h264 qpel mc, size 16 patch by (Romain Dolbeau <dolbeau at caps-entreprise dot com>)
[libav.git] / libavcodec / ppc / dsputil_h264_altivec.c
CommitLineData
a6a12a8a
RD
1/*
2 * Copyright (c) 2004 Romain Dolbeau <romain@dolbeau.org>
3 *
4 * This library is free software; you can redistribute it and/or
5 * modify it under the terms of the GNU Lesser General Public
6 * License as published by the Free Software Foundation; either
7 * version 2 of the License, or (at your option) any later version.
8 *
9 * This library is distributed in the hope that it will be useful,
10 * but WITHOUT ANY WARRANTY; without even the implied warranty of
11 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
12 * Lesser General Public License for more details.
13 *
14 * You should have received a copy of the GNU Lesser General Public
15 * License along with this library; if not, write to the Free Software
16 * Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA
17 */
18
19#include "../dsputil.h"
20
21#include "gcc_fixes.h"
22
23#include "dsputil_altivec.h"
24
25#define PUT_OP_U8_ALTIVEC(d, s, dst) d = s
26#define AVG_OP_U8_ALTIVEC(d, s, dst) d = vec_avg(dst, s)
27
28#define OP_U8_ALTIVEC PUT_OP_U8_ALTIVEC
29#define PREFIX_h264_chroma_mc8_altivec put_h264_chroma_mc8_altivec
30#define PREFIX_h264_chroma_mc8_num altivec_put_h264_chroma_mc8_num
31#define PREFIX_h264_qpel16_h_lowpass_altivec put_h264_qpel16_h_lowpass_altivec
32#define PREFIX_h264_qpel16_h_lowpass_num altivec_put_h264_qpel16_h_lowpass_num
33#define PREFIX_h264_qpel16_v_lowpass_altivec put_h264_qpel16_v_lowpass_altivec
34#define PREFIX_h264_qpel16_v_lowpass_num altivec_put_h264_qpel16_v_lowpass_num
35#define PREFIX_h264_qpel16_hv_lowpass_altivec put_h264_qpel16_hv_lowpass_altivec
36#define PREFIX_h264_qpel16_hv_lowpass_num altivec_put_h264_qpel16_hv_lowpass_num
37#include "dsputil_h264_template_altivec.c"
38#undef OP_U8_ALTIVEC
39#undef PREFIX_h264_chroma_mc8_altivec
40#undef PREFIX_h264_chroma_mc8_num
41#undef PREFIX_h264_qpel16_h_lowpass_altivec
42#undef PREFIX_h264_qpel16_h_lowpass_num
43#undef PREFIX_h264_qpel16_v_lowpass_altivec
44#undef PREFIX_h264_qpel16_v_lowpass_num
45#undef PREFIX_h264_qpel16_hv_lowpass_altivec
46#undef PREFIX_h264_qpel16_hv_lowpass_num
47
48#define OP_U8_ALTIVEC AVG_OP_U8_ALTIVEC
49#define PREFIX_h264_chroma_mc8_altivec avg_h264_chroma_mc8_altivec
50#define PREFIX_h264_chroma_mc8_num altivec_avg_h264_chroma_mc8_num
51#define PREFIX_h264_qpel16_h_lowpass_altivec avg_h264_qpel16_h_lowpass_altivec
52#define PREFIX_h264_qpel16_h_lowpass_num altivec_avg_h264_qpel16_h_lowpass_num
53#define PREFIX_h264_qpel16_v_lowpass_altivec avg_h264_qpel16_v_lowpass_altivec
54#define PREFIX_h264_qpel16_v_lowpass_num altivec_avg_h264_qpel16_v_lowpass_num
55#define PREFIX_h264_qpel16_hv_lowpass_altivec avg_h264_qpel16_hv_lowpass_altivec
56#define PREFIX_h264_qpel16_hv_lowpass_num altivec_avg_h264_qpel16_hv_lowpass_num
57#include "dsputil_h264_template_altivec.c"
58#undef OP_U8_ALTIVEC
59#undef PREFIX_h264_chroma_mc8_altivec
60#undef PREFIX_h264_chroma_mc8_num
61#undef PREFIX_h264_qpel16_h_lowpass_altivec
62#undef PREFIX_h264_qpel16_h_lowpass_num
63#undef PREFIX_h264_qpel16_v_lowpass_altivec
64#undef PREFIX_h264_qpel16_v_lowpass_num
65#undef PREFIX_h264_qpel16_hv_lowpass_altivec
66#undef PREFIX_h264_qpel16_hv_lowpass_num
67
68#define H264_MC(OPNAME, SIZE, CODETYPE) \
69static void OPNAME ## h264_qpel ## SIZE ## _mc00_ ## CODETYPE (uint8_t *dst, uint8_t *src, int stride){\
70 OPNAME ## pixels ## SIZE ## _ ## CODETYPE(dst, src, stride, SIZE);\
71}\
72\
73static void OPNAME ## h264_qpel ## SIZE ## _mc10_ ## CODETYPE(uint8_t *dst, uint8_t *src, int stride){ \
74 uint64_t temp[SIZE*SIZE/8] __align16;\
75 uint8_t * const half= (uint8_t*)temp;\
76 put_h264_qpel ## SIZE ## _h_lowpass_ ## CODETYPE(half, src, SIZE, stride);\
77 OPNAME ## pixels ## SIZE ## _l2_ ## CODETYPE(dst, src, half, stride, stride, SIZE);\
78}\
79\
80static void OPNAME ## h264_qpel ## SIZE ## _mc20_ ## CODETYPE(uint8_t *dst, uint8_t *src, int stride){\
81 OPNAME ## h264_qpel ## SIZE ## _h_lowpass_ ## CODETYPE(dst, src, stride, stride);\
82}\
83\
84static void OPNAME ## h264_qpel ## SIZE ## _mc30_ ## CODETYPE(uint8_t *dst, uint8_t *src, int stride){\
85 uint64_t temp[SIZE*SIZE/8] __align16;\
86 uint8_t * const half= (uint8_t*)temp;\
87 put_h264_qpel ## SIZE ## _h_lowpass_ ## CODETYPE(half, src, SIZE, stride);\
88 OPNAME ## pixels ## SIZE ## _l2_ ## CODETYPE(dst, src+1, half, stride, stride, SIZE);\
89}\
90\
91static void OPNAME ## h264_qpel ## SIZE ## _mc01_ ## CODETYPE(uint8_t *dst, uint8_t *src, int stride){\
92 uint64_t temp[SIZE*SIZE/8] __align16;\
93 uint8_t * const half= (uint8_t*)temp;\
94 put_h264_qpel ## SIZE ## _v_lowpass_ ## CODETYPE(half, src, SIZE, stride);\
95 OPNAME ## pixels ## SIZE ## _l2_ ## CODETYPE(dst, src, half, stride, stride, SIZE);\
96}\
97\
98static void OPNAME ## h264_qpel ## SIZE ## _mc02_ ## CODETYPE(uint8_t *dst, uint8_t *src, int stride){\
99 OPNAME ## h264_qpel ## SIZE ## _v_lowpass_ ## CODETYPE(dst, src, stride, stride);\
100}\
101\
102static void OPNAME ## h264_qpel ## SIZE ## _mc03_ ## CODETYPE(uint8_t *dst, uint8_t *src, int stride){\
103 uint64_t temp[SIZE*SIZE/8] __align16;\
104 uint8_t * const half= (uint8_t*)temp;\
105 put_h264_qpel ## SIZE ## _v_lowpass_ ## CODETYPE(half, src, SIZE, stride);\
106 OPNAME ## pixels ## SIZE ## _l2_ ## CODETYPE(dst, src+stride, half, stride, stride, SIZE);\
107}\
108\
109static void OPNAME ## h264_qpel ## SIZE ## _mc11_ ## CODETYPE(uint8_t *dst, uint8_t *src, int stride){\
110 uint64_t temp[SIZE*SIZE/4] __align16;\
111 uint8_t * const halfH= (uint8_t*)temp;\
112 uint8_t * const halfV= ((uint8_t*)temp) + SIZE*SIZE;\
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){\
119 uint64_t temp[SIZE*SIZE/4] __align16;\
120 uint8_t * const halfH= (uint8_t*)temp;\
121 uint8_t * const halfV= ((uint8_t*)temp) + SIZE*SIZE;\
122 put_h264_qpel ## SIZE ## _h_lowpass_ ## CODETYPE(halfH, src, SIZE, stride);\
123 put_h264_qpel ## SIZE ## _v_lowpass_ ## CODETYPE(halfV, src+1, SIZE, stride);\
124 OPNAME ## pixels ## SIZE ## _l2_ ## CODETYPE(dst, halfH, halfV, stride, SIZE, SIZE);\
125}\
126\
127static void OPNAME ## h264_qpel ## SIZE ## _mc13_ ## CODETYPE(uint8_t *dst, uint8_t *src, int stride){\
128 uint64_t temp[SIZE*SIZE/4] __align16;\
129 uint8_t * const halfH= (uint8_t*)temp;\
130 uint8_t * const halfV= ((uint8_t*)temp) + SIZE*SIZE;\
131 put_h264_qpel ## SIZE ## _h_lowpass_ ## CODETYPE(halfH, src + stride, SIZE, stride);\
132 put_h264_qpel ## SIZE ## _v_lowpass_ ## CODETYPE(halfV, src, SIZE, stride);\
133 OPNAME ## pixels ## SIZE ## _l2_ ## CODETYPE(dst, halfH, halfV, stride, SIZE, SIZE);\
134}\
135\
136static void OPNAME ## h264_qpel ## SIZE ## _mc33_ ## CODETYPE(uint8_t *dst, uint8_t *src, int stride){\
137 uint64_t temp[SIZE*SIZE/4] __align16;\
138 uint8_t * const halfH= (uint8_t*)temp;\
139 uint8_t * const halfV= ((uint8_t*)temp) + 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\
145static void OPNAME ## h264_qpel ## SIZE ## _mc22_ ## CODETYPE(uint8_t *dst, uint8_t *src, int stride){\
146 uint64_t temp[SIZE*(SIZE+8)/4] __align16;\
147 int16_t * const tmp= (int16_t*)temp;\
148 OPNAME ## h264_qpel ## SIZE ## _hv_lowpass_ ## CODETYPE(dst, tmp, src, stride, SIZE, stride);\
149}\
150\
151static void OPNAME ## h264_qpel ## SIZE ## _mc21_ ## CODETYPE(uint8_t *dst, uint8_t *src, int stride){\
152 uint64_t temp[SIZE*(SIZE+8)/4 + SIZE*SIZE/4] __align16;\
153 uint8_t * const halfH= (uint8_t*)temp;\
154 uint8_t * const halfHV= ((uint8_t*)temp) + SIZE*SIZE;\
155 int16_t * const tmp= ((int16_t*)temp) + SIZE*SIZE;\
156 put_h264_qpel ## SIZE ## _h_lowpass_ ## CODETYPE(halfH, src, SIZE, stride);\
157 put_h264_qpel ## SIZE ## _hv_lowpass_ ## CODETYPE(halfHV, tmp, src, SIZE, SIZE, stride);\
158 OPNAME ## pixels ## SIZE ## _l2_ ## CODETYPE(dst, halfH, halfHV, stride, SIZE, SIZE);\
159}\
160\
161static void OPNAME ## h264_qpel ## SIZE ## _mc23_ ## CODETYPE(uint8_t *dst, uint8_t *src, int stride){\
162 uint64_t temp[SIZE*(SIZE+8)/4 + SIZE*SIZE/4] __align16;\
163 uint8_t * const halfH= (uint8_t*)temp;\
164 uint8_t * const halfHV= ((uint8_t*)temp) + SIZE*SIZE;\
165 int16_t * const tmp= ((int16_t*)temp) + SIZE*SIZE;\
166 put_h264_qpel ## SIZE ## _h_lowpass_ ## CODETYPE(halfH, src + stride, SIZE, stride);\
167 put_h264_qpel ## SIZE ## _hv_lowpass_ ## CODETYPE(halfHV, tmp, src, SIZE, SIZE, stride);\
168 OPNAME ## pixels ## SIZE ## _l2_ ## CODETYPE(dst, halfH, halfHV, stride, SIZE, SIZE);\
169}\
170\
171static void OPNAME ## h264_qpel ## SIZE ## _mc12_ ## CODETYPE(uint8_t *dst, uint8_t *src, int stride){\
172 uint64_t temp[SIZE*(SIZE+8)/4 + SIZE*SIZE/4] __align16;\
173 uint8_t * const halfV= (uint8_t*)temp;\
174 uint8_t * const halfHV= ((uint8_t*)temp) + SIZE*SIZE;\
175 int16_t * const tmp= ((int16_t*)temp) + SIZE*SIZE;\
176 put_h264_qpel ## SIZE ## _v_lowpass_ ## CODETYPE(halfV, src, SIZE, stride);\
177 put_h264_qpel ## SIZE ## _hv_lowpass_ ## CODETYPE(halfHV, tmp, src, SIZE, SIZE, stride);\
178 OPNAME ## pixels ## SIZE ## _l2_ ## CODETYPE(dst, halfV, halfHV, stride, SIZE, SIZE);\
179}\
180\
181static void OPNAME ## h264_qpel ## SIZE ## _mc32_ ## CODETYPE(uint8_t *dst, uint8_t *src, int stride){\
182 uint64_t temp[SIZE*(SIZE+8)/4 + SIZE*SIZE/4] __align16;\
183 uint8_t * const halfV= (uint8_t*)temp;\
184 uint8_t * const halfHV= ((uint8_t*)temp) + SIZE*SIZE;\
185 int16_t * const tmp= ((int16_t*)temp) + SIZE*SIZE;\
186 put_h264_qpel ## SIZE ## _v_lowpass_ ## CODETYPE(halfV, src+1, SIZE, stride);\
187 put_h264_qpel ## SIZE ## _hv_lowpass_ ## CODETYPE(halfHV, tmp, src, SIZE, SIZE, stride);\
188 OPNAME ## pixels ## SIZE ## _l2_ ## CODETYPE(dst, halfV, halfHV, stride, SIZE, SIZE);\
189}\
190
191
192/* from dsputil.c */
193static inline void put_pixels8_l2(uint8_t * dst, const uint8_t * src1, const uint8_t * src2, int dst_stride, int src_stride1, int src_stride2, int h) {
194 int i;
195 for (i = 0; i < h; i++) {
196 uint32_t a, b;
197 a = (((const struct unaligned_32 *) (&src1[i * src_stride1]))->l);
198 b = (((const struct unaligned_32 *) (&src2[i * src_stride2]))->l);
199 *((uint32_t *) & dst[i * dst_stride]) = rnd_avg32(a, b);
200 a = (((const struct unaligned_32 *) (&src1[i * src_stride1 + 4]))->l);
201 b = (((const struct unaligned_32 *) (&src2[i * src_stride2 + 4]))->l);
202 *((uint32_t *) & dst[i * dst_stride + 4]) = rnd_avg32(a, b);
203 }
204} static inline void avg_pixels8_l2(uint8_t * dst, const uint8_t * src1, const uint8_t * src2, int dst_stride, int src_stride1, int src_stride2, int h) {
205 int i;
206 for (i = 0; i < h; i++) {
207 uint32_t a, b;
208 a = (((const struct unaligned_32 *) (&src1[i * src_stride1]))->l);
209 b = (((const struct unaligned_32 *) (&src2[i * src_stride2]))->l);
210 *((uint32_t *) & dst[i * dst_stride]) = rnd_avg32(*((uint32_t *) & dst[i * dst_stride]), rnd_avg32(a, b));
211 a = (((const struct unaligned_32 *) (&src1[i * src_stride1 + 4]))->l);
212 b = (((const struct unaligned_32 *) (&src2[i * src_stride2 + 4]))->l);
213 *((uint32_t *) & dst[i * dst_stride + 4]) = rnd_avg32(*((uint32_t *) & dst[i * dst_stride + 4]), rnd_avg32(a, b));
214 }
215} static inline void put_pixels16_l2(uint8_t * dst, const uint8_t * src1, const uint8_t * src2, int dst_stride, int src_stride1, int src_stride2, int h) {
216 put_pixels8_l2(dst, src1, src2, dst_stride, src_stride1, src_stride2, h);
217 put_pixels8_l2(dst + 8, src1 + 8, src2 + 8, dst_stride, src_stride1, src_stride2, h);
218} static inline void avg_pixels16_l2(uint8_t * dst, const uint8_t * src1, const uint8_t * src2, int dst_stride, int src_stride1, int src_stride2, int h) {
219 avg_pixels8_l2(dst, src1, src2, dst_stride, src_stride1, src_stride2, h);
220 avg_pixels8_l2(dst + 8, src1 + 8, src2 + 8, dst_stride, src_stride1, src_stride2, h);
221}
222
223/* UNIMPLEMENTED YET !! */
224#define put_pixels16_l2_altivec(d,s1,s2,ds,s1s,h) put_pixels16_l2(d,s1,s2,ds,s1s,16,h)
225#define avg_pixels16_l2_altivec(d,s1,s2,ds,s1s,h) avg_pixels16_l2(d,s1,s2,ds,s1s,16,h)
226
227H264_MC(put_, 16, altivec)
228 H264_MC(avg_, 16, altivec)
229
230void dsputil_h264_init_ppc(DSPContext* c, AVCodecContext *avctx) {
231
232#ifdef HAVE_ALTIVEC
233 if (has_altivec()) {
234 c->put_h264_chroma_pixels_tab[0] = put_h264_chroma_mc8_altivec;
235 c->avg_h264_chroma_pixels_tab[0] = avg_h264_chroma_mc8_altivec;
236
237#define dspfunc(PFX, IDX, NUM) \
238 c->PFX ## _pixels_tab[IDX][ 0] = PFX ## NUM ## _mc00_altivec; \
239 c->PFX ## _pixels_tab[IDX][ 1] = PFX ## NUM ## _mc10_altivec; \
240 c->PFX ## _pixels_tab[IDX][ 2] = PFX ## NUM ## _mc20_altivec; \
241 c->PFX ## _pixels_tab[IDX][ 3] = PFX ## NUM ## _mc30_altivec; \
242 c->PFX ## _pixels_tab[IDX][ 4] = PFX ## NUM ## _mc01_altivec; \
243 c->PFX ## _pixels_tab[IDX][ 5] = PFX ## NUM ## _mc11_altivec; \
244 c->PFX ## _pixels_tab[IDX][ 6] = PFX ## NUM ## _mc21_altivec; \
245 c->PFX ## _pixels_tab[IDX][ 7] = PFX ## NUM ## _mc31_altivec; \
246 c->PFX ## _pixels_tab[IDX][ 8] = PFX ## NUM ## _mc02_altivec; \
247 c->PFX ## _pixels_tab[IDX][ 9] = PFX ## NUM ## _mc12_altivec; \
248 c->PFX ## _pixels_tab[IDX][10] = PFX ## NUM ## _mc22_altivec; \
249 c->PFX ## _pixels_tab[IDX][11] = PFX ## NUM ## _mc32_altivec; \
250 c->PFX ## _pixels_tab[IDX][12] = PFX ## NUM ## _mc03_altivec; \
251 c->PFX ## _pixels_tab[IDX][13] = PFX ## NUM ## _mc13_altivec; \
252 c->PFX ## _pixels_tab[IDX][14] = PFX ## NUM ## _mc23_altivec; \
253 c->PFX ## _pixels_tab[IDX][15] = PFX ## NUM ## _mc33_altivec
254
255 dspfunc(put_h264_qpel, 0, 16);
256 dspfunc(avg_h264_qpel, 0, 16);
257#undef dspfunc
258
259 } else
260#endif /* HAVE_ALTIVEC */
261 {
262 // Non-AltiVec PPC optimisations
263
264 // ... pending ...
265 }
266}