2 * aligned/packed access motion
4 * Copyright (c) 2001-2003 BERO <bero@geocities.co.jp>
6 * This library is free software; you can redistribute it and/or
7 * modify it under the terms of the GNU Lesser General Public
8 * License as published by the Free Software Foundation; either
9 * version 2 of the License, or (at your option) any later version.
11 * This library is distributed in the hope that it will be useful,
12 * but WITHOUT ANY WARRANTY; without even the implied warranty of
13 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
14 * Lesser General Public License for more details.
16 * You should have received a copy of the GNU Lesser General Public
17 * License along with this library; if not, write to the Free Software
18 * Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA
22 #include "../avcodec.h"
23 #include "../dsputil.h"
26 #define LP(p) *(uint32_t*)(p)
29 #define UNPACK(ph,pl,tt0,tt1) do { \
30 uint32_t t0,t1; t0=tt0;t1=tt1; \
31 ph = ( (t0 & ~BYTE_VEC32(0x03))>>2) + ( (t1 & ~BYTE_VEC32(0x03))>>2); \
32 pl = (t0 & BYTE_VEC32(0x03)) + (t1 & BYTE_VEC32(0x03)); } while(0)
34 #define rnd_PACK(ph,pl,nph,npl) ph + nph + (((pl + npl + BYTE_VEC32(0x02))>>2) & BYTE_VEC32(0x03))
35 #define no_rnd_PACK(ph,pl,nph,npl) ph + nph + (((pl + npl + BYTE_VEC32(0x01))>>2) & BYTE_VEC32(0x03))
38 #define MERGE1(a,b,ofs) (ofs==0)?a:( ((a)>>(8*ofs))|((b)<<(32-8*ofs)) )
39 #define MERGE2(a,b,ofs) (ofs==3)?b:( ((a)>>(8*(ofs+1)))|((b)<<(32-8*(ofs+1))) )
41 #define MERGE1(a,b,ofs) (ofs==0)?a:( ((a)<<(8*ofs))|((b)>>(32-8*ofs)) )
42 #define MERGE2(a,b,ofs) (ofs==3)?b:( ((a)<<(8+8*ofs))|((b)>>(32-8-8*ofs)) )
46 #define put(d,s) d = s
47 #define avg(d,s) d = rnd_avg32(s,d)
52 OP(LP(dest),MERGE1(LP(ref),LP(ref+4),ofs)); \
59 OP(LP(dest),LP(ref)); \
67 static void put_pixels4_c(uint8_t *dest,const uint8_t *ref, const int stride,int height)
70 case 0: OP_C40(); return;
71 case 1: OP_C4(1); return;
72 case 2: OP_C4(2); return;
73 case 3: OP_C4(3); return;
80 static void avg_pixels4_c(uint8_t *dest,const uint8_t *ref, const int stride,int height)
83 case 0: OP_C40(); return;
84 case 1: OP_C4(1); return;
85 case 2: OP_C4(2); return;
86 case 3: OP_C4(3); return;
92 #define OP_C(ofs,sz,avg2) \
99 OP(LP(dest+0), MERGE1(t0,t1,ofs)); \
101 OP(LP(dest+4), MERGE1(t1,t0,ofs)); \
104 OP(LP(dest+8), MERGE1(t0,t1,ofs)); \
106 OP(LP(dest+12), MERGE1(t1,t0,ofs)); \
114 #define OP_C0(sz,avg2) \
117 OP(LP(dest+0), LP(ref+0)); \
118 OP(LP(dest+4), LP(ref+4)); \
120 OP(LP(dest+8), LP(ref+8)); \
121 OP(LP(dest+12), LP(ref+12)); \
128 #define OP_X(ofs,sz,avg2) \
135 OP(LP(dest+0), avg2(MERGE1(t0,t1,ofs),MERGE2(t0,t1,ofs))); \
137 OP(LP(dest+4), avg2(MERGE1(t1,t0,ofs),MERGE2(t1,t0,ofs))); \
140 OP(LP(dest+8), avg2(MERGE1(t0,t1,ofs),MERGE2(t0,t1,ofs))); \
142 OP(LP(dest+12), avg2(MERGE1(t1,t0,ofs),MERGE2(t1,t0,ofs))); \
150 #define OP_Y0(sz,avg2) \
152 uint32_t t0,t1,t2,t3,t; \
164 OP(LP(dest+0), avg2(t0,t)); t0 = t; \
166 OP(LP(dest+4), avg2(t1,t)); t1 = t; \
169 OP(LP(dest+8), avg2(t2,t)); t2 = t; \
171 OP(LP(dest+12), avg2(t3,t)); t3 = t; \
177 #define OP_Y(ofs,sz,avg2) \
179 uint32_t t0,t1,t2,t3,t,w0,w1; \
184 t0 = MERGE1(w0,w1,ofs); \
186 t1 = MERGE1(w1,w0,ofs); \
189 t2 = MERGE1(w0,w1,ofs); \
191 t3 = MERGE1(w1,w0,ofs); \
198 t = MERGE1(w0,w1,ofs); \
199 OP(LP(dest+0), avg2(t0,t)); t0 = t; \
201 t = MERGE1(w1,w0,ofs); \
202 OP(LP(dest+4), avg2(t1,t)); t1 = t; \
205 t = MERGE1(w0,w1,ofs); \
206 OP(LP(dest+8), avg2(t2,t)); t2 = t; \
208 t = MERGE1(w1,w0,ofs); \
209 OP(LP(dest+12), avg2(t3,t)); t3 = t; \
215 #define OP_X0(sz,avg2) OP_X(0,sz,avg2)
216 #define OP_XY0(sz,PACK) OP_XY(0,sz,PACK)
217 #define OP_XY(ofs,sz,PACK) \
219 uint32_t t2,t3,w0,w1; \
220 uint32_t a0,a1,a2,a3,a4,a5,a6,a7; \
225 UNPACK(a0,a1,MERGE1(w0,w1,ofs),MERGE2(w0,w1,ofs)); \
227 UNPACK(a2,a3,MERGE1(w1,w0,ofs),MERGE2(w1,w0,ofs)); \
230 UNPACK(a4,a5,MERGE1(w0,w1,ofs),MERGE2(w0,w1,ofs)); \
232 UNPACK(a6,a7,MERGE1(w1,w0,ofs),MERGE2(w1,w0,ofs)); \
238 UNPACK(t2,t3,MERGE1(w0,w1,ofs),MERGE2(w0,w1,ofs)); \
239 OP(LP(dest+0),PACK(a0,a1,t2,t3)); \
242 UNPACK(t2,t3,MERGE1(w1,w0,ofs),MERGE2(w1,w0,ofs)); \
243 OP(LP(dest+4),PACK(a2,a3,t2,t3)); \
247 UNPACK(t2,t3,MERGE1(w0,w1,ofs),MERGE2(w0,w1,ofs)); \
248 OP(LP(dest+8),PACK(a4,a5,t2,t3)); \
251 UNPACK(t2,t3,MERGE1(w1,w0,ofs),MERGE2(w1,w0,ofs)); \
252 OP(LP(dest+12),PACK(a6,a7,t2,t3)); \
259 #define DEFFUNC(op,rnd,xy,sz,OP_N,avgfunc) \
260 static void op##_##rnd##_pixels##sz##_##xy (uint8_t * dest, const uint8_t * ref, \
261 const int stride, int height) \
263 switch((int)ref&3) { \
264 case 0:OP_N##0(sz,rnd##_##avgfunc); return; \
265 case 1:OP_N(1,sz,rnd##_##avgfunc); return; \
266 case 2:OP_N(2,sz,rnd##_##avgfunc); return; \
267 case 3:OP_N(3,sz,rnd##_##avgfunc); return; \
273 DEFFUNC(put, rnd,o,8,OP_C,avg2)
274 DEFFUNC(put, rnd,x,8,OP_X,avg2)
275 DEFFUNC(put,no_rnd,x,8,OP_X,avg2)
276 DEFFUNC(put, rnd,y,8,OP_Y,avg2)
277 DEFFUNC(put,no_rnd,y,8,OP_Y,avg2)
278 DEFFUNC(put, rnd,xy,8,OP_XY,PACK)
279 DEFFUNC(put,no_rnd,xy,8,OP_XY,PACK)
280 DEFFUNC(put, rnd,o,16,OP_C,avg2)
281 DEFFUNC(put, rnd,x,16,OP_X,avg2)
282 DEFFUNC(put,no_rnd,x,16,OP_X,avg2)
283 DEFFUNC(put, rnd,y,16,OP_Y,avg2)
284 DEFFUNC(put,no_rnd,y,16,OP_Y,avg2)
285 DEFFUNC(put, rnd,xy,16,OP_XY,PACK)
286 DEFFUNC(put,no_rnd,xy,16,OP_XY,PACK)
291 DEFFUNC(avg, rnd,o,8,OP_C,avg2)
292 DEFFUNC(avg, rnd,x,8,OP_X,avg2)
293 DEFFUNC(avg,no_rnd,x,8,OP_X,avg2)
294 DEFFUNC(avg, rnd,y,8,OP_Y,avg2)
295 DEFFUNC(avg,no_rnd,y,8,OP_Y,avg2)
296 DEFFUNC(avg, rnd,xy,8,OP_XY,PACK)
297 DEFFUNC(avg,no_rnd,xy,8,OP_XY,PACK)
298 DEFFUNC(avg, rnd,o,16,OP_C,avg2)
299 DEFFUNC(avg, rnd,x,16,OP_X,avg2)
300 DEFFUNC(avg,no_rnd,x,16,OP_X,avg2)
301 DEFFUNC(avg, rnd,y,16,OP_Y,avg2)
302 DEFFUNC(avg,no_rnd,y,16,OP_Y,avg2)
303 DEFFUNC(avg, rnd,xy,16,OP_XY,PACK)
304 DEFFUNC(avg,no_rnd,xy,16,OP_XY,PACK)
308 #define put_no_rnd_pixels8_o put_rnd_pixels8_o
309 #define put_no_rnd_pixels16_o put_rnd_pixels16_o
310 #define avg_no_rnd_pixels8_o avg_rnd_pixels8_o
311 #define avg_no_rnd_pixels16_o avg_rnd_pixels16_o
313 #define put_pixels8_c put_rnd_pixels8_o
314 #define put_pixels16_c put_rnd_pixels16_o
315 #define avg_pixels8_c avg_rnd_pixels8_o
316 #define avg_pixels16_c avg_rnd_pixels16_o
317 #define put_no_rnd_pixels8_c put_rnd_pixels8_o
318 #define put_no_rnd_pixels16_c put_rnd_pixels16_o
319 #define avg_no_rnd_pixels8_c avg_rnd_pixels8_o
320 #define avg_no_rnd_pixels16_c avg_rnd_pixels16_o
330 void dsputil_init_align(DSPContext* c, AVCodecContext *avctx)
332 c->put_pixels_tab[0][0] = put_rnd_pixels16_o;
333 c->put_pixels_tab[0][1] = put_rnd_pixels16_x;
334 c->put_pixels_tab[0][2] = put_rnd_pixels16_y;
335 c->put_pixels_tab[0][3] = put_rnd_pixels16_xy;
336 c->put_pixels_tab[1][0] = put_rnd_pixels8_o;
337 c->put_pixels_tab[1][1] = put_rnd_pixels8_x;
338 c->put_pixels_tab[1][2] = put_rnd_pixels8_y;
339 c->put_pixels_tab[1][3] = put_rnd_pixels8_xy;
341 c->put_no_rnd_pixels_tab[0][0] = put_no_rnd_pixels16_o;
342 c->put_no_rnd_pixels_tab[0][1] = put_no_rnd_pixels16_x;
343 c->put_no_rnd_pixels_tab[0][2] = put_no_rnd_pixels16_y;
344 c->put_no_rnd_pixels_tab[0][3] = put_no_rnd_pixels16_xy;
345 c->put_no_rnd_pixels_tab[1][0] = put_no_rnd_pixels8_o;
346 c->put_no_rnd_pixels_tab[1][1] = put_no_rnd_pixels8_x;
347 c->put_no_rnd_pixels_tab[1][2] = put_no_rnd_pixels8_y;
348 c->put_no_rnd_pixels_tab[1][3] = put_no_rnd_pixels8_xy;
350 c->avg_pixels_tab[0][0] = avg_rnd_pixels16_o;
351 c->avg_pixels_tab[0][1] = avg_rnd_pixels16_x;
352 c->avg_pixels_tab[0][2] = avg_rnd_pixels16_y;
353 c->avg_pixels_tab[0][3] = avg_rnd_pixels16_xy;
354 c->avg_pixels_tab[1][0] = avg_rnd_pixels8_o;
355 c->avg_pixels_tab[1][1] = avg_rnd_pixels8_x;
356 c->avg_pixels_tab[1][2] = avg_rnd_pixels8_y;
357 c->avg_pixels_tab[1][3] = avg_rnd_pixels8_xy;
359 c->avg_no_rnd_pixels_tab[0][0] = avg_no_rnd_pixels16_o;
360 c->avg_no_rnd_pixels_tab[0][1] = avg_no_rnd_pixels16_x;
361 c->avg_no_rnd_pixels_tab[0][2] = avg_no_rnd_pixels16_y;
362 c->avg_no_rnd_pixels_tab[0][3] = avg_no_rnd_pixels16_xy;
363 c->avg_no_rnd_pixels_tab[1][0] = avg_no_rnd_pixels8_o;
364 c->avg_no_rnd_pixels_tab[1][1] = avg_no_rnd_pixels8_x;
365 c->avg_no_rnd_pixels_tab[1][2] = avg_no_rnd_pixels8_y;
366 c->avg_no_rnd_pixels_tab[1][3] = avg_no_rnd_pixels8_xy;
370 #define dspfunc(PFX, IDX, NUM) \
371 c->PFX ## _pixels_tab[IDX][ 0] = PFX ## NUM ## _mc00_c; \
372 c->PFX ## _pixels_tab[IDX][ 1] = PFX ## NUM ## _mc10_c; \
373 c->PFX ## _pixels_tab[IDX][ 2] = PFX ## NUM ## _mc20_c; \
374 c->PFX ## _pixels_tab[IDX][ 3] = PFX ## NUM ## _mc30_c; \
375 c->PFX ## _pixels_tab[IDX][ 4] = PFX ## NUM ## _mc01_c; \
376 c->PFX ## _pixels_tab[IDX][ 5] = PFX ## NUM ## _mc11_c; \
377 c->PFX ## _pixels_tab[IDX][ 6] = PFX ## NUM ## _mc21_c; \
378 c->PFX ## _pixels_tab[IDX][ 7] = PFX ## NUM ## _mc31_c; \
379 c->PFX ## _pixels_tab[IDX][ 8] = PFX ## NUM ## _mc02_c; \
380 c->PFX ## _pixels_tab[IDX][ 9] = PFX ## NUM ## _mc12_c; \
381 c->PFX ## _pixels_tab[IDX][10] = PFX ## NUM ## _mc22_c; \
382 c->PFX ## _pixels_tab[IDX][11] = PFX ## NUM ## _mc32_c; \
383 c->PFX ## _pixels_tab[IDX][12] = PFX ## NUM ## _mc03_c; \
384 c->PFX ## _pixels_tab[IDX][13] = PFX ## NUM ## _mc13_c; \
385 c->PFX ## _pixels_tab[IDX][14] = PFX ## NUM ## _mc23_c; \
386 c->PFX ## _pixels_tab[IDX][15] = PFX ## NUM ## _mc33_c
388 dspfunc(put_qpel, 0, 16);
389 dspfunc(put_no_rnd_qpel, 0, 16);
391 dspfunc(avg_qpel, 0, 16);
392 /* dspfunc(avg_no_rnd_qpel, 0, 16); */
394 dspfunc(put_qpel, 1, 8);
395 dspfunc(put_no_rnd_qpel, 1, 8);
397 dspfunc(avg_qpel, 1, 8);
398 /* dspfunc(avg_no_rnd_qpel, 1, 8); */
400 dspfunc(put_h264_qpel, 0, 16);
401 dspfunc(put_h264_qpel, 1, 8);
402 dspfunc(put_h264_qpel, 2, 4);
403 dspfunc(avg_h264_qpel, 0, 16);
404 dspfunc(avg_h264_qpel, 1, 8);
405 dspfunc(avg_h264_qpel, 2, 4);
408 c->put_h264_chroma_pixels_tab[0]= put_h264_chroma_mc8_c;
409 c->put_h264_chroma_pixels_tab[1]= put_h264_chroma_mc4_c;
410 c->put_h264_chroma_pixels_tab[2]= put_h264_chroma_mc2_c;
411 c->avg_h264_chroma_pixels_tab[0]= avg_h264_chroma_mc8_c;
412 c->avg_h264_chroma_pixels_tab[1]= avg_h264_chroma_mc4_c;
413 c->avg_h264_chroma_pixels_tab[2]= avg_h264_chroma_mc2_c;
415 c->put_mspel_pixels_tab[0]= put_mspel8_mc00_c;
416 c->put_mspel_pixels_tab[1]= put_mspel8_mc10_c;
417 c->put_mspel_pixels_tab[2]= put_mspel8_mc20_c;
418 c->put_mspel_pixels_tab[3]= put_mspel8_mc30_c;
419 c->put_mspel_pixels_tab[4]= put_mspel8_mc02_c;
420 c->put_mspel_pixels_tab[5]= put_mspel8_mc12_c;
421 c->put_mspel_pixels_tab[6]= put_mspel8_mc22_c;
422 c->put_mspel_pixels_tab[7]= put_mspel8_mc32_c;