Add Alpha assembly for pix_abs16x16. Optimized for pca56, no large win
[libav.git] / libavcodec / dsputil.h
CommitLineData
ff4ec49e
FB
1/*
2 * DSP utils
3 * Copyright (c) 2000, 2001, 2002 Fabrice Bellard.
4 *
5 * This library is free software; you can redistribute it and/or
6 * modify it under the terms of the GNU Lesser General Public
7 * License as published by the Free Software Foundation; either
8 * version 2 of the License, or (at your option) any later version.
9 *
10 * This library is distributed in the hope that it will be useful,
11 * but WITHOUT ANY WARRANTY; without even the implied warranty of
12 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
13 * Lesser General Public License for more details.
14 *
15 * You should have received a copy of the GNU Lesser General Public
16 * License along with this library; if not, write to the Free Software
17 * Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA
18 */
de6d9b64
FB
19#ifndef DSPUTIL_H
20#define DSPUTIL_H
21
22#include "common.h"
43f1708f 23#include "avcodec.h"
de6d9b64 24
44eb4951 25//#define DEBUG
de6d9b64
FB
26/* dct code */
27typedef short DCTELEM;
28
03c94ede 29void fdct_ifast (DCTELEM *data);
28db7fce 30void ff_jpeg_fdct_islow (DCTELEM *data);
de6d9b64
FB
31
32void j_rev_dct (DCTELEM *data);
33
3f09f52a 34void ff_fdct_mmx(DCTELEM *block);
de6d9b64 35
e0eac44e
FB
36/* encoding scans */
37extern UINT8 ff_alternate_horizontal_scan[64];
38extern UINT8 ff_alternate_vertical_scan[64];
39extern UINT8 zigzag_direct[64];
40
5a240838
MN
41/* permutation table */
42extern UINT8 permutation[64];
43
de6d9b64
FB
44/* pixel operations */
45#define MAX_NEG_CROP 384
46
47/* temporary */
48extern UINT32 squareTbl[512];
0cfa9713 49extern UINT8 cropTbl[256 + 2 * MAX_NEG_CROP];
de6d9b64
FB
50
51void dsputil_init(void);
52
b7c27ee6
MN
53/* minimum alignment rules ;)
54if u notice errors in the align stuff, need more alignment for some asm code for some cpu
55or need to use a function with less aligned data then send a mail to the ffmpeg-dev list, ...
56
57!warning these alignments might not match reallity, (missing attribute((align)) stuff somewhere possible)
58i (michael) didnt check them, these are just the alignents which i think could be reached easily ...
de6d9b64 59
b7c27ee6
MN
60!future video codecs might need functions with less strict alignment
61*/
62
63/* pixel ops : interface with DCT */
64extern void (*ff_idct)(DCTELEM *block/*align 16*/);
65extern void (*ff_idct_put)(UINT8 *dest/*align 8*/, int line_size, DCTELEM *block/*align 16*/);
66extern void (*ff_idct_add)(UINT8 *dest/*align 8*/, int line_size, DCTELEM *block/*align 16*/);
67extern void (*get_pixels)(DCTELEM *block/*align 16*/, const UINT8 *pixels/*align 8*/, int line_size);
68extern void (*diff_pixels)(DCTELEM *block/*align 16*/, const UINT8 *s1/*align 8*/, const UINT8 *s2/*align 8*/, int stride);
69extern void (*put_pixels_clamped)(const DCTELEM *block/*align 16*/, UINT8 *pixels/*align 8*/, int line_size);
70extern void (*add_pixels_clamped)(const DCTELEM *block/*align 16*/, UINT8 *pixels/*align 8*/, int line_size);
71extern void (*gmc1)(UINT8 *dst/*align 8*/, UINT8 *src/*align 1*/, int srcStride, int h, int x16, int y16, int rounder);
72extern void (*clear_blocks)(DCTELEM *blocks/*align 16*/);
3aa102be
MN
73extern int (*pix_sum)(UINT8 * pix, int line_size);
74extern int (*pix_norm1)(UINT8 * pix, int line_size);
75
44eb4951 76
de6d9b64
FB
77
78void get_pixels_c(DCTELEM *block, const UINT8 *pixels, int line_size);
9dbcbd92 79void diff_pixels_c(DCTELEM *block, const UINT8 *s1, const UINT8 *s2, int stride);
de6d9b64
FB
80void put_pixels_clamped_c(const DCTELEM *block, UINT8 *pixels, int line_size);
81void add_pixels_clamped_c(const DCTELEM *block, UINT8 *pixels, int line_size);
649c00c9 82void clear_blocks_c(DCTELEM *blocks);
de6d9b64
FB
83
84/* add and put pixel (decoding) */
b7c27ee6
MN
85// blocksizes for op_pixels_func are 8x4,8x8 16x8 16x16
86typedef void (*op_pixels_func)(UINT8 *block/*align width (8 or 16)*/, const UINT8 *pixels/*align 1*/, int line_size, int h);
87typedef void (*qpel_mc_func)(UINT8 *dst/*align width (8 or 16)*/, UINT8 *src/*align 1*/, int stride);
b3184779
MN
88
89extern op_pixels_func put_pixels_tab[2][4];
90extern op_pixels_func avg_pixels_tab[2][4];
91extern op_pixels_func put_no_rnd_pixels_tab[2][4];
92extern op_pixels_func avg_no_rnd_pixels_tab[2][4];
93extern qpel_mc_func put_qpel_pixels_tab[2][16];
94extern qpel_mc_func avg_qpel_pixels_tab[2][16];
95extern qpel_mc_func put_no_rnd_qpel_pixels_tab[2][16];
96extern qpel_mc_func avg_no_rnd_qpel_pixels_tab[2][16];
97
98#define CALL_2X_PIXELS(a, b, n)\
99static void a(uint8_t *block, const uint8_t *pixels, int line_size, int h){\
100 b(block , pixels , line_size, h);\
101 b(block+n, pixels+n, line_size, h);\
102}
44eb4951 103
de6d9b64
FB
104/* motion estimation */
105
b7c27ee6 106typedef int (*op_pixels_abs_func)(UINT8 *blk1/*align width (8 or 16)*/, UINT8 *blk2/*align 1*/, int line_size);
de6d9b64
FB
107
108extern op_pixels_abs_func pix_abs16x16;
109extern op_pixels_abs_func pix_abs16x16_x2;
110extern op_pixels_abs_func pix_abs16x16_y2;
111extern op_pixels_abs_func pix_abs16x16_xy2;
ba6802de
MN
112extern op_pixels_abs_func pix_abs8x8;
113extern op_pixels_abs_func pix_abs8x8_x2;
114extern op_pixels_abs_func pix_abs8x8_y2;
115extern op_pixels_abs_func pix_abs8x8_xy2;
116
117int pix_abs16x16_c(UINT8 *blk1, UINT8 *blk2, int lx);
118int pix_abs16x16_x2_c(UINT8 *blk1, UINT8 *blk2, int lx);
119int pix_abs16x16_y2_c(UINT8 *blk1, UINT8 *blk2, int lx);
120int pix_abs16x16_xy2_c(UINT8 *blk1, UINT8 *blk2, int lx);
de6d9b64 121
e0eac44e
FB
122static inline int block_permute_op(int j)
123{
5a240838 124 return permutation[j];
e0eac44e
FB
125}
126
127void block_permute(INT16 *block);
128
3d03c0a2 129#if defined(HAVE_MMX)
de6d9b64
FB
130
131#define MM_MMX 0x0001 /* standard MMX */
132#define MM_3DNOW 0x0004 /* AMD 3DNOW */
133#define MM_MMXEXT 0x0002 /* SSE integer functions or AMD MMX ext */
134#define MM_SSE 0x0008 /* SSE functions */
135#define MM_SSE2 0x0010 /* PIV SSE2 functions */
136
137extern int mm_flags;
138
139int mm_support(void);
140
141static inline void emms(void)
142{
fb16b7e7
FB
143 __asm __volatile ("emms;":::"memory");
144}
145
146#define emms_c() \
147{\
148 if (mm_flags & MM_MMX)\
149 emms();\
de6d9b64
FB
150}
151
152#define __align8 __attribute__ ((aligned (8)))
153
154void dsputil_init_mmx(void);
57060b1e 155void dsputil_set_bit_exact_mmx(void);
de6d9b64 156
3d03c0a2
FB
157#elif defined(ARCH_ARMV4L)
158
159#define emms_c()
160
161/* This is to use 4 bytes read to the IDCT pointers for some 'zero'
162 line ptimizations */
163#define __align8 __attribute__ ((aligned (4)))
164
165void dsputil_init_armv4l(void);
166
c34270f5
FB
167#elif defined(HAVE_MLIB)
168
169#define emms_c()
170
171/* SPARC/VIS IDCT needs 8-byte aligned DCT blocks */
172#define __align8 __attribute__ ((aligned (8)))
173
174void dsputil_init_mlib(void);
175
1e98dffb
NK
176#elif defined(ARCH_ALPHA)
177
178#define emms_c()
179#define __align8 __attribute__ ((aligned (8)))
180
181void dsputil_init_alpha(void);
182
59925ef2
BF
183#elif defined(ARCH_POWERPC)
184
185#define emms_c()
186#define __align8 __attribute__ ((aligned (16)))
187
ab6c65f6 188void dsputil_init_ppc(void);
59925ef2 189
d46aba26
LS
190#elif defined(HAVE_MMI)
191
192#define emms_c()
193
194#define __align8 __attribute__ ((aligned (16)))
195
196void dsputil_init_mmi(void);
197
de6d9b64
FB
198#else
199
fb16b7e7
FB
200#define emms_c()
201
de6d9b64
FB
202#define __align8
203
204#endif
205
6d4985bb
FB
206#ifdef __GNUC__
207
208struct unaligned_64 { uint64_t l; } __attribute__((packed));
209struct unaligned_32 { uint32_t l; } __attribute__((packed));
210
211#define LD32(a) (((const struct unaligned_32 *) (a))->l)
212#define LD64(a) (((const struct unaligned_64 *) (a))->l)
213
214#define ST32(a, b) (((struct unaligned_32 *) (a))->l) = (b)
215
216#else /* __GNUC__ */
217
218#define LD32(a) (*((uint32_t*)(a)))
219#define LD64(a) (*((uint64_t*)(a)))
220
221#define ST32(a, b) *((uint32_t*)(a)) = (b)
222
223#endif /* !__GNUC__ */
224
43f1708f
J
225/* PSNR */
226void get_psnr(UINT8 *orig_image[3], UINT8 *coded_image[3],
227 int orig_linesize[3], int coded_linesize,
228 AVCodecContext *avctx);
229
de6d9b64 230#endif