yading@10: /* yading@10: * Simple IDCT yading@10: * yading@10: * Copyright (c) 2001 Michael Niedermayer 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: /** yading@10: * @file yading@10: * simpleidct in C. yading@10: */ yading@10: yading@10: #include "libavutil/intreadwrite.h" yading@10: #include "avcodec.h" yading@10: #include "mathops.h" yading@10: #include "simple_idct.h" yading@10: yading@10: #define BIT_DEPTH 8 yading@10: #include "simple_idct_template.c" yading@10: #undef BIT_DEPTH yading@10: yading@10: #define BIT_DEPTH 10 yading@10: #include "simple_idct_template.c" yading@10: #undef BIT_DEPTH yading@10: yading@10: /* 2x4x8 idct */ yading@10: yading@10: #define CN_SHIFT 12 yading@10: #define C_FIX(x) ((int)((x) * (1 << CN_SHIFT) + 0.5)) yading@10: #define C1 C_FIX(0.6532814824) yading@10: #define C2 C_FIX(0.2705980501) yading@10: yading@10: /* row idct is multiple by 16 * sqrt(2.0), col idct4 is normalized, yading@10: and the butterfly must be multiplied by 0.5 * sqrt(2.0) */ yading@10: #define C_SHIFT (4+1+12) yading@10: yading@10: static inline void idct4col_put(uint8_t *dest, int line_size, const int16_t *col) yading@10: { yading@10: int c0, c1, c2, c3, a0, a1, a2, a3; yading@10: yading@10: a0 = col[8*0]; yading@10: a1 = col[8*2]; yading@10: a2 = col[8*4]; yading@10: a3 = col[8*6]; yading@10: c0 = ((a0 + a2) << (CN_SHIFT - 1)) + (1 << (C_SHIFT - 1)); yading@10: c2 = ((a0 - a2) << (CN_SHIFT - 1)) + (1 << (C_SHIFT - 1)); yading@10: c1 = a1 * C1 + a3 * C2; yading@10: c3 = a1 * C2 - a3 * C1; yading@10: dest[0] = av_clip_uint8((c0 + c1) >> C_SHIFT); yading@10: dest += line_size; yading@10: dest[0] = av_clip_uint8((c2 + c3) >> C_SHIFT); yading@10: dest += line_size; yading@10: dest[0] = av_clip_uint8((c2 - c3) >> C_SHIFT); yading@10: dest += line_size; yading@10: dest[0] = av_clip_uint8((c0 - c1) >> C_SHIFT); yading@10: } yading@10: yading@10: #define BF(k) \ yading@10: {\ yading@10: int a0, a1;\ yading@10: a0 = ptr[k];\ yading@10: a1 = ptr[8 + k];\ yading@10: ptr[k] = a0 + a1;\ yading@10: ptr[8 + k] = a0 - a1;\ yading@10: } yading@10: yading@10: /* only used by DV codec. The input must be interlaced. 128 is added yading@10: to the pixels before clamping to avoid systematic error yading@10: (1024*sqrt(2)) offset would be needed otherwise. */ yading@10: /* XXX: I think a 1.0/sqrt(2) normalization should be needed to yading@10: compensate the extra butterfly stage - I don't have the full DV yading@10: specification */ yading@10: void ff_simple_idct248_put(uint8_t *dest, int line_size, int16_t *block) yading@10: { yading@10: int i; yading@10: int16_t *ptr; yading@10: yading@10: /* butterfly */ yading@10: ptr = block; yading@10: for(i=0;i<4;i++) { yading@10: BF(0); yading@10: BF(1); yading@10: BF(2); yading@10: BF(3); yading@10: BF(4); yading@10: BF(5); yading@10: BF(6); yading@10: BF(7); yading@10: ptr += 2 * 8; yading@10: } yading@10: yading@10: /* IDCT8 on each line */ yading@10: for(i=0; i<8; i++) { yading@10: idctRowCondDC_8(block + i*8, 0); yading@10: } yading@10: yading@10: /* IDCT4 and store */ yading@10: for(i=0;i<8;i++) { yading@10: idct4col_put(dest + i, 2 * line_size, block + i); yading@10: idct4col_put(dest + line_size + i, 2 * line_size, block + 8 + i); yading@10: } yading@10: } yading@10: yading@10: /* 8x4 & 4x8 WMV2 IDCT */ yading@10: #undef CN_SHIFT yading@10: #undef C_SHIFT yading@10: #undef C_FIX yading@10: #undef C1 yading@10: #undef C2 yading@10: #define CN_SHIFT 12 yading@10: #define C_FIX(x) ((int)((x) * 1.414213562 * (1 << CN_SHIFT) + 0.5)) yading@10: #define C1 C_FIX(0.6532814824) yading@10: #define C2 C_FIX(0.2705980501) yading@10: #define C3 C_FIX(0.5) yading@10: #define C_SHIFT (4+1+12) yading@10: static inline void idct4col_add(uint8_t *dest, int line_size, const int16_t *col) yading@10: { yading@10: int c0, c1, c2, c3, a0, a1, a2, a3; yading@10: yading@10: a0 = col[8*0]; yading@10: a1 = col[8*1]; yading@10: a2 = col[8*2]; yading@10: a3 = col[8*3]; yading@10: c0 = (a0 + a2)*C3 + (1 << (C_SHIFT - 1)); yading@10: c2 = (a0 - a2)*C3 + (1 << (C_SHIFT - 1)); yading@10: c1 = a1 * C1 + a3 * C2; yading@10: c3 = a1 * C2 - a3 * C1; yading@10: dest[0] = av_clip_uint8(dest[0] + ((c0 + c1) >> C_SHIFT)); yading@10: dest += line_size; yading@10: dest[0] = av_clip_uint8(dest[0] + ((c2 + c3) >> C_SHIFT)); yading@10: dest += line_size; yading@10: dest[0] = av_clip_uint8(dest[0] + ((c2 - c3) >> C_SHIFT)); yading@10: dest += line_size; yading@10: dest[0] = av_clip_uint8(dest[0] + ((c0 - c1) >> C_SHIFT)); yading@10: } yading@10: yading@10: #define RN_SHIFT 15 yading@10: #define R_FIX(x) ((int)((x) * 1.414213562 * (1 << RN_SHIFT) + 0.5)) yading@10: #define R1 R_FIX(0.6532814824) yading@10: #define R2 R_FIX(0.2705980501) yading@10: #define R3 R_FIX(0.5) yading@10: #define R_SHIFT 11 yading@10: static inline void idct4row(int16_t *row) yading@10: { yading@10: int c0, c1, c2, c3, a0, a1, a2, a3; yading@10: yading@10: a0 = row[0]; yading@10: a1 = row[1]; yading@10: a2 = row[2]; yading@10: a3 = row[3]; yading@10: c0 = (a0 + a2)*R3 + (1 << (R_SHIFT - 1)); yading@10: c2 = (a0 - a2)*R3 + (1 << (R_SHIFT - 1)); yading@10: c1 = a1 * R1 + a3 * R2; yading@10: c3 = a1 * R2 - a3 * R1; yading@10: row[0]= (c0 + c1) >> R_SHIFT; yading@10: row[1]= (c2 + c3) >> R_SHIFT; yading@10: row[2]= (c2 - c3) >> R_SHIFT; yading@10: row[3]= (c0 - c1) >> R_SHIFT; yading@10: } yading@10: yading@10: void ff_simple_idct84_add(uint8_t *dest, int line_size, int16_t *block) yading@10: { yading@10: int i; yading@10: yading@10: /* IDCT8 on each line */ yading@10: for(i=0; i<4; i++) { yading@10: idctRowCondDC_8(block + i*8, 0); yading@10: } yading@10: yading@10: /* IDCT4 and store */ yading@10: for(i=0;i<8;i++) { yading@10: idct4col_add(dest + i, line_size, block + i); yading@10: } yading@10: } yading@10: yading@10: void ff_simple_idct48_add(uint8_t *dest, int line_size, int16_t *block) yading@10: { yading@10: int i; yading@10: yading@10: /* IDCT4 on each line */ yading@10: for(i=0; i<8; i++) { yading@10: idct4row(block + i*8); yading@10: } yading@10: yading@10: /* IDCT8 and store */ yading@10: for(i=0; i<4; i++){ yading@10: idctSparseColAdd_8(dest + i, line_size, block + i); yading@10: } yading@10: } yading@10: yading@10: void ff_simple_idct44_add(uint8_t *dest, int line_size, int16_t *block) yading@10: { yading@10: int i; yading@10: yading@10: /* IDCT4 on each line */ yading@10: for(i=0; i<4; i++) { yading@10: idct4row(block + i*8); yading@10: } yading@10: yading@10: /* IDCT4 and store */ yading@10: for(i=0; i<4; i++){ yading@10: idct4col_add(dest + i, line_size, block + i); yading@10: } yading@10: } yading@10: yading@10: void ff_prores_idct(int16_t *block, const int16_t *qmat) yading@10: { yading@10: int i; yading@10: yading@10: for (i = 0; i < 64; i++) yading@10: block[i] *= qmat[i]; yading@10: yading@10: for (i = 0; i < 8; i++) yading@10: idctRowCondDC_10(block + i*8, 2); yading@10: yading@10: for (i = 0; i < 8; i++) yading@10: idctSparseCol_10(block + i); yading@10: }