FFmpeg: libavcodec/ivi_dsp.c Source File

00001 /*
00002  * DSP functions for Indeo Video Interactive codecs (Indeo4 and Indeo5)
00003  *
00004  * Copyright (c) 2009-2011 Maxim Poliakovski
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 
00029 #include "avcodec.h"
00030 #include "dsputil.h"
00031 #include "dwt.h"
00032 #include "ivi_common.h"
00033 #include "ivi_dsp.h"
00034 
00035 void ff_ivi_recompose53(const IVIPlaneDesc *plane, uint8_t *dst,
00036                         const int dst_pitch)
00037 {
00038     int             x, y, indx;
00039     int32_t         p0, p1, p2, p3, tmp0, tmp1, tmp2;
00040     int32_t         b0_1, b0_2, b1_1, b1_2, b1_3, b2_1, b2_2, b2_3, b2_4, b2_5, b2_6;
00041     int32_t         b3_1, b3_2, b3_3, b3_4, b3_5, b3_6, b3_7, b3_8, b3_9;
00042     int32_t         pitch, back_pitch;
00043     const IDWTELEM *b0_ptr, *b1_ptr, *b2_ptr, *b3_ptr;
00044     const int       num_bands = 4;
00045 
00046     /* all bands should have the same pitch */
00047     pitch = plane->bands[0].pitch;
00048 
00049     /* pixels at the position "y-1" will be set to pixels at the "y" for the 1st iteration */
00050     back_pitch = 0;
00051 
00052     /* get pointers to the wavelet bands */
00053     b0_ptr = plane->bands[0].buf;
00054     b1_ptr = plane->bands[1].buf;
00055     b2_ptr = plane->bands[2].buf;
00056     b3_ptr = plane->bands[3].buf;
00057 
00058     for (y = 0; y < plane->height; y += 2) {
00059 
00060         if (y+2 >= plane->height)
00061             pitch= 0;
00062         /* load storage variables with values */
00063         if (num_bands > 0) {
00064             b0_1 = b0_ptr[0];
00065             b0_2 = b0_ptr[pitch];
00066         }
00067 
00068         if (num_bands > 1) {
00069             b1_1 = b1_ptr[back_pitch];
00070             b1_2 = b1_ptr[0];
00071             b1_3 = b1_1 - b1_2*6 + b1_ptr[pitch];
00072         }
00073 
00074         if (num_bands > 2) {
00075             b2_2 = b2_ptr[0];     // b2[x,  y  ]
00076             b2_3 = b2_2;          // b2[x+1,y  ] = b2[x,y]
00077             b2_5 = b2_ptr[pitch]; // b2[x  ,y+1]
00078             b2_6 = b2_5;          // b2[x+1,y+1] = b2[x,y+1]
00079         }
00080 
00081         if (num_bands > 3) {
00082             b3_2 = b3_ptr[back_pitch]; // b3[x  ,y-1]
00083             b3_3 = b3_2;               // b3[x+1,y-1] = b3[x  ,y-1]
00084             b3_5 = b3_ptr[0];          // b3[x  ,y  ]
00085             b3_6 = b3_5;               // b3[x+1,y  ] = b3[x  ,y  ]
00086             b3_8 = b3_2 - b3_5*6 + b3_ptr[pitch];
00087             b3_9 = b3_8;
00088         }
00089 
00090         for (x = 0, indx = 0; x < plane->width; x+=2, indx++) {
00091             if (x+2 >= plane->width) {
00092                 b0_ptr --;
00093                 b1_ptr --;
00094                 b2_ptr --;
00095                 b3_ptr --;
00096             }
00097 
00098             /* some values calculated in the previous iterations can */
00099             /* be reused in the next ones, so do appropriate copying */
00100             b2_1 = b2_2; // b2[x-1,y  ] = b2[x,  y  ]
00101             b2_2 = b2_3; // b2[x  ,y  ] = b2[x+1,y  ]
00102             b2_4 = b2_5; // b2[x-1,y+1] = b2[x  ,y+1]
00103             b2_5 = b2_6; // b2[x  ,y+1] = b2[x+1,y+1]
00104             b3_1 = b3_2; // b3[x-1,y-1] = b3[x  ,y-1]
00105             b3_2 = b3_3; // b3[x  ,y-1] = b3[x+1,y-1]
00106             b3_4 = b3_5; // b3[x-1,y  ] = b3[x  ,y  ]
00107             b3_5 = b3_6; // b3[x  ,y  ] = b3[x+1,y  ]
00108             b3_7 = b3_8; // vert_HPF(x-1)
00109             b3_8 = b3_9; // vert_HPF(x  )
00110 
00111             p0 = p1 = p2 = p3 = 0;
00112 
00113             /* process the LL-band by applying LPF both vertically and horizontally */
00114             if (num_bands > 0) {
00115                 tmp0 = b0_1;
00116                 tmp2 = b0_2;
00117                 b0_1 = b0_ptr[indx+1];
00118                 b0_2 = b0_ptr[pitch+indx+1];
00119                 tmp1 = tmp0 + b0_1;
00120 
00121                 p0 =  tmp0 << 4;
00122                 p1 =  tmp1 << 3;
00123                 p2 = (tmp0 + tmp2) << 3;
00124                 p3 = (tmp1 + tmp2 + b0_2) << 2;
00125             }
00126 
00127             /* process the HL-band by applying HPF vertically and LPF horizontally */
00128             if (num_bands > 1) {
00129                 tmp0 = b1_2;
00130                 tmp1 = b1_1;
00131                 b1_2 = b1_ptr[indx+1];
00132                 b1_1 = b1_ptr[back_pitch+indx+1];
00133 
00134                 tmp2 = tmp1 - tmp0*6 + b1_3;
00135                 b1_3 = b1_1 - b1_2*6 + b1_ptr[pitch+indx+1];
00136 
00137                 p0 += (tmp0 + tmp1) << 3;
00138                 p1 += (tmp0 + tmp1 + b1_1 + b1_2) << 2;
00139                 p2 +=  tmp2 << 2;
00140                 p3 += (tmp2 + b1_3) << 1;
00141             }
00142 
00143             /* process the LH-band by applying LPF vertically and HPF horizontally */
00144             if (num_bands > 2) {
00145                 b2_3 = b2_ptr[indx+1];
00146                 b2_6 = b2_ptr[pitch+indx+1];
00147 
00148                 tmp0 = b2_1 + b2_2;
00149                 tmp1 = b2_1 - b2_2*6 + b2_3;
00150 
00151                 p0 += tmp0 << 3;
00152                 p1 += tmp1 << 2;
00153                 p2 += (tmp0 + b2_4 + b2_5) << 2;
00154                 p3 += (tmp1 + b2_4 - b2_5*6 + b2_6) << 1;
00155             }
00156 
00157             /* process the HH-band by applying HPF both vertically and horizontally */
00158             if (num_bands > 3) {
00159                 b3_6 = b3_ptr[indx+1];            // b3[x+1,y  ]
00160                 b3_3 = b3_ptr[back_pitch+indx+1]; // b3[x+1,y-1]
00161 
00162                 tmp0 = b3_1 + b3_4;
00163                 tmp1 = b3_2 + b3_5;
00164                 tmp2 = b3_3 + b3_6;
00165 
00166                 b3_9 = b3_3 - b3_6*6 + b3_ptr[pitch+indx+1];
00167 
00168                 p0 += (tmp0 + tmp1) << 2;
00169                 p1 += (tmp0 - tmp1*6 + tmp2) << 1;
00170                 p2 += (b3_7 + b3_8) << 1;
00171                 p3 +=  b3_7 - b3_8*6 + b3_9;
00172             }
00173 
00174             /* output four pixels */
00175             dst[x]             = av_clip_uint8((p0 >> 6) + 128);
00176             dst[x+1]           = av_clip_uint8((p1 >> 6) + 128);
00177             dst[dst_pitch+x]   = av_clip_uint8((p2 >> 6) + 128);
00178             dst[dst_pitch+x+1] = av_clip_uint8((p3 >> 6) + 128);
00179         }// for x
00180 
00181         dst += dst_pitch << 1;
00182 
00183         back_pitch = -pitch;
00184 
00185         b0_ptr += pitch + 1;
00186         b1_ptr += pitch + 1;
00187         b2_ptr += pitch + 1;
00188         b3_ptr += pitch + 1;
00189     }
00190 }
00191 
00192 void ff_ivi_recompose_haar(const IVIPlaneDesc *plane, uint8_t *dst,
00193                            const int dst_pitch)
00194 {
00195     int             x, y, indx, b0, b1, b2, b3, p0, p1, p2, p3;
00196     const IDWTELEM *b0_ptr, *b1_ptr, *b2_ptr, *b3_ptr;
00197     int32_t         pitch;
00198 
00199     /* all bands should have the same pitch */
00200     pitch = plane->bands[0].pitch;
00201 
00202     /* get pointers to the wavelet bands */
00203     b0_ptr = plane->bands[0].buf;
00204     b1_ptr = plane->bands[1].buf;
00205     b2_ptr = plane->bands[2].buf;
00206     b3_ptr = plane->bands[3].buf;
00207 
00208     for (y = 0; y < plane->height; y += 2) {
00209         for (x = 0, indx = 0; x < plane->width; x += 2, indx++) {
00210             /* load coefficients */
00211             b0 = b0_ptr[indx]; //should be: b0 = (num_bands > 0) ? b0_ptr[indx] : 0;
00212             b1 = b1_ptr[indx]; //should be: b1 = (num_bands > 1) ? b1_ptr[indx] : 0;
00213             b2 = b2_ptr[indx]; //should be: b2 = (num_bands > 2) ? b2_ptr[indx] : 0;
00214             b3 = b3_ptr[indx]; //should be: b3 = (num_bands > 3) ? b3_ptr[indx] : 0;
00215 
00216             /* haar wavelet recomposition */
00217             p0 = (b0 + b1 + b2 + b3 + 2) >> 2;
00218             p1 = (b0 + b1 - b2 - b3 + 2) >> 2;
00219             p2 = (b0 - b1 + b2 - b3 + 2) >> 2;
00220             p3 = (b0 - b1 - b2 + b3 + 2) >> 2;
00221 
00222             /* bias, convert and output four pixels */
00223             dst[x]                 = av_clip_uint8(p0 + 128);
00224             dst[x + 1]             = av_clip_uint8(p1 + 128);
00225             dst[dst_pitch + x]     = av_clip_uint8(p2 + 128);
00226             dst[dst_pitch + x + 1] = av_clip_uint8(p3 + 128);
00227         }// for x
00228 
00229         dst += dst_pitch << 1;
00230 
00231         b0_ptr += pitch;
00232         b1_ptr += pitch;
00233         b2_ptr += pitch;
00234         b3_ptr += pitch;
00235     }// for y
00236 }
00237 
00239 #define IVI_HAAR_BFLY(s1, s2, o1, o2, t) \
00240     t  = (s1 - s2) >> 1;\
00241     o1 = (s1 + s2) >> 1;\
00242     o2 = t;\
00243 
00244 
00245 #define INV_HAAR8(s1, s5, s3, s7, s2, s4, s6, s8,\
00246                   d1, d2, d3, d4, d5, d6, d7, d8,\
00247                   t0, t1, t2, t3, t4, t5, t6, t7, t8) {\
00248     t1 = s1 << 1; t5 = s5 << 1;\
00249     IVI_HAAR_BFLY(t1, t5, t1, t5, t0); IVI_HAAR_BFLY(t1, s3, t1, t3, t0);\
00250     IVI_HAAR_BFLY(t5, s7, t5, t7, t0); IVI_HAAR_BFLY(t1, s2, t1, t2, t0);\
00251     IVI_HAAR_BFLY(t3, s4, t3, t4, t0); IVI_HAAR_BFLY(t5, s6, t5, t6, t0);\
00252     IVI_HAAR_BFLY(t7, s8, t7, t8, t0);\
00253     d1 = COMPENSATE(t1);\
00254     d2 = COMPENSATE(t2);\
00255     d3 = COMPENSATE(t3);\
00256     d4 = COMPENSATE(t4);\
00257     d5 = COMPENSATE(t5);\
00258     d6 = COMPENSATE(t6);\
00259     d7 = COMPENSATE(t7);\
00260     d8 = COMPENSATE(t8); }
00261 
00263 #define INV_HAAR4(s1, s3, s5, s7) {\
00264     HAAR_BFLY(s1, s5);  HAAR_BFLY(s1, s3);  HAAR_BFLY(s5, s7);\
00265     s1 = COMPENSATE(s1);\
00266     s3 = COMPENSATE(s3);\
00267     s5 = COMPENSATE(s5);\
00268     s7 = COMPENSATE(s7); }
00269 
00270 void ff_ivi_inverse_haar_8x8(const int32_t *in, int16_t *out, uint32_t pitch,
00271                              const uint8_t *flags)
00272 {
00273     int     i, shift, sp1, sp2, sp3, sp4;
00274     const int32_t *src;
00275     int32_t *dst;
00276     int     tmp[64];
00277     int     t0, t1, t2, t3, t4, t5, t6, t7, t8;
00278 
00279     /* apply the InvHaar8 to all columns */
00280 #define COMPENSATE(x) (x)
00281     src = in;
00282     dst = tmp;
00283     for (i = 0; i < 8; i++) {
00284         if (flags[i]) {
00285             /* pre-scaling */
00286             shift = !(i & 4);
00287             sp1 = src[ 0] << shift;
00288             sp2 = src[ 8] << shift;
00289             sp3 = src[16] << shift;
00290             sp4 = src[24] << shift;
00291             INV_HAAR8(    sp1,     sp2,     sp3,     sp4,
00292                       src[32], src[40], src[48], src[56],
00293                       dst[ 0], dst[ 8], dst[16], dst[24],
00294                       dst[32], dst[40], dst[48], dst[56],
00295                       t0, t1, t2, t3, t4, t5, t6, t7, t8);
00296         } else
00297             dst[ 0] = dst[ 8] = dst[16] = dst[24] =
00298             dst[32] = dst[40] = dst[48] = dst[56] = 0;
00299 
00300         src++;
00301         dst++;
00302     }
00303 #undef  COMPENSATE
00304 
00305     /* apply the InvHaar8 to all rows */
00306 #define COMPENSATE(x) (x)
00307     src = tmp;
00308     for (i = 0; i < 8; i++) {
00309         if (   !src[0] && !src[1] && !src[2] && !src[3]
00310             && !src[4] && !src[5] && !src[6] && !src[7]) {
00311             memset(out, 0, 8 * sizeof(out[0]));
00312         } else {
00313             INV_HAAR8(src[0], src[1], src[2], src[3],
00314                       src[4], src[5], src[6], src[7],
00315                       out[0], out[1], out[2], out[3],
00316                       out[4], out[5], out[6], out[7],
00317                       t0, t1, t2, t3, t4, t5, t6, t7, t8);
00318         }
00319         src += 8;
00320         out += pitch;
00321     }
00322 #undef  COMPENSATE
00323 }
00324 
00325 void ff_ivi_dc_haar_2d(const int32_t *in, int16_t *out, uint32_t pitch,
00326                        int blk_size)
00327 {
00328     int     x, y;
00329     int16_t dc_coeff;
00330 
00331     dc_coeff = (*in + 0) >> 3;
00332 
00333     for (y = 0; y < blk_size; out += pitch, y++) {
00334         for (x = 0; x < blk_size; x++)
00335             out[x] = dc_coeff;
00336     }
00337 }
00338 
00340 #define IVI_SLANT_BFLY(s1, s2, o1, o2, t) \
00341     t  = s1 - s2;\
00342     o1 = s1 + s2;\
00343     o2 = t;\
00344 
00345 
00346 #define IVI_IREFLECT(s1, s2, o1, o2, t) \
00347     t  = ((s1 + s2*2 + 2) >> 2) + s1;\
00348     o2 = ((s1*2 - s2 + 2) >> 2) - s2;\
00349     o1 = t;\
00350 
00351 
00352 #define IVI_SLANT_PART4(s1, s2, o1, o2, t) \
00353     t  = s2 + ((s1*4  - s2 + 4) >> 3);\
00354     o2 = s1 + ((-s1 - s2*4 + 4) >> 3);\
00355     o1 = t;\
00356 
00357 
00358 #define IVI_INV_SLANT8(s1, s4, s8, s5, s2, s6, s3, s7,\
00359                        d1, d2, d3, d4, d5, d6, d7, d8,\
00360                        t0, t1, t2, t3, t4, t5, t6, t7, t8) {\
00361     IVI_SLANT_PART4(s4, s5, t4, t5, t0);\
00362 \
00363     IVI_SLANT_BFLY(s1, t5, t1, t5, t0); IVI_SLANT_BFLY(s2, s6, t2, t6, t0);\
00364     IVI_SLANT_BFLY(s7, s3, t7, t3, t0); IVI_SLANT_BFLY(t4, s8, t4, t8, t0);\
00365 \
00366     IVI_SLANT_BFLY(t1, t2, t1, t2, t0); IVI_IREFLECT  (t4, t3, t4, t3, t0);\
00367     IVI_SLANT_BFLY(t5, t6, t5, t6, t0); IVI_IREFLECT  (t8, t7, t8, t7, t0);\
00368     IVI_SLANT_BFLY(t1, t4, t1, t4, t0); IVI_SLANT_BFLY(t2, t3, t2, t3, t0);\
00369     IVI_SLANT_BFLY(t5, t8, t5, t8, t0); IVI_SLANT_BFLY(t6, t7, t6, t7, t0);\
00370     d1 = COMPENSATE(t1);\
00371     d2 = COMPENSATE(t2);\
00372     d3 = COMPENSATE(t3);\
00373     d4 = COMPENSATE(t4);\
00374     d5 = COMPENSATE(t5);\
00375     d6 = COMPENSATE(t6);\
00376     d7 = COMPENSATE(t7);\
00377     d8 = COMPENSATE(t8);}
00378 
00380 #define IVI_INV_SLANT4(s1, s4, s2, s3, d1, d2, d3, d4, t0, t1, t2, t3, t4) {\
00381     IVI_SLANT_BFLY(s1, s2, t1, t2, t0); IVI_IREFLECT  (s4, s3, t4, t3, t0);\
00382 \
00383     IVI_SLANT_BFLY(t1, t4, t1, t4, t0); IVI_SLANT_BFLY(t2, t3, t2, t3, t0);\
00384     d1 = COMPENSATE(t1);\
00385     d2 = COMPENSATE(t2);\
00386     d3 = COMPENSATE(t3);\
00387     d4 = COMPENSATE(t4);}
00388 
00389 void ff_ivi_inverse_slant_8x8(const int32_t *in, int16_t *out, uint32_t pitch, const uint8_t *flags)
00390 {
00391     int     i;
00392     const int32_t *src;
00393     int32_t *dst;
00394     int     tmp[64];
00395     int     t0, t1, t2, t3, t4, t5, t6, t7, t8;
00396 
00397 #define COMPENSATE(x) (x)
00398     src = in;
00399     dst = tmp;
00400     for (i = 0; i < 8; i++) {
00401         if (flags[i]) {
00402             IVI_INV_SLANT8(src[0], src[8], src[16], src[24], src[32], src[40], src[48], src[56],
00403                            dst[0], dst[8], dst[16], dst[24], dst[32], dst[40], dst[48], dst[56],
00404                            t0, t1, t2, t3, t4, t5, t6, t7, t8);
00405         } else
00406             dst[0] = dst[8] = dst[16] = dst[24] = dst[32] = dst[40] = dst[48] = dst[56] = 0;
00407 
00408             src++;
00409             dst++;
00410     }
00411 #undef COMPENSATE
00412 
00413 #define COMPENSATE(x) ((x + 1)>>1)
00414     src = tmp;
00415     for (i = 0; i < 8; i++) {
00416         if (!src[0] && !src[1] && !src[2] && !src[3] && !src[4] && !src[5] && !src[6] && !src[7]) {
00417             memset(out, 0, 8*sizeof(out[0]));
00418         } else {
00419             IVI_INV_SLANT8(src[0], src[1], src[2], src[3], src[4], src[5], src[6], src[7],
00420                            out[0], out[1], out[2], out[3], out[4], out[5], out[6], out[7],
00421                            t0, t1, t2, t3, t4, t5, t6, t7, t8);
00422         }
00423         src += 8;
00424         out += pitch;
00425     }
00426 #undef COMPENSATE
00427 }
00428 
00429 void ff_ivi_inverse_slant_4x4(const int32_t *in, int16_t *out, uint32_t pitch, const uint8_t *flags)
00430 {
00431     int     i;
00432     const int32_t *src;
00433     int32_t *dst;
00434     int     tmp[16];
00435     int     t0, t1, t2, t3, t4;
00436 
00437 #define COMPENSATE(x) (x)
00438     src = in;
00439     dst = tmp;
00440     for (i = 0; i < 4; i++) {
00441         if (flags[i]) {
00442             IVI_INV_SLANT4(src[0], src[4], src[8], src[12],
00443                            dst[0], dst[4], dst[8], dst[12],
00444                            t0, t1, t2, t3, t4);
00445         } else
00446             dst[0] = dst[4] = dst[8] = dst[12] = 0;
00447 
00448             src++;
00449             dst++;
00450     }
00451 #undef COMPENSATE
00452 
00453 #define COMPENSATE(x) ((x + 1)>>1)
00454     src = tmp;
00455     for (i = 0; i < 4; i++) {
00456         if (!src[0] && !src[1] && !src[2] && !src[3]) {
00457             out[0] = out[1] = out[2] = out[3] = 0;
00458         } else {
00459             IVI_INV_SLANT4(src[0], src[1], src[2], src[3],
00460                            out[0], out[1], out[2], out[3],
00461                            t0, t1, t2, t3, t4);
00462         }
00463         src += 4;
00464         out += pitch;
00465     }
00466 #undef COMPENSATE
00467 }
00468 
00469 void ff_ivi_dc_slant_2d(const int32_t *in, int16_t *out, uint32_t pitch, int blk_size)
00470 {
00471     int     x, y;
00472     int16_t dc_coeff;
00473 
00474     dc_coeff = (*in + 1) >> 1;
00475 
00476     for (y = 0; y < blk_size; out += pitch, y++) {
00477         for (x = 0; x < blk_size; x++)
00478             out[x] = dc_coeff;
00479     }
00480 }
00481 
00482 void ff_ivi_row_slant8(const int32_t *in, int16_t *out, uint32_t pitch, const uint8_t *flags)
00483 {
00484     int     i;
00485     int     t0, t1, t2, t3, t4, t5, t6, t7, t8;
00486 
00487 #define COMPENSATE(x) ((x + 1)>>1)
00488     for (i = 0; i < 8; i++) {
00489         if (!in[0] && !in[1] && !in[2] && !in[3] && !in[4] && !in[5] && !in[6] && !in[7]) {
00490             memset(out, 0, 8*sizeof(out[0]));
00491         } else {
00492             IVI_INV_SLANT8( in[0],  in[1],  in[2],  in[3],  in[4],  in[5],  in[6],  in[7],
00493                            out[0], out[1], out[2], out[3], out[4], out[5], out[6], out[7],
00494                            t0, t1, t2, t3, t4, t5, t6, t7, t8);
00495         }
00496         in += 8;
00497         out += pitch;
00498     }
00499 #undef COMPENSATE
00500 }
00501 
00502 void ff_ivi_dc_row_slant(const int32_t *in, int16_t *out, uint32_t pitch, int blk_size)
00503 {
00504     int     x, y;
00505     int16_t dc_coeff;
00506 
00507     dc_coeff = (*in + 1) >> 1;
00508 
00509     for (x = 0; x < blk_size; x++)
00510         out[x] = dc_coeff;
00511 
00512     out += pitch;
00513 
00514     for (y = 1; y < blk_size; out += pitch, y++) {
00515         for (x = 0; x < blk_size; x++)
00516             out[x] = 0;
00517     }
00518 }
00519 
00520 void ff_ivi_col_slant8(const int32_t *in, int16_t *out, uint32_t pitch, const uint8_t *flags)
00521 {
00522     int     i, row2, row4, row8;
00523     int     t0, t1, t2, t3, t4, t5, t6, t7, t8;
00524 
00525     row2 = pitch << 1;
00526     row4 = pitch << 2;
00527     row8 = pitch << 3;
00528 
00529 #define COMPENSATE(x) ((x + 1)>>1)
00530     for (i = 0; i < 8; i++) {
00531         if (flags[i]) {
00532             IVI_INV_SLANT8(in[0], in[8], in[16], in[24], in[32], in[40], in[48], in[56],
00533                            out[0], out[pitch], out[row2], out[row2 + pitch], out[row4],
00534                            out[row4 + pitch],  out[row4 + row2], out[row8 - pitch],
00535                            t0, t1, t2, t3, t4, t5, t6, t7, t8);
00536         } else {
00537             out[0] = out[pitch] = out[row2] = out[row2 + pitch] = out[row4] =
00538             out[row4 + pitch] =  out[row4 + row2] = out[row8 - pitch] = 0;
00539         }
00540 
00541         in++;
00542         out++;
00543     }
00544 #undef COMPENSATE
00545 }
00546 
00547 void ff_ivi_dc_col_slant(const int32_t *in, int16_t *out, uint32_t pitch, int blk_size)
00548 {
00549     int     x, y;
00550     int16_t dc_coeff;
00551 
00552     dc_coeff = (*in + 1) >> 1;
00553 
00554     for (y = 0; y < blk_size; out += pitch, y++) {
00555         out[0] = dc_coeff;
00556         for (x = 1; x < blk_size; x++)
00557             out[x] = 0;
00558     }
00559 }
00560 
00561 void ff_ivi_put_pixels_8x8(const int32_t *in, int16_t *out, uint32_t pitch,
00562                            const uint8_t *flags)
00563 {
00564     int     x, y;
00565 
00566     for (y = 0; y < 8; out += pitch, in += 8, y++)
00567         for (x = 0; x < 8; x++)
00568             out[x] = in[x];
00569 }
00570 
00571 void ff_ivi_put_dc_pixel_8x8(const int32_t *in, int16_t *out, uint32_t pitch,
00572                              int blk_size)
00573 {
00574     int     y;
00575 
00576     out[0] = in[0];
00577     memset(out + 1, 0, 7*sizeof(out[0]));
00578     out += pitch;
00579 
00580     for (y = 1; y < 8; out += pitch, y++)
00581         memset(out, 0, 8*sizeof(out[0]));
00582 }
00583 
00584 #define IVI_MC_TEMPLATE(size, suffix, OP) \
00585 void ff_ivi_mc_ ## size ##x## size ## suffix (int16_t *buf, const int16_t *ref_buf, \
00586                                               uint32_t pitch, int mc_type) \
00587 { \
00588     int     i, j; \
00589     const int16_t *wptr; \
00590 \
00591     switch (mc_type) { \
00592     case 0: /* fullpel (no interpolation) */ \
00593         for (i = 0; i < size; i++, buf += pitch, ref_buf += pitch) { \
00594             for (j = 0; j < size; j++) {\
00595                 OP(buf[j], ref_buf[j]); \
00596             } \
00597         } \
00598         break; \
00599     case 1: /* horizontal halfpel interpolation */ \
00600         for (i = 0; i < size; i++, buf += pitch, ref_buf += pitch) \
00601             for (j = 0; j < size; j++) \
00602                 OP(buf[j], (ref_buf[j] + ref_buf[j+1]) >> 1); \
00603         break; \
00604     case 2: /* vertical halfpel interpolation */ \
00605         wptr = ref_buf + pitch; \
00606         for (i = 0; i < size; i++, buf += pitch, wptr += pitch, ref_buf += pitch) \
00607             for (j = 0; j < size; j++) \
00608                 OP(buf[j], (ref_buf[j] + wptr[j]) >> 1); \
00609         break; \
00610     case 3: /* vertical and horizontal halfpel interpolation */ \
00611         wptr = ref_buf + pitch; \
00612         for (i = 0; i < size; i++, buf += pitch, wptr += pitch, ref_buf += pitch) \
00613             for (j = 0; j < size; j++) \
00614                 OP(buf[j], (ref_buf[j] + ref_buf[j+1] + wptr[j] + wptr[j+1]) >> 2); \
00615         break; \
00616     } \
00617 } \
00618 
00619 #define OP_PUT(a, b)  (a) = (b)
00620 #define OP_ADD(a, b)  (a) += (b)
00621 
00622 IVI_MC_TEMPLATE(8, _no_delta, OP_PUT)
00623 IVI_MC_TEMPLATE(8, _delta,    OP_ADD)
00624 IVI_MC_TEMPLATE(4, _no_delta, OP_PUT)
00625 IVI_MC_TEMPLATE(4, _delta,    OP_ADD)