FFmpeg
 All Data Structures Namespaces Files Functions Variables Typedefs Enumerations Enumerator Macros Groups Pages
mpegvideoencdsp.c
Go to the documentation of this file.
1 /*
2  * This file is part of FFmpeg.
3  *
4  * FFmpeg 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 2.1 of the License, or (at your option) any later version.
8  *
9  * FFmpeg 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 FFmpeg; if not, write to the Free Software
16  * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
17  */
18 
19 #include <assert.h>
20 #include <stdint.h>
21 #include <string.h>
22 
23 #include "config.h"
24 #include "libavutil/avassert.h"
25 #include "libavutil/attributes.h"
26 #include "libavutil/imgutils.h"
27 #include "avcodec.h"
28 #include "me_cmp.h"
29 #include "mpegvideoencdsp.h"
30 
31 static int try_8x8basis_c(int16_t rem[64], int16_t weight[64],
32  int16_t basis[64], int scale)
33 {
34  int i;
35  unsigned int sum = 0;
36 
37  for (i = 0; i < 8 * 8; i++) {
38  int b = rem[i] + ((basis[i] * scale +
39  (1 << (BASIS_SHIFT - RECON_SHIFT - 1))) >>
41  int w = weight[i];
42  b >>= RECON_SHIFT;
43  av_assert2(-512 < b && b < 512);
44 
45  sum += (w * b) * (w * b) >> 4;
46  }
47  return sum >> 2;
48 }
49 
50 static void add_8x8basis_c(int16_t rem[64], int16_t basis[64], int scale)
51 {
52  int i;
53 
54  for (i = 0; i < 8 * 8; i++)
55  rem[i] += (basis[i] * scale +
56  (1 << (BASIS_SHIFT - RECON_SHIFT - 1))) >>
58 }
59 
60 static int pix_sum_c(uint8_t *pix, int line_size)
61 {
62  int s = 0, i, j;
63 
64  for (i = 0; i < 16; i++) {
65  for (j = 0; j < 16; j += 8) {
66  s += pix[0];
67  s += pix[1];
68  s += pix[2];
69  s += pix[3];
70  s += pix[4];
71  s += pix[5];
72  s += pix[6];
73  s += pix[7];
74  pix += 8;
75  }
76  pix += line_size - 16;
77  }
78  return s;
79 }
80 
81 static int pix_norm1_c(uint8_t *pix, int line_size)
82 {
83  int s = 0, i, j;
84  uint32_t *sq = ff_square_tab + 256;
85 
86  for (i = 0; i < 16; i++) {
87  for (j = 0; j < 16; j += 8) {
88 #if 0
89  s += sq[pix[0]];
90  s += sq[pix[1]];
91  s += sq[pix[2]];
92  s += sq[pix[3]];
93  s += sq[pix[4]];
94  s += sq[pix[5]];
95  s += sq[pix[6]];
96  s += sq[pix[7]];
97 #else
98 #if HAVE_FAST_64BIT
99  register uint64_t x = *(uint64_t *) pix;
100  s += sq[x & 0xff];
101  s += sq[(x >> 8) & 0xff];
102  s += sq[(x >> 16) & 0xff];
103  s += sq[(x >> 24) & 0xff];
104  s += sq[(x >> 32) & 0xff];
105  s += sq[(x >> 40) & 0xff];
106  s += sq[(x >> 48) & 0xff];
107  s += sq[(x >> 56) & 0xff];
108 #else
109  register uint32_t x = *(uint32_t *) pix;
110  s += sq[x & 0xff];
111  s += sq[(x >> 8) & 0xff];
112  s += sq[(x >> 16) & 0xff];
113  s += sq[(x >> 24) & 0xff];
114  x = *(uint32_t *) (pix + 4);
115  s += sq[x & 0xff];
116  s += sq[(x >> 8) & 0xff];
117  s += sq[(x >> 16) & 0xff];
118  s += sq[(x >> 24) & 0xff];
119 #endif
120 #endif
121  pix += 8;
122  }
123  pix += line_size - 16;
124  }
125  return s;
126 }
127 
128 /* draw the edges of width 'w' of an image of size width, height */
129 // FIXME: Check that this is OK for MPEG-4 interlaced.
130 static void draw_edges_8_c(uint8_t *buf, int wrap, int width, int height,
131  int w, int h, int sides)
132 {
133  uint8_t *ptr = buf, *last_line;
134  int i;
135 
136  /* left and right */
137  for (i = 0; i < height; i++) {
138  memset(ptr - w, ptr[0], w);
139  memset(ptr + width, ptr[width - 1], w);
140  ptr += wrap;
141  }
142 
143  /* top and bottom + corners */
144  buf -= w;
145  last_line = buf + (height - 1) * wrap;
146  if (sides & EDGE_TOP)
147  for (i = 0; i < h; i++)
148  // top
149  memcpy(buf - (i + 1) * wrap, buf, width + w + w);
150  if (sides & EDGE_BOTTOM)
151  for (i = 0; i < h; i++)
152  // bottom
153  memcpy(last_line + (i + 1) * wrap, last_line, width + w + w);
154 }
155 
156 /* 2x2 -> 1x1 */
157 static void shrink22(uint8_t *dst, int dst_wrap,
158  const uint8_t *src, int src_wrap,
159  int width, int height)
160 {
161  int w;
162  const uint8_t *s1, *s2;
163  uint8_t *d;
164 
165  for (; height > 0; height--) {
166  s1 = src;
167  s2 = s1 + src_wrap;
168  d = dst;
169  for (w = width; w >= 4; w -= 4) {
170  d[0] = (s1[0] + s1[1] + s2[0] + s2[1] + 2) >> 2;
171  d[1] = (s1[2] + s1[3] + s2[2] + s2[3] + 2) >> 2;
172  d[2] = (s1[4] + s1[5] + s2[4] + s2[5] + 2) >> 2;
173  d[3] = (s1[6] + s1[7] + s2[6] + s2[7] + 2) >> 2;
174  s1 += 8;
175  s2 += 8;
176  d += 4;
177  }
178  for (; w > 0; w--) {
179  d[0] = (s1[0] + s1[1] + s2[0] + s2[1] + 2) >> 2;
180  s1 += 2;
181  s2 += 2;
182  d++;
183  }
184  src += 2 * src_wrap;
185  dst += dst_wrap;
186  }
187 }
188 
189 /* 4x4 -> 1x1 */
190 static void shrink44(uint8_t *dst, int dst_wrap,
191  const uint8_t *src, int src_wrap,
192  int width, int height)
193 {
194  int w;
195  const uint8_t *s1, *s2, *s3, *s4;
196  uint8_t *d;
197 
198  for (; height > 0; height--) {
199  s1 = src;
200  s2 = s1 + src_wrap;
201  s3 = s2 + src_wrap;
202  s4 = s3 + src_wrap;
203  d = dst;
204  for (w = width; w > 0; w--) {
205  d[0] = (s1[0] + s1[1] + s1[2] + s1[3] +
206  s2[0] + s2[1] + s2[2] + s2[3] +
207  s3[0] + s3[1] + s3[2] + s3[3] +
208  s4[0] + s4[1] + s4[2] + s4[3] + 8) >> 4;
209  s1 += 4;
210  s2 += 4;
211  s3 += 4;
212  s4 += 4;
213  d++;
214  }
215  src += 4 * src_wrap;
216  dst += dst_wrap;
217  }
218 }
219 
220 /* 8x8 -> 1x1 */
221 static void shrink88(uint8_t *dst, int dst_wrap,
222  const uint8_t *src, int src_wrap,
223  int width, int height)
224 {
225  int w, i;
226 
227  for (; height > 0; height--) {
228  for(w = width;w > 0; w--) {
229  int tmp = 0;
230  for (i = 0; i < 8; i++) {
231  tmp += src[0] + src[1] + src[2] + src[3] +
232  src[4] + src[5] + src[6] + src[7];
233  src += src_wrap;
234  }
235  *(dst++) = (tmp + 32) >> 6;
236  src += 8 - 8 * src_wrap;
237  }
238  src += 8 * src_wrap - 8 * width;
239  dst += dst_wrap - width;
240  }
241 }
242 
244  AVCodecContext *avctx)
245 {
248 
249  c->shrink[0] = av_image_copy_plane;
250  c->shrink[1] = shrink22;
251  c->shrink[2] = shrink44;
252  c->shrink[3] = shrink88;
253 
254  c->pix_sum = pix_sum_c;
255  c->pix_norm1 = pix_norm1_c;
256 
258 
259  if (ARCH_ARM)
260  ff_mpegvideoencdsp_init_arm(c, avctx);
261  if (ARCH_PPC)
262  ff_mpegvideoencdsp_init_ppc(c, avctx);
263  if (ARCH_X86)
264  ff_mpegvideoencdsp_init_x86(c, avctx);
265  if (ARCH_MIPS)
267 }
int(* try_8x8basis)(int16_t rem[64], int16_t weight[64], int16_t basis[64], int scale)
const char * s
Definition: avisynth_c.h:631
#define RECON_SHIFT
static int pix_norm1_c(uint8_t *pix, int line_size)
misc image utilities
void(* shrink[4])(uint8_t *dst, int dst_wrap, const uint8_t *src, int src_wrap, int width, int height)
void(* add_8x8basis)(int16_t rem[64], int16_t basis[64], int scale)
const char * b
Definition: vf_curves.c:109
av_cold void ff_mpegvideoencdsp_init_ppc(MpegvideoEncDSPContext *c, AVCodecContext *avctx)
#define EDGE_TOP
static void shrink22(uint8_t *dst, int dst_wrap, const uint8_t *src, int src_wrap, int width, int height)
uint32_t ff_square_tab[512]
Definition: me_cmp.c:32
Macro definitions for various function/variable attributes.
av_cold void ff_mpegvideoencdsp_init_mips(MpegvideoEncDSPContext *c, AVCodecContext *avctx)
static void shrink88(uint8_t *dst, int dst_wrap, const uint8_t *src, int src_wrap, int width, int height)
av_cold void ff_mpegvideoencdsp_init(MpegvideoEncDSPContext *c, AVCodecContext *avctx)
uint8_t
#define av_cold
Definition: attributes.h:82
#define av_assert2(cond)
assert() equivalent, that does lie in speed critical code.
Definition: avassert.h:63
#define height
static void draw_edges_8_c(uint8_t *buf, int wrap, int width, int height, int w, int h, int sides)
#define s2
Definition: regdef.h:39
#define wrap(func)
Definition: neontest.h:65
simple assert() macros that are a bit more flexible than ISO C assert().
#define width
int(* pix_sum)(uint8_t *pix, int line_size)
#define s4
Definition: regdef.h:41
#define src
Definition: vp9dsp.c:530
#define s3
Definition: regdef.h:40
static void add_8x8basis_c(int16_t rem[64], int16_t basis[64], int scale)
Libavcodec external API header.
av_cold void ff_mpegvideoencdsp_init_arm(MpegvideoEncDSPContext *c, AVCodecContext *avctx)
main external API structure.
Definition: avcodec.h:1649
void * buf
Definition: avisynth_c.h:553
int(* pix_norm1)(uint8_t *pix, int line_size)
static int pix_sum_c(uint8_t *pix, int line_size)
#define s1
Definition: regdef.h:38
static int weight(int i, int blen, int offset)
Definition: diracdec.c:1429
#define BASIS_SHIFT
void ff_mpegvideoencdsp_init_x86(MpegvideoEncDSPContext *c, AVCodecContext *avctx)
static double c[64]
static int16_t basis[64][64]
void(* draw_edges)(uint8_t *buf, int wrap, int width, int height, int w, int h, int sides)
static uint8_t tmp[8]
Definition: des.c:38
static void shrink44(uint8_t *dst, int dst_wrap, const uint8_t *src, int src_wrap, int width, int height)
#define EDGE_BOTTOM
static int try_8x8basis_c(int16_t rem[64], int16_t weight[64], int16_t basis[64], int scale)
void av_image_copy_plane(uint8_t *dst, int dst_linesize, const uint8_t *src, int src_linesize, int bytewidth, int height)
Copy image plane from src to dst.
Definition: imgutils.c:287