FFmpeg
Loading...
Searching...
No Matches
hevcdsp_mmi.c
Go to the documentation of this file.
1/*
2 * Copyright (c) 2019 Shiyou Yin (yinshiyou-hf@loongson.cn)
3 *
4 * This file is part of FFmpeg.
5 *
6 * FFmpeg is free software; you can redistribute it and/or
7 * modify it under the terms of the GNU Lesser General Public
8 * License as published by the Free Software Foundation; either
9 * version 2.1 of the License, or (at your option) any later version.
10 *
11 * FFmpeg is distributed in the hope that it will be useful,
12 * but WITHOUT ANY WARRANTY; without even the implied warranty of
13 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
14 * Lesser General Public License for more details.
15 *
16 * You should have received a copy of the GNU Lesser General Public
17 * License along with FFmpeg; if not, write to the Free Software
18 * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
19 */
20
21#include "libavcodec/hevc/hevcdec.h"
22#include "libavcodec/bit_depth_template.c"
23#include "libavcodec/mips/hevcdsp_mips.h"
24#include "libavutil/intfloat.h"
25#include "libavutil/mips/mmiutils.h"
26
27#define PUT_HEVC_QPEL_H(w, x_step, src_step, dst_step) \
28void ff_hevc_put_hevc_qpel_h##w##_8_mmi(int16_t *dst, const uint8_t *_src, \
29 ptrdiff_t _srcstride, \
30 int height, intptr_t mx, \
31 intptr_t my, int width) \
32{ \
33 int x, y; \
34 const pixel *src = (const pixel*)_src - 3; \
35 ptrdiff_t srcstride = _srcstride / sizeof(pixel); \
36 double ftmp[15]; \
37 uint64_t rtmp[1]; \
38 const int8_t *filter = ff_hevc_qpel_filters[mx]; \
39 DECLARE_VAR_ALL64; \
40 \
41 x = x_step; \
42 y = height; \
43 __asm__ volatile( \
44 MMI_LDC1(%[ftmp1], %[filter], 0x00) \
45 "li %[rtmp0], 0x08 \n\t" \
46 "dmtc1 %[rtmp0], %[ftmp0] \n\t" \
47 "punpckhbh %[ftmp2], %[ftmp0], %[ftmp1] \n\t" \
48 "punpcklbh %[ftmp1], %[ftmp0], %[ftmp1] \n\t" \
49 "psrah %[ftmp1], %[ftmp1], %[ftmp0] \n\t" \
50 "psrah %[ftmp2], %[ftmp2], %[ftmp0] \n\t" \
51 "pxor %[ftmp0], %[ftmp0], %[ftmp0] \n\t" \
52 \
53 "1: \n\t" \
54 "2: \n\t" \
55 MMI_ULDC1(%[ftmp3], %[src], 0x00) \
56 MMI_ULDC1(%[ftmp4], %[src], 0x01) \
57 MMI_ULDC1(%[ftmp5], %[src], 0x02) \
58 MMI_ULDC1(%[ftmp6], %[src], 0x03) \
59 "punpcklbh %[ftmp7], %[ftmp3], %[ftmp0] \n\t" \
60 "punpckhbh %[ftmp8], %[ftmp3], %[ftmp0] \n\t" \
61 "pmullh %[ftmp7], %[ftmp7], %[ftmp1] \n\t" \
62 "pmullh %[ftmp8], %[ftmp8], %[ftmp2] \n\t" \
63 "paddh %[ftmp3], %[ftmp7], %[ftmp8] \n\t" \
64 "punpcklbh %[ftmp7], %[ftmp4], %[ftmp0] \n\t" \
65 "punpckhbh %[ftmp8], %[ftmp4], %[ftmp0] \n\t" \
66 "pmullh %[ftmp7], %[ftmp7], %[ftmp1] \n\t" \
67 "pmullh %[ftmp8], %[ftmp8], %[ftmp2] \n\t" \
68 "paddh %[ftmp4], %[ftmp7], %[ftmp8] \n\t" \
69 "punpcklbh %[ftmp7], %[ftmp5], %[ftmp0] \n\t" \
70 "punpckhbh %[ftmp8], %[ftmp5], %[ftmp0] \n\t" \
71 "pmullh %[ftmp7], %[ftmp7], %[ftmp1] \n\t" \
72 "pmullh %[ftmp8], %[ftmp8], %[ftmp2] \n\t" \
73 "paddh %[ftmp5], %[ftmp7], %[ftmp8] \n\t" \
74 "punpcklbh %[ftmp7], %[ftmp6], %[ftmp0] \n\t" \
75 "punpckhbh %[ftmp8], %[ftmp6], %[ftmp0] \n\t" \
76 "pmullh %[ftmp7], %[ftmp7], %[ftmp1] \n\t" \
77 "pmullh %[ftmp8], %[ftmp8], %[ftmp2] \n\t" \
78 "paddh %[ftmp6], %[ftmp7], %[ftmp8] \n\t" \
79 TRANSPOSE_4H(%[ftmp3], %[ftmp4], %[ftmp5], %[ftmp6], \
80 %[ftmp7], %[ftmp8], %[ftmp9], %[ftmp10]) \
81 "paddh %[ftmp3], %[ftmp3], %[ftmp4] \n\t" \
82 "paddh %[ftmp5], %[ftmp5], %[ftmp6] \n\t" \
83 "paddh %[ftmp3], %[ftmp3], %[ftmp5] \n\t" \
84 MMI_USDC1(%[ftmp3], %[dst], 0x00) \
85 \
86 "daddi %[x], %[x], -0x01 \n\t" \
87 PTR_ADDIU "%[src], %[src], 0x04 \n\t" \
88 PTR_ADDIU "%[dst], %[dst], 0x08 \n\t" \
89 "bnez %[x], 2b \n\t" \
90 \
91 "daddi %[y], %[y], -0x01 \n\t" \
92 "li %[x], " #x_step " \n\t" \
93 PTR_ADDIU "%[src], %[src], " #src_step " \n\t" \
94 PTR_ADDIU "%[dst], %[dst], " #dst_step " \n\t" \
95 PTR_ADDU "%[src], %[src], %[stride] \n\t" \
96 PTR_ADDIU "%[dst], %[dst], 0x80 \n\t" \
97 "bnez %[y], 1b \n\t" \
98 : RESTRICT_ASM_ALL64 \
99 [ftmp0]"=&f"(ftmp[0]), [ftmp1]"=&f"(ftmp[1]), \
100 [ftmp2]"=&f"(ftmp[2]), [ftmp3]"=&f"(ftmp[3]), \
101 [ftmp4]"=&f"(ftmp[4]), [ftmp5]"=&f"(ftmp[5]), \
102 [ftmp6]"=&f"(ftmp[6]), [ftmp7]"=&f"(ftmp[7]), \
103 [ftmp8]"=&f"(ftmp[8]), [ftmp9]"=&f"(ftmp[9]), \
104 [ftmp10]"=&f"(ftmp[10]), [rtmp0]"=&r"(rtmp[0]), \
105 [src]"+&r"(src), [dst]"+&r"(dst), [y]"+&r"(y), \
106 [x]"+&r"(x) \
107 : [filter]"r"(filter), [stride]"r"(srcstride) \
108 : "memory" \
109 ); \
110}
111
112PUT_HEVC_QPEL_H(4, 1, -4, -8);
113PUT_HEVC_QPEL_H(8, 2, -8, -16);
114PUT_HEVC_QPEL_H(12, 3, -12, -24);
115PUT_HEVC_QPEL_H(16, 4, -16, -32);
116PUT_HEVC_QPEL_H(24, 6, -24, -48);
117PUT_HEVC_QPEL_H(32, 8, -32, -64);
118PUT_HEVC_QPEL_H(48, 12, -48, -96);
119PUT_HEVC_QPEL_H(64, 16, -64, -128);
120
121#define PUT_HEVC_QPEL_HV(w, x_step, src_step, dst_step) \
122void ff_hevc_put_hevc_qpel_hv##w##_8_mmi(int16_t *dst, const uint8_t *_src,\
123 ptrdiff_t _srcstride, \
124 int height, intptr_t mx, \
125 intptr_t my, int width) \
126{ \
127 int x, y; \
128 const int8_t *filter; \
129 const pixel *src = (const pixel*)_src; \
130 ptrdiff_t srcstride = _srcstride / sizeof(pixel); \
131 int16_t tmp_array[(MAX_PB_SIZE + QPEL_EXTRA) * MAX_PB_SIZE]; \
132 int16_t *tmp = tmp_array; \
133 double ftmp[15]; \
134 uint64_t rtmp[1]; \
135 DECLARE_VAR_ALL64; \
136 \
137 src -= (QPEL_EXTRA_BEFORE * srcstride + 3); \
138 filter = ff_hevc_qpel_filters[mx]; \
139 x = x_step; \
140 y = height + QPEL_EXTRA; \
141 __asm__ volatile( \
142 MMI_LDC1(%[ftmp1], %[filter], 0x00) \
143 "li %[rtmp0], 0x08 \n\t" \
144 "dmtc1 %[rtmp0], %[ftmp0] \n\t" \
145 "punpckhbh %[ftmp2], %[ftmp0], %[ftmp1] \n\t" \
146 "punpcklbh %[ftmp1], %[ftmp0], %[ftmp1] \n\t" \
147 "psrah %[ftmp1], %[ftmp1], %[ftmp0] \n\t" \
148 "psrah %[ftmp2], %[ftmp2], %[ftmp0] \n\t" \
149 "pxor %[ftmp0], %[ftmp0], %[ftmp0] \n\t" \
150 \
151 "1: \n\t" \
152 "2: \n\t" \
153 MMI_ULDC1(%[ftmp3], %[src], 0x00) \
154 MMI_ULDC1(%[ftmp4], %[src], 0x01) \
155 MMI_ULDC1(%[ftmp5], %[src], 0x02) \
156 MMI_ULDC1(%[ftmp6], %[src], 0x03) \
157 "punpcklbh %[ftmp7], %[ftmp3], %[ftmp0] \n\t" \
158 "punpckhbh %[ftmp8], %[ftmp3], %[ftmp0] \n\t" \
159 "pmullh %[ftmp7], %[ftmp7], %[ftmp1] \n\t" \
160 "pmullh %[ftmp8], %[ftmp8], %[ftmp2] \n\t" \
161 "paddh %[ftmp3], %[ftmp7], %[ftmp8] \n\t" \
162 "punpcklbh %[ftmp7], %[ftmp4], %[ftmp0] \n\t" \
163 "punpckhbh %[ftmp8], %[ftmp4], %[ftmp0] \n\t" \
164 "pmullh %[ftmp7], %[ftmp7], %[ftmp1] \n\t" \
165 "pmullh %[ftmp8], %[ftmp8], %[ftmp2] \n\t" \
166 "paddh %[ftmp4], %[ftmp7], %[ftmp8] \n\t" \
167 "punpcklbh %[ftmp7], %[ftmp5], %[ftmp0] \n\t" \
168 "punpckhbh %[ftmp8], %[ftmp5], %[ftmp0] \n\t" \
169 "pmullh %[ftmp7], %[ftmp7], %[ftmp1] \n\t" \
170 "pmullh %[ftmp8], %[ftmp8], %[ftmp2] \n\t" \
171 "paddh %[ftmp5], %[ftmp7], %[ftmp8] \n\t" \
172 "punpcklbh %[ftmp7], %[ftmp6], %[ftmp0] \n\t" \
173 "punpckhbh %[ftmp8], %[ftmp6], %[ftmp0] \n\t" \
174 "pmullh %[ftmp7], %[ftmp7], %[ftmp1] \n\t" \
175 "pmullh %[ftmp8], %[ftmp8], %[ftmp2] \n\t" \
176 "paddh %[ftmp6], %[ftmp7], %[ftmp8] \n\t" \
177 TRANSPOSE_4H(%[ftmp3], %[ftmp4], %[ftmp5], %[ftmp6], \
178 %[ftmp7], %[ftmp8], %[ftmp9], %[ftmp10]) \
179 "paddh %[ftmp3], %[ftmp3], %[ftmp4] \n\t" \
180 "paddh %[ftmp5], %[ftmp5], %[ftmp6] \n\t" \
181 "paddh %[ftmp3], %[ftmp3], %[ftmp5] \n\t" \
182 MMI_USDC1(%[ftmp3], %[tmp], 0x00) \
183 \
184 "daddi %[x], %[x], -0x01 \n\t" \
185 PTR_ADDIU "%[src], %[src], 0x04 \n\t" \
186 PTR_ADDIU "%[tmp], %[tmp], 0x08 \n\t" \
187 "bnez %[x], 2b \n\t" \
188 \
189 "daddi %[y], %[y], -0x01 \n\t" \
190 "li %[x], " #x_step " \n\t" \
191 PTR_ADDIU "%[src], %[src], " #src_step " \n\t" \
192 PTR_ADDIU "%[tmp], %[tmp], " #dst_step " \n\t" \
193 PTR_ADDU "%[src], %[src], %[stride] \n\t" \
194 PTR_ADDIU "%[tmp], %[tmp], 0x80 \n\t" \
195 "bnez %[y], 1b \n\t" \
196 : RESTRICT_ASM_ALL64 \
197 [ftmp0]"=&f"(ftmp[0]), [ftmp1]"=&f"(ftmp[1]), \
198 [ftmp2]"=&f"(ftmp[2]), [ftmp3]"=&f"(ftmp[3]), \
199 [ftmp4]"=&f"(ftmp[4]), [ftmp5]"=&f"(ftmp[5]), \
200 [ftmp6]"=&f"(ftmp[6]), [ftmp7]"=&f"(ftmp[7]), \
201 [ftmp8]"=&f"(ftmp[8]), [ftmp9]"=&f"(ftmp[9]), \
202 [ftmp10]"=&f"(ftmp[10]), [rtmp0]"=&r"(rtmp[0]), \
203 [src]"+&r"(src), [tmp]"+&r"(tmp), [y]"+&r"(y), \
204 [x]"+&r"(x) \
205 : [filter]"r"(filter), [stride]"r"(srcstride) \
206 : "memory" \
207 ); \
208 \
209 tmp = tmp_array + QPEL_EXTRA_BEFORE * 4 -12; \
210 filter = ff_hevc_qpel_filters[my]; \
211 x = x_step; \
212 y = height; \
213 __asm__ volatile( \
214 MMI_LDC1(%[ftmp1], %[filter], 0x00) \
215 "li %[rtmp0], 0x08 \n\t" \
216 "dmtc1 %[rtmp0], %[ftmp0] \n\t" \
217 "punpckhbh %[ftmp2], %[ftmp0], %[ftmp1] \n\t" \
218 "punpcklbh %[ftmp1], %[ftmp0], %[ftmp1] \n\t" \
219 "psrah %[ftmp1], %[ftmp1], %[ftmp0] \n\t" \
220 "psrah %[ftmp2], %[ftmp2], %[ftmp0] \n\t" \
221 "li %[rtmp0], 0x06 \n\t" \
222 "dmtc1 %[rtmp0], %[ftmp0] \n\t" \
223 \
224 "1: \n\t" \
225 "2: \n\t" \
226 MMI_ULDC1(%[ftmp3], %[tmp], 0x00) \
227 PTR_ADDIU "%[tmp], %[tmp], 0x80 \n\t" \
228 MMI_ULDC1(%[ftmp4], %[tmp], 0x00) \
229 PTR_ADDIU "%[tmp], %[tmp], 0x80 \n\t" \
230 MMI_ULDC1(%[ftmp5], %[tmp], 0x00) \
231 PTR_ADDIU "%[tmp], %[tmp], 0x80 \n\t" \
232 MMI_ULDC1(%[ftmp6], %[tmp], 0x00) \
233 PTR_ADDIU "%[tmp], %[tmp], 0x80 \n\t" \
234 MMI_ULDC1(%[ftmp7], %[tmp], 0x00) \
235 PTR_ADDIU "%[tmp], %[tmp], 0x80 \n\t" \
236 MMI_ULDC1(%[ftmp8], %[tmp], 0x00) \
237 PTR_ADDIU "%[tmp], %[tmp], 0x80 \n\t" \
238 MMI_ULDC1(%[ftmp9], %[tmp], 0x00) \
239 PTR_ADDIU "%[tmp], %[tmp], 0x80 \n\t" \
240 MMI_ULDC1(%[ftmp10], %[tmp], 0x00) \
241 PTR_ADDIU "%[tmp], %[tmp], -0x380 \n\t" \
242 TRANSPOSE_4H(%[ftmp3], %[ftmp4], %[ftmp5], %[ftmp6], \
243 %[ftmp11], %[ftmp12], %[ftmp13], %[ftmp14]) \
244 TRANSPOSE_4H(%[ftmp7], %[ftmp8], %[ftmp9], %[ftmp10], \
245 %[ftmp11], %[ftmp12], %[ftmp13], %[ftmp14]) \
246 "pmaddhw %[ftmp11], %[ftmp3], %[ftmp1] \n\t" \
247 "pmaddhw %[ftmp12], %[ftmp7], %[ftmp2] \n\t" \
248 "pmaddhw %[ftmp13], %[ftmp4], %[ftmp1] \n\t" \
249 "pmaddhw %[ftmp14], %[ftmp8], %[ftmp2] \n\t" \
250 "paddw %[ftmp11], %[ftmp11], %[ftmp12] \n\t" \
251 "paddw %[ftmp13], %[ftmp13], %[ftmp14] \n\t" \
252 TRANSPOSE_2W(%[ftmp11], %[ftmp13], %[ftmp3], %[ftmp4]) \
253 "paddw %[ftmp3], %[ftmp3], %[ftmp4] \n\t" \
254 "psraw %[ftmp3], %[ftmp3], %[ftmp0] \n\t" \
255 "pmaddhw %[ftmp11], %[ftmp5], %[ftmp1] \n\t" \
256 "pmaddhw %[ftmp12], %[ftmp9], %[ftmp2] \n\t" \
257 "pmaddhw %[ftmp13], %[ftmp6], %[ftmp1] \n\t" \
258 "pmaddhw %[ftmp14], %[ftmp10], %[ftmp2] \n\t" \
259 "paddw %[ftmp11], %[ftmp11], %[ftmp12] \n\t" \
260 "paddw %[ftmp13], %[ftmp13], %[ftmp14] \n\t" \
261 TRANSPOSE_2W(%[ftmp11], %[ftmp13], %[ftmp5], %[ftmp6]) \
262 "paddw %[ftmp5], %[ftmp5], %[ftmp6] \n\t" \
263 "psraw %[ftmp5], %[ftmp5], %[ftmp0] \n\t" \
264 "packsswh %[ftmp3], %[ftmp3], %[ftmp5] \n\t" \
265 MMI_USDC1(%[ftmp3], %[dst], 0x00) \
266 \
267 "daddi %[x], %[x], -0x01 \n\t" \
268 PTR_ADDIU "%[dst], %[dst], 0x08 \n\t" \
269 PTR_ADDIU "%[tmp], %[tmp], 0x08 \n\t" \
270 "bnez %[x], 2b \n\t" \
271 \
272 "daddi %[y], %[y], -0x01 \n\t" \
273 "li %[x], " #x_step " \n\t" \
274 PTR_ADDIU "%[dst], %[dst], " #dst_step " \n\t" \
275 PTR_ADDIU "%[tmp], %[tmp], " #dst_step " \n\t" \
276 PTR_ADDIU "%[dst], %[dst], 0x80 \n\t" \
277 PTR_ADDIU "%[tmp], %[tmp], 0x80 \n\t" \
278 "bnez %[y], 1b \n\t" \
279 : RESTRICT_ASM_ALL64 \
280 [ftmp0]"=&f"(ftmp[0]), [ftmp1]"=&f"(ftmp[1]), \
281 [ftmp2]"=&f"(ftmp[2]), [ftmp3]"=&f"(ftmp[3]), \
282 [ftmp4]"=&f"(ftmp[4]), [ftmp5]"=&f"(ftmp[5]), \
283 [ftmp6]"=&f"(ftmp[6]), [ftmp7]"=&f"(ftmp[7]), \
284 [ftmp8]"=&f"(ftmp[8]), [ftmp9]"=&f"(ftmp[9]), \
285 [ftmp10]"=&f"(ftmp[10]), [ftmp11]"=&f"(ftmp[11]), \
286 [ftmp12]"=&f"(ftmp[12]), [ftmp13]"=&f"(ftmp[13]), \
287 [ftmp14]"=&f"(ftmp[14]), [rtmp0]"=&r"(rtmp[0]), \
288 [dst]"+&r"(dst), [tmp]"+&r"(tmp), [y]"+&r"(y), \
289 [x]"+&r"(x) \
290 : [filter]"r"(filter), [stride]"r"(srcstride) \
291 : "memory" \
292 ); \
293}
294
295PUT_HEVC_QPEL_HV(4, 1, -4, -8);
296PUT_HEVC_QPEL_HV(8, 2, -8, -16);
297PUT_HEVC_QPEL_HV(12, 3, -12, -24);
298PUT_HEVC_QPEL_HV(16, 4, -16, -32);
299PUT_HEVC_QPEL_HV(24, 6, -24, -48);
300PUT_HEVC_QPEL_HV(32, 8, -32, -64);
301PUT_HEVC_QPEL_HV(48, 12, -48, -96);
302PUT_HEVC_QPEL_HV(64, 16, -64, -128);
303
304#define PUT_HEVC_QPEL_BI_H(w, x_step, src_step, src2_step, dst_step) \
305void ff_hevc_put_hevc_qpel_bi_h##w##_8_mmi(uint8_t *_dst, \
306 ptrdiff_t _dststride, \
307 const uint8_t *_src, \
308 ptrdiff_t _srcstride, \
309 const int16_t *src2, int height, \
310 intptr_t mx, intptr_t my, \
311 int width) \
312{ \
313 int x, y; \
314 const pixel *src = (const pixel*)_src - 3; \
315 ptrdiff_t srcstride = _srcstride / sizeof(pixel); \
316 pixel *dst = (pixel *)_dst; \
317 ptrdiff_t dststride = _dststride / sizeof(pixel); \
318 const int8_t *filter = ff_hevc_qpel_filters[mx]; \
319 double ftmp[20]; \
320 uint64_t rtmp[1]; \
321 union av_intfloat64 shift; \
322 union av_intfloat64 offset; \
323 DECLARE_VAR_ALL64; \
324 DECLARE_VAR_LOW32; \
325 shift.i = 7; \
326 offset.i = 64; \
327 \
328 x = width >> 2; \
329 y = height; \
330 __asm__ volatile( \
331 MMI_LDC1(%[ftmp1], %[filter], 0x00) \
332 "li %[rtmp0], 0x08 \n\t" \
333 "dmtc1 %[rtmp0], %[ftmp0] \n\t" \
334 "punpckhbh %[ftmp2], %[ftmp0], %[ftmp1] \n\t" \
335 "punpcklbh %[ftmp1], %[ftmp0], %[ftmp1] \n\t" \
336 "psrah %[ftmp1], %[ftmp1], %[ftmp0] \n\t" \
337 "psrah %[ftmp2], %[ftmp2], %[ftmp0] \n\t" \
338 "pxor %[ftmp0], %[ftmp0], %[ftmp0] \n\t" \
339 "punpcklhw %[offset], %[offset], %[offset] \n\t" \
340 "punpcklwd %[offset], %[offset], %[offset] \n\t" \
341 \
342 "1: \n\t" \
343 "li %[x], " #x_step " \n\t" \
344 "2: \n\t" \
345 MMI_ULDC1(%[ftmp3], %[src], 0x00) \
346 MMI_ULDC1(%[ftmp4], %[src], 0x01) \
347 MMI_ULDC1(%[ftmp5], %[src], 0x02) \
348 MMI_ULDC1(%[ftmp6], %[src], 0x03) \
349 "punpcklbh %[ftmp7], %[ftmp3], %[ftmp0] \n\t" \
350 "punpckhbh %[ftmp8], %[ftmp3], %[ftmp0] \n\t" \
351 "pmullh %[ftmp7], %[ftmp7], %[ftmp1] \n\t" \
352 "pmullh %[ftmp8], %[ftmp8], %[ftmp2] \n\t" \
353 "paddh %[ftmp3], %[ftmp7], %[ftmp8] \n\t" \
354 "punpcklbh %[ftmp7], %[ftmp4], %[ftmp0] \n\t" \
355 "punpckhbh %[ftmp8], %[ftmp4], %[ftmp0] \n\t" \
356 "pmullh %[ftmp7], %[ftmp7], %[ftmp1] \n\t" \
357 "pmullh %[ftmp8], %[ftmp8], %[ftmp2] \n\t" \
358 "paddh %[ftmp4], %[ftmp7], %[ftmp8] \n\t" \
359 "punpcklbh %[ftmp7], %[ftmp5], %[ftmp0] \n\t" \
360 "punpckhbh %[ftmp8], %[ftmp5], %[ftmp0] \n\t" \
361 "pmullh %[ftmp7], %[ftmp7], %[ftmp1] \n\t" \
362 "pmullh %[ftmp8], %[ftmp8], %[ftmp2] \n\t" \
363 "paddh %[ftmp5], %[ftmp7], %[ftmp8] \n\t" \
364 "punpcklbh %[ftmp7], %[ftmp6], %[ftmp0] \n\t" \
365 "punpckhbh %[ftmp8], %[ftmp6], %[ftmp0] \n\t" \
366 "pmullh %[ftmp7], %[ftmp7], %[ftmp1] \n\t" \
367 "pmullh %[ftmp8], %[ftmp8], %[ftmp2] \n\t" \
368 "paddh %[ftmp6], %[ftmp7], %[ftmp8] \n\t" \
369 TRANSPOSE_4H(%[ftmp3], %[ftmp4], %[ftmp5], %[ftmp6], \
370 %[ftmp7], %[ftmp8], %[ftmp9], %[ftmp10]) \
371 "paddh %[ftmp3], %[ftmp3], %[ftmp4] \n\t" \
372 "paddh %[ftmp5], %[ftmp5], %[ftmp6] \n\t" \
373 "paddh %[ftmp3], %[ftmp3], %[ftmp5] \n\t" \
374 "paddh %[ftmp3], %[ftmp3], %[offset] \n\t" \
375 MMI_ULDC1(%[ftmp4], %[src2], 0x00) \
376 "li %[rtmp0], 0x10 \n\t" \
377 "dmtc1 %[rtmp0], %[ftmp8] \n\t" \
378 "punpcklhw %[ftmp5], %[ftmp0], %[ftmp3] \n\t" \
379 "punpckhhw %[ftmp6], %[ftmp0], %[ftmp3] \n\t" \
380 "punpckhhw %[ftmp3], %[ftmp0], %[ftmp4] \n\t" \
381 "punpcklhw %[ftmp4], %[ftmp0], %[ftmp4] \n\t" \
382 "psraw %[ftmp5], %[ftmp5], %[ftmp8] \n\t" \
383 "psraw %[ftmp6], %[ftmp6], %[ftmp8] \n\t" \
384 "psraw %[ftmp3], %[ftmp3], %[ftmp8] \n\t" \
385 "psraw %[ftmp4], %[ftmp4], %[ftmp8] \n\t" \
386 "paddw %[ftmp5], %[ftmp5], %[ftmp4] \n\t" \
387 "paddw %[ftmp6], %[ftmp6], %[ftmp3] \n\t" \
388 "psraw %[ftmp5], %[ftmp5], %[shift] \n\t" \
389 "psraw %[ftmp6], %[ftmp6], %[shift] \n\t" \
390 "packsswh %[ftmp5], %[ftmp5], %[ftmp6] \n\t" \
391 "pcmpgth %[ftmp7], %[ftmp5], %[ftmp0] \n\t" \
392 "pand %[ftmp3], %[ftmp5], %[ftmp7] \n\t" \
393 "packushb %[ftmp3], %[ftmp3], %[ftmp3] \n\t" \
394 MMI_USWC1(%[ftmp3], %[dst], 0x00) \
395 \
396 "daddi %[x], %[x], -0x01 \n\t" \
397 PTR_ADDIU "%[src], %[src], 0x04 \n\t" \
398 PTR_ADDIU "%[dst], %[dst], 0x04 \n\t" \
399 PTR_ADDIU "%[src2], %[src2], 0x08 \n\t" \
400 "bnez %[x], 2b \n\t" \
401 \
402 "daddi %[y], %[y], -0x01 \n\t" \
403 PTR_ADDIU "%[src], %[src], " #src_step " \n\t" \
404 PTR_ADDIU "%[dst], %[dst], " #dst_step " \n\t" \
405 PTR_ADDIU "%[src2], %[src2], " #src2_step " \n\t" \
406 PTR_ADDU "%[src], %[src], %[src_stride] \n\t" \
407 PTR_ADDU "%[dst], %[dst], %[dst_stride] \n\t" \
408 PTR_ADDIU "%[src2], %[src2], 0x80 \n\t" \
409 "bnez %[y], 1b \n\t" \
410 : RESTRICT_ASM_ALL64 RESTRICT_ASM_LOW32 \
411 [ftmp0]"=&f"(ftmp[0]), [ftmp1]"=&f"(ftmp[1]), \
412 [ftmp2]"=&f"(ftmp[2]), [ftmp3]"=&f"(ftmp[3]), \
413 [ftmp4]"=&f"(ftmp[4]), [ftmp5]"=&f"(ftmp[5]), \
414 [ftmp6]"=&f"(ftmp[6]), [ftmp7]"=&f"(ftmp[7]), \
415 [ftmp8]"=&f"(ftmp[8]), [ftmp9]"=&f"(ftmp[9]), \
416 [ftmp10]"=&f"(ftmp[10]), [ftmp11]"=&f"(ftmp[11]), \
417 [ftmp12]"=&f"(ftmp[12]), [src2]"+&r"(src2), \
418 [dst]"+&r"(dst), [src]"+&r"(src), [y]"+&r"(y), [x]"=&r"(x), \
419 [offset]"+&f"(offset.f), [rtmp0]"=&r"(rtmp[0]) \
420 : [src_stride]"r"(srcstride), [dst_stride]"r"(dststride), \
421 [filter]"r"(filter), [shift]"f"(shift.f) \
422 : "memory" \
423 ); \
424}
425
426PUT_HEVC_QPEL_BI_H(4, 1, -4, -8, -4);
427PUT_HEVC_QPEL_BI_H(8, 2, -8, -16, -8);
428PUT_HEVC_QPEL_BI_H(12, 3, -12, -24, -12);
429PUT_HEVC_QPEL_BI_H(16, 4, -16, -32, -16);
430PUT_HEVC_QPEL_BI_H(24, 6, -24, -48, -24);
431PUT_HEVC_QPEL_BI_H(32, 8, -32, -64, -32);
432PUT_HEVC_QPEL_BI_H(48, 12, -48, -96, -48);
433PUT_HEVC_QPEL_BI_H(64, 16, -64, -128, -64);
434
435#define PUT_HEVC_QPEL_BI_HV(w, x_step, src_step, src2_step, dst_step) \
436void ff_hevc_put_hevc_qpel_bi_hv##w##_8_mmi(uint8_t *_dst, \
437 ptrdiff_t _dststride, \
438 const uint8_t *_src, \
439 ptrdiff_t _srcstride, \
440 const int16_t *src2, int height, \
441 intptr_t mx, intptr_t my, \
442 int width) \
443{ \
444 int x, y; \
445 const int8_t *filter; \
446 pixel *src = (pixel*)_src; \
447 ptrdiff_t srcstride = _srcstride / sizeof(pixel); \
448 pixel *dst = (pixel *)_dst; \
449 ptrdiff_t dststride = _dststride / sizeof(pixel); \
450 int16_t tmp_array[(MAX_PB_SIZE + QPEL_EXTRA) * MAX_PB_SIZE]; \
451 int16_t *tmp = tmp_array; \
452 double ftmp[20]; \
453 uint64_t rtmp[1]; \
454 union av_intfloat64 shift; \
455 union av_intfloat64 offset; \
456 DECLARE_VAR_ALL64; \
457 DECLARE_VAR_LOW32; \
458 shift.i = 7; \
459 offset.i = 64; \
460 \
461 src -= (QPEL_EXTRA_BEFORE * srcstride + 3); \
462 filter = ff_hevc_qpel_filters[mx]; \
463 x = width >> 2; \
464 y = height + QPEL_EXTRA; \
465 __asm__ volatile( \
466 MMI_LDC1(%[ftmp1], %[filter], 0x00) \
467 "li %[rtmp0], 0x08 \n\t" \
468 "dmtc1 %[rtmp0], %[ftmp0] \n\t" \
469 "punpckhbh %[ftmp2], %[ftmp0], %[ftmp1] \n\t" \
470 "punpcklbh %[ftmp1], %[ftmp0], %[ftmp1] \n\t" \
471 "psrah %[ftmp1], %[ftmp1], %[ftmp0] \n\t" \
472 "psrah %[ftmp2], %[ftmp2], %[ftmp0] \n\t" \
473 "pxor %[ftmp0], %[ftmp0], %[ftmp0] \n\t" \
474 \
475 "1: \n\t" \
476 "2: \n\t" \
477 MMI_ULDC1(%[ftmp3], %[src], 0x00) \
478 MMI_ULDC1(%[ftmp4], %[src], 0x01) \
479 MMI_ULDC1(%[ftmp5], %[src], 0x02) \
480 MMI_ULDC1(%[ftmp6], %[src], 0x03) \
481 "punpcklbh %[ftmp7], %[ftmp3], %[ftmp0] \n\t" \
482 "punpckhbh %[ftmp8], %[ftmp3], %[ftmp0] \n\t" \
483 "pmullh %[ftmp7], %[ftmp7], %[ftmp1] \n\t" \
484 "pmullh %[ftmp8], %[ftmp8], %[ftmp2] \n\t" \
485 "paddh %[ftmp3], %[ftmp7], %[ftmp8] \n\t" \
486 "punpcklbh %[ftmp7], %[ftmp4], %[ftmp0] \n\t" \
487 "punpckhbh %[ftmp8], %[ftmp4], %[ftmp0] \n\t" \
488 "pmullh %[ftmp7], %[ftmp7], %[ftmp1] \n\t" \
489 "pmullh %[ftmp8], %[ftmp8], %[ftmp2] \n\t" \
490 "paddh %[ftmp4], %[ftmp7], %[ftmp8] \n\t" \
491 "punpcklbh %[ftmp7], %[ftmp5], %[ftmp0] \n\t" \
492 "punpckhbh %[ftmp8], %[ftmp5], %[ftmp0] \n\t" \
493 "pmullh %[ftmp7], %[ftmp7], %[ftmp1] \n\t" \
494 "pmullh %[ftmp8], %[ftmp8], %[ftmp2] \n\t" \
495 "paddh %[ftmp5], %[ftmp7], %[ftmp8] \n\t" \
496 "punpcklbh %[ftmp7], %[ftmp6], %[ftmp0] \n\t" \
497 "punpckhbh %[ftmp8], %[ftmp6], %[ftmp0] \n\t" \
498 "pmullh %[ftmp7], %[ftmp7], %[ftmp1] \n\t" \
499 "pmullh %[ftmp8], %[ftmp8], %[ftmp2] \n\t" \
500 "paddh %[ftmp6], %[ftmp7], %[ftmp8] \n\t" \
501 TRANSPOSE_4H(%[ftmp3], %[ftmp4], %[ftmp5], %[ftmp6], \
502 %[ftmp7], %[ftmp8], %[ftmp9], %[ftmp10]) \
503 "paddh %[ftmp3], %[ftmp3], %[ftmp4] \n\t" \
504 "paddh %[ftmp5], %[ftmp5], %[ftmp6] \n\t" \
505 "paddh %[ftmp3], %[ftmp3], %[ftmp5] \n\t" \
506 MMI_USDC1(%[ftmp3], %[tmp], 0x00) \
507 \
508 "daddi %[x], %[x], -0x01 \n\t" \
509 PTR_ADDIU "%[src], %[src], 0x04 \n\t" \
510 PTR_ADDIU "%[tmp], %[tmp], 0x08 \n\t" \
511 "bnez %[x], 2b \n\t" \
512 \
513 "daddi %[y], %[y], -0x01 \n\t" \
514 "li %[x], " #x_step " \n\t" \
515 PTR_ADDIU "%[src], %[src], " #src_step " \n\t" \
516 PTR_ADDIU "%[tmp], %[tmp], " #src2_step " \n\t" \
517 PTR_ADDU "%[src], %[src], %[stride] \n\t" \
518 PTR_ADDIU "%[tmp], %[tmp], 0x80 \n\t" \
519 "bnez %[y], 1b \n\t" \
520 : RESTRICT_ASM_ALL64 \
521 [ftmp0]"=&f"(ftmp[0]), [ftmp1]"=&f"(ftmp[1]), \
522 [ftmp2]"=&f"(ftmp[2]), [ftmp3]"=&f"(ftmp[3]), \
523 [ftmp4]"=&f"(ftmp[4]), [ftmp5]"=&f"(ftmp[5]), \
524 [ftmp6]"=&f"(ftmp[6]), [ftmp7]"=&f"(ftmp[7]), \
525 [ftmp8]"=&f"(ftmp[8]), [ftmp9]"=&f"(ftmp[9]), \
526 [ftmp10]"=&f"(ftmp[10]), [rtmp0]"=&r"(rtmp[0]), \
527 [src]"+&r"(src), [tmp]"+&r"(tmp), [y]"+&r"(y), \
528 [x]"+&r"(x) \
529 : [filter]"r"(filter), [stride]"r"(srcstride) \
530 : "memory" \
531 ); \
532 \
533 tmp = tmp_array; \
534 filter = ff_hevc_qpel_filters[my]; \
535 x = width >> 2; \
536 y = height; \
537 __asm__ volatile( \
538 MMI_LDC1(%[ftmp1], %[filter], 0x00) \
539 "li %[rtmp0], 0x08 \n\t" \
540 "dmtc1 %[rtmp0], %[ftmp0] \n\t" \
541 "punpckhbh %[ftmp2], %[ftmp0], %[ftmp1] \n\t" \
542 "punpcklbh %[ftmp1], %[ftmp0], %[ftmp1] \n\t" \
543 "psrah %[ftmp1], %[ftmp1], %[ftmp0] \n\t" \
544 "psrah %[ftmp2], %[ftmp2], %[ftmp0] \n\t" \
545 "li %[rtmp0], 0x06 \n\t" \
546 "dmtc1 %[rtmp0], %[ftmp0] \n\t" \
547 "punpcklwd %[offset], %[offset], %[offset] \n\t" \
548 \
549 "1: \n\t" \
550 "li %[x], " #x_step " \n\t" \
551 "2: \n\t" \
552 MMI_ULDC1(%[ftmp3], %[tmp], 0x00) \
553 PTR_ADDIU "%[tmp], %[tmp], 0x80 \n\t" \
554 MMI_ULDC1(%[ftmp4], %[tmp], 0x00) \
555 PTR_ADDIU "%[tmp], %[tmp], 0x80 \n\t" \
556 MMI_ULDC1(%[ftmp5], %[tmp], 0x00) \
557 PTR_ADDIU "%[tmp], %[tmp], 0x80 \n\t" \
558 MMI_ULDC1(%[ftmp6], %[tmp], 0x00) \
559 PTR_ADDIU "%[tmp], %[tmp], 0x80 \n\t" \
560 MMI_ULDC1(%[ftmp7], %[tmp], 0x00) \
561 PTR_ADDIU "%[tmp], %[tmp], 0x80 \n\t" \
562 MMI_ULDC1(%[ftmp8], %[tmp], 0x00) \
563 PTR_ADDIU "%[tmp], %[tmp], 0x80 \n\t" \
564 MMI_ULDC1(%[ftmp9], %[tmp], 0x00) \
565 PTR_ADDIU "%[tmp], %[tmp], 0x80 \n\t" \
566 MMI_ULDC1(%[ftmp10], %[tmp], 0x00) \
567 PTR_ADDIU "%[tmp], %[tmp], -0x380 \n\t" \
568 TRANSPOSE_4H(%[ftmp3], %[ftmp4], %[ftmp5], %[ftmp6], \
569 %[ftmp11], %[ftmp12], %[ftmp13], %[ftmp14]) \
570 TRANSPOSE_4H(%[ftmp7], %[ftmp8], %[ftmp9], %[ftmp10], \
571 %[ftmp11], %[ftmp12], %[ftmp13], %[ftmp14]) \
572 "pmaddhw %[ftmp11], %[ftmp3], %[ftmp1] \n\t" \
573 "pmaddhw %[ftmp12], %[ftmp7], %[ftmp2] \n\t" \
574 "pmaddhw %[ftmp13], %[ftmp4], %[ftmp1] \n\t" \
575 "pmaddhw %[ftmp14], %[ftmp8], %[ftmp2] \n\t" \
576 "paddw %[ftmp11], %[ftmp11], %[ftmp12] \n\t" \
577 "paddw %[ftmp13], %[ftmp13], %[ftmp14] \n\t" \
578 TRANSPOSE_2W(%[ftmp11], %[ftmp13], %[ftmp3], %[ftmp4]) \
579 "paddw %[ftmp3], %[ftmp3], %[ftmp4] \n\t" \
580 "psraw %[ftmp3], %[ftmp3], %[ftmp0] \n\t" \
581 "pmaddhw %[ftmp11], %[ftmp5], %[ftmp1] \n\t" \
582 "pmaddhw %[ftmp12], %[ftmp9], %[ftmp2] \n\t" \
583 "pmaddhw %[ftmp13], %[ftmp6], %[ftmp1] \n\t" \
584 "pmaddhw %[ftmp14], %[ftmp10], %[ftmp2] \n\t" \
585 "paddw %[ftmp11], %[ftmp11], %[ftmp12] \n\t" \
586 "paddw %[ftmp13], %[ftmp13], %[ftmp14] \n\t" \
587 TRANSPOSE_2W(%[ftmp11], %[ftmp13], %[ftmp5], %[ftmp6]) \
588 "paddw %[ftmp5], %[ftmp5], %[ftmp6] \n\t" \
589 "psraw %[ftmp5], %[ftmp5], %[ftmp0] \n\t" \
590 "packsswh %[ftmp3], %[ftmp3], %[ftmp5] \n\t" \
591 MMI_ULDC1(%[ftmp4], %[src2], 0x00) \
592 "pxor %[ftmp7], %[ftmp7], %[ftmp7] \n\t" \
593 "li %[rtmp0], 0x10 \n\t" \
594 "dmtc1 %[rtmp0], %[ftmp8] \n\t" \
595 "punpcklhw %[ftmp5], %[ftmp7], %[ftmp3] \n\t" \
596 "punpckhhw %[ftmp6], %[ftmp7], %[ftmp3] \n\t" \
597 "punpckhhw %[ftmp3], %[ftmp7], %[ftmp4] \n\t" \
598 "punpcklhw %[ftmp4], %[ftmp7], %[ftmp4] \n\t" \
599 "psraw %[ftmp5], %[ftmp5], %[ftmp8] \n\t" \
600 "psraw %[ftmp6], %[ftmp6], %[ftmp8] \n\t" \
601 "psraw %[ftmp3], %[ftmp3], %[ftmp8] \n\t" \
602 "psraw %[ftmp4], %[ftmp4], %[ftmp8] \n\t" \
603 "paddw %[ftmp5], %[ftmp5], %[ftmp4] \n\t" \
604 "paddw %[ftmp6], %[ftmp6], %[ftmp3] \n\t" \
605 "paddw %[ftmp5], %[ftmp5], %[offset] \n\t" \
606 "paddw %[ftmp6], %[ftmp6], %[offset] \n\t" \
607 "psraw %[ftmp5], %[ftmp5], %[shift] \n\t" \
608 "psraw %[ftmp6], %[ftmp6], %[shift] \n\t" \
609 "packsswh %[ftmp5], %[ftmp5], %[ftmp6] \n\t" \
610 "pcmpgth %[ftmp7], %[ftmp5], %[ftmp7] \n\t" \
611 "pand %[ftmp3], %[ftmp5], %[ftmp7] \n\t" \
612 "packushb %[ftmp3], %[ftmp3], %[ftmp3] \n\t" \
613 MMI_USWC1(%[ftmp3], %[dst], 0x00) \
614 \
615 "daddi %[x], %[x], -0x01 \n\t" \
616 PTR_ADDIU "%[src2], %[src2], 0x08 \n\t" \
617 PTR_ADDIU "%[tmp], %[tmp], 0x08 \n\t" \
618 PTR_ADDIU "%[dst], %[dst], 0x04 \n\t" \
619 "bnez %[x], 2b \n\t" \
620 \
621 "daddi %[y], %[y], -0x01 \n\t" \
622 PTR_ADDIU "%[src2], %[src2], " #src2_step " \n\t" \
623 PTR_ADDIU "%[tmp], %[tmp], " #src2_step " \n\t" \
624 PTR_ADDIU "%[dst], %[dst], " #dst_step " \n\t" \
625 PTR_ADDIU "%[src2], %[src2], 0x80 \n\t" \
626 PTR_ADDU "%[dst], %[dst], %[stride] \n\t" \
627 PTR_ADDIU "%[tmp], %[tmp], 0x80 \n\t" \
628 "bnez %[y], 1b \n\t" \
629 : RESTRICT_ASM_ALL64 RESTRICT_ASM_LOW32 \
630 [ftmp0]"=&f"(ftmp[0]), [ftmp1]"=&f"(ftmp[1]), \
631 [ftmp2]"=&f"(ftmp[2]), [ftmp3]"=&f"(ftmp[3]), \
632 [ftmp4]"=&f"(ftmp[4]), [ftmp5]"=&f"(ftmp[5]), \
633 [ftmp6]"=&f"(ftmp[6]), [ftmp7]"=&f"(ftmp[7]), \
634 [ftmp8]"=&f"(ftmp[8]), [ftmp9]"=&f"(ftmp[9]), \
635 [ftmp10]"=&f"(ftmp[10]), [ftmp11]"=&f"(ftmp[11]), \
636 [ftmp12]"=&f"(ftmp[12]), [ftmp13]"=&f"(ftmp[13]), \
637 [ftmp14]"=&f"(ftmp[14]), [src2]"+&r"(src2), \
638 [dst]"+&r"(dst), [tmp]"+&r"(tmp), [y]"+&r"(y), [x]"=&r"(x), \
639 [offset]"+&f"(offset.f), [rtmp0]"=&r"(rtmp[0]) \
640 : [filter]"r"(filter), [stride]"r"(dststride), \
641 [shift]"f"(shift.f) \
642 : "memory" \
643 ); \
644}
645
646PUT_HEVC_QPEL_BI_HV(4, 1, -4, -8, -4);
647PUT_HEVC_QPEL_BI_HV(8, 2, -8, -16, -8);
648PUT_HEVC_QPEL_BI_HV(12, 3, -12, -24, -12);
649PUT_HEVC_QPEL_BI_HV(16, 4, -16, -32, -16);
650PUT_HEVC_QPEL_BI_HV(24, 6, -24, -48, -24);
651PUT_HEVC_QPEL_BI_HV(32, 8, -32, -64, -32);
652PUT_HEVC_QPEL_BI_HV(48, 12, -48, -96, -48);
653PUT_HEVC_QPEL_BI_HV(64, 16, -64, -128, -64);
654
655#define PUT_HEVC_EPEL_BI_HV(w, x_step, src_step, src2_step, dst_step) \
656void ff_hevc_put_hevc_epel_bi_hv##w##_8_mmi(uint8_t *_dst, \
657 ptrdiff_t _dststride, \
658 const uint8_t *_src, \
659 ptrdiff_t _srcstride, \
660 const int16_t *src2, int height, \
661 intptr_t mx, intptr_t my, \
662 int width) \
663{ \
664 int x, y; \
665 pixel *src = (pixel *)_src; \
666 ptrdiff_t srcstride = _srcstride / sizeof(pixel); \
667 pixel *dst = (pixel *)_dst; \
668 ptrdiff_t dststride = _dststride / sizeof(pixel); \
669 const int8_t *filter = ff_hevc_epel_filters[mx]; \
670 int16_t tmp_array[(MAX_PB_SIZE + EPEL_EXTRA) * MAX_PB_SIZE]; \
671 int16_t *tmp = tmp_array; \
672 double ftmp[12]; \
673 uint64_t rtmp[1]; \
674 union av_intfloat64 shift; \
675 union av_intfloat64 offset; \
676 DECLARE_VAR_ALL64; \
677 DECLARE_VAR_LOW32; \
678 shift.i = 7; \
679 offset.i = 64; \
680 \
681 src -= (EPEL_EXTRA_BEFORE * srcstride + 1); \
682 x = width >> 2; \
683 y = height + EPEL_EXTRA; \
684 __asm__ volatile( \
685 MMI_LWC1(%[ftmp1], %[filter], 0x00) \
686 "li %[rtmp0], 0x08 \n\t" \
687 "dmtc1 %[rtmp0], %[ftmp0] \n\t" \
688 "punpcklbh %[ftmp1], %[ftmp0], %[ftmp1] \n\t" \
689 "psrah %[ftmp1], %[ftmp1], %[ftmp0] \n\t" \
690 "pxor %[ftmp0], %[ftmp0], %[ftmp0] \n\t" \
691 \
692 "1: \n\t" \
693 "2: \n\t" \
694 MMI_ULWC1(%[ftmp2], %[src], 0x00) \
695 MMI_ULWC1(%[ftmp3], %[src], 0x01) \
696 MMI_ULWC1(%[ftmp4], %[src], 0x02) \
697 MMI_ULWC1(%[ftmp5], %[src], 0x03) \
698 "punpcklbh %[ftmp2], %[ftmp2], %[ftmp0] \n\t" \
699 "pmullh %[ftmp2], %[ftmp2], %[ftmp1] \n\t" \
700 "punpcklbh %[ftmp3], %[ftmp3], %[ftmp0] \n\t" \
701 "pmullh %[ftmp3], %[ftmp3], %[ftmp1] \n\t" \
702 "punpcklbh %[ftmp4], %[ftmp4], %[ftmp0] \n\t" \
703 "pmullh %[ftmp4], %[ftmp4], %[ftmp1] \n\t" \
704 "punpcklbh %[ftmp5], %[ftmp5], %[ftmp0] \n\t" \
705 "pmullh %[ftmp5], %[ftmp5], %[ftmp1] \n\t" \
706 TRANSPOSE_4H(%[ftmp2], %[ftmp3], %[ftmp4], %[ftmp5], \
707 %[ftmp6], %[ftmp7], %[ftmp8], %[ftmp9]) \
708 "paddh %[ftmp2], %[ftmp2], %[ftmp3] \n\t" \
709 "paddh %[ftmp4], %[ftmp4], %[ftmp5] \n\t" \
710 "paddh %[ftmp2], %[ftmp2], %[ftmp4] \n\t" \
711 MMI_USDC1(%[ftmp2], %[tmp], 0x00) \
712 \
713 "daddi %[x], %[x], -0x01 \n\t" \
714 PTR_ADDIU "%[src], %[src], 0x04 \n\t" \
715 PTR_ADDIU "%[tmp], %[tmp], 0x08 \n\t" \
716 "bnez %[x], 2b \n\t" \
717 \
718 "daddi %[y], %[y], -0x01 \n\t" \
719 "li %[x], " #x_step " \n\t" \
720 PTR_ADDIU "%[src], %[src], " #src_step " \n\t" \
721 PTR_ADDIU "%[tmp], %[tmp], " #src2_step " \n\t" \
722 PTR_ADDU "%[src], %[src], %[stride] \n\t" \
723 PTR_ADDIU "%[tmp], %[tmp], 0x80 \n\t" \
724 "bnez %[y], 1b \n\t" \
725 : RESTRICT_ASM_ALL64 \
726 [ftmp0]"=&f"(ftmp[0]), [ftmp1]"=&f"(ftmp[1]), \
727 [ftmp2]"=&f"(ftmp[2]), [ftmp3]"=&f"(ftmp[3]), \
728 [ftmp4]"=&f"(ftmp[4]), [ftmp5]"=&f"(ftmp[5]), \
729 [ftmp6]"=&f"(ftmp[6]), [ftmp7]"=&f"(ftmp[7]), \
730 [ftmp8]"=&f"(ftmp[8]), [ftmp9]"=&f"(ftmp[9]), \
731 [rtmp0]"=&r"(rtmp[0]), \
732 [src]"+&r"(src), [tmp]"+&r"(tmp), [y]"+&r"(y), \
733 [x]"+&r"(x) \
734 : [filter]"r"(filter), [stride]"r"(srcstride) \
735 : "memory" \
736 ); \
737 \
738 tmp = tmp_array; \
739 filter = ff_hevc_epel_filters[my]; \
740 x = width >> 2; \
741 y = height; \
742 __asm__ volatile( \
743 MMI_LWC1(%[ftmp1], %[filter], 0x00) \
744 "li %[rtmp0], 0x08 \n\t" \
745 "dmtc1 %[rtmp0], %[ftmp0] \n\t" \
746 "punpcklbh %[ftmp1], %[ftmp0], %[ftmp1] \n\t" \
747 "psrah %[ftmp1], %[ftmp1], %[ftmp0] \n\t" \
748 "li %[rtmp0], 0x06 \n\t" \
749 "dmtc1 %[rtmp0], %[ftmp0] \n\t" \
750 "punpcklwd %[offset], %[offset], %[offset] \n\t" \
751 "pxor %[ftmp2], %[ftmp2], %[ftmp2] \n\t" \
752 \
753 "1: \n\t" \
754 "li %[x], " #x_step " \n\t" \
755 "2: \n\t" \
756 MMI_ULDC1(%[ftmp3], %[tmp], 0x00) \
757 PTR_ADDIU "%[tmp], %[tmp], 0x80 \n\t" \
758 MMI_ULDC1(%[ftmp4], %[tmp], 0x00) \
759 PTR_ADDIU "%[tmp], %[tmp], 0x80 \n\t" \
760 MMI_ULDC1(%[ftmp5], %[tmp], 0x00) \
761 PTR_ADDIU "%[tmp], %[tmp], 0x80 \n\t" \
762 MMI_ULDC1(%[ftmp6], %[tmp], 0x00) \
763 PTR_ADDIU "%[tmp], %[tmp], -0x180 \n\t" \
764 TRANSPOSE_4H(%[ftmp3], %[ftmp4], %[ftmp5], %[ftmp6], \
765 %[ftmp7], %[ftmp8], %[ftmp9], %[ftmp10]) \
766 "pmaddhw %[ftmp7], %[ftmp3], %[ftmp1] \n\t" \
767 "pmaddhw %[ftmp8], %[ftmp4], %[ftmp1] \n\t" \
768 TRANSPOSE_2W(%[ftmp7], %[ftmp8], %[ftmp3], %[ftmp4]) \
769 "paddw %[ftmp3], %[ftmp3], %[ftmp4] \n\t" \
770 "psraw %[ftmp3], %[ftmp3], %[ftmp0] \n\t" \
771 "pmaddhw %[ftmp7], %[ftmp5], %[ftmp1] \n\t" \
772 "pmaddhw %[ftmp8], %[ftmp6], %[ftmp1] \n\t" \
773 TRANSPOSE_2W(%[ftmp7], %[ftmp8], %[ftmp5], %[ftmp6]) \
774 "paddw %[ftmp5], %[ftmp5], %[ftmp6] \n\t" \
775 "psraw %[ftmp5], %[ftmp5], %[ftmp0] \n\t" \
776 "packsswh %[ftmp3], %[ftmp3], %[ftmp5] \n\t" \
777 MMI_ULDC1(%[ftmp4], %[src2], 0x00) \
778 "li %[rtmp0], 0x10 \n\t" \
779 "dmtc1 %[rtmp0], %[ftmp8] \n\t" \
780 "punpcklhw %[ftmp5], %[ftmp2], %[ftmp3] \n\t" \
781 "punpckhhw %[ftmp6], %[ftmp2], %[ftmp3] \n\t" \
782 "punpckhhw %[ftmp3], %[ftmp2], %[ftmp4] \n\t" \
783 "punpcklhw %[ftmp4], %[ftmp2], %[ftmp4] \n\t" \
784 "psraw %[ftmp5], %[ftmp5], %[ftmp8] \n\t" \
785 "psraw %[ftmp6], %[ftmp6], %[ftmp8] \n\t" \
786 "psraw %[ftmp3], %[ftmp3], %[ftmp8] \n\t" \
787 "psraw %[ftmp4], %[ftmp4], %[ftmp8] \n\t" \
788 "paddw %[ftmp5], %[ftmp5], %[ftmp4] \n\t" \
789 "paddw %[ftmp6], %[ftmp6], %[ftmp3] \n\t" \
790 "paddw %[ftmp5], %[ftmp5], %[offset] \n\t" \
791 "paddw %[ftmp6], %[ftmp6], %[offset] \n\t" \
792 "psraw %[ftmp5], %[ftmp5], %[shift] \n\t" \
793 "psraw %[ftmp6], %[ftmp6], %[shift] \n\t" \
794 "packsswh %[ftmp5], %[ftmp5], %[ftmp6] \n\t" \
795 "pcmpgth %[ftmp7], %[ftmp5], %[ftmp2] \n\t" \
796 "pand %[ftmp3], %[ftmp5], %[ftmp7] \n\t" \
797 "packushb %[ftmp3], %[ftmp3], %[ftmp3] \n\t" \
798 MMI_USWC1(%[ftmp3], %[dst], 0x0) \
799 \
800 "daddi %[x], %[x], -0x01 \n\t" \
801 PTR_ADDIU "%[src2], %[src2], 0x08 \n\t" \
802 PTR_ADDIU "%[tmp], %[tmp], 0x08 \n\t" \
803 PTR_ADDIU "%[dst], %[dst], 0x04 \n\t" \
804 "bnez %[x], 2b \n\t" \
805 \
806 "daddi %[y], %[y], -0x01 \n\t" \
807 PTR_ADDIU "%[src2], %[src2], " #src2_step " \n\t" \
808 PTR_ADDIU "%[tmp], %[tmp], " #src2_step " \n\t" \
809 PTR_ADDIU "%[dst], %[dst], " #dst_step " \n\t" \
810 PTR_ADDIU "%[src2], %[src2], 0x80 \n\t" \
811 PTR_ADDU "%[dst], %[dst], %[stride] \n\t" \
812 PTR_ADDIU "%[tmp], %[tmp], 0x80 \n\t" \
813 "bnez %[y], 1b \n\t" \
814 : RESTRICT_ASM_LOW32 RESTRICT_ASM_ALL64 \
815 [ftmp0]"=&f"(ftmp[0]), [ftmp1]"=&f"(ftmp[1]), \
816 [ftmp2]"=&f"(ftmp[2]), [ftmp3]"=&f"(ftmp[3]), \
817 [ftmp4]"=&f"(ftmp[4]), [ftmp5]"=&f"(ftmp[5]), \
818 [ftmp6]"=&f"(ftmp[6]), [ftmp7]"=&f"(ftmp[7]), \
819 [ftmp8]"=&f"(ftmp[8]), [ftmp9]"=&f"(ftmp[9]), \
820 [ftmp10]"=&f"(ftmp[10]), [src2]"+&r"(src2), \
821 [dst]"+&r"(dst), [tmp]"+&r"(tmp), [y]"+&r"(y), [x]"=&r"(x), \
822 [offset]"+&f"(offset.f), [rtmp0]"=&r"(rtmp[0]) \
823 : [filter]"r"(filter), [stride]"r"(dststride), \
824 [shift]"f"(shift.f) \
825 : "memory" \
826 ); \
827}
828
829PUT_HEVC_EPEL_BI_HV(4, 1, -4, -8, -4);
830PUT_HEVC_EPEL_BI_HV(8, 2, -8, -16, -8);
831PUT_HEVC_EPEL_BI_HV(12, 3, -12, -24, -12);
832PUT_HEVC_EPEL_BI_HV(16, 4, -16, -32, -16);
833PUT_HEVC_EPEL_BI_HV(24, 6, -24, -48, -24);
834PUT_HEVC_EPEL_BI_HV(32, 8, -32, -64, -32);
835
836#define PUT_HEVC_PEL_BI_PIXELS(w, x_step, src_step, dst_step, src2_step) \
837void ff_hevc_put_hevc_pel_bi_pixels##w##_8_mmi(uint8_t *_dst, \
838 ptrdiff_t _dststride, \
839 const uint8_t *_src, \
840 ptrdiff_t _srcstride, \
841 const int16_t *src2, int height, \
842 intptr_t mx, intptr_t my, \
843 int width) \
844{ \
845 int x, y; \
846 pixel *src = (pixel *)_src; \
847 ptrdiff_t srcstride = _srcstride / sizeof(pixel); \
848 pixel *dst = (pixel *)_dst; \
849 ptrdiff_t dststride = _dststride / sizeof(pixel); \
850 double ftmp[12]; \
851 uint64_t rtmp[1]; \
852 union av_intfloat64 shift; \
853 DECLARE_VAR_ALL64; \
854 shift.i = 7; \
855 \
856 y = height; \
857 x = width >> 3; \
858 __asm__ volatile( \
859 "pxor %[ftmp0], %[ftmp0], %[ftmp0] \n\t" \
860 "li %[rtmp0], 0x06 \n\t" \
861 "dmtc1 %[rtmp0], %[ftmp1] \n\t" \
862 "li %[rtmp0], 0x10 \n\t" \
863 "dmtc1 %[rtmp0], %[ftmp10] \n\t" \
864 "li %[rtmp0], 0x40 \n\t" \
865 "dmtc1 %[rtmp0], %[offset] \n\t" \
866 "punpcklhw %[offset], %[offset], %[offset] \n\t" \
867 "punpcklwd %[offset], %[offset], %[offset] \n\t" \
868 \
869 "1: \n\t" \
870 "2: \n\t" \
871 MMI_ULDC1(%[ftmp5], %[src], 0x00) \
872 MMI_ULDC1(%[ftmp2], %[src2], 0x00) \
873 MMI_ULDC1(%[ftmp3], %[src2], 0x08) \
874 "punpcklbh %[ftmp4], %[ftmp5], %[ftmp0] \n\t" \
875 "punpckhbh %[ftmp5], %[ftmp5], %[ftmp0] \n\t" \
876 "psllh %[ftmp4], %[ftmp4], %[ftmp1] \n\t" \
877 "psllh %[ftmp5], %[ftmp5], %[ftmp1] \n\t" \
878 "paddh %[ftmp4], %[ftmp4], %[offset] \n\t" \
879 "paddh %[ftmp5], %[ftmp5], %[offset] \n\t" \
880 "punpcklhw %[ftmp6], %[ftmp4], %[ftmp0] \n\t" \
881 "punpckhhw %[ftmp7], %[ftmp4], %[ftmp0] \n\t" \
882 "punpcklhw %[ftmp8], %[ftmp5], %[ftmp0] \n\t" \
883 "punpckhhw %[ftmp9], %[ftmp5], %[ftmp0] \n\t" \
884 "punpcklhw %[ftmp4], %[ftmp0], %[ftmp3] \n\t" \
885 "punpckhhw %[ftmp5], %[ftmp0], %[ftmp3] \n\t" \
886 "punpckhhw %[ftmp3], %[ftmp0], %[ftmp2] \n\t" \
887 "punpcklhw %[ftmp2], %[ftmp0], %[ftmp2] \n\t" \
888 "psraw %[ftmp2], %[ftmp2], %[ftmp10] \n\t" \
889 "psraw %[ftmp3], %[ftmp3], %[ftmp10] \n\t" \
890 "psraw %[ftmp4], %[ftmp4], %[ftmp10] \n\t" \
891 "psraw %[ftmp5], %[ftmp5], %[ftmp10] \n\t" \
892 "paddw %[ftmp2], %[ftmp2], %[ftmp6] \n\t" \
893 "paddw %[ftmp3], %[ftmp3], %[ftmp7] \n\t" \
894 "paddw %[ftmp4], %[ftmp4], %[ftmp8] \n\t" \
895 "paddw %[ftmp5], %[ftmp5], %[ftmp9] \n\t" \
896 "psraw %[ftmp2], %[ftmp2], %[shift] \n\t" \
897 "psraw %[ftmp3], %[ftmp3], %[shift] \n\t" \
898 "psraw %[ftmp4], %[ftmp4], %[shift] \n\t" \
899 "psraw %[ftmp5], %[ftmp5], %[shift] \n\t" \
900 "packsswh %[ftmp2], %[ftmp2], %[ftmp3] \n\t" \
901 "packsswh %[ftmp4], %[ftmp4], %[ftmp5] \n\t" \
902 "pcmpgth %[ftmp3], %[ftmp2], %[ftmp0] \n\t" \
903 "pcmpgth %[ftmp5], %[ftmp4], %[ftmp0] \n\t" \
904 "pand %[ftmp2], %[ftmp2], %[ftmp3] \n\t" \
905 "pand %[ftmp4], %[ftmp4], %[ftmp5] \n\t" \
906 "packushb %[ftmp2], %[ftmp2], %[ftmp4] \n\t" \
907 MMI_USDC1(%[ftmp2], %[dst], 0x0) \
908 \
909 "daddi %[x], %[x], -0x01 \n\t" \
910 PTR_ADDIU "%[src], %[src], 0x08 \n\t" \
911 PTR_ADDIU "%[dst], %[dst], 0x08 \n\t" \
912 PTR_ADDIU "%[src2], %[src2], 0x10 \n\t" \
913 "bnez %[x], 2b \n\t" \
914 \
915 PTR_ADDIU "%[src], %[src], " #src_step " \n\t" \
916 PTR_ADDIU "%[dst], %[dst], " #dst_step " \n\t" \
917 PTR_ADDIU "%[src2], %[src2], " #src2_step " \n\t" \
918 "li %[x], " #x_step " \n\t" \
919 "daddi %[y], %[y], -0x01 \n\t" \
920 PTR_ADDU "%[src], %[src], %[srcstride] \n\t" \
921 PTR_ADDU "%[dst], %[dst], %[dststride] \n\t" \
922 PTR_ADDIU "%[src2], %[src2], 0x80 \n\t" \
923 "bnez %[y], 1b \n\t" \
924 : RESTRICT_ASM_ALL64 \
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]), \
929 [ftmp8]"=&f"(ftmp[8]), [ftmp9]"=&f"(ftmp[9]), \
930 [ftmp10]"=&f"(ftmp[10]), [offset]"=&f"(ftmp[11]), \
931 [src2]"+&r"(src2), [dst]"+&r"(dst), [src]"+&r"(src), \
932 [x]"+&r"(x), [y]"+&r"(y), [rtmp0]"=&r"(rtmp[0]) \
933 : [dststride]"r"(dststride), [shift]"f"(shift.f), \
934 [srcstride]"r"(srcstride) \
935 : "memory" \
936 ); \
937} \
938
939PUT_HEVC_PEL_BI_PIXELS(8, 1, -8, -8, -16);
940PUT_HEVC_PEL_BI_PIXELS(16, 2, -16, -16, -32);
941PUT_HEVC_PEL_BI_PIXELS(24, 3, -24, -24, -48);
942PUT_HEVC_PEL_BI_PIXELS(32, 4, -32, -32, -64);
943PUT_HEVC_PEL_BI_PIXELS(48, 6, -48, -48, -96);
944PUT_HEVC_PEL_BI_PIXELS(64, 8, -64, -64, -128);
945
946#define PUT_HEVC_QPEL_UNI_HV(w, x_step, src_step, dst_step, tmp_step) \
947void ff_hevc_put_hevc_qpel_uni_hv##w##_8_mmi(uint8_t *_dst, \
948 ptrdiff_t _dststride, \
949 const uint8_t *_src, \
950 ptrdiff_t _srcstride, \
951 int height, \
952 intptr_t mx, intptr_t my, \
953 int width) \
954{ \
955 int x, y; \
956 const int8_t *filter; \
957 pixel *src = (pixel*)_src; \
958 ptrdiff_t srcstride = _srcstride / sizeof(pixel); \
959 pixel *dst = (pixel *)_dst; \
960 ptrdiff_t dststride = _dststride / sizeof(pixel); \
961 int16_t tmp_array[(MAX_PB_SIZE + QPEL_EXTRA) * MAX_PB_SIZE]; \
962 int16_t *tmp = tmp_array; \
963 double ftmp[20]; \
964 uint64_t rtmp[1]; \
965 union av_intfloat64 shift; \
966 union av_intfloat64 offset; \
967 DECLARE_VAR_ALL64; \
968 DECLARE_VAR_LOW32; \
969 shift.i = 6; \
970 offset.i = 32; \
971 \
972 src -= (QPEL_EXTRA_BEFORE * srcstride + 3); \
973 filter = ff_hevc_qpel_filters[mx]; \
974 x = width >> 2; \
975 y = height + QPEL_EXTRA; \
976 __asm__ volatile( \
977 MMI_LDC1(%[ftmp1], %[filter], 0x00) \
978 "li %[rtmp0], 0x08 \n\t" \
979 "dmtc1 %[rtmp0], %[ftmp0] \n\t" \
980 "punpckhbh %[ftmp2], %[ftmp0], %[ftmp1] \n\t" \
981 "punpcklbh %[ftmp1], %[ftmp0], %[ftmp1] \n\t" \
982 "psrah %[ftmp1], %[ftmp1], %[ftmp0] \n\t" \
983 "psrah %[ftmp2], %[ftmp2], %[ftmp0] \n\t" \
984 "pxor %[ftmp0], %[ftmp0], %[ftmp0] \n\t" \
985 \
986 "1: \n\t" \
987 "2: \n\t" \
988 MMI_ULDC1(%[ftmp3], %[src], 0x00) \
989 MMI_ULDC1(%[ftmp4], %[src], 0x01) \
990 MMI_ULDC1(%[ftmp5], %[src], 0x02) \
991 MMI_ULDC1(%[ftmp6], %[src], 0x03) \
992 "punpcklbh %[ftmp7], %[ftmp3], %[ftmp0] \n\t" \
993 "punpckhbh %[ftmp8], %[ftmp3], %[ftmp0] \n\t" \
994 "pmullh %[ftmp7], %[ftmp7], %[ftmp1] \n\t" \
995 "pmullh %[ftmp8], %[ftmp8], %[ftmp2] \n\t" \
996 "paddh %[ftmp3], %[ftmp7], %[ftmp8] \n\t" \
997 "punpcklbh %[ftmp7], %[ftmp4], %[ftmp0] \n\t" \
998 "punpckhbh %[ftmp8], %[ftmp4], %[ftmp0] \n\t" \
999 "pmullh %[ftmp7], %[ftmp7], %[ftmp1] \n\t" \
1000 "pmullh %[ftmp8], %[ftmp8], %[ftmp2] \n\t" \
1001 "paddh %[ftmp4], %[ftmp7], %[ftmp8] \n\t" \
1002 "punpcklbh %[ftmp7], %[ftmp5], %[ftmp0] \n\t" \
1003 "punpckhbh %[ftmp8], %[ftmp5], %[ftmp0] \n\t" \
1004 "pmullh %[ftmp7], %[ftmp7], %[ftmp1] \n\t" \
1005 "pmullh %[ftmp8], %[ftmp8], %[ftmp2] \n\t" \
1006 "paddh %[ftmp5], %[ftmp7], %[ftmp8] \n\t" \
1007 "punpcklbh %[ftmp7], %[ftmp6], %[ftmp0] \n\t" \
1008 "punpckhbh %[ftmp8], %[ftmp6], %[ftmp0] \n\t" \
1009 "pmullh %[ftmp7], %[ftmp7], %[ftmp1] \n\t" \
1010 "pmullh %[ftmp8], %[ftmp8], %[ftmp2] \n\t" \
1011 "paddh %[ftmp6], %[ftmp7], %[ftmp8] \n\t" \
1012 TRANSPOSE_4H(%[ftmp3], %[ftmp4], %[ftmp5], %[ftmp6], \
1013 %[ftmp7], %[ftmp8], %[ftmp9], %[ftmp10]) \
1014 "paddh %[ftmp3], %[ftmp3], %[ftmp4] \n\t" \
1015 "paddh %[ftmp5], %[ftmp5], %[ftmp6] \n\t" \
1016 "paddh %[ftmp3], %[ftmp3], %[ftmp5] \n\t" \
1017 MMI_USDC1(%[ftmp3], %[tmp], 0x0) \
1018 \
1019 "daddi %[x], %[x], -0x01 \n\t" \
1020 PTR_ADDIU "%[src], %[src], 0x04 \n\t" \
1021 PTR_ADDIU "%[tmp], %[tmp], 0x08 \n\t" \
1022 "bnez %[x], 2b \n\t" \
1023 \
1024 "daddi %[y], %[y], -0x01 \n\t" \
1025 "li %[x], " #x_step " \n\t" \
1026 PTR_ADDIU "%[src], %[src], " #src_step " \n\t" \
1027 PTR_ADDIU "%[tmp], %[tmp], " #tmp_step " \n\t" \
1028 PTR_ADDU "%[src], %[src], %[stride] \n\t" \
1029 PTR_ADDIU "%[tmp], %[tmp], 0x80 \n\t" \
1030 "bnez %[y], 1b \n\t" \
1031 : RESTRICT_ASM_ALL64 \
1032 [ftmp0]"=&f"(ftmp[0]), [ftmp1]"=&f"(ftmp[1]), \
1033 [ftmp2]"=&f"(ftmp[2]), [ftmp3]"=&f"(ftmp[3]), \
1034 [ftmp4]"=&f"(ftmp[4]), [ftmp5]"=&f"(ftmp[5]), \
1035 [ftmp6]"=&f"(ftmp[6]), [ftmp7]"=&f"(ftmp[7]), \
1036 [ftmp8]"=&f"(ftmp[8]), [ftmp9]"=&f"(ftmp[9]), \
1037 [ftmp10]"=&f"(ftmp[10]), [rtmp0]"=&r"(rtmp[0]), \
1038 [src]"+&r"(src), [tmp]"+&r"(tmp), [y]"+&r"(y), \
1039 [x]"+&r"(x) \
1040 : [filter]"r"(filter), [stride]"r"(srcstride) \
1041 : "memory" \
1042 ); \
1043 \
1044 tmp = tmp_array; \
1045 filter = ff_hevc_qpel_filters[my]; \
1046 x = width >> 2; \
1047 y = height; \
1048 __asm__ volatile( \
1049 MMI_LDC1(%[ftmp1], %[filter], 0x00) \
1050 "li %[rtmp0], 0x08 \n\t" \
1051 "dmtc1 %[rtmp0], %[ftmp0] \n\t" \
1052 "punpckhbh %[ftmp2], %[ftmp0], %[ftmp1] \n\t" \
1053 "punpcklbh %[ftmp1], %[ftmp0], %[ftmp1] \n\t" \
1054 "psrah %[ftmp1], %[ftmp1], %[ftmp0] \n\t" \
1055 "psrah %[ftmp2], %[ftmp2], %[ftmp0] \n\t" \
1056 "li %[rtmp0], 0x06 \n\t" \
1057 "dmtc1 %[rtmp0], %[ftmp0] \n\t" \
1058 "punpcklhw %[offset], %[offset], %[offset] \n\t" \
1059 "punpcklwd %[offset], %[offset], %[offset] \n\t" \
1060 \
1061 "1: \n\t" \
1062 "li %[x], " #x_step " \n\t" \
1063 "2: \n\t" \
1064 MMI_ULDC1(%[ftmp3], %[tmp], 0x00) \
1065 PTR_ADDIU "%[tmp], %[tmp], 0x80 \n\t" \
1066 MMI_ULDC1(%[ftmp4], %[tmp], 0x00) \
1067 PTR_ADDIU "%[tmp], %[tmp], 0x80 \n\t" \
1068 MMI_ULDC1(%[ftmp5], %[tmp], 0x00) \
1069 PTR_ADDIU "%[tmp], %[tmp], 0x80 \n\t" \
1070 MMI_ULDC1(%[ftmp6], %[tmp], 0x00) \
1071 PTR_ADDIU "%[tmp], %[tmp], 0x80 \n\t" \
1072 MMI_ULDC1(%[ftmp7], %[tmp], 0x00) \
1073 PTR_ADDIU "%[tmp], %[tmp], 0x80 \n\t" \
1074 MMI_ULDC1(%[ftmp8], %[tmp], 0x00) \
1075 PTR_ADDIU "%[tmp], %[tmp], 0x80 \n\t" \
1076 MMI_ULDC1(%[ftmp9], %[tmp], 0x00) \
1077 PTR_ADDIU "%[tmp], %[tmp], 0x80 \n\t" \
1078 MMI_ULDC1(%[ftmp10], %[tmp], 0x00) \
1079 PTR_ADDIU "%[tmp], %[tmp], -0x380 \n\t" \
1080 TRANSPOSE_4H(%[ftmp3], %[ftmp4], %[ftmp5], %[ftmp6], \
1081 %[ftmp11], %[ftmp12], %[ftmp13], %[ftmp14]) \
1082 TRANSPOSE_4H(%[ftmp7], %[ftmp8], %[ftmp9], %[ftmp10], \
1083 %[ftmp11], %[ftmp12], %[ftmp13], %[ftmp14]) \
1084 "pmaddhw %[ftmp11], %[ftmp3], %[ftmp1] \n\t" \
1085 "pmaddhw %[ftmp12], %[ftmp7], %[ftmp2] \n\t" \
1086 "pmaddhw %[ftmp13], %[ftmp4], %[ftmp1] \n\t" \
1087 "pmaddhw %[ftmp14], %[ftmp8], %[ftmp2] \n\t" \
1088 "paddw %[ftmp11], %[ftmp11], %[ftmp12] \n\t" \
1089 "paddw %[ftmp13], %[ftmp13], %[ftmp14] \n\t" \
1090 TRANSPOSE_2W(%[ftmp11], %[ftmp13], %[ftmp3], %[ftmp4]) \
1091 "paddw %[ftmp3], %[ftmp3], %[ftmp4] \n\t" \
1092 "psraw %[ftmp3], %[ftmp3], %[ftmp0] \n\t" \
1093 "pmaddhw %[ftmp11], %[ftmp5], %[ftmp1] \n\t" \
1094 "pmaddhw %[ftmp12], %[ftmp9], %[ftmp2] \n\t" \
1095 "pmaddhw %[ftmp13], %[ftmp6], %[ftmp1] \n\t" \
1096 "pmaddhw %[ftmp14], %[ftmp10], %[ftmp2] \n\t" \
1097 "paddw %[ftmp11], %[ftmp11], %[ftmp12] \n\t" \
1098 "paddw %[ftmp13], %[ftmp13], %[ftmp14] \n\t" \
1099 TRANSPOSE_2W(%[ftmp11], %[ftmp13], %[ftmp5], %[ftmp6]) \
1100 "paddw %[ftmp5], %[ftmp5], %[ftmp6] \n\t" \
1101 "psraw %[ftmp5], %[ftmp5], %[ftmp0] \n\t" \
1102 "packsswh %[ftmp3], %[ftmp3], %[ftmp5] \n\t" \
1103 "paddh %[ftmp3], %[ftmp3], %[offset] \n\t" \
1104 "psrah %[ftmp3], %[ftmp3], %[shift] \n\t" \
1105 "pxor %[ftmp7], %[ftmp7], %[ftmp7] \n\t" \
1106 "pcmpgth %[ftmp7], %[ftmp3], %[ftmp7] \n\t" \
1107 "pand %[ftmp3], %[ftmp3], %[ftmp7] \n\t" \
1108 "packushb %[ftmp3], %[ftmp3], %[ftmp3] \n\t" \
1109 MMI_USWC1(%[ftmp3], %[dst], 0x00) \
1110 \
1111 "daddi %[x], %[x], -0x01 \n\t" \
1112 PTR_ADDIU "%[tmp], %[tmp], 0x08 \n\t" \
1113 PTR_ADDIU "%[dst], %[dst], 0x04 \n\t" \
1114 "bnez %[x], 2b \n\t" \
1115 \
1116 "daddi %[y], %[y], -0x01 \n\t" \
1117 PTR_ADDIU "%[tmp], %[tmp], " #tmp_step " \n\t" \
1118 PTR_ADDIU "%[dst], %[dst], " #dst_step " \n\t" \
1119 PTR_ADDU "%[dst], %[dst], %[stride] \n\t" \
1120 PTR_ADDIU "%[tmp], %[tmp], 0x80 \n\t" \
1121 "bnez %[y], 1b \n\t" \
1122 : RESTRICT_ASM_ALL64 RESTRICT_ASM_LOW32 \
1123 [ftmp0]"=&f"(ftmp[0]), [ftmp1]"=&f"(ftmp[1]), \
1124 [ftmp2]"=&f"(ftmp[2]), [ftmp3]"=&f"(ftmp[3]), \
1125 [ftmp4]"=&f"(ftmp[4]), [ftmp5]"=&f"(ftmp[5]), \
1126 [ftmp6]"=&f"(ftmp[6]), [ftmp7]"=&f"(ftmp[7]), \
1127 [ftmp8]"=&f"(ftmp[8]), [ftmp9]"=&f"(ftmp[9]), \
1128 [ftmp10]"=&f"(ftmp[10]), [ftmp11]"=&f"(ftmp[11]), \
1129 [ftmp12]"=&f"(ftmp[12]), [ftmp13]"=&f"(ftmp[13]), \
1130 [ftmp14]"=&f"(ftmp[14]), \
1131 [dst]"+&r"(dst), [tmp]"+&r"(tmp), [y]"+&r"(y), [x]"=&r"(x), \
1132 [offset]"+&f"(offset.f), [rtmp0]"=&r"(rtmp[0]) \
1133 : [filter]"r"(filter), [stride]"r"(dststride), \
1134 [shift]"f"(shift.f) \
1135 : "memory" \
1136 ); \
1137}
1138
1139PUT_HEVC_QPEL_UNI_HV(4, 1, -4, -4, -8);
1140PUT_HEVC_QPEL_UNI_HV(8, 2, -8, -8, -16);
1141PUT_HEVC_QPEL_UNI_HV(12, 3, -12, -12, -24);
1142PUT_HEVC_QPEL_UNI_HV(16, 4, -16, -16, -32);
1143PUT_HEVC_QPEL_UNI_HV(24, 6, -24, -24, -48);
1144PUT_HEVC_QPEL_UNI_HV(32, 8, -32, -32, -64);
1145PUT_HEVC_QPEL_UNI_HV(48, 12, -48, -48, -96);
1146PUT_HEVC_QPEL_UNI_HV(64, 16, -64, -64, -128);
bit_depth_template.c
hevcdsp_mips.h
PUT_HEVC_EPEL_BI_HV
#define PUT_HEVC_EPEL_BI_HV(w, x_step, src_step, src2_step, dst_step)
Definition hevcdsp_mmi.c:655
PUT_HEVC_QPEL_BI_HV
#define PUT_HEVC_QPEL_BI_HV(w, x_step, src_step, src2_step, dst_step)
Definition hevcdsp_mmi.c:435
PUT_HEVC_QPEL_UNI_HV
#define PUT_HEVC_QPEL_UNI_HV(w, x_step, src_step, dst_step, tmp_step)
Definition hevcdsp_mmi.c:946
PUT_HEVC_QPEL_HV
#define PUT_HEVC_QPEL_HV(w, x_step, src_step, dst_step)
Definition hevcdsp_mmi.c:121
PUT_HEVC_PEL_BI_PIXELS
#define PUT_HEVC_PEL_BI_PIXELS(w, x_step, src_step, dst_step, src2_step)
Definition hevcdsp_mmi.c:836
PUT_HEVC_QPEL_BI_H
#define PUT_HEVC_QPEL_BI_H(w, x_step, src_step, src2_step, dst_step)
Definition hevcdsp_mmi.c:304
PUT_HEVC_QPEL_H
#define PUT_HEVC_QPEL_H(w, x_step, src_step, dst_step)
Definition hevcdsp_mmi.c:27
Generated on Mon Aug 24 2026 19:20:57 for FFmpeg by doxygen 1.13.2