Skip to content

F
ffmpeg.wasm-core
Commit 7a556ebc authored by Michael Niedermayer's avatar Michael Niedermayer
Browse files
Options

Merge commit '0c15a9aa' 

* commit '0c15a9aa':
  sh4: Remove dubious aligned dsputil code

Conflicts:
	libavcodec/sh4/dsputil_align.c
	libavcodec/sh4/h264chroma_init.c
	libavcodec/sh4/hpeldsp.c
	libavcodec/sh4/qpel.c

If someone wants to maintain the sh4 code in ffmpeg, wants to
add more optimizations, or volunteers to maintain any of
what is removed here and can confirm that they are faster.
Then please contact us!
Merged-by: 's avatarMichael Niedermayer <michaelni@gmx.at>
parents 430d69c9 0c15a9aa
Expand all Show whitespace changes
Inline Side-by-side
Showing with 1 addition and 1666 deletions
libavcodec/h264chroma.c
View file @ 7a556ebc
......@@ -47,8 +47,6 @@ void ff_h264chroma_init(H264ChromaContext *c, int bit_depth)
ff_h264chroma_init_arm(c, bit_depth);
if (ARCH_PPC)
ff_h264chroma_init_ppc(c, bit_depth);
if (ARCH_SH4)
ff_h264chroma_init_sh4(c, bit_depth);
if (ARCH_X86)
ff_h264chroma_init_x86(c, bit_depth);
}
libavcodec/h264chroma.h
View file @ 7a556ebc
......@@ -32,7 +32,6 @@ void ff_h264chroma_init(H264ChromaContext *c, int bit_depth);
void ff_h264chroma_init_arm(H264ChromaContext *c, int bit_depth);
void ff_h264chroma_init_ppc(H264ChromaContext *c, int bit_depth);
void ff_h264chroma_init_sh4(H264ChromaContext *c, int bit_depth);
void ff_h264chroma_init_x86(H264ChromaContext *c, int bit_depth);
#endif /* AVCODEC_H264CHROMA_H */
libavcodec/hpeldsp.c
View file @ 7a556ebc
......@@ -62,8 +62,6 @@ av_cold void ff_hpeldsp_init(HpelDSPContext *c, int flags)
ff_hpeldsp_init_bfin(c, flags);
if (ARCH_PPC)
ff_hpeldsp_init_ppc(c, flags);
if (ARCH_SH4)
ff_hpeldsp_init_sh4(c, flags);
if (HAVE_VIS)
ff_hpeldsp_init_vis(c, flags);
if (ARCH_X86)
......
libavcodec/hpeldsp.h
View file @ 7a556ebc
......@@ -98,7 +98,6 @@ void ff_hpeldsp_init_alpha(HpelDSPContext *c, int flags);
void ff_hpeldsp_init_arm(HpelDSPContext *c, int flags);
void ff_hpeldsp_init_bfin(HpelDSPContext *c, int flags);
void ff_hpeldsp_init_ppc(HpelDSPContext *c, int flags);
void ff_hpeldsp_init_sh4(HpelDSPContext *c, int flags);
void ff_hpeldsp_init_vis(HpelDSPContext *c, int flags);
void ff_hpeldsp_init_x86(HpelDSPContext *c, int flags);
......
libavcodec/sh4/Makefile
View file @ 7a556ebc
OBJS += sh4/dsputil_align.o \
sh4/dsputil_sh4.o \
OBJS += sh4/dsputil_sh4.o \
sh4/idct_sh4.o \
OBJS-$(CONFIG_H264CHROMA) += sh4/h264chroma_init.o \
OBJS-$(CONFIG_HPELDSP) += sh4/hpeldsp.o
libavcodec/sh4/dsputil_align.c deleted 100644 → 0
View file @ 430d69c9
/*
* 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 "libavutil/attributes.h"
#include "libavcodec/avcodec.h"
#include "libavcodec/dsputil.h"
#include "libavcodec/rnd_avg.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_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 put_pixels8_c ff_put_rnd_pixels8_o
#define put_pixels16_c ff_put_rnd_pixels16_o
#define avg_pixels8_c ff_avg_rnd_pixels8_o
#define avg_pixels16_c ff_avg_rnd_pixels16_o
#define put_no_rnd_pixels8_c ff_put_rnd_pixels8_o
#define put_no_rnd_pixels16_c ff_put_rnd_pixels16_o
#define avg_no_rnd_pixels16_c ff_avg_rnd_pixels16_o
#if CONFIG_HPELDSP
#include "qpel.c"
#endif
av_cold void ff_dsputil_init_align(DSPContext *c, AVCodecContext *avctx)
{
#if CONFIG_HPELDSP
#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); */
#undef dspfunc
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
}
libavcodec/sh4/dsputil_sh4.c
View file @ 7a556ebc
......@@ -94,7 +94,6 @@ av_cold 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;
......
libavcodec/sh4/dsputil_sh4.h
View file @ 7a556ebc
......@@ -24,15 +24,5 @@
#include "libavcodec/hpeldsp.h"
void ff_idct_sh4(int16_t *block);
void ff_dsputil_init_align(DSPContext* c, AVCodecContext *avctx);
void ff_put_rnd_pixels8_o(uint8_t *dest, const uint8_t *ref,
const ptrdiff_t stride, int height);
void ff_put_rnd_pixels16_o(uint8_t *dest, const uint8_t *ref,
const ptrdiff_t stride, int height);
void ff_avg_rnd_pixels8_o (uint8_t *dest, const uint8_t *ref,
const ptrdiff_t stride, int height);
void ff_avg_rnd_pixels16_o(uint8_t *dest, const uint8_t *ref,
const ptrdiff_t stride, int height);
#endif /* AVCODEC_SH4_DSPUTIL_SH4_H */
libavcodec/sh4/h264chroma_init.c deleted 100644 → 0
View file @ 430d69c9
/*
* 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 <assert.h>
#include <stdint.h>
#include "libavutil/attributes.h"
#include "libavcodec/h264chroma.h"
#define H264_CHROMA_MC(OPNAME, OP)\
static void OPNAME ## h264_chroma_mc2_sh4(uint8_t *dst/*align 8*/, uint8_t *src/*align 1*/, int stride, int h, int x, int y){\
const int A=(8-x)*(8-y);\
const int B=( x)*(8-y);\
const int C=(8-x)*( y);\
const int D=( x)*( y);\
\
assert(x<8 && y<8 && x>=0 && y>=0);\
\
do {\
int t0,t1,t2,t3; \
uint8_t *s0 = src; \
uint8_t *s1 = src+stride; \
t0 = *s0++; t2 = *s1++; \
t1 = *s0++; t3 = *s1++; \
OP(dst[0], (A*t0 + B*t1 + C*t2 + D*t3));\
t0 = *s0++; t2 = *s1++; \
OP(dst[1], (A*t1 + B*t0 + C*t3 + D*t2));\
dst+= stride;\
src+= stride;\
}while(--h);\
}\
\
static void OPNAME ## h264_chroma_mc4_sh4(uint8_t *dst/*align 8*/, uint8_t *src/*align 1*/, int stride, int h, int x, int y){\
const int A=(8-x)*(8-y);\
const int B=( x)*(8-y);\
const int C=(8-x)*( y);\
const int D=( x)*( y);\
\
assert(x<8 && y<8 && x>=0 && y>=0);\
\
do {\
int t0,t1,t2,t3; \
uint8_t *s0 = src; \
uint8_t *s1 = src+stride; \
t0 = *s0++; t2 = *s1++; \
t1 = *s0++; t3 = *s1++; \
OP(dst[0], (A*t0 + B*t1 + C*t2 + D*t3));\
t0 = *s0++; t2 = *s1++; \
OP(dst[1], (A*t1 + B*t0 + C*t3 + D*t2));\
t1 = *s0++; t3 = *s1++; \
OP(dst[2], (A*t0 + B*t1 + C*t2 + D*t3));\
t0 = *s0++; t2 = *s1++; \
OP(dst[3], (A*t1 + B*t0 + C*t3 + D*t2));\
dst+= stride;\
src+= stride;\
}while(--h);\
}\
\
static void OPNAME ## h264_chroma_mc8_sh4(uint8_t *dst/*align 8*/, uint8_t *src/*align 1*/, int stride, int h, int x, int y){\
const int A=(8-x)*(8-y);\
const int B=( x)*(8-y);\
const int C=(8-x)*( y);\
const int D=( x)*( y);\
\
assert(x<8 && y<8 && x>=0 && y>=0);\
\
do {\
int t0,t1,t2,t3; \
uint8_t *s0 = src; \
uint8_t *s1 = src+stride; \
t0 = *s0++; t2 = *s1++; \
t1 = *s0++; t3 = *s1++; \
OP(dst[0], (A*t0 + B*t1 + C*t2 + D*t3));\
t0 = *s0++; t2 = *s1++; \
OP(dst[1], (A*t1 + B*t0 + C*t3 + D*t2));\
t1 = *s0++; t3 = *s1++; \
OP(dst[2], (A*t0 + B*t1 + C*t2 + D*t3));\
t0 = *s0++; t2 = *s1++; \
OP(dst[3], (A*t1 + B*t0 + C*t3 + D*t2));\
t1 = *s0++; t3 = *s1++; \
OP(dst[4], (A*t0 + B*t1 + C*t2 + D*t3));\
t0 = *s0++; t2 = *s1++; \
OP(dst[5], (A*t1 + B*t0 + C*t3 + D*t2));\
t1 = *s0++; t3 = *s1++; \
OP(dst[6], (A*t0 + B*t1 + C*t2 + D*t3));\
t0 = *s0++; t2 = *s1++; \
OP(dst[7], (A*t1 + B*t0 + C*t3 + D*t2));\
dst+= stride;\
src+= stride;\
}while(--h);\
}
#define op_avg(a, b) a = (((a)+(((b) + 32)>>6)+1)>>1)
#define op_put(a, b) a = (((b) + 32)>>6)
H264_CHROMA_MC(put_ , op_put)
H264_CHROMA_MC(avg_ , op_avg)
#undef op_avg
#undef op_put
av_cold void ff_h264chroma_init_sh4(H264ChromaContext *c, int bit_depth)
{
const int high_bit_depth = bit_depth > 8;
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;
}
}
libavcodec/sh4/hpeldsp.c deleted 100644 → 0
View file @ 430d69c9
This diff is collapsed.
libavcodec/sh4/qpel.c deleted 100644 → 0
View file @ 430d69c9
This diff is collapsed.
Markdown is supported
0% or
You are about to add 0 people to the discussion. Proceed with caution.
Finish editing this message first!
Please register or to comment