1 /*
2
3 Copyright (c) 2009, 2010, 2011, 2013 STMicroelectronics
4 Written by Christophe Lyon
5
6 Permission is hereby granted, free of charge, to any person obtaining a copy
7 of this software and associated documentation files (the "Software"), to deal
8 in the Software without restriction, including without limitation the rights
9 to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
10 copies of the Software, and to permit persons to whom the Software is
11 furnished to do so, subject to the following conditions:
12
13 The above copyright notice and this permission notice shall be included in
14 all copies or substantial portions of the Software.
15
16 THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
17 IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
18 FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
19 AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
20 LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
21 OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN
22 THE SOFTWARE.
23
24 */
25
26 #if defined(__arm__) || defined(__aarch64__)
27 #include <arm_neon.h>
28 #else
29 #include "stm-arm-neon.h"
30 #endif
31
32 #include "stm-arm-neon-ref.h"
33
exec_vstX_lane(void)34 void exec_vstX_lane (void)
35 {
36 /* In this case, input variables are arrays of vectors */
37 #define DECL_VSTX_LANE(T1, W, N, X) \
38 VECT_ARRAY_TYPE(T1, W, N, X) VECT_ARRAY_VAR(vector, T1, W, N, X); \
39 VECT_ARRAY_TYPE(T1, W, N, X) VECT_ARRAY_VAR(vector_src, T1, W, N, X); \
40 VECT_VAR_DECL(result_bis_##X, T1, W, N)[X * N]
41
42 /* We need to use a temporary result buffer (result_bis), because
43 the one used for other tests is not large enough. A subset of the
44 result data is moved from result_bis to result, and it is this
45 subset which is used to check the actual behaviour. The next
46 macro enables to move another chunk of data from result_bis to
47 result. */
48 #define TEST_VSTX_LANE(Q, T1, T2, W, N, X, L) \
49 memset (VECT_VAR(buffer_src, T1, W, N), 0xAA, \
50 sizeof(VECT_VAR(buffer_src, T1, W, N))); \
51 memset (VECT_VAR(result_bis_##X, T1, W, N), 0, \
52 sizeof(VECT_VAR(result_bis_##X, T1, W, N))); \
53 \
54 VECT_ARRAY_VAR(vector_src, T1, W, N, X) = \
55 vld##X##Q##_##T2##W(VECT_VAR(buffer_src, T1, W, N)); \
56 \
57 VECT_ARRAY_VAR(vector, T1, W, N, X) = \
58 /* Use dedicated init buffer, of size X */ \
59 vld##X##Q##_lane_##T2##W(VECT_VAR(buffer_vld##X##_lane, T1, W, X), \
60 VECT_ARRAY_VAR(vector_src, T1, W, N, X), \
61 L); \
62 vst##X##Q##_lane_##T2##W(VECT_VAR(result_bis_##X, T1, W, N), \
63 VECT_ARRAY_VAR(vector, T1, W, N, X), \
64 L); \
65 memcpy(VECT_VAR(result, T1, W, N), VECT_VAR(result_bis_##X, T1, W, N), \
66 sizeof(VECT_VAR(result, T1, W, N)));
67
68 /* Overwrite "result" with the contents of "result_bis"[Y] */
69 #define TEST_EXTRA_CHUNK(T1, W, N, X, Y) \
70 memcpy(VECT_VAR(result, T1, W, N), \
71 &(VECT_VAR(result_bis_##X, T1, W, N)[Y*N]), \
72 sizeof(VECT_VAR(result, T1, W, N)));
73
74 /* With ARM RVCT, we need to declare variables before any executable
75 statement */
76
77 /* We need all variants in 64 bits, but there is no 64x2 variant */
78 #define DECL_ALL_VSTX_LANE(X) \
79 DECL_VSTX_LANE(int, 8, 8, X); \
80 DECL_VSTX_LANE(int, 16, 4, X); \
81 DECL_VSTX_LANE(int, 32, 2, X); \
82 DECL_VSTX_LANE(uint, 8, 8, X); \
83 DECL_VSTX_LANE(uint, 16, 4, X); \
84 DECL_VSTX_LANE(uint, 32, 2, X); \
85 DECL_VSTX_LANE(poly, 8, 8, X); \
86 DECL_VSTX_LANE(poly, 16, 4, X); \
87 DECL_VSTX_LANE(float, 32, 2, X); \
88 DECL_VSTX_LANE(int, 16, 8, X); \
89 DECL_VSTX_LANE(int, 32, 4, X); \
90 DECL_VSTX_LANE(uint, 16, 8, X); \
91 DECL_VSTX_LANE(uint, 32, 4, X); \
92 DECL_VSTX_LANE(poly, 16, 8, X); \
93 DECL_VSTX_LANE(float, 32, 4, X)
94
95 #if defined(__ARM_FP16_FORMAT_IEEE) && ( ((__ARM_FP & 0x2) != 0) || ((__ARM_NEON_FP16_INTRINSICS & 1) != 0) )
96 #define DECL_ALL_VSTX_LANE_FP16(X) \
97 DECL_VSTX_LANE(float, 16, 4, X); \
98 DECL_VSTX_LANE(float, 16, 8, X)
99 #endif
100
101 #define DUMMY_ARRAY(V, T, W, N, L) VECT_VAR_DECL(V,T,W,N)[N*L]
102
103 /* Use the same lanes regardless of the size of the array (X), for
104 simplicity */
105 #define TEST_ALL_VSTX_LANE(X) \
106 TEST_VSTX_LANE(, int, s, 8, 8, X, 7); \
107 TEST_VSTX_LANE(, int, s, 16, 4, X, 2); \
108 TEST_VSTX_LANE(, int, s, 32, 2, X, 0); \
109 TEST_VSTX_LANE(, float, f, 32, 2, X, 0); \
110 TEST_VSTX_LANE(, uint, u, 8, 8, X, 4); \
111 TEST_VSTX_LANE(, uint, u, 16, 4, X, 3); \
112 TEST_VSTX_LANE(, uint, u, 32, 2, X, 1); \
113 TEST_VSTX_LANE(, poly, p, 8, 8, X, 4); \
114 TEST_VSTX_LANE(, poly, p, 16, 4, X, 3); \
115 TEST_VSTX_LANE(q, int, s, 16, 8, X, 6); \
116 TEST_VSTX_LANE(q, int, s, 32, 4, X, 2); \
117 TEST_VSTX_LANE(q, uint, u, 16, 8, X, 5); \
118 TEST_VSTX_LANE(q, uint, u, 32, 4, X, 0); \
119 TEST_VSTX_LANE(q, poly, p, 16, 8, X, 5); \
120 TEST_VSTX_LANE(q, float, f, 32, 4, X, 2)
121
122 #if defined(__ARM_FP16_FORMAT_IEEE) && ( ((__ARM_FP & 0x2) != 0) || ((__ARM_NEON_FP16_INTRINSICS & 1) != 0) )
123 #define TEST_ALL_VSTX_LANE_FP16(X) \
124 TEST_VSTX_LANE(, float, f, 16, 4, X, 3); \
125 TEST_VSTX_LANE(q, float, f, 16, 8, X, 6)
126 #endif
127
128 #define TEST_ALL_EXTRA_CHUNKS(X, Y) \
129 TEST_EXTRA_CHUNK(int, 8, 8, X, Y); \
130 TEST_EXTRA_CHUNK(int, 16, 4, X, Y); \
131 TEST_EXTRA_CHUNK(int, 32, 2, X, Y); \
132 TEST_EXTRA_CHUNK(uint, 8, 8, X, Y); \
133 TEST_EXTRA_CHUNK(uint, 16, 4, X, Y); \
134 TEST_EXTRA_CHUNK(uint, 32, 2, X, Y); \
135 TEST_EXTRA_CHUNK(poly, 8, 8, X, Y); \
136 TEST_EXTRA_CHUNK(poly, 16, 4, X, Y); \
137 TEST_EXTRA_CHUNK(float, 32, 2, X, Y); \
138 TEST_EXTRA_CHUNK(int, 16, 8, X, Y); \
139 TEST_EXTRA_CHUNK(int, 32, 4, X, Y); \
140 TEST_EXTRA_CHUNK(uint, 16, 8, X, Y); \
141 TEST_EXTRA_CHUNK(uint, 32, 4, X, Y); \
142 TEST_EXTRA_CHUNK(poly, 16, 8, X, Y); \
143 TEST_EXTRA_CHUNK(float, 32, 4, X, Y)
144
145 #if defined(__ARM_FP16_FORMAT_IEEE) && ( ((__ARM_FP & 0x2) != 0) || ((__ARM_NEON_FP16_INTRINSICS & 1) != 0) )
146 #define TEST_ALL_EXTRA_CHUNKS_FP16(X, Y) \
147 TEST_EXTRA_CHUNK(float, 16, 4, X, Y); \
148 TEST_EXTRA_CHUNK(float, 16, 8, X, Y)
149 #endif
150
151 /* Declare the temporary buffers / variables */
152 DECL_ALL_VSTX_LANE(2);
153 DECL_ALL_VSTX_LANE(3);
154 DECL_ALL_VSTX_LANE(4);
155 #if defined(__ARM_FP16_FORMAT_IEEE) && ( ((__ARM_FP & 0x2) != 0) || ((__ARM_NEON_FP16_INTRINSICS & 1) != 0) )
156 DECL_ALL_VSTX_LANE_FP16(2);
157 DECL_ALL_VSTX_LANE_FP16(3);
158 DECL_ALL_VSTX_LANE_FP16(4);
159 #endif
160
161 /* Define dummy input arrays, large enough for x4 vectors */
162 DUMMY_ARRAY(buffer_src, int, 8, 8, 4);
163 DUMMY_ARRAY(buffer_src, int, 16, 4, 4);
164 DUMMY_ARRAY(buffer_src, int, 32, 2, 4);
165 DUMMY_ARRAY(buffer_src, uint, 8, 8, 4);
166 DUMMY_ARRAY(buffer_src, uint, 16, 4, 4);
167 DUMMY_ARRAY(buffer_src, uint, 32, 2, 4);
168 DUMMY_ARRAY(buffer_src, poly, 8, 8, 4);
169 DUMMY_ARRAY(buffer_src, poly, 16, 4, 4);
170 DUMMY_ARRAY(buffer_src, float, 32, 2, 4);
171 DUMMY_ARRAY(buffer_src, int, 16, 8, 4);
172 DUMMY_ARRAY(buffer_src, int, 32, 4, 4);
173 DUMMY_ARRAY(buffer_src, uint, 16, 8, 4);
174 DUMMY_ARRAY(buffer_src, uint, 32, 4, 4);
175 DUMMY_ARRAY(buffer_src, poly, 16, 8, 4);
176 DUMMY_ARRAY(buffer_src, float, 32, 4, 4);
177 #if defined(__ARM_FP16_FORMAT_IEEE) && ( ((__ARM_FP & 0x2) != 0) || ((__ARM_NEON_FP16_INTRINSICS & 1) != 0) )
178 DUMMY_ARRAY(buffer_src, float, 16, 4, 4);
179 DUMMY_ARRAY(buffer_src, float, 16, 8, 4);
180 #endif
181
182 /* Check vst2_lane/vst2q_lane */
183 clean_results ();
184 #define TEST_MSG "VST2_LANE/VST2Q_LANE"
185 TEST_ALL_VSTX_LANE(2);
186 #if defined(__ARM_FP16_FORMAT_IEEE) && ( ((__ARM_FP & 0x2) != 0) || ((__ARM_NEON_FP16_INTRINSICS & 1) != 0) )
187 TEST_ALL_VSTX_LANE_FP16(2);
188 #endif
189 dump_results_hex2 (TEST_MSG, " chunk 0");
190
191 TEST_ALL_EXTRA_CHUNKS(2, 1);
192 #if defined(__ARM_FP16_FORMAT_IEEE) && ( ((__ARM_FP & 0x2) != 0) || ((__ARM_NEON_FP16_INTRINSICS & 1) != 0) )
193 TEST_ALL_EXTRA_CHUNKS_FP16(2, 1);
194 #endif
195 dump_results_hex2 (TEST_MSG, " chunk 1");
196
197 /* Check vst3_lane/vst3q_lane */
198 clean_results ();
199 #undef TEST_MSG
200 #define TEST_MSG "VST3_LANE/VST3Q_LANE"
201 TEST_ALL_VSTX_LANE(3);
202 #if defined(__ARM_FP16_FORMAT_IEEE) && ( ((__ARM_FP & 0x2) != 0) || ((__ARM_NEON_FP16_INTRINSICS & 1) != 0) )
203 TEST_ALL_VSTX_LANE_FP16(3);
204 #endif
205 dump_results_hex2 (TEST_MSG, " chunk 0");
206
207 TEST_ALL_EXTRA_CHUNKS(3, 1);
208 #if defined(__ARM_FP16_FORMAT_IEEE) && ( ((__ARM_FP & 0x2) != 0) || ((__ARM_NEON_FP16_INTRINSICS & 1) != 0) )
209 TEST_ALL_EXTRA_CHUNKS_FP16(3, 1);
210 #endif
211 dump_results_hex2 (TEST_MSG, " chunk 1");
212 TEST_ALL_EXTRA_CHUNKS(3, 2);
213 #if defined(__ARM_FP16_FORMAT_IEEE) && ( ((__ARM_FP & 0x2) != 0) || ((__ARM_NEON_FP16_INTRINSICS & 1) != 0) )
214 TEST_ALL_EXTRA_CHUNKS_FP16(3, 2);
215 #endif
216 dump_results_hex2 (TEST_MSG, " chunk 2");
217
218 /* Check vst4_lane/vst4q_lane */
219 clean_results ();
220 #undef TEST_MSG
221 #define TEST_MSG "VST4_LANE/VST4Q_LANE"
222 TEST_ALL_VSTX_LANE(4);
223 #if defined(__ARM_FP16_FORMAT_IEEE) && ( ((__ARM_FP & 0x2) != 0) || ((__ARM_NEON_FP16_INTRINSICS & 1) != 0) )
224 TEST_ALL_VSTX_LANE_FP16(4);
225 #endif
226 dump_results_hex2 (TEST_MSG, " chunk 0");
227
228 TEST_ALL_EXTRA_CHUNKS(4, 1);
229 #if defined(__ARM_FP16_FORMAT_IEEE) && ( ((__ARM_FP & 0x2) != 0) || ((__ARM_NEON_FP16_INTRINSICS & 1) != 0) )
230 TEST_ALL_EXTRA_CHUNKS_FP16(4, 1);
231 #endif
232 dump_results_hex2 (TEST_MSG, " chunk 1");
233 TEST_ALL_EXTRA_CHUNKS(4, 2);
234 #if defined(__ARM_FP16_FORMAT_IEEE) && ( ((__ARM_FP & 0x2) != 0) || ((__ARM_NEON_FP16_INTRINSICS & 1) != 0) )
235 TEST_ALL_EXTRA_CHUNKS_FP16(4, 2);
236 #endif
237 dump_results_hex2 (TEST_MSG, " chunk 2");
238 TEST_ALL_EXTRA_CHUNKS(4, 3);
239 #if defined(__ARM_FP16_FORMAT_IEEE) && ( ((__ARM_FP & 0x2) != 0) || ((__ARM_NEON_FP16_INTRINSICS & 1) != 0) )
240 TEST_ALL_EXTRA_CHUNKS_FP16(4, 3);
241 #endif
242 dump_results_hex2 (TEST_MSG, " chunk 3");
243 }
244