1/*
2 * This file is part of FFmpeg.
3 *
4 * FFmpeg is free software; you can redistribute it and/or
5 * modify it under the terms of the GNU Lesser General Public
6 * License as published by the Free Software Foundation; either
7 * version 2.1 of the License, or (at your option) any later version.
8 *
9 * FFmpeg is distributed in the hope that it will be useful,
10 * but WITHOUT ANY WARRANTY; without even the implied warranty of
11 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
12 * Lesser General Public License for more details.
13 *
14 * You should have received a copy of the GNU Lesser General Public
15 * License along with FFmpeg; if not, write to the Free Software
16 * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
17 */
18
19#include "config.h"
20
21#include <stdint.h>
22
23#include "libavutil/attributes.h"
24#include "libavutil/cpu.h"
25#include "libavutil/ppc/cpu.h"
26#include "libavutil/ppc/util_altivec.h"
27
28#include "libavcodec/mpegvideoencdsp.h"
29
30#if HAVE_ALTIVEC
31
32#if HAVE_VSX
33static int pix_norm1_altivec(uint8_t *pix, int line_size)
34{
35    int i, s = 0;
36    const vector unsigned int zero =
37        (const vector unsigned int) vec_splat_u32(0);
38    vector unsigned int sv = (vector unsigned int) vec_splat_u32(0);
39    vector signed int sum;
40
41    for (i = 0; i < 16; i++) {
42        /* Read the potentially unaligned pixels. */
43        //vector unsigned char pixl = vec_ld(0,  pix);
44        //vector unsigned char pixr = vec_ld(15, pix);
45        //vector unsigned char pixv = vec_perm(pixl, pixr, perm);
46        vector unsigned char pixv = vec_vsx_ld(0,  pix);
47
48        /* Square the values, and add them to our sum. */
49        sv = vec_msum(pixv, pixv, sv);
50
51        pix += line_size;
52    }
53    /* Sum up the four partial sums, and put the result into s. */
54    sum = vec_sums((vector signed int) sv, (vector signed int) zero);
55    sum = vec_splat(sum, 3);
56    vec_ste(sum, 0, &s);
57    return s;
58}
59#else
60static int pix_norm1_altivec(uint8_t *pix, int line_size)
61{
62    int i, s = 0;
63    const vector unsigned int zero =
64        (const vector unsigned int) vec_splat_u32(0);
65    vector unsigned char perm = vec_lvsl(0, pix);
66    vector unsigned int sv = (vector unsigned int) vec_splat_u32(0);
67    vector signed int sum;
68
69    for (i = 0; i < 16; i++) {
70        /* Read the potentially unaligned pixels. */
71        vector unsigned char pixl = vec_ld(0,  pix);
72        vector unsigned char pixr = vec_ld(15, pix);
73        vector unsigned char pixv = vec_perm(pixl, pixr, perm);
74
75        /* Square the values, and add them to our sum. */
76        sv = vec_msum(pixv, pixv, sv);
77
78        pix += line_size;
79    }
80    /* Sum up the four partial sums, and put the result into s. */
81    sum = vec_sums((vector signed int) sv, (vector signed int) zero);
82    sum = vec_splat(sum, 3);
83    vec_ste(sum, 0, &s);
84
85    return s;
86}
87#endif /* HAVE_VSX */
88
89#if HAVE_VSX
90static int pix_sum_altivec(uint8_t *pix, int line_size)
91{
92    int i, s;
93    const vector unsigned int zero =
94        (const vector unsigned int) vec_splat_u32(0);
95    vector unsigned int sad = (vector unsigned int) vec_splat_u32(0);
96    vector signed int sumdiffs;
97
98    for (i = 0; i < 16; i++) {
99        /* Read the potentially unaligned 16 pixels into t1. */
100        //vector unsigned char pixl = vec_ld(0,  pix);
101        //vector unsigned char pixr = vec_ld(15, pix);
102        //vector unsigned char t1   = vec_perm(pixl, pixr, perm);
103        vector unsigned char t1   = vec_vsx_ld(0,  pix);
104
105        /* Add each 4 pixel group together and put 4 results into sad. */
106        sad = vec_sum4s(t1, sad);
107
108        pix += line_size;
109    }
110
111    /* Sum up the four partial sums, and put the result into s. */
112    sumdiffs = vec_sums((vector signed int) sad, (vector signed int) zero);
113    sumdiffs = vec_splat(sumdiffs, 3);
114    vec_ste(sumdiffs, 0, &s);
115    return s;
116}
117#else
118static int pix_sum_altivec(uint8_t *pix, int line_size)
119{
120    int i, s;
121    const vector unsigned int zero =
122        (const vector unsigned int) vec_splat_u32(0);
123    vector unsigned char perm = vec_lvsl(0, pix);
124    vector unsigned int sad = (vector unsigned int) vec_splat_u32(0);
125    vector signed int sumdiffs;
126
127    for (i = 0; i < 16; i++) {
128        /* Read the potentially unaligned 16 pixels into t1. */
129        vector unsigned char pixl = vec_ld(0,  pix);
130        vector unsigned char pixr = vec_ld(15, pix);
131        vector unsigned char t1   = vec_perm(pixl, pixr, perm);
132
133        /* Add each 4 pixel group together and put 4 results into sad. */
134        sad = vec_sum4s(t1, sad);
135
136        pix += line_size;
137    }
138
139    /* Sum up the four partial sums, and put the result into s. */
140    sumdiffs = vec_sums((vector signed int) sad, (vector signed int) zero);
141    sumdiffs = vec_splat(sumdiffs, 3);
142    vec_ste(sumdiffs, 0, &s);
143
144    return s;
145}
146
147#endif /* HAVE_VSX */
148
149#endif /* HAVE_ALTIVEC */
150
151av_cold void ff_mpegvideoencdsp_init_ppc(MpegvideoEncDSPContext *c,
152                                         AVCodecContext *avctx)
153{
154#if HAVE_ALTIVEC
155    if (!PPC_ALTIVEC(av_get_cpu_flags()))
156        return;
157
158    c->pix_norm1 = pix_norm1_altivec;
159    c->pix_sum   = pix_sum_altivec;
160#endif /* HAVE_ALTIVEC */
161}
162