qpel.c

从FFMPEG转换而来的H264解码程序,VC下编译..

/*****************************************************************************
 *
 *  XVID MPEG-4 VIDEO CODEC
 *  - QPel interpolation -
 *
 *  Copyright(C) 2003 Pascal Massimino <[e-mail]>
 *
 *  This program is free software ; you can redistribute it and/or modify
 *  it under the terms of the GNU General Public License as published by
 *  the Free Software Foundation ; either version 2 of the License, or
 *  (at your option) any later version.
 *
 *  This program 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 General Public License for more details.
 *
 *  You should have received a copy of the GNU General Public License
 *  along with this program ; if not, write to the Free Software
 *  Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA  02111-1307 USA
 *
 * $Id: qpel.c,v 1.8 2005/11/22 10:23:01 suxen_drol Exp $
 *
 ****************************************************************************/

#ifndef XVID_AUTO_INCLUDE

#include <stdio.h>

#include "../portab.h"
#include "qpel.h"

/* Quarterpel FIR definition
 ****************************************************************************/

static const int32_t FIR_Tab_8[9][8] = {
	{ 14, -3,  2, -1,  0,  0,  0,  0 },
	{ 23, 19, -6,  3, -1,  0,  0,  0 },
	{ -7, 20, 20, -6,  3, -1,  0,  0 },
	{  3, -6, 20, 20, -6,  3, -1,  0 },
	{ -1,  3, -6, 20, 20, -6,  3, -1 },
	{  0, -1,  3, -6, 20, 20, -6,  3 },
	{  0,  0, -1,  3, -6, 20, 20, -7 },
	{  0,  0,  0, -1,  3, -6, 19, 23 },
	{  0,  0,  0,  0, -1,  2, -3, 14 }
};

static const int32_t FIR_Tab_16[17][16] = {
	{ 14, -3,  2, -1,  0,  0,  0,  0,  0,  0,  0,  0,  0,  0,  0,  0 },
	{ 23, 19, -6,  3, -1,  0,  0,  0,  0,  0,  0,  0,  0,  0,  0,  0 },
	{ -7, 20, 20, -6,  3, -1,  0,  0,  0,  0,  0,  0,  0,  0,  0,  0 },
	{  3, -6, 20, 20, -6,  3, -1,  0,  0,  0,  0,  0,  0,  0,  0,  0 },
	{ -1,  3, -6, 20, 20, -6,  3, -1,  0,  0,  0,  0,  0,  0,  0,  0 },
	{  0, -1,  3, -6, 20, 20, -6,  3, -1,  0,  0,  0,  0,  0,  0,  0 },
	{  0,  0, -1,  3, -6, 20, 20, -6,  3, -1,  0,  0,  0,  0,  0,  0 },
	{  0,  0,  0, -1,  3, -6, 20, 20, -6,  3, -1,  0,  0,  0,  0,  0 },
	{  0,  0,  0,  0, -1,  3, -6, 20, 20, -6,  3, -1,  0,  0,  0,  0 },
	{  0,  0,  0,  0,  0, -1,  3, -6, 20, 20, -6,  3, -1,  0,  0,  0 },
	{  0,  0,  0,  0,  0,  0, -1,  3, -6, 20, 20, -6,  3, -1,  0,  0 },
	{  0,  0,  0,  0,  0,  0,  0, -1,  3, -6, 20, 20, -6,  3, -1,  0 },
	{  0,  0,  0,  0,  0,  0,  0,  0, -1,  3, -6, 20, 20, -6,  3, -1 },
	{  0,  0,  0,  0,  0,  0,  0,  0,  0, -1,  3, -6, 20, 20, -6,  3 },
	{  0,  0,  0,  0,  0,  0,  0,  0,  0,  0, -1,  3, -6, 20, 20, -7 },
	{  0,  0,  0,  0,  0,  0,  0,  0,  0,  0,  0, -1,  3, -6, 19, 23 },
	{  0,  0,  0,  0,  0,  0,  0,  0,  0,  0,  0,  0, -1,  2, -3, 14 }
};

/* Implementation
 ****************************************************************************/

#define XVID_AUTO_INCLUDE
/* First auto include this file to generate reference code for SIMD versions
 * This set of functions are good for educational purpose, because they're
 * straightforward to understand, use loops and so on... But obviously they
 * sux when it comes to speed */
#define REFERENCE_CODE

/* 16x? filters */

#define SIZE  16
#define TABLE FIR_Tab_16

#define STORE(d,s)  (d) = (s)
#define FUNC_H      H_Pass_16_C_ref
#define FUNC_V      V_Pass_16_C_ref
#define FUNC_HA     H_Pass_Avrg_16_C_ref
#define FUNC_VA     V_Pass_Avrg_16_C_ref
#define FUNC_HA_UP  H_Pass_Avrg_Up_16_C_ref
#define FUNC_VA_UP  V_Pass_Avrg_Up_16_C_ref

#include "qpel.c"   /* self-include ourself */

/* note: B-frame always uses Rnd=0... */
#define STORE(d,s)  (d) = ( (s)+(d)+1 ) >> 1
#define FUNC_H      H_Pass_16_Add_C_ref
#define FUNC_V      V_Pass_16_Add_C_ref
#define FUNC_HA     H_Pass_Avrg_16_Add_C_ref
#define FUNC_VA     V_Pass_Avrg_16_Add_C_ref
#define FUNC_HA_UP  H_Pass_Avrg_Up_16_Add_C_ref
#define FUNC_VA_UP  V_Pass_Avrg_Up_16_Add_C_ref

#include "qpel.c"   /* self-include ourself */

#undef SIZE
#undef TABLE

/* 8x? filters */

#define SIZE  8
#define TABLE FIR_Tab_8

#define STORE(d,s)  (d) = (s)
#define FUNC_H      H_Pass_8_C_ref
#define FUNC_V      V_Pass_8_C_ref
#define FUNC_HA     H_Pass_Avrg_8_C_ref
#define FUNC_VA     V_Pass_Avrg_8_C_ref
#define FUNC_HA_UP  H_Pass_Avrg_Up_8_C_ref
#define FUNC_VA_UP  V_Pass_Avrg_Up_8_C_ref

#include "qpel.c"   /* self-include ourself */

/* note: B-frame always uses Rnd=0... */
#define STORE(d,s)  (d) = ( (s)+(d)+1 ) >> 1
#define FUNC_H      H_Pass_8_Add_C_ref
#define FUNC_V      V_Pass_8_Add_C_ref
#define FUNC_HA     H_Pass_Avrg_8_Add_C_ref
#define FUNC_VA     V_Pass_Avrg_8_Add_C_ref
#define FUNC_HA_UP  H_Pass_Avrg_Up_8_Add_C_ref
#define FUNC_VA_UP  V_Pass_Avrg_Up_8_Add_C_ref

#include "qpel.c"   /* self-include ourself */

#undef SIZE
#undef TABLE

/* Then we define more optimized C version where loops are unrolled, where
 * FIR coeffcients are not read from memory but are hardcoded in instructions
 * They should be faster */
#undef REFERENCE_CODE

/* 16x? filters */

#define SIZE  16

#define STORE(d,s)  (d) = (s)
#define FUNC_H      H_Pass_16_C
#define FUNC_V      V_Pass_16_C
#define FUNC_HA     H_Pass_Avrg_16_C
#define FUNC_VA     V_Pass_Avrg_16_C
#define FUNC_HA_UP  H_Pass_Avrg_Up_16_C
#define FUNC_VA_UP  V_Pass_Avrg_Up_16_C

#include "qpel.c"   /* self-include ourself */

/* note: B-frame always uses Rnd=0... */
#define STORE(d,s)  (d) = ( (s)+(d)+1 ) >> 1
#define FUNC_H      H_Pass_16_Add_C
#define FUNC_V      V_Pass_16_Add_C
#define FUNC_HA     H_Pass_Avrg_16_Add_C
#define FUNC_VA     V_Pass_Avrg_16_Add_C
#define FUNC_HA_UP  H_Pass_Avrg_Up_16_Add_C
#define FUNC_VA_UP  V_Pass_Avrg_Up_16_Add_C

#include "qpel.c"   /* self-include ourself */

#undef SIZE
#undef TABLE

/* 8x? filters */

#define SIZE  8
#define TABLE FIR_Tab_8

#define STORE(d,s)  (d) = (s)
#define FUNC_H      H_Pass_8_C
#define FUNC_V      V_Pass_8_C
#define FUNC_HA     H_Pass_Avrg_8_C
#define FUNC_VA     V_Pass_Avrg_8_C
#define FUNC_HA_UP  H_Pass_Avrg_Up_8_C
#define FUNC_VA_UP  V_Pass_Avrg_Up_8_C

#include "qpel.c"   /* self-include ourself */

/* note: B-frame always uses Rnd=0... */
#define STORE(d,s)  (d) = ( (s)+(d)+1 ) >> 1
#define FUNC_H      H_Pass_8_Add_C
#define FUNC_V      V_Pass_8_Add_C
#define FUNC_HA     H_Pass_Avrg_8_Add_C
#define FUNC_VA     V_Pass_Avrg_8_Add_C
#define FUNC_HA_UP  H_Pass_Avrg_Up_8_Add_C
#define FUNC_VA_UP  V_Pass_Avrg_Up_8_Add_C

#include "qpel.c"   /* self-include ourself */

#undef SIZE
#undef TABLE
#undef XVID_AUTO_INCLUDE

/* Global scope hooks
 ****************************************************************************/

XVID_QP_FUNCS *xvid_QP_Funcs = NULL;
XVID_QP_FUNCS *xvid_QP_Add_Funcs = NULL;

/* Reference plain C impl. declaration
 ****************************************************************************/

XVID_QP_FUNCS xvid_QP_Funcs_C_ref = {
	H_Pass_16_C_ref, H_Pass_Avrg_16_C_ref, H_Pass_Avrg_Up_16_C_ref,
	V_Pass_16_C_ref, V_Pass_Avrg_16_C_ref, V_Pass_Avrg_Up_16_C_ref,

	H_Pass_8_C_ref, H_Pass_Avrg_8_C_ref, H_Pass_Avrg_Up_8_C_ref,
	V_Pass_8_C_ref, V_Pass_Avrg_8_C_ref, V_Pass_Avrg_Up_8_C_ref
};

XVID_QP_FUNCS xvid_QP_Add_Funcs_C_ref = {
	H_Pass_16_Add_C_ref, H_Pass_Avrg_16_Add_C_ref, H_Pass_Avrg_Up_16_Add_C_ref,
	V_Pass_16_Add_C_ref, V_Pass_Avrg_16_Add_C_ref, V_Pass_Avrg_Up_16_Add_C_ref,

	H_Pass_8_Add_C_ref, H_Pass_Avrg_8_Add_C_ref, H_Pass_Avrg_Up_8_Add_C_ref,
	V_Pass_8_Add_C_ref, V_Pass_Avrg_8_Add_C_ref, V_Pass_Avrg_Up_8_Add_C_ref
};

/* Plain C impl. declaration (faster than ref one)
 ****************************************************************************/

XVID_QP_FUNCS xvid_QP_Funcs_C = {
	H_Pass_16_C, H_Pass_Avrg_16_C, H_Pass_Avrg_Up_16_C,
	V_Pass_16_C, V_Pass_Avrg_16_C, V_Pass_Avrg_Up_16_C,

	H_Pass_8_C, H_Pass_Avrg_8_C, H_Pass_Avrg_Up_8_C,
	V_Pass_8_C, V_Pass_Avrg_8_C, V_Pass_Avrg_Up_8_C
};

XVID_QP_FUNCS xvid_QP_Add_Funcs_C = {
	H_Pass_16_Add_C, H_Pass_Avrg_16_Add_C, H_Pass_Avrg_Up_16_Add_C,
	V_Pass_16_Add_C, V_Pass_Avrg_16_Add_C, V_Pass_Avrg_Up_16_Add_C,

	H_Pass_8_Add_C, H_Pass_Avrg_8_Add_C, H_Pass_Avrg_Up_8_Add_C,
	V_Pass_8_Add_C, V_Pass_Avrg_8_Add_C, V_Pass_Avrg_Up_8_Add_C
};

/* mmx impl. declaration (see. qpel_mmx.asm
 ****************************************************************************/

#ifdef ARCH_IS_IA32
extern XVID_QP_PASS_SIGNATURE(xvid_H_Pass_16_mmx);
extern XVID_QP_PASS_SIGNATURE(xvid_H_Pass_Avrg_16_mmx);
extern XVID_QP_PASS_SIGNATURE(xvid_H_Pass_Avrg_Up_16_mmx);
extern XVID_QP_PASS_SIGNATURE(xvid_V_Pass_16_mmx);
extern XVID_QP_PASS_SIGNATURE(xvid_V_Pass_Avrg_16_mmx);
extern XVID_QP_PASS_SIGNATURE(xvid_V_Pass_Avrg_Up_16_mmx);

extern XVID_QP_PASS_SIGNATURE(xvid_H_Pass_8_mmx);
extern XVID_QP_PASS_SIGNATURE(xvid_H_Pass_Avrg_8_mmx);
extern XVID_QP_PASS_SIGNATURE(xvid_H_Pass_Avrg_Up_8_mmx);
extern XVID_QP_PASS_SIGNATURE(xvid_V_Pass_8_mmx);
extern XVID_QP_PASS_SIGNATURE(xvid_V_Pass_Avrg_8_mmx);
extern XVID_QP_PASS_SIGNATURE(xvid_V_Pass_Avrg_Up_8_mmx);

extern XVID_QP_PASS_SIGNATURE(xvid_H_Pass_Add_16_mmx);
extern XVID_QP_PASS_SIGNATURE(xvid_H_Pass_Avrg_Add_16_mmx);
extern XVID_QP_PASS_SIGNATURE(xvid_H_Pass_Avrg_Up_Add_16_mmx);
extern XVID_QP_PASS_SIGNATURE(xvid_V_Pass_Add_16_mmx);
extern XVID_QP_PASS_SIGNATURE(xvid_V_Pass_Avrg_Add_16_mmx);
extern XVID_QP_PASS_SIGNATURE(xvid_V_Pass_Avrg_Up_Add_16_mmx);

extern XVID_QP_PASS_SIGNATURE(xvid_H_Pass_8_Add_mmx);
extern XVID_QP_PASS_SIGNATURE(xvid_H_Pass_Avrg_8_Add_mmx);
extern XVID_QP_PASS_SIGNATURE(xvid_H_Pass_Avrg_Up_8_Add_mmx);
extern XVID_QP_PASS_SIGNATURE(xvid_V_Pass_8_Add_mmx);
extern XVID_QP_PASS_SIGNATURE(xvid_V_Pass_Avrg_8_Add_mmx);
extern XVID_QP_PASS_SIGNATURE(xvid_V_Pass_Avrg_Up_8_Add_mmx);

XVID_QP_FUNCS xvid_QP_Funcs_mmx = {
	xvid_H_Pass_16_mmx, xvid_H_Pass_Avrg_16_mmx, xvid_H_Pass_Avrg_Up_16_mmx,
	xvid_V_Pass_16_mmx, xvid_V_Pass_Avrg_16_mmx, xvid_V_Pass_Avrg_Up_16_mmx,

	xvid_H_Pass_8_mmx, xvid_H_Pass_Avrg_8_mmx, xvid_H_Pass_Avrg_Up_8_mmx,
	xvid_V_Pass_8_mmx, xvid_V_Pass_Avrg_8_mmx, xvid_V_Pass_Avrg_Up_8_mmx
};

XVID_QP_FUNCS xvid_QP_Add_Funcs_mmx = {
	xvid_H_Pass_Add_16_mmx, xvid_H_Pass_Avrg_Add_16_mmx, xvid_H_Pass_Avrg_Up_Add_16_mmx,
	xvid_V_Pass_Add_16_mmx, xvid_V_Pass_Avrg_Add_16_mmx, xvid_V_Pass_Avrg_Up_Add_16_mmx,

	xvid_H_Pass_8_Add_mmx, xvid_H_Pass_Avrg_8_Add_mmx, xvid_H_Pass_Avrg_Up_8_Add_mmx,
	xvid_V_Pass_8_Add_mmx, xvid_V_Pass_Avrg_8_Add_mmx, xvid_V_Pass_Avrg_Up_8_Add_mmx,
};
#endif /* ARCH_IS_IA32 */


/* altivec impl. declaration (see qpel_altivec.c)
 ****************************************************************************/

#ifdef ARCH_IS_PPC

extern XVID_QP_PASS_SIGNATURE(H_Pass_16_Altivec_C);
extern XVID_QP_PASS_SIGNATURE(H_Pass_Avrg_16_Altivec_C);
extern XVID_QP_PASS_SIGNATURE(H_Pass_Avrg_Up_16_Altivec_C);
extern XVID_QP_PASS_SIGNATURE(V_Pass_16_Altivec_C);
extern XVID_QP_PASS_SIGNATURE(V_Pass_Avrg_16_Altivec_C);
extern XVID_QP_PASS_SIGNATURE(V_Pass_Avrg_Up_16_Altivec_C);

extern XVID_QP_PASS_SIGNATURE(H_Pass_8_Altivec_C);
extern XVID_QP_PASS_SIGNATURE(H_Pass_Avrg_8_Altivec_C);
extern XVID_QP_PASS_SIGNATURE(H_Pass_Avrg_Up_8_Altivec_C);
extern XVID_QP_PASS_SIGNATURE(V_Pass_8_Altivec_C);
extern XVID_QP_PASS_SIGNATURE(V_Pass_Avrg_8_Altivec_C);
extern XVID_QP_PASS_SIGNATURE(V_Pass_Avrg_Up_8_Altivec_C);


extern XVID_QP_PASS_SIGNATURE(H_Pass_16_Add_Altivec_C);
extern XVID_QP_PASS_SIGNATURE(H_Pass_Avrg_16_Add_Altivec_C);
extern XVID_QP_PASS_SIGNATURE(H_Pass_Avrg_Up_16_Add_Altivec_C);
extern XVID_QP_PASS_SIGNATURE(V_Pass_16_Add_Altivec_C);
extern XVID_QP_PASS_SIGNATURE(V_Pass_Avrg_16_Add_Altivec_C);
extern XVID_QP_PASS_SIGNATURE(V_Pass_Avrg_Up_16_Add_Altivec_C);

extern XVID_QP_PASS_SIGNATURE(H_Pass_8_Add_Altivec_C);
extern XVID_QP_PASS_SIGNATURE(H_Pass_Avrg_8_Add_Altivec_C);
extern XVID_QP_PASS_SIGNATURE(H_Pass_Avrg_Up_8_Add_Altivec_C);
extern XVID_QP_PASS_SIGNATURE(V_Pass_8_Add_Altivec_C);
extern XVID_QP_PASS_SIGNATURE(V_Pass_Avrg_8_Add_Altivec_C);
extern XVID_QP_PASS_SIGNATURE(V_Pass_Avrg_Up_8_Add_Altivec_C);

XVID_QP_FUNCS xvid_QP_Funcs_Altivec_C = {
	H_Pass_16_Altivec_C, H_Pass_Avrg_16_Altivec_C, H_Pass_Avrg_Up_16_Altivec_C,
	V_Pass_16_Altivec_C, V_Pass_Avrg_16_Altivec_C, V_Pass_Avrg_Up_16_Altivec_C,

	H_Pass_8_Altivec_C, H_Pass_Avrg_8_Altivec_C, H_Pass_Avrg_Up_8_Altivec_C,
	V_Pass_8_Altivec_C, V_Pass_Avrg_8_Altivec_C, V_Pass_Avrg_Up_8_Altivec_C
};

XVID_QP_FUNCS xvid_QP_Add_Funcs_Altivec_C = {
	H_Pass_16_Add_Altivec_C, H_Pass_Avrg_16_Add_Altivec_C, H_Pass_Avrg_Up_16_Add_Altivec_C,
	V_Pass_16_Add_Altivec_C, V_Pass_Avrg_16_Add_Altivec_C, V_Pass_Avrg_Up_16_Add_Altivec_C,

	H_Pass_8_Add_Altivec_C, H_Pass_Avrg_8_Add_Altivec_C, H_Pass_Avrg_Up_8_Add_Altivec_C,
	V_Pass_8_Add_Altivec_C, V_Pass_Avrg_8_Add_Altivec_C, V_Pass_Avrg_Up_8_Add_Altivec_C
};

#endif /* ARCH_IS_PPC */

/* mmx impl. (for 64bit bus) declaration (see. qpel_mmx.asm
 ****************************************************************************/

#ifdef ARCH_IS_X86_64
extern XVID_QP_PASS_SIGNATURE(xvid_H_Pass_16_x86_64);
extern XVID_QP_PASS_SIGNATURE(xvid_H_Pass_Avrg_16_x86_64);
extern XVID_QP_PASS_SIGNATURE(xvid_H_Pass_Avrg_Up_16_x86_64);
extern XVID_QP_PASS_SIGNATURE(xvid_V_Pass_16_x86_64);
extern XVID_QP_PASS_SIGNATURE(xvid_V_Pass_Avrg_16_x86_64);
extern XVID_QP_PASS_SIGNATURE(xvid_V_Pass_Avrg_Up_16_x86_64);

extern XVID_QP_PASS_SIGNATURE(xvid_H_Pass_8_x86_64);
extern XVID_QP_PASS_SIGNATURE(xvid_H_Pass_Avrg_8_x86_64);
extern XVID_QP_PASS_SIGNATURE(xvid_H_Pass_Avrg_Up_8_x86_64);
extern XVID_QP_PASS_SIGNATURE(xvid_V_Pass_8_x86_64);
extern XVID_QP_PASS_SIGNATURE(xvid_V_Pass_Avrg_8_x86_64);
extern XVID_QP_PASS_SIGNATURE(xvid_V_Pass_Avrg_Up_8_x86_64);

extern XVID_QP_PASS_SIGNATURE(xvid_H_Pass_Add_16_x86_64);
extern XVID_QP_PASS_SIGNATURE(xvid_H_Pass_Avrg_Add_16_x86_64);
extern XVID_QP_PASS_SIGNATURE(xvid_H_Pass_Avrg_Up_Add_16_x86_64);
extern XVID_QP_PASS_SIGNATURE(xvid_V_Pass_Add_16_x86_64);
extern XVID_QP_PASS_SIGNATURE(xvid_V_Pass_Avrg_Add_16_x86_64);
extern XVID_QP_PASS_SIGNATURE(xvid_V_Pass_Avrg_Up_Add_16_x86_64);

extern XVID_QP_PASS_SIGNATURE(xvid_H_Pass_8_Add_x86_64);
extern XVID_QP_PASS_SIGNATURE(xvid_H_Pass_Avrg_8_Add_x86_64);