summaryrefslogtreecommitdiff
path: root/libavcodec/alpha/mpegvideo_alpha.c
blob: 3c32af157de4a20afe1ec7252d54739fcefec59c (plain)
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
/*
 * Alpha optimized DSP utils
 * Copyright (c) 2002 Falk Hueffner <falk@debian.org>
 *
 * This file is part of FFmpeg.
 *
 * FFmpeg is free software; you can redistribute it and/or
 * modify it under the terms of the GNU Lesser General Public
 * License as published by the Free Software Foundation; either
 * version 2.1 of the License, or (at your option) any later version.
 *
 * FFmpeg is distributed in the hope that it will be useful,
 * but WITHOUT ANY WARRANTY; without even the implied warranty of
 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
 * Lesser General Public License for more details.
 *
 * You should have received a copy of the GNU Lesser General Public
 * License along with FFmpeg; if not, write to the Free Software
 * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
 */

#include "libavcodec/dsputil.h"
#include "libavcodec/mpegvideo.h"
#include "asm.h"

static void dct_unquantize_h263_intra_axp(MpegEncContext *s, DCTELEM *block,
                                    int n, int qscale)
{
    int i, n_coeffs;
    uint64_t qmul, qadd;
    uint64_t correction;
    DCTELEM *orig_block = block;
    DCTELEM block0;             /* might not be used uninitialized */

    qadd = WORD_VEC((qscale - 1) | 1);
    qmul = qscale << 1;
    /* This mask kills spill from negative subwords to the next subword.  */
    correction = WORD_VEC((qmul - 1) + 1); /* multiplication / addition */

    if (!s->h263_aic) {
        if (n < 4)
            block0 = block[0] * s->y_dc_scale;
        else
            block0 = block[0] * s->c_dc_scale;
    } else {
        qadd = 0;
    }
    n_coeffs = 63; // does not always use zigzag table

    for(i = 0; i <= n_coeffs; block += 4, i += 4) {
        uint64_t levels, negmask, zeros, add;

        levels = ldq(block);
        if (levels == 0)
            continue;

#ifdef __alpha_max__
        /* I don't think the speed difference justifies runtime
           detection.  */
        negmask = maxsw4(levels, -1); /* negative -> ffff (-1) */
        negmask = minsw4(negmask, 0); /* positive -> 0000 (0) */
#else
        negmask = cmpbge(WORD_VEC(0x7fff), levels);
        negmask &= (negmask >> 1) | (1 << 7);
        negmask = zap(-1, negmask);
#endif

        zeros = cmpbge(0, levels);
        zeros &= zeros >> 1;
        /* zeros |= zeros << 1 is not needed since qadd <= 255, so
           zapping the lower byte suffices.  */

        levels *= qmul;
        levels -= correction & (negmask << 16);

        /* Negate qadd for negative levels.  */
        add = qadd ^ negmask;
        add += WORD_VEC(0x0001) & negmask;
        /* Set qadd to 0 for levels == 0.  */
        add = zap(add, zeros);

        levels += add;

        stq(levels, block);
    }

    if (s->mb_intra && !s->h263_aic)
        orig_block[0] = block0;
}

static void dct_unquantize_h263_inter_axp(MpegEncContext *s, DCTELEM *block,
                                    int n, int qscale)
{
    int i, n_coeffs;
    uint64_t qmul, qadd;
    uint64_t correction;

    qadd = WORD_VEC((qscale - 1) | 1);
    qmul = qscale << 1;
    /* This mask kills spill from negative subwords to the next subword.  */
    correction = WORD_VEC((qmul - 1) + 1); /* multiplication / addition */

    n_coeffs = s->intra_scantable.raster_end[s->block_last_index[n]];

    for(i = 0; i <= n_coeffs; block += 4, i += 4) {
        uint64_t levels, negmask, zeros, add;

        levels = ldq(block);
        if (levels == 0)
            continue;

#ifdef __alpha_max__
        /* I don't think the speed difference justifies runtime
           detection.  */
        negmask = maxsw4(levels, -1); /* negative -> ffff (-1) */
        negmask = minsw4(negmask, 0); /* positive -> 0000 (0) */
#else
        negmask = cmpbge(WORD_VEC(0x7fff), levels);
        negmask &= (negmask >> 1) | (1 << 7);
        negmask = zap(-1, negmask);
#endif

        zeros = cmpbge(0, levels);
        zeros &= zeros >> 1;
        /* zeros |= zeros << 1 is not needed since qadd <= 255, so
           zapping the lower byte suffices.  */

        levels *= qmul;
        levels -= correction & (negmask << 16);

        /* Negate qadd for negative levels.  */
        add = qadd ^ negmask;
        add += WORD_VEC(0x0001) & negmask;
        /* Set qadd to 0 for levels == 0.  */
        add = zap(add, zeros);

        levels += add;

        stq(levels, block);
    }
}

void MPV_common_init_axp(MpegEncContext *s)
{
    s->dct_unquantize_h263_intra = dct_unquantize_h263_intra_axp;
    s->dct_unquantize_h263_inter = dct_unquantize_h263_inter_axp;
}