2 * aligned/packed access motion
4 * Copyright (c) 2001-2003 BERO <bero@geocities.co.jp>
6 * This file is part of FFmpeg.
8 * FFmpeg is free software; you can redistribute it and/or
9 * modify it under the terms of the GNU Lesser General Public
10 * License as published by the Free Software Foundation; either
11 * version 2.1 of the License, or (at your option) any later version.
13 * FFmpeg is distributed in the hope that it will be useful,
14 * but WITHOUT ANY WARRANTY; without even the implied warranty of
15 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
16 * Lesser General Public License for more details.
18 * You should have received a copy of the GNU Lesser General Public
19 * License along with FFmpeg; if not, write to the Free Software
20 * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
24 #include "libavcodec/avcodec.h"
25 #include "libavcodec/dsputil.h"
26 #include "dsputil_sh4.h"
29 #define LP(p) *(uint32_t*)(p)
30 #define LPC(p) *(const uint32_t*)(p)
33 #define UNPACK(ph,pl,tt0,tt1) do { \
34 uint32_t t0,t1; t0=tt0;t1=tt1; \
35 ph = ( (t0 & ~BYTE_VEC32(0x03))>>2) + ( (t1 & ~BYTE_VEC32(0x03))>>2); \
36 pl = (t0 & BYTE_VEC32(0x03)) + (t1 & BYTE_VEC32(0x03)); } while(0)
38 #define rnd_PACK(ph,pl,nph,npl) ph + nph + (((pl + npl + BYTE_VEC32(0x02))>>2) & BYTE_VEC32(0x03))
39 #define no_rnd_PACK(ph,pl,nph,npl) ph + nph + (((pl + npl + BYTE_VEC32(0x01))>>2) & BYTE_VEC32(0x03))
42 #define MERGE1(a,b,ofs) (ofs==0)?a:( ((a)>>(8*ofs))|((b)<<(32-8*ofs)) )
43 #define MERGE2(a,b,ofs) (ofs==3)?b:( ((a)>>(8*(ofs+1)))|((b)<<(32-8*(ofs+1))) )
45 #define MERGE1(a,b,ofs) (ofs==0)?a:( ((a)<<(8*ofs))|((b)>>(32-8*ofs)) )
46 #define MERGE2(a,b,ofs) (ofs==3)?b:( ((a)<<(8+8*ofs))|((b)>>(32-8-8*ofs)) )
50 #define put(d,s) d = s
51 #define avg(d,s) d = rnd_avg32(s,d)
56 OP(LP(dest),MERGE1(LPC(ref),LPC(ref+4),ofs)); \
63 OP(LP(dest),LPC(ref)); \
71 static void put_pixels4_c(uint8_t *dest,const uint8_t *ref, const int stride,int height)
74 case 0: OP_C40(); return;
75 case 1: OP_C4(1); return;
76 case 2: OP_C4(2); return;
77 case 3: OP_C4(3); return;
84 static void avg_pixels4_c(uint8_t *dest,const uint8_t *ref, const int stride,int height)
87 case 0: OP_C40(); return;
88 case 1: OP_C4(1); return;
89 case 2: OP_C4(2); return;
90 case 3: OP_C4(3); return;
96 #define OP_C(ofs,sz,avg2) \
103 OP(LP(dest+0), MERGE1(t0,t1,ofs)); \
105 OP(LP(dest+4), MERGE1(t1,t0,ofs)); \
108 OP(LP(dest+8), MERGE1(t0,t1,ofs)); \
110 OP(LP(dest+12), MERGE1(t1,t0,ofs)); \
118 #define OP_C0(sz,avg2) \
121 OP(LP(dest+0), LPC(ref+0)); \
122 OP(LP(dest+4), LPC(ref+4)); \
124 OP(LP(dest+8), LPC(ref+8)); \
125 OP(LP(dest+12), LPC(ref+12)); \
132 #define OP_X(ofs,sz,avg2) \
139 OP(LP(dest+0), avg2(MERGE1(t0,t1,ofs),MERGE2(t0,t1,ofs))); \
141 OP(LP(dest+4), avg2(MERGE1(t1,t0,ofs),MERGE2(t1,t0,ofs))); \
144 OP(LP(dest+8), avg2(MERGE1(t0,t1,ofs),MERGE2(t0,t1,ofs))); \
146 OP(LP(dest+12), avg2(MERGE1(t1,t0,ofs),MERGE2(t1,t0,ofs))); \
154 #define OP_Y0(sz,avg2) \
156 uint32_t t0,t1,t2,t3,t; \
168 OP(LP(dest+0), avg2(t0,t)); t0 = t; \
170 OP(LP(dest+4), avg2(t1,t)); t1 = t; \
173 OP(LP(dest+8), avg2(t2,t)); t2 = t; \
175 OP(LP(dest+12), avg2(t3,t)); t3 = t; \
181 #define OP_Y(ofs,sz,avg2) \
183 uint32_t t0,t1,t2,t3,t,w0,w1; \
188 t0 = MERGE1(w0,w1,ofs); \
190 t1 = MERGE1(w1,w0,ofs); \
193 t2 = MERGE1(w0,w1,ofs); \
195 t3 = MERGE1(w1,w0,ofs); \
202 t = MERGE1(w0,w1,ofs); \
203 OP(LP(dest+0), avg2(t0,t)); t0 = t; \
205 t = MERGE1(w1,w0,ofs); \
206 OP(LP(dest+4), avg2(t1,t)); t1 = t; \
209 t = MERGE1(w0,w1,ofs); \
210 OP(LP(dest+8), avg2(t2,t)); t2 = t; \
212 t = MERGE1(w1,w0,ofs); \
213 OP(LP(dest+12), avg2(t3,t)); t3 = t; \
219 #define OP_X0(sz,avg2) OP_X(0,sz,avg2)
220 #define OP_XY0(sz,PACK) OP_XY(0,sz,PACK)
221 #define OP_XY(ofs,sz,PACK) \
223 uint32_t t2,t3,w0,w1; \
224 uint32_t a0,a1,a2,a3,a4,a5,a6,a7; \
229 UNPACK(a0,a1,MERGE1(w0,w1,ofs),MERGE2(w0,w1,ofs)); \
231 UNPACK(a2,a3,MERGE1(w1,w0,ofs),MERGE2(w1,w0,ofs)); \
234 UNPACK(a4,a5,MERGE1(w0,w1,ofs),MERGE2(w0,w1,ofs)); \
236 UNPACK(a6,a7,MERGE1(w1,w0,ofs),MERGE2(w1,w0,ofs)); \
242 UNPACK(t2,t3,MERGE1(w0,w1,ofs),MERGE2(w0,w1,ofs)); \
243 OP(LP(dest+0),PACK(a0,a1,t2,t3)); \
246 UNPACK(t2,t3,MERGE1(w1,w0,ofs),MERGE2(w1,w0,ofs)); \
247 OP(LP(dest+4),PACK(a2,a3,t2,t3)); \
251 UNPACK(t2,t3,MERGE1(w0,w1,ofs),MERGE2(w0,w1,ofs)); \
252 OP(LP(dest+8),PACK(a4,a5,t2,t3)); \
255 UNPACK(t2,t3,MERGE1(w1,w0,ofs),MERGE2(w1,w0,ofs)); \
256 OP(LP(dest+12),PACK(a6,a7,t2,t3)); \
263 #define DEFFUNC(op,rnd,xy,sz,OP_N,avgfunc) \
264 static void op##_##rnd##_pixels##sz##_##xy (uint8_t * dest, const uint8_t * ref, \
265 const int stride, int height) \
267 switch((int)ref&3) { \
268 case 0:OP_N##0(sz,rnd##_##avgfunc); return; \
269 case 1:OP_N(1,sz,rnd##_##avgfunc); return; \
270 case 2:OP_N(2,sz,rnd##_##avgfunc); return; \
271 case 3:OP_N(3,sz,rnd##_##avgfunc); return; \
277 DEFFUNC(put, rnd,o,8,OP_C,avg32)
278 DEFFUNC(put, rnd,x,8,OP_X,avg32)
279 DEFFUNC(put,no_rnd,x,8,OP_X,avg32)
280 DEFFUNC(put, rnd,y,8,OP_Y,avg32)
281 DEFFUNC(put,no_rnd,y,8,OP_Y,avg32)
282 DEFFUNC(put, rnd,xy,8,OP_XY,PACK)
283 DEFFUNC(put,no_rnd,xy,8,OP_XY,PACK)
284 DEFFUNC(put, rnd,o,16,OP_C,avg32)
285 DEFFUNC(put, rnd,x,16,OP_X,avg32)
286 DEFFUNC(put,no_rnd,x,16,OP_X,avg32)
287 DEFFUNC(put, rnd,y,16,OP_Y,avg32)
288 DEFFUNC(put,no_rnd,y,16,OP_Y,avg32)
289 DEFFUNC(put, rnd,xy,16,OP_XY,PACK)
290 DEFFUNC(put,no_rnd,xy,16,OP_XY,PACK)
295 DEFFUNC(avg, rnd,o,8,OP_C,avg32)
296 DEFFUNC(avg, rnd,x,8,OP_X,avg32)
297 DEFFUNC(avg,no_rnd,x,8,OP_X,avg32)
298 DEFFUNC(avg, rnd,y,8,OP_Y,avg32)
299 DEFFUNC(avg,no_rnd,y,8,OP_Y,avg32)
300 DEFFUNC(avg, rnd,xy,8,OP_XY,PACK)
301 DEFFUNC(avg,no_rnd,xy,8,OP_XY,PACK)
302 DEFFUNC(avg, rnd,o,16,OP_C,avg32)
303 DEFFUNC(avg, rnd,x,16,OP_X,avg32)
304 DEFFUNC(avg,no_rnd,x,16,OP_X,avg32)
305 DEFFUNC(avg, rnd,y,16,OP_Y,avg32)
306 DEFFUNC(avg,no_rnd,y,16,OP_Y,avg32)
307 DEFFUNC(avg, rnd,xy,16,OP_XY,PACK)
308 DEFFUNC(avg,no_rnd,xy,16,OP_XY,PACK)
312 #define put_no_rnd_pixels8_o put_rnd_pixels8_o
313 #define put_no_rnd_pixels16_o put_rnd_pixels16_o
314 #define avg_no_rnd_pixels8_o avg_rnd_pixels8_o
315 #define avg_no_rnd_pixels16_o avg_rnd_pixels16_o
317 #define put_pixels8_c put_rnd_pixels8_o
318 #define put_pixels16_c put_rnd_pixels16_o
319 #define avg_pixels8_c avg_rnd_pixels8_o
320 #define avg_pixels16_c avg_rnd_pixels16_o
321 #define put_no_rnd_pixels8_c put_rnd_pixels8_o
322 #define put_no_rnd_pixels16_c put_rnd_pixels16_o
323 #define avg_no_rnd_pixels8_c avg_rnd_pixels8_o
324 #define avg_no_rnd_pixels16_c avg_rnd_pixels16_o
334 void dsputil_init_align(DSPContext* c, AVCodecContext *avctx)
336 const int high_bit_depth = avctx->bits_per_raw_sample > 8;
338 if (!high_bit_depth) {
339 c->put_pixels_tab[0][0] = put_rnd_pixels16_o;
340 c->put_pixels_tab[0][1] = put_rnd_pixels16_x;
341 c->put_pixels_tab[0][2] = put_rnd_pixels16_y;
342 c->put_pixels_tab[0][3] = put_rnd_pixels16_xy;
343 c->put_pixels_tab[1][0] = put_rnd_pixels8_o;
344 c->put_pixels_tab[1][1] = put_rnd_pixels8_x;
345 c->put_pixels_tab[1][2] = put_rnd_pixels8_y;
346 c->put_pixels_tab[1][3] = put_rnd_pixels8_xy;
348 c->put_no_rnd_pixels_tab[0][0] = put_no_rnd_pixels16_o;
349 c->put_no_rnd_pixels_tab[0][1] = put_no_rnd_pixels16_x;
350 c->put_no_rnd_pixels_tab[0][2] = put_no_rnd_pixels16_y;
351 c->put_no_rnd_pixels_tab[0][3] = put_no_rnd_pixels16_xy;
352 c->put_no_rnd_pixels_tab[1][0] = put_no_rnd_pixels8_o;
353 c->put_no_rnd_pixels_tab[1][1] = put_no_rnd_pixels8_x;
354 c->put_no_rnd_pixels_tab[1][2] = put_no_rnd_pixels8_y;
355 c->put_no_rnd_pixels_tab[1][3] = put_no_rnd_pixels8_xy;
357 c->avg_pixels_tab[0][0] = avg_rnd_pixels16_o;
358 c->avg_pixels_tab[0][1] = avg_rnd_pixels16_x;
359 c->avg_pixels_tab[0][2] = avg_rnd_pixels16_y;
360 c->avg_pixels_tab[0][3] = avg_rnd_pixels16_xy;
361 c->avg_pixels_tab[1][0] = avg_rnd_pixels8_o;
362 c->avg_pixels_tab[1][1] = avg_rnd_pixels8_x;
363 c->avg_pixels_tab[1][2] = avg_rnd_pixels8_y;
364 c->avg_pixels_tab[1][3] = avg_rnd_pixels8_xy;
366 c->avg_no_rnd_pixels_tab[0][0] = avg_no_rnd_pixels16_o;
367 c->avg_no_rnd_pixels_tab[0][1] = avg_no_rnd_pixels16_x;
368 c->avg_no_rnd_pixels_tab[0][2] = avg_no_rnd_pixels16_y;
369 c->avg_no_rnd_pixels_tab[0][3] = avg_no_rnd_pixels16_xy;
370 c->avg_no_rnd_pixels_tab[1][0] = avg_no_rnd_pixels8_o;
371 c->avg_no_rnd_pixels_tab[1][1] = avg_no_rnd_pixels8_x;
372 c->avg_no_rnd_pixels_tab[1][2] = avg_no_rnd_pixels8_y;
373 c->avg_no_rnd_pixels_tab[1][3] = avg_no_rnd_pixels8_xy;
378 #define dspfunc(PFX, IDX, NUM) \
379 c->PFX ## _pixels_tab[IDX][ 0] = PFX ## NUM ## _mc00_sh4; \
380 c->PFX ## _pixels_tab[IDX][ 1] = PFX ## NUM ## _mc10_sh4; \
381 c->PFX ## _pixels_tab[IDX][ 2] = PFX ## NUM ## _mc20_sh4; \
382 c->PFX ## _pixels_tab[IDX][ 3] = PFX ## NUM ## _mc30_sh4; \
383 c->PFX ## _pixels_tab[IDX][ 4] = PFX ## NUM ## _mc01_sh4; \
384 c->PFX ## _pixels_tab[IDX][ 5] = PFX ## NUM ## _mc11_sh4; \
385 c->PFX ## _pixels_tab[IDX][ 6] = PFX ## NUM ## _mc21_sh4; \
386 c->PFX ## _pixels_tab[IDX][ 7] = PFX ## NUM ## _mc31_sh4; \
387 c->PFX ## _pixels_tab[IDX][ 8] = PFX ## NUM ## _mc02_sh4; \
388 c->PFX ## _pixels_tab[IDX][ 9] = PFX ## NUM ## _mc12_sh4; \
389 c->PFX ## _pixels_tab[IDX][10] = PFX ## NUM ## _mc22_sh4; \
390 c->PFX ## _pixels_tab[IDX][11] = PFX ## NUM ## _mc32_sh4; \
391 c->PFX ## _pixels_tab[IDX][12] = PFX ## NUM ## _mc03_sh4; \
392 c->PFX ## _pixels_tab[IDX][13] = PFX ## NUM ## _mc13_sh4; \
393 c->PFX ## _pixels_tab[IDX][14] = PFX ## NUM ## _mc23_sh4; \
394 c->PFX ## _pixels_tab[IDX][15] = PFX ## NUM ## _mc33_sh4
396 dspfunc(put_qpel, 0, 16);
397 dspfunc(put_no_rnd_qpel, 0, 16);
399 dspfunc(avg_qpel, 0, 16);
400 /* dspfunc(avg_no_rnd_qpel, 0, 16); */
402 dspfunc(put_qpel, 1, 8);
403 dspfunc(put_no_rnd_qpel, 1, 8);
405 dspfunc(avg_qpel, 1, 8);
406 /* dspfunc(avg_no_rnd_qpel, 1, 8); */
408 if (!high_bit_depth) {
409 dspfunc(put_h264_qpel, 0, 16);
410 dspfunc(put_h264_qpel, 1, 8);
411 dspfunc(put_h264_qpel, 2, 4);
412 dspfunc(avg_h264_qpel, 0, 16);
413 dspfunc(avg_h264_qpel, 1, 8);
414 dspfunc(avg_h264_qpel, 2, 4);
418 if (!high_bit_depth) {
419 c->put_h264_chroma_pixels_tab[0]= put_h264_chroma_mc8_sh4;
420 c->put_h264_chroma_pixels_tab[1]= put_h264_chroma_mc4_sh4;
421 c->put_h264_chroma_pixels_tab[2]= put_h264_chroma_mc2_sh4;
422 c->avg_h264_chroma_pixels_tab[0]= avg_h264_chroma_mc8_sh4;
423 c->avg_h264_chroma_pixels_tab[1]= avg_h264_chroma_mc4_sh4;
424 c->avg_h264_chroma_pixels_tab[2]= avg_h264_chroma_mc2_sh4;
427 c->put_mspel_pixels_tab[0]= put_mspel8_mc00_sh4;
428 c->put_mspel_pixels_tab[1]= put_mspel8_mc10_sh4;
429 c->put_mspel_pixels_tab[2]= put_mspel8_mc20_sh4;
430 c->put_mspel_pixels_tab[3]= put_mspel8_mc30_sh4;
431 c->put_mspel_pixels_tab[4]= put_mspel8_mc02_sh4;
432 c->put_mspel_pixels_tab[5]= put_mspel8_mc12_sh4;
433 c->put_mspel_pixels_tab[6]= put_mspel8_mc22_sh4;
434 c->put_mspel_pixels_tab[7]= put_mspel8_mc32_sh4;