2
2
* Alpha optimized DSP utils
3
3
* Copyright (c) 2002 Falk Hueffner <falk@debian.org>
5
* This library is free software; you can redistribute it and/or
5
* This file is part of FFmpeg.
7
* FFmpeg is free software; you can redistribute it and/or
6
8
* modify it under the terms of the GNU Lesser General Public
7
9
* License as published by the Free Software Foundation; either
8
* version 2 of the License, or (at your option) any later version.
10
* version 2.1 of the License, or (at your option) any later version.
10
* This library is distributed in the hope that it will be useful,
12
* FFmpeg is distributed in the hope that it will be useful,
11
13
* but WITHOUT ANY WARRANTY; without even the implied warranty of
12
14
* MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
13
15
* Lesser General Public License for more details.
15
17
* You should have received a copy of the GNU Lesser General Public
16
* License along with this library; if not, write to the Free Software
17
* Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA
18
* License along with FFmpeg; if not, write to the Free Software
19
* Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
21
23
#include "../dsputil.h"
22
24
#include "../mpegvideo.h"
24
static void dct_unquantize_h263_axp(MpegEncContext *s, DCTELEM *block,
26
static void dct_unquantize_h263_intra_axp(MpegEncContext *s, DCTELEM *block,
28
30
uint64_t qmul, qadd;
29
31
uint64_t correction;
30
32
DCTELEM *orig_block = block;
33
DCTELEM block0; /* might not be used uninitialized */
33
35
qadd = WORD_VEC((qscale - 1) | 1);
34
36
qmul = qscale << 1;
35
/* This mask kills spill from negative subwords to the next subword. */
37
/* This mask kills spill from negative subwords to the next subword. */
36
38
correction = WORD_VEC((qmul - 1) + 1); /* multiplication / addition */
41
block0 = block[0] * s->y_dc_scale;
43
block0 = block[0] * s->c_dc_scale;
47
n_coeffs = 63; // does not always use zigzag table
42
block0 = block[0] * s->y_dc_scale;
44
block0 = block[0] * s->c_dc_scale;
49
n_coeffs = s->intra_scantable.raster_end[s->block_last_index[n]];
48
n_coeffs = 63; // does not always use zigzag table
52
50
for(i = 0; i <= n_coeffs; block += 4, i += 4) {
53
51
uint64_t levels, negmask, zeros, add;
90
88
orig_block[0] = block0;
91
static void dct_unquantize_h263_inter_axp(MpegEncContext *s, DCTELEM *block,
98
qadd = WORD_VEC((qscale - 1) | 1);
100
/* This mask kills spill from negative subwords to the next subword. */
101
correction = WORD_VEC((qmul - 1) + 1); /* multiplication / addition */
103
n_coeffs = s->intra_scantable.raster_end[s->block_last_index[n]];
105
for(i = 0; i <= n_coeffs; block += 4, i += 4) {
106
uint64_t levels, negmask, zeros, add;
113
/* I don't think the speed difference justifies runtime
115
negmask = maxsw4(levels, -1); /* negative -> ffff (-1) */
116
negmask = minsw4(negmask, 0); /* positive -> 0000 (0) */
118
negmask = cmpbge(WORD_VEC(0x7fff), levels);
119
negmask &= (negmask >> 1) | (1 << 7);
120
negmask = zap(-1, negmask);
123
zeros = cmpbge(0, levels);
125
/* zeros |= zeros << 1 is not needed since qadd <= 255, so
126
zapping the lower byte suffices. */
129
levels -= correction & (negmask << 16);
131
/* Negate qadd for negative levels. */
132
add = qadd ^ negmask;
133
add += WORD_VEC(0x0001) & negmask;
134
/* Set qadd to 0 for levels == 0. */
135
add = zap(add, zeros);
93
143
void MPV_common_init_axp(MpegEncContext *s)
95
s->dct_unquantize_h263 = dct_unquantize_h263_axp;
145
s->dct_unquantize_h263_intra = dct_unquantize_h263_intra_axp;
146
s->dct_unquantize_h263_inter = dct_unquantize_h263_inter_axp;