2 * Copyright (c) 2015 Shivraj Patil (Shivraj.Patil@imgtec.com)
4 * This file is part of FFmpeg.
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.
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.
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
21 #include "libavcodec/vp9dsp.h"
22 #include "libavutil/mips/generic_macros_msa.h"
23 #include "vp9dsp_mips.h"
25 static const uint8_t mc_filt_mask_arr[16 * 3] = {
27 0, 1, 1, 2, 2, 3, 3, 4, 4, 5, 5, 6, 6, 7, 7, 8,
29 0, 1, 1, 2, 2, 3, 3, 4, 16, 17, 17, 18, 18, 19, 19, 20,
31 8, 9, 9, 10, 10, 11, 11, 12, 24, 25, 25, 26, 26, 27, 27, 28
34 static const int8_t vp9_bilinear_filters_msa[15][2] = {
52 #define FILT_8TAP_DPADD_S_H(vec0, vec1, vec2, vec3, \
53 filt0, filt1, filt2, filt3) \
57 tmp0 = __msa_dotp_s_h((v16i8) vec0, (v16i8) filt0); \
58 tmp0 = __msa_dpadd_s_h(tmp0, (v16i8) vec1, (v16i8) filt1); \
59 tmp1 = __msa_dotp_s_h((v16i8) vec2, (v16i8) filt2); \
60 tmp1 = __msa_dpadd_s_h(tmp1, (v16i8) vec3, (v16i8) filt3); \
61 tmp0 = __msa_adds_s_h(tmp0, tmp1); \
66 #define HORIZ_8TAP_FILT(src0, src1, mask0, mask1, mask2, mask3, \
67 filt_h0, filt_h1, filt_h2, filt_h3) \
69 v16i8 vec0_m, vec1_m, vec2_m, vec3_m; \
72 VSHF_B4_SB(src0, src1, mask0, mask1, mask2, mask3, \
73 vec0_m, vec1_m, vec2_m, vec3_m); \
74 hz_out_m = FILT_8TAP_DPADD_S_H(vec0_m, vec1_m, vec2_m, vec3_m, \
75 filt_h0, filt_h1, filt_h2, filt_h3); \
77 hz_out_m = __msa_srari_h(hz_out_m, 7); \
78 hz_out_m = __msa_sat_s_h(hz_out_m, 7); \
83 #define HORIZ_8TAP_4WID_4VECS_FILT(src0, src1, src2, src3, \
84 mask0, mask1, mask2, mask3, \
85 filt0, filt1, filt2, filt3, \
88 v16i8 vec0_m, vec1_m, vec2_m, vec3_m, vec4_m, vec5_m, vec6_m, vec7_m; \
89 v8i16 res0_m, res1_m, res2_m, res3_m; \
91 VSHF_B2_SB(src0, src1, src2, src3, mask0, mask0, vec0_m, vec1_m); \
92 DOTP_SB2_SH(vec0_m, vec1_m, filt0, filt0, res0_m, res1_m); \
93 VSHF_B2_SB(src0, src1, src2, src3, mask1, mask1, vec2_m, vec3_m); \
94 DPADD_SB2_SH(vec2_m, vec3_m, filt1, filt1, res0_m, res1_m); \
95 VSHF_B2_SB(src0, src1, src2, src3, mask2, mask2, vec4_m, vec5_m); \
96 DOTP_SB2_SH(vec4_m, vec5_m, filt2, filt2, res2_m, res3_m); \
97 VSHF_B2_SB(src0, src1, src2, src3, mask3, mask3, vec6_m, vec7_m); \
98 DPADD_SB2_SH(vec6_m, vec7_m, filt3, filt3, res2_m, res3_m); \
99 ADDS_SH2_SH(res0_m, res2_m, res1_m, res3_m, out0, out1); \
102 #define HORIZ_8TAP_8WID_4VECS_FILT(src0, src1, src2, src3, \
103 mask0, mask1, mask2, mask3, \
104 filt0, filt1, filt2, filt3, \
105 out0, out1, out2, out3) \
107 v16i8 vec0_m, vec1_m, vec2_m, vec3_m, vec4_m, vec5_m, vec6_m, vec7_m; \
108 v8i16 res0_m, res1_m, res2_m, res3_m, res4_m, res5_m, res6_m, res7_m; \
110 VSHF_B2_SB(src0, src0, src1, src1, mask0, mask0, vec0_m, vec1_m); \
111 VSHF_B2_SB(src2, src2, src3, src3, mask0, mask0, vec2_m, vec3_m); \
112 DOTP_SB4_SH(vec0_m, vec1_m, vec2_m, vec3_m, filt0, filt0, filt0, filt0, \
113 res0_m, res1_m, res2_m, res3_m); \
114 VSHF_B2_SB(src0, src0, src1, src1, mask2, mask2, vec0_m, vec1_m); \
115 VSHF_B2_SB(src2, src2, src3, src3, mask2, mask2, vec2_m, vec3_m); \
116 DOTP_SB4_SH(vec0_m, vec1_m, vec2_m, vec3_m, filt2, filt2, filt2, filt2, \
117 res4_m, res5_m, res6_m, res7_m); \
118 VSHF_B2_SB(src0, src0, src1, src1, mask1, mask1, vec4_m, vec5_m); \
119 VSHF_B2_SB(src2, src2, src3, src3, mask1, mask1, vec6_m, vec7_m); \
120 DPADD_SB4_SH(vec4_m, vec5_m, vec6_m, vec7_m, filt1, filt1, filt1, filt1, \
121 res0_m, res1_m, res2_m, res3_m); \
122 VSHF_B2_SB(src0, src0, src1, src1, mask3, mask3, vec4_m, vec5_m); \
123 VSHF_B2_SB(src2, src2, src3, src3, mask3, mask3, vec6_m, vec7_m); \
124 DPADD_SB4_SH(vec4_m, vec5_m, vec6_m, vec7_m, filt3, filt3, filt3, filt3, \
125 res4_m, res5_m, res6_m, res7_m); \
126 ADDS_SH4_SH(res0_m, res4_m, res1_m, res5_m, res2_m, res6_m, res3_m, \
127 res7_m, out0, out1, out2, out3); \
130 #define PCKEV_XORI128_AVG_ST_UB(in0, in1, dst, pdst) \
134 tmp_m = PCKEV_XORI128_UB(in1, in0); \
135 tmp_m = __msa_aver_u_b(tmp_m, (v16u8) dst); \
136 ST_UB(tmp_m, (pdst)); \
139 #define PCKEV_AVG_ST_UB(in0, in1, dst, pdst) \
143 tmp_m = (v16u8) __msa_pckev_b((v16i8) in0, (v16i8) in1); \
144 tmp_m = __msa_aver_u_b(tmp_m, (v16u8) dst); \
145 ST_UB(tmp_m, (pdst)); \
148 #define PCKEV_AVG_ST8x4_UB(in1, dst0, in2, dst1, in3, dst2, in4, dst3, \
151 v16u8 tmp0_m, tmp1_m, tmp2_m, tmp3_m; \
152 uint8_t *pdst_m = (uint8_t *) (pdst); \
154 PCKEV_B2_UB(in2, in1, in4, in3, tmp0_m, tmp1_m); \
155 PCKEV_D2_UB(dst1, dst0, dst3, dst2, tmp2_m, tmp3_m); \
156 AVER_UB2_UB(tmp0_m, tmp2_m, tmp1_m, tmp3_m, tmp0_m, tmp1_m); \
157 ST8x4_UB(tmp0_m, tmp1_m, pdst_m, stride); \
160 static void common_hz_8t_4x4_msa(const uint8_t *src, int32_t src_stride,
161 uint8_t *dst, int32_t dst_stride,
162 const int8_t *filter)
164 v16u8 mask0, mask1, mask2, mask3, out;
165 v16i8 src0, src1, src2, src3, filt0, filt1, filt2, filt3;
166 v8i16 filt, out0, out1;
168 mask0 = LD_UB(&mc_filt_mask_arr[16]);
171 /* rearranging filter */
172 filt = LD_SH(filter);
173 SPLATI_H4_SB(filt, 0, 1, 2, 3, filt0, filt1, filt2, filt3);
179 LD_SB4(src, src_stride, src0, src1, src2, src3);
180 XORI_B4_128_SB(src0, src1, src2, src3);
181 HORIZ_8TAP_4WID_4VECS_FILT(src0, src1, src2, src3, mask0, mask1, mask2,
182 mask3, filt0, filt1, filt2, filt3, out0, out1);
183 SRARI_H2_SH(out0, out1, 7);
184 SAT_SH2_SH(out0, out1, 7);
185 out = PCKEV_XORI128_UB(out0, out1);
186 ST4x4_UB(out, out, 0, 1, 2, 3, dst, dst_stride);
189 static void common_hz_8t_4x8_msa(const uint8_t *src, int32_t src_stride,
190 uint8_t *dst, int32_t dst_stride,
191 const int8_t *filter)
193 v16i8 filt0, filt1, filt2, filt3;
194 v16i8 src0, src1, src2, src3;
195 v16u8 mask0, mask1, mask2, mask3, out;
196 v8i16 filt, out0, out1, out2, out3;
198 mask0 = LD_UB(&mc_filt_mask_arr[16]);
201 /* rearranging filter */
202 filt = LD_SH(filter);
203 SPLATI_H4_SB(filt, 0, 1, 2, 3, filt0, filt1, filt2, filt3);
209 LD_SB4(src, src_stride, src0, src1, src2, src3);
210 XORI_B4_128_SB(src0, src1, src2, src3);
211 src += (4 * src_stride);
212 HORIZ_8TAP_4WID_4VECS_FILT(src0, src1, src2, src3, mask0, mask1, mask2,
213 mask3, filt0, filt1, filt2, filt3, out0, out1);
214 LD_SB4(src, src_stride, src0, src1, src2, src3);
215 XORI_B4_128_SB(src0, src1, src2, src3);
216 HORIZ_8TAP_4WID_4VECS_FILT(src0, src1, src2, src3, mask0, mask1, mask2,
217 mask3, filt0, filt1, filt2, filt3, out2, out3);
218 SRARI_H4_SH(out0, out1, out2, out3, 7);
219 SAT_SH4_SH(out0, out1, out2, out3, 7);
220 out = PCKEV_XORI128_UB(out0, out1);
221 ST4x4_UB(out, out, 0, 1, 2, 3, dst, dst_stride);
222 dst += (4 * dst_stride);
223 out = PCKEV_XORI128_UB(out2, out3);
224 ST4x4_UB(out, out, 0, 1, 2, 3, dst, dst_stride);
227 static void common_hz_8t_4x16_msa(const uint8_t *src, int32_t src_stride,
228 uint8_t *dst, int32_t dst_stride,
229 const int8_t *filter)
231 v16u8 mask0, mask1, mask2, mask3, out;
232 v16i8 src0, src1, src2, src3, filt0, filt1, filt2, filt3;
233 v8i16 filt, out0, out1, out2, out3;
235 mask0 = LD_UB(&mc_filt_mask_arr[16]);
238 /* rearranging filter */
239 filt = LD_SH(filter);
240 SPLATI_H4_SB(filt, 0, 1, 2, 3, filt0, filt1, filt2, filt3);
246 LD_SB4(src, src_stride, src0, src1, src2, src3);
247 XORI_B4_128_SB(src0, src1, src2, src3);
248 src += (4 * src_stride);
249 HORIZ_8TAP_4WID_4VECS_FILT(src0, src1, src2, src3, mask0, mask1, mask2,
250 mask3, filt0, filt1, filt2, filt3, out0, out1);
251 LD_SB4(src, src_stride, src0, src1, src2, src3);
252 XORI_B4_128_SB(src0, src1, src2, src3);
253 src += (4 * src_stride);
254 HORIZ_8TAP_4WID_4VECS_FILT(src0, src1, src2, src3, mask0, mask1, mask2,
255 mask3, filt0, filt1, filt2, filt3, out2, out3);
256 SRARI_H4_SH(out0, out1, out2, out3, 7);
257 SAT_SH4_SH(out0, out1, out2, out3, 7);
258 out = PCKEV_XORI128_UB(out0, out1);
259 ST4x4_UB(out, out, 0, 1, 2, 3, dst, dst_stride);
260 dst += (4 * dst_stride);
261 out = PCKEV_XORI128_UB(out2, out3);
262 ST4x4_UB(out, out, 0, 1, 2, 3, dst, dst_stride);
263 dst += (4 * dst_stride);
265 LD_SB4(src, src_stride, src0, src1, src2, src3);
266 XORI_B4_128_SB(src0, src1, src2, src3);
267 src += (4 * src_stride);
268 HORIZ_8TAP_4WID_4VECS_FILT(src0, src1, src2, src3, mask0, mask1, mask2,
269 mask3, filt0, filt1, filt2, filt3, out0, out1);
270 LD_SB4(src, src_stride, src0, src1, src2, src3);
271 XORI_B4_128_SB(src0, src1, src2, src3);
272 src += (4 * src_stride);
273 HORIZ_8TAP_4WID_4VECS_FILT(src0, src1, src2, src3, mask0, mask1, mask2,
274 mask3, filt0, filt1, filt2, filt3, out2, out3);
276 SRARI_H4_SH(out0, out1, out2, out3, 7);
277 SAT_SH4_SH(out0, out1, out2, out3, 7);
278 out = PCKEV_XORI128_UB(out0, out1);
279 ST4x4_UB(out, out, 0, 1, 2, 3, dst, dst_stride);
280 dst += (4 * dst_stride);
281 out = PCKEV_XORI128_UB(out2, out3);
282 ST4x4_UB(out, out, 0, 1, 2, 3, dst, dst_stride);
285 static void common_hz_8t_4w_msa(const uint8_t *src, int32_t src_stride,
286 uint8_t *dst, int32_t dst_stride,
287 const int8_t *filter, int32_t height)
290 common_hz_8t_4x4_msa(src, src_stride, dst, dst_stride, filter);
291 } else if (8 == height) {
292 common_hz_8t_4x8_msa(src, src_stride, dst, dst_stride, filter);
293 } else if (16 == height) {
294 common_hz_8t_4x16_msa(src, src_stride, dst, dst_stride, filter);
298 static void common_hz_8t_8x4_msa(const uint8_t *src, int32_t src_stride,
299 uint8_t *dst, int32_t dst_stride,
300 const int8_t *filter)
302 v16i8 src0, src1, src2, src3, filt0, filt1, filt2, filt3;
303 v16u8 mask0, mask1, mask2, mask3, tmp0, tmp1;
304 v8i16 filt, out0, out1, out2, out3;
306 mask0 = LD_UB(&mc_filt_mask_arr[0]);
309 /* rearranging filter */
310 filt = LD_SH(filter);
311 SPLATI_H4_SB(filt, 0, 1, 2, 3, filt0, filt1, filt2, filt3);
317 LD_SB4(src, src_stride, src0, src1, src2, src3);
318 XORI_B4_128_SB(src0, src1, src2, src3);
319 HORIZ_8TAP_8WID_4VECS_FILT(src0, src1, src2, src3, mask0, mask1, mask2,
320 mask3, filt0, filt1, filt2, filt3, out0, out1,
322 SRARI_H4_SH(out0, out1, out2, out3, 7);
323 SAT_SH4_SH(out0, out1, out2, out3, 7);
324 tmp0 = PCKEV_XORI128_UB(out0, out1);
325 tmp1 = PCKEV_XORI128_UB(out2, out3);
326 ST8x4_UB(tmp0, tmp1, dst, dst_stride);
329 static void common_hz_8t_8x8mult_msa(const uint8_t *src, int32_t src_stride,
330 uint8_t *dst, int32_t dst_stride,
331 const int8_t *filter, int32_t height)
334 v16i8 src0, src1, src2, src3, filt0, filt1, filt2, filt3;
335 v16u8 mask0, mask1, mask2, mask3, tmp0, tmp1;
336 v8i16 filt, out0, out1, out2, out3;
338 mask0 = LD_UB(&mc_filt_mask_arr[0]);
341 /* rearranging filter */
342 filt = LD_SH(filter);
343 SPLATI_H4_SB(filt, 0, 1, 2, 3, filt0, filt1, filt2, filt3);
349 for (loop_cnt = (height >> 2); loop_cnt--;) {
350 LD_SB4(src, src_stride, src0, src1, src2, src3);
351 XORI_B4_128_SB(src0, src1, src2, src3);
352 src += (4 * src_stride);
353 HORIZ_8TAP_8WID_4VECS_FILT(src0, src1, src2, src3, mask0, mask1, mask2,
354 mask3, filt0, filt1, filt2, filt3, out0,
356 SRARI_H4_SH(out0, out1, out2, out3, 7);
357 SAT_SH4_SH(out0, out1, out2, out3, 7);
358 tmp0 = PCKEV_XORI128_UB(out0, out1);
359 tmp1 = PCKEV_XORI128_UB(out2, out3);
360 ST8x4_UB(tmp0, tmp1, dst, dst_stride);
361 dst += (4 * dst_stride);
365 static void common_hz_8t_8w_msa(const uint8_t *src, int32_t src_stride,
366 uint8_t *dst, int32_t dst_stride,
367 const int8_t *filter, int32_t height)
370 common_hz_8t_8x4_msa(src, src_stride, dst, dst_stride, filter);
372 common_hz_8t_8x8mult_msa(src, src_stride, dst, dst_stride, filter,
377 static void common_hz_8t_16w_msa(const uint8_t *src, int32_t src_stride,
378 uint8_t *dst, int32_t dst_stride,
379 const int8_t *filter, int32_t height)
382 v16i8 src0, src1, src2, src3, filt0, filt1, filt2, filt3;
383 v16u8 mask0, mask1, mask2, mask3, out;
384 v8i16 filt, out0, out1, out2, out3;
386 mask0 = LD_UB(&mc_filt_mask_arr[0]);
389 /* rearranging filter */
390 filt = LD_SH(filter);
391 SPLATI_H4_SB(filt, 0, 1, 2, 3, filt0, filt1, filt2, filt3);
397 for (loop_cnt = (height >> 1); loop_cnt--;) {
398 LD_SB2(src, src_stride, src0, src2);
399 LD_SB2(src + 8, src_stride, src1, src3);
400 XORI_B4_128_SB(src0, src1, src2, src3);
401 src += (2 * src_stride);
402 HORIZ_8TAP_8WID_4VECS_FILT(src0, src1, src2, src3, mask0, mask1, mask2,
403 mask3, filt0, filt1, filt2, filt3, out0,
405 SRARI_H4_SH(out0, out1, out2, out3, 7);
406 SAT_SH4_SH(out0, out1, out2, out3, 7);
407 out = PCKEV_XORI128_UB(out0, out1);
410 out = PCKEV_XORI128_UB(out2, out3);
416 static void common_hz_8t_32w_msa(const uint8_t *src, int32_t src_stride,
417 uint8_t *dst, int32_t dst_stride,
418 const int8_t *filter, int32_t height)
421 v16i8 src0, src1, src2, src3, filt0, filt1, filt2, filt3;
422 v16u8 mask0, mask1, mask2, mask3, out;
423 v8i16 filt, out0, out1, out2, out3;
425 mask0 = LD_UB(&mc_filt_mask_arr[0]);
428 /* rearranging filter */
429 filt = LD_SH(filter);
430 SPLATI_H4_SB(filt, 0, 1, 2, 3, filt0, filt1, filt2, filt3);
436 for (loop_cnt = (height >> 1); loop_cnt--;) {
438 src2 = LD_SB(src + 16);
439 src3 = LD_SB(src + 24);
440 src1 = __msa_sldi_b(src2, src0, 8);
442 XORI_B4_128_SB(src0, src1, src2, src3);
443 HORIZ_8TAP_8WID_4VECS_FILT(src0, src1, src2, src3, mask0, mask1, mask2,
444 mask3, filt0, filt1, filt2, filt3, out0,
446 SRARI_H4_SH(out0, out1, out2, out3, 7);
447 SAT_SH4_SH(out0, out1, out2, out3, 7);
450 src2 = LD_SB(src + 16);
451 src3 = LD_SB(src + 24);
452 src1 = __msa_sldi_b(src2, src0, 8);
455 out = PCKEV_XORI128_UB(out0, out1);
457 out = PCKEV_XORI128_UB(out2, out3);
458 ST_UB(out, dst + 16);
461 XORI_B4_128_SB(src0, src1, src2, src3);
462 HORIZ_8TAP_8WID_4VECS_FILT(src0, src1, src2, src3, mask0, mask1, mask2,
463 mask3, filt0, filt1, filt2, filt3, out0,
465 SRARI_H4_SH(out0, out1, out2, out3, 7);
466 SAT_SH4_SH(out0, out1, out2, out3, 7);
467 out = PCKEV_XORI128_UB(out0, out1);
469 out = PCKEV_XORI128_UB(out2, out3);
470 ST_UB(out, dst + 16);
475 static void common_hz_8t_64w_msa(const uint8_t *src, int32_t src_stride,
476 uint8_t *dst, int32_t dst_stride,
477 const int8_t *filter, int32_t height)
480 v16i8 src0, src1, src2, src3, filt0, filt1, filt2, filt3;
481 v16u8 mask0, mask1, mask2, mask3, out;
482 v8i16 filt, out0, out1, out2, out3;
484 mask0 = LD_UB(&mc_filt_mask_arr[0]);
487 /* rearranging filter */
488 filt = LD_SH(filter);
489 SPLATI_H4_SB(filt, 0, 1, 2, 3, filt0, filt1, filt2, filt3);
495 for (loop_cnt = height; loop_cnt--;) {
497 src2 = LD_SB(src + 16);
498 src3 = LD_SB(src + 24);
499 src1 = __msa_sldi_b(src2, src0, 8);
501 XORI_B4_128_SB(src0, src1, src2, src3);
502 HORIZ_8TAP_8WID_4VECS_FILT(src0, src1, src2, src3, mask0, mask1,
503 mask2, mask3, filt0, filt1, filt2, filt3,
504 out0, out1, out2, out3);
505 SRARI_H4_SH(out0, out1, out2, out3, 7);
506 SAT_SH4_SH(out0, out1, out2, out3, 7);
507 out = PCKEV_XORI128_UB(out0, out1);
509 out = PCKEV_XORI128_UB(out2, out3);
510 ST_UB(out, dst + 16);
512 src0 = LD_SB(src + 32);
513 src2 = LD_SB(src + 48);
514 src3 = LD_SB(src + 56);
515 src1 = __msa_sldi_b(src2, src0, 8);
518 XORI_B4_128_SB(src0, src1, src2, src3);
519 HORIZ_8TAP_8WID_4VECS_FILT(src0, src1, src2, src3, mask0, mask1,
520 mask2, mask3, filt0, filt1, filt2, filt3,
521 out0, out1, out2, out3);
522 SRARI_H4_SH(out0, out1, out2, out3, 7);
523 SAT_SH4_SH(out0, out1, out2, out3, 7);
524 out = PCKEV_XORI128_UB(out0, out1);
525 ST_UB(out, dst + 32);
526 out = PCKEV_XORI128_UB(out2, out3);
527 ST_UB(out, dst + 48);
532 static void common_vt_8t_4w_msa(const uint8_t *src, int32_t src_stride,
533 uint8_t *dst, int32_t dst_stride,
534 const int8_t *filter, int32_t height)
537 v16i8 src0, src1, src2, src3, src4, src5, src6, src7, src8, src9, src10;
538 v16i8 src10_r, src32_r, src54_r, src76_r, src98_r, src21_r, src43_r;
539 v16i8 src65_r, src87_r, src109_r, src2110, src4332, src6554, src8776;
540 v16i8 src10998, filt0, filt1, filt2, filt3;
542 v8i16 filt, out10, out32;
544 src -= (3 * src_stride);
546 filt = LD_SH(filter);
547 SPLATI_H4_SB(filt, 0, 1, 2, 3, filt0, filt1, filt2, filt3);
549 LD_SB7(src, src_stride, src0, src1, src2, src3, src4, src5, src6);
550 src += (7 * src_stride);
552 ILVR_B4_SB(src1, src0, src3, src2, src5, src4, src2, src1, src10_r, src32_r,
554 ILVR_B2_SB(src4, src3, src6, src5, src43_r, src65_r);
555 ILVR_D3_SB(src21_r, src10_r, src43_r, src32_r, src65_r, src54_r, src2110,
557 XORI_B3_128_SB(src2110, src4332, src6554);
559 for (loop_cnt = (height >> 2); loop_cnt--;) {
560 LD_SB4(src, src_stride, src7, src8, src9, src10);
561 src += (4 * src_stride);
563 ILVR_B4_SB(src7, src6, src8, src7, src9, src8, src10, src9, src76_r,
564 src87_r, src98_r, src109_r);
565 ILVR_D2_SB(src87_r, src76_r, src109_r, src98_r, src8776, src10998);
566 XORI_B2_128_SB(src8776, src10998);
567 out10 = FILT_8TAP_DPADD_S_H(src2110, src4332, src6554, src8776, filt0,
568 filt1, filt2, filt3);
569 out32 = FILT_8TAP_DPADD_S_H(src4332, src6554, src8776, src10998, filt0,
570 filt1, filt2, filt3);
571 SRARI_H2_SH(out10, out32, 7);
572 SAT_SH2_SH(out10, out32, 7);
573 out = PCKEV_XORI128_UB(out10, out32);
574 ST4x4_UB(out, out, 0, 1, 2, 3, dst, dst_stride);
575 dst += (4 * dst_stride);
584 static void common_vt_8t_8w_msa(const uint8_t *src, int32_t src_stride,
585 uint8_t *dst, int32_t dst_stride,
586 const int8_t *filter, int32_t height)
589 v16i8 src0, src1, src2, src3, src4, src5, src6, src7, src8, src9, src10;
590 v16i8 src10_r, src32_r, src54_r, src76_r, src98_r, src21_r, src43_r;
591 v16i8 src65_r, src87_r, src109_r, filt0, filt1, filt2, filt3;
593 v8i16 filt, out0_r, out1_r, out2_r, out3_r;
595 src -= (3 * src_stride);
597 filt = LD_SH(filter);
598 SPLATI_H4_SB(filt, 0, 1, 2, 3, filt0, filt1, filt2, filt3);
600 LD_SB7(src, src_stride, src0, src1, src2, src3, src4, src5, src6);
601 XORI_B7_128_SB(src0, src1, src2, src3, src4, src5, src6);
602 src += (7 * src_stride);
603 ILVR_B4_SB(src1, src0, src3, src2, src5, src4, src2, src1, src10_r, src32_r,
605 ILVR_B2_SB(src4, src3, src6, src5, src43_r, src65_r);
607 for (loop_cnt = (height >> 2); loop_cnt--;) {
608 LD_SB4(src, src_stride, src7, src8, src9, src10);
609 XORI_B4_128_SB(src7, src8, src9, src10);
610 src += (4 * src_stride);
612 ILVR_B4_SB(src7, src6, src8, src7, src9, src8, src10, src9, src76_r,
613 src87_r, src98_r, src109_r);
614 out0_r = FILT_8TAP_DPADD_S_H(src10_r, src32_r, src54_r, src76_r, filt0,
615 filt1, filt2, filt3);
616 out1_r = FILT_8TAP_DPADD_S_H(src21_r, src43_r, src65_r, src87_r, filt0,
617 filt1, filt2, filt3);
618 out2_r = FILT_8TAP_DPADD_S_H(src32_r, src54_r, src76_r, src98_r, filt0,
619 filt1, filt2, filt3);
620 out3_r = FILT_8TAP_DPADD_S_H(src43_r, src65_r, src87_r, src109_r, filt0,
621 filt1, filt2, filt3);
622 SRARI_H4_SH(out0_r, out1_r, out2_r, out3_r, 7);
623 SAT_SH4_SH(out0_r, out1_r, out2_r, out3_r, 7);
624 tmp0 = PCKEV_XORI128_UB(out0_r, out1_r);
625 tmp1 = PCKEV_XORI128_UB(out2_r, out3_r);
626 ST8x4_UB(tmp0, tmp1, dst, dst_stride);
627 dst += (4 * dst_stride);
639 static void common_vt_8t_16w_msa(const uint8_t *src, int32_t src_stride,
640 uint8_t *dst, int32_t dst_stride,
641 const int8_t *filter, int32_t height)
644 v16i8 src0, src1, src2, src3, src4, src5, src6, src7, src8, src9, src10;
645 v16i8 filt0, filt1, filt2, filt3;
646 v16i8 src10_r, src32_r, src54_r, src76_r, src98_r, src21_r, src43_r;
647 v16i8 src65_r, src87_r, src109_r, src10_l, src32_l, src54_l, src76_l;
648 v16i8 src98_l, src21_l, src43_l, src65_l, src87_l, src109_l;
649 v16u8 tmp0, tmp1, tmp2, tmp3;
650 v8i16 filt, out0_r, out1_r, out2_r, out3_r, out0_l, out1_l, out2_l, out3_l;
652 src -= (3 * src_stride);
654 filt = LD_SH(filter);
655 SPLATI_H4_SB(filt, 0, 1, 2, 3, filt0, filt1, filt2, filt3);
657 LD_SB7(src, src_stride, src0, src1, src2, src3, src4, src5, src6);
658 XORI_B7_128_SB(src0, src1, src2, src3, src4, src5, src6);
659 src += (7 * src_stride);
660 ILVR_B4_SB(src1, src0, src3, src2, src5, src4, src2, src1, src10_r, src32_r,
662 ILVR_B2_SB(src4, src3, src6, src5, src43_r, src65_r);
663 ILVL_B4_SB(src1, src0, src3, src2, src5, src4, src2, src1, src10_l, src32_l,
665 ILVL_B2_SB(src4, src3, src6, src5, src43_l, src65_l);
667 for (loop_cnt = (height >> 2); loop_cnt--;) {
668 LD_SB4(src, src_stride, src7, src8, src9, src10);
669 XORI_B4_128_SB(src7, src8, src9, src10);
670 src += (4 * src_stride);
672 ILVR_B4_SB(src7, src6, src8, src7, src9, src8, src10, src9, src76_r,
673 src87_r, src98_r, src109_r);
674 ILVL_B4_SB(src7, src6, src8, src7, src9, src8, src10, src9, src76_l,
675 src87_l, src98_l, src109_l);
676 out0_r = FILT_8TAP_DPADD_S_H(src10_r, src32_r, src54_r, src76_r, filt0,
677 filt1, filt2, filt3);
678 out1_r = FILT_8TAP_DPADD_S_H(src21_r, src43_r, src65_r, src87_r, filt0,
679 filt1, filt2, filt3);
680 out2_r = FILT_8TAP_DPADD_S_H(src32_r, src54_r, src76_r, src98_r, filt0,
681 filt1, filt2, filt3);
682 out3_r = FILT_8TAP_DPADD_S_H(src43_r, src65_r, src87_r, src109_r, filt0,
683 filt1, filt2, filt3);
684 out0_l = FILT_8TAP_DPADD_S_H(src10_l, src32_l, src54_l, src76_l, filt0,
685 filt1, filt2, filt3);
686 out1_l = FILT_8TAP_DPADD_S_H(src21_l, src43_l, src65_l, src87_l, filt0,
687 filt1, filt2, filt3);
688 out2_l = FILT_8TAP_DPADD_S_H(src32_l, src54_l, src76_l, src98_l, filt0,
689 filt1, filt2, filt3);
690 out3_l = FILT_8TAP_DPADD_S_H(src43_l, src65_l, src87_l, src109_l, filt0,
691 filt1, filt2, filt3);
692 SRARI_H4_SH(out0_r, out1_r, out2_r, out3_r, 7);
693 SRARI_H4_SH(out0_l, out1_l, out2_l, out3_l, 7);
694 SAT_SH4_SH(out0_r, out1_r, out2_r, out3_r, 7);
695 SAT_SH4_SH(out0_l, out1_l, out2_l, out3_l, 7);
696 PCKEV_B4_UB(out0_l, out0_r, out1_l, out1_r, out2_l, out2_r, out3_l,
697 out3_r, tmp0, tmp1, tmp2, tmp3);
698 XORI_B4_128_UB(tmp0, tmp1, tmp2, tmp3);
699 ST_UB4(tmp0, tmp1, tmp2, tmp3, dst, dst_stride);
700 dst += (4 * dst_stride);
718 static void common_vt_8t_16w_mult_msa(const uint8_t *src, int32_t src_stride,
719 uint8_t *dst, int32_t dst_stride,
720 const int8_t *filter, int32_t height,
723 const uint8_t *src_tmp;
725 uint32_t loop_cnt, cnt;
726 v16i8 src0, src1, src2, src3, src4, src5, src6, src7, src8, src9, src10;
727 v16i8 filt0, filt1, filt2, filt3;
728 v16i8 src10_r, src32_r, src54_r, src76_r, src98_r, src21_r, src43_r;
729 v16i8 src65_r, src87_r, src109_r, src10_l, src32_l, src54_l, src76_l;
730 v16i8 src98_l, src21_l, src43_l, src65_l, src87_l, src109_l;
731 v16u8 tmp0, tmp1, tmp2, tmp3;
732 v8i16 filt, out0_r, out1_r, out2_r, out3_r, out0_l, out1_l, out2_l, out3_l;
734 src -= (3 * src_stride);
736 filt = LD_SH(filter);
737 SPLATI_H4_SB(filt, 0, 1, 2, 3, filt0, filt1, filt2, filt3);
739 for (cnt = (width >> 4); cnt--;) {
743 LD_SB7(src_tmp, src_stride, src0, src1, src2, src3, src4, src5, src6);
744 XORI_B7_128_SB(src0, src1, src2, src3, src4, src5, src6);
745 src_tmp += (7 * src_stride);
746 ILVR_B4_SB(src1, src0, src3, src2, src5, src4, src2, src1, src10_r,
747 src32_r, src54_r, src21_r);
748 ILVR_B2_SB(src4, src3, src6, src5, src43_r, src65_r);
749 ILVL_B4_SB(src1, src0, src3, src2, src5, src4, src2, src1, src10_l,
750 src32_l, src54_l, src21_l);
751 ILVL_B2_SB(src4, src3, src6, src5, src43_l, src65_l);
753 for (loop_cnt = (height >> 2); loop_cnt--;) {
754 LD_SB4(src_tmp, src_stride, src7, src8, src9, src10);
755 XORI_B4_128_SB(src7, src8, src9, src10);
756 src_tmp += (4 * src_stride);
757 ILVR_B4_SB(src7, src6, src8, src7, src9, src8, src10, src9, src76_r,
758 src87_r, src98_r, src109_r);
759 ILVL_B4_SB(src7, src6, src8, src7, src9, src8, src10, src9, src76_l,
760 src87_l, src98_l, src109_l);
761 out0_r = FILT_8TAP_DPADD_S_H(src10_r, src32_r, src54_r, src76_r,
762 filt0, filt1, filt2, filt3);
763 out1_r = FILT_8TAP_DPADD_S_H(src21_r, src43_r, src65_r, src87_r,
764 filt0, filt1, filt2, filt3);
765 out2_r = FILT_8TAP_DPADD_S_H(src32_r, src54_r, src76_r, src98_r,
766 filt0, filt1, filt2, filt3);
767 out3_r = FILT_8TAP_DPADD_S_H(src43_r, src65_r, src87_r, src109_r,
768 filt0, filt1, filt2, filt3);
769 out0_l = FILT_8TAP_DPADD_S_H(src10_l, src32_l, src54_l, src76_l,
770 filt0, filt1, filt2, filt3);
771 out1_l = FILT_8TAP_DPADD_S_H(src21_l, src43_l, src65_l, src87_l,
772 filt0, filt1, filt2, filt3);
773 out2_l = FILT_8TAP_DPADD_S_H(src32_l, src54_l, src76_l, src98_l,
774 filt0, filt1, filt2, filt3);
775 out3_l = FILT_8TAP_DPADD_S_H(src43_l, src65_l, src87_l, src109_l,
776 filt0, filt1, filt2, filt3);
777 SRARI_H4_SH(out0_r, out1_r, out2_r, out3_r, 7);
778 SRARI_H4_SH(out0_l, out1_l, out2_l, out3_l, 7);
779 SAT_SH4_SH(out0_r, out1_r, out2_r, out3_r, 7);
780 SAT_SH4_SH(out0_l, out1_l, out2_l, out3_l, 7);
781 PCKEV_B4_UB(out0_l, out0_r, out1_l, out1_r, out2_l, out2_r, out3_l,
782 out3_r, tmp0, tmp1, tmp2, tmp3);
783 XORI_B4_128_UB(tmp0, tmp1, tmp2, tmp3);
784 ST_UB4(tmp0, tmp1, tmp2, tmp3, dst_tmp, dst_stride);
785 dst_tmp += (4 * dst_stride);
807 static void common_vt_8t_32w_msa(const uint8_t *src, int32_t src_stride,
808 uint8_t *dst, int32_t dst_stride,
809 const int8_t *filter, int32_t height)
811 common_vt_8t_16w_mult_msa(src, src_stride, dst, dst_stride, filter, height,
815 static void common_vt_8t_64w_msa(const uint8_t *src, int32_t src_stride,
816 uint8_t *dst, int32_t dst_stride,
817 const int8_t *filter, int32_t height)
819 common_vt_8t_16w_mult_msa(src, src_stride, dst, dst_stride, filter, height,
823 static void common_hv_8ht_8vt_4w_msa(const uint8_t *src, int32_t src_stride,
824 uint8_t *dst, int32_t dst_stride,
825 const int8_t *filter_horiz,
826 const int8_t *filter_vert,
830 v16i8 src0, src1, src2, src3, src4, src5, src6, src7, src8, src9, src10;
831 v16i8 filt_hz0, filt_hz1, filt_hz2, filt_hz3;
832 v16u8 mask0, mask1, mask2, mask3, out;
833 v8i16 hz_out0, hz_out1, hz_out2, hz_out3, hz_out4, hz_out5, hz_out6;
834 v8i16 hz_out7, hz_out8, hz_out9, tmp0, tmp1, out0, out1, out2, out3, out4;
835 v8i16 filt, filt_vt0, filt_vt1, filt_vt2, filt_vt3;
837 mask0 = LD_UB(&mc_filt_mask_arr[16]);
838 src -= (3 + 3 * src_stride);
840 /* rearranging filter */
841 filt = LD_SH(filter_horiz);
842 SPLATI_H4_SB(filt, 0, 1, 2, 3, filt_hz0, filt_hz1, filt_hz2, filt_hz3);
848 LD_SB7(src, src_stride, src0, src1, src2, src3, src4, src5, src6);
849 XORI_B7_128_SB(src0, src1, src2, src3, src4, src5, src6);
850 src += (7 * src_stride);
852 hz_out0 = HORIZ_8TAP_FILT(src0, src1, mask0, mask1, mask2, mask3, filt_hz0,
853 filt_hz1, filt_hz2, filt_hz3);
854 hz_out2 = HORIZ_8TAP_FILT(src2, src3, mask0, mask1, mask2, mask3, filt_hz0,
855 filt_hz1, filt_hz2, filt_hz3);
856 hz_out4 = HORIZ_8TAP_FILT(src4, src5, mask0, mask1, mask2, mask3, filt_hz0,
857 filt_hz1, filt_hz2, filt_hz3);
858 hz_out5 = HORIZ_8TAP_FILT(src5, src6, mask0, mask1, mask2, mask3, filt_hz0,
859 filt_hz1, filt_hz2, filt_hz3);
860 SLDI_B2_SH(hz_out2, hz_out4, hz_out0, hz_out2, hz_out1, hz_out3, 8);
862 filt = LD_SH(filter_vert);
863 SPLATI_H4_SH(filt, 0, 1, 2, 3, filt_vt0, filt_vt1, filt_vt2, filt_vt3);
865 ILVEV_B2_SH(hz_out0, hz_out1, hz_out2, hz_out3, out0, out1);
866 out2 = (v8i16) __msa_ilvev_b((v16i8) hz_out5, (v16i8) hz_out4);
868 for (loop_cnt = (height >> 2); loop_cnt--;) {
869 LD_SB4(src, src_stride, src7, src8, src9, src10);
870 XORI_B4_128_SB(src7, src8, src9, src10);
871 src += (4 * src_stride);
873 hz_out7 = HORIZ_8TAP_FILT(src7, src8, mask0, mask1, mask2, mask3,
874 filt_hz0, filt_hz1, filt_hz2, filt_hz3);
875 hz_out6 = (v8i16) __msa_sldi_b((v16i8) hz_out7, (v16i8) hz_out5, 8);
876 out3 = (v8i16) __msa_ilvev_b((v16i8) hz_out7, (v16i8) hz_out6);
877 tmp0 = FILT_8TAP_DPADD_S_H(out0, out1, out2, out3, filt_vt0, filt_vt1,
880 hz_out9 = HORIZ_8TAP_FILT(src9, src10, mask0, mask1, mask2, mask3,
881 filt_hz0, filt_hz1, filt_hz2, filt_hz3);
882 hz_out8 = (v8i16) __msa_sldi_b((v16i8) hz_out9, (v16i8) hz_out7, 8);
883 out4 = (v8i16) __msa_ilvev_b((v16i8) hz_out9, (v16i8) hz_out8);
884 tmp1 = FILT_8TAP_DPADD_S_H(out1, out2, out3, out4, filt_vt0, filt_vt1,
886 SRARI_H2_SH(tmp0, tmp1, 7);
887 SAT_SH2_SH(tmp0, tmp1, 7);
888 out = PCKEV_XORI128_UB(tmp0, tmp1);
889 ST4x4_UB(out, out, 0, 1, 2, 3, dst, dst_stride);
890 dst += (4 * dst_stride);
899 static void common_hv_8ht_8vt_8w_msa(const uint8_t *src, int32_t src_stride,
900 uint8_t *dst, int32_t dst_stride,
901 const int8_t *filter_horiz,
902 const int8_t *filter_vert,
906 v16i8 src0, src1, src2, src3, src4, src5, src6, src7, src8, src9, src10;
907 v16i8 filt_hz0, filt_hz1, filt_hz2, filt_hz3;
908 v16u8 mask0, mask1, mask2, mask3, vec0, vec1;
909 v8i16 filt, filt_vt0, filt_vt1, filt_vt2, filt_vt3;
910 v8i16 hz_out0, hz_out1, hz_out2, hz_out3, hz_out4, hz_out5, hz_out6;
911 v8i16 hz_out7, hz_out8, hz_out9, hz_out10, tmp0, tmp1, tmp2, tmp3;
912 v8i16 out0, out1, out2, out3, out4, out5, out6, out7, out8, out9;
914 mask0 = LD_UB(&mc_filt_mask_arr[0]);
915 src -= (3 + 3 * src_stride);
917 /* rearranging filter */
918 filt = LD_SH(filter_horiz);
919 SPLATI_H4_SB(filt, 0, 1, 2, 3, filt_hz0, filt_hz1, filt_hz2, filt_hz3);
925 LD_SB7(src, src_stride, src0, src1, src2, src3, src4, src5, src6);
926 src += (7 * src_stride);
928 XORI_B7_128_SB(src0, src1, src2, src3, src4, src5, src6);
929 hz_out0 = HORIZ_8TAP_FILT(src0, src0, mask0, mask1, mask2, mask3, filt_hz0,
930 filt_hz1, filt_hz2, filt_hz3);
931 hz_out1 = HORIZ_8TAP_FILT(src1, src1, mask0, mask1, mask2, mask3, filt_hz0,
932 filt_hz1, filt_hz2, filt_hz3);
933 hz_out2 = HORIZ_8TAP_FILT(src2, src2, mask0, mask1, mask2, mask3, filt_hz0,
934 filt_hz1, filt_hz2, filt_hz3);
935 hz_out3 = HORIZ_8TAP_FILT(src3, src3, mask0, mask1, mask2, mask3, filt_hz0,
936 filt_hz1, filt_hz2, filt_hz3);
937 hz_out4 = HORIZ_8TAP_FILT(src4, src4, mask0, mask1, mask2, mask3, filt_hz0,
938 filt_hz1, filt_hz2, filt_hz3);
939 hz_out5 = HORIZ_8TAP_FILT(src5, src5, mask0, mask1, mask2, mask3, filt_hz0,
940 filt_hz1, filt_hz2, filt_hz3);
941 hz_out6 = HORIZ_8TAP_FILT(src6, src6, mask0, mask1, mask2, mask3, filt_hz0,
942 filt_hz1, filt_hz2, filt_hz3);
944 filt = LD_SH(filter_vert);
945 SPLATI_H4_SH(filt, 0, 1, 2, 3, filt_vt0, filt_vt1, filt_vt2, filt_vt3);
947 ILVEV_B2_SH(hz_out0, hz_out1, hz_out2, hz_out3, out0, out1);
948 ILVEV_B2_SH(hz_out4, hz_out5, hz_out1, hz_out2, out2, out4);
949 ILVEV_B2_SH(hz_out3, hz_out4, hz_out5, hz_out6, out5, out6);
951 for (loop_cnt = (height >> 2); loop_cnt--;) {
952 LD_SB4(src, src_stride, src7, src8, src9, src10);
953 src += (4 * src_stride);
955 XORI_B4_128_SB(src7, src8, src9, src10);
957 hz_out7 = HORIZ_8TAP_FILT(src7, src7, mask0, mask1, mask2, mask3,
958 filt_hz0, filt_hz1, filt_hz2, filt_hz3);
959 out3 = (v8i16) __msa_ilvev_b((v16i8) hz_out7, (v16i8) hz_out6);
960 tmp0 = FILT_8TAP_DPADD_S_H(out0, out1, out2, out3, filt_vt0, filt_vt1,
963 hz_out8 = HORIZ_8TAP_FILT(src8, src8, mask0, mask1, mask2, mask3,
964 filt_hz0, filt_hz1, filt_hz2, filt_hz3);
965 out7 = (v8i16) __msa_ilvev_b((v16i8) hz_out8, (v16i8) hz_out7);
966 tmp1 = FILT_8TAP_DPADD_S_H(out4, out5, out6, out7, filt_vt0, filt_vt1,
969 hz_out9 = HORIZ_8TAP_FILT(src9, src9, mask0, mask1, mask2, mask3,
970 filt_hz0, filt_hz1, filt_hz2, filt_hz3);
971 out8 = (v8i16) __msa_ilvev_b((v16i8) hz_out9, (v16i8) hz_out8);
972 tmp2 = FILT_8TAP_DPADD_S_H(out1, out2, out3, out8, filt_vt0,
973 filt_vt1, filt_vt2, filt_vt3);
975 hz_out10 = HORIZ_8TAP_FILT(src10, src10, mask0, mask1, mask2, mask3,
976 filt_hz0, filt_hz1, filt_hz2, filt_hz3);
977 out9 = (v8i16) __msa_ilvev_b((v16i8) hz_out10, (v16i8) hz_out9);
978 tmp3 = FILT_8TAP_DPADD_S_H(out5, out6, out7, out9, filt_vt0, filt_vt1,
980 SRARI_H4_SH(tmp0, tmp1, tmp2, tmp3, 7);
981 SAT_SH4_SH(tmp0, tmp1, tmp2, tmp3, 7);
982 vec0 = PCKEV_XORI128_UB(tmp0, tmp1);
983 vec1 = PCKEV_XORI128_UB(tmp2, tmp3);
984 ST8x4_UB(vec0, vec1, dst, dst_stride);
985 dst += (4 * dst_stride);
997 static void common_hv_8ht_8vt_16w_msa(const uint8_t *src, int32_t src_stride,
998 uint8_t *dst, int32_t dst_stride,
999 const int8_t *filter_horiz,
1000 const int8_t *filter_vert,
1003 int32_t multiple8_cnt;
1005 for (multiple8_cnt = 2; multiple8_cnt--;) {
1006 common_hv_8ht_8vt_8w_msa(src, src_stride, dst, dst_stride, filter_horiz,
1007 filter_vert, height);
1014 static void common_hv_8ht_8vt_32w_msa(const uint8_t *src, int32_t src_stride,
1015 uint8_t *dst, int32_t dst_stride,
1016 const int8_t *filter_horiz,
1017 const int8_t *filter_vert,
1020 int32_t multiple8_cnt;
1022 for (multiple8_cnt = 4; multiple8_cnt--;) {
1023 common_hv_8ht_8vt_8w_msa(src, src_stride, dst, dst_stride, filter_horiz,
1024 filter_vert, height);
1031 static void common_hv_8ht_8vt_64w_msa(const uint8_t *src, int32_t src_stride,
1032 uint8_t *dst, int32_t dst_stride,
1033 const int8_t *filter_horiz,
1034 const int8_t *filter_vert,
1037 int32_t multiple8_cnt;
1039 for (multiple8_cnt = 8; multiple8_cnt--;) {
1040 common_hv_8ht_8vt_8w_msa(src, src_stride, dst, dst_stride, filter_horiz,
1041 filter_vert, height);
1048 static void common_hz_8t_and_aver_dst_4x4_msa(const uint8_t *src,
1050 uint8_t *dst, int32_t dst_stride,
1051 const int8_t *filter)
1053 v16i8 src0, src1, src2, src3, filt0, filt1, filt2, filt3;
1054 v16u8 dst0, dst1, dst2, dst3, res2, res3;
1055 v16u8 mask0, mask1, mask2, mask3;
1056 v8i16 filt, res0, res1;
1058 mask0 = LD_UB(&mc_filt_mask_arr[16]);
1061 /* rearranging filter */
1062 filt = LD_SH(filter);
1063 SPLATI_H4_SB(filt, 0, 1, 2, 3, filt0, filt1, filt2, filt3);
1069 LD_SB4(src, src_stride, src0, src1, src2, src3);
1070 XORI_B4_128_SB(src0, src1, src2, src3);
1071 HORIZ_8TAP_4WID_4VECS_FILT(src0, src1, src2, src3, mask0, mask1, mask2,
1072 mask3, filt0, filt1, filt2, filt3, res0, res1);
1073 LD_UB4(dst, dst_stride, dst0, dst1, dst2, dst3);
1074 SRARI_H2_SH(res0, res1, 7);
1075 SAT_SH2_SH(res0, res1, 7);
1076 PCKEV_B2_UB(res0, res0, res1, res1, res2, res3);
1077 ILVR_W2_UB(dst1, dst0, dst3, dst2, dst0, dst2);
1078 XORI_B2_128_UB(res2, res3);
1079 AVER_UB2_UB(res2, dst0, res3, dst2, res2, res3);
1080 ST4x4_UB(res2, res3, 0, 1, 0, 1, dst, dst_stride);
1083 static void common_hz_8t_and_aver_dst_4x8_msa(const uint8_t *src,
1085 uint8_t *dst, int32_t dst_stride,
1086 const int8_t *filter)
1088 v16i8 src0, src1, src2, src3, filt0, filt1, filt2, filt3;
1089 v16u8 mask0, mask1, mask2, mask3, res0, res1, res2, res3;
1090 v16u8 dst0, dst1, dst2, dst3, dst4, dst5, dst6, dst7;
1091 v8i16 filt, vec0, vec1, vec2, vec3;
1093 mask0 = LD_UB(&mc_filt_mask_arr[16]);
1096 /* rearranging filter */
1097 filt = LD_SH(filter);
1098 SPLATI_H4_SB(filt, 0, 1, 2, 3, filt0, filt1, filt2, filt3);
1104 LD_SB4(src, src_stride, src0, src1, src2, src3);
1105 XORI_B4_128_SB(src0, src1, src2, src3);
1106 src += (4 * src_stride);
1107 LD_UB8(dst, dst_stride, dst0, dst1, dst2, dst3, dst4, dst5, dst6, dst7);
1108 HORIZ_8TAP_4WID_4VECS_FILT(src0, src1, src2, src3, mask0, mask1, mask2,
1109 mask3, filt0, filt1, filt2, filt3, vec0, vec1);
1110 LD_SB4(src, src_stride, src0, src1, src2, src3);
1111 XORI_B4_128_SB(src0, src1, src2, src3);
1112 HORIZ_8TAP_4WID_4VECS_FILT(src0, src1, src2, src3, mask0, mask1, mask2,
1113 mask3, filt0, filt1, filt2, filt3, vec2, vec3);
1114 SRARI_H4_SH(vec0, vec1, vec2, vec3, 7);
1115 SAT_SH4_SH(vec0, vec1, vec2, vec3, 7);
1116 PCKEV_B4_UB(vec0, vec0, vec1, vec1, vec2, vec2, vec3, vec3,
1117 res0, res1, res2, res3);
1118 ILVR_D2_UB(res1, res0, res3, res2, res0, res2);
1119 XORI_B2_128_UB(res0, res2);
1120 ILVR_W4_UB(dst1, dst0, dst3, dst2, dst5, dst4, dst7, dst6,
1121 dst0, dst2, dst4, dst6);
1122 ILVR_D2_UB(dst2, dst0, dst6, dst4, dst0, dst4);
1123 AVER_UB2_UB(res0, dst0, res2, dst4, res0, res2);
1124 ST4x8_UB(res0, res2, dst, dst_stride);
1127 static void common_hz_8t_and_aver_dst_4w_msa(const uint8_t *src,
1129 uint8_t *dst, int32_t dst_stride,
1130 const int8_t *filter,
1134 common_hz_8t_and_aver_dst_4x4_msa(src, src_stride, dst, dst_stride,
1136 } else if (8 == height) {
1137 common_hz_8t_and_aver_dst_4x8_msa(src, src_stride, dst, dst_stride,
1142 static void common_hz_8t_and_aver_dst_8w_msa(const uint8_t *src,
1144 uint8_t *dst, int32_t dst_stride,
1145 const int8_t *filter,
1149 v16i8 src0, src1, src2, src3, filt0, filt1, filt2, filt3;
1150 v16u8 mask0, mask1, mask2, mask3, dst0, dst1, dst2, dst3;
1151 v8i16 filt, out0, out1, out2, out3;
1153 mask0 = LD_UB(&mc_filt_mask_arr[0]);
1156 /* rearranging filter */
1157 filt = LD_SH(filter);
1158 SPLATI_H4_SB(filt, 0, 1, 2, 3, filt0, filt1, filt2, filt3);
1164 for (loop_cnt = (height >> 2); loop_cnt--;) {
1165 LD_SB4(src, src_stride, src0, src1, src2, src3);
1166 XORI_B4_128_SB(src0, src1, src2, src3);
1167 src += (4 * src_stride);
1168 HORIZ_8TAP_8WID_4VECS_FILT(src0, src1, src2, src3, mask0, mask1, mask2,
1169 mask3, filt0, filt1, filt2, filt3, out0,
1171 LD_UB4(dst, dst_stride, dst0, dst1, dst2, dst3);
1172 SRARI_H4_SH(out0, out1, out2, out3, 7);
1173 SAT_SH4_SH(out0, out1, out2, out3, 7);
1174 CONVERT_UB_AVG_ST8x4_UB(out0, out1, out2, out3, dst0, dst1, dst2, dst3,
1176 dst += (4 * dst_stride);
1180 static void common_hz_8t_and_aver_dst_16w_msa(const uint8_t *src,
1182 uint8_t *dst, int32_t dst_stride,
1183 const int8_t *filter,
1187 v16i8 src0, src1, src2, src3, filt0, filt1, filt2, filt3;
1188 v16u8 mask0, mask1, mask2, mask3, dst0, dst1;
1189 v8i16 filt, out0, out1, out2, out3;
1190 v8i16 vec0, vec1, vec2, vec3, vec4, vec5, vec6, vec7;
1191 v8i16 vec8, vec9, vec10, vec11, vec12, vec13, vec14, vec15;
1193 mask0 = LD_UB(&mc_filt_mask_arr[0]);
1196 /* rearranging filter */
1197 filt = LD_SH(filter);
1198 SPLATI_H4_SB(filt, 0, 1, 2, 3, filt0, filt1, filt2, filt3);
1204 for (loop_cnt = height >> 1; loop_cnt--;) {
1205 LD_SB2(src, src_stride, src0, src2);
1206 LD_SB2(src + 8, src_stride, src1, src3);
1207 src += (2 * src_stride);
1209 XORI_B4_128_SB(src0, src1, src2, src3);
1210 VSHF_B4_SH(src0, src0, mask0, mask1, mask2, mask3, vec0, vec4, vec8,
1212 VSHF_B4_SH(src1, src1, mask0, mask1, mask2, mask3, vec1, vec5, vec9,
1214 VSHF_B4_SH(src2, src2, mask0, mask1, mask2, mask3, vec2, vec6, vec10,
1216 VSHF_B4_SH(src3, src3, mask0, mask1, mask2, mask3, vec3, vec7, vec11,
1218 DOTP_SB4_SH(vec0, vec1, vec2, vec3, filt0, filt0, filt0, filt0, vec0,
1220 DOTP_SB4_SH(vec8, vec9, vec10, vec11, filt2, filt2, filt2, filt2, vec8,
1221 vec9, vec10, vec11);
1222 DPADD_SB4_SH(vec4, vec5, vec6, vec7, filt1, filt1, filt1, filt1, vec0,
1224 DPADD_SB4_SH(vec12, vec13, vec14, vec15, filt3, filt3, filt3, filt3,
1225 vec8, vec9, vec10, vec11);
1226 ADDS_SH4_SH(vec0, vec8, vec1, vec9, vec2, vec10, vec3, vec11, out0,
1228 LD_UB2(dst, dst_stride, dst0, dst1);
1229 SRARI_H4_SH(out0, out1, out2, out3, 7);
1230 SAT_SH4_SH(out0, out1, out2, out3, 7);
1231 PCKEV_XORI128_AVG_ST_UB(out1, out0, dst0, dst);
1233 PCKEV_XORI128_AVG_ST_UB(out3, out2, dst1, dst);
1238 static void common_hz_8t_and_aver_dst_32w_msa(const uint8_t *src,
1240 uint8_t *dst, int32_t dst_stride,
1241 const int8_t *filter,
1245 v16i8 src0, src1, src2, src3, filt0, filt1, filt2, filt3;
1246 v16u8 dst1, dst2, mask0, mask1, mask2, mask3;
1247 v8i16 filt, out0, out1, out2, out3;
1248 v8i16 vec0, vec1, vec2, vec3, vec4, vec5, vec6, vec7;
1249 v8i16 vec8, vec9, vec10, vec11, vec12, vec13, vec14, vec15;
1251 mask0 = LD_UB(&mc_filt_mask_arr[0]);
1254 /* rearranging filter */
1255 filt = LD_SH(filter);
1256 SPLATI_H4_SB(filt, 0, 1, 2, 3, filt0, filt1, filt2, filt3);
1262 for (loop_cnt = height; loop_cnt--;) {
1264 src2 = LD_SB(src + 16);
1265 src3 = LD_SB(src + 24);
1266 src1 = __msa_sldi_b(src2, src0, 8);
1269 XORI_B4_128_SB(src0, src1, src2, src3);
1270 VSHF_B4_SH(src0, src0, mask0, mask1, mask2, mask3, vec0, vec4, vec8,
1272 VSHF_B4_SH(src1, src1, mask0, mask1, mask2, mask3, vec1, vec5, vec9,
1274 VSHF_B4_SH(src2, src2, mask0, mask1, mask2, mask3, vec2, vec6, vec10,
1276 VSHF_B4_SH(src3, src3, mask0, mask1, mask2, mask3, vec3, vec7, vec11,
1278 DOTP_SB4_SH(vec0, vec1, vec2, vec3, filt0, filt0, filt0, filt0, vec0,
1280 DOTP_SB4_SH(vec8, vec9, vec10, vec11, filt2, filt2, filt2, filt2, vec8,
1281 vec9, vec10, vec11);
1282 DPADD_SB4_SH(vec4, vec5, vec6, vec7, filt1, filt1, filt1, filt1, vec0,
1284 DPADD_SB4_SH(vec12, vec13, vec14, vec15, filt3, filt3, filt3, filt3,
1285 vec8, vec9, vec10, vec11);
1286 ADDS_SH4_SH(vec0, vec8, vec1, vec9, vec2, vec10, vec3, vec11, out0,
1288 SRARI_H4_SH(out0, out1, out2, out3, 7);
1289 SAT_SH4_SH(out0, out1, out2, out3, 7);
1290 LD_UB2(dst, 16, dst1, dst2);
1291 PCKEV_XORI128_AVG_ST_UB(out1, out0, dst1, dst);
1292 PCKEV_XORI128_AVG_ST_UB(out3, out2, dst2, dst + 16);
1297 static void common_hz_8t_and_aver_dst_64w_msa(const uint8_t *src,
1299 uint8_t *dst, int32_t dst_stride,
1300 const int8_t *filter,
1303 uint32_t loop_cnt, cnt;
1304 v16i8 src0, src1, src2, src3, filt0, filt1, filt2, filt3;
1305 v16u8 dst1, dst2, mask0, mask1, mask2, mask3;
1306 v8i16 filt, out0, out1, out2, out3;
1307 v8i16 vec0, vec1, vec2, vec3, vec4, vec5, vec6, vec7;
1308 v8i16 vec8, vec9, vec10, vec11, vec12, vec13, vec14, vec15;
1310 mask0 = LD_UB(&mc_filt_mask_arr[0]);
1313 /* rearranging filter */
1314 filt = LD_SH(filter);
1315 SPLATI_H4_SB(filt, 0, 1, 2, 3, filt0, filt1, filt2, filt3);
1321 for (loop_cnt = height; loop_cnt--;) {
1322 for (cnt = 0; cnt < 2; ++cnt) {
1323 src0 = LD_SB(&src[cnt << 5]);
1324 src2 = LD_SB(&src[16 + (cnt << 5)]);
1325 src3 = LD_SB(&src[24 + (cnt << 5)]);
1326 src1 = __msa_sldi_b(src2, src0, 8);
1328 XORI_B4_128_SB(src0, src1, src2, src3);
1329 VSHF_B4_SH(src0, src0, mask0, mask1, mask2, mask3, vec0, vec4, vec8,
1331 VSHF_B4_SH(src1, src1, mask0, mask1, mask2, mask3, vec1, vec5, vec9,
1333 VSHF_B4_SH(src2, src2, mask0, mask1, mask2, mask3, vec2, vec6,
1335 VSHF_B4_SH(src3, src3, mask0, mask1, mask2, mask3, vec3, vec7,
1337 DOTP_SB4_SH(vec0, vec1, vec2, vec3, filt0, filt0, filt0, filt0,
1338 vec0, vec1, vec2, vec3);
1339 DOTP_SB4_SH(vec8, vec9, vec10, vec11, filt2, filt2, filt2, filt2,
1340 vec8, vec9, vec10, vec11);
1341 DPADD_SB4_SH(vec4, vec5, vec6, vec7, filt1, filt1, filt1, filt1,
1342 vec0, vec1, vec2, vec3);
1343 DPADD_SB4_SH(vec12, vec13, vec14, vec15, filt3, filt3, filt3, filt3,
1344 vec8, vec9, vec10, vec11);
1345 ADDS_SH4_SH(vec0, vec8, vec1, vec9, vec2, vec10, vec3, vec11, out0,
1347 SRARI_H4_SH(out0, out1, out2, out3, 7);
1348 SAT_SH4_SH(out0, out1, out2, out3, 7);
1349 LD_UB2(&dst[cnt << 5], 16, dst1, dst2);
1350 PCKEV_XORI128_AVG_ST_UB(out1, out0, dst1, &dst[cnt << 5]);
1351 PCKEV_XORI128_AVG_ST_UB(out3, out2, dst2, &dst[16 + (cnt << 5)]);
1359 static void common_vt_8t_and_aver_dst_4w_msa(const uint8_t *src,
1361 uint8_t *dst, int32_t dst_stride,
1362 const int8_t *filter,
1366 v16i8 src0, src1, src2, src3, src4, src5, src6, src7, src8, src9, src10;
1367 v16u8 dst0, dst1, dst2, dst3, out;
1368 v16i8 src10_r, src32_r, src54_r, src76_r, src98_r, src21_r, src43_r;
1369 v16i8 src65_r, src87_r, src109_r, src2110, src4332, src6554, src8776;
1370 v16i8 src10998, filt0, filt1, filt2, filt3;
1371 v8i16 filt, out10, out32;
1373 src -= (3 * src_stride);
1375 filt = LD_SH(filter);
1376 SPLATI_H4_SB(filt, 0, 1, 2, 3, filt0, filt1, filt2, filt3);
1378 LD_SB7(src, src_stride, src0, src1, src2, src3, src4, src5, src6);
1379 src += (7 * src_stride);
1381 ILVR_B4_SB(src1, src0, src3, src2, src5, src4, src2, src1, src10_r, src32_r,
1383 ILVR_B2_SB(src4, src3, src6, src5, src43_r, src65_r);
1384 ILVR_D3_SB(src21_r, src10_r, src43_r, src32_r, src65_r, src54_r, src2110,
1386 XORI_B3_128_SB(src2110, src4332, src6554);
1388 for (loop_cnt = (height >> 2); loop_cnt--;) {
1389 LD_SB4(src, src_stride, src7, src8, src9, src10);
1390 src += (4 * src_stride);
1392 LD_UB4(dst, dst_stride, dst0, dst1, dst2, dst3);
1393 ILVR_B4_SB(src7, src6, src8, src7, src9, src8, src10, src9, src76_r,
1394 src87_r, src98_r, src109_r);
1395 ILVR_D2_SB(src87_r, src76_r, src109_r, src98_r, src8776, src10998);
1396 XORI_B2_128_SB(src8776, src10998);
1397 out10 = FILT_8TAP_DPADD_S_H(src2110, src4332, src6554, src8776, filt0,
1398 filt1, filt2, filt3);
1399 out32 = FILT_8TAP_DPADD_S_H(src4332, src6554, src8776, src10998, filt0,
1400 filt1, filt2, filt3);
1401 SRARI_H2_SH(out10, out32, 7);
1402 SAT_SH2_SH(out10, out32, 7);
1403 out = PCKEV_XORI128_UB(out10, out32);
1404 ILVR_W2_UB(dst1, dst0, dst3, dst2, dst0, dst2);
1406 dst0 = (v16u8) __msa_ilvr_d((v2i64) dst2, (v2i64) dst0);
1407 out = __msa_aver_u_b(out, dst0);
1409 ST4x4_UB(out, out, 0, 1, 2, 3, dst, dst_stride);
1410 dst += (4 * dst_stride);
1419 static void common_vt_8t_and_aver_dst_8w_msa(const uint8_t *src,
1421 uint8_t *dst, int32_t dst_stride,
1422 const int8_t *filter,
1426 v16i8 src0, src1, src2, src3, src4, src5, src6, src7, src8, src9, src10;
1427 v16u8 dst0, dst1, dst2, dst3;
1428 v16i8 src10_r, src32_r, src54_r, src76_r, src98_r, src21_r, src43_r;
1429 v16i8 src65_r, src87_r, src109_r, filt0, filt1, filt2, filt3;
1430 v8i16 filt, out0, out1, out2, out3;
1432 src -= (3 * src_stride);
1434 filt = LD_SH(filter);
1435 SPLATI_H4_SB(filt, 0, 1, 2, 3, filt0, filt1, filt2, filt3);
1437 LD_SB7(src, src_stride, src0, src1, src2, src3, src4, src5, src6);
1438 src += (7 * src_stride);
1440 XORI_B7_128_SB(src0, src1, src2, src3, src4, src5, src6);
1441 ILVR_B4_SB(src1, src0, src3, src2, src5, src4, src2, src1, src10_r, src32_r,
1443 ILVR_B2_SB(src4, src3, src6, src5, src43_r, src65_r);
1445 for (loop_cnt = (height >> 2); loop_cnt--;) {
1446 LD_SB4(src, src_stride, src7, src8, src9, src10);
1447 src += (4 * src_stride);
1449 LD_UB4(dst, dst_stride, dst0, dst1, dst2, dst3);
1450 XORI_B4_128_SB(src7, src8, src9, src10);
1451 ILVR_B4_SB(src7, src6, src8, src7, src9, src8, src10, src9, src76_r,
1452 src87_r, src98_r, src109_r);
1453 out0 = FILT_8TAP_DPADD_S_H(src10_r, src32_r, src54_r, src76_r, filt0,
1454 filt1, filt2, filt3);
1455 out1 = FILT_8TAP_DPADD_S_H(src21_r, src43_r, src65_r, src87_r, filt0,
1456 filt1, filt2, filt3);
1457 out2 = FILT_8TAP_DPADD_S_H(src32_r, src54_r, src76_r, src98_r, filt0,
1458 filt1, filt2, filt3);
1459 out3 = FILT_8TAP_DPADD_S_H(src43_r, src65_r, src87_r, src109_r, filt0,
1460 filt1, filt2, filt3);
1461 SRARI_H4_SH(out0, out1, out2, out3, 7);
1462 SAT_SH4_SH(out0, out1, out2, out3, 7);
1463 CONVERT_UB_AVG_ST8x4_UB(out0, out1, out2, out3, dst0, dst1, dst2, dst3,
1465 dst += (4 * dst_stride);
1477 static void common_vt_8t_and_aver_dst_16w_mult_msa(const uint8_t *src,
1481 const int8_t *filter,
1485 const uint8_t *src_tmp;
1487 uint32_t loop_cnt, cnt;
1488 v16i8 src0, src1, src2, src3, src4, src5, src6, src7, src8, src9, src10;
1489 v16i8 src10_r, src32_r, src54_r, src76_r, src98_r, src21_r, src43_r;
1490 v16i8 src65_r, src87_r, src109_r, src10_l, src32_l, src54_l, src76_l;
1491 v16i8 src98_l, src21_l, src43_l, src65_l, src87_l, src109_l;
1492 v16i8 filt0, filt1, filt2, filt3;
1493 v16u8 dst0, dst1, dst2, dst3, tmp0, tmp1, tmp2, tmp3;
1494 v8i16 out0_r, out1_r, out2_r, out3_r, out0_l, out1_l, out2_l, out3_l, filt;
1496 src -= (3 * src_stride);
1498 filt = LD_SH(filter);
1499 SPLATI_H4_SB(filt, 0, 1, 2, 3, filt0, filt1, filt2, filt3);
1501 for (cnt = (width >> 4); cnt--;) {
1505 LD_SB7(src_tmp, src_stride, src0, src1, src2, src3, src4, src5, src6);
1506 XORI_B7_128_SB(src0, src1, src2, src3, src4, src5, src6);
1507 src_tmp += (7 * src_stride);
1509 ILVR_B4_SB(src1, src0, src3, src2, src5, src4, src2, src1, src10_r,
1510 src32_r, src54_r, src21_r);
1511 ILVR_B2_SB(src4, src3, src6, src5, src43_r, src65_r);
1512 ILVL_B4_SB(src1, src0, src3, src2, src5, src4, src2, src1, src10_l,
1513 src32_l, src54_l, src21_l);
1514 ILVL_B2_SB(src4, src3, src6, src5, src43_l, src65_l);
1516 for (loop_cnt = (height >> 2); loop_cnt--;) {
1517 LD_SB4(src_tmp, src_stride, src7, src8, src9, src10);
1518 src_tmp += (4 * src_stride);
1520 LD_UB4(dst_tmp, dst_stride, dst0, dst1, dst2, dst3);
1521 XORI_B4_128_SB(src7, src8, src9, src10);
1522 ILVR_B4_SB(src7, src6, src8, src7, src9, src8, src10, src9, src76_r,
1523 src87_r, src98_r, src109_r);
1524 ILVL_B4_SB(src7, src6, src8, src7, src9, src8, src10, src9, src76_l,
1525 src87_l, src98_l, src109_l);
1526 out0_r = FILT_8TAP_DPADD_S_H(src10_r, src32_r, src54_r, src76_r,
1527 filt0, filt1, filt2, filt3);
1528 out1_r = FILT_8TAP_DPADD_S_H(src21_r, src43_r, src65_r, src87_r,
1529 filt0, filt1, filt2, filt3);
1530 out2_r = FILT_8TAP_DPADD_S_H(src32_r, src54_r, src76_r, src98_r,
1531 filt0, filt1, filt2, filt3);
1532 out3_r = FILT_8TAP_DPADD_S_H(src43_r, src65_r, src87_r, src109_r,
1533 filt0, filt1, filt2, filt3);
1534 out0_l = FILT_8TAP_DPADD_S_H(src10_l, src32_l, src54_l, src76_l,
1535 filt0, filt1, filt2, filt3);
1536 out1_l = FILT_8TAP_DPADD_S_H(src21_l, src43_l, src65_l, src87_l,
1537 filt0, filt1, filt2, filt3);
1538 out2_l = FILT_8TAP_DPADD_S_H(src32_l, src54_l, src76_l, src98_l,
1539 filt0, filt1, filt2, filt3);
1540 out3_l = FILT_8TAP_DPADD_S_H(src43_l, src65_l, src87_l, src109_l,
1541 filt0, filt1, filt2, filt3);
1542 SRARI_H4_SH(out0_r, out1_r, out2_r, out3_r, 7);
1543 SRARI_H4_SH(out0_l, out1_l, out2_l, out3_l, 7);
1544 SAT_SH4_SH(out0_r, out1_r, out2_r, out3_r, 7);
1545 SAT_SH4_SH(out0_l, out1_l, out2_l, out3_l, 7);
1546 PCKEV_B4_UB(out0_l, out0_r, out1_l, out1_r, out2_l, out2_r, out3_l,
1547 out3_r, tmp0, tmp1, tmp2, tmp3);
1548 XORI_B4_128_UB(tmp0, tmp1, tmp2, tmp3);
1549 AVER_UB4_UB(tmp0, dst0, tmp1, dst1, tmp2, dst2, tmp3, dst3,
1550 dst0, dst1, dst2, dst3);
1551 ST_UB4(dst0, dst1, dst2, dst3, dst_tmp, dst_stride);
1552 dst_tmp += (4 * dst_stride);
1574 static void common_vt_8t_and_aver_dst_16w_msa(const uint8_t *src,
1576 uint8_t *dst, int32_t dst_stride,
1577 const int8_t *filter,
1580 common_vt_8t_and_aver_dst_16w_mult_msa(src, src_stride, dst, dst_stride,
1581 filter, height, 16);
1584 static void common_vt_8t_and_aver_dst_32w_msa(const uint8_t *src,
1586 uint8_t *dst, int32_t dst_stride,
1587 const int8_t *filter,
1590 common_vt_8t_and_aver_dst_16w_mult_msa(src, src_stride, dst, dst_stride,
1591 filter, height, 32);
1594 static void common_vt_8t_and_aver_dst_64w_msa(const uint8_t *src,
1596 uint8_t *dst, int32_t dst_stride,
1597 const int8_t *filter,
1600 common_vt_8t_and_aver_dst_16w_mult_msa(src, src_stride, dst, dst_stride,
1601 filter, height, 64);
1604 static void common_hv_8ht_8vt_and_aver_dst_4w_msa(const uint8_t *src,
1608 const int8_t *filter_horiz,
1609 const int8_t *filter_vert,
1613 v16i8 src0, src1, src2, src3, src4, src5, src6, src7, src8, src9, src10;
1614 v16u8 dst0, dst1, dst2, dst3, mask0, mask1, mask2, mask3, tmp0, tmp1;
1615 v16i8 filt_hz0, filt_hz1, filt_hz2, filt_hz3;
1616 v8i16 hz_out0, hz_out1, hz_out2, hz_out3, hz_out4, hz_out5, hz_out6;
1617 v8i16 hz_out7, hz_out8, hz_out9, res0, res1, vec0, vec1, vec2, vec3, vec4;
1618 v8i16 filt, filt_vt0, filt_vt1, filt_vt2, filt_vt3;
1620 mask0 = LD_UB(&mc_filt_mask_arr[16]);
1621 src -= (3 + 3 * src_stride);
1623 /* rearranging filter */
1624 filt = LD_SH(filter_horiz);
1625 SPLATI_H4_SB(filt, 0, 1, 2, 3, filt_hz0, filt_hz1, filt_hz2, filt_hz3);
1631 LD_SB7(src, src_stride, src0, src1, src2, src3, src4, src5, src6);
1632 XORI_B7_128_SB(src0, src1, src2, src3, src4, src5, src6);
1633 src += (7 * src_stride);
1635 hz_out0 = HORIZ_8TAP_FILT(src0, src1, mask0, mask1, mask2, mask3, filt_hz0,
1636 filt_hz1, filt_hz2, filt_hz3);
1637 hz_out2 = HORIZ_8TAP_FILT(src2, src3, mask0, mask1, mask2, mask3, filt_hz0,
1638 filt_hz1, filt_hz2, filt_hz3);
1639 hz_out4 = HORIZ_8TAP_FILT(src4, src5, mask0, mask1, mask2, mask3, filt_hz0,
1640 filt_hz1, filt_hz2, filt_hz3);
1641 hz_out5 = HORIZ_8TAP_FILT(src5, src6, mask0, mask1, mask2, mask3, filt_hz0,
1642 filt_hz1, filt_hz2, filt_hz3);
1643 SLDI_B2_SH(hz_out2, hz_out4, hz_out0, hz_out2, hz_out1, hz_out3, 8);
1645 filt = LD_SH(filter_vert);
1646 SPLATI_H4_SH(filt, 0, 1, 2, 3, filt_vt0, filt_vt1, filt_vt2, filt_vt3);
1648 ILVEV_B2_SH(hz_out0, hz_out1, hz_out2, hz_out3, vec0, vec1);
1649 vec2 = (v8i16) __msa_ilvev_b((v16i8) hz_out5, (v16i8) hz_out4);
1651 for (loop_cnt = (height >> 2); loop_cnt--;) {
1652 LD_SB4(src, src_stride, src7, src8, src9, src10);
1653 XORI_B4_128_SB(src7, src8, src9, src10);
1654 src += (4 * src_stride);
1656 LD_UB4(dst, dst_stride, dst0, dst1, dst2, dst3);
1657 hz_out7 = HORIZ_8TAP_FILT(src7, src8, mask0, mask1, mask2, mask3,
1658 filt_hz0, filt_hz1, filt_hz2, filt_hz3);
1659 hz_out6 = (v8i16) __msa_sldi_b((v16i8) hz_out7, (v16i8) hz_out5, 8);
1660 vec3 = (v8i16) __msa_ilvev_b((v16i8) hz_out7, (v16i8) hz_out6);
1661 res0 = FILT_8TAP_DPADD_S_H(vec0, vec1, vec2, vec3, filt_vt0, filt_vt1,
1662 filt_vt2, filt_vt3);
1664 hz_out9 = HORIZ_8TAP_FILT(src9, src10, mask0, mask1, mask2, mask3,
1665 filt_hz0, filt_hz1, filt_hz2, filt_hz3);
1666 hz_out8 = (v8i16) __msa_sldi_b((v16i8) hz_out9, (v16i8) hz_out7, 8);
1667 vec4 = (v8i16) __msa_ilvev_b((v16i8) hz_out9, (v16i8) hz_out8);
1668 res1 = FILT_8TAP_DPADD_S_H(vec1, vec2, vec3, vec4, filt_vt0, filt_vt1,
1669 filt_vt2, filt_vt3);
1670 ILVR_W2_UB(dst1, dst0, dst3, dst2, dst0, dst2);
1672 SRARI_H2_SH(res0, res1, 7);
1673 SAT_SH2_SH(res0, res1, 7);
1674 PCKEV_B2_UB(res0, res0, res1, res1, tmp0, tmp1);
1675 XORI_B2_128_UB(tmp0, tmp1);
1676 AVER_UB2_UB(tmp0, dst0, tmp1, dst2, tmp0, tmp1);
1677 ST4x4_UB(tmp0, tmp1, 0, 1, 0, 1, dst, dst_stride);
1678 dst += (4 * dst_stride);
1687 static void common_hv_8ht_8vt_and_aver_dst_8w_msa(const uint8_t *src,
1691 const int8_t *filter_horiz,
1692 const int8_t *filter_vert,
1696 v16i8 src0, src1, src2, src3, src4, src5, src6, src7, src8, src9, src10;
1697 v16i8 filt_hz0, filt_hz1, filt_hz2, filt_hz3;
1698 v8i16 filt, filt_vt0, filt_vt1, filt_vt2, filt_vt3;
1699 v16u8 dst0, dst1, dst2, dst3, mask0, mask1, mask2, mask3;
1700 v8i16 hz_out0, hz_out1, hz_out2, hz_out3, hz_out4, hz_out5, hz_out6;
1701 v8i16 hz_out7, hz_out8, hz_out9, hz_out10, tmp0, tmp1, tmp2, tmp3;
1702 v8i16 out0, out1, out2, out3, out4, out5, out6, out7, out8, out9;
1704 mask0 = LD_UB(&mc_filt_mask_arr[0]);
1705 src -= (3 + 3 * src_stride);
1707 /* rearranging filter */
1708 filt = LD_SH(filter_horiz);
1709 SPLATI_H4_SB(filt, 0, 1, 2, 3, filt_hz0, filt_hz1, filt_hz2, filt_hz3);
1715 LD_SB7(src, src_stride, src0, src1, src2, src3, src4, src5, src6);
1716 src += (7 * src_stride);
1718 XORI_B7_128_SB(src0, src1, src2, src3, src4, src5, src6);
1719 hz_out0 = HORIZ_8TAP_FILT(src0, src0, mask0, mask1, mask2, mask3, filt_hz0,
1720 filt_hz1, filt_hz2, filt_hz3);
1721 hz_out1 = HORIZ_8TAP_FILT(src1, src1, mask0, mask1, mask2, mask3, filt_hz0,
1722 filt_hz1, filt_hz2, filt_hz3);
1723 hz_out2 = HORIZ_8TAP_FILT(src2, src2, mask0, mask1, mask2, mask3, filt_hz0,
1724 filt_hz1, filt_hz2, filt_hz3);
1725 hz_out3 = HORIZ_8TAP_FILT(src3, src3, mask0, mask1, mask2, mask3, filt_hz0,
1726 filt_hz1, filt_hz2, filt_hz3);
1727 hz_out4 = HORIZ_8TAP_FILT(src4, src4, mask0, mask1, mask2, mask3, filt_hz0,
1728 filt_hz1, filt_hz2, filt_hz3);
1729 hz_out5 = HORIZ_8TAP_FILT(src5, src5, mask0, mask1, mask2, mask3, filt_hz0,
1730 filt_hz1, filt_hz2, filt_hz3);
1731 hz_out6 = HORIZ_8TAP_FILT(src6, src6, mask0, mask1, mask2, mask3, filt_hz0,
1732 filt_hz1, filt_hz2, filt_hz3);
1734 filt = LD_SH(filter_vert);
1735 SPLATI_H4_SH(filt, 0, 1, 2, 3, filt_vt0, filt_vt1, filt_vt2, filt_vt3);
1737 ILVEV_B2_SH(hz_out0, hz_out1, hz_out2, hz_out3, out0, out1);
1738 ILVEV_B2_SH(hz_out4, hz_out5, hz_out1, hz_out2, out2, out4);
1739 ILVEV_B2_SH(hz_out3, hz_out4, hz_out5, hz_out6, out5, out6);
1741 for (loop_cnt = (height >> 2); loop_cnt--;) {
1742 LD_SB4(src, src_stride, src7, src8, src9, src10);
1743 XORI_B4_128_SB(src7, src8, src9, src10);
1744 src += (4 * src_stride);
1746 LD_UB4(dst, dst_stride, dst0, dst1, dst2, dst3);
1748 hz_out7 = HORIZ_8TAP_FILT(src7, src7, mask0, mask1, mask2, mask3,
1749 filt_hz0, filt_hz1, filt_hz2, filt_hz3);
1750 out3 = (v8i16) __msa_ilvev_b((v16i8) hz_out7, (v16i8) hz_out6);
1751 tmp0 = FILT_8TAP_DPADD_S_H(out0, out1, out2, out3, filt_vt0, filt_vt1,
1752 filt_vt2, filt_vt3);
1754 hz_out8 = HORIZ_8TAP_FILT(src8, src8, mask0, mask1, mask2, mask3,
1755 filt_hz0, filt_hz1, filt_hz2, filt_hz3);
1756 out7 = (v8i16) __msa_ilvev_b((v16i8) hz_out8, (v16i8) hz_out7);
1757 tmp1 = FILT_8TAP_DPADD_S_H(out4, out5, out6, out7, filt_vt0, filt_vt1,
1758 filt_vt2, filt_vt3);
1760 hz_out9 = HORIZ_8TAP_FILT(src9, src9, mask0, mask1, mask2, mask3,
1761 filt_hz0, filt_hz1, filt_hz2, filt_hz3);
1762 out8 = (v8i16) __msa_ilvev_b((v16i8) hz_out9, (v16i8) hz_out8);
1763 tmp2 = FILT_8TAP_DPADD_S_H(out1, out2, out3, out8, filt_vt0, filt_vt1,
1764 filt_vt2, filt_vt3);
1766 hz_out10 = HORIZ_8TAP_FILT(src10, src10, mask0, mask1, mask2, mask3,
1767 filt_hz0, filt_hz1, filt_hz2, filt_hz3);
1768 out9 = (v8i16) __msa_ilvev_b((v16i8) hz_out10, (v16i8) hz_out9);
1769 tmp3 = FILT_8TAP_DPADD_S_H(out5, out6, out7, out9, filt_vt0, filt_vt1,
1770 filt_vt2, filt_vt3);
1772 SRARI_H4_SH(tmp0, tmp1, tmp2, tmp3, 7);
1773 SAT_SH4_SH(tmp0, tmp1, tmp2, tmp3, 7);
1774 CONVERT_UB_AVG_ST8x4_UB(tmp0, tmp1, tmp2, tmp3, dst0, dst1, dst2, dst3,
1776 dst += (4 * dst_stride);
1788 static void common_hv_8ht_8vt_and_aver_dst_16w_msa(const uint8_t *src,
1792 const int8_t *filter_horiz,
1793 const int8_t *filter_vert,
1796 int32_t multiple8_cnt;
1798 for (multiple8_cnt = 2; multiple8_cnt--;) {
1799 common_hv_8ht_8vt_and_aver_dst_8w_msa(src, src_stride, dst, dst_stride,
1800 filter_horiz, filter_vert,
1808 static void common_hv_8ht_8vt_and_aver_dst_32w_msa(const uint8_t *src,
1812 const int8_t *filter_horiz,
1813 const int8_t *filter_vert,
1816 int32_t multiple8_cnt;
1818 for (multiple8_cnt = 4; multiple8_cnt--;) {
1819 common_hv_8ht_8vt_and_aver_dst_8w_msa(src, src_stride, dst, dst_stride,
1820 filter_horiz, filter_vert,
1828 static void common_hv_8ht_8vt_and_aver_dst_64w_msa(const uint8_t *src,
1832 const int8_t *filter_horiz,
1833 const int8_t *filter_vert,
1836 int32_t multiple8_cnt;
1838 for (multiple8_cnt = 8; multiple8_cnt--;) {
1839 common_hv_8ht_8vt_and_aver_dst_8w_msa(src, src_stride, dst, dst_stride,
1840 filter_horiz, filter_vert,
1848 static void common_hz_2t_4x4_msa(const uint8_t *src, int32_t src_stride,
1849 uint8_t *dst, int32_t dst_stride,
1850 const int8_t *filter)
1852 v16i8 src0, src1, src2, src3, mask;
1853 v16u8 filt0, vec0, vec1, res0, res1;
1854 v8u16 vec2, vec3, filt;
1856 mask = LD_SB(&mc_filt_mask_arr[16]);
1858 /* rearranging filter */
1859 filt = LD_UH(filter);
1860 filt0 = (v16u8) __msa_splati_h((v8i16) filt, 0);
1862 LD_SB4(src, src_stride, src0, src1, src2, src3);
1863 VSHF_B2_UB(src0, src1, src2, src3, mask, mask, vec0, vec1);
1864 DOTP_UB2_UH(vec0, vec1, filt0, filt0, vec2, vec3);
1865 SRARI_H2_UH(vec2, vec3, 7);
1866 PCKEV_B2_UB(vec2, vec2, vec3, vec3, res0, res1);
1867 ST4x4_UB(res0, res1, 0, 1, 0, 1, dst, dst_stride);
1870 static void common_hz_2t_4x8_msa(const uint8_t *src, int32_t src_stride,
1871 uint8_t *dst, int32_t dst_stride,
1872 const int8_t *filter)
1874 v16u8 vec0, vec1, vec2, vec3, filt0;
1875 v16i8 src0, src1, src2, src3, src4, src5, src6, src7, mask;
1876 v16i8 res0, res1, res2, res3;
1877 v8u16 vec4, vec5, vec6, vec7, filt;
1879 mask = LD_SB(&mc_filt_mask_arr[16]);
1881 /* rearranging filter */
1882 filt = LD_UH(filter);
1883 filt0 = (v16u8) __msa_splati_h((v8i16) filt, 0);
1885 LD_SB8(src, src_stride, src0, src1, src2, src3, src4, src5, src6, src7);
1886 VSHF_B2_UB(src0, src1, src2, src3, mask, mask, vec0, vec1);
1887 VSHF_B2_UB(src4, src5, src6, src7, mask, mask, vec2, vec3);
1888 DOTP_UB4_UH(vec0, vec1, vec2, vec3, filt0, filt0, filt0, filt0,
1889 vec4, vec5, vec6, vec7);
1890 SRARI_H4_UH(vec4, vec5, vec6, vec7, 7);
1891 PCKEV_B4_SB(vec4, vec4, vec5, vec5, vec6, vec6, vec7, vec7,
1892 res0, res1, res2, res3);
1893 ST4x4_UB(res0, res1, 0, 1, 0, 1, dst, dst_stride);
1894 dst += (4 * dst_stride);
1895 ST4x4_UB(res2, res3, 0, 1, 0, 1, dst, dst_stride);
1898 void ff_put_bilin_4h_msa(uint8_t *dst, ptrdiff_t dst_stride,
1899 const uint8_t *src, ptrdiff_t src_stride,
1900 int height, int mx, int my)
1902 const int8_t *filter = vp9_bilinear_filters_msa[mx - 1];
1905 common_hz_2t_4x4_msa(src, src_stride, dst, dst_stride, filter);
1906 } else if (8 == height) {
1907 common_hz_2t_4x8_msa(src, src_stride, dst, dst_stride, filter);
1911 static void common_hz_2t_8x4_msa(const uint8_t *src, int32_t src_stride,
1912 uint8_t *dst, int32_t dst_stride,
1913 const int8_t *filter)
1916 v16i8 src0, src1, src2, src3, mask;
1917 v8u16 vec0, vec1, vec2, vec3, filt;
1919 mask = LD_SB(&mc_filt_mask_arr[0]);
1921 /* rearranging filter */
1922 filt = LD_UH(filter);
1923 filt0 = (v16u8) __msa_splati_h((v8i16) filt, 0);
1925 LD_SB4(src, src_stride, src0, src1, src2, src3);
1926 VSHF_B2_UH(src0, src0, src1, src1, mask, mask, vec0, vec1);
1927 VSHF_B2_UH(src2, src2, src3, src3, mask, mask, vec2, vec3);
1928 DOTP_UB4_UH(vec0, vec1, vec2, vec3, filt0, filt0, filt0, filt0,
1929 vec0, vec1, vec2, vec3);
1930 SRARI_H4_UH(vec0, vec1, vec2, vec3, 7);
1931 PCKEV_B2_SB(vec1, vec0, vec3, vec2, src0, src1);
1932 ST8x4_UB(src0, src1, dst, dst_stride);
1935 static void common_hz_2t_8x8mult_msa(const uint8_t *src, int32_t src_stride,
1936 uint8_t *dst, int32_t dst_stride,
1937 const int8_t *filter, int32_t height)
1940 v16i8 src0, src1, src2, src3, mask, out0, out1;
1941 v8u16 vec0, vec1, vec2, vec3, filt;
1943 mask = LD_SB(&mc_filt_mask_arr[0]);
1945 /* rearranging filter */
1946 filt = LD_UH(filter);
1947 filt0 = (v16u8) __msa_splati_h((v8i16) filt, 0);
1949 LD_SB4(src, src_stride, src0, src1, src2, src3);
1950 src += (4 * src_stride);
1952 VSHF_B2_UH(src0, src0, src1, src1, mask, mask, vec0, vec1);
1953 VSHF_B2_UH(src2, src2, src3, src3, mask, mask, vec2, vec3);
1954 DOTP_UB4_UH(vec0, vec1, vec2, vec3, filt0, filt0, filt0, filt0,
1955 vec0, vec1, vec2, vec3);
1956 SRARI_H4_UH(vec0, vec1, vec2, vec3, 7);
1957 LD_SB4(src, src_stride, src0, src1, src2, src3);
1958 src += (4 * src_stride);
1960 PCKEV_B2_SB(vec1, vec0, vec3, vec2, out0, out1);
1961 ST8x4_UB(out0, out1, dst, dst_stride);
1962 dst += (4 * dst_stride);
1964 VSHF_B2_UH(src0, src0, src1, src1, mask, mask, vec0, vec1);
1965 VSHF_B2_UH(src2, src2, src3, src3, mask, mask, vec2, vec3);
1966 DOTP_UB4_UH(vec0, vec1, vec2, vec3, filt0, filt0, filt0, filt0,
1967 vec0, vec1, vec2, vec3);
1968 SRARI_H4_UH(vec0, vec1, vec2, vec3, 7);
1969 PCKEV_B2_SB(vec1, vec0, vec3, vec2, out0, out1);
1970 ST8x4_UB(out0, out1, dst, dst_stride);
1971 dst += (4 * dst_stride);
1974 LD_SB4(src, src_stride, src0, src1, src2, src3);
1975 src += (4 * src_stride);
1977 VSHF_B2_UH(src0, src0, src1, src1, mask, mask, vec0, vec1);
1978 VSHF_B2_UH(src2, src2, src3, src3, mask, mask, vec2, vec3);
1979 DOTP_UB4_UH(vec0, vec1, vec2, vec3, filt0, filt0, filt0, filt0,
1980 vec0, vec1, vec2, vec3);
1981 SRARI_H4_UH(vec0, vec1, vec2, vec3, 7);
1982 LD_SB4(src, src_stride, src0, src1, src2, src3);
1983 src += (4 * src_stride);
1985 PCKEV_B2_SB(vec1, vec0, vec3, vec2, out0, out1);
1986 ST8x4_UB(out0, out1, dst, dst_stride);
1988 VSHF_B2_UH(src0, src0, src1, src1, mask, mask, vec0, vec1);
1989 VSHF_B2_UH(src2, src2, src3, src3, mask, mask, vec2, vec3);
1990 DOTP_UB4_UH(vec0, vec1, vec2, vec3, filt0, filt0, filt0, filt0,
1991 vec0, vec1, vec2, vec3);
1992 SRARI_H4_UH(vec0, vec1, vec2, vec3, 7);
1993 PCKEV_B2_SB(vec1, vec0, vec3, vec2, out0, out1);
1994 ST8x4_UB(out0, out1, dst + 4 * dst_stride, dst_stride);
1998 void ff_put_bilin_8h_msa(uint8_t *dst, ptrdiff_t dst_stride,
1999 const uint8_t *src, ptrdiff_t src_stride,
2000 int height, int mx, int my)
2002 const int8_t *filter = vp9_bilinear_filters_msa[mx - 1];
2005 common_hz_2t_8x4_msa(src, src_stride, dst, dst_stride, filter);
2007 common_hz_2t_8x8mult_msa(src, src_stride, dst, dst_stride, filter,
2012 void ff_put_bilin_16h_msa(uint8_t *dst, ptrdiff_t dst_stride,
2013 const uint8_t *src, ptrdiff_t src_stride,
2014 int height, int mx, int my)
2017 const int8_t *filter = vp9_bilinear_filters_msa[mx - 1];
2018 v16i8 src0, src1, src2, src3, src4, src5, src6, src7, mask;
2019 v16u8 filt0, vec0, vec1, vec2, vec3, vec4, vec5, vec6, vec7;
2020 v8u16 out0, out1, out2, out3, out4, out5, out6, out7, filt;
2022 mask = LD_SB(&mc_filt_mask_arr[0]);
2024 loop_cnt = (height >> 2) - 1;
2026 /* rearranging filter */
2027 filt = LD_UH(filter);
2028 filt0 = (v16u8) __msa_splati_h((v8i16) filt, 0);
2030 LD_SB4(src, src_stride, src0, src2, src4, src6);
2031 LD_SB4(src + 8, src_stride, src1, src3, src5, src7);
2032 src += (4 * src_stride);
2034 VSHF_B2_UB(src0, src0, src1, src1, mask, mask, vec0, vec1);
2035 VSHF_B2_UB(src2, src2, src3, src3, mask, mask, vec2, vec3);
2036 VSHF_B2_UB(src4, src4, src5, src5, mask, mask, vec4, vec5);
2037 VSHF_B2_UB(src6, src6, src7, src7, mask, mask, vec6, vec7);
2038 DOTP_UB4_UH(vec0, vec1, vec2, vec3, filt0, filt0, filt0, filt0,
2039 out0, out1, out2, out3);
2040 DOTP_UB4_UH(vec4, vec5, vec6, vec7, filt0, filt0, filt0, filt0,
2041 out4, out5, out6, out7);
2042 SRARI_H4_UH(out0, out1, out2, out3, 7);
2043 SRARI_H4_UH(out4, out5, out6, out7, 7);
2044 PCKEV_ST_SB(out0, out1, dst);
2046 PCKEV_ST_SB(out2, out3, dst);
2048 PCKEV_ST_SB(out4, out5, dst);
2050 PCKEV_ST_SB(out6, out7, dst);
2053 for (; loop_cnt--;) {
2054 LD_SB4(src, src_stride, src0, src2, src4, src6);
2055 LD_SB4(src + 8, src_stride, src1, src3, src5, src7);
2056 src += (4 * src_stride);
2058 VSHF_B2_UB(src0, src0, src1, src1, mask, mask, vec0, vec1);
2059 VSHF_B2_UB(src2, src2, src3, src3, mask, mask, vec2, vec3);
2060 VSHF_B2_UB(src4, src4, src5, src5, mask, mask, vec4, vec5);
2061 VSHF_B2_UB(src6, src6, src7, src7, mask, mask, vec6, vec7);
2062 DOTP_UB4_UH(vec0, vec1, vec2, vec3, filt0, filt0, filt0, filt0,
2063 out0, out1, out2, out3);
2064 DOTP_UB4_UH(vec4, vec5, vec6, vec7, filt0, filt0, filt0, filt0,
2065 out4, out5, out6, out7);
2066 SRARI_H4_UH(out0, out1, out2, out3, 7);
2067 SRARI_H4_UH(out4, out5, out6, out7, 7);
2068 PCKEV_ST_SB(out0, out1, dst);
2070 PCKEV_ST_SB(out2, out3, dst);
2072 PCKEV_ST_SB(out4, out5, dst);
2074 PCKEV_ST_SB(out6, out7, dst);
2079 void ff_put_bilin_32h_msa(uint8_t *dst, ptrdiff_t dst_stride,
2080 const uint8_t *src, ptrdiff_t src_stride,
2081 int height, int mx, int my)
2084 const int8_t *filter = vp9_bilinear_filters_msa[mx - 1];
2085 v16i8 src0, src1, src2, src3, src4, src5, src6, src7, mask;
2086 v16u8 filt0, vec0, vec1, vec2, vec3, vec4, vec5, vec6, vec7;
2087 v8u16 out0, out1, out2, out3, out4, out5, out6, out7, filt;
2089 mask = LD_SB(&mc_filt_mask_arr[0]);
2091 /* rearranging filter */
2092 filt = LD_UH(filter);
2093 filt0 = (v16u8) __msa_splati_h((v8i16) filt, 0);
2095 for (loop_cnt = height >> 1; loop_cnt--;) {
2097 src2 = LD_SB(src + 16);
2098 src3 = LD_SB(src + 24);
2099 src1 = __msa_sldi_b(src2, src0, 8);
2102 src6 = LD_SB(src + 16);
2103 src7 = LD_SB(src + 24);
2104 src5 = __msa_sldi_b(src6, src4, 8);
2107 VSHF_B2_UB(src0, src0, src1, src1, mask, mask, vec0, vec1);
2108 VSHF_B2_UB(src2, src2, src3, src3, mask, mask, vec2, vec3);
2109 VSHF_B2_UB(src4, src4, src5, src5, mask, mask, vec4, vec5);
2110 VSHF_B2_UB(src6, src6, src7, src7, mask, mask, vec6, vec7);
2111 DOTP_UB4_UH(vec0, vec1, vec2, vec3, filt0, filt0, filt0, filt0,
2112 out0, out1, out2, out3);
2113 DOTP_UB4_UH(vec4, vec5, vec6, vec7, filt0, filt0, filt0, filt0,
2114 out4, out5, out6, out7);
2115 SRARI_H4_UH(out0, out1, out2, out3, 7);
2116 SRARI_H4_UH(out4, out5, out6, out7, 7);
2117 PCKEV_ST_SB(out0, out1, dst);
2118 PCKEV_ST_SB(out2, out3, dst + 16);
2120 PCKEV_ST_SB(out4, out5, dst);
2121 PCKEV_ST_SB(out6, out7, dst + 16);
2126 void ff_put_bilin_64h_msa(uint8_t *dst, ptrdiff_t dst_stride,
2127 const uint8_t *src, ptrdiff_t src_stride,
2128 int height, int mx, int my)
2131 const int8_t *filter = vp9_bilinear_filters_msa[mx - 1];
2132 v16i8 src0, src1, src2, src3, src4, src5, src6, src7, mask;
2133 v16u8 filt0, vec0, vec1, vec2, vec3, vec4, vec5, vec6, vec7;
2134 v8u16 out0, out1, out2, out3, out4, out5, out6, out7, filt;
2136 mask = LD_SB(&mc_filt_mask_arr[0]);
2138 /* rearranging filter */
2139 filt = LD_UH(filter);
2140 filt0 = (v16u8) __msa_splati_h((v8i16) filt, 0);
2142 for (loop_cnt = height; loop_cnt--;) {
2144 src2 = LD_SB(src + 16);
2145 src4 = LD_SB(src + 32);
2146 src6 = LD_SB(src + 48);
2147 src7 = LD_SB(src + 56);
2148 SLDI_B3_SB(src2, src4, src6, src0, src2, src4, src1, src3, src5, 8);
2151 VSHF_B2_UB(src0, src0, src1, src1, mask, mask, vec0, vec1);
2152 VSHF_B2_UB(src2, src2, src3, src3, mask, mask, vec2, vec3);
2153 VSHF_B2_UB(src4, src4, src5, src5, mask, mask, vec4, vec5);
2154 VSHF_B2_UB(src6, src6, src7, src7, mask, mask, vec6, vec7);
2155 DOTP_UB4_UH(vec0, vec1, vec2, vec3, filt0, filt0, filt0, filt0,
2156 out0, out1, out2, out3);
2157 DOTP_UB4_UH(vec4, vec5, vec6, vec7, filt0, filt0, filt0, filt0,
2158 out4, out5, out6, out7);
2159 SRARI_H4_UH(out0, out1, out2, out3, 7);
2160 SRARI_H4_UH(out4, out5, out6, out7, 7);
2161 PCKEV_ST_SB(out0, out1, dst);
2162 PCKEV_ST_SB(out2, out3, dst + 16);
2163 PCKEV_ST_SB(out4, out5, dst + 32);
2164 PCKEV_ST_SB(out6, out7, dst + 48);
2169 static void common_vt_2t_4x4_msa(const uint8_t *src, int32_t src_stride,
2170 uint8_t *dst, int32_t dst_stride,
2171 const int8_t *filter)
2173 v16i8 src0, src1, src2, src3, src4;
2174 v16i8 src10_r, src32_r, src21_r, src43_r, src2110, src4332;
2179 filt = LD_SH(filter);
2180 filt0 = (v16u8) __msa_splati_h(filt, 0);
2182 LD_SB5(src, src_stride, src0, src1, src2, src3, src4);
2183 src += (5 * src_stride);
2185 ILVR_B4_SB(src1, src0, src2, src1, src3, src2, src4, src3,
2186 src10_r, src21_r, src32_r, src43_r);
2187 ILVR_D2_SB(src21_r, src10_r, src43_r, src32_r, src2110, src4332);
2188 DOTP_UB2_UH(src2110, src4332, filt0, filt0, tmp0, tmp1);
2189 SRARI_H2_UH(tmp0, tmp1, 7);
2190 SAT_UH2_UH(tmp0, tmp1, 7);
2191 src2110 = __msa_pckev_b((v16i8) tmp1, (v16i8) tmp0);
2192 ST4x4_UB(src2110, src2110, 0, 1, 2, 3, dst, dst_stride);
2195 static void common_vt_2t_4x8_msa(const uint8_t *src, int32_t src_stride,
2196 uint8_t *dst, int32_t dst_stride,
2197 const int8_t *filter)
2199 v16i8 src0, src1, src2, src3, src4, src5, src6, src7, src8;
2200 v16i8 src10_r, src32_r, src54_r, src76_r, src21_r, src43_r;
2201 v16i8 src65_r, src87_r, src2110, src4332, src6554, src8776;
2202 v8u16 tmp0, tmp1, tmp2, tmp3;
2206 filt = LD_SH(filter);
2207 filt0 = (v16u8) __msa_splati_h(filt, 0);
2209 LD_SB8(src, src_stride, src0, src1, src2, src3, src4, src5, src6, src7);
2210 src += (8 * src_stride);
2215 ILVR_B4_SB(src1, src0, src2, src1, src3, src2, src4, src3, src10_r, src21_r,
2217 ILVR_B4_SB(src5, src4, src6, src5, src7, src6, src8, src7, src54_r, src65_r,
2219 ILVR_D4_SB(src21_r, src10_r, src43_r, src32_r, src65_r, src54_r,
2220 src87_r, src76_r, src2110, src4332, src6554, src8776);
2221 DOTP_UB4_UH(src2110, src4332, src6554, src8776, filt0, filt0, filt0, filt0,
2222 tmp0, tmp1, tmp2, tmp3);
2223 SRARI_H4_UH(tmp0, tmp1, tmp2, tmp3, 7);
2224 SAT_UH4_UH(tmp0, tmp1, tmp2, tmp3, 7);
2225 PCKEV_B2_SB(tmp1, tmp0, tmp3, tmp2, src2110, src4332);
2226 ST4x4_UB(src2110, src2110, 0, 1, 2, 3, dst, dst_stride);
2227 ST4x4_UB(src4332, src4332, 0, 1, 2, 3, dst + 4 * dst_stride, dst_stride);
2230 void ff_put_bilin_4v_msa(uint8_t *dst, ptrdiff_t dst_stride,
2231 const uint8_t *src, ptrdiff_t src_stride,
2232 int height, int mx, int my)
2234 const int8_t *filter = vp9_bilinear_filters_msa[my - 1];
2237 common_vt_2t_4x4_msa(src, src_stride, dst, dst_stride, filter);
2238 } else if (8 == height) {
2239 common_vt_2t_4x8_msa(src, src_stride, dst, dst_stride, filter);
2243 static void common_vt_2t_8x4_msa(const uint8_t *src, int32_t src_stride,
2244 uint8_t *dst, int32_t dst_stride,
2245 const int8_t *filter)
2247 v16u8 src0, src1, src2, src3, src4, vec0, vec1, vec2, vec3, filt0;
2249 v8u16 tmp0, tmp1, tmp2, tmp3;
2252 /* rearranging filter_y */
2253 filt = LD_SH(filter);
2254 filt0 = (v16u8) __msa_splati_h(filt, 0);
2256 LD_UB5(src, src_stride, src0, src1, src2, src3, src4);
2257 ILVR_B2_UB(src1, src0, src2, src1, vec0, vec1);
2258 ILVR_B2_UB(src3, src2, src4, src3, vec2, vec3);
2259 DOTP_UB4_UH(vec0, vec1, vec2, vec3, filt0, filt0, filt0, filt0,
2260 tmp0, tmp1, tmp2, tmp3);
2261 SRARI_H4_UH(tmp0, tmp1, tmp2, tmp3, 7);
2262 SAT_UH4_UH(tmp0, tmp1, tmp2, tmp3, 7);
2263 PCKEV_B2_SB(tmp1, tmp0, tmp3, tmp2, out0, out1);
2264 ST8x4_UB(out0, out1, dst, dst_stride);
2267 static void common_vt_2t_8x8mult_msa(const uint8_t *src, int32_t src_stride,
2268 uint8_t *dst, int32_t dst_stride,
2269 const int8_t *filter, int32_t height)
2272 v16u8 src0, src1, src2, src3, src4, src5, src6, src7, src8;
2273 v16u8 vec0, vec1, vec2, vec3, vec4, vec5, vec6, vec7, filt0;
2275 v8u16 tmp0, tmp1, tmp2, tmp3;
2278 /* rearranging filter_y */
2279 filt = LD_SH(filter);
2280 filt0 = (v16u8) __msa_splati_h(filt, 0);
2285 for (loop_cnt = (height >> 3); loop_cnt--;) {
2286 LD_UB8(src, src_stride, src1, src2, src3, src4, src5, src6, src7, src8);
2287 src += (8 * src_stride);
2289 ILVR_B4_UB(src1, src0, src2, src1, src3, src2, src4, src3,
2290 vec0, vec1, vec2, vec3);
2291 ILVR_B4_UB(src5, src4, src6, src5, src7, src6, src8, src7,
2292 vec4, vec5, vec6, vec7);
2293 DOTP_UB4_UH(vec0, vec1, vec2, vec3, filt0, filt0, filt0, filt0,
2294 tmp0, tmp1, tmp2, tmp3);
2295 SRARI_H4_UH(tmp0, tmp1, tmp2, tmp3, 7);
2296 SAT_UH4_UH(tmp0, tmp1, tmp2, tmp3, 7);
2297 PCKEV_B2_SB(tmp1, tmp0, tmp3, tmp2, out0, out1);
2298 ST8x4_UB(out0, out1, dst, dst_stride);
2299 dst += (4 * dst_stride);
2301 DOTP_UB4_UH(vec4, vec5, vec6, vec7, filt0, filt0, filt0, filt0,
2302 tmp0, tmp1, tmp2, tmp3);
2303 SRARI_H4_UH(tmp0, tmp1, tmp2, tmp3, 7);
2304 SAT_UH4_UH(tmp0, tmp1, tmp2, tmp3, 7);
2305 PCKEV_B2_SB(tmp1, tmp0, tmp3, tmp2, out0, out1);
2306 ST8x4_UB(out0, out1, dst, dst_stride);
2307 dst += (4 * dst_stride);
2313 void ff_put_bilin_8v_msa(uint8_t *dst, ptrdiff_t dst_stride,
2314 const uint8_t *src, ptrdiff_t src_stride,
2315 int height, int mx, int my)
2317 const int8_t *filter = vp9_bilinear_filters_msa[my - 1];
2320 common_vt_2t_8x4_msa(src, src_stride, dst, dst_stride, filter);
2322 common_vt_2t_8x8mult_msa(src, src_stride, dst, dst_stride, filter,
2327 void ff_put_bilin_16v_msa(uint8_t *dst, ptrdiff_t dst_stride,
2328 const uint8_t *src, ptrdiff_t src_stride,
2329 int height, int mx, int my)
2332 const int8_t *filter = vp9_bilinear_filters_msa[my - 1];
2333 v16u8 src0, src1, src2, src3, src4;
2334 v16u8 vec0, vec1, vec2, vec3, vec4, vec5, vec6, vec7, filt0;
2335 v8u16 tmp0, tmp1, tmp2, tmp3;
2338 /* rearranging filter_y */
2339 filt = LD_SH(filter);
2340 filt0 = (v16u8) __msa_splati_h(filt, 0);
2345 for (loop_cnt = (height >> 2); loop_cnt--;) {
2346 LD_UB4(src, src_stride, src1, src2, src3, src4);
2347 src += (4 * src_stride);
2349 ILVR_B2_UB(src1, src0, src2, src1, vec0, vec2);
2350 ILVL_B2_UB(src1, src0, src2, src1, vec1, vec3);
2351 DOTP_UB2_UH(vec0, vec1, filt0, filt0, tmp0, tmp1);
2352 SRARI_H2_UH(tmp0, tmp1, 7);
2353 SAT_UH2_UH(tmp0, tmp1, 7);
2354 PCKEV_ST_SB(tmp0, tmp1, dst);
2357 ILVR_B2_UB(src3, src2, src4, src3, vec4, vec6);
2358 ILVL_B2_UB(src3, src2, src4, src3, vec5, vec7);
2359 DOTP_UB2_UH(vec2, vec3, filt0, filt0, tmp2, tmp3);
2360 SRARI_H2_UH(tmp2, tmp3, 7);
2361 SAT_UH2_UH(tmp2, tmp3, 7);
2362 PCKEV_ST_SB(tmp2, tmp3, dst);
2365 DOTP_UB2_UH(vec4, vec5, filt0, filt0, tmp0, tmp1);
2366 SRARI_H2_UH(tmp0, tmp1, 7);
2367 SAT_UH2_UH(tmp0, tmp1, 7);
2368 PCKEV_ST_SB(tmp0, tmp1, dst);
2371 DOTP_UB2_UH(vec6, vec7, filt0, filt0, tmp2, tmp3);
2372 SRARI_H2_UH(tmp2, tmp3, 7);
2373 SAT_UH2_UH(tmp2, tmp3, 7);
2374 PCKEV_ST_SB(tmp2, tmp3, dst);
2381 void ff_put_bilin_32v_msa(uint8_t *dst, ptrdiff_t dst_stride,
2382 const uint8_t *src, ptrdiff_t src_stride,
2383 int height, int mx, int my)
2386 const int8_t *filter = vp9_bilinear_filters_msa[my - 1];
2387 v16u8 src0, src1, src2, src3, src4, src5, src6, src7, src8, src9;
2388 v16u8 vec0, vec1, vec2, vec3, vec4, vec5, vec6, vec7, filt0;
2389 v8u16 tmp0, tmp1, tmp2, tmp3;
2392 /* rearranging filter_y */
2393 filt = LD_SH(filter);
2394 filt0 = (v16u8) __msa_splati_h(filt, 0);
2397 src5 = LD_UB(src + 16);
2400 for (loop_cnt = (height >> 2); loop_cnt--;) {
2401 LD_UB4(src, src_stride, src1, src2, src3, src4);
2402 ILVR_B2_UB(src1, src0, src2, src1, vec0, vec2);
2403 ILVL_B2_UB(src1, src0, src2, src1, vec1, vec3);
2405 LD_UB4(src + 16, src_stride, src6, src7, src8, src9);
2406 src += (4 * src_stride);
2408 DOTP_UB2_UH(vec0, vec1, filt0, filt0, tmp0, tmp1);
2409 SRARI_H2_UH(tmp0, tmp1, 7);
2410 SAT_UH2_UH(tmp0, tmp1, 7);
2411 PCKEV_ST_SB(tmp0, tmp1, dst);
2412 DOTP_UB2_UH(vec2, vec3, filt0, filt0, tmp2, tmp3);
2413 SRARI_H2_UH(tmp2, tmp3, 7);
2414 SAT_UH2_UH(tmp2, tmp3, 7);
2415 PCKEV_ST_SB(tmp2, tmp3, dst + dst_stride);
2417 ILVR_B2_UB(src3, src2, src4, src3, vec4, vec6);
2418 ILVL_B2_UB(src3, src2, src4, src3, vec5, vec7);
2419 DOTP_UB2_UH(vec4, vec5, filt0, filt0, tmp0, tmp1);
2420 SRARI_H2_UH(tmp0, tmp1, 7);
2421 SAT_UH2_UH(tmp0, tmp1, 7);
2422 PCKEV_ST_SB(tmp0, tmp1, dst + 2 * dst_stride);
2424 DOTP_UB2_UH(vec6, vec7, filt0, filt0, tmp2, tmp3);
2425 SRARI_H2_UH(tmp2, tmp3, 7);
2426 SAT_UH2_UH(tmp2, tmp3, 7);
2427 PCKEV_ST_SB(tmp2, tmp3, dst + 3 * dst_stride);
2429 ILVR_B2_UB(src6, src5, src7, src6, vec0, vec2);
2430 ILVL_B2_UB(src6, src5, src7, src6, vec1, vec3);
2431 DOTP_UB2_UH(vec0, vec1, filt0, filt0, tmp0, tmp1);
2432 SRARI_H2_UH(tmp0, tmp1, 7);
2433 SAT_UH2_UH(tmp0, tmp1, 7);
2434 PCKEV_ST_SB(tmp0, tmp1, dst + 16);
2436 DOTP_UB2_UH(vec2, vec3, filt0, filt0, tmp2, tmp3);
2437 SRARI_H2_UH(tmp2, tmp3, 7);
2438 SAT_UH2_UH(tmp2, tmp3, 7);
2439 PCKEV_ST_SB(tmp2, tmp3, dst + 16 + dst_stride);
2441 ILVR_B2_UB(src8, src7, src9, src8, vec4, vec6);
2442 ILVL_B2_UB(src8, src7, src9, src8, vec5, vec7);
2443 DOTP_UB2_UH(vec4, vec5, filt0, filt0, tmp0, tmp1);
2444 SRARI_H2_UH(tmp0, tmp1, 7);
2445 SAT_UH2_UH(tmp0, tmp1, 7);
2446 PCKEV_ST_SB(tmp0, tmp1, dst + 16 + 2 * dst_stride);
2448 DOTP_UB2_UH(vec6, vec7, filt0, filt0, tmp2, tmp3);
2449 SRARI_H2_UH(tmp2, tmp3, 7);
2450 SAT_UH2_UH(tmp2, tmp3, 7);
2451 PCKEV_ST_SB(tmp2, tmp3, dst + 16 + 3 * dst_stride);
2452 dst += (4 * dst_stride);
2459 void ff_put_bilin_64v_msa(uint8_t *dst, ptrdiff_t dst_stride,
2460 const uint8_t *src, ptrdiff_t src_stride,
2461 int height, int mx, int my)
2464 const int8_t *filter = vp9_bilinear_filters_msa[my - 1];
2465 v16u8 src0, src1, src2, src3, src4, src5, src6, src7, src8, src9, src10;
2466 v16u8 src11, vec0, vec1, vec2, vec3, vec4, vec5, vec6, vec7, filt0;
2467 v8u16 tmp0, tmp1, tmp2, tmp3, tmp4, tmp5, tmp6, tmp7;
2470 /* rearranging filter_y */
2471 filt = LD_SH(filter);
2472 filt0 = (v16u8) __msa_splati_h(filt, 0);
2474 LD_UB4(src, 16, src0, src3, src6, src9);
2477 for (loop_cnt = (height >> 1); loop_cnt--;) {
2478 LD_UB2(src, src_stride, src1, src2);
2479 LD_UB2(src + 16, src_stride, src4, src5);
2480 LD_UB2(src + 32, src_stride, src7, src8);
2481 LD_UB2(src + 48, src_stride, src10, src11);
2482 src += (2 * src_stride);
2484 ILVR_B2_UB(src1, src0, src2, src1, vec0, vec2);
2485 ILVL_B2_UB(src1, src0, src2, src1, vec1, vec3);
2486 DOTP_UB2_UH(vec0, vec1, filt0, filt0, tmp0, tmp1);
2487 SRARI_H2_UH(tmp0, tmp1, 7);
2488 SAT_UH2_UH(tmp0, tmp1, 7);
2489 PCKEV_ST_SB(tmp0, tmp1, dst);
2491 DOTP_UB2_UH(vec2, vec3, filt0, filt0, tmp2, tmp3);
2492 SRARI_H2_UH(tmp2, tmp3, 7);
2493 SAT_UH2_UH(tmp2, tmp3, 7);
2494 PCKEV_ST_SB(tmp2, tmp3, dst + dst_stride);
2496 ILVR_B2_UB(src4, src3, src5, src4, vec4, vec6);
2497 ILVL_B2_UB(src4, src3, src5, src4, vec5, vec7);
2498 DOTP_UB2_UH(vec4, vec5, filt0, filt0, tmp4, tmp5);
2499 SRARI_H2_UH(tmp4, tmp5, 7);
2500 SAT_UH2_UH(tmp4, tmp5, 7);
2501 PCKEV_ST_SB(tmp4, tmp5, dst + 16);
2503 DOTP_UB2_UH(vec6, vec7, filt0, filt0, tmp6, tmp7);
2504 SRARI_H2_UH(tmp6, tmp7, 7);
2505 SAT_UH2_UH(tmp6, tmp7, 7);
2506 PCKEV_ST_SB(tmp6, tmp7, dst + 16 + dst_stride);
2508 ILVR_B2_UB(src7, src6, src8, src7, vec0, vec2);
2509 ILVL_B2_UB(src7, src6, src8, src7, vec1, vec3);
2510 DOTP_UB2_UH(vec0, vec1, filt0, filt0, tmp0, tmp1);
2511 SRARI_H2_UH(tmp0, tmp1, 7);
2512 SAT_UH2_UH(tmp0, tmp1, 7);
2513 PCKEV_ST_SB(tmp0, tmp1, dst + 32);
2515 DOTP_UB2_UH(vec2, vec3, filt0, filt0, tmp2, tmp3);
2516 SRARI_H2_UH(tmp2, tmp3, 7);
2517 SAT_UH2_UH(tmp2, tmp3, 7);
2518 PCKEV_ST_SB(tmp2, tmp3, dst + 32 + dst_stride);
2520 ILVR_B2_UB(src10, src9, src11, src10, vec4, vec6);
2521 ILVL_B2_UB(src10, src9, src11, src10, vec5, vec7);
2522 DOTP_UB2_UH(vec4, vec5, filt0, filt0, tmp4, tmp5);
2523 SRARI_H2_UH(tmp4, tmp5, 7);
2524 SAT_UH2_UH(tmp4, tmp5, 7);
2525 PCKEV_ST_SB(tmp4, tmp5, dst + 48);
2527 DOTP_UB2_UH(vec6, vec7, filt0, filt0, tmp6, tmp7);
2528 SRARI_H2_UH(tmp6, tmp7, 7);
2529 SAT_UH2_UH(tmp6, tmp7, 7);
2530 PCKEV_ST_SB(tmp6, tmp7, dst + 48 + dst_stride);
2531 dst += (2 * dst_stride);
2540 static void common_hv_2ht_2vt_4x4_msa(const uint8_t *src, int32_t src_stride,
2541 uint8_t *dst, int32_t dst_stride,
2542 const int8_t *filter_horiz, const int8_t *filter_vert)
2544 v16i8 src0, src1, src2, src3, src4, mask;
2545 v16u8 filt_vt, filt_hz, vec0, vec1, res0, res1;
2546 v8u16 hz_out0, hz_out1, hz_out2, hz_out3, hz_out4, filt, tmp0, tmp1;
2548 mask = LD_SB(&mc_filt_mask_arr[16]);
2550 /* rearranging filter */
2551 filt = LD_UH(filter_horiz);
2552 filt_hz = (v16u8) __msa_splati_h((v8i16) filt, 0);
2554 filt = LD_UH(filter_vert);
2555 filt_vt = (v16u8) __msa_splati_h((v8i16) filt, 0);
2557 LD_SB5(src, src_stride, src0, src1, src2, src3, src4);
2558 hz_out0 = HORIZ_2TAP_FILT_UH(src0, src1, mask, filt_hz, 7);
2559 hz_out2 = HORIZ_2TAP_FILT_UH(src2, src3, mask, filt_hz, 7);
2560 hz_out4 = HORIZ_2TAP_FILT_UH(src4, src4, mask, filt_hz, 7);
2561 hz_out1 = (v8u16) __msa_sldi_b((v16i8) hz_out2, (v16i8) hz_out0, 8);
2562 hz_out3 = (v8u16) __msa_pckod_d((v2i64) hz_out4, (v2i64) hz_out2);
2564 ILVEV_B2_UB(hz_out0, hz_out1, hz_out2, hz_out3, vec0, vec1);
2565 DOTP_UB2_UH(vec0, vec1, filt_vt, filt_vt, tmp0, tmp1);
2566 SRARI_H2_UH(tmp0, tmp1, 7);
2567 SAT_UH2_UH(tmp0, tmp1, 7);
2568 PCKEV_B2_UB(tmp0, tmp0, tmp1, tmp1, res0, res1);
2569 ST4x4_UB(res0, res1, 0, 1, 0, 1, dst, dst_stride);
2572 static void common_hv_2ht_2vt_4x8_msa(const uint8_t *src, int32_t src_stride,
2573 uint8_t *dst, int32_t dst_stride,
2574 const int8_t *filter_horiz, const int8_t *filter_vert)
2576 v16i8 src0, src1, src2, src3, src4, src5, src6, src7, src8, mask;
2577 v16i8 res0, res1, res2, res3;
2578 v16u8 filt_hz, filt_vt, vec0, vec1, vec2, vec3;
2579 v8u16 hz_out0, hz_out1, hz_out2, hz_out3, hz_out4, hz_out5, hz_out6;
2580 v8u16 hz_out7, hz_out8, vec4, vec5, vec6, vec7, filt;
2582 mask = LD_SB(&mc_filt_mask_arr[16]);
2584 /* rearranging filter */
2585 filt = LD_UH(filter_horiz);
2586 filt_hz = (v16u8) __msa_splati_h((v8i16) filt, 0);
2588 filt = LD_UH(filter_vert);
2589 filt_vt = (v16u8) __msa_splati_h((v8i16) filt, 0);
2591 LD_SB8(src, src_stride, src0, src1, src2, src3, src4, src5, src6, src7);
2592 src += (8 * src_stride);
2595 hz_out0 = HORIZ_2TAP_FILT_UH(src0, src1, mask, filt_hz, 7);
2596 hz_out2 = HORIZ_2TAP_FILT_UH(src2, src3, mask, filt_hz, 7);
2597 hz_out4 = HORIZ_2TAP_FILT_UH(src4, src5, mask, filt_hz, 7);
2598 hz_out6 = HORIZ_2TAP_FILT_UH(src6, src7, mask, filt_hz, 7);
2599 hz_out8 = HORIZ_2TAP_FILT_UH(src8, src8, mask, filt_hz, 7);
2600 SLDI_B3_UH(hz_out2, hz_out4, hz_out6, hz_out0, hz_out2, hz_out4, hz_out1,
2601 hz_out3, hz_out5, 8);
2602 hz_out7 = (v8u16) __msa_pckod_d((v2i64) hz_out8, (v2i64) hz_out6);
2604 ILVEV_B2_UB(hz_out0, hz_out1, hz_out2, hz_out3, vec0, vec1);
2605 ILVEV_B2_UB(hz_out4, hz_out5, hz_out6, hz_out7, vec2, vec3);
2606 DOTP_UB4_UH(vec0, vec1, vec2, vec3, filt_vt, filt_vt, filt_vt, filt_vt,
2607 vec4, vec5, vec6, vec7);
2608 SRARI_H4_UH(vec4, vec5, vec6, vec7, 7);
2609 SAT_UH4_UH(vec4, vec5, vec6, vec7, 7);
2610 PCKEV_B4_SB(vec4, vec4, vec5, vec5, vec6, vec6, vec7, vec7,
2611 res0, res1, res2, res3);
2612 ST4x4_UB(res0, res1, 0, 1, 0, 1, dst, dst_stride);
2613 dst += (4 * dst_stride);
2614 ST4x4_UB(res2, res3, 0, 1, 0, 1, dst, dst_stride);
2617 void ff_put_bilin_4hv_msa(uint8_t *dst, ptrdiff_t dst_stride,
2618 const uint8_t *src, ptrdiff_t src_stride,
2619 int height, int mx, int my)
2621 const int8_t *filter_horiz = vp9_bilinear_filters_msa[mx - 1];
2622 const int8_t *filter_vert = vp9_bilinear_filters_msa[my - 1];
2625 common_hv_2ht_2vt_4x4_msa(src, src_stride, dst, dst_stride,
2626 filter_horiz, filter_vert);
2627 } else if (8 == height) {
2628 common_hv_2ht_2vt_4x8_msa(src, src_stride, dst, dst_stride,
2629 filter_horiz, filter_vert);
2633 static void common_hv_2ht_2vt_8x4_msa(const uint8_t *src, int32_t src_stride,
2634 uint8_t *dst, int32_t dst_stride,
2635 const int8_t *filter_horiz, const int8_t *filter_vert)
2637 v16i8 src0, src1, src2, src3, src4, mask, out0, out1;
2638 v16u8 filt_hz, filt_vt, vec0, vec1, vec2, vec3;
2639 v8u16 hz_out0, hz_out1, tmp0, tmp1, tmp2, tmp3;
2642 mask = LD_SB(&mc_filt_mask_arr[0]);
2644 /* rearranging filter */
2645 filt = LD_SH(filter_horiz);
2646 filt_hz = (v16u8) __msa_splati_h(filt, 0);
2648 filt = LD_SH(filter_vert);
2649 filt_vt = (v16u8) __msa_splati_h(filt, 0);
2651 LD_SB5(src, src_stride, src0, src1, src2, src3, src4);
2653 hz_out0 = HORIZ_2TAP_FILT_UH(src0, src0, mask, filt_hz, 7);
2654 hz_out1 = HORIZ_2TAP_FILT_UH(src1, src1, mask, filt_hz, 7);
2655 vec0 = (v16u8) __msa_ilvev_b((v16i8) hz_out1, (v16i8) hz_out0);
2656 tmp0 = __msa_dotp_u_h(vec0, filt_vt);
2658 hz_out0 = HORIZ_2TAP_FILT_UH(src2, src2, mask, filt_hz, 7);
2659 vec1 = (v16u8) __msa_ilvev_b((v16i8) hz_out0, (v16i8) hz_out1);
2660 tmp1 = __msa_dotp_u_h(vec1, filt_vt);
2662 hz_out1 = HORIZ_2TAP_FILT_UH(src3, src3, mask, filt_hz, 7);
2663 vec2 = (v16u8) __msa_ilvev_b((v16i8) hz_out1, (v16i8) hz_out0);
2664 tmp2 = __msa_dotp_u_h(vec2, filt_vt);
2666 hz_out0 = HORIZ_2TAP_FILT_UH(src4, src4, mask, filt_hz, 7);
2667 vec3 = (v16u8) __msa_ilvev_b((v16i8) hz_out0, (v16i8) hz_out1);
2668 tmp3 = __msa_dotp_u_h(vec3, filt_vt);
2670 SRARI_H4_UH(tmp0, tmp1, tmp2, tmp3, 7);
2671 SAT_UH4_UH(tmp0, tmp1, tmp2, tmp3, 7);
2672 PCKEV_B2_SB(tmp1, tmp0, tmp3, tmp2, out0, out1);
2673 ST8x4_UB(out0, out1, dst, dst_stride);
2676 static void common_hv_2ht_2vt_8x8mult_msa(const uint8_t *src, int32_t src_stride,
2677 uint8_t *dst, int32_t dst_stride,
2678 const int8_t *filter_horiz, const int8_t *filter_vert,
2682 v16i8 src0, src1, src2, src3, src4, mask, out0, out1;
2683 v16u8 filt_hz, filt_vt, vec0;
2684 v8u16 hz_out0, hz_out1, tmp1, tmp2, tmp3, tmp4, tmp5, tmp6, tmp7, tmp8;
2687 mask = LD_SB(&mc_filt_mask_arr[0]);
2689 /* rearranging filter */
2690 filt = LD_SH(filter_horiz);
2691 filt_hz = (v16u8) __msa_splati_h(filt, 0);
2693 filt = LD_SH(filter_vert);
2694 filt_vt = (v16u8) __msa_splati_h(filt, 0);
2699 hz_out0 = HORIZ_2TAP_FILT_UH(src0, src0, mask, filt_hz, 7);
2701 for (loop_cnt = (height >> 3); loop_cnt--;) {
2702 LD_SB4(src, src_stride, src1, src2, src3, src4);
2703 src += (4 * src_stride);
2705 hz_out1 = HORIZ_2TAP_FILT_UH(src1, src1, mask, filt_hz, 7);
2706 vec0 = (v16u8) __msa_ilvev_b((v16i8) hz_out1, (v16i8) hz_out0);
2707 tmp1 = __msa_dotp_u_h(vec0, filt_vt);
2709 hz_out0 = HORIZ_2TAP_FILT_UH(src2, src2, mask, filt_hz, 7);
2710 vec0 = (v16u8) __msa_ilvev_b((v16i8) hz_out0, (v16i8) hz_out1);
2711 tmp2 = __msa_dotp_u_h(vec0, filt_vt);
2713 SRARI_H2_UH(tmp1, tmp2, 7);
2714 SAT_UH2_UH(tmp1, tmp2, 7);
2716 hz_out1 = HORIZ_2TAP_FILT_UH(src3, src3, mask, filt_hz, 7);
2717 vec0 = (v16u8) __msa_ilvev_b((v16i8) hz_out1, (v16i8) hz_out0);
2718 tmp3 = __msa_dotp_u_h(vec0, filt_vt);
2720 hz_out0 = HORIZ_2TAP_FILT_UH(src4, src4, mask, filt_hz, 7);
2721 LD_SB4(src, src_stride, src1, src2, src3, src4);
2722 src += (4 * src_stride);
2723 vec0 = (v16u8) __msa_ilvev_b((v16i8) hz_out0, (v16i8) hz_out1);
2724 tmp4 = __msa_dotp_u_h(vec0, filt_vt);
2726 SRARI_H2_UH(tmp3, tmp4, 7);
2727 SAT_UH2_UH(tmp3, tmp4, 7);
2728 PCKEV_B2_SB(tmp2, tmp1, tmp4, tmp3, out0, out1);
2729 ST8x4_UB(out0, out1, dst, dst_stride);
2730 dst += (4 * dst_stride);
2732 hz_out1 = HORIZ_2TAP_FILT_UH(src1, src1, mask, filt_hz, 7);
2733 vec0 = (v16u8) __msa_ilvev_b((v16i8) hz_out1, (v16i8) hz_out0);
2734 tmp5 = __msa_dotp_u_h(vec0, filt_vt);
2736 hz_out0 = HORIZ_2TAP_FILT_UH(src2, src2, mask, filt_hz, 7);
2737 vec0 = (v16u8) __msa_ilvev_b((v16i8) hz_out0, (v16i8) hz_out1);
2738 tmp6 = __msa_dotp_u_h(vec0, filt_vt);
2740 hz_out1 = HORIZ_2TAP_FILT_UH(src3, src3, mask, filt_hz, 7);
2741 vec0 = (v16u8) __msa_ilvev_b((v16i8) hz_out1, (v16i8) hz_out0);
2742 tmp7 = __msa_dotp_u_h(vec0, filt_vt);
2744 hz_out0 = HORIZ_2TAP_FILT_UH(src4, src4, mask, filt_hz, 7);
2745 vec0 = (v16u8) __msa_ilvev_b((v16i8) hz_out0, (v16i8) hz_out1);
2746 tmp8 = __msa_dotp_u_h(vec0, filt_vt);
2748 SRARI_H4_UH(tmp5, tmp6, tmp7, tmp8, 7);
2749 SAT_UH4_UH(tmp5, tmp6, tmp7, tmp8, 7);
2750 PCKEV_B2_SB(tmp6, tmp5, tmp8, tmp7, out0, out1);
2751 ST8x4_UB(out0, out1, dst, dst_stride);
2752 dst += (4 * dst_stride);
2756 void ff_put_bilin_8hv_msa(uint8_t *dst, ptrdiff_t dst_stride,
2757 const uint8_t *src, ptrdiff_t src_stride,
2758 int height, int mx, int my)
2760 const int8_t *filter_horiz = vp9_bilinear_filters_msa[mx - 1];
2761 const int8_t *filter_vert = vp9_bilinear_filters_msa[my - 1];
2764 common_hv_2ht_2vt_8x4_msa(src, src_stride, dst, dst_stride,
2765 filter_horiz, filter_vert);
2767 common_hv_2ht_2vt_8x8mult_msa(src, src_stride, dst, dst_stride,
2768 filter_horiz, filter_vert, height);
2772 void ff_put_bilin_16hv_msa(uint8_t *dst, ptrdiff_t dst_stride,
2773 const uint8_t *src, ptrdiff_t src_stride,
2774 int height, int mx, int my)
2777 const int8_t *filter_horiz = vp9_bilinear_filters_msa[mx - 1];
2778 const int8_t *filter_vert = vp9_bilinear_filters_msa[my - 1];
2779 v16i8 src0, src1, src2, src3, src4, src5, src6, src7, mask;
2780 v16u8 filt_hz, filt_vt, vec0, vec1;
2781 v8u16 tmp1, tmp2, hz_out0, hz_out1, hz_out2, hz_out3;
2784 mask = LD_SB(&mc_filt_mask_arr[0]);
2786 /* rearranging filter */
2787 filt = LD_SH(filter_horiz);
2788 filt_hz = (v16u8) __msa_splati_h(filt, 0);
2790 filt = LD_SH(filter_vert);
2791 filt_vt = (v16u8) __msa_splati_h(filt, 0);
2793 LD_SB2(src, 8, src0, src1);
2796 hz_out0 = HORIZ_2TAP_FILT_UH(src0, src0, mask, filt_hz, 7);
2797 hz_out2 = HORIZ_2TAP_FILT_UH(src1, src1, mask, filt_hz, 7);
2800 for (loop_cnt = (height >> 2); loop_cnt--;) {
2801 LD_SB4(src, src_stride, src0, src2, src4, src6);
2802 LD_SB4(src + 8, src_stride, src1, src3, src5, src7);
2803 src += (4 * src_stride);
2805 hz_out1 = HORIZ_2TAP_FILT_UH(src0, src0, mask, filt_hz, 7);
2806 hz_out3 = HORIZ_2TAP_FILT_UH(src1, src1, mask, filt_hz, 7);
2807 ILVEV_B2_UB(hz_out0, hz_out1, hz_out2, hz_out3, vec0, vec1);
2808 DOTP_UB2_UH(vec0, vec1, filt_vt, filt_vt, tmp1, tmp2);
2809 SRARI_H2_UH(tmp1, tmp2, 7);
2810 SAT_UH2_UH(tmp1, tmp2, 7);
2811 PCKEV_ST_SB(tmp1, tmp2, dst);
2814 hz_out0 = HORIZ_2TAP_FILT_UH(src2, src2, mask, filt_hz, 7);
2815 hz_out2 = HORIZ_2TAP_FILT_UH(src3, src3, mask, filt_hz, 7);
2816 ILVEV_B2_UB(hz_out1, hz_out0, hz_out3, hz_out2, vec0, vec1);
2817 DOTP_UB2_UH(vec0, vec1, filt_vt, filt_vt, tmp1, tmp2);
2818 SRARI_H2_UH(tmp1, tmp2, 7);
2819 SAT_UH2_UH(tmp1, tmp2, 7);
2820 PCKEV_ST_SB(tmp1, tmp2, dst);
2823 hz_out1 = HORIZ_2TAP_FILT_UH(src4, src4, mask, filt_hz, 7);
2824 hz_out3 = HORIZ_2TAP_FILT_UH(src5, src5, mask, filt_hz, 7);
2825 ILVEV_B2_UB(hz_out0, hz_out1, hz_out2, hz_out3, vec0, vec1);
2826 DOTP_UB2_UH(vec0, vec1, filt_vt, filt_vt, tmp1, tmp2);
2827 SRARI_H2_UH(tmp1, tmp2, 7);
2828 SAT_UH2_UH(tmp1, tmp2, 7);
2829 PCKEV_ST_SB(tmp1, tmp2, dst);
2832 hz_out0 = HORIZ_2TAP_FILT_UH(src6, src6, mask, filt_hz, 7);
2833 hz_out2 = HORIZ_2TAP_FILT_UH(src7, src7, mask, filt_hz, 7);
2834 ILVEV_B2_UB(hz_out1, hz_out0, hz_out3, hz_out2, vec0, vec1);
2835 DOTP_UB2_UH(vec0, vec1, filt_vt, filt_vt, tmp1, tmp2);
2836 SRARI_H2_UH(tmp1, tmp2, 7);
2837 SAT_UH2_UH(tmp1, tmp2, 7);
2838 PCKEV_ST_SB(tmp1, tmp2, dst);
2843 void ff_put_bilin_32hv_msa(uint8_t *dst, ptrdiff_t dst_stride,
2844 const uint8_t *src, ptrdiff_t src_stride,
2845 int height, int mx, int my)
2847 int32_t multiple8_cnt;
2849 for (multiple8_cnt = 2; multiple8_cnt--;) {
2850 ff_put_bilin_16hv_msa(dst, dst_stride, src, src_stride, height, mx, my);
2857 void ff_put_bilin_64hv_msa(uint8_t *dst, ptrdiff_t dst_stride,
2858 const uint8_t *src, ptrdiff_t src_stride,
2859 int height, int mx, int my)
2861 int32_t multiple8_cnt;
2863 for (multiple8_cnt = 4; multiple8_cnt--;) {
2864 ff_put_bilin_16hv_msa(dst, dst_stride, src, src_stride, height, mx, my);
2871 static void common_hz_2t_and_aver_dst_4x4_msa(const uint8_t *src,
2873 uint8_t *dst, int32_t dst_stride,
2874 const int8_t *filter)
2876 v16i8 src0, src1, src2, src3, mask;
2877 v16u8 filt0, dst0, dst1, dst2, dst3, vec0, vec1, res0, res1;
2878 v8u16 vec2, vec3, filt;
2880 mask = LD_SB(&mc_filt_mask_arr[16]);
2882 /* rearranging filter */
2883 filt = LD_UH(filter);
2884 filt0 = (v16u8) __msa_splati_h((v8i16) filt, 0);
2886 LD_SB4(src, src_stride, src0, src1, src2, src3);
2887 LD_UB4(dst, dst_stride, dst0, dst1, dst2, dst3);
2888 VSHF_B2_UB(src0, src1, src2, src3, mask, mask, vec0, vec1);
2889 DOTP_UB2_UH(vec0, vec1, filt0, filt0, vec2, vec3);
2890 SRARI_H2_UH(vec2, vec3, 7);
2891 PCKEV_B2_UB(vec2, vec2, vec3, vec3, res0, res1);
2892 ILVR_W2_UB(dst1, dst0, dst3, dst2, dst0, dst2);
2893 AVER_UB2_UB(res0, dst0, res1, dst2, res0, res1);
2894 ST4x4_UB(res0, res1, 0, 1, 0, 1, dst, dst_stride);
2897 static void common_hz_2t_and_aver_dst_4x8_msa(const uint8_t *src,
2899 uint8_t *dst, int32_t dst_stride,
2900 const int8_t *filter)
2902 v16i8 src0, src1, src2, src3, src4, src5, src6, src7, mask;
2903 v16u8 filt0, vec0, vec1, vec2, vec3, res0, res1, res2, res3;
2904 v16u8 dst0, dst1, dst2, dst3, dst4, dst5, dst6, dst7;
2905 v8u16 vec4, vec5, vec6, vec7, filt;
2907 mask = LD_SB(&mc_filt_mask_arr[16]);
2909 /* rearranging filter */
2910 filt = LD_UH(filter);
2911 filt0 = (v16u8) __msa_splati_h((v8i16) filt, 0);
2913 LD_SB8(src, src_stride, src0, src1, src2, src3, src4, src5, src6, src7);
2914 LD_UB8(dst, dst_stride, dst0, dst1, dst2, dst3, dst4, dst5, dst6, dst7);
2915 VSHF_B2_UB(src0, src1, src2, src3, mask, mask, vec0, vec1);
2916 VSHF_B2_UB(src4, src5, src6, src7, mask, mask, vec2, vec3);
2917 DOTP_UB4_UH(vec0, vec1, vec2, vec3, filt0, filt0, filt0, filt0, vec4, vec5,
2919 SRARI_H4_UH(vec4, vec5, vec6, vec7, 7);
2920 PCKEV_B4_UB(vec4, vec4, vec5, vec5, vec6, vec6, vec7, vec7, res0, res1,
2922 ILVR_W4_UB(dst1, dst0, dst3, dst2, dst5, dst4, dst7, dst6, dst0, dst2,
2924 AVER_UB4_UB(res0, dst0, res1, dst2, res2, dst4, res3, dst6, res0, res1,
2926 ST4x4_UB(res0, res1, 0, 1, 0, 1, dst, dst_stride);
2927 dst += (4 * dst_stride);
2928 ST4x4_UB(res2, res3, 0, 1, 0, 1, dst, dst_stride);
2931 void ff_avg_bilin_4h_msa(uint8_t *dst, ptrdiff_t dst_stride,
2932 const uint8_t *src, ptrdiff_t src_stride,
2933 int height, int mx, int my)
2935 const int8_t *filter = vp9_bilinear_filters_msa[mx - 1];
2938 common_hz_2t_and_aver_dst_4x4_msa(src, src_stride, dst, dst_stride,
2940 } else if (8 == height) {
2941 common_hz_2t_and_aver_dst_4x8_msa(src, src_stride, dst, dst_stride,
2946 static void common_hz_2t_and_aver_dst_8x4_msa(const uint8_t *src,
2948 uint8_t *dst, int32_t dst_stride,
2949 const int8_t *filter)
2951 v16i8 src0, src1, src2, src3, mask;
2952 v16u8 filt0, dst0, dst1, dst2, dst3;
2953 v8u16 vec0, vec1, vec2, vec3, filt;
2955 mask = LD_SB(&mc_filt_mask_arr[0]);
2957 /* rearranging filter */
2958 filt = LD_UH(filter);
2959 filt0 = (v16u8) __msa_splati_h((v8i16) filt, 0);
2961 LD_SB4(src, src_stride, src0, src1, src2, src3);
2962 VSHF_B2_UH(src0, src0, src1, src1, mask, mask, vec0, vec1);
2963 VSHF_B2_UH(src2, src2, src3, src3, mask, mask, vec2, vec3);
2964 DOTP_UB4_UH(vec0, vec1, vec2, vec3, filt0, filt0, filt0, filt0,
2965 vec0, vec1, vec2, vec3);
2966 SRARI_H4_UH(vec0, vec1, vec2, vec3, 7);
2967 LD_UB4(dst, dst_stride, dst0, dst1, dst2, dst3);
2968 PCKEV_AVG_ST8x4_UB(vec0, dst0, vec1, dst1, vec2, dst2, vec3, dst3,
2972 static void common_hz_2t_and_aver_dst_8x8mult_msa(const uint8_t *src,
2976 const int8_t *filter,
2979 v16i8 src0, src1, src2, src3, mask;
2980 v16u8 filt0, dst0, dst1, dst2, dst3;
2981 v8u16 vec0, vec1, vec2, vec3, filt;
2983 mask = LD_SB(&mc_filt_mask_arr[0]);
2985 /* rearranging filter */
2986 filt = LD_UH(filter);
2987 filt0 = (v16u8) __msa_splati_h((v8i16) filt, 0);
2989 LD_SB4(src, src_stride, src0, src1, src2, src3);
2990 src += (4 * src_stride);
2991 VSHF_B2_UH(src0, src0, src1, src1, mask, mask, vec0, vec1);
2992 VSHF_B2_UH(src2, src2, src3, src3, mask, mask, vec2, vec3);
2993 DOTP_UB4_UH(vec0, vec1, vec2, vec3, filt0, filt0, filt0, filt0, vec0, vec1,
2995 SRARI_H4_UH(vec0, vec1, vec2, vec3, 7);
2996 LD_UB4(dst, dst_stride, dst0, dst1, dst2, dst3);
2997 LD_SB4(src, src_stride, src0, src1, src2, src3);
2998 src += (4 * src_stride);
2999 PCKEV_AVG_ST8x4_UB(vec0, dst0, vec1, dst1, vec2, dst2, vec3, dst3,
3001 dst += (4 * dst_stride);
3003 VSHF_B2_UH(src0, src0, src1, src1, mask, mask, vec0, vec1);
3004 VSHF_B2_UH(src2, src2, src3, src3, mask, mask, vec2, vec3);
3005 DOTP_UB4_UH(vec0, vec1, vec2, vec3, filt0, filt0, filt0, filt0, vec0, vec1,
3007 SRARI_H4_UH(vec0, vec1, vec2, vec3, 7);
3008 LD_UB4(dst, dst_stride, dst0, dst1, dst2, dst3);
3009 PCKEV_AVG_ST8x4_UB(vec0, dst0, vec1, dst1, vec2, dst2, vec3, dst3,
3011 dst += (4 * dst_stride);
3014 LD_SB4(src, src_stride, src0, src1, src2, src3);
3015 src += (4 * src_stride);
3017 VSHF_B2_UH(src0, src0, src1, src1, mask, mask, vec0, vec1);
3018 VSHF_B2_UH(src2, src2, src3, src3, mask, mask, vec2, vec3);
3019 DOTP_UB4_UH(vec0, vec1, vec2, vec3, filt0, filt0, filt0, filt0, vec0,
3021 SRARI_H4_UH(vec0, vec1, vec2, vec3, 7);
3022 LD_UB4(dst, dst_stride, dst0, dst1, dst2, dst3);
3023 LD_SB4(src, src_stride, src0, src1, src2, src3);
3024 PCKEV_AVG_ST8x4_UB(vec0, dst0, vec1, dst1, vec2, dst2, vec3, dst3,
3026 dst += (4 * dst_stride);
3028 VSHF_B2_UH(src0, src0, src1, src1, mask, mask, vec0, vec1);
3029 VSHF_B2_UH(src2, src2, src3, src3, mask, mask, vec2, vec3);
3030 DOTP_UB4_UH(vec0, vec1, vec2, vec3, filt0, filt0, filt0, filt0, vec0,
3032 SRARI_H4_UH(vec0, vec1, vec2, vec3, 7);
3033 LD_UB4(dst, dst_stride, dst0, dst1, dst2, dst3);
3034 PCKEV_AVG_ST8x4_UB(vec0, dst0, vec1, dst1, vec2, dst2, vec3, dst3,
3039 void ff_avg_bilin_8h_msa(uint8_t *dst, ptrdiff_t dst_stride,
3040 const uint8_t *src, ptrdiff_t src_stride,
3041 int height, int mx, int my)
3043 const int8_t *filter = vp9_bilinear_filters_msa[mx - 1];
3046 common_hz_2t_and_aver_dst_8x4_msa(src, src_stride, dst, dst_stride,
3049 common_hz_2t_and_aver_dst_8x8mult_msa(src, src_stride, dst, dst_stride,
3054 void ff_avg_bilin_16h_msa(uint8_t *dst, ptrdiff_t dst_stride,
3055 const uint8_t *src, ptrdiff_t src_stride,
3056 int height, int mx, int my)
3059 const int8_t *filter = vp9_bilinear_filters_msa[mx - 1];
3060 v16i8 src0, src1, src2, src3, src4, src5, src6, src7, mask;
3061 v16u8 filt0, dst0, dst1, dst2, dst3;
3062 v16u8 vec0, vec1, vec2, vec3, vec4, vec5, vec6, vec7;
3063 v8u16 res0, res1, res2, res3, res4, res5, res6, res7, filt;
3065 mask = LD_SB(&mc_filt_mask_arr[0]);
3067 /* rearranging filter */
3068 filt = LD_UH(filter);
3069 filt0 = (v16u8) __msa_splati_h((v8i16) filt, 0);
3071 LD_SB4(src, src_stride, src0, src2, src4, src6);
3072 LD_SB4(src + 8, src_stride, src1, src3, src5, src7);
3073 src += (4 * src_stride);
3075 VSHF_B2_UB(src0, src0, src1, src1, mask, mask, vec0, vec1);
3076 VSHF_B2_UB(src2, src2, src3, src3, mask, mask, vec2, vec3);
3077 VSHF_B2_UB(src4, src4, src5, src5, mask, mask, vec4, vec5);
3078 VSHF_B2_UB(src6, src6, src7, src7, mask, mask, vec6, vec7);
3079 DOTP_UB4_UH(vec0, vec1, vec2, vec3, filt0, filt0, filt0, filt0, res0, res1,
3081 DOTP_UB4_UH(vec4, vec5, vec6, vec7, filt0, filt0, filt0, filt0, res4, res5,
3083 SRARI_H4_UH(res0, res1, res2, res3, 7);
3084 SRARI_H4_UH(res4, res5, res6, res7, 7);
3085 LD_UB4(dst, dst_stride, dst0, dst1, dst2, dst3);
3086 PCKEV_AVG_ST_UB(res1, res0, dst0, dst);
3088 PCKEV_AVG_ST_UB(res3, res2, dst1, dst);
3090 PCKEV_AVG_ST_UB(res5, res4, dst2, dst);
3092 PCKEV_AVG_ST_UB(res7, res6, dst3, dst);
3095 for (loop_cnt = (height >> 2) - 1; loop_cnt--;) {
3096 LD_SB4(src, src_stride, src0, src2, src4, src6);
3097 LD_SB4(src + 8, src_stride, src1, src3, src5, src7);
3098 src += (4 * src_stride);
3100 VSHF_B2_UB(src0, src0, src1, src1, mask, mask, vec0, vec1);
3101 VSHF_B2_UB(src2, src2, src3, src3, mask, mask, vec2, vec3);
3102 VSHF_B2_UB(src4, src4, src5, src5, mask, mask, vec4, vec5);
3103 VSHF_B2_UB(src6, src6, src7, src7, mask, mask, vec6, vec7);
3104 DOTP_UB4_UH(vec0, vec1, vec2, vec3, filt0, filt0, filt0, filt0, res0,
3106 DOTP_UB4_UH(vec4, vec5, vec6, vec7, filt0, filt0, filt0, filt0, res4,
3108 SRARI_H4_UH(res0, res1, res2, res3, 7);
3109 SRARI_H4_UH(res4, res5, res6, res7, 7);
3110 LD_UB4(dst, dst_stride, dst0, dst1, dst2, dst3);
3111 PCKEV_AVG_ST_UB(res1, res0, dst0, dst);
3113 PCKEV_AVG_ST_UB(res3, res2, dst1, dst);
3115 PCKEV_AVG_ST_UB(res5, res4, dst2, dst);
3117 PCKEV_AVG_ST_UB(res7, res6, dst3, dst);
3122 void ff_avg_bilin_32h_msa(uint8_t *dst, ptrdiff_t dst_stride,
3123 const uint8_t *src, ptrdiff_t src_stride,
3124 int height, int mx, int my)
3127 const int8_t *filter = vp9_bilinear_filters_msa[mx - 1];
3128 v16i8 src0, src1, src2, src3, src4, src5, src6, src7, mask;
3129 v16u8 filt0, dst0, dst1, dst2, dst3;
3130 v16u8 vec0, vec1, vec2, vec3, vec4, vec5, vec6, vec7;
3131 v8u16 res0, res1, res2, res3, res4, res5, res6, res7, filt;
3133 mask = LD_SB(&mc_filt_mask_arr[0]);
3135 /* rearranging filter */
3136 filt = LD_UH(filter);
3137 filt0 = (v16u8) __msa_splati_h((v8i16) filt, 0);
3139 for (loop_cnt = (height >> 1); loop_cnt--;) {
3141 src2 = LD_SB(src + 16);
3142 src3 = LD_SB(src + 24);
3143 src1 = __msa_sldi_b(src2, src0, 8);
3146 src6 = LD_SB(src + 16);
3147 src7 = LD_SB(src + 24);
3148 src5 = __msa_sldi_b(src6, src4, 8);
3151 VSHF_B2_UB(src0, src0, src1, src1, mask, mask, vec0, vec1);
3152 VSHF_B2_UB(src2, src2, src3, src3, mask, mask, vec2, vec3);
3153 VSHF_B2_UB(src4, src4, src5, src5, mask, mask, vec4, vec5);
3154 VSHF_B2_UB(src6, src6, src7, src7, mask, mask, vec6, vec7);
3155 DOTP_UB4_UH(vec0, vec1, vec2, vec3, filt0, filt0, filt0, filt0,
3156 res0, res1, res2, res3);
3157 DOTP_UB4_UH(vec4, vec5, vec6, vec7, filt0, filt0, filt0, filt0,
3158 res4, res5, res6, res7);
3159 SRARI_H4_UH(res0, res1, res2, res3, 7);
3160 SRARI_H4_UH(res4, res5, res6, res7, 7);
3161 LD_UB2(dst, 16, dst0, dst1);
3162 PCKEV_AVG_ST_UB(res1, res0, dst0, dst);
3163 PCKEV_AVG_ST_UB(res3, res2, dst1, (dst + 16));
3165 LD_UB2(dst, 16, dst2, dst3);
3166 PCKEV_AVG_ST_UB(res5, res4, dst2, dst);
3167 PCKEV_AVG_ST_UB(res7, res6, dst3, (dst + 16));
3172 void ff_avg_bilin_64h_msa(uint8_t *dst, ptrdiff_t dst_stride,
3173 const uint8_t *src, ptrdiff_t src_stride,
3174 int height, int mx, int my)
3177 const int8_t *filter = vp9_bilinear_filters_msa[mx - 1];
3178 v16i8 src0, src1, src2, src3, src4, src5, src6, src7, mask;
3179 v16u8 filt0, dst0, dst1, dst2, dst3;
3180 v16u8 vec0, vec1, vec2, vec3, vec4, vec5, vec6, vec7;
3181 v8u16 out0, out1, out2, out3, out4, out5, out6, out7, filt;
3183 mask = LD_SB(&mc_filt_mask_arr[0]);
3185 /* rearranging filter */
3186 filt = LD_UH(filter);
3187 filt0 = (v16u8) __msa_splati_h((v8i16) filt, 0);
3189 for (loop_cnt = height; loop_cnt--;) {
3190 LD_SB4(src, 16, src0, src2, src4, src6);
3191 src7 = LD_SB(src + 56);
3192 SLDI_B3_SB(src2, src4, src6, src0, src2, src4, src1, src3, src5, 8);
3195 VSHF_B2_UB(src0, src0, src1, src1, mask, mask, vec0, vec1);
3196 VSHF_B2_UB(src2, src2, src3, src3, mask, mask, vec2, vec3);
3197 VSHF_B2_UB(src4, src4, src5, src5, mask, mask, vec4, vec5);
3198 VSHF_B2_UB(src6, src6, src7, src7, mask, mask, vec6, vec7);
3199 DOTP_UB4_UH(vec0, vec1, vec2, vec3, filt0, filt0, filt0, filt0,
3200 out0, out1, out2, out3);
3201 DOTP_UB4_UH(vec4, vec5, vec6, vec7, filt0, filt0, filt0, filt0,
3202 out4, out5, out6, out7);
3203 SRARI_H4_UH(out0, out1, out2, out3, 7);
3204 SRARI_H4_UH(out4, out5, out6, out7, 7);
3205 LD_UB4(dst, 16, dst0, dst1, dst2, dst3);
3206 PCKEV_AVG_ST_UB(out1, out0, dst0, dst);
3207 PCKEV_AVG_ST_UB(out3, out2, dst1, dst + 16);
3208 PCKEV_AVG_ST_UB(out5, out4, dst2, dst + 32);
3209 PCKEV_AVG_ST_UB(out7, out6, dst3, dst + 48);
3214 static void common_vt_2t_and_aver_dst_4x4_msa(const uint8_t *src,
3216 uint8_t *dst, int32_t dst_stride,
3217 const int8_t *filter)
3219 v16i8 src0, src1, src2, src3, src4;
3220 v16u8 dst0, dst1, dst2, dst3, out, filt0, src2110, src4332;
3221 v16i8 src10_r, src32_r, src21_r, src43_r;
3225 filt = LD_SH(filter);
3226 filt0 = (v16u8) __msa_splati_h(filt, 0);
3228 LD_SB4(src, src_stride, src0, src1, src2, src3);
3229 src += (4 * src_stride);
3234 LD_UB4(dst, dst_stride, dst0, dst1, dst2, dst3);
3235 ILVR_W2_UB(dst1, dst0, dst3, dst2, dst0, dst1);
3236 dst0 = (v16u8) __msa_ilvr_d((v2i64) dst1, (v2i64) dst0);
3237 ILVR_B4_SB(src1, src0, src2, src1, src3, src2, src4, src3,
3238 src10_r, src21_r, src32_r, src43_r);
3239 ILVR_D2_UB(src21_r, src10_r, src43_r, src32_r, src2110, src4332);
3240 DOTP_UB2_UH(src2110, src4332, filt0, filt0, tmp0, tmp1);
3241 SRARI_H2_UH(tmp0, tmp1, 7);
3242 SAT_UH2_UH(tmp0, tmp1, 7);
3244 out = (v16u8) __msa_pckev_b((v16i8) tmp1, (v16i8) tmp0);
3245 out = __msa_aver_u_b(out, dst0);
3247 ST4x4_UB(out, out, 0, 1, 2, 3, dst, dst_stride);
3250 static void common_vt_2t_and_aver_dst_4x8_msa(const uint8_t *src,
3252 uint8_t *dst, int32_t dst_stride,
3253 const int8_t *filter)
3255 v16u8 dst0, dst1, dst2, dst3, dst4, dst5, dst6, dst7;
3256 v16i8 src0, src1, src2, src3, src4, src5, src6, src7, src8, src87_r;
3257 v16i8 src10_r, src32_r, src54_r, src76_r, src21_r, src43_r, src65_r;
3258 v16u8 src2110, src4332, src6554, src8776, filt0;
3259 v8u16 tmp0, tmp1, tmp2, tmp3;
3262 filt = LD_SH(filter);
3263 filt0 = (v16u8) __msa_splati_h(filt, 0);
3265 LD_SB8(src, src_stride, src0, src1, src2, src3, src4, src5, src6, src7);
3266 src += (8 * src_stride);
3269 LD_UB8(dst, dst_stride, dst0, dst1, dst2, dst3, dst4, dst5, dst6, dst7);
3270 ILVR_W4_UB(dst1, dst0, dst3, dst2, dst5, dst4, dst7, dst6, dst0, dst1,
3272 ILVR_D2_UB(dst1, dst0, dst3, dst2, dst0, dst1);
3273 ILVR_B4_SB(src1, src0, src2, src1, src3, src2, src4, src3, src10_r, src21_r,
3275 ILVR_B4_SB(src5, src4, src6, src5, src7, src6, src8, src7, src54_r, src65_r,
3277 ILVR_D4_UB(src21_r, src10_r, src43_r, src32_r, src65_r, src54_r,
3278 src87_r, src76_r, src2110, src4332, src6554, src8776);
3279 DOTP_UB4_UH(src2110, src4332, src6554, src8776, filt0, filt0, filt0, filt0,
3280 tmp0, tmp1, tmp2, tmp3);
3281 SRARI_H4_UH(tmp0, tmp1, tmp2, tmp3, 7);
3282 SAT_UH4_UH(tmp0, tmp1, tmp2, tmp3, 7);
3283 PCKEV_B2_UB(tmp1, tmp0, tmp3, tmp2, src2110, src4332);
3284 AVER_UB2_UB(src2110, dst0, src4332, dst1, src2110, src4332);
3285 ST4x4_UB(src2110, src2110, 0, 1, 2, 3, dst, dst_stride);
3286 dst += (4 * dst_stride);
3287 ST4x4_UB(src4332, src4332, 0, 1, 2, 3, dst, dst_stride);
3290 void ff_avg_bilin_4v_msa(uint8_t *dst, ptrdiff_t dst_stride,
3291 const uint8_t *src, ptrdiff_t src_stride,
3292 int height, int mx, int my)
3294 const int8_t *filter = vp9_bilinear_filters_msa[my - 1];
3297 common_vt_2t_and_aver_dst_4x4_msa(src, src_stride, dst, dst_stride,
3299 } else if (8 == height) {
3300 common_vt_2t_and_aver_dst_4x8_msa(src, src_stride, dst, dst_stride,
3305 static void common_vt_2t_and_aver_dst_8x4_msa(const uint8_t *src,
3309 const int8_t *filter)
3311 v16u8 src0, src1, src2, src3, src4;
3312 v16u8 dst0, dst1, dst2, dst3, vec0, vec1, vec2, vec3, filt0;
3313 v8u16 tmp0, tmp1, tmp2, tmp3;
3316 /* rearranging filter_y */
3317 filt = LD_SH(filter);
3318 filt0 = (v16u8) __msa_splati_h(filt, 0);
3320 LD_UB5(src, src_stride, src0, src1, src2, src3, src4);
3321 LD_UB4(dst, dst_stride, dst0, dst1, dst2, dst3);
3322 ILVR_B2_UB(src1, src0, src2, src1, vec0, vec1);
3323 ILVR_B2_UB(src3, src2, src4, src3, vec2, vec3);
3324 DOTP_UB4_UH(vec0, vec1, vec2, vec3, filt0, filt0, filt0, filt0,
3325 tmp0, tmp1, tmp2, tmp3);
3326 SRARI_H4_UH(tmp0, tmp1, tmp2, tmp3, 7);
3327 SAT_UH4_UH(tmp0, tmp1, tmp2, tmp3, 7);
3328 PCKEV_AVG_ST8x4_UB(tmp0, dst0, tmp1, dst1, tmp2, dst2, tmp3, dst3,
3332 static void common_vt_2t_and_aver_dst_8x8mult_msa(const uint8_t *src,
3336 const int8_t *filter,
3340 v16u8 src0, src1, src2, src3, src4, src5, src6, src7, src8;
3341 v16u8 dst1, dst2, dst3, dst4, dst5, dst6, dst7, dst8;
3342 v16u8 vec0, vec1, vec2, vec3, vec4, vec5, vec6, vec7, filt0;
3343 v8u16 tmp0, tmp1, tmp2, tmp3;
3346 /* rearranging filter_y */
3347 filt = LD_SH(filter);
3348 filt0 = (v16u8) __msa_splati_h(filt, 0);
3353 for (loop_cnt = (height >> 3); loop_cnt--;) {
3354 LD_UB8(src, src_stride, src1, src2, src3, src4, src5, src6, src7, src8);
3355 src += (8 * src_stride);
3356 LD_UB8(dst, dst_stride, dst1, dst2, dst3, dst4, dst5, dst6, dst7, dst8);
3358 ILVR_B4_UB(src1, src0, src2, src1, src3, src2, src4, src3,
3359 vec0, vec1, vec2, vec3);
3360 ILVR_B4_UB(src5, src4, src6, src5, src7, src6, src8, src7,
3361 vec4, vec5, vec6, vec7);
3362 DOTP_UB4_UH(vec0, vec1, vec2, vec3, filt0, filt0, filt0, filt0,
3363 tmp0, tmp1, tmp2, tmp3);
3364 SRARI_H4_UH(tmp0, tmp1, tmp2, tmp3, 7);
3365 SAT_UH4_UH(tmp0, tmp1, tmp2, tmp3, 7);
3366 PCKEV_AVG_ST8x4_UB(tmp0, dst1, tmp1, dst2, tmp2, dst3, tmp3,
3367 dst4, dst, dst_stride);
3368 dst += (4 * dst_stride);
3370 DOTP_UB4_UH(vec4, vec5, vec6, vec7, filt0, filt0, filt0, filt0,
3371 tmp0, tmp1, tmp2, tmp3);
3372 SRARI_H4_UH(tmp0, tmp1, tmp2, tmp3, 7);
3373 SAT_UH4_UH(tmp0, tmp1, tmp2, tmp3, 7);
3374 PCKEV_AVG_ST8x4_UB(tmp0, dst5, tmp1, dst6, tmp2, dst7, tmp3,
3375 dst8, dst, dst_stride);
3376 dst += (4 * dst_stride);
3382 void ff_avg_bilin_8v_msa(uint8_t *dst, ptrdiff_t dst_stride,
3383 const uint8_t *src, ptrdiff_t src_stride,
3384 int height, int mx, int my)
3386 const int8_t *filter = vp9_bilinear_filters_msa[my - 1];
3389 common_vt_2t_and_aver_dst_8x4_msa(src, src_stride, dst, dst_stride,
3392 common_vt_2t_and_aver_dst_8x8mult_msa(src, src_stride, dst, dst_stride,
3397 void ff_avg_bilin_16v_msa(uint8_t *dst, ptrdiff_t dst_stride,
3398 const uint8_t *src, ptrdiff_t src_stride,
3399 int height, int mx, int my)
3402 const int8_t *filter = vp9_bilinear_filters_msa[my - 1];
3403 v16u8 src0, src1, src2, src3, src4, dst0, dst1, dst2, dst3, filt0;
3404 v16u8 vec0, vec1, vec2, vec3, vec4, vec5, vec6, vec7;
3405 v8u16 tmp0, tmp1, tmp2, tmp3, filt;
3407 /* rearranging filter_y */
3408 filt = LD_UH(filter);
3409 filt0 = (v16u8) __msa_splati_h((v8i16) filt, 0);
3414 for (loop_cnt = (height >> 2); loop_cnt--;) {
3415 LD_UB4(src, src_stride, src1, src2, src3, src4);
3416 src += (4 * src_stride);
3418 LD_UB4(dst, dst_stride, dst0, dst1, dst2, dst3);
3419 ILVR_B2_UB(src1, src0, src2, src1, vec0, vec2);
3420 ILVL_B2_UB(src1, src0, src2, src1, vec1, vec3);
3421 DOTP_UB2_UH(vec0, vec1, filt0, filt0, tmp0, tmp1);
3422 SRARI_H2_UH(tmp0, tmp1, 7);
3423 SAT_UH2_UH(tmp0, tmp1, 7);
3424 PCKEV_AVG_ST_UB(tmp1, tmp0, dst0, dst);
3427 ILVR_B2_UB(src3, src2, src4, src3, vec4, vec6);
3428 ILVL_B2_UB(src3, src2, src4, src3, vec5, vec7);
3429 DOTP_UB2_UH(vec2, vec3, filt0, filt0, tmp2, tmp3);
3430 SRARI_H2_UH(tmp2, tmp3, 7);
3431 SAT_UH2_UH(tmp2, tmp3, 7);
3432 PCKEV_AVG_ST_UB(tmp3, tmp2, dst1, dst);
3435 DOTP_UB2_UH(vec4, vec5, filt0, filt0, tmp0, tmp1);
3436 SRARI_H2_UH(tmp0, tmp1, 7);
3437 SAT_UH2_UH(tmp0, tmp1, 7);
3438 PCKEV_AVG_ST_UB(tmp1, tmp0, dst2, dst);
3441 DOTP_UB2_UH(vec6, vec7, filt0, filt0, tmp2, tmp3);
3442 SRARI_H2_UH(tmp2, tmp3, 7);
3443 SAT_UH2_UH(tmp2, tmp3, 7);
3444 PCKEV_AVG_ST_UB(tmp3, tmp2, dst3, dst);
3451 void ff_avg_bilin_32v_msa(uint8_t *dst, ptrdiff_t dst_stride,
3452 const uint8_t *src, ptrdiff_t src_stride,
3453 int height, int mx, int my)
3456 const int8_t *filter = vp9_bilinear_filters_msa[my - 1];
3457 v16u8 src0, src1, src2, src3, src4, src5, src6, src7, src8, src9;
3458 v16u8 dst0, dst1, dst2, dst3, dst4, dst5, dst6, dst7;
3459 v16u8 vec0, vec1, vec2, vec3, vec4, vec5, vec6, vec7, filt0;
3460 v8u16 tmp0, tmp1, tmp2, tmp3, filt;
3462 /* rearranging filter_y */
3463 filt = LD_UH(filter);
3464 filt0 = (v16u8) __msa_splati_h((v8i16) filt, 0);
3466 LD_UB2(src, 16, src0, src5);
3469 for (loop_cnt = (height >> 2); loop_cnt--;) {
3470 LD_UB4(src, src_stride, src1, src2, src3, src4);
3471 LD_UB4(dst, dst_stride, dst0, dst1, dst2, dst3);
3472 ILVR_B2_UB(src1, src0, src2, src1, vec0, vec2);
3473 ILVL_B2_UB(src1, src0, src2, src1, vec1, vec3);
3475 LD_UB4(src + 16, src_stride, src6, src7, src8, src9);
3476 LD_UB4(dst + 16, dst_stride, dst4, dst5, dst6, dst7);
3477 src += (4 * src_stride);
3479 DOTP_UB2_UH(vec0, vec1, filt0, filt0, tmp0, tmp1);
3480 SRARI_H2_UH(tmp0, tmp1, 7);
3481 SAT_UH2_UH(tmp0, tmp1, 7);
3482 PCKEV_AVG_ST_UB(tmp1, tmp0, dst0, dst);
3484 DOTP_UB2_UH(vec2, vec3, filt0, filt0, tmp2, tmp3);
3485 SRARI_H2_UH(tmp2, tmp3, 7);
3486 SAT_UH2_UH(tmp2, tmp3, 7);
3487 PCKEV_AVG_ST_UB(tmp3, tmp2, dst1, dst + dst_stride);
3489 ILVR_B2_UB(src3, src2, src4, src3, vec4, vec6);
3490 ILVL_B2_UB(src3, src2, src4, src3, vec5, vec7);
3491 DOTP_UB2_UH(vec4, vec5, filt0, filt0, tmp0, tmp1);
3492 SRARI_H2_UH(tmp0, tmp1, 7);
3493 SAT_UH2_UH(tmp0, tmp1, 7);
3494 PCKEV_AVG_ST_UB(tmp1, tmp0, dst2, dst + 2 * dst_stride);
3496 DOTP_UB2_UH(vec6, vec7, filt0, filt0, tmp2, tmp3);
3497 SRARI_H2_UH(tmp2, tmp3, 7);
3498 SAT_UH2_UH(tmp2, tmp3, 7);
3499 PCKEV_AVG_ST_UB(tmp3, tmp2, dst3, dst + 3 * dst_stride);
3501 ILVR_B2_UB(src6, src5, src7, src6, vec0, vec2);
3502 ILVL_B2_UB(src6, src5, src7, src6, vec1, vec3);
3503 DOTP_UB2_UH(vec0, vec1, filt0, filt0, tmp0, tmp1);
3504 SRARI_H2_UH(tmp0, tmp1, 7);
3505 SAT_UH2_UH(tmp0, tmp1, 7);
3506 PCKEV_AVG_ST_UB(tmp1, tmp0, dst4, dst + 16);
3508 DOTP_UB2_UH(vec2, vec3, filt0, filt0, tmp2, tmp3);
3509 SRARI_H2_UH(tmp2, tmp3, 7);
3510 SAT_UH2_UH(tmp2, tmp3, 7);
3511 PCKEV_AVG_ST_UB(tmp3, tmp2, dst5, dst + 16 + dst_stride);
3513 ILVR_B2_UB(src8, src7, src9, src8, vec4, vec6);
3514 ILVL_B2_UB(src8, src7, src9, src8, vec5, vec7);
3515 DOTP_UB2_UH(vec4, vec5, filt0, filt0, tmp0, tmp1);
3516 SRARI_H2_UH(tmp0, tmp1, 7);
3517 SAT_UH2_UH(tmp0, tmp1, 7);
3518 PCKEV_AVG_ST_UB(tmp1, tmp0, dst6, dst + 16 + 2 * dst_stride);
3520 DOTP_UB2_UH(vec6, vec7, filt0, filt0, tmp2, tmp3);
3521 SRARI_H2_UH(tmp2, tmp3, 7);
3522 SAT_UH2_UH(tmp2, tmp3, 7);
3523 PCKEV_AVG_ST_UB(tmp3, tmp2, dst7, dst + 16 + 3 * dst_stride);
3524 dst += (4 * dst_stride);
3531 void ff_avg_bilin_64v_msa(uint8_t *dst, ptrdiff_t dst_stride,
3532 const uint8_t *src, ptrdiff_t src_stride,
3533 int height, int mx, int my)
3536 const int8_t *filter = vp9_bilinear_filters_msa[my - 1];
3537 v16u8 src0, src1, src2, src3, src4, src5;
3538 v16u8 src6, src7, src8, src9, src10, src11, filt0;
3539 v16u8 dst0, dst1, dst2, dst3, dst4, dst5, dst6, dst7;
3540 v16u8 vec0, vec1, vec2, vec3, vec4, vec5, vec6, vec7;
3541 v8u16 tmp0, tmp1, tmp2, tmp3, tmp4, tmp5, tmp6, tmp7;
3544 /* rearranging filter_y */
3545 filt = LD_UH(filter);
3546 filt0 = (v16u8) __msa_splati_h((v8i16) filt, 0);
3548 LD_UB4(src, 16, src0, src3, src6, src9);
3551 for (loop_cnt = (height >> 1); loop_cnt--;) {
3552 LD_UB2(src, src_stride, src1, src2);
3553 LD_UB2(dst, dst_stride, dst0, dst1);
3554 LD_UB2(src + 16, src_stride, src4, src5);
3555 LD_UB2(dst + 16, dst_stride, dst2, dst3);
3556 LD_UB2(src + 32, src_stride, src7, src8);
3557 LD_UB2(dst + 32, dst_stride, dst4, dst5);
3558 LD_UB2(src + 48, src_stride, src10, src11);
3559 LD_UB2(dst + 48, dst_stride, dst6, dst7);
3560 src += (2 * src_stride);
3562 ILVR_B2_UB(src1, src0, src2, src1, vec0, vec2);
3563 ILVL_B2_UB(src1, src0, src2, src1, vec1, vec3);
3564 DOTP_UB2_UH(vec0, vec1, filt0, filt0, tmp0, tmp1);
3565 SRARI_H2_UH(tmp0, tmp1, 7);
3566 SAT_UH2_UH(tmp0, tmp1, 7);
3567 PCKEV_AVG_ST_UB(tmp1, tmp0, dst0, dst);
3569 DOTP_UB2_UH(vec2, vec3, filt0, filt0, tmp2, tmp3);
3570 SRARI_H2_UH(tmp2, tmp3, 7);
3571 SAT_UH2_UH(tmp2, tmp3, 7);
3572 PCKEV_AVG_ST_UB(tmp3, tmp2, dst1, dst + dst_stride);
3574 ILVR_B2_UB(src4, src3, src5, src4, vec4, vec6);
3575 ILVL_B2_UB(src4, src3, src5, src4, vec5, vec7);
3576 DOTP_UB2_UH(vec4, vec5, filt0, filt0, tmp4, tmp5);
3577 SRARI_H2_UH(tmp4, tmp5, 7);
3578 SAT_UH2_UH(tmp4, tmp5, 7);
3579 PCKEV_AVG_ST_UB(tmp5, tmp4, dst2, dst + 16);
3581 DOTP_UB2_UH(vec6, vec7, filt0, filt0, tmp6, tmp7);
3582 SRARI_H2_UH(tmp6, tmp7, 7);
3583 SAT_UH2_UH(tmp6, tmp7, 7);
3584 PCKEV_AVG_ST_UB(tmp7, tmp6, dst3, dst + 16 + dst_stride);
3586 ILVR_B2_UB(src7, src6, src8, src7, vec0, vec2);
3587 ILVL_B2_UB(src7, src6, src8, src7, vec1, vec3);
3588 DOTP_UB2_UH(vec0, vec1, filt0, filt0, tmp0, tmp1);
3589 SRARI_H2_UH(tmp0, tmp1, 7);
3590 SAT_UH2_UH(tmp0, tmp1, 7);
3591 PCKEV_AVG_ST_UB(tmp1, tmp0, dst4, dst + 32);
3593 DOTP_UB2_UH(vec2, vec3, filt0, filt0, tmp2, tmp3);
3594 SRARI_H2_UH(tmp2, tmp3, 7);
3595 SAT_UH2_UH(tmp2, tmp3, 7);
3596 PCKEV_AVG_ST_UB(tmp3, tmp2, dst5, dst + 32 + dst_stride);
3598 ILVR_B2_UB(src10, src9, src11, src10, vec4, vec6);
3599 ILVL_B2_UB(src10, src9, src11, src10, vec5, vec7);
3600 DOTP_UB2_UH(vec4, vec5, filt0, filt0, tmp4, tmp5);
3601 SRARI_H2_UH(tmp4, tmp5, 7);
3602 SAT_UH2_UH(tmp4, tmp5, 7);
3603 PCKEV_AVG_ST_UB(tmp5, tmp4, dst6, (dst + 48));
3605 DOTP_UB2_UH(vec6, vec7, filt0, filt0, tmp6, tmp7);
3606 SRARI_H2_UH(tmp6, tmp7, 7);
3607 SAT_UH2_UH(tmp6, tmp7, 7);
3608 PCKEV_AVG_ST_UB(tmp7, tmp6, dst7, dst + 48 + dst_stride);
3609 dst += (2 * dst_stride);
3618 static void common_hv_2ht_2vt_and_aver_dst_4x4_msa(const uint8_t *src,
3622 const int8_t *filter_horiz,
3623 const int8_t *filter_vert)
3625 v16i8 src0, src1, src2, src3, src4, mask;
3626 v16u8 filt_hz, filt_vt, vec0, vec1;
3627 v16u8 dst0, dst1, dst2, dst3, res0, res1;
3628 v8u16 hz_out0, hz_out1, hz_out2, hz_out3, hz_out4, tmp0, tmp1, filt;
3630 mask = LD_SB(&mc_filt_mask_arr[16]);
3632 /* rearranging filter */
3633 filt = LD_UH(filter_horiz);
3634 filt_hz = (v16u8) __msa_splati_h((v8i16) filt, 0);
3636 filt = LD_UH(filter_vert);
3637 filt_vt = (v16u8) __msa_splati_h((v8i16) filt, 0);
3639 LD_SB5(src, src_stride, src0, src1, src2, src3, src4);
3641 hz_out0 = HORIZ_2TAP_FILT_UH(src0, src1, mask, filt_hz, 7);
3642 hz_out2 = HORIZ_2TAP_FILT_UH(src2, src3, mask, filt_hz, 7);
3643 hz_out4 = HORIZ_2TAP_FILT_UH(src4, src4, mask, filt_hz, 7);
3644 hz_out1 = (v8u16) __msa_sldi_b((v16i8) hz_out2, (v16i8) hz_out0, 8);
3645 hz_out3 = (v8u16) __msa_pckod_d((v2i64) hz_out4, (v2i64) hz_out2);
3646 ILVEV_B2_UB(hz_out0, hz_out1, hz_out2, hz_out3, vec0, vec1);
3648 LD_UB4(dst, dst_stride, dst0, dst1, dst2, dst3);
3649 ILVR_W2_UB(dst1, dst0, dst3, dst2, dst0, dst2);
3650 DOTP_UB2_UH(vec0, vec1, filt_vt, filt_vt, tmp0, tmp1);
3651 SRARI_H2_UH(tmp0, tmp1, 7);
3652 SAT_UH2_UH(tmp0, tmp1, 7);
3653 PCKEV_B2_UB(tmp0, tmp0, tmp1, tmp1, res0, res1);
3654 AVER_UB2_UB(res0, dst0, res1, dst2, res0, res1);
3655 ST4x4_UB(res0, res1, 0, 1, 0, 1, dst, dst_stride);
3658 static void common_hv_2ht_2vt_and_aver_dst_4x8_msa(const uint8_t *src,
3662 const int8_t *filter_horiz,
3663 const int8_t *filter_vert)
3665 v16i8 src0, src1, src2, src3, src4, src5, src6, src7, src8, mask;
3666 v16u8 filt_hz, filt_vt, vec0, vec1, vec2, vec3, res0, res1, res2, res3;
3667 v16u8 dst0, dst1, dst2, dst3, dst4, dst5, dst6, dst7;
3668 v8u16 hz_out0, hz_out1, hz_out2, hz_out3, hz_out4, hz_out5, hz_out6;
3669 v8u16 hz_out7, hz_out8, tmp0, tmp1, tmp2, tmp3;
3672 mask = LD_SB(&mc_filt_mask_arr[16]);
3674 /* rearranging filter */
3675 filt = LD_SH(filter_horiz);
3676 filt_hz = (v16u8) __msa_splati_h(filt, 0);
3678 filt = LD_SH(filter_vert);
3679 filt_vt = (v16u8) __msa_splati_h(filt, 0);
3681 LD_SB8(src, src_stride, src0, src1, src2, src3, src4, src5, src6, src7);
3682 src += (8 * src_stride);
3685 hz_out0 = HORIZ_2TAP_FILT_UH(src0, src1, mask, filt_hz, 7);
3686 hz_out2 = HORIZ_2TAP_FILT_UH(src2, src3, mask, filt_hz, 7);
3687 hz_out4 = HORIZ_2TAP_FILT_UH(src4, src5, mask, filt_hz, 7);
3688 hz_out6 = HORIZ_2TAP_FILT_UH(src6, src7, mask, filt_hz, 7);
3689 hz_out8 = HORIZ_2TAP_FILT_UH(src8, src8, mask, filt_hz, 7);
3690 SLDI_B3_UH(hz_out2, hz_out4, hz_out6, hz_out0, hz_out2, hz_out4, hz_out1,
3691 hz_out3, hz_out5, 8);
3692 hz_out7 = (v8u16) __msa_pckod_d((v2i64) hz_out8, (v2i64) hz_out6);
3694 LD_UB8(dst, dst_stride, dst0, dst1, dst2, dst3, dst4, dst5, dst6, dst7);
3695 ILVR_W4_UB(dst1, dst0, dst3, dst2, dst5, dst4, dst7, dst6, dst0, dst2,
3697 ILVEV_B2_UB(hz_out0, hz_out1, hz_out2, hz_out3, vec0, vec1);
3698 ILVEV_B2_UB(hz_out4, hz_out5, hz_out6, hz_out7, vec2, vec3);
3699 DOTP_UB4_UH(vec0, vec1, vec2, vec3, filt_vt, filt_vt, filt_vt, filt_vt,
3700 tmp0, tmp1, tmp2, tmp3);
3701 SRARI_H4_UH(tmp0, tmp1, tmp2, tmp3, 7);
3702 SAT_UH4_UH(tmp0, tmp1, tmp2, tmp3, 7);
3703 PCKEV_B4_UB(tmp0, tmp0, tmp1, tmp1, tmp2, tmp2, tmp3, tmp3, res0, res1,
3705 AVER_UB4_UB(res0, dst0, res1, dst2, res2, dst4, res3, dst6, res0, res1,
3707 ST4x4_UB(res0, res1, 0, 1, 0, 1, dst, dst_stride);
3708 dst += (4 * dst_stride);
3709 ST4x4_UB(res2, res3, 0, 1, 0, 1, dst, dst_stride);
3712 void ff_avg_bilin_4hv_msa(uint8_t *dst, ptrdiff_t dst_stride,
3713 const uint8_t *src, ptrdiff_t src_stride,
3714 int height, int mx, int my)
3716 const int8_t *filter_horiz = vp9_bilinear_filters_msa[mx - 1];
3717 const int8_t *filter_vert = vp9_bilinear_filters_msa[my - 1];
3720 common_hv_2ht_2vt_and_aver_dst_4x4_msa(src, src_stride, dst, dst_stride,
3721 filter_horiz, filter_vert);
3722 } else if (8 == height) {
3723 common_hv_2ht_2vt_and_aver_dst_4x8_msa(src, src_stride, dst, dst_stride,
3724 filter_horiz, filter_vert);
3728 static void common_hv_2ht_2vt_and_aver_dst_8x4_msa(const uint8_t *src,
3732 const int8_t *filter_horiz,
3733 const int8_t *filter_vert)
3735 v16i8 src0, src1, src2, src3, src4, mask;
3736 v16u8 filt_hz, filt_vt, dst0, dst1, dst2, dst3, vec0, vec1, vec2, vec3;
3737 v8u16 hz_out0, hz_out1, tmp0, tmp1, tmp2, tmp3;
3740 mask = LD_SB(&mc_filt_mask_arr[0]);
3742 /* rearranging filter */
3743 filt = LD_SH(filter_horiz);
3744 filt_hz = (v16u8) __msa_splati_h(filt, 0);
3746 filt = LD_SH(filter_vert);
3747 filt_vt = (v16u8) __msa_splati_h(filt, 0);
3749 LD_SB5(src, src_stride, src0, src1, src2, src3, src4);
3750 src += (5 * src_stride);
3752 LD_UB4(dst, dst_stride, dst0, dst1, dst2, dst3);
3753 hz_out0 = HORIZ_2TAP_FILT_UH(src0, src0, mask, filt_hz, 7);
3754 hz_out1 = HORIZ_2TAP_FILT_UH(src1, src1, mask, filt_hz, 7);
3755 vec0 = (v16u8) __msa_ilvev_b((v16i8) hz_out1, (v16i8) hz_out0);
3756 tmp0 = __msa_dotp_u_h(vec0, filt_vt);
3758 hz_out0 = HORIZ_2TAP_FILT_UH(src2, src2, mask, filt_hz, 7);
3759 vec1 = (v16u8) __msa_ilvev_b((v16i8) hz_out0, (v16i8) hz_out1);
3760 tmp1 = __msa_dotp_u_h(vec1, filt_vt);
3762 hz_out1 = HORIZ_2TAP_FILT_UH(src3, src3, mask, filt_hz, 7);
3763 vec2 = (v16u8) __msa_ilvev_b((v16i8) hz_out1, (v16i8) hz_out0);
3764 tmp2 = __msa_dotp_u_h(vec2, filt_vt);
3766 hz_out0 = HORIZ_2TAP_FILT_UH(src4, src4, mask, filt_hz, 7);
3767 vec3 = (v16u8) __msa_ilvev_b((v16i8) hz_out0, (v16i8) hz_out1);
3768 tmp3 = __msa_dotp_u_h(vec3, filt_vt);
3770 SRARI_H4_UH(tmp0, tmp1, tmp2, tmp3, 7);
3771 SAT_UH4_UH(tmp0, tmp1, tmp2, tmp3, 7);
3772 PCKEV_AVG_ST8x4_UB(tmp0, dst0, tmp1, dst1, tmp2, dst2, tmp3, dst3,
3776 static void common_hv_2ht_2vt_and_aver_dst_8x8mult_msa(const uint8_t *src,
3780 const int8_t *filter_horiz,
3781 const int8_t *filter_vert,
3785 v16i8 src0, src1, src2, src3, src4, mask;
3786 v16u8 filt_hz, filt_vt, vec0, dst0, dst1, dst2, dst3;
3787 v8u16 hz_out0, hz_out1, tmp0, tmp1, tmp2, tmp3;
3790 mask = LD_SB(&mc_filt_mask_arr[0]);
3792 /* rearranging filter */
3793 filt = LD_SH(filter_horiz);
3794 filt_hz = (v16u8) __msa_splati_h(filt, 0);
3796 filt = LD_SH(filter_vert);
3797 filt_vt = (v16u8) __msa_splati_h(filt, 0);
3802 hz_out0 = HORIZ_2TAP_FILT_UH(src0, src0, mask, filt_hz, 7);
3804 for (loop_cnt = (height >> 2); loop_cnt--;) {
3805 LD_SB4(src, src_stride, src1, src2, src3, src4);
3806 src += (4 * src_stride);
3808 hz_out1 = HORIZ_2TAP_FILT_UH(src1, src1, mask, filt_hz, 7);
3809 vec0 = (v16u8) __msa_ilvev_b((v16i8) hz_out1, (v16i8) hz_out0);
3810 tmp0 = __msa_dotp_u_h(vec0, filt_vt);
3812 hz_out0 = HORIZ_2TAP_FILT_UH(src2, src2, mask, filt_hz, 7);
3813 vec0 = (v16u8) __msa_ilvev_b((v16i8) hz_out0, (v16i8) hz_out1);
3814 tmp1 = __msa_dotp_u_h(vec0, filt_vt);
3816 SRARI_H2_UH(tmp0, tmp1, 7);
3817 SAT_UH2_UH(tmp0, tmp1, 7);
3819 hz_out1 = HORIZ_2TAP_FILT_UH(src3, src3, mask, filt_hz, 7);
3820 vec0 = (v16u8) __msa_ilvev_b((v16i8) hz_out1, (v16i8) hz_out0);
3821 tmp2 = __msa_dotp_u_h(vec0, filt_vt);
3823 hz_out0 = HORIZ_2TAP_FILT_UH(src4, src4, mask, filt_hz, 7);
3824 vec0 = (v16u8) __msa_ilvev_b((v16i8) hz_out0, (v16i8) hz_out1);
3825 tmp3 = __msa_dotp_u_h(vec0, filt_vt);
3827 SRARI_H2_UH(tmp2, tmp3, 7);
3828 SAT_UH2_UH(tmp2, tmp3, 7);
3829 LD_UB4(dst, dst_stride, dst0, dst1, dst2, dst3);
3830 PCKEV_AVG_ST8x4_UB(tmp0, dst0, tmp1, dst1, tmp2, dst2, tmp3,
3831 dst3, dst, dst_stride);
3832 dst += (4 * dst_stride);
3836 void ff_avg_bilin_8hv_msa(uint8_t *dst, ptrdiff_t dst_stride,
3837 const uint8_t *src, ptrdiff_t src_stride,
3838 int height, int mx, int my)
3840 const int8_t *filter_horiz = vp9_bilinear_filters_msa[mx - 1];
3841 const int8_t *filter_vert = vp9_bilinear_filters_msa[my - 1];
3844 common_hv_2ht_2vt_and_aver_dst_8x4_msa(src, src_stride, dst, dst_stride,
3845 filter_horiz, filter_vert);
3847 common_hv_2ht_2vt_and_aver_dst_8x8mult_msa(src, src_stride,
3849 filter_horiz, filter_vert,
3854 void ff_avg_bilin_16hv_msa(uint8_t *dst, ptrdiff_t dst_stride,
3855 const uint8_t *src, ptrdiff_t src_stride,
3856 int height, int mx, int my)
3859 const int8_t *filter_horiz = vp9_bilinear_filters_msa[mx - 1];
3860 const int8_t *filter_vert = vp9_bilinear_filters_msa[my - 1];
3861 v16i8 src0, src1, src2, src3, src4, src5, src6, src7, mask;
3862 v16u8 filt_hz, filt_vt, vec0, vec1, dst0, dst1, dst2, dst3;
3863 v8u16 hz_out0, hz_out1, hz_out2, hz_out3, tmp0, tmp1;
3866 mask = LD_SB(&mc_filt_mask_arr[0]);
3868 /* rearranging filter */
3869 filt = LD_SH(filter_horiz);
3870 filt_hz = (v16u8) __msa_splati_h(filt, 0);
3872 filt = LD_SH(filter_vert);
3873 filt_vt = (v16u8) __msa_splati_h(filt, 0);
3875 LD_SB2(src, 8, src0, src1);
3878 hz_out0 = HORIZ_2TAP_FILT_UH(src0, src0, mask, filt_hz, 7);
3879 hz_out2 = HORIZ_2TAP_FILT_UH(src1, src1, mask, filt_hz, 7);
3881 for (loop_cnt = (height >> 2); loop_cnt--;) {
3882 LD_SB4(src, src_stride, src0, src2, src4, src6);
3883 LD_SB4(src + 8, src_stride, src1, src3, src5, src7);
3884 src += (4 * src_stride);
3885 LD_UB4(dst, dst_stride, dst0, dst1, dst2, dst3);
3887 hz_out1 = HORIZ_2TAP_FILT_UH(src0, src0, mask, filt_hz, 7);
3888 hz_out3 = HORIZ_2TAP_FILT_UH(src1, src1, mask, filt_hz, 7);
3889 ILVEV_B2_UB(hz_out0, hz_out1, hz_out2, hz_out3, vec0, vec1);
3890 DOTP_UB2_UH(vec0, vec1, filt_vt, filt_vt, tmp0, tmp1);
3891 SRARI_H2_UH(tmp0, tmp1, 7);
3892 SAT_UH2_UH(tmp0, tmp1, 7);
3893 PCKEV_AVG_ST_UB(tmp1, tmp0, dst0, dst);
3896 hz_out0 = HORIZ_2TAP_FILT_UH(src2, src2, mask, filt_hz, 7);
3897 hz_out2 = HORIZ_2TAP_FILT_UH(src3, src3, mask, filt_hz, 7);
3898 ILVEV_B2_UB(hz_out1, hz_out0, hz_out3, hz_out2, vec0, vec1);
3899 DOTP_UB2_UH(vec0, vec1, filt_vt, filt_vt, tmp0, tmp1);
3900 SRARI_H2_UH(tmp0, tmp1, 7);
3901 SAT_UH2_UH(tmp0, tmp1, 7);
3902 PCKEV_AVG_ST_UB(tmp1, tmp0, dst1, dst);
3905 hz_out1 = HORIZ_2TAP_FILT_UH(src4, src4, mask, filt_hz, 7);
3906 hz_out3 = HORIZ_2TAP_FILT_UH(src5, src5, mask, filt_hz, 7);
3907 ILVEV_B2_UB(hz_out0, hz_out1, hz_out2, hz_out3, vec0, vec1);
3908 DOTP_UB2_UH(vec0, vec1, filt_vt, filt_vt, tmp0, tmp1);
3909 SRARI_H2_UH(tmp0, tmp1, 7);
3910 SAT_UH2_UH(tmp0, tmp1, 7);
3911 PCKEV_AVG_ST_UB(tmp1, tmp0, dst2, dst);
3914 hz_out0 = HORIZ_2TAP_FILT_UH(src6, src6, mask, filt_hz, 7);
3915 hz_out2 = HORIZ_2TAP_FILT_UH(src7, src7, mask, filt_hz, 7);
3916 ILVEV_B2_UB(hz_out1, hz_out0, hz_out3, hz_out2, vec0, vec1);
3917 DOTP_UB2_UH(vec0, vec1, filt_vt, filt_vt, tmp0, tmp1);
3918 SRARI_H2_UH(tmp0, tmp1, 7);
3919 SAT_UH2_UH(tmp0, tmp1, 7);
3920 PCKEV_AVG_ST_UB(tmp1, tmp0, dst3, dst);
3925 void ff_avg_bilin_32hv_msa(uint8_t *dst, ptrdiff_t dst_stride,
3926 const uint8_t *src, ptrdiff_t src_stride,
3927 int height, int mx, int my)
3929 int32_t multiple8_cnt;
3931 for (multiple8_cnt = 2; multiple8_cnt--;) {
3932 ff_avg_bilin_16hv_msa(dst, dst_stride, src, src_stride, height, mx, my);
3939 void ff_avg_bilin_64hv_msa(uint8_t *dst, ptrdiff_t dst_stride,
3940 const uint8_t *src, ptrdiff_t src_stride,
3941 int height, int mx, int my)
3943 int32_t multiple8_cnt;
3945 for (multiple8_cnt = 4; multiple8_cnt--;) {
3946 ff_avg_bilin_16hv_msa(dst, dst_stride, src, src_stride, height, mx, my);
3953 static void copy_width8_msa(const uint8_t *src, int32_t src_stride,
3954 uint8_t *dst, int32_t dst_stride,
3958 uint64_t out0, out1, out2, out3, out4, out5, out6, out7;
3959 v16u8 src0, src1, src2, src3, src4, src5, src6, src7;
3961 if (0 == height % 12) {
3962 for (cnt = (height / 12); cnt--;) {
3963 LD_UB8(src, src_stride,
3964 src0, src1, src2, src3, src4, src5, src6, src7);
3965 src += (8 * src_stride);
3967 out0 = __msa_copy_u_d((v2i64) src0, 0);
3968 out1 = __msa_copy_u_d((v2i64) src1, 0);
3969 out2 = __msa_copy_u_d((v2i64) src2, 0);
3970 out3 = __msa_copy_u_d((v2i64) src3, 0);
3971 out4 = __msa_copy_u_d((v2i64) src4, 0);
3972 out5 = __msa_copy_u_d((v2i64) src5, 0);
3973 out6 = __msa_copy_u_d((v2i64) src6, 0);
3974 out7 = __msa_copy_u_d((v2i64) src7, 0);
3976 SD4(out0, out1, out2, out3, dst, dst_stride);
3977 dst += (4 * dst_stride);
3978 SD4(out4, out5, out6, out7, dst, dst_stride);
3979 dst += (4 * dst_stride);
3981 LD_UB4(src, src_stride, src0, src1, src2, src3);
3982 src += (4 * src_stride);
3984 out0 = __msa_copy_u_d((v2i64) src0, 0);
3985 out1 = __msa_copy_u_d((v2i64) src1, 0);
3986 out2 = __msa_copy_u_d((v2i64) src2, 0);
3987 out3 = __msa_copy_u_d((v2i64) src3, 0);
3989 SD4(out0, out1, out2, out3, dst, dst_stride);
3990 dst += (4 * dst_stride);
3992 } else if (0 == height % 8) {
3993 for (cnt = height >> 3; cnt--;) {
3994 LD_UB8(src, src_stride,
3995 src0, src1, src2, src3, src4, src5, src6, src7);
3996 src += (8 * src_stride);
3998 out0 = __msa_copy_u_d((v2i64) src0, 0);
3999 out1 = __msa_copy_u_d((v2i64) src1, 0);
4000 out2 = __msa_copy_u_d((v2i64) src2, 0);
4001 out3 = __msa_copy_u_d((v2i64) src3, 0);
4002 out4 = __msa_copy_u_d((v2i64) src4, 0);
4003 out5 = __msa_copy_u_d((v2i64) src5, 0);
4004 out6 = __msa_copy_u_d((v2i64) src6, 0);
4005 out7 = __msa_copy_u_d((v2i64) src7, 0);
4007 SD4(out0, out1, out2, out3, dst, dst_stride);
4008 dst += (4 * dst_stride);
4009 SD4(out4, out5, out6, out7, dst, dst_stride);
4010 dst += (4 * dst_stride);
4012 } else if (0 == height % 4) {
4013 for (cnt = (height / 4); cnt--;) {
4014 LD_UB4(src, src_stride, src0, src1, src2, src3);
4015 src += (4 * src_stride);
4016 out0 = __msa_copy_u_d((v2i64) src0, 0);
4017 out1 = __msa_copy_u_d((v2i64) src1, 0);
4018 out2 = __msa_copy_u_d((v2i64) src2, 0);
4019 out3 = __msa_copy_u_d((v2i64) src3, 0);
4021 SD4(out0, out1, out2, out3, dst, dst_stride);
4022 dst += (4 * dst_stride);
4024 } else if (0 == height % 2) {
4025 for (cnt = (height / 2); cnt--;) {
4026 LD_UB2(src, src_stride, src0, src1);
4027 src += (2 * src_stride);
4028 out0 = __msa_copy_u_d((v2i64) src0, 0);
4029 out1 = __msa_copy_u_d((v2i64) src1, 0);
4039 static void copy_16multx8mult_msa(const uint8_t *src, int32_t src_stride,
4040 uint8_t *dst, int32_t dst_stride,
4041 int32_t height, int32_t width)
4043 int32_t cnt, loop_cnt;
4044 const uint8_t *src_tmp;
4046 v16u8 src0, src1, src2, src3, src4, src5, src6, src7;
4048 for (cnt = (width >> 4); cnt--;) {
4052 for (loop_cnt = (height >> 3); loop_cnt--;) {
4053 LD_UB8(src_tmp, src_stride,
4054 src0, src1, src2, src3, src4, src5, src6, src7);
4055 src_tmp += (8 * src_stride);
4057 ST_UB8(src0, src1, src2, src3, src4, src5, src6, src7,
4058 dst_tmp, dst_stride);
4059 dst_tmp += (8 * dst_stride);
4067 static void copy_width16_msa(const uint8_t *src, int32_t src_stride,
4068 uint8_t *dst, int32_t dst_stride,
4072 v16u8 src0, src1, src2, src3, src4, src5, src6, src7;
4074 if (0 == height % 12) {
4075 for (cnt = (height / 12); cnt--;) {
4076 LD_UB8(src, src_stride,
4077 src0, src1, src2, src3, src4, src5, src6, src7);
4078 src += (8 * src_stride);
4079 ST_UB8(src0, src1, src2, src3, src4, src5, src6, src7,
4081 dst += (8 * dst_stride);
4083 LD_UB4(src, src_stride, src0, src1, src2, src3);
4084 src += (4 * src_stride);
4085 ST_UB4(src0, src1, src2, src3, dst, dst_stride);
4086 dst += (4 * dst_stride);
4088 } else if (0 == height % 8) {
4089 copy_16multx8mult_msa(src, src_stride, dst, dst_stride, height, 16);
4090 } else if (0 == height % 4) {
4091 for (cnt = (height >> 2); cnt--;) {
4092 LD_UB4(src, src_stride, src0, src1, src2, src3);
4093 src += (4 * src_stride);
4095 ST_UB4(src0, src1, src2, src3, dst, dst_stride);
4096 dst += (4 * dst_stride);
4101 static void copy_width32_msa(const uint8_t *src, int32_t src_stride,
4102 uint8_t *dst, int32_t dst_stride,
4106 v16u8 src0, src1, src2, src3, src4, src5, src6, src7;
4108 if (0 == height % 12) {
4109 for (cnt = (height / 12); cnt--;) {
4110 LD_UB4(src, src_stride, src0, src1, src2, src3);
4111 LD_UB4(src + 16, src_stride, src4, src5, src6, src7);
4112 src += (4 * src_stride);
4113 ST_UB4(src0, src1, src2, src3, dst, dst_stride);
4114 ST_UB4(src4, src5, src6, src7, dst + 16, dst_stride);
4115 dst += (4 * dst_stride);
4117 LD_UB4(src, src_stride, src0, src1, src2, src3);
4118 LD_UB4(src + 16, src_stride, src4, src5, src6, src7);
4119 src += (4 * src_stride);
4120 ST_UB4(src0, src1, src2, src3, dst, dst_stride);
4121 ST_UB4(src4, src5, src6, src7, dst + 16, dst_stride);
4122 dst += (4 * dst_stride);
4124 LD_UB4(src, src_stride, src0, src1, src2, src3);
4125 LD_UB4(src + 16, src_stride, src4, src5, src6, src7);
4126 src += (4 * src_stride);
4127 ST_UB4(src0, src1, src2, src3, dst, dst_stride);
4128 ST_UB4(src4, src5, src6, src7, dst + 16, dst_stride);
4129 dst += (4 * dst_stride);
4131 } else if (0 == height % 8) {
4132 copy_16multx8mult_msa(src, src_stride, dst, dst_stride, height, 32);
4133 } else if (0 == height % 4) {
4134 for (cnt = (height >> 2); cnt--;) {
4135 LD_UB4(src, src_stride, src0, src1, src2, src3);
4136 LD_UB4(src + 16, src_stride, src4, src5, src6, src7);
4137 src += (4 * src_stride);
4138 ST_UB4(src0, src1, src2, src3, dst, dst_stride);
4139 ST_UB4(src4, src5, src6, src7, dst + 16, dst_stride);
4140 dst += (4 * dst_stride);
4145 static void copy_width64_msa(const uint8_t *src, int32_t src_stride,
4146 uint8_t *dst, int32_t dst_stride,
4149 copy_16multx8mult_msa(src, src_stride, dst, dst_stride, height, 64);
4152 static void avg_width4_msa(const uint8_t *src, int32_t src_stride,
4153 uint8_t *dst, int32_t dst_stride,
4157 uint32_t out0, out1, out2, out3;
4158 v16u8 src0, src1, src2, src3;
4159 v16u8 dst0, dst1, dst2, dst3;
4161 if (0 == (height % 4)) {
4162 for (cnt = (height / 4); cnt--;) {
4163 LD_UB4(src, src_stride, src0, src1, src2, src3);
4164 src += (4 * src_stride);
4166 LD_UB4(dst, dst_stride, dst0, dst1, dst2, dst3);
4168 AVER_UB4_UB(src0, dst0, src1, dst1, src2, dst2, src3, dst3,
4169 dst0, dst1, dst2, dst3);
4171 out0 = __msa_copy_u_w((v4i32) dst0, 0);
4172 out1 = __msa_copy_u_w((v4i32) dst1, 0);
4173 out2 = __msa_copy_u_w((v4i32) dst2, 0);
4174 out3 = __msa_copy_u_w((v4i32) dst3, 0);
4175 SW4(out0, out1, out2, out3, dst, dst_stride);
4176 dst += (4 * dst_stride);
4178 } else if (0 == (height % 2)) {
4179 for (cnt = (height / 2); cnt--;) {
4180 LD_UB2(src, src_stride, src0, src1);
4181 src += (2 * src_stride);
4183 LD_UB2(dst, dst_stride, dst0, dst1);
4185 AVER_UB2_UB(src0, dst0, src1, dst1, dst0, dst1);
4187 out0 = __msa_copy_u_w((v4i32) dst0, 0);
4188 out1 = __msa_copy_u_w((v4i32) dst1, 0);
4197 static void avg_width8_msa(const uint8_t *src, int32_t src_stride,
4198 uint8_t *dst, int32_t dst_stride,
4202 uint64_t out0, out1, out2, out3;
4203 v16u8 src0, src1, src2, src3;
4204 v16u8 dst0, dst1, dst2, dst3;
4206 for (cnt = (height / 4); cnt--;) {
4207 LD_UB4(src, src_stride, src0, src1, src2, src3);
4208 src += (4 * src_stride);
4209 LD_UB4(dst, dst_stride, dst0, dst1, dst2, dst3);
4211 AVER_UB4_UB(src0, dst0, src1, dst1, src2, dst2, src3, dst3,
4212 dst0, dst1, dst2, dst3);
4214 out0 = __msa_copy_u_d((v2i64) dst0, 0);
4215 out1 = __msa_copy_u_d((v2i64) dst1, 0);
4216 out2 = __msa_copy_u_d((v2i64) dst2, 0);
4217 out3 = __msa_copy_u_d((v2i64) dst3, 0);
4218 SD4(out0, out1, out2, out3, dst, dst_stride);
4219 dst += (4 * dst_stride);
4223 static void avg_width16_msa(const uint8_t *src, int32_t src_stride,
4224 uint8_t *dst, int32_t dst_stride,
4228 v16u8 src0, src1, src2, src3, src4, src5, src6, src7;
4229 v16u8 dst0, dst1, dst2, dst3, dst4, dst5, dst6, dst7;
4231 for (cnt = (height / 8); cnt--;) {
4232 LD_UB8(src, src_stride, src0, src1, src2, src3, src4, src5, src6, src7);
4233 src += (8 * src_stride);
4234 LD_UB8(dst, dst_stride, dst0, dst1, dst2, dst3, dst4, dst5, dst6, dst7);
4236 AVER_UB4_UB(src0, dst0, src1, dst1, src2, dst2, src3, dst3,
4237 dst0, dst1, dst2, dst3);
4238 AVER_UB4_UB(src4, dst4, src5, dst5, src6, dst6, src7, dst7,
4239 dst4, dst5, dst6, dst7);
4240 ST_UB8(dst0, dst1, dst2, dst3, dst4, dst5, dst6, dst7, dst, dst_stride);
4241 dst += (8 * dst_stride);
4245 static void avg_width32_msa(const uint8_t *src, int32_t src_stride,
4246 uint8_t *dst, int32_t dst_stride,
4250 uint8_t *dst_dup = dst;
4251 v16u8 src0, src1, src2, src3, src4, src5, src6, src7;
4252 v16u8 src8, src9, src10, src11, src12, src13, src14, src15;
4253 v16u8 dst0, dst1, dst2, dst3, dst4, dst5, dst6, dst7;
4254 v16u8 dst8, dst9, dst10, dst11, dst12, dst13, dst14, dst15;
4256 for (cnt = (height / 8); cnt--;) {
4257 LD_UB4(src, src_stride, src0, src2, src4, src6);
4258 LD_UB4(src + 16, src_stride, src1, src3, src5, src7);
4259 src += (4 * src_stride);
4260 LD_UB4(dst_dup, dst_stride, dst0, dst2, dst4, dst6);
4261 LD_UB4(dst_dup + 16, dst_stride, dst1, dst3, dst5, dst7);
4262 dst_dup += (4 * dst_stride);
4263 LD_UB4(src, src_stride, src8, src10, src12, src14);
4264 LD_UB4(src + 16, src_stride, src9, src11, src13, src15);
4265 src += (4 * src_stride);
4266 LD_UB4(dst_dup, dst_stride, dst8, dst10, dst12, dst14);
4267 LD_UB4(dst_dup + 16, dst_stride, dst9, dst11, dst13, dst15);
4268 dst_dup += (4 * dst_stride);
4270 AVER_UB4_UB(src0, dst0, src1, dst1, src2, dst2, src3, dst3,
4271 dst0, dst1, dst2, dst3);
4272 AVER_UB4_UB(src4, dst4, src5, dst5, src6, dst6, src7, dst7,
4273 dst4, dst5, dst6, dst7);
4274 AVER_UB4_UB(src8, dst8, src9, dst9, src10, dst10, src11, dst11,
4275 dst8, dst9, dst10, dst11);
4276 AVER_UB4_UB(src12, dst12, src13, dst13, src14, dst14, src15, dst15,
4277 dst12, dst13, dst14, dst15);
4279 ST_UB4(dst0, dst2, dst4, dst6, dst, dst_stride);
4280 ST_UB4(dst1, dst3, dst5, dst7, dst + 16, dst_stride);
4281 dst += (4 * dst_stride);
4282 ST_UB4(dst8, dst10, dst12, dst14, dst, dst_stride);
4283 ST_UB4(dst9, dst11, dst13, dst15, dst + 16, dst_stride);
4284 dst += (4 * dst_stride);
4288 static void avg_width64_msa(const uint8_t *src, int32_t src_stride,
4289 uint8_t *dst, int32_t dst_stride,
4293 uint8_t *dst_dup = dst;
4294 v16u8 src0, src1, src2, src3, src4, src5, src6, src7;
4295 v16u8 src8, src9, src10, src11, src12, src13, src14, src15;
4296 v16u8 dst0, dst1, dst2, dst3, dst4, dst5, dst6, dst7;
4297 v16u8 dst8, dst9, dst10, dst11, dst12, dst13, dst14, dst15;
4299 for (cnt = (height / 4); cnt--;) {
4300 LD_UB4(src, 16, src0, src1, src2, src3);
4302 LD_UB4(src, 16, src4, src5, src6, src7);
4304 LD_UB4(src, 16, src8, src9, src10, src11);
4306 LD_UB4(src, 16, src12, src13, src14, src15);
4309 LD_UB4(dst_dup, 16, dst0, dst1, dst2, dst3);
4310 dst_dup += dst_stride;
4311 LD_UB4(dst_dup, 16, dst4, dst5, dst6, dst7);
4312 dst_dup += dst_stride;
4313 LD_UB4(dst_dup, 16, dst8, dst9, dst10, dst11);
4314 dst_dup += dst_stride;
4315 LD_UB4(dst_dup, 16, dst12, dst13, dst14, dst15);
4316 dst_dup += dst_stride;
4318 AVER_UB4_UB(src0, dst0, src1, dst1, src2, dst2, src3, dst3,
4319 dst0, dst1, dst2, dst3);
4320 AVER_UB4_UB(src4, dst4, src5, dst5, src6, dst6, src7, dst7,
4321 dst4, dst5, dst6, dst7);
4322 AVER_UB4_UB(src8, dst8, src9, dst9, src10, dst10, src11, dst11,
4323 dst8, dst9, dst10, dst11);
4324 AVER_UB4_UB(src12, dst12, src13, dst13, src14, dst14, src15, dst15,
4325 dst12, dst13, dst14, dst15);
4327 ST_UB4(dst0, dst1, dst2, dst3, dst, 16);
4329 ST_UB4(dst4, dst5, dst6, dst7, dst, 16);
4331 ST_UB4(dst8, dst9, dst10, dst11, dst, 16);
4333 ST_UB4(dst12, dst13, dst14, dst15, dst, 16);
4338 static const int8_t vp9_subpel_filters_msa[3][15][8] = {
4339 [FILTER_8TAP_REGULAR] = {
4340 {0, 1, -5, 126, 8, -3, 1, 0},
4341 {-1, 3, -10, 122, 18, -6, 2, 0},
4342 {-1, 4, -13, 118, 27, -9, 3, -1},
4343 {-1, 4, -16, 112, 37, -11, 4, -1},
4344 {-1, 5, -18, 105, 48, -14, 4, -1},
4345 {-1, 5, -19, 97, 58, -16, 5, -1},
4346 {-1, 6, -19, 88, 68, -18, 5, -1},
4347 {-1, 6, -19, 78, 78, -19, 6, -1},
4348 {-1, 5, -18, 68, 88, -19, 6, -1},
4349 {-1, 5, -16, 58, 97, -19, 5, -1},
4350 {-1, 4, -14, 48, 105, -18, 5, -1},
4351 {-1, 4, -11, 37, 112, -16, 4, -1},
4352 {-1, 3, -9, 27, 118, -13, 4, -1},
4353 {0, 2, -6, 18, 122, -10, 3, -1},
4354 {0, 1, -3, 8, 126, -5, 1, 0},
4355 }, [FILTER_8TAP_SHARP] = {
4356 {-1, 3, -7, 127, 8, -3, 1, 0},
4357 {-2, 5, -13, 125, 17, -6, 3, -1},
4358 {-3, 7, -17, 121, 27, -10, 5, -2},
4359 {-4, 9, -20, 115, 37, -13, 6, -2},
4360 {-4, 10, -23, 108, 48, -16, 8, -3},
4361 {-4, 10, -24, 100, 59, -19, 9, -3},
4362 {-4, 11, -24, 90, 70, -21, 10, -4},
4363 {-4, 11, -23, 80, 80, -23, 11, -4},
4364 {-4, 10, -21, 70, 90, -24, 11, -4},
4365 {-3, 9, -19, 59, 100, -24, 10, -4},
4366 {-3, 8, -16, 48, 108, -23, 10, -4},
4367 {-2, 6, -13, 37, 115, -20, 9, -4},
4368 {-2, 5, -10, 27, 121, -17, 7, -3},
4369 {-1, 3, -6, 17, 125, -13, 5, -2},
4370 {0, 1, -3, 8, 127, -7, 3, -1},
4371 }, [FILTER_8TAP_SMOOTH] = {
4372 {-3, -1, 32, 64, 38, 1, -3, 0},
4373 {-2, -2, 29, 63, 41, 2, -3, 0},
4374 {-2, -2, 26, 63, 43, 4, -4, 0},
4375 {-2, -3, 24, 62, 46, 5, -4, 0},
4376 {-2, -3, 21, 60, 49, 7, -4, 0},
4377 {-1, -4, 18, 59, 51, 9, -4, 0},
4378 {-1, -4, 16, 57, 53, 12, -4, -1},
4379 {-1, -4, 14, 55, 55, 14, -4, -1},
4380 {-1, -4, 12, 53, 57, 16, -4, -1},
4381 {0, -4, 9, 51, 59, 18, -4, -1},
4382 {0, -4, 7, 49, 60, 21, -3, -2},
4383 {0, -4, 5, 46, 62, 24, -3, -2},
4384 {0, -4, 4, 43, 63, 26, -2, -2},
4385 {0, -3, 2, 41, 63, 29, -2, -2},
4386 {0, -3, 1, 38, 64, 32, -1, -3},
4390 #define VP9_8TAP_MIPS_MSA_FUNC(SIZE, type, type_idx) \
4391 void ff_put_8tap_##type##_##SIZE##h_msa(uint8_t *dst, ptrdiff_t dststride, \
4392 const uint8_t *src, \
4393 ptrdiff_t srcstride, \
4394 int h, int mx, int my) \
4396 const int8_t *filter = vp9_subpel_filters_msa[type_idx][mx-1]; \
4398 common_hz_8t_##SIZE##w_msa(src, srcstride, dst, dststride, filter, h); \
4401 void ff_put_8tap_##type##_##SIZE##v_msa(uint8_t *dst, ptrdiff_t dststride, \
4402 const uint8_t *src, \
4403 ptrdiff_t srcstride, \
4404 int h, int mx, int my) \
4406 const int8_t *filter = vp9_subpel_filters_msa[type_idx][my-1]; \
4408 common_vt_8t_##SIZE##w_msa(src, srcstride, dst, dststride, filter, h); \
4411 void ff_put_8tap_##type##_##SIZE##hv_msa(uint8_t *dst, ptrdiff_t dststride, \
4412 const uint8_t *src, \
4413 ptrdiff_t srcstride, \
4414 int h, int mx, int my) \
4416 const uint8_t *hfilter = vp9_subpel_filters_msa[type_idx][mx-1]; \
4417 const uint8_t *vfilter = vp9_subpel_filters_msa[type_idx][my-1]; \
4419 common_hv_8ht_8vt_##SIZE##w_msa(src, srcstride, dst, dststride, hfilter, \
4423 void ff_avg_8tap_##type##_##SIZE##h_msa(uint8_t *dst, ptrdiff_t dststride, \
4424 const uint8_t *src, \
4425 ptrdiff_t srcstride, \
4426 int h, int mx, int my) \
4428 const int8_t *filter = vp9_subpel_filters_msa[type_idx][mx-1]; \
4430 common_hz_8t_and_aver_dst_##SIZE##w_msa(src, srcstride, dst, \
4431 dststride, filter, h); \
4434 void ff_avg_8tap_##type##_##SIZE##v_msa(uint8_t *dst, ptrdiff_t dststride, \
4435 const uint8_t *src, \
4436 ptrdiff_t srcstride, \
4437 int h, int mx, int my) \
4439 const int8_t *filter = vp9_subpel_filters_msa[type_idx][my-1]; \
4441 common_vt_8t_and_aver_dst_##SIZE##w_msa(src, srcstride, dst, dststride, \
4445 void ff_avg_8tap_##type##_##SIZE##hv_msa(uint8_t *dst, ptrdiff_t dststride, \
4446 const uint8_t *src, \
4447 ptrdiff_t srcstride, \
4448 int h, int mx, int my) \
4450 const uint8_t *hfilter = vp9_subpel_filters_msa[type_idx][mx-1]; \
4451 const uint8_t *vfilter = vp9_subpel_filters_msa[type_idx][my-1]; \
4453 common_hv_8ht_8vt_and_aver_dst_##SIZE##w_msa(src, srcstride, dst, \
4454 dststride, hfilter, \
4458 #define VP9_COPY_AVG_MIPS_MSA_FUNC(SIZE) \
4459 void ff_copy##SIZE##_msa(uint8_t *dst, ptrdiff_t dststride, \
4460 const uint8_t *src, ptrdiff_t srcstride, \
4461 int h, int mx, int my) \
4464 copy_width##SIZE##_msa(src, srcstride, dst, dststride, h); \
4467 void ff_avg##SIZE##_msa(uint8_t *dst, ptrdiff_t dststride, \
4468 const uint8_t *src, ptrdiff_t srcstride, \
4469 int h, int mx, int my) \
4472 avg_width##SIZE##_msa(src, srcstride, dst, dststride, h); \
4475 #define VP9_AVG_MIPS_MSA_FUNC(SIZE) \
4476 void ff_avg##SIZE##_msa(uint8_t *dst, ptrdiff_t dststride, \
4477 const uint8_t *src, ptrdiff_t srcstride, \
4478 int h, int mx, int my) \
4481 avg_width##SIZE##_msa(src, srcstride, dst, dststride, h); \
4484 VP9_8TAP_MIPS_MSA_FUNC(64, regular, FILTER_8TAP_REGULAR);
4485 VP9_8TAP_MIPS_MSA_FUNC(32, regular, FILTER_8TAP_REGULAR);
4486 VP9_8TAP_MIPS_MSA_FUNC(16, regular, FILTER_8TAP_REGULAR);
4487 VP9_8TAP_MIPS_MSA_FUNC(8, regular, FILTER_8TAP_REGULAR);
4488 VP9_8TAP_MIPS_MSA_FUNC(4, regular, FILTER_8TAP_REGULAR);
4490 VP9_8TAP_MIPS_MSA_FUNC(64, sharp, FILTER_8TAP_SHARP);
4491 VP9_8TAP_MIPS_MSA_FUNC(32, sharp, FILTER_8TAP_SHARP);
4492 VP9_8TAP_MIPS_MSA_FUNC(16, sharp, FILTER_8TAP_SHARP);
4493 VP9_8TAP_MIPS_MSA_FUNC(8, sharp, FILTER_8TAP_SHARP);
4494 VP9_8TAP_MIPS_MSA_FUNC(4, sharp, FILTER_8TAP_SHARP);
4496 VP9_8TAP_MIPS_MSA_FUNC(64, smooth, FILTER_8TAP_SMOOTH);
4497 VP9_8TAP_MIPS_MSA_FUNC(32, smooth, FILTER_8TAP_SMOOTH);
4498 VP9_8TAP_MIPS_MSA_FUNC(16, smooth, FILTER_8TAP_SMOOTH);
4499 VP9_8TAP_MIPS_MSA_FUNC(8, smooth, FILTER_8TAP_SMOOTH);
4500 VP9_8TAP_MIPS_MSA_FUNC(4, smooth, FILTER_8TAP_SMOOTH);
4502 VP9_COPY_AVG_MIPS_MSA_FUNC(64);
4503 VP9_COPY_AVG_MIPS_MSA_FUNC(32);
4504 VP9_COPY_AVG_MIPS_MSA_FUNC(16);
4505 VP9_COPY_AVG_MIPS_MSA_FUNC(8);
4506 VP9_AVG_MIPS_MSA_FUNC(4);
4508 #undef VP9_8TAP_MIPS_MSA_FUNC
4509 #undef VP9_COPY_AVG_MIPS_MSA_FUNC
4510 #undef VP9_AVG_MIPS_MSA_FUNC