1/**
2 * @file
3 * MMX-optimized functions for the VP6 decoder
4 *
5 * Copyright (C) 2009  Sebastien Lucas <sebastien.lucas@gmail.com>
6 *
7 * This file is part of FFmpeg.
8 *
9 * FFmpeg is free software; you can redistribute it and/or
10 * modify it under the terms of the GNU Lesser General Public
11 * License as published by the Free Software Foundation; either
12 * version 2.1 of the License, or (at your option) any later version.
13 *
14 * FFmpeg is distributed in the hope that it will be useful,
15 * but WITHOUT ANY WARRANTY; without even the implied warranty of
16 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
17 * Lesser General Public License for more details.
18 *
19 * You should have received a copy of the GNU Lesser General Public
20 * License along with FFmpeg; if not, write to the Free Software
21 * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
22 */
23
24#include "libavutil/x86_cpu.h"
25#include "libavcodec/dsputil.h"
26#include "dsputil_mmx.h"
27#include "vp6dsp_mmx.h"
28
29
30#define DIAG4_MMX(in1,in2,in3,in4)                                            \
31    "movq  "#in1"(%0), %%mm0             \n\t"                                \
32    "movq  "#in2"(%0), %%mm1             \n\t"                                \
33    "movq       %%mm0, %%mm3             \n\t"                                \
34    "movq       %%mm1, %%mm4             \n\t"                                \
35    "punpcklbw  %%mm7, %%mm0             \n\t"                                \
36    "punpcklbw  %%mm7, %%mm1             \n\t"                                \
37    "punpckhbw  %%mm7, %%mm3             \n\t"                                \
38    "punpckhbw  %%mm7, %%mm4             \n\t"                                \
39    "pmullw     0(%2), %%mm0             \n\t" /* src[x-8 ] * biweight [0] */ \
40    "pmullw     8(%2), %%mm1             \n\t" /* src[x   ] * biweight [1] */ \
41    "pmullw     0(%2), %%mm3             \n\t" /* src[x-8 ] * biweight [0] */ \
42    "pmullw     8(%2), %%mm4             \n\t" /* src[x   ] * biweight [1] */ \
43    "paddw      %%mm1, %%mm0             \n\t"                                \
44    "paddw      %%mm4, %%mm3             \n\t"                                \
45    "movq  "#in3"(%0), %%mm1             \n\t"                                \
46    "movq  "#in4"(%0), %%mm2             \n\t"                                \
47    "movq       %%mm1, %%mm4             \n\t"                                \
48    "movq       %%mm2, %%mm5             \n\t"                                \
49    "punpcklbw  %%mm7, %%mm1             \n\t"                                \
50    "punpcklbw  %%mm7, %%mm2             \n\t"                                \
51    "punpckhbw  %%mm7, %%mm4             \n\t"                                \
52    "punpckhbw  %%mm7, %%mm5             \n\t"                                \
53    "pmullw    16(%2), %%mm1             \n\t" /* src[x+8 ] * biweight [2] */ \
54    "pmullw    24(%2), %%mm2             \n\t" /* src[x+16] * biweight [3] */ \
55    "pmullw    16(%2), %%mm4             \n\t" /* src[x+8 ] * biweight [2] */ \
56    "pmullw    24(%2), %%mm5             \n\t" /* src[x+16] * biweight [3] */ \
57    "paddw      %%mm2, %%mm1             \n\t"                                \
58    "paddw      %%mm5, %%mm4             \n\t"                                \
59    "paddsw     %%mm1, %%mm0             \n\t"                                \
60    "paddsw     %%mm4, %%mm3             \n\t"                                \
61    "paddsw     %%mm6, %%mm0             \n\t" /* Add 64 */                   \
62    "paddsw     %%mm6, %%mm3             \n\t" /* Add 64 */                   \
63    "psraw         $7, %%mm0             \n\t"                                \
64    "psraw         $7, %%mm3             \n\t"                                \
65    "packuswb   %%mm3, %%mm0             \n\t"                                \
66    "movq       %%mm0,  (%1)             \n\t"
67
68void ff_vp6_filter_diag4_mmx(uint8_t *dst, uint8_t *src, int stride,
69                             const int16_t *h_weights, const int16_t *v_weights)
70{
71    uint8_t tmp[8*11], *t = tmp;
72    int16_t weights[4*4];
73    int i;
74    src -= stride;
75
76    for (i=0; i<4*4; i++)
77        weights[i] = h_weights[i>>2];
78
79    __asm__ volatile(
80    "pxor %%mm7, %%mm7                   \n\t"
81    "movq "MANGLE(ff_pw_64)", %%mm6      \n\t"
82    "1:                                  \n\t"
83    DIAG4_MMX(-1,0,1,2)
84    "add  $8, %1                         \n\t"
85    "add  %3, %0                         \n\t"
86    "decl %4                             \n\t"
87    "jnz 1b                              \n\t"
88    : "+r"(src), "+r"(t)
89    : "r"(weights), "r"((x86_reg)stride), "r"(11)
90    : "memory");
91
92    t = tmp + 8;
93    for (i=0; i<4*4; i++)
94        weights[i] = v_weights[i>>2];
95
96    __asm__ volatile(
97    "pxor %%mm7, %%mm7                   \n\t"
98    "movq "MANGLE(ff_pw_64)", %%mm6      \n\t"
99    "1:                                  \n\t"
100    DIAG4_MMX(-8,0,8,16)
101    "add  $8, %0                         \n\t"
102    "add  %3, %1                         \n\t"
103    "decl %4                             \n\t"
104    "jnz 1b                              \n\t"
105    : "+r"(t), "+r"(dst)
106    : "r"(weights), "r"((x86_reg)stride), "r"(8)
107    : "memory");
108}
109