FFmpeg
h264chroma_msa.c
Go to the documentation of this file.
1 /*
2  * Copyright (c) 2015 - 2017 Shivraj Patil (Shivraj.Patil@imgtec.com)
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 
22 #include "h264chroma_mips.h"
23 
24 static const uint8_t chroma_mask_arr[16 * 5] = {
25  0, 1, 1, 2, 2, 3, 3, 4, 16, 17, 17, 18, 18, 19, 19, 20,
26  0, 2, 2, 4, 4, 6, 6, 8, 16, 18, 18, 20, 20, 22, 22, 24,
27  0, 1, 1, 2, 2, 3, 3, 4, 4, 5, 5, 6, 6, 7, 7, 8,
28  0, 1, 1, 2, 16, 17, 17, 18, 4, 5, 5, 6, 6, 7, 7, 8,
29  0, 1, 1, 2, 16, 17, 17, 18, 16, 17, 17, 18, 18, 19, 19, 20
30 };
31 
32 static void avc_chroma_hz_2x2_msa(uint8_t *src, uint8_t *dst, int32_t stride,
33  uint32_t coeff0, uint32_t coeff1)
34 {
35  uint16_t out0, out1;
36  v16i8 src0, src1;
37  v8u16 res_r;
38  v8i16 res;
39  v16i8 mask;
40  v16i8 coeff_vec0 = __msa_fill_b(coeff0);
41  v16i8 coeff_vec1 = __msa_fill_b(coeff1);
42  v16u8 coeff_vec = (v16u8) __msa_ilvr_b(coeff_vec0, coeff_vec1);
43 
44  mask = LD_SB(&chroma_mask_arr[0]);
45 
47 
48  src0 = __msa_vshf_b(mask, src1, src0);
49  res_r = __msa_dotp_u_h((v16u8) src0, coeff_vec);
50  res_r <<= 3;
51  res_r = (v8u16) __msa_srari_h((v8i16) res_r, 6);
52  res_r = __msa_sat_u_h(res_r, 7);
53  res = (v8i16) __msa_pckev_b((v16i8) res_r, (v16i8) res_r);
54 
55  out0 = __msa_copy_u_h(res, 0);
56  out1 = __msa_copy_u_h(res, 2);
57 
58  SH(out0, dst);
59  dst += stride;
60  SH(out1, dst);
61 }
62 
63 static void avc_chroma_hz_2x4_msa(uint8_t *src, uint8_t *dst, int32_t stride,
64  uint32_t coeff0, uint32_t coeff1)
65 {
66  v16u8 src0, src1, src2, src3;
67  v8u16 res_r;
68  v8i16 res;
69  v16i8 mask;
70  v16i8 coeff_vec0 = __msa_fill_b(coeff0);
71  v16i8 coeff_vec1 = __msa_fill_b(coeff1);
72  v16u8 coeff_vec = (v16u8) __msa_ilvr_b(coeff_vec0, coeff_vec1);
73 
74  mask = LD_SB(&chroma_mask_arr[64]);
75 
76  LD_UB4(src, stride, src0, src1, src2, src3);
77 
78  VSHF_B2_UB(src0, src1, src2, src3, mask, mask, src0, src2);
79 
80  src0 = (v16u8) __msa_ilvr_d((v2i64) src2, (v2i64) src0);
81 
82  res_r = __msa_dotp_u_h(src0, coeff_vec);
83  res_r <<= 3;
84  res_r = (v8u16) __msa_srari_h((v8i16) res_r, 6);
85  res_r = __msa_sat_u_h(res_r, 7);
86  res = (v8i16) __msa_pckev_b((v16i8) res_r, (v16i8) res_r);
87 
88  ST_H4(res, 0, 1, 2, 3, dst, stride);
89 }
90 
91 static void avc_chroma_hz_2w_msa(uint8_t *src, uint8_t *dst, int32_t stride,
92  uint32_t coeff0, uint32_t coeff1,
94 {
95  if (2 == height) {
96  avc_chroma_hz_2x2_msa(src, dst, stride, coeff0, coeff1);
97  } else if (4 == height) {
98  avc_chroma_hz_2x4_msa(src, dst, stride, coeff0, coeff1);
99  }
100 }
101 
102 static void avc_chroma_hz_4x2_msa(uint8_t *src, uint8_t *dst, int32_t stride,
103  uint32_t coeff0, uint32_t coeff1)
104 {
105  v16i8 src0, src1;
106  v8u16 res_r;
107  v4i32 res;
108  v16i8 mask;
109  v16i8 coeff_vec0 = __msa_fill_b(coeff0);
110  v16i8 coeff_vec1 = __msa_fill_b(coeff1);
111  v16u8 coeff_vec = (v16u8) __msa_ilvr_b(coeff_vec0, coeff_vec1);
112 
113  mask = LD_SB(&chroma_mask_arr[0]);
114 
115  LD_SB2(src, stride, src0, src1);
116 
117  src0 = __msa_vshf_b(mask, src1, src0);
118  res_r = __msa_dotp_u_h((v16u8) src0, coeff_vec);
119  res_r <<= 3;
120  res_r = (v8u16) __msa_srari_h((v8i16) res_r, 6);
121  res_r = __msa_sat_u_h(res_r, 7);
122  res = (v4i32) __msa_pckev_b((v16i8) res_r, (v16i8) res_r);
123 
124  ST_W2(res, 0, 1, dst, stride);
125 }
126 
127 static void avc_chroma_hz_4x4_msa(uint8_t *src, uint8_t *dst, int32_t stride,
128  uint32_t coeff0, uint32_t coeff1)
129 {
130  v16u8 src0, src1, src2, src3, out;
131  v8u16 res0_r, res1_r;
132  v16i8 mask;
133  v16i8 coeff_vec0 = __msa_fill_b(coeff0);
134  v16i8 coeff_vec1 = __msa_fill_b(coeff1);
135  v16u8 coeff_vec = (v16u8) __msa_ilvr_b(coeff_vec0, coeff_vec1);
136 
137  mask = LD_SB(&chroma_mask_arr[0]);
138 
139  LD_UB4(src, stride, src0, src1, src2, src3);
140  VSHF_B2_UB(src0, src1, src2, src3, mask, mask, src0, src2);
141  DOTP_UB2_UH(src0, src2, coeff_vec, coeff_vec, res0_r, res1_r);
142  res0_r <<= 3;
143  res1_r <<= 3;
144  SRARI_H2_UH(res0_r, res1_r, 6);
145  SAT_UH2_UH(res0_r, res1_r, 7);
146  out = (v16u8) __msa_pckev_b((v16i8) res1_r, (v16i8) res0_r);
147  ST_W4(out, 0, 1, 2, 3, dst, stride);
148 }
149 
150 static void avc_chroma_hz_4x8_msa(uint8_t *src, uint8_t *dst, int32_t stride,
151  uint32_t coeff0, uint32_t coeff1)
152 {
153  v16u8 src0, src1, src2, src3, src4, src5, src6, src7, out0, out1;
154  v16i8 mask;
155  v8u16 res0, res1, res2, res3;
156  v16i8 coeff_vec0 = __msa_fill_b(coeff0);
157  v16i8 coeff_vec1 = __msa_fill_b(coeff1);
158  v16u8 coeff_vec = (v16u8) __msa_ilvr_b(coeff_vec0, coeff_vec1);
159 
160  mask = LD_SB(&chroma_mask_arr[0]);
161 
162  LD_UB8(src, stride, src0, src1, src2, src3, src4, src5, src6, src7);
163  VSHF_B2_UB(src0, src1, src2, src3, mask, mask, src0, src2);
164  VSHF_B2_UB(src4, src5, src6, src7, mask, mask, src4, src6);
165  DOTP_UB2_UH(src0, src2, coeff_vec, coeff_vec, res0, res1);
166  DOTP_UB2_UH(src4, src6, coeff_vec, coeff_vec, res2, res3);
167  SLLI_4V(res0, res1, res2, res3, 3);
168  SRARI_H4_UH(res0, res1, res2, res3, 6);
169  SAT_UH4_UH(res0, res1, res2, res3, 7);
170  PCKEV_B2_UB(res1, res0, res3, res2, out0, out1);
171  ST_W8(out0, out1, 0, 1, 2, 3, 0, 1, 2, 3, dst, stride);
172 }
173 
174 static void avc_chroma_hz_4w_msa(uint8_t *src, uint8_t *dst, int32_t stride,
175  uint32_t coeff0, uint32_t coeff1,
176  int32_t height)
177 {
178  if (2 == height) {
179  avc_chroma_hz_4x2_msa(src, dst, stride, coeff0, coeff1);
180  } else if (4 == height) {
181  avc_chroma_hz_4x4_msa(src, dst, stride, coeff0, coeff1);
182  } else if (8 == height) {
183  avc_chroma_hz_4x8_msa(src, dst, stride, coeff0, coeff1);
184  }
185 }
186 
187 static void avc_chroma_hz_8x4_msa(uint8_t *src, uint8_t *dst, int32_t stride,
188  uint32_t coeff0, uint32_t coeff1)
189 {
190  v16u8 src0, src1, src2, src3, out0, out1;
191  v8u16 res0, res1, res2, res3;
192  v16i8 mask;
193  v16i8 coeff_vec0 = __msa_fill_b(coeff0);
194  v16i8 coeff_vec1 = __msa_fill_b(coeff1);
195  v16u8 coeff_vec = (v16u8) __msa_ilvr_b(coeff_vec0, coeff_vec1);
196 
197  mask = LD_SB(&chroma_mask_arr[32]);
198  LD_UB4(src, stride, src0, src1, src2, src3);
200  VSHF_B2_UB(src2, src2, src3, src3, mask, mask, src2, src3);
201  DOTP_UB4_UH(src0, src1, src2, src3, coeff_vec, coeff_vec, coeff_vec,
202  coeff_vec, res0, res1, res2, res3);
203  SLLI_4V(res0, res1, res2, res3, 3);
204  SRARI_H4_UH(res0, res1, res2, res3, 6);
205  SAT_UH4_UH(res0, res1, res2, res3, 7);
206  PCKEV_B2_UB(res1, res0, res3, res2, out0, out1);
207  ST_D4(out0, out1, 0, 1, 0, 1, dst, stride);
208 }
209 
210 static void avc_chroma_hz_8x8_msa(uint8_t *src, uint8_t *dst, int32_t stride,
211  uint32_t coeff0, uint32_t coeff1)
212 {
213  v16u8 src0, src1, src2, src3, src4, src5, src6, src7;
214  v16u8 out0, out1, out2, out3;
215  v8u16 res0, res1, res2, res3, res4, res5, res6, res7;
216  v16i8 mask;
217  v16i8 coeff_vec0 = __msa_fill_b(coeff0);
218  v16i8 coeff_vec1 = __msa_fill_b(coeff1);
219  v16u8 coeff_vec = (v16u8) __msa_ilvr_b(coeff_vec0, coeff_vec1);
220 
221  mask = LD_SB(&chroma_mask_arr[32]);
222 
223  LD_UB8(src, stride, src0, src1, src2, src3, src4, src5, src6, src7);
225  VSHF_B2_UB(src2, src2, src3, src3, mask, mask, src2, src3);
226  VSHF_B2_UB(src4, src4, src5, src5, mask, mask, src4, src5);
227  VSHF_B2_UB(src6, src6, src7, src7, mask, mask, src6, src7);
228  DOTP_UB4_UH(src0, src1, src2, src3, coeff_vec, coeff_vec, coeff_vec,
229  coeff_vec, res0, res1, res2, res3);
230  DOTP_UB4_UH(src4, src5, src6, src7, coeff_vec, coeff_vec, coeff_vec,
231  coeff_vec, res4, res5, res6, res7);
232  SLLI_4V(res0, res1, res2, res3, 3);
233  SLLI_4V(res4, res5, res6, res7, 3);
234  SRARI_H4_UH(res0, res1, res2, res3, 6);
235  SRARI_H4_UH(res4, res5, res6, res7, 6);
236  SAT_UH4_UH(res0, res1, res2, res3, 7);
237  SAT_UH4_UH(res4, res5, res6, res7, 7);
238  PCKEV_B2_UB(res1, res0, res3, res2, out0, out1);
239  PCKEV_B2_UB(res5, res4, res7, res6, out2, out3);
240  ST_D8(out0, out1, out2, out3, 0, 1, 0, 1, 0, 1, 0, 1, dst, stride);
241 }
242 
243 static void avc_chroma_hz_nonmult_msa(uint8_t *src, uint8_t *dst,
244  int32_t stride, uint32_t coeff0,
245  uint32_t coeff1, int32_t height)
246 {
247  uint32_t row;
248  v16u8 src0, src1, src2, src3, out0, out1;
249  v8u16 res0, res1, res2, res3;
250  v16i8 mask;
251  v16i8 coeff_vec0 = __msa_fill_b(coeff0);
252  v16i8 coeff_vec1 = __msa_fill_b(coeff1);
253  v16u8 coeff_vec = (v16u8) __msa_ilvr_b(coeff_vec0, coeff_vec1);
254 
255  mask = LD_SB(&chroma_mask_arr[32]);
256 
257  for (row = height >> 2; row--;) {
258  LD_UB4(src, stride, src0, src1, src2, src3);
259  src += (4 * stride);
260 
262  VSHF_B2_UB(src2, src2, src3, src3, mask, mask, src2, src3);
263  DOTP_UB4_UH(src0, src1, src2, src3, coeff_vec, coeff_vec, coeff_vec,
264  coeff_vec, res0, res1, res2, res3);
265  SLLI_4V(res0, res1, res2, res3, 3);
266  SRARI_H4_UH(res0, res1, res2, res3, 6);
267  SAT_UH4_UH(res0, res1, res2, res3, 7);
268  PCKEV_B2_UB(res1, res0, res3, res2, out0, out1);
269  ST_D4(out0, out1, 0, 1, 0, 1, dst, stride);
270  dst += (4 * stride);
271  }
272 
273  if (0 != (height % 4)) {
274  for (row = (height % 4); row--;) {
275  src0 = LD_UB(src);
276  src += stride;
277 
278  src0 = (v16u8) __msa_vshf_b(mask, (v16i8) src0, (v16i8) src0);
279 
280  res0 = __msa_dotp_u_h(src0, coeff_vec);
281  res0 <<= 3;
282  res0 = (v8u16) __msa_srari_h((v8i16) res0, 6);
283  res0 = __msa_sat_u_h(res0, 7);
284  res0 = (v8u16) __msa_pckev_b((v16i8) res0, (v16i8) res0);
285 
286  ST_D1(res0, 0, dst);
287  dst += stride;
288  }
289  }
290 }
291 
292 static void avc_chroma_hz_8w_msa(uint8_t *src, uint8_t *dst, int32_t stride,
293  uint32_t coeff0, uint32_t coeff1,
294  int32_t height)
295 {
296  if (4 == height) {
297  avc_chroma_hz_8x4_msa(src, dst, stride, coeff0, coeff1);
298  } else if (8 == height) {
299  avc_chroma_hz_8x8_msa(src, dst, stride, coeff0, coeff1);
300  } else {
301  avc_chroma_hz_nonmult_msa(src, dst, stride, coeff0, coeff1, height);
302  }
303 }
304 
305 static void avc_chroma_vt_2x2_msa(uint8_t *src, uint8_t *dst, int32_t stride,
306  uint32_t coeff0, uint32_t coeff1)
307 {
308  uint16_t out0, out1;
309  v16i8 src0, src1, src2;
310  v16u8 tmp0, tmp1;
311  v8i16 res;
312  v8u16 res_r;
313  v16i8 coeff_vec0 = __msa_fill_b(coeff0);
314  v16i8 coeff_vec1 = __msa_fill_b(coeff1);
315  v16u8 coeff_vec = (v16u8) __msa_ilvr_b(coeff_vec0, coeff_vec1);
316 
317  LD_SB3(src, stride, src0, src1, src2);
318 
319  ILVR_B2_UB(src1, src0, src2, src1, tmp0, tmp1);
320 
321  tmp0 = (v16u8) __msa_ilvr_d((v2i64) tmp1, (v2i64) tmp0);
322 
323  res_r = __msa_dotp_u_h(tmp0, coeff_vec);
324  res_r <<= 3;
325  res_r = (v8u16) __msa_srari_h((v8i16) res_r, 6);
326  res_r = __msa_sat_u_h(res_r, 7);
327  res = (v8i16) __msa_pckev_b((v16i8) res_r, (v16i8) res_r);
328 
329  out0 = __msa_copy_u_h(res, 0);
330  out1 = __msa_copy_u_h(res, 2);
331 
332  SH(out0, dst);
333  dst += stride;
334  SH(out1, dst);
335 }
336 
337 static void avc_chroma_vt_2x4_msa(uint8_t *src, uint8_t *dst, int32_t stride,
338  uint32_t coeff0, uint32_t coeff1)
339 {
340  v16u8 src0, src1, src2, src3, src4;
341  v16u8 tmp0, tmp1, tmp2, tmp3;
342  v8i16 res;
343  v8u16 res_r;
344  v16i8 coeff_vec0 = __msa_fill_b(coeff0);
345  v16i8 coeff_vec1 = __msa_fill_b(coeff1);
346  v16u8 coeff_vec = (v16u8) __msa_ilvr_b(coeff_vec0, coeff_vec1);
347 
348  LD_UB5(src, stride, src0, src1, src2, src3, src4);
349  ILVR_B4_UB(src1, src0, src2, src1, src3, src2, src4, src3,
350  tmp0, tmp1, tmp2, tmp3);
351  ILVR_W2_UB(tmp1, tmp0, tmp3, tmp2, tmp0, tmp2);
352 
353  tmp0 = (v16u8) __msa_ilvr_d((v2i64) tmp2, (v2i64) tmp0);
354 
355  res_r = __msa_dotp_u_h(tmp0, coeff_vec);
356  res_r <<= 3;
357  res_r = (v8u16) __msa_srari_h((v8i16) res_r, 6);
358  res_r = __msa_sat_u_h(res_r, 7);
359 
360  res = (v8i16) __msa_pckev_b((v16i8) res_r, (v16i8) res_r);
361 
362  ST_H4(res, 0, 1, 2, 3, dst, stride);
363 }
364 
365 static void avc_chroma_vt_2w_msa(uint8_t *src, uint8_t *dst, int32_t stride,
366  uint32_t coeff0, uint32_t coeff1,
367  int32_t height)
368 {
369  if (2 == height) {
370  avc_chroma_vt_2x2_msa(src, dst, stride, coeff0, coeff1);
371  } else if (4 == height) {
372  avc_chroma_vt_2x4_msa(src, dst, stride, coeff0, coeff1);
373  }
374 }
375 
376 static void avc_chroma_vt_4x2_msa(uint8_t *src, uint8_t *dst, int32_t stride,
377  uint32_t coeff0, uint32_t coeff1)
378 {
379  v16u8 src0, src1, src2;
380  v16u8 tmp0, tmp1;
381  v4i32 res;
382  v8u16 res_r;
383  v16i8 coeff_vec0 = __msa_fill_b(coeff0);
384  v16i8 coeff_vec1 = __msa_fill_b(coeff1);
385  v16u8 coeff_vec = (v16u8) __msa_ilvr_b(coeff_vec0, coeff_vec1);
386 
387  LD_UB3(src, stride, src0, src1, src2);
388  ILVR_B2_UB(src1, src0, src2, src1, tmp0, tmp1);
389 
390  tmp0 = (v16u8) __msa_ilvr_d((v2i64) tmp1, (v2i64) tmp0);
391  res_r = __msa_dotp_u_h(tmp0, coeff_vec);
392  res_r <<= 3;
393  res_r = (v8u16) __msa_srari_h((v8i16) res_r, 6);
394  res_r = __msa_sat_u_h(res_r, 7);
395  res = (v4i32) __msa_pckev_b((v16i8) res_r, (v16i8) res_r);
396 
397  ST_W2(res, 0, 1, dst, stride);
398 }
399 
400 static void avc_chroma_vt_4x4_msa(uint8_t *src, uint8_t *dst, int32_t stride,
401  uint32_t coeff0, uint32_t coeff1)
402 {
403  v16u8 src0, src1, src2, src3, src4;
404  v16u8 tmp0, tmp1, tmp2, tmp3;
405  v16u8 out;
406  v8u16 res0_r, res1_r;
407  v16i8 coeff_vec0 = __msa_fill_b(coeff0);
408  v16i8 coeff_vec1 = __msa_fill_b(coeff1);
409  v16u8 coeff_vec = (v16u8) __msa_ilvr_b(coeff_vec0, coeff_vec1);
410 
411  LD_UB5(src, stride, src0, src1, src2, src3, src4);
412  ILVR_B4_UB(src1, src0, src2, src1, src3, src2, src4, src3, tmp0, tmp1, tmp2,
413  tmp3);
414  ILVR_D2_UB(tmp1, tmp0, tmp3, tmp2, tmp0, tmp2);
415  DOTP_UB2_UH(tmp0, tmp2, coeff_vec, coeff_vec, res0_r, res1_r);
416  res0_r <<= 3;
417  res1_r <<= 3;
418  SRARI_H2_UH(res0_r, res1_r, 6);
419  SAT_UH2_UH(res0_r, res1_r, 7);
420  out = (v16u8) __msa_pckev_b((v16i8) res1_r, (v16i8) res0_r);
421  ST_W4(out, 0, 1, 2, 3, dst, stride);
422 }
423 
424 static void avc_chroma_vt_4x8_msa(uint8_t *src, uint8_t *dst, int32_t stride,
425  uint32_t coeff0, uint32_t coeff1)
426 {
427  v16u8 src0, src1, src2, src3, src4, src5, src6, src7, src8;
428  v16u8 tmp0, tmp1, tmp2, tmp3, tmp4, tmp5, tmp6, tmp7, out0, out1;
429  v8u16 res0, res1, res2, res3;
430  v16i8 coeff_vec0 = __msa_fill_b(coeff0);
431  v16i8 coeff_vec1 = __msa_fill_b(coeff1);
432  v16u8 coeff_vec = (v16u8) __msa_ilvr_b(coeff_vec0, coeff_vec1);
433 
434  LD_UB5(src, stride, src0, src1, src2, src3, src4);
435  src += (5 * stride);
436  LD_UB4(src, stride, src5, src6, src7, src8);
437  ILVR_B4_UB(src1, src0, src2, src1, src3, src2, src4, src3, tmp0, tmp1, tmp2,
438  tmp3);
439  ILVR_B4_UB(src5, src4, src6, src5, src7, src6, src8, src7, tmp4, tmp5, tmp6,
440  tmp7);
441  ILVR_D2_UB(tmp1, tmp0, tmp3, tmp2, tmp0, tmp2);
442  ILVR_D2_UB(tmp5, tmp4, tmp7, tmp6, tmp4, tmp6);
443  DOTP_UB2_UH(tmp0, tmp2, coeff_vec, coeff_vec, res0, res1);
444  DOTP_UB2_UH(tmp4, tmp6, coeff_vec, coeff_vec, res2, res3);
445  SLLI_4V(res0, res1, res2, res3, 3);
446  SRARI_H4_UH(res0, res1, res2, res3, 6);
447  SAT_UH4_UH(res0, res1, res2, res3, 7);
448  PCKEV_B2_UB(res1, res0, res3, res2, out0, out1);
449  ST_W8(out0, out1, 0, 1, 2, 3, 0, 1, 2, 3, dst, stride);
450 }
451 
452 static void avc_chroma_vt_4w_msa(uint8_t *src, uint8_t *dst, int32_t stride,
453  uint32_t coeff0, uint32_t coeff1,
454  int32_t height)
455 {
456  if (2 == height) {
457  avc_chroma_vt_4x2_msa(src, dst, stride, coeff0, coeff1);
458  } else if (4 == height) {
459  avc_chroma_vt_4x4_msa(src, dst, stride, coeff0, coeff1);
460  } else if (8 == height) {
461  avc_chroma_vt_4x8_msa(src, dst, stride, coeff0, coeff1);
462  }
463 }
464 
465 static void avc_chroma_vt_8x4_msa(uint8_t *src, uint8_t *dst, int32_t stride,
466  uint32_t coeff0, uint32_t coeff1)
467 {
468  v16u8 src0, src1, src2, src3, src4, out0, out1;
469  v8u16 res0, res1, res2, res3;
470  v16i8 coeff_vec0 = __msa_fill_b(coeff0);
471  v16i8 coeff_vec1 = __msa_fill_b(coeff1);
472  v16u8 coeff_vec = (v16u8) __msa_ilvr_b(coeff_vec0, coeff_vec1);
473 
474  LD_UB5(src, stride, src0, src1, src2, src3, src4);
475  ILVR_B4_UB(src1, src0, src2, src1, src3, src2, src4, src3, src0, src1, src2,
476  src3);
477  DOTP_UB4_UH(src0, src1, src2, src3, coeff_vec, coeff_vec, coeff_vec,
478  coeff_vec, res0, res1, res2, res3);
479  SLLI_4V(res0, res1, res2, res3, 3);
480  SRARI_H4_UH(res0, res1, res2, res3, 6);
481  SAT_UH4_UH(res0, res1, res2, res3, 7);
482  PCKEV_B2_UB(res1, res0, res3, res2, out0, out1);
483  ST_D4(out0, out1, 0, 1, 0, 1, dst, stride);
484 }
485 
486 static void avc_chroma_vt_8x8_msa(uint8_t *src, uint8_t *dst, int32_t stride,
487  uint32_t coeff0, uint32_t coeff1)
488 {
489  v16u8 src0, src1, src2, src3, src4, src5, src6, src7, src8;
490  v16u8 out0, out1, out2, out3;
491  v8u16 res0, res1, res2, res3, res4, res5, res6, res7;
492  v16i8 coeff_vec0 = __msa_fill_b(coeff0);
493  v16i8 coeff_vec1 = __msa_fill_b(coeff1);
494  v16u8 coeff_vec = (v16u8) __msa_ilvr_b(coeff_vec0, coeff_vec1);
495 
496  LD_UB5(src, stride, src0, src1, src2, src3, src4);
497  src += (5 * stride);
498  LD_UB4(src, stride, src5, src6, src7, src8);
499  ILVR_B4_UB(src1, src0, src2, src1, src3, src2, src4, src3, src0, src1, src2,
500  src3);
501  ILVR_B4_UB(src5, src4, src6, src5, src7, src6, src8, src7, src4, src5, src6,
502  src7);
503  DOTP_UB4_UH(src0, src1, src2, src3, coeff_vec, coeff_vec, coeff_vec,
504  coeff_vec, res0, res1, res2, res3);
505  DOTP_UB4_UH(src4, src5, src6, src7, coeff_vec, coeff_vec, coeff_vec,
506  coeff_vec, res4, res5, res6, res7);
507  SLLI_4V(res0, res1, res2, res3, 3);
508  SLLI_4V(res4, res5, res6, res7, 3);
509  SRARI_H4_UH(res0, res1, res2, res3, 6);
510  SRARI_H4_UH(res4, res5, res6, res7, 6);
511  SAT_UH4_UH(res0, res1, res2, res3, 7);
512  SAT_UH4_UH(res0, res1, res2, res3, 7);
513  PCKEV_B2_UB(res1, res0, res3, res2, out0, out1);
514  PCKEV_B2_UB(res5, res4, res7, res6, out2, out3);
515  ST_D8(out0, out1, out2, out3, 0, 1, 0, 1, 0, 1, 0, 1, dst, stride);
516 }
517 
518 static void avc_chroma_vt_8w_msa(uint8_t *src, uint8_t *dst, int32_t stride,
519  uint32_t coeff0, uint32_t coeff1,
520  int32_t height)
521 {
522  if (4 == height) {
523  avc_chroma_vt_8x4_msa(src, dst, stride, coeff0, coeff1);
524  } else if (8 == height) {
525  avc_chroma_vt_8x8_msa(src, dst, stride, coeff0, coeff1);
526  }
527 }
528 
529 static void avc_chroma_hv_2x2_msa(uint8_t *src, uint8_t *dst, int32_t stride,
530  uint32_t coef_hor0, uint32_t coef_hor1,
531  uint32_t coef_ver0, uint32_t coef_ver1)
532 {
533  uint16_t out0, out1;
534  v16u8 src0, src1, src2;
535  v8u16 res_hz0, res_hz1, res_vt0, res_vt1;
536  v8i16 res_vert;
537  v16i8 mask;
538  v16i8 coeff_hz_vec0 = __msa_fill_b(coef_hor0);
539  v16i8 coeff_hz_vec1 = __msa_fill_b(coef_hor1);
540  v16u8 coeff_hz_vec = (v16u8) __msa_ilvr_b(coeff_hz_vec0, coeff_hz_vec1);
541  v8u16 coeff_vt_vec0 = (v8u16) __msa_fill_h(coef_ver0);
542  v8u16 coeff_vt_vec1 = (v8u16) __msa_fill_h(coef_ver1);
543 
544  mask = LD_SB(&chroma_mask_arr[48]);
545 
546  LD_UB3(src, stride, src0, src1, src2);
547  VSHF_B2_UB(src0, src1, src1, src2, mask, mask, src0, src1);
548  DOTP_UB2_UH(src0, src1, coeff_hz_vec, coeff_hz_vec, res_hz0, res_hz1);
549  MUL2(res_hz0, coeff_vt_vec1, res_hz1, coeff_vt_vec0, res_vt0, res_vt1);
550 
551  res_vt0 += res_vt1;
552  res_vt0 = (v8u16) __msa_srari_h((v8i16) res_vt0, 6);
553  res_vt0 = __msa_sat_u_h(res_vt0, 7);
554  res_vert = (v8i16) __msa_pckev_b((v16i8) res_vt0, (v16i8) res_vt0);
555 
556  out0 = __msa_copy_u_h(res_vert, 0);
557  out1 = __msa_copy_u_h(res_vert, 1);
558 
559  SH(out0, dst);
560  dst += stride;
561  SH(out1, dst);
562 }
563 
564 static void avc_chroma_hv_2x4_msa(uint8_t *src, uint8_t *dst, int32_t stride,
565  uint32_t coef_hor0, uint32_t coef_hor1,
566  uint32_t coef_ver0, uint32_t coef_ver1)
567 {
568  v16u8 src0, src1, src2, src3, src4;
569  v16u8 tmp0, tmp1, tmp2, tmp3;
570  v8u16 res_hz0, res_hz1, res_vt0, res_vt1;
571  v8i16 res;
572  v16i8 mask;
573  v16i8 coeff_hz_vec0 = __msa_fill_b(coef_hor0);
574  v16i8 coeff_hz_vec1 = __msa_fill_b(coef_hor1);
575  v16u8 coeff_hz_vec = (v16u8) __msa_ilvr_b(coeff_hz_vec0, coeff_hz_vec1);
576  v8u16 coeff_vt_vec0 = (v8u16) __msa_fill_h(coef_ver0);
577  v8u16 coeff_vt_vec1 = (v8u16) __msa_fill_h(coef_ver1);
578 
579  mask = LD_SB(&chroma_mask_arr[48]);
580 
581  LD_UB5(src, stride, src0, src1, src2, src3, src4);
582 
583  VSHF_B2_UB(src0, src1, src2, src3, mask, mask, tmp0, tmp1);
584  VSHF_B2_UB(src1, src2, src3, src4, mask, mask, tmp2, tmp3);
585  ILVR_D2_UB(tmp1, tmp0, tmp3, tmp2, src0, src1);
586  DOTP_UB2_UH(src0, src1, coeff_hz_vec, coeff_hz_vec, res_hz0, res_hz1);
587  MUL2(res_hz0, coeff_vt_vec1, res_hz1, coeff_vt_vec0, res_vt0, res_vt1);
588 
589  res_vt0 += res_vt1;
590  res_vt0 = (v8u16) __msa_srari_h((v8i16) res_vt0, 6);
591  res_vt0 = __msa_sat_u_h(res_vt0, 7);
592 
593  res = (v8i16) __msa_pckev_b((v16i8) res_vt0, (v16i8) res_vt0);
594 
595  ST_H4(res, 0, 1, 2, 3, dst, stride);
596 }
597 
598 static void avc_chroma_hv_2w_msa(uint8_t *src, uint8_t *dst, int32_t stride,
599  uint32_t coef_hor0, uint32_t coef_hor1,
600  uint32_t coef_ver0, uint32_t coef_ver1,
601  int32_t height)
602 {
603  if (2 == height) {
604  avc_chroma_hv_2x2_msa(src, dst, stride, coef_hor0, coef_hor1, coef_ver0,
605  coef_ver1);
606  } else if (4 == height) {
607  avc_chroma_hv_2x4_msa(src, dst, stride, coef_hor0, coef_hor1, coef_ver0,
608  coef_ver1);
609  }
610 }
611 
612 static void avc_chroma_hv_4x2_msa(uint8_t *src, uint8_t *dst, int32_t stride,
613  uint32_t coef_hor0, uint32_t coef_hor1,
614  uint32_t coef_ver0, uint32_t coef_ver1)
615 {
616  v16u8 src0, src1, src2;
617  v8u16 res_hz0, res_hz1, res_vt0, res_vt1;
618  v16i8 mask;
619  v4i32 res;
620  v16i8 coeff_hz_vec0 = __msa_fill_b(coef_hor0);
621  v16i8 coeff_hz_vec1 = __msa_fill_b(coef_hor1);
622  v16u8 coeff_hz_vec = (v16u8) __msa_ilvr_b(coeff_hz_vec0, coeff_hz_vec1);
623  v8u16 coeff_vt_vec0 = (v8u16) __msa_fill_h(coef_ver0);
624  v8u16 coeff_vt_vec1 = (v8u16) __msa_fill_h(coef_ver1);
625 
626  mask = LD_SB(&chroma_mask_arr[0]);
627  LD_UB3(src, stride, src0, src1, src2);
628  VSHF_B2_UB(src0, src1, src1, src2, mask, mask, src0, src1);
629  DOTP_UB2_UH(src0, src1, coeff_hz_vec, coeff_hz_vec, res_hz0, res_hz1);
630  MUL2(res_hz0, coeff_vt_vec1, res_hz1, coeff_vt_vec0, res_vt0, res_vt1);
631 
632  res_vt0 += res_vt1;
633  res_vt0 = (v8u16) __msa_srari_h((v8i16) res_vt0, 6);
634  res_vt0 = __msa_sat_u_h(res_vt0, 7);
635  res = (v4i32) __msa_pckev_b((v16i8) res_vt0, (v16i8) res_vt0);
636 
637  ST_W2(res, 0, 1, dst, stride);
638 }
639 
640 static void avc_chroma_hv_4x4_msa(uint8_t *src, uint8_t *dst, int32_t stride,
641  uint32_t coef_hor0, uint32_t coef_hor1,
642  uint32_t coef_ver0, uint32_t coef_ver1)
643 {
644  v16u8 src0, src1, src2, src3, src4;
645  v8u16 res_hz0, res_hz1, res_hz2, res_hz3;
646  v8u16 res_vt0, res_vt1, res_vt2, res_vt3;
647  v16i8 mask;
648  v16i8 coeff_hz_vec0 = __msa_fill_b(coef_hor0);
649  v16i8 coeff_hz_vec1 = __msa_fill_b(coef_hor1);
650  v16u8 coeff_hz_vec = (v16u8) __msa_ilvr_b(coeff_hz_vec0, coeff_hz_vec1);
651  v8u16 coeff_vt_vec0 = (v8u16) __msa_fill_h(coef_ver0);
652  v8u16 coeff_vt_vec1 = (v8u16) __msa_fill_h(coef_ver1);
653  v4i32 res0, res1;
654 
655  mask = LD_SB(&chroma_mask_arr[0]);
656 
657  LD_UB5(src, stride, src0, src1, src2, src3, src4);
658  VSHF_B2_UB(src0, src1, src1, src2, mask, mask, src0, src1);
659  VSHF_B2_UB(src2, src3, src3, src4, mask, mask, src2, src3);
660  DOTP_UB4_UH(src0, src1, src2, src3, coeff_hz_vec, coeff_hz_vec,
661  coeff_hz_vec, coeff_hz_vec, res_hz0, res_hz1, res_hz2,
662  res_hz3);
663  MUL4(res_hz0, coeff_vt_vec1, res_hz1, coeff_vt_vec0, res_hz2, coeff_vt_vec1,
664  res_hz3, coeff_vt_vec0, res_vt0, res_vt1, res_vt2, res_vt3);
665  ADD2(res_vt0, res_vt1, res_vt2, res_vt3, res_vt0, res_vt1);
666  SRARI_H2_UH(res_vt0, res_vt1, 6);
667  SAT_UH2_UH(res_vt0, res_vt1, 7);
668  PCKEV_B2_SW(res_vt0, res_vt0, res_vt1, res_vt1, res0, res1);
669  ST_W2(res0, 0, 1, dst, stride);
670  ST_W2(res1, 0, 1, dst + 2 * stride, stride);
671 }
672 
673 static void avc_chroma_hv_4x8_msa(uint8_t *src, uint8_t *dst, int32_t stride,
674  uint32_t coef_hor0, uint32_t coef_hor1,
675  uint32_t coef_ver0, uint32_t coef_ver1)
676 {
677  v16u8 src0, src1, src2, src3, src4, src5, src6, src7, src8, res0, res1;
678  v8u16 res_hz0, res_hz1, res_hz2, res_hz3, res_hz4, res_hz5, res_hz6, res_hz7;
679  v8u16 res_vt0, res_vt1, res_vt2, res_vt3, res_vt4, res_vt5, res_vt6, res_vt7;
680  v16i8 mask;
681  v16i8 coeff_hz_vec0 = __msa_fill_b(coef_hor0);
682  v16i8 coeff_hz_vec1 = __msa_fill_b(coef_hor1);
683  v16u8 coeff_hz_vec = (v16u8) __msa_ilvr_b(coeff_hz_vec0, coeff_hz_vec1);
684  v8u16 coeff_vt_vec0 = (v8u16) __msa_fill_h(coef_ver0);
685  v8u16 coeff_vt_vec1 = (v8u16) __msa_fill_h(coef_ver1);
686 
687  mask = LD_SB(&chroma_mask_arr[0]);
688 
689  LD_UB5(src, stride, src0, src1, src2, src3, src4);
690  src += (5 * stride);
691  LD_UB4(src, stride, src5, src6, src7, src8);
692 
693  VSHF_B2_UB(src0, src1, src1, src2, mask, mask, src0, src1);
694  VSHF_B2_UB(src2, src3, src3, src4, mask, mask, src2, src3);
695  VSHF_B2_UB(src4, src5, src5, src6, mask, mask, src4, src5);
696  VSHF_B2_UB(src6, src7, src7, src8, mask, mask, src6, src7);
697  DOTP_UB4_UH(src0, src1, src2, src3, coeff_hz_vec, coeff_hz_vec,
698  coeff_hz_vec, coeff_hz_vec, res_hz0, res_hz1, res_hz2, res_hz3);
699  DOTP_UB4_UH(src4, src5, src6, src7, coeff_hz_vec, coeff_hz_vec,
700  coeff_hz_vec, coeff_hz_vec, res_hz4, res_hz5, res_hz6, res_hz7);
701  MUL4(res_hz0, coeff_vt_vec1, res_hz1, coeff_vt_vec0, res_hz2, coeff_vt_vec1,
702  res_hz3, coeff_vt_vec0, res_vt0, res_vt1, res_vt2, res_vt3);
703  MUL4(res_hz4, coeff_vt_vec1, res_hz5, coeff_vt_vec0, res_hz6, coeff_vt_vec1,
704  res_hz7, coeff_vt_vec0, res_vt4, res_vt5, res_vt6, res_vt7);
705  ADD2(res_vt0, res_vt1, res_vt2, res_vt3, res_vt0, res_vt1);
706  ADD2(res_vt4, res_vt5, res_vt6, res_vt7, res_vt2, res_vt3);
707  SRARI_H4_UH(res_vt0, res_vt1, res_vt2, res_vt3, 6);
708  SAT_UH4_UH(res_vt0, res_vt1, res_vt2, res_vt3, 7);
709  PCKEV_B2_UB(res_vt1, res_vt0, res_vt3, res_vt2, res0, res1);
710  ST_W8(res0, res1, 0, 1, 2, 3, 0, 1, 2, 3, dst, stride);
711 }
712 
713 static void avc_chroma_hv_4w_msa(uint8_t *src, uint8_t *dst, int32_t stride,
714  uint32_t coef_hor0, uint32_t coef_hor1,
715  uint32_t coef_ver0, uint32_t coef_ver1,
716  int32_t height)
717 {
718  if (2 == height) {
719  avc_chroma_hv_4x2_msa(src, dst, stride, coef_hor0, coef_hor1, coef_ver0,
720  coef_ver1);
721  } else if (4 == height) {
722  avc_chroma_hv_4x4_msa(src, dst, stride, coef_hor0, coef_hor1, coef_ver0,
723  coef_ver1);
724  } else if (8 == height) {
725  avc_chroma_hv_4x8_msa(src, dst, stride, coef_hor0, coef_hor1, coef_ver0,
726  coef_ver1);
727  }
728 }
729 
730 static void avc_chroma_hv_8x4_msa(uint8_t *src, uint8_t *dst, int32_t stride,
731  uint32_t coef_hor0, uint32_t coef_hor1,
732  uint32_t coef_ver0, uint32_t coef_ver1)
733 {
734  v16u8 src0, src1, src2, src3, src4, out0, out1;
735  v8u16 res_hz0, res_hz1, res_hz2, res_hz3, res_hz4;
736  v8u16 res_vt0, res_vt1, res_vt2, res_vt3;
737  v16i8 mask;
738  v16i8 coeff_hz_vec0 = __msa_fill_b(coef_hor0);
739  v16i8 coeff_hz_vec1 = __msa_fill_b(coef_hor1);
740  v16u8 coeff_hz_vec = (v16u8) __msa_ilvr_b(coeff_hz_vec0, coeff_hz_vec1);
741  v8u16 coeff_vt_vec0 = (v8u16) __msa_fill_h(coef_ver0);
742  v8u16 coeff_vt_vec1 = (v8u16) __msa_fill_h(coef_ver1);
743 
744  mask = LD_SB(&chroma_mask_arr[32]);
745 
746  src0 = LD_UB(src);
747  src += stride;
748 
749  src0 = (v16u8) __msa_vshf_b(mask, (v16i8) src0, (v16i8) src0);
750  res_hz0 = __msa_dotp_u_h(src0, coeff_hz_vec);
751 
752  LD_UB4(src, stride, src1, src2, src3, src4);
753  src += (4 * stride);
754 
755  VSHF_B2_UB(src1, src1, src2, src2, mask, mask, src1, src2);
756  VSHF_B2_UB(src3, src3, src4, src4, mask, mask, src3, src4);
757  DOTP_UB4_UH(src1, src2, src3, src4, coeff_hz_vec, coeff_hz_vec,
758  coeff_hz_vec, coeff_hz_vec, res_hz1, res_hz2, res_hz3, res_hz4);
759  MUL4(res_hz1, coeff_vt_vec0, res_hz2, coeff_vt_vec0, res_hz3, coeff_vt_vec0,
760  res_hz4, coeff_vt_vec0, res_vt0, res_vt1, res_vt2, res_vt3);
761 
762  res_vt0 += (res_hz0 * coeff_vt_vec1);
763  res_vt1 += (res_hz1 * coeff_vt_vec1);
764  res_vt2 += (res_hz2 * coeff_vt_vec1);
765  res_vt3 += (res_hz3 * coeff_vt_vec1);
766 
767  SRARI_H4_UH(res_vt0, res_vt1, res_vt2, res_vt3, 6);
768  SAT_UH4_UH(res_vt0, res_vt1, res_vt2, res_vt3, 7);
769  PCKEV_B2_UB(res_vt1, res_vt0, res_vt3, res_vt2, out0, out1);
770  ST_D4(out0, out1, 0, 1, 0, 1, dst, stride);
771 }
772 
773 static void avc_chroma_hv_8x8_msa(uint8_t *src, uint8_t *dst, int32_t stride,
774  uint32_t coef_hor0, uint32_t coef_hor1,
775  uint32_t coef_ver0, uint32_t coef_ver1)
776 {
777  v16u8 src0, src1, src2, src3, src4, src5, src6, src7, src8;
778  v16u8 out0, out1, out2, out3;
779  v8u16 res_hz0, res_hz1, res_hz2, res_hz3, res_hz4;
780  v8u16 res_hz5, res_hz6, res_hz7, res_hz8;
781  v8u16 res_vt0, res_vt1, res_vt2, res_vt3;
782  v8u16 res_vt4, res_vt5, res_vt6, res_vt7;
783  v16i8 mask;
784  v16i8 coeff_hz_vec0 = __msa_fill_b(coef_hor0);
785  v16i8 coeff_hz_vec1 = __msa_fill_b(coef_hor1);
786  v16u8 coeff_hz_vec = (v16u8) __msa_ilvr_b(coeff_hz_vec0, coeff_hz_vec1);
787  v8u16 coeff_vt_vec0 = (v8u16) __msa_fill_h(coef_ver0);
788  v8u16 coeff_vt_vec1 = (v8u16) __msa_fill_h(coef_ver1);
789 
790  mask = LD_SB(&chroma_mask_arr[32]);
791 
792  LD_UB5(src, stride, src0, src1, src2, src3, src4);
793  src += (5 * stride);
794  LD_UB4(src, stride, src5, src6, src7, src8);
795  src0 = (v16u8) __msa_vshf_b(mask, (v16i8) src0, (v16i8) src0);
796  VSHF_B2_UB(src1, src1, src2, src2, mask, mask, src1, src2);
797  VSHF_B2_UB(src3, src3, src4, src4, mask, mask, src3, src4);
798  VSHF_B2_UB(src5, src5, src6, src6, mask, mask, src5, src6);
799  VSHF_B2_UB(src7, src7, src8, src8, mask, mask, src7, src8);
800  res_hz0 = __msa_dotp_u_h(src0, coeff_hz_vec);
801  DOTP_UB4_UH(src1, src2, src3, src4, coeff_hz_vec, coeff_hz_vec,
802  coeff_hz_vec, coeff_hz_vec, res_hz1, res_hz2, res_hz3,
803  res_hz4);
804  DOTP_UB4_UH(src5, src6, src7, src8, coeff_hz_vec, coeff_hz_vec,
805  coeff_hz_vec, coeff_hz_vec, res_hz5, res_hz6, res_hz7, res_hz8);
806  MUL4(res_hz1, coeff_vt_vec0, res_hz2, coeff_vt_vec0, res_hz3,
807  coeff_vt_vec0, res_hz4, coeff_vt_vec0, res_vt0, res_vt1, res_vt2,
808  res_vt3);
809  MUL4(res_hz5, coeff_vt_vec0, res_hz6, coeff_vt_vec0, res_hz7,
810  coeff_vt_vec0, res_hz8, coeff_vt_vec0, res_vt4, res_vt5, res_vt6,
811  res_vt7);
812  res_vt0 += (res_hz0 * coeff_vt_vec1);
813  res_vt1 += (res_hz1 * coeff_vt_vec1);
814  res_vt2 += (res_hz2 * coeff_vt_vec1);
815  res_vt3 += (res_hz3 * coeff_vt_vec1);
816  res_vt4 += (res_hz4 * coeff_vt_vec1);
817  res_vt5 += (res_hz5 * coeff_vt_vec1);
818  res_vt6 += (res_hz6 * coeff_vt_vec1);
819  res_vt7 += (res_hz7 * coeff_vt_vec1);
820  SRARI_H4_UH(res_vt0, res_vt1, res_vt2, res_vt3, 6);
821  SRARI_H4_UH(res_vt4, res_vt5, res_vt6, res_vt7, 6);
822  SAT_UH4_UH(res_vt0, res_vt1, res_vt2, res_vt3, 7);
823  SAT_UH4_UH(res_vt4, res_vt5, res_vt6, res_vt7, 7);
824  PCKEV_B2_UB(res_vt1, res_vt0, res_vt3, res_vt2, out0, out1);
825  PCKEV_B2_UB(res_vt5, res_vt4, res_vt7, res_vt6, out2, out3);
826  ST_D8(out0, out1, out2, out3, 0, 1, 0, 1, 0, 1, 0, 1, dst, stride);
827 }
828 
829 static void avc_chroma_hv_8w_msa(uint8_t *src, uint8_t *dst, int32_t stride,
830  uint32_t coef_hor0, uint32_t coef_hor1,
831  uint32_t coef_ver0, uint32_t coef_ver1,
832  int32_t height)
833 {
834  if (4 == height) {
835  avc_chroma_hv_8x4_msa(src, dst, stride, coef_hor0, coef_hor1, coef_ver0,
836  coef_ver1);
837  } else if (8 == height) {
838  avc_chroma_hv_8x8_msa(src, dst, stride, coef_hor0, coef_hor1, coef_ver0,
839  coef_ver1);
840  }
841 }
842 
843 static void avc_chroma_hz_and_aver_dst_2x2_msa(uint8_t *src, uint8_t *dst,
844  int32_t stride, uint32_t coeff0,
845  uint32_t coeff1)
846 {
847  uint16_t out0, out1;
848  v16i8 src0, src1;
849  v16u8 dst_data = { 0 };
850  v8u16 res_r;
851  v16u8 res;
852  v16i8 mask;
853  v16i8 coeff_vec0 = __msa_fill_b(coeff0);
854  v16i8 coeff_vec1 = __msa_fill_b(coeff1);
855  v16u8 coeff_vec = (v16u8) __msa_ilvr_b(coeff_vec0, coeff_vec1);
856 
857  mask = LD_SB(&chroma_mask_arr[0]);
858 
859  LD_SB2(src, stride, src0, src1);
860 
861  out0 = LH(dst);
862  out1 = LH(dst + stride);
863 
864  dst_data = (v16u8) __msa_insert_h((v8i16) dst_data, 0, out0);
865  dst_data = (v16u8) __msa_insert_h((v8i16) dst_data, 2, out1);
866 
867  src0 = __msa_vshf_b(mask, src1, src0);
868 
869  res_r = __msa_dotp_u_h((v16u8) src0, coeff_vec);
870  res_r <<= 3;
871  res_r = (v8u16) __msa_srari_h((v8i16) res_r, 6);
872  res_r = __msa_sat_u_h(res_r, 7);
873 
874  res = (v16u8) __msa_pckev_b((v16i8) res_r, (v16i8) res_r);
875  dst_data = __msa_aver_u_b(res, dst_data);
876 
877  out0 = __msa_copy_u_h((v8i16) dst_data, 0);
878  out1 = __msa_copy_u_h((v8i16) dst_data, 2);
879 
880  SH(out0, dst);
881  dst += stride;
882  SH(out1, dst);
883 }
884 
885 static void avc_chroma_hz_and_aver_dst_2x4_msa(uint8_t *src, uint8_t *dst,
886  int32_t stride, uint32_t coeff0,
887  uint32_t coeff1)
888 {
889  uint16_t tp0, tp1, tp2, tp3;
890  v16u8 src0, src1, src2, src3;
891  v16u8 dst0, dst_data = { 0 };
892  v8u16 res_r;
893  v16i8 mask;
894  v16i8 coeff_vec0 = __msa_fill_b(coeff0);
895  v16i8 coeff_vec1 = __msa_fill_b(coeff1);
896  v16u8 coeff_vec = (v16u8) __msa_ilvr_b(coeff_vec0, coeff_vec1);
897 
898  mask = LD_SB(&chroma_mask_arr[64]);
899 
900  LD_UB4(src, stride, src0, src1, src2, src3);
901  tp0 = LH(dst);
902  tp1 = LH(dst + stride);
903  tp2 = LH(dst + 2 * stride);
904  tp3 = LH(dst + 3 * stride);
905  dst_data = (v16u8) __msa_insert_h((v8i16) dst_data, 0, tp0);
906  dst_data = (v16u8) __msa_insert_h((v8i16) dst_data, 1, tp1);
907  dst_data = (v16u8) __msa_insert_h((v8i16) dst_data, 2, tp2);
908  dst_data = (v16u8) __msa_insert_h((v8i16) dst_data, 3, tp3);
909 
910  VSHF_B2_UB(src0, src1, src2, src3, mask, mask, src0, src2);
911 
912  src0 = (v16u8) __msa_ilvr_d((v2i64) src2, (v2i64) src0);
913 
914  res_r = __msa_dotp_u_h(src0, coeff_vec);
915  res_r <<= 3;
916  res_r = (v8u16) __msa_srari_h((v8i16) res_r, 6);
917  res_r = __msa_sat_u_h(res_r, 7);
918 
919  dst0 = (v16u8) __msa_pckev_b((v16i8) res_r, (v16i8) res_r);
920  dst0 = __msa_aver_u_b(dst0, dst_data);
921 
922  ST_H4(dst0, 0, 1, 2, 3, dst, stride);
923 }
924 
925 static void avc_chroma_hz_and_aver_dst_2w_msa(uint8_t *src, uint8_t *dst,
926  int32_t stride, uint32_t coeff0,
927  uint32_t coeff1, int32_t height)
928 {
929  if (2 == height) {
930  avc_chroma_hz_and_aver_dst_2x2_msa(src, dst, stride, coeff0, coeff1);
931  } else if (4 == height) {
932  avc_chroma_hz_and_aver_dst_2x4_msa(src, dst, stride, coeff0, coeff1);
933  }
934 }
935 
936 static void avc_chroma_hz_and_aver_dst_4x2_msa(uint8_t *src, uint8_t *dst,
937  int32_t stride, uint32_t coeff0,
938  uint32_t coeff1)
939 {
940  uint32_t load0, load1;
941  v16i8 src0, src1;
942  v16u8 dst_data = { 0 };
943  v8u16 res_r;
944  v16i8 res, mask;
945  v16i8 coeff_vec0 = __msa_fill_b(coeff0);
946  v16i8 coeff_vec1 = __msa_fill_b(coeff1);
947  v16u8 coeff_vec = (v16u8) __msa_ilvr_b(coeff_vec0, coeff_vec1);
948 
949  mask = LD_SB(&chroma_mask_arr[0]);
950 
951  LD_SB2(src, stride, src0, src1);
952 
953  LW2(dst, stride, load0, load1);
954 
955  INSERT_W2_UB(load0, load1, dst_data);
956 
957  src0 = __msa_vshf_b(mask, src1, src0);
958 
959  res_r = __msa_dotp_u_h((v16u8) src0, coeff_vec);
960  res_r <<= 3;
961  res_r = (v8u16) __msa_srari_h((v8i16) res_r, 6);
962  res_r = __msa_sat_u_h(res_r, 7);
963  res = __msa_pckev_b((v16i8) res_r, (v16i8) res_r);
964  dst_data = __msa_aver_u_b((v16u8) res, dst_data);
965 
966  ST_W2(dst_data, 0, 1, dst, stride);
967 }
968 
969 static void avc_chroma_hz_and_aver_dst_4x4_msa(uint8_t *src, uint8_t *dst,
970  int32_t stride, uint32_t coeff0,
971  uint32_t coeff1)
972 {
973  uint32_t tp0, tp1, tp2, tp3;
974  v16u8 src0, src1, src2, src3;
975  v16u8 out, dst_data = { 0 };
976  v16i8 mask;
977  v8u16 res0_r, res1_r;
978  v16i8 coeff_vec0 = __msa_fill_b(coeff0);
979  v16i8 coeff_vec1 = __msa_fill_b(coeff1);
980  v16u8 coeff_vec = (v16u8) __msa_ilvr_b(coeff_vec0, coeff_vec1);
981 
982  mask = LD_SB(&chroma_mask_arr[0]);
983 
984  LD_UB4(src, stride, src0, src1, src2, src3);
985  LW4(dst, stride, tp0, tp1, tp2, tp3);
986  INSERT_W4_UB(tp0, tp1, tp2, tp3, dst_data);
987  VSHF_B2_UB(src0, src1, src2, src3, mask, mask, src0, src2);
988  DOTP_UB2_UH(src0, src2, coeff_vec, coeff_vec, res0_r, res1_r);
989  res0_r <<= 3;
990  res1_r <<= 3;
991  SRARI_H2_UH(res0_r, res1_r, 6);
992  SAT_UH2_UH(res0_r, res1_r, 7);
993  out = (v16u8) __msa_pckev_b((v16i8) res1_r, (v16i8) res0_r);
994  out = __msa_aver_u_b(out, dst_data);
995  ST_W4(out, 0, 1, 2, 3, dst, stride);
996 }
997 
998 static void avc_chroma_hz_and_aver_dst_4x8_msa(uint8_t *src, uint8_t *dst,
999  int32_t stride, uint32_t coeff0,
1000  uint32_t coeff1)
1001 {
1002  uint32_t tp0, tp1, tp2, tp3;
1003  v16u8 src0, src1, src2, src3, src4, src5, src6, src7, out0, out1;
1004  v16u8 dst0 = { 0 }, dst1 = { 0 };
1005  v16i8 mask;
1006  v8u16 res0, res1, res2, res3;
1007  v16i8 coeff_vec0 = __msa_fill_b(coeff0);
1008  v16i8 coeff_vec1 = __msa_fill_b(coeff1);
1009  v16u8 coeff_vec = (v16u8) __msa_ilvr_b(coeff_vec0, coeff_vec1);
1010 
1011  mask = LD_SB(&chroma_mask_arr[0]);
1012 
1013  LD_UB8(src, stride, src0, src1, src2, src3, src4, src5, src6, src7);
1014  LW4(dst, stride, tp0, tp1, tp2, tp3);
1015  INSERT_W4_UB(tp0, tp1, tp2, tp3, dst0);
1016  LW4(dst + 4 * stride, stride, tp0, tp1, tp2, tp3);
1017  INSERT_W4_UB(tp0, tp1, tp2, tp3, dst1);
1018  VSHF_B2_UB(src0, src1, src2, src3, mask, mask, src0, src2);
1019  VSHF_B2_UB(src4, src5, src6, src7, mask, mask, src4, src6);
1020  DOTP_UB2_UH(src0, src2, coeff_vec, coeff_vec, res0, res1);
1021  DOTP_UB2_UH(src4, src6, coeff_vec, coeff_vec, res2, res3);
1022  SLLI_4V(res0, res1, res2, res3, 3);
1023  SRARI_H4_UH(res0, res1, res2, res3, 6);
1024  SAT_UH4_UH(res0, res1, res2, res3, 7);
1025  PCKEV_B2_UB(res1, res0, res3, res2, out0, out1);
1026  AVER_UB2_UB(out0, dst0, out1, dst1, out0, out1);
1027  ST_W8(out0, out1, 0, 1, 2, 3, 0, 1, 2, 3, dst, stride);
1028 }
1029 
1030 static void avc_chroma_hz_and_aver_dst_4w_msa(uint8_t *src, uint8_t *dst,
1031  int32_t stride, uint32_t coeff0,
1032  uint32_t coeff1, int32_t height)
1033 {
1034  if (2 == height) {
1035  avc_chroma_hz_and_aver_dst_4x2_msa(src, dst, stride, coeff0, coeff1);
1036  } else if (4 == height) {
1037  avc_chroma_hz_and_aver_dst_4x4_msa(src, dst, stride, coeff0, coeff1);
1038  } else if (8 == height) {
1039  avc_chroma_hz_and_aver_dst_4x8_msa(src, dst, stride, coeff0, coeff1);
1040  }
1041 }
1042 
1043 static void avc_chroma_hz_and_aver_dst_8x4_msa(uint8_t *src, uint8_t *dst,
1044  int32_t stride, uint32_t coeff0,
1045  uint32_t coeff1)
1046 {
1047  uint64_t tp0, tp1, tp2, tp3;
1048  v16u8 src0, src1, src2, src3, out0, out1;
1049  v16u8 dst0 = { 0 }, dst1 = { 0 };
1050  v8u16 res0, res1, res2, res3;
1051  v16i8 mask;
1052  v16i8 coeff_vec0 = __msa_fill_b(coeff0);
1053  v16i8 coeff_vec1 = __msa_fill_b(coeff1);
1054  v16u8 coeff_vec = (v16u8) __msa_ilvr_b(coeff_vec0, coeff_vec1);
1055 
1056  mask = LD_SB(&chroma_mask_arr[32]);
1057  LD_UB4(src, stride, src0, src1, src2, src3);
1058  LD4(dst, stride, tp0, tp1, tp2, tp3);
1059  INSERT_D2_UB(tp0, tp1, dst0);
1060  INSERT_D2_UB(tp2, tp3, dst1);
1062  VSHF_B2_UB(src2, src2, src3, src3, mask, mask, src2, src3);
1063  DOTP_UB4_UH(src0, src1, src2, src3, coeff_vec, coeff_vec, coeff_vec,
1064  coeff_vec, res0, res1, res2, res3);
1065  SLLI_4V(res0, res1, res2, res3, 3);
1066  SRARI_H4_UH(res0, res1, res2, res3, 6);
1067  SAT_UH4_UH(res0, res1, res2, res3, 7);
1068  PCKEV_B2_UB(res1, res0, res3, res2, out0, out1);
1069  AVER_UB2_UB(out0, dst0, out1, dst1, dst0, dst1);
1070  ST_D4(dst0, dst1, 0, 1, 0, 1, dst, stride);
1071 }
1072 
1073 static void avc_chroma_hz_and_aver_dst_8x8_msa(uint8_t *src, uint8_t *dst,
1074  int32_t stride, uint32_t coeff0,
1075  uint32_t coeff1)
1076 {
1077  uint64_t tp0, tp1, tp2, tp3;
1078  v16u8 src0, src1, src2, src3, src4, src5, src6, src7;
1079  v16u8 out0, out1, out2, out3;
1080  v16u8 dst0 = { 0 }, dst1 = { 0 }, dst2 = { 0 }, dst3 = { 0 };
1081  v8u16 res0, res1, res2, res3, res4, res5, res6, res7;
1082  v16i8 mask;
1083  v16i8 coeff_vec0 = __msa_fill_b(coeff0);
1084  v16i8 coeff_vec1 = __msa_fill_b(coeff1);
1085  v16u8 coeff_vec = (v16u8) __msa_ilvr_b(coeff_vec0, coeff_vec1);
1086 
1087  mask = LD_SB(&chroma_mask_arr[32]);
1088 
1089  LD_UB8(src, stride, src0, src1, src2, src3, src4, src5, src6, src7);
1090  LD4(dst, stride, tp0, tp1, tp2, tp3);
1091  INSERT_D2_UB(tp0, tp1, dst0);
1092  INSERT_D2_UB(tp2, tp3, dst1);
1093  LD4(dst + 4 * stride, stride, tp0, tp1, tp2, tp3);
1094  INSERT_D2_UB(tp0, tp1, dst2);
1095  INSERT_D2_UB(tp2, tp3, dst3);
1097  VSHF_B2_UB(src2, src2, src3, src3, mask, mask, src2, src3);
1098  VSHF_B2_UB(src4, src4, src5, src5, mask, mask, src4, src5);
1099  VSHF_B2_UB(src6, src6, src7, src7, mask, mask, src6, src7);
1100  DOTP_UB4_UH(src0, src1, src2, src3, coeff_vec, coeff_vec, coeff_vec,
1101  coeff_vec, res0, res1, res2, res3);
1102  DOTP_UB4_UH(src4, src5, src6, src7, coeff_vec, coeff_vec, coeff_vec,
1103  coeff_vec, res4, res5, res6, res7);
1104  SLLI_4V(res0, res1, res2, res3, 3);
1105  SLLI_4V(res4, res5, res6, res7, 3);
1106  SRARI_H4_UH(res0, res1, res2, res3, 6);
1107  SRARI_H4_UH(res4, res5, res6, res7, 6);
1108  SAT_UH4_UH(res0, res1, res2, res3, 7);
1109  SAT_UH4_UH(res4, res5, res6, res7, 7);
1110  PCKEV_B2_UB(res1, res0, res3, res2, out0, out1);
1111  PCKEV_B2_UB(res5, res4, res7, res6, out2, out3);
1112  AVER_UB2_UB(out0, dst0, out1, dst1, out0, out1);
1113  AVER_UB2_UB(out2, dst2, out3, dst3, out2, out3);
1114  ST_D8(out0, out1, out2, out3, 0, 1, 0, 1, 0, 1, 0, 1, dst, stride);
1115 }
1116 
1117 static void avc_chroma_hz_and_aver_dst_8w_msa(uint8_t *src, uint8_t *dst,
1118  int32_t stride, uint32_t coeff0,
1119  uint32_t coeff1, int32_t height)
1120 {
1121  if (4 == height) {
1122  avc_chroma_hz_and_aver_dst_8x4_msa(src, dst, stride, coeff0, coeff1);
1123  } else if (8 == height) {
1124  avc_chroma_hz_and_aver_dst_8x8_msa(src, dst, stride, coeff0, coeff1);
1125  }
1126 }
1127 
1128 static void avc_chroma_vt_and_aver_dst_2x2_msa(uint8_t *src, uint8_t *dst,
1129  int32_t stride, uint32_t coeff0,
1130  uint32_t coeff1)
1131 {
1132  uint16_t out0, out1;
1133  v16i8 src0, src1, src2, tmp0, tmp1, res;
1134  v16u8 dst_data = { 0 };
1135  v8i16 out;
1136  v8u16 res_r;
1137  v16i8 coeff_vec0 = __msa_fill_b(coeff0);
1138  v16i8 coeff_vec1 = __msa_fill_b(coeff1);
1139  v16u8 coeff_vec = (v16u8) __msa_ilvr_b(coeff_vec0, coeff_vec1);
1140 
1141  LD_SB3(src, stride, src0, src1, src2);
1142  out0 = LH(dst);
1143  out1 = LH(dst + stride);
1144 
1145  dst_data = (v16u8) __msa_insert_h((v8i16) dst_data, 0, out0);
1146  dst_data = (v16u8) __msa_insert_h((v8i16) dst_data, 2, out1);
1147 
1148  ILVR_B2_SB(src1, src0, src2, src1, tmp0, tmp1);
1149 
1150  tmp0 = (v16i8) __msa_ilvr_d((v2i64) tmp1, (v2i64) tmp0);
1151  res_r = __msa_dotp_u_h((v16u8) tmp0, coeff_vec);
1152  res_r <<= 3;
1153  res_r = (v8u16) __msa_srari_h((v8i16) res_r, 6);
1154  res_r = __msa_sat_u_h(res_r, 7);
1155  res = __msa_pckev_b((v16i8) res_r, (v16i8) res_r);
1156  out = (v8i16) __msa_aver_u_b((v16u8) res, dst_data);
1157  out0 = __msa_copy_u_h(out, 0);
1158  out1 = __msa_copy_u_h(out, 2);
1159 
1160  SH(out0, dst);
1161  dst += stride;
1162  SH(out1, dst);
1163 }
1164 
1165 static void avc_chroma_vt_and_aver_dst_2x4_msa(uint8_t *src, uint8_t *dst,
1166  int32_t stride, uint32_t coeff0,
1167  uint32_t coeff1)
1168 {
1169  uint16_t tp0, tp1, tp2, tp3;
1170  v16i8 src0, src1, src2, src3, src4;
1171  v16u8 tmp0, tmp1, tmp2, tmp3;
1172  v8u16 res_r;
1173  v8i16 res;
1174  v16i8 coeff_vec0 = __msa_fill_b(coeff0);
1175  v16i8 coeff_vec1 = __msa_fill_b(coeff1);
1176  v16u8 coeff_vec = (v16u8) __msa_ilvr_b(coeff_vec0, coeff_vec1);
1177  v16u8 dst_data = { 0 };
1178 
1179  LD_SB5(src, stride, src0, src1, src2, src3, src4);
1180 
1181  tp0 = LH(dst);
1182  tp1 = LH(dst + stride);
1183  tp2 = LH(dst + 2 * stride);
1184  tp3 = LH(dst + 3 * stride);
1185  dst_data = (v16u8) __msa_insert_h((v8i16) dst_data, 0, tp0);
1186  dst_data = (v16u8) __msa_insert_h((v8i16) dst_data, 1, tp1);
1187  dst_data = (v16u8) __msa_insert_h((v8i16) dst_data, 2, tp2);
1188  dst_data = (v16u8) __msa_insert_h((v8i16) dst_data, 3, tp3);
1189 
1190  ILVR_B4_UB(src1, src0, src2, src1, src3, src2, src4, src3,
1191  tmp0, tmp1, tmp2, tmp3);
1192  ILVR_W2_UB(tmp1, tmp0, tmp3, tmp2, tmp0, tmp2);
1193 
1194  tmp0 = (v16u8) __msa_ilvr_d((v2i64) tmp2, (v2i64) tmp0);
1195 
1196  res_r = __msa_dotp_u_h(tmp0, coeff_vec);
1197  res_r <<= 3;
1198  res_r = (v8u16) __msa_srari_h((v8i16) res_r, 6);
1199  res_r = __msa_sat_u_h(res_r, 7);
1200 
1201  res = (v8i16) __msa_pckev_b((v16i8) res_r, (v16i8) res_r);
1202  res = (v8i16) __msa_aver_u_b((v16u8) res, dst_data);
1203 
1204  ST_H4(res, 0, 1, 2, 3, dst, stride);
1205 }
1206 
1207 static void avc_chroma_vt_and_aver_dst_2w_msa(uint8_t *src, uint8_t *dst,
1208  int32_t stride, uint32_t coeff0,
1209  uint32_t coeff1, int32_t height)
1210 {
1211  if (2 == height) {
1212  avc_chroma_vt_and_aver_dst_2x2_msa(src, dst, stride, coeff0, coeff1);
1213  } else if (4 == height) {
1214  avc_chroma_vt_and_aver_dst_2x4_msa(src, dst, stride, coeff0, coeff1);
1215  }
1216 }
1217 
1218 static void avc_chroma_vt_and_aver_dst_4x2_msa(uint8_t *src, uint8_t *dst,
1219  int32_t stride, uint32_t coeff0,
1220  uint32_t coeff1)
1221 {
1222  uint32_t load0, load1;
1223  v16u8 src0, src1, src2, tmp0, tmp1;
1224  v16u8 dst_data = { 0 };
1225  v8u16 res_r;
1226  v16u8 res;
1227  v16i8 coeff_vec0 = __msa_fill_b(coeff0);
1228  v16i8 coeff_vec1 = __msa_fill_b(coeff1);
1229  v16u8 coeff_vec = (v16u8) __msa_ilvr_b(coeff_vec0, coeff_vec1);
1230 
1231  LD_UB3(src, stride, src0, src1, src2);
1232 
1233  LW2(dst, stride, load0, load1);
1234 
1235  INSERT_W2_UB(load0, load1, dst_data);
1236  ILVR_B2_UB(src1, src0, src2, src1, tmp0, tmp1);
1237 
1238  tmp0 = (v16u8) __msa_ilvr_d((v2i64) tmp1, (v2i64) tmp0);
1239 
1240  res_r = __msa_dotp_u_h(tmp0, coeff_vec);
1241  res_r <<= 3;
1242  res_r = (v8u16) __msa_srari_h((v8i16) res_r, 6);
1243  res_r = __msa_sat_u_h(res_r, 7);
1244  res = (v16u8) __msa_pckev_b((v16i8) res_r, (v16i8) res_r);
1245  res = __msa_aver_u_b(res, dst_data);
1246 
1247  ST_W2(res, 0, 1, dst, stride);
1248 }
1249 
1250 static void avc_chroma_vt_and_aver_dst_4x4_msa(uint8_t *src, uint8_t *dst,
1251  int32_t stride, uint32_t coeff0,
1252  uint32_t coeff1)
1253 {
1254  uint32_t tp0, tp1, tp2, tp3;
1255  v16u8 src0, src1, src2, src3, src4;
1256  v16u8 tmp0, tmp1, tmp2, tmp3;
1257  v16u8 dst0 = { 0 };
1258  v8u16 res0_r, res1_r;
1259  v16u8 out;
1260  v16i8 coeff_vec0 = __msa_fill_b(coeff0);
1261  v16i8 coeff_vec1 = __msa_fill_b(coeff1);
1262  v16u8 coeff_vec = (v16u8) __msa_ilvr_b(coeff_vec0, coeff_vec1);
1263 
1264  LD_UB5(src, stride, src0, src1, src2, src3, src4);
1265  LW4(dst, stride, tp0, tp1, tp2, tp3);
1266  INSERT_W4_UB(tp0, tp1, tp2, tp3, dst0);
1267  ILVR_B4_UB(src1, src0, src2, src1, src3, src2, src4, src3, tmp0, tmp1, tmp2,
1268  tmp3);
1269  ILVR_D2_UB(tmp1, tmp0, tmp3, tmp2, tmp0, tmp2);
1270  DOTP_UB2_UH(tmp0, tmp2, coeff_vec, coeff_vec, res0_r, res1_r);
1271  res0_r <<= 3;
1272  res1_r <<= 3;
1273  SRARI_H2_UH(res0_r, res1_r, 6);
1274  SAT_UH2_UH(res0_r, res1_r, 7);
1275  out = (v16u8) __msa_pckev_b((v16i8) res1_r, (v16i8) res0_r);
1276  out = __msa_aver_u_b(out, dst0);
1277  ST_W4(out, 0, 1, 2, 3, dst, stride);
1278 }
1279 
1280 static void avc_chroma_vt_and_aver_dst_4x8_msa(uint8_t *src, uint8_t *dst,
1281  int32_t stride, uint32_t coeff0,
1282  uint32_t coeff1)
1283 {
1284  uint32_t tp0, tp1, tp2, tp3;
1285  v16u8 src0, src1, src2, src3, src4, src5, src6, src7, src8;
1286  v16u8 tmp0, tmp1, tmp2, tmp3, tmp4, tmp5, tmp6, tmp7, out0, out1;
1287  v16u8 dst0 = { 0 }, dst1 = { 0 };
1288  v8u16 res0, res1, res2, res3;
1289  v16i8 coeff_vec0 = __msa_fill_b(coeff0);
1290  v16i8 coeff_vec1 = __msa_fill_b(coeff1);
1291  v16u8 coeff_vec = (v16u8) __msa_ilvr_b(coeff_vec0, coeff_vec1);
1292 
1293  LD_UB5(src, stride, src0, src1, src2, src3, src4);
1294  src += (5 * stride);
1295  LD_UB4(src, stride, src5, src6, src7, src8);
1296  LW4(dst, stride, tp0, tp1, tp2, tp3);
1297  INSERT_W4_UB(tp0, tp1, tp2, tp3, dst0);
1298  LW4(dst + 4 * stride, stride, tp0, tp1, tp2, tp3);
1299  INSERT_W4_UB(tp0, tp1, tp2, tp3, dst1);
1300  ILVR_B4_UB(src1, src0, src2, src1, src3, src2, src4, src3, tmp0, tmp1, tmp2,
1301  tmp3);
1302  ILVR_B4_UB(src5, src4, src6, src5, src7, src6, src8, src7, tmp4, tmp5, tmp6,
1303  tmp7);
1304  ILVR_D2_UB(tmp1, tmp0, tmp3, tmp2, tmp0, tmp2);
1305  ILVR_D2_UB(tmp5, tmp4, tmp7, tmp6, tmp4, tmp6);
1306  DOTP_UB2_UH(tmp0, tmp2, coeff_vec, coeff_vec, res0, res1);
1307  DOTP_UB2_UH(tmp4, tmp6, coeff_vec, coeff_vec, res2, res3);
1308  SLLI_4V(res0, res1, res2, res3, 3);
1309  SRARI_H4_UH(res0, res1, res2, res3, 6);
1310  SAT_UH4_UH(res0, res1, res2, res3, 7);
1311  PCKEV_B2_UB(res1, res0, res3, res2, out0, out1);
1312  AVER_UB2_UB(out0, dst0, out1, dst1, out0, out1);
1313  ST_W8(out0, out1, 0, 1, 2, 3, 0, 1, 2, 3, dst, stride);
1314 }
1315 
1316 static void avc_chroma_vt_and_aver_dst_4w_msa(uint8_t *src, uint8_t *dst,
1317  int32_t stride, uint32_t coeff0,
1318  uint32_t coeff1, int32_t height)
1319 {
1320  if (2 == height) {
1321  avc_chroma_vt_and_aver_dst_4x2_msa(src, dst, stride, coeff0, coeff1);
1322  } else if (4 == height) {
1323  avc_chroma_vt_and_aver_dst_4x4_msa(src, dst, stride, coeff0, coeff1);
1324  } else if (8 == height) {
1325  avc_chroma_vt_and_aver_dst_4x8_msa(src, dst, stride, coeff0, coeff1);
1326  }
1327 }
1328 
1329 static void avc_chroma_vt_and_aver_dst_8x4_msa(uint8_t *src, uint8_t *dst,
1330  int32_t stride, uint32_t coeff0,
1331  uint32_t coeff1)
1332 {
1333  uint64_t tp0, tp1, tp2, tp3;
1334  v16u8 src0, src1, src2, src3, src4;
1335  v16u8 out0, out1;
1336  v8u16 res0, res1, res2, res3;
1337  v16u8 dst0 = { 0 }, dst1 = { 0 };
1338  v16i8 coeff_vec0 = __msa_fill_b(coeff0);
1339  v16i8 coeff_vec1 = __msa_fill_b(coeff1);
1340  v16u8 coeff_vec = (v16u8) __msa_ilvr_b(coeff_vec0, coeff_vec1);
1341 
1342  LD_UB5(src, stride, src0, src1, src2, src3, src4);
1343  LD4(dst, stride, tp0, tp1, tp2, tp3);
1344  INSERT_D2_UB(tp0, tp1, dst0);
1345  INSERT_D2_UB(tp2, tp3, dst1);
1346  ILVR_B4_UB(src1, src0, src2, src1, src3, src2, src4, src3,
1347  src0, src1, src2, src3);
1348  DOTP_UB4_UH(src0, src1, src2, src3, coeff_vec, coeff_vec, coeff_vec,
1349  coeff_vec, res0, res1, res2, res3);
1350  SLLI_4V(res0, res1, res2, res3, 3);
1351  SRARI_H4_UH(res0, res1, res2, res3, 6);
1352  SAT_UH4_UH(res0, res1, res2, res3, 7);
1353  PCKEV_B2_UB(res1, res0, res3, res2, out0, out1);
1354  AVER_UB2_UB(out0, dst0, out1, dst1, out0, out1);
1355  ST_D4(out0, out1, 0, 1, 0, 1, dst, stride);
1356 }
1357 
1358 static void avc_chroma_vt_and_aver_dst_8x8_msa(uint8_t *src, uint8_t *dst,
1359  int32_t stride, uint32_t coeff0,
1360  uint32_t coeff1)
1361 {
1362  uint64_t tp0, tp1, tp2, tp3;
1363  v16u8 src0, src1, src2, src3, src4, src5, src6, src7, src8;
1364  v16u8 out0, out1, out2, out3;
1365  v16u8 dst0 = { 0 }, dst1 = { 0 }, dst2 = { 0 }, dst3 = { 0 };
1366  v8u16 res0, res1, res2, res3, res4, res5, res6, res7;
1367  v16i8 coeff_vec0 = __msa_fill_b(coeff0);
1368  v16i8 coeff_vec1 = __msa_fill_b(coeff1);
1369  v16u8 coeff_vec = (v16u8) __msa_ilvr_b(coeff_vec0, coeff_vec1);
1370 
1371  LD_UB5(src, stride, src0, src1, src2, src3, src4);
1372  src += (5 * stride);
1373  LD_UB4(src, stride, src5, src6, src7, src8);
1374  LD4(dst, stride, tp0, tp1, tp2, tp3);
1375  INSERT_D2_UB(tp0, tp1, dst0);
1376  INSERT_D2_UB(tp2, tp3, dst1);
1377  LD4(dst + 4 * stride, stride, tp0, tp1, tp2, tp3);
1378  INSERT_D2_UB(tp0, tp1, dst2);
1379  INSERT_D2_UB(tp2, tp3, dst3);
1380  ILVR_B4_UB(src1, src0, src2, src1, src3, src2, src4, src3,
1381  src0, src1, src2, src3);
1382  ILVR_B4_UB(src5, src4, src6, src5, src7, src6, src8, src7,
1383  src4, src5, src6, src7);
1384  DOTP_UB4_UH(src0, src1, src2, src3, coeff_vec, coeff_vec, coeff_vec,
1385  coeff_vec, res0, res1, res2, res3);
1386  DOTP_UB4_UH(src4, src5, src6, src7, coeff_vec, coeff_vec, coeff_vec,
1387  coeff_vec, res4, res5, res6, res7);
1388  SLLI_4V(res0, res1, res2, res3, 3);
1389  SLLI_4V(res4, res5, res6, res7, 3);
1390  SRARI_H4_UH(res0, res1, res2, res3, 6);
1391  SRARI_H4_UH(res4, res5, res6, res7, 6);
1392  SAT_UH4_UH(res0, res1, res2, res3, 7);
1393  SAT_UH4_UH(res0, res1, res2, res3, 7);
1394  PCKEV_B2_UB(res1, res0, res3, res2, out0, out1);
1395  PCKEV_B2_UB(res5, res4, res7, res6, out2, out3);
1396  AVER_UB2_UB(out0, dst0, out1, dst1, out0, out1);
1397  AVER_UB2_UB(out2, dst2, out3, dst3, out2, out3);
1398  ST_D8(out0, out1, out2, out3, 0, 1, 0, 1, 0, 1, 0, 1, dst, stride);
1399 }
1400 
1401 static void avc_chroma_vt_and_aver_dst_8w_msa(uint8_t *src, uint8_t *dst,
1402  int32_t stride, uint32_t coeff0,
1403  uint32_t coeff1, int32_t height)
1404 {
1405  if (4 == height) {
1406  avc_chroma_vt_and_aver_dst_8x4_msa(src, dst, stride, coeff0, coeff1);
1407  } else if (8 == height) {
1408  avc_chroma_vt_and_aver_dst_8x8_msa(src, dst, stride, coeff0, coeff1);
1409  }
1410 }
1411 
1412 static void avc_chroma_hv_and_aver_dst_2x2_msa(uint8_t *src, uint8_t *dst,
1413  int32_t stride,
1414  uint32_t coef_hor0,
1415  uint32_t coef_hor1,
1416  uint32_t coef_ver0,
1417  uint32_t coef_ver1)
1418 {
1419  uint16_t out0, out1;
1420  v16u8 dst0 = { 0 };
1421  v16u8 src0, src1, src2;
1422  v8u16 res_hz0, res_hz1, res_vt0, res_vt1;
1423  v16i8 res, mask;
1424  v16i8 coeff_hz_vec0 = __msa_fill_b(coef_hor0);
1425  v16i8 coeff_hz_vec1 = __msa_fill_b(coef_hor1);
1426  v16u8 coeff_hz_vec = (v16u8) __msa_ilvr_b(coeff_hz_vec0, coeff_hz_vec1);
1427  v8u16 coeff_vt_vec0 = (v8u16) __msa_fill_h(coef_ver0);
1428  v8u16 coeff_vt_vec1 = (v8u16) __msa_fill_h(coef_ver1);
1429 
1430  mask = LD_SB(&chroma_mask_arr[48]);
1431 
1432  LD_UB3(src, stride, src0, src1, src2);
1433  out0 = LH(dst);
1434  out1 = LH(dst + stride);
1435  dst0 = (v16u8) __msa_insert_h((v8i16) dst0, 0, out0);
1436  dst0 = (v16u8) __msa_insert_h((v8i16) dst0, 1, out1);
1437  VSHF_B2_UB(src0, src1, src1, src2, mask, mask, src0, src1);
1438  DOTP_UB2_UH(src0, src1, coeff_hz_vec, coeff_hz_vec, res_hz0, res_hz1);
1439  MUL2(res_hz0, coeff_vt_vec1, res_hz1, coeff_vt_vec0, res_vt0, res_vt1);
1440 
1441  res_vt0 += res_vt1;
1442  res_vt0 = (v8u16) __msa_srari_h((v8i16) res_vt0, 6);
1443  res_vt0 = __msa_sat_u_h(res_vt0, 7);
1444  res = __msa_pckev_b((v16i8) res_vt0, (v16i8) res_vt0);
1445  dst0 = __msa_aver_u_b((v16u8) res, dst0);
1446  out0 = __msa_copy_u_h((v8i16) dst0, 0);
1447  out1 = __msa_copy_u_h((v8i16) dst0, 1);
1448 
1449  SH(out0, dst);
1450  dst += stride;
1451  SH(out1, dst);
1452 }
1453 
1454 static void avc_chroma_hv_and_aver_dst_2x4_msa(uint8_t *src, uint8_t *dst,
1455  int32_t stride,
1456  uint32_t coef_hor0,
1457  uint32_t coef_hor1,
1458  uint32_t coef_ver0,
1459  uint32_t coef_ver1)
1460 {
1461  uint16_t tp0, tp1, tp2, tp3;
1462  v16u8 src0, src1, src2, src3, src4;
1463  v16u8 tmp0, tmp1, tmp2, tmp3;
1464  v16u8 dst0 = { 0 };
1465  v8u16 res_hz0, res_hz1, res_vt0, res_vt1;
1466  v16i8 res, mask;
1467  v16i8 coeff_hz_vec0 = __msa_fill_b(coef_hor0);
1468  v16i8 coeff_hz_vec1 = __msa_fill_b(coef_hor1);
1469  v16u8 coeff_hz_vec = (v16u8) __msa_ilvr_b(coeff_hz_vec0, coeff_hz_vec1);
1470  v8u16 coeff_vt_vec0 = (v8u16) __msa_fill_h(coef_ver0);
1471  v8u16 coeff_vt_vec1 = (v8u16) __msa_fill_h(coef_ver1);
1472 
1473  mask = LD_SB(&chroma_mask_arr[48]);
1474 
1475  LD_UB5(src, stride, src0, src1, src2, src3, src4);
1476  tp0 = LH(dst);
1477  tp1 = LH(dst + stride);
1478  tp2 = LH(dst + 2 * stride);
1479  tp3 = LH(dst + 3 * stride);
1480  dst0 = (v16u8) __msa_insert_h((v8i16) dst0, 0, tp0);
1481  dst0 = (v16u8) __msa_insert_h((v8i16) dst0, 1, tp1);
1482  dst0 = (v16u8) __msa_insert_h((v8i16) dst0, 2, tp2);
1483  dst0 = (v16u8) __msa_insert_h((v8i16) dst0, 3, tp3);
1484  VSHF_B2_UB(src0, src1, src2, src3, mask, mask, tmp0, tmp1);
1485  VSHF_B2_UB(src1, src2, src3, src4, mask, mask, tmp2, tmp3);
1486  ILVR_D2_UB(tmp1, tmp0, tmp3, tmp2, src0, src1);
1487  DOTP_UB2_UH(src0, src1, coeff_hz_vec, coeff_hz_vec, res_hz0, res_hz1);
1488  MUL2(res_hz0, coeff_vt_vec1, res_hz1, coeff_vt_vec0, res_vt0, res_vt1);
1489 
1490  res_vt0 += res_vt1;
1491  res_vt0 = (v8u16) __msa_srari_h((v8i16) res_vt0, 6);
1492  res_vt0 = __msa_sat_u_h(res_vt0, 7);
1493  res = __msa_pckev_b((v16i8) res_vt0, (v16i8) res_vt0);
1494  dst0 = __msa_aver_u_b((v16u8) res, dst0);
1495 
1496  ST_H4(dst0, 0, 1, 2, 3, dst, stride);
1497 }
1498 
1499 static void avc_chroma_hv_and_aver_dst_2w_msa(uint8_t *src, uint8_t *dst,
1500  int32_t stride,
1501  uint32_t coef_hor0,
1502  uint32_t coef_hor1,
1503  uint32_t coef_ver0,
1504  uint32_t coef_ver1,
1505  int32_t height)
1506 {
1507  if (2 == height) {
1509  coef_hor1, coef_ver0, coef_ver1);
1510  } else if (4 == height) {
1512  coef_hor1, coef_ver0, coef_ver1);
1513  }
1514 }
1515 
1516 static void avc_chroma_hv_and_aver_dst_4x2_msa(uint8_t *src, uint8_t *dst,
1517  int32_t stride,
1518  uint32_t coef_hor0,
1519  uint32_t coef_hor1,
1520  uint32_t coef_ver0,
1521  uint32_t coef_ver1)
1522 {
1523  uint32_t tp0, tp1;
1524  v16u8 src0, src1, src2;
1525  v16u8 dst0, dst_data = { 0 };
1526  v8u16 res_hz0, res_hz1, res_vt0, res_vt1;
1527  v16i8 mask;
1528  v16i8 coeff_hz_vec0 = __msa_fill_b(coef_hor0);
1529  v16i8 coeff_hz_vec1 = __msa_fill_b(coef_hor1);
1530  v16u8 coeff_hz_vec = (v16u8) __msa_ilvr_b(coeff_hz_vec0, coeff_hz_vec1);
1531  v8u16 coeff_vt_vec0 = (v8u16) __msa_fill_h(coef_ver0);
1532  v8u16 coeff_vt_vec1 = (v8u16) __msa_fill_h(coef_ver1);
1533 
1534  mask = LD_SB(&chroma_mask_arr[0]);
1535 
1536  LD_UB3(src, stride, src0, src1, src2);
1537  LW2(dst, stride, tp0, tp1);
1538  INSERT_W2_UB(tp0, tp1, dst_data);
1539  VSHF_B2_UB(src0, src1, src1, src2, mask, mask, src0, src1);
1540  DOTP_UB2_UH(src0, src1, coeff_hz_vec, coeff_hz_vec, res_hz0, res_hz1);
1541  MUL2(res_hz0, coeff_vt_vec1, res_hz1, coeff_vt_vec0, res_vt0, res_vt1);
1542 
1543  res_vt0 += res_vt1;
1544  res_vt0 = (v8u16) __msa_srari_h((v8i16) res_vt0, 6);
1545  res_vt0 = __msa_sat_u_h(res_vt0, 7);
1546  dst0 = (v16u8) __msa_pckev_b((v16i8) res_vt0, (v16i8) res_vt0);
1547  dst0 = __msa_aver_u_b(dst0, dst_data);
1548 
1549  ST_W2(dst0, 0, 1, dst, stride);
1550 }
1551 
1552 static void avc_chroma_hv_and_aver_dst_4x4_msa(uint8_t *src, uint8_t *dst,
1553  int32_t stride,
1554  uint32_t coef_hor0,
1555  uint32_t coef_hor1,
1556  uint32_t coef_ver0,
1557  uint32_t coef_ver1)
1558 {
1559  uint32_t tp0, tp1, tp2, tp3;
1560  v16u8 src0, src1, src2, src3, src4;
1561  v16u8 out, dst_data = { 0 };
1562  v8u16 res_hz0, res_hz1, res_hz2, res_hz3;
1563  v8u16 res_vt0, res_vt1, res_vt2, res_vt3;
1564  v16i8 mask;
1565  v16i8 coeff_hz_vec0 = __msa_fill_b(coef_hor0);
1566  v16i8 coeff_hz_vec1 = __msa_fill_b(coef_hor1);
1567  v16u8 coeff_hz_vec = (v16u8) __msa_ilvr_b(coeff_hz_vec0, coeff_hz_vec1);
1568  v8u16 coeff_vt_vec0 = (v8u16) __msa_fill_h(coef_ver0);
1569  v8u16 coeff_vt_vec1 = (v8u16) __msa_fill_h(coef_ver1);
1570 
1571  mask = LD_SB(&chroma_mask_arr[0]);
1572 
1573  LD_UB5(src, stride, src0, src1, src2, src3, src4);
1574  LW4(dst, stride, tp0, tp1, tp2, tp3);
1575  INSERT_W4_UB(tp0, tp1, tp2, tp3, dst_data);
1576  VSHF_B2_UB(src0, src1, src1, src2, mask, mask, src0, src1);
1577  VSHF_B2_UB(src2, src3, src3, src4, mask, mask, src2, src3);
1578  DOTP_UB4_UH(src0, src1, src2, src3, coeff_hz_vec, coeff_hz_vec,
1579  coeff_hz_vec, coeff_hz_vec, res_hz0, res_hz1, res_hz2,
1580  res_hz3);
1581  MUL4(res_hz0, coeff_vt_vec1, res_hz1, coeff_vt_vec0, res_hz2, coeff_vt_vec1,
1582  res_hz3, coeff_vt_vec0, res_vt0, res_vt1, res_vt2, res_vt3);
1583  ADD2(res_vt0, res_vt1, res_vt2, res_vt3, res_vt0, res_vt1);
1584  SRARI_H2_UH(res_vt0, res_vt1, 6);
1585  SAT_UH2_UH(res_vt0, res_vt1, 7);
1586  out = (v16u8) __msa_pckev_b((v16i8) res_vt1, (v16i8) res_vt0);
1587  out = __msa_aver_u_b(out, dst_data);
1588  ST_W4(out, 0, 1, 2, 3, dst, stride);
1589 }
1590 
1591 static void avc_chroma_hv_and_aver_dst_4x8_msa(uint8_t *src, uint8_t *dst,
1592  int32_t stride,
1593  uint32_t coef_hor0,
1594  uint32_t coef_hor1,
1595  uint32_t coef_ver0,
1596  uint32_t coef_ver1)
1597 {
1598  uint32_t tp0, tp1, tp2, tp3;
1599  v16u8 src0, src1, src2, src3, src4, src5, src6, src7, src8, res0, res1;
1600  v16u8 dst0 = { 0 }, dst1 = { 0 };
1601  v8u16 res_hz0, res_hz1, res_hz2, res_hz3, res_hz4, res_hz5, res_hz6, res_hz7;
1602  v8u16 res_vt0, res_vt1, res_vt2, res_vt3, res_vt4, res_vt5, res_vt6, res_vt7;
1603  v16i8 mask;
1604  v16i8 coeff_hz_vec0 = __msa_fill_b(coef_hor0);
1605  v16i8 coeff_hz_vec1 = __msa_fill_b(coef_hor1);
1606  v16u8 coeff_hz_vec = (v16u8) __msa_ilvr_b(coeff_hz_vec0, coeff_hz_vec1);
1607  v8u16 coeff_vt_vec0 = (v8u16) __msa_fill_h(coef_ver0);
1608  v8u16 coeff_vt_vec1 = (v8u16) __msa_fill_h(coef_ver1);
1609 
1610  mask = LD_SB(&chroma_mask_arr[0]);
1611 
1612  LD_UB5(src, stride, src0, src1, src2, src3, src4);
1613  src += (5 * stride);
1614  LD_UB4(src, stride, src5, src6, src7, src8);
1615  LW4(dst, stride, tp0, tp1, tp2, tp3);
1616  INSERT_W4_UB(tp0, tp1, tp2, tp3, dst0);
1617  LW4(dst + 4 * stride, stride, tp0, tp1, tp2, tp3);
1618  INSERT_W4_UB(tp0, tp1, tp2, tp3, dst1);
1619  VSHF_B2_UB(src0, src1, src1, src2, mask, mask, src0, src1);
1620  VSHF_B2_UB(src2, src3, src3, src4, mask, mask, src2, src3);
1621  VSHF_B2_UB(src4, src5, src5, src6, mask, mask, src4, src5);
1622  VSHF_B2_UB(src6, src7, src7, src8, mask, mask, src6, src7);
1623  DOTP_UB4_UH(src0, src1, src2, src3, coeff_hz_vec, coeff_hz_vec,
1624  coeff_hz_vec, coeff_hz_vec, res_hz0, res_hz1, res_hz2, res_hz3);
1625  DOTP_UB4_UH(src4, src5, src6, src7, coeff_hz_vec, coeff_hz_vec,
1626  coeff_hz_vec, coeff_hz_vec, res_hz4, res_hz5, res_hz6, res_hz7);
1627  MUL4(res_hz0, coeff_vt_vec1, res_hz1, coeff_vt_vec0, res_hz2, coeff_vt_vec1,
1628  res_hz3, coeff_vt_vec0, res_vt0, res_vt1, res_vt2, res_vt3);
1629  MUL4(res_hz4, coeff_vt_vec1, res_hz5, coeff_vt_vec0, res_hz6, coeff_vt_vec1,
1630  res_hz7, coeff_vt_vec0, res_vt4, res_vt5, res_vt6, res_vt7);
1631  ADD2(res_vt0, res_vt1, res_vt2, res_vt3, res_vt0, res_vt1);
1632  ADD2(res_vt4, res_vt5, res_vt6, res_vt7, res_vt2, res_vt3);
1633  SRARI_H4_UH(res_vt0, res_vt1, res_vt2, res_vt3, 6);
1634  SAT_UH4_UH(res_vt0, res_vt1, res_vt2, res_vt3, 7);
1635  PCKEV_B2_UB(res_vt1, res_vt0, res_vt3, res_vt2, res0, res1);
1636  AVER_UB2_UB(res0, dst0, res1, dst1, res0, res1);
1637  ST_W8(res0, res1, 0, 1, 2, 3, 0, 1, 2, 3, dst, stride);
1638 }
1639 
1640 static void avc_chroma_hv_and_aver_dst_4w_msa(uint8_t *src, uint8_t *dst,
1641  int32_t stride,
1642  uint32_t coef_hor0,
1643  uint32_t coef_hor1,
1644  uint32_t coef_ver0,
1645  uint32_t coef_ver1,
1646  int32_t height)
1647 {
1648  if (2 == height) {
1650  coef_hor1, coef_ver0, coef_ver1);
1651  } else if (4 == height) {
1653  coef_hor1, coef_ver0, coef_ver1);
1654  } else if (8 == height) {
1656  coef_hor1, coef_ver0, coef_ver1);
1657  }
1658 }
1659 
1660 static void avc_chroma_hv_and_aver_dst_8x4_msa(uint8_t *src, uint8_t *dst,
1661  int32_t stride,
1662  uint32_t coef_hor0,
1663  uint32_t coef_hor1,
1664  uint32_t coef_ver0,
1665  uint32_t coef_ver1)
1666 {
1667  uint64_t tp0, tp1, tp2, tp3;
1668  v16u8 src0, src1, src2, src3, src4, out0, out1;
1669  v8u16 res_hz0, res_hz1, res_hz2;
1670  v8u16 res_hz3, res_hz4;
1671  v8u16 res_vt0, res_vt1, res_vt2, res_vt3;
1672  v16u8 dst0 = { 0 }, dst1 = { 0 };
1673  v16i8 mask;
1674  v16i8 coeff_hz_vec0 = __msa_fill_b(coef_hor0);
1675  v16i8 coeff_hz_vec1 = __msa_fill_b(coef_hor1);
1676  v16u8 coeff_hz_vec = (v16u8) __msa_ilvr_b(coeff_hz_vec0, coeff_hz_vec1);
1677  v8u16 coeff_vt_vec0 = (v8u16) __msa_fill_h(coef_ver0);
1678  v8u16 coeff_vt_vec1 = (v8u16) __msa_fill_h(coef_ver1);
1679 
1680  mask = LD_SB(&chroma_mask_arr[32]);
1681 
1682  src0 = LD_UB(src);
1683  src += stride;
1684  src0 = (v16u8) __msa_vshf_b(mask, (v16i8) src0, (v16i8) src0);
1685  res_hz0 = __msa_dotp_u_h(src0, coeff_hz_vec);
1686  LD_UB4(src, stride, src1, src2, src3, src4);
1687  src += (4 * stride);
1688  LD4(dst, stride, tp0, tp1, tp2, tp3);
1689  INSERT_D2_UB(tp0, tp1, dst0);
1690  INSERT_D2_UB(tp2, tp3, dst1);
1691  VSHF_B2_UB(src1, src1, src2, src2, mask, mask, src1, src2);
1692  VSHF_B2_UB(src3, src3, src4, src4, mask, mask, src3, src4);
1693  DOTP_UB4_UH(src1, src2, src3, src4, coeff_hz_vec, coeff_hz_vec,
1694  coeff_hz_vec, coeff_hz_vec, res_hz1, res_hz2, res_hz3, res_hz4);
1695  MUL4(res_hz1, coeff_vt_vec0, res_hz2, coeff_vt_vec0, res_hz3, coeff_vt_vec0,
1696  res_hz4, coeff_vt_vec0, res_vt0, res_vt1, res_vt2, res_vt3);
1697  res_vt0 += (res_hz0 * coeff_vt_vec1);
1698  res_vt1 += (res_hz1 * coeff_vt_vec1);
1699  res_vt2 += (res_hz2 * coeff_vt_vec1);
1700  res_vt3 += (res_hz3 * coeff_vt_vec1);
1701  SRARI_H4_UH(res_vt0, res_vt1, res_vt2, res_vt3, 6);
1702  SAT_UH4_UH(res_vt0, res_vt1, res_vt2, res_vt3, 7);
1703  PCKEV_B2_UB(res_vt1, res_vt0, res_vt3, res_vt2, out0, out1);
1704  AVER_UB2_UB(out0, dst0, out1, dst1, out0, out1);
1705  ST_D4(out0, out1, 0, 1, 0, 1, dst, stride);
1706 }
1707 
1708 static void avc_chroma_hv_and_aver_dst_8x8_msa(uint8_t *src, uint8_t *dst,
1709  int32_t stride,
1710  uint32_t coef_hor0,
1711  uint32_t coef_hor1,
1712  uint32_t coef_ver0,
1713  uint32_t coef_ver1)
1714 {
1715  uint64_t tp0, tp1, tp2, tp3;
1716  v16u8 src0, src1, src2, src3, src4, src5, src6, src7, src8;
1717  v16u8 out0, out1, out2, out3;
1718  v16u8 dst0 = { 0 }, dst1 = { 0 }, dst2 = { 0 }, dst3 = { 0 };
1719  v8u16 res_hz0, res_hz1, res_hz2, res_hz3, res_hz4;
1720  v8u16 res_hz5, res_hz6, res_hz7, res_hz8;
1721  v8u16 res_vt0, res_vt1, res_vt2, res_vt3;
1722  v8u16 res_vt4, res_vt5, res_vt6, res_vt7;
1723  v16i8 mask;
1724  v16i8 coeff_hz_vec0 = __msa_fill_b(coef_hor0);
1725  v16i8 coeff_hz_vec1 = __msa_fill_b(coef_hor1);
1726  v16u8 coeff_hz_vec = (v16u8) __msa_ilvr_b(coeff_hz_vec0, coeff_hz_vec1);
1727  v8u16 coeff_vt_vec0 = (v8u16) __msa_fill_h(coef_ver0);
1728  v8u16 coeff_vt_vec1 = (v8u16) __msa_fill_h(coef_ver1);
1729 
1730  mask = LD_SB(&chroma_mask_arr[32]);
1731 
1732  LD_UB5(src, stride, src0, src1, src2, src3, src4);
1733  src += (5 * stride);
1734  LD_UB4(src, stride, src5, src6, src7, src8);
1735  src0 = (v16u8) __msa_vshf_b(mask, (v16i8) src0, (v16i8) src0);
1736  VSHF_B2_UB(src1, src1, src2, src2, mask, mask, src1, src2);
1737  VSHF_B2_UB(src3, src3, src4, src4, mask, mask, src3, src4);
1738  VSHF_B2_UB(src5, src5, src6, src6, mask, mask, src5, src6);
1739  VSHF_B2_UB(src7, src7, src8, src8, mask, mask, src7, src8);
1740  res_hz0 = __msa_dotp_u_h(src0, coeff_hz_vec);
1741  DOTP_UB4_UH(src1, src2, src3, src4, coeff_hz_vec, coeff_hz_vec,
1742  coeff_hz_vec, coeff_hz_vec, res_hz1, res_hz2, res_hz3,
1743  res_hz4);
1744  DOTP_UB4_UH(src5, src6, src7, src8, coeff_hz_vec, coeff_hz_vec,
1745  coeff_hz_vec, coeff_hz_vec, res_hz5, res_hz6, res_hz7, res_hz8);
1746  MUL4(res_hz1, coeff_vt_vec0, res_hz2, coeff_vt_vec0, res_hz3,
1747  coeff_vt_vec0, res_hz4, coeff_vt_vec0, res_vt0, res_vt1, res_vt2,
1748  res_vt3);
1749  MUL4(res_hz5, coeff_vt_vec0, res_hz6, coeff_vt_vec0, res_hz7,
1750  coeff_vt_vec0, res_hz8, coeff_vt_vec0, res_vt4, res_vt5, res_vt6,
1751  res_vt7);
1752  LD4(dst, stride, tp0, tp1, tp2, tp3);
1753  INSERT_D2_UB(tp0, tp1, dst0);
1754  INSERT_D2_UB(tp2, tp3, dst1);
1755  LD4(dst + 4 * stride, stride, tp0, tp1, tp2, tp3);
1756  INSERT_D2_UB(tp0, tp1, dst2);
1757  INSERT_D2_UB(tp2, tp3, dst3);
1758  res_vt0 += (res_hz0 * coeff_vt_vec1);
1759  res_vt1 += (res_hz1 * coeff_vt_vec1);
1760  res_vt2 += (res_hz2 * coeff_vt_vec1);
1761  res_vt3 += (res_hz3 * coeff_vt_vec1);
1762  res_vt4 += (res_hz4 * coeff_vt_vec1);
1763  res_vt5 += (res_hz5 * coeff_vt_vec1);
1764  res_vt6 += (res_hz6 * coeff_vt_vec1);
1765  res_vt7 += (res_hz7 * coeff_vt_vec1);
1766  SRARI_H4_UH(res_vt0, res_vt1, res_vt2, res_vt3, 6);
1767  SRARI_H4_UH(res_vt4, res_vt5, res_vt6, res_vt7, 6);
1768  SAT_UH4_UH(res_vt0, res_vt1, res_vt2, res_vt3, 7);
1769  SAT_UH4_UH(res_vt4, res_vt5, res_vt6, res_vt7, 7);
1770  PCKEV_B2_UB(res_vt1, res_vt0, res_vt3, res_vt2, out0, out1);
1771  PCKEV_B2_UB(res_vt5, res_vt4, res_vt7, res_vt6, out2, out3);
1772  AVER_UB2_UB(out0, dst0, out1, dst1, out0, out1);
1773  AVER_UB2_UB(out2, dst2, out3, dst3, out2, out3);
1774  ST_D8(out0, out1, out2, out3, 0, 1, 0, 1, 0, 1, 0, 1, dst, stride);
1775 }
1776 
1777 static void avc_chroma_hv_and_aver_dst_8w_msa(uint8_t *src, uint8_t *dst,
1778  int32_t stride,
1779  uint32_t coef_hor0,
1780  uint32_t coef_hor1,
1781  uint32_t coef_ver0,
1782  uint32_t coef_ver1,
1783  int32_t height)
1784 {
1785  if (4 == height) {
1787  coef_hor1, coef_ver0, coef_ver1);
1788  } else if (8 == height) {
1790  coef_hor1, coef_ver0, coef_ver1);
1791  }
1792 }
1793 
1794 static void copy_width4_msa(uint8_t *src, uint8_t *dst, int32_t stride,
1795  int32_t height)
1796 {
1797  uint32_t tp0, tp1, tp2, tp3, tp4, tp5, tp6, tp7;
1798 
1799  if (8 == height) {
1800  LW4(src, stride, tp0, tp1, tp2, tp3);
1801  src += 4 * stride;
1802  LW4(src, stride, tp4, tp5, tp6, tp7);
1803  SW4(tp0, tp1, tp2, tp3, dst, stride);
1804  dst += 4 * stride;
1805  SW4(tp4, tp5, tp6, tp7, dst, stride);
1806  } else if (4 == height) {
1807  LW4(src, stride, tp0, tp1, tp2, tp3);
1808  SW4(tp0, tp1, tp2, tp3, dst, stride);
1809  } else if (2 == height) {
1810  LW2(src, stride, tp0, tp1);
1811  SW(tp0, dst);
1812  dst += stride;
1813  SW(tp1, dst);
1814  }
1815 }
1816 
1817 static void copy_width8_msa(uint8_t *src, uint8_t *dst, int32_t stride,
1818  int32_t height)
1819 {
1820  uint64_t src0, src1, src2, src3, src4, src5, src6, src7;
1821 
1822  if (8 == height) {
1823  LD4(src, stride, src0, src1, src2, src3);
1824  src += 4 * stride;
1825  LD4(src, stride, src4, src5, src6, src7);
1826  SD4(src0, src1, src2, src3, dst, stride);
1827  dst += 4 * stride;
1828  SD4(src4, src5, src6, src7, dst, stride);
1829  } else if (4 == height) {
1830  LD4(src, stride, src0, src1, src2, src3);
1831  SD4(src0, src1, src2, src3, dst, stride);
1832  }
1833 }
1834 
1835 static void avg_width4_msa(uint8_t *src, uint8_t *dst, int32_t stride,
1836  int32_t height)
1837 {
1838  uint32_t tp0, tp1, tp2, tp3;
1839  v16u8 src0 = { 0 }, src1 = { 0 }, dst0 = { 0 }, dst1 = { 0 };
1840 
1841  if (8 == height) {
1842  LW4(src, stride, tp0, tp1, tp2, tp3);
1843  src += 4 * stride;
1844  INSERT_W4_UB(tp0, tp1, tp2, tp3, src0);
1845  LW4(src, stride, tp0, tp1, tp2, tp3);
1846  INSERT_W4_UB(tp0, tp1, tp2, tp3, src1);
1847  LW4(dst, stride, tp0, tp1, tp2, tp3);
1848  INSERT_W4_UB(tp0, tp1, tp2, tp3, dst0);
1849  LW4(dst + 4 * stride, stride, tp0, tp1, tp2, tp3);
1850  INSERT_W4_UB(tp0, tp1, tp2, tp3, dst1);
1851  AVER_UB2_UB(src0, dst0, src1, dst1, dst0, dst1);
1852  ST_W8(dst0, dst1, 0, 1, 2, 3, 0, 1, 2, 3, dst, stride);
1853  } else if (4 == height) {
1854  LW4(src, stride, tp0, tp1, tp2, tp3);
1855  INSERT_W4_UB(tp0, tp1, tp2, tp3, src0);
1856  LW4(dst, stride, tp0, tp1, tp2, tp3);
1857  INSERT_W4_UB(tp0, tp1, tp2, tp3, dst0);
1858  dst0 = __msa_aver_u_b(src0, dst0);
1859  ST_W4(dst0, 0, 1, 2, 3, dst, stride);
1860  } else if (2 == height) {
1861  LW2(src, stride, tp0, tp1);
1862  INSERT_W2_UB(tp0, tp1, src0);
1863  LW2(dst, stride, tp0, tp1);
1864  INSERT_W2_UB(tp0, tp1, dst0);
1865  dst0 = __msa_aver_u_b(src0, dst0);
1866  ST_W2(dst0, 0, 1, dst, stride);
1867  }
1868 }
1869 
1870 static void avg_width8_msa(uint8_t *src, uint8_t *dst, int32_t stride,
1871  int32_t height)
1872 {
1873  uint64_t tp0, tp1, tp2, tp3, tp4, tp5, tp6, tp7;
1874  v16u8 src0 = { 0 }, src1 = { 0 }, src2 = { 0 }, src3 = { 0 };
1875  v16u8 dst0 = { 0 }, dst1 = { 0 }, dst2 = { 0 }, dst3 = { 0 };
1876 
1877  if (8 == height) {
1878  LD4(src, stride, tp0, tp1, tp2, tp3);
1879  src += 4 * stride;
1880  LD4(src, stride, tp4, tp5, tp6, tp7);
1881  INSERT_D2_UB(tp0, tp1, src0);
1882  INSERT_D2_UB(tp2, tp3, src1);
1883  INSERT_D2_UB(tp4, tp5, src2);
1884  INSERT_D2_UB(tp6, tp7, src3);
1885  LD4(dst, stride, tp0, tp1, tp2, tp3);
1886  LD4(dst + 4 * stride, stride, tp4, tp5, tp6, tp7);
1887  INSERT_D2_UB(tp0, tp1, dst0);
1888  INSERT_D2_UB(tp2, tp3, dst1);
1889  INSERT_D2_UB(tp4, tp5, dst2);
1890  INSERT_D2_UB(tp6, tp7, dst3);
1891  AVER_UB4_UB(src0, dst0, src1, dst1, src2, dst2, src3, dst3, dst0, dst1,
1892  dst2, dst3);
1893  ST_D8(dst0, dst1, dst2, dst3, 0, 1, 0, 1, 0, 1, 0, 1, dst, stride);
1894  } else if (4 == height) {
1895  LD4(src, stride, tp0, tp1, tp2, tp3);
1896  INSERT_D2_UB(tp0, tp1, src0);
1897  INSERT_D2_UB(tp2, tp3, src1);
1898  LD4(dst, stride, tp0, tp1, tp2, tp3);
1899  INSERT_D2_UB(tp0, tp1, dst0);
1900  INSERT_D2_UB(tp2, tp3, dst1);
1901  AVER_UB2_UB(src0, dst0, src1, dst1, dst0, dst1);
1902  ST_D4(dst0, dst1, 0, 1, 0, 1, dst, stride);
1903  }
1904 }
1905 
1906 void ff_put_h264_chroma_mc8_msa(uint8_t *dst, uint8_t *src,
1907  ptrdiff_t stride, int height, int x, int y)
1908 {
1909  av_assert2(x < 8 && y < 8 && x >= 0 && y >= 0);
1910 
1911  if (x && y) {
1912  avc_chroma_hv_8w_msa(src, dst, stride, x, (8 - x), y, (8 - y), height);
1913  } else if (x) {
1914  avc_chroma_hz_8w_msa(src, dst, stride, x, (8 - x), height);
1915  } else if (y) {
1916  avc_chroma_vt_8w_msa(src, dst, stride, y, (8 - y), height);
1917  } else {
1918  copy_width8_msa(src, dst, stride, height);
1919  }
1920 }
1921 
1922 void ff_put_h264_chroma_mc4_msa(uint8_t *dst, uint8_t *src,
1923  ptrdiff_t stride, int height, int x, int y)
1924 {
1925  av_assert2(x < 8 && y < 8 && x >= 0 && y >= 0);
1926 
1927  if (x && y) {
1928  avc_chroma_hv_4w_msa(src, dst, stride, x, (8 - x), y, (8 - y), height);
1929  } else if (x) {
1930  avc_chroma_hz_4w_msa(src, dst, stride, x, (8 - x), height);
1931  } else if (y) {
1932  avc_chroma_vt_4w_msa(src, dst, stride, y, (8 - y), height);
1933  } else {
1934  copy_width4_msa(src, dst, stride, height);
1935  }
1936 }
1937 
1938 void ff_put_h264_chroma_mc2_msa(uint8_t *dst, uint8_t *src,
1939  ptrdiff_t stride, int height, int x, int y)
1940 {
1941  int32_t cnt;
1942 
1943  av_assert2(x < 8 && y < 8 && x >= 0 && y >= 0);
1944 
1945  if (x && y) {
1946  avc_chroma_hv_2w_msa(src, dst, stride, x, (8 - x), y, (8 - y), height);
1947  } else if (x) {
1948  avc_chroma_hz_2w_msa(src, dst, stride, x, (8 - x), height);
1949  } else if (y) {
1950  avc_chroma_vt_2w_msa(src, dst, stride, y, (8 - y), height);
1951  } else {
1952  for (cnt = height; cnt--;) {
1953  *((uint16_t *) dst) = *((uint16_t *) src);
1954 
1955  src += stride;
1956  dst += stride;
1957  }
1958  }
1959 }
1960 
1961 void ff_avg_h264_chroma_mc8_msa(uint8_t *dst, uint8_t *src,
1962  ptrdiff_t stride, int height, int x, int y)
1963 {
1964  av_assert2(x < 8 && y < 8 && x >= 0 && y >= 0);
1965 
1966 
1967  if (x && y) {
1968  avc_chroma_hv_and_aver_dst_8w_msa(src, dst, stride, x, (8 - x), y,
1969  (8 - y), height);
1970  } else if (x) {
1971  avc_chroma_hz_and_aver_dst_8w_msa(src, dst, stride, x, (8 - x), height);
1972  } else if (y) {
1973  avc_chroma_vt_and_aver_dst_8w_msa(src, dst, stride, y, (8 - y), height);
1974  } else {
1975  avg_width8_msa(src, dst, stride, height);
1976  }
1977 }
1978 
1979 void ff_avg_h264_chroma_mc4_msa(uint8_t *dst, uint8_t *src,
1980  ptrdiff_t stride, int height, int x, int y)
1981 {
1982  av_assert2(x < 8 && y < 8 && x >= 0 && y >= 0);
1983 
1984  if (x && y) {
1985  avc_chroma_hv_and_aver_dst_4w_msa(src, dst, stride, x, (8 - x), y,
1986  (8 - y), height);
1987  } else if (x) {
1988  avc_chroma_hz_and_aver_dst_4w_msa(src, dst, stride, x, (8 - x), height);
1989  } else if (y) {
1990  avc_chroma_vt_and_aver_dst_4w_msa(src, dst, stride, y, (8 - y), height);
1991  } else {
1992  avg_width4_msa(src, dst, stride, height);
1993  }
1994 }
1995 
1996 void ff_avg_h264_chroma_mc2_msa(uint8_t *dst, uint8_t *src,
1997  ptrdiff_t stride, int height, int x, int y)
1998 {
1999  int32_t cnt;
2000 
2001  av_assert2(x < 8 && y < 8 && x >= 0 && y >= 0);
2002 
2003  if (x && y) {
2004  avc_chroma_hv_and_aver_dst_2w_msa(src, dst, stride, x, (8 - x), y,
2005  (8 - y), height);
2006  } else if (x) {
2007  avc_chroma_hz_and_aver_dst_2w_msa(src, dst, stride, x, (8 - x), height);
2008  } else if (y) {
2009  avc_chroma_vt_and_aver_dst_2w_msa(src, dst, stride, y, (8 - y), height);
2010  } else {
2011  for (cnt = height; cnt--;) {
2012  dst[0] = (dst[0] + src[0] + 1) >> 1;
2013  dst[1] = (dst[1] + src[1] + 1) >> 1;
2014 
2015  src += stride;
2016  dst += stride;
2017  }
2018  }
2019 }
stride
int stride
Definition: mace.c:144
avc_chroma_vt_and_aver_dst_2x2_msa
static void avc_chroma_vt_and_aver_dst_2x2_msa(uint8_t *src, uint8_t *dst, int32_t stride, uint32_t coeff0, uint32_t coeff1)
Definition: h264chroma_msa.c:1128
LD_UB8
#define LD_UB8(...)
Definition: generic_macros_msa.h:335
ff_avg_h264_chroma_mc4_msa
void ff_avg_h264_chroma_mc4_msa(uint8_t *dst, uint8_t *src, ptrdiff_t stride, int height, int x, int y)
Definition: h264chroma_msa.c:1979
out
FILE * out
Definition: movenc.c:54
avc_chroma_hv_and_aver_dst_4x8_msa
static void avc_chroma_hv_and_aver_dst_4x8_msa(uint8_t *src, uint8_t *dst, int32_t stride, uint32_t coef_hor0, uint32_t coef_hor1, uint32_t coef_ver0, uint32_t coef_ver1)
Definition: h264chroma_msa.c:1591
avc_chroma_hv_and_aver_dst_2x2_msa
static void avc_chroma_hv_and_aver_dst_2x2_msa(uint8_t *src, uint8_t *dst, int32_t stride, uint32_t coef_hor0, uint32_t coef_hor1, uint32_t coef_ver0, uint32_t coef_ver1)
Definition: h264chroma_msa.c:1412
avc_chroma_hv_4x4_msa
static void avc_chroma_hv_4x4_msa(uint8_t *src, uint8_t *dst, int32_t stride, uint32_t coef_hor0, uint32_t coef_hor1, uint32_t coef_ver0, uint32_t coef_ver1)
Definition: h264chroma_msa.c:640
h264chroma_mips.h
avc_chroma_hz_2w_msa
static void avc_chroma_hz_2w_msa(uint8_t *src, uint8_t *dst, int32_t stride, uint32_t coeff0, uint32_t coeff1, int32_t height)
Definition: h264chroma_msa.c:91
VSHF_B2_UB
#define VSHF_B2_UB(...)
Definition: generic_macros_msa.h:661
AVER_UB2_UB
#define AVER_UB2_UB(...)
Definition: generic_macros_msa.h:595
avc_chroma_hz_8x4_msa
static void avc_chroma_hz_8x4_msa(uint8_t *src, uint8_t *dst, int32_t stride, uint32_t coeff0, uint32_t coeff1)
Definition: h264chroma_msa.c:187
DOTP_UB2_UH
#define DOTP_UB2_UH(...)
Definition: generic_macros_msa.h:740
avc_chroma_vt_4x2_msa
static void avc_chroma_vt_4x2_msa(uint8_t *src, uint8_t *dst, int32_t stride, uint32_t coeff0, uint32_t coeff1)
Definition: h264chroma_msa.c:376
ST_D1
#define ST_D1(in, idx, pdst)
Definition: generic_macros_msa.h:485
avc_chroma_hz_2x4_msa
static void avc_chroma_hz_2x4_msa(uint8_t *src, uint8_t *dst, int32_t stride, uint32_t coeff0, uint32_t coeff1)
Definition: h264chroma_msa.c:63
LD_UB5
#define LD_UB5(...)
Definition: generic_macros_msa.h:307
AVER_UB4_UB
#define AVER_UB4_UB(...)
Definition: generic_macros_msa.h:603
avc_chroma_hz_4x8_msa
static void avc_chroma_hz_4x8_msa(uint8_t *src, uint8_t *dst, int32_t stride, uint32_t coeff0, uint32_t coeff1)
Definition: h264chroma_msa.c:150
SRARI_H4_UH
#define SRARI_H4_UH(...)
Definition: generic_macros_msa.h:2066
avc_chroma_hv_and_aver_dst_8x8_msa
static void avc_chroma_hv_and_aver_dst_8x8_msa(uint8_t *src, uint8_t *dst, int32_t stride, uint32_t coef_hor0, uint32_t coef_hor1, uint32_t coef_ver0, uint32_t coef_ver1)
Definition: h264chroma_msa.c:1708
avc_chroma_hv_and_aver_dst_2w_msa
static void avc_chroma_hv_and_aver_dst_2w_msa(uint8_t *src, uint8_t *dst, int32_t stride, uint32_t coef_hor0, uint32_t coef_hor1, uint32_t coef_ver0, uint32_t coef_ver1, int32_t height)
Definition: h264chroma_msa.c:1499
SH
#define SH(val, pdst)
Definition: generic_macros_msa.h:154
avc_chroma_hv_4x8_msa
static void avc_chroma_hv_4x8_msa(uint8_t *src, uint8_t *dst, int32_t stride, uint32_t coef_hor0, uint32_t coef_hor1, uint32_t coef_ver0, uint32_t coef_ver1)
Definition: h264chroma_msa.c:673
avc_chroma_hz_and_aver_dst_4w_msa
static void avc_chroma_hz_and_aver_dst_4w_msa(uint8_t *src, uint8_t *dst, int32_t stride, uint32_t coeff0, uint32_t coeff1, int32_t height)
Definition: h264chroma_msa.c:1030
ff_put_h264_chroma_mc8_msa
void ff_put_h264_chroma_mc8_msa(uint8_t *dst, uint8_t *src, ptrdiff_t stride, int height, int x, int y)
Definition: h264chroma_msa.c:1906
avc_chroma_hz_4w_msa
static void avc_chroma_hz_4w_msa(uint8_t *src, uint8_t *dst, int32_t stride, uint32_t coeff0, uint32_t coeff1, int32_t height)
Definition: h264chroma_msa.c:174
ADD2
#define ADD2(in0, in1, in2, in3, out0, out1)
Definition: generic_macros_msa.h:2118
generic_macros_msa.h
copy_width8_msa
static void copy_width8_msa(uint8_t *src, uint8_t *dst, int32_t stride, int32_t height)
Definition: h264chroma_msa.c:1817
ST_W4
#define ST_W4(in, idx0, idx1, idx2, idx3, pdst, stride)
Definition: vp8_lpf_lsx.c:234
LD_SB
#define LD_SB(...)
Definition: generic_macros_msa.h:33
ILVR_W2_UB
#define ILVR_W2_UB(...)
Definition: generic_macros_msa.h:1416
LD_UB
#define LD_UB(...)
Definition: generic_macros_msa.h:32
LD_SB5
#define LD_SB5(...)
Definition: generic_macros_msa.h:308
avc_chroma_vt_and_aver_dst_4x4_msa
static void avc_chroma_vt_and_aver_dst_4x4_msa(uint8_t *src, uint8_t *dst, int32_t stride, uint32_t coeff0, uint32_t coeff1)
Definition: h264chroma_msa.c:1250
LH
#define LH(psrc)
Definition: generic_macros_msa.h:89
LD_UB3
#define LD_UB3(...)
Definition: generic_macros_msa.h:288
avg_width4_msa
static void avg_width4_msa(uint8_t *src, uint8_t *dst, int32_t stride, int32_t height)
Definition: h264chroma_msa.c:1835
chroma_mask_arr
static const uint8_t chroma_mask_arr[16 *5]
Definition: h264chroma_msa.c:24
SW
#define SW(val, pdst)
Definition: generic_macros_msa.h:167
mask
static const uint16_t mask[17]
Definition: lzw.c:38
MUL2
#define MUL2(in0, in1, in2, in3, out0, out1)
Definition: generic_macros_msa.h:2101
ILVR_B4_UB
#define ILVR_B4_UB(...)
Definition: generic_macros_msa.h:1359
avg_width8_msa
static void avg_width8_msa(uint8_t *src, uint8_t *dst, int32_t stride, int32_t height)
Definition: h264chroma_msa.c:1870
SAT_UH2_UH
#define SAT_UH2_UH(...)
Definition: generic_macros_msa.h:1567
avc_chroma_hv_2x2_msa
static void avc_chroma_hv_2x2_msa(uint8_t *src, uint8_t *dst, int32_t stride, uint32_t coef_hor0, uint32_t coef_hor1, uint32_t coef_ver0, uint32_t coef_ver1)
Definition: h264chroma_msa.c:529
avc_chroma_hz_nonmult_msa
static void avc_chroma_hz_nonmult_msa(uint8_t *src, uint8_t *dst, int32_t stride, uint32_t coeff0, uint32_t coeff1, int32_t height)
Definition: h264chroma_msa.c:243
DOTP_UB4_UH
#define DOTP_UB4_UH(...)
Definition: generic_macros_msa.h:749
avc_chroma_hz_4x2_msa
static void avc_chroma_hz_4x2_msa(uint8_t *src, uint8_t *dst, int32_t stride, uint32_t coeff0, uint32_t coeff1)
Definition: h264chroma_msa.c:102
PCKEV_B2_SW
#define PCKEV_B2_SW(...)
Definition: generic_macros_msa.h:1722
avc_chroma_vt_4w_msa
static void avc_chroma_vt_4w_msa(uint8_t *src, uint8_t *dst, int32_t stride, uint32_t coeff0, uint32_t coeff1, int32_t height)
Definition: h264chroma_msa.c:452
ST_D8
#define ST_D8(in0, in1, in2, in3, idx0, idx1, idx2, idx3, idx4, idx5, idx6, idx7, pdst, stride)
Definition: generic_macros_msa.h:511
copy_width4_msa
static void copy_width4_msa(uint8_t *src, uint8_t *dst, int32_t stride, int32_t height)
Definition: h264chroma_msa.c:1794
avc_chroma_hz_and_aver_dst_8x4_msa
static void avc_chroma_hz_and_aver_dst_8x4_msa(uint8_t *src, uint8_t *dst, int32_t stride, uint32_t coeff0, uint32_t coeff1)
Definition: h264chroma_msa.c:1043
avc_chroma_hv_4w_msa
static void avc_chroma_hv_4w_msa(uint8_t *src, uint8_t *dst, int32_t stride, uint32_t coef_hor0, uint32_t coef_hor1, uint32_t coef_ver0, uint32_t coef_ver1, int32_t height)
Definition: h264chroma_msa.c:713
avc_chroma_vt_and_aver_dst_8w_msa
static void avc_chroma_vt_and_aver_dst_8w_msa(uint8_t *src, uint8_t *dst, int32_t stride, uint32_t coeff0, uint32_t coeff1, int32_t height)
Definition: h264chroma_msa.c:1401
PCKEV_B2_UB
#define PCKEV_B2_UB(...)
Definition: generic_macros_msa.h:1720
avc_chroma_vt_and_aver_dst_2x4_msa
static void avc_chroma_vt_and_aver_dst_2x4_msa(uint8_t *src, uint8_t *dst, int32_t stride, uint32_t coeff0, uint32_t coeff1)
Definition: h264chroma_msa.c:1165
src
#define src
Definition: vp8dsp.c:255
avc_chroma_hv_8w_msa
static void avc_chroma_hv_8w_msa(uint8_t *src, uint8_t *dst, int32_t stride, uint32_t coef_hor0, uint32_t coef_hor1, uint32_t coef_ver0, uint32_t coef_ver1, int32_t height)
Definition: h264chroma_msa.c:829
avc_chroma_hv_and_aver_dst_4x4_msa
static void avc_chroma_hv_and_aver_dst_4x4_msa(uint8_t *src, uint8_t *dst, int32_t stride, uint32_t coef_hor0, uint32_t coef_hor1, uint32_t coef_ver0, uint32_t coef_ver1)
Definition: h264chroma_msa.c:1552
avc_chroma_hv_and_aver_dst_8w_msa
static void avc_chroma_hv_and_aver_dst_8w_msa(uint8_t *src, uint8_t *dst, int32_t stride, uint32_t coef_hor0, uint32_t coef_hor1, uint32_t coef_ver0, uint32_t coef_ver1, int32_t height)
Definition: h264chroma_msa.c:1777
avc_chroma_vt_8x4_msa
static void avc_chroma_vt_8x4_msa(uint8_t *src, uint8_t *dst, int32_t stride, uint32_t coeff0, uint32_t coeff1)
Definition: h264chroma_msa.c:465
avc_chroma_hz_8w_msa
static void avc_chroma_hz_8w_msa(uint8_t *src, uint8_t *dst, int32_t stride, uint32_t coeff0, uint32_t coeff1, int32_t height)
Definition: h264chroma_msa.c:292
ST_W2
#define ST_W2(in, idx0, idx1, pdst, stride)
Definition: generic_macros_msa.h:450
avc_chroma_vt_2x2_msa
static void avc_chroma_vt_2x2_msa(uint8_t *src, uint8_t *dst, int32_t stride, uint32_t coeff0, uint32_t coeff1)
Definition: h264chroma_msa.c:305
avc_chroma_hz_and_aver_dst_4x2_msa
static void avc_chroma_hz_and_aver_dst_4x2_msa(uint8_t *src, uint8_t *dst, int32_t stride, uint32_t coeff0, uint32_t coeff1)
Definition: h264chroma_msa.c:936
avc_chroma_hz_and_aver_dst_2x2_msa
static void avc_chroma_hz_and_aver_dst_2x2_msa(uint8_t *src, uint8_t *dst, int32_t stride, uint32_t coeff0, uint32_t coeff1)
Definition: h264chroma_msa.c:843
SLLI_4V
#define SLLI_4V(in0, in1, in2, in3, shift)
Definition: generic_macros_msa.h:1921
ff_avg_h264_chroma_mc2_msa
void ff_avg_h264_chroma_mc2_msa(uint8_t *dst, uint8_t *src, ptrdiff_t stride, int height, int x, int y)
Definition: h264chroma_msa.c:1996
avc_chroma_hz_and_aver_dst_4x4_msa
static void avc_chroma_hz_and_aver_dst_4x4_msa(uint8_t *src, uint8_t *dst, int32_t stride, uint32_t coeff0, uint32_t coeff1)
Definition: h264chroma_msa.c:969
ILVR_D2_UB
#define ILVR_D2_UB(...)
Definition: generic_macros_msa.h:1443
avc_chroma_hv_2x4_msa
static void avc_chroma_hv_2x4_msa(uint8_t *src, uint8_t *dst, int32_t stride, uint32_t coef_hor0, uint32_t coef_hor1, uint32_t coef_ver0, uint32_t coef_ver1)
Definition: h264chroma_msa.c:564
SD4
#define SD4(in0, in1, in2, in3, pdst, stride)
Definition: generic_macros_msa.h:256
LD_UB4
#define LD_UB4(...)
Definition: generic_macros_msa.h:296
avc_chroma_hz_4x4_msa
static void avc_chroma_hz_4x4_msa(uint8_t *src, uint8_t *dst, int32_t stride, uint32_t coeff0, uint32_t coeff1)
Definition: h264chroma_msa.c:127
avc_chroma_vt_4x8_msa
static void avc_chroma_vt_4x8_msa(uint8_t *src, uint8_t *dst, int32_t stride, uint32_t coeff0, uint32_t coeff1)
Definition: h264chroma_msa.c:424
ILVR_B2_SB
#define ILVR_B2_SB(...)
Definition: generic_macros_msa.h:1338
avc_chroma_hv_2w_msa
static void avc_chroma_hv_2w_msa(uint8_t *src, uint8_t *dst, int32_t stride, uint32_t coef_hor0, uint32_t coef_hor1, uint32_t coef_ver0, uint32_t coef_ver1, int32_t height)
Definition: h264chroma_msa.c:598
height
#define height
avc_chroma_vt_and_aver_dst_4x2_msa
static void avc_chroma_vt_and_aver_dst_4x2_msa(uint8_t *src, uint8_t *dst, int32_t stride, uint32_t coeff0, uint32_t coeff1)
Definition: h264chroma_msa.c:1218
avc_chroma_hv_and_aver_dst_4x2_msa
static void avc_chroma_hv_and_aver_dst_4x2_msa(uint8_t *src, uint8_t *dst, int32_t stride, uint32_t coef_hor0, uint32_t coef_hor1, uint32_t coef_ver0, uint32_t coef_ver1)
Definition: h264chroma_msa.c:1516
SW4
#define SW4(in0, in1, in2, in3, pdst, stride)
Definition: generic_macros_msa.h:241
avc_chroma_vt_and_aver_dst_8x4_msa
static void avc_chroma_vt_and_aver_dst_8x4_msa(uint8_t *src, uint8_t *dst, int32_t stride, uint32_t coeff0, uint32_t coeff1)
Definition: h264chroma_msa.c:1329
avc_chroma_vt_and_aver_dst_8x8_msa
static void avc_chroma_vt_and_aver_dst_8x8_msa(uint8_t *src, uint8_t *dst, int32_t stride, uint32_t coeff0, uint32_t coeff1)
Definition: h264chroma_msa.c:1358
LW4
#define LW4(psrc, stride, out0, out1, out2, out3)
Definition: generic_macros_msa.h:202
src0
#define src0
Definition: h264pred.c:139
avc_chroma_hz_and_aver_dst_8x8_msa
static void avc_chroma_hz_and_aver_dst_8x8_msa(uint8_t *src, uint8_t *dst, int32_t stride, uint32_t coeff0, uint32_t coeff1)
Definition: h264chroma_msa.c:1073
src1
#define src1
Definition: h264pred.c:140
av_assert2
#define av_assert2(cond)
assert() equivalent, that does lie in speed critical code.
Definition: avassert.h:64
SRARI_H2_UH
#define SRARI_H2_UH(...)
Definition: generic_macros_msa.h:2058
avc_chroma_hz_8x8_msa
static void avc_chroma_hz_8x8_msa(uint8_t *src, uint8_t *dst, int32_t stride, uint32_t coeff0, uint32_t coeff1)
Definition: h264chroma_msa.c:210
avc_chroma_vt_2x4_msa
static void avc_chroma_vt_2x4_msa(uint8_t *src, uint8_t *dst, int32_t stride, uint32_t coeff0, uint32_t coeff1)
Definition: h264chroma_msa.c:337
ST_H4
#define ST_H4(in, idx0, idx1, idx2, idx3, pdst, stride)
Definition: generic_macros_msa.h:417
avc_chroma_vt_4x4_msa
static void avc_chroma_vt_4x4_msa(uint8_t *src, uint8_t *dst, int32_t stride, uint32_t coeff0, uint32_t coeff1)
Definition: h264chroma_msa.c:400
avc_chroma_vt_8x8_msa
static void avc_chroma_vt_8x8_msa(uint8_t *src, uint8_t *dst, int32_t stride, uint32_t coeff0, uint32_t coeff1)
Definition: h264chroma_msa.c:486
avc_chroma_hz_and_aver_dst_2w_msa
static void avc_chroma_hz_and_aver_dst_2w_msa(uint8_t *src, uint8_t *dst, int32_t stride, uint32_t coeff0, uint32_t coeff1, int32_t height)
Definition: h264chroma_msa.c:925
ST_W8
#define ST_W8(in0, in1, idx0, idx1, idx2, idx3, idx4, idx5, idx6, idx7, pdst, stride)
Definition: generic_macros_msa.h:470
LD4
#define LD4(psrc, stride, out0, out1, out2, out3)
Definition: generic_macros_msa.h:228
avc_chroma_hz_and_aver_dst_2x4_msa
static void avc_chroma_hz_and_aver_dst_2x4_msa(uint8_t *src, uint8_t *dst, int32_t stride, uint32_t coeff0, uint32_t coeff1)
Definition: h264chroma_msa.c:885
INSERT_W4_UB
#define INSERT_W4_UB(...)
Definition: generic_macros_msa.h:1153
avc_chroma_hv_and_aver_dst_8x4_msa
static void avc_chroma_hv_and_aver_dst_8x4_msa(uint8_t *src, uint8_t *dst, int32_t stride, uint32_t coef_hor0, uint32_t coef_hor1, uint32_t coef_ver0, uint32_t coef_ver1)
Definition: h264chroma_msa.c:1660
LD_SB3
#define LD_SB3(...)
Definition: generic_macros_msa.h:289
ILVR_B2_UB
#define ILVR_B2_UB(...)
Definition: generic_macros_msa.h:1337
ff_put_h264_chroma_mc2_msa
void ff_put_h264_chroma_mc2_msa(uint8_t *dst, uint8_t *src, ptrdiff_t stride, int height, int x, int y)
Definition: h264chroma_msa.c:1938
avc_chroma_vt_and_aver_dst_2w_msa
static void avc_chroma_vt_and_aver_dst_2w_msa(uint8_t *src, uint8_t *dst, int32_t stride, uint32_t coeff0, uint32_t coeff1, int32_t height)
Definition: h264chroma_msa.c:1207
ST_D4
#define ST_D4(in0, in1, idx0, idx1, idx2, idx3, pdst, stride)
Definition: generic_macros_msa.h:499
avc_chroma_hz_and_aver_dst_4x8_msa
static void avc_chroma_hz_and_aver_dst_4x8_msa(uint8_t *src, uint8_t *dst, int32_t stride, uint32_t coeff0, uint32_t coeff1)
Definition: h264chroma_msa.c:998
avc_chroma_vt_8w_msa
static void avc_chroma_vt_8w_msa(uint8_t *src, uint8_t *dst, int32_t stride, uint32_t coeff0, uint32_t coeff1, int32_t height)
Definition: h264chroma_msa.c:518
ff_put_h264_chroma_mc4_msa
void ff_put_h264_chroma_mc4_msa(uint8_t *dst, uint8_t *src, ptrdiff_t stride, int height, int x, int y)
Definition: h264chroma_msa.c:1922
avc_chroma_hv_4x2_msa
static void avc_chroma_hv_4x2_msa(uint8_t *src, uint8_t *dst, int32_t stride, uint32_t coef_hor0, uint32_t coef_hor1, uint32_t coef_ver0, uint32_t coef_ver1)
Definition: h264chroma_msa.c:612
avc_chroma_hv_and_aver_dst_2x4_msa
static void avc_chroma_hv_and_aver_dst_2x4_msa(uint8_t *src, uint8_t *dst, int32_t stride, uint32_t coef_hor0, uint32_t coef_hor1, uint32_t coef_ver0, uint32_t coef_ver1)
Definition: h264chroma_msa.c:1454
INSERT_W2_UB
#define INSERT_W2_UB(...)
Definition: generic_macros_msa.h:1143
avc_chroma_hv_8x8_msa
static void avc_chroma_hv_8x8_msa(uint8_t *src, uint8_t *dst, int32_t stride, uint32_t coef_hor0, uint32_t coef_hor1, uint32_t coef_ver0, uint32_t coef_ver1)
Definition: h264chroma_msa.c:773
avc_chroma_hv_8x4_msa
static void avc_chroma_hv_8x4_msa(uint8_t *src, uint8_t *dst, int32_t stride, uint32_t coef_hor0, uint32_t coef_hor1, uint32_t coef_ver0, uint32_t coef_ver1)
Definition: h264chroma_msa.c:730
MUL4
#define MUL4(in0, in1, in2, in3, in4, in5, in6, in7, out0, out1, out2, out3)
Definition: generic_macros_msa.h:2106
int32_t
int32_t
Definition: audioconvert.c:56
avc_chroma_hz_2x2_msa
static void avc_chroma_hz_2x2_msa(uint8_t *src, uint8_t *dst, int32_t stride, uint32_t coeff0, uint32_t coeff1)
Definition: h264chroma_msa.c:32
avc_chroma_vt_and_aver_dst_4w_msa
static void avc_chroma_vt_and_aver_dst_4w_msa(uint8_t *src, uint8_t *dst, int32_t stride, uint32_t coeff0, uint32_t coeff1, int32_t height)
Definition: h264chroma_msa.c:1316
avc_chroma_hz_and_aver_dst_8w_msa
static void avc_chroma_hz_and_aver_dst_8w_msa(uint8_t *src, uint8_t *dst, int32_t stride, uint32_t coeff0, uint32_t coeff1, int32_t height)
Definition: h264chroma_msa.c:1117
INSERT_D2_UB
#define INSERT_D2_UB(...)
Definition: generic_macros_msa.h:1169
avc_chroma_vt_and_aver_dst_4x8_msa
static void avc_chroma_vt_and_aver_dst_4x8_msa(uint8_t *src, uint8_t *dst, int32_t stride, uint32_t coeff0, uint32_t coeff1)
Definition: h264chroma_msa.c:1280
avc_chroma_vt_2w_msa
static void avc_chroma_vt_2w_msa(uint8_t *src, uint8_t *dst, int32_t stride, uint32_t coeff0, uint32_t coeff1, int32_t height)
Definition: h264chroma_msa.c:365
avc_chroma_hv_and_aver_dst_4w_msa
static void avc_chroma_hv_and_aver_dst_4w_msa(uint8_t *src, uint8_t *dst, int32_t stride, uint32_t coef_hor0, uint32_t coef_hor1, uint32_t coef_ver0, uint32_t coef_ver1, int32_t height)
Definition: h264chroma_msa.c:1640
ff_avg_h264_chroma_mc8_msa
void ff_avg_h264_chroma_mc8_msa(uint8_t *dst, uint8_t *src, ptrdiff_t stride, int height, int x, int y)
Definition: h264chroma_msa.c:1961
LD_SB2
#define LD_SB2(...)
Definition: generic_macros_msa.h:278
LW2
#define LW2(psrc, stride, out0, out1)
Definition: generic_macros_msa.h:210
SAT_UH4_UH
#define SAT_UH4_UH(...)
Definition: generic_macros_msa.h:1575