doxygen: Replace '\' by '@' in Doxygen markup tags.
[libav.git] / libavcodec / ppc / h264_altivec.c
CommitLineData
a6a12a8a
RD
1/*
2 * Copyright (c) 2004 Romain Dolbeau <romain@dolbeau.org>
3 *
2912e87a 4 * This file is part of Libav.
b78e7197 5 *
2912e87a 6 * Libav is free software; you can redistribute it and/or
a6a12a8a
RD
7 * modify it under the terms of the GNU Lesser General Public
8 * License as published by the Free Software Foundation; either
b78e7197 9 * version 2.1 of the License, or (at your option) any later version.
a6a12a8a 10 *
2912e87a 11 * Libav is distributed in the hope that it will be useful,
a6a12a8a
RD
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
2912e87a 17 * License along with Libav; if not, write to the Free Software
5509bffa 18 * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
a6a12a8a 19 */
115329f1 20
c6c98d08 21#include "libavutil/cpu.h"
245976da 22#include "libavcodec/dsputil.h"
337e3fd9 23#include "libavcodec/h264data.h"
4693b031 24#include "libavcodec/h264dsp.h"
a6a12a8a 25
3035cb67 26#include "dsputil_altivec.h"
89523bee 27#include "util_altivec.h"
3813dcc9 28#include "types_altivec.h"
a6a12a8a
RD
29
30#define PUT_OP_U8_ALTIVEC(d, s, dst) d = s
31#define AVG_OP_U8_ALTIVEC(d, s, dst) d = vec_avg(dst, s)
32
33#define OP_U8_ALTIVEC PUT_OP_U8_ALTIVEC
34#define PREFIX_h264_chroma_mc8_altivec put_h264_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
b5f7e6eb 42#include "h264_template_altivec.c"
a6a12a8a
RD
43#undef OP_U8_ALTIVEC
44#undef PREFIX_h264_chroma_mc8_altivec
45#undef PREFIX_h264_chroma_mc8_num
46#undef PREFIX_h264_qpel16_h_lowpass_altivec
47#undef PREFIX_h264_qpel16_h_lowpass_num
48#undef PREFIX_h264_qpel16_v_lowpass_altivec
49#undef PREFIX_h264_qpel16_v_lowpass_num
50#undef PREFIX_h264_qpel16_hv_lowpass_altivec
51#undef PREFIX_h264_qpel16_hv_lowpass_num
52
53#define OP_U8_ALTIVEC AVG_OP_U8_ALTIVEC
54#define PREFIX_h264_chroma_mc8_altivec avg_h264_chroma_mc8_altivec
55#define PREFIX_h264_chroma_mc8_num altivec_avg_h264_chroma_mc8_num
56#define PREFIX_h264_qpel16_h_lowpass_altivec avg_h264_qpel16_h_lowpass_altivec
57#define PREFIX_h264_qpel16_h_lowpass_num altivec_avg_h264_qpel16_h_lowpass_num
58#define PREFIX_h264_qpel16_v_lowpass_altivec avg_h264_qpel16_v_lowpass_altivec
59#define PREFIX_h264_qpel16_v_lowpass_num altivec_avg_h264_qpel16_v_lowpass_num
60#define PREFIX_h264_qpel16_hv_lowpass_altivec avg_h264_qpel16_hv_lowpass_altivec
61#define PREFIX_h264_qpel16_hv_lowpass_num altivec_avg_h264_qpel16_hv_lowpass_num
b5f7e6eb 62#include "h264_template_altivec.c"
a6a12a8a
RD
63#undef OP_U8_ALTIVEC
64#undef PREFIX_h264_chroma_mc8_altivec
65#undef PREFIX_h264_chroma_mc8_num
66#undef PREFIX_h264_qpel16_h_lowpass_altivec
67#undef PREFIX_h264_qpel16_h_lowpass_num
68#undef PREFIX_h264_qpel16_v_lowpass_altivec
69#undef PREFIX_h264_qpel16_v_lowpass_num
70#undef PREFIX_h264_qpel16_hv_lowpass_altivec
71#undef PREFIX_h264_qpel16_hv_lowpass_num
72
73#define H264_MC(OPNAME, SIZE, CODETYPE) \
74static void OPNAME ## h264_qpel ## SIZE ## _mc00_ ## CODETYPE (uint8_t *dst, uint8_t *src, int stride){\
75 OPNAME ## pixels ## SIZE ## _ ## CODETYPE(dst, src, stride, SIZE);\
76}\
77\
78static void OPNAME ## h264_qpel ## SIZE ## _mc10_ ## CODETYPE(uint8_t *dst, uint8_t *src, int stride){ \
84dc2d8a 79 DECLARE_ALIGNED(16, uint8_t, half)[SIZE*SIZE];\
a6a12a8a
RD
80 put_h264_qpel ## SIZE ## _h_lowpass_ ## CODETYPE(half, src, SIZE, stride);\
81 OPNAME ## pixels ## SIZE ## _l2_ ## CODETYPE(dst, src, half, stride, stride, SIZE);\
82}\
83\
84static void OPNAME ## h264_qpel ## SIZE ## _mc20_ ## CODETYPE(uint8_t *dst, uint8_t *src, int stride){\
85 OPNAME ## h264_qpel ## SIZE ## _h_lowpass_ ## CODETYPE(dst, src, stride, stride);\
86}\
87\
88static void OPNAME ## h264_qpel ## SIZE ## _mc30_ ## CODETYPE(uint8_t *dst, uint8_t *src, int stride){\
84dc2d8a 89 DECLARE_ALIGNED(16, uint8_t, half)[SIZE*SIZE];\
a6a12a8a
RD
90 put_h264_qpel ## SIZE ## _h_lowpass_ ## CODETYPE(half, src, SIZE, stride);\
91 OPNAME ## pixels ## SIZE ## _l2_ ## CODETYPE(dst, src+1, half, stride, stride, SIZE);\
92}\
93\
94static void OPNAME ## h264_qpel ## SIZE ## _mc01_ ## CODETYPE(uint8_t *dst, uint8_t *src, int stride){\
84dc2d8a 95 DECLARE_ALIGNED(16, uint8_t, half)[SIZE*SIZE];\
a6a12a8a
RD
96 put_h264_qpel ## SIZE ## _v_lowpass_ ## CODETYPE(half, src, SIZE, stride);\
97 OPNAME ## pixels ## SIZE ## _l2_ ## CODETYPE(dst, src, half, stride, stride, SIZE);\
98}\
99\
100static void OPNAME ## h264_qpel ## SIZE ## _mc02_ ## CODETYPE(uint8_t *dst, uint8_t *src, int stride){\
101 OPNAME ## h264_qpel ## SIZE ## _v_lowpass_ ## CODETYPE(dst, src, stride, stride);\
102}\
103\
104static void OPNAME ## h264_qpel ## SIZE ## _mc03_ ## CODETYPE(uint8_t *dst, uint8_t *src, int stride){\
84dc2d8a 105 DECLARE_ALIGNED(16, uint8_t, half)[SIZE*SIZE];\
a6a12a8a
RD
106 put_h264_qpel ## SIZE ## _v_lowpass_ ## CODETYPE(half, src, SIZE, stride);\
107 OPNAME ## pixels ## SIZE ## _l2_ ## CODETYPE(dst, src+stride, half, stride, stride, SIZE);\
108}\
109\
110static void OPNAME ## h264_qpel ## SIZE ## _mc11_ ## CODETYPE(uint8_t *dst, uint8_t *src, int stride){\
84dc2d8a
MR
111 DECLARE_ALIGNED(16, uint8_t, halfH)[SIZE*SIZE];\
112 DECLARE_ALIGNED(16, uint8_t, halfV)[SIZE*SIZE];\
a6a12a8a
RD
113 put_h264_qpel ## SIZE ## _h_lowpass_ ## CODETYPE(halfH, src, SIZE, stride);\
114 put_h264_qpel ## SIZE ## _v_lowpass_ ## CODETYPE(halfV, src, SIZE, stride);\
115 OPNAME ## pixels ## SIZE ## _l2_ ## CODETYPE(dst, halfH, halfV, stride, SIZE, SIZE);\
116}\
117\
118static void OPNAME ## h264_qpel ## SIZE ## _mc31_ ## CODETYPE(uint8_t *dst, uint8_t *src, int stride){\
84dc2d8a
MR
119 DECLARE_ALIGNED(16, uint8_t, halfH)[SIZE*SIZE];\
120 DECLARE_ALIGNED(16, uint8_t, halfV)[SIZE*SIZE];\
a6a12a8a
RD
121 put_h264_qpel ## SIZE ## _h_lowpass_ ## CODETYPE(halfH, src, SIZE, stride);\
122 put_h264_qpel ## SIZE ## _v_lowpass_ ## CODETYPE(halfV, src+1, SIZE, stride);\
123 OPNAME ## pixels ## SIZE ## _l2_ ## CODETYPE(dst, halfH, halfV, stride, SIZE, SIZE);\
124}\
125\
126static void OPNAME ## h264_qpel ## SIZE ## _mc13_ ## CODETYPE(uint8_t *dst, uint8_t *src, int stride){\
84dc2d8a
MR
127 DECLARE_ALIGNED(16, uint8_t, halfH)[SIZE*SIZE];\
128 DECLARE_ALIGNED(16, uint8_t, halfV)[SIZE*SIZE];\
a6a12a8a
RD
129 put_h264_qpel ## SIZE ## _h_lowpass_ ## CODETYPE(halfH, src + stride, SIZE, stride);\
130 put_h264_qpel ## SIZE ## _v_lowpass_ ## CODETYPE(halfV, src, SIZE, stride);\
131 OPNAME ## pixels ## SIZE ## _l2_ ## CODETYPE(dst, halfH, halfV, stride, SIZE, SIZE);\
132}\
133\
134static void OPNAME ## h264_qpel ## SIZE ## _mc33_ ## CODETYPE(uint8_t *dst, uint8_t *src, int stride){\
84dc2d8a
MR
135 DECLARE_ALIGNED(16, uint8_t, halfH)[SIZE*SIZE];\
136 DECLARE_ALIGNED(16, uint8_t, halfV)[SIZE*SIZE];\
a6a12a8a
RD
137 put_h264_qpel ## SIZE ## _h_lowpass_ ## CODETYPE(halfH, src + stride, SIZE, stride);\
138 put_h264_qpel ## SIZE ## _v_lowpass_ ## CODETYPE(halfV, src+1, SIZE, stride);\
139 OPNAME ## pixels ## SIZE ## _l2_ ## CODETYPE(dst, halfH, halfV, stride, SIZE, SIZE);\
140}\
141\
142static void OPNAME ## h264_qpel ## SIZE ## _mc22_ ## CODETYPE(uint8_t *dst, uint8_t *src, int stride){\
84dc2d8a 143 DECLARE_ALIGNED(16, int16_t, tmp)[SIZE*(SIZE+8)];\
a6a12a8a
RD
144 OPNAME ## h264_qpel ## SIZE ## _hv_lowpass_ ## CODETYPE(dst, tmp, src, stride, SIZE, stride);\
145}\
146\
147static void OPNAME ## h264_qpel ## SIZE ## _mc21_ ## CODETYPE(uint8_t *dst, uint8_t *src, int stride){\
84dc2d8a
MR
148 DECLARE_ALIGNED(16, uint8_t, halfH)[SIZE*SIZE];\
149 DECLARE_ALIGNED(16, uint8_t, halfHV)[SIZE*SIZE];\
150 DECLARE_ALIGNED(16, int16_t, tmp)[SIZE*(SIZE+8)];\
a6a12a8a
RD
151 put_h264_qpel ## SIZE ## _h_lowpass_ ## CODETYPE(halfH, src, SIZE, stride);\
152 put_h264_qpel ## SIZE ## _hv_lowpass_ ## CODETYPE(halfHV, tmp, src, SIZE, SIZE, stride);\
153 OPNAME ## pixels ## SIZE ## _l2_ ## CODETYPE(dst, halfH, halfHV, stride, SIZE, SIZE);\
154}\
155\
156static void OPNAME ## h264_qpel ## SIZE ## _mc23_ ## CODETYPE(uint8_t *dst, uint8_t *src, int stride){\
84dc2d8a
MR
157 DECLARE_ALIGNED(16, uint8_t, halfH)[SIZE*SIZE];\
158 DECLARE_ALIGNED(16, uint8_t, halfHV)[SIZE*SIZE];\
159 DECLARE_ALIGNED(16, int16_t, tmp)[SIZE*(SIZE+8)];\
a6a12a8a
RD
160 put_h264_qpel ## SIZE ## _h_lowpass_ ## CODETYPE(halfH, src + stride, SIZE, stride);\
161 put_h264_qpel ## SIZE ## _hv_lowpass_ ## CODETYPE(halfHV, tmp, src, SIZE, SIZE, stride);\
162 OPNAME ## pixels ## SIZE ## _l2_ ## CODETYPE(dst, halfH, halfHV, stride, SIZE, SIZE);\
163}\
164\
165static void OPNAME ## h264_qpel ## SIZE ## _mc12_ ## CODETYPE(uint8_t *dst, uint8_t *src, int stride){\
84dc2d8a
MR
166 DECLARE_ALIGNED(16, uint8_t, halfV)[SIZE*SIZE];\
167 DECLARE_ALIGNED(16, uint8_t, halfHV)[SIZE*SIZE];\
168 DECLARE_ALIGNED(16, int16_t, tmp)[SIZE*(SIZE+8)];\
a6a12a8a
RD
169 put_h264_qpel ## SIZE ## _v_lowpass_ ## CODETYPE(halfV, src, SIZE, stride);\
170 put_h264_qpel ## SIZE ## _hv_lowpass_ ## CODETYPE(halfHV, tmp, src, SIZE, SIZE, stride);\
171 OPNAME ## pixels ## SIZE ## _l2_ ## CODETYPE(dst, halfV, halfHV, stride, SIZE, SIZE);\
172}\
173\
174static void OPNAME ## h264_qpel ## SIZE ## _mc32_ ## CODETYPE(uint8_t *dst, uint8_t *src, int stride){\
84dc2d8a
MR
175 DECLARE_ALIGNED(16, uint8_t, halfV)[SIZE*SIZE];\
176 DECLARE_ALIGNED(16, uint8_t, halfHV)[SIZE*SIZE];\
177 DECLARE_ALIGNED(16, int16_t, tmp)[SIZE*(SIZE+8)];\
a6a12a8a
RD
178 put_h264_qpel ## SIZE ## _v_lowpass_ ## CODETYPE(halfV, src+1, SIZE, stride);\
179 put_h264_qpel ## SIZE ## _hv_lowpass_ ## CODETYPE(halfHV, tmp, src, SIZE, SIZE, stride);\
180 OPNAME ## pixels ## SIZE ## _l2_ ## CODETYPE(dst, halfV, halfHV, stride, SIZE, SIZE);\
181}\
182
0d18f798
LB
183static inline void put_pixels16_l2_altivec( uint8_t * dst, const uint8_t * src1,
184 const uint8_t * src2, int dst_stride,
185 int src_stride1, int h)
186{
187 int i;
a6b4448c 188 vec_u8 a, b, d, tmp1, tmp2, mask, mask_, edges, align;
a6a12a8a 189
0d18f798
LB
190 mask_ = vec_lvsl(0, src2);
191
192 for (i = 0; i < h; i++) {
193
194 tmp1 = vec_ld(i * src_stride1, src1);
195 mask = vec_lvsl(i * src_stride1, src1);
196 tmp2 = vec_ld(i * src_stride1 + 15, src1);
197
198 a = vec_perm(tmp1, tmp2, mask);
199
200 tmp1 = vec_ld(i * 16, src2);
201 tmp2 = vec_ld(i * 16 + 15, src2);
202
203 b = vec_perm(tmp1, tmp2, mask_);
204
205 tmp1 = vec_ld(0, dst);
206 mask = vec_lvsl(0, dst);
207 tmp2 = vec_ld(15, dst);
208
209 d = vec_avg(a, b);
210
211 edges = vec_perm(tmp2, tmp1, mask);
212
213 align = vec_lvsr(0, dst);
214
7e821457 215 tmp2 = vec_perm(d, edges, align);
27303c8a 216 tmp1 = vec_perm(edges, d, align);
0d18f798 217
cb243ea2 218 vec_st(tmp2, 15, dst);
27303c8a 219 vec_st(tmp1, 0 , dst);
0d18f798
LB
220
221 dst += dst_stride;
222 }
223}
224
225static inline void avg_pixels16_l2_altivec( uint8_t * dst, const uint8_t * src1,
226 const uint8_t * src2, int dst_stride,
227 int src_stride1, int h)
228{
229 int i;
a6b4448c 230 vec_u8 a, b, d, tmp1, tmp2, mask, mask_, edges, align;
0d18f798
LB
231
232 mask_ = vec_lvsl(0, src2);
233
234 for (i = 0; i < h; i++) {
235
236 tmp1 = vec_ld(i * src_stride1, src1);
237 mask = vec_lvsl(i * src_stride1, src1);
238 tmp2 = vec_ld(i * src_stride1 + 15, src1);
239
240 a = vec_perm(tmp1, tmp2, mask);
241
242 tmp1 = vec_ld(i * 16, src2);
243 tmp2 = vec_ld(i * 16 + 15, src2);
244
245 b = vec_perm(tmp1, tmp2, mask_);
246
247 tmp1 = vec_ld(0, dst);
248 mask = vec_lvsl(0, dst);
249 tmp2 = vec_ld(15, dst);
250
251 d = vec_avg(vec_perm(tmp1, tmp2, mask), vec_avg(a, b));
252
253 edges = vec_perm(tmp2, tmp1, mask);
254
255 align = vec_lvsr(0, dst);
256
7e821457 257 tmp2 = vec_perm(d, edges, align);
27303c8a 258 tmp1 = vec_perm(edges, d, align);
0d18f798 259
cb243ea2 260 vec_st(tmp2, 15, dst);
27303c8a 261 vec_st(tmp1, 0 , dst);
0d18f798
LB
262
263 dst += dst_stride;
264 }
a6a12a8a
RD
265}
266
0d18f798 267/* Implemented but could be faster
a6a12a8a
RD
268#define put_pixels16_l2_altivec(d,s1,s2,ds,s1s,h) put_pixels16_l2(d,s1,s2,ds,s1s,16,h)
269#define avg_pixels16_l2_altivec(d,s1,s2,ds,s1s,h) avg_pixels16_l2(d,s1,s2,ds,s1s,16,h)
0d18f798 270 */
a6a12a8a 271
e3905ce0
DB
272H264_MC(put_, 16, altivec)
273H264_MC(avg_, 16, altivec)
a6a12a8a 274
3813dcc9
GP
275
276/****************************************************************************
277 * IDCT transform:
278 ****************************************************************************/
279
830bf1f2
DB
280#define VEC_1D_DCT(vb0,vb1,vb2,vb3,va0,va1,va2,va3) \
281 /* 1st stage */ \
282 vz0 = vec_add(vb0,vb2); /* temp[0] = Y[0] + Y[2] */ \
283 vz1 = vec_sub(vb0,vb2); /* temp[1] = Y[0] - Y[2] */ \
284 vz2 = vec_sra(vb1,vec_splat_u16(1)); \
285 vz2 = vec_sub(vz2,vb3); /* temp[2] = Y[1].1/2 - Y[3] */ \
286 vz3 = vec_sra(vb3,vec_splat_u16(1)); \
287 vz3 = vec_add(vb1,vz3); /* temp[3] = Y[1] + Y[3].1/2 */ \
288 /* 2nd stage: output */ \
289 va0 = vec_add(vz0,vz3); /* x[0] = temp[0] + temp[3] */ \
290 va1 = vec_add(vz1,vz2); /* x[1] = temp[1] + temp[2] */ \
291 va2 = vec_sub(vz1,vz2); /* x[2] = temp[1] - temp[2] */ \
292 va3 = vec_sub(vz0,vz3) /* x[3] = temp[0] - temp[3] */
5dda2539
LB
293
294#define VEC_TRANSPOSE_4(a0,a1,a2,a3,b0,b1,b2,b3) \
295 b0 = vec_mergeh( a0, a0 ); \
296 b1 = vec_mergeh( a1, a0 ); \
297 b2 = vec_mergeh( a2, a0 ); \
298 b3 = vec_mergeh( a3, a0 ); \
299 a0 = vec_mergeh( b0, b2 ); \
300 a1 = vec_mergel( b0, b2 ); \
301 a2 = vec_mergeh( b1, b3 ); \
302 a3 = vec_mergel( b1, b3 ); \
303 b0 = vec_mergeh( a0, a2 ); \
304 b1 = vec_mergel( a0, a2 ); \
305 b2 = vec_mergeh( a1, a3 ); \
306 b3 = vec_mergel( a1, a3 )
307
308#define VEC_LOAD_U8_ADD_S16_STORE_U8(va) \
309 vdst_orig = vec_ld(0, dst); \
310 vdst = vec_perm(vdst_orig, zero_u8v, vdst_mask); \
a6b4448c 311 vdst_ss = (vec_s16) vec_mergeh(zero_u8v, vdst); \
5dda2539
LB
312 va = vec_add(va, vdst_ss); \
313 va_u8 = vec_packsu(va, zero_s16v); \
a6b4448c 314 va_u32 = vec_splat((vec_u32)va_u8, 0); \
5dda2539
LB
315 vec_ste(va_u32, element, (uint32_t*)dst);
316
317static void ff_h264_idct_add_altivec(uint8_t *dst, DCTELEM *block, int stride)
318{
a6b4448c
LB
319 vec_s16 va0, va1, va2, va3;
320 vec_s16 vz0, vz1, vz2, vz3;
321 vec_s16 vtmp0, vtmp1, vtmp2, vtmp3;
322 vec_u8 va_u8;
323 vec_u32 va_u32;
324 vec_s16 vdst_ss;
325 const vec_u16 v6us = vec_splat_u16(6);
326 vec_u8 vdst, vdst_orig;
327 vec_u8 vdst_mask = vec_lvsl(0, dst);
5dda2539
LB
328 int element = ((unsigned long)dst & 0xf) >> 2;
329 LOAD_ZERO;
330
331 block[0] += 32; /* add 32 as a DC-level for rounding */
332
333 vtmp0 = vec_ld(0,block);
334 vtmp1 = vec_sld(vtmp0, vtmp0, 8);
335 vtmp2 = vec_ld(16,block);
336 vtmp3 = vec_sld(vtmp2, vtmp2, 8);
337
338 VEC_1D_DCT(vtmp0,vtmp1,vtmp2,vtmp3,va0,va1,va2,va3);
339 VEC_TRANSPOSE_4(va0,va1,va2,va3,vtmp0,vtmp1,vtmp2,vtmp3);
340 VEC_1D_DCT(vtmp0,vtmp1,vtmp2,vtmp3,va0,va1,va2,va3);
341
342 va0 = vec_sra(va0,v6us);
343 va1 = vec_sra(va1,v6us);
344 va2 = vec_sra(va2,v6us);
345 va3 = vec_sra(va3,v6us);
346
347 VEC_LOAD_U8_ADD_S16_STORE_U8(va0);
348 dst += stride;
349 VEC_LOAD_U8_ADD_S16_STORE_U8(va1);
350 dst += stride;
351 VEC_LOAD_U8_ADD_S16_STORE_U8(va2);
352 dst += stride;
353 VEC_LOAD_U8_ADD_S16_STORE_U8(va3);
354}
355
3813dcc9
GP
356#define IDCT8_1D_ALTIVEC(s0, s1, s2, s3, s4, s5, s6, s7, d0, d1, d2, d3, d4, d5, d6, d7) {\
357 /* a0 = SRC(0) + SRC(4); */ \
a6b4448c 358 vec_s16 a0v = vec_add(s0, s4); \
3813dcc9 359 /* a2 = SRC(0) - SRC(4); */ \
a6b4448c 360 vec_s16 a2v = vec_sub(s0, s4); \
3813dcc9 361 /* a4 = (SRC(2)>>1) - SRC(6); */ \
a6b4448c 362 vec_s16 a4v = vec_sub(vec_sra(s2, onev), s6); \
3813dcc9 363 /* a6 = (SRC(6)>>1) + SRC(2); */ \
a6b4448c 364 vec_s16 a6v = vec_add(vec_sra(s6, onev), s2); \
3813dcc9 365 /* b0 = a0 + a6; */ \
a6b4448c 366 vec_s16 b0v = vec_add(a0v, a6v); \
3813dcc9 367 /* b2 = a2 + a4; */ \
a6b4448c 368 vec_s16 b2v = vec_add(a2v, a4v); \
3813dcc9 369 /* b4 = a2 - a4; */ \
a6b4448c 370 vec_s16 b4v = vec_sub(a2v, a4v); \
3813dcc9 371 /* b6 = a0 - a6; */ \
a6b4448c 372 vec_s16 b6v = vec_sub(a0v, a6v); \
3813dcc9
GP
373 /* a1 = SRC(5) - SRC(3) - SRC(7) - (SRC(7)>>1); */ \
374 /* a1 = (SRC(5)-SRC(3)) - (SRC(7) + (SRC(7)>>1)); */ \
a6b4448c 375 vec_s16 a1v = vec_sub( vec_sub(s5, s3), vec_add(s7, vec_sra(s7, onev)) ); \
3813dcc9
GP
376 /* a3 = SRC(7) + SRC(1) - SRC(3) - (SRC(3)>>1); */ \
377 /* a3 = (SRC(7)+SRC(1)) - (SRC(3) + (SRC(3)>>1)); */ \
a6b4448c 378 vec_s16 a3v = vec_sub( vec_add(s7, s1), vec_add(s3, vec_sra(s3, onev)) );\
3813dcc9
GP
379 /* a5 = SRC(7) - SRC(1) + SRC(5) + (SRC(5)>>1); */ \
380 /* a5 = (SRC(7)-SRC(1)) + SRC(5) + (SRC(5)>>1); */ \
a6b4448c 381 vec_s16 a5v = vec_add( vec_sub(s7, s1), vec_add(s5, vec_sra(s5, onev)) );\
3813dcc9 382 /* a7 = SRC(5)+SRC(3) + SRC(1) + (SRC(1)>>1); */ \
a6b4448c 383 vec_s16 a7v = vec_add( vec_add(s5, s3), vec_add(s1, vec_sra(s1, onev)) );\
3813dcc9 384 /* b1 = (a7>>2) + a1; */ \
a6b4448c 385 vec_s16 b1v = vec_add( vec_sra(a7v, twov), a1v); \
3813dcc9 386 /* b3 = a3 + (a5>>2); */ \
a6b4448c 387 vec_s16 b3v = vec_add(a3v, vec_sra(a5v, twov)); \
3813dcc9 388 /* b5 = (a3>>2) - a5; */ \
a6b4448c 389 vec_s16 b5v = vec_sub( vec_sra(a3v, twov), a5v); \
3813dcc9 390 /* b7 = a7 - (a1>>2); */ \
a6b4448c 391 vec_s16 b7v = vec_sub( a7v, vec_sra(a1v, twov)); \
3813dcc9
GP
392 /* DST(0, b0 + b7); */ \
393 d0 = vec_add(b0v, b7v); \
394 /* DST(1, b2 + b5); */ \
395 d1 = vec_add(b2v, b5v); \
396 /* DST(2, b4 + b3); */ \
397 d2 = vec_add(b4v, b3v); \
398 /* DST(3, b6 + b1); */ \
399 d3 = vec_add(b6v, b1v); \
400 /* DST(4, b6 - b1); */ \
401 d4 = vec_sub(b6v, b1v); \
402 /* DST(5, b4 - b3); */ \
403 d5 = vec_sub(b4v, b3v); \
404 /* DST(6, b2 - b5); */ \
405 d6 = vec_sub(b2v, b5v); \
406 /* DST(7, b0 - b7); */ \
407 d7 = vec_sub(b0v, b7v); \
408}
409
410#define ALTIVEC_STORE_SUM_CLIP(dest, idctv, perm_ldv, perm_stv, sel) { \
411 /* unaligned load */ \
a6b4448c
LB
412 vec_u8 hv = vec_ld( 0, dest ); \
413 vec_u8 lv = vec_ld( 7, dest ); \
414 vec_u8 dstv = vec_perm( hv, lv, (vec_u8)perm_ldv ); \
415 vec_s16 idct_sh6 = vec_sra(idctv, sixv); \
416 vec_u16 dst16 = (vec_u16)vec_mergeh(zero_u8v, dstv); \
417 vec_s16 idstsum = vec_adds(idct_sh6, (vec_s16)dst16); \
418 vec_u8 idstsum8 = vec_packsu(zero_s16v, idstsum); \
419 vec_u8 edgehv; \
3813dcc9 420 /* unaligned store */ \
a6b4448c
LB
421 vec_u8 bodyv = vec_perm( idstsum8, idstsum8, perm_stv );\
422 vec_u8 edgelv = vec_perm( sel, zero_u8v, perm_stv ); \
3813dcc9
GP
423 lv = vec_sel( lv, bodyv, edgelv ); \
424 vec_st( lv, 7, dest ); \
425 hv = vec_ld( 0, dest ); \
426 edgehv = vec_perm( zero_u8v, sel, perm_stv ); \
427 hv = vec_sel( hv, bodyv, edgehv ); \
428 vec_st( hv, 0, dest ); \
429 }
430
ddb8c2c0 431static void ff_h264_idct8_add_altivec( uint8_t *dst, DCTELEM *dct, int stride ) {
a6b4448c
LB
432 vec_s16 s0, s1, s2, s3, s4, s5, s6, s7;
433 vec_s16 d0, d1, d2, d3, d4, d5, d6, d7;
434 vec_s16 idct0, idct1, idct2, idct3, idct4, idct5, idct6, idct7;
3813dcc9 435
a6b4448c
LB
436 vec_u8 perm_ldv = vec_lvsl(0, dst);
437 vec_u8 perm_stv = vec_lvsr(8, dst);
3813dcc9 438
a6b4448c
LB
439 const vec_u16 onev = vec_splat_u16(1);
440 const vec_u16 twov = vec_splat_u16(2);
441 const vec_u16 sixv = vec_splat_u16(6);
3813dcc9 442
a6b4448c 443 const vec_u8 sel = (vec_u8) {0,0,0,0,0,0,0,0,-1,-1,-1,-1,-1,-1,-1,-1};
3813dcc9
GP
444 LOAD_ZERO;
445
446 dct[0] += 32; // rounding for the >>6 at the end
447
448 s0 = vec_ld(0x00, (int16_t*)dct);
449 s1 = vec_ld(0x10, (int16_t*)dct);
450 s2 = vec_ld(0x20, (int16_t*)dct);
451 s3 = vec_ld(0x30, (int16_t*)dct);
452 s4 = vec_ld(0x40, (int16_t*)dct);
453 s5 = vec_ld(0x50, (int16_t*)dct);
454 s6 = vec_ld(0x60, (int16_t*)dct);
455 s7 = vec_ld(0x70, (int16_t*)dct);
456
457 IDCT8_1D_ALTIVEC(s0, s1, s2, s3, s4, s5, s6, s7,
458 d0, d1, d2, d3, d4, d5, d6, d7);
459
460 TRANSPOSE8( d0, d1, d2, d3, d4, d5, d6, d7 );
461
462 IDCT8_1D_ALTIVEC(d0, d1, d2, d3, d4, d5, d6, d7,
463 idct0, idct1, idct2, idct3, idct4, idct5, idct6, idct7);
464
465 ALTIVEC_STORE_SUM_CLIP(&dst[0*stride], idct0, perm_ldv, perm_stv, sel);
466 ALTIVEC_STORE_SUM_CLIP(&dst[1*stride], idct1, perm_ldv, perm_stv, sel);
467 ALTIVEC_STORE_SUM_CLIP(&dst[2*stride], idct2, perm_ldv, perm_stv, sel);
468 ALTIVEC_STORE_SUM_CLIP(&dst[3*stride], idct3, perm_ldv, perm_stv, sel);
469 ALTIVEC_STORE_SUM_CLIP(&dst[4*stride], idct4, perm_ldv, perm_stv, sel);
470 ALTIVEC_STORE_SUM_CLIP(&dst[5*stride], idct5, perm_ldv, perm_stv, sel);
471 ALTIVEC_STORE_SUM_CLIP(&dst[6*stride], idct6, perm_ldv, perm_stv, sel);
472 ALTIVEC_STORE_SUM_CLIP(&dst[7*stride], idct7, perm_ldv, perm_stv, sel);
473}
474
c663cb0d
DC
475static av_always_inline void h264_idct_dc_add_internal(uint8_t *dst, DCTELEM *block, int stride, int size)
476{
477 vec_s16 dc16;
478 vec_u8 dcplus, dcminus, v0, v1, v2, v3, aligner;
479 LOAD_ZERO;
84dc2d8a 480 DECLARE_ALIGNED(16, int, dc);
c663cb0d
DC
481 int i;
482
483 dc = (block[0] + 32) >> 6;
484 dc16 = vec_splat((vec_s16) vec_lde(0, &dc), 1);
485
486 if (size == 4)
487 dc16 = vec_sld(dc16, zero_s16v, 8);
488 dcplus = vec_packsu(dc16, zero_s16v);
489 dcminus = vec_packsu(vec_sub(zero_s16v, dc16), zero_s16v);
490
491 aligner = vec_lvsr(0, dst);
492 dcplus = vec_perm(dcplus, dcplus, aligner);
493 dcminus = vec_perm(dcminus, dcminus, aligner);
494
495 for (i = 0; i < size; i += 4) {
496 v0 = vec_ld(0, dst+0*stride);
497 v1 = vec_ld(0, dst+1*stride);
498 v2 = vec_ld(0, dst+2*stride);
499 v3 = vec_ld(0, dst+3*stride);
500
501 v0 = vec_adds(v0, dcplus);
502 v1 = vec_adds(v1, dcplus);
503 v2 = vec_adds(v2, dcplus);
504 v3 = vec_adds(v3, dcplus);
505
506 v0 = vec_subs(v0, dcminus);
507 v1 = vec_subs(v1, dcminus);
508 v2 = vec_subs(v2, dcminus);
509 v3 = vec_subs(v3, dcminus);
510
511 vec_st(v0, 0, dst+0*stride);
512 vec_st(v1, 0, dst+1*stride);
513 vec_st(v2, 0, dst+2*stride);
514 vec_st(v3, 0, dst+3*stride);
515
516 dst += 4*stride;
337e3fd9
GP
517 }
518}
519
c663cb0d
DC
520static void h264_idct_dc_add_altivec(uint8_t *dst, DCTELEM *block, int stride)
521{
522 h264_idct_dc_add_internal(dst, block, stride, 4);
523}
524
525static void ff_h264_idct8_dc_add_altivec(uint8_t *dst, DCTELEM *block, int stride)
526{
527 h264_idct_dc_add_internal(dst, block, stride, 8);
528}
529
c90b9442 530static void ff_h264_idct_add16_altivec(uint8_t *dst, const int *block_offset, DCTELEM *block, int stride, const uint8_t nnzc[15*8]){
b06688ff
DC
531 int i;
532 for(i=0; i<16; i++){
533 int nnz = nnzc[ scan8[i] ];
534 if(nnz){
535 if(nnz==1 && block[i*16]) h264_idct_dc_add_altivec(dst + block_offset[i], block + i*16, stride);
536 else ff_h264_idct_add_altivec(dst + block_offset[i], block + i*16, stride);
537 }
538 }
539}
540
c90b9442 541static void ff_h264_idct_add16intra_altivec(uint8_t *dst, const int *block_offset, DCTELEM *block, int stride, const uint8_t nnzc[15*8]){
b06688ff
DC
542 int i;
543 for(i=0; i<16; i++){
544 if(nnzc[ scan8[i] ]) ff_h264_idct_add_altivec(dst + block_offset[i], block + i*16, stride);
545 else if(block[i*16]) h264_idct_dc_add_altivec(dst + block_offset[i], block + i*16, stride);
546 }
547}
548
c90b9442 549static void ff_h264_idct8_add4_altivec(uint8_t *dst, const int *block_offset, DCTELEM *block, int stride, const uint8_t nnzc[15*8]){
337e3fd9
GP
550 int i;
551 for(i=0; i<16; i+=4){
552 int nnz = nnzc[ scan8[i] ];
553 if(nnz){
554 if(nnz==1 && block[i*16]) ff_h264_idct8_dc_add_altivec(dst + block_offset[i], block + i*16, stride);
555 else ff_h264_idct8_add_altivec (dst + block_offset[i], block + i*16, stride);
556 }
557 }
558}
559
c90b9442
JGG
560static void ff_h264_idct_add8_altivec(uint8_t **dest, const int *block_offset, DCTELEM *block, int stride, const uint8_t nnzc[15*8]){
561 int i, j;
562 for (j = 1; j < 3; j++) {
563 for(i = j * 16; i < j * 16 + 4; i++){
564 if(nnzc[ scan8[i] ])
565 ff_h264_idct_add_altivec(dest[j-1] + block_offset[i], block + i*16, stride);
566 else if(block[i*16])
567 h264_idct_dc_add_altivec(dest[j-1] + block_offset[i], block + i*16, stride);
568 }
b06688ff
DC
569 }
570}
571
f057cc03 572#define transpose4x16(r0, r1, r2, r3) { \
a6b4448c
LB
573 register vec_u8 r4; \
574 register vec_u8 r5; \
575 register vec_u8 r6; \
576 register vec_u8 r7; \
f057cc03
GB
577 \
578 r4 = vec_mergeh(r0, r2); /*0, 2 set 0*/ \
579 r5 = vec_mergel(r0, r2); /*0, 2 set 1*/ \
580 r6 = vec_mergeh(r1, r3); /*1, 3 set 0*/ \
581 r7 = vec_mergel(r1, r3); /*1, 3 set 1*/ \
582 \
583 r0 = vec_mergeh(r4, r6); /*all set 0*/ \
584 r1 = vec_mergel(r4, r6); /*all set 1*/ \
585 r2 = vec_mergeh(r5, r7); /*all set 2*/ \
586 r3 = vec_mergel(r5, r7); /*all set 3*/ \
587}
588
da1fce39 589static inline void write16x4(uint8_t *dst, int dst_stride,
a6b4448c
LB
590 register vec_u8 r0, register vec_u8 r1,
591 register vec_u8 r2, register vec_u8 r3) {
84dc2d8a 592 DECLARE_ALIGNED(16, unsigned char, result)[64];
f057cc03
GB
593 uint32_t *src_int = (uint32_t *)result, *dst_int = (uint32_t *)dst;
594 int int_dst_stride = dst_stride/4;
595
596 vec_st(r0, 0, result);
597 vec_st(r1, 16, result);
598 vec_st(r2, 32, result);
599 vec_st(r3, 48, result);
600 /* FIXME: there has to be a better way!!!! */
601 *dst_int = *src_int;
602 *(dst_int+ int_dst_stride) = *(src_int + 1);
603 *(dst_int+ 2*int_dst_stride) = *(src_int + 2);
604 *(dst_int+ 3*int_dst_stride) = *(src_int + 3);
605 *(dst_int+ 4*int_dst_stride) = *(src_int + 4);
606 *(dst_int+ 5*int_dst_stride) = *(src_int + 5);
607 *(dst_int+ 6*int_dst_stride) = *(src_int + 6);
608 *(dst_int+ 7*int_dst_stride) = *(src_int + 7);
609 *(dst_int+ 8*int_dst_stride) = *(src_int + 8);
610 *(dst_int+ 9*int_dst_stride) = *(src_int + 9);
611 *(dst_int+10*int_dst_stride) = *(src_int + 10);
612 *(dst_int+11*int_dst_stride) = *(src_int + 11);
613 *(dst_int+12*int_dst_stride) = *(src_int + 12);
614 *(dst_int+13*int_dst_stride) = *(src_int + 13);
615 *(dst_int+14*int_dst_stride) = *(src_int + 14);
616 *(dst_int+15*int_dst_stride) = *(src_int + 15);
617}
618
ff159e78
DB
619/** @brief performs a 6x16 transpose of data in src, and stores it to dst
620 @todo FIXME: see if we can't spare some vec_lvsl() by them factorizing
f057cc03
GB
621 out of unaligned_load() */
622#define readAndTranspose16x6(src, src_stride, r8, r9, r10, r11, r12, r13) {\
a6b4448c
LB
623 register vec_u8 r0 = unaligned_load(0, src); \
624 register vec_u8 r1 = unaligned_load( src_stride, src); \
625 register vec_u8 r2 = unaligned_load(2* src_stride, src); \
626 register vec_u8 r3 = unaligned_load(3* src_stride, src); \
627 register vec_u8 r4 = unaligned_load(4* src_stride, src); \
628 register vec_u8 r5 = unaligned_load(5* src_stride, src); \
629 register vec_u8 r6 = unaligned_load(6* src_stride, src); \
630 register vec_u8 r7 = unaligned_load(7* src_stride, src); \
631 register vec_u8 r14 = unaligned_load(14*src_stride, src); \
632 register vec_u8 r15 = unaligned_load(15*src_stride, src); \
f057cc03
GB
633 \
634 r8 = unaligned_load( 8*src_stride, src); \
635 r9 = unaligned_load( 9*src_stride, src); \
636 r10 = unaligned_load(10*src_stride, src); \
637 r11 = unaligned_load(11*src_stride, src); \
638 r12 = unaligned_load(12*src_stride, src); \
639 r13 = unaligned_load(13*src_stride, src); \
640 \
641 /*Merge first pairs*/ \
642 r0 = vec_mergeh(r0, r8); /*0, 8*/ \
643 r1 = vec_mergeh(r1, r9); /*1, 9*/ \
644 r2 = vec_mergeh(r2, r10); /*2,10*/ \
645 r3 = vec_mergeh(r3, r11); /*3,11*/ \
646 r4 = vec_mergeh(r4, r12); /*4,12*/ \
647 r5 = vec_mergeh(r5, r13); /*5,13*/ \
648 r6 = vec_mergeh(r6, r14); /*6,14*/ \
649 r7 = vec_mergeh(r7, r15); /*7,15*/ \
650 \
651 /*Merge second pairs*/ \
652 r8 = vec_mergeh(r0, r4); /*0,4, 8,12 set 0*/ \
653 r9 = vec_mergel(r0, r4); /*0,4, 8,12 set 1*/ \
654 r10 = vec_mergeh(r1, r5); /*1,5, 9,13 set 0*/ \
655 r11 = vec_mergel(r1, r5); /*1,5, 9,13 set 1*/ \
656 r12 = vec_mergeh(r2, r6); /*2,6,10,14 set 0*/ \
657 r13 = vec_mergel(r2, r6); /*2,6,10,14 set 1*/ \
658 r14 = vec_mergeh(r3, r7); /*3,7,11,15 set 0*/ \
659 r15 = vec_mergel(r3, r7); /*3,7,11,15 set 1*/ \
660 \
661 /*Third merge*/ \
e3905ce0
DB
662 r0 = vec_mergeh(r8, r12); /*0,2,4,6,8,10,12,14 set 0*/ \
663 r1 = vec_mergel(r8, r12); /*0,2,4,6,8,10,12,14 set 1*/ \
664 r2 = vec_mergeh(r9, r13); /*0,2,4,6,8,10,12,14 set 2*/ \
f057cc03
GB
665 r4 = vec_mergeh(r10, r14); /*1,3,5,7,9,11,13,15 set 0*/ \
666 r5 = vec_mergel(r10, r14); /*1,3,5,7,9,11,13,15 set 1*/ \
667 r6 = vec_mergeh(r11, r15); /*1,3,5,7,9,11,13,15 set 2*/ \
668 /* Don't need to compute 3 and 7*/ \
669 \
670 /*Final merge*/ \
671 r8 = vec_mergeh(r0, r4); /*all set 0*/ \
672 r9 = vec_mergel(r0, r4); /*all set 1*/ \
673 r10 = vec_mergeh(r1, r5); /*all set 2*/ \
674 r11 = vec_mergel(r1, r5); /*all set 3*/ \
675 r12 = vec_mergeh(r2, r6); /*all set 4*/ \
676 r13 = vec_mergel(r2, r6); /*all set 5*/ \
677 /* Don't need to compute 14 and 15*/ \
678 \
679}
680
681// out: o = |x-y| < a
a6b4448c
LB
682static inline vec_u8 diff_lt_altivec ( register vec_u8 x,
683 register vec_u8 y,
684 register vec_u8 a) {
685
686 register vec_u8 diff = vec_subs(x, y);
687 register vec_u8 diffneg = vec_subs(y, x);
688 register vec_u8 o = vec_or(diff, diffneg); /* |x-y| */
689 o = (vec_u8)vec_cmplt(o, a);
f057cc03
GB
690 return o;
691}
692
a6b4448c
LB
693static inline vec_u8 h264_deblock_mask ( register vec_u8 p0,
694 register vec_u8 p1,
695 register vec_u8 q0,
696 register vec_u8 q1,
697 register vec_u8 alpha,
698 register vec_u8 beta) {
f057cc03 699
a6b4448c
LB
700 register vec_u8 mask;
701 register vec_u8 tempmask;
f057cc03
GB
702
703 mask = diff_lt_altivec(p0, q0, alpha);
704 tempmask = diff_lt_altivec(p1, p0, beta);
705 mask = vec_and(mask, tempmask);
706 tempmask = diff_lt_altivec(q1, q0, beta);
707 mask = vec_and(mask, tempmask);
708
709 return mask;
710}
711
22fa38f0 712// out: newp1 = clip((p2 + ((p0 + q0 + 1) >> 1)) >> 1, p1-tc0, p1+tc0)
a6b4448c
LB
713static inline vec_u8 h264_deblock_q1(register vec_u8 p0,
714 register vec_u8 p1,
715 register vec_u8 p2,
716 register vec_u8 q0,
717 register vec_u8 tc0) {
718
719 register vec_u8 average = vec_avg(p0, q0);
720 register vec_u8 temp;
721 register vec_u8 uncliped;
722 register vec_u8 ones;
723 register vec_u8 max;
724 register vec_u8 min;
725 register vec_u8 newp1;
963eca22
GP
726
727 temp = vec_xor(average, p2);
728 average = vec_avg(average, p2); /*avg(p2, avg(p0, q0)) */
729 ones = vec_splat_u8(1);
730 temp = vec_and(temp, ones); /*(p2^avg(p0, q0)) & 1 */
731 uncliped = vec_subs(average, temp); /*(p2+((p0+q0+1)>>1))>>1 */
732 max = vec_adds(p1, tc0);
733 min = vec_subs(p1, tc0);
22fa38f0
GB
734 newp1 = vec_max(min, uncliped);
735 newp1 = vec_min(max, newp1);
736 return newp1;
f057cc03
GB
737}
738
739#define h264_deblock_p0_q0(p0, p1, q0, q1, tc0masked) { \
740 \
a6b4448c 741 const vec_u8 A0v = vec_sl(vec_splat_u8(10), vec_splat_u8(4)); \
f057cc03 742 \
a6b4448c
LB
743 register vec_u8 pq0bit = vec_xor(p0,q0); \
744 register vec_u8 q1minus; \
745 register vec_u8 p0minus; \
746 register vec_u8 stage1; \
747 register vec_u8 stage2; \
748 register vec_u8 vec160; \
749 register vec_u8 delta; \
750 register vec_u8 deltaneg; \
f057cc03 751 \
f4a02f6e 752 q1minus = vec_nor(q1, q1); /* 255 - q1 */ \
f057cc03
GB
753 stage1 = vec_avg(p1, q1minus); /* (p1 - q1 + 256)>>1 */ \
754 stage2 = vec_sr(stage1, vec_splat_u8(1)); /* (p1 - q1 + 256)>>2 = 64 + (p1 - q1) >> 2 */ \
f4a02f6e 755 p0minus = vec_nor(p0, p0); /* 255 - p0 */ \
f057cc03
GB
756 stage1 = vec_avg(q0, p0minus); /* (q0 - p0 + 256)>>1 */ \
757 pq0bit = vec_and(pq0bit, vec_splat_u8(1)); \
758 stage2 = vec_avg(stage2, pq0bit); /* 32 + ((q0 - p0)&1 + (p1 - q1) >> 2 + 1) >> 1 */ \
759 stage2 = vec_adds(stage2, stage1); /* 160 + ((p0 - q0) + (p1 - q1) >> 2 + 1) >> 1 */ \
760 vec160 = vec_ld(0, &A0v); \
761 deltaneg = vec_subs(vec160, stage2); /* -d */ \
762 delta = vec_subs(stage2, vec160); /* d */ \
763 deltaneg = vec_min(tc0masked, deltaneg); \
764 delta = vec_min(tc0masked, delta); \
765 p0 = vec_subs(p0, deltaneg); \
766 q0 = vec_subs(q0, delta); \
767 p0 = vec_adds(p0, delta); \
768 q0 = vec_adds(q0, deltaneg); \
769}
770
771#define h264_loop_filter_luma_altivec(p2, p1, p0, q0, q1, q2, alpha, beta, tc0) { \
84dc2d8a 772 DECLARE_ALIGNED(16, unsigned char, temp)[16]; \
a6b4448c
LB
773 register vec_u8 alphavec; \
774 register vec_u8 betavec; \
775 register vec_u8 mask; \
776 register vec_u8 p1mask; \
777 register vec_u8 q1mask; \
e970d98c 778 register vector signed char tc0vec; \
a6b4448c
LB
779 register vec_u8 finaltc0; \
780 register vec_u8 tc0masked; \
781 register vec_u8 newp1; \
782 register vec_u8 newq1; \
f057cc03
GB
783 \
784 temp[0] = alpha; \
785 temp[1] = beta; \
786 alphavec = vec_ld(0, temp); \
787 betavec = vec_splat(alphavec, 0x1); \
788 alphavec = vec_splat(alphavec, 0x0); \
789 mask = h264_deblock_mask(p0, p1, q0, q1, alphavec, betavec); /*if in block */ \
790 \
791 *((int *)temp) = *((int *)tc0); \
e970d98c 792 tc0vec = vec_ld(0, (signed char*)temp); \
f057cc03
GB
793 tc0vec = vec_mergeh(tc0vec, tc0vec); \
794 tc0vec = vec_mergeh(tc0vec, tc0vec); \
22fa38f0 795 mask = vec_and(mask, vec_cmpgt(tc0vec, vec_splat_s8(-1))); /* if tc0[i] >= 0 */ \
a6b4448c 796 finaltc0 = vec_and((vec_u8)tc0vec, mask); /* tc = tc0 */ \
f057cc03
GB
797 \
798 p1mask = diff_lt_altivec(p2, p0, betavec); \
830bf1f2 799 p1mask = vec_and(p1mask, mask); /* if ( |p2 - p0| < beta) */ \
a6b4448c 800 tc0masked = vec_and(p1mask, (vec_u8)tc0vec); \
f057cc03 801 finaltc0 = vec_sub(finaltc0, p1mask); /* tc++ */ \
22fa38f0 802 newp1 = h264_deblock_q1(p0, p1, p2, q0, tc0masked); \
f057cc03
GB
803 /*end if*/ \
804 \
805 q1mask = diff_lt_altivec(q2, q0, betavec); \
806 q1mask = vec_and(q1mask, mask); /* if ( |q2 - q0| < beta ) */\
a6b4448c 807 tc0masked = vec_and(q1mask, (vec_u8)tc0vec); \
f057cc03 808 finaltc0 = vec_sub(finaltc0, q1mask); /* tc++ */ \
22fa38f0 809 newq1 = h264_deblock_q1(p0, q1, q2, q0, tc0masked); \
f057cc03
GB
810 /*end if*/ \
811 \
812 h264_deblock_p0_q0(p0, p1, q0, q1, finaltc0); \
22fa38f0
GB
813 p1 = newp1; \
814 q1 = newq1; \
f057cc03
GB
815}
816
817static void h264_v_loop_filter_luma_altivec(uint8_t *pix, int stride, int alpha, int beta, int8_t *tc0) {
818
830bf1f2 819 if ((tc0[0] & tc0[1] & tc0[2] & tc0[3]) >= 0) {
a6b4448c
LB
820 register vec_u8 p2 = vec_ld(-3*stride, pix);
821 register vec_u8 p1 = vec_ld(-2*stride, pix);
822 register vec_u8 p0 = vec_ld(-1*stride, pix);
823 register vec_u8 q0 = vec_ld(0, pix);
824 register vec_u8 q1 = vec_ld(stride, pix);
825 register vec_u8 q2 = vec_ld(2*stride, pix);
f057cc03
GB
826 h264_loop_filter_luma_altivec(p2, p1, p0, q0, q1, q2, alpha, beta, tc0);
827 vec_st(p1, -2*stride, pix);
828 vec_st(p0, -1*stride, pix);
829 vec_st(q0, 0, pix);
830 vec_st(q1, stride, pix);
831 }
832}
833
834static void h264_h_loop_filter_luma_altivec(uint8_t *pix, int stride, int alpha, int beta, int8_t *tc0) {
835
a6b4448c 836 register vec_u8 line0, line1, line2, line3, line4, line5;
830bf1f2 837 if ((tc0[0] & tc0[1] & tc0[2] & tc0[3]) < 0)
f057cc03
GB
838 return;
839 readAndTranspose16x6(pix-3, stride, line0, line1, line2, line3, line4, line5);
840 h264_loop_filter_luma_altivec(line0, line1, line2, line3, line4, line5, alpha, beta, tc0);
841 transpose4x16(line1, line2, line3, line4);
842 write16x4(pix-2, stride, line1, line2, line3, line4);
843}
844
8b2bc85f 845static av_always_inline
c2d33742
RB
846void weight_h264_W_altivec(uint8_t *block, int stride, int height,
847 int log2_denom, int weight, int offset, int w)
8b2bc85f
DC
848{
849 int y, aligned;
850 vec_u8 vblock;
851 vec_s16 vtemp, vweight, voffset, v0, v1;
852 vec_u16 vlog2_denom;
84dc2d8a 853 DECLARE_ALIGNED(16, int32_t, temp)[4];
8b2bc85f
DC
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
c2d33742 868 for (y = 0; y < height; y++) {
8b2bc85f
DC
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}
df7fb435
DC
890
891static av_always_inline
c2d33742
RB
892void biweight_h264_W_altivec(uint8_t *dst, uint8_t *src, int stride, int height,
893 int log2_denom, int weightd, int weights, int offset, int w)
df7fb435
DC
894{
895 int y, dst_aligned, src_aligned;
896 vec_u8 vsrc, vdst;
54b90958
DC
897 vec_s16 vtemp, vweights, vweightd, voffset, v0, v1, v2, v3;
898 vec_u16 vlog2_denom;
84dc2d8a 899 DECLARE_ALIGNED(16, int32_t, temp)[4];
df7fb435
DC
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
62865344
DC
908 vtemp = (vec_s16)vec_ld(0, temp);
909 vlog2_denom = (vec_u16)vec_splat(vtemp, 1);
df7fb435
DC
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
c2d33742 916 for (y = 0; y < height; y++) {
df7fb435
DC
917 vdst = vec_ld(0, dst);
918 vsrc = vec_ld(0, src);
919
62865344
DC
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);
df7fb435
DC
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) {
54b90958
DC
933 v0 = vec_mladd(v0, vweightd, zero_s16v);
934 v2 = vec_mladd(v2, vweights, zero_s16v);
df7fb435
DC
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) {
54b90958
DC
941 v1 = vec_mladd(v1, vweightd, zero_s16v);
942 v3 = vec_mladd(v3, vweights, zero_s16v);
df7fb435
DC
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
c2d33742
RB
956#define H264_WEIGHT(W) \
957static void ff_weight_h264_pixels ## W ## _altivec(uint8_t *block, int stride, int height, \
958 int log2_denom, int weight, int offset){ \
b034c95c 959 weight_h264_W_altivec(block, stride, height, log2_denom, weight, offset, W); \
8b2bc85f 960}\
c2d33742
RB
961static void ff_biweight_h264_pixels ## W ## _altivec(uint8_t *dst, uint8_t *src, int stride, int height, \
962 int log2_denom, int weightd, int weights, int offset){ \
b034c95c 963 biweight_h264_W_altivec(dst, src, stride, height, log2_denom, weightd, weights, offset, W); \
df7fb435
DC
964}
965
c2d33742
RB
966H264_WEIGHT(16)
967H264_WEIGHT( 8)
df7fb435 968
a6a12a8a 969void dsputil_h264_init_ppc(DSPContext* c, AVCodecContext *avctx) {
a617c6aa 970 const int high_bit_depth = avctx->bits_per_raw_sample > 8;
115329f1 971
c6c98d08 972 if (av_get_cpu_flags() & AV_CPU_FLAG_ALTIVEC) {
19a0729b 973 if (!high_bit_depth) {
830bf1f2 974 c->put_h264_chroma_pixels_tab[0] = put_h264_chroma_mc8_altivec;
830bf1f2 975 c->avg_h264_chroma_pixels_tab[0] = avg_h264_chroma_mc8_altivec;
a6a12a8a
RD
976
977#define dspfunc(PFX, IDX, NUM) \
830bf1f2
DB
978 c->PFX ## _pixels_tab[IDX][ 0] = PFX ## NUM ## _mc00_altivec; \
979 c->PFX ## _pixels_tab[IDX][ 1] = PFX ## NUM ## _mc10_altivec; \
980 c->PFX ## _pixels_tab[IDX][ 2] = PFX ## NUM ## _mc20_altivec; \
981 c->PFX ## _pixels_tab[IDX][ 3] = PFX ## NUM ## _mc30_altivec; \
982 c->PFX ## _pixels_tab[IDX][ 4] = PFX ## NUM ## _mc01_altivec; \
983 c->PFX ## _pixels_tab[IDX][ 5] = PFX ## NUM ## _mc11_altivec; \
984 c->PFX ## _pixels_tab[IDX][ 6] = PFX ## NUM ## _mc21_altivec; \
985 c->PFX ## _pixels_tab[IDX][ 7] = PFX ## NUM ## _mc31_altivec; \
986 c->PFX ## _pixels_tab[IDX][ 8] = PFX ## NUM ## _mc02_altivec; \
987 c->PFX ## _pixels_tab[IDX][ 9] = PFX ## NUM ## _mc12_altivec; \
988 c->PFX ## _pixels_tab[IDX][10] = PFX ## NUM ## _mc22_altivec; \
989 c->PFX ## _pixels_tab[IDX][11] = PFX ## NUM ## _mc32_altivec; \
990 c->PFX ## _pixels_tab[IDX][12] = PFX ## NUM ## _mc03_altivec; \
991 c->PFX ## _pixels_tab[IDX][13] = PFX ## NUM ## _mc13_altivec; \
992 c->PFX ## _pixels_tab[IDX][14] = PFX ## NUM ## _mc23_altivec; \
993 c->PFX ## _pixels_tab[IDX][15] = PFX ## NUM ## _mc33_altivec
994
995 dspfunc(put_h264_qpel, 0, 16);
996 dspfunc(avg_h264_qpel, 0, 16);
a6a12a8a 997#undef dspfunc
4693b031 998 }
19a0729b 999 }
4693b031
MR
1000}
1001
76741b0e 1002void ff_h264dsp_init_ppc(H264DSPContext *c, const int bit_depth, const int chroma_format_idc)
4693b031 1003{
c6c98d08 1004 if (av_get_cpu_flags() & AV_CPU_FLAG_ALTIVEC) {
19a0729b 1005 if (bit_depth == 8) {
4693b031 1006 c->h264_idct_add = ff_h264_idct_add_altivec;
76741b0e
BC
1007 if (chroma_format_idc == 1)
1008 c->h264_idct_add8 = ff_h264_idct_add8_altivec;
4693b031
MR
1009 c->h264_idct_add16 = ff_h264_idct_add16_altivec;
1010 c->h264_idct_add16intra = ff_h264_idct_add16intra_altivec;
1011 c->h264_idct_dc_add= h264_idct_dc_add_altivec;
1012 c->h264_idct8_dc_add = ff_h264_idct8_dc_add_altivec;
1013 c->h264_idct8_add = ff_h264_idct8_add_altivec;
1014 c->h264_idct8_add4 = ff_h264_idct8_add4_altivec;
1015 c->h264_v_loop_filter_luma= h264_v_loop_filter_luma_altivec;
1016 c->h264_h_loop_filter_luma= h264_h_loop_filter_luma_altivec;
df7fb435 1017
c2d33742
RB
1018 c->weight_h264_pixels_tab[0] = ff_weight_h264_pixels16_altivec;
1019 c->weight_h264_pixels_tab[1] = ff_weight_h264_pixels8_altivec;
1020 c->biweight_h264_pixels_tab[0] = ff_biweight_h264_pixels16_altivec;
1021 c->biweight_h264_pixels_tab[1] = ff_biweight_h264_pixels8_altivec;
830bf1f2 1022 }
19a0729b 1023 }
a6a12a8a 1024}