nengel@2: /* nengel@2: * Simple IDCT nengel@2: * nengel@2: * Copyright (c) 2001 Michael Niedermayer nengel@2: * nengel@2: * This file is part of FFmpeg. nengel@2: * nengel@2: * FFmpeg is free software; you can redistribute it and/or nengel@2: * modify it under the terms of the GNU Lesser General Public nengel@2: * License as published by the Free Software Foundation; either nengel@2: * version 2.1 of the License, or (at your option) any later version. nengel@2: * nengel@2: * FFmpeg is distributed in the hope that it will be useful, nengel@2: * but WITHOUT ANY WARRANTY; without even the implied warranty of nengel@2: * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU nengel@2: * Lesser General Public License for more details. nengel@2: * nengel@2: * You should have received a copy of the GNU Lesser General Public nengel@2: * License along with FFmpeg; if not, write to the Free Software nengel@2: * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA nengel@2: */ nengel@2: nengel@2: /** nengel@2: * @file nengel@2: * simpleidct in C. nengel@2: */ nengel@2: nengel@2: /* nengel@2: based upon some outcommented c code from mpeg2dec (idct_mmx.c nengel@2: written by Aaron Holtzman ) nengel@2: */ nengel@2: #include "avcodec.h" nengel@2: #include "dsputil.h" nengel@2: #include "mathops.h" nengel@2: #include "simple_idct.h" nengel@2: nengel@2: #if 0 nengel@2: #define W1 2841 /* 2048*sqrt (2)*cos (1*pi/16) */ nengel@2: #define W2 2676 /* 2048*sqrt (2)*cos (2*pi/16) */ nengel@2: #define W3 2408 /* 2048*sqrt (2)*cos (3*pi/16) */ nengel@2: #define W4 2048 /* 2048*sqrt (2)*cos (4*pi/16) */ nengel@2: #define W5 1609 /* 2048*sqrt (2)*cos (5*pi/16) */ nengel@2: #define W6 1108 /* 2048*sqrt (2)*cos (6*pi/16) */ nengel@2: #define W7 565 /* 2048*sqrt (2)*cos (7*pi/16) */ nengel@2: #define ROW_SHIFT 8 nengel@2: #define COL_SHIFT 17 nengel@2: #else nengel@2: #define W1 22725 //cos(i*M_PI/16)*sqrt(2)*(1<<14) + 0.5 nengel@2: #define W2 21407 //cos(i*M_PI/16)*sqrt(2)*(1<<14) + 0.5 nengel@2: #define W3 19266 //cos(i*M_PI/16)*sqrt(2)*(1<<14) + 0.5 nengel@2: #define W4 16383 //cos(i*M_PI/16)*sqrt(2)*(1<<14) + 0.5 nengel@2: #define W5 12873 //cos(i*M_PI/16)*sqrt(2)*(1<<14) + 0.5 nengel@2: #define W6 8867 //cos(i*M_PI/16)*sqrt(2)*(1<<14) + 0.5 nengel@2: #define W7 4520 //cos(i*M_PI/16)*sqrt(2)*(1<<14) + 0.5 nengel@2: #define ROW_SHIFT 11 nengel@2: #define COL_SHIFT 20 // 6 nengel@2: #endif nengel@2: nengel@2: static inline void idctRowCondDC (DCTELEM * row) nengel@2: { nengel@2: int a0, a1, a2, a3, b0, b1, b2, b3; nengel@2: uint64_t temp; nengel@2: nengel@2: #if HAVE_BIGENDIAN nengel@2: #define ROW0_MASK 0xffff000000000000LL nengel@2: #else nengel@2: #define ROW0_MASK 0xffffLL nengel@2: #endif nengel@2: if(sizeof(DCTELEM)==2){ nengel@2: if ( ((((uint64_t *)row)[0] & ~ROW0_MASK) | nengel@2: ((uint64_t *)row)[1]) == 0) { nengel@2: temp = (row[0] << 3) & 0xffff; nengel@2: temp += temp << 16; nengel@2: temp += temp << 32; nengel@2: ((uint64_t *)row)[0] = temp; nengel@2: ((uint64_t *)row)[1] = temp; nengel@2: return; nengel@2: } nengel@2: }else{ nengel@2: if (!(row[1]|row[2]|row[3]|row[4]|row[5]|row[6]|row[7])) { nengel@2: row[0]=row[1]=row[2]=row[3]=row[4]=row[5]=row[6]=row[7]= row[0] << 3; nengel@2: return; nengel@2: } nengel@2: } nengel@2: nengel@2: a0 = (W4 * row[0]) + (1 << (ROW_SHIFT - 1)); nengel@2: a1 = a0; nengel@2: a2 = a0; nengel@2: a3 = a0; nengel@2: nengel@2: /* no need to optimize : gcc does it */ nengel@2: a0 += W2 * row[2]; nengel@2: a1 += W6 * row[2]; nengel@2: a2 -= W6 * row[2]; nengel@2: a3 -= W2 * row[2]; nengel@2: nengel@2: b0 = MUL16(W1, row[1]); nengel@2: MAC16(b0, W3, row[3]); nengel@2: b1 = MUL16(W3, row[1]); nengel@2: MAC16(b1, -W7, row[3]); nengel@2: b2 = MUL16(W5, row[1]); nengel@2: MAC16(b2, -W1, row[3]); nengel@2: b3 = MUL16(W7, row[1]); nengel@2: MAC16(b3, -W5, row[3]); nengel@2: nengel@2: temp = ((uint64_t*)row)[1]; nengel@2: nengel@2: if (temp != 0) { nengel@2: a0 += W4*row[4] + W6*row[6]; nengel@2: a1 += - W4*row[4] - W2*row[6]; nengel@2: a2 += - W4*row[4] + W2*row[6]; nengel@2: a3 += W4*row[4] - W6*row[6]; nengel@2: nengel@2: MAC16(b0, W5, row[5]); nengel@2: MAC16(b0, W7, row[7]); nengel@2: nengel@2: MAC16(b1, -W1, row[5]); nengel@2: MAC16(b1, -W5, row[7]); nengel@2: nengel@2: MAC16(b2, W7, row[5]); nengel@2: MAC16(b2, W3, row[7]); nengel@2: nengel@2: MAC16(b3, W3, row[5]); nengel@2: MAC16(b3, -W1, row[7]); nengel@2: } nengel@2: nengel@2: row[0] = (a0 + b0) >> ROW_SHIFT; nengel@2: row[7] = (a0 - b0) >> ROW_SHIFT; nengel@2: row[1] = (a1 + b1) >> ROW_SHIFT; nengel@2: row[6] = (a1 - b1) >> ROW_SHIFT; nengel@2: row[2] = (a2 + b2) >> ROW_SHIFT; nengel@2: row[5] = (a2 - b2) >> ROW_SHIFT; nengel@2: row[3] = (a3 + b3) >> ROW_SHIFT; nengel@2: row[4] = (a3 - b3) >> ROW_SHIFT; nengel@2: } nengel@2: nengel@2: static inline void idctSparseColPut (uint8_t *dest, int line_size, nengel@2: DCTELEM * col) nengel@2: { nengel@2: int a0, a1, a2, a3, b0, b1, b2, b3; nengel@2: uint8_t *cm = ff_cropTbl + MAX_NEG_CROP; nengel@2: nengel@2: /* XXX: I did that only to give same values as previous code */ nengel@2: a0 = W4 * (col[8*0] + ((1<<(COL_SHIFT-1))/W4)); nengel@2: a1 = a0; nengel@2: a2 = a0; nengel@2: a3 = a0; nengel@2: nengel@2: a0 += + W2*col[8*2]; nengel@2: a1 += + W6*col[8*2]; nengel@2: a2 += - W6*col[8*2]; nengel@2: a3 += - W2*col[8*2]; nengel@2: nengel@2: b0 = MUL16(W1, col[8*1]); nengel@2: b1 = MUL16(W3, col[8*1]); nengel@2: b2 = MUL16(W5, col[8*1]); nengel@2: b3 = MUL16(W7, col[8*1]); nengel@2: nengel@2: MAC16(b0, + W3, col[8*3]); nengel@2: MAC16(b1, - W7, col[8*3]); nengel@2: MAC16(b2, - W1, col[8*3]); nengel@2: MAC16(b3, - W5, col[8*3]); nengel@2: nengel@2: if(col[8*4]){ nengel@2: a0 += + W4*col[8*4]; nengel@2: a1 += - W4*col[8*4]; nengel@2: a2 += - W4*col[8*4]; nengel@2: a3 += + W4*col[8*4]; nengel@2: } nengel@2: nengel@2: if (col[8*5]) { nengel@2: MAC16(b0, + W5, col[8*5]); nengel@2: MAC16(b1, - W1, col[8*5]); nengel@2: MAC16(b2, + W7, col[8*5]); nengel@2: MAC16(b3, + W3, col[8*5]); nengel@2: } nengel@2: nengel@2: if(col[8*6]){ nengel@2: a0 += + W6*col[8*6]; nengel@2: a1 += - W2*col[8*6]; nengel@2: a2 += + W2*col[8*6]; nengel@2: a3 += - W6*col[8*6]; nengel@2: } nengel@2: nengel@2: if (col[8*7]) { nengel@2: MAC16(b0, + W7, col[8*7]); nengel@2: MAC16(b1, - W5, col[8*7]); nengel@2: MAC16(b2, + W3, col[8*7]); nengel@2: MAC16(b3, - W1, col[8*7]); nengel@2: } nengel@2: nengel@2: dest[0] = cm[(a0 + b0) >> COL_SHIFT]; nengel@2: dest += line_size; nengel@2: dest[0] = cm[(a1 + b1) >> COL_SHIFT]; nengel@2: dest += line_size; nengel@2: dest[0] = cm[(a2 + b2) >> COL_SHIFT]; nengel@2: dest += line_size; nengel@2: dest[0] = cm[(a3 + b3) >> COL_SHIFT]; nengel@2: dest += line_size; nengel@2: dest[0] = cm[(a3 - b3) >> COL_SHIFT]; nengel@2: dest += line_size; nengel@2: dest[0] = cm[(a2 - b2) >> COL_SHIFT]; nengel@2: dest += line_size; nengel@2: dest[0] = cm[(a1 - b1) >> COL_SHIFT]; nengel@2: dest += line_size; nengel@2: dest[0] = cm[(a0 - b0) >> COL_SHIFT]; nengel@2: } nengel@2: nengel@2: static inline void idctSparseColAdd (uint8_t *dest, int line_size, nengel@2: DCTELEM * col) nengel@2: { nengel@2: int a0, a1, a2, a3, b0, b1, b2, b3; nengel@2: uint8_t *cm = ff_cropTbl + MAX_NEG_CROP; nengel@2: nengel@2: /* XXX: I did that only to give same values as previous code */ nengel@2: a0 = W4 * (col[8*0] + ((1<<(COL_SHIFT-1))/W4)); nengel@2: a1 = a0; nengel@2: a2 = a0; nengel@2: a3 = a0; nengel@2: nengel@2: a0 += + W2*col[8*2]; nengel@2: a1 += + W6*col[8*2]; nengel@2: a2 += - W6*col[8*2]; nengel@2: a3 += - W2*col[8*2]; nengel@2: nengel@2: b0 = MUL16(W1, col[8*1]); nengel@2: b1 = MUL16(W3, col[8*1]); nengel@2: b2 = MUL16(W5, col[8*1]); nengel@2: b3 = MUL16(W7, col[8*1]); nengel@2: nengel@2: MAC16(b0, + W3, col[8*3]); nengel@2: MAC16(b1, - W7, col[8*3]); nengel@2: MAC16(b2, - W1, col[8*3]); nengel@2: MAC16(b3, - W5, col[8*3]); nengel@2: nengel@2: if(col[8*4]){ nengel@2: a0 += + W4*col[8*4]; nengel@2: a1 += - W4*col[8*4]; nengel@2: a2 += - W4*col[8*4]; nengel@2: a3 += + W4*col[8*4]; nengel@2: } nengel@2: nengel@2: if (col[8*5]) { nengel@2: MAC16(b0, + W5, col[8*5]); nengel@2: MAC16(b1, - W1, col[8*5]); nengel@2: MAC16(b2, + W7, col[8*5]); nengel@2: MAC16(b3, + W3, col[8*5]); nengel@2: } nengel@2: nengel@2: if(col[8*6]){ nengel@2: a0 += + W6*col[8*6]; nengel@2: a1 += - W2*col[8*6]; nengel@2: a2 += + W2*col[8*6]; nengel@2: a3 += - W6*col[8*6]; nengel@2: } nengel@2: nengel@2: if (col[8*7]) { nengel@2: MAC16(b0, + W7, col[8*7]); nengel@2: MAC16(b1, - W5, col[8*7]); nengel@2: MAC16(b2, + W3, col[8*7]); nengel@2: MAC16(b3, - W1, col[8*7]); nengel@2: } nengel@2: nengel@2: dest[0] = cm[dest[0] + ((a0 + b0) >> COL_SHIFT)]; nengel@2: dest += line_size; nengel@2: dest[0] = cm[dest[0] + ((a1 + b1) >> COL_SHIFT)]; nengel@2: dest += line_size; nengel@2: dest[0] = cm[dest[0] + ((a2 + b2) >> COL_SHIFT)]; nengel@2: dest += line_size; nengel@2: dest[0] = cm[dest[0] + ((a3 + b3) >> COL_SHIFT)]; nengel@2: dest += line_size; nengel@2: dest[0] = cm[dest[0] + ((a3 - b3) >> COL_SHIFT)]; nengel@2: dest += line_size; nengel@2: dest[0] = cm[dest[0] + ((a2 - b2) >> COL_SHIFT)]; nengel@2: dest += line_size; nengel@2: dest[0] = cm[dest[0] + ((a1 - b1) >> COL_SHIFT)]; nengel@2: dest += line_size; nengel@2: dest[0] = cm[dest[0] + ((a0 - b0) >> COL_SHIFT)]; nengel@2: } nengel@2: nengel@2: static inline void idctSparseCol (DCTELEM * col) nengel@2: { nengel@2: int a0, a1, a2, a3, b0, b1, b2, b3; nengel@2: nengel@2: /* XXX: I did that only to give same values as previous code */ nengel@2: a0 = W4 * (col[8*0] + ((1<<(COL_SHIFT-1))/W4)); nengel@2: a1 = a0; nengel@2: a2 = a0; nengel@2: a3 = a0; nengel@2: nengel@2: a0 += + W2*col[8*2]; nengel@2: a1 += + W6*col[8*2]; nengel@2: a2 += - W6*col[8*2]; nengel@2: a3 += - W2*col[8*2]; nengel@2: nengel@2: b0 = MUL16(W1, col[8*1]); nengel@2: b1 = MUL16(W3, col[8*1]); nengel@2: b2 = MUL16(W5, col[8*1]); nengel@2: b3 = MUL16(W7, col[8*1]); nengel@2: nengel@2: MAC16(b0, + W3, col[8*3]); nengel@2: MAC16(b1, - W7, col[8*3]); nengel@2: MAC16(b2, - W1, col[8*3]); nengel@2: MAC16(b3, - W5, col[8*3]); nengel@2: nengel@2: if(col[8*4]){ nengel@2: a0 += + W4*col[8*4]; nengel@2: a1 += - W4*col[8*4]; nengel@2: a2 += - W4*col[8*4]; nengel@2: a3 += + W4*col[8*4]; nengel@2: } nengel@2: nengel@2: if (col[8*5]) { nengel@2: MAC16(b0, + W5, col[8*5]); nengel@2: MAC16(b1, - W1, col[8*5]); nengel@2: MAC16(b2, + W7, col[8*5]); nengel@2: MAC16(b3, + W3, col[8*5]); nengel@2: } nengel@2: nengel@2: if(col[8*6]){ nengel@2: a0 += + W6*col[8*6]; nengel@2: a1 += - W2*col[8*6]; nengel@2: a2 += + W2*col[8*6]; nengel@2: a3 += - W6*col[8*6]; nengel@2: } nengel@2: nengel@2: if (col[8*7]) { nengel@2: MAC16(b0, + W7, col[8*7]); nengel@2: MAC16(b1, - W5, col[8*7]); nengel@2: MAC16(b2, + W3, col[8*7]); nengel@2: MAC16(b3, - W1, col[8*7]); nengel@2: } nengel@2: nengel@2: col[0 ] = ((a0 + b0) >> COL_SHIFT); nengel@2: col[8 ] = ((a1 + b1) >> COL_SHIFT); nengel@2: col[16] = ((a2 + b2) >> COL_SHIFT); nengel@2: col[24] = ((a3 + b3) >> COL_SHIFT); nengel@2: col[32] = ((a3 - b3) >> COL_SHIFT); nengel@2: col[40] = ((a2 - b2) >> COL_SHIFT); nengel@2: col[48] = ((a1 - b1) >> COL_SHIFT); nengel@2: col[56] = ((a0 - b0) >> COL_SHIFT); nengel@2: } nengel@2: nengel@2: void ff_simple_idct_put(uint8_t *dest, int line_size, DCTELEM *block) nengel@2: { nengel@2: int i; nengel@2: for(i=0; i<8; i++) nengel@2: idctRowCondDC(block + i*8); nengel@2: nengel@2: for(i=0; i<8; i++) nengel@2: idctSparseColPut(dest + i, line_size, block + i); nengel@2: } nengel@2: nengel@2: void ff_simple_idct_add(uint8_t *dest, int line_size, DCTELEM *block) nengel@2: { nengel@2: int i; nengel@2: for(i=0; i<8; i++) nengel@2: idctRowCondDC(block + i*8); nengel@2: nengel@2: for(i=0; i<8; i++) nengel@2: idctSparseColAdd(dest + i, line_size, block + i); nengel@2: } nengel@2: nengel@2: void ff_simple_idct(DCTELEM *block) nengel@2: { nengel@2: int i; nengel@2: for(i=0; i<8; i++) nengel@2: idctRowCondDC(block + i*8); nengel@2: nengel@2: for(i=0; i<8; i++) nengel@2: idctSparseCol(block + i); nengel@2: }