Updated FFMPEG to version 1.1.2, using this project: http://sourceforge.net/projects/ffmpeg4android/

This commit is contained in:
Sergii Pylypenko
2013-02-21 18:29:51 +02:00
parent 758a9658d2
commit fff7a99a41
3492 changed files with 886704 additions and 5414 deletions

View File

@@ -0,0 +1,3 @@
OBJS += sh4/dsputil_align.o \
sh4/dsputil_sh4.o \
sh4/idct_sh4.o \

View File

@@ -0,0 +1,439 @@
/*
* aligned/packed access motion
*
* Copyright (c) 2001-2003 BERO <bero@geocities.co.jp>
*
* This file is part of FFmpeg.
*
* FFmpeg is free software; you can redistribute it and/or
* modify it under the terms of the GNU Lesser General Public
* License as published by the Free Software Foundation; either
* version 2.1 of the License, or (at your option) any later version.
*
* FFmpeg 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
* Lesser General Public License for more details.
*
* You should have received a copy of the GNU Lesser General Public
* License along with FFmpeg; if not, write to the Free Software
* Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
*/
#include "libavcodec/avcodec.h"
#include "libavcodec/dsputil.h"
#include "dsputil_sh4.h"
#define LP(p) *(uint32_t*)(p)
#define LPC(p) *(const uint32_t*)(p)
#define UNPACK(ph,pl,tt0,tt1) do { \
uint32_t t0,t1; t0=tt0;t1=tt1; \
ph = ( (t0 & ~BYTE_VEC32(0x03))>>2) + ( (t1 & ~BYTE_VEC32(0x03))>>2); \
pl = (t0 & BYTE_VEC32(0x03)) + (t1 & BYTE_VEC32(0x03)); } while(0)
#define rnd_PACK(ph,pl,nph,npl) ph + nph + (((pl + npl + BYTE_VEC32(0x02))>>2) & BYTE_VEC32(0x03))
#define no_rnd_PACK(ph,pl,nph,npl) ph + nph + (((pl + npl + BYTE_VEC32(0x01))>>2) & BYTE_VEC32(0x03))
/* little-endian */
#define MERGE1(a,b,ofs) (ofs==0)?a:( ((a)>>(8*ofs))|((b)<<(32-8*ofs)) )
#define MERGE2(a,b,ofs) (ofs==3)?b:( ((a)>>(8*(ofs+1)))|((b)<<(32-8*(ofs+1))) )
/* big
#define MERGE1(a,b,ofs) (ofs==0)?a:( ((a)<<(8*ofs))|((b)>>(32-8*ofs)) )
#define MERGE2(a,b,ofs) (ofs==3)?b:( ((a)<<(8+8*ofs))|((b)>>(32-8-8*ofs)) )
*/
#define put(d,s) d = s
#define avg(d,s) d = rnd_avg32(s,d)
#define OP_C4(ofs) \
ref-=ofs; \
do { \
OP(LP(dest),MERGE1(LPC(ref),LPC(ref+4),ofs)); \
ref+=stride; \
dest+=stride; \
} while(--height)
#define OP_C40() \
do { \
OP(LP(dest),LPC(ref)); \
ref+=stride; \
dest+=stride; \
} while(--height)
#define OP put
static void put_pixels4_c(uint8_t *dest,const uint8_t *ref, const int stride,int height)
{
switch((int)ref&3){
case 0: OP_C40(); return;
case 1: OP_C4(1); return;
case 2: OP_C4(2); return;
case 3: OP_C4(3); return;
}
}
#undef OP
#define OP avg
static void avg_pixels4_c(uint8_t *dest,const uint8_t *ref, const int stride,int height)
{
switch((int)ref&3){
case 0: OP_C40(); return;
case 1: OP_C4(1); return;
case 2: OP_C4(2); return;
case 3: OP_C4(3); return;
}
}
#undef OP
#define OP_C(ofs,sz,avg2) \
{ \
ref-=ofs; \
do { \
uint32_t t0,t1; \
t0 = LPC(ref+0); \
t1 = LPC(ref+4); \
OP(LP(dest+0), MERGE1(t0,t1,ofs)); \
t0 = LPC(ref+8); \
OP(LP(dest+4), MERGE1(t1,t0,ofs)); \
if (sz==16) { \
t1 = LPC(ref+12); \
OP(LP(dest+8), MERGE1(t0,t1,ofs)); \
t0 = LPC(ref+16); \
OP(LP(dest+12), MERGE1(t1,t0,ofs)); \
} \
ref+=stride; \
dest+= stride; \
} while(--height); \
}
/* aligned */
#define OP_C0(sz,avg2) \
{ \
do { \
OP(LP(dest+0), LPC(ref+0)); \
OP(LP(dest+4), LPC(ref+4)); \
if (sz==16) { \
OP(LP(dest+8), LPC(ref+8)); \
OP(LP(dest+12), LPC(ref+12)); \
} \
ref+=stride; \
dest+= stride; \
} while(--height); \
}
#define OP_X(ofs,sz,avg2) \
{ \
ref-=ofs; \
do { \
uint32_t t0,t1; \
t0 = LPC(ref+0); \
t1 = LPC(ref+4); \
OP(LP(dest+0), avg2(MERGE1(t0,t1,ofs),MERGE2(t0,t1,ofs))); \
t0 = LPC(ref+8); \
OP(LP(dest+4), avg2(MERGE1(t1,t0,ofs),MERGE2(t1,t0,ofs))); \
if (sz==16) { \
t1 = LPC(ref+12); \
OP(LP(dest+8), avg2(MERGE1(t0,t1,ofs),MERGE2(t0,t1,ofs))); \
t0 = LPC(ref+16); \
OP(LP(dest+12), avg2(MERGE1(t1,t0,ofs),MERGE2(t1,t0,ofs))); \
} \
ref+=stride; \
dest+= stride; \
} while(--height); \
}
/* aligned */
#define OP_Y0(sz,avg2) \
{ \
uint32_t t0,t1,t2,t3,t; \
\
t0 = LPC(ref+0); \
t1 = LPC(ref+4); \
if (sz==16) { \
t2 = LPC(ref+8); \
t3 = LPC(ref+12); \
} \
do { \
ref += stride; \
\
t = LPC(ref+0); \
OP(LP(dest+0), avg2(t0,t)); t0 = t; \
t = LPC(ref+4); \
OP(LP(dest+4), avg2(t1,t)); t1 = t; \
if (sz==16) { \
t = LPC(ref+8); \
OP(LP(dest+8), avg2(t2,t)); t2 = t; \
t = LPC(ref+12); \
OP(LP(dest+12), avg2(t3,t)); t3 = t; \
} \
dest+= stride; \
} while(--height); \
}
#define OP_Y(ofs,sz,avg2) \
{ \
uint32_t t0,t1,t2,t3,t,w0,w1; \
\
ref-=ofs; \
w0 = LPC(ref+0); \
w1 = LPC(ref+4); \
t0 = MERGE1(w0,w1,ofs); \
w0 = LPC(ref+8); \
t1 = MERGE1(w1,w0,ofs); \
if (sz==16) { \
w1 = LPC(ref+12); \
t2 = MERGE1(w0,w1,ofs); \
w0 = LPC(ref+16); \
t3 = MERGE1(w1,w0,ofs); \
} \
do { \
ref += stride; \
\
w0 = LPC(ref+0); \
w1 = LPC(ref+4); \
t = MERGE1(w0,w1,ofs); \
OP(LP(dest+0), avg2(t0,t)); t0 = t; \
w0 = LPC(ref+8); \
t = MERGE1(w1,w0,ofs); \
OP(LP(dest+4), avg2(t1,t)); t1 = t; \
if (sz==16) { \
w1 = LPC(ref+12); \
t = MERGE1(w0,w1,ofs); \
OP(LP(dest+8), avg2(t2,t)); t2 = t; \
w0 = LPC(ref+16); \
t = MERGE1(w1,w0,ofs); \
OP(LP(dest+12), avg2(t3,t)); t3 = t; \
} \
dest+=stride; \
} while(--height); \
}
#define OP_X0(sz,avg2) OP_X(0,sz,avg2)
#define OP_XY0(sz,PACK) OP_XY(0,sz,PACK)
#define OP_XY(ofs,sz,PACK) \
{ \
uint32_t t2,t3,w0,w1; \
uint32_t a0,a1,a2,a3,a4,a5,a6,a7; \
\
ref -= ofs; \
w0 = LPC(ref+0); \
w1 = LPC(ref+4); \
UNPACK(a0,a1,MERGE1(w0,w1,ofs),MERGE2(w0,w1,ofs)); \
w0 = LPC(ref+8); \
UNPACK(a2,a3,MERGE1(w1,w0,ofs),MERGE2(w1,w0,ofs)); \
if (sz==16) { \
w1 = LPC(ref+12); \
UNPACK(a4,a5,MERGE1(w0,w1,ofs),MERGE2(w0,w1,ofs)); \
w0 = LPC(ref+16); \
UNPACK(a6,a7,MERGE1(w1,w0,ofs),MERGE2(w1,w0,ofs)); \
} \
do { \
ref+=stride; \
w0 = LPC(ref+0); \
w1 = LPC(ref+4); \
UNPACK(t2,t3,MERGE1(w0,w1,ofs),MERGE2(w0,w1,ofs)); \
OP(LP(dest+0),PACK(a0,a1,t2,t3)); \
a0 = t2; a1 = t3; \
w0 = LPC(ref+8); \
UNPACK(t2,t3,MERGE1(w1,w0,ofs),MERGE2(w1,w0,ofs)); \
OP(LP(dest+4),PACK(a2,a3,t2,t3)); \
a2 = t2; a3 = t3; \
if (sz==16) { \
w1 = LPC(ref+12); \
UNPACK(t2,t3,MERGE1(w0,w1,ofs),MERGE2(w0,w1,ofs)); \
OP(LP(dest+8),PACK(a4,a5,t2,t3)); \
a4 = t2; a5 = t3; \
w0 = LPC(ref+16); \
UNPACK(t2,t3,MERGE1(w1,w0,ofs),MERGE2(w1,w0,ofs)); \
OP(LP(dest+12),PACK(a6,a7,t2,t3)); \
a6 = t2; a7 = t3; \
} \
dest+=stride; \
} while(--height); \
}
#define DEFFUNC(op,rnd,xy,sz,OP_N,avgfunc) \
static void op##_##rnd##_pixels##sz##_##xy (uint8_t * dest, const uint8_t * ref, \
const int stride, int height) \
{ \
switch((int)ref&3) { \
case 0:OP_N##0(sz,rnd##_##avgfunc); return; \
case 1:OP_N(1,sz,rnd##_##avgfunc); return; \
case 2:OP_N(2,sz,rnd##_##avgfunc); return; \
case 3:OP_N(3,sz,rnd##_##avgfunc); return; \
} \
}
#define OP put
DEFFUNC(put, rnd,o,8,OP_C,avg32)
DEFFUNC(put, rnd,x,8,OP_X,avg32)
DEFFUNC(put,no_rnd,x,8,OP_X,avg32)
DEFFUNC(put, rnd,y,8,OP_Y,avg32)
DEFFUNC(put,no_rnd,y,8,OP_Y,avg32)
DEFFUNC(put, rnd,xy,8,OP_XY,PACK)
DEFFUNC(put,no_rnd,xy,8,OP_XY,PACK)
DEFFUNC(put, rnd,o,16,OP_C,avg32)
DEFFUNC(put, rnd,x,16,OP_X,avg32)
DEFFUNC(put,no_rnd,x,16,OP_X,avg32)
DEFFUNC(put, rnd,y,16,OP_Y,avg32)
DEFFUNC(put,no_rnd,y,16,OP_Y,avg32)
DEFFUNC(put, rnd,xy,16,OP_XY,PACK)
DEFFUNC(put,no_rnd,xy,16,OP_XY,PACK)
#undef OP
#define OP avg
DEFFUNC(avg, rnd,o,8,OP_C,avg32)
DEFFUNC(avg, rnd,x,8,OP_X,avg32)
DEFFUNC(avg,no_rnd,x,8,OP_X,avg32)
DEFFUNC(avg, rnd,y,8,OP_Y,avg32)
DEFFUNC(avg,no_rnd,y,8,OP_Y,avg32)
DEFFUNC(avg, rnd,xy,8,OP_XY,PACK)
DEFFUNC(avg,no_rnd,xy,8,OP_XY,PACK)
DEFFUNC(avg, rnd,o,16,OP_C,avg32)
DEFFUNC(avg, rnd,x,16,OP_X,avg32)
DEFFUNC(avg,no_rnd,x,16,OP_X,avg32)
DEFFUNC(avg, rnd,y,16,OP_Y,avg32)
DEFFUNC(avg,no_rnd,y,16,OP_Y,avg32)
DEFFUNC(avg, rnd,xy,16,OP_XY,PACK)
DEFFUNC(avg,no_rnd,xy,16,OP_XY,PACK)
#undef OP
#define put_no_rnd_pixels8_o put_rnd_pixels8_o
#define put_no_rnd_pixels16_o put_rnd_pixels16_o
#define avg_no_rnd_pixels8_o avg_rnd_pixels8_o
#define avg_no_rnd_pixels16_o avg_rnd_pixels16_o
#define put_pixels8_c put_rnd_pixels8_o
#define put_pixels16_c put_rnd_pixels16_o
#define avg_pixels8_c avg_rnd_pixels8_o
#define avg_pixels16_c avg_rnd_pixels16_o
#define put_no_rnd_pixels8_c put_rnd_pixels8_o
#define put_no_rnd_pixels16_c put_rnd_pixels16_o
#define avg_no_rnd_pixels8_c avg_rnd_pixels8_o
#define avg_no_rnd_pixels16_c avg_rnd_pixels16_o
#define QPEL
#ifdef QPEL
#include "qpel.c"
#endif
void ff_dsputil_init_align(DSPContext* c, AVCodecContext *avctx)
{
const int high_bit_depth = avctx->bits_per_raw_sample > 8;
if (!high_bit_depth) {
c->put_pixels_tab[0][0] = put_rnd_pixels16_o;
c->put_pixels_tab[0][1] = put_rnd_pixels16_x;
c->put_pixels_tab[0][2] = put_rnd_pixels16_y;
c->put_pixels_tab[0][3] = put_rnd_pixels16_xy;
c->put_pixels_tab[1][0] = put_rnd_pixels8_o;
c->put_pixels_tab[1][1] = put_rnd_pixels8_x;
c->put_pixels_tab[1][2] = put_rnd_pixels8_y;
c->put_pixels_tab[1][3] = put_rnd_pixels8_xy;
c->put_no_rnd_pixels_tab[0][0] = put_no_rnd_pixels16_o;
c->put_no_rnd_pixels_tab[0][1] = put_no_rnd_pixels16_x;
c->put_no_rnd_pixels_tab[0][2] = put_no_rnd_pixels16_y;
c->put_no_rnd_pixels_tab[0][3] = put_no_rnd_pixels16_xy;
c->put_no_rnd_pixels_tab[1][0] = put_no_rnd_pixels8_o;
c->put_no_rnd_pixels_tab[1][1] = put_no_rnd_pixels8_x;
c->put_no_rnd_pixels_tab[1][2] = put_no_rnd_pixels8_y;
c->put_no_rnd_pixels_tab[1][3] = put_no_rnd_pixels8_xy;
c->avg_pixels_tab[0][0] = avg_rnd_pixels16_o;
c->avg_pixels_tab[0][1] = avg_rnd_pixels16_x;
c->avg_pixels_tab[0][2] = avg_rnd_pixels16_y;
c->avg_pixels_tab[0][3] = avg_rnd_pixels16_xy;
c->avg_pixels_tab[1][0] = avg_rnd_pixels8_o;
c->avg_pixels_tab[1][1] = avg_rnd_pixels8_x;
c->avg_pixels_tab[1][2] = avg_rnd_pixels8_y;
c->avg_pixels_tab[1][3] = avg_rnd_pixels8_xy;
c->avg_no_rnd_pixels_tab[0][0] = avg_no_rnd_pixels16_o;
c->avg_no_rnd_pixels_tab[0][1] = avg_no_rnd_pixels16_x;
c->avg_no_rnd_pixels_tab[0][2] = avg_no_rnd_pixels16_y;
c->avg_no_rnd_pixels_tab[0][3] = avg_no_rnd_pixels16_xy;
c->avg_no_rnd_pixels_tab[1][0] = avg_no_rnd_pixels8_o;
c->avg_no_rnd_pixels_tab[1][1] = avg_no_rnd_pixels8_x;
c->avg_no_rnd_pixels_tab[1][2] = avg_no_rnd_pixels8_y;
c->avg_no_rnd_pixels_tab[1][3] = avg_no_rnd_pixels8_xy;
}
#ifdef QPEL
#define dspfunc(PFX, IDX, NUM) \
c->PFX ## _pixels_tab[IDX][ 0] = PFX ## NUM ## _mc00_sh4; \
c->PFX ## _pixels_tab[IDX][ 1] = PFX ## NUM ## _mc10_sh4; \
c->PFX ## _pixels_tab[IDX][ 2] = PFX ## NUM ## _mc20_sh4; \
c->PFX ## _pixels_tab[IDX][ 3] = PFX ## NUM ## _mc30_sh4; \
c->PFX ## _pixels_tab[IDX][ 4] = PFX ## NUM ## _mc01_sh4; \
c->PFX ## _pixels_tab[IDX][ 5] = PFX ## NUM ## _mc11_sh4; \
c->PFX ## _pixels_tab[IDX][ 6] = PFX ## NUM ## _mc21_sh4; \
c->PFX ## _pixels_tab[IDX][ 7] = PFX ## NUM ## _mc31_sh4; \
c->PFX ## _pixels_tab[IDX][ 8] = PFX ## NUM ## _mc02_sh4; \
c->PFX ## _pixels_tab[IDX][ 9] = PFX ## NUM ## _mc12_sh4; \
c->PFX ## _pixels_tab[IDX][10] = PFX ## NUM ## _mc22_sh4; \
c->PFX ## _pixels_tab[IDX][11] = PFX ## NUM ## _mc32_sh4; \
c->PFX ## _pixels_tab[IDX][12] = PFX ## NUM ## _mc03_sh4; \
c->PFX ## _pixels_tab[IDX][13] = PFX ## NUM ## _mc13_sh4; \
c->PFX ## _pixels_tab[IDX][14] = PFX ## NUM ## _mc23_sh4; \
c->PFX ## _pixels_tab[IDX][15] = PFX ## NUM ## _mc33_sh4
dspfunc(put_qpel, 0, 16);
dspfunc(put_no_rnd_qpel, 0, 16);
dspfunc(avg_qpel, 0, 16);
/* dspfunc(avg_no_rnd_qpel, 0, 16); */
dspfunc(put_qpel, 1, 8);
dspfunc(put_no_rnd_qpel, 1, 8);
dspfunc(avg_qpel, 1, 8);
/* dspfunc(avg_no_rnd_qpel, 1, 8); */
if (!high_bit_depth) {
dspfunc(put_h264_qpel, 0, 16);
dspfunc(put_h264_qpel, 1, 8);
dspfunc(put_h264_qpel, 2, 4);
dspfunc(avg_h264_qpel, 0, 16);
dspfunc(avg_h264_qpel, 1, 8);
dspfunc(avg_h264_qpel, 2, 4);
}
#undef dspfunc
if (!high_bit_depth) {
c->put_h264_chroma_pixels_tab[0]= put_h264_chroma_mc8_sh4;
c->put_h264_chroma_pixels_tab[1]= put_h264_chroma_mc4_sh4;
c->put_h264_chroma_pixels_tab[2]= put_h264_chroma_mc2_sh4;
c->avg_h264_chroma_pixels_tab[0]= avg_h264_chroma_mc8_sh4;
c->avg_h264_chroma_pixels_tab[1]= avg_h264_chroma_mc4_sh4;
c->avg_h264_chroma_pixels_tab[2]= avg_h264_chroma_mc2_sh4;
}
c->put_mspel_pixels_tab[0]= put_mspel8_mc00_sh4;
c->put_mspel_pixels_tab[1]= put_mspel8_mc10_sh4;
c->put_mspel_pixels_tab[2]= put_mspel8_mc20_sh4;
c->put_mspel_pixels_tab[3]= put_mspel8_mc30_sh4;
c->put_mspel_pixels_tab[4]= put_mspel8_mc02_sh4;
c->put_mspel_pixels_tab[5]= put_mspel8_mc12_sh4;
c->put_mspel_pixels_tab[6]= put_mspel8_mc22_sh4;
c->put_mspel_pixels_tab[7]= put_mspel8_mc32_sh4;
c->gmc1 = gmc1_c;
#endif
}

View File

@@ -0,0 +1,107 @@
/*
* sh4 dsputil
*
* Copyright (c) 2003 BERO <bero@geocities.co.jp>
*
* This file is part of FFmpeg.
*
* FFmpeg is free software; you can redistribute it and/or
* modify it under the terms of the GNU Lesser General Public
* License as published by the Free Software Foundation; either
* version 2.1 of the License, or (at your option) any later version.
*
* FFmpeg 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
* Lesser General Public License for more details.
*
* You should have received a copy of the GNU Lesser General Public
* License along with FFmpeg; if not, write to the Free Software
* Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
*/
#include "libavcodec/avcodec.h"
#include "libavcodec/dsputil.h"
#include "dsputil_sh4.h"
#include "sh4.h"
static void memzero_align8(void *dst,size_t size)
{
int fpscr;
fp_single_enter(fpscr);
dst = (char *)dst + size;
size /= 32;
__asm__ volatile (
" fldi0 fr0\n"
" fldi0 fr1\n"
" fschg\n" // double
"1: \n" \
" dt %1\n"
" fmov dr0,@-%0\n"
" fmov dr0,@-%0\n"
" fmov dr0,@-%0\n"
" bf.s 1b\n"
" fmov dr0,@-%0\n"
" fschg" //back to single
: "+r"(dst),"+r"(size) :: "memory" );
fp_single_leave(fpscr);
}
static void clear_blocks_sh4(DCTELEM *blocks)
{
memzero_align8(blocks,sizeof(DCTELEM)*6*64);
}
static void idct_put(uint8_t *dest, int line_size, DCTELEM *block)
{
int i;
uint8_t *cm = ff_cropTbl + MAX_NEG_CROP;
ff_idct_sh4(block);
for(i=0;i<8;i++) {
dest[0] = cm[block[0]];
dest[1] = cm[block[1]];
dest[2] = cm[block[2]];
dest[3] = cm[block[3]];
dest[4] = cm[block[4]];
dest[5] = cm[block[5]];
dest[6] = cm[block[6]];
dest[7] = cm[block[7]];
dest+=line_size;
block+=8;
}
}
static void idct_add(uint8_t *dest, int line_size, DCTELEM *block)
{
int i;
uint8_t *cm = ff_cropTbl + MAX_NEG_CROP;
ff_idct_sh4(block);
for(i=0;i<8;i++) {
dest[0] = cm[dest[0]+block[0]];
dest[1] = cm[dest[1]+block[1]];
dest[2] = cm[dest[2]+block[2]];
dest[3] = cm[dest[3]+block[3]];
dest[4] = cm[dest[4]+block[4]];
dest[5] = cm[dest[5]+block[5]];
dest[6] = cm[dest[6]+block[6]];
dest[7] = cm[dest[7]+block[7]];
dest+=line_size;
block+=8;
}
}
void ff_dsputil_init_sh4(DSPContext* c, AVCodecContext *avctx)
{
const int idct_algo= avctx->idct_algo;
const int high_bit_depth = avctx->bits_per_raw_sample > 8;
ff_dsputil_init_align(c,avctx);
if (!high_bit_depth)
c->clear_blocks = clear_blocks_sh4;
if (avctx->bits_per_raw_sample <= 8 &&
(idct_algo==FF_IDCT_AUTO || idct_algo==FF_IDCT_SH4)) {
c->idct_put = idct_put;
c->idct_add = idct_add;
c->idct = ff_idct_sh4;
c->idct_permutation_type= FF_NO_IDCT_PERM;
}
}

View File

@@ -0,0 +1,28 @@
/*
* This file is part of FFmpeg.
*
* FFmpeg is free software; you can redistribute it and/or
* modify it under the terms of the GNU Lesser General Public
* License as published by the Free Software Foundation; either
* version 2.1 of the License, or (at your option) any later version.
*
* FFmpeg 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
* Lesser General Public License for more details.
*
* You should have received a copy of the GNU Lesser General Public
* License along with FFmpeg; if not, write to the Free Software
* Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
*/
#ifndef AVCODEC_SH4_DSPUTIL_SH4_H
#define AVCODEC_SH4_DSPUTIL_SH4_H
#include "libavcodec/avcodec.h"
#include "libavcodec/dsputil.h"
void ff_idct_sh4(DCTELEM *block);
void ff_dsputil_init_align(DSPContext* c, AVCodecContext *avctx);
#endif /* AVCODEC_SH4_DSPUTIL_SH4_H */

View File

@@ -0,0 +1,211 @@
/*
* idct for sh4
*
* Copyright (c) 2001-2003 BERO <bero@geocities.co.jp>
*
* This file is part of FFmpeg.
*
* FFmpeg is free software; you can redistribute it and/or
* modify it under the terms of the GNU Lesser General Public
* License as published by the Free Software Foundation; either
* version 2.1 of the License, or (at your option) any later version.
*
* FFmpeg 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
* Lesser General Public License for more details.
*
* You should have received a copy of the GNU Lesser General Public
* License along with FFmpeg; if not, write to the Free Software
* Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
*/
#include "libavcodec/dsputil.h"
#include "dsputil_sh4.h"
#include "sh4.h"
#define c1 1.38703984532214752434 /* sqrt(2)*cos(1*pi/16) */
#define c2 1.30656296487637657577 /* sqrt(2)*cos(2*pi/16) */
#define c3 1.17587560241935884520 /* sqrt(2)*cos(3*pi/16) */
#define c4 1.00000000000000000000 /* sqrt(2)*cos(4*pi/16) */
#define c5 0.78569495838710234903 /* sqrt(2)*cos(5*pi/16) */
#define c6 0.54119610014619712324 /* sqrt(2)*cos(6*pi/16) */
#define c7 0.27589937928294311353 /* sqrt(2)*cos(7*pi/16) */
static const float even_table[] __attribute__ ((aligned(8))) = {
c4, c4, c4, c4,
c2, c6,-c6,-c2,
c4,-c4,-c4, c4,
c6,-c2, c2,-c6
};
static const float odd_table[] __attribute__ ((aligned(8))) = {
c1, c3, c5, c7,
c3,-c7,-c1,-c5,
c5,-c1, c7, c3,
c7,-c5, c3,-c1
};
#undef c1
#undef c2
#undef c3
#undef c4
#undef c5
#undef c6
#undef c7
#define load_matrix(table) \
do { \
const float *t = table; \
__asm__ volatile( \
" fschg\n" \
" fmov @%0+,xd0\n" \
" fmov @%0+,xd2\n" \
" fmov @%0+,xd4\n" \
" fmov @%0+,xd6\n" \
" fmov @%0+,xd8\n" \
" fmov @%0+,xd10\n" \
" fmov @%0+,xd12\n" \
" fmov @%0+,xd14\n" \
" fschg\n" \
: "+r"(t) \
); \
} while (0)
#define ftrv() \
__asm__ volatile("ftrv xmtrx,fv0" \
: "+f"(fr0),"+f"(fr1),"+f"(fr2),"+f"(fr3));
#define DEFREG \
register float fr0 __asm__("fr0"); \
register float fr1 __asm__("fr1"); \
register float fr2 __asm__("fr2"); \
register float fr3 __asm__("fr3")
#define DESCALE(x,n) (x)*(1.0f/(1<<(n)))
/* this code work worse on gcc cvs. 3.2.3 work fine */
//optimized
void ff_idct_sh4(DCTELEM *block)
{
DEFREG;
int i;
float tblock[8*8],*fblock;
int ofs1,ofs2,ofs3;
int fpscr;
fp_single_enter(fpscr);
/* row */
/* even part */
load_matrix(even_table);
fblock = tblock+4;
i = 8;
do {
fr0 = block[0];
fr1 = block[2];
fr2 = block[4];
fr3 = block[6];
block+=8;
ftrv();
*--fblock = fr3;
*--fblock = fr2;
*--fblock = fr1;
*--fblock = fr0;
fblock+=8+4;
} while(--i);
block-=8*8;
fblock-=8*8+4;
load_matrix(odd_table);
i = 8;
do {
float t0,t1,t2,t3;
fr0 = block[1];
fr1 = block[3];
fr2 = block[5];
fr3 = block[7];
block+=8;
ftrv();
t0 = *fblock++;
t1 = *fblock++;
t2 = *fblock++;
t3 = *fblock++;
fblock+=4;
*--fblock = t0 - fr0;
*--fblock = t1 - fr1;
*--fblock = t2 - fr2;
*--fblock = t3 - fr3;
*--fblock = t3 + fr3;
*--fblock = t2 + fr2;
*--fblock = t1 + fr1;
*--fblock = t0 + fr0;
fblock+=8;
} while(--i);
block-=8*8;
fblock-=8*8;
/* col */
/* even part */
load_matrix(even_table);
ofs1 = sizeof(float)*2*8;
ofs2 = sizeof(float)*4*8;
ofs3 = sizeof(float)*6*8;
i = 8;
#define OA(fblock,ofs) *(float*)((char*)fblock + ofs)
do {
fr0 = OA(fblock, 0);
fr1 = OA(fblock,ofs1);
fr2 = OA(fblock,ofs2);
fr3 = OA(fblock,ofs3);
ftrv();
OA(fblock,0 ) = fr0;
OA(fblock,ofs1) = fr1;
OA(fblock,ofs2) = fr2;
OA(fblock,ofs3) = fr3;
fblock++;
} while(--i);
fblock-=8;
load_matrix(odd_table);
i=8;
do {
float t0,t1,t2,t3;
t0 = OA(fblock, 0); /* [8*0] */
t1 = OA(fblock,ofs1); /* [8*2] */
t2 = OA(fblock,ofs2); /* [8*4] */
t3 = OA(fblock,ofs3); /* [8*6] */
fblock+=8;
fr0 = OA(fblock, 0); /* [8*1] */
fr1 = OA(fblock,ofs1); /* [8*3] */
fr2 = OA(fblock,ofs2); /* [8*5] */
fr3 = OA(fblock,ofs3); /* [8*7] */
fblock+=-8+1;
ftrv();
block[8*0] = DESCALE(t0 + fr0,3);
block[8*7] = DESCALE(t0 - fr0,3);
block[8*1] = DESCALE(t1 + fr1,3);
block[8*6] = DESCALE(t1 - fr1,3);
block[8*2] = DESCALE(t2 + fr2,3);
block[8*5] = DESCALE(t2 - fr2,3);
block[8*3] = DESCALE(t3 + fr3,3);
block[8*4] = DESCALE(t3 - fr3,3);
block++;
} while(--i);
fp_single_leave(fpscr);
}

File diff suppressed because it is too large Load Diff

View File

@@ -0,0 +1,44 @@
/*
* Copyright (c) 2008 Mans Rullgard <mans@mansr.com>
*
* This file is part of FFmpeg.
*
* FFmpeg is free software; you can redistribute it and/or
* modify it under the terms of the GNU Lesser General Public
* License as published by the Free Software Foundation; either
* version 2.1 of the License, or (at your option) any later version.
*
* FFmpeg 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
* Lesser General Public License for more details.
*
* You should have received a copy of the GNU Lesser General Public
* License along with FFmpeg; if not, write to the Free Software
* Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
*/
#ifndef AVCODEC_SH4_SH4_H
#define AVCODEC_SH4_SH4_H
#ifdef __SH4__
# define fp_single_enter(fpscr) \
do { \
__asm__ volatile ("sts fpscr, %0 \n\t" \
"and %1, %0 \n\t" \
"lds %0, fpscr \n\t" \
: "=&r"(fpscr) : "r"(~(1<<19))); \
} while (0)
# define fp_single_leave(fpscr) \
do { \
__asm__ volatile ("or %1, %0 \n\t" \
"lds %0, fpscr \n\t" \
: "+r"(fpscr) : "r"(1<<19)); \
} while (0)
#else
# define fp_single_enter(fpscr) ((void)fpscr)
# define fp_single_leave(fpscr)
#endif
#endif /* AVCODEC_SH4_SH4_H */