FFmpeg: libavcodec/sh4/dsputil

00001 /*
00002  * aligned/packed access motion
00003  *
00004  * Copyright (c) 2001-2003 BERO <bero@geocities.co.jp>
00005  *
00006  * This file is part of FFmpeg.
00007  *
00008  * FFmpeg is free software; you can redistribute it and/or
00009  * modify it under the terms of the GNU Lesser General Public
00010  * License as published by the Free Software Foundation; either
00011  * version 2.1 of the License, or (at your option) any later version.
00012  *
00013  * FFmpeg is distributed in the hope that it will be useful,
00014  * but WITHOUT ANY WARRANTY; without even the implied warranty of
00015  * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
00016  * Lesser General Public License for more details.
00017  *
00018  * You should have received a copy of the GNU Lesser General Public
00019  * License along with FFmpeg; if not, write to the Free Software
00020  * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
00021  */
00022 
00023 
00024 #include "libavcodec/avcodec.h"
00025 #include "libavcodec/dsputil.h"
00026 #include "dsputil_sh4.h"
00027 
00028 
00029 #define         LP(p)           *(uint32_t*)(p)
00030 #define         LPC(p)          *(const uint32_t*)(p)
00031 
00032 
00033 #define         UNPACK(ph,pl,tt0,tt1) do { \
00034         uint32_t t0,t1; t0=tt0;t1=tt1; \
00035         ph = ( (t0 & ~BYTE_VEC32(0x03))>>2) + ( (t1 & ~BYTE_VEC32(0x03))>>2); \
00036         pl = (t0 & BYTE_VEC32(0x03)) + (t1 & BYTE_VEC32(0x03)); } while(0)
00037 
00038 #define         rnd_PACK(ph,pl,nph,npl) ph + nph + (((pl + npl + BYTE_VEC32(0x02))>>2) & BYTE_VEC32(0x03))
00039 #define         no_rnd_PACK(ph,pl,nph,npl)      ph + nph + (((pl + npl + BYTE_VEC32(0x01))>>2) & BYTE_VEC32(0x03))
00040 
00041 /* little endian */
00042 #define         MERGE1(a,b,ofs) (ofs==0)?a:( ((a)>>(8*ofs))|((b)<<(32-8*ofs)) )
00043 #define         MERGE2(a,b,ofs) (ofs==3)?b:( ((a)>>(8*(ofs+1)))|((b)<<(32-8*(ofs+1))) )
00044 /* big
00045 #define         MERGE1(a,b,ofs) (ofs==0)?a:( ((a)<<(8*ofs))|((b)>>(32-8*ofs)) )
00046 #define         MERGE2(a,b,ofs) (ofs==3)?b:( ((a)<<(8+8*ofs))|((b)>>(32-8-8*ofs)) )
00047 */
00048 
00049 
00050 #define         put(d,s)        d = s
00051 #define         avg(d,s)        d = rnd_avg32(s,d)
00052 
00053 #define         OP_C4(ofs) \
00054         ref-=ofs; \
00055         do { \
00056                 OP(LP(dest),MERGE1(LPC(ref),LPC(ref+4),ofs)); \
00057                 ref+=stride; \
00058                 dest+=stride; \
00059         } while(--height)
00060 
00061 #define        OP_C40() \
00062         do { \
00063                 OP(LP(dest),LPC(ref)); \
00064                 ref+=stride; \
00065                 dest+=stride; \
00066         } while(--height)
00067 
00068 
00069 #define         OP      put
00070 
00071 static void put_pixels4_c(uint8_t *dest,const uint8_t *ref, const int stride,int height)
00072 {
00073         switch((int)ref&3){
00074         case 0: OP_C40(); return;
00075         case 1: OP_C4(1); return;
00076         case 2: OP_C4(2); return;
00077         case 3: OP_C4(3); return;
00078         }
00079 }
00080 
00081 #undef          OP
00082 #define         OP      avg
00083 
00084 static void avg_pixels4_c(uint8_t *dest,const uint8_t *ref, const int stride,int height)
00085 {
00086         switch((int)ref&3){
00087         case 0: OP_C40(); return;
00088         case 1: OP_C4(1); return;
00089         case 2: OP_C4(2); return;
00090         case 3: OP_C4(3); return;
00091         }
00092 }
00093 
00094 #undef          OP
00095 
00096 #define         OP_C(ofs,sz,avg2) \
00097 { \
00098         ref-=ofs; \
00099         do { \
00100                 uint32_t        t0,t1; \
00101                 t0 = LPC(ref+0); \
00102                 t1 = LPC(ref+4); \
00103                 OP(LP(dest+0), MERGE1(t0,t1,ofs)); \
00104                 t0 = LPC(ref+8); \
00105                 OP(LP(dest+4), MERGE1(t1,t0,ofs)); \
00106 if (sz==16) { \
00107                 t1 = LPC(ref+12); \
00108                 OP(LP(dest+8), MERGE1(t0,t1,ofs)); \
00109                 t0 = LPC(ref+16); \
00110                 OP(LP(dest+12), MERGE1(t1,t0,ofs)); \
00111 } \
00112                 ref+=stride; \
00113                 dest+= stride; \
00114         } while(--height); \
00115 }
00116 
00117 /* aligned */
00118 #define         OP_C0(sz,avg2) \
00119 { \
00120         do { \
00121                 OP(LP(dest+0), LPC(ref+0)); \
00122                 OP(LP(dest+4), LPC(ref+4)); \
00123 if (sz==16) { \
00124                 OP(LP(dest+8), LPC(ref+8)); \
00125                 OP(LP(dest+12), LPC(ref+12)); \
00126 } \
00127                 ref+=stride; \
00128                 dest+= stride; \
00129         } while(--height); \
00130 }
00131 
00132 #define         OP_X(ofs,sz,avg2) \
00133 { \
00134         ref-=ofs; \
00135         do { \
00136                 uint32_t        t0,t1; \
00137                 t0 = LPC(ref+0); \
00138                 t1 = LPC(ref+4); \
00139                 OP(LP(dest+0), avg2(MERGE1(t0,t1,ofs),MERGE2(t0,t1,ofs))); \
00140                 t0 = LPC(ref+8); \
00141                 OP(LP(dest+4), avg2(MERGE1(t1,t0,ofs),MERGE2(t1,t0,ofs))); \
00142 if (sz==16) { \
00143                 t1 = LPC(ref+12); \
00144                 OP(LP(dest+8), avg2(MERGE1(t0,t1,ofs),MERGE2(t0,t1,ofs))); \
00145                 t0 = LPC(ref+16); \
00146                 OP(LP(dest+12), avg2(MERGE1(t1,t0,ofs),MERGE2(t1,t0,ofs))); \
00147 } \
00148                 ref+=stride; \
00149                 dest+= stride; \
00150         } while(--height); \
00151 }
00152 
00153 /* aligned */
00154 #define         OP_Y0(sz,avg2) \
00155 { \
00156         uint32_t t0,t1,t2,t3,t; \
00157 \
00158         t0 = LPC(ref+0); \
00159         t1 = LPC(ref+4); \
00160 if (sz==16) { \
00161         t2 = LPC(ref+8); \
00162         t3 = LPC(ref+12); \
00163 } \
00164         do { \
00165                 ref += stride; \
00166 \
00167                 t = LPC(ref+0); \
00168                 OP(LP(dest+0), avg2(t0,t)); t0 = t; \
00169                 t = LPC(ref+4); \
00170                 OP(LP(dest+4), avg2(t1,t)); t1 = t; \
00171 if (sz==16) { \
00172                 t = LPC(ref+8); \
00173                 OP(LP(dest+8), avg2(t2,t)); t2 = t; \
00174                 t = LPC(ref+12); \
00175                 OP(LP(dest+12), avg2(t3,t)); t3 = t; \
00176 } \
00177                 dest+= stride; \
00178         } while(--height); \
00179 }
00180 
00181 #define         OP_Y(ofs,sz,avg2) \
00182 { \
00183         uint32_t t0,t1,t2,t3,t,w0,w1; \
00184 \
00185         ref-=ofs; \
00186         w0 = LPC(ref+0); \
00187         w1 = LPC(ref+4); \
00188         t0 = MERGE1(w0,w1,ofs); \
00189         w0 = LPC(ref+8); \
00190         t1 = MERGE1(w1,w0,ofs); \
00191 if (sz==16) { \
00192         w1 = LPC(ref+12); \
00193         t2 = MERGE1(w0,w1,ofs); \
00194         w0 = LPC(ref+16); \
00195         t3 = MERGE1(w1,w0,ofs); \
00196 } \
00197         do { \
00198                 ref += stride; \
00199 \
00200                 w0 = LPC(ref+0); \
00201                 w1 = LPC(ref+4); \
00202                 t = MERGE1(w0,w1,ofs); \
00203                 OP(LP(dest+0), avg2(t0,t)); t0 = t; \
00204                 w0 = LPC(ref+8); \
00205                 t = MERGE1(w1,w0,ofs); \
00206                 OP(LP(dest+4), avg2(t1,t)); t1 = t; \
00207 if (sz==16) { \
00208                 w1 = LPC(ref+12); \
00209                 t = MERGE1(w0,w1,ofs); \
00210                 OP(LP(dest+8), avg2(t2,t)); t2 = t; \
00211                 w0 = LPC(ref+16); \
00212                 t = MERGE1(w1,w0,ofs); \
00213                 OP(LP(dest+12), avg2(t3,t)); t3 = t; \
00214 } \
00215                 dest+=stride; \
00216         } while(--height); \
00217 }
00218 
00219 #define OP_X0(sz,avg2) OP_X(0,sz,avg2)
00220 #define OP_XY0(sz,PACK) OP_XY(0,sz,PACK)
00221 #define         OP_XY(ofs,sz,PACK) \
00222 { \
00223         uint32_t        t2,t3,w0,w1; \
00224         uint32_t        a0,a1,a2,a3,a4,a5,a6,a7; \
00225 \
00226         ref -= ofs; \
00227         w0 = LPC(ref+0); \
00228         w1 = LPC(ref+4); \
00229         UNPACK(a0,a1,MERGE1(w0,w1,ofs),MERGE2(w0,w1,ofs)); \
00230         w0 = LPC(ref+8); \
00231         UNPACK(a2,a3,MERGE1(w1,w0,ofs),MERGE2(w1,w0,ofs)); \
00232 if (sz==16) { \
00233         w1 = LPC(ref+12); \
00234         UNPACK(a4,a5,MERGE1(w0,w1,ofs),MERGE2(w0,w1,ofs)); \
00235         w0 = LPC(ref+16); \
00236         UNPACK(a6,a7,MERGE1(w1,w0,ofs),MERGE2(w1,w0,ofs)); \
00237 } \
00238         do { \
00239                 ref+=stride; \
00240                 w0 = LPC(ref+0); \
00241                 w1 = LPC(ref+4); \
00242                 UNPACK(t2,t3,MERGE1(w0,w1,ofs),MERGE2(w0,w1,ofs)); \
00243                 OP(LP(dest+0),PACK(a0,a1,t2,t3)); \
00244                 a0 = t2; a1 = t3; \
00245                 w0 = LPC(ref+8); \
00246                 UNPACK(t2,t3,MERGE1(w1,w0,ofs),MERGE2(w1,w0,ofs)); \
00247                 OP(LP(dest+4),PACK(a2,a3,t2,t3)); \
00248                 a2 = t2; a3 = t3; \
00249 if (sz==16) { \
00250                 w1 = LPC(ref+12); \
00251                 UNPACK(t2,t3,MERGE1(w0,w1,ofs),MERGE2(w0,w1,ofs)); \
00252                 OP(LP(dest+8),PACK(a4,a5,t2,t3)); \
00253                 a4 = t2; a5 = t3; \
00254                 w0 = LPC(ref+16); \
00255                 UNPACK(t2,t3,MERGE1(w1,w0,ofs),MERGE2(w1,w0,ofs)); \
00256                 OP(LP(dest+12),PACK(a6,a7,t2,t3)); \
00257                 a6 = t2; a7 = t3; \
00258 } \
00259                 dest+=stride; \
00260         } while(--height); \
00261 }
00262 
00263 #define         DEFFUNC(op,rnd,xy,sz,OP_N,avgfunc) \
00264 static void op##_##rnd##_pixels##sz##_##xy (uint8_t * dest, const uint8_t * ref, \
00265                                 const int stride, int height) \
00266 { \
00267         switch((int)ref&3) { \
00268         case 0:OP_N##0(sz,rnd##_##avgfunc); return; \
00269         case 1:OP_N(1,sz,rnd##_##avgfunc); return; \
00270         case 2:OP_N(2,sz,rnd##_##avgfunc); return; \
00271         case 3:OP_N(3,sz,rnd##_##avgfunc); return; \
00272         } \
00273 }
00274 
00275 #define OP put
00276 
00277 DEFFUNC(put,   rnd,o,8,OP_C,avg32)
00278 DEFFUNC(put,   rnd,x,8,OP_X,avg32)
00279 DEFFUNC(put,no_rnd,x,8,OP_X,avg32)
00280 DEFFUNC(put,   rnd,y,8,OP_Y,avg32)
00281 DEFFUNC(put,no_rnd,y,8,OP_Y,avg32)
00282 DEFFUNC(put,   rnd,xy,8,OP_XY,PACK)
00283 DEFFUNC(put,no_rnd,xy,8,OP_XY,PACK)
00284 DEFFUNC(put,   rnd,o,16,OP_C,avg32)
00285 DEFFUNC(put,   rnd,x,16,OP_X,avg32)
00286 DEFFUNC(put,no_rnd,x,16,OP_X,avg32)
00287 DEFFUNC(put,   rnd,y,16,OP_Y,avg32)
00288 DEFFUNC(put,no_rnd,y,16,OP_Y,avg32)
00289 DEFFUNC(put,   rnd,xy,16,OP_XY,PACK)
00290 DEFFUNC(put,no_rnd,xy,16,OP_XY,PACK)
00291 
00292 #undef OP
00293 #define OP avg
00294 
00295 DEFFUNC(avg,   rnd,o,8,OP_C,avg32)
00296 DEFFUNC(avg,   rnd,x,8,OP_X,avg32)
00297 DEFFUNC(avg,no_rnd,x,8,OP_X,avg32)
00298 DEFFUNC(avg,   rnd,y,8,OP_Y,avg32)
00299 DEFFUNC(avg,no_rnd,y,8,OP_Y,avg32)
00300 DEFFUNC(avg,   rnd,xy,8,OP_XY,PACK)
00301 DEFFUNC(avg,no_rnd,xy,8,OP_XY,PACK)
00302 DEFFUNC(avg,   rnd,o,16,OP_C,avg32)
00303 DEFFUNC(avg,   rnd,x,16,OP_X,avg32)
00304 DEFFUNC(avg,no_rnd,x,16,OP_X,avg32)
00305 DEFFUNC(avg,   rnd,y,16,OP_Y,avg32)
00306 DEFFUNC(avg,no_rnd,y,16,OP_Y,avg32)
00307 DEFFUNC(avg,   rnd,xy,16,OP_XY,PACK)
00308 DEFFUNC(avg,no_rnd,xy,16,OP_XY,PACK)
00309 
00310 #undef OP
00311 
00312 #define         put_no_rnd_pixels8_o     put_rnd_pixels8_o
00313 #define         put_no_rnd_pixels16_o    put_rnd_pixels16_o
00314 #define         avg_no_rnd_pixels8_o     avg_rnd_pixels8_o
00315 #define         avg_no_rnd_pixels16_o    avg_rnd_pixels16_o
00316 
00317 #define         put_pixels8_c            put_rnd_pixels8_o
00318 #define         put_pixels16_c           put_rnd_pixels16_o
00319 #define         avg_pixels8_c            avg_rnd_pixels8_o
00320 #define         avg_pixels16_c           avg_rnd_pixels16_o
00321 #define         put_no_rnd_pixels8_c     put_rnd_pixels8_o
00322 #define         put_no_rnd_pixels16_c    put_rnd_pixels16_o
00323 #define         avg_no_rnd_pixels8_c     avg_rnd_pixels8_o
00324 #define         avg_no_rnd_pixels16_c    avg_rnd_pixels16_o
00325 
00326 #define         QPEL
00327 
00328 #ifdef QPEL
00329 
00330 #include "qpel.c"
00331 
00332 #endif
00333 
00334 void dsputil_init_align(DSPContext* c, AVCodecContext *avctx)
00335 {
00336         const int high_bit_depth = avctx->codec_id == CODEC_ID_H264 && avctx->bits_per_raw_sample > 8;
00337 
00338         if (!high_bit_depth) {
00339         c->put_pixels_tab[0][0] = put_rnd_pixels16_o;
00340         c->put_pixels_tab[0][1] = put_rnd_pixels16_x;
00341         c->put_pixels_tab[0][2] = put_rnd_pixels16_y;
00342         c->put_pixels_tab[0][3] = put_rnd_pixels16_xy;
00343         c->put_pixels_tab[1][0] = put_rnd_pixels8_o;
00344         c->put_pixels_tab[1][1] = put_rnd_pixels8_x;
00345         c->put_pixels_tab[1][2] = put_rnd_pixels8_y;
00346         c->put_pixels_tab[1][3] = put_rnd_pixels8_xy;
00347 
00348         c->put_no_rnd_pixels_tab[0][0] = put_no_rnd_pixels16_o;
00349         c->put_no_rnd_pixels_tab[0][1] = put_no_rnd_pixels16_x;
00350         c->put_no_rnd_pixels_tab[0][2] = put_no_rnd_pixels16_y;
00351         c->put_no_rnd_pixels_tab[0][3] = put_no_rnd_pixels16_xy;
00352         c->put_no_rnd_pixels_tab[1][0] = put_no_rnd_pixels8_o;
00353         c->put_no_rnd_pixels_tab[1][1] = put_no_rnd_pixels8_x;
00354         c->put_no_rnd_pixels_tab[1][2] = put_no_rnd_pixels8_y;
00355         c->put_no_rnd_pixels_tab[1][3] = put_no_rnd_pixels8_xy;
00356 
00357         c->avg_pixels_tab[0][0] = avg_rnd_pixels16_o;
00358         c->avg_pixels_tab[0][1] = avg_rnd_pixels16_x;
00359         c->avg_pixels_tab[0][2] = avg_rnd_pixels16_y;
00360         c->avg_pixels_tab[0][3] = avg_rnd_pixels16_xy;
00361         c->avg_pixels_tab[1][0] = avg_rnd_pixels8_o;
00362         c->avg_pixels_tab[1][1] = avg_rnd_pixels8_x;
00363         c->avg_pixels_tab[1][2] = avg_rnd_pixels8_y;
00364         c->avg_pixels_tab[1][3] = avg_rnd_pixels8_xy;
00365 
00366         c->avg_no_rnd_pixels_tab[0][0] = avg_no_rnd_pixels16_o;
00367         c->avg_no_rnd_pixels_tab[0][1] = avg_no_rnd_pixels16_x;
00368         c->avg_no_rnd_pixels_tab[0][2] = avg_no_rnd_pixels16_y;
00369         c->avg_no_rnd_pixels_tab[0][3] = avg_no_rnd_pixels16_xy;
00370         c->avg_no_rnd_pixels_tab[1][0] = avg_no_rnd_pixels8_o;
00371         c->avg_no_rnd_pixels_tab[1][1] = avg_no_rnd_pixels8_x;
00372         c->avg_no_rnd_pixels_tab[1][2] = avg_no_rnd_pixels8_y;
00373         c->avg_no_rnd_pixels_tab[1][3] = avg_no_rnd_pixels8_xy;
00374         }
00375 
00376 #ifdef QPEL
00377 
00378 #define dspfunc(PFX, IDX, NUM) \
00379     c->PFX ## _pixels_tab[IDX][ 0] = PFX ## NUM ## _mc00_sh4; \
00380     c->PFX ## _pixels_tab[IDX][ 1] = PFX ## NUM ## _mc10_sh4; \
00381     c->PFX ## _pixels_tab[IDX][ 2] = PFX ## NUM ## _mc20_sh4; \
00382     c->PFX ## _pixels_tab[IDX][ 3] = PFX ## NUM ## _mc30_sh4; \
00383     c->PFX ## _pixels_tab[IDX][ 4] = PFX ## NUM ## _mc01_sh4; \
00384     c->PFX ## _pixels_tab[IDX][ 5] = PFX ## NUM ## _mc11_sh4; \
00385     c->PFX ## _pixels_tab[IDX][ 6] = PFX ## NUM ## _mc21_sh4; \
00386     c->PFX ## _pixels_tab[IDX][ 7] = PFX ## NUM ## _mc31_sh4; \
00387     c->PFX ## _pixels_tab[IDX][ 8] = PFX ## NUM ## _mc02_sh4; \
00388     c->PFX ## _pixels_tab[IDX][ 9] = PFX ## NUM ## _mc12_sh4; \
00389     c->PFX ## _pixels_tab[IDX][10] = PFX ## NUM ## _mc22_sh4; \
00390     c->PFX ## _pixels_tab[IDX][11] = PFX ## NUM ## _mc32_sh4; \
00391     c->PFX ## _pixels_tab[IDX][12] = PFX ## NUM ## _mc03_sh4; \
00392     c->PFX ## _pixels_tab[IDX][13] = PFX ## NUM ## _mc13_sh4; \
00393     c->PFX ## _pixels_tab[IDX][14] = PFX ## NUM ## _mc23_sh4; \
00394     c->PFX ## _pixels_tab[IDX][15] = PFX ## NUM ## _mc33_sh4
00395 
00396     dspfunc(put_qpel, 0, 16);
00397     dspfunc(put_no_rnd_qpel, 0, 16);
00398 
00399     dspfunc(avg_qpel, 0, 16);
00400     /* dspfunc(avg_no_rnd_qpel, 0, 16); */
00401 
00402     dspfunc(put_qpel, 1, 8);
00403     dspfunc(put_no_rnd_qpel, 1, 8);
00404 
00405     dspfunc(avg_qpel, 1, 8);
00406     /* dspfunc(avg_no_rnd_qpel, 1, 8); */
00407 
00408     if (!high_bit_depth) {
00409     dspfunc(put_h264_qpel, 0, 16);
00410     dspfunc(put_h264_qpel, 1, 8);
00411     dspfunc(put_h264_qpel, 2, 4);
00412     dspfunc(avg_h264_qpel, 0, 16);
00413     dspfunc(avg_h264_qpel, 1, 8);
00414     dspfunc(avg_h264_qpel, 2, 4);
00415     }
00416 
00417 #undef dspfunc
00418     if (!high_bit_depth) {
00419     c->put_h264_chroma_pixels_tab[0]= put_h264_chroma_mc8_sh4;
00420     c->put_h264_chroma_pixels_tab[1]= put_h264_chroma_mc4_sh4;
00421     c->put_h264_chroma_pixels_tab[2]= put_h264_chroma_mc2_sh4;
00422     c->avg_h264_chroma_pixels_tab[0]= avg_h264_chroma_mc8_sh4;
00423     c->avg_h264_chroma_pixels_tab[1]= avg_h264_chroma_mc4_sh4;
00424     c->avg_h264_chroma_pixels_tab[2]= avg_h264_chroma_mc2_sh4;
00425     }
00426 
00427     c->put_mspel_pixels_tab[0]= put_mspel8_mc00_sh4;
00428     c->put_mspel_pixels_tab[1]= put_mspel8_mc10_sh4;
00429     c->put_mspel_pixels_tab[2]= put_mspel8_mc20_sh4;
00430     c->put_mspel_pixels_tab[3]= put_mspel8_mc30_sh4;
00431     c->put_mspel_pixels_tab[4]= put_mspel8_mc02_sh4;
00432     c->put_mspel_pixels_tab[5]= put_mspel8_mc12_sh4;
00433     c->put_mspel_pixels_tab[6]= put_mspel8_mc22_sh4;
00434     c->put_mspel_pixels_tab[7]= put_mspel8_mc32_sh4;
00435 
00436     c->gmc1 = gmc1_c;
00437     c->gmc = gmc_c;
00438 
00439 #endif
00440 }
libavcodec/sh4/dsputil_align.c
