2 * Copyright (c) 2000, 2001 Fabrice Bellard
3 * Copyright (c) 2002-2004 Michael Niedermayer <michaelni@gmx.at>
5 * This file is part of FFmpeg.
7 * FFmpeg is free software; you can redistribute it and/or
8 * modify it under the terms of the GNU Lesser General Public
9 * License as published by the Free Software Foundation; either
10 * version 2.1 of the License, or (at your option) any later version.
12 * FFmpeg is distributed in the hope that it will be useful,
13 * but WITHOUT ANY WARRANTY; without even the implied warranty of
14 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
15 * Lesser General Public License for more details.
17 * You should have received a copy of the GNU Lesser General Public
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
23 #include "libavutil/attributes.h"
24 #include "libavutil/cpu.h"
25 #include "libavutil/x86/cpu.h"
26 #include "libavcodec/avcodec.h"
27 #include "libavcodec/dsputil.h"
28 #include "libavcodec/simple_idct.h"
29 #include "dsputil_x86.h"
30 #include "idct_xvid.h"
32 /* Input permutation for the simple_idct_mmx */
33 static const uint8_t simple_mmx_permutation[64] = {
34 0x00, 0x08, 0x04, 0x09, 0x01, 0x0C, 0x05, 0x0D,
35 0x10, 0x18, 0x14, 0x19, 0x11, 0x1C, 0x15, 0x1D,
36 0x20, 0x28, 0x24, 0x29, 0x21, 0x2C, 0x25, 0x2D,
37 0x12, 0x1A, 0x16, 0x1B, 0x13, 0x1E, 0x17, 0x1F,
38 0x02, 0x0A, 0x06, 0x0B, 0x03, 0x0E, 0x07, 0x0F,
39 0x30, 0x38, 0x34, 0x39, 0x31, 0x3C, 0x35, 0x3D,
40 0x22, 0x2A, 0x26, 0x2B, 0x23, 0x2E, 0x27, 0x2F,
41 0x32, 0x3A, 0x36, 0x3B, 0x33, 0x3E, 0x37, 0x3F,
44 static const uint8_t idct_sse2_row_perm[8] = { 0, 4, 1, 5, 2, 6, 3, 7 };
46 av_cold int ff_init_scantable_permutation_x86(uint8_t *idct_permutation,
47 int idct_permutation_type)
51 switch (idct_permutation_type) {
52 case FF_SIMPLE_IDCT_PERM:
53 for (i = 0; i < 64; i++)
54 idct_permutation[i] = simple_mmx_permutation[i];
56 case FF_SSE2_IDCT_PERM:
57 for (i = 0; i < 64; i++)
58 idct_permutation[i] = (i & 0x38) | idct_sse2_row_perm[i & 7];
65 static av_cold void dsputil_init_mmx(DSPContext *c, AVCodecContext *avctx,
66 int cpu_flags, unsigned high_bit_depth)
69 c->put_pixels_clamped = ff_put_pixels_clamped_mmx;
70 c->add_pixels_clamped = ff_add_pixels_clamped_mmx;
72 if (!high_bit_depth) {
73 c->draw_edges = ff_draw_edges_mmx;
76 if (avctx->lowres == 0 && !high_bit_depth) {
77 switch (avctx->idct_algo) {
79 case FF_IDCT_SIMPLEAUTO:
80 case FF_IDCT_SIMPLEMMX:
81 c->idct_put = ff_simple_idct_put_mmx;
82 c->idct_add = ff_simple_idct_add_mmx;
83 c->idct = ff_simple_idct_mmx;
84 c->idct_permutation_type = FF_SIMPLE_IDCT_PERM;
87 c->idct_put = ff_idct_xvid_mmx_put;
88 c->idct_add = ff_idct_xvid_mmx_add;
89 c->idct = ff_idct_xvid_mmx;
94 #endif /* HAVE_MMX_INLINE */
97 c->put_signed_pixels_clamped = ff_put_signed_pixels_clamped_mmx;
98 #endif /* HAVE_MMX_EXTERNAL */
101 static av_cold void dsputil_init_mmxext(DSPContext *c, AVCodecContext *avctx,
102 int cpu_flags, unsigned high_bit_depth)
104 #if HAVE_MMXEXT_INLINE
105 if (!high_bit_depth && avctx->idct_algo == FF_IDCT_XVIDMMX && avctx->lowres == 0) {
106 c->idct_put = ff_idct_xvid_mmxext_put;
107 c->idct_add = ff_idct_xvid_mmxext_add;
108 c->idct = ff_idct_xvid_mmxext;
110 #endif /* HAVE_MMXEXT_INLINE */
113 static av_cold void dsputil_init_sse2(DSPContext *c, AVCodecContext *avctx,
114 int cpu_flags, unsigned high_bit_depth)
117 if (!high_bit_depth && avctx->idct_algo == FF_IDCT_XVIDMMX && avctx->lowres == 0) {
118 c->idct_put = ff_idct_xvid_sse2_put;
119 c->idct_add = ff_idct_xvid_sse2_add;
120 c->idct = ff_idct_xvid_sse2;
121 c->idct_permutation_type = FF_SSE2_IDCT_PERM;
123 #endif /* HAVE_SSE2_INLINE */
125 #if HAVE_SSE2_EXTERNAL
126 c->put_signed_pixels_clamped = ff_put_signed_pixels_clamped_sse2;
127 #endif /* HAVE_SSE2_EXTERNAL */
130 av_cold void ff_dsputil_init_x86(DSPContext *c, AVCodecContext *avctx,
131 unsigned high_bit_depth)
133 int cpu_flags = av_get_cpu_flags();
135 if (X86_MMX(cpu_flags))
136 dsputil_init_mmx(c, avctx, cpu_flags, high_bit_depth);
138 if (X86_MMXEXT(cpu_flags))
139 dsputil_init_mmxext(c, avctx, cpu_flags, high_bit_depth);
141 if (X86_SSE2(cpu_flags))
142 dsputil_init_sse2(c, avctx, cpu_flags, high_bit_depth);
145 ff_dsputilenc_init_mmx(c, avctx, high_bit_depth);