2 * VC-1 and WMV3 - DSP functions Loongson MMI-optimized
4 * Copyright (c) 2016 Zhou Xiaoyong <zhouxiaoyong@loongson.cn>
6 * This file is part of FFmpeg.
8 * FFmpeg is free software; you can redistribute it and/or
9 * modify it under the terms of the GNU Lesser General Public
10 * License as published by the Free Software Foundation; either
11 * version 2.1 of the License, or (at your option) any later version.
13 * FFmpeg is distributed in the hope that it will be useful,
14 * but WITHOUT ANY WARRANTY; without even the implied warranty of
15 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
16 * Lesser General Public License for more details.
18 * You should have received a copy of the GNU Lesser General Public
19 * License along with FFmpeg; if not, write to the Free Software
20 * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
23 #include "libavutil/avassert.h"
24 #include "libavcodec/vc1dsp.h"
25 #include "constants.h"
26 #include "vc1dsp_mips.h"
27 #include "hpeldsp_mips.h"
28 #include "libavutil/mem_internal.h"
29 #include "libavutil/mips/mmiutils.h"
31 #define VC1_INV_TRANCS_8_TYPE1(o1, o2, r1, r2, r3, r4, c0) \
32 "li %[tmp0], "#r1" \n\t" \
33 "mtc1 %[tmp0], %[ftmp13] \n\t" \
34 "punpcklwd %[ftmp13], %[ftmp13], %[ftmp13] \n\t" \
35 "li %[tmp0], "#r2" \n\t" \
36 "mtc1 %[tmp0], %[ftmp14] \n\t" \
37 "punpcklwd %[ftmp14], %[ftmp14], %[ftmp14] \n\t" \
38 "pmaddhw %[ftmp1], %[ftmp5], %[ftmp13] \n\t" \
39 "pmaddhw %[ftmp2], %[ftmp7], %[ftmp14] \n\t" \
40 "paddw %[ftmp1], %[ftmp1], %[ftmp2] \n\t" \
41 "pmaddhw %[ftmp2], %[ftmp6], %[ftmp13] \n\t" \
42 "pmaddhw %[ftmp3], %[ftmp8], %[ftmp14] \n\t" \
43 "paddw %[ftmp2], %[ftmp2], %[ftmp3] \n\t" \
45 "li %[tmp0], "#r3" \n\t" \
46 "mtc1 %[tmp0], %[ftmp13] \n\t" \
47 "punpcklwd %[ftmp13], %[ftmp13], %[ftmp13] \n\t" \
48 "li %[tmp0], "#r4" \n\t" \
49 "mtc1 %[tmp0], %[ftmp14] \n\t" \
50 "punpcklwd %[ftmp14], %[ftmp14], %[ftmp14] \n\t" \
51 "pmaddhw %[ftmp3], %[ftmp9], %[ftmp13] \n\t" \
52 "pmaddhw %[ftmp4], %[ftmp11], %[ftmp14] \n\t" \
53 "paddw %[ftmp3], %[ftmp3], %[ftmp4] \n\t" \
54 "pmaddhw %[ftmp4], %[ftmp10], %[ftmp13] \n\t" \
55 "pmaddhw %[ftmp13], %[ftmp12], %[ftmp14] \n\t" \
56 "paddw %[ftmp4], %[ftmp4], %[ftmp13] \n\t" \
58 "paddw %[ftmp1], %[ftmp1], "#c0" \n\t" \
59 "paddw %[ftmp2], %[ftmp2], "#c0" \n\t" \
60 "paddw %[ftmp13], %[ftmp1], %[ftmp3] \n\t" \
61 "psubw %[ftmp14], %[ftmp1], %[ftmp3] \n\t" \
62 "paddw %[ftmp1], %[ftmp2], %[ftmp4] \n\t" \
63 "psubw %[ftmp3], %[ftmp2], %[ftmp4] \n\t" \
64 "psraw %[ftmp13], %[ftmp13], %[ftmp0] \n\t" \
65 "psraw %[ftmp1], %[ftmp1], %[ftmp0] \n\t" \
66 "psraw %[ftmp14], %[ftmp14], %[ftmp0] \n\t" \
67 "psraw %[ftmp3], %[ftmp3], %[ftmp0] \n\t" \
68 "punpcklhw %[ftmp2], %[ftmp13], %[ftmp1] \n\t" \
69 "punpckhhw %[ftmp4], %[ftmp13], %[ftmp1] \n\t" \
70 "punpcklhw "#o1", %[ftmp2], %[ftmp4] \n\t" \
71 "punpcklhw %[ftmp2], %[ftmp14], %[ftmp3] \n\t" \
72 "punpckhhw %[ftmp4], %[ftmp14], %[ftmp3] \n\t" \
73 "punpcklhw "#o2", %[ftmp2], %[ftmp4] \n\t"
75 #define VC1_INV_TRANCS_8_TYPE2(o1, o2, r1, r2, r3, r4, c0, c1) \
76 "li %[tmp0], "#r1" \n\t" \
77 "mtc1 %[tmp0], %[ftmp13] \n\t" \
78 "punpcklwd %[ftmp13], %[ftmp13], %[ftmp13] \n\t" \
79 "li %[tmp0], "#r2" \n\t" \
80 "mtc1 %[tmp0], %[ftmp14] \n\t" \
81 "punpcklwd %[ftmp14], %[ftmp14], %[ftmp14] \n\t" \
82 "pmaddhw %[ftmp1], %[ftmp5], %[ftmp13] \n\t" \
83 "pmaddhw %[ftmp2], %[ftmp7], %[ftmp14] \n\t" \
84 "paddw %[ftmp1], %[ftmp1], %[ftmp2] \n\t" \
85 "pmaddhw %[ftmp2], %[ftmp6], %[ftmp13] \n\t" \
86 "pmaddhw %[ftmp3], %[ftmp8], %[ftmp14] \n\t" \
87 "paddw %[ftmp2], %[ftmp2], %[ftmp3] \n\t" \
89 "li %[tmp0], "#r3" \n\t" \
90 "mtc1 %[tmp0], %[ftmp13] \n\t" \
91 "punpcklwd %[ftmp13], %[ftmp13], %[ftmp13] \n\t" \
92 "li %[tmp0], "#r4" \n\t" \
93 "mtc1 %[tmp0], %[ftmp14] \n\t" \
94 "punpcklwd %[ftmp14], %[ftmp14], %[ftmp14] \n\t" \
95 "pmaddhw %[ftmp3], %[ftmp9], %[ftmp13] \n\t" \
96 "pmaddhw %[ftmp4], %[ftmp11], %[ftmp14] \n\t" \
97 "paddw %[ftmp3], %[ftmp3], %[ftmp4] \n\t" \
98 "pmaddhw %[ftmp4], %[ftmp10], %[ftmp13] \n\t" \
99 "pmaddhw %[ftmp13], %[ftmp12], %[ftmp14] \n\t" \
100 "paddw %[ftmp4], %[ftmp4], %[ftmp13] \n\t" \
102 "paddw %[ftmp13], %[ftmp1], %[ftmp3] \n\t" \
103 "psubw %[ftmp14], %[ftmp1], %[ftmp3] \n\t" \
104 "paddw %[ftmp14], %[ftmp14], "#c1" \n\t" \
105 "paddw %[ftmp1], %[ftmp2], %[ftmp4] \n\t" \
106 "psubw %[ftmp3], %[ftmp2], %[ftmp4] \n\t" \
107 "paddw %[ftmp3], %[ftmp3], "#c1" \n\t" \
108 "paddw %[ftmp13], %[ftmp13], "#c0" \n\t" \
109 "paddw %[ftmp14], %[ftmp14], "#c0" \n\t" \
110 "paddw %[ftmp1], %[ftmp1], "#c0" \n\t" \
111 "paddw %[ftmp3], %[ftmp3], "#c0" \n\t" \
112 "psraw %[ftmp13], %[ftmp13], %[ftmp0] \n\t" \
113 "psraw %[ftmp1], %[ftmp1], %[ftmp0] \n\t" \
114 "psraw %[ftmp14], %[ftmp14], %[ftmp0] \n\t" \
115 "psraw %[ftmp3], %[ftmp3], %[ftmp0] \n\t" \
116 "punpcklhw %[ftmp2], %[ftmp13], %[ftmp1] \n\t" \
117 "punpckhhw %[ftmp4], %[ftmp13], %[ftmp1] \n\t" \
118 "punpcklhw "#o1", %[ftmp2], %[ftmp4] \n\t" \
119 "punpcklhw %[ftmp2], %[ftmp14], %[ftmp3] \n\t" \
120 "punpckhhw %[ftmp4], %[ftmp14], %[ftmp3] \n\t" \
121 "punpcklhw "#o2", %[ftmp2], %[ftmp4] \n\t"
123 /* Do inverse transform on 8x8 block */
124 void ff_vc1_inv_trans_8x8_dc_mmi(uint8_t *dest, ptrdiff_t linesize, int16_t *block)
131 dc = (3 * dc + 1) >> 1;
132 dc = (3 * dc + 16) >> 5;
135 "xor %[ftmp0], %[ftmp0], %[ftmp0] \n\t"
136 "pshufh %[dc], %[dc], %[ftmp0] \n\t"
137 "li %[count], 0x02 \n\t"
140 MMI_LDC1(%[ftmp1], %[dest], 0x00)
141 PTR_ADDU "%[addr0], %[dest], %[linesize] \n\t"
142 MMI_LDC1(%[ftmp2], %[addr0], 0x00)
143 PTR_ADDU "%[addr0], %[addr0], %[linesize] \n\t"
144 MMI_LDC1(%[ftmp3], %[addr0], 0x00)
145 PTR_ADDU "%[addr0], %[addr0], %[linesize] \n\t"
146 MMI_LDC1(%[ftmp4], %[addr0], 0x00)
148 "punpckhbh %[ftmp5], %[ftmp1], %[ftmp0] \n\t"
149 "punpcklbh %[ftmp1], %[ftmp1], %[ftmp0] \n\t"
150 "punpckhbh %[ftmp6], %[ftmp2], %[ftmp0] \n\t"
151 "punpcklbh %[ftmp2], %[ftmp2], %[ftmp0] \n\t"
152 "punpckhbh %[ftmp7], %[ftmp3], %[ftmp0] \n\t"
153 "punpcklbh %[ftmp3], %[ftmp3], %[ftmp0] \n\t"
154 "punpckhbh %[ftmp8], %[ftmp4], %[ftmp0] \n\t"
155 "punpcklbh %[ftmp4], %[ftmp4], %[ftmp0] \n\t"
157 "paddsh %[ftmp1], %[ftmp1], %[dc] \n\t"
158 "paddsh %[ftmp2], %[ftmp2], %[dc] \n\t"
159 "paddsh %[ftmp3], %[ftmp3], %[dc] \n\t"
160 "paddsh %[ftmp4], %[ftmp4], %[dc] \n\t"
161 "paddsh %[ftmp5], %[ftmp5], %[dc] \n\t"
162 "paddsh %[ftmp6], %[ftmp6], %[dc] \n\t"
163 "paddsh %[ftmp7], %[ftmp7], %[dc] \n\t"
164 "paddsh %[ftmp8], %[ftmp8], %[dc] \n\t"
166 "packushb %[ftmp1], %[ftmp1], %[ftmp5] \n\t"
167 "packushb %[ftmp2], %[ftmp2], %[ftmp6] \n\t"
168 "packushb %[ftmp3], %[ftmp3], %[ftmp7] \n\t"
169 "packushb %[ftmp4], %[ftmp4], %[ftmp8] \n\t"
171 MMI_SDC1(%[ftmp1], %[dest], 0x00)
172 PTR_ADDU "%[addr0], %[dest], %[linesize] \n\t"
173 MMI_SDC1(%[ftmp2], %[addr0], 0x00)
174 PTR_ADDU "%[addr0], %[addr0], %[linesize] \n\t"
175 MMI_SDC1(%[ftmp3], %[addr0], 0x00)
176 PTR_ADDU "%[addr0], %[addr0], %[linesize] \n\t"
177 MMI_SDC1(%[ftmp4], %[addr0], 0x00)
179 "addiu %[count], %[count], -0x01 \n\t"
180 PTR_ADDU "%[dest], %[addr0], %[linesize] \n\t"
181 "bnez %[count], 1b \n\t"
182 : [ftmp0]"=&f"(ftmp[0]), [ftmp1]"=&f"(ftmp[1]),
183 [ftmp2]"=&f"(ftmp[2]), [ftmp3]"=&f"(ftmp[3]),
184 [ftmp4]"=&f"(ftmp[4]), [ftmp5]"=&f"(ftmp[5]),
185 [ftmp6]"=&f"(ftmp[6]), [ftmp7]"=&f"(ftmp[7]),
186 [ftmp8]"=&f"(ftmp[8]),
187 [addr0]"=&r"(addr[0]),
188 [count]"=&r"(count), [dest]"+&r"(dest)
189 : [linesize]"r"((mips_reg)linesize),
195 #if _MIPS_SIM != _ABIO32
196 void ff_vc1_inv_trans_8x8_mmi(int16_t block[64])
198 DECLARE_ALIGNED(16, int16_t, temp[64]);
199 DECLARE_ALIGNED(8, const uint64_t, ff_pw_1_local) = {0x0000000100000001ULL};
200 DECLARE_ALIGNED(8, const uint64_t, ff_pw_4_local) = {0x0000000400000004ULL};
201 DECLARE_ALIGNED(8, const uint64_t, ff_pw_64_local)= {0x0000004000000040ULL};
206 /* 1st loop: start */
207 "li %[tmp0], 0x03 \n\t"
208 "mtc1 %[tmp0], %[ftmp0] \n\t"
211 MMI_LDC1(%[ftmp1], %[block], 0x00)
212 MMI_LDC1(%[ftmp11], %[block], 0x10)
213 MMI_LDC1(%[ftmp2], %[block], 0x20)
214 MMI_LDC1(%[ftmp12], %[block], 0x30)
215 MMI_LDC1(%[ftmp3], %[block], 0x40)
216 MMI_LDC1(%[ftmp13], %[block], 0x50)
217 MMI_LDC1(%[ftmp4], %[block], 0x60)
218 MMI_LDC1(%[ftmp14], %[block], 0x70)
219 "punpcklhw %[ftmp5], %[ftmp1], %[ftmp2] \n\t"
220 "punpckhhw %[ftmp6], %[ftmp1], %[ftmp2] \n\t"
221 "punpcklhw %[ftmp7], %[ftmp3], %[ftmp4] \n\t"
222 "punpckhhw %[ftmp8], %[ftmp3], %[ftmp4] \n\t"
224 "punpcklhw %[ftmp9], %[ftmp11], %[ftmp12] \n\t"
225 "punpckhhw %[ftmp10], %[ftmp11], %[ftmp12] \n\t"
226 "punpcklhw %[ftmp11], %[ftmp13], %[ftmp14] \n\t"
227 "punpckhhw %[ftmp12], %[ftmp13], %[ftmp14] \n\t"
229 /* ftmp15:dst03,dst02,dst01,dst00 ftmp22:dst73,dst72,dst71,dst70 */
230 VC1_INV_TRANCS_8_TYPE1(%[ftmp15], %[ftmp22], 0x0010000c, 0x0006000c,
231 0x000f0010, 0x00040009, %[ff_pw_4])
233 /* ftmp16:dst13,dst12,dst11,dst10 ftmp21:dst63,dst62,dst61,dst60 */
234 VC1_INV_TRANCS_8_TYPE1(%[ftmp16], %[ftmp21], 0x0006000c, 0xfff0fff4,
235 0xfffc000f, 0xfff7fff0, %[ff_pw_4])
237 /* ftmp17:dst23,dst22,dst21,dst20 ftmp20:dst53,dst52,dst51,dst50 */
238 VC1_INV_TRANCS_8_TYPE1(%[ftmp17], %[ftmp20], 0xfffa000c, 0x0010fff4,
239 0xfff00009, 0x000f0004, %[ff_pw_4])
241 /* ftmp18:dst33,dst32,dst31,dst30 ftmp19:dst43,dst42,dst41,dst40 */
242 VC1_INV_TRANCS_8_TYPE1(%[ftmp18], %[ftmp19], 0xfff0000c, 0xfffa000c,
243 0xfff70004, 0xfff0000f, %[ff_pw_4])
245 TRANSPOSE_4H(%[ftmp15], %[ftmp16], %[ftmp17], %[ftmp18],
246 %[ftmp1], %[ftmp2], %[ftmp3], %[ftmp4])
248 TRANSPOSE_4H(%[ftmp19], %[ftmp20], %[ftmp21], %[ftmp22],
249 %[ftmp1], %[ftmp2], %[ftmp3], %[ftmp4])
251 MMI_SDC1(%[ftmp15], %[temp], 0x00)
252 MMI_SDC1(%[ftmp19], %[temp], 0x08)
253 MMI_SDC1(%[ftmp16], %[temp], 0x10)
254 MMI_SDC1(%[ftmp20], %[temp], 0x18)
255 MMI_SDC1(%[ftmp17], %[temp], 0x20)
256 MMI_SDC1(%[ftmp21], %[temp], 0x28)
257 MMI_SDC1(%[ftmp18], %[temp], 0x30)
258 MMI_SDC1(%[ftmp22], %[temp], 0x38)
261 MMI_LDC1(%[ftmp1], %[block], 0x08)
262 MMI_LDC1(%[ftmp11], %[block], 0x18)
263 MMI_LDC1(%[ftmp2], %[block], 0x28)
264 MMI_LDC1(%[ftmp12], %[block], 0x38)
265 MMI_LDC1(%[ftmp3], %[block], 0x48)
266 MMI_LDC1(%[ftmp13], %[block], 0x58)
267 MMI_LDC1(%[ftmp4], %[block], 0x68)
268 MMI_LDC1(%[ftmp14], %[block], 0x78)
269 "punpcklhw %[ftmp5], %[ftmp1], %[ftmp2] \n\t"
270 "punpckhhw %[ftmp6], %[ftmp1], %[ftmp2] \n\t"
271 "punpcklhw %[ftmp7], %[ftmp3], %[ftmp4] \n\t"
272 "punpckhhw %[ftmp8], %[ftmp3], %[ftmp4] \n\t"
274 "punpcklhw %[ftmp9], %[ftmp11], %[ftmp12] \n\t"
275 "punpckhhw %[ftmp10], %[ftmp11], %[ftmp12] \n\t"
276 "punpcklhw %[ftmp11], %[ftmp13], %[ftmp14] \n\t"
277 "punpckhhw %[ftmp12], %[ftmp13], %[ftmp14] \n\t"
279 /* ftmp15:dst03,dst02,dst01,dst00 ftmp22:dst73,dst72,dst71,dst70 */
280 VC1_INV_TRANCS_8_TYPE1(%[ftmp15], %[ftmp22], 0x0010000c, 0x0006000c,
281 0x000f0010, 0x00040009, %[ff_pw_4])
283 /* ftmp16:dst13,dst12,dst11,dst10 ftmp21:dst63,dst62,dst61,dst60 */
284 VC1_INV_TRANCS_8_TYPE1(%[ftmp16], %[ftmp21], 0x0006000c, 0xfff0fff4,
285 0xfffc000f, 0xfff7fff0, %[ff_pw_4])
287 /* ftmp17:dst23,dst22,dst21,dst20 ftmp20:dst53,dst52,dst51,dst50 */
288 VC1_INV_TRANCS_8_TYPE1(%[ftmp17], %[ftmp20], 0xfffa000c, 0x0010fff4,
289 0xfff00009, 0x000f0004, %[ff_pw_4])
291 /* ftmp18:dst33,dst32,dst31,dst30 ftmp19:dst43,dst42,dst41,dst40 */
292 VC1_INV_TRANCS_8_TYPE1(%[ftmp18], %[ftmp19], 0xfff0000c, 0xfffa000c,
293 0xfff70004, 0xfff0000f, %[ff_pw_4])
295 TRANSPOSE_4H(%[ftmp15], %[ftmp16], %[ftmp17], %[ftmp18],
296 %[ftmp1], %[ftmp2], %[ftmp3], %[ftmp4])
298 TRANSPOSE_4H(%[ftmp19], %[ftmp20], %[ftmp21], %[ftmp22],
299 %[ftmp1], %[ftmp2], %[ftmp3], %[ftmp4])
301 MMI_SDC1(%[ftmp19], %[temp], 0x48)
302 MMI_SDC1(%[ftmp20], %[temp], 0x58)
303 MMI_SDC1(%[ftmp21], %[temp], 0x68)
304 MMI_SDC1(%[ftmp22], %[temp], 0x78)
307 /* 2nd loop: start */
308 "li %[tmp0], 0x07 \n\t"
309 "mtc1 %[tmp0], %[ftmp0] \n\t"
312 MMI_LDC1(%[ftmp1], %[temp], 0x00)
313 MMI_LDC1(%[ftmp11], %[temp], 0x10)
314 MMI_LDC1(%[ftmp2], %[temp], 0x20)
315 MMI_LDC1(%[ftmp12], %[temp], 0x30)
316 "punpcklhw %[ftmp5], %[ftmp1], %[ftmp2] \n\t"
317 "punpckhhw %[ftmp6], %[ftmp1], %[ftmp2] \n\t"
318 "punpcklhw %[ftmp7], %[ftmp15], %[ftmp17] \n\t"
319 "punpckhhw %[ftmp8], %[ftmp15], %[ftmp17] \n\t"
321 "punpcklhw %[ftmp9], %[ftmp11], %[ftmp12] \n\t"
322 "punpckhhw %[ftmp10], %[ftmp11], %[ftmp12] \n\t"
323 "punpcklhw %[ftmp11], %[ftmp16], %[ftmp18] \n\t"
324 "punpckhhw %[ftmp12], %[ftmp16], %[ftmp18] \n\t"
326 /* ftmp15:dst03,dst02,dst01,dst00 ftmp22:dst73,dst72,dst71,dst70 */
327 VC1_INV_TRANCS_8_TYPE2(%[ftmp15], %[ftmp22], 0x0010000c, 0x0006000c,
328 0x000f0010, 0x00040009, %[ff_pw_64], %[ff_pw_1])
330 /* ftmp16:dst13,dst12,dst11,dst10 ftmp21:dst63,dst62,dst61,dst60 */
331 VC1_INV_TRANCS_8_TYPE2(%[ftmp16], %[ftmp21], 0x0006000c, 0xfff0fff4,
332 0xfffc000f, 0xfff7fff0, %[ff_pw_64], %[ff_pw_1])
334 /* ftmp17:dst23,dst22,dst21,dst20 ftmp20:dst53,dst52,dst51,dst50 */
335 VC1_INV_TRANCS_8_TYPE2(%[ftmp17], %[ftmp20], 0xfffa000c, 0x0010fff4,
336 0xfff00009, 0x000f0004, %[ff_pw_64], %[ff_pw_1])
338 /* ftmp18:dst33,dst32,dst31,dst30 ftmp19:dst43,dst42,dst41,dst40 */
339 VC1_INV_TRANCS_8_TYPE2(%[ftmp18], %[ftmp19], 0xfff0000c, 0xfffa000c,
340 0xfff70004, 0xfff0000f, %[ff_pw_64], %[ff_pw_1])
342 MMI_SDC1(%[ftmp15], %[block], 0x00)
343 MMI_SDC1(%[ftmp16], %[block], 0x10)
344 MMI_SDC1(%[ftmp17], %[block], 0x20)
345 MMI_SDC1(%[ftmp18], %[block], 0x30)
346 MMI_SDC1(%[ftmp19], %[block], 0x40)
347 MMI_SDC1(%[ftmp20], %[block], 0x50)
348 MMI_SDC1(%[ftmp21], %[block], 0x60)
349 MMI_SDC1(%[ftmp22], %[block], 0x70)
352 MMI_LDC1(%[ftmp1], %[temp], 0x08)
353 MMI_LDC1(%[ftmp11], %[temp], 0x18)
354 MMI_LDC1(%[ftmp2], %[temp], 0x28)
355 MMI_LDC1(%[ftmp12], %[temp], 0x38)
356 MMI_LDC1(%[ftmp3], %[temp], 0x48)
357 MMI_LDC1(%[ftmp13], %[temp], 0x58)
358 MMI_LDC1(%[ftmp4], %[temp], 0x68)
359 MMI_LDC1(%[ftmp14], %[temp], 0x78)
360 "punpcklhw %[ftmp5], %[ftmp1], %[ftmp2] \n\t"
361 "punpckhhw %[ftmp6], %[ftmp1], %[ftmp2] \n\t"
362 "punpcklhw %[ftmp7], %[ftmp3], %[ftmp4] \n\t"
363 "punpckhhw %[ftmp8], %[ftmp3], %[ftmp4] \n\t"
365 "punpcklhw %[ftmp9], %[ftmp11], %[ftmp12] \n\t"
366 "punpckhhw %[ftmp10], %[ftmp11], %[ftmp12] \n\t"
367 "punpcklhw %[ftmp11], %[ftmp13], %[ftmp14] \n\t"
368 "punpckhhw %[ftmp12], %[ftmp13], %[ftmp14] \n\t"
370 /* ftmp15:dst03,dst02,dst01,dst00 ftmp22:dst73,dst72,dst71,dst70 */
371 VC1_INV_TRANCS_8_TYPE2(%[ftmp15], %[ftmp22], 0x0010000c, 0x0006000c,
372 0x000f0010, 0x00040009, %[ff_pw_64], %[ff_pw_1])
374 /* ftmp16:dst13,dst12,dst11,dst10 ftmp21:dst63,dst62,dst61,dst60 */
375 VC1_INV_TRANCS_8_TYPE2(%[ftmp16], %[ftmp21], 0x0006000c, 0xfff0fff4,
376 0xfffc000f, 0xfff7fff0, %[ff_pw_64], %[ff_pw_1])
378 /* ftmp17:dst23,dst22,dst21,dst20 ftmp20:dst53,dst52,dst51,dst50 */
379 VC1_INV_TRANCS_8_TYPE2(%[ftmp17], %[ftmp20], 0xfffa000c, 0x0010fff4,
380 0xfff00009, 0x000f0004, %[ff_pw_64], %[ff_pw_1])
382 /* ftmp18:dst33,dst32,dst31,dst30 ftmp19:dst43,dst42,dst41,dst40 */
383 VC1_INV_TRANCS_8_TYPE2(%[ftmp18], %[ftmp19], 0xfff0000c, 0xfffa000c,
384 0xfff70004, 0xfff0000f, %[ff_pw_64], %[ff_pw_1])
386 MMI_SDC1(%[ftmp15], %[block], 0x08)
387 MMI_SDC1(%[ftmp16], %[block], 0x18)
388 MMI_SDC1(%[ftmp17], %[block], 0x28)
389 MMI_SDC1(%[ftmp18], %[block], 0x38)
390 MMI_SDC1(%[ftmp19], %[block], 0x48)
391 MMI_SDC1(%[ftmp20], %[block], 0x58)
392 MMI_SDC1(%[ftmp21], %[block], 0x68)
393 MMI_SDC1(%[ftmp22], %[block], 0x78)
395 : [ftmp0]"=&f"(ftmp[0]), [ftmp1]"=&f"(ftmp[1]),
396 [ftmp2]"=&f"(ftmp[2]), [ftmp3]"=&f"(ftmp[3]),
397 [ftmp4]"=&f"(ftmp[4]), [ftmp5]"=&f"(ftmp[5]),
398 [ftmp6]"=&f"(ftmp[6]), [ftmp7]"=&f"(ftmp[7]),
399 [ftmp8]"=&f"(ftmp[8]), [ftmp9]"=&f"(ftmp[9]),
400 [ftmp10]"=&f"(ftmp[10]), [ftmp11]"=&f"(ftmp[11]),
401 [ftmp12]"=&f"(ftmp[12]), [ftmp13]"=&f"(ftmp[13]),
402 [ftmp14]"=&f"(ftmp[14]), [ftmp15]"=&f"(ftmp[15]),
403 [ftmp16]"=&f"(ftmp[16]), [ftmp17]"=&f"(ftmp[17]),
404 [ftmp18]"=&f"(ftmp[18]), [ftmp19]"=&f"(ftmp[19]),
405 [ftmp20]"=&f"(ftmp[20]), [ftmp21]"=&f"(ftmp[21]),
406 [ftmp22]"=&f"(ftmp[22]),
408 : [ff_pw_1]"f"(ff_pw_1_local), [ff_pw_64]"f"(ff_pw_64_local),
409 [ff_pw_4]"f"(ff_pw_4_local), [block]"r"(block),
416 /* Do inverse transform on 8x4 part of block */
417 void ff_vc1_inv_trans_8x4_dc_mmi(uint8_t *dest, ptrdiff_t linesize, int16_t *block)
422 dc = ( 3 * dc + 1) >> 1;
423 dc = (17 * dc + 64) >> 7;
426 "xor %[ftmp0], %[ftmp0], %[ftmp0] \n\t"
427 "pshufh %[dc], %[dc], %[ftmp0] \n\t"
429 MMI_LDC1(%[ftmp1], %[dest0], 0x00)
430 MMI_LDC1(%[ftmp2], %[dest1], 0x00)
431 MMI_LDC1(%[ftmp3], %[dest2], 0x00)
432 MMI_LDC1(%[ftmp4], %[dest3], 0x00)
434 "punpckhbh %[ftmp5], %[ftmp1], %[ftmp0] \n\t"
435 "punpcklbh %[ftmp1], %[ftmp1], %[ftmp0] \n\t"
436 "punpckhbh %[ftmp6], %[ftmp2], %[ftmp0] \n\t"
437 "punpcklbh %[ftmp2], %[ftmp2], %[ftmp0] \n\t"
438 "punpckhbh %[ftmp7], %[ftmp3], %[ftmp0] \n\t"
439 "punpcklbh %[ftmp3], %[ftmp3], %[ftmp0] \n\t"
440 "punpckhbh %[ftmp8], %[ftmp4], %[ftmp0] \n\t"
441 "punpcklbh %[ftmp4], %[ftmp4], %[ftmp0] \n\t"
443 "paddsh %[ftmp1], %[ftmp1], %[dc] \n\t"
444 "paddsh %[ftmp2], %[ftmp2], %[dc] \n\t"
445 "paddsh %[ftmp3], %[ftmp3], %[dc] \n\t"
446 "paddsh %[ftmp4], %[ftmp4], %[dc] \n\t"
447 "paddsh %[ftmp5], %[ftmp5], %[dc] \n\t"
448 "paddsh %[ftmp6], %[ftmp6], %[dc] \n\t"
449 "paddsh %[ftmp7], %[ftmp7], %[dc] \n\t"
450 "paddsh %[ftmp8], %[ftmp8], %[dc] \n\t"
452 "packushb %[ftmp1], %[ftmp1], %[ftmp5] \n\t"
453 "packushb %[ftmp2], %[ftmp2], %[ftmp6] \n\t"
454 "packushb %[ftmp3], %[ftmp3], %[ftmp7] \n\t"
455 "packushb %[ftmp4], %[ftmp4], %[ftmp8] \n\t"
457 MMI_SDC1(%[ftmp1], %[dest0], 0x00)
458 MMI_SDC1(%[ftmp2], %[dest1], 0x00)
459 MMI_SDC1(%[ftmp3], %[dest2], 0x00)
460 MMI_SDC1(%[ftmp4], %[dest3], 0x00)
461 : [ftmp0]"=&f"(ftmp[0]), [ftmp1]"=&f"(ftmp[1]),
462 [ftmp2]"=&f"(ftmp[2]), [ftmp3]"=&f"(ftmp[3]),
463 [ftmp4]"=&f"(ftmp[4]), [ftmp5]"=&f"(ftmp[5]),
464 [ftmp6]"=&f"(ftmp[6]), [ftmp7]"=&f"(ftmp[7]),
465 [ftmp8]"=&f"(ftmp[8])
466 : [dest0]"r"(dest+0*linesize), [dest1]"r"(dest+1*linesize),
467 [dest2]"r"(dest+2*linesize), [dest3]"r"(dest+3*linesize),
473 #if _MIPS_SIM != _ABIO32
474 void ff_vc1_inv_trans_8x4_mmi(uint8_t *dest, ptrdiff_t linesize, int16_t *block)
476 int16_t *src = block;
477 int16_t *dst = block;
481 DECLARE_ALIGNED(16, const uint64_t, ff_pw_4_local) = {0x0000000400000004ULL};
482 DECLARE_ALIGNED(16, const uint64_t, ff_pw_64_local)= {0x0000004000000040ULL};
483 int16_t coeff[64] = {12, 16, 16, 15, 12, 9, 6, 4,
484 12, 15, 6, -4, -12, -16, -16, -9,
485 12, 9, -6, -16, -12, 4, 16, 15,
486 12, 4, -16, -9, 12, 15, -6, -16,
487 12, -4, -16, 9, 12, -15, -6, 16,
488 12, -9, -6, 16, -12, -4, 16, -15,
489 12, -15, 6, 4, -12, 16, -16, 9,
490 12, -16, 16, -15, 12, -9, 6, -4};
494 "li %[tmp0], 0x03 \n\t"
495 "mtc1 %[tmp0], %[ftmp0] \n\t"
498 MMI_LDC1(%[ftmp1], %[src], 0x00)
499 MMI_LDC1(%[ftmp2], %[src], 0x08)
501 /* ftmp11: dst1,dst0 */
502 MMI_LDC1(%[ftmp3], %[coeff], 0x00)
503 MMI_LDC1(%[ftmp4], %[coeff], 0x08)
504 MMI_LDC1(%[ftmp5], %[coeff], 0x10)
505 MMI_LDC1(%[ftmp6], %[coeff], 0x18)
506 "pmaddhw %[ftmp7], %[ftmp1], %[ftmp3] \n\t"
507 "pmaddhw %[ftmp8], %[ftmp2], %[ftmp4] \n\t"
508 "paddw %[ftmp9], %[ftmp7], %[ftmp8] \n\t"
509 "pmaddhw %[ftmp7], %[ftmp1], %[ftmp5] \n\t"
510 "pmaddhw %[ftmp8], %[ftmp2], %[ftmp6] \n\t"
511 "paddw %[ftmp10], %[ftmp7], %[ftmp8] \n\t"
512 "punpcklwd %[ftmp7], %[ftmp9], %[ftmp10] \n\t"
513 "punpckhwd %[ftmp8], %[ftmp9], %[ftmp10] \n\t"
514 "paddw %[ftmp11], %[ftmp7], %[ftmp8] \n\t"
515 "paddw %[ftmp11], %[ftmp11], %[ff_pw_4] \n\t"
517 /* ftmp12: dst3,dst2 */
518 MMI_LDC1(%[ftmp3], %[coeff], 0x20)
519 MMI_LDC1(%[ftmp4], %[coeff], 0x28)
520 MMI_LDC1(%[ftmp5], %[coeff], 0x30)
521 MMI_LDC1(%[ftmp6], %[coeff], 0x38)
522 "pmaddhw %[ftmp7], %[ftmp1], %[ftmp3] \n\t"
523 "pmaddhw %[ftmp8], %[ftmp2], %[ftmp4] \n\t"
524 "paddw %[ftmp9], %[ftmp7], %[ftmp8] \n\t"
525 "pmaddhw %[ftmp7], %[ftmp1], %[ftmp5] \n\t"
526 "pmaddhw %[ftmp8], %[ftmp2], %[ftmp6] \n\t"
527 "paddw %[ftmp10], %[ftmp7], %[ftmp8] \n\t"
528 "punpcklwd %[ftmp7], %[ftmp9], %[ftmp10] \n\t"
529 "punpckhwd %[ftmp8], %[ftmp9], %[ftmp10] \n\t"
530 "paddw %[ftmp12], %[ftmp7], %[ftmp8] \n\t"
531 "paddw %[ftmp12], %[ftmp12], %[ff_pw_4] \n\t"
533 /* ftmp13: dst5,dst4 */
534 MMI_LDC1(%[ftmp3], %[coeff], 0x40)
535 MMI_LDC1(%[ftmp4], %[coeff], 0x48)
536 MMI_LDC1(%[ftmp5], %[coeff], 0x50)
537 MMI_LDC1(%[ftmp6], %[coeff], 0x58)
538 "pmaddhw %[ftmp7], %[ftmp1], %[ftmp3] \n\t"
539 "pmaddhw %[ftmp8], %[ftmp2], %[ftmp4] \n\t"
540 "paddw %[ftmp9], %[ftmp7], %[ftmp8] \n\t"
541 "pmaddhw %[ftmp7], %[ftmp1], %[ftmp5] \n\t"
542 "pmaddhw %[ftmp8], %[ftmp2], %[ftmp6] \n\t"
543 "paddw %[ftmp10], %[ftmp7], %[ftmp8] \n\t"
544 "punpcklwd %[ftmp7], %[ftmp9], %[ftmp10] \n\t"
545 "punpckhwd %[ftmp8], %[ftmp9], %[ftmp10] \n\t"
546 "paddw %[ftmp13], %[ftmp7], %[ftmp8] \n\t"
547 "paddw %[ftmp13], %[ftmp13], %[ff_pw_4] \n\t"
549 /* ftmp14: dst7,dst6 */
550 MMI_LDC1(%[ftmp3], %[coeff], 0x60)
551 MMI_LDC1(%[ftmp4], %[coeff], 0x68)
552 MMI_LDC1(%[ftmp5], %[coeff], 0x70)
553 MMI_LDC1(%[ftmp6], %[coeff], 0x78)
554 "pmaddhw %[ftmp7], %[ftmp1], %[ftmp3] \n\t"
555 "pmaddhw %[ftmp8], %[ftmp2], %[ftmp4] \n\t"
556 "paddw %[ftmp9], %[ftmp7], %[ftmp8] \n\t"
557 "pmaddhw %[ftmp7], %[ftmp1], %[ftmp5] \n\t"
558 "pmaddhw %[ftmp8], %[ftmp2], %[ftmp6] \n\t"
559 "paddw %[ftmp10], %[ftmp7], %[ftmp8] \n\t"
560 "punpcklwd %[ftmp7], %[ftmp9], %[ftmp10] \n\t"
561 "punpckhwd %[ftmp8], %[ftmp9], %[ftmp10] \n\t"
562 "paddw %[ftmp14], %[ftmp7], %[ftmp8] \n\t"
563 "paddw %[ftmp14], %[ftmp14], %[ff_pw_4] \n\t"
565 /* ftmp9: dst3,dst2,dst1,dst0 ftmp10: dst7,dst6,dst5,dst4 */
566 "psraw %[ftmp11], %[ftmp11], %[ftmp0] \n\t"
567 "psraw %[ftmp12], %[ftmp12], %[ftmp0] \n\t"
568 "psraw %[ftmp13], %[ftmp13], %[ftmp0] \n\t"
569 "psraw %[ftmp14], %[ftmp14], %[ftmp0] \n\t"
570 "punpcklhw %[ftmp7], %[ftmp11], %[ftmp12] \n\t"
571 "punpckhhw %[ftmp8], %[ftmp11], %[ftmp12] \n\t"
572 "punpcklhw %[ftmp9], %[ftmp7], %[ftmp8] \n\t"
573 "punpcklhw %[ftmp7], %[ftmp13], %[ftmp14] \n\t"
574 "punpckhhw %[ftmp8], %[ftmp13], %[ftmp14] \n\t"
575 "punpcklhw %[ftmp10], %[ftmp7], %[ftmp8] \n\t"
576 MMI_SDC1(%[ftmp9], %[dst], 0x00)
577 MMI_SDC1(%[ftmp10], %[dst], 0x08)
579 PTR_ADDIU "%[src], %[src], 0x10 \n\t"
580 PTR_ADDIU "%[dst], %[dst], 0x10 \n\t"
581 "addiu %[count], %[count], -0x01 \n\t"
582 "bnez %[count], 1b \n\t"
583 : [ftmp0]"=&f"(ftmp[0]), [ftmp1]"=&f"(ftmp[1]),
584 [ftmp2]"=&f"(ftmp[2]), [ftmp3]"=&f"(ftmp[3]),
585 [ftmp4]"=&f"(ftmp[4]), [ftmp5]"=&f"(ftmp[5]),
586 [ftmp6]"=&f"(ftmp[6]), [ftmp7]"=&f"(ftmp[7]),
587 [ftmp8]"=&f"(ftmp[8]), [ftmp9]"=&f"(ftmp[9]),
588 [ftmp10]"=&f"(ftmp[10]), [ftmp11]"=&f"(ftmp[11]),
589 [ftmp12]"=&f"(ftmp[12]), [ftmp13]"=&f"(ftmp[13]),
590 [ftmp14]"=&f"(ftmp[14]), [tmp0]"=&r"(tmp[0]),
591 [src]"+&r"(src), [dst]"+&r"(dst), [count]"+&r"(count)
592 : [ff_pw_4]"f"(ff_pw_4_local), [coeff]"r"(coeff)
600 "li %[tmp0], 0x44 \n\t"
601 "mtc1 %[tmp0], %[ftmp15] \n\t"
604 "li %[tmp0], 0x07 \n\t"
605 "mtc1 %[tmp0], %[ftmp0] \n\t"
606 MMI_LDC1(%[ftmp1], %[src], 0x00)
607 MMI_LDC1(%[ftmp2], %[src], 0x10)
608 MMI_LDC1(%[ftmp3], %[src], 0x20)
609 MMI_LDC1(%[ftmp4], %[src], 0x30)
610 "punpcklhw %[ftmp5], %[ftmp1], %[ftmp2] \n\t"
611 "punpckhhw %[ftmp6], %[ftmp1], %[ftmp2] \n\t"
612 "punpcklhw %[ftmp7], %[ftmp3], %[ftmp4] \n\t"
613 "punpckhhw %[ftmp8], %[ftmp3], %[ftmp4] \n\t"
615 /* ftmp11: dst03,dst02,dst01,dst00 */
616 "li %[tmp0], 0x00160011 \n\t"
617 "mtc1 %[tmp0], %[ftmp3] \n\t"
618 "pshufh %[ftmp3], %[ftmp3], %[ftmp15] \n\t"
619 "li %[tmp0], 0x000a0011 \n\t"
620 "mtc1 %[tmp0], %[ftmp4] \n\t"
621 "pshufh %[ftmp4], %[ftmp4], %[ftmp15] \n\t"
622 "pmaddhw %[ftmp1], %[ftmp5], %[ftmp3] \n\t"
623 "pmaddhw %[ftmp2], %[ftmp7], %[ftmp4] \n\t"
624 "paddw %[ftmp9], %[ftmp1], %[ftmp2] \n\t"
625 "pmaddhw %[ftmp1], %[ftmp6], %[ftmp3] \n\t"
626 "pmaddhw %[ftmp2], %[ftmp8], %[ftmp4] \n\t"
627 "paddw %[ftmp10], %[ftmp1], %[ftmp2] \n\t"
628 "paddw %[ftmp9], %[ftmp9], %[ff_pw_64] \n\t"
629 "paddw %[ftmp10], %[ftmp10], %[ff_pw_64] \n\t"
630 "psraw %[ftmp9], %[ftmp9], %[ftmp0] \n\t"
631 "psraw %[ftmp10], %[ftmp10], %[ftmp0] \n\t"
632 "punpcklhw %[ftmp1], %[ftmp9], %[ftmp10] \n\t"
633 "punpckhhw %[ftmp2], %[ftmp9], %[ftmp10] \n\t"
634 "punpcklhw %[ftmp11], %[ftmp1], %[ftmp2] \n\t"
636 /* ftmp12: dst13,dst12,dst11,dst10 */
637 "li %[tmp0], 0x000a0011 \n\t"
638 "mtc1 %[tmp0], %[ftmp3] \n\t"
639 "pshufh %[ftmp3], %[ftmp3], %[ftmp15] \n\t"
640 "li %[tmp0], 0xffeaffef \n\t"
641 "mtc1 %[tmp0], %[ftmp4] \n\t"
642 "pshufh %[ftmp4], %[ftmp4], %[ftmp15] \n\t"
643 "pmaddhw %[ftmp1], %[ftmp5], %[ftmp3] \n\t"
644 "pmaddhw %[ftmp2], %[ftmp7], %[ftmp4] \n\t"
645 "paddw %[ftmp9], %[ftmp1], %[ftmp2] \n\t"
646 "pmaddhw %[ftmp1], %[ftmp6], %[ftmp3] \n\t"
647 "pmaddhw %[ftmp2], %[ftmp8], %[ftmp4] \n\t"
648 "paddw %[ftmp10], %[ftmp1], %[ftmp2] \n\t"
649 "paddw %[ftmp9], %[ftmp9], %[ff_pw_64] \n\t"
650 "paddw %[ftmp10], %[ftmp10], %[ff_pw_64] \n\t"
651 "psraw %[ftmp9], %[ftmp9], %[ftmp0] \n\t"
652 "psraw %[ftmp10], %[ftmp10], %[ftmp0] \n\t"
653 "punpcklhw %[ftmp1], %[ftmp9], %[ftmp10] \n\t"
654 "punpckhhw %[ftmp2], %[ftmp9], %[ftmp10] \n\t"
655 "punpcklhw %[ftmp12], %[ftmp1], %[ftmp2] \n\t"
657 /* ftmp13: dst23,dst22,dst21,dst20 */
658 "li %[tmp0], 0xfff60011 \n\t"
659 "mtc1 %[tmp0], %[ftmp3] \n\t"
660 "pshufh %[ftmp3], %[ftmp3], %[ftmp15] \n\t"
661 "li %[tmp0], 0x0016ffef \n\t"
662 "mtc1 %[tmp0], %[ftmp4] \n\t"
663 "pshufh %[ftmp4], %[ftmp4], %[ftmp15] \n\t"
664 "pmaddhw %[ftmp1], %[ftmp5], %[ftmp3] \n\t"
665 "pmaddhw %[ftmp2], %[ftmp7], %[ftmp4] \n\t"
666 "paddw %[ftmp9], %[ftmp1], %[ftmp2] \n\t"
667 "pmaddhw %[ftmp1], %[ftmp6], %[ftmp3] \n\t"
668 "pmaddhw %[ftmp2], %[ftmp8], %[ftmp4] \n\t"
669 "paddw %[ftmp10], %[ftmp1], %[ftmp2] \n\t"
670 "paddw %[ftmp9], %[ftmp9], %[ff_pw_64] \n\t"
671 "paddw %[ftmp10], %[ftmp10], %[ff_pw_64] \n\t"
672 "psraw %[ftmp9], %[ftmp9], %[ftmp0] \n\t"
673 "psraw %[ftmp10], %[ftmp10], %[ftmp0] \n\t"
674 "punpcklhw %[ftmp1], %[ftmp9], %[ftmp10] \n\t"
675 "punpckhhw %[ftmp2], %[ftmp9], %[ftmp10] \n\t"
676 "punpcklhw %[ftmp13], %[ftmp1], %[ftmp2] \n\t"
678 /* ftmp14: dst33,dst32,dst31,dst30 */
679 "li %[tmp0], 0xffea0011 \n\t"
680 "mtc1 %[tmp0], %[ftmp3] \n\t"
681 "pshufh %[ftmp3], %[ftmp3], %[ftmp15] \n\t"
682 "li %[tmp0], 0xfff60011 \n\t"
683 "mtc1 %[tmp0], %[ftmp4] \n\t"
684 "pshufh %[ftmp4], %[ftmp4], %[ftmp15] \n\t"
685 "pmaddhw %[ftmp1], %[ftmp5], %[ftmp3] \n\t"
686 "pmaddhw %[ftmp2], %[ftmp7], %[ftmp4] \n\t"
687 "paddw %[ftmp9], %[ftmp1], %[ftmp2] \n\t"
688 "pmaddhw %[ftmp1], %[ftmp6], %[ftmp3] \n\t"
689 "pmaddhw %[ftmp2], %[ftmp8], %[ftmp4] \n\t"
690 "paddw %[ftmp10], %[ftmp1], %[ftmp2] \n\t"
691 "paddw %[ftmp9], %[ftmp9], %[ff_pw_64] \n\t"
692 "paddw %[ftmp10], %[ftmp10], %[ff_pw_64] \n\t"
693 "psraw %[ftmp9], %[ftmp9], %[ftmp0] \n\t"
694 "psraw %[ftmp10], %[ftmp10], %[ftmp0] \n\t"
695 "punpcklhw %[ftmp1], %[ftmp9], %[ftmp10] \n\t"
696 "punpckhhw %[ftmp2], %[ftmp9], %[ftmp10] \n\t"
697 "punpcklhw %[ftmp14], %[ftmp1], %[ftmp2] \n\t"
699 MMI_LWC1(%[ftmp1], %[dest], 0x00)
700 PTR_ADDU "%[tmp0], %[dest], %[linesize] \n\t"
701 MMI_LWC1(%[ftmp2], %[tmp0], 0x00)
702 PTR_ADDU "%[tmp0], %[tmp0], %[linesize] \n\t"
703 MMI_LWC1(%[ftmp3], %[tmp0], 0x00)
704 PTR_ADDU "%[tmp0], %[tmp0], %[linesize] \n\t"
705 MMI_LWC1(%[ftmp4], %[tmp0], 0x00)
706 "xor %[ftmp0], %[ftmp0], %[ftmp0] \n\t"
707 "punpcklbh %[ftmp1], %[ftmp1], %[ftmp0] \n\t"
708 "punpcklbh %[ftmp2], %[ftmp2], %[ftmp0] \n\t"
709 "punpcklbh %[ftmp3], %[ftmp3], %[ftmp0] \n\t"
710 "punpcklbh %[ftmp4], %[ftmp4], %[ftmp0] \n\t"
711 "paddh %[ftmp1], %[ftmp1], %[ftmp11] \n\t"
712 "paddh %[ftmp2], %[ftmp2], %[ftmp12] \n\t"
713 "paddh %[ftmp3], %[ftmp3], %[ftmp13] \n\t"
714 "paddh %[ftmp4], %[ftmp4], %[ftmp14] \n\t"
715 "packushb %[ftmp1], %[ftmp1], %[ftmp0] \n\t"
716 "packushb %[ftmp2], %[ftmp2], %[ftmp0] \n\t"
717 "packushb %[ftmp3], %[ftmp3], %[ftmp0] \n\t"
718 "packushb %[ftmp4], %[ftmp4], %[ftmp0] \n\t"
719 MMI_SWC1(%[ftmp1], %[dest], 0x00)
720 PTR_ADDU "%[tmp0], %[dest], %[linesize] \n\t"
721 MMI_SWC1(%[ftmp2], %[tmp0], 0x00)
722 PTR_ADDU "%[tmp0], %[tmp0], %[linesize] \n\t"
723 MMI_SWC1(%[ftmp3], %[tmp0], 0x00)
724 PTR_ADDU "%[tmp0], %[tmp0], %[linesize] \n\t"
725 MMI_SWC1(%[ftmp4], %[tmp0], 0x00)
728 "li %[tmp0], 0x07 \n\t"
729 "mtc1 %[tmp0], %[ftmp0] \n\t"
730 MMI_LDC1(%[ftmp1], %[src], 0x08)
731 MMI_LDC1(%[ftmp2], %[src], 0x18)
732 MMI_LDC1(%[ftmp3], %[src], 0x28)
733 MMI_LDC1(%[ftmp4], %[src], 0x38)
734 "punpcklhw %[ftmp5], %[ftmp1], %[ftmp2] \n\t"
735 "punpckhhw %[ftmp6], %[ftmp1], %[ftmp2] \n\t"
736 "punpcklhw %[ftmp7], %[ftmp3], %[ftmp4] \n\t"
737 "punpckhhw %[ftmp8], %[ftmp3], %[ftmp4] \n\t"
739 /* ftmp11: dst03,dst02,dst01,dst00 */
740 "li %[tmp0], 0x00160011 \n\t"
741 "mtc1 %[tmp0], %[ftmp3] \n\t"
742 "pshufh %[ftmp3], %[ftmp3], %[ftmp15] \n\t"
743 "li %[tmp0], 0x000a0011 \n\t"
744 "mtc1 %[tmp0], %[ftmp4] \n\t"
745 "pshufh %[ftmp4], %[ftmp4], %[ftmp15] \n\t"
746 "pmaddhw %[ftmp1], %[ftmp5], %[ftmp3] \n\t"
747 "pmaddhw %[ftmp2], %[ftmp7], %[ftmp4] \n\t"
748 "paddw %[ftmp9], %[ftmp1], %[ftmp2] \n\t"
749 "pmaddhw %[ftmp1], %[ftmp6], %[ftmp3] \n\t"
750 "pmaddhw %[ftmp2], %[ftmp8], %[ftmp4] \n\t"
751 "paddw %[ftmp10], %[ftmp1], %[ftmp2] \n\t"
752 "paddw %[ftmp9], %[ftmp9], %[ff_pw_64] \n\t"
753 "paddw %[ftmp10], %[ftmp10], %[ff_pw_64] \n\t"
754 "psraw %[ftmp9], %[ftmp9], %[ftmp0] \n\t"
755 "psraw %[ftmp10], %[ftmp10], %[ftmp0] \n\t"
756 "punpcklhw %[ftmp1], %[ftmp9], %[ftmp10] \n\t"
757 "punpckhhw %[ftmp2], %[ftmp9], %[ftmp10] \n\t"
758 "punpcklhw %[ftmp11], %[ftmp1], %[ftmp2] \n\t"
760 /* ftmp12: dst13,dst12,dst11,dst10 */
761 "li %[tmp0], 0x000a0011 \n\t"
762 "mtc1 %[tmp0], %[ftmp3] \n\t"
763 "pshufh %[ftmp3], %[ftmp3], %[ftmp15] \n\t"
764 "li %[tmp0], 0xffeaffef \n\t"
765 "mtc1 %[tmp0], %[ftmp4] \n\t"
766 "pshufh %[ftmp4], %[ftmp4], %[ftmp15] \n\t"
767 "pmaddhw %[ftmp1], %[ftmp5], %[ftmp3] \n\t"
768 "pmaddhw %[ftmp2], %[ftmp7], %[ftmp4] \n\t"
769 "paddw %[ftmp9], %[ftmp1], %[ftmp2] \n\t"
770 "pmaddhw %[ftmp1], %[ftmp6], %[ftmp3] \n\t"
771 "pmaddhw %[ftmp2], %[ftmp8], %[ftmp4] \n\t"
772 "paddw %[ftmp10], %[ftmp1], %[ftmp2] \n\t"
773 "paddw %[ftmp9], %[ftmp9], %[ff_pw_64] \n\t"
774 "paddw %[ftmp10], %[ftmp10], %[ff_pw_64] \n\t"
775 "psraw %[ftmp9], %[ftmp9], %[ftmp0] \n\t"
776 "psraw %[ftmp10], %[ftmp10], %[ftmp0] \n\t"
777 "punpcklhw %[ftmp1], %[ftmp9], %[ftmp10] \n\t"
778 "punpckhhw %[ftmp2], %[ftmp9], %[ftmp10] \n\t"
779 "punpcklhw %[ftmp12], %[ftmp1], %[ftmp2] \n\t"
781 /* ftmp13: dst23,dst22,dst21,dst20 */
782 "li %[tmp0], 0xfff60011 \n\t"
783 "mtc1 %[tmp0], %[ftmp3] \n\t"
784 "pshufh %[ftmp3], %[ftmp3], %[ftmp15] \n\t"
785 "li %[tmp0], 0x0016ffef \n\t"
786 "mtc1 %[tmp0], %[ftmp4] \n\t"
787 "pshufh %[ftmp4], %[ftmp4], %[ftmp15] \n\t"
788 "pmaddhw %[ftmp1], %[ftmp5], %[ftmp3] \n\t"
789 "pmaddhw %[ftmp2], %[ftmp7], %[ftmp4] \n\t"
790 "paddw %[ftmp9], %[ftmp1], %[ftmp2] \n\t"
791 "pmaddhw %[ftmp1], %[ftmp6], %[ftmp3] \n\t"
792 "pmaddhw %[ftmp2], %[ftmp8], %[ftmp4] \n\t"
793 "paddw %[ftmp10], %[ftmp1], %[ftmp2] \n\t"
794 "paddw %[ftmp9], %[ftmp9], %[ff_pw_64] \n\t"
795 "paddw %[ftmp10], %[ftmp10], %[ff_pw_64] \n\t"
796 "psraw %[ftmp9], %[ftmp9], %[ftmp0] \n\t"
797 "psraw %[ftmp10], %[ftmp10], %[ftmp0] \n\t"
798 "punpcklhw %[ftmp1], %[ftmp9], %[ftmp10] \n\t"
799 "punpckhhw %[ftmp2], %[ftmp9], %[ftmp10] \n\t"
800 "punpcklhw %[ftmp13], %[ftmp1], %[ftmp2] \n\t"
802 /* ftmp14: dst33,dst32,dst31,dst30 */
803 "li %[tmp0], 0xffea0011 \n\t"
804 "mtc1 %[tmp0], %[ftmp3] \n\t"
805 "pshufh %[ftmp3], %[ftmp3], %[ftmp15] \n\t"
806 "li %[tmp0], 0xfff60011 \n\t"
807 "mtc1 %[tmp0], %[ftmp4] \n\t"
808 "pshufh %[ftmp4], %[ftmp4], %[ftmp15] \n\t"
809 "pmaddhw %[ftmp1], %[ftmp5], %[ftmp3] \n\t"
810 "pmaddhw %[ftmp2], %[ftmp7], %[ftmp4] \n\t"
811 "paddw %[ftmp9], %[ftmp1], %[ftmp2] \n\t"
812 "pmaddhw %[ftmp1], %[ftmp6], %[ftmp3] \n\t"
813 "pmaddhw %[ftmp2], %[ftmp8], %[ftmp4] \n\t"
814 "paddw %[ftmp10], %[ftmp1], %[ftmp2] \n\t"
815 "paddw %[ftmp9], %[ftmp9], %[ff_pw_64] \n\t"
816 "paddw %[ftmp10], %[ftmp10], %[ff_pw_64] \n\t"
817 "psraw %[ftmp9], %[ftmp9], %[ftmp0] \n\t"
818 "psraw %[ftmp10], %[ftmp10], %[ftmp0] \n\t"
819 "punpcklhw %[ftmp1], %[ftmp9], %[ftmp10] \n\t"
820 "punpckhhw %[ftmp2], %[ftmp9], %[ftmp10] \n\t"
821 "punpcklhw %[ftmp14], %[ftmp1], %[ftmp2] \n\t"
823 MMI_LWC1(%[ftmp1], %[dest], 0x04)
824 PTR_ADDU "%[tmp0], %[dest], %[linesize] \n\t"
825 MMI_LWC1(%[ftmp2], %[tmp0], 0x04)
826 PTR_ADDU "%[tmp0], %[tmp0], %[linesize] \n\t"
827 MMI_LWC1(%[ftmp3], %[tmp0], 0x04)
828 PTR_ADDU "%[tmp0], %[tmp0], %[linesize] \n\t"
829 MMI_LWC1(%[ftmp4], %[tmp0], 0x04)
830 "xor %[ftmp0], %[ftmp0], %[ftmp0] \n\t"
831 "punpcklbh %[ftmp1], %[ftmp1], %[ftmp0] \n\t"
832 "punpcklbh %[ftmp2], %[ftmp2], %[ftmp0] \n\t"
833 "punpcklbh %[ftmp3], %[ftmp3], %[ftmp0] \n\t"
834 "punpcklbh %[ftmp4], %[ftmp4], %[ftmp0] \n\t"
835 "paddh %[ftmp1], %[ftmp1], %[ftmp11] \n\t"
836 "paddh %[ftmp2], %[ftmp2], %[ftmp12] \n\t"
837 "paddh %[ftmp3], %[ftmp3], %[ftmp13] \n\t"
838 "paddh %[ftmp4], %[ftmp4], %[ftmp14] \n\t"
839 "packushb %[ftmp1], %[ftmp1], %[ftmp0] \n\t"
840 "packushb %[ftmp2], %[ftmp2], %[ftmp0] \n\t"
841 "packushb %[ftmp3], %[ftmp3], %[ftmp0] \n\t"
842 "packushb %[ftmp4], %[ftmp4], %[ftmp0] \n\t"
843 MMI_SWC1(%[ftmp1], %[dest], 0x04)
844 PTR_ADDU "%[tmp0], %[dest], %[linesize] \n\t"
845 MMI_SWC1(%[ftmp2], %[tmp0], 0x04)
846 PTR_ADDU "%[tmp0], %[tmp0], %[linesize] \n\t"
847 MMI_SWC1(%[ftmp3], %[tmp0], 0x04)
848 PTR_ADDU "%[tmp0], %[tmp0], %[linesize] \n\t"
849 MMI_SWC1(%[ftmp4], %[tmp0], 0x04)
851 : [ftmp0]"=&f"(ftmp[0]), [ftmp1]"=&f"(ftmp[1]),
852 [ftmp2]"=&f"(ftmp[2]), [ftmp3]"=&f"(ftmp[3]),
853 [ftmp4]"=&f"(ftmp[4]), [ftmp5]"=&f"(ftmp[5]),
854 [ftmp6]"=&f"(ftmp[6]), [ftmp7]"=&f"(ftmp[7]),
855 [ftmp8]"=&f"(ftmp[8]), [ftmp9]"=&f"(ftmp[9]),
856 [ftmp10]"=&f"(ftmp[10]), [ftmp11]"=&f"(ftmp[11]),
857 [ftmp12]"=&f"(ftmp[12]), [ftmp13]"=&f"(ftmp[13]),
858 [ftmp14]"=&f"(ftmp[14]), [ftmp15]"=&f"(ftmp[15]),
860 : [ff_pw_64]"f"(ff_pw_64_local),
861 [src]"r"(src), [dest]"r"(dest), [linesize]"r"(linesize)
867 /* Do inverse transform on 4x8 parts of block */
868 void ff_vc1_inv_trans_4x8_dc_mmi(uint8_t *dest, ptrdiff_t linesize, int16_t *block)
874 dc = (17 * dc + 4) >> 3;
875 dc = (12 * dc + 64) >> 7;
878 "xor %[ftmp0], %[ftmp0], %[ftmp0] \n\t"
879 "pshufh %[dc], %[dc], %[ftmp0] \n\t"
881 MMI_LWC1(%[ftmp1], %[dest0], 0x00)
882 MMI_LWC1(%[ftmp2], %[dest1], 0x00)
883 MMI_LWC1(%[ftmp3], %[dest2], 0x00)
884 MMI_LWC1(%[ftmp4], %[dest3], 0x00)
885 MMI_LWC1(%[ftmp5], %[dest4], 0x00)
886 MMI_LWC1(%[ftmp6], %[dest5], 0x00)
887 MMI_LWC1(%[ftmp7], %[dest6], 0x00)
888 MMI_LWC1(%[ftmp8], %[dest7], 0x00)
890 "punpcklbh %[ftmp1], %[ftmp1], %[ftmp0] \n\t"
891 "punpcklbh %[ftmp2], %[ftmp2], %[ftmp0] \n\t"
892 "punpcklbh %[ftmp3], %[ftmp3], %[ftmp0] \n\t"
893 "punpcklbh %[ftmp4], %[ftmp4], %[ftmp0] \n\t"
894 "punpcklbh %[ftmp5], %[ftmp5], %[ftmp0] \n\t"
895 "punpcklbh %[ftmp6], %[ftmp6], %[ftmp0] \n\t"
896 "punpcklbh %[ftmp7], %[ftmp7], %[ftmp0] \n\t"
897 "punpcklbh %[ftmp8], %[ftmp8], %[ftmp0] \n\t"
899 "paddsh %[ftmp1], %[ftmp1], %[dc] \n\t"
900 "paddsh %[ftmp2], %[ftmp2], %[dc] \n\t"
901 "paddsh %[ftmp3], %[ftmp3], %[dc] \n\t"
902 "paddsh %[ftmp4], %[ftmp4], %[dc] \n\t"
903 "paddsh %[ftmp5], %[ftmp5], %[dc] \n\t"
904 "paddsh %[ftmp6], %[ftmp6], %[dc] \n\t"
905 "paddsh %[ftmp7], %[ftmp7], %[dc] \n\t"
906 "paddsh %[ftmp8], %[ftmp8], %[dc] \n\t"
908 "packushb %[ftmp1], %[ftmp1], %[ftmp0] \n\t"
909 "packushb %[ftmp2], %[ftmp2], %[ftmp0] \n\t"
910 "packushb %[ftmp3], %[ftmp3], %[ftmp0] \n\t"
911 "packushb %[ftmp4], %[ftmp4], %[ftmp0] \n\t"
912 "packushb %[ftmp5], %[ftmp5], %[ftmp0] \n\t"
913 "packushb %[ftmp6], %[ftmp6], %[ftmp0] \n\t"
914 "packushb %[ftmp7], %[ftmp7], %[ftmp0] \n\t"
915 "packushb %[ftmp8], %[ftmp8], %[ftmp0] \n\t"
917 MMI_SWC1(%[ftmp1], %[dest0], 0x00)
918 MMI_SWC1(%[ftmp2], %[dest1], 0x00)
919 MMI_SWC1(%[ftmp3], %[dest2], 0x00)
920 MMI_SWC1(%[ftmp4], %[dest3], 0x00)
921 MMI_SWC1(%[ftmp5], %[dest4], 0x00)
922 MMI_SWC1(%[ftmp6], %[dest5], 0x00)
923 MMI_SWC1(%[ftmp7], %[dest6], 0x00)
924 MMI_SWC1(%[ftmp8], %[dest7], 0x00)
925 : [ftmp0]"=&f"(ftmp[0]), [ftmp1]"=&f"(ftmp[1]),
926 [ftmp2]"=&f"(ftmp[2]), [ftmp3]"=&f"(ftmp[3]),
927 [ftmp4]"=&f"(ftmp[4]), [ftmp5]"=&f"(ftmp[5]),
928 [ftmp6]"=&f"(ftmp[6]), [ftmp7]"=&f"(ftmp[7]),
930 [ftmp8]"=&f"(ftmp[8])
931 : [dest0]"r"(dest+0*linesize), [dest1]"r"(dest+1*linesize),
932 [dest2]"r"(dest+2*linesize), [dest3]"r"(dest+3*linesize),
933 [dest4]"r"(dest+4*linesize), [dest5]"r"(dest+5*linesize),
934 [dest6]"r"(dest+6*linesize), [dest7]"r"(dest+7*linesize),
940 #if _MIPS_SIM != _ABIO32
941 void ff_vc1_inv_trans_4x8_mmi(uint8_t *dest, ptrdiff_t linesize, int16_t *block)
943 int16_t *src = block;
944 int16_t *dst = block;
946 uint32_t count = 8, tmp[1];
947 int16_t coeff[16] = {17, 22, 17, 10,
951 DECLARE_ALIGNED(8, const uint64_t, ff_pw_1_local) = {0x0000000100000001ULL};
952 DECLARE_ALIGNED(8, const uint64_t, ff_pw_4_local) = {0x0000000400000004ULL};
953 DECLARE_ALIGNED(8, const uint64_t, ff_pw_64_local)= {0x0000004000000040ULL};
958 "li %[tmp0], 0x03 \n\t"
959 "mtc1 %[tmp0], %[ftmp0] \n\t"
961 MMI_LDC1(%[ftmp2], %[coeff], 0x00)
962 MMI_LDC1(%[ftmp3], %[coeff], 0x08)
963 MMI_LDC1(%[ftmp4], %[coeff], 0x10)
964 MMI_LDC1(%[ftmp5], %[coeff], 0x18)
966 /* ftmp8: dst3,dst2,dst1,dst0 */
967 MMI_LDC1(%[ftmp1], %[src], 0x00)
968 "pmaddhw %[ftmp6], %[ftmp2], %[ftmp1] \n\t"
969 "pmaddhw %[ftmp7], %[ftmp3], %[ftmp1] \n\t"
970 "pmaddhw %[ftmp8], %[ftmp4], %[ftmp1] \n\t"
971 "pmaddhw %[ftmp9], %[ftmp5], %[ftmp1] \n\t"
972 "punpcklwd %[ftmp10], %[ftmp6], %[ftmp7] \n\t"
973 "punpckhwd %[ftmp11], %[ftmp6], %[ftmp7] \n\t"
974 "punpcklwd %[ftmp6], %[ftmp8], %[ftmp9] \n\t"
975 "punpckhwd %[ftmp7], %[ftmp8], %[ftmp9] \n\t"
976 "paddw %[ftmp8], %[ftmp10], %[ftmp11] \n\t"
977 "paddw %[ftmp9], %[ftmp6], %[ftmp7] \n\t"
978 "paddw %[ftmp8], %[ftmp8], %[ff_pw_4] \n\t"
979 "paddw %[ftmp9], %[ftmp9], %[ff_pw_4] \n\t"
980 "psraw %[ftmp8], %[ftmp8], %[ftmp0] \n\t"
981 "psraw %[ftmp9], %[ftmp9], %[ftmp0] \n\t"
982 "punpcklhw %[ftmp6], %[ftmp8], %[ftmp9] \n\t"
983 "punpckhhw %[ftmp7], %[ftmp8], %[ftmp9] \n\t"
984 "punpcklhw %[ftmp8], %[ftmp6], %[ftmp7] \n\t"
985 MMI_SDC1(%[ftmp8], %[dst], 0x00)
987 PTR_ADDIU "%[src], %[src], 0x10 \n\t"
988 PTR_ADDIU "%[dst], %[dst], 0x10 \n\t"
989 "addiu %[count], %[count], -0x01 \n\t"
990 "bnez %[count], 1b \n\t"
991 : [ftmp0]"=&f"(ftmp[0]), [ftmp1]"=&f"(ftmp[1]),
992 [ftmp2]"=&f"(ftmp[2]), [ftmp3]"=&f"(ftmp[3]),
993 [ftmp4]"=&f"(ftmp[4]), [ftmp5]"=&f"(ftmp[5]),
994 [ftmp6]"=&f"(ftmp[6]), [ftmp7]"=&f"(ftmp[7]),
995 [ftmp8]"=&f"(ftmp[8]), [ftmp9]"=&f"(ftmp[9]),
996 [ftmp10]"=&f"(ftmp[10]), [ftmp11]"=&f"(ftmp[11]),
997 [tmp0]"=&r"(tmp[0]), [count]"+&r"(count),
998 [src]"+&r"(src), [dst]"+&r"(dst)
999 : [ff_pw_4]"f"(ff_pw_4_local), [coeff]"r"(coeff)
1007 "li %[tmp0], 0x07 \n\t"
1008 "mtc1 %[tmp0], %[ftmp0] \n\t"
1010 MMI_LDC1(%[ftmp1], %[src], 0x00)
1011 MMI_LDC1(%[ftmp2], %[src], 0x20)
1012 MMI_LDC1(%[ftmp3], %[src], 0x40)
1013 MMI_LDC1(%[ftmp4], %[src], 0x60)
1014 "punpcklhw %[ftmp5], %[ftmp1], %[ftmp2] \n\t"
1015 "punpckhhw %[ftmp6], %[ftmp1], %[ftmp2] \n\t"
1016 "punpcklhw %[ftmp7], %[ftmp3], %[ftmp4] \n\t"
1017 "punpckhhw %[ftmp8], %[ftmp3], %[ftmp4] \n\t"
1019 MMI_LDC1(%[ftmp1], %[src], 0x10)
1020 MMI_LDC1(%[ftmp2], %[src], 0x30)
1021 MMI_LDC1(%[ftmp3], %[src], 0x50)
1022 MMI_LDC1(%[ftmp4], %[src], 0x70)
1023 "punpcklhw %[ftmp9], %[ftmp1], %[ftmp2] \n\t"
1024 "punpckhhw %[ftmp10], %[ftmp1], %[ftmp2] \n\t"
1025 "punpcklhw %[ftmp11], %[ftmp3], %[ftmp4] \n\t"
1026 "punpckhhw %[ftmp12], %[ftmp3], %[ftmp4] \n\t"
1028 /* ftmp15:dst03,dst02,dst01,dst00 ftmp22:dst73,dst72,dst71,dst70 */
1029 VC1_INV_TRANCS_8_TYPE2(%[ftmp15], %[ftmp22], 0x0010000c, 0x0006000c,
1030 0x000f0010, 0x00040009, %[ff_pw_64], %[ff_pw_1])
1032 /* ftmp16:dst13,dst12,dst11,dst10 ftmp21:dst63,dst62,dst61,dst60 */
1033 VC1_INV_TRANCS_8_TYPE2(%[ftmp16], %[ftmp21], 0x0006000c, 0xfff0fff4,
1034 0xfffc000f, 0xfff7fff0, %[ff_pw_64], %[ff_pw_1])
1036 /* ftmp17:dst23,dst22,dst21,dst20 ftmp20:dst53,dst52,dst51,dst50 */
1037 VC1_INV_TRANCS_8_TYPE2(%[ftmp17], %[ftmp20], 0xfffa000c, 0x0010fff4,
1038 0xfff00009, 0x000f0004, %[ff_pw_64], %[ff_pw_1])
1040 /* ftmp18:dst33,dst32,dst31,dst30 ftmp19:dst43,dst42,dst41,dst40 */
1041 VC1_INV_TRANCS_8_TYPE2(%[ftmp18], %[ftmp19], 0xfff0000c, 0xfffa000c,
1042 0xfff70004, 0xfff0000f, %[ff_pw_64], %[ff_pw_1])
1044 MMI_LWC1(%[ftmp1], %[dest], 0x00)
1045 PTR_ADDU "%[tmp0], %[dest], %[linesize] \n\t"
1046 MMI_LWC1(%[ftmp2], %[tmp0], 0x00)
1047 PTR_ADDU "%[tmp0], %[tmp0], %[linesize] \n\t"
1048 MMI_LWC1(%[ftmp3], %[tmp0], 0x00)
1049 PTR_ADDU "%[tmp0], %[tmp0], %[linesize] \n\t"
1050 MMI_LWC1(%[ftmp4], %[tmp0], 0x00)
1051 PTR_ADDU "%[tmp0], %[tmp0], %[linesize] \n\t"
1052 MMI_LWC1(%[ftmp5], %[tmp0], 0x00)
1053 PTR_ADDU "%[tmp0], %[tmp0], %[linesize] \n\t"
1054 MMI_LWC1(%[ftmp6], %[tmp0], 0x00)
1055 PTR_ADDU "%[tmp0], %[tmp0], %[linesize] \n\t"
1056 MMI_LWC1(%[ftmp7], %[tmp0], 0x00)
1057 PTR_ADDU "%[tmp0], %[tmp0], %[linesize] \n\t"
1058 MMI_LWC1(%[ftmp8], %[tmp0], 0x00)
1059 "xor %[ftmp0], %[ftmp0], %[ftmp0] \n\t"
1060 "punpcklbh %[ftmp1], %[ftmp1], %[ftmp0] \n\t"
1061 "punpcklbh %[ftmp2], %[ftmp2], %[ftmp0] \n\t"
1062 "punpcklbh %[ftmp3], %[ftmp3], %[ftmp0] \n\t"
1063 "punpcklbh %[ftmp4], %[ftmp4], %[ftmp0] \n\t"
1064 "punpcklbh %[ftmp5], %[ftmp5], %[ftmp0] \n\t"
1065 "punpcklbh %[ftmp6], %[ftmp6], %[ftmp0] \n\t"
1066 "punpcklbh %[ftmp7], %[ftmp7], %[ftmp0] \n\t"
1067 "punpcklbh %[ftmp8], %[ftmp8], %[ftmp0] \n\t"
1069 "paddh %[ftmp1], %[ftmp1], %[ftmp15] \n\t"
1070 "paddh %[ftmp2], %[ftmp2], %[ftmp16] \n\t"
1071 "paddh %[ftmp3], %[ftmp3], %[ftmp17] \n\t"
1072 "paddh %[ftmp4], %[ftmp4], %[ftmp18] \n\t"
1073 "paddh %[ftmp5], %[ftmp5], %[ftmp19] \n\t"
1074 "paddh %[ftmp6], %[ftmp6], %[ftmp20] \n\t"
1075 "paddh %[ftmp7], %[ftmp7], %[ftmp21] \n\t"
1076 "paddh %[ftmp8], %[ftmp8], %[ftmp22] \n\t"
1078 "packushb %[ftmp1], %[ftmp1], %[ftmp0] \n\t"
1079 "packushb %[ftmp2], %[ftmp2], %[ftmp0] \n\t"
1080 "packushb %[ftmp3], %[ftmp3], %[ftmp0] \n\t"
1081 "packushb %[ftmp4], %[ftmp4], %[ftmp0] \n\t"
1082 "packushb %[ftmp5], %[ftmp5], %[ftmp0] \n\t"
1083 "packushb %[ftmp6], %[ftmp6], %[ftmp0] \n\t"
1084 "packushb %[ftmp7], %[ftmp7], %[ftmp0] \n\t"
1085 "packushb %[ftmp8], %[ftmp8], %[ftmp0] \n\t"
1087 MMI_SWC1(%[ftmp1], %[dest], 0x00)
1088 PTR_ADDU "%[tmp0], %[dest], %[linesize] \n\t"
1089 MMI_SWC1(%[ftmp2], %[tmp0], 0x00)
1090 PTR_ADDU "%[tmp0], %[tmp0], %[linesize] \n\t"
1091 MMI_SWC1(%[ftmp3], %[tmp0], 0x00)
1092 PTR_ADDU "%[tmp0], %[tmp0], %[linesize] \n\t"
1093 MMI_SWC1(%[ftmp4], %[tmp0], 0x00)
1094 PTR_ADDU "%[tmp0], %[tmp0], %[linesize] \n\t"
1095 MMI_SWC1(%[ftmp5], %[tmp0], 0x00)
1096 PTR_ADDU "%[tmp0], %[tmp0], %[linesize] \n\t"
1097 MMI_SWC1(%[ftmp6], %[tmp0], 0x00)
1098 PTR_ADDU "%[tmp0], %[tmp0], %[linesize] \n\t"
1099 MMI_SWC1(%[ftmp7], %[tmp0], 0x00)
1100 PTR_ADDU "%[tmp0], %[tmp0], %[linesize] \n\t"
1101 MMI_SWC1(%[ftmp8], %[tmp0], 0x00)
1103 : [ftmp0]"=&f"(ftmp[0]), [ftmp1]"=&f"(ftmp[1]),
1104 [ftmp2]"=&f"(ftmp[2]), [ftmp3]"=&f"(ftmp[3]),
1105 [ftmp4]"=&f"(ftmp[4]), [ftmp5]"=&f"(ftmp[5]),
1106 [ftmp6]"=&f"(ftmp[6]), [ftmp7]"=&f"(ftmp[7]),
1107 [ftmp8]"=&f"(ftmp[8]), [ftmp9]"=&f"(ftmp[9]),
1108 [ftmp10]"=&f"(ftmp[10]), [ftmp11]"=&f"(ftmp[11]),
1109 [ftmp12]"=&f"(ftmp[12]), [ftmp13]"=&f"(ftmp[13]),
1110 [ftmp14]"=&f"(ftmp[14]), [ftmp15]"=&f"(ftmp[15]),
1111 [ftmp16]"=&f"(ftmp[16]), [ftmp17]"=&f"(ftmp[17]),
1112 [ftmp18]"=&f"(ftmp[18]), [ftmp19]"=&f"(ftmp[19]),
1113 [ftmp20]"=&f"(ftmp[20]), [ftmp21]"=&f"(ftmp[21]),
1114 [ftmp22]"=&f"(ftmp[22]),
1116 : [ff_pw_1]"f"(ff_pw_1_local), [ff_pw_64]"f"(ff_pw_64_local),
1117 [src]"r"(src), [dest]"r"(dest), [linesize]"r"(linesize)
1123 /* Do inverse transform on 4x4 part of block */
1124 void ff_vc1_inv_trans_4x4_dc_mmi(uint8_t *dest, ptrdiff_t linesize, int16_t *block)
1130 dc = (17 * dc + 4) >> 3;
1131 dc = (17 * dc + 64) >> 7;
1134 "xor %[ftmp0], %[ftmp0], %[ftmp0] \n\t"
1135 "pshufh %[dc], %[dc], %[ftmp0] \n\t"
1137 MMI_LWC1(%[ftmp1], %[dest0], 0x00)
1138 MMI_LWC1(%[ftmp2], %[dest1], 0x00)
1139 MMI_LWC1(%[ftmp3], %[dest2], 0x00)
1140 MMI_LWC1(%[ftmp4], %[dest3], 0x00)
1142 "punpcklbh %[ftmp1], %[ftmp1], %[ftmp0] \n\t"
1143 "punpcklbh %[ftmp2], %[ftmp2], %[ftmp0] \n\t"
1144 "punpcklbh %[ftmp3], %[ftmp3], %[ftmp0] \n\t"
1145 "punpcklbh %[ftmp4], %[ftmp4], %[ftmp0] \n\t"
1147 "paddsh %[ftmp1], %[ftmp1], %[dc] \n\t"
1148 "paddsh %[ftmp2], %[ftmp2], %[dc] \n\t"
1149 "paddsh %[ftmp3], %[ftmp3], %[dc] \n\t"
1150 "paddsh %[ftmp4], %[ftmp4], %[dc] \n\t"
1152 "packushb %[ftmp1], %[ftmp1], %[ftmp0] \n\t"
1153 "packushb %[ftmp2], %[ftmp2], %[ftmp0] \n\t"
1154 "packushb %[ftmp3], %[ftmp3], %[ftmp0] \n\t"
1155 "packushb %[ftmp4], %[ftmp4], %[ftmp0] \n\t"
1157 MMI_SWC1(%[ftmp1], %[dest0], 0x00)
1158 MMI_SWC1(%[ftmp2], %[dest1], 0x00)
1159 MMI_SWC1(%[ftmp3], %[dest2], 0x00)
1160 MMI_SWC1(%[ftmp4], %[dest3], 0x00)
1161 : [ftmp0]"=&f"(ftmp[0]), [ftmp1]"=&f"(ftmp[1]),
1162 [ftmp2]"=&f"(ftmp[2]), [ftmp3]"=&f"(ftmp[3]),
1164 [ftmp4]"=&f"(ftmp[4])
1165 : [dest0]"r"(dest+0*linesize), [dest1]"r"(dest+1*linesize),
1166 [dest2]"r"(dest+2*linesize), [dest3]"r"(dest+3*linesize),
1172 void ff_vc1_inv_trans_4x4_mmi(uint8_t *dest, ptrdiff_t linesize, int16_t *block)
1174 int16_t *src = block;
1175 int16_t *dst = block;
1177 uint32_t count = 4, tmp[1];
1178 int16_t coeff[16] = {17, 22, 17, 10,
1182 DECLARE_ALIGNED(8, const uint64_t, ff_pw_4_local) = {0x0000000400000004ULL};
1183 DECLARE_ALIGNED(8, const uint64_t, ff_pw_64_local)= {0x0000004000000040ULL};
1187 "li %[tmp0], 0x03 \n\t"
1188 "mtc1 %[tmp0], %[ftmp0] \n\t"
1189 MMI_LDC1(%[ftmp2], %[coeff], 0x00)
1190 MMI_LDC1(%[ftmp3], %[coeff], 0x08)
1191 MMI_LDC1(%[ftmp4], %[coeff], 0x10)
1192 MMI_LDC1(%[ftmp5], %[coeff], 0x18)
1194 /* ftmp8: dst3,dst2,dst1,dst0 */
1195 MMI_LDC1(%[ftmp1], %[src], 0x00)
1196 "pmaddhw %[ftmp6], %[ftmp2], %[ftmp1] \n\t"
1197 "pmaddhw %[ftmp7], %[ftmp3], %[ftmp1] \n\t"
1198 "pmaddhw %[ftmp8], %[ftmp4], %[ftmp1] \n\t"
1199 "pmaddhw %[ftmp9], %[ftmp5], %[ftmp1] \n\t"
1200 "punpcklwd %[ftmp10], %[ftmp6], %[ftmp7] \n\t"
1201 "punpckhwd %[ftmp11], %[ftmp6], %[ftmp7] \n\t"
1202 "punpcklwd %[ftmp6], %[ftmp8], %[ftmp9] \n\t"
1203 "punpckhwd %[ftmp7], %[ftmp8], %[ftmp9] \n\t"
1204 "paddw %[ftmp8], %[ftmp10], %[ftmp11] \n\t"
1205 "paddw %[ftmp9], %[ftmp6], %[ftmp7] \n\t"
1206 "paddw %[ftmp8], %[ftmp8], %[ff_pw_4] \n\t"
1207 "paddw %[ftmp9], %[ftmp9], %[ff_pw_4] \n\t"
1208 "psraw %[ftmp8], %[ftmp8], %[ftmp0] \n\t"
1209 "psraw %[ftmp9], %[ftmp9], %[ftmp0] \n\t"
1210 "punpcklhw %[ftmp6], %[ftmp8], %[ftmp9] \n\t"
1211 "punpckhhw %[ftmp7], %[ftmp8], %[ftmp9] \n\t"
1212 "punpcklhw %[ftmp8], %[ftmp6], %[ftmp7] \n\t"
1213 MMI_SDC1(%[ftmp8], %[dst], 0x00)
1215 PTR_ADDIU "%[src], %[src], 0x10 \n\t"
1216 PTR_ADDIU "%[dst], %[dst], 0x10 \n\t"
1217 "addiu %[count], %[count], -0x01 \n\t"
1218 "bnez %[count], 1b \n\t"
1219 : [ftmp0]"=&f"(ftmp[0]), [ftmp1]"=&f"(ftmp[1]),
1220 [ftmp2]"=&f"(ftmp[2]), [ftmp3]"=&f"(ftmp[3]),
1221 [ftmp4]"=&f"(ftmp[4]), [ftmp5]"=&f"(ftmp[5]),
1222 [ftmp6]"=&f"(ftmp[6]), [ftmp7]"=&f"(ftmp[7]),
1223 [ftmp8]"=&f"(ftmp[8]), [ftmp9]"=&f"(ftmp[9]),
1224 [ftmp10]"=&f"(ftmp[10]), [ftmp11]"=&f"(ftmp[11]),
1225 [tmp0]"=&r"(tmp[0]), [count]"+&r"(count),
1226 [src]"+&r"(src), [dst]"+&r"(dst)
1227 : [ff_pw_4]"f"(ff_pw_4_local), [coeff]"r"(coeff)
1235 "li %[tmp0], 0x07 \n\t"
1236 "mtc1 %[tmp0], %[ftmp0] \n\t"
1237 "li %[tmp0], 0x44 \n\t"
1238 "mtc1 %[tmp0], %[ftmp15] \n\t"
1240 MMI_LDC1(%[ftmp1], %[src], 0x00)
1241 MMI_LDC1(%[ftmp2], %[src], 0x10)
1242 MMI_LDC1(%[ftmp3], %[src], 0x20)
1243 MMI_LDC1(%[ftmp4], %[src], 0x30)
1244 "punpcklhw %[ftmp5], %[ftmp1], %[ftmp2] \n\t"
1245 "punpckhhw %[ftmp6], %[ftmp1], %[ftmp2] \n\t"
1246 "punpcklhw %[ftmp7], %[ftmp3], %[ftmp4] \n\t"
1247 "punpckhhw %[ftmp8], %[ftmp3], %[ftmp4] \n\t"
1249 /* ftmp11: dst03,dst02,dst01,dst00 */
1250 "li %[tmp0], 0x00160011 \n\t"
1251 "mtc1 %[tmp0], %[ftmp3] \n\t"
1252 "pshufh %[ftmp3], %[ftmp3], %[ftmp15] \n\t"
1253 "li %[tmp0], 0x000a0011 \n\t"
1254 "mtc1 %[tmp0], %[ftmp4] \n\t"
1255 "pshufh %[ftmp4], %[ftmp4], %[ftmp15] \n\t"
1256 "pmaddhw %[ftmp1], %[ftmp5], %[ftmp3] \n\t"
1257 "pmaddhw %[ftmp2], %[ftmp7], %[ftmp4] \n\t"
1258 "paddw %[ftmp9], %[ftmp1], %[ftmp2] \n\t"
1259 "pmaddhw %[ftmp1], %[ftmp6], %[ftmp3] \n\t"
1260 "pmaddhw %[ftmp2], %[ftmp8], %[ftmp4] \n\t"
1261 "paddw %[ftmp10], %[ftmp1], %[ftmp2] \n\t"
1262 "paddw %[ftmp9], %[ftmp9], %[ff_pw_64] \n\t"
1263 "paddw %[ftmp10], %[ftmp10], %[ff_pw_64] \n\t"
1264 "psraw %[ftmp9], %[ftmp9], %[ftmp0] \n\t"
1265 "psraw %[ftmp10], %[ftmp10], %[ftmp0] \n\t"
1266 "punpcklhw %[ftmp1], %[ftmp9], %[ftmp10] \n\t"
1267 "punpckhhw %[ftmp2], %[ftmp9], %[ftmp10] \n\t"
1268 "punpcklhw %[ftmp11], %[ftmp1], %[ftmp2] \n\t"
1270 /* ftmp12: dst13,dst12,dst11,dst10 */
1271 "li %[tmp0], 0x000a0011 \n\t"
1272 "mtc1 %[tmp0], %[ftmp3] \n\t"
1273 "pshufh %[ftmp3], %[ftmp3], %[ftmp15] \n\t"
1274 "li %[tmp0], 0xffeaffef \n\t"
1275 "mtc1 %[tmp0], %[ftmp4] \n\t"
1276 "pshufh %[ftmp4], %[ftmp4], %[ftmp15] \n\t"
1277 "pmaddhw %[ftmp1], %[ftmp5], %[ftmp3] \n\t"
1278 "pmaddhw %[ftmp2], %[ftmp7], %[ftmp4] \n\t"
1279 "paddw %[ftmp9], %[ftmp1], %[ftmp2] \n\t"
1280 "pmaddhw %[ftmp1], %[ftmp6], %[ftmp3] \n\t"
1281 "pmaddhw %[ftmp2], %[ftmp8], %[ftmp4] \n\t"
1282 "paddw %[ftmp10], %[ftmp1], %[ftmp2] \n\t"
1283 "paddw %[ftmp9], %[ftmp9], %[ff_pw_64] \n\t"
1284 "paddw %[ftmp10], %[ftmp10], %[ff_pw_64] \n\t"
1285 "psraw %[ftmp9], %[ftmp9], %[ftmp0] \n\t"
1286 "psraw %[ftmp10], %[ftmp10], %[ftmp0] \n\t"
1287 "punpcklhw %[ftmp1], %[ftmp9], %[ftmp10] \n\t"
1288 "punpckhhw %[ftmp2], %[ftmp9], %[ftmp10] \n\t"
1289 "punpcklhw %[ftmp12], %[ftmp1], %[ftmp2] \n\t"
1291 /* ftmp13: dst23,dst22,dst21,dst20 */
1292 "li %[tmp0], 0xfff60011 \n\t"
1293 "mtc1 %[tmp0], %[ftmp3] \n\t"
1294 "pshufh %[ftmp3], %[ftmp3], %[ftmp15] \n\t"
1295 "li %[tmp0], 0x0016ffef \n\t"
1296 "mtc1 %[tmp0], %[ftmp4] \n\t"
1297 "pshufh %[ftmp4], %[ftmp4], %[ftmp15] \n\t"
1298 "pmaddhw %[ftmp1], %[ftmp5], %[ftmp3] \n\t"
1299 "pmaddhw %[ftmp2], %[ftmp7], %[ftmp4] \n\t"
1300 "paddw %[ftmp9], %[ftmp1], %[ftmp2] \n\t"
1301 "pmaddhw %[ftmp1], %[ftmp6], %[ftmp3] \n\t"
1302 "pmaddhw %[ftmp2], %[ftmp8], %[ftmp4] \n\t"
1303 "paddw %[ftmp10], %[ftmp1], %[ftmp2] \n\t"
1304 "paddw %[ftmp9], %[ftmp9], %[ff_pw_64] \n\t"
1305 "paddw %[ftmp10], %[ftmp10], %[ff_pw_64] \n\t"
1306 "psraw %[ftmp9], %[ftmp9], %[ftmp0] \n\t"
1307 "psraw %[ftmp10], %[ftmp10], %[ftmp0] \n\t"
1308 "punpcklhw %[ftmp1], %[ftmp9], %[ftmp10] \n\t"
1309 "punpckhhw %[ftmp2], %[ftmp9], %[ftmp10] \n\t"
1310 "punpcklhw %[ftmp13], %[ftmp1], %[ftmp2] \n\t"
1312 /* ftmp14: dst33,dst32,dst31,dst30 */
1313 "li %[tmp0], 0xffea0011 \n\t"
1314 "mtc1 %[tmp0], %[ftmp3] \n\t"
1315 "pshufh %[ftmp3], %[ftmp3], %[ftmp15] \n\t"
1316 "li %[tmp0], 0xfff60011 \n\t"
1317 "mtc1 %[tmp0], %[ftmp4] \n\t"
1318 "pshufh %[ftmp4], %[ftmp4], %[ftmp15] \n\t"
1319 "pmaddhw %[ftmp1], %[ftmp5], %[ftmp3] \n\t"
1320 "pmaddhw %[ftmp2], %[ftmp7], %[ftmp4] \n\t"
1321 "paddw %[ftmp9], %[ftmp1], %[ftmp2] \n\t"
1322 "pmaddhw %[ftmp1], %[ftmp6], %[ftmp3] \n\t"
1323 "pmaddhw %[ftmp2], %[ftmp8], %[ftmp4] \n\t"
1324 "paddw %[ftmp10], %[ftmp1], %[ftmp2] \n\t"
1325 "paddw %[ftmp9], %[ftmp9], %[ff_pw_64] \n\t"
1326 "paddw %[ftmp10], %[ftmp10], %[ff_pw_64] \n\t"
1327 "psraw %[ftmp9], %[ftmp9], %[ftmp0] \n\t"
1328 "psraw %[ftmp10], %[ftmp10], %[ftmp0] \n\t"
1329 "punpcklhw %[ftmp1], %[ftmp9], %[ftmp10] \n\t"
1330 "punpckhhw %[ftmp2], %[ftmp9], %[ftmp10] \n\t"
1331 "punpcklhw %[ftmp14], %[ftmp1], %[ftmp2] \n\t"
1333 MMI_LWC1(%[ftmp1], %[dest], 0x00)
1334 PTR_ADDU "%[tmp0], %[dest], %[linesize] \n\t"
1335 MMI_LWC1(%[ftmp2], %[tmp0], 0x00)
1336 PTR_ADDU "%[tmp0], %[tmp0], %[linesize] \n\t"
1337 MMI_LWC1(%[ftmp3], %[tmp0], 0x00)
1338 PTR_ADDU "%[tmp0], %[tmp0], %[linesize] \n\t"
1339 MMI_LWC1(%[ftmp4], %[tmp0], 0x00)
1340 "xor %[ftmp0], %[ftmp0], %[ftmp0] \n\t"
1341 "punpcklbh %[ftmp1], %[ftmp1], %[ftmp0] \n\t"
1342 "punpcklbh %[ftmp2], %[ftmp2], %[ftmp0] \n\t"
1343 "punpcklbh %[ftmp3], %[ftmp3], %[ftmp0] \n\t"
1344 "punpcklbh %[ftmp4], %[ftmp4], %[ftmp0] \n\t"
1345 "paddh %[ftmp1], %[ftmp1], %[ftmp11] \n\t"
1346 "paddh %[ftmp2], %[ftmp2], %[ftmp12] \n\t"
1347 "paddh %[ftmp3], %[ftmp3], %[ftmp13] \n\t"
1348 "paddh %[ftmp4], %[ftmp4], %[ftmp14] \n\t"
1349 "packushb %[ftmp1], %[ftmp1], %[ftmp0] \n\t"
1350 "packushb %[ftmp2], %[ftmp2], %[ftmp0] \n\t"
1351 "packushb %[ftmp3], %[ftmp3], %[ftmp0] \n\t"
1352 "packushb %[ftmp4], %[ftmp4], %[ftmp0] \n\t"
1354 MMI_SWC1(%[ftmp1], %[dest], 0x00)
1355 PTR_ADDU "%[tmp0], %[dest], %[linesize] \n\t"
1356 MMI_SWC1(%[ftmp2], %[tmp0], 0x00)
1357 PTR_ADDU "%[tmp0], %[tmp0], %[linesize] \n\t"
1358 MMI_SWC1(%[ftmp3], %[tmp0], 0x00)
1359 PTR_ADDU "%[tmp0], %[tmp0], %[linesize] \n\t"
1360 MMI_SWC1(%[ftmp4], %[tmp0], 0x00)
1362 : [ftmp0]"=&f"(ftmp[0]), [ftmp1]"=&f"(ftmp[1]),
1363 [ftmp2]"=&f"(ftmp[2]), [ftmp3]"=&f"(ftmp[3]),
1364 [ftmp4]"=&f"(ftmp[4]), [ftmp5]"=&f"(ftmp[5]),
1365 [ftmp6]"=&f"(ftmp[6]), [ftmp7]"=&f"(ftmp[7]),
1366 [ftmp8]"=&f"(ftmp[8]), [ftmp9]"=&f"(ftmp[9]),
1367 [ftmp10]"=&f"(ftmp[10]), [ftmp11]"=&f"(ftmp[11]),
1368 [ftmp12]"=&f"(ftmp[12]), [ftmp13]"=&f"(ftmp[13]),
1369 [ftmp14]"=&f"(ftmp[14]), [ftmp15]"=&f"(ftmp[15]),
1371 : [ff_pw_64]"f"(ff_pw_64_local),
1372 [src]"r"(src), [dest]"r"(dest), [linesize]"r"(linesize)
1377 /* Apply overlap transform to horizontal edge */
1378 void ff_vc1_h_overlap_mmi(uint8_t *src, int stride)
1384 for (i = 0; i < 8; i++) {
1389 d1 = (a - d + 3 + rnd) >> 3;
1390 d2 = (a - d + b - c + 4 - rnd) >> 3;
1393 src[-1] = av_clip_uint8(b - d2);
1394 src[0] = av_clip_uint8(c + d2);
1401 void ff_vc1_h_s_overlap_mmi(int16_t *left, int16_t *right, int left_stride, int right_stride, int flags)
1406 int rnd1 = flags & 2 ? 3 : 4;
1407 int rnd2 = 7 - rnd1;
1408 for (i = 0; i < 8; i++) {
1416 left[6] = ((a << 3) - d1 + rnd1) >> 3;
1417 left[7] = ((b << 3) - d2 + rnd2) >> 3;
1418 right[0] = ((c << 3) + d2 + rnd1) >> 3;
1419 right[1] = ((d << 3) + d1 + rnd2) >> 3;
1421 right += right_stride;
1422 left += left_stride;
1430 /* Apply overlap transform to vertical edge */
1431 void ff_vc1_v_overlap_mmi(uint8_t *src, int stride)
1437 for (i = 0; i < 8; i++) {
1438 a = src[-2 * stride];
1442 d1 = (a - d + 3 + rnd) >> 3;
1443 d2 = (a - d + b - c + 4 - rnd) >> 3;
1445 src[-2 * stride] = a - d1;
1446 src[-stride] = av_clip_uint8(b - d2);
1447 src[0] = av_clip_uint8(c + d2);
1448 src[stride] = d + d1;
1454 void ff_vc1_v_s_overlap_mmi(int16_t *top, int16_t *bottom)
1459 int rnd1 = 4, rnd2 = 3;
1460 for (i = 0; i < 8; i++) {
1468 top[48] = ((a << 3) - d1 + rnd1) >> 3;
1469 top[56] = ((b << 3) - d2 + rnd2) >> 3;
1470 bottom[0] = ((c << 3) + d2 + rnd1) >> 3;
1471 bottom[8] = ((d << 3) + d1 + rnd2) >> 3;
1481 * VC-1 in-loop deblocking filter for one line
1482 * @param src source block type
1483 * @param stride block stride
1484 * @param pq block quantizer
1485 * @return whether other 3 pairs should be filtered or not
1488 static av_always_inline int vc1_filter_line(uint8_t *src, int stride, int pq)
1490 int a0 = (2 * (src[-2 * stride] - src[1 * stride]) -
1491 5 * (src[-1 * stride] - src[0 * stride]) + 4) >> 3;
1492 int a0_sign = a0 >> 31; /* Store sign */
1494 a0 = (a0 ^ a0_sign) - a0_sign; /* a0 = FFABS(a0); */
1496 int a1 = FFABS((2 * (src[-4 * stride] - src[-1 * stride]) -
1497 5 * (src[-3 * stride] - src[-2 * stride]) + 4) >> 3);
1498 int a2 = FFABS((2 * (src[ 0 * stride] - src[ 3 * stride]) -
1499 5 * (src[ 1 * stride] - src[ 2 * stride]) + 4) >> 3);
1500 if (a1 < a0 || a2 < a0) {
1501 int clip = src[-1 * stride] - src[0 * stride];
1502 int clip_sign = clip >> 31;
1504 clip = ((clip ^ clip_sign) - clip_sign) >> 1;
1506 int a3 = FFMIN(a1, a2);
1507 int d = 5 * (a3 - a0);
1508 int d_sign = (d >> 31);
1510 d = ((d ^ d_sign) - d_sign) >> 3;
1513 if (d_sign ^ clip_sign)
1517 d = (d ^ d_sign) - d_sign; /* Restore sign */
1518 src[-1 * stride] = av_clip_uint8(src[-1 * stride] - d);
1519 src[ 0 * stride] = av_clip_uint8(src[ 0 * stride] + d);
1529 * VC-1 in-loop deblocking filter
1530 * @param src source block type
1531 * @param step distance between horizontally adjacent elements
1532 * @param stride distance between vertically adjacent elements
1533 * @param len edge length to filter (4 or 8 pixels)
1534 * @param pq block quantizer
1537 static inline void vc1_loop_filter(uint8_t *src, int step, int stride,
1543 for (i = 0; i < len; i += 4) {
1544 filt3 = vc1_filter_line(src + 2 * step, stride, pq);
1546 vc1_filter_line(src + 0 * step, stride, pq);
1547 vc1_filter_line(src + 1 * step, stride, pq);
1548 vc1_filter_line(src + 3 * step, stride, pq);
1554 void ff_vc1_v_loop_filter4_mmi(uint8_t *src, int stride, int pq)
1556 vc1_loop_filter(src, 1, stride, 4, pq);
1559 void ff_vc1_h_loop_filter4_mmi(uint8_t *src, int stride, int pq)
1561 vc1_loop_filter(src, stride, 1, 4, pq);
1564 void ff_vc1_v_loop_filter8_mmi(uint8_t *src, int stride, int pq)
1566 vc1_loop_filter(src, 1, stride, 8, pq);
1569 void ff_vc1_h_loop_filter8_mmi(uint8_t *src, int stride, int pq)
1571 vc1_loop_filter(src, stride, 1, 8, pq);
1574 void ff_vc1_v_loop_filter16_mmi(uint8_t *src, int stride, int pq)
1576 vc1_loop_filter(src, 1, stride, 16, pq);
1579 void ff_vc1_h_loop_filter16_mmi(uint8_t *src, int stride, int pq)
1581 vc1_loop_filter(src, stride, 1, 16, pq);
1584 void ff_put_vc1_mspel_mc00_mmi(uint8_t *dst, const uint8_t *src,
1585 ptrdiff_t stride, int rnd)
1587 ff_put_pixels8_8_mmi(dst, src, stride, 8);
1589 void ff_put_vc1_mspel_mc00_16_mmi(uint8_t *dst, const uint8_t *src,
1590 ptrdiff_t stride, int rnd)
1592 ff_put_pixels16_8_mmi(dst, src, stride, 16);
1594 void ff_avg_vc1_mspel_mc00_mmi(uint8_t *dst, const uint8_t *src,
1595 ptrdiff_t stride, int rnd)
1597 ff_avg_pixels8_8_mmi(dst, src, stride, 8);
1599 void ff_avg_vc1_mspel_mc00_16_mmi(uint8_t *dst, const uint8_t *src,
1600 ptrdiff_t stride, int rnd)
1602 ff_avg_pixels16_8_mmi(dst, src, stride, 16);
1605 #define OP_PUT(S, D)
1606 #define OP_AVG(S, D) \
1607 "ldc1 $f16, "#S" \n\t" \
1608 "pavgb "#D", "#D", $f16 \n\t"
1610 /** Add rounder from $f14 to $f6 and pack result at destination */
1611 #define NORMALIZE_MMI(SHIFT) \
1612 "paddh $f6, $f6, $f14 \n\t" /* +bias-r */ \
1613 "paddh $f8, $f8, $f14 \n\t" /* +bias-r */ \
1614 "psrah $f6, $f6, "SHIFT" \n\t" \
1615 "psrah $f8, $f8, "SHIFT" \n\t"
1617 #define TRANSFER_DO_PACK(OP) \
1618 "packushb $f6, $f6, $f8 \n\t" \
1620 "sdc1 $f6, 0x00(%[dst]) \n\t"
1622 #define TRANSFER_DONT_PACK(OP) \
1623 OP(0(%[dst]), $f6) \
1624 OP(8(%[dst]), $f8) \
1625 "sdc1 $f6, 0x00(%[dst]) \n\t" \
1626 "sdc1 $f8, 0x08(%[dst]) \n\t"
1628 /** @see MSPEL_FILTER13_CORE for use as UNPACK macro */
1629 #define DO_UNPACK(reg) \
1630 "punpcklbh "reg", "reg", $f0 \n\t"
1631 #define DONT_UNPACK(reg)
1633 /** Compute the rounder 32-r or 8-r and unpacks it to $f14 */
1634 #define LOAD_ROUNDER_MMI(ROUND) \
1635 "lwc1 $f14, "ROUND" \n\t" \
1636 "punpcklhw $f14, $f14, $f14 \n\t" \
1637 "punpcklwd $f14, $f14, $f14 \n\t"
1640 #define SHIFT2_LINE(OFF, R0, R1, R2, R3) \
1641 "paddh "#R1", "#R1", "#R2" \n\t" \
1642 PTR_ADDU "$9, %[src], %[stride1] \n\t" \
1643 MMI_ULWC1(R0, $9, 0x00) \
1644 "pmullh "#R1", "#R1", $f6 \n\t" \
1645 "punpcklbh "#R0", "#R0", $f0 \n\t" \
1646 PTR_ADDU "$9, %[src], %[stride] \n\t" \
1647 MMI_ULWC1(R3, $9, 0x00) \
1648 "psubh "#R1", "#R1", "#R0" \n\t" \
1649 "punpcklbh "#R3", "#R3", $f0 \n\t" \
1650 "paddh "#R1", "#R1", $f14 \n\t" \
1651 "psubh "#R1", "#R1", "#R3" \n\t" \
1652 "psrah "#R1", "#R1", %[shift] \n\t" \
1653 MMI_SDC1(R1, %[dst], OFF) \
1654 PTR_ADDU "%[src], %[src], %[stride] \n\t"
1656 /** Sacrificing $f12 makes it possible to pipeline loads from src */
1657 static void vc1_put_ver_16b_shift2_mmi(int16_t *dst,
1658 const uint8_t *src, mips_reg stride,
1659 int rnd, int64_t shift)
1665 "xor $f0, $f0, $f0 \n\t"
1667 LOAD_ROUNDER_MMI("%[rnd]")
1668 "ldc1 $f12, %[ff_pw_9] \n\t"
1670 MMI_ULWC1($f4, %[src], 0x00)
1671 PTR_ADDU "%[src], %[src], %[stride] \n\t"
1672 MMI_ULWC1($f6, %[src], 0x00)
1673 "punpcklbh $f4, $f4, $f0 \n\t"
1674 "punpcklbh $f6, $f6, $f0 \n\t"
1675 SHIFT2_LINE( 0, $f2, $f4, $f6, $f8)
1676 SHIFT2_LINE( 24, $f4, $f6, $f8, $f2)
1677 SHIFT2_LINE( 48, $f6, $f8, $f2, $f4)
1678 SHIFT2_LINE( 72, $f8, $f2, $f4, $f6)
1679 SHIFT2_LINE( 96, $f2, $f4, $f6, $f8)
1680 SHIFT2_LINE(120, $f4, $f6, $f8, $f2)
1681 SHIFT2_LINE(144, $f6, $f8, $f2, $f4)
1682 SHIFT2_LINE(168, $f8, $f2, $f4, $f6)
1683 PTR_SUBU "%[src], %[src], %[stride2] \n\t"
1684 PTR_ADDIU "%[dst], %[dst], 0x08 \n\t"
1685 "addiu $8, $8, -0x01 \n\t"
1687 : RESTRICT_ASM_LOW32 RESTRICT_ASM_ADDRT
1688 [src]"+r"(src), [dst]"+r"(dst)
1689 : [stride]"r"(stride), [stride1]"r"(-2*stride),
1690 [shift]"f"(shift), [rnd]"m"(rnd),
1691 [stride2]"r"(9*stride-4), [ff_pw_9]"m"(ff_pw_9)
1692 : "$8", "$9", "$f0", "$f2", "$f4", "$f6", "$f8", "$f10", "$f12",
1693 "$f14", "$f16", "memory"
1698 * Data is already unpacked, so some operations can directly be made from
1701 #define VC1_HOR_16B_SHIFT2(OP, OPNAME) \
1702 static void OPNAME ## vc1_hor_16b_shift2_mmi(uint8_t *dst, mips_reg stride, \
1703 const int16_t *src, int rnd) \
1706 DECLARE_VAR_ALL64; \
1707 DECLARE_VAR_ADDRT; \
1710 rnd -= (-1+9+9-1)*1024; /* Add -1024 bias */ \
1713 LOAD_ROUNDER_MMI("%[rnd]") \
1714 "ldc1 $f12, %[ff_pw_128] \n\t" \
1715 "ldc1 $f10, %[ff_pw_9] \n\t" \
1717 MMI_ULDC1($f2, %[src], 0x00) \
1718 MMI_ULDC1($f4, %[src], 0x08) \
1719 MMI_ULDC1($f6, %[src], 0x02) \
1720 MMI_ULDC1($f8, %[src], 0x0a) \
1721 MMI_ULDC1($f0, %[src], 0x06) \
1722 "paddh $f2, $f2, $f0 \n\t" \
1723 MMI_ULDC1($f0, %[src], 0x0e) \
1724 "paddh $f4, $f4, $f0 \n\t" \
1725 MMI_ULDC1($f0, %[src], 0x04) \
1726 "paddh $f6, $f6, $f0 \n\t" \
1727 MMI_ULDC1($f0, %[src], 0x0b) \
1728 "paddh $f8, $f8, $f0 \n\t" \
1729 "pmullh $f6, $f6, $f10 \n\t" \
1730 "pmullh $f8, $f8, $f10 \n\t" \
1731 "psubh $f6, $f6, $f2 \n\t" \
1732 "psubh $f8, $f8, $f4 \n\t" \
1733 "li $8, 0x07 \n\t" \
1734 "mtc1 $8, $f16 \n\t" \
1735 NORMALIZE_MMI("$f16") \
1737 "paddh $f6, $f6, $f12 \n\t" \
1738 "paddh $f8, $f8, $f12 \n\t" \
1739 TRANSFER_DO_PACK(OP) \
1740 "addiu %[h], %[h], -0x01 \n\t" \
1741 PTR_ADDIU "%[src], %[src], 0x18 \n\t" \
1742 PTR_ADDU "%[dst], %[dst], %[stride] \n\t" \
1743 "bnez %[h], 1b \n\t" \
1744 : RESTRICT_ASM_ALL64 RESTRICT_ASM_ADDRT \
1746 [src]"+r"(src), [dst]"+r"(dst) \
1747 : [stride]"r"(stride), [rnd]"m"(rnd), \
1748 [ff_pw_9]"m"(ff_pw_9), [ff_pw_128]"m"(ff_pw_128) \
1749 : "$8", "$f0", "$f2", "$f4", "$f6", "$f8", "$f10", "$f12", "$f14", \
1754 VC1_HOR_16B_SHIFT2(OP_PUT, put_)
1755 VC1_HOR_16B_SHIFT2(OP_AVG, avg_)
1758 * Purely vertical or horizontal 1/2 shift interpolation.
1759 * Sacrify $f12 for *9 factor.
1761 #define VC1_SHIFT2(OP, OPNAME)\
1762 static void OPNAME ## vc1_shift2_mmi(uint8_t *dst, const uint8_t *src, \
1763 mips_reg stride, int rnd, \
1766 DECLARE_VAR_LOW32; \
1767 DECLARE_VAR_ADDRT; \
1772 "xor $f0, $f0, $f0 \n\t" \
1773 "li $10, 0x08 \n\t" \
1774 LOAD_ROUNDER_MMI("%[rnd]") \
1775 "ldc1 $f12, %[ff_pw_9] \n\t" \
1777 MMI_ULWC1($f6, %[src], 0x00) \
1778 MMI_ULWC1($f8, %[src], 0x04) \
1779 PTR_ADDU "$9, %[src], %[offset] \n\t" \
1780 MMI_ULWC1($f2, $9, 0x00) \
1781 MMI_ULWC1($f4, $9, 0x04) \
1782 PTR_ADDU "%[src], %[src], %[offset] \n\t" \
1783 "punpcklbh $f6, $f6, $f0 \n\t" \
1784 "punpcklbh $f8, $f8, $f0 \n\t" \
1785 "punpcklbh $f2, $f2, $f0 \n\t" \
1786 "punpcklbh $f4, $f4, $f0 \n\t" \
1787 "paddh $f6, $f6, $f2 \n\t" \
1788 "paddh $f8, $f8, $f4 \n\t" \
1789 PTR_ADDU "$9, %[src], %[offset_x2n] \n\t" \
1790 MMI_ULWC1($f2, $9, 0x00) \
1791 MMI_ULWC1($f4, $9, 0x04) \
1792 "pmullh $f6, $f6, $f12 \n\t" /* 0,9,9,0*/ \
1793 "pmullh $f8, $f8, $f12 \n\t" /* 0,9,9,0*/ \
1794 "punpcklbh $f2, $f2, $f0 \n\t" \
1795 "punpcklbh $f4, $f4, $f0 \n\t" \
1796 "psubh $f6, $f6, $f2 \n\t" /*-1,9,9,0*/ \
1797 "psubh $f8, $f8, $f4 \n\t" /*-1,9,9,0*/ \
1798 PTR_ADDU "$9, %[src], %[offset] \n\t" \
1799 MMI_ULWC1($f2, $9, 0x00) \
1800 MMI_ULWC1($f4, $9, 0x04) \
1801 "punpcklbh $f2, $f2, $f0 \n\t" \
1802 "punpcklbh $f4, $f4, $f0 \n\t" \
1803 "psubh $f6, $f6, $f2 \n\t" /*-1,9,9,-1*/ \
1804 "psubh $f8, $f8, $f4 \n\t" /*-1,9,9,-1*/ \
1805 "li $8, 0x04 \n\t" \
1806 "mtc1 $8, $f16 \n\t" \
1807 NORMALIZE_MMI("$f16") \
1808 "packushb $f6, $f6, $f8 \n\t" \
1810 "sdc1 $f6, 0x00(%[dst]) \n\t" \
1811 "addiu $10, $10, -0x01 \n\t" \
1812 PTR_ADDU "%[src], %[src], %[stride1] \n\t" \
1813 PTR_ADDU "%[dst], %[dst], %[stride] \n\t" \
1814 "bnez $10, 1b \n\t" \
1815 : RESTRICT_ASM_LOW32 RESTRICT_ASM_ADDRT \
1816 [src]"+r"(src), [dst]"+r"(dst) \
1817 : [offset]"r"(offset), [offset_x2n]"r"(-2*offset), \
1818 [stride]"r"(stride), [rnd]"m"(rnd), \
1819 [stride1]"r"(stride-offset), \
1820 [ff_pw_9]"m"(ff_pw_9) \
1821 : "$8", "$9", "$10", "$f0", "$f2", "$f4", "$f6", "$f8", "$f10", \
1822 "$f12", "$f14", "$f16", "memory" \
1826 VC1_SHIFT2(OP_PUT, put_)
1827 VC1_SHIFT2(OP_AVG, avg_)
1830 * Core of the 1/4 and 3/4 shift bicubic interpolation.
1832 * @param UNPACK Macro unpacking arguments from 8 to 16bits (can be empty).
1833 * @param LOAD "MMI_ULWC1" or "MMI_ULDC1", if data read is already unpacked.
1834 * @param M "1" for MMI_ULWC1, "2" for MMI_ULDC1.
1835 * @param A1 Stride address of 1st tap (beware of unpacked/packed).
1836 * @param A2 Stride address of 2nd tap
1837 * @param A3 Stride address of 3rd tap
1838 * @param A4 Stride address of 4th tap
1840 #define MSPEL_FILTER13_CORE(UNPACK, LOAD, M, A1, A2, A3, A4) \
1841 PTR_ADDU "$9, %[src], "#A1" \n\t" \
1842 LOAD($f2, $9, M*0) \
1843 LOAD($f4, $9, M*4) \
1846 "pmullh $f2, $f2, %[ff_pw_3] \n\t" \
1847 "pmullh $f4, $f4, %[ff_pw_3] \n\t" \
1848 PTR_ADDU "$9, %[src], "#A2" \n\t" \
1849 LOAD($f6, $9, M*0) \
1850 LOAD($f8, $9, M*4) \
1853 "pmullh $f6, $f6, $f12 \n\t" /* *18 */ \
1854 "pmullh $f8, $f8, $f12 \n\t" /* *18 */ \
1855 "psubh $f6, $f6, $f2 \n\t" /* *18, -3 */ \
1856 "psubh $f8, $f8, $f4 \n\t" /* *18, -3 */ \
1857 PTR_ADDU "$9, %[src], "#A4" \n\t" \
1858 LOAD($f2, $9, M*0) \
1859 LOAD($f4, $9, M*4) \
1862 "li $8, 0x02 \n\t" \
1863 "mtc1 $8, $f16 \n\t" \
1864 "psllh $f2, $f2, $f16 \n\t" /* 4* */ \
1865 "psllh $f4, $f4, $f16 \n\t" /* 4* */ \
1866 "psubh $f6, $f6, $f2 \n\t" /* -4,18,-3 */ \
1867 "psubh $f8, $f8, $f4 \n\t" /* -4,18,-3 */ \
1868 PTR_ADDU "$9, %[src], "#A3" \n\t" \
1869 LOAD($f2, $9, M*0) \
1870 LOAD($f4, $9, M*4) \
1873 "pmullh $f2, $f2, $f10 \n\t" /* *53 */ \
1874 "pmullh $f4, $f4, $f10 \n\t" /* *53 */ \
1875 "paddh $f6, $f6, $f2 \n\t" /* 4,53,18,-3 */ \
1876 "paddh $f8, $f8, $f4 \n\t" /* 4,53,18,-3 */
1879 * Macro to build the vertical 16bits version of vc1_put_shift[13].
1880 * Here, offset=src_stride. Parameters passed A1 to A4 must use
1881 * %3 (src_stride), %4 (2*src_stride) and %5 (3*src_stride).
1883 * @param NAME Either 1 or 3
1884 * @see MSPEL_FILTER13_CORE for information on A1->A4
1886 #define MSPEL_FILTER13_VER_16B(NAME, A1, A2, A3, A4) \
1888 vc1_put_ver_16b_ ## NAME ## _mmi(int16_t *dst, const uint8_t *src, \
1889 mips_reg src_stride, \
1890 int rnd, int64_t shift) \
1893 DECLARE_VAR_LOW32; \
1894 DECLARE_VAR_ADDRT; \
1896 src -= src_stride; \
1899 "xor $f0, $f0, $f0 \n\t" \
1900 LOAD_ROUNDER_MMI("%[rnd]") \
1901 "ldc1 $f10, %[ff_pw_53] \n\t" \
1902 "ldc1 $f12, %[ff_pw_18] \n\t" \
1905 MSPEL_FILTER13_CORE(DO_UNPACK, MMI_ULWC1, 1, A1, A2, A3, A4) \
1906 NORMALIZE_MMI("%[shift]") \
1907 TRANSFER_DONT_PACK(OP_PUT) \
1908 /* Last 3 (in fact 4) bytes on the line */ \
1909 PTR_ADDU "$9, %[src], "#A1" \n\t" \
1910 MMI_ULWC1($f2, $9, 0x08) \
1912 "mov.d $f6, $f2 \n\t" \
1913 "paddh $f2, $f2, $f2 \n\t" \
1914 "paddh $f2, $f2, $f6 \n\t" /* 3* */ \
1915 PTR_ADDU "$9, %[src], "#A2" \n\t" \
1916 MMI_ULWC1($f6, $9, 0x08) \
1918 "pmullh $f6, $f6, $f12 \n\t" /* *18 */ \
1919 "psubh $f6, $f6, $f2 \n\t" /* *18,-3 */ \
1920 PTR_ADDU "$9, %[src], "#A3" \n\t" \
1921 MMI_ULWC1($f2, $9, 0x08) \
1923 "pmullh $f2, $f2, $f10 \n\t" /* *53 */ \
1924 "paddh $f6, $f6, $f2 \n\t" /* *53,18,-3 */ \
1925 PTR_ADDU "$9, %[src], "#A4" \n\t" \
1926 MMI_ULWC1($f2, $9, 0x08) \
1928 "li $8, 0x02 \n\t" \
1929 "mtc1 $8, $f16 \n\t" \
1930 "psllh $f2, $f2, $f16 \n\t" /* 4* */ \
1931 "psubh $f6, $f6, $f2 \n\t" \
1932 "paddh $f6, $f6, $f14 \n\t" \
1933 "li $8, 0x06 \n\t" \
1934 "mtc1 $8, $f16 \n\t" \
1935 "psrah $f6, $f6, $f16 \n\t" \
1936 "sdc1 $f6, 0x10(%[dst]) \n\t" \
1937 "addiu %[h], %[h], -0x01 \n\t" \
1938 PTR_ADDU "%[src], %[src], %[stride_x1] \n\t" \
1939 PTR_ADDIU "%[dst], %[dst], 0x18 \n\t" \
1940 "bnez %[h], 1b \n\t" \
1941 : RESTRICT_ASM_LOW32 RESTRICT_ASM_ADDRT \
1943 [src]"+r"(src), [dst]"+r"(dst) \
1944 : [stride_x1]"r"(src_stride), [stride_x2]"r"(2*src_stride), \
1945 [stride_x3]"r"(3*src_stride), \
1946 [rnd]"m"(rnd), [shift]"f"(shift), \
1947 [ff_pw_53]"m"(ff_pw_53), [ff_pw_18]"m"(ff_pw_18), \
1948 [ff_pw_3]"f"(ff_pw_3) \
1949 : "$8", "$9", "$f0", "$f2", "$f4", "$f6", "$f8", "$f10", "$f12", \
1950 "$f14", "$f16", "memory" \
1955 * Macro to build the horizontal 16bits version of vc1_put_shift[13].
1956 * Here, offset=16bits, so parameters passed A1 to A4 should be simple.
1958 * @param NAME Either 1 or 3
1959 * @see MSPEL_FILTER13_CORE for information on A1->A4
1961 #define MSPEL_FILTER13_HOR_16B(NAME, A1, A2, A3, A4, OP, OPNAME) \
1963 OPNAME ## vc1_hor_16b_ ## NAME ## _mmi(uint8_t *dst, mips_reg stride, \
1964 const int16_t *src, int rnd) \
1967 DECLARE_VAR_ALL64; \
1968 DECLARE_VAR_ADDRT; \
1971 rnd -= (-4+58+13-3)*256; /* Add -256 bias */ \
1974 "xor $f0, $f0, $f0 \n\t" \
1975 LOAD_ROUNDER_MMI("%[rnd]") \
1976 "ldc1 $f10, %[ff_pw_53] \n\t" \
1977 "ldc1 $f12, %[ff_pw_18] \n\t" \
1980 MSPEL_FILTER13_CORE(DONT_UNPACK, MMI_ULDC1, 2, A1, A2, A3, A4) \
1981 "li $8, 0x07 \n\t" \
1982 "mtc1 $8, $f16 \n\t" \
1983 NORMALIZE_MMI("$f16") \
1985 "paddh $f6, $f6, %[ff_pw_128] \n\t" \
1986 "paddh $f8, $f8, %[ff_pw_128] \n\t" \
1987 TRANSFER_DO_PACK(OP) \
1988 "addiu %[h], %[h], -0x01 \n\t" \
1989 PTR_ADDU "%[src], %[src], 0x18 \n\t" \
1990 PTR_ADDU "%[dst], %[dst], %[stride] \n\t" \
1991 "bnez %[h], 1b \n\t" \
1992 : RESTRICT_ASM_ALL64 RESTRICT_ASM_ADDRT \
1994 [src]"+r"(src), [dst]"+r"(dst) \
1995 : [stride]"r"(stride), [rnd]"m"(rnd), \
1996 [ff_pw_53]"m"(ff_pw_53), [ff_pw_18]"m"(ff_pw_18), \
1997 [ff_pw_3]"f"(ff_pw_3), [ff_pw_128]"f"(ff_pw_128) \
1998 : "$8", "$9", "$f0", "$f2", "$f4", "$f6", "$f8", "$f10", "$f12", \
1999 "$f14", "$f16", "memory" \
2004 * Macro to build the 8bits, any direction, version of vc1_put_shift[13].
2005 * Here, offset=src_stride. Parameters passed A1 to A4 must use
2006 * %3 (offset), %4 (2*offset) and %5 (3*offset).
2008 * @param NAME Either 1 or 3
2009 * @see MSPEL_FILTER13_CORE for information on A1->A4
2011 #define MSPEL_FILTER13_8B(NAME, A1, A2, A3, A4, OP, OPNAME) \
2013 OPNAME ## vc1_## NAME ## _mmi(uint8_t *dst, const uint8_t *src, \
2014 mips_reg stride, int rnd, mips_reg offset) \
2017 DECLARE_VAR_LOW32; \
2018 DECLARE_VAR_ADDRT; \
2023 __asm__ volatile ( \
2024 "xor $f0, $f0, $f0 \n\t" \
2025 LOAD_ROUNDER_MMI("%[rnd]") \
2026 "ldc1 $f10, %[ff_pw_53] \n\t" \
2027 "ldc1 $f12, %[ff_pw_18] \n\t" \
2030 MSPEL_FILTER13_CORE(DO_UNPACK, MMI_ULWC1, 1, A1, A2, A3, A4) \
2031 "li $8, 0x06 \n\t" \
2032 "mtc1 $8, $f16 \n\t" \
2033 NORMALIZE_MMI("$f16") \
2034 TRANSFER_DO_PACK(OP) \
2035 "addiu %[h], %[h], -0x01 \n\t" \
2036 PTR_ADDU "%[src], %[src], %[stride] \n\t" \
2037 PTR_ADDU "%[dst], %[dst], %[stride] \n\t" \
2038 "bnez %[h], 1b \n\t" \
2039 : RESTRICT_ASM_LOW32 RESTRICT_ASM_ADDRT \
2041 [src]"+r"(src), [dst]"+r"(dst) \
2042 : [offset_x1]"r"(offset), [offset_x2]"r"(2*offset), \
2043 [offset_x3]"r"(3*offset), [stride]"r"(stride), \
2045 [ff_pw_53]"m"(ff_pw_53), [ff_pw_18]"m"(ff_pw_18), \
2046 [ff_pw_3]"f"(ff_pw_3) \
2047 : "$8", "$9", "$f0", "$f2", "$f4", "$f6", "$f8", "$f10", "$f12", \
2048 "$f14", "$f16", "memory" \
2053 /** 1/4 shift bicubic interpolation */
2054 MSPEL_FILTER13_8B(shift1, %[offset_x3], %[offset_x2], %[offset_x1], $0, OP_PUT, put_)
2055 MSPEL_FILTER13_8B(shift1, %[offset_x3], %[offset_x2], %[offset_x1], $0, OP_AVG, avg_)
2056 MSPEL_FILTER13_VER_16B(shift1, %[stride_x3], %[stride_x2], %[stride_x1], $0)
2057 MSPEL_FILTER13_HOR_16B(shift1, 6, 4, 2, 0, OP_PUT, put_)
2058 MSPEL_FILTER13_HOR_16B(shift1, 6, 4, 2, 0, OP_AVG, avg_)
2060 /** 3/4 shift bicubic interpolation */
2061 MSPEL_FILTER13_8B(shift3, $0, %[offset_x1], %[offset_x2], %[offset_x3], OP_PUT, put_)
2062 MSPEL_FILTER13_8B(shift3, $0, %[offset_x1], %[offset_x2], %[offset_x3], OP_AVG, avg_)
2063 MSPEL_FILTER13_VER_16B(shift3, $0, %[stride_x1], %[stride_x2], %[stride_x3])
2064 MSPEL_FILTER13_HOR_16B(shift3, 0, 2, 4, 6, OP_PUT, put_)
2065 MSPEL_FILTER13_HOR_16B(shift3, 0, 2, 4, 6, OP_AVG, avg_)
2067 typedef void (*vc1_mspel_mc_filter_ver_16bits)
2068 (int16_t *dst, const uint8_t *src, mips_reg src_stride, int rnd,
2070 typedef void (*vc1_mspel_mc_filter_hor_16bits)
2071 (uint8_t *dst, mips_reg dst_stride, const int16_t *src, int rnd);
2072 typedef void (*vc1_mspel_mc_filter_8bits)
2073 (uint8_t *dst, const uint8_t *src, mips_reg stride, int rnd,
2077 * Interpolate fractional pel values by applying proper vertical then
2078 * horizontal filter.
2080 * @param dst Destination buffer for interpolated pels.
2081 * @param src Source buffer.
2082 * @param stride Stride for both src and dst buffers.
2083 * @param hmode Horizontal filter (expressed in quarter pixels shift).
2084 * @param hmode Vertical filter.
2085 * @param rnd Rounding bias.
2087 #define VC1_MSPEL_MC(OP) \
2088 static void OP ## vc1_mspel_mc(uint8_t *dst, const uint8_t *src, int stride,\
2089 int hmode, int vmode, int rnd) \
2091 static const vc1_mspel_mc_filter_ver_16bits vc1_put_shift_ver_16bits[] =\
2092 { NULL, vc1_put_ver_16b_shift1_mmi, \
2093 vc1_put_ver_16b_shift2_mmi, \
2094 vc1_put_ver_16b_shift3_mmi }; \
2095 static const vc1_mspel_mc_filter_hor_16bits vc1_put_shift_hor_16bits[] =\
2096 { NULL, OP ## vc1_hor_16b_shift1_mmi, \
2097 OP ## vc1_hor_16b_shift2_mmi, \
2098 OP ## vc1_hor_16b_shift3_mmi }; \
2099 static const vc1_mspel_mc_filter_8bits vc1_put_shift_8bits[] = \
2100 { NULL, OP ## vc1_shift1_mmi, \
2101 OP ## vc1_shift2_mmi, \
2102 OP ## vc1_shift3_mmi }; \
2104 if (vmode) { /* Vertical filter to apply */ \
2105 if (hmode) { /* Horizontal filter to apply, output to tmp */ \
2106 static const int shift_value[] = { 0, 5, 1, 5 }; \
2107 int shift = (shift_value[hmode]+shift_value[vmode])>>1; \
2109 LOCAL_ALIGNED(16, int16_t, tmp, [12*8]); \
2111 r = (1<<(shift-1)) + rnd-1; \
2112 vc1_put_shift_ver_16bits[vmode](tmp, src-1, stride, r, shift); \
2114 vc1_put_shift_hor_16bits[hmode](dst, stride, tmp+1, 64-rnd); \
2117 else { /* No horizontal filter, output 8 lines to dst */ \
2118 vc1_put_shift_8bits[vmode](dst, src, stride, 1-rnd, stride); \
2123 /* Horizontal mode with no vertical mode */ \
2124 vc1_put_shift_8bits[hmode](dst, src, stride, rnd, 1); \
2126 static void OP ## vc1_mspel_mc_16(uint8_t *dst, const uint8_t *src, \
2127 int stride, int hmode, int vmode, int rnd)\
2129 OP ## vc1_mspel_mc(dst + 0, src + 0, stride, hmode, vmode, rnd); \
2130 OP ## vc1_mspel_mc(dst + 8, src + 8, stride, hmode, vmode, rnd); \
2131 dst += 8*stride; src += 8*stride; \
2132 OP ## vc1_mspel_mc(dst + 0, src + 0, stride, hmode, vmode, rnd); \
2133 OP ## vc1_mspel_mc(dst + 8, src + 8, stride, hmode, vmode, rnd); \
2139 /** Macro to ease bicubic filter interpolation functions declarations */
2140 #define DECLARE_FUNCTION(a, b) \
2141 void ff_put_vc1_mspel_mc ## a ## b ## _mmi(uint8_t *dst, \
2142 const uint8_t *src, \
2146 put_vc1_mspel_mc(dst, src, stride, a, b, rnd); \
2148 void ff_avg_vc1_mspel_mc ## a ## b ## _mmi(uint8_t *dst, \
2149 const uint8_t *src, \
2153 avg_vc1_mspel_mc(dst, src, stride, a, b, rnd); \
2155 void ff_put_vc1_mspel_mc ## a ## b ## _16_mmi(uint8_t *dst, \
2156 const uint8_t *src, \
2160 put_vc1_mspel_mc_16(dst, src, stride, a, b, rnd); \
2162 void ff_avg_vc1_mspel_mc ## a ## b ## _16_mmi(uint8_t *dst, \
2163 const uint8_t *src, \
2167 avg_vc1_mspel_mc_16(dst, src, stride, a, b, rnd); \
2170 DECLARE_FUNCTION(0, 1)
2171 DECLARE_FUNCTION(0, 2)
2172 DECLARE_FUNCTION(0, 3)
2174 DECLARE_FUNCTION(1, 0)
2175 DECLARE_FUNCTION(1, 1)
2176 DECLARE_FUNCTION(1, 2)
2177 DECLARE_FUNCTION(1, 3)
2179 DECLARE_FUNCTION(2, 0)
2180 DECLARE_FUNCTION(2, 1)
2181 DECLARE_FUNCTION(2, 2)
2182 DECLARE_FUNCTION(2, 3)
2184 DECLARE_FUNCTION(3, 0)
2185 DECLARE_FUNCTION(3, 1)
2186 DECLARE_FUNCTION(3, 2)
2187 DECLARE_FUNCTION(3, 3)
2189 #define CHROMA_MC_8_MMI \
2190 "punpckhbh %[ftmp5], %[ftmp1], %[ftmp0] \n\t" \
2191 "punpcklbh %[ftmp1], %[ftmp1], %[ftmp0] \n\t" \
2192 "punpckhbh %[ftmp6], %[ftmp2], %[ftmp0] \n\t" \
2193 "punpcklbh %[ftmp2], %[ftmp2], %[ftmp0] \n\t" \
2194 "punpckhbh %[ftmp7], %[ftmp3], %[ftmp0] \n\t" \
2195 "punpcklbh %[ftmp3], %[ftmp3], %[ftmp0] \n\t" \
2196 "punpckhbh %[ftmp8], %[ftmp4], %[ftmp0] \n\t" \
2197 "punpcklbh %[ftmp4], %[ftmp4], %[ftmp0] \n\t" \
2199 "pmullh %[ftmp1], %[ftmp1], %[A] \n\t" \
2200 "pmullh %[ftmp5], %[ftmp5], %[A] \n\t" \
2201 "pmullh %[ftmp2], %[ftmp2], %[B] \n\t" \
2202 "pmullh %[ftmp6], %[ftmp6], %[B] \n\t" \
2203 "pmullh %[ftmp3], %[ftmp3], %[C] \n\t" \
2204 "pmullh %[ftmp7], %[ftmp7], %[C] \n\t" \
2205 "pmullh %[ftmp4], %[ftmp4], %[D] \n\t" \
2206 "pmullh %[ftmp8], %[ftmp8], %[D] \n\t" \
2208 "paddh %[ftmp1], %[ftmp1], %[ftmp2] \n\t" \
2209 "paddh %[ftmp3], %[ftmp3], %[ftmp4] \n\t" \
2210 "paddh %[ftmp1], %[ftmp1], %[ftmp3] \n\t" \
2211 "paddh %[ftmp1], %[ftmp1], %[ff_pw_28] \n\t" \
2213 "paddh %[ftmp5], %[ftmp5], %[ftmp6] \n\t" \
2214 "paddh %[ftmp7], %[ftmp7], %[ftmp8] \n\t" \
2215 "paddh %[ftmp5], %[ftmp5], %[ftmp7] \n\t" \
2216 "paddh %[ftmp5], %[ftmp5], %[ff_pw_28] \n\t" \
2218 "psrlh %[ftmp1], %[ftmp1], %[ftmp9] \n\t" \
2219 "psrlh %[ftmp5], %[ftmp5], %[ftmp9] \n\t" \
2220 "packushb %[ftmp1], %[ftmp1], %[ftmp5] \n\t"
2223 #define CHROMA_MC_4_MMI \
2224 "punpcklbh %[ftmp1], %[ftmp1], %[ftmp0] \n\t" \
2225 "punpcklbh %[ftmp2], %[ftmp2], %[ftmp0] \n\t" \
2226 "punpcklbh %[ftmp3], %[ftmp3], %[ftmp0] \n\t" \
2227 "punpcklbh %[ftmp4], %[ftmp4], %[ftmp0] \n\t" \
2229 "pmullh %[ftmp1], %[ftmp1], %[A] \n\t" \
2230 "pmullh %[ftmp2], %[ftmp2], %[B] \n\t" \
2231 "pmullh %[ftmp3], %[ftmp3], %[C] \n\t" \
2232 "pmullh %[ftmp4], %[ftmp4], %[D] \n\t" \
2234 "paddh %[ftmp1], %[ftmp1], %[ftmp2] \n\t" \
2235 "paddh %[ftmp3], %[ftmp3], %[ftmp4] \n\t" \
2236 "paddh %[ftmp1], %[ftmp1], %[ftmp3] \n\t" \
2237 "paddh %[ftmp1], %[ftmp1], %[ff_pw_28] \n\t" \
2239 "psrlh %[ftmp1], %[ftmp1], %[ftmp5] \n\t" \
2240 "packushb %[ftmp1], %[ftmp1], %[ftmp0] \n\t"
2243 void ff_put_no_rnd_vc1_chroma_mc8_mmi(uint8_t *dst /* align 8 */,
2244 uint8_t *src /* align 1 */,
2245 ptrdiff_t stride, int h, int x, int y)
2247 const int A = (8 - x) * (8 - y);
2248 const int B = (x) * (8 - y);
2249 const int C = (8 - x) * (y);
2250 const int D = (x) * (y);
2256 av_assert2(x < 8 && y < 8 && x >= 0 && y >= 0);
2259 "li %[tmp0], 0x06 \n\t"
2260 "xor %[ftmp0], %[ftmp0], %[ftmp0] \n\t"
2261 "mtc1 %[tmp0], %[ftmp9] \n\t"
2262 "pshufh %[A], %[A], %[ftmp0] \n\t"
2263 "pshufh %[B], %[B], %[ftmp0] \n\t"
2264 "pshufh %[C], %[C], %[ftmp0] \n\t"
2265 "pshufh %[D], %[D], %[ftmp0] \n\t"
2268 MMI_ULDC1(%[ftmp1], %[src], 0x00)
2269 MMI_ULDC1(%[ftmp2], %[src], 0x01)
2270 PTR_ADDU "%[src], %[src], %[stride] \n\t"
2271 MMI_ULDC1(%[ftmp3], %[src], 0x00)
2272 MMI_ULDC1(%[ftmp4], %[src], 0x01)
2276 MMI_SDC1(%[ftmp1], %[dst], 0x00)
2277 "addiu %[h], %[h], -0x01 \n\t"
2278 PTR_ADDU "%[dst], %[dst], %[stride] \n\t"
2279 "bnez %[h], 1b \n\t"
2280 : [ftmp0]"=&f"(ftmp[0]), [ftmp1]"=&f"(ftmp[1]),
2281 [ftmp2]"=&f"(ftmp[2]), [ftmp3]"=&f"(ftmp[3]),
2282 [ftmp4]"=&f"(ftmp[4]), [ftmp5]"=&f"(ftmp[5]),
2283 [ftmp6]"=&f"(ftmp[6]), [ftmp7]"=&f"(ftmp[7]),
2284 [ftmp8]"=&f"(ftmp[8]), [ftmp9]"=&f"(ftmp[9]),
2287 [tmp0]"=&r"(tmp[0]),
2288 [src]"+&r"(src), [dst]"+&r"(dst),
2290 : [stride]"r"((mips_reg)stride),
2291 [A]"f"(A), [B]"f"(B),
2292 [C]"f"(C), [D]"f"(D),
2293 [ff_pw_28]"f"(ff_pw_28)
2298 void ff_put_no_rnd_vc1_chroma_mc4_mmi(uint8_t *dst /* align 8 */,
2299 uint8_t *src /* align 1 */,
2300 ptrdiff_t stride, int h, int x, int y)
2302 const int A = (8 - x) * (8 - y);
2303 const int B = (x) * (8 - y);
2304 const int C = (8 - x) * (y);
2305 const int D = (x) * (y);
2311 av_assert2(x < 8 && y < 8 && x >= 0 && y >= 0);
2314 "li %[tmp0], 0x06 \n\t"
2315 "xor %[ftmp0], %[ftmp0], %[ftmp0] \n\t"
2316 "mtc1 %[tmp0], %[ftmp5] \n\t"
2317 "pshufh %[A], %[A], %[ftmp0] \n\t"
2318 "pshufh %[B], %[B], %[ftmp0] \n\t"
2319 "pshufh %[C], %[C], %[ftmp0] \n\t"
2320 "pshufh %[D], %[D], %[ftmp0] \n\t"
2323 MMI_ULWC1(%[ftmp1], %[src], 0x00)
2324 MMI_ULWC1(%[ftmp2], %[src], 0x01)
2325 PTR_ADDU "%[src], %[src], %[stride] \n\t"
2326 MMI_ULWC1(%[ftmp3], %[src], 0x00)
2327 MMI_ULWC1(%[ftmp4], %[src], 0x01)
2331 MMI_SWC1(%[ftmp1], %[dst], 0x00)
2332 "addiu %[h], %[h], -0x01 \n\t"
2333 PTR_ADDU "%[dst], %[dst], %[stride] \n\t"
2334 "bnez %[h], 1b \n\t"
2335 : [ftmp0]"=&f"(ftmp[0]), [ftmp1]"=&f"(ftmp[1]),
2336 [ftmp2]"=&f"(ftmp[2]), [ftmp3]"=&f"(ftmp[3]),
2337 [ftmp4]"=&f"(ftmp[4]), [ftmp5]"=&f"(ftmp[5]),
2338 [tmp0]"=&r"(tmp[0]),
2341 [src]"+&r"(src), [dst]"+&r"(dst),
2343 : [stride]"r"((mips_reg)stride),
2344 [A]"f"(A), [B]"f"(B),
2345 [C]"f"(C), [D]"f"(D),
2346 [ff_pw_28]"f"(ff_pw_28)
2351 void ff_avg_no_rnd_vc1_chroma_mc8_mmi(uint8_t *dst /* align 8 */,
2352 uint8_t *src /* align 1 */,
2353 ptrdiff_t stride, int h, int x, int y)
2355 const int A = (8 - x) * (8 - y);
2356 const int B = (x) * (8 - y);
2357 const int C = (8 - x) * (y);
2358 const int D = (x) * (y);
2364 av_assert2(x < 8 && y < 8 && x >= 0 && y >= 0);
2367 "li %[tmp0], 0x06 \n\t"
2368 "xor %[ftmp0], %[ftmp0], %[ftmp0] \n\t"
2369 "mtc1 %[tmp0], %[ftmp9] \n\t"
2370 "pshufh %[A], %[A], %[ftmp0] \n\t"
2371 "pshufh %[B], %[B], %[ftmp0] \n\t"
2372 "pshufh %[C], %[C], %[ftmp0] \n\t"
2373 "pshufh %[D], %[D], %[ftmp0] \n\t"
2376 MMI_ULDC1(%[ftmp1], %[src], 0x00)
2377 MMI_ULDC1(%[ftmp2], %[src], 0x01)
2378 PTR_ADDU "%[src], %[src], %[stride] \n\t"
2379 MMI_ULDC1(%[ftmp3], %[src], 0x00)
2380 MMI_ULDC1(%[ftmp4], %[src], 0x01)
2384 MMI_LDC1(%[ftmp2], %[dst], 0x00)
2385 "pavgb %[ftmp1], %[ftmp1], %[ftmp2] \n\t"
2387 MMI_SDC1(%[ftmp1], %[dst], 0x00)
2388 "addiu %[h], %[h], -0x01 \n\t"
2389 PTR_ADDU "%[dst], %[dst], %[stride] \n\t"
2390 "bnez %[h], 1b \n\t"
2391 : [ftmp0]"=&f"(ftmp[0]), [ftmp1]"=&f"(ftmp[1]),
2392 [ftmp2]"=&f"(ftmp[2]), [ftmp3]"=&f"(ftmp[3]),
2393 [ftmp4]"=&f"(ftmp[4]), [ftmp5]"=&f"(ftmp[5]),
2394 [ftmp6]"=&f"(ftmp[6]), [ftmp7]"=&f"(ftmp[7]),
2395 [ftmp8]"=&f"(ftmp[8]), [ftmp9]"=&f"(ftmp[9]),
2396 [tmp0]"=&r"(tmp[0]),
2399 [src]"+&r"(src), [dst]"+&r"(dst),
2401 : [stride]"r"((mips_reg)stride),
2402 [A]"f"(A), [B]"f"(B),
2403 [C]"f"(C), [D]"f"(D),
2404 [ff_pw_28]"f"(ff_pw_28)
2409 void ff_avg_no_rnd_vc1_chroma_mc4_mmi(uint8_t *dst /* align 8 */,
2410 uint8_t *src /* align 1 */,
2411 ptrdiff_t stride, int h, int x, int y)
2413 const int A = (8 - x) * (8 - y);
2414 const int B = ( x) * (8 - y);
2415 const int C = (8 - x) * ( y);
2416 const int D = ( x) * ( y);
2422 av_assert2(x < 8 && y < 8 && x >= 0 && y >= 0);
2425 "li %[tmp0], 0x06 \n\t"
2426 "xor %[ftmp0], %[ftmp0], %[ftmp0] \n\t"
2427 "mtc1 %[tmp0], %[ftmp5] \n\t"
2428 "pshufh %[A], %[A], %[ftmp0] \n\t"
2429 "pshufh %[B], %[B], %[ftmp0] \n\t"
2430 "pshufh %[C], %[C], %[ftmp0] \n\t"
2431 "pshufh %[D], %[D], %[ftmp0] \n\t"
2434 MMI_ULWC1(%[ftmp1], %[src], 0x00)
2435 MMI_ULWC1(%[ftmp2], %[src], 0x01)
2436 PTR_ADDU "%[src], %[src], %[stride] \n\t"
2437 MMI_ULWC1(%[ftmp3], %[src], 0x00)
2438 MMI_ULWC1(%[ftmp4], %[src], 0x01)
2442 MMI_LWC1(%[ftmp2], %[dst], 0x00)
2443 "pavgb %[ftmp1], %[ftmp1], %[ftmp2] \n\t"
2445 MMI_SWC1(%[ftmp1], %[dst], 0x00)
2446 "addiu %[h], %[h], -0x01 \n\t"
2447 PTR_ADDU "%[dst], %[dst], %[stride] \n\t"
2448 "bnez %[h], 1b \n\t"
2449 : [ftmp0]"=&f"(ftmp[0]), [ftmp1]"=&f"(ftmp[1]),
2450 [ftmp2]"=&f"(ftmp[2]), [ftmp3]"=&f"(ftmp[3]),
2451 [ftmp4]"=&f"(ftmp[4]), [ftmp5]"=&f"(ftmp[5]),
2452 [tmp0]"=&r"(tmp[0]),
2455 [src]"+&r"(src), [dst]"+&r"(dst),
2457 : [stride]"r"((mips_reg)stride),
2458 [A]"f"(A), [B]"f"(B),
2459 [C]"f"(C), [D]"f"(D),
2460 [ff_pw_28]"f"(ff_pw_28)