 16, 4));
133
134 CHECK_FP (TEST_MSG, float, 16, 4, PRIx16, expected3_static, "");
135
136 #undef TEST_MSG
137 #define TEST_MSG "VMULXQ_N (FP16)"
138 clean_results ();
139
140 DECL_VARIABLE (vsrc_1, float, 16, 8);
141 VECT_VAR_DECL (buf_src_1, float, 16, 8) [] = {A, B, C, D, I, J, K, L};
142 VLOAD (vsrc_1, buf_src_1, q, float, f, 16, 8);
143 DECL_VARIABLE (vector_res, float, 16, 8)
144 = vmulxq_n_f16 (VECT_VAR (vsrc_1, float, 16, 8), elemE);
145 vst1q_f16 (VECT_VAR (result, float, 16, 8),
146 VECT_VAR (vector_res, float, 16, 8));
147
148 CHECK_FP (TEST_MSG, float, 16, 8, PRIx16, expected0_static, "");
149
150 VECT_VAR (vector_res, float, 16, 8)
151 = vmulxq_n_f16 (VECT_VAR (vsrc_1, float, 16, 8), elemF);
152 vst1q_f16 (VECT_VAR (result, float, 16, 8),
153 VECT_VAR (vector_res, float, 16, 8));
154
155 CHECK_FP (TEST_MSG, float, 16, 8, PRIx16, expected1_static, "");
156
157 VECT_VAR (vector_res, float, 16, 8)
158 = vmulxq_n_f16 (VECT_VAR (vsrc_1, float, 16, 8), elemG);
159 vst1q_f16 (VECT_VAR (result, float, 16, 8),
160 VECT_VAR (vector_res, float, 16, 8));
161
162 CHECK_FP (TEST_MSG, float, 16, 8, PRIx16, expected2_static, "");
163
164 VECT_VAR (vector_res, float, 16, 8)
165 = vmulxq_n_f16 (VECT_VAR (vsrc_1, float, 16, 8), elemH);
166 vst1q_f16 (VECT_VAR (result, float, 16, 8),
167 VECT_VAR (vector_res, float, 16, 8));
168
169 CHECK_FP (TEST_MSG, float, 16, 8, PRIx16, expected3_static, "");
170 }
171
172 int
173 main (void)
174 {
175 exec_vmulx_n_f16 ();
176 return 0;
177 }