?  search:
summary | log | graphiclog1 | graphiclog2 | commit | commitdiff | tree | refs | edit | fork
blame | history | raw | HEAD
Move some variable declarations to inside of loops.
[ffmpeg-lucabe.git] / libavcodec / ppc / h264_altivec.c
blob29a9c7315ab36686003654e0595dc6b38a85cab8
1 /*
2 * Copyright (c) 2004 Romain Dolbeau <romain@dolbeau.org>
3 *
4 * This file is part of FFmpeg.
5 *
6 * FFmpeg is free software; you can redistribute it and/or
7 * modify it under the terms of the GNU Lesser General Public
8 * License as published by the Free Software Foundation; either
9 * version 2.1 of the License, or (at your option) any later version.
10 *
11 * FFmpeg is distributed in the hope that it will be useful,
12 * but WITHOUT ANY WARRANTY; without even the implied warranty of
13 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
14 * Lesser General Public License for more details.
15 *
16 * You should have received a copy of the GNU Lesser General Public
17 * License along with FFmpeg; if not, write to the Free Software
18 * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
19 */
20
21 #include "libavcodec/dsputil.h"
22 #include "libavcodec/h264data.h"
23
24 #include "dsputil_ppc.h"
25 #include "dsputil_altivec.h"
26 #include "util_altivec.h"
27 #include "types_altivec.h"
28
29 #define PUT_OP_U8_ALTIVEC(d, s, dst) d = s
30 #define AVG_OP_U8_ALTIVEC(d, s, dst) d = vec_avg(dst, s)
31
32 #define OP_U8_ALTIVEC PUT_OP_U8_ALTIVEC
33 #define PREFIX_h264_chroma_mc8_altivec put_h264_chroma_mc8_altivec
34 #define PREFIX_no_rnd_vc1_chroma_mc8_altivec put_no_rnd_vc1_chroma_mc8_altivec
35 #define PREFIX_h264_chroma_mc8_num altivec_put_h264_chroma_mc8_num
36 #define PREFIX_h264_qpel16_h_lowpass_altivec put_h264_qpel16_h_lowpass_altivec
37 #define PREFIX_h264_qpel16_h_lowpass_num altivec_put_h264_qpel16_h_lowpass_num
38 #define PREFIX_h264_qpel16_v_lowpass_altivec put_h264_qpel16_v_lowpass_altivec
39 #define PREFIX_h264_qpel16_v_lowpass_num altivec_put_h264_qpel16_v_lowpass_num
40 #define PREFIX_h264_qpel16_hv_lowpass_altivec put_h264_qpel16_hv_lowpass_altivec
41 #define PREFIX_h264_qpel16_hv_lowpass_num altivec_put_h264_qpel16_hv_lowpass_num
42 #include "h264_template_altivec.c"
43 #undef OP_U8_ALTIVEC
44 #undef PREFIX_h264_chroma_mc8_altivec
45 #undef PREFIX_no_rnd_vc1_chroma_mc8_altivec
46 #undef PREFIX_h264_chroma_mc8_num
47 #undef PREFIX_h264_qpel16_h_lowpass_altivec
48 #undef PREFIX_h264_qpel16_h_lowpass_num
49 #undef PREFIX_h264_qpel16_v_lowpass_altivec
50 #undef PREFIX_h264_qpel16_v_lowpass_num
51 #undef PREFIX_h264_qpel16_hv_lowpass_altivec
52 #undef PREFIX_h264_qpel16_hv_lowpass_num
53
54 #define OP_U8_ALTIVEC AVG_OP_U8_ALTIVEC
55 #define PREFIX_h264_chroma_mc8_altivec avg_h264_chroma_mc8_altivec
56 #define PREFIX_no_rnd_vc1_chroma_mc8_altivec avg_no_rnd_vc1_chroma_mc8_altivec
57 #define PREFIX_h264_chroma_mc8_num altivec_avg_h264_chroma_mc8_num
58 #define PREFIX_h264_qpel16_h_lowpass_altivec avg_h264_qpel16_h_lowpass_altivec
59 #define PREFIX_h264_qpel16_h_lowpass_num altivec_avg_h264_qpel16_h_lowpass_num
60 #define PREFIX_h264_qpel16_v_lowpass_altivec avg_h264_qpel16_v_lowpass_altivec
61 #define PREFIX_h264_qpel16_v_lowpass_num altivec_avg_h264_qpel16_v_lowpass_num
62 #define PREFIX_h264_qpel16_hv_lowpass_altivec avg_h264_qpel16_hv_lowpass_altivec
63 #define PREFIX_h264_qpel16_hv_lowpass_num altivec_avg_h264_qpel16_hv_lowpass_num
64 #include "h264_template_altivec.c"
65 #undef OP_U8_ALTIVEC
66 #undef PREFIX_h264_chroma_mc8_altivec
67 #undef PREFIX_no_rnd_vc1_chroma_mc8_altivec
68 #undef PREFIX_h264_chroma_mc8_num
69 #undef PREFIX_h264_qpel16_h_lowpass_altivec
70 #undef PREFIX_h264_qpel16_h_lowpass_num
71 #undef PREFIX_h264_qpel16_v_lowpass_altivec
72 #undef PREFIX_h264_qpel16_v_lowpass_num
73 #undef PREFIX_h264_qpel16_hv_lowpass_altivec
74 #undef PREFIX_h264_qpel16_hv_lowpass_num
75
76 #define H264_MC(OPNAME, SIZE, CODETYPE) \
77 static void OPNAME ## h264_qpel ## SIZE ## _mc00_ ## CODETYPE (uint8_t *dst, uint8_t *src, int stride){\
78 OPNAME ## pixels ## SIZE ## _ ## CODETYPE(dst, src, stride, SIZE);\
79 }\
80 \
81 static void OPNAME ## h264_qpel ## SIZE ## _mc10_ ## CODETYPE(uint8_t *dst, uint8_t *src, int stride){ \
82 DECLARE_ALIGNED_16(uint8_t, half[SIZE*SIZE]);\
83 put_h264_qpel ## SIZE ## _h_lowpass_ ## CODETYPE(half, src, SIZE, stride);\
84 OPNAME ## pixels ## SIZE ## _l2_ ## CODETYPE(dst, src, half, stride, stride, SIZE);\
85 }\
86 \
87 static void OPNAME ## h264_qpel ## SIZE ## _mc20_ ## CODETYPE(uint8_t *dst, uint8_t *src, int stride){\
88 OPNAME ## h264_qpel ## SIZE ## _h_lowpass_ ## CODETYPE(dst, src, stride, stride);\
89 }\
90 \
91 static void OPNAME ## h264_qpel ## SIZE ## _mc30_ ## CODETYPE(uint8_t *dst, uint8_t *src, int stride){\
92 DECLARE_ALIGNED_16(uint8_t, half[SIZE*SIZE]);\
93 put_h264_qpel ## SIZE ## _h_lowpass_ ## CODETYPE(half, src, SIZE, stride);\
94 OPNAME ## pixels ## SIZE ## _l2_ ## CODETYPE(dst, src+1, half, stride, stride, SIZE);\
95 }\
96 \
97 static void OPNAME ## h264_qpel ## SIZE ## _mc01_ ## CODETYPE(uint8_t *dst, uint8_t *src, int stride){\
98 DECLARE_ALIGNED_16(uint8_t, half[SIZE*SIZE]);\
99 put_h264_qpel ## SIZE ## _v_lowpass_ ## CODETYPE(half, src, SIZE, stride);\
100 OPNAME ## pixels ## SIZE ## _l2_ ## CODETYPE(dst, src, half, stride, stride, SIZE);\
101 }\
102 \
103 static void OPNAME ## h264_qpel ## SIZE ## _mc02_ ## CODETYPE(uint8_t *dst, uint8_t *src, int stride){\
104 OPNAME ## h264_qpel ## SIZE ## _v_lowpass_ ## CODETYPE(dst, src, stride, stride);\
105 }\
106 \
107 static void OPNAME ## h264_qpel ## SIZE ## _mc03_ ## CODETYPE(uint8_t *dst, uint8_t *src, int stride){\
108 DECLARE_ALIGNED_16(uint8_t, half[SIZE*SIZE]);\
109 put_h264_qpel ## SIZE ## _v_lowpass_ ## CODETYPE(half, src, SIZE, stride);\
110 OPNAME ## pixels ## SIZE ## _l2_ ## CODETYPE(dst, src+stride, half, stride, stride, SIZE);\
111 }\
112 \
113 static void OPNAME ## h264_qpel ## SIZE ## _mc11_ ## CODETYPE(uint8_t *dst, uint8_t *src, int stride){\
114 DECLARE_ALIGNED_16(uint8_t, halfH[SIZE*SIZE]);\
115 DECLARE_ALIGNED_16(uint8_t, halfV[SIZE*SIZE]);\
116 put_h264_qpel ## SIZE ## _h_lowpass_ ## CODETYPE(halfH, src, SIZE, stride);\
117 put_h264_qpel ## SIZE ## _v_lowpass_ ## CODETYPE(halfV, src, SIZE, stride);\
118 OPNAME ## pixels ## SIZE ## _l2_ ## CODETYPE(dst, halfH, halfV, stride, SIZE, SIZE);\
119 }\
120 \
121 static void OPNAME ## h264_qpel ## SIZE ## _mc31_ ## CODETYPE(uint8_t *dst, uint8_t *src, int stride){\
122 DECLARE_ALIGNED_16(uint8_t, halfH[SIZE*SIZE]);\
123 DECLARE_ALIGNED_16(uint8_t, halfV[SIZE*SIZE]);\
124 put_h264_qpel ## SIZE ## _h_lowpass_ ## CODETYPE(halfH, src, SIZE, stride);\
125 put_h264_qpel ## SIZE ## _v_lowpass_ ## CODETYPE(halfV, src+1, SIZE, stride);\
126 OPNAME ## pixels ## SIZE ## _l2_ ## CODETYPE(dst, halfH, halfV, stride, SIZE, SIZE);\
127 }\
128 \
129 static void OPNAME ## h264_qpel ## SIZE ## _mc13_ ## CODETYPE(uint8_t *dst, uint8_t *src, int stride){\
130 DECLARE_ALIGNED_16(uint8_t, halfH[SIZE*SIZE]);\
131 DECLARE_ALIGNED_16(uint8_t, halfV[SIZE*SIZE]);\
132 put_h264_qpel ## SIZE ## _h_lowpass_ ## CODETYPE(halfH, src + stride, SIZE, stride);\
133 put_h264_qpel ## SIZE ## _v_lowpass_ ## CODETYPE(halfV, src, SIZE, stride);\
134 OPNAME ## pixels ## SIZE ## _l2_ ## CODETYPE(dst, halfH, halfV, stride, SIZE, SIZE);\
135 }\
136 \
137 static void OPNAME ## h264_qpel ## SIZE ## _mc33_ ## CODETYPE(uint8_t *dst, uint8_t *src, int stride){\
138 DECLARE_ALIGNED_16(uint8_t, halfH[SIZE*SIZE]);\
139 DECLARE_ALIGNED_16(uint8_t, halfV[SIZE*SIZE]);\
140 put_h264_qpel ## SIZE ## _h_lowpass_ ## CODETYPE(halfH, src + stride, SIZE, stride);\
141 put_h264_qpel ## SIZE ## _v_lowpass_ ## CODETYPE(halfV, src+1, SIZE, stride);\
142 OPNAME ## pixels ## SIZE ## _l2_ ## CODETYPE(dst, halfH, halfV, stride, SIZE, SIZE);\
143 }\
144 \
145 static void OPNAME ## h264_qpel ## SIZE ## _mc22_ ## CODETYPE(uint8_t *dst, uint8_t *src, int stride){\
146 DECLARE_ALIGNED_16(int16_t, tmp[SIZE*(SIZE+8)]);\
147 OPNAME ## h264_qpel ## SIZE ## _hv_lowpass_ ## CODETYPE(dst, tmp, src, stride, SIZE, stride);\
148 }\
149 \
150 static void OPNAME ## h264_qpel ## SIZE ## _mc21_ ## CODETYPE(uint8_t *dst, uint8_t *src, int stride){\
151 DECLARE_ALIGNED_16(uint8_t, halfH[SIZE*SIZE]);\
152 DECLARE_ALIGNED_16(uint8_t, halfHV[SIZE*SIZE]);\
153 DECLARE_ALIGNED_16(int16_t, tmp[SIZE*(SIZE+8)]);\
154 put_h264_qpel ## SIZE ## _h_lowpass_ ## CODETYPE(halfH, src, SIZE, stride);\
155 put_h264_qpel ## SIZE ## _hv_lowpass_ ## CODETYPE(halfHV, tmp, src, SIZE, SIZE, stride);\
156 OPNAME ## pixels ## SIZE ## _l2_ ## CODETYPE(dst, halfH, halfHV, stride, SIZE, SIZE);\
157 }\
158 \
159 static void OPNAME ## h264_qpel ## SIZE ## _mc23_ ## CODETYPE(uint8_t *dst, uint8_t *src, int stride){\
160 DECLARE_ALIGNED_16(uint8_t, halfH[SIZE*SIZE]);\
161 DECLARE_ALIGNED_16(uint8_t, halfHV[SIZE*SIZE]);\
162 DECLARE_ALIGNED_16(int16_t, tmp[SIZE*(SIZE+8)]);\
163 put_h264_qpel ## SIZE ## _h_lowpass_ ## CODETYPE(halfH, src + stride, SIZE, stride);\
164 put_h264_qpel ## SIZE ## _hv_lowpass_ ## CODETYPE(halfHV, tmp, src, SIZE, SIZE, stride);\
165 OPNAME ## pixels ## SIZE ## _l2_ ## CODETYPE(dst, halfH, halfHV, stride, SIZE, SIZE);\
166 }\
167 \
168 static void OPNAME ## h264_qpel ## SIZE ## _mc12_ ## CODETYPE(uint8_t *dst, uint8_t *src, int stride){\
169 DECLARE_ALIGNED_16(uint8_t, halfV[SIZE*SIZE]);\
170 DECLARE_ALIGNED_16(uint8_t, halfHV[SIZE*SIZE]);\
171 DECLARE_ALIGNED_16(int16_t, tmp[SIZE*(SIZE+8)]);\
172 put_h264_qpel ## SIZE ## _v_lowpass_ ## CODETYPE(halfV, src, SIZE, stride);\
173 put_h264_qpel ## SIZE ## _hv_lowpass_ ## CODETYPE(halfHV, tmp, src, SIZE, SIZE, stride);\
174 OPNAME ## pixels ## SIZE ## _l2_ ## CODETYPE(dst, halfV, halfHV, stride, SIZE, SIZE);\
175 }\
176 \
177 static void OPNAME ## h264_qpel ## SIZE ## _mc32_ ## CODETYPE(uint8_t *dst, uint8_t *src, int stride){\
178 DECLARE_ALIGNED_16(uint8_t, halfV[SIZE*SIZE]);\
179 DECLARE_ALIGNED_16(uint8_t, halfHV[SIZE*SIZE]);\
180 DECLARE_ALIGNED_16(int16_t, tmp[SIZE*(SIZE+8)]);\
181 put_h264_qpel ## SIZE ## _v_lowpass_ ## CODETYPE(halfV, src+1, SIZE, stride);\
182 put_h264_qpel ## SIZE ## _hv_lowpass_ ## CODETYPE(halfHV, tmp, src, SIZE, SIZE, stride);\
183 OPNAME ## pixels ## SIZE ## _l2_ ## CODETYPE(dst, halfV, halfHV, stride, SIZE, SIZE);\
184 }\
185
186 static inline void put_pixels16_l2_altivec( uint8_t * dst, const uint8_t * src1,
187 const uint8_t * src2, int dst_stride,
188 int src_stride1, int h)
189 {
190 int i;
191 vec_u8 a, b, d, tmp1, tmp2, mask, mask_, edges, align;
192
193 mask_ = vec_lvsl(0, src2);
194
195 for (i = 0; i < h; i++) {
196
197 tmp1 = vec_ld(i * src_stride1, src1);
198 mask = vec_lvsl(i * src_stride1, src1);
199 tmp2 = vec_ld(i * src_stride1 + 15, src1);
200
201 a = vec_perm(tmp1, tmp2, mask);
202
203 tmp1 = vec_ld(i * 16, src2);
204 tmp2 = vec_ld(i * 16 + 15, src2);
205
206 b = vec_perm(tmp1, tmp2, mask_);
207
208 tmp1 = vec_ld(0, dst);
209 mask = vec_lvsl(0, dst);
210 tmp2 = vec_ld(15, dst);
211
212 d = vec_avg(a, b);
213
214 edges = vec_perm(tmp2, tmp1, mask);
215
216 align = vec_lvsr(0, dst);
217
218 tmp2 = vec_perm(d, edges, align);
219 tmp1 = vec_perm(edges, d, align);
220
221 vec_st(tmp2, 15, dst);
222 vec_st(tmp1, 0 , dst);
223
224 dst += dst_stride;
225 }
226 }
227
228 static inline void avg_pixels16_l2_altivec( uint8_t * dst, const uint8_t * src1,
229 const uint8_t * src2, int dst_stride,
230 int src_stride1, int h)
231 {
232 int i;
233 vec_u8 a, b, d, tmp1, tmp2, mask, mask_, edges, align;
234
235 mask_ = vec_lvsl(0, src2);
236
237 for (i = 0; i < h; i++) {
238
239 tmp1 = vec_ld(i * src_stride1, src1);
240 mask = vec_lvsl(i * src_stride1, src1);
241 tmp2 = vec_ld(i * src_stride1 + 15, src1);
242
243 a = vec_perm(tmp1, tmp2, mask);
244
245 tmp1 = vec_ld(i * 16, src2);
246 tmp2 = vec_ld(i * 16 + 15, src2);
247
248 b = vec_perm(tmp1, tmp2, mask_);
249
250 tmp1 = vec_ld(0, dst);
251 mask = vec_lvsl(0, dst);
252 tmp2 = vec_ld(15, dst);
253
254 d = vec_avg(vec_perm(tmp1, tmp2, mask), vec_avg(a, b));
255
256 edges = vec_perm(tmp2, tmp1, mask);
257
258 align = vec_lvsr(0, dst);
259
260 tmp2 = vec_perm(d, edges, align);
261 tmp1 = vec_perm(edges, d, align);
262
263 vec_st(tmp2, 15, dst);
264 vec_st(tmp1, 0 , dst);
265
266 dst += dst_stride;
267 }
268 }
269
270 /* Implemented but could be faster
271 #define put_pixels16_l2_altivec(d,s1,s2,ds,s1s,h) put_pixels16_l2(d,s1,s2,ds,s1s,16,h)
272 #define avg_pixels16_l2_altivec(d,s1,s2,ds,s1s,h) avg_pixels16_l2(d,s1,s2,ds,s1s,16,h)
273 */
274
275 H264_MC(put_, 16, altivec)
276 H264_MC(avg_, 16, altivec)
277
278
279 /****************************************************************************
280 * IDCT transform:
281 ****************************************************************************/
282
283 #define VEC_1D_DCT(vb0,vb1,vb2,vb3,va0,va1,va2,va3) \
284 /* 1st stage */ \
285 vz0 = vec_add(vb0,vb2); /* temp[0] = Y[0] + Y[2] */ \
286 vz1 = vec_sub(vb0,vb2); /* temp[1] = Y[0] - Y[2] */ \
287 vz2 = vec_sra(vb1,vec_splat_u16(1)); \
288 vz2 = vec_sub(vz2,vb3); /* temp[2] = Y[1].1/2 - Y[3] */ \
289 vz3 = vec_sra(vb3,vec_splat_u16(1)); \
290 vz3 = vec_add(vb1,vz3); /* temp[3] = Y[1] + Y[3].1/2 */ \
291 /* 2nd stage: output */ \
292 va0 = vec_add(vz0,vz3); /* x[0] = temp[0] + temp[3] */ \
293 va1 = vec_add(vz1,vz2); /* x[1] = temp[1] + temp[2] */ \
294 va2 = vec_sub(vz1,vz2); /* x[2] = temp[1] - temp[2] */ \
295 va3 = vec_sub(vz0,vz3) /* x[3] = temp[0] - temp[3] */
296
297 #define VEC_TRANSPOSE_4(a0,a1,a2,a3,b0,b1,b2,b3) \
298 b0 = vec_mergeh( a0, a0 ); \
299 b1 = vec_mergeh( a1, a0 ); \
300 b2 = vec_mergeh( a2, a0 ); \
301 b3 = vec_mergeh( a3, a0 ); \
302 a0 = vec_mergeh( b0, b2 ); \
303 a1 = vec_mergel( b0, b2 ); \
304 a2 = vec_mergeh( b1, b3 ); \
305 a3 = vec_mergel( b1, b3 ); \
306 b0 = vec_mergeh( a0, a2 ); \
307 b1 = vec_mergel( a0, a2 ); \
308 b2 = vec_mergeh( a1, a3 ); \
309 b3 = vec_mergel( a1, a3 )
310
311 #define VEC_LOAD_U8_ADD_S16_STORE_U8(va) \
312 vdst_orig = vec_ld(0, dst); \
313 vdst = vec_perm(vdst_orig, zero_u8v, vdst_mask); \
314 vdst_ss = (vec_s16) vec_mergeh(zero_u8v, vdst); \
315 va = vec_add(va, vdst_ss); \
316 va_u8 = vec_packsu(va, zero_s16v); \
317 va_u32 = vec_splat((vec_u32)va_u8, 0); \
318 vec_ste(va_u32, element, (uint32_t*)dst);
319
320 static void ff_h264_idct_add_altivec(uint8_t *dst, DCTELEM *block, int stride)
321 {
322 vec_s16 va0, va1, va2, va3;
323 vec_s16 vz0, vz1, vz2, vz3;
324 vec_s16 vtmp0, vtmp1, vtmp2, vtmp3;
325 vec_u8 va_u8;
326 vec_u32 va_u32;
327 vec_s16 vdst_ss;
328 const vec_u16 v6us = vec_splat_u16(6);
329 vec_u8 vdst, vdst_orig;
330 vec_u8 vdst_mask = vec_lvsl(0, dst);
331 int element = ((unsigned long)dst & 0xf) >> 2;
332 LOAD_ZERO;
333
334 block[0] += 32; /* add 32 as a DC-level for rounding */
335
336 vtmp0 = vec_ld(0,block);
337 vtmp1 = vec_sld(vtmp0, vtmp0, 8);
338 vtmp2 = vec_ld(16,block);
339 vtmp3 = vec_sld(vtmp2, vtmp2, 8);
340
341 VEC_1D_DCT(vtmp0,vtmp1,vtmp2,vtmp3,va0,va1,va2,va3);
342 VEC_TRANSPOSE_4(va0,va1,va2,va3,vtmp0,vtmp1,vtmp2,vtmp3);
343 VEC_1D_DCT(vtmp0,vtmp1,vtmp2,vtmp3,va0,va1,va2,va3);
344
345 va0 = vec_sra(va0,v6us);
346 va1 = vec_sra(va1,v6us);
347 va2 = vec_sra(va2,v6us);
348 va3 = vec_sra(va3,v6us);
349
350 VEC_LOAD_U8_ADD_S16_STORE_U8(va0);
351 dst += stride;
352 VEC_LOAD_U8_ADD_S16_STORE_U8(va1);
353 dst += stride;
354 VEC_LOAD_U8_ADD_S16_STORE_U8(va2);
355 dst += stride;
356 VEC_LOAD_U8_ADD_S16_STORE_U8(va3);
357 }
358
359 #define IDCT8_1D_ALTIVEC(s0, s1, s2, s3, s4, s5, s6, s7, d0, d1, d2, d3, d4, d5, d6, d7) {\
360 /* a0 = SRC(0) + SRC(4); */ \
361 vec_s16 a0v = vec_add(s0, s4); \
362 /* a2 = SRC(0) - SRC(4); */ \
363 vec_s16 a2v = vec_sub(s0, s4); \
364 /* a4 = (SRC(2)>>1) - SRC(6); */ \
365 vec_s16 a4v = vec_sub(vec_sra(s2, onev), s6); \
366 /* a6 = (SRC(6)>>1) + SRC(2); */ \
367 vec_s16 a6v = vec_add(vec_sra(s6, onev), s2); \
368 /* b0 = a0 + a6; */ \
369 vec_s16 b0v = vec_add(a0v, a6v); \
370 /* b2 = a2 + a4; */ \
371 vec_s16 b2v = vec_add(a2v, a4v); \
372 /* b4 = a2 - a4; */ \
373 vec_s16 b4v = vec_sub(a2v, a4v); \
374 /* b6 = a0 - a6; */ \
375 vec_s16 b6v = vec_sub(a0v, a6v); \
376 /* a1 = SRC(5) - SRC(3) - SRC(7) - (SRC(7)>>1); */ \
377 /* a1 = (SRC(5)-SRC(3)) - (SRC(7) + (SRC(7)>>1)); */ \
378 vec_s16 a1v = vec_sub( vec_sub(s5, s3), vec_add(s7, vec_sra(s7, onev)) ); \
379 /* a3 = SRC(7) + SRC(1) - SRC(3) - (SRC(3)>>1); */ \
380 /* a3 = (SRC(7)+SRC(1)) - (SRC(3) + (SRC(3)>>1)); */ \
381 vec_s16 a3v = vec_sub( vec_add(s7, s1), vec_add(s3, vec_sra(s3, onev)) );\
382 /* a5 = SRC(7) - SRC(1) + SRC(5) + (SRC(5)>>1); */ \
383 /* a5 = (SRC(7)-SRC(1)) + SRC(5) + (SRC(5)>>1); */ \
384 vec_s16 a5v = vec_add( vec_sub(s7, s1), vec_add(s5, vec_sra(s5, onev)) );\
385 /* a7 = SRC(5)+SRC(3) + SRC(1) + (SRC(1)>>1); */ \
386 vec_s16 a7v = vec_add( vec_add(s5, s3), vec_add(s1, vec_sra(s1, onev)) );\
387 /* b1 = (a7>>2) + a1; */ \
388 vec_s16 b1v = vec_add( vec_sra(a7v, twov), a1v); \
389 /* b3 = a3 + (a5>>2); */ \
390 vec_s16 b3v = vec_add(a3v, vec_sra(a5v, twov)); \
391 /* b5 = (a3>>2) - a5; */ \
392 vec_s16 b5v = vec_sub( vec_sra(a3v, twov), a5v); \
393 /* b7 = a7 - (a1>>2); */ \
394 vec_s16 b7v = vec_sub( a7v, vec_sra(a1v, twov)); \
395 /* DST(0, b0 + b7); */ \
396 d0 = vec_add(b0v, b7v); \
397 /* DST(1, b2 + b5); */ \
398 d1 = vec_add(b2v, b5v); \
399 /* DST(2, b4 + b3); */ \
400 d2 = vec_add(b4v, b3v); \
401 /* DST(3, b6 + b1); */ \
402 d3 = vec_add(b6v, b1v); \
403 /* DST(4, b6 - b1); */ \
404 d4 = vec_sub(b6v, b1v); \
405 /* DST(5, b4 - b3); */ \
406 d5 = vec_sub(b4v, b3v); \
407 /* DST(6, b2 - b5); */ \
408 d6 = vec_sub(b2v, b5v); \
409 /* DST(7, b0 - b7); */ \
410 d7 = vec_sub(b0v, b7v); \
411 }
412
413 #define ALTIVEC_STORE_SUM_CLIP(dest, idctv, perm_ldv, perm_stv, sel) { \
414 /* unaligned load */ \
415 vec_u8 hv = vec_ld( 0, dest ); \
416 vec_u8 lv = vec_ld( 7, dest ); \
417 vec_u8 dstv = vec_perm( hv, lv, (vec_u8)perm_ldv ); \
418 vec_s16 idct_sh6 = vec_sra(idctv, sixv); \
419 vec_u16 dst16 = (vec_u16)vec_mergeh(zero_u8v, dstv); \
420 vec_s16 idstsum = vec_adds(idct_sh6, (vec_s16)dst16); \
421 vec_u8 idstsum8 = vec_packsu(zero_s16v, idstsum); \
422 vec_u8 edgehv; \
423 /* unaligned store */ \
424 vec_u8 bodyv = vec_perm( idstsum8, idstsum8, perm_stv );\
425 vec_u8 edgelv = vec_perm( sel, zero_u8v, perm_stv ); \
426 lv = vec_sel( lv, bodyv, edgelv ); \
427 vec_st( lv, 7, dest ); \
428 hv = vec_ld( 0, dest ); \
429 edgehv = vec_perm( zero_u8v, sel, perm_stv ); \
430 hv = vec_sel( hv, bodyv, edgehv ); \
431 vec_st( hv, 0, dest ); \
432 }
433
434 void ff_h264_idct8_add_altivec( uint8_t *dst, DCTELEM *dct, int stride ) {
435 vec_s16 s0, s1, s2, s3, s4, s5, s6, s7;
436 vec_s16 d0, d1, d2, d3, d4, d5, d6, d7;
437 vec_s16 idct0, idct1, idct2, idct3, idct4, idct5, idct6, idct7;
438
439 vec_u8 perm_ldv = vec_lvsl(0, dst);
440 vec_u8 perm_stv = vec_lvsr(8, dst);
441
442 const vec_u16 onev = vec_splat_u16(1);
443 const vec_u16 twov = vec_splat_u16(2);
444 const vec_u16 sixv = vec_splat_u16(6);
445
446 const vec_u8 sel = (vec_u8) {0,0,0,0,0,0,0,0,-1,-1,-1,-1,-1,-1,-1,-1};
447 LOAD_ZERO;
448
449 dct[0] += 32; // rounding for the >>6 at the end
450
451 s0 = vec_ld(0x00, (int16_t*)dct);
452 s1 = vec_ld(0x10, (int16_t*)dct);
453 s2 = vec_ld(0x20, (int16_t*)dct);
454 s3 = vec_ld(0x30, (int16_t*)dct);
455 s4 = vec_ld(0x40, (int16_t*)dct);
456 s5 = vec_ld(0x50, (int16_t*)dct);
457 s6 = vec_ld(0x60, (int16_t*)dct);
458 s7 = vec_ld(0x70, (int16_t*)dct);
459
460 IDCT8_1D_ALTIVEC(s0, s1, s2, s3, s4, s5, s6, s7,
461 d0, d1, d2, d3, d4, d5, d6, d7);
462
463 TRANSPOSE8( d0, d1, d2, d3, d4, d5, d6, d7 );
464
465 IDCT8_1D_ALTIVEC(d0, d1, d2, d3, d4, d5, d6, d7,
466 idct0, idct1, idct2, idct3, idct4, idct5, idct6, idct7);
467
468 ALTIVEC_STORE_SUM_CLIP(&dst[0*stride], idct0, perm_ldv, perm_stv, sel);
469 ALTIVEC_STORE_SUM_CLIP(&dst[1*stride], idct1, perm_ldv, perm_stv, sel);
470 ALTIVEC_STORE_SUM_CLIP(&dst[2*stride], idct2, perm_ldv, perm_stv, sel);
471 ALTIVEC_STORE_SUM_CLIP(&dst[3*stride], idct3, perm_ldv, perm_stv, sel);
472 ALTIVEC_STORE_SUM_CLIP(&dst[4*stride], idct4, perm_ldv, perm_stv, sel);
473 ALTIVEC_STORE_SUM_CLIP(&dst[5*stride], idct5, perm_ldv, perm_stv, sel);
474 ALTIVEC_STORE_SUM_CLIP(&dst[6*stride], idct6, perm_ldv, perm_stv, sel);
475 ALTIVEC_STORE_SUM_CLIP(&dst[7*stride], idct7, perm_ldv, perm_stv, sel);
476 }
477
478 static av_always_inline void h264_idct_dc_add_internal(uint8_t *dst, DCTELEM *block, int stride, int size)
479 {
480 vec_s16 dc16;
481 vec_u8 dcplus, dcminus, v0, v1, v2, v3, aligner;
482 LOAD_ZERO;
483 DECLARE_ALIGNED_16(int, dc);
484 int i;
485
486 dc = (block[0] + 32) >> 6;
487 dc16 = vec_splat((vec_s16) vec_lde(0, &dc), 1);
488
489 if (size == 4)
490 dc16 = vec_sld(dc16, zero_s16v, 8);
491 dcplus = vec_packsu(dc16, zero_s16v);
492 dcminus = vec_packsu(vec_sub(zero_s16v, dc16), zero_s16v);
493
494 aligner = vec_lvsr(0, dst);
495 dcplus = vec_perm(dcplus, dcplus, aligner);
496 dcminus = vec_perm(dcminus, dcminus, aligner);
497
498 for (i = 0; i < size; i += 4) {
499 v0 = vec_ld(0, dst+0*stride);
500 v1 = vec_ld(0, dst+1*stride);
501 v2 = vec_ld(0, dst+2*stride);
502 v3 = vec_ld(0, dst+3*stride);
503
504 v0 = vec_adds(v0, dcplus);
505 v1 = vec_adds(v1, dcplus);
506 v2 = vec_adds(v2, dcplus);
507 v3 = vec_adds(v3, dcplus);
508
509 v0 = vec_subs(v0, dcminus);
510 v1 = vec_subs(v1, dcminus);
511 v2 = vec_subs(v2, dcminus);
512 v3 = vec_subs(v3, dcminus);
513
514 vec_st(v0, 0, dst+0*stride);
515 vec_st(v1, 0, dst+1*stride);
516 vec_st(v2, 0, dst+2*stride);
517 vec_st(v3, 0, dst+3*stride);
518
519 dst += 4*stride;
520 }
521 }
522
523 static void h264_idct_dc_add_altivec(uint8_t *dst, DCTELEM *block, int stride)
524 {
525 h264_idct_dc_add_internal(dst, block, stride, 4);
526 }
527
528 static void ff_h264_idct8_dc_add_altivec(uint8_t *dst, DCTELEM *block, int stride)
529 {
530 h264_idct_dc_add_internal(dst, block, stride, 8);
531 }
532
533 static void ff_h264_idct_add16_altivec(uint8_t *dst, const int *block_offset, DCTELEM *block, int stride, const uint8_t nnzc[6*8]){
534 int i;
535 for(i=0; i<16; i++){
536 int nnz = nnzc[ scan8[i] ];
537 if(nnz){
538 if(nnz==1 && block[i*16]) h264_idct_dc_add_altivec(dst + block_offset[i], block + i*16, stride);
539 else ff_h264_idct_add_altivec(dst + block_offset[i], block + i*16, stride);
540 }
541 }
542 }
543
544 static void ff_h264_idct_add16intra_altivec(uint8_t *dst, const int *block_offset, DCTELEM *block, int stride, const uint8_t nnzc[6*8]){
545 int i;
546 for(i=0; i<16; i++){
547 if(nnzc[ scan8[i] ]) ff_h264_idct_add_altivec(dst + block_offset[i], block + i*16, stride);
548 else if(block[i*16]) h264_idct_dc_add_altivec(dst + block_offset[i], block + i*16, stride);
549 }
550 }
551
552 static void ff_h264_idct8_add4_altivec(uint8_t *dst, const int *block_offset, DCTELEM *block, int stride, const uint8_t nnzc[6*8]){
553 int i;
554 for(i=0; i<16; i+=4){
555 int nnz = nnzc[ scan8[i] ];
556 if(nnz){
557 if(nnz==1 && block[i*16]) ff_h264_idct8_dc_add_altivec(dst + block_offset[i], block + i*16, stride);
558 else ff_h264_idct8_add_altivec (dst + block_offset[i], block + i*16, stride);
559 }
560 }
561 }
562
563 static void ff_h264_idct_add8_altivec(uint8_t **dest, const int *block_offset, DCTELEM *block, int stride, const uint8_t nnzc[6*8]){
564 int i;
565 for(i=16; i<16+8; i++){
566 if(nnzc[ scan8[i] ])
567 ff_h264_idct_add_altivec(dest[(i&4)>>2] + block_offset[i], block + i*16, stride);
568 else if(block[i*16])
569 h264_idct_dc_add_altivec(dest[(i&4)>>2] + block_offset[i], block + i*16, stride);
570 }
571 }
572
573 #define transpose4x16(r0, r1, r2, r3) { \
574 register vec_u8 r4; \
575 register vec_u8 r5; \
576 register vec_u8 r6; \
577 register vec_u8 r7; \
578 \
579 r4 = vec_mergeh(r0, r2); /*0, 2 set 0*/ \
580 r5 = vec_mergel(r0, r2); /*0, 2 set 1*/ \
581 r6 = vec_mergeh(r1, r3); /*1, 3 set 0*/ \
582 r7 = vec_mergel(r1, r3); /*1, 3 set 1*/ \
583 \
584 r0 = vec_mergeh(r4, r6); /*all set 0*/ \
585 r1 = vec_mergel(r4, r6); /*all set 1*/ \
586 r2 = vec_mergeh(r5, r7); /*all set 2*/ \
587 r3 = vec_mergel(r5, r7); /*all set 3*/ \
588 }
589
590 static inline void write16x4(uint8_t *dst, int dst_stride,
591 register vec_u8 r0, register vec_u8 r1,
592 register vec_u8 r2, register vec_u8 r3) {
593 DECLARE_ALIGNED_16(unsigned char, result[64]);
594 uint32_t *src_int = (uint32_t *)result, *dst_int = (uint32_t *)dst;
595 int int_dst_stride = dst_stride/4;
596
597 vec_st(r0, 0, result);
598 vec_st(r1, 16, result);
599 vec_st(r2, 32, result);
600 vec_st(r3, 48, result);
601 /* FIXME: there has to be a better way!!!! */
602 *dst_int = *src_int;
603 *(dst_int+ int_dst_stride) = *(src_int + 1);
604 *(dst_int+ 2*int_dst_stride) = *(src_int + 2);
605 *(dst_int+ 3*int_dst_stride) = *(src_int + 3);
606 *(dst_int+ 4*int_dst_stride) = *(src_int + 4);
607 *(dst_int+ 5*int_dst_stride) = *(src_int + 5);
608 *(dst_int+ 6*int_dst_stride) = *(src_int + 6);
609 *(dst_int+ 7*int_dst_stride) = *(src_int + 7);
610 *(dst_int+ 8*int_dst_stride) = *(src_int + 8);
611 *(dst_int+ 9*int_dst_stride) = *(src_int + 9);
612 *(dst_int+10*int_dst_stride) = *(src_int + 10);
613 *(dst_int+11*int_dst_stride) = *(src_int + 11);
614 *(dst_int+12*int_dst_stride) = *(src_int + 12);
615 *(dst_int+13*int_dst_stride) = *(src_int + 13);
616 *(dst_int+14*int_dst_stride) = *(src_int + 14);
617 *(dst_int+15*int_dst_stride) = *(src_int + 15);
618 }
619
620 /** \brief performs a 6x16 transpose of data in src, and stores it to dst
621 \todo FIXME: see if we can't spare some vec_lvsl() by them factorizing
622 out of unaligned_load() */
623 #define readAndTranspose16x6(src, src_stride, r8, r9, r10, r11, r12, r13) {\
624 register vec_u8 r0 = unaligned_load(0, src); \
625 register vec_u8 r1 = unaligned_load( src_stride, src); \
626 register vec_u8 r2 = unaligned_load(2* src_stride, src); \
627 register vec_u8 r3 = unaligned_load(3* src_stride, src); \
628 register vec_u8 r4 = unaligned_load(4* src_stride, src); \
629 register vec_u8 r5 = unaligned_load(5* src_stride, src); \
630 register vec_u8 r6 = unaligned_load(6* src_stride, src); \
631 register vec_u8 r7 = unaligned_load(7* src_stride, src); \
632 register vec_u8 r14 = unaligned_load(14*src_stride, src); \
633 register vec_u8 r15 = unaligned_load(15*src_stride, src); \
634 \
635 r8 = unaligned_load( 8*src_stride, src); \
636 r9 = unaligned_load( 9*src_stride, src); \
637 r10 = unaligned_load(10*src_stride, src); \
638 r11 = unaligned_load(11*src_stride, src); \
639 r12 = unaligned_load(12*src_stride, src); \
640 r13 = unaligned_load(13*src_stride, src); \
641 \
642 /*Merge first pairs*/ \
643 r0 = vec_mergeh(r0, r8); /*0, 8*/ \
644 r1 = vec_mergeh(r1, r9); /*1, 9*/ \
645 r2 = vec_mergeh(r2, r10); /*2,10*/ \
646 r3 = vec_mergeh(r3, r11); /*3,11*/ \
647 r4 = vec_mergeh(r4, r12); /*4,12*/ \
648 r5 = vec_mergeh(r5, r13); /*5,13*/ \
649 r6 = vec_mergeh(r6, r14); /*6,14*/ \
650 r7 = vec_mergeh(r7, r15); /*7,15*/ \
651 \
652 /*Merge second pairs*/ \
653 r8 = vec_mergeh(r0, r4); /*0,4, 8,12 set 0*/ \
654 r9 = vec_mergel(r0, r4); /*0,4, 8,12 set 1*/ \
655 r10 = vec_mergeh(r1, r5); /*1,5, 9,13 set 0*/ \
656 r11 = vec_mergel(r1, r5); /*1,5, 9,13 set 1*/ \
657 r12 = vec_mergeh(r2, r6); /*2,6,10,14 set 0*/ \
658 r13 = vec_mergel(r2, r6); /*2,6,10,14 set 1*/ \
659 r14 = vec_mergeh(r3, r7); /*3,7,11,15 set 0*/ \
660 r15 = vec_mergel(r3, r7); /*3,7,11,15 set 1*/ \
661 \
662 /*Third merge*/ \
663 r0 = vec_mergeh(r8, r12); /*0,2,4,6,8,10,12,14 set 0*/ \
664 r1 = vec_mergel(r8, r12); /*0,2,4,6,8,10,12,14 set 1*/ \
665 r2 = vec_mergeh(r9, r13); /*0,2,4,6,8,10,12,14 set 2*/ \
666 r4 = vec_mergeh(r10, r14); /*1,3,5,7,9,11,13,15 set 0*/ \
667 r5 = vec_mergel(r10, r14); /*1,3,5,7,9,11,13,15 set 1*/ \
668 r6 = vec_mergeh(r11, r15); /*1,3,5,7,9,11,13,15 set 2*/ \
669 /* Don't need to compute 3 and 7*/ \
670 \
671 /*Final merge*/ \
672 r8 = vec_mergeh(r0, r4); /*all set 0*/ \
673 r9 = vec_mergel(r0, r4); /*all set 1*/ \
674 r10 = vec_mergeh(r1, r5); /*all set 2*/ \
675 r11 = vec_mergel(r1, r5); /*all set 3*/ \
676 r12 = vec_mergeh(r2, r6); /*all set 4*/ \
677 r13 = vec_mergel(r2, r6); /*all set 5*/ \
678 /* Don't need to compute 14 and 15*/ \
679 \
680 }
681
682 // out: o = |x-y| < a
683 static inline vec_u8 diff_lt_altivec ( register vec_u8 x,
684 register vec_u8 y,
685 register vec_u8 a) {
686
687 register vec_u8 diff = vec_subs(x, y);
688 register vec_u8 diffneg = vec_subs(y, x);
689 register vec_u8 o = vec_or(diff, diffneg); /* |x-y| */
690 o = (vec_u8)vec_cmplt(o, a);
691 return o;
692 }
693
694 static inline vec_u8 h264_deblock_mask ( register vec_u8 p0,
695 register vec_u8 p1,
696 register vec_u8 q0,
697 register vec_u8 q1,
698 register vec_u8 alpha,
699 register vec_u8 beta) {
700
701 register vec_u8 mask;
702 register vec_u8 tempmask;
703
704 mask = diff_lt_altivec(p0, q0, alpha);
705 tempmask = diff_lt_altivec(p1, p0, beta);
706 mask = vec_and(mask, tempmask);
707 tempmask = diff_lt_altivec(q1, q0, beta);
708 mask = vec_and(mask, tempmask);
709
710 return mask;
711 }
712
713 // out: newp1 = clip((p2 + ((p0 + q0 + 1) >> 1)) >> 1, p1-tc0, p1+tc0)
714 static inline vec_u8 h264_deblock_q1(register vec_u8 p0,
715 register vec_u8 p1,
716 register vec_u8 p2,
717 register vec_u8 q0,
718 register vec_u8 tc0) {
719
720 register vec_u8 average = vec_avg(p0, q0);
721 register vec_u8 temp;
722 register vec_u8 uncliped;
723 register vec_u8 ones;
724 register vec_u8 max;
725 register vec_u8 min;
726 register vec_u8 newp1;
727
728 temp = vec_xor(average, p2);
729 average = vec_avg(average, p2); /*avg(p2, avg(p0, q0)) */
730 ones = vec_splat_u8(1);
731 temp = vec_and(temp, ones); /*(p2^avg(p0, q0)) & 1 */
732 uncliped = vec_subs(average, temp); /*(p2+((p0+q0+1)>>1))>>1 */
733 max = vec_adds(p1, tc0);
734 min = vec_subs(p1, tc0);
735 newp1 = vec_max(min, uncliped);
736 newp1 = vec_min(max, newp1);
737 return newp1;
738 }
739
740 #define h264_deblock_p0_q0(p0, p1, q0, q1, tc0masked) { \
741 \
742 const vec_u8 A0v = vec_sl(vec_splat_u8(10), vec_splat_u8(4)); \
743 \
744 register vec_u8 pq0bit = vec_xor(p0,q0); \
745 register vec_u8 q1minus; \
746 register vec_u8 p0minus; \
747 register vec_u8 stage1; \
748 register vec_u8 stage2; \
749 register vec_u8 vec160; \
750 register vec_u8 delta; \
751 register vec_u8 deltaneg; \
752 \
753 q1minus = vec_nor(q1, q1); /* 255 - q1 */ \
754 stage1 = vec_avg(p1, q1minus); /* (p1 - q1 + 256)>>1 */ \
755 stage2 = vec_sr(stage1, vec_splat_u8(1)); /* (p1 - q1 + 256)>>2 = 64 + (p1 - q1) >> 2 */ \
756 p0minus = vec_nor(p0, p0); /* 255 - p0 */ \
757 stage1 = vec_avg(q0, p0minus); /* (q0 - p0 + 256)>>1 */ \
758 pq0bit = vec_and(pq0bit, vec_splat_u8(1)); \
759 stage2 = vec_avg(stage2, pq0bit); /* 32 + ((q0 - p0)&1 + (p1 - q1) >> 2 + 1) >> 1 */ \
760 stage2 = vec_adds(stage2, stage1); /* 160 + ((p0 - q0) + (p1 - q1) >> 2 + 1) >> 1 */ \
761 vec160 = vec_ld(0, &A0v); \
762 deltaneg = vec_subs(vec160, stage2); /* -d */ \
763 delta = vec_subs(stage2, vec160); /* d */ \
764 deltaneg = vec_min(tc0masked, deltaneg); \
765 delta = vec_min(tc0masked, delta); \
766 p0 = vec_subs(p0, deltaneg); \
767 q0 = vec_subs(q0, delta); \
768 p0 = vec_adds(p0, delta); \
769 q0 = vec_adds(q0, deltaneg); \
770 }
771
772 #define h264_loop_filter_luma_altivec(p2, p1, p0, q0, q1, q2, alpha, beta, tc0) { \
773 DECLARE_ALIGNED_16(unsigned char, temp[16]); \
774 register vec_u8 alphavec; \
775 register vec_u8 betavec; \
776 register vec_u8 mask; \
777 register vec_u8 p1mask; \
778 register vec_u8 q1mask; \
779 register vector signed char tc0vec; \
780 register vec_u8 finaltc0; \
781 register vec_u8 tc0masked; \
782 register vec_u8 newp1; \
783 register vec_u8 newq1; \
784 \
785 temp[0] = alpha; \
786 temp[1] = beta; \
787 alphavec = vec_ld(0, temp); \
788 betavec = vec_splat(alphavec, 0x1); \
789 alphavec = vec_splat(alphavec, 0x0); \
790 mask = h264_deblock_mask(p0, p1, q0, q1, alphavec, betavec); /*if in block */ \
791 \
792 *((int *)temp) = *((int *)tc0); \
793 tc0vec = vec_ld(0, (signed char*)temp); \
794 tc0vec = vec_mergeh(tc0vec, tc0vec); \
795 tc0vec = vec_mergeh(tc0vec, tc0vec); \
796 mask = vec_and(mask, vec_cmpgt(tc0vec, vec_splat_s8(-1))); /* if tc0[i] >= 0 */ \
797 finaltc0 = vec_and((vec_u8)tc0vec, mask); /* tc = tc0 */ \
798 \
799 p1mask = diff_lt_altivec(p2, p0, betavec); \
800 p1mask = vec_and(p1mask, mask); /* if ( |p2 - p0| < beta) */ \
801 tc0masked = vec_and(p1mask, (vec_u8)tc0vec); \
802 finaltc0 = vec_sub(finaltc0, p1mask); /* tc++ */ \
803 newp1 = h264_deblock_q1(p0, p1, p2, q0, tc0masked); \
804 /*end if*/ \
805 \
806 q1mask = diff_lt_altivec(q2, q0, betavec); \
807 q1mask = vec_and(q1mask, mask); /* if ( |q2 - q0| < beta ) */\
808 tc0masked = vec_and(q1mask, (vec_u8)tc0vec); \
809 finaltc0 = vec_sub(finaltc0, q1mask); /* tc++ */ \
810 newq1 = h264_deblock_q1(p0, q1, q2, q0, tc0masked); \
811 /*end if*/ \
812 \
813 h264_deblock_p0_q0(p0, p1, q0, q1, finaltc0); \
814 p1 = newp1; \
815 q1 = newq1; \
816 }
817
818 static void h264_v_loop_filter_luma_altivec(uint8_t *pix, int stride, int alpha, int beta, int8_t *tc0) {
819
820 if ((tc0[0] & tc0[1] & tc0[2] & tc0[3]) >= 0) {
821 register vec_u8 p2 = vec_ld(-3*stride, pix);
822 register vec_u8 p1 = vec_ld(-2*stride, pix);
823 register vec_u8 p0 = vec_ld(-1*stride, pix);
824 register vec_u8 q0 = vec_ld(0, pix);
825 register vec_u8 q1 = vec_ld(stride, pix);
826 register vec_u8 q2 = vec_ld(2*stride, pix);
827 h264_loop_filter_luma_altivec(p2, p1, p0, q0, q1, q2, alpha, beta, tc0);
828 vec_st(p1, -2*stride, pix);
829 vec_st(p0, -1*stride, pix);
830 vec_st(q0, 0, pix);
831 vec_st(q1, stride, pix);
832 }
833 }
834
835 static void h264_h_loop_filter_luma_altivec(uint8_t *pix, int stride, int alpha, int beta, int8_t *tc0) {
836
837 register vec_u8 line0, line1, line2, line3, line4, line5;
838 if ((tc0[0] & tc0[1] & tc0[2] & tc0[3]) < 0)
839 return;
840 readAndTranspose16x6(pix-3, stride, line0, line1, line2, line3, line4, line5);
841 h264_loop_filter_luma_altivec(line0, line1, line2, line3, line4, line5, alpha, beta, tc0);
842 transpose4x16(line1, line2, line3, line4);
843 write16x4(pix-2, stride, line1, line2, line3, line4);
844 }
845
846 static av_always_inline
847 void weight_h264_WxH_altivec(uint8_t *block, int stride, int log2_denom, int weight, int offset, int w, int h)
848 {
849 int y, aligned;
850 vec_u8 vblock;
851 vec_s16 vtemp, vweight, voffset, v0, v1;
852 vec_u16 vlog2_denom;
853 DECLARE_ALIGNED_16(int32_t, temp[4]);
854 LOAD_ZERO;
855
856 offset <<= log2_denom;
857 if(log2_denom) offset += 1<<(log2_denom-1);
858 temp[0] = log2_denom;
859 temp[1] = weight;
860 temp[2] = offset;
861
862 vtemp = (vec_s16)vec_ld(0, temp);
863 vlog2_denom = (vec_u16)vec_splat(vtemp, 1);
864 vweight = vec_splat(vtemp, 3);
865 voffset = vec_splat(vtemp, 5);
866 aligned = !((unsigned long)block & 0xf);
867
868 for (y=0; y<h; y++) {
869 vblock = vec_ld(0, block);
870
871 v0 = (vec_s16)vec_mergeh(zero_u8v, vblock);
872 v1 = (vec_s16)vec_mergel(zero_u8v, vblock);
873
874 if (w == 16 || aligned) {
875 v0 = vec_mladd(v0, vweight, zero_s16v);
876 v0 = vec_adds(v0, voffset);
877 v0 = vec_sra(v0, vlog2_denom);
878 }
879 if (w == 16 || !aligned) {
880 v1 = vec_mladd(v1, vweight, zero_s16v);
881 v1 = vec_adds(v1, voffset);
882 v1 = vec_sra(v1, vlog2_denom);
883 }
884 vblock = vec_packsu(v0, v1);
885 vec_st(vblock, 0, block);
886
887 block += stride;
888 }
889 }
890
891 static av_always_inline
892 void biweight_h264_WxH_altivec(uint8_t *dst, uint8_t *src, int stride, int log2_denom,
893 int weightd, int weights, int offset, int w, int h)
894 {
895 int y, dst_aligned, src_aligned;
896 vec_u8 vsrc, vdst;
897 vec_s16 vtemp, vweights, vweightd, voffset, v0, v1, v2, v3;
898 vec_u16 vlog2_denom;
899 DECLARE_ALIGNED_16(int32_t, temp[4]);
900 LOAD_ZERO;
901
902 offset = ((offset + 1) | 1) << log2_denom;
903 temp[0] = log2_denom+1;
904 temp[1] = weights;
905 temp[2] = weightd;
906 temp[3] = offset;
907
908 vtemp = (vec_s16)vec_ld(0, temp);
909 vlog2_denom = (vec_u16)vec_splat(vtemp, 1);
910 vweights = vec_splat(vtemp, 3);
911 vweightd = vec_splat(vtemp, 5);
912 voffset = vec_splat(vtemp, 7);
913 dst_aligned = !((unsigned long)dst & 0xf);
914 src_aligned = !((unsigned long)src & 0xf);
915
916 for (y=0; y<h; y++) {
917 vdst = vec_ld(0, dst);
918 vsrc = vec_ld(0, src);
919
920 v0 = (vec_s16)vec_mergeh(zero_u8v, vdst);
921 v1 = (vec_s16)vec_mergel(zero_u8v, vdst);
922 v2 = (vec_s16)vec_mergeh(zero_u8v, vsrc);
923 v3 = (vec_s16)vec_mergel(zero_u8v, vsrc);
924
925 if (w == 8) {
926 if (src_aligned)
927 v3 = v2;
928 else
929 v2 = v3;
930 }
931
932 if (w == 16 || dst_aligned) {
933 v0 = vec_mladd(v0, vweightd, zero_s16v);
934 v2 = vec_mladd(v2, vweights, zero_s16v);
935
936 v0 = vec_adds(v0, voffset);
937 v0 = vec_adds(v0, v2);
938 v0 = vec_sra(v0, vlog2_denom);
939 }
940 if (w == 16 || !dst_aligned) {
941 v1 = vec_mladd(v1, vweightd, zero_s16v);
942 v3 = vec_mladd(v3, vweights, zero_s16v);
943
944 v1 = vec_adds(v1, voffset);
945 v1 = vec_adds(v1, v3);
946 v1 = vec_sra(v1, vlog2_denom);
947 }
948 vdst = vec_packsu(v0, v1);
949 vec_st(vdst, 0, dst);
950
951 dst += stride;
952 src += stride;
953 }
954 }
955
956 #define H264_WEIGHT(W,H) \
957 static void ff_weight_h264_pixels ## W ## x ## H ## _altivec(uint8_t *block, int stride, int log2_denom, int weight, int offset){ \
958 weight_h264_WxH_altivec(block, stride, log2_denom, weight, offset, W, H); \
959 }\
960 static void ff_biweight_h264_pixels ## W ## x ## H ## _altivec(uint8_t *dst, uint8_t *src, int stride, int log2_denom, int weightd, int weights, int offset){ \
961 biweight_h264_WxH_altivec(dst, src, stride, log2_denom, weightd, weights, offset, W, H); \
962 }
963
964 H264_WEIGHT(16,16)
965 H264_WEIGHT(16, 8)
966 H264_WEIGHT( 8,16)
967 H264_WEIGHT( 8, 8)
968 H264_WEIGHT( 8, 4)
969
970 void dsputil_h264_init_ppc(DSPContext* c, AVCodecContext *avctx) {
971
972 if (has_altivec()) {
973 c->put_h264_chroma_pixels_tab[0] = put_h264_chroma_mc8_altivec;
974 c->avg_h264_chroma_pixels_tab[0] = avg_h264_chroma_mc8_altivec;
975 c->put_no_rnd_vc1_chroma_pixels_tab[0] = put_no_rnd_vc1_chroma_mc8_altivec;
976 c->avg_no_rnd_vc1_chroma_pixels_tab[0] = avg_no_rnd_vc1_chroma_mc8_altivec;
977 c->h264_idct_add = ff_h264_idct_add_altivec;
978 c->h264_idct_add8 = ff_h264_idct_add8_altivec;
979 c->h264_idct_add16 = ff_h264_idct_add16_altivec;
980 c->h264_idct_add16intra = ff_h264_idct_add16intra_altivec;
981 c->h264_idct_dc_add= h264_idct_dc_add_altivec;
982 c->h264_idct8_dc_add = ff_h264_idct8_dc_add_altivec;
983 c->h264_idct8_add = ff_h264_idct8_add_altivec;
984 c->h264_idct8_add4 = ff_h264_idct8_add4_altivec;
985 c->h264_v_loop_filter_luma= h264_v_loop_filter_luma_altivec;
986 c->h264_h_loop_filter_luma= h264_h_loop_filter_luma_altivec;
987
988 #define dspfunc(PFX, IDX, NUM) \
989 c->PFX ## _pixels_tab[IDX][ 0] = PFX ## NUM ## _mc00_altivec; \
990 c->PFX ## _pixels_tab[IDX][ 1] = PFX ## NUM ## _mc10_altivec; \
991 c->PFX ## _pixels_tab[IDX][ 2] = PFX ## NUM ## _mc20_altivec; \
992 c->PFX ## _pixels_tab[IDX][ 3] = PFX ## NUM ## _mc30_altivec; \
993 c->PFX ## _pixels_tab[IDX][ 4] = PFX ## NUM ## _mc01_altivec; \
994 c->PFX ## _pixels_tab[IDX][ 5] = PFX ## NUM ## _mc11_altivec; \
995 c->PFX ## _pixels_tab[IDX][ 6] = PFX ## NUM ## _mc21_altivec; \
996 c->PFX ## _pixels_tab[IDX][ 7] = PFX ## NUM ## _mc31_altivec; \
997 c->PFX ## _pixels_tab[IDX][ 8] = PFX ## NUM ## _mc02_altivec; \
998 c->PFX ## _pixels_tab[IDX][ 9] = PFX ## NUM ## _mc12_altivec; \
999 c->PFX ## _pixels_tab[IDX][10] = PFX ## NUM ## _mc22_altivec; \
1000 c->PFX ## _pixels_tab[IDX][11] = PFX ## NUM ## _mc32_altivec; \
1001 c->PFX ## _pixels_tab[IDX][12] = PFX ## NUM ## _mc03_altivec; \
1002 c->PFX ## _pixels_tab[IDX][13] = PFX ## NUM ## _mc13_altivec; \
1003 c->PFX ## _pixels_tab[IDX][14] = PFX ## NUM ## _mc23_altivec; \
1004 c->PFX ## _pixels_tab[IDX][15] = PFX ## NUM ## _mc33_altivec
1005
1006 dspfunc(put_h264_qpel, 0, 16);
1007 dspfunc(avg_h264_qpel, 0, 16);
1008 #undef dspfunc
1009
1010 c->weight_h264_pixels_tab[0] = ff_weight_h264_pixels16x16_altivec;
1011 c->weight_h264_pixels_tab[1] = ff_weight_h264_pixels16x8_altivec;
1012 c->weight_h264_pixels_tab[2] = ff_weight_h264_pixels8x16_altivec;
1013 c->weight_h264_pixels_tab[3] = ff_weight_h264_pixels8x8_altivec;
1014 c->weight_h264_pixels_tab[4] = ff_weight_h264_pixels8x4_altivec;
1015 c->biweight_h264_pixels_tab[0] = ff_biweight_h264_pixels16x16_altivec;
1016 c->biweight_h264_pixels_tab[1] = ff_biweight_h264_pixels16x8_altivec;
1017 c->biweight_h264_pixels_tab[2] = ff_biweight_h264_pixels8x16_altivec;
1018 c->biweight_h264_pixels_tab[3] = ff_biweight_h264_pixels8x8_altivec;
1019 c->biweight_h264_pixels_tab[4] = ff_biweight_h264_pixels8x4_altivec;
1020 }
1021 }
lucabe's unfinished and experimental patches for ffmpeg