falkTX
/
FFmpeg
mirror of https://github.com/falkTX/FFmpeg.git
1
0
Fork 0
Code Issues 0 Releases 338 Wiki Activity
Browse Source

dsputil: Split motion estimation compare bits off into their own context

tags/n2.4
Diego Biurrun 11 years ago
parent
a578b0407d
commit
2d60444331
31 changed files with 184 additions and 215 deletions
Split View
Diff Options
Show Stats Download Patch File Download Diff File
  1. +9
    -9
      configure
  2. +1
    -1
      libavcodec/Makefile
  3. +3
    -2
      libavcodec/ac3enc.c
  4. +2
    -2
      libavcodec/ac3enc.h
  5. +2
    -2
      libavcodec/arm/Makefile
  6. +0
    -0
      libavcodec/arm/me_cmp_armv6.S
  7. +2
    -3
      libavcodec/arm/me_cmp_init_arm.c
  8. +1
    -1
      libavcodec/dv.h
  9. +5
    -5
      libavcodec/dvenc.c
  10. +5
    -5
      libavcodec/error_resilience.c
  11. +2
    -2
      libavcodec/error_resilience.h
  12. +4
    -4
      libavcodec/h264.c
  13. +2
    -2
      libavcodec/h264.h
  14. +1
    -2
      libavcodec/h264_slice.c
  15. +9
    -19
      libavcodec/me_cmp.c
  16. +12
    -24
      libavcodec/me_cmp.h
  17. +22
    -21
      libavcodec/motion_est.c
  18. +32
    -33
      libavcodec/motion_est_template.c
  19. +1
    -1
      libavcodec/mpeg4videoenc.c
  20. +2
    -2
      libavcodec/mpegvideo.c
  21. +2
    -2
      libavcodec/mpegvideo.h
  22. +44
    -52
      libavcodec/mpegvideo_enc.c
  23. +1
    -1
      libavcodec/mpegvideoencdsp.c
  24. +1
    -1
      libavcodec/ppc/Makefile
  25. +2
    -2
      libavcodec/ppc/me_cmp.c
  26. +5
    -5
      libavcodec/svq1enc.c
  27. +2
    -2
      libavcodec/svq1enc.h
  28. +3
    -3
      libavcodec/utils.c
  29. +2
    -2
      libavcodec/x86/Makefile
  30. +1
    -1
      libavcodec/x86/me_cmp.asm
  31. +4
    -4
      libavcodec/x86/me_cmp_init.c

+ 9
- 9
configure View File

@@ -1533,7 +1533,6 @@ CONFIG_EXTRA="
blockdsp
bswapdsp
cabac
dsputil
dvprofile
fdctdsp
gcrypt
@@ -1552,6 +1551,7 @@ CONFIG_EXTRA="
intrax8
lgplv3
lpc
me_cmp
mpeg_er
mpegaudio
mpegaudiodsp
@@ -1707,24 +1707,24 @@ threads_if_any="$THREADS_LIST"

# subsystems
dct_select="rdft"
dsputil_select="fdctdsp idctdsp pixblockdsp"
error_resilience_select="dsputil"
error_resilience_select="me_cmp"
intrax8_select="error_resilience"
mdct_select="fft"
rdft_select="fft"
me_cmp_select="fdctdsp idctdsp pixblockdsp"
mpeg_er_select="error_resilience"
mpegaudio_select="mpegaudiodsp"
mpegaudiodsp_select="dct"
mpegvideo_select="blockdsp dsputil hpeldsp idctdsp videodsp"
mpegvideoenc_select="dsputil mpegvideo pixblockdsp qpeldsp"
mpegvideo_select="blockdsp hpeldsp idctdsp me_cmp videodsp"
mpegvideoenc_select="me_cmp mpegvideo pixblockdsp qpeldsp"

# decoders / encoders
aac_decoder_select="mdct sinewin"
aac_encoder_select="audio_frame_queue mdct sinewin"
aac_latm_decoder_select="aac_decoder aac_latm_parser"
ac3_decoder_select="ac3_parser ac3dsp bswapdsp mdct"
ac3_encoder_select="ac3dsp audiodsp dsputil mdct"
ac3_fixed_encoder_select="ac3dsp audiodsp dsputil mdct"
ac3_encoder_select="ac3dsp audiodsp mdct me_cmp"
ac3_fixed_encoder_select="ac3dsp audiodsp mdct me_cmp"
aic_decoder_select="golomb idctdsp"
alac_encoder_select="lpc"
als_decoder_select="bswapdsp"
@@ -1752,7 +1752,7 @@ dca_decoder_select="mdct"
dnxhd_decoder_select="blockdsp idctdsp"
dnxhd_encoder_select="aandcttables blockdsp fdctdsp idctdsp mpegvideoenc pixblockdsp"
dvvideo_decoder_select="dvprofile idctdsp"
dvvideo_encoder_select="dsputil dvprofile fdctdsp pixblockdsp"
dvvideo_encoder_select="dvprofile fdctdsp me_cmp pixblockdsp"
dxa_decoder_deps="zlib"
eac3_decoder_select="ac3_decoder"
eac3_encoder_select="ac3_encoder"
@@ -1856,7 +1856,7 @@ shorten_decoder_select="golomb"
sipr_decoder_select="lsp"
sp5x_decoder_select="mjpeg_decoder"
svq1_decoder_select="hpeldsp"
svq1_encoder_select="aandcttables dsputil hpeldsp mpegvideoenc"
svq1_encoder_select="aandcttables hpeldsp me_cmp mpegvideoenc"
svq3_decoder_select="h264_decoder hpeldsp tpeldsp"
svq3_decoder_suggest="zlib"
tak_decoder_select="audiodsp"


+ 1
- 1
libavcodec/Makefile View File

@@ -35,7 +35,6 @@ OBJS-$(CONFIG_BLOCKDSP) += blockdsp.o
OBJS-$(CONFIG_BSWAPDSP) += bswapdsp.o
OBJS-$(CONFIG_CABAC) += cabac.o
OBJS-$(CONFIG_DCT) += dct.o dct32_fixed.o dct32_float.o
OBJS-$(CONFIG_DSPUTIL) += dsputil.o
OBJS-$(CONFIG_DXVA2) += dxva2.o
OBJS-$(CONFIG_ERROR_RESILIENCE) += error_resilience.o
OBJS-$(CONFIG_FDCTDSP) += fdctdsp.o faandct.o \
@@ -60,6 +59,7 @@ OBJS-$(CONFIG_LIBXVID) += libxvid_rc.o
OBJS-$(CONFIG_LPC) += lpc.o
OBJS-$(CONFIG_LSP) += lsp.o
OBJS-$(CONFIG_MDCT) += mdct_fixed.o mdct_float.o
OBJS-$(CONFIG_ME_CMP) += me_cmp.o
OBJS-$(CONFIG_MPEG_ER) += mpeg_er.o
OBJS-$(CONFIG_MPEGAUDIO) += mpegaudio.o mpegaudiodata.o \
mpegaudiodecheader.o


+ 3
- 2
libavcodec/ac3enc.c View File

@@ -36,6 +36,7 @@
#include "libavutil/internal.h"
#include "libavutil/opt.h"
#include "avcodec.h"
#include "me_cmp.h"
#include "put_bits.h"
#include "audiodsp.h"
#include "ac3dsp.h"
@@ -379,7 +380,7 @@ static void compute_exp_strategy(AC3EncodeContext *s)
exp_strategy[blk] = EXP_NEW;
continue;
}
exp_diff = s->dsp.sad[0](NULL, exp, exp - AC3_MAX_COEFS, 16, 16);
exp_diff = s->mecc.sad[0](NULL, exp, exp - AC3_MAX_COEFS, 16, 16);
exp_strategy[blk] = EXP_REUSE;
if (ch == CPL_CH && exp_diff > (EXP_DIFF_THRESHOLD * (s->blocks[blk].end_freq[ch] - s->start_freq[ch]) / AC3_MAX_COEFS))
exp_strategy[blk] = EXP_NEW;
@@ -2482,7 +2483,7 @@ av_cold int ff_ac3_encode_init(AVCodecContext *avctx)
goto init_fail;

ff_audiodsp_init(&s->adsp);
ff_dsputil_init(&s->dsp, avctx);
ff_me_cmp_init(&s->mecc, avctx);
ff_ac3dsp_init(&s->ac3dsp, avctx->flags & CODEC_FLAG_BITEXACT);

dprint_options(s);


+ 2
- 2
libavcodec/ac3enc.h View File

@@ -35,9 +35,9 @@
#include "ac3.h"
#include "ac3dsp.h"
#include "avcodec.h"
#include "dsputil.h"
#include "fft.h"
#include "mathops.h"
#include "me_cmp.h"
#include "put_bits.h"
#include "audiodsp.h"

@@ -162,9 +162,9 @@ typedef struct AC3EncodeContext {
AC3EncOptions options; ///< encoding options
AVCodecContext *avctx; ///< parent AVCodecContext
PutBitContext pb; ///< bitstream writer context
DSPContext dsp;
AudioDSPContext adsp;
AVFloatDSPContext fdsp;
MECmpContext mecc;
AC3DSPContext ac3dsp; ///< AC-3 optimized functions
FFTContext mdct; ///< FFT context for MDCT calculation
const SampleType *mdct_window; ///< MDCT window function array


+ 2
- 2
libavcodec/arm/Makefile View File

@@ -6,7 +6,6 @@ OBJS-$(CONFIG_AC3DSP) += arm/ac3dsp_init_arm.o \
arm/ac3dsp_arm.o
OBJS-$(CONFIG_AUDIODSP) += arm/audiodsp_init_arm.o
OBJS-$(CONFIG_BLOCKDSP) += arm/blockdsp_init_arm.o
OBJS-$(CONFIG_DSPUTIL) += arm/dsputil_init_arm.o
OBJS-$(CONFIG_FFT) += arm/fft_init_arm.o \
arm/fft_fixed_init_arm.o
OBJS-$(CONFIG_H264CHROMA) += arm/h264chroma_init_arm.o
@@ -19,6 +18,7 @@ OBJS-$(CONFIG_IDCTDSP) += arm/idctdsp_init_arm.o \
arm/idctdsp_arm.o \
arm/jrevdct_arm.o \
arm/simple_idct_arm.o
OBJS-$(CONFIG_ME_CMP) += arm/me_cmp_init_arm.o
OBJS-$(CONFIG_MPEGAUDIODSP) += arm/mpegaudiodsp_init_arm.o
OBJS-$(CONFIG_MPEGVIDEO) += arm/mpegvideo_arm.o
OBJS-$(CONFIG_MPEGVIDEOENC) += arm/mpegvideoencdsp_init_arm.o
@@ -53,13 +53,13 @@ ARMV5TE-OBJS-$(CONFIG_VIDEODSP) += arm/videodsp_init_armv5te.o \
ARMV5TE-OBJS-$(CONFIG_MLP_DECODER) += arm/mlpdsp_armv5te.o

ARMV6-OBJS-$(CONFIG_AC3DSP) += arm/ac3dsp_armv6.o
ARMV6-OBJS-$(CONFIG_DSPUTIL) += arm/dsputil_armv6.o
ARMV6-OBJS-$(CONFIG_H264DSP) += arm/h264dsp_armv6.o
ARMV6-OBJS-$(CONFIG_HPELDSP) += arm/hpeldsp_init_armv6.o \
arm/hpeldsp_armv6.o
ARMV6-OBJS-$(CONFIG_IDCTDSP) += arm/idctdsp_init_armv6.o \
arm/idctdsp_armv6.o \
arm/simple_idct_armv6.o
ARMV6-OBJS-$(CONFIG_ME_CMP) += arm/me_cmp_armv6.o
ARMV6-OBJS-$(CONFIG_MPEGAUDIODSP) += arm/mpegaudiodsp_fixed_armv6.o
ARMV6-OBJS-$(CONFIG_MPEGVIDEOENC) += arm/mpegvideoencdsp_armv6.o
ARMV6-OBJS-$(CONFIG_PIXBLOCKDSP) += arm/pixblockdsp_armv6.o


libavcodec/arm/dsputil_armv6.S → libavcodec/arm/me_cmp_armv6.S View File


libavcodec/arm/dsputil_init_arm.c → libavcodec/arm/me_cmp_init_arm.c View File

@@ -22,7 +22,7 @@
#include "libavutil/cpu.h"
#include "libavutil/arm/cpu.h"
#include "libavcodec/avcodec.h"
#include "libavcodec/dsputil.h"
#include "libavcodec/me_cmp.h"
#include "libavcodec/mpegvideo.h"

int ff_pix_abs16_armv6(MpegEncContext *s, uint8_t *blk1, uint8_t *blk2,
@@ -38,8 +38,7 @@ int ff_pix_abs8_armv6(MpegEncContext *s, uint8_t *blk1, uint8_t *blk2,
int ff_sse16_armv6(MpegEncContext *s, uint8_t *blk1, uint8_t *blk2,
int line_size, int h);


av_cold void ff_dsputil_init_arm(DSPContext *c, AVCodecContext *avctx)
av_cold void ff_me_cmp_init_arm(MECmpContext *c, AVCodecContext *avctx)
{
int cpu_flags = av_get_cpu_flags();


+ 1
- 1
libavcodec/dv.h View File

@@ -28,7 +28,7 @@
#define AVCODEC_DV_H

#include "avcodec.h"
#include "dsputil.h"
#include "me_cmp.h"
#include "get_bits.h"
#include "dv_profile.h"



+ 5
- 5
libavcodec/dvenc.c View File

@@ -28,9 +28,9 @@
#include "libavutil/pixdesc.h"
#include "config.h"
#include "avcodec.h"
#include "dsputil.h"
#include "fdctdsp.h"
#include "internal.h"
#include "me_cmp.h"
#include "pixblockdsp.h"
#include "put_bits.h"
#include "dv.h"
@@ -40,8 +40,8 @@
static av_cold int dvvideo_encode_init(AVCodecContext *avctx)
{
DVVideoContext *s = avctx->priv_data;
DSPContext dsp;
FDCTDSPContext fdsp;
MECmpContext mecc;
PixblockDSPContext pdsp;
int ret;

@@ -65,13 +65,13 @@ static av_cold int dvvideo_encode_init(AVCodecContext *avctx)

dv_vlc_map_tableinit();

ff_dsputil_init(&dsp, avctx);
ff_fdctdsp_init(&fdsp, avctx);
ff_me_cmp_init(&mecc, avctx);
ff_pixblockdsp_init(&pdsp, avctx);
ff_set_cmp(&dsp, dsp.ildct_cmp, avctx->ildct_cmp);
ff_set_cmp(&mecc, mecc.ildct_cmp, avctx->ildct_cmp);

s->get_pixels = pdsp.get_pixels;
s->ildct_cmp = dsp.ildct_cmp[5];
s->ildct_cmp = mecc.ildct_cmp[5];

s->fdct[0] = fdsp.fdct;
s->fdct[1] = fdsp.fdct248;


+ 5
- 5
libavcodec/error_resilience.c View File

@@ -715,11 +715,11 @@ FF_ENABLE_DEPRECATION_WARNINGS
} else {
ff_thread_await_progress(s->last_pic.tf, mb_y, 0);
}
is_intra_likely += s->dsp->sad[0](NULL, last_mb_ptr, mb_ptr,
linesize[0], 16);
is_intra_likely -= s->dsp->sad[0](NULL, last_mb_ptr,
last_mb_ptr + linesize[0] * 16,
linesize[0], 16);
is_intra_likely += s->mecc->sad[0](NULL, last_mb_ptr, mb_ptr,
linesize[0], 16);
is_intra_likely -= s->mecc->sad[0](NULL, last_mb_ptr,
last_mb_ptr + linesize[0] * 16,
linesize[0], 16);
} else {
if (IS_INTRA(s->cur_pic.mb_type[mb_xy]))
is_intra_likely++;


+ 2
- 2
libavcodec/error_resilience.h View File

@@ -23,7 +23,7 @@
#include <stdint.h>

#include "avcodec.h"
#include "dsputil.h"
#include "me_cmp.h"
#include "thread.h"

///< current MB is the first after a resync marker
@@ -52,7 +52,7 @@ typedef struct ERPicture {

typedef struct ERContext {
AVCodecContext *avctx;
DSPContext *dsp;
MECmpContext *mecc;

int *mb_index2xy;
int mb_num;


+ 4
- 4
libavcodec/h264.c View File

@@ -33,7 +33,6 @@
#include "internal.h"
#include "cabac.h"
#include "cabac_functions.h"
#include "dsputil.h"
#include "error_resilience.h"
#include "avcodec.h"
#include "h264.h"
@@ -42,6 +41,7 @@
#include "h264_mvpred.h"
#include "golomb.h"
#include "mathops.h"
#include "me_cmp.h"
#include "mpegutils.h"
#include "rectangle.h"
#include "svq3.h"
@@ -490,7 +490,7 @@ int ff_h264_context_init(H264Context *h)
if (CONFIG_ERROR_RESILIENCE) {
/* init ER */
er->avctx = h->avctx;
er->dsp = &h->dsp;
er->mecc = &h->mecc;
er->decode_mb = h264_er_decode_mb;
er->opaque = h;
er->quarter_sample = 1;
@@ -620,7 +620,7 @@ av_cold int ff_h264_decode_init(AVCodecContext *avctx)

/* needed so that IDCT permutation is known early */
if (CONFIG_ERROR_RESILIENCE)
ff_dsputil_init(&h->dsp, h->avctx);
ff_me_cmp_init(&h->mecc, h->avctx);
ff_videodsp_init(&h->vdsp, 8);

memset(h->pps.scaling_matrix4, 16, 6 * 16 * sizeof(uint8_t));
@@ -1234,7 +1234,7 @@ int ff_h264_set_parameter_from_sps(H264Context *h)
ff_h264_pred_init(&h->hpc, h->avctx->codec_id, h->sps.bit_depth_luma,
h->sps.chroma_format_idc);
if (CONFIG_ERROR_RESILIENCE)
ff_dsputil_init(&h->dsp, h->avctx);
ff_me_cmp_init(&h->mecc, h->avctx);
ff_videodsp_init(&h->vdsp, h->sps.bit_depth_luma);
} else {
av_log(h->avctx, AV_LOG_ERROR, "Unsupported bit depth %d\n",


+ 2
- 2
libavcodec/h264.h View File

@@ -30,13 +30,13 @@

#include "libavutil/intreadwrite.h"
#include "cabac.h"
#include "dsputil.h"
#include "error_resilience.h"
#include "get_bits.h"
#include "h264chroma.h"
#include "h264dsp.h"
#include "h264pred.h"
#include "h264qpel.h"
#include "me_cmp.h"
#include "mpegutils.h"
#include "parser.h"
#include "qpeldsp.h"
@@ -302,7 +302,7 @@ typedef struct H264Picture {
*/
typedef struct H264Context {
AVCodecContext *avctx;
DSPContext dsp;
MECmpContext mecc;
VideoDSPContext vdsp;
H264DSPContext h264dsp;
H264ChromaContext h264chroma;


+ 1
- 2
libavcodec/h264_slice.c View File

@@ -31,7 +31,6 @@
#include "internal.h"
#include "cabac.h"
#include "cabac_functions.h"
#include "dsputil.h"
#include "error_resilience.h"
#include "avcodec.h"
#include "h264.h"
@@ -1119,7 +1118,7 @@ static int h264_slice_header_init(H264Context *h, int reinit)
if (!c)
return AVERROR(ENOMEM);
c->avctx = h->avctx;
c->dsp = h->dsp;
c->mecc = h->mecc;
c->vdsp = h->vdsp;
c->h264dsp = h->h264dsp;
c->h264qpel = h->h264qpel;


libavcodec/dsputil.c → libavcodec/me_cmp.c View File

@@ -1,8 +1,4 @@
/*
* DSP utils
* Copyright (c) 2000, 2001 Fabrice Bellard
* Copyright (c) 2002-2004 Michael Niedermayer <michaelni@gmx.at>
*
* This file is part of Libav.
*
* Libav is free software; you can redistribute it and/or
@@ -20,16 +16,11 @@
* Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
*/

/**
* @file
* DSP utils
*/

#include "libavutil/attributes.h"
#include "avcodec.h"
#include "copy_block.h"
#include "dsputil.h"
#include "simple_idct.h"
#include "me_cmp.h"
#include "mpegvideo.h"
#include "config.h"

@@ -369,7 +360,7 @@ static int zero_cmp(MpegEncContext *s, uint8_t *a, uint8_t *b,
return 0;
}

void ff_set_cmp(DSPContext *c, me_cmp_func *cmp, int type)
void ff_set_cmp(MECmpContext *c, me_cmp_func *cmp, int type)
{
int i;

@@ -551,7 +542,7 @@ static int dct_sad8x8_c(MpegEncContext *s, uint8_t *src1,

s->pdsp.diff_pixels(temp, src1, src2, stride);
s->fdsp.fdct(temp);
return s->dsp.sum_abs_dctelem(temp);
return s->mecc.sum_abs_dctelem(temp);
}

#if CONFIG_GPL
@@ -722,7 +713,7 @@ static int rd8x8_c(MpegEncContext *s, uint8_t *src1, uint8_t *src2,

s->idsp.idct_add(lsrc2, 8, temp);

distortion = s->dsp.sse[1](NULL, lsrc2, lsrc1, 8, 8);
distortion = s->mecc.sse[1](NULL, lsrc2, lsrc1, 8, 8);

return distortion + ((bits * s->qscale * s->qscale * 109 + 64) >> 7);
}
@@ -891,8 +882,7 @@ WRAPPER8_16_SQ(quant_psnr8x8_c, quant_psnr16_c)
WRAPPER8_16_SQ(rd8x8_c, rd16_c)
WRAPPER8_16_SQ(bit8x8_c, bit16_c)

/* init static data */
av_cold void ff_dsputil_static_init(void)
av_cold void ff_me_cmp_init_static(void)
{
int i;

@@ -900,7 +890,7 @@ av_cold void ff_dsputil_static_init(void)
ff_square_tab[i] = (i - 256) * (i - 256);
}

av_cold void ff_dsputil_init(DSPContext *c, AVCodecContext *avctx)
av_cold void ff_me_cmp_init(MECmpContext *c, AVCodecContext *avctx)
{
c->sum_abs_dctelem = sum_abs_dctelem_c;

@@ -944,9 +934,9 @@ av_cold void ff_dsputil_init(DSPContext *c, AVCodecContext *avctx)
c->nsse[1] = nsse8_c;

if (ARCH_ARM)
ff_dsputil_init_arm(c, avctx);
ff_me_cmp_init_arm(c, avctx);
if (ARCH_PPC)
ff_dsputil_init_ppc(c, avctx);
ff_me_cmp_init_ppc(c, avctx);
if (ARCH_X86)
ff_dsputil_init_x86(c, avctx);
ff_me_cmp_init_x86(c, avctx);
}

libavcodec/dsputil.h → libavcodec/me_cmp.h View File

@@ -1,8 +1,4 @@
/*
* DSP utils
* Copyright (c) 2000, 2001, 2002 Fabrice Bellard
* Copyright (c) 2002-2004 Michael Niedermayer <michaelni@gmx.at>
*
* This file is part of Libav.
*
* Libav is free software; you can redistribute it and/or
@@ -20,15 +16,10 @@
* Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
*/

/**
* @file
* DSP utils.
* Note, many functions in here may use MMX which trashes the FPU state, it is
* absolutely necessary to call emms_c() between DSP & float/double code.
*/
#ifndef AVCODEC_ME_CMP_H
#define AVCODEC_ME_CMP_H

#ifndef AVCODEC_DSPUTIL_H
#define AVCODEC_DSPUTIL_H
#include <stdint.h>

#include "avcodec.h"

@@ -44,10 +35,7 @@ typedef int (*me_cmp_func)(struct MpegEncContext *c,
uint8_t *blk1 /* align width (8 or 16) */,
uint8_t *blk2 /* align 1 */, int line_size, int h);

/**
* DSPContext.
*/
typedef struct DSPContext {
typedef struct MECmpContext {
int (*sum_abs_dctelem)(int16_t *block /* align 16 */);

me_cmp_func sad[6]; /* identical to pix_absAxA except additional void * */
@@ -71,15 +59,15 @@ typedef struct DSPContext {
me_cmp_func frame_skip_cmp[6]; // only width 8 used

me_cmp_func pix_abs[2][4];
} DSPContext;
} MECmpContext;

void ff_dsputil_static_init(void);
void ff_dsputil_init(DSPContext *p, AVCodecContext *avctx);
void ff_me_cmp_init_static(void);

void ff_set_cmp(DSPContext *c, me_cmp_func *cmp, int type);
void ff_me_cmp_init(MECmpContext *c, AVCodecContext *avctx);
void ff_me_cmp_init_arm(MECmpContext *c, AVCodecContext *avctx);
void ff_me_cmp_init_ppc(MECmpContext *c, AVCodecContext *avctx);
void ff_me_cmp_init_x86(MECmpContext *c, AVCodecContext *avctx);

void ff_dsputil_init_arm(DSPContext *c, AVCodecContext *avctx);
void ff_dsputil_init_ppc(DSPContext *c, AVCodecContext *avctx);
void ff_dsputil_init_x86(DSPContext *c, AVCodecContext *avctx);
void ff_set_cmp(MECmpContext *c, me_cmp_func *cmp, int type);

#endif /* AVCODEC_DSPUTIL_H */
#endif /* AVCODEC_ME_CMP_H */

+ 22
- 21
libavcodec/motion_est.c View File

@@ -317,10 +317,10 @@ int ff_init_me(MpegEncContext *s){
av_log(s->avctx, AV_LOG_INFO, "ME_MAP size may be a little small for the selected diamond size\n");
}

ff_set_cmp(&s->dsp, s->dsp.me_pre_cmp, c->avctx->me_pre_cmp);
ff_set_cmp(&s->dsp, s->dsp.me_cmp, c->avctx->me_cmp);
ff_set_cmp(&s->dsp, s->dsp.me_sub_cmp, c->avctx->me_sub_cmp);
ff_set_cmp(&s->dsp, s->dsp.mb_cmp, c->avctx->mb_cmp);
ff_set_cmp(&s->mecc, s->mecc.me_pre_cmp, c->avctx->me_pre_cmp);
ff_set_cmp(&s->mecc, s->mecc.me_cmp, c->avctx->me_cmp);
ff_set_cmp(&s->mecc, s->mecc.me_sub_cmp, c->avctx->me_sub_cmp);
ff_set_cmp(&s->mecc, s->mecc.mb_cmp, c->avctx->mb_cmp);

c->flags = get_flags(c, 0, c->avctx->me_cmp &FF_CMP_CHROMA);
c->sub_flags= get_flags(c, 0, c->avctx->me_sub_cmp&FF_CMP_CHROMA);
@@ -361,12 +361,10 @@ int ff_init_me(MpegEncContext *s){
/* 8x8 fullpel search would need a 4x4 chroma compare, which we do
* not have yet, and even if we had, the motion estimation code
* does not expect it. */
if((c->avctx->me_cmp&FF_CMP_CHROMA)/* && !s->dsp.me_cmp[2]*/){
s->dsp.me_cmp[2]= zero_cmp;
}
if((c->avctx->me_sub_cmp&FF_CMP_CHROMA) && !s->dsp.me_sub_cmp[2]){
s->dsp.me_sub_cmp[2]= zero_cmp;
}
if ((c->avctx->me_cmp & FF_CMP_CHROMA) /* && !s->mecc.me_cmp[2] */)
s->mecc.me_cmp[2] = zero_cmp;
if ((c->avctx->me_sub_cmp & FF_CMP_CHROMA) && !s->mecc.me_sub_cmp[2])
s->mecc.me_sub_cmp[2] = zero_cmp;
c->hpel_put[2][0]= c->hpel_put[2][1]=
c->hpel_put[2][2]= c->hpel_put[2][3]= zero_hpel;

@@ -379,7 +377,7 @@ int ff_init_me(MpegEncContext *s){

#define CHECK_SAD_HALF_MV(suffix, x, y) \
{\
d= s->dsp.pix_abs[size][(x?1:0)+(y?2:0)](NULL, pix, ptr+((x)>>1), stride, h);\
d = s->mecc.pix_abs[size][(x ? 1 : 0) + (y ? 2 : 0)](NULL, pix, ptr + ((x) >> 1), stride, h); \
d += (mv_penalty[pen_x + x] + mv_penalty[pen_y + y])*penalty_factor;\
COPY3_IF_LT(dminh, d, dx, x, dy, y)\
}
@@ -615,7 +613,7 @@ static inline int h263_mv4_search(MpegEncContext *s, int mx, int my, int shift)

dmin4= c->sub_motion_search(s, &mx4, &my4, dmin4, block, block, size, h);

if(s->dsp.me_sub_cmp[0] != s->dsp.mb_cmp[0]){
if (s->mecc.me_sub_cmp[0] != s->mecc.mb_cmp[0]) {
int dxy;
const int offset= ((block&1) + (block>>1)*stride)*8;
uint8_t *dest_y = c->scratchpad + offset;
@@ -657,8 +655,11 @@ static inline int h263_mv4_search(MpegEncContext *s, int mx, int my, int shift)
if(same)
return INT_MAX;

if(s->dsp.me_sub_cmp[0] != s->dsp.mb_cmp[0]){
dmin_sum += s->dsp.mb_cmp[0](s, s->new_picture.f->data[0] + s->mb_x*16 + s->mb_y*16*stride, c->scratchpad, stride, 16);
if (s->mecc.me_sub_cmp[0] != s->mecc.mb_cmp[0]) {
dmin_sum += s->mecc.mb_cmp[0](s,
s->new_picture.f->data[0] +
s->mb_x * 16 + s->mb_y * 16 * stride,
c->scratchpad, stride, 16);
}

if(c->avctx->mb_cmp&FF_CMP_CHROMA){
@@ -680,8 +681,8 @@ static inline int h263_mv4_search(MpegEncContext *s, int mx, int my, int shift)
s->hdsp.put_pixels_tab [1][dxy](c->scratchpad + 8, s->last_picture.f->data[2] + offset, s->uvlinesize, 8);
}

dmin_sum += s->dsp.mb_cmp[1](s, s->new_picture.f->data[1] + s->mb_x*8 + s->mb_y*8*s->uvlinesize, c->scratchpad , s->uvlinesize, 8);
dmin_sum += s->dsp.mb_cmp[1](s, s->new_picture.f->data[2] + s->mb_x*8 + s->mb_y*8*s->uvlinesize, c->scratchpad+8, s->uvlinesize, 8);
dmin_sum += s->mecc.mb_cmp[1](s, s->new_picture.f->data[1] + s->mb_x * 8 + s->mb_y * 8 * s->uvlinesize, c->scratchpad, s->uvlinesize, 8);
dmin_sum += s->mecc.mb_cmp[1](s, s->new_picture.f->data[2] + s->mb_x * 8 + s->mb_y * 8 * s->uvlinesize, c->scratchpad + 8, s->uvlinesize, 8);
}

c->pred_x= mx;
@@ -777,7 +778,7 @@ static int interlaced_search(MpegEncContext *s, int ref_index,
mv_table[xy][0]= mx_i;
mv_table[xy][1]= my_i;

if(s->dsp.me_sub_cmp[0] != s->dsp.mb_cmp[0]){
if (s->mecc.me_sub_cmp[0] != s->mecc.mb_cmp[0]) {
int dxy;

//FIXME chroma ME
@@ -789,7 +790,7 @@ static int interlaced_search(MpegEncContext *s, int ref_index,
}else{
s->hdsp.put_pixels_tab [size][dxy](c->scratchpad, ref , stride, h);
}
dmin= s->dsp.mb_cmp[size](s, c->src[block][0], c->scratchpad, stride, h);
dmin = s->mecc.mb_cmp[size](s, c->src[block][0], c->scratchpad, stride, h);
dmin+= (mv_penalty[mx_i-c->pred_x] + mv_penalty[my_i-c->pred_y] + 1)*c->mb_penalty_factor;
}else
dmin+= c->mb_penalty_factor; //field_select bits
@@ -940,7 +941,7 @@ void ff_estimate_p_frame_motion(MpegEncContext * s,
/* At this point (mx,my) are full-pell and the relative displacement */
ppix = c->ref[0][0] + (my * s->linesize) + mx;

vard = s->dsp.sse[0](NULL, pix, ppix, s->linesize, 16);
vard = s->mecc.sse[0](NULL, pix, ppix, s->linesize, 16);

pic->mc_mb_var[s->mb_stride * mb_y + mb_x] = (vard+128)>>8;
c->mc_mb_var_sum_temp += (vard+128)>>8;
@@ -1037,7 +1038,7 @@ void ff_estimate_p_frame_motion(MpegEncContext * s,
*(uint32_t*)(&c->scratchpad[i*s->linesize+12]) = mean;
}

intra_score= s->dsp.mb_cmp[0](s, c->scratchpad, pix, s->linesize, 16);
intra_score= s->mecc.mb_cmp[0](s, c->scratchpad, pix, s->linesize, 16);
}
intra_score += c->mb_penalty_factor*16;

@@ -1237,7 +1238,7 @@ static inline int check_bidir_mv(MpegEncContext * s,

fbmin = (mv_penalty_f[motion_fx-pred_fx] + mv_penalty_f[motion_fy-pred_fy])*c->mb_penalty_factor
+(mv_penalty_b[motion_bx-pred_bx] + mv_penalty_b[motion_by-pred_by])*c->mb_penalty_factor
+ s->dsp.mb_cmp[size](s, src_data[0], dest_y, stride, h); //FIXME new_pic
+ s->mecc.mb_cmp[size](s, src_data[0], dest_y, stride, h); // FIXME new_pic

if(c->avctx->mb_cmp&FF_CMP_CHROMA){
}


+ 32
- 33
libavcodec/motion_est_template.c View File

@@ -63,8 +63,8 @@ static int hpel_motion_search(MpegEncContext * s,

//FIXME factorize

cmp_sub= s->dsp.me_sub_cmp[size];
chroma_cmp_sub= s->dsp.me_sub_cmp[size+1];
cmp_sub = s->mecc.me_sub_cmp[size];
chroma_cmp_sub = s->mecc.me_sub_cmp[size + 1];

if(c->skip){ //FIXME move out of hpel?
*mx_ptr = 0;
@@ -166,7 +166,6 @@ static inline int get_mb_score(MpegEncContext *s, int mx, int my,
int src_index, int ref_index, int size,
int h, int add_rate)
{
// const int check_luma= s->dsp.me_sub_cmp != s->dsp.mb_cmp;
MotionEstContext * const c= &s->me;
const int penalty_factor= c->mb_penalty_factor;
const int flags= c->mb_flags;
@@ -179,8 +178,8 @@ static inline int get_mb_score(MpegEncContext *s, int mx, int my,

//FIXME factorize

cmp_sub= s->dsp.mb_cmp[size];
chroma_cmp_sub= s->dsp.mb_cmp[size+1];
cmp_sub = s->mecc.mb_cmp[size];
chroma_cmp_sub = s->mecc.mb_cmp[size + 1];

// assert(!c->skip);
// assert(c->avctx->me_sub_cmp != c->avctx->mb_cmp);
@@ -226,12 +225,12 @@ static int qpel_motion_search(MpegEncContext * s,
LOAD_COMMON
int flags= c->sub_flags;

cmpf= s->dsp.me_cmp[size];
chroma_cmpf= s->dsp.me_cmp[size+1]; //factorize FIXME
cmpf = s->mecc.me_cmp[size];
chroma_cmpf = s->mecc.me_cmp[size + 1]; // FIXME: factorize
//FIXME factorize

cmp_sub= s->dsp.me_sub_cmp[size];
chroma_cmp_sub= s->dsp.me_sub_cmp[size+1];
cmp_sub = s->mecc.me_sub_cmp[size];
chroma_cmp_sub = s->mecc.me_sub_cmp[size + 1];

if(c->skip){ //FIXME somehow move up (benchmark)
*mx_ptr = 0;
@@ -427,8 +426,8 @@ static av_always_inline int small_diamond_search(MpegEncContext * s, int *best,
LOAD_COMMON2
unsigned map_generation = c->map_generation;

cmpf= s->dsp.me_cmp[size];
chroma_cmpf= s->dsp.me_cmp[size+1];
cmpf = s->mecc.me_cmp[size];
chroma_cmpf = s->mecc.me_cmp[size + 1];

{ /* ensure that the best point is in the MAP as h/qpel refinement needs it */
const unsigned key = (best[1]<<ME_MAP_MV_BITS) + best[0] + map_generation;
@@ -468,8 +467,8 @@ static int funny_diamond_search(MpegEncContext * s, int *best, int dmin,
LOAD_COMMON2
unsigned map_generation = c->map_generation;

cmpf= s->dsp.me_cmp[size];
chroma_cmpf= s->dsp.me_cmp[size+1];
cmpf = s->mecc.me_cmp[size];
chroma_cmpf = s->mecc.me_cmp[size + 1];

for(dia_size=1; dia_size<=4; dia_size++){
int dir;
@@ -511,8 +510,8 @@ static int hex_search(MpegEncContext * s, int *best, int dmin,
int x,y,d;
const int dec= dia_size & (dia_size-1);

cmpf= s->dsp.me_cmp[size];
chroma_cmpf= s->dsp.me_cmp[size+1];
cmpf = s->mecc.me_cmp[size];
chroma_cmpf = s->mecc.me_cmp[size + 1];

for(;dia_size; dia_size= dec ? dia_size-1 : dia_size>>1){
do{
@@ -548,8 +547,8 @@ static int l2s_dia_search(MpegEncContext * s, int *best, int dmin,
static const int hex[8][2]={{-2, 0}, {-1,-1}, { 0,-2}, { 1,-1},
{ 2, 0}, { 1, 1}, { 0, 2}, {-1, 1}};

cmpf= s->dsp.me_cmp[size];
chroma_cmpf= s->dsp.me_cmp[size+1];
cmpf = s->mecc.me_cmp[size];
chroma_cmpf = s->mecc.me_cmp[size + 1];

for(; dia_size; dia_size= dec ? dia_size-1 : dia_size>>1){
do{
@@ -587,8 +586,8 @@ static int umh_search(MpegEncContext * s, int *best, int dmin,
{-2, 3}, { 0, 4}, { 2, 3},
{-2,-3}, { 0,-4}, { 2,-3},};

cmpf= s->dsp.me_cmp[size];
chroma_cmpf= s->dsp.me_cmp[size+1];
cmpf = s->mecc.me_cmp[size];
chroma_cmpf = s->mecc.me_cmp[size + 1];

x= best[0];
y= best[1];
@@ -630,8 +629,8 @@ static int full_search(MpegEncContext * s, int *best, int dmin,
int x,y, d;
const int dia_size= c->dia_size&0xFF;

cmpf= s->dsp.me_cmp[size];
chroma_cmpf= s->dsp.me_cmp[size+1];
cmpf = s->mecc.me_cmp[size];
chroma_cmpf = s->mecc.me_cmp[size + 1];

for(y=FFMAX(-dia_size, ymin); y<=FFMIN(dia_size,ymax); y++){
for(x=FFMAX(-dia_size, xmin); x<=FFMIN(dia_size,xmax); x++){
@@ -694,8 +693,8 @@ static int sab_diamond_search(MpegEncContext * s, int *best, int dmin,
LOAD_COMMON2
unsigned map_generation = c->map_generation;

cmpf= s->dsp.me_cmp[size];
chroma_cmpf= s->dsp.me_cmp[size+1];
cmpf = s->mecc.me_cmp[size];
chroma_cmpf = s->mecc.me_cmp[size + 1];

/*Note j<MAX_SAB_SIZE is needed if MAX_SAB_SIZE < ME_MAP_SIZE as j can
become larger due to MVs overflowing their ME_MAP_MV_BITS bits space in map
@@ -779,8 +778,8 @@ static int var_diamond_search(MpegEncContext * s, int *best, int dmin,
LOAD_COMMON2
unsigned map_generation = c->map_generation;

cmpf= s->dsp.me_cmp[size];
chroma_cmpf= s->dsp.me_cmp[size+1];
cmpf = s->mecc.me_cmp[size];
chroma_cmpf = s->mecc.me_cmp[size + 1];

for(dia_size=1; dia_size<=c->dia_size; dia_size++){
int dir, start, end;
@@ -880,12 +879,12 @@ static av_always_inline int epzs_motion_search_internal(MpegEncContext * s, int

if(c->pre_pass){
penalty_factor= c->pre_penalty_factor;
cmpf= s->dsp.me_pre_cmp[size];
chroma_cmpf= s->dsp.me_pre_cmp[size+1];
cmpf = s->mecc.me_pre_cmp[size];
chroma_cmpf = s->mecc.me_pre_cmp[size + 1];
}else{
penalty_factor= c->penalty_factor;
cmpf= s->dsp.me_cmp[size];
chroma_cmpf= s->dsp.me_cmp[size+1];
cmpf = s->mecc.me_cmp[size];
chroma_cmpf = s->mecc.me_cmp[size + 1];
}

map_generation= update_map_generation(c);
@@ -1009,8 +1008,8 @@ static int epzs_motion_search4(MpegEncContext * s,
int flags= c->flags;
LOAD_COMMON2

cmpf= s->dsp.me_cmp[size];
chroma_cmpf= s->dsp.me_cmp[size+1];
cmpf = s->mecc.me_cmp[size];
chroma_cmpf = s->mecc.me_cmp[size + 1];

map_generation= update_map_generation(c);

@@ -1068,8 +1067,8 @@ static int epzs_motion_search2(MpegEncContext * s,
int flags= c->flags;
LOAD_COMMON2

cmpf= s->dsp.me_cmp[size];
chroma_cmpf= s->dsp.me_cmp[size+1];
cmpf = s->mecc.me_cmp[size];
chroma_cmpf = s->mecc.me_cmp[size + 1];

map_generation= update_map_generation(c);



+ 1
- 1
libavcodec/mpeg4videoenc.c View File

@@ -689,7 +689,7 @@ void ff_mpeg4_encode_mb(MpegEncContext *s, int16_t block[6][64],
b_pic = pic->f->data[0] + offset;
if (!pic->shared)
b_pic += INPLACE_OFFSET;
diff = s->dsp.sad[0](NULL, p_pic, b_pic, s->linesize, 16);
diff = s->mecc.sad[0](NULL, p_pic, b_pic, s->linesize, 16);
if (diff > s->qscale * 70) { // FIXME check that 70 is optimal
s->mb_skipped = 0;
break;


+ 2
- 2
libavcodec/mpegvideo.c View File

@@ -378,9 +378,9 @@ static void mpeg_er_decode_mb(void *opaque, int ref, int mv_dir, int mv_type,
av_cold int ff_dct_common_init(MpegEncContext *s)
{
ff_blockdsp_init(&s->bdsp, s->avctx);
ff_dsputil_init(&s->dsp, s->avctx);
ff_hpeldsp_init(&s->hdsp, s->avctx->flags);
ff_idctdsp_init(&s->idsp, s->avctx);
ff_me_cmp_init(&s->mecc, s->avctx);
ff_mpegvideodsp_init(&s->mdsp);
ff_videodsp_init(&s->vdsp, s->avctx->bits_per_raw_sample);

@@ -1051,7 +1051,7 @@ static int init_er(MpegEncContext *s)
int i;

er->avctx = s->avctx;
er->dsp = &s->dsp;
er->mecc = &s->mecc;

er->mb_index2xy = s->mb_index2xy;
er->mb_num = s->mb_num;


+ 2
- 2
libavcodec/mpegvideo.h View File

@@ -30,13 +30,13 @@

#include "avcodec.h"
#include "blockdsp.h"
#include "dsputil.h"
#include "error_resilience.h"
#include "fdctdsp.h"
#include "get_bits.h"
#include "h263dsp.h"
#include "hpeldsp.h"
#include "idctdsp.h"
#include "me_cmp.h"
#include "mpegvideodsp.h"
#include "mpegvideoencdsp.h"
#include "pixblockdsp.h"
@@ -356,10 +356,10 @@ typedef struct MpegEncContext {
int h263_long_vectors; ///< use horrible h263v1 long vector mode

BlockDSPContext bdsp;
DSPContext dsp; ///< pointers for accelerated dsp functions
FDCTDSPContext fdsp;
HpelDSPContext hdsp;
IDCTDSPContext idsp;
MECmpContext mecc;
MpegVideoDSPContext mdsp;
MpegvideoEncDSPContext mpvencdsp;
PixblockDSPContext pdsp;


+ 44
- 52
libavcodec/mpegvideo_enc.c View File

@@ -702,6 +702,7 @@ av_cold int ff_MPV_encode_init(AVCodecContext *avctx)
ff_MPV_encode_init_x86(s);

ff_fdctdsp_init(&s->fdsp, avctx);
ff_me_cmp_init(&s->mecc, avctx);
ff_mpegvideoencdsp_init(&s->mpvencdsp, avctx);
ff_pixblockdsp_init(&s->pdsp, avctx);
ff_qpeldsp_init(&s->qdsp);
@@ -744,8 +745,8 @@ av_cold int ff_MPV_encode_init(AVCodecContext *avctx)

s->quant_precision = 5;

ff_set_cmp(&s->dsp, s->dsp.ildct_cmp, s->avctx->ildct_cmp);
ff_set_cmp(&s->dsp, s->dsp.frame_skip_cmp, s->avctx->frame_skip_cmp);
ff_set_cmp(&s->mecc, s->mecc.ildct_cmp, s->avctx->ildct_cmp);
ff_set_cmp(&s->mecc, s->mecc.frame_skip_cmp, s->avctx->frame_skip_cmp);

if (CONFIG_H261_ENCODER && s->out_format == FMT_H261)
ff_h261_encode_init(s);
@@ -895,8 +896,8 @@ static int get_intra_count(MpegEncContext *s, uint8_t *src,
for (y = 0; y < h; y += 16) {
for (x = 0; x < w; x += 16) {
int offset = x + y * stride;
int sad = s->dsp.sad[0](NULL, src + offset, ref + offset, stride,
16);
int sad = s->mecc.sad[0](NULL, src + offset, ref + offset,
stride, 16);
int mean = (s->mpvencdsp.pix_sum(src + offset, stride) + 128) >> 8;
int sae = get_sae(src + offset, mean, stride);

@@ -1053,7 +1054,7 @@ static int skip_check(MpegEncContext *s, Picture *p, Picture *ref)
int off = p->shared ? 0 : 16;
uint8_t *dptr = p->f->data[plane] + 8 * (x + y * stride) + off;
uint8_t *rptr = ref->f->data[plane] + 8 * (x + y * stride);
int v = s->dsp.frame_skip_cmp[1](s, dptr, rptr, stride, 8);
int v = s->mecc.frame_skip_cmp[1](s, dptr, rptr, stride, 8);

switch (s->avctx->frame_skip_exp) {
case 0: score = FFMAX(score, v); break;
@@ -1923,16 +1924,15 @@ static av_always_inline void encode_mb_internal(MpegEncContext *s,
int progressive_score, interlaced_score;

s->interlaced_dct = 0;
progressive_score = s->dsp.ildct_cmp[4](s, ptr_y,
NULL, wrap_y, 8) +
s->dsp.ildct_cmp[4](s, ptr_y + wrap_y * 8,
NULL, wrap_y, 8) - 400;
progressive_score = s->mecc.ildct_cmp[4](s, ptr_y, NULL, wrap_y, 8) +
s->mecc.ildct_cmp[4](s, ptr_y + wrap_y * 8,
NULL, wrap_y, 8) - 400;

if (progressive_score > 0) {
interlaced_score = s->dsp.ildct_cmp[4](s, ptr_y,
NULL, wrap_y * 2, 8) +
s->dsp.ildct_cmp[4](s, ptr_y + wrap_y,
NULL, wrap_y * 2, 8);
interlaced_score = s->mecc.ildct_cmp[4](s, ptr_y,
NULL, wrap_y * 2, 8) +
s->mecc.ildct_cmp[4](s, ptr_y + wrap_y,
NULL, wrap_y * 2, 8);
if (progressive_score > interlaced_score) {
s->interlaced_dct = 1;

@@ -1996,23 +1996,20 @@ static av_always_inline void encode_mb_internal(MpegEncContext *s,
int progressive_score, interlaced_score;

s->interlaced_dct = 0;
progressive_score = s->dsp.ildct_cmp[0](s, dest_y,
ptr_y, wrap_y,
8) +
s->dsp.ildct_cmp[0](s, dest_y + wrap_y * 8,
ptr_y + wrap_y * 8, wrap_y,
8) - 400;
progressive_score = s->mecc.ildct_cmp[0](s, dest_y, ptr_y, wrap_y, 8) +
s->mecc.ildct_cmp[0](s, dest_y + wrap_y * 8,
ptr_y + wrap_y * 8,
wrap_y, 8) - 400;

if (s->avctx->ildct_cmp == FF_CMP_VSSE)
progressive_score -= 400;

if (progressive_score > 0) {
interlaced_score = s->dsp.ildct_cmp[0](s, dest_y,
ptr_y,
wrap_y * 2, 8) +
s->dsp.ildct_cmp[0](s, dest_y + wrap_y,
ptr_y + wrap_y,
wrap_y * 2, 8);
interlaced_score = s->mecc.ildct_cmp[0](s, dest_y, ptr_y,
wrap_y * 2, 8) +
s->mecc.ildct_cmp[0](s, dest_y + wrap_y,
ptr_y + wrap_y,
wrap_y * 2, 8);

if (progressive_score > interlaced_score) {
s->interlaced_dct = 1;
@@ -2049,33 +2046,28 @@ static av_always_inline void encode_mb_internal(MpegEncContext *s,
if (s->current_picture.mc_mb_var[s->mb_stride * mb_y + mb_x] <
2 * s->qscale * s->qscale) {
// FIXME optimize
if (s->dsp.sad[1](NULL, ptr_y , dest_y,
wrap_y, 8) < 20 * s->qscale)
if (s->mecc.sad[1](NULL, ptr_y, dest_y, wrap_y, 8) < 20 * s->qscale)
skip_dct[0] = 1;
if (s->dsp.sad[1](NULL, ptr_y + 8,
dest_y + 8, wrap_y, 8) < 20 * s->qscale)
if (s->mecc.sad[1](NULL, ptr_y + 8, dest_y + 8, wrap_y, 8) < 20 * s->qscale)
skip_dct[1] = 1;
if (s->dsp.sad[1](NULL, ptr_y + dct_offset,
dest_y + dct_offset, wrap_y, 8) < 20 * s->qscale)
if (s->mecc.sad[1](NULL, ptr_y + dct_offset, dest_y + dct_offset,
wrap_y, 8) < 20 * s->qscale)
skip_dct[2] = 1;
if (s->dsp.sad[1](NULL, ptr_y + dct_offset + 8,
dest_y + dct_offset + 8,
wrap_y, 8) < 20 * s->qscale)
if (s->mecc.sad[1](NULL, ptr_y + dct_offset + 8, dest_y + dct_offset + 8,
wrap_y, 8) < 20 * s->qscale)
skip_dct[3] = 1;
if (s->dsp.sad[1](NULL, ptr_cb, dest_cb,
wrap_c, 8) < 20 * s->qscale)
if (s->mecc.sad[1](NULL, ptr_cb, dest_cb, wrap_c, 8) < 20 * s->qscale)
skip_dct[4] = 1;
if (s->dsp.sad[1](NULL, ptr_cr, dest_cr,
wrap_c, 8) < 20 * s->qscale)
if (s->mecc.sad[1](NULL, ptr_cr, dest_cr, wrap_c, 8) < 20 * s->qscale)
skip_dct[5] = 1;
if (!s->chroma_y_shift) { /* 422 */
if (s->dsp.sad[1](NULL, ptr_cb + (dct_offset >> 1),
dest_cb + (dct_offset >> 1),
wrap_c, 8) < 20 * s->qscale)
if (s->mecc.sad[1](NULL, ptr_cb + (dct_offset >> 1),
dest_cb + (dct_offset >> 1),
wrap_c, 8) < 20 * s->qscale)
skip_dct[6] = 1;
if (s->dsp.sad[1](NULL, ptr_cr + (dct_offset >> 1),
dest_cr + (dct_offset >> 1),
wrap_c, 8) < 20 * s->qscale)
if (s->mecc.sad[1](NULL, ptr_cr + (dct_offset >> 1),
dest_cr + (dct_offset >> 1),
wrap_c, 8) < 20 * s->qscale)
skip_dct[7] = 1;
}
}
@@ -2340,9 +2332,9 @@ static int sse(MpegEncContext *s, uint8_t *src1, uint8_t *src2, int w, int h, in
int x,y;

if(w==16 && h==16)
return s->dsp.sse[0](NULL, src1, src2, stride, 16);
return s->mecc.sse[0](NULL, src1, src2, stride, 16);
else if(w==8 && h==8)
return s->dsp.sse[1](NULL, src1, src2, stride, 8);
return s->mecc.sse[1](NULL, src1, src2, stride, 8);

for(y=0; y<h; y++){
for(x=0; x<w; x++){
@@ -2364,13 +2356,13 @@ static int sse_mb(MpegEncContext *s){

if(w==16 && h==16)
if(s->avctx->mb_cmp == FF_CMP_NSSE){
return s->dsp.nsse[0](s, s->new_picture.f->data[0] + s->mb_x*16 + s->mb_y*s->linesize*16, s->dest[0], s->linesize, 16)
+s->dsp.nsse[1](s, s->new_picture.f->data[1] + s->mb_x*8 + s->mb_y*s->uvlinesize*8,s->dest[1], s->uvlinesize, 8)
+s->dsp.nsse[1](s, s->new_picture.f->data[2] + s->mb_x*8 + s->mb_y*s->uvlinesize*8,s->dest[2], s->uvlinesize, 8);
return s->mecc.nsse[0](s, s->new_picture.f->data[0] + s->mb_x * 16 + s->mb_y * s->linesize * 16, s->dest[0], s->linesize, 16) +
s->mecc.nsse[1](s, s->new_picture.f->data[1] + s->mb_x * 8 + s->mb_y * s->uvlinesize * 8, s->dest[1], s->uvlinesize, 8) +
s->mecc.nsse[1](s, s->new_picture.f->data[2] + s->mb_x * 8 + s->mb_y * s->uvlinesize * 8, s->dest[2], s->uvlinesize, 8);
}else{
return s->dsp.sse[0](NULL, s->new_picture.f->data[0] + s->mb_x*16 + s->mb_y*s->linesize*16, s->dest[0], s->linesize, 16)
+s->dsp.sse[1](NULL, s->new_picture.f->data[1] + s->mb_x*8 + s->mb_y*s->uvlinesize*8,s->dest[1], s->uvlinesize, 8)
+s->dsp.sse[1](NULL, s->new_picture.f->data[2] + s->mb_x*8 + s->mb_y*s->uvlinesize*8,s->dest[2], s->uvlinesize, 8);
return s->mecc.sse[0](NULL, s->new_picture.f->data[0] + s->mb_x * 16 + s->mb_y * s->linesize * 16, s->dest[0], s->linesize, 16) +
s->mecc.sse[1](NULL, s->new_picture.f->data[1] + s->mb_x * 8 + s->mb_y * s->uvlinesize * 8, s->dest[1], s->uvlinesize, 8) +
s->mecc.sse[1](NULL, s->new_picture.f->data[2] + s->mb_x * 8 + s->mb_y * s->uvlinesize * 8, s->dest[2], s->uvlinesize, 8);
}
else
return sse(s, s->new_picture.f->data[0] + s->mb_x*16 + s->mb_y*s->linesize*16, s->dest[0], w, h, s->linesize)


+ 1
- 1
libavcodec/mpegvideoencdsp.c View File

@@ -24,8 +24,8 @@
#include "libavutil/attributes.h"
#include "libavutil/imgutils.h"
#include "avcodec.h"
#include "dsputil.h"
#include "imgconvert.h"
#include "me_cmp.h"
#include "mpegvideoencdsp.h"

static int try_8x8basis_c(int16_t rem[64], int16_t weight[64],


+ 1
- 1
libavcodec/ppc/Makefile View File

@@ -2,7 +2,6 @@ OBJS += ppc/fmtconvert_altivec.o \

OBJS-$(CONFIG_AUDIODSP) += ppc/audiodsp.o
OBJS-$(CONFIG_BLOCKDSP) += ppc/blockdsp.o
OBJS-$(CONFIG_DSPUTIL) += ppc/dsputil_altivec.o
OBJS-$(CONFIG_FFT) += ppc/fft_altivec.o
OBJS-$(CONFIG_H264CHROMA) += ppc/h264chroma_init.o
OBJS-$(CONFIG_H264DSP) += ppc/h264dsp.o
@@ -11,6 +10,7 @@ OBJS-$(CONFIG_HPELDSP) += ppc/hpeldsp_altivec.o
OBJS-$(CONFIG_HUFFYUVDSP) += ppc/huffyuvdsp_altivec.o
OBJS-$(CONFIG_FDCTDSP) += ppc/fdctdsp.o
OBJS-$(CONFIG_IDCTDSP) += ppc/idctdsp.o
OBJS-$(CONFIG_ME_CMP) += ppc/me_cmp.o
OBJS-$(CONFIG_MPEGAUDIODSP) += ppc/mpegaudiodsp_altivec.o
OBJS-$(CONFIG_MPEGVIDEO) += ppc/mpegvideo_altivec.o \
ppc/mpegvideodsp.o


libavcodec/ppc/dsputil_altivec.c → libavcodec/ppc/me_cmp.c View File

@@ -31,8 +31,8 @@
#include "libavutil/ppc/types_altivec.h"
#include "libavutil/ppc/util_altivec.h"
#include "libavcodec/avcodec.h"
#include "libavcodec/dsputil.h"
#include "libavcodec/mpegvideo.h"
#include "libavcodec/me_cmp.h"

#if HAVE_ALTIVEC
static int sad16_x2_altivec(MpegEncContext *v, uint8_t *pix1, uint8_t *pix2,
@@ -744,7 +744,7 @@ static int hadamard8_diff16_altivec(MpegEncContext *s, uint8_t *dst,
}
#endif /* HAVE_ALTIVEC */

av_cold void ff_dsputil_init_ppc(DSPContext *c, AVCodecContext *avctx)
av_cold void ff_me_cmp_init_ppc(MECmpContext *c, AVCodecContext *avctx)
{
#if HAVE_ALTIVEC
if (!PPC_ALTIVEC(av_get_cpu_flags()))

+ 5
- 5
libavcodec/svq1enc.c View File

@@ -27,8 +27,8 @@
*/

#include "avcodec.h"
#include "dsputil.h"
#include "hpeldsp.h"
#include "me_cmp.h"
#include "mpegvideo.h"
#include "h263.h"
#include "internal.h"
@@ -306,7 +306,7 @@ static int svq1_encode_plane(SVQ1EncContext *s, int plane,
s->m.current_picture.motion_val[0] = s->motion_val8[plane] + 2;
s->m.p_mv_table = s->motion_val16[plane] +
s->m.mb_stride + 1;
s->m.dsp = s->dsp; // move
s->m.mecc = s->mecc; // move
ff_init_me(&s->m);

s->m.me.dia_size = s->avctx->dia_size;
@@ -431,8 +431,8 @@ static int svq1_encode_plane(SVQ1EncContext *s, int plane,
best = score[1] <= score[0];

vlc = ff_svq1_block_type_vlc[SVQ1_BLOCK_SKIP];
score[2] = s->dsp.sse[0](NULL, src + 16 * x, ref,
stride, 16);
score[2] = s->mecc.sse[0](NULL, src + 16 * x, ref,
stride, 16);
score[2] += vlc[1] * lambda;
if (score[2] < score[best] && mx == 0 && my == 0) {
best = 2;
@@ -509,8 +509,8 @@ static av_cold int svq1_encode_init(AVCodecContext *avctx)
SVQ1EncContext *const s = avctx->priv_data;
int ret;

ff_dsputil_init(&s->dsp, avctx);
ff_hpeldsp_init(&s->hdsp, avctx->flags);
ff_me_cmp_init(&s->mecc, avctx);
ff_mpegvideoencdsp_init(&s->m.mpvencdsp, avctx);

avctx->coded_frame = av_frame_alloc();


+ 2
- 2
libavcodec/svq1enc.h View File

@@ -25,9 +25,9 @@

#include "libavutil/frame.h"
#include "avcodec.h"
#include "dsputil.h"
#include "get_bits.h"
#include "hpeldsp.h"
#include "me_cmp.h"
#include "mpegvideo.h"
#include "put_bits.h"

@@ -37,7 +37,7 @@ typedef struct SVQ1EncContext {
* of MpegEncContext, so this will be removed then. */
MpegEncContext m;
AVCodecContext *avctx;
DSPContext dsp;
MECmpContext mecc;
HpelDSPContext hdsp;
AVFrame *current_picture;
AVFrame *last_picture;


+ 3
- 3
libavcodec/utils.c View File

@@ -39,8 +39,8 @@
#include "libavutil/samplefmt.h"
#include "libavutil/dict.h"
#include "avcodec.h"
#include "dsputil.h"
#include "libavutil/opt.h"
#include "me_cmp.h"
#include "mpegvideo.h"
#include "thread.h"
#include "internal.h"
@@ -100,8 +100,8 @@ static av_cold void avcodec_init(void)
return;
initialized = 1;

if (CONFIG_DSPUTIL)
ff_dsputil_static_init();
if (CONFIG_ME_CMP)
ff_me_cmp_init_static();
}

int av_codec_is_encoder(const AVCodec *codec)


+ 2
- 2
libavcodec/x86/Makefile View File

@@ -6,7 +6,6 @@ OBJS-$(CONFIG_AUDIODSP) += x86/audiodsp_init.o
OBJS-$(CONFIG_BLOCKDSP) += x86/blockdsp.o
OBJS-$(CONFIG_BSWAPDSP) += x86/bswapdsp_init.o
OBJS-$(CONFIG_DCT) += x86/dct_init.o
OBJS-$(CONFIG_DSPUTIL) += x86/dsputil_init.o
OBJS-$(CONFIG_FDCTDSP) += x86/fdctdsp_init.o
OBJS-$(CONFIG_FFT) += x86/fft_init.o
OBJS-$(CONFIG_H263DSP) += x86/h263dsp_init.o
@@ -19,6 +18,7 @@ OBJS-$(CONFIG_HUFFYUVDSP) += x86/huffyuvdsp_init.o
OBJS-$(CONFIG_HUFFYUVENCDSP) += x86/huffyuvencdsp_mmx.o
OBJS-$(CONFIG_IDCTDSP) += x86/idctdsp_init.o
OBJS-$(CONFIG_LPC) += x86/lpc.o
OBJS-$(CONFIG_ME_CMP) += x86/me_cmp_init.o
OBJS-$(CONFIG_MPEGAUDIODSP) += x86/mpegaudiodsp.o
OBJS-$(CONFIG_MPEGVIDEO) += x86/mpegvideo.o \
x86/mpegvideodsp.o
@@ -70,7 +70,6 @@ YASM-OBJS-$(CONFIG_AUDIODSP) += x86/audiodsp.o
YASM-OBJS-$(CONFIG_BSWAPDSP) += x86/bswapdsp.o
YASM-OBJS-$(CONFIG_DCT) += x86/dct32.o
YASM-OBJS-$(CONFIG_DNXHD_ENCODER) += x86/dnxhdenc.o
YASM-OBJS-$(CONFIG_DSPUTIL) += x86/dsputilenc.o
YASM-OBJS-$(CONFIG_FFT) += x86/fft.o
YASM-OBJS-$(CONFIG_H263DSP) += x86/h263_loopfilter.o
YASM-OBJS-$(CONFIG_H264CHROMA) += x86/h264_chromamc.o \
@@ -90,6 +89,7 @@ YASM-OBJS-$(CONFIG_H264QPEL) += x86/h264_qpel_8bit.o \
YASM-OBJS-$(CONFIG_HPELDSP) += x86/fpel.o \
x86/hpeldsp.o
YASM-OBJS-$(CONFIG_HUFFYUVDSP) += x86/huffyuvdsp.o
YASM-OBJS-$(CONFIG_ME_CMP) += x86/me_cmp.o
YASM-OBJS-$(CONFIG_MPEGAUDIODSP) += x86/imdct36.o
YASM-OBJS-$(CONFIG_MPEGVIDEOENC) += x86/mpegvideoencdsp.o
YASM-OBJS-$(CONFIG_PIXBLOCKDSP) += x86/pixblockdsp.o


libavcodec/x86/dsputilenc.asm → libavcodec/x86/me_cmp.asm View File

@@ -1,5 +1,5 @@
;*****************************************************************************
;* MMX optimized DSP utils
;* SIMD-optimized motion compensation estimation
;*****************************************************************************
;* Copyright (c) 2000, 2001 Fabrice Bellard
;* Copyright (c) 2002-2004 Michael Niedermayer <michaelni@gmx.at>

libavcodec/x86/dsputil_init.c → libavcodec/x86/me_cmp_init.c View File

@@ -1,5 +1,5 @@
/*
* MMX optimized DSP utils
* SIMD-optimized motion estimation
* Copyright (c) 2000, 2001 Fabrice Bellard
* Copyright (c) 2002-2004 Michael Niedermayer <michaelni@gmx.at>
*
@@ -26,7 +26,7 @@
#include "libavutil/cpu.h"
#include "libavutil/x86/asm.h"
#include "libavutil/x86/cpu.h"
#include "libavcodec/dsputil.h"
#include "libavcodec/me_cmp.h"
#include "libavcodec/mpegvideo.h"

#if HAVE_INLINE_ASM
@@ -411,7 +411,7 @@ static int nsse16_mmx(MpegEncContext *c, uint8_t *pix1, uint8_t *pix2,
int score1, score2;

if (c)
score1 = c->dsp.sse[0](c, pix1, pix2, line_size, h);
score1 = c->mecc.sse[0](c, pix1, pix2, line_size, h);
else
score1 = sse16_mmx(c, pix1, pix2, line_size, h);
score2 = hf_noise16_mmx(pix1, line_size, h) -
@@ -1225,7 +1225,7 @@ hadamard_func(mmxext)
hadamard_func(sse2)
hadamard_func(ssse3)

av_cold void ff_dsputil_init_x86(DSPContext *c, AVCodecContext *avctx)
av_cold void ff_me_cmp_init_x86(MECmpContext *c, AVCodecContext *avctx)
{
int cpu_flags = av_get_cpu_flags();


Write Preview
Loading…
Cancel
Save