OLD | NEW |
| (Empty) |
1 /* | |
2 * Copyright (c) 2013 The WebM project authors. All Rights Reserved. | |
3 * | |
4 * Use of this source code is governed by a BSD-style license | |
5 * that can be found in the LICENSE file in the root of the source | |
6 * tree. An additional intellectual property rights grant can be found | |
7 * in the file PATENTS. All contributing project authors may | |
8 * be found in the AUTHORS file in the root of the source tree. | |
9 */ | |
10 | |
11 #include <assert.h> | |
12 | |
13 #include "./vpx_config.h" | |
14 #include "./vp9_rtcd.h" | |
15 #include "vp9/common/vp9_common.h" | |
16 #include "vp9/common/vp9_blockd.h" | |
17 #include "vp9/common/mips/dspr2/vp9_common_dspr2.h" | |
18 #include "vpx_dsp/txfm_common.h" | |
19 #include "vpx_ports/mem.h" | |
20 | |
21 #if HAVE_DSPR2 | |
22 void vp9_idct32_cols_add_blk_dspr2(int16_t *input, uint8_t *dest, | |
23 int dest_stride) { | |
24 int16_t step1_0, step1_1, step1_2, step1_3, step1_4, step1_5, step1_6; | |
25 int16_t step1_7, step1_8, step1_9, step1_10, step1_11, step1_12, step1_13; | |
26 int16_t step1_14, step1_15, step1_16, step1_17, step1_18, step1_19; | |
27 int16_t step1_20, step1_21, step1_22, step1_23, step1_24, step1_25, step1_26; | |
28 int16_t step1_27, step1_28, step1_29, step1_30, step1_31; | |
29 int16_t step2_0, step2_1, step2_2, step2_3, step2_4, step2_5, step2_6; | |
30 int16_t step2_7, step2_8, step2_9, step2_10, step2_11, step2_12, step2_13; | |
31 int16_t step2_14, step2_15, step2_16, step2_17, step2_18, step2_19, step2_20; | |
32 int16_t step2_21, step2_22, step2_23, step2_24, step2_25, step2_26, step2_27; | |
33 int16_t step2_28, step2_29, step2_30, step2_31; | |
34 int16_t step3_8, step3_9, step3_10, step3_11, step3_12, step3_13, step3_14; | |
35 int16_t step3_15, step3_16, step3_17, step3_18, step3_19, step3_20, step3_21; | |
36 int16_t step3_22, step3_23, step3_24, step3_25, step3_26, step3_27; | |
37 int16_t step3_28, step3_29, step3_30, step3_31; | |
38 int temp0, temp1, temp2, temp3; | |
39 int load1, load2, load3, load4; | |
40 int result1, result2; | |
41 int i, temp21; | |
42 uint8_t *dest_pix, *dest_pix1; | |
43 const int const_2_power_13 = 8192; | |
44 uint8_t *cm = vpx_ff_cropTbl; | |
45 | |
46 /* prefetch vpx_ff_cropTbl */ | |
47 prefetch_load(vpx_ff_cropTbl); | |
48 prefetch_load(vpx_ff_cropTbl + 32); | |
49 prefetch_load(vpx_ff_cropTbl + 64); | |
50 prefetch_load(vpx_ff_cropTbl + 96); | |
51 prefetch_load(vpx_ff_cropTbl + 128); | |
52 prefetch_load(vpx_ff_cropTbl + 160); | |
53 prefetch_load(vpx_ff_cropTbl + 192); | |
54 prefetch_load(vpx_ff_cropTbl + 224); | |
55 | |
56 for (i = 0; i < 32; ++i) { | |
57 dest_pix = dest + i; | |
58 dest_pix1 = dest + i + 31 * dest_stride; | |
59 | |
60 __asm__ __volatile__ ( | |
61 "lh %[load1], 2(%[input]) \n\t" | |
62 "lh %[load2], 62(%[input]) \n\t" | |
63 "lh %[load3], 34(%[input]) \n\t" | |
64 "lh %[load4], 30(%[input]) \n\t" | |
65 | |
66 "mtlo %[const_2_power_13], $ac1 \n\t" | |
67 "mthi $zero, $ac1 \n\t" | |
68 "mtlo %[const_2_power_13], $ac3 \n\t" | |
69 "mthi $zero, $ac3 \n\t" | |
70 | |
71 "madd $ac1, %[load1], %[cospi_31_64] \n\t" | |
72 "msub $ac1, %[load2], %[cospi_1_64] \n\t" | |
73 "extp %[temp0], $ac1, 31 \n\t" | |
74 | |
75 "madd $ac3, %[load1], %[cospi_1_64] \n\t" | |
76 "madd $ac3, %[load2], %[cospi_31_64] \n\t" | |
77 "extp %[temp3], $ac3, 31 \n\t" | |
78 | |
79 "mtlo %[const_2_power_13], $ac1 \n\t" | |
80 "mthi $zero, $ac1 \n\t" | |
81 "mtlo %[const_2_power_13], $ac2 \n\t" | |
82 "mthi $zero, $ac2 \n\t" | |
83 | |
84 "madd $ac2, %[load3], %[cospi_15_64] \n\t" | |
85 "msub $ac2, %[load4], %[cospi_17_64] \n\t" | |
86 "extp %[temp1], $ac2, 31 \n\t" | |
87 | |
88 "madd $ac1, %[load3], %[cospi_17_64] \n\t" | |
89 "madd $ac1, %[load4], %[cospi_15_64] \n\t" | |
90 "extp %[temp2], $ac1, 31 \n\t" | |
91 | |
92 "mtlo %[const_2_power_13], $ac1 \n\t" | |
93 "mthi $zero, $ac1 \n\t" | |
94 "mtlo %[const_2_power_13], $ac3 \n\t" | |
95 "mthi $zero, $ac3 \n\t" | |
96 | |
97 "sub %[load1], %[temp3], %[temp2] \n\t" | |
98 "sub %[load2], %[temp0], %[temp1] \n\t" | |
99 | |
100 "madd $ac1, %[load1], %[cospi_28_64] \n\t" | |
101 "msub $ac1, %[load2], %[cospi_4_64] \n\t" | |
102 "madd $ac3, %[load1], %[cospi_4_64] \n\t" | |
103 "madd $ac3, %[load2], %[cospi_28_64] \n\t" | |
104 | |
105 "extp %[step1_17], $ac1, 31 \n\t" | |
106 "extp %[step1_30], $ac3, 31 \n\t" | |
107 "add %[step1_16], %[temp0], %[temp1] \n\t" | |
108 "add %[step1_31], %[temp2], %[temp3] \n\t" | |
109 | |
110 : [load1] "=&r" (load1), [load2] "=&r" (load2), [load3] "=&r" (load3), | |
111 [load4] "=&r" (load4), [temp0] "=&r" (temp0), [temp1] "=&r" (temp1), | |
112 [temp2] "=&r" (temp2), [temp3] "=&r" (temp3), | |
113 [step1_16] "=r" (step1_16), [step1_17] "=r" (step1_17), | |
114 [step1_30] "=r" (step1_30), [step1_31] "=r" (step1_31) | |
115 : [const_2_power_13] "r" (const_2_power_13), [input] "r" (input), | |
116 [cospi_31_64] "r" (cospi_31_64), [cospi_1_64] "r" (cospi_1_64), | |
117 [cospi_4_64] "r" (cospi_4_64), [cospi_17_64] "r" (cospi_17_64), | |
118 [cospi_15_64] "r" (cospi_15_64), [cospi_28_64] "r" (cospi_28_64) | |
119 ); | |
120 | |
121 __asm__ __volatile__ ( | |
122 "lh %[load1], 18(%[input]) \n\t" | |
123 "lh %[load2], 46(%[input]) \n\t" | |
124 "lh %[load3], 50(%[input]) \n\t" | |
125 "lh %[load4], 14(%[input]) \n\t" | |
126 | |
127 "mtlo %[const_2_power_13], $ac1 \n\t" | |
128 "mthi $zero, $ac1 \n\t" | |
129 "mtlo %[const_2_power_13], $ac3 \n\t" | |
130 "mthi $zero, $ac3 \n\t" | |
131 | |
132 "madd $ac1, %[load1], %[cospi_23_64] \n\t" | |
133 "msub $ac1, %[load2], %[cospi_9_64] \n\t" | |
134 "extp %[temp0], $ac1, 31 \n\t" | |
135 | |
136 "madd $ac3, %[load1], %[cospi_9_64] \n\t" | |
137 "madd $ac3, %[load2], %[cospi_23_64] \n\t" | |
138 "extp %[temp3], $ac3, 31 \n\t" | |
139 | |
140 "mtlo %[const_2_power_13], $ac1 \n\t" | |
141 "mthi $zero, $ac1 \n\t" | |
142 "mtlo %[const_2_power_13], $ac2 \n\t" | |
143 "mthi $zero, $ac2 \n\t" | |
144 | |
145 "madd $ac2, %[load3], %[cospi_7_64] \n\t" | |
146 "msub $ac2, %[load4], %[cospi_25_64] \n\t" | |
147 "extp %[temp1], $ac2, 31 \n\t" | |
148 | |
149 "madd $ac1, %[load3], %[cospi_25_64] \n\t" | |
150 "madd $ac1, %[load4], %[cospi_7_64] \n\t" | |
151 "extp %[temp2], $ac1, 31 \n\t" | |
152 | |
153 "mtlo %[const_2_power_13], $ac1 \n\t" | |
154 "mthi $zero, $ac1 \n\t" | |
155 "mtlo %[const_2_power_13], $ac3 \n\t" | |
156 "mthi $zero, $ac3 \n\t" | |
157 | |
158 "sub %[load1], %[temp1], %[temp0] \n\t" | |
159 "sub %[load2], %[temp2], %[temp3] \n\t" | |
160 | |
161 "msub $ac1, %[load1], %[cospi_28_64] \n\t" | |
162 "msub $ac1, %[load2], %[cospi_4_64] \n\t" | |
163 "msub $ac3, %[load1], %[cospi_4_64] \n\t" | |
164 "madd $ac3, %[load2], %[cospi_28_64] \n\t" | |
165 | |
166 "extp %[step1_18], $ac1, 31 \n\t" | |
167 "extp %[step1_29], $ac3, 31 \n\t" | |
168 "add %[step1_19], %[temp0], %[temp1] \n\t" | |
169 "add %[step1_28], %[temp2], %[temp3] \n\t" | |
170 | |
171 : [load1] "=&r" (load1), [load2] "=&r" (load2), [load3] "=&r" (load3), | |
172 [load4] "=&r" (load4), [temp0] "=&r" (temp0), [temp1] "=&r" (temp1), | |
173 [temp2] "=&r" (temp2), [temp3] "=&r" (temp3), | |
174 [step1_18] "=r" (step1_18), [step1_19] "=r" (step1_19), | |
175 [step1_28] "=r" (step1_28), [step1_29] "=r" (step1_29) | |
176 : [const_2_power_13] "r" (const_2_power_13), [input] "r" (input), | |
177 [cospi_23_64] "r" (cospi_23_64), [cospi_9_64] "r" (cospi_9_64), | |
178 [cospi_4_64] "r" (cospi_4_64), [cospi_7_64] "r" (cospi_7_64), | |
179 [cospi_25_64] "r" (cospi_25_64), [cospi_28_64] "r" (cospi_28_64) | |
180 ); | |
181 | |
182 __asm__ __volatile__ ( | |
183 "lh %[load1], 10(%[input]) \n\t" | |
184 "lh %[load2], 54(%[input]) \n\t" | |
185 "lh %[load3], 42(%[input]) \n\t" | |
186 "lh %[load4], 22(%[input]) \n\t" | |
187 | |
188 "mtlo %[const_2_power_13], $ac1 \n\t" | |
189 "mthi $zero, $ac1 \n\t" | |
190 "mtlo %[const_2_power_13], $ac3 \n\t" | |
191 "mthi $zero, $ac3 \n\t" | |
192 | |
193 "madd $ac1, %[load1], %[cospi_27_64] \n\t" | |
194 "msub $ac1, %[load2], %[cospi_5_64] \n\t" | |
195 "extp %[temp0], $ac1, 31 \n\t" | |
196 | |
197 "madd $ac3, %[load1], %[cospi_5_64] \n\t" | |
198 "madd $ac3, %[load2], %[cospi_27_64] \n\t" | |
199 "extp %[temp3], $ac3, 31 \n\t" | |
200 | |
201 "mtlo %[const_2_power_13], $ac1 \n\t" | |
202 "mthi $zero, $ac1 \n\t" | |
203 "mtlo %[const_2_power_13], $ac2 \n\t" | |
204 "mthi $zero, $ac2 \n\t" | |
205 | |
206 "madd $ac2, %[load3], %[cospi_11_64] \n\t" | |
207 "msub $ac2, %[load4], %[cospi_21_64] \n\t" | |
208 "extp %[temp1], $ac2, 31 \n\t" | |
209 | |
210 "madd $ac1, %[load3], %[cospi_21_64] \n\t" | |
211 "madd $ac1, %[load4], %[cospi_11_64] \n\t" | |
212 "extp %[temp2], $ac1, 31 \n\t" | |
213 | |
214 "mtlo %[const_2_power_13], $ac1 \n\t" | |
215 "mthi $zero, $ac1 \n\t" | |
216 "mtlo %[const_2_power_13], $ac3 \n\t" | |
217 "mthi $zero, $ac3 \n\t" | |
218 | |
219 "sub %[load1], %[temp0], %[temp1] \n\t" | |
220 "sub %[load2], %[temp3], %[temp2] \n\t" | |
221 | |
222 "madd $ac1, %[load2], %[cospi_12_64] \n\t" | |
223 "msub $ac1, %[load1], %[cospi_20_64] \n\t" | |
224 "madd $ac3, %[load1], %[cospi_12_64] \n\t" | |
225 "madd $ac3, %[load2], %[cospi_20_64] \n\t" | |
226 | |
227 "extp %[step1_21], $ac1, 31 \n\t" | |
228 "extp %[step1_26], $ac3, 31 \n\t" | |
229 "add %[step1_20], %[temp0], %[temp1] \n\t" | |
230 "add %[step1_27], %[temp2], %[temp3] \n\t" | |
231 | |
232 : [load1] "=&r" (load1), [load2] "=&r" (load2), [load3] "=&r" (load3), | |
233 [load4] "=&r" (load4), [temp0] "=&r" (temp0), [temp1] "=&r" (temp1), | |
234 [temp2] "=&r" (temp2), [temp3] "=&r" (temp3), | |
235 [step1_20] "=r" (step1_20), [step1_21] "=r" (step1_21), | |
236 [step1_26] "=r" (step1_26), [step1_27] "=r" (step1_27) | |
237 : [const_2_power_13] "r" (const_2_power_13), [input] "r" (input), | |
238 [cospi_27_64] "r" (cospi_27_64), [cospi_5_64] "r" (cospi_5_64), | |
239 [cospi_11_64] "r" (cospi_11_64), [cospi_21_64] "r" (cospi_21_64), | |
240 [cospi_12_64] "r" (cospi_12_64), [cospi_20_64] "r" (cospi_20_64) | |
241 ); | |
242 | |
243 __asm__ __volatile__ ( | |
244 "lh %[load1], 26(%[input]) \n\t" | |
245 "lh %[load2], 38(%[input]) \n\t" | |
246 "lh %[load3], 58(%[input]) \n\t" | |
247 "lh %[load4], 6(%[input]) \n\t" | |
248 | |
249 "mtlo %[const_2_power_13], $ac1 \n\t" | |
250 "mthi $zero, $ac1 \n\t" | |
251 "mtlo %[const_2_power_13], $ac3 \n\t" | |
252 "mthi $zero, $ac3 \n\t" | |
253 | |
254 "madd $ac1, %[load1], %[cospi_19_64] \n\t" | |
255 "msub $ac1, %[load2], %[cospi_13_64] \n\t" | |
256 "extp %[temp0], $ac1, 31 \n\t" | |
257 "madd $ac3, %[load1], %[cospi_13_64] \n\t" | |
258 "madd $ac3, %[load2], %[cospi_19_64] \n\t" | |
259 "extp %[temp3], $ac3, 31 \n\t" | |
260 | |
261 "mtlo %[const_2_power_13], $ac1 \n\t" | |
262 "mthi $zero, $ac1 \n\t" | |
263 "mtlo %[const_2_power_13], $ac2 \n\t" | |
264 "mthi $zero, $ac2 \n\t" | |
265 | |
266 "madd $ac2, %[load3], %[cospi_3_64] \n\t" | |
267 "msub $ac2, %[load4], %[cospi_29_64] \n\t" | |
268 "extp %[temp1], $ac2, 31 \n\t" | |
269 "madd $ac1, %[load3], %[cospi_29_64] \n\t" | |
270 "madd $ac1, %[load4], %[cospi_3_64] \n\t" | |
271 "extp %[temp2], $ac1, 31 \n\t" | |
272 | |
273 "mtlo %[const_2_power_13], $ac1 \n\t" | |
274 "mthi $zero, $ac1 \n\t" | |
275 "mtlo %[const_2_power_13], $ac3 \n\t" | |
276 "mthi $zero, $ac3 \n\t" | |
277 | |
278 "sub %[load1], %[temp1], %[temp0] \n\t" | |
279 "sub %[load2], %[temp2], %[temp3] \n\t" | |
280 "msub $ac1, %[load1], %[cospi_12_64] \n\t" | |
281 "msub $ac1, %[load2], %[cospi_20_64] \n\t" | |
282 "msub $ac3, %[load1], %[cospi_20_64] \n\t" | |
283 "madd $ac3, %[load2], %[cospi_12_64] \n\t" | |
284 "extp %[step1_22], $ac1, 31 \n\t" | |
285 "extp %[step1_25], $ac3, 31 \n\t" | |
286 "add %[step1_23], %[temp0], %[temp1] \n\t" | |
287 "add %[step1_24], %[temp2], %[temp3] \n\t" | |
288 | |
289 : [load1] "=&r" (load1), [load2] "=&r" (load2), [load3] "=&r" (load3), | |
290 [load4] "=&r" (load4), [temp0] "=&r" (temp0), [temp1] "=&r" (temp1), | |
291 [temp2] "=&r" (temp2), [temp3] "=&r" (temp3), | |
292 [step1_22] "=r" (step1_22), [step1_23] "=r" (step1_23), | |
293 [step1_24] "=r" (step1_24), [step1_25] "=r" (step1_25) | |
294 : [const_2_power_13] "r" (const_2_power_13), [input] "r" (input), | |
295 [cospi_19_64] "r" (cospi_19_64), [cospi_13_64] "r" (cospi_13_64), | |
296 [cospi_3_64] "r" (cospi_3_64), [cospi_29_64] "r" (cospi_29_64), | |
297 [cospi_12_64] "r" (cospi_12_64), [cospi_20_64] "r" (cospi_20_64) | |
298 ); | |
299 | |
300 __asm__ __volatile__ ( | |
301 "lh %[load1], 4(%[input]) \n\t" | |
302 "lh %[load2], 60(%[input]) \n\t" | |
303 "lh %[load3], 36(%[input]) \n\t" | |
304 "lh %[load4], 28(%[input]) \n\t" | |
305 | |
306 "mtlo %[const_2_power_13], $ac1 \n\t" | |
307 "mthi $zero, $ac1 \n\t" | |
308 "mtlo %[const_2_power_13], $ac3 \n\t" | |
309 "mthi $zero, $ac3 \n\t" | |
310 | |
311 "madd $ac1, %[load1], %[cospi_30_64] \n\t" | |
312 "msub $ac1, %[load2], %[cospi_2_64] \n\t" | |
313 "extp %[temp0], $ac1, 31 \n\t" | |
314 "madd $ac3, %[load1], %[cospi_2_64] \n\t" | |
315 "madd $ac3, %[load2], %[cospi_30_64] \n\t" | |
316 "extp %[temp3], $ac3, 31 \n\t" | |
317 | |
318 "mtlo %[const_2_power_13], $ac1 \n\t" | |
319 "mthi $zero, $ac1 \n\t" | |
320 "mtlo %[const_2_power_13], $ac2 \n\t" | |
321 "mthi $zero, $ac2 \n\t" | |
322 | |
323 "madd $ac2, %[load3], %[cospi_14_64] \n\t" | |
324 "msub $ac2, %[load4], %[cospi_18_64] \n\t" | |
325 "extp %[temp1], $ac2, 31 \n\t" | |
326 "madd $ac1, %[load3], %[cospi_18_64] \n\t" | |
327 "madd $ac1, %[load4], %[cospi_14_64] \n\t" | |
328 "extp %[temp2], $ac1, 31 \n\t" | |
329 | |
330 "mtlo %[const_2_power_13], $ac1 \n\t" | |
331 "mthi $zero, $ac1 \n\t" | |
332 "mtlo %[const_2_power_13], $ac3 \n\t" | |
333 "mthi $zero, $ac3 \n\t" | |
334 | |
335 "sub %[load1], %[temp0], %[temp1] \n\t" | |
336 "sub %[load2], %[temp3], %[temp2] \n\t" | |
337 "msub $ac1, %[load1], %[cospi_8_64] \n\t" | |
338 "madd $ac1, %[load2], %[cospi_24_64] \n\t" | |
339 "madd $ac3, %[load1], %[cospi_24_64] \n\t" | |
340 "madd $ac3, %[load2], %[cospi_8_64] \n\t" | |
341 "extp %[step2_9], $ac1, 31 \n\t" | |
342 "extp %[step2_14], $ac3, 31 \n\t" | |
343 "add %[step2_8], %[temp0], %[temp1] \n\t" | |
344 "add %[step2_15], %[temp2], %[temp3] \n\t" | |
345 | |
346 : [load1] "=&r" (load1), [load2] "=&r" (load2), [load3] "=&r" (load3), | |
347 [load4] "=&r" (load4), [temp0] "=&r" (temp0), [temp1] "=&r" (temp1), | |
348 [temp2] "=&r" (temp2), [temp3] "=&r" (temp3), | |
349 [step2_8] "=r" (step2_8), [step2_9] "=r" (step2_9), | |
350 [step2_14] "=r" (step2_14), [step2_15] "=r" (step2_15) | |
351 : [const_2_power_13] "r" (const_2_power_13), [input] "r" (input), | |
352 [cospi_30_64] "r" (cospi_30_64), [cospi_2_64] "r" (cospi_2_64), | |
353 [cospi_14_64] "r" (cospi_14_64), [cospi_18_64] "r" (cospi_18_64), | |
354 [cospi_8_64] "r" (cospi_8_64), [cospi_24_64] "r" (cospi_24_64) | |
355 ); | |
356 | |
357 __asm__ __volatile__ ( | |
358 "lh %[load1], 20(%[input]) \n\t" | |
359 "lh %[load2], 44(%[input]) \n\t" | |
360 "lh %[load3], 52(%[input]) \n\t" | |
361 "lh %[load4], 12(%[input]) \n\t" | |
362 | |
363 "mtlo %[const_2_power_13], $ac1 \n\t" | |
364 "mthi $zero, $ac1 \n\t" | |
365 "mtlo %[const_2_power_13], $ac3 \n\t" | |
366 "mthi $zero, $ac3 \n\t" | |
367 | |
368 "madd $ac1, %[load1], %[cospi_22_64] \n\t" | |
369 "msub $ac1, %[load2], %[cospi_10_64] \n\t" | |
370 "extp %[temp0], $ac1, 31 \n\t" | |
371 "madd $ac3, %[load1], %[cospi_10_64] \n\t" | |
372 "madd $ac3, %[load2], %[cospi_22_64] \n\t" | |
373 "extp %[temp3], $ac3, 31 \n\t" | |
374 | |
375 "mtlo %[const_2_power_13], $ac1 \n\t" | |
376 "mthi $zero, $ac1 \n\t" | |
377 "mtlo %[const_2_power_13], $ac2 \n\t" | |
378 "mthi $zero, $ac2 \n\t" | |
379 | |
380 "madd $ac2, %[load3], %[cospi_6_64] \n\t" | |
381 "msub $ac2, %[load4], %[cospi_26_64] \n\t" | |
382 "extp %[temp1], $ac2, 31 \n\t" | |
383 "madd $ac1, %[load3], %[cospi_26_64] \n\t" | |
384 "madd $ac1, %[load4], %[cospi_6_64] \n\t" | |
385 "extp %[temp2], $ac1, 31 \n\t" | |
386 | |
387 "mtlo %[const_2_power_13], $ac1 \n\t" | |
388 "mthi $zero, $ac1 \n\t" | |
389 "mtlo %[const_2_power_13], $ac3 \n\t" | |
390 "mthi $zero, $ac3 \n\t" | |
391 | |
392 "sub %[load1], %[temp1], %[temp0] \n\t" | |
393 "sub %[load2], %[temp2], %[temp3] \n\t" | |
394 "msub $ac1, %[load1], %[cospi_24_64] \n\t" | |
395 "msub $ac1, %[load2], %[cospi_8_64] \n\t" | |
396 "madd $ac3, %[load2], %[cospi_24_64] \n\t" | |
397 "msub $ac3, %[load1], %[cospi_8_64] \n\t" | |
398 "extp %[step2_10], $ac1, 31 \n\t" | |
399 "extp %[step2_13], $ac3, 31 \n\t" | |
400 "add %[step2_11], %[temp0], %[temp1] \n\t" | |
401 "add %[step2_12], %[temp2], %[temp3] \n\t" | |
402 | |
403 : [load1] "=&r" (load1), [load2] "=&r" (load2), [load3] "=&r" (load3), | |
404 [load4] "=&r" (load4), [temp0] "=&r" (temp0), [temp1] "=&r" (temp1), | |
405 [temp2] "=&r" (temp2), [temp3] "=&r" (temp3), | |
406 [step2_10] "=r" (step2_10), [step2_11] "=r" (step2_11), | |
407 [step2_12] "=r" (step2_12), [step2_13] "=r" (step2_13) | |
408 : [const_2_power_13] "r" (const_2_power_13), [input] "r" (input), | |
409 [cospi_22_64] "r" (cospi_22_64), [cospi_10_64] "r" (cospi_10_64), | |
410 [cospi_6_64] "r" (cospi_6_64), [cospi_26_64] "r" (cospi_26_64), | |
411 [cospi_8_64] "r" (cospi_8_64), [cospi_24_64] "r" (cospi_24_64) | |
412 ); | |
413 | |
414 __asm__ __volatile__ ( | |
415 "mtlo %[const_2_power_13], $ac0 \n\t" | |
416 "mthi $zero, $ac0 \n\t" | |
417 "sub %[temp0], %[step2_14], %[step2_13] \n\t" | |
418 "sub %[temp0], %[temp0], %[step2_9] \n\t" | |
419 "add %[temp0], %[temp0], %[step2_10] \n\t" | |
420 "madd $ac0, %[temp0], %[cospi_16_64] \n\t" | |
421 "mtlo %[const_2_power_13], $ac1 \n\t" | |
422 "mthi $zero, $ac1 \n\t" | |
423 "sub %[temp1], %[step2_14], %[step2_13] \n\t" | |
424 "add %[temp1], %[temp1], %[step2_9] \n\t" | |
425 "sub %[temp1], %[temp1], %[step2_10] \n\t" | |
426 "madd $ac1, %[temp1], %[cospi_16_64] \n\t" | |
427 "mtlo %[const_2_power_13], $ac2 \n\t" | |
428 "mthi $zero, $ac2 \n\t" | |
429 "sub %[temp0], %[step2_15], %[step2_12] \n\t" | |
430 "sub %[temp0], %[temp0], %[step2_8] \n\t" | |
431 "add %[temp0], %[temp0], %[step2_11] \n\t" | |
432 "madd $ac2, %[temp0], %[cospi_16_64] \n\t" | |
433 "mtlo %[const_2_power_13], $ac3 \n\t" | |
434 "mthi $zero, $ac3 \n\t" | |
435 "sub %[temp1], %[step2_15], %[step2_12] \n\t" | |
436 "add %[temp1], %[temp1], %[step2_8] \n\t" | |
437 "sub %[temp1], %[temp1], %[step2_11] \n\t" | |
438 "madd $ac3, %[temp1], %[cospi_16_64] \n\t" | |
439 | |
440 "add %[step3_8], %[step2_8], %[step2_11] \n\t" | |
441 "add %[step3_9], %[step2_9], %[step2_10] \n\t" | |
442 "add %[step3_14], %[step2_13], %[step2_14] \n\t" | |
443 "add %[step3_15], %[step2_12], %[step2_15] \n\t" | |
444 "extp %[step3_10], $ac0, 31 \n\t" | |
445 "extp %[step3_13], $ac1, 31 \n\t" | |
446 "extp %[step3_11], $ac2, 31 \n\t" | |
447 "extp %[step3_12], $ac3, 31 \n\t" | |
448 | |
449 : [temp0] "=&r" (temp0), [temp1] "=&r" (temp1), | |
450 [step3_8] "=r" (step3_8), [step3_9] "=r" (step3_9), | |
451 [step3_10] "=r" (step3_10), [step3_11] "=r" (step3_11), | |
452 [step3_12] "=r" (step3_12), [step3_13] "=r" (step3_13), | |
453 [step3_14] "=r" (step3_14), [step3_15] "=r" (step3_15) | |
454 : [const_2_power_13] "r" (const_2_power_13), [step2_8] "r" (step2_8), | |
455 [step2_9] "r" (step2_9), [step2_10] "r" (step2_10), | |
456 [step2_11] "r" (step2_11), [step2_12] "r" (step2_12), | |
457 [step2_13] "r" (step2_13), [step2_14] "r" (step2_14), | |
458 [step2_15] "r" (step2_15), [cospi_16_64] "r" (cospi_16_64) | |
459 ); | |
460 | |
461 step2_18 = step1_17 - step1_18; | |
462 step2_29 = step1_30 - step1_29; | |
463 | |
464 __asm__ __volatile__ ( | |
465 "mtlo %[const_2_power_13], $ac0 \n\t" | |
466 "mthi $zero, $ac0 \n\t" | |
467 "msub $ac0, %[step2_18], %[cospi_8_64] \n\t" | |
468 "madd $ac0, %[step2_29], %[cospi_24_64] \n\t" | |
469 "extp %[step3_18], $ac0, 31 \n\t" | |
470 | |
471 : [step3_18] "=r" (step3_18) | |
472 : [const_2_power_13] "r" (const_2_power_13), | |
473 [step2_18] "r" (step2_18), [step2_29] "r" (step2_29), | |
474 [cospi_24_64] "r" (cospi_24_64), [cospi_8_64] "r" (cospi_8_64) | |
475 ); | |
476 | |
477 temp21 = step2_18 * cospi_24_64 + step2_29 * cospi_8_64; | |
478 step3_29 = (temp21 + DCT_CONST_ROUNDING) >> DCT_CONST_BITS; | |
479 | |
480 step2_19 = step1_16 - step1_19; | |
481 step2_28 = step1_31 - step1_28; | |
482 | |
483 __asm__ __volatile__ ( | |
484 "mtlo %[const_2_power_13], $ac0 \n\t" | |
485 "mthi $zero, $ac0 \n\t" | |
486 "msub $ac0, %[step2_19], %[cospi_8_64] \n\t" | |
487 "madd $ac0, %[step2_28], %[cospi_24_64] \n\t" | |
488 "extp %[step3_19], $ac0, 31 \n\t" | |
489 | |
490 : [step3_19] "=r" (step3_19) | |
491 : [const_2_power_13] "r" (const_2_power_13), | |
492 [step2_19] "r" (step2_19), [step2_28] "r" (step2_28), | |
493 [cospi_24_64] "r" (cospi_24_64), [cospi_8_64] "r" (cospi_8_64) | |
494 ); | |
495 | |
496 temp21 = step2_19 * cospi_24_64 + step2_28 * cospi_8_64; | |
497 step3_28 = (temp21 + DCT_CONST_ROUNDING) >> DCT_CONST_BITS; | |
498 | |
499 step3_16 = step1_16 + step1_19; | |
500 step3_17 = step1_17 + step1_18; | |
501 step3_30 = step1_29 + step1_30; | |
502 step3_31 = step1_28 + step1_31; | |
503 | |
504 step2_20 = step1_23 - step1_20; | |
505 step2_27 = step1_24 - step1_27; | |
506 | |
507 __asm__ __volatile__ ( | |
508 "mtlo %[const_2_power_13], $ac0 \n\t" | |
509 "mthi $zero, $ac0 \n\t" | |
510 "msub $ac0, %[step2_20], %[cospi_24_64] \n\t" | |
511 "msub $ac0, %[step2_27], %[cospi_8_64] \n\t" | |
512 "extp %[step3_20], $ac0, 31 \n\t" | |
513 | |
514 : [step3_20] "=r" (step3_20) | |
515 : [const_2_power_13] "r" (const_2_power_13), | |
516 [step2_20] "r" (step2_20), [step2_27] "r" (step2_27), | |
517 [cospi_24_64] "r" (cospi_24_64), [cospi_8_64] "r" (cospi_8_64) | |
518 ); | |
519 | |
520 temp21 = -step2_20 * cospi_8_64 + step2_27 * cospi_24_64; | |
521 step3_27 = (temp21 + DCT_CONST_ROUNDING) >> DCT_CONST_BITS; | |
522 | |
523 step2_21 = step1_22 - step1_21; | |
524 step2_26 = step1_25 - step1_26; | |
525 | |
526 __asm__ __volatile__ ( | |
527 "mtlo %[const_2_power_13], $ac1 \n\t" | |
528 "mthi $zero, $ac1 \n\t" | |
529 "msub $ac1, %[step2_21], %[cospi_24_64] \n\t" | |
530 "msub $ac1, %[step2_26], %[cospi_8_64] \n\t" | |
531 "extp %[step3_21], $ac1, 31 \n\t" | |
532 | |
533 : [step3_21] "=r" (step3_21) | |
534 : [const_2_power_13] "r" (const_2_power_13), | |
535 [step2_21] "r" (step2_21), [step2_26] "r" (step2_26), | |
536 [cospi_24_64] "r" (cospi_24_64), [cospi_8_64] "r" (cospi_8_64) | |
537 ); | |
538 | |
539 temp21 = -step2_21 * cospi_8_64 + step2_26 * cospi_24_64; | |
540 step3_26 = (temp21 + DCT_CONST_ROUNDING) >> DCT_CONST_BITS; | |
541 | |
542 step3_22 = step1_21 + step1_22; | |
543 step3_23 = step1_20 + step1_23; | |
544 step3_24 = step1_24 + step1_27; | |
545 step3_25 = step1_25 + step1_26; | |
546 | |
547 step2_16 = step3_16 + step3_23; | |
548 step2_17 = step3_17 + step3_22; | |
549 step2_18 = step3_18 + step3_21; | |
550 step2_19 = step3_19 + step3_20; | |
551 step2_20 = step3_19 - step3_20; | |
552 step2_21 = step3_18 - step3_21; | |
553 step2_22 = step3_17 - step3_22; | |
554 step2_23 = step3_16 - step3_23; | |
555 | |
556 step2_24 = step3_31 - step3_24; | |
557 step2_25 = step3_30 - step3_25; | |
558 step2_26 = step3_29 - step3_26; | |
559 step2_27 = step3_28 - step3_27; | |
560 step2_28 = step3_28 + step3_27; | |
561 step2_29 = step3_29 + step3_26; | |
562 step2_30 = step3_30 + step3_25; | |
563 step2_31 = step3_31 + step3_24; | |
564 | |
565 __asm__ __volatile__ ( | |
566 "lh %[load1], 0(%[input]) \n\t" | |
567 "lh %[load2], 32(%[input]) \n\t" | |
568 "lh %[load3], 16(%[input]) \n\t" | |
569 "lh %[load4], 48(%[input]) \n\t" | |
570 | |
571 "mtlo %[const_2_power_13], $ac1 \n\t" | |
572 "mthi $zero, $ac1 \n\t" | |
573 "mtlo %[const_2_power_13], $ac2 \n\t" | |
574 "mthi $zero, $ac2 \n\t" | |
575 "add %[result1], %[load1], %[load2] \n\t" | |
576 "sub %[result2], %[load1], %[load2] \n\t" | |
577 "madd $ac1, %[result1], %[cospi_16_64] \n\t" | |
578 "madd $ac2, %[result2], %[cospi_16_64] \n\t" | |
579 "extp %[temp0], $ac1, 31 \n\t" | |
580 "extp %[temp1], $ac2, 31 \n\t" | |
581 | |
582 "mtlo %[const_2_power_13], $ac3 \n\t" | |
583 "mthi $zero, $ac3 \n\t" | |
584 "madd $ac3, %[load3], %[cospi_24_64] \n\t" | |
585 "msub $ac3, %[load4], %[cospi_8_64] \n\t" | |
586 "extp %[temp2], $ac3, 31 \n\t" | |
587 "mtlo %[const_2_power_13], $ac1 \n\t" | |
588 "mthi $zero, $ac1 \n\t" | |
589 "madd $ac1, %[load3], %[cospi_8_64] \n\t" | |
590 "madd $ac1, %[load4], %[cospi_24_64] \n\t" | |
591 "extp %[temp3], $ac1, 31 \n\t" | |
592 "add %[step1_0], %[temp0], %[temp3] \n\t" | |
593 "add %[step1_1], %[temp1], %[temp2] \n\t" | |
594 "sub %[step1_2], %[temp1], %[temp2] \n\t" | |
595 "sub %[step1_3], %[temp0], %[temp3] \n\t" | |
596 | |
597 : [load1] "=&r" (load1), [load2] "=&r" (load2), | |
598 [load3] "=&r" (load3), [load4] "=&r" (load4), | |
599 [result1] "=&r" (result1), [result2] "=&r" (result2), | |
600 [temp0] "=&r" (temp0), [temp1] "=&r" (temp1), | |
601 [temp2] "=&r" (temp2), [temp3] "=&r" (temp3), | |
602 [step1_0] "=r" (step1_0), [step1_1] "=r" (step1_1), | |
603 [step1_2] "=r" (step1_2), [step1_3] "=r" (step1_3) | |
604 : [const_2_power_13] "r" (const_2_power_13), [input] "r" (input), | |
605 [cospi_24_64] "r" (cospi_24_64), [cospi_8_64] "r" (cospi_8_64), | |
606 [cospi_16_64] "r" (cospi_16_64) | |
607 ); | |
608 | |
609 __asm__ __volatile__ ( | |
610 "lh %[load1], 8(%[input]) \n\t" | |
611 "lh %[load2], 56(%[input]) \n\t" | |
612 "lh %[load3], 40(%[input]) \n\t" | |
613 "lh %[load4], 24(%[input]) \n\t" | |
614 | |
615 "mtlo %[const_2_power_13], $ac1 \n\t" | |
616 "mthi $zero, $ac1 \n\t" | |
617 "mtlo %[const_2_power_13], $ac3 \n\t" | |
618 "mthi $zero, $ac3 \n\t" | |
619 | |
620 "madd $ac1, %[load1], %[cospi_28_64] \n\t" | |
621 "msub $ac1, %[load2], %[cospi_4_64] \n\t" | |
622 "extp %[temp0], $ac1, 31 \n\t" | |
623 "madd $ac3, %[load1], %[cospi_4_64] \n\t" | |
624 "madd $ac3, %[load2], %[cospi_28_64] \n\t" | |
625 "extp %[temp3], $ac3, 31 \n\t" | |
626 | |
627 "mtlo %[const_2_power_13], $ac1 \n\t" | |
628 "mthi $zero, $ac1 \n\t" | |
629 "mtlo %[const_2_power_13], $ac2 \n\t" | |
630 "mthi $zero, $ac2 \n\t" | |
631 | |
632 "madd $ac2, %[load3], %[cospi_12_64] \n\t" | |
633 "msub $ac2, %[load4], %[cospi_20_64] \n\t" | |
634 "extp %[temp1], $ac2, 31 \n\t" | |
635 "madd $ac1, %[load3], %[cospi_20_64] \n\t" | |
636 "madd $ac1, %[load4], %[cospi_12_64] \n\t" | |
637 "extp %[temp2], $ac1, 31 \n\t" | |
638 | |
639 "mtlo %[const_2_power_13], $ac1 \n\t" | |
640 "mthi $zero, $ac1 \n\t" | |
641 "mtlo %[const_2_power_13], $ac3 \n\t" | |
642 "mthi $zero, $ac3 \n\t" | |
643 | |
644 "sub %[load1], %[temp3], %[temp2] \n\t" | |
645 "sub %[load1], %[load1], %[temp0] \n\t" | |
646 "add %[load1], %[load1], %[temp1] \n\t" | |
647 "sub %[load2], %[temp0], %[temp1] \n\t" | |
648 "sub %[load2], %[load2], %[temp2] \n\t" | |
649 "add %[load2], %[load2], %[temp3] \n\t" | |
650 "madd $ac1, %[load1], %[cospi_16_64] \n\t" | |
651 "madd $ac3, %[load2], %[cospi_16_64] \n\t" | |
652 | |
653 "extp %[step1_5], $ac1, 31 \n\t" | |
654 "extp %[step1_6], $ac3, 31 \n\t" | |
655 "add %[step1_4], %[temp0], %[temp1] \n\t" | |
656 "add %[step1_7], %[temp3], %[temp2] \n\t" | |
657 | |
658 : [load1] "=&r" (load1), [load2] "=&r" (load2), | |
659 [load3] "=&r" (load3), [load4] "=&r" (load4), | |
660 [temp0] "=&r" (temp0), [temp1] "=&r" (temp1), | |
661 [temp2] "=&r" (temp2), [temp3] "=&r" (temp3), | |
662 [step1_4] "=r" (step1_4), [step1_5] "=r" (step1_5), | |
663 [step1_6] "=r" (step1_6), [step1_7] "=r" (step1_7) | |
664 : [const_2_power_13] "r" (const_2_power_13), [input] "r" (input), | |
665 [cospi_20_64] "r" (cospi_20_64), [cospi_12_64] "r" (cospi_12_64), | |
666 [cospi_4_64] "r" (cospi_4_64), [cospi_28_64] "r" (cospi_28_64), | |
667 [cospi_16_64] "r" (cospi_16_64) | |
668 ); | |
669 | |
670 step2_0 = step1_0 + step1_7; | |
671 step2_1 = step1_1 + step1_6; | |
672 step2_2 = step1_2 + step1_5; | |
673 step2_3 = step1_3 + step1_4; | |
674 step2_4 = step1_3 - step1_4; | |
675 step2_5 = step1_2 - step1_5; | |
676 step2_6 = step1_1 - step1_6; | |
677 step2_7 = step1_0 - step1_7; | |
678 | |
679 // stage 7 | |
680 step1_0 = step2_0 + step3_15; | |
681 step1_1 = step2_1 + step3_14; | |
682 step1_2 = step2_2 + step3_13; | |
683 step1_3 = step2_3 + step3_12; | |
684 step1_4 = step2_4 + step3_11; | |
685 step1_5 = step2_5 + step3_10; | |
686 step1_6 = step2_6 + step3_9; | |
687 step1_7 = step2_7 + step3_8; | |
688 step1_8 = step2_7 - step3_8; | |
689 step1_9 = step2_6 - step3_9; | |
690 step1_10 = step2_5 - step3_10; | |
691 step1_11 = step2_4 - step3_11; | |
692 step1_12 = step2_3 - step3_12; | |
693 step1_13 = step2_2 - step3_13; | |
694 step1_14 = step2_1 - step3_14; | |
695 step1_15 = step2_0 - step3_15; | |
696 | |
697 __asm__ __volatile__ ( | |
698 "sub %[temp0], %[step2_27], %[step2_20] \n\t" | |
699 "mtlo %[const_2_power_13], $ac0 \n\t" | |
700 "mthi $zero, $ac0 \n\t" | |
701 "madd $ac0, %[temp0], %[cospi_16_64] \n\t" | |
702 "extp %[step1_20], $ac0, 31 \n\t" | |
703 | |
704 : [temp0] "=&r" (temp0), [step1_20] "=r" (step1_20) | |
705 : [const_2_power_13] "r" (const_2_power_13), [step2_20] "r" (step2_20), | |
706 [step2_27] "r" (step2_27), [cospi_16_64] "r" (cospi_16_64) | |
707 ); | |
708 | |
709 temp21 = (step2_20 + step2_27) * cospi_16_64; | |
710 step1_27 = (temp21 + DCT_CONST_ROUNDING) >> DCT_CONST_BITS; | |
711 | |
712 __asm__ __volatile__ ( | |
713 "sub %[temp0], %[step2_26], %[step2_21] \n\t" | |
714 "mtlo %[const_2_power_13], $ac0 \n\t" | |
715 "mthi $zero, $ac0 \n\t" | |
716 "madd $ac0, %[temp0], %[cospi_16_64] \n\t" | |
717 "extp %[step1_21], $ac0, 31 \n\t" | |
718 | |
719 : [temp0] "=&r" (temp0), [step1_21] "=r" (step1_21) | |
720 : [const_2_power_13] "r" (const_2_power_13), [step2_26] "r" (step2_26), | |
721 [step2_21] "r" (step2_21), [cospi_16_64] "r" (cospi_16_64) | |
722 ); | |
723 | |
724 temp21 = (step2_21 + step2_26) * cospi_16_64; | |
725 step1_26 = (temp21 + DCT_CONST_ROUNDING) >> DCT_CONST_BITS; | |
726 | |
727 __asm__ __volatile__ ( | |
728 "sub %[temp0], %[step2_25], %[step2_22] \n\t" | |
729 "mtlo %[const_2_power_13], $ac0 \n\t" | |
730 "mthi $zero, $ac0 \n\t" | |
731 "madd $ac0, %[temp0], %[cospi_16_64] \n\t" | |
732 "extp %[step1_22], $ac0, 31 \n\t" | |
733 | |
734 : [temp0] "=&r" (temp0), [step1_22] "=r" (step1_22) | |
735 : [const_2_power_13] "r" (const_2_power_13), [step2_25] "r" (step2_25), | |
736 [step2_22] "r" (step2_22), [cospi_16_64] "r" (cospi_16_64) | |
737 ); | |
738 | |
739 temp21 = (step2_22 + step2_25) * cospi_16_64; | |
740 step1_25 = (temp21 + DCT_CONST_ROUNDING) >> DCT_CONST_BITS; | |
741 | |
742 __asm__ __volatile__ ( | |
743 "sub %[temp0], %[step2_24], %[step2_23] \n\t" | |
744 "mtlo %[const_2_power_13], $ac0 \n\t" | |
745 "mthi $zero, $ac0 \n\t" | |
746 "madd $ac0, %[temp0], %[cospi_16_64] \n\t" | |
747 "extp %[step1_23], $ac0, 31 \n\t" | |
748 | |
749 : [temp0] "=&r" (temp0), [step1_23] "=r" (step1_23) | |
750 : [const_2_power_13] "r" (const_2_power_13), [step2_24] "r" (step2_24), | |
751 [step2_23] "r" (step2_23), [cospi_16_64] "r" (cospi_16_64) | |
752 ); | |
753 | |
754 temp21 = (step2_23 + step2_24) * cospi_16_64; | |
755 step1_24 = (temp21 + DCT_CONST_ROUNDING) >> DCT_CONST_BITS; | |
756 | |
757 __asm__ __volatile__ ( | |
758 "lbu %[temp2], 0(%[dest_pix]) \n\t" | |
759 "add %[temp0], %[step1_0], %[step2_31] \n\t" | |
760 "addi %[temp0], %[temp0], 32 \n\t" | |
761 "sra %[temp0], %[temp0], 6 \n\t" | |
762 "add %[temp2], %[temp2], %[temp0] \n\t" | |
763 "lbux %[temp0], %[temp2](%[cm]) \n\t" | |
764 "add %[temp1], %[step1_1], %[step2_30] \n\t" | |
765 "sb %[temp0], 0(%[dest_pix]) \n\t" | |
766 "addu %[dest_pix], %[dest_pix], %[dest_stride] \n\t" | |
767 "lbu %[temp3], 0(%[dest_pix]) \n\t" | |
768 "addi %[temp1], %[temp1], 32 \n\t" | |
769 "sra %[temp1], %[temp1], 6 \n\t" | |
770 "add %[temp3], %[temp3], %[temp1] \n\t" | |
771 "lbux %[temp1], %[temp3](%[cm]) \n\t" | |
772 "sb %[temp1], 0(%[dest_pix]) \n\t" | |
773 "addu %[dest_pix], %[dest_pix], %[dest_stride] \n\t" | |
774 | |
775 "lbu %[temp2], 0(%[dest_pix]) \n\t" | |
776 "add %[temp0], %[step1_2], %[step2_29] \n\t" | |
777 "addi %[temp0], %[temp0], 32 \n\t" | |
778 "sra %[temp0], %[temp0], 6 \n\t" | |
779 "add %[temp2], %[temp2], %[temp0] \n\t" | |
780 "lbux %[temp0], %[temp2](%[cm]) \n\t" | |
781 "add %[temp1], %[step1_3], %[step2_28] \n\t" | |
782 "sb %[temp0], 0(%[dest_pix]) \n\t" | |
783 "addu %[dest_pix], %[dest_pix], %[dest_stride] \n\t" | |
784 "lbu %[temp3], 0(%[dest_pix]) \n\t" | |
785 "addi %[temp1], %[temp1], 32 \n\t" | |
786 "sra %[temp1], %[temp1], 6 \n\t" | |
787 "add %[temp3], %[temp3], %[temp1] \n\t" | |
788 "lbux %[temp1], %[temp3](%[cm]) \n\t" | |
789 "sb %[temp1], 0(%[dest_pix]) \n\t" | |
790 "addu %[dest_pix], %[dest_pix], %[dest_stride] \n\t" | |
791 | |
792 : [temp0] "=&r" (temp0), [temp1] "=&r" (temp1), [temp2] "=&r" (temp2), | |
793 [temp3] "=&r" (temp3), [dest_pix] "+r" (dest_pix) | |
794 : [cm] "r" (cm), [dest_stride] "r" (dest_stride), | |
795 [step1_0] "r" (step1_0), [step1_1] "r" (step1_1), | |
796 [step1_2] "r" (step1_2), [step1_3] "r" (step1_3), | |
797 [step2_28] "r" (step2_28), [step2_29] "r" (step2_29), | |
798 [step2_30] "r" (step2_30), [step2_31] "r" (step2_31) | |
799 ); | |
800 | |
801 step3_12 = ROUND_POWER_OF_TWO((step1_3 - step2_28), 6); | |
802 step3_13 = ROUND_POWER_OF_TWO((step1_2 - step2_29), 6); | |
803 step3_14 = ROUND_POWER_OF_TWO((step1_1 - step2_30), 6); | |
804 step3_15 = ROUND_POWER_OF_TWO((step1_0 - step2_31), 6); | |
805 | |
806 __asm__ __volatile__ ( | |
807 "lbu %[temp2], 0(%[dest_pix1]) \n\t" | |
808 "add %[temp2], %[temp2], %[step3_15] \n\t" | |
809 "lbux %[temp0], %[temp2](%[cm]) \n\t" | |
810 "sb %[temp0], 0(%[dest_pix1]) \n\t" | |
811 "subu %[dest_pix1], %[dest_pix1], %[dest_stride] \n\t" | |
812 "lbu %[temp3], 0(%[dest_pix1]) \n\t" | |
813 "add %[temp3], %[temp3], %[step3_14] \n\t" | |
814 "lbux %[temp1], %[temp3](%[cm]) \n\t" | |
815 "sb %[temp1], 0(%[dest_pix1]) \n\t" | |
816 "subu %[dest_pix1], %[dest_pix1], %[dest_stride] \n\t" | |
817 | |
818 "lbu %[temp2], 0(%[dest_pix1]) \n\t" | |
819 "add %[temp2], %[temp2], %[step3_13] \n\t" | |
820 "lbux %[temp0], %[temp2](%[cm]) \n\t" | |
821 "sb %[temp0], 0(%[dest_pix1]) \n\t" | |
822 "subu %[dest_pix1], %[dest_pix1], %[dest_stride] \n\t" | |
823 "lbu %[temp3], 0(%[dest_pix1]) \n\t" | |
824 "add %[temp3], %[temp3], %[step3_12] \n\t" | |
825 "lbux %[temp1], %[temp3](%[cm]) \n\t" | |
826 "sb %[temp1], 0(%[dest_pix1]) \n\t" | |
827 "subu %[dest_pix1], %[dest_pix1], %[dest_stride] \n\t" | |
828 | |
829 : [temp0] "=&r" (temp0), [temp1] "=&r" (temp1), [temp2] "=&r" (temp2), | |
830 [temp3] "=&r" (temp3), [dest_pix1] "+r" (dest_pix1) | |
831 : [cm] "r" (cm), [dest_stride] "r" (dest_stride), | |
832 [step3_12] "r" (step3_12), [step3_13] "r" (step3_13), | |
833 [step3_14] "r" (step3_14), [step3_15] "r" (step3_15) | |
834 ); | |
835 | |
836 __asm__ __volatile__ ( | |
837 "lbu %[temp2], 0(%[dest_pix]) \n\t" | |
838 "add %[temp0], %[step1_4], %[step1_27] \n\t" | |
839 "addi %[temp0], %[temp0], 32 \n\t" | |
840 "sra %[temp0], %[temp0], 6 \n\t" | |
841 "add %[temp2], %[temp2], %[temp0] \n\t" | |
842 "lbux %[temp0], %[temp2](%[cm]) \n\t" | |
843 "add %[temp1], %[step1_5], %[step1_26] \n\t" | |
844 "sb %[temp0], 0(%[dest_pix]) \n\t" | |
845 "addu %[dest_pix], %[dest_pix], %[dest_stride] \n\t" | |
846 "lbu %[temp3], 0(%[dest_pix]) \n\t" | |
847 "addi %[temp1], %[temp1], 32 \n\t" | |
848 "sra %[temp1], %[temp1], 6 \n\t" | |
849 "add %[temp3], %[temp3], %[temp1] \n\t" | |
850 "lbux %[temp1], %[temp3](%[cm]) \n\t" | |
851 "sb %[temp1], 0(%[dest_pix]) \n\t" | |
852 "addu %[dest_pix], %[dest_pix], %[dest_stride] \n\t" | |
853 | |
854 "lbu %[temp2], 0(%[dest_pix]) \n\t" | |
855 "add %[temp0], %[step1_6], %[step1_25] \n\t" | |
856 "addi %[temp0], %[temp0], 32 \n\t" | |
857 "sra %[temp0], %[temp0], 6 \n\t" | |
858 "add %[temp2], %[temp2], %[temp0] \n\t" | |
859 "lbux %[temp0], %[temp2](%[cm]) \n\t" | |
860 "add %[temp1], %[step1_7], %[step1_24] \n\t" | |
861 "sb %[temp0], 0(%[dest_pix]) \n\t" | |
862 "addu %[dest_pix], %[dest_pix], %[dest_stride] \n\t" | |
863 "lbu %[temp3], 0(%[dest_pix]) \n\t" | |
864 "addi %[temp1], %[temp1], 32 \n\t" | |
865 "sra %[temp1], %[temp1], 6 \n\t" | |
866 "add %[temp3], %[temp3], %[temp1] \n\t" | |
867 "lbux %[temp1], %[temp3](%[cm]) \n\t" | |
868 "sb %[temp1], 0(%[dest_pix]) \n\t" | |
869 "addu %[dest_pix], %[dest_pix], %[dest_stride] \n\t" | |
870 | |
871 : [temp0] "=&r" (temp0), [temp1] "=&r" (temp1), [temp2] "=&r" (temp2), | |
872 [temp3] "=&r" (temp3), [dest_pix] "+r" (dest_pix) | |
873 : [cm] "r" (cm), [dest_stride] "r" (dest_stride), | |
874 [step1_4] "r" (step1_4), [step1_5] "r" (step1_5), | |
875 [step1_6] "r" (step1_6), [step1_7] "r" (step1_7), | |
876 [step1_24] "r" (step1_24), [step1_25] "r" (step1_25), | |
877 [step1_26] "r" (step1_26), [step1_27] "r" (step1_27) | |
878 ); | |
879 | |
880 step3_12 = ROUND_POWER_OF_TWO((step1_7 - step1_24), 6); | |
881 step3_13 = ROUND_POWER_OF_TWO((step1_6 - step1_25), 6); | |
882 step3_14 = ROUND_POWER_OF_TWO((step1_5 - step1_26), 6); | |
883 step3_15 = ROUND_POWER_OF_TWO((step1_4 - step1_27), 6); | |
884 | |
885 __asm__ __volatile__ ( | |
886 "lbu %[temp2], 0(%[dest_pix1]) \n\t" | |
887 "add %[temp2], %[temp2], %[step3_15] \n\t" | |
888 "lbux %[temp0], %[temp2](%[cm]) \n\t" | |
889 "sb %[temp0], 0(%[dest_pix1]) \n\t" | |
890 "subu %[dest_pix1], %[dest_pix1], %[dest_stride] \n\t" | |
891 "lbu %[temp3], 0(%[dest_pix1]) \n\t" | |
892 "add %[temp3], %[temp3], %[step3_14] \n\t" | |
893 "lbux %[temp1], %[temp3](%[cm]) \n\t" | |
894 "sb %[temp1], 0(%[dest_pix1]) \n\t" | |
895 "subu %[dest_pix1], %[dest_pix1], %[dest_stride] \n\t" | |
896 | |
897 "lbu %[temp2], 0(%[dest_pix1]) \n\t" | |
898 "add %[temp2], %[temp2], %[step3_13] \n\t" | |
899 "lbux %[temp0], %[temp2](%[cm]) \n\t" | |
900 "sb %[temp0], 0(%[dest_pix1]) \n\t" | |
901 "subu %[dest_pix1], %[dest_pix1], %[dest_stride] \n\t" | |
902 "lbu %[temp3], 0(%[dest_pix1]) \n\t" | |
903 "add %[temp3], %[temp3], %[step3_12] \n\t" | |
904 "lbux %[temp1], %[temp3](%[cm]) \n\t" | |
905 "sb %[temp1], 0(%[dest_pix1]) \n\t" | |
906 "subu %[dest_pix1], %[dest_pix1], %[dest_stride] \n\t" | |
907 | |
908 : [temp0] "=&r" (temp0), [temp1] "=&r" (temp1), [temp2] "=&r" (temp2), | |
909 [temp3] "=&r" (temp3), [dest_pix1] "+r" (dest_pix1) | |
910 : [cm] "r" (cm), [dest_stride] "r" (dest_stride), | |
911 [step3_12] "r" (step3_12), [step3_13] "r" (step3_13), | |
912 [step3_14] "r" (step3_14), [step3_15] "r" (step3_15) | |
913 ); | |
914 | |
915 __asm__ __volatile__ ( | |
916 "lbu %[temp2], 0(%[dest_pix]) \n\t" | |
917 "add %[temp0], %[step1_8], %[step1_23] \n\t" | |
918 "addi %[temp0], %[temp0], 32 \n\t" | |
919 "sra %[temp0], %[temp0], 6 \n\t" | |
920 "add %[temp2], %[temp2], %[temp0] \n\t" | |
921 "lbux %[temp0], %[temp2](%[cm]) \n\t" | |
922 "add %[temp1], %[step1_9], %[step1_22] \n\t" | |
923 "sb %[temp0], 0(%[dest_pix]) \n\t" | |
924 "addu %[dest_pix], %[dest_pix], %[dest_stride] \n\t" | |
925 "lbu %[temp3], 0(%[dest_pix]) \n\t" | |
926 "addi %[temp1], %[temp1], 32 \n\t" | |
927 "sra %[temp1], %[temp1], 6 \n\t" | |
928 "add %[temp3], %[temp3], %[temp1] \n\t" | |
929 "lbux %[temp1], %[temp3](%[cm]) \n\t" | |
930 "sb %[temp1], 0(%[dest_pix]) \n\t" | |
931 "addu %[dest_pix], %[dest_pix], %[dest_stride] \n\t" | |
932 | |
933 "lbu %[temp2], 0(%[dest_pix]) \n\t" | |
934 "add %[temp0], %[step1_10], %[step1_21] \n\t" | |
935 "addi %[temp0], %[temp0], 32 \n\t" | |
936 "sra %[temp0], %[temp0], 6 \n\t" | |
937 "add %[temp2], %[temp2], %[temp0] \n\t" | |
938 "lbux %[temp0], %[temp2](%[cm]) \n\t" | |
939 "add %[temp1], %[step1_11], %[step1_20] \n\t" | |
940 "sb %[temp0], 0(%[dest_pix]) \n\t" | |
941 "addu %[dest_pix], %[dest_pix], %[dest_stride] \n\t" | |
942 "lbu %[temp3], 0(%[dest_pix]) \n\t" | |
943 "addi %[temp1], %[temp1], 32 \n\t" | |
944 "sra %[temp1], %[temp1], 6 \n\t" | |
945 "add %[temp3], %[temp3], %[temp1] \n\t" | |
946 "lbux %[temp1], %[temp3](%[cm]) \n\t" | |
947 "sb %[temp1], 0(%[dest_pix]) \n\t" | |
948 "addu %[dest_pix], %[dest_pix], %[dest_stride] \n\t" | |
949 | |
950 : [temp0] "=&r" (temp0), [temp1] "=&r" (temp1), [temp2] "=&r" (temp2), | |
951 [temp3] "=&r" (temp3), [dest_pix] "+r" (dest_pix) | |
952 : [cm] "r" (cm), [dest_stride] "r" (dest_stride), | |
953 [step1_8] "r" (step1_8), [step1_9] "r" (step1_9), | |
954 [step1_10] "r" (step1_10), [step1_11] "r" (step1_11), | |
955 [step1_20] "r" (step1_20), [step1_21] "r" (step1_21), | |
956 [step1_22] "r" (step1_22), [step1_23] "r" (step1_23) | |
957 ); | |
958 | |
959 step3_12 = ROUND_POWER_OF_TWO((step1_11 - step1_20), 6); | |
960 step3_13 = ROUND_POWER_OF_TWO((step1_10 - step1_21), 6); | |
961 step3_14 = ROUND_POWER_OF_TWO((step1_9 - step1_22), 6); | |
962 step3_15 = ROUND_POWER_OF_TWO((step1_8 - step1_23), 6); | |
963 | |
964 __asm__ __volatile__ ( | |
965 "lbu %[temp2], 0(%[dest_pix1]) \n\t" | |
966 "add %[temp2], %[temp2], %[step3_15] \n\t" | |
967 "lbux %[temp0], %[temp2](%[cm]) \n\t" | |
968 "sb %[temp0], 0(%[dest_pix1]) \n\t" | |
969 "subu %[dest_pix1], %[dest_pix1], %[dest_stride] \n\t" | |
970 "lbu %[temp3], 0(%[dest_pix1]) \n\t" | |
971 "add %[temp3], %[temp3], %[step3_14] \n\t" | |
972 "lbux %[temp1], %[temp3](%[cm]) \n\t" | |
973 "sb %[temp1], 0(%[dest_pix1]) \n\t" | |
974 "subu %[dest_pix1], %[dest_pix1], %[dest_stride] \n\t" | |
975 | |
976 "lbu %[temp2], 0(%[dest_pix1]) \n\t" | |
977 "add %[temp2], %[temp2], %[step3_13] \n\t" | |
978 "lbux %[temp0], %[temp2](%[cm]) \n\t" | |
979 "sb %[temp0], 0(%[dest_pix1]) \n\t" | |
980 "subu %[dest_pix1], %[dest_pix1], %[dest_stride] \n\t" | |
981 "lbu %[temp3], 0(%[dest_pix1]) \n\t" | |
982 "add %[temp3], %[temp3], %[step3_12] \n\t" | |
983 "lbux %[temp1], %[temp3](%[cm]) \n\t" | |
984 "sb %[temp1], 0(%[dest_pix1]) \n\t" | |
985 "subu %[dest_pix1], %[dest_pix1], %[dest_stride] \n\t" | |
986 | |
987 : [temp0] "=&r" (temp0), [temp1] "=&r" (temp1), [temp2] "=&r" (temp2), | |
988 [temp3] "=&r" (temp3), [dest_pix1] "+r" (dest_pix1) | |
989 : [cm] "r" (cm), [dest_stride] "r" (dest_stride), | |
990 [step3_12] "r" (step3_12), [step3_13] "r" (step3_13), | |
991 [step3_14] "r" (step3_14), [step3_15] "r" (step3_15) | |
992 ); | |
993 | |
994 __asm__ __volatile__ ( | |
995 "lbu %[temp2], 0(%[dest_pix]) \n\t" | |
996 "add %[temp0], %[step1_12], %[step2_19] \n\t" | |
997 "addi %[temp0], %[temp0], 32 \n\t" | |
998 "sra %[temp0], %[temp0], 6 \n\t" | |
999 "add %[temp2], %[temp2], %[temp0] \n\t" | |
1000 "lbux %[temp0], %[temp2](%[cm]) \n\t" | |
1001 "add %[temp1], %[step1_13], %[step2_18] \n\t" | |
1002 "sb %[temp0], 0(%[dest_pix]) \n\t" | |
1003 "addu %[dest_pix], %[dest_pix], %[dest_stride] \n\t" | |
1004 "lbu %[temp3], 0(%[dest_pix]) \n\t" | |
1005 "addi %[temp1], %[temp1], 32 \n\t" | |
1006 "sra %[temp1], %[temp1], 6 \n\t" | |
1007 "add %[temp3], %[temp3], %[temp1] \n\t" | |
1008 "lbux %[temp1], %[temp3](%[cm]) \n\t" | |
1009 "sb %[temp1], 0(%[dest_pix]) \n\t" | |
1010 "addu %[dest_pix], %[dest_pix], %[dest_stride] \n\t" | |
1011 | |
1012 "lbu %[temp2], 0(%[dest_pix]) \n\t" | |
1013 "add %[temp0], %[step1_14], %[step2_17] \n\t" | |
1014 "addi %[temp0], %[temp0], 32 \n\t" | |
1015 "sra %[temp0], %[temp0], 6 \n\t" | |
1016 "add %[temp2], %[temp2], %[temp0] \n\t" | |
1017 "lbux %[temp0], %[temp2](%[cm]) \n\t" | |
1018 "add %[temp1], %[step1_15], %[step2_16] \n\t" | |
1019 "sb %[temp0], 0(%[dest_pix]) \n\t" | |
1020 "addu %[dest_pix], %[dest_pix], %[dest_stride] \n\t" | |
1021 "lbu %[temp3], 0(%[dest_pix]) \n\t" | |
1022 "addi %[temp1], %[temp1], 32 \n\t" | |
1023 "sra %[temp1], %[temp1], 6 \n\t" | |
1024 "add %[temp3], %[temp3], %[temp1] \n\t" | |
1025 "lbux %[temp1], %[temp3](%[cm]) \n\t" | |
1026 "sb %[temp1], 0(%[dest_pix]) \n\t" | |
1027 | |
1028 : [temp0] "=&r" (temp0), [temp1] "=&r" (temp1), [temp2] "=&r" (temp2), | |
1029 [temp3] "=&r" (temp3), [dest_pix] "+r" (dest_pix) | |
1030 : [cm] "r" (cm), [dest_stride] "r" (dest_stride), | |
1031 [step1_12] "r" (step1_12), [step1_13] "r" (step1_13), | |
1032 [step1_14] "r" (step1_14), [step1_15] "r" (step1_15), | |
1033 [step2_16] "r" (step2_16), [step2_17] "r" (step2_17), | |
1034 [step2_18] "r" (step2_18), [step2_19] "r" (step2_19) | |
1035 ); | |
1036 | |
1037 step3_12 = ROUND_POWER_OF_TWO((step1_15 - step2_16), 6); | |
1038 step3_13 = ROUND_POWER_OF_TWO((step1_14 - step2_17), 6); | |
1039 step3_14 = ROUND_POWER_OF_TWO((step1_13 - step2_18), 6); | |
1040 step3_15 = ROUND_POWER_OF_TWO((step1_12 - step2_19), 6); | |
1041 | |
1042 __asm__ __volatile__ ( | |
1043 "lbu %[temp2], 0(%[dest_pix1]) \n\t" | |
1044 "add %[temp2], %[temp2], %[step3_15] \n\t" | |
1045 "lbux %[temp0], %[temp2](%[cm]) \n\t" | |
1046 "sb %[temp0], 0(%[dest_pix1]) \n\t" | |
1047 "subu %[dest_pix1], %[dest_pix1], %[dest_stride] \n\t" | |
1048 "lbu %[temp3], 0(%[dest_pix1]) \n\t" | |
1049 "add %[temp3], %[temp3], %[step3_14] \n\t" | |
1050 "lbux %[temp1], %[temp3](%[cm]) \n\t" | |
1051 "sb %[temp1], 0(%[dest_pix1]) \n\t" | |
1052 "subu %[dest_pix1], %[dest_pix1], %[dest_stride] \n\t" | |
1053 | |
1054 "lbu %[temp2], 0(%[dest_pix1]) \n\t" | |
1055 "add %[temp2], %[temp2], %[step3_13] \n\t" | |
1056 "lbux %[temp0], %[temp2](%[cm]) \n\t" | |
1057 "sb %[temp0], 0(%[dest_pix1]) \n\t" | |
1058 "subu %[dest_pix1], %[dest_pix1], %[dest_stride] \n\t" | |
1059 "lbu %[temp3], 0(%[dest_pix1]) \n\t" | |
1060 "add %[temp3], %[temp3], %[step3_12] \n\t" | |
1061 "lbux %[temp1], %[temp3](%[cm]) \n\t" | |
1062 "sb %[temp1], 0(%[dest_pix1]) \n\t" | |
1063 | |
1064 : [temp0] "=&r" (temp0), [temp1] "=&r" (temp1), [temp2] "=&r" (temp2), | |
1065 [temp3] "=&r" (temp3), [dest_pix1] "+r" (dest_pix1) | |
1066 : [cm] "r" (cm), [dest_stride] "r" (dest_stride), | |
1067 [step3_12] "r" (step3_12), [step3_13] "r" (step3_13), | |
1068 [step3_14] "r" (step3_14), [step3_15] "r" (step3_15) | |
1069 ); | |
1070 | |
1071 input += 32; | |
1072 } | |
1073 } | |
1074 #endif // #if HAVE_DSPR2 | |
OLD | NEW |