1 /**************************************************************************
2 *
3 * Copyright © 2010 Luca Barbieri
4 *
5 * Permission is hereby granted, free of charge, to any person obtaining a
6 * copy of this software and associated documentation files (the "Software"),
7 * to deal in the Software without restriction, including without limitation
8 * the rights to use, copy, modify, merge, publish, distribute, sublicense,
9 * and/or sell copies of the Software, and to permit persons to whom the
10 * Software is furnished to do so, subject to the following conditions:
11 *
12 * The above copyright notice and this permission notice (including the next
13 * paragraph) shall be included in all copies or substantial portions of the
14 * 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
19 * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
20 * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING
21 * FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER
22 * DEALINGS IN THE SOFTWARE.
23 *
24 **************************************************************************/
25
26 #include <stdio.h>
27 #include "translate/translate.h"
28 #include "util/u_memory.h"
29 #include "util/u_format.h"
30 #include "util/u_half.h"
31 #include "util/u_cpu_detect.h"
32 #include "rtasm/rtasm_cpu.h"
33
34 /* don't use this for serious use */
rand_double()35 static double rand_double()
36 {
37 const double rm = (double)RAND_MAX + 1;
38 double div = 1;
39 double v = 0;
40 unsigned i;
41 for(i = 0; i < 4; ++i)
42 {
43 div *= rm;
44 v += (double)rand() / div;
45 }
46 return v;
47 }
48
main(int argc,char ** argv)49 int main(int argc, char** argv)
50 {
51 struct translate *(*create_fn)(const struct translate_key *key) = 0;
52
53 struct translate_key key;
54 unsigned output_format;
55 unsigned input_format;
56 unsigned buffer_size = 4096;
57 unsigned char* buffer[5];
58 unsigned char* byte_buffer;
59 float* float_buffer;
60 double* double_buffer;
61 uint16_t *half_buffer;
62 unsigned * elts;
63 unsigned count = 4;
64 unsigned i, j, k;
65 unsigned passed = 0;
66 unsigned total = 0;
67 const float error = 0.03125;
68
69 create_fn = 0;
70
71 util_cpu_detect();
72
73 if(argc <= 1)
74 {}
75 else if (!strcmp(argv[1], "generic"))
76 create_fn = translate_generic_create;
77 else if (!strcmp(argv[1], "x86"))
78 create_fn = translate_sse2_create;
79 else if (!strcmp(argv[1], "nosse"))
80 {
81 util_cpu_caps.has_sse = 0;
82 util_cpu_caps.has_sse2 = 0;
83 util_cpu_caps.has_sse3 = 0;
84 util_cpu_caps.has_sse4_1 = 0;
85 create_fn = translate_sse2_create;
86 }
87 else if (!strcmp(argv[1], "sse"))
88 {
89 if(!util_cpu_caps.has_sse || !rtasm_cpu_has_sse())
90 {
91 printf("Error: CPU doesn't support SSE (test with qemu)\n");
92 return 2;
93 }
94 util_cpu_caps.has_sse2 = 0;
95 util_cpu_caps.has_sse3 = 0;
96 util_cpu_caps.has_sse4_1 = 0;
97 create_fn = translate_sse2_create;
98 }
99 else if (!strcmp(argv[1], "sse2"))
100 {
101 if(!util_cpu_caps.has_sse2 || !rtasm_cpu_has_sse())
102 {
103 printf("Error: CPU doesn't support SSE2 (test with qemu)\n");
104 return 2;
105 }
106 util_cpu_caps.has_sse3 = 0;
107 util_cpu_caps.has_sse4_1 = 0;
108 create_fn = translate_sse2_create;
109 }
110 else if (!strcmp(argv[1], "sse3"))
111 {
112 if(!util_cpu_caps.has_sse3 || !rtasm_cpu_has_sse())
113 {
114 printf("Error: CPU doesn't support SSE3 (test with qemu)\n");
115 return 2;
116 }
117 util_cpu_caps.has_sse4_1 = 0;
118 create_fn = translate_sse2_create;
119 }
120 else if (!strcmp(argv[1], "sse4.1"))
121 {
122 if(!util_cpu_caps.has_sse4_1 || !rtasm_cpu_has_sse())
123 {
124 printf("Error: CPU doesn't support SSE4.1 (test with qemu)\n");
125 return 2;
126 }
127 create_fn = translate_sse2_create;
128 }
129
130 if (!create_fn)
131 {
132 printf("Usage: ./translate_test [generic|x86|nosse|sse|sse2|sse3|sse4.1]\n");
133 return 2;
134 }
135
136 for (i = 1; i < Elements(buffer); ++i)
137 buffer[i] = align_malloc(buffer_size, 4096);
138
139 byte_buffer = align_malloc(buffer_size, 4096);
140 float_buffer = align_malloc(buffer_size, 4096);
141 double_buffer = align_malloc(buffer_size, 4096);
142 half_buffer = align_malloc(buffer_size, 4096);
143
144 elts = align_malloc(count * sizeof *elts, 4096);
145
146 key.nr_elements = 1;
147 key.element[0].input_buffer = 0;
148 key.element[0].input_offset = 0;
149 key.element[0].output_offset = 0;
150 key.element[0].type = TRANSLATE_ELEMENT_NORMAL;
151 key.element[0].instance_divisor = 0;
152
153 srand(4359025);
154
155 /* avoid negative values that work badly when converted to unsigned format*/
156 for (i = 0; i < buffer_size; ++i)
157 byte_buffer[i] = rand() & 0x7f7f7f7f;
158
159 for (i = 0; i < buffer_size / sizeof(float); ++i)
160 float_buffer[i] = (float)rand_double();
161
162 for (i = 0; i < buffer_size / sizeof(double); ++i)
163 double_buffer[i] = rand_double();
164
165 for (i = 0; i < buffer_size / sizeof(double); ++i)
166 half_buffer[i] = util_float_to_half((float) rand_double());
167
168 for (i = 0; i < count; ++i)
169 elts[i] = i;
170
171 for (output_format = 1; output_format < PIPE_FORMAT_COUNT; ++output_format)
172 {
173 const struct util_format_description* output_format_desc = util_format_description(output_format);
174 unsigned output_format_size;
175 unsigned output_normalized = 0;
176
177 if (!output_format_desc
178 || !output_format_desc->fetch_rgba_float
179 || !output_format_desc->pack_rgba_float
180 || output_format_desc->colorspace != UTIL_FORMAT_COLORSPACE_RGB
181 || output_format_desc->layout != UTIL_FORMAT_LAYOUT_PLAIN
182 || !translate_is_output_format_supported(output_format))
183 continue;
184
185 for(i = 0; i < output_format_desc->nr_channels; ++i)
186 {
187 if(output_format_desc->channel[i].type != UTIL_FORMAT_TYPE_FLOAT)
188 output_normalized |= (1 << output_format_desc->channel[i].normalized);
189 }
190
191 output_format_size = util_format_get_stride(output_format, 1);
192
193 for (input_format = 1; input_format < PIPE_FORMAT_COUNT; ++input_format)
194 {
195 const struct util_format_description* input_format_desc = util_format_description(input_format);
196 unsigned input_format_size;
197 struct translate* translate[2];
198 unsigned fail = 0;
199 unsigned used_generic = 0;
200 unsigned input_normalized = 0;
201 boolean input_is_float = FALSE;
202
203 if (!input_format_desc
204 || !input_format_desc->fetch_rgba_float
205 || !input_format_desc->pack_rgba_float
206 || input_format_desc->colorspace != UTIL_FORMAT_COLORSPACE_RGB
207 || input_format_desc->layout != UTIL_FORMAT_LAYOUT_PLAIN
208 || !translate_is_output_format_supported(input_format))
209 continue;
210
211 input_format_size = util_format_get_stride(input_format, 1);
212
213 for(i = 0; i < input_format_desc->nr_channels; ++i)
214 {
215 if(input_format_desc->channel[i].type == UTIL_FORMAT_TYPE_FLOAT)
216 {
217 input_is_float = 1;
218 input_normalized |= 1 << 1;
219 }
220 else
221 input_normalized |= (1 << input_format_desc->channel[i].normalized);
222 }
223
224 if(((input_normalized | output_normalized) == 3)
225 || ((input_normalized & 1) && (output_normalized & 1)
226 && input_format_size * output_format_desc->nr_channels > output_format_size * input_format_desc->nr_channels))
227 continue;
228
229 key.element[0].input_format = input_format;
230 key.element[0].output_format = output_format;
231 key.output_stride = output_format_size;
232 translate[0] = create_fn(&key);
233 if (!translate[0])
234 continue;
235
236 key.element[0].input_format = output_format;
237 key.element[0].output_format = input_format;
238 key.output_stride = input_format_size;
239 translate[1] = create_fn(&key);
240 if(!translate[1])
241 {
242 used_generic = 1;
243 translate[1] = translate_generic_create(&key);
244 if(!translate[1])
245 continue;
246 }
247
248 for(i = 1; i < 5; ++i)
249 memset(buffer[i], 0xcd - (0x22 * i), 4096);
250
251 if(input_is_float && input_format_desc->channel[0].size == 32)
252 buffer[0] = (unsigned char*)float_buffer;
253 else if(input_is_float && input_format_desc->channel[0].size == 64)
254 buffer[0] = (unsigned char*)double_buffer;
255 else if(input_is_float && input_format_desc->channel[0].size == 16)
256 buffer[0] = (unsigned char*)half_buffer;
257 else if(input_is_float)
258 abort();
259 else
260 buffer[0] = byte_buffer;
261
262 translate[0]->set_buffer(translate[0], 0, buffer[0], input_format_size, count - 1);
263 translate[0]->run_elts(translate[0], elts, count, 0, buffer[1]);
264 translate[1]->set_buffer(translate[1], 0, buffer[1], output_format_size, count - 1);
265 translate[1]->run_elts(translate[1], elts, count, 0, buffer[2]);
266 translate[0]->set_buffer(translate[0], 0, buffer[2], input_format_size, count - 1);
267 translate[0]->run_elts(translate[0], elts, count, 0, buffer[3]);
268 translate[1]->set_buffer(translate[1], 0, buffer[3], output_format_size, count - 1);
269 translate[1]->run_elts(translate[1], elts, count, 0, buffer[4]);
270
271 for (i = 0; i < count; ++i)
272 {
273 float a[4];
274 float b[4];
275 input_format_desc->fetch_rgba_float(a, buffer[2] + i * input_format_size, 0, 0);
276 input_format_desc->fetch_rgba_float(b, buffer[4] + i * input_format_size, 0, 0);
277
278 for (j = 0; j < count; ++j)
279 {
280 float d = a[j] - b[j];
281 if (d > error || d < -error)
282 {
283 fail = 1;
284 break;
285 }
286 }
287 }
288
289 printf("%s%s: %s -> %s -> %s -> %s -> %s\n",
290 fail ? "FAIL" : "PASS",
291 used_generic ? "[GENERIC]" : "",
292 input_format_desc->name, output_format_desc->name, input_format_desc->name, output_format_desc->name, input_format_desc->name);
293
294 if (1)
295 {
296 for (i = 0; i < Elements(buffer); ++i)
297 {
298 unsigned format_size = (i & 1) ? output_format_size : input_format_size;
299 printf("%c ", (i == 2 || i == 4) ? '*' : ' ');
300 for (j = 0; j < count; ++j)
301 {
302 for (k = 0; k < format_size; ++k)
303 {
304 printf("%02x", buffer[i][j * format_size + k]);
305 }
306 printf(" ");
307 }
308 printf("\n");
309 }
310 }
311
312 if (!fail)
313 ++passed;
314 ++total;
315
316 if(translate[1])
317 translate[1]->release(translate[1]);
318 translate[0]->release(translate[0]);
319 }
320 }
321
322 printf("%u/%u tests passed for translate_%s\n", passed, total, argv[1]);
323 return passed != total;
324 }
325