summaryrefslogtreecommitdiff
path: root/ffmpeg/libavcodec/sh4
diff options
context:
space:
mode:
Diffstat (limited to 'ffmpeg/libavcodec/sh4')
-rw-r--r--ffmpeg/libavcodec/sh4/Makefile7
-rw-r--r--ffmpeg/libavcodec/sh4/dsputil_align.c312
-rw-r--r--ffmpeg/libavcodec/sh4/dsputil_sh4.c108
-rw-r--r--ffmpeg/libavcodec/sh4/dsputil_sh4.h34
-rw-r--r--ffmpeg/libavcodec/sh4/h264chroma_init.c132
-rw-r--r--ffmpeg/libavcodec/sh4/hpeldsp.c347
-rw-r--r--ffmpeg/libavcodec/sh4/idct_sh4.c210
-rw-r--r--ffmpeg/libavcodec/sh4/qpel.c862
-rw-r--r--ffmpeg/libavcodec/sh4/sh4.h44
9 files changed, 2056 insertions, 0 deletions
diff --git a/ffmpeg/libavcodec/sh4/Makefile b/ffmpeg/libavcodec/sh4/Makefile
new file mode 100644
index 0000000..a24b298
--- /dev/null
+++ b/ffmpeg/libavcodec/sh4/Makefile
@@ -0,0 +1,7 @@
+OBJS += sh4/dsputil_align.o \
+ sh4/dsputil_sh4.o \
+ sh4/idct_sh4.o \
+
+OBJS-$(CONFIG_H264CHROMA) += sh4/h264chroma_init.o \
+
+OBJS-$(CONFIG_HPELDSP) += sh4/hpeldsp.o
diff --git a/ffmpeg/libavcodec/sh4/dsputil_align.c b/ffmpeg/libavcodec/sh4/dsputil_align.c
new file mode 100644
index 0000000..d6bd479
--- /dev/null
+++ b/ffmpeg/libavcodec/sh4/dsputil_align.c
@@ -0,0 +1,312 @@
+/*
+ * 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 DEFFUNC(op,rnd,xy,sz,OP_N,avgfunc) \
+static void op##_##rnd##_pixels##sz##_##xy (uint8_t * dest, const uint8_t * ref, \
+ const ptrdiff_t 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 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_H264QPEL
+
+#include "qpel.c"
+
+#endif
+
+av_cold void ff_dsputil_init_align(DSPContext *c, AVCodecContext *avctx)
+{
+ const int high_bit_depth = avctx->bits_per_raw_sample > 8;
+
+#if CONFIG_H264QPEL
+
+#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
+}
diff --git a/ffmpeg/libavcodec/sh4/dsputil_sh4.c b/ffmpeg/libavcodec/sh4/dsputil_sh4.c
new file mode 100644
index 0000000..6e19fa0
--- /dev/null
+++ b/ffmpeg/libavcodec/sh4/dsputil_sh4.c
@@ -0,0 +1,108 @@
+/*
+ * 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 "libavutil/attributes.h"
+#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(int16_t *blocks)
+{
+ memzero_align8(blocks,sizeof(int16_t)*6*64);
+}
+
+static void idct_put(uint8_t *dest, int line_size, int16_t *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, int16_t *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;
+ }
+}
+
+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;
+ 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;
+ }
+}
diff --git a/ffmpeg/libavcodec/sh4/dsputil_sh4.h b/ffmpeg/libavcodec/sh4/dsputil_sh4.h
new file mode 100644
index 0000000..bd97a5b
--- /dev/null
+++ b/ffmpeg/libavcodec/sh4/dsputil_sh4.h
@@ -0,0 +1,34 @@
+/*
+ * 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"
+#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 int stride, int height);
+void ff_put_rnd_pixels16_o(uint8_t * dest, const uint8_t * ref, const int stride, int height);
+void ff_avg_rnd_pixels8_o (uint8_t * dest, const uint8_t * ref, const int stride, int height);
+void ff_avg_rnd_pixels16_o(uint8_t * dest, const uint8_t * ref, const int stride, int height);
+
+#endif /* AVCODEC_SH4_DSPUTIL_SH4_H */
diff --git a/ffmpeg/libavcodec/sh4/h264chroma_init.c b/ffmpeg/libavcodec/sh4/h264chroma_init.c
new file mode 100644
index 0000000..d15f0ae
--- /dev/null
+++ b/ffmpeg/libavcodec/sh4/h264chroma_init.c
@@ -0,0 +1,132 @@
+/*
+ * 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;
+ }
+}
diff --git a/ffmpeg/libavcodec/sh4/hpeldsp.c b/ffmpeg/libavcodec/sh4/hpeldsp.c
new file mode 100644
index 0000000..b524003
--- /dev/null
+++ b/ffmpeg/libavcodec/sh4/hpeldsp.c
@@ -0,0 +1,347 @@
+/*
+ * 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 "libavcodec/bit_depth_template.c" // for BYTE_VEC32
+#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(prefix, op,rnd,xy,sz,OP_N,avgfunc) \
+prefix 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( ,ff_put,rnd,o,8,OP_C,avg32)
+DEFFUNC(static,put, rnd,x,8,OP_X,avg32)
+DEFFUNC(static,put,no_rnd,x,8,OP_X,avg32)
+DEFFUNC(static,put, rnd,y,8,OP_Y,avg32)
+DEFFUNC(static,put,no_rnd,y,8,OP_Y,avg32)
+DEFFUNC(static,put, rnd,xy,8,OP_XY,PACK)
+DEFFUNC(static,put,no_rnd,xy,8,OP_XY,PACK)
+DEFFUNC( ,ff_put,rnd,o,16,OP_C,avg32)
+DEFFUNC(static,put, rnd,x,16,OP_X,avg32)
+DEFFUNC(static,put,no_rnd,x,16,OP_X,avg32)
+DEFFUNC(static,put, rnd,y,16,OP_Y,avg32)
+DEFFUNC(static,put,no_rnd,y,16,OP_Y,avg32)
+DEFFUNC(static,put, rnd,xy,16,OP_XY,PACK)
+DEFFUNC(static,put,no_rnd,xy,16,OP_XY,PACK)
+
+#undef OP
+#define OP avg
+
+DEFFUNC( ,ff_avg,rnd,o,8,OP_C,avg32)
+DEFFUNC(static,avg, rnd,x,8,OP_X,avg32)
+DEFFUNC(static,avg, rnd,y,8,OP_Y,avg32)
+DEFFUNC(static,avg, rnd,xy,8,OP_XY,PACK)
+DEFFUNC( ,ff_avg,rnd,o,16,OP_C,avg32)
+DEFFUNC(static,avg, rnd,x,16,OP_X,avg32)
+DEFFUNC(static,avg,no_rnd,x,16,OP_X,avg32)
+DEFFUNC(static,avg, rnd,y,16,OP_Y,avg32)
+DEFFUNC(static,avg,no_rnd,y,16,OP_Y,avg32)
+DEFFUNC(static,avg, rnd,xy,16,OP_XY,PACK)
+DEFFUNC(static,avg,no_rnd,xy,16,OP_XY,PACK)
+
+#undef OP
+
+#define ff_put_no_rnd_pixels8_o ff_put_rnd_pixels8_o
+#define ff_put_no_rnd_pixels16_o ff_put_rnd_pixels16_o
+#define ff_avg_no_rnd_pixels16_o ff_avg_rnd_pixels16_o
+
+void ff_hpeldsp_init_sh4(HpelDSPContext* c, int flags)
+{
+ c->put_pixels_tab[0][0] = ff_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] = ff_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] = ff_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] = ff_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] = ff_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] = ff_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] = ff_avg_no_rnd_pixels16_o;
+ c->avg_no_rnd_pixels_tab[1] = avg_no_rnd_pixels16_x;
+ c->avg_no_rnd_pixels_tab[2] = avg_no_rnd_pixels16_y;
+ c->avg_no_rnd_pixels_tab[3] = avg_no_rnd_pixels16_xy;
+}
diff --git a/ffmpeg/libavcodec/sh4/idct_sh4.c b/ffmpeg/libavcodec/sh4/idct_sh4.c
new file mode 100644
index 0000000..d4e8711
--- /dev/null
+++ b/ffmpeg/libavcodec/sh4/idct_sh4.c
@@ -0,0 +1,210 @@
+/*
+ * 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 "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(int16_t *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);
+}
diff --git a/ffmpeg/libavcodec/sh4/qpel.c b/ffmpeg/libavcodec/sh4/qpel.c
new file mode 100644
index 0000000..1a5a4cd
--- /dev/null
+++ b/ffmpeg/libavcodec/sh4/qpel.c
@@ -0,0 +1,862 @@
+/*
+ * This is optimized for sh, which have post increment addressing (*p++).
+ * Some CPU may be index (p[n]) faster than post increment (*p++).
+ *
+ * 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/common.h"
+#include "libavcodec/copy_block.h"
+#include "libavcodec/rnd_avg.h"
+
+#define PIXOP2(OPNAME, OP) \
+\
+static inline void OPNAME ## _pixels4_l2_aligned(uint8_t *dst, const uint8_t *src1, const uint8_t *src2, int dst_stride, int src_stride1, int src_stride2, int h) \
+{\
+ do {\
+ OP(LP(dst ),rnd_avg32(LPC(src1 ),LPC(src2 )) ); \
+ src1+=src_stride1; \
+ src2+=src_stride2; \
+ dst+=dst_stride; \
+ } while(--h); \
+}\
+\
+static inline void OPNAME ## _pixels4_l2_aligned2(uint8_t *dst, const uint8_t *src1, const uint8_t *src2, int dst_stride, int src_stride1, int src_stride2, int h) \
+{\
+ do {\
+ OP(LP(dst ),rnd_avg32(AV_RN32(src1 ),LPC(src2 )) ); \
+ src1+=src_stride1; \
+ src2+=src_stride2; \
+ dst+=dst_stride; \
+ } while(--h); \
+}\
+\
+static inline void OPNAME ## _no_rnd_pixels16_l2_aligned2(uint8_t *dst, const uint8_t *src1, const uint8_t *src2, int dst_stride, int src_stride1, int src_stride2, int h) \
+{\
+ do {\
+ OP(LP(dst ),no_rnd_avg32(AV_RN32(src1 ),LPC(src2 )) ); \
+ OP(LP(dst+4),no_rnd_avg32(AV_RN32(src1+4),LPC(src2+4)) ); \
+ OP(LP(dst+8),no_rnd_avg32(AV_RN32(src1+8),LPC(src2+8)) ); \
+ OP(LP(dst+12),no_rnd_avg32(AV_RN32(src1+12),LPC(src2+12)) ); \
+ src1+=src_stride1; \
+ src2+=src_stride2; \
+ dst+=dst_stride; \
+ } while(--h); \
+}\
+\
+static inline void OPNAME ## _pixels16_l2_aligned2(uint8_t *dst, const uint8_t *src1, const uint8_t *src2, int dst_stride, int src_stride1, int src_stride2, int h) \
+{\
+ do {\
+ OP(LP(dst ),rnd_avg32(AV_RN32(src1 ),LPC(src2 )) ); \
+ OP(LP(dst+4),rnd_avg32(AV_RN32(src1+4),LPC(src2+4)) ); \
+ OP(LP(dst+8),rnd_avg32(AV_RN32(src1+8),LPC(src2+8)) ); \
+ OP(LP(dst+12),rnd_avg32(AV_RN32(src1+12),LPC(src2+12)) ); \
+ src1+=src_stride1; \
+ src2+=src_stride2; \
+ dst+=dst_stride; \
+ } while(--h); \
+}\
+\
+static inline void OPNAME ## _no_rnd_pixels8_l2_aligned2(uint8_t *dst, const uint8_t *src1, const uint8_t *src2, int dst_stride, int src_stride1, int src_stride2, int h) \
+{\
+ do { /* onlye src2 aligned */\
+ OP(LP(dst ),no_rnd_avg32(AV_RN32(src1 ),LPC(src2 )) ); \
+ OP(LP(dst+4),no_rnd_avg32(AV_RN32(src1+4),LPC(src2+4)) ); \
+ src1+=src_stride1; \
+ src2+=src_stride2; \
+ dst+=dst_stride; \
+ } while(--h); \
+}\
+\
+static inline void OPNAME ## _pixels8_l2_aligned2(uint8_t *dst, const uint8_t *src1, const uint8_t *src2, int dst_stride, int src_stride1, int src_stride2, int h) \
+{\
+ do {\
+ OP(LP(dst ),rnd_avg32(AV_RN32(src1 ),LPC(src2 )) ); \
+ OP(LP(dst+4),rnd_avg32(AV_RN32(src1+4),LPC(src2+4)) ); \
+ src1+=src_stride1; \
+ src2+=src_stride2; \
+ dst+=dst_stride; \
+ } while(--h); \
+}\
+\
+static inline void OPNAME ## _no_rnd_pixels8_l2_aligned(uint8_t *dst, const uint8_t *src1, const uint8_t *src2, int dst_stride, int src_stride1, int src_stride2, int h) \
+{\
+ do {\
+ OP(LP(dst ),no_rnd_avg32(LPC(src1 ),LPC(src2 )) ); \
+ OP(LP(dst+4),no_rnd_avg32(LPC(src1+4),LPC(src2+4)) ); \
+ src1+=src_stride1; \
+ src2+=src_stride2; \
+ dst+=dst_stride; \
+ } while(--h); \
+}\
+\
+static inline void OPNAME ## _pixels8_l2_aligned(uint8_t *dst, const uint8_t *src1, const uint8_t *src2, int dst_stride, int src_stride1, int src_stride2, int h) \
+{\
+ do {\
+ OP(LP(dst ),rnd_avg32(LPC(src1 ),LPC(src2 )) ); \
+ OP(LP(dst+4),rnd_avg32(LPC(src1+4),LPC(src2+4)) ); \
+ src1+=src_stride1; \
+ src2+=src_stride2; \
+ dst+=dst_stride; \
+ } while(--h); \
+}\
+\
+static inline void OPNAME ## _no_rnd_pixels16_l2_aligned(uint8_t *dst, const uint8_t *src1, const uint8_t *src2, int dst_stride, int src_stride1, int src_stride2, int h) \
+{\
+ do {\
+ OP(LP(dst ),no_rnd_avg32(LPC(src1 ),LPC(src2 )) ); \
+ OP(LP(dst+4),no_rnd_avg32(LPC(src1+4),LPC(src2+4)) ); \
+ OP(LP(dst+8),no_rnd_avg32(LPC(src1+8),LPC(src2+8)) ); \
+ OP(LP(dst+12),no_rnd_avg32(LPC(src1+12),LPC(src2+12)) ); \
+ src1+=src_stride1; \
+ src2+=src_stride2; \
+ dst+=dst_stride; \
+ } while(--h); \
+}\
+\
+static inline void OPNAME ## _pixels16_l2_aligned(uint8_t *dst, const uint8_t *src1, const uint8_t *src2, int dst_stride, int src_stride1, int src_stride2, int h) \
+{\
+ do {\
+ OP(LP(dst ),rnd_avg32(LPC(src1 ),LPC(src2 )) ); \
+ OP(LP(dst+4),rnd_avg32(LPC(src1+4),LPC(src2+4)) ); \
+ OP(LP(dst+8),rnd_avg32(LPC(src1+8),LPC(src2+8)) ); \
+ OP(LP(dst+12),rnd_avg32(LPC(src1+12),LPC(src2+12)) ); \
+ src1+=src_stride1; \
+ src2+=src_stride2; \
+ dst+=dst_stride; \
+ } while(--h); \
+}\
+\
+static inline void OPNAME ## _no_rnd_pixels16_l2_aligned1(uint8_t *dst, const uint8_t *src1, const uint8_t *src2, int dst_stride, int src_stride1, int src_stride2, int h) \
+{ OPNAME ## _no_rnd_pixels16_l2_aligned2(dst,src2,src1,dst_stride,src_stride2,src_stride1,h); } \
+\
+static inline void OPNAME ## _pixels16_l2_aligned1(uint8_t *dst, const uint8_t *src1, const uint8_t *src2, int dst_stride, int src_stride1, int src_stride2, int h) \
+{ OPNAME ## _pixels16_l2_aligned2(dst,src2,src1,dst_stride,src_stride2,src_stride1,h); } \
+\
+static inline void OPNAME ## _no_rnd_pixels8_l2_aligned1(uint8_t *dst, const uint8_t *src1, const uint8_t *src2, int dst_stride, int src_stride1, int src_stride2, int h) \
+{ OPNAME ## _no_rnd_pixels8_l2_aligned2(dst,src2,src1,dst_stride,src_stride2,src_stride1,h); } \
+\
+static inline void OPNAME ## _pixels8_l2_aligned1(uint8_t *dst, const uint8_t *src1, const uint8_t *src2, int dst_stride, int src_stride1, int src_stride2, int h) \
+{ OPNAME ## _pixels8_l2_aligned2(dst,src2,src1,dst_stride,src_stride2,src_stride1,h); } \
+\
+static inline void OPNAME ## _pixels8_l4_aligned(uint8_t *dst, const uint8_t *src1, uint8_t *src2, uint8_t *src3, uint8_t *src4,int dst_stride, int src_stride1, int src_stride2,int src_stride3,int src_stride4, int h){\
+ do { \
+ uint32_t a0,a1,a2,a3; \
+ UNPACK(a0,a1,LPC(src1),LPC(src2)); \
+ UNPACK(a2,a3,LPC(src3),LPC(src4)); \
+ OP(LP(dst),rnd_PACK(a0,a1,a2,a3)); \
+ UNPACK(a0,a1,LPC(src1+4),LPC(src2+4)); \
+ UNPACK(a2,a3,LPC(src3+4),LPC(src4+4)); \
+ OP(LP(dst+4),rnd_PACK(a0,a1,a2,a3)); \
+ src1+=src_stride1;\
+ src2+=src_stride2;\
+ src3+=src_stride3;\
+ src4+=src_stride4;\
+ dst+=dst_stride;\
+ } while(--h); \
+} \
+\
+static inline void OPNAME ## _no_rnd_pixels8_l4_aligned(uint8_t *dst, const uint8_t *src1, uint8_t *src2, uint8_t *src3, uint8_t *src4,int dst_stride, int src_stride1, int src_stride2,int src_stride3,int src_stride4, int h){\
+ do { \
+ uint32_t a0,a1,a2,a3; \
+ UNPACK(a0,a1,LPC(src1),LPC(src2)); \
+ UNPACK(a2,a3,LPC(src3),LPC(src4)); \
+ OP(LP(dst),no_rnd_PACK(a0,a1,a2,a3)); \
+ UNPACK(a0,a1,LPC(src1+4),LPC(src2+4)); \
+ UNPACK(a2,a3,LPC(src3+4),LPC(src4+4)); \
+ OP(LP(dst+4),no_rnd_PACK(a0,a1,a2,a3)); \
+ src1+=src_stride1;\
+ src2+=src_stride2;\
+ src3+=src_stride3;\
+ src4+=src_stride4;\
+ dst+=dst_stride;\
+ } while(--h); \
+} \
+\
+static inline void OPNAME ## _pixels8_l4_aligned0(uint8_t *dst, const uint8_t *src1, uint8_t *src2, uint8_t *src3, uint8_t *src4,int dst_stride, int src_stride1, int src_stride2,int src_stride3,int src_stride4, int h){\
+ do { \
+ uint32_t a0,a1,a2,a3; /* src1 only not aligned */\
+ UNPACK(a0,a1,AV_RN32(src1),LPC(src2)); \
+ UNPACK(a2,a3,LPC(src3),LPC(src4)); \
+ OP(LP(dst),rnd_PACK(a0,a1,a2,a3)); \
+ UNPACK(a0,a1,AV_RN32(src1+4),LPC(src2+4)); \
+ UNPACK(a2,a3,LPC(src3+4),LPC(src4+4)); \
+ OP(LP(dst+4),rnd_PACK(a0,a1,a2,a3)); \
+ src1+=src_stride1;\
+ src2+=src_stride2;\
+ src3+=src_stride3;\
+ src4+=src_stride4;\
+ dst+=dst_stride;\
+ } while(--h); \
+} \
+\
+static inline void OPNAME ## _no_rnd_pixels8_l4_aligned0(uint8_t *dst, const uint8_t *src1, uint8_t *src2, uint8_t *src3, uint8_t *src4,int dst_stride, int src_stride1, int src_stride2,int src_stride3,int src_stride4, int h){\
+ do { \
+ uint32_t a0,a1,a2,a3; \
+ UNPACK(a0,a1,AV_RN32(src1),LPC(src2)); \
+ UNPACK(a2,a3,LPC(src3),LPC(src4)); \
+ OP(LP(dst),no_rnd_PACK(a0,a1,a2,a3)); \
+ UNPACK(a0,a1,AV_RN32(src1+4),LPC(src2+4)); \
+ UNPACK(a2,a3,LPC(src3+4),LPC(src4+4)); \
+ OP(LP(dst+4),no_rnd_PACK(a0,a1,a2,a3)); \
+ src1+=src_stride1;\
+ src2+=src_stride2;\
+ src3+=src_stride3;\
+ src4+=src_stride4;\
+ dst+=dst_stride;\
+ } while(--h); \
+} \
+\
+static inline void OPNAME ## _pixels16_l4_aligned(uint8_t *dst, const uint8_t *src1, uint8_t *src2, uint8_t *src3, uint8_t *src4,int dst_stride, int src_stride1, int src_stride2,int src_stride3,int src_stride4, int h){\
+ do { \
+ uint32_t a0,a1,a2,a3; \
+ UNPACK(a0,a1,LPC(src1),LPC(src2)); \
+ UNPACK(a2,a3,LPC(src3),LPC(src4)); \
+ OP(LP(dst),rnd_PACK(a0,a1,a2,a3)); \
+ UNPACK(a0,a1,LPC(src1+4),LPC(src2+4)); \
+ UNPACK(a2,a3,LPC(src3+4),LPC(src4+4)); \
+ OP(LP(dst+8),rnd_PACK(a0,a1,a2,a3)); \
+ UNPACK(a0,a1,LPC(src1+8),LPC(src2+8)); \
+ UNPACK(a2,a3,LPC(src3+8),LPC(src4+8)); \
+ OP(LP(dst+8),rnd_PACK(a0,a1,a2,a3)); \
+ UNPACK(a0,a1,LPC(src1+12),LPC(src2+12)); \
+ UNPACK(a2,a3,LPC(src3+12),LPC(src4+12)); \
+ OP(LP(dst+12),rnd_PACK(a0,a1,a2,a3)); \
+ src1+=src_stride1;\
+ src2+=src_stride2;\
+ src3+=src_stride3;\
+ src4+=src_stride4;\
+ dst+=dst_stride;\
+ } while(--h); \
+} \
+\
+static inline void OPNAME ## _no_rnd_pixels16_l4_aligned(uint8_t *dst, const uint8_t *src1, uint8_t *src2, uint8_t *src3, uint8_t *src4,int dst_stride, int src_stride1, int src_stride2,int src_stride3,int src_stride4, int h){\
+ do { \
+ uint32_t a0,a1,a2,a3; \
+ UNPACK(a0,a1,LPC(src1),LPC(src2)); \
+ UNPACK(a2,a3,LPC(src3),LPC(src4)); \
+ OP(LP(dst),no_rnd_PACK(a0,a1,a2,a3)); \
+ UNPACK(a0,a1,LPC(src1+4),LPC(src2+4)); \
+ UNPACK(a2,a3,LPC(src3+4),LPC(src4+4)); \
+ OP(LP(dst+4),no_rnd_PACK(a0,a1,a2,a3)); \
+ UNPACK(a0,a1,LPC(src1+8),LPC(src2+8)); \
+ UNPACK(a2,a3,LPC(src3+8),LPC(src4+8)); \
+ OP(LP(dst+8),no_rnd_PACK(a0,a1,a2,a3)); \
+ UNPACK(a0,a1,LPC(src1+12),LPC(src2+12)); \
+ UNPACK(a2,a3,LPC(src3+12),LPC(src4+12)); \
+ OP(LP(dst+12),no_rnd_PACK(a0,a1,a2,a3)); \
+ src1+=src_stride1;\
+ src2+=src_stride2;\
+ src3+=src_stride3;\
+ src4+=src_stride4;\
+ dst+=dst_stride;\
+ } while(--h); \
+} \
+\
+static inline void OPNAME ## _pixels16_l4_aligned0(uint8_t *dst, const uint8_t *src1, uint8_t *src2, uint8_t *src3, uint8_t *src4,int dst_stride, int src_stride1, int src_stride2,int src_stride3,int src_stride4, int h){\
+ do { /* src1 is unaligned */\
+ uint32_t a0,a1,a2,a3; \
+ UNPACK(a0,a1,AV_RN32(src1),LPC(src2)); \
+ UNPACK(a2,a3,LPC(src3),LPC(src4)); \
+ OP(LP(dst),rnd_PACK(a0,a1,a2,a3)); \
+ UNPACK(a0,a1,AV_RN32(src1+4),LPC(src2+4)); \
+ UNPACK(a2,a3,LPC(src3+4),LPC(src4+4)); \
+ OP(LP(dst+8),rnd_PACK(a0,a1,a2,a3)); \
+ UNPACK(a0,a1,AV_RN32(src1+8),LPC(src2+8)); \
+ UNPACK(a2,a3,LPC(src3+8),LPC(src4+8)); \
+ OP(LP(dst+8),rnd_PACK(a0,a1,a2,a3)); \
+ UNPACK(a0,a1,AV_RN32(src1+12),LPC(src2+12)); \
+ UNPACK(a2,a3,LPC(src3+12),LPC(src4+12)); \
+ OP(LP(dst+12),rnd_PACK(a0,a1,a2,a3)); \
+ src1+=src_stride1;\
+ src2+=src_stride2;\
+ src3+=src_stride3;\
+ src4+=src_stride4;\
+ dst+=dst_stride;\
+ } while(--h); \
+} \
+\
+static inline void OPNAME ## _no_rnd_pixels16_l4_aligned0(uint8_t *dst, const uint8_t *src1, uint8_t *src2, uint8_t *src3, uint8_t *src4,int dst_stride, int src_stride1, int src_stride2,int src_stride3,int src_stride4, int h){\
+ do { \
+ uint32_t a0,a1,a2,a3; \
+ UNPACK(a0,a1,AV_RN32(src1),LPC(src2)); \
+ UNPACK(a2,a3,LPC(src3),LPC(src4)); \
+ OP(LP(dst),no_rnd_PACK(a0,a1,a2,a3)); \
+ UNPACK(a0,a1,AV_RN32(src1+4),LPC(src2+4)); \
+ UNPACK(a2,a3,LPC(src3+4),LPC(src4+4)); \
+ OP(LP(dst+4),no_rnd_PACK(a0,a1,a2,a3)); \
+ UNPACK(a0,a1,AV_RN32(src1+8),LPC(src2+8)); \
+ UNPACK(a2,a3,LPC(src3+8),LPC(src4+8)); \
+ OP(LP(dst+8),no_rnd_PACK(a0,a1,a2,a3)); \
+ UNPACK(a0,a1,AV_RN32(src1+12),LPC(src2+12)); \
+ UNPACK(a2,a3,LPC(src3+12),LPC(src4+12)); \
+ OP(LP(dst+12),no_rnd_PACK(a0,a1,a2,a3)); \
+ src1+=src_stride1;\
+ src2+=src_stride2;\
+ src3+=src_stride3;\
+ src4+=src_stride4;\
+ dst+=dst_stride;\
+ } while(--h); \
+} \
+\
+
+#define op_avg(a, b) a = rnd_avg32(a,b)
+#define op_put(a, b) a = b
+
+PIXOP2(avg, op_avg)
+PIXOP2(put, op_put)
+#undef op_avg
+#undef op_put
+
+#define avg2(a,b) ((a+b+1)>>1)
+#define avg4(a,b,c,d) ((a+b+c+d+2)>>2)
+
+
+static void gmc1_c(uint8_t *dst, uint8_t *src, int stride, int h, int x16, int y16, int rounder)
+{
+ const int A=(16-x16)*(16-y16);
+ const int B=( x16)*(16-y16);
+ const int C=(16-x16)*( y16);
+ const int D=( x16)*( y16);
+
+ do {
+ int t0,t1,t2,t3;
+ uint8_t *s0 = src;
+ uint8_t *s1 = src+stride;
+ t0 = *s0++; t2 = *s1++;
+ t1 = *s0++; t3 = *s1++;
+ dst[0]= (A*t0 + B*t1 + C*t2 + D*t3 + rounder)>>8;
+ t0 = *s0++; t2 = *s1++;
+ dst[1]= (A*t1 + B*t0 + C*t3 + D*t2 + rounder)>>8;
+ t1 = *s0++; t3 = *s1++;
+ dst[2]= (A*t0 + B*t1 + C*t2 + D*t3 + rounder)>>8;
+ t0 = *s0++; t2 = *s1++;
+ dst[3]= (A*t1 + B*t0 + C*t3 + D*t2 + rounder)>>8;
+ t1 = *s0++; t3 = *s1++;
+ dst[4]= (A*t0 + B*t1 + C*t2 + D*t3 + rounder)>>8;
+ t0 = *s0++; t2 = *s1++;
+ dst[5]= (A*t1 + B*t0 + C*t3 + D*t2 + rounder)>>8;
+ t1 = *s0++; t3 = *s1++;
+ dst[6]= (A*t0 + B*t1 + C*t2 + D*t3 + rounder)>>8;
+ t0 = *s0++; t2 = *s1++;
+ dst[7]= (A*t1 + B*t0 + C*t3 + D*t2 + rounder)>>8;
+ dst+= stride;
+ src+= stride;
+ }while(--h);
+}
+
+#define QPEL_MC(r, OPNAME, RND, OP) \
+static void OPNAME ## mpeg4_qpel8_h_lowpass(uint8_t *dst, uint8_t *src, int dstStride, int srcStride, int h){\
+ uint8_t *cm = ff_cropTbl + MAX_NEG_CROP;\
+ do {\
+ uint8_t *s = src; \
+ int src0,src1,src2,src3,src4,src5,src6,src7,src8;\
+ src0= *s++;\
+ src1= *s++;\
+ src2= *s++;\
+ src3= *s++;\
+ src4= *s++;\
+ OP(dst[0], (src0+src1)*20 - (src0+src2)*6 + (src1+src3)*3 - (src2+src4));\
+ src5= *s++;\
+ OP(dst[1], (src1+src2)*20 - (src0+src3)*6 + (src0+src4)*3 - (src1+src5));\
+ src6= *s++;\
+ OP(dst[2], (src2+src3)*20 - (src1+src4)*6 + (src0+src5)*3 - (src0+src6));\
+ src7= *s++;\
+ OP(dst[3], (src3+src4)*20 - (src2+src5)*6 + (src1+src6)*3 - (src0+src7));\
+ src8= *s++;\
+ OP(dst[4], (src4+src5)*20 - (src3+src6)*6 + (src2+src7)*3 - (src1+src8));\
+ OP(dst[5], (src5+src6)*20 - (src4+src7)*6 + (src3+src8)*3 - (src2+src8));\
+ OP(dst[6], (src6+src7)*20 - (src5+src8)*6 + (src4+src8)*3 - (src3+src7));\
+ OP(dst[7], (src7+src8)*20 - (src6+src8)*6 + (src5+src7)*3 - (src4+src6));\
+ dst+=dstStride;\
+ src+=srcStride;\
+ }while(--h);\
+}\
+\
+static void OPNAME ## mpeg4_qpel8_v_lowpass(uint8_t *dst, uint8_t *src, int dstStride, int srcStride){\
+ uint8_t *cm = ff_cropTbl + MAX_NEG_CROP;\
+ int w=8;\
+ do{\
+ uint8_t *s = src, *d=dst;\
+ int src0,src1,src2,src3,src4,src5,src6,src7,src8;\
+ src0 = *s; s+=srcStride; \
+ src1 = *s; s+=srcStride; \
+ src2 = *s; s+=srcStride; \
+ src3 = *s; s+=srcStride; \
+ src4 = *s; s+=srcStride; \
+ OP(*d, (src0+src1)*20 - (src0+src2)*6 + (src1+src3)*3 - (src2+src4));d+=dstStride;\
+ src5 = *s; s+=srcStride; \
+ OP(*d, (src1+src2)*20 - (src0+src3)*6 + (src0+src4)*3 - (src1+src5));d+=dstStride;\
+ src6 = *s; s+=srcStride; \
+ OP(*d, (src2+src3)*20 - (src1+src4)*6 + (src0+src5)*3 - (src0+src6));d+=dstStride;\
+ src7 = *s; s+=srcStride; \
+ OP(*d, (src3+src4)*20 - (src2+src5)*6 + (src1+src6)*3 - (src0+src7));d+=dstStride;\
+ src8 = *s; \
+ OP(*d, (src4+src5)*20 - (src3+src6)*6 + (src2+src7)*3 - (src1+src8));d+=dstStride;\
+ OP(*d, (src5+src6)*20 - (src4+src7)*6 + (src3+src8)*3 - (src2+src8));d+=dstStride;\
+ OP(*d, (src6+src7)*20 - (src5+src8)*6 + (src4+src8)*3 - (src3+src7));d+=dstStride;\
+ OP(*d, (src7+src8)*20 - (src6+src8)*6 + (src5+src7)*3 - (src4+src6));\
+ dst++;\
+ src++;\
+ }while(--w);\
+}\
+\
+static void OPNAME ## mpeg4_qpel16_h_lowpass(uint8_t *dst, uint8_t *src, int dstStride, int srcStride, int h){\
+ uint8_t *cm = ff_cropTbl + MAX_NEG_CROP;\
+ do {\
+ uint8_t *s = src;\
+ int src0,src1,src2,src3,src4,src5,src6,src7,src8;\
+ int src9,src10,src11,src12,src13,src14,src15,src16;\
+ src0= *s++;\
+ src1= *s++;\
+ src2= *s++;\
+ src3= *s++;\
+ src4= *s++;\
+ OP(dst[ 0], (src0 +src1 )*20 - (src0 +src2 )*6 + (src1 +src3 )*3 - (src2 +src4 ));\
+ src5= *s++;\
+ OP(dst[ 1], (src1 +src2 )*20 - (src0 +src3 )*6 + (src0 +src4 )*3 - (src1 +src5 ));\
+ src6= *s++;\
+ OP(dst[ 2], (src2 +src3 )*20 - (src1 +src4 )*6 + (src0 +src5 )*3 - (src0 +src6 ));\
+ src7= *s++;\
+ OP(dst[ 3], (src3 +src4 )*20 - (src2 +src5 )*6 + (src1 +src6 )*3 - (src0 +src7 ));\
+ src8= *s++;\
+ OP(dst[ 4], (src4 +src5 )*20 - (src3 +src6 )*6 + (src2 +src7 )*3 - (src1 +src8 ));\
+ src9= *s++;\
+ OP(dst[ 5], (src5 +src6 )*20 - (src4 +src7 )*6 + (src3 +src8 )*3 - (src2 +src9 ));\
+ src10= *s++;\
+ OP(dst[ 6], (src6 +src7 )*20 - (src5 +src8 )*6 + (src4 +src9 )*3 - (src3 +src10));\
+ src11= *s++;\
+ OP(dst[ 7], (src7 +src8 )*20 - (src6 +src9 )*6 + (src5 +src10)*3 - (src4 +src11));\
+ src12= *s++;\
+ OP(dst[ 8], (src8 +src9 )*20 - (src7 +src10)*6 + (src6 +src11)*3 - (src5 +src12));\
+ src13= *s++;\
+ OP(dst[ 9], (src9 +src10)*20 - (src8 +src11)*6 + (src7 +src12)*3 - (src6 +src13));\
+ src14= *s++;\
+ OP(dst[10], (src10+src11)*20 - (src9 +src12)*6 + (src8 +src13)*3 - (src7 +src14));\
+ src15= *s++;\
+ OP(dst[11], (src11+src12)*20 - (src10+src13)*6 + (src9 +src14)*3 - (src8 +src15));\
+ src16= *s++;\
+ OP(dst[12], (src12+src13)*20 - (src11+src14)*6 + (src10+src15)*3 - (src9 +src16));\
+ OP(dst[13], (src13+src14)*20 - (src12+src15)*6 + (src11+src16)*3 - (src10+src16));\
+ OP(dst[14], (src14+src15)*20 - (src13+src16)*6 + (src12+src16)*3 - (src11+src15));\
+ OP(dst[15], (src15+src16)*20 - (src14+src16)*6 + (src13+src15)*3 - (src12+src14));\
+ dst+=dstStride;\
+ src+=srcStride;\
+ }while(--h);\
+}\
+\
+static void OPNAME ## mpeg4_qpel16_v_lowpass(uint8_t *dst, uint8_t *src, int dstStride, int srcStride){\
+ uint8_t *cm = ff_cropTbl + MAX_NEG_CROP;\
+ int w=16;\
+ do {\
+ uint8_t *s = src, *d=dst;\
+ int src0,src1,src2,src3,src4,src5,src6,src7,src8;\
+ int src9,src10,src11,src12,src13,src14,src15,src16;\
+ src0 = *s; s+=srcStride; \
+ src1 = *s; s+=srcStride; \
+ src2 = *s; s+=srcStride; \
+ src3 = *s; s+=srcStride; \
+ src4 = *s; s+=srcStride; \
+ OP(*d, (src0 +src1 )*20 - (src0 +src2 )*6 + (src1 +src3 )*3 - (src2 +src4 ));d+=dstStride;\
+ src5 = *s; s+=srcStride; \
+ OP(*d, (src1 +src2 )*20 - (src0 +src3 )*6 + (src0 +src4 )*3 - (src1 +src5 ));d+=dstStride;\
+ src6 = *s; s+=srcStride; \
+ OP(*d, (src2 +src3 )*20 - (src1 +src4 )*6 + (src0 +src5 )*3 - (src0 +src6 ));d+=dstStride;\
+ src7 = *s; s+=srcStride; \
+ OP(*d, (src3 +src4 )*20 - (src2 +src5 )*6 + (src1 +src6 )*3 - (src0 +src7 ));d+=dstStride;\
+ src8 = *s; s+=srcStride; \
+ OP(*d, (src4 +src5 )*20 - (src3 +src6 )*6 + (src2 +src7 )*3 - (src1 +src8 ));d+=dstStride;\
+ src9 = *s; s+=srcStride; \
+ OP(*d, (src5 +src6 )*20 - (src4 +src7 )*6 + (src3 +src8 )*3 - (src2 +src9 ));d+=dstStride;\
+ src10 = *s; s+=srcStride; \
+ OP(*d, (src6 +src7 )*20 - (src5 +src8 )*6 + (src4 +src9 )*3 - (src3 +src10));d+=dstStride;\
+ src11 = *s; s+=srcStride; \
+ OP(*d, (src7 +src8 )*20 - (src6 +src9 )*6 + (src5 +src10)*3 - (src4 +src11));d+=dstStride;\
+ src12 = *s; s+=srcStride; \
+ OP(*d, (src8 +src9 )*20 - (src7 +src10)*6 + (src6 +src11)*3 - (src5 +src12));d+=dstStride;\
+ src13 = *s; s+=srcStride; \
+ OP(*d, (src9 +src10)*20 - (src8 +src11)*6 + (src7 +src12)*3 - (src6 +src13));d+=dstStride;\
+ src14 = *s; s+=srcStride; \
+ OP(*d, (src10+src11)*20 - (src9 +src12)*6 + (src8 +src13)*3 - (src7 +src14));d+=dstStride;\
+ src15 = *s; s+=srcStride; \
+ OP(*d, (src11+src12)*20 - (src10+src13)*6 + (src9 +src14)*3 - (src8 +src15));d+=dstStride;\
+ src16 = *s; \
+ OP(*d, (src12+src13)*20 - (src11+src14)*6 + (src10+src15)*3 - (src9 +src16));d+=dstStride;\
+ OP(*d, (src13+src14)*20 - (src12+src15)*6 + (src11+src16)*3 - (src10+src16));d+=dstStride;\
+ OP(*d, (src14+src15)*20 - (src13+src16)*6 + (src12+src16)*3 - (src11+src15));d+=dstStride;\
+ OP(*d, (src15+src16)*20 - (src14+src16)*6 + (src13+src15)*3 - (src12+src14));\
+ dst++;\
+ src++;\
+ }while(--w);\
+}\
+\
+static void OPNAME ## qpel8_mc00_sh4 (uint8_t *dst, uint8_t *src, int stride){\
+ OPNAME ## pixels8_c(dst, src, stride, 8);\
+}\
+\
+static void OPNAME ## qpel8_mc10_sh4(uint8_t *dst, uint8_t *src, int stride){\
+ uint8_t half[64];\
+ put ## RND ## mpeg4_qpel8_h_lowpass(half, src, 8, stride, 8);\
+ OPNAME ## pixels8_l2_aligned2(dst, src, half, stride, stride, 8, 8);\
+}\
+\
+static void OPNAME ## qpel8_mc20_sh4(uint8_t *dst, uint8_t *src, int stride){\
+ OPNAME ## mpeg4_qpel8_h_lowpass(dst, src, stride, stride, 8);\
+}\
+\
+static void OPNAME ## qpel8_mc30_sh4(uint8_t *dst, uint8_t *src, int stride){\
+ uint8_t half[64];\
+ put ## RND ## mpeg4_qpel8_h_lowpass(half, src, 8, stride, 8);\
+ OPNAME ## pixels8_l2_aligned2(dst, src+1, half, stride, stride, 8, 8);\
+}\
+\
+static void OPNAME ## qpel8_mc01_sh4(uint8_t *dst, uint8_t *src, int stride){\
+ uint8_t full[16*9];\
+ uint8_t half[64];\
+ copy_block9(full, src, 16, stride, 9);\
+ put ## RND ## mpeg4_qpel8_v_lowpass(half, full, 8, 16);\
+ OPNAME ## pixels8_l2_aligned(dst, full, half, stride, 16, 8, 8);\
+}\
+\
+static void OPNAME ## qpel8_mc02_sh4(uint8_t *dst, uint8_t *src, int stride){\
+ uint8_t full[16*9];\
+ copy_block9(full, src, 16, stride, 9);\
+ OPNAME ## mpeg4_qpel8_v_lowpass(dst, full, stride, 16);\
+}\
+\
+static void OPNAME ## qpel8_mc03_sh4(uint8_t *dst, uint8_t *src, int stride){\
+ uint8_t full[16*9];\
+ uint8_t half[64];\
+ copy_block9(full, src, 16, stride, 9);\
+ put ## RND ## mpeg4_qpel8_v_lowpass(half, full, 8, 16);\
+ OPNAME ## pixels8_l2_aligned(dst, full+16, half, stride, 16, 8, 8);\
+}\
+static void OPNAME ## qpel8_mc11_sh4(uint8_t *dst, uint8_t *src, int stride){\
+ uint8_t full[16*9];\
+ uint8_t halfH[72];\
+ uint8_t halfHV[64];\
+ copy_block9(full, src, 16, stride, 9);\
+ put ## RND ## mpeg4_qpel8_h_lowpass(halfH, full, 8, 16, 9);\
+ put ## RND ## pixels8_l2_aligned(halfH, halfH, full, 8, 8, 16, 9);\
+ put ## RND ## mpeg4_qpel8_v_lowpass(halfHV, halfH, 8, 8);\
+ OPNAME ## pixels8_l2_aligned(dst, halfH, halfHV, stride, 8, 8, 8);\
+}\
+static void OPNAME ## qpel8_mc31_sh4(uint8_t *dst, uint8_t *src, int stride){\
+ uint8_t full[16*9];\
+ uint8_t halfH[72];\
+ uint8_t halfHV[64];\
+ copy_block9(full, src, 16, stride, 9);\
+ put ## RND ## mpeg4_qpel8_h_lowpass(halfH, full, 8, 16, 9);\
+ put ## RND ## pixels8_l2_aligned1(halfH, halfH, full+1, 8, 8, 16, 9);\
+ put ## RND ## mpeg4_qpel8_v_lowpass(halfHV, halfH, 8, 8);\
+ OPNAME ## pixels8_l2_aligned(dst, halfH, halfHV, stride, 8, 8, 8);\
+}\
+static void OPNAME ## qpel8_mc13_sh4(uint8_t *dst, uint8_t *src, int stride){\
+ uint8_t full[16*9];\
+ uint8_t halfH[72];\
+ uint8_t halfHV[64];\
+ copy_block9(full, src, 16, stride, 9);\
+ put ## RND ## mpeg4_qpel8_h_lowpass(halfH, full, 8, 16, 9);\
+ put ## RND ## pixels8_l2_aligned(halfH, halfH, full, 8, 8, 16, 9);\
+ put ## RND ## mpeg4_qpel8_v_lowpass(halfHV, halfH, 8, 8);\
+ OPNAME ## pixels8_l2_aligned(dst, halfH+8, halfHV, stride, 8, 8, 8);\
+}\
+static void OPNAME ## qpel8_mc33_sh4(uint8_t *dst, uint8_t *src, int stride){\
+ uint8_t full[16*9];\
+ uint8_t halfH[72];\
+ uint8_t halfHV[64];\
+ copy_block9(full, src, 16, stride, 9);\
+ put ## RND ## mpeg4_qpel8_h_lowpass(halfH, full, 8, 16, 9);\
+ put ## RND ## pixels8_l2_aligned1(halfH, halfH, full+1, 8, 8, 16, 9);\
+ put ## RND ## mpeg4_qpel8_v_lowpass(halfHV, halfH, 8, 8);\
+ OPNAME ## pixels8_l2_aligned(dst, halfH+8, halfHV, stride, 8, 8, 8);\
+}\
+static void OPNAME ## qpel8_mc21_sh4(uint8_t *dst, uint8_t *src, int stride){\
+ uint8_t halfH[72];\
+ uint8_t halfHV[64];\
+ put ## RND ## mpeg4_qpel8_h_lowpass(halfH, src, 8, stride, 9);\
+ put ## RND ## mpeg4_qpel8_v_lowpass(halfHV, halfH, 8, 8);\
+ OPNAME ## pixels8_l2_aligned(dst, halfH, halfHV, stride, 8, 8, 8);\
+}\
+static void OPNAME ## qpel8_mc23_sh4(uint8_t *dst, uint8_t *src, int stride){\
+ uint8_t halfH[72];\
+ uint8_t halfHV[64];\
+ put ## RND ## mpeg4_qpel8_h_lowpass(halfH, src, 8, stride, 9);\
+ put ## RND ## mpeg4_qpel8_v_lowpass(halfHV, halfH, 8, 8);\
+ OPNAME ## pixels8_l2_aligned(dst, halfH+8, halfHV, stride, 8, 8, 8);\
+}\
+static void OPNAME ## qpel8_mc12_sh4(uint8_t *dst, uint8_t *src, int stride){\
+ uint8_t full[16*9];\
+ uint8_t halfH[72];\
+ copy_block9(full, src, 16, stride, 9);\
+ put ## RND ## mpeg4_qpel8_h_lowpass(halfH, full, 8, 16, 9);\
+ put ## RND ## pixels8_l2_aligned(halfH, halfH, full, 8, 8, 16, 9);\
+ OPNAME ## mpeg4_qpel8_v_lowpass(dst, halfH, stride, 8);\
+}\
+static void OPNAME ## qpel8_mc32_sh4(uint8_t *dst, uint8_t *src, int stride){\
+ uint8_t full[16*9];\
+ uint8_t halfH[72];\
+ copy_block9(full, src, 16, stride, 9);\
+ put ## RND ## mpeg4_qpel8_h_lowpass(halfH, full, 8, 16, 9);\
+ put ## RND ## pixels8_l2_aligned1(halfH, halfH, full+1, 8, 8, 16, 9);\
+ OPNAME ## mpeg4_qpel8_v_lowpass(dst, halfH, stride, 8);\
+}\
+static void OPNAME ## qpel8_mc22_sh4(uint8_t *dst, uint8_t *src, int stride){\
+ uint8_t halfH[72];\
+ put ## RND ## mpeg4_qpel8_h_lowpass(halfH, src, 8, stride, 9);\
+ OPNAME ## mpeg4_qpel8_v_lowpass(dst, halfH, stride, 8);\
+}\
+static void OPNAME ## qpel16_mc00_sh4 (uint8_t *dst, uint8_t *src, int stride){\
+ OPNAME ## pixels16_c(dst, src, stride, 16);\
+}\
+\
+static void OPNAME ## qpel16_mc10_sh4(uint8_t *dst, uint8_t *src, int stride){\
+ uint8_t half[256];\
+ put ## RND ## mpeg4_qpel16_h_lowpass(half, src, 16, stride, 16);\
+ OPNAME ## pixels16_l2_aligned2(dst, src, half, stride, stride, 16, 16);\
+}\
+\
+static void OPNAME ## qpel16_mc20_sh4(uint8_t *dst, uint8_t *src, int stride){\
+ OPNAME ## mpeg4_qpel16_h_lowpass(dst, src, stride, stride, 16);\
+}\
+\
+static void OPNAME ## qpel16_mc30_sh4(uint8_t *dst, uint8_t *src, int stride){\
+ uint8_t half[256];\
+ put ## RND ## mpeg4_qpel16_h_lowpass(half, src, 16, stride, 16);\
+ OPNAME ## pixels16_l2_aligned2(dst, src+1, half, stride, stride, 16, 16);\
+}\
+\
+static void OPNAME ## qpel16_mc01_sh4(uint8_t *dst, uint8_t *src, int stride){\
+ uint8_t full[24*17];\
+ uint8_t half[256];\
+ copy_block17(full, src, 24, stride, 17);\
+ put ## RND ## mpeg4_qpel16_v_lowpass(half, full, 16, 24);\
+ OPNAME ## pixels16_l2_aligned(dst, full, half, stride, 24, 16, 16);\
+}\
+\
+static void OPNAME ## qpel16_mc02_sh4(uint8_t *dst, uint8_t *src, int stride){\
+ uint8_t full[24*17];\
+ copy_block17(full, src, 24, stride, 17);\
+ OPNAME ## mpeg4_qpel16_v_lowpass(dst, full, stride, 24);\
+}\
+\
+static void OPNAME ## qpel16_mc03_sh4(uint8_t *dst, uint8_t *src, int stride){\
+ uint8_t full[24*17];\
+ uint8_t half[256];\
+ copy_block17(full, src, 24, stride, 17);\
+ put ## RND ## mpeg4_qpel16_v_lowpass(half, full, 16, 24);\
+ OPNAME ## pixels16_l2_aligned(dst, full+24, half, stride, 24, 16, 16);\
+}\
+static void OPNAME ## qpel16_mc11_sh4(uint8_t *dst, uint8_t *src, int stride){\
+ uint8_t full[24*17];\
+ uint8_t halfH[272];\
+ uint8_t halfHV[256];\
+ copy_block17(full, src, 24, stride, 17);\
+ put ## RND ## mpeg4_qpel16_h_lowpass(halfH, full, 16, 24, 17);\
+ put ## RND ## pixels16_l2_aligned(halfH, halfH, full, 16, 16, 24, 17);\
+ put ## RND ## mpeg4_qpel16_v_lowpass(halfHV, halfH, 16, 16);\
+ OPNAME ## pixels16_l2_aligned(dst, halfH, halfHV, stride, 16, 16, 16);\
+}\
+static void OPNAME ## qpel16_mc31_sh4(uint8_t *dst, uint8_t *src, int stride){\
+ uint8_t full[24*17];\
+ uint8_t halfH[272];\
+ uint8_t halfHV[256];\
+ copy_block17(full, src, 24, stride, 17);\
+ put ## RND ## mpeg4_qpel16_h_lowpass(halfH, full, 16, 24, 17);\
+ put ## RND ## pixels16_l2_aligned1(halfH, halfH, full+1, 16, 16, 24, 17);\
+ put ## RND ## mpeg4_qpel16_v_lowpass(halfHV, halfH, 16, 16);\
+ OPNAME ## pixels16_l2_aligned(dst, halfH, halfHV, stride, 16, 16, 16);\
+}\
+static void OPNAME ## qpel16_mc13_sh4(uint8_t *dst, uint8_t *src, int stride){\
+ uint8_t full[24*17];\
+ uint8_t halfH[272];\
+ uint8_t halfHV[256];\
+ copy_block17(full, src, 24, stride, 17);\
+ put ## RND ## mpeg4_qpel16_h_lowpass(halfH, full, 16, 24, 17);\
+ put ## RND ## pixels16_l2_aligned(halfH, halfH, full, 16, 16, 24, 17);\
+ put ## RND ## mpeg4_qpel16_v_lowpass(halfHV, halfH, 16, 16);\
+ OPNAME ## pixels16_l2_aligned(dst, halfH+16, halfHV, stride, 16, 16, 16);\
+}\
+static void OPNAME ## qpel16_mc33_sh4(uint8_t *dst, uint8_t *src, int stride){\
+ uint8_t full[24*17];\
+ uint8_t halfH[272];\
+ uint8_t halfHV[256];\
+ copy_block17(full, src, 24, stride, 17);\
+ put ## RND ## mpeg4_qpel16_h_lowpass(halfH, full, 16, 24, 17);\
+ put ## RND ## pixels16_l2_aligned1(halfH, halfH, full+1, 16, 16, 24, 17);\
+ put ## RND ## mpeg4_qpel16_v_lowpass(halfHV, halfH, 16, 16);\
+ OPNAME ## pixels16_l2_aligned(dst, halfH+16, halfHV, stride, 16, 16, 16);\
+}\
+static void OPNAME ## qpel16_mc21_sh4(uint8_t *dst, uint8_t *src, int stride){\
+ uint8_t halfH[272];\
+ uint8_t halfHV[256];\
+ put ## RND ## mpeg4_qpel16_h_lowpass(halfH, src, 16, stride, 17);\
+ put ## RND ## mpeg4_qpel16_v_lowpass(halfHV, halfH, 16, 16);\
+ OPNAME ## pixels16_l2_aligned(dst, halfH, halfHV, stride, 16, 16, 16);\
+}\
+static void OPNAME ## qpel16_mc23_sh4(uint8_t *dst, uint8_t *src, int stride){\
+ uint8_t halfH[272];\
+ uint8_t halfHV[256];\
+ put ## RND ## mpeg4_qpel16_h_lowpass(halfH, src, 16, stride, 17);\
+ put ## RND ## mpeg4_qpel16_v_lowpass(halfHV, halfH, 16, 16);\
+ OPNAME ## pixels16_l2_aligned(dst, halfH+16, halfHV, stride, 16, 16, 16);\
+}\
+static void OPNAME ## qpel16_mc12_sh4(uint8_t *dst, uint8_t *src, int stride){\
+ uint8_t full[24*17];\
+ uint8_t halfH[272];\
+ copy_block17(full, src, 24, stride, 17);\
+ put ## RND ## mpeg4_qpel16_h_lowpass(halfH, full, 16, 24, 17);\
+ put ## RND ## pixels16_l2_aligned(halfH, halfH, full, 16, 16, 24, 17);\
+ OPNAME ## mpeg4_qpel16_v_lowpass(dst, halfH, stride, 16);\
+}\
+static void OPNAME ## qpel16_mc32_sh4(uint8_t *dst, uint8_t *src, int stride){\
+ uint8_t full[24*17];\
+ uint8_t halfH[272];\
+ copy_block17(full, src, 24, stride, 17);\
+ put ## RND ## mpeg4_qpel16_h_lowpass(halfH, full, 16, 24, 17);\
+ put ## RND ## pixels16_l2_aligned1(halfH, halfH, full+1, 16, 16, 24, 17);\
+ OPNAME ## mpeg4_qpel16_v_lowpass(dst, halfH, stride, 16);\
+}\
+static void OPNAME ## qpel16_mc22_sh4(uint8_t *dst, uint8_t *src, int stride){\
+ uint8_t halfH[272];\
+ put ## RND ## mpeg4_qpel16_h_lowpass(halfH, src, 16, stride, 17);\
+ OPNAME ## mpeg4_qpel16_v_lowpass(dst, halfH, stride, 16);\
+}
+
+#define op_avg(a, b) a = (((a)+cm[((b) + 16)>>5]+1)>>1)
+#define op_avg_no_rnd(a, b) a = (((a)+cm[((b) + 15)>>5])>>1)
+#define op_put(a, b) a = cm[((b) + 16)>>5]
+#define op_put_no_rnd(a, b) a = cm[((b) + 15)>>5]
+
+QPEL_MC(0, put_ , _ , op_put)
+QPEL_MC(1, put_no_rnd_, _no_rnd_, op_put_no_rnd)
+QPEL_MC(0, avg_ , _ , op_avg)
+//QPEL_MC(1, avg_no_rnd , _ , op_avg)
+#undef op_avg
+#undef op_avg_no_rnd
+#undef op_put
+#undef op_put_no_rnd
+
+static void wmv2_mspel8_h_lowpass(uint8_t *dst, uint8_t *src, int dstStride, int srcStride, int h){
+ uint8_t *cm = ff_cropTbl + MAX_NEG_CROP;
+
+ do{
+ int src_1,src0,src1,src2,src3,src4,src5,src6,src7,src8,src9;
+ uint8_t *s = src;
+ src_1 = s[-1];
+ src0 = *s++;
+ src1 = *s++;
+ src2 = *s++;
+ dst[0]= cm[(9*(src0 + src1) - (src_1 + src2) + 8)>>4];
+ src3 = *s++;
+ dst[1]= cm[(9*(src1 + src2) - (src0 + src3) + 8)>>4];
+ src4 = *s++;
+ dst[2]= cm[(9*(src2 + src3) - (src1 + src4) + 8)>>4];
+ src5 = *s++;
+ dst[3]= cm[(9*(src3 + src4) - (src2 + src5) + 8)>>4];
+ src6 = *s++;
+ dst[4]= cm[(9*(src4 + src5) - (src3 + src6) + 8)>>4];
+ src7 = *s++;
+ dst[5]= cm[(9*(src5 + src6) - (src4 + src7) + 8)>>4];
+ src8 = *s++;
+ dst[6]= cm[(9*(src6 + src7) - (src5 + src8) + 8)>>4];
+ src9 = *s++;
+ dst[7]= cm[(9*(src7 + src8) - (src6 + src9) + 8)>>4];
+ dst+=dstStride;
+ src+=srcStride;
+ }while(--h);
+}
+
+static void wmv2_mspel8_v_lowpass(uint8_t *dst, uint8_t *src, int dstStride, int srcStride, int w){
+ uint8_t *cm = ff_cropTbl + MAX_NEG_CROP;
+
+ do{
+ int src_1,src0,src1,src2,src3,src4,src5,src6,src7,src8,src9;
+ uint8_t *s = src,*d = dst;
+ src_1 = *(s-srcStride);
+ src0 = *s; s+=srcStride;
+ src1 = *s; s+=srcStride;
+ src2 = *s; s+=srcStride;
+ *d= cm[(9*(src0 + src1) - (src_1 + src2) + 8)>>4]; d+=dstStride;
+ src3 = *s; s+=srcStride;
+ *d= cm[(9*(src1 + src2) - (src0 + src3) + 8)>>4]; d+=dstStride;
+ src4 = *s; s+=srcStride;
+ *d= cm[(9*(src2 + src3) - (src1 + src4) + 8)>>4]; d+=dstStride;
+ src5 = *s; s+=srcStride;
+ *d= cm[(9*(src3 + src4) - (src2 + src5) + 8)>>4]; d+=dstStride;
+ src6 = *s; s+=srcStride;
+ *d= cm[(9*(src4 + src5) - (src3 + src6) + 8)>>4]; d+=dstStride;
+ src7 = *s; s+=srcStride;
+ *d= cm[(9*(src5 + src6) - (src4 + src7) + 8)>>4]; d+=dstStride;
+ src8 = *s; s+=srcStride;
+ *d= cm[(9*(src6 + src7) - (src5 + src8) + 8)>>4]; d+=dstStride;
+ src9 = *s;
+ *d= cm[(9*(src7 + src8) - (src6 + src9) + 8)>>4]; d+=dstStride;
+ src++;
+ dst++;
+ }while(--w);
+}
+
+static void put_mspel8_mc00_sh4 (uint8_t *dst, uint8_t *src, int stride){
+ put_pixels8_c(dst, src, stride, 8);
+}
+
+static void put_mspel8_mc10_sh4(uint8_t *dst, uint8_t *src, int stride){
+ uint8_t half[64];
+ wmv2_mspel8_h_lowpass(half, src, 8, stride, 8);
+ put_pixels8_l2_aligned2(dst, src, half, stride, stride, 8, 8);
+}
+
+static void put_mspel8_mc20_sh4(uint8_t *dst, uint8_t *src, int stride){
+ wmv2_mspel8_h_lowpass(dst, src, stride, stride, 8);
+}
+
+static void put_mspel8_mc30_sh4(uint8_t *dst, uint8_t *src, int stride){
+ uint8_t half[64];
+ wmv2_mspel8_h_lowpass(half, src, 8, stride, 8);
+ put_pixels8_l2_aligned2(dst, src+1, half, stride, stride, 8, 8);
+}
+
+static void put_mspel8_mc02_sh4(uint8_t *dst, uint8_t *src, int stride){
+ wmv2_mspel8_v_lowpass(dst, src, stride, stride, 8);
+}
+
+static void put_mspel8_mc12_sh4(uint8_t *dst, uint8_t *src, int stride){
+ uint8_t halfH[88];
+ uint8_t halfV[64];
+ uint8_t halfHV[64];
+ wmv2_mspel8_h_lowpass(halfH, src-stride, 8, stride, 11);
+ wmv2_mspel8_v_lowpass(halfV, src, 8, stride, 8);
+ wmv2_mspel8_v_lowpass(halfHV, halfH+8, 8, 8, 8);
+ put_pixels8_l2_aligned(dst, halfV, halfHV, stride, 8, 8, 8);
+}
+static void put_mspel8_mc32_sh4(uint8_t *dst, uint8_t *src, int stride){
+ uint8_t halfH[88];
+ uint8_t halfV[64];
+ uint8_t halfHV[64];
+ wmv2_mspel8_h_lowpass(halfH, src-stride, 8, stride, 11);
+ wmv2_mspel8_v_lowpass(halfV, src+1, 8, stride, 8);
+ wmv2_mspel8_v_lowpass(halfHV, halfH+8, 8, 8, 8);
+ put_pixels8_l2_aligned(dst, halfV, halfHV, stride, 8, 8, 8);
+}
+static void put_mspel8_mc22_sh4(uint8_t *dst, uint8_t *src, int stride){
+ uint8_t halfH[88];
+ wmv2_mspel8_h_lowpass(halfH, src-stride, 8, stride, 11);
+ wmv2_mspel8_v_lowpass(dst, halfH+8, stride, 8, 8);
+}
diff --git a/ffmpeg/libavcodec/sh4/sh4.h b/ffmpeg/libavcodec/sh4/sh4.h
new file mode 100644
index 0000000..5d46540
--- /dev/null
+++ b/ffmpeg/libavcodec/sh4/sh4.h
@@ -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 */
generated by cgit v1.2.3 (git 2.39.1) at 2025-12-06 16:20:47 +0000