aboutsummaryrefslogtreecommitdiffstats
path: root/libavcodec/ppc/mpegvideo_altivec.c
blob: a09932bdf83a465993acb91aac68a6e997173a64 (plain) (blame)
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
/*
 * Copyright (c) 2002 Dieter Shirley
 *
 * dct_unquantize_h263_altivec:
 * Copyright (c) 2003 Romain Dolbeau <romain@dolbeau.org>
 *
 * This file is part of Libav.
 *
 * Libav 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.
 *
 * Libav 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 Libav; if not, write to the Free Software
 * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
 */

#include <stdlib.h>
#include <stdio.h>

#include "config.h"
#include "libavutil/attributes.h"
#include "libavutil/cpu.h"
#include "libavutil/ppc/cpu.h"
#include "libavutil/ppc/types_altivec.h"
#include "libavutil/ppc/util_altivec.h"
#include "libavcodec/mpegvideo.h"

#if HAVE_ALTIVEC && HAVE_BIGENDIAN

/* AltiVec version of dct_unquantize_h263
   this code assumes `block' is 16 bytes-aligned */
static void dct_unquantize_h263_altivec(MpegEncContext *s,
                                 int16_t *block, int n, int qscale)
{
    int i, level, qmul, qadd;
    int nCoeffs;

    assert(s->block_last_index[n]>=0);

    qadd = (qscale - 1) | 1;
    qmul = qscale << 1;

    if (s->mb_intra) {
        if (!s->h263_aic) {
            if (n < 4)
                block[0] = block[0] * s->y_dc_scale;
            else
                block[0] = block[0] * s->c_dc_scale;
        }else
            qadd = 0;
        i = 1;
        nCoeffs= 63; //does not always use zigzag table
    } else {
        i = 0;
        nCoeffs= s->intra_scantable.raster_end[ s->block_last_index[n] ];
    }

    {
        register const vector signed short vczero = (const vector signed short)vec_splat_s16(0);
        DECLARE_ALIGNED(16, short, qmul8) = qmul;
        DECLARE_ALIGNED(16, short, qadd8) = qadd;
        register vector signed short blockv, qmulv, qaddv, nqaddv, temp1;
        register vector bool short blockv_null, blockv_neg;
        register short backup_0 = block[0];
        register int j = 0;

        qmulv = vec_splat((vec_s16)vec_lde(0, &qmul8), 0);
        qaddv = vec_splat((vec_s16)vec_lde(0, &qadd8), 0);
        nqaddv = vec_sub(vczero, qaddv);

        // vectorize all the 16 bytes-aligned blocks
        // of 8 elements
        for(; (j + 7) <= nCoeffs ; j+=8) {
            blockv = vec_ld(j << 1, block);
            blockv_neg = vec_cmplt(blockv, vczero);
            blockv_null = vec_cmpeq(blockv, vczero);
            // choose between +qadd or -qadd as the third operand
            temp1 = vec_sel(qaddv, nqaddv, blockv_neg);
            // multiply & add (block{i,i+7} * qmul [+-] qadd)
            temp1 = vec_mladd(blockv, qmulv, temp1);
            // put 0 where block[{i,i+7} used to have 0
            blockv = vec_sel(temp1, blockv, blockv_null);
            vec_st(blockv, j << 1, block);
        }

        // if nCoeffs isn't a multiple of 8, finish the job
        // using good old scalar units.
        // (we could do it using a truncated vector,
        // but I'm not sure it's worth the hassle)
        for(; j <= nCoeffs ; j++) {
            level = block[j];
            if (level) {
                if (level < 0) {
                    level = level * qmul - qadd;
                } else {
                    level = level * qmul + qadd;
                }
                block[j] = level;
            }
        }

        if (i == 1) {
            // cheat. this avoid special-casing the first iteration
            block[0] = backup_0;
        }
    }
}

#endif /* HAVE_ALTIVEC && HAVE_BIGENDIAN */

av_cold void ff_mpv_common_init_ppc(MpegEncContext *s)
{
#if HAVE_ALTIVEC && HAVE_BIGENDIAN
    if (!PPC_ALTIVEC(av_get_cpu_flags()))
        return;

    if ((s->avctx->dct_algo == FF_DCT_AUTO) ||
        (s->avctx->dct_algo == FF_DCT_ALTIVEC)) {
        s->dct_unquantize_h263_intra = dct_unquantize_h263_altivec;
        s->dct_unquantize_h263_inter = dct_unquantize_h263_altivec;
    }
#endif /* HAVE_ALTIVEC && HAVE_BIGENDIAN */
}