yading@10: /*
yading@10:  * aligned/packed access motion
yading@10:  *
yading@10:  * Copyright (c) 2001-2003 BERO <bero@geocities.co.jp>
yading@10:  *
yading@10:  * This file is part of FFmpeg.
yading@10:  *
yading@10:  * FFmpeg is free software; you can redistribute it and/or
yading@10:  * modify it under the terms of the GNU Lesser General Public
yading@10:  * License as published by the Free Software Foundation; either
yading@10:  * version 2.1 of the License, or (at your option) any later version.
yading@10:  *
yading@10:  * FFmpeg is distributed in the hope that it will be useful,
yading@10:  * but WITHOUT ANY WARRANTY; without even the implied warranty of
yading@10:  * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
yading@10:  * Lesser General Public License for more details.
yading@10:  *
yading@10:  * You should have received a copy of the GNU Lesser General Public
yading@10:  * License along with FFmpeg; if not, write to the Free Software
yading@10:  * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
yading@10:  */
yading@10: 
yading@10: #include "libavutil/attributes.h"
yading@10: #include "libavcodec/avcodec.h"
yading@10: #include "libavcodec/dsputil.h"
yading@10: #include "libavcodec/bit_depth_template.c" // for BYTE_VEC32
yading@10: #include "libavcodec/hpeldsp.h"
yading@10: #include "libavcodec/rnd_avg.h"
yading@10: #include "dsputil_sh4.h"
yading@10: 
yading@10: 
yading@10: #define         LP(p)           *(uint32_t*)(p)
yading@10: #define         LPC(p)          *(const uint32_t*)(p)
yading@10: 
yading@10: 
yading@10: #define         UNPACK(ph,pl,tt0,tt1) do { \
yading@10:         uint32_t t0,t1; t0=tt0;t1=tt1; \
yading@10:         ph = ( (t0 & ~BYTE_VEC32(0x03))>>2) + ( (t1 & ~BYTE_VEC32(0x03))>>2); \
yading@10:         pl = (t0 & BYTE_VEC32(0x03)) + (t1 & BYTE_VEC32(0x03)); } while(0)
yading@10: 
yading@10: #define         rnd_PACK(ph,pl,nph,npl) ph + nph + (((pl + npl + BYTE_VEC32(0x02))>>2) & BYTE_VEC32(0x03))
yading@10: #define         no_rnd_PACK(ph,pl,nph,npl)      ph + nph + (((pl + npl + BYTE_VEC32(0x01))>>2) & BYTE_VEC32(0x03))
yading@10: 
yading@10: /* little-endian */
yading@10: #define         MERGE1(a,b,ofs) (ofs==0)?a:( ((a)>>(8*ofs))|((b)<<(32-8*ofs)) )
yading@10: #define         MERGE2(a,b,ofs) (ofs==3)?b:( ((a)>>(8*(ofs+1)))|((b)<<(32-8*(ofs+1))) )
yading@10: /* big
yading@10: #define         MERGE1(a,b,ofs) (ofs==0)?a:( ((a)<<(8*ofs))|((b)>>(32-8*ofs)) )
yading@10: #define         MERGE2(a,b,ofs) (ofs==3)?b:( ((a)<<(8+8*ofs))|((b)>>(32-8-8*ofs)) )
yading@10: */
yading@10: 
yading@10: 
yading@10: #define         put(d,s)        d = s
yading@10: #define         avg(d,s)        d = rnd_avg32(s,d)
yading@10: 
yading@10: #define         OP_C4(ofs) \
yading@10:         ref-=ofs; \
yading@10:         do { \
yading@10:                 OP(LP(dest),MERGE1(LPC(ref),LPC(ref+4),ofs)); \
yading@10:                 ref+=stride; \
yading@10:                 dest+=stride; \
yading@10:         } while(--height)
yading@10: 
yading@10: #define        OP_C40() \
yading@10:         do { \
yading@10:                 OP(LP(dest),LPC(ref)); \
yading@10:                 ref+=stride; \
yading@10:                 dest+=stride; \
yading@10:         } while(--height)
yading@10: 
yading@10: 
yading@10: #define         OP      put
yading@10: 
yading@10: static void put_pixels4_c(uint8_t *dest, const uint8_t *ref,
yading@10:                           const int stride, int height)
yading@10: {
yading@10:         switch((int)ref&3){
yading@10:         case 0: OP_C40(); return;
yading@10:         case 1: OP_C4(1); return;
yading@10:         case 2: OP_C4(2); return;
yading@10:         case 3: OP_C4(3); return;
yading@10:         }
yading@10: }
yading@10: 
yading@10: #undef          OP
yading@10: #define         OP      avg
yading@10: 
yading@10: static void avg_pixels4_c(uint8_t *dest, const uint8_t *ref,
yading@10:                           const int stride, int height)
yading@10: {
yading@10:         switch((int)ref&3){
yading@10:         case 0: OP_C40(); return;
yading@10:         case 1: OP_C4(1); return;
yading@10:         case 2: OP_C4(2); return;
yading@10:         case 3: OP_C4(3); return;
yading@10:         }
yading@10: }
yading@10: 
yading@10: #undef          OP
yading@10: 
yading@10: #define         OP_C(ofs,sz,avg2) \
yading@10: { \
yading@10:         ref-=ofs; \
yading@10:         do { \
yading@10:                 uint32_t        t0,t1; \
yading@10:                 t0 = LPC(ref+0); \
yading@10:                 t1 = LPC(ref+4); \
yading@10:                 OP(LP(dest+0), MERGE1(t0,t1,ofs)); \
yading@10:                 t0 = LPC(ref+8); \
yading@10:                 OP(LP(dest+4), MERGE1(t1,t0,ofs)); \
yading@10: if (sz==16) { \
yading@10:                 t1 = LPC(ref+12); \
yading@10:                 OP(LP(dest+8), MERGE1(t0,t1,ofs)); \
yading@10:                 t0 = LPC(ref+16); \
yading@10:                 OP(LP(dest+12), MERGE1(t1,t0,ofs)); \
yading@10: } \
yading@10:                 ref+=stride; \
yading@10:                 dest+= stride; \
yading@10:         } while(--height); \
yading@10: }
yading@10: 
yading@10: /* aligned */
yading@10: #define         OP_C0(sz,avg2) \
yading@10: { \
yading@10:         do { \
yading@10:                 OP(LP(dest+0), LPC(ref+0)); \
yading@10:                 OP(LP(dest+4), LPC(ref+4)); \
yading@10: if (sz==16) { \
yading@10:                 OP(LP(dest+8), LPC(ref+8)); \
yading@10:                 OP(LP(dest+12), LPC(ref+12)); \
yading@10: } \
yading@10:                 ref+=stride; \
yading@10:                 dest+= stride; \
yading@10:         } while(--height); \
yading@10: }
yading@10: 
yading@10: #define         OP_X(ofs,sz,avg2) \
yading@10: { \
yading@10:         ref-=ofs; \
yading@10:         do { \
yading@10:                 uint32_t        t0,t1; \
yading@10:                 t0 = LPC(ref+0); \
yading@10:                 t1 = LPC(ref+4); \
yading@10:                 OP(LP(dest+0), avg2(MERGE1(t0,t1,ofs),MERGE2(t0,t1,ofs))); \
yading@10:                 t0 = LPC(ref+8); \
yading@10:                 OP(LP(dest+4), avg2(MERGE1(t1,t0,ofs),MERGE2(t1,t0,ofs))); \
yading@10: if (sz==16) { \
yading@10:                 t1 = LPC(ref+12); \
yading@10:                 OP(LP(dest+8), avg2(MERGE1(t0,t1,ofs),MERGE2(t0,t1,ofs))); \
yading@10:                 t0 = LPC(ref+16); \
yading@10:                 OP(LP(dest+12), avg2(MERGE1(t1,t0,ofs),MERGE2(t1,t0,ofs))); \
yading@10: } \
yading@10:                 ref+=stride; \
yading@10:                 dest+= stride; \
yading@10:         } while(--height); \
yading@10: }
yading@10: 
yading@10: /* aligned */
yading@10: #define         OP_Y0(sz,avg2) \
yading@10: { \
yading@10:         uint32_t t0,t1,t2,t3,t; \
yading@10: \
yading@10:         t0 = LPC(ref+0); \
yading@10:         t1 = LPC(ref+4); \
yading@10: if (sz==16) { \
yading@10:         t2 = LPC(ref+8); \
yading@10:         t3 = LPC(ref+12); \
yading@10: } \
yading@10:         do { \
yading@10:                 ref += stride; \
yading@10: \
yading@10:                 t = LPC(ref+0); \
yading@10:                 OP(LP(dest+0), avg2(t0,t)); t0 = t; \
yading@10:                 t = LPC(ref+4); \
yading@10:                 OP(LP(dest+4), avg2(t1,t)); t1 = t; \
yading@10: if (sz==16) { \
yading@10:                 t = LPC(ref+8); \
yading@10:                 OP(LP(dest+8), avg2(t2,t)); t2 = t; \
yading@10:                 t = LPC(ref+12); \
yading@10:                 OP(LP(dest+12), avg2(t3,t)); t3 = t; \
yading@10: } \
yading@10:                 dest+= stride; \
yading@10:         } while(--height); \
yading@10: }
yading@10: 
yading@10: #define         OP_Y(ofs,sz,avg2) \
yading@10: { \
yading@10:         uint32_t t0,t1,t2,t3,t,w0,w1; \
yading@10: \
yading@10:         ref-=ofs; \
yading@10:         w0 = LPC(ref+0); \
yading@10:         w1 = LPC(ref+4); \
yading@10:         t0 = MERGE1(w0,w1,ofs); \
yading@10:         w0 = LPC(ref+8); \
yading@10:         t1 = MERGE1(w1,w0,ofs); \
yading@10: if (sz==16) { \
yading@10:         w1 = LPC(ref+12); \
yading@10:         t2 = MERGE1(w0,w1,ofs); \
yading@10:         w0 = LPC(ref+16); \
yading@10:         t3 = MERGE1(w1,w0,ofs); \
yading@10: } \
yading@10:         do { \
yading@10:                 ref += stride; \
yading@10: \
yading@10:                 w0 = LPC(ref+0); \
yading@10:                 w1 = LPC(ref+4); \
yading@10:                 t = MERGE1(w0,w1,ofs); \
yading@10:                 OP(LP(dest+0), avg2(t0,t)); t0 = t; \
yading@10:                 w0 = LPC(ref+8); \
yading@10:                 t = MERGE1(w1,w0,ofs); \
yading@10:                 OP(LP(dest+4), avg2(t1,t)); t1 = t; \
yading@10: if (sz==16) { \
yading@10:                 w1 = LPC(ref+12); \
yading@10:                 t = MERGE1(w0,w1,ofs); \
yading@10:                 OP(LP(dest+8), avg2(t2,t)); t2 = t; \
yading@10:                 w0 = LPC(ref+16); \
yading@10:                 t = MERGE1(w1,w0,ofs); \
yading@10:                 OP(LP(dest+12), avg2(t3,t)); t3 = t; \
yading@10: } \
yading@10:                 dest+=stride; \
yading@10:         } while(--height); \
yading@10: }
yading@10: 
yading@10: #define OP_X0(sz,avg2) OP_X(0,sz,avg2)
yading@10: #define OP_XY0(sz,PACK) OP_XY(0,sz,PACK)
yading@10: #define         OP_XY(ofs,sz,PACK) \
yading@10: { \
yading@10:         uint32_t        t2,t3,w0,w1; \
yading@10:         uint32_t        a0,a1,a2,a3,a4,a5,a6,a7; \
yading@10: \
yading@10:         ref -= ofs; \
yading@10:         w0 = LPC(ref+0); \
yading@10:         w1 = LPC(ref+4); \
yading@10:         UNPACK(a0,a1,MERGE1(w0,w1,ofs),MERGE2(w0,w1,ofs)); \
yading@10:         w0 = LPC(ref+8); \
yading@10:         UNPACK(a2,a3,MERGE1(w1,w0,ofs),MERGE2(w1,w0,ofs)); \
yading@10: if (sz==16) { \
yading@10:         w1 = LPC(ref+12); \
yading@10:         UNPACK(a4,a5,MERGE1(w0,w1,ofs),MERGE2(w0,w1,ofs)); \
yading@10:         w0 = LPC(ref+16); \
yading@10:         UNPACK(a6,a7,MERGE1(w1,w0,ofs),MERGE2(w1,w0,ofs)); \
yading@10: } \
yading@10:         do { \
yading@10:                 ref+=stride; \
yading@10:                 w0 = LPC(ref+0); \
yading@10:                 w1 = LPC(ref+4); \
yading@10:                 UNPACK(t2,t3,MERGE1(w0,w1,ofs),MERGE2(w0,w1,ofs)); \
yading@10:                 OP(LP(dest+0),PACK(a0,a1,t2,t3)); \
yading@10:                 a0 = t2; a1 = t3; \
yading@10:                 w0 = LPC(ref+8); \
yading@10:                 UNPACK(t2,t3,MERGE1(w1,w0,ofs),MERGE2(w1,w0,ofs)); \
yading@10:                 OP(LP(dest+4),PACK(a2,a3,t2,t3)); \
yading@10:                 a2 = t2; a3 = t3; \
yading@10: if (sz==16) { \
yading@10:                 w1 = LPC(ref+12); \
yading@10:                 UNPACK(t2,t3,MERGE1(w0,w1,ofs),MERGE2(w0,w1,ofs)); \
yading@10:                 OP(LP(dest+8),PACK(a4,a5,t2,t3)); \
yading@10:                 a4 = t2; a5 = t3; \
yading@10:                 w0 = LPC(ref+16); \
yading@10:                 UNPACK(t2,t3,MERGE1(w1,w0,ofs),MERGE2(w1,w0,ofs)); \
yading@10:                 OP(LP(dest+12),PACK(a6,a7,t2,t3)); \
yading@10:                 a6 = t2; a7 = t3; \
yading@10: } \
yading@10:                 dest+=stride; \
yading@10:         } while(--height); \
yading@10: }
yading@10: 
yading@10: #define         DEFFUNC(prefix, op, rnd, xy, sz, OP_N, avgfunc) \
yading@10: prefix void op##_##rnd##_pixels##sz##_##xy(uint8_t *dest, const uint8_t *ref, \
yading@10:                                            const ptrdiff_t stride, int height) \
yading@10: { \
yading@10:         switch((int)ref&3) { \
yading@10:         case 0:OP_N##0(sz,rnd##_##avgfunc); return; \
yading@10:         case 1:OP_N(1,sz,rnd##_##avgfunc); return; \
yading@10:         case 2:OP_N(2,sz,rnd##_##avgfunc); return; \
yading@10:         case 3:OP_N(3,sz,rnd##_##avgfunc); return; \
yading@10:         } \
yading@10: }
yading@10: 
yading@10: #define OP put
yading@10: 
yading@10: DEFFUNC(      ,ff_put,rnd,o,8,OP_C,avg32)
yading@10: DEFFUNC(static,put,   rnd,x,8,OP_X,avg32)
yading@10: DEFFUNC(static,put,no_rnd,x,8,OP_X,avg32)
yading@10: DEFFUNC(static,put,   rnd,y,8,OP_Y,avg32)
yading@10: DEFFUNC(static,put,no_rnd,y,8,OP_Y,avg32)
yading@10: DEFFUNC(static,put,   rnd,xy,8,OP_XY,PACK)
yading@10: DEFFUNC(static,put,no_rnd,xy,8,OP_XY,PACK)
yading@10: DEFFUNC(      ,ff_put,rnd,o,16,OP_C,avg32)
yading@10: DEFFUNC(static,put,   rnd,x,16,OP_X,avg32)
yading@10: DEFFUNC(static,put,no_rnd,x,16,OP_X,avg32)
yading@10: DEFFUNC(static,put,   rnd,y,16,OP_Y,avg32)
yading@10: DEFFUNC(static,put,no_rnd,y,16,OP_Y,avg32)
yading@10: DEFFUNC(static,put,   rnd,xy,16,OP_XY,PACK)
yading@10: DEFFUNC(static,put,no_rnd,xy,16,OP_XY,PACK)
yading@10: 
yading@10: #undef OP
yading@10: #define OP avg
yading@10: 
yading@10: DEFFUNC(      ,ff_avg,rnd,o,8,OP_C,avg32)
yading@10: DEFFUNC(static,avg,   rnd,x,8,OP_X,avg32)
yading@10: DEFFUNC(static,avg,   rnd,y,8,OP_Y,avg32)
yading@10: DEFFUNC(static,avg,   rnd,xy,8,OP_XY,PACK)
yading@10: DEFFUNC(      ,ff_avg,rnd,o,16,OP_C,avg32)
yading@10: DEFFUNC(static,avg,   rnd,x,16,OP_X,avg32)
yading@10: DEFFUNC(static,avg,no_rnd,x,16,OP_X,avg32)
yading@10: DEFFUNC(static,avg,   rnd,y,16,OP_Y,avg32)
yading@10: DEFFUNC(static,avg,no_rnd,y,16,OP_Y,avg32)
yading@10: DEFFUNC(static,avg,   rnd,xy,16,OP_XY,PACK)
yading@10: DEFFUNC(static,avg,no_rnd,xy,16,OP_XY,PACK)
yading@10: 
yading@10: #undef OP
yading@10: 
yading@10: #define         ff_put_no_rnd_pixels8_o     ff_put_rnd_pixels8_o
yading@10: #define         ff_put_no_rnd_pixels16_o    ff_put_rnd_pixels16_o
yading@10: #define         ff_avg_no_rnd_pixels16_o    ff_avg_rnd_pixels16_o
yading@10: 
yading@10: av_cold void ff_hpeldsp_init_sh4(HpelDSPContext *c, int flags)
yading@10: {
yading@10:     c->put_pixels_tab[0][0] = ff_put_rnd_pixels16_o;
yading@10:     c->put_pixels_tab[0][1] = put_rnd_pixels16_x;
yading@10:     c->put_pixels_tab[0][2] = put_rnd_pixels16_y;
yading@10:     c->put_pixels_tab[0][3] = put_rnd_pixels16_xy;
yading@10:     c->put_pixels_tab[1][0] = ff_put_rnd_pixels8_o;
yading@10:     c->put_pixels_tab[1][1] = put_rnd_pixels8_x;
yading@10:     c->put_pixels_tab[1][2] = put_rnd_pixels8_y;
yading@10:     c->put_pixels_tab[1][3] = put_rnd_pixels8_xy;
yading@10: 
yading@10:     c->put_no_rnd_pixels_tab[0][0] = ff_put_no_rnd_pixels16_o;
yading@10:     c->put_no_rnd_pixels_tab[0][1] = put_no_rnd_pixels16_x;
yading@10:     c->put_no_rnd_pixels_tab[0][2] = put_no_rnd_pixels16_y;
yading@10:     c->put_no_rnd_pixels_tab[0][3] = put_no_rnd_pixels16_xy;
yading@10:     c->put_no_rnd_pixels_tab[1][0] = ff_put_no_rnd_pixels8_o;
yading@10:     c->put_no_rnd_pixels_tab[1][1] = put_no_rnd_pixels8_x;
yading@10:     c->put_no_rnd_pixels_tab[1][2] = put_no_rnd_pixels8_y;
yading@10:     c->put_no_rnd_pixels_tab[1][3] = put_no_rnd_pixels8_xy;
yading@10: 
yading@10:     c->avg_pixels_tab[0][0] = ff_avg_rnd_pixels16_o;
yading@10:     c->avg_pixels_tab[0][1] = avg_rnd_pixels16_x;
yading@10:     c->avg_pixels_tab[0][2] = avg_rnd_pixels16_y;
yading@10:     c->avg_pixels_tab[0][3] = avg_rnd_pixels16_xy;
yading@10:     c->avg_pixels_tab[1][0] = ff_avg_rnd_pixels8_o;
yading@10:     c->avg_pixels_tab[1][1] = avg_rnd_pixels8_x;
yading@10:     c->avg_pixels_tab[1][2] = avg_rnd_pixels8_y;
yading@10:     c->avg_pixels_tab[1][3] = avg_rnd_pixels8_xy;
yading@10: 
yading@10:     c->avg_no_rnd_pixels_tab[0] = ff_avg_no_rnd_pixels16_o;
yading@10:     c->avg_no_rnd_pixels_tab[1] = avg_no_rnd_pixels16_x;
yading@10:     c->avg_no_rnd_pixels_tab[2] = avg_no_rnd_pixels16_y;
yading@10:     c->avg_no_rnd_pixels_tab[3] = avg_no_rnd_pixels16_xy;
yading@10: }