summaryrefslogtreecommitdiff
path: root/src/libffmpeg/libavcodec/dsputil.h
diff options
context:
space:
mode:
Diffstat (limited to 'src/libffmpeg/libavcodec/dsputil.h')
-rw-r--r--src/libffmpeg/libavcodec/dsputil.h108
1 files changed, 59 insertions, 49 deletions
diff --git a/src/libffmpeg/libavcodec/dsputil.h b/src/libffmpeg/libavcodec/dsputil.h
index 6126c78b5..d8346d509 100644
--- a/src/libffmpeg/libavcodec/dsputil.h
+++ b/src/libffmpeg/libavcodec/dsputil.h
@@ -1,6 +1,7 @@
/*
* DSP utils
* Copyright (c) 2000, 2001, 2002 Fabrice Bellard.
+ * Copyright (c) 2002-2004 Michael Niedermayer <michaelni@gmx.at>
*
* This library is free software; you can redistribute it and/or
* modify it under the terms of the GNU Lesser General Public
@@ -29,28 +30,31 @@
#include "common.h"
#include "avcodec.h"
-#include "xineutils.h"
#if defined(ARCH_X86)
-#define HAVE_MMX 1
+#define HAVE_MMX 1
#endif
-#undef DEBUG
-
+//#define DEBUG
/* dct code */
typedef short DCTELEM;
void fdct_ifast (DCTELEM *data);
+void fdct_ifast248 (DCTELEM *data);
void ff_jpeg_fdct_islow (DCTELEM *data);
+void ff_fdct248_islow (DCTELEM *data);
void j_rev_dct (DCTELEM *data);
void ff_fdct_mmx(DCTELEM *block);
+void ff_fdct_mmx2(DCTELEM *block);
+void ff_fdct_sse2(DCTELEM *block);
/* encoding scans */
extern const uint8_t ff_alternate_horizontal_scan[64];
extern const uint8_t ff_alternate_vertical_scan[64];
extern const uint8_t ff_zigzag_direct[64];
+extern const uint8_t ff_zigzag248_direct[64];
/* pixel operations */
#define MAX_NEG_CROP 384
@@ -80,6 +84,7 @@ void clear_blocks_c(DCTELEM *blocks);
/* add and put pixel (decoding) */
// blocksizes for op_pixels_func are 8x4,8x8 16x8 16x16
+//h for op_pixels_func is limited to {width/2, width} but never larger than 16 and never smaller then 4
typedef void (*op_pixels_func)(uint8_t *block/*align width (8 or 16)*/, const uint8_t *pixels/*align 1*/, int line_size, int h);
typedef void (*tpel_mc_func)(uint8_t *block/*align width (8 or 16)*/, const uint8_t *pixels/*align 1*/, int line_size, int w, int h);
typedef void (*qpel_mc_func)(uint8_t *dst/*align width (8 or 16)*/, uint8_t *src/*align 1*/, int stride);
@@ -110,10 +115,9 @@ static void a(uint8_t *block, const uint8_t *pixels, int line_size, int h){\
}
/* motion estimation */
-
-typedef int (*op_pixels_abs_func)(uint8_t *blk1/*align width (8 or 16)*/, uint8_t *blk2/*align 1*/, int line_size)/* __attribute__ ((const))*/;
-
-typedef int (*me_cmp_func)(void /*MpegEncContext*/ *s, uint8_t *blk1/*align width (8 or 16)*/, uint8_t *blk2/*align 1*/, int line_size)/* __attribute__ ((const))*/;
+// h is limited to {width/2, width, 2*width} but never larger than 16 and never smaller then 2
+// allthough currently h<4 is not used as functions with width <8 are not used and neither implemented
+typedef int (*me_cmp_func)(void /*MpegEncContext*/ *s, uint8_t *blk1/*align width (8 or 16)*/, uint8_t *blk2/*align 1*/, int line_size, int h)/* __attribute__ ((const))*/;
/**
@@ -137,25 +141,28 @@ typedef struct DSPContext {
void (*clear_blocks)(DCTELEM *blocks/*align 16*/);
int (*pix_sum)(uint8_t * pix, int line_size);
int (*pix_norm1)(uint8_t * pix, int line_size);
- me_cmp_func sad[2]; /* identical to pix_absAxA except additional void * */
- me_cmp_func sse[2];
- me_cmp_func hadamard8_diff[2];
- me_cmp_func dct_sad[2];
- me_cmp_func quant_psnr[2];
- me_cmp_func bit[2];
- me_cmp_func rd[2];
- int (*hadamard8_abs )(uint8_t *src, int stride, int mean);
-
- me_cmp_func me_pre_cmp[11];
- me_cmp_func me_cmp[11];
- me_cmp_func me_sub_cmp[11];
- me_cmp_func mb_cmp[11];
-
- /* maybe create an array for 16/8/4/2 functions */
+// 16x16 8x8 4x4 2x2 16x8 8x4 4x2 8x16 4x8 2x4
+
+ me_cmp_func sad[5]; /* identical to pix_absAxA except additional void * */
+ me_cmp_func sse[5];
+ me_cmp_func hadamard8_diff[5];
+ me_cmp_func dct_sad[5];
+ me_cmp_func quant_psnr[5];
+ me_cmp_func bit[5];
+ me_cmp_func rd[5];
+ me_cmp_func vsad[5];
+ me_cmp_func vsse[5];
+
+ me_cmp_func me_pre_cmp[5];
+ me_cmp_func me_cmp[5];
+ me_cmp_func me_sub_cmp[5];
+ me_cmp_func mb_cmp[5];
+ me_cmp_func ildct_cmp[5]; //only width 16 used
+
/**
* Halfpel motion compensation with rounding (a+b+1)>>1.
* this is an array[4][4] of motion compensation funcions for 4
- * horizontal blocksizes (2,4,8,16) and the 4 halfpel positions<br>
+ * horizontal blocksizes (8,16) and the 4 halfpel positions<br>
* *pixels_tab[ 0->16xH 1->8xH ][ xhalfpel + 2*yhalfpel ]
* @param block destination where the result is stored
* @param pixels source
@@ -167,7 +174,7 @@ typedef struct DSPContext {
/**
* Halfpel motion compensation with rounding (a+b+1)>>1.
* This is an array[4][4] of motion compensation functions for 4
- * horizontal blocksizes (2,4,8,16) and the 4 halfpel positions<br>
+ * horizontal blocksizes (8,16) and the 4 halfpel positions<br>
* *pixels_tab[ 0->16xH 1->8xH ][ xhalfpel + 2*yhalfpel ]
* @param block destination into which the result is averaged (a+b+1)>>1
* @param pixels source
@@ -227,14 +234,7 @@ typedef struct DSPContext {
qpel_mc_func put_h264_qpel_pixels_tab[3][16];
qpel_mc_func avg_h264_qpel_pixels_tab[3][16];
- op_pixels_abs_func pix_abs16x16;
- op_pixels_abs_func pix_abs16x16_x2;
- op_pixels_abs_func pix_abs16x16_y2;
- op_pixels_abs_func pix_abs16x16_xy2;
- op_pixels_abs_func pix_abs8x8;
- op_pixels_abs_func pix_abs8x8_x2;
- op_pixels_abs_func pix_abs8x8_y2;
- op_pixels_abs_func pix_abs8x8_xy2;
+ me_cmp_func pix_abs[2][4];
/* huffyuv specific */
void (*add_bytes)(uint8_t *dst/*align 16*/, uint8_t *src/*align 16*/, int w);
@@ -246,8 +246,12 @@ typedef struct DSPContext {
void (*sub_hfyu_median_prediction)(uint8_t *dst, uint8_t *src1, uint8_t *src2, int w, int *left, int *left_top);
void (*bswap_buf)(uint32_t *dst, uint32_t *src, int w);
+ void (*h263_v_loop_filter)(uint8_t *src, int stride, int qscale);
+ void (*h263_h_loop_filter)(uint8_t *src, int stride, int qscale);
+
/* (I)DCT */
void (*fdct)(DCTELEM *block/* align 16*/);
+ void (*fdct248)(DCTELEM *block/* align 16*/);
/* IDCT really*/
void (*idct)(DCTELEM *block/* align 16*/);
@@ -295,6 +299,8 @@ void dsputil_init(DSPContext* p, AVCodecContext *avctx);
*/
void ff_block_permute(DCTELEM *block, uint8_t *permutation, const uint8_t *scantable, int last);
+void ff_set_cmp(DSPContext* c, me_cmp_func *cmp, int type);
+
#define BYTE_VEC32(c) ((c)*0x01010101UL)
static inline uint32_t rnd_avg32(uint32_t a, uint32_t b)
@@ -322,25 +328,21 @@ int mm_support(void);
#undef emms_c
-#if 0
#define MM_MMX 0x0001 /* standard MMX */
#define MM_3DNOW 0x0004 /* AMD 3DNOW */
#define MM_MMXEXT 0x0002 /* SSE integer functions or AMD MMX ext */
#define MM_SSE 0x0008 /* SSE functions */
#define MM_SSE2 0x0010 /* PIV SSE2 functions */
-#endif
extern int mm_flags;
void add_pixels_clamped_mmx(const DCTELEM *block, uint8_t *pixels, int line_size);
void put_pixels_clamped_mmx(const DCTELEM *block, uint8_t *pixels, int line_size);
-#if 0
static inline void emms(void)
{
__asm __volatile ("emms;":::"memory");
}
-#endif
#define emms_c() \
@@ -382,7 +384,9 @@ void dsputil_init_alpha(DSPContext* c, AVCodecContext *avctx);
extern int mm_flags;
#if defined(HAVE_ALTIVEC) && !defined(CONFIG_DARWIN)
+#define pixel altivec_pixel
#include <altivec.h>
+#undef pixel
#endif
#define __align8 __attribute__ ((aligned (16)))
@@ -483,12 +487,24 @@ void ff_mdct_calc(MDCTContext *s, FFTSample *out,
const FFTSample *input, FFTSample *tmp);
void ff_mdct_end(MDCTContext *s);
-#define WARPER88_1616(name8, name16)\
-static int name16(void /*MpegEncContext*/ *s, uint8_t *dst, uint8_t *src, int stride){\
- return name8(s, dst , src , stride)\
- +name8(s, dst+8 , src+8 , stride)\
- +name8(s, dst +8*stride, src +8*stride, stride)\
- +name8(s, dst+8+8*stride, src+8+8*stride, stride);\
+#define WARPER8_16(name8, name16)\
+static int name16(void /*MpegEncContext*/ *s, uint8_t *dst, uint8_t *src, int stride, int h){\
+ return name8(s, dst , src , stride, h)\
+ +name8(s, dst+8 , src+8 , stride, h);\
+}
+
+#define WARPER8_16_SQ(name8, name16)\
+static int name16(void /*MpegEncContext*/ *s, uint8_t *dst, uint8_t *src, int stride, int h){\
+ int score=0;\
+ score +=name8(s, dst , src , stride, 8);\
+ score +=name8(s, dst+8 , src+8 , stride, 8);\
+ if(h==16){\
+ dst += 8*stride;\
+ src += 8*stride;\
+ score +=name8(s, dst , src , stride, 8);\
+ score +=name8(s, dst+8 , src+8 , stride, 8);\
+ }\
+ return score;\
}
#ifndef HAVE_LRINTF
@@ -506,10 +522,4 @@ static inline long int lrintf(float x)
}
#endif
-#if defined(CONFIG_OS2) || defined(CONFIG_SUNOS)
-static inline float floorf(float f) {
- return floor(f);
-}
-#endif
-
#endif
generated by cgit v1.2.3 (git 2.25.1) at 2025-04-27 13:43:43 +0000