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