2 * BlackFin MPEGVIDEO OPTIMIZATIONS
4 * Copyright (C) 2007 Marc Hoffman <mmh@pleasantst.com>
6 * This file is part of FFmpeg.
8 * FFmpeg 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.
13 * FFmpeg 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.
18 * You should have received a copy of the GNU Lesser General Public
19 * License along with FFmpeg; if not, write to the Free Software
20 * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
23 #include "libavcodec/avcodec.h"
24 #include "libavcodec/dsputil.h"
25 #include "libavcodec/mpegvideo.h"
26 #include "dsputil_bfin.h"
29 extern void ff_bfin_fdct (DCTELEM
*block
) attribute_l1_text
;
32 static int dct_quantize_bfin (MpegEncContext
*s
,
33 DCTELEM
*block
, int n
,
34 int qscale
, int *overflow
)
36 int last_non_zero
, q
, start_i
;
39 const uint8_t *scantable
= s
->intra_scantable
.scantable
;
49 s
->denoise_dct(s
, block
);
61 /* For AIC we skip quant/dequant of INTRADC */
64 /* note: block[0] is assumed to be positive */
65 dc
= block
[0] = (block
[0] + (q
>> 1)) / q
;
68 bias
= s
->q_intra_matrix16
[qscale
][1];
69 qmat
= s
->q_intra_matrix16
[qscale
][0];
74 bias
= s
->q_inter_matrix16
[qscale
][1];
75 qmat
= s
->q_inter_matrix16
[qscale
][0];
82 /* for(i=start_i; i<64; i++) { */
83 /* sign = (block[i]>>15)|1; */
84 /* level = ((abs(block[i])+bias[0])*qmat[i])>>16; */
85 /* if (level < 0) level = 0; */
87 /* level = level * sign; */
88 /* block[i] = level; */
94 "r0=r1>>>15 (v); \n\t"
95 "lsetup (0f,1f) lc0=%3; \n\t"
97 " r1=abs r1 (v) || r2=[%2++];\n\t"
99 " r1=max(r1,%6) (v); \n\t"
100 " r1.h=(a1 =r1.h*r2.h), r1.l=(a0 =r1.l*r2.l) (tfu); \n\t"
102 " r0.h=(a1 =r1.h*r0.h), r0.l=(a0 =r1.l*r0.l) (is) || r1=[%1++];\n\t"
103 "1: r0=r1>>>15 (v) || [i2++]=r0;\n\t"
108 : "b" (block
), "b" (qmat
), "a" (32), "d" (0x00010001), "d" (bias
[0]*0x10001), "d" (0)
109 : "R0","R1","R2", "I2");
110 if (start_i
== 1) block
[0] = dc
;
118 ("r0=b[%1--] (x); \n\t"
119 "lsetup (0f,1f) lc0=%3; \n\t" /* for(i=63; i>=start_i; i--) { */
120 "0: p0=r0; \n\t" /* j = scantable[i]; */
121 " p0=%2+(p0<<1); \n\t" /* if (block[j]) { */
122 " r0=w[p0]; \n\t" /* last_non_zero = i; */
123 " cc=r0==0; \n\t" /* break; */
124 " if !cc jump 2f; \n\t" /* } */
125 "1: r0=b[%1--] (x); \n\t" /* } */
131 : "=d" (last_non_zero
)
132 : "a" (scantable
+63), "a" (block
), "a" (63), "d" (last_non_zero
)
137 *overflow
= s
->max_qcoeff
< max
; //overflow might have happened
141 /* we need this permutation so that we correct the IDCT, we only permute the !=0 elements */
142 if (s
->dsp
.idct_permutation_type
!= FF_NO_IDCT_PERM
)
143 ff_block_permute(block
, s
->dsp
.idct_permutation
, scantable
, last_non_zero
);
145 return last_non_zero
;
148 void MPV_common_init_bfin (MpegEncContext
*s
)
150 s
->dct_quantize
= dct_quantize_bfin
;