Debian Codegraph @

gcc/testsuite/gcc.target/aarch64/advsimd-intrinsics/vfmash_lane_f16_1.c

raw
1 /* { dg-require-effective-target arm_v8_2a_fp16_scalar_hw } */
2 /* { dg-add-options arm_v8_2a_fp16_neon } */
3 /* { dg-skip-if "" { arm*-*-* } } */
4
5 #include <arm_neon.h>
6 #include "arm-neon-ref.h"
7 #include "compute-ref-data.h"
8
9 #define FP16_C(a) ((__fp16) a)
10 #define A0 FP16_C (123.4)
11 #define B0 FP16_C (-5.8)
12 #define C0 FP16_C (-3.8)
13 #define D0 FP16_C (10)
14
15 #define A1 FP16_C (12.4)
16 #define B1 FP16_C (-5.8)
17 #define C1 FP16_C (90.8)
18 #define D1 FP16_C (24)
19
20 #define A2 FP16_C (23.4)
21 #define B2 FP16_C (-5.8)
22 #define C2 FP16_C (8.9)
23 #define D2 FP16_C (4)
24
25 #define E0 FP16_C (3.4)
26 #define F0 FP16_C (-55.8)
27 #define G0 FP16_C (-31.8)
28 #define H0 FP16_C (2)
29
30 #define E1 FP16_C (123.4)
31 #define F1 FP16_C (-5.8)
32 #define G1 FP16_C (-3.8)
33 #define H1 FP16_C (102)
34
35 #define E2 FP16_C (4.9)
36 #define F2 FP16_C (-15.8)
37 #define G2 FP16_C (39.8)
38 #define H2 FP16_C (49)
39
40 extern void abort ();
41
42 float16_t src1[8] = { A0, B0, C0, D0, E0, F0, G0, H0 };
43 float16_t src2[8] = { A1, B1, C1, D1, E1, F1, G1, H1 };
44 VECT_VAR_DECL (src3, float, 16, 4) [] = { A2, B2, C2, D2 };
45 VECT_VAR_DECL (src3, float, 16, 8) [] = { A2, B2, C2, D2, E2, F2, G2, H2 };
46
47 /* Expected results for vfmah_lane_f16. */
48 uint16_t expected[4] = { 0x5E76 /* A0 + A1 * A2. */,
49 0x4EF6 /* B0 + B1 * B2. */,
50 0x6249 /* C0 + C1 * C2. */,
51 0x56A0 /* D0 + D1 * D2. */ };
52
53 /* Expected results for vfmah_laneq_f16. */
54 uint16_t expected_laneq[8] = { 0x5E76 /* A0 + A1 * A2. */,
55 0x4EF6 /* B0 + B1 * B2. */,
56 0x6249 /* C0 + C1 * C2. */,
57 0x56A0 /* D0 + D1 * D2. */,
58 0x60BF /* E0 + E1 * E2. */,
59 0x507A /* F0 + F1 * F2. */,
60 0xD9B9 /* G0 + G1 * G2. */,
61 0x6CE2 /* H0 + H1 * H2. */ };
62
63 /* Expected results for vfmsh_lane_f16. */
64 uint16_t expected_fms[4] = { 0xD937 /* A0 + -A1 * A2. */,
65 0xD0EE /* B0 + -B1 * B2. */,
66 0xE258 /* C0 + -C1 * C2. */,
67 0xD560 /* D0 + -D1 * D2. */ };
68
69 /* Expected results for vfmsh_laneq_f16. */
70 uint16_t expected_fms_laneq[8] = { 0xD937 /* A0 + -A1 * A2. */,
71 0xD0EE /* B0 + -B1 * B2. */,
72 0xE258 /* C0 + -C1 * C2. */,
73 0xD560 /* D0 + -D1 * D2. */,
74 0xE0B2 /* E0 + -E1 * E2. */,
75 0xD89C /* F0 + -F1 * F2. */,
76 0x5778 /* G0 + -G1 * G2. */,
77 0xECE1 /* H0 + -H1 * H2. */ };
78
79 void exec_vfmash_lane_f16 (void)
80 {
81 #define CHECK_LANE(N) \
82 ret = vfmah_lane_f16 (src1[N], src2[N], VECT_VAR (vsrc3, float, 16, 4), N);\
83 if (*(uint16_t *) &ret != expected[N])\
84 abort ();
85
86 DECL_VARIABLE(vsrc3, float, 16, 4);
87 VLOAD (vsrc3, src3, , float, f, 16, 4);
88 float16_t ret;
89 CHECK_LANE(0)
90 CHECK_LANE(1)
91 CHECK_LANE(2)
92 CHECK_LANE(3)
93
94 #undef CHECK_LANE
95 #define CHECK_LANE(N) \
96 ret = vfmah_laneq_f16 (src1[N], src2[N], VECT_VAR (vsrc3, float, 16, 8), N);\
97 if (*(uint16_t *) &ret != expected_laneq[N]) \
98 abort ();
99
100 DECL_VARIABLE(vsrc3, float, 16, 8);
101 VLOAD (vsrc3, src3, q, float, f, 16, 8);
102 CHECK_LANE(0)
103 CHECK_LANE(1)
104 CHECK_LANE(2)
105 CHECK_LANE(3)
106 CHECK_LANE(4)
107 CHECK_LANE(5)
108 CHECK_LANE(6)
109 CHECK_LANE(7)
110
111 #undef CHECK_LANE
112 #define CHECK_LANE(N) \
113 ret = vfmsh_lane_f16 (src1[N], src2[N], VECT_VAR (vsrc3, float, 16, 4), N);\
114 if (*(uint16_t *) &ret != expected_fms[N])\
115 abort ();
116
117 CHECK_LANE(0)
118 CHECK_LANE(1)
119 CHECK_LANE(2)
120
121 #undef CHECK_LANE
122 #define CHECK_LANE(N) \
123 ret = vfmsh_laneq_f16 (src1[N], src2[N], VECT_VAR (vsrc3, float, 16, 8), N);\
124 if (*(uint16_t *) &ret != expected_fms_laneq[N]) \
125 abort ();
126
127 CHECK_LANE(0)
128 CHECK_LANE(1)
129 CHECK_LANE(2)
130 CHECK_LANE(3)
131 CHECK_LANE(4)
132 CHECK_LANE(5)
133 CHECK_LANE(6)
134 CHECK_LANE(7)
135 }
136
137 int
138 main (void)
139 {
140 exec_vfmash_lane_f16 ();
141 return 0;
142 }