mpegvideo_bfin.c
Go to the documentation of this file.
1 /*
2  * BlackFin MPEGVIDEO OPTIMIZATIONS
3  *
4  * Copyright (C) 2007 Marc Hoffman <mmh@pleasantst.com>
5  *
6  * This file is part of Libav.
7  *
8  * Libav is free software; you can redistribute it and/or
9  * modify it under the terms of the GNU Lesser General Public
10  * License as published by the Free Software Foundation; either
11  * version 2.1 of the License, or (at your option) any later version.
12  *
13  * Libav is distributed in the hope that it will be useful,
14  * but WITHOUT ANY WARRANTY; without even the implied warranty of
15  * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
16  * Lesser General Public License for more details.
17  *
18  * You should have received a copy of the GNU Lesser General Public
19  * License along with Libav; if not, write to the Free Software
20  * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
21  */
22 
23 #include "libavcodec/avcodec.h"
24 #include "libavcodec/dsputil.h"
25 #include "libavcodec/mpegvideo.h"
26 #include "dsputil_bfin.h"
27 
29  DCTELEM *block, int n,
30  int qscale, int *overflow)
31 {
32  int last_non_zero, q, start_i;
33  const short *qmat;
34  short *bias;
35  const uint8_t *scantable= s->intra_scantable.scantable;
36  short dc;
37  int max=0;
38 
39  PROF("fdct",0);
40  s->dsp.fdct(block);
41  EPROF();
42 
43  PROF("denoise",1);
44  if(s->dct_error_sum)
45  s->denoise_dct(s, block);
46  EPROF();
47 
48  PROF("quant-init",2);
49  if (s->mb_intra) {
50  if (!s->h263_aic) {
51  if (n < 4)
52  q = s->y_dc_scale;
53  else
54  q = s->c_dc_scale;
55  q = q << 3;
56  } else
57  /* For AIC we skip quant/dequant of INTRADC */
58  q = 1 << 3;
59 
60  /* note: block[0] is assumed to be positive */
61  dc = block[0] = (block[0] + (q >> 1)) / q;
62  start_i = 1;
63  last_non_zero = 0;
64  bias = s->q_intra_matrix16[qscale][1];
65  qmat = s->q_intra_matrix16[qscale][0];
66 
67  } else {
68  start_i = 0;
69  last_non_zero = -1;
70  bias = s->q_inter_matrix16[qscale][1];
71  qmat = s->q_inter_matrix16[qscale][0];
72 
73  }
74  EPROF();
75 
76  PROF("quantize",4);
77 
78  /* for(i=start_i; i<64; i++) { */
79  /* sign = (block[i]>>15)|1; */
80  /* level = ((abs(block[i])+bias[0])*qmat[i])>>16; */
81  /* if (level < 0) level = 0; */
82  /* max |= level; */
83  /* level = level * sign; */
84  /* block[i] = level; */
85  /* } */
86 
87  __asm__ volatile
88  ("i2=%1;\n\t"
89  "r1=[%1++]; \n\t"
90  "r0=r1>>>15 (v); \n\t"
91  "lsetup (0f,1f) lc0=%3; \n\t"
92  "0: r0=r0|%4; \n\t"
93  " r1=abs r1 (v) || r2=[%2++];\n\t"
94  " r1=r1+|+%5; \n\t"
95  " r1=max(r1,%6) (v); \n\t"
96  " r1.h=(a1 =r1.h*r2.h), r1.l=(a0 =r1.l*r2.l) (tfu); \n\t"
97  " %0=%0|r1; \n\t"
98  " r0.h=(a1 =r1.h*r0.h), r0.l=(a0 =r1.l*r0.l) (is) || r1=[%1++];\n\t"
99  "1: r0=r1>>>15 (v) || [i2++]=r0;\n\t"
100  "r1=%0>>16; \n\t"
101  "%0=%0|r1; \n\t"
102  "%0.h=0; \n\t"
103  : "=&d" (max)
104  : "b" (block), "b" (qmat), "a" (32), "d" (0x00010001), "d" (bias[0]*0x10001), "d" (0)
105  : "R0","R1","R2", "I2");
106  if (start_i == 1) block[0] = dc;
107 
108  EPROF();
109 
110 
111  PROF("zzscan",5);
112 
113  __asm__ volatile
114  ("r0=b[%1--] (x); \n\t"
115  "lsetup (0f,1f) lc0=%3; \n\t" /* for(i=63; i>=start_i; i--) { */
116  "0: p0=r0; \n\t" /* j = scantable[i]; */
117  " p0=%2+(p0<<1); \n\t" /* if (block[j]) { */
118  " r0=w[p0]; \n\t" /* last_non_zero = i; */
119  " cc=r0==0; \n\t" /* break; */
120  " if !cc jump 2f; \n\t" /* } */
121  "1: r0=b[%1--] (x); \n\t" /* } */
122  " %0=%4; \n\t"
123  " jump 3f; \n\t"
124  "2: %0=lc0; \n\t"
125  "3:\n\t"
126 
127  : "=d" (last_non_zero)
128  : "a" (scantable+63), "a" (block), "a" (63), "d" (last_non_zero)
129  : "P0","R0");
130 
131  EPROF();
132 
133  *overflow= s->max_qcoeff < max; //overflow might have happened
134 
135  bfprof();
136 
137  /* we need this permutation so that we correct the IDCT, we only permute the !=0 elements */
139  ff_block_permute(block, s->dsp.idct_permutation, scantable, last_non_zero);
140 
141  return last_non_zero;
142 }
143 
145 {
146 /* s->dct_quantize= dct_quantize_bfin; */
147 }
148