X-Git-Url: https://git.sesse.net/?a=blobdiff_plain;f=libavcodec%2Fsimple_idct.c;h=7bf0836aa4b8d3ef14ea60b5af4d069773215c36;hb=1472b7dd2824513894f779125c298f0ad991ad04;hp=de3d9265cfdfedd3d36fdea553d0d306e7215121;hpb=ff4ec49e648bead601b3b8c618f6d19f3548d368;p=ffmpeg diff --git a/libavcodec/simple_idct.c b/libavcodec/simple_idct.c index de3d9265cfd..7bf0836aa4b 100644 --- a/libavcodec/simple_idct.c +++ b/libavcodec/simple_idct.c @@ -3,26 +3,34 @@ * * Copyright (c) 2001 Michael Niedermayer * - * This library is free software; you can redistribute it and/or + * This file is part of FFmpeg. + * + * FFmpeg is free software; you can redistribute it and/or * modify it under the terms of the GNU Lesser General Public * License as published by the Free Software Foundation; either - * version 2 of the License, or (at your option) any later version. + * version 2.1 of the License, or (at your option) any later version. * - * This library is distributed in the hope that it will be useful, + * FFmpeg is distributed in the hope that it will be useful, * but WITHOUT ANY WARRANTY; without even the implied warranty of * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU * Lesser General Public License for more details. * * You should have received a copy of the GNU Lesser General Public - * License along with this library; if not, write to the Free Software - * Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA + * License along with FFmpeg; if not, write to the Free Software + * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA + */ + +/** + * @file simple_idct.c + * simpleidct in C. */ + /* based upon some outcommented c code from mpeg2dec (idct_mmx.c - written by Aaron Holtzman ) + written by Aaron Holtzman ) */ #include "avcodec.h" - +#include "dsputil.h" #include "simple_idct.h" #if 0 @@ -47,543 +55,547 @@ #define COL_SHIFT 20 // 6 #endif -/* 8x8 Matrix used to do a trivial (slow) 8 point IDCT */ -static int coeff[64]={ - W4, W4, W4, W4, W4, W4, W4, W4, - W1, W3, W5, W7,-W7,-W5,-W3,-W1, - W2, W6,-W6,-W2,-W2,-W6, W6, W2, - W3,-W7,-W1,-W5, W5, W1, W7,-W3, - W4,-W4,-W4, W4, W4,-W4,-W4, W4, - W5,-W1, W7, W3,-W3,-W7, W1,-W5, - W6,-W2, W2,-W6,-W6, W2,-W2, W6, - W7,-W5, W3,-W1, W1,-W3, W5,-W7 -}; - -static int inline idctRowCondZ (int16_t * row) +#if defined(ARCH_POWERPC_405) + +/* signed 16x16 -> 32 multiply add accumulate */ +#define MAC16(rt, ra, rb) \ + asm ("maclhw %0, %2, %3" : "=r" (rt) : "0" (rt), "r" (ra), "r" (rb)); + +/* signed 16x16 -> 32 multiply */ +#define MUL16(rt, ra, rb) \ + asm ("mullhw %0, %1, %2" : "=r" (rt) : "r" (ra), "r" (rb)); + +#else + +/* signed 16x16 -> 32 multiply add accumulate */ +#define MAC16(rt, ra, rb) rt += (ra) * (rb) + +/* signed 16x16 -> 32 multiply */ +#define MUL16(rt, ra, rb) rt = (ra) * (rb) + +#endif + +static inline void idctRowCondDC (DCTELEM * row) { - int a0, a1, a2, a3, b0, b1, b2, b3; - - if( !( ((uint32_t*)row)[0]|((uint32_t*)row)[1] |((uint32_t*)row)[2] |((uint32_t*)row)[3])) { -/* row[0] = row[1] = row[2] = row[3] = row[4] = - row[5] = row[6] = row[7] = 0;*/ - return 0; - } - - if(!( ((uint32_t*)row)[2] |((uint32_t*)row)[3] )){ - a0 = W4*row[0] + W2*row[2] + (1<<(ROW_SHIFT-1)); - a1 = W4*row[0] + W6*row[2] + (1<<(ROW_SHIFT-1)); - a2 = W4*row[0] - W6*row[2] + (1<<(ROW_SHIFT-1)); - a3 = W4*row[0] - W2*row[2] + (1<<(ROW_SHIFT-1)); - - b0 = W1*row[1] + W3*row[3]; - b1 = W3*row[1] - W7*row[3]; - b2 = W5*row[1] - W1*row[3]; - b3 = W7*row[1] - W5*row[3]; - }else{ - a0 = W4*row[0] + W2*row[2] + W4*row[4] + W6*row[6] + (1<<(ROW_SHIFT-1)); - a1 = W4*row[0] + W6*row[2] - W4*row[4] - W2*row[6] + (1<<(ROW_SHIFT-1)); - a2 = W4*row[0] - W6*row[2] - W4*row[4] + W2*row[6] + (1<<(ROW_SHIFT-1)); - a3 = W4*row[0] - W2*row[2] + W4*row[4] - W6*row[6] + (1<<(ROW_SHIFT-1)); - - b0 = W1*row[1] + W3*row[3] + W5*row[5] + W7*row[7]; - b1 = W3*row[1] - W7*row[3] - W1*row[5] - W5*row[7]; - b2 = W5*row[1] - W1*row[3] + W7*row[5] + W3*row[7]; - b3 = W7*row[1] - W5*row[3] + W3*row[5] - W1*row[7]; - } - - row[0] = (a0 + b0) >> ROW_SHIFT; - row[1] = (a1 + b1) >> ROW_SHIFT; - row[2] = (a2 + b2) >> ROW_SHIFT; - row[3] = (a3 + b3) >> ROW_SHIFT; - row[4] = (a3 - b3) >> ROW_SHIFT; - row[5] = (a2 - b2) >> ROW_SHIFT; - row[6] = (a1 - b1) >> ROW_SHIFT; - row[7] = (a0 - b0) >> ROW_SHIFT; - - return 1; + int a0, a1, a2, a3, b0, b1, b2, b3; +#ifdef HAVE_FAST_64BIT + uint64_t temp; +#else + uint32_t temp; +#endif + +#ifdef HAVE_FAST_64BIT +#ifdef WORDS_BIGENDIAN +#define ROW0_MASK 0xffff000000000000LL +#else +#define ROW0_MASK 0xffffLL +#endif + if(sizeof(DCTELEM)==2){ + if ( ((((uint64_t *)row)[0] & ~ROW0_MASK) | + ((uint64_t *)row)[1]) == 0) { + temp = (row[0] << 3) & 0xffff; + temp += temp << 16; + temp += temp << 32; + ((uint64_t *)row)[0] = temp; + ((uint64_t *)row)[1] = temp; + return; + } + }else{ + if (!(row[1]|row[2]|row[3]|row[4]|row[5]|row[6]|row[7])) { + row[0]=row[1]=row[2]=row[3]=row[4]=row[5]=row[6]=row[7]= row[0] << 3; + return; + } + } +#else + if(sizeof(DCTELEM)==2){ + if (!(((uint32_t*)row)[1] | + ((uint32_t*)row)[2] | + ((uint32_t*)row)[3] | + row[1])) { + temp = (row[0] << 3) & 0xffff; + temp += temp << 16; + ((uint32_t*)row)[0]=((uint32_t*)row)[1] = + ((uint32_t*)row)[2]=((uint32_t*)row)[3] = temp; + return; + } + }else{ + if (!(row[1]|row[2]|row[3]|row[4]|row[5]|row[6]|row[7])) { + row[0]=row[1]=row[2]=row[3]=row[4]=row[5]=row[6]=row[7]= row[0] << 3; + return; + } + } +#endif + + a0 = (W4 * row[0]) + (1 << (ROW_SHIFT - 1)); + a1 = a0; + a2 = a0; + a3 = a0; + + /* no need to optimize : gcc does it */ + a0 += W2 * row[2]; + a1 += W6 * row[2]; + a2 -= W6 * row[2]; + a3 -= W2 * row[2]; + + MUL16(b0, W1, row[1]); + MAC16(b0, W3, row[3]); + MUL16(b1, W3, row[1]); + MAC16(b1, -W7, row[3]); + MUL16(b2, W5, row[1]); + MAC16(b2, -W1, row[3]); + MUL16(b3, W7, row[1]); + MAC16(b3, -W5, row[3]); + +#ifdef HAVE_FAST_64BIT + temp = ((uint64_t*)row)[1]; +#else + temp = ((uint32_t*)row)[2] | ((uint32_t*)row)[3]; +#endif + if (temp != 0) { + a0 += W4*row[4] + W6*row[6]; + a1 += - W4*row[4] - W2*row[6]; + a2 += - W4*row[4] + W2*row[6]; + a3 += W4*row[4] - W6*row[6]; + + MAC16(b0, W5, row[5]); + MAC16(b0, W7, row[7]); + + MAC16(b1, -W1, row[5]); + MAC16(b1, -W5, row[7]); + + MAC16(b2, W7, row[5]); + MAC16(b2, W3, row[7]); + + MAC16(b3, W3, row[5]); + MAC16(b3, -W1, row[7]); + } + + row[0] = (a0 + b0) >> ROW_SHIFT; + row[7] = (a0 - b0) >> ROW_SHIFT; + row[1] = (a1 + b1) >> ROW_SHIFT; + row[6] = (a1 - b1) >> ROW_SHIFT; + row[2] = (a2 + b2) >> ROW_SHIFT; + row[5] = (a2 - b2) >> ROW_SHIFT; + row[3] = (a3 + b3) >> ROW_SHIFT; + row[4] = (a3 - b3) >> ROW_SHIFT; } -#ifdef ARCH_ALPHA -/* 0: all entries 0, 1: only first entry nonzero, 2: otherwise */ -static int inline idctRowCondDC(int16_t *row) +static inline void idctSparseColPut (uint8_t *dest, int line_size, + DCTELEM * col) { - int_fast32_t a0, a1, a2, a3, b0, b1, b2, b3; - uint64_t *lrow = (uint64_t *) row; - - if (lrow[1] == 0) { - if (lrow[0] == 0) - return 0; - if ((lrow[0] & ~0xffffULL) == 0) { - uint64_t v; - - a0 = W4 * row[0]; - a0 += 1 << (ROW_SHIFT - 1); - a0 >>= ROW_SHIFT; - v = (uint16_t) a0; - v += v << 16; - v += v << 32; - lrow[0] = v; - lrow[1] = v; - - return 1; - } - } - - a0 = W4 * row[0]; - a1 = W4 * row[0]; - a2 = W4 * row[0]; - a3 = W4 * row[0]; - - if (row[2]) { - a0 += W2 * row[2]; - a1 += W6 * row[2]; - a2 -= W6 * row[2]; - a3 -= W2 * row[2]; - } - - if (row[4]) { - a0 += W4 * row[4]; - a1 -= W4 * row[4]; - a2 -= W4 * row[4]; - a3 += W4 * row[4]; - } - - if (row[6]) { - a0 += W6 * row[6]; - a1 -= W2 * row[6]; - a2 += W2 * row[6]; - a3 -= W6 * row[6]; - } - - a0 += 1 << (ROW_SHIFT - 1); - a1 += 1 << (ROW_SHIFT - 1); - a2 += 1 << (ROW_SHIFT - 1); - a3 += 1 << (ROW_SHIFT - 1); - - if (row[1]) { - b0 = W1 * row[1]; - b1 = W3 * row[1]; - b2 = W5 * row[1]; - b3 = W7 * row[1]; - } else { - b0 = 0; - b1 = 0; - b2 = 0; - b3 = 0; - } - - if (row[3]) { - b0 += W3 * row[3]; - b1 -= W7 * row[3]; - b2 -= W1 * row[3]; - b3 -= W5 * row[3]; - } - - if (row[5]) { - b0 += W5 * row[5]; - b1 -= W1 * row[5]; - b2 += W7 * row[5]; - b3 += W3 * row[5]; - } - - if (row[7]) { - b0 += W7 * row[7]; - b1 -= W5 * row[7]; - b2 += W3 * row[7]; - b3 -= W1 * row[7]; - } - - row[0] = (a0 + b0) >> ROW_SHIFT; - row[1] = (a1 + b1) >> ROW_SHIFT; - row[2] = (a2 + b2) >> ROW_SHIFT; - row[3] = (a3 + b3) >> ROW_SHIFT; - row[4] = (a3 - b3) >> ROW_SHIFT; - row[5] = (a2 - b2) >> ROW_SHIFT; - row[6] = (a1 - b1) >> ROW_SHIFT; - row[7] = (a0 - b0) >> ROW_SHIFT; - - return 2; + int a0, a1, a2, a3, b0, b1, b2, b3; + uint8_t *cm = ff_cropTbl + MAX_NEG_CROP; + + /* XXX: I did that only to give same values as previous code */ + a0 = W4 * (col[8*0] + ((1<<(COL_SHIFT-1))/W4)); + a1 = a0; + a2 = a0; + a3 = a0; + + a0 += + W2*col[8*2]; + a1 += + W6*col[8*2]; + a2 += - W6*col[8*2]; + a3 += - W2*col[8*2]; + + MUL16(b0, W1, col[8*1]); + MUL16(b1, W3, col[8*1]); + MUL16(b2, W5, col[8*1]); + MUL16(b3, W7, col[8*1]); + + MAC16(b0, + W3, col[8*3]); + MAC16(b1, - W7, col[8*3]); + MAC16(b2, - W1, col[8*3]); + MAC16(b3, - W5, col[8*3]); + + if(col[8*4]){ + a0 += + W4*col[8*4]; + a1 += - W4*col[8*4]; + a2 += - W4*col[8*4]; + a3 += + W4*col[8*4]; + } + + if (col[8*5]) { + MAC16(b0, + W5, col[8*5]); + MAC16(b1, - W1, col[8*5]); + MAC16(b2, + W7, col[8*5]); + MAC16(b3, + W3, col[8*5]); + } + + if(col[8*6]){ + a0 += + W6*col[8*6]; + a1 += - W2*col[8*6]; + a2 += + W2*col[8*6]; + a3 += - W6*col[8*6]; + } + + if (col[8*7]) { + MAC16(b0, + W7, col[8*7]); + MAC16(b1, - W5, col[8*7]); + MAC16(b2, + W3, col[8*7]); + MAC16(b3, - W1, col[8*7]); + } + + dest[0] = cm[(a0 + b0) >> COL_SHIFT]; + dest += line_size; + dest[0] = cm[(a1 + b1) >> COL_SHIFT]; + dest += line_size; + dest[0] = cm[(a2 + b2) >> COL_SHIFT]; + dest += line_size; + dest[0] = cm[(a3 + b3) >> COL_SHIFT]; + dest += line_size; + dest[0] = cm[(a3 - b3) >> COL_SHIFT]; + dest += line_size; + dest[0] = cm[(a2 - b2) >> COL_SHIFT]; + dest += line_size; + dest[0] = cm[(a1 - b1) >> COL_SHIFT]; + dest += line_size; + dest[0] = cm[(a0 - b0) >> COL_SHIFT]; } -#else /* not ARCH_ALPHA */ -static int inline idctRowCondDC (int16_t * row) + +static inline void idctSparseColAdd (uint8_t *dest, int line_size, + DCTELEM * col) { - int a0, a1, a2, a3, b0, b1, b2, b3; - - if( !( ((uint32_t*)row)[1] |((uint32_t*)row)[2] |((uint32_t*)row)[3]| row[1])) { -// row[0] = row[1] = row[2] = row[3] = row[4] = row[5] = row[6] = row[7] = row[0]<<3; - uint16_t temp= row[0]<<3; - ((uint32_t*)row)[0]=((uint32_t*)row)[1]= - ((uint32_t*)row)[2]=((uint32_t*)row)[3]= temp + (temp<<16); - return 0; - } - - if(!( ((uint32_t*)row)[2] |((uint32_t*)row)[3] )){ - a0 = W4*row[0] + W2*row[2] + (1<<(ROW_SHIFT-1)); - a1 = W4*row[0] + W6*row[2] + (1<<(ROW_SHIFT-1)); - a2 = W4*row[0] - W6*row[2] + (1<<(ROW_SHIFT-1)); - a3 = W4*row[0] - W2*row[2] + (1<<(ROW_SHIFT-1)); - - b0 = W1*row[1] + W3*row[3]; - b1 = W3*row[1] - W7*row[3]; - b2 = W5*row[1] - W1*row[3]; - b3 = W7*row[1] - W5*row[3]; - }else{ - a0 = W4*row[0] + W2*row[2] + W4*row[4] + W6*row[6] + (1<<(ROW_SHIFT-1)); - a1 = W4*row[0] + W6*row[2] - W4*row[4] - W2*row[6] + (1<<(ROW_SHIFT-1)); - a2 = W4*row[0] - W6*row[2] - W4*row[4] + W2*row[6] + (1<<(ROW_SHIFT-1)); - a3 = W4*row[0] - W2*row[2] + W4*row[4] - W6*row[6] + (1<<(ROW_SHIFT-1)); - - b0 = W1*row[1] + W3*row[3] + W5*row[5] + W7*row[7]; - b1 = W3*row[1] - W7*row[3] - W1*row[5] - W5*row[7]; - b2 = W5*row[1] - W1*row[3] + W7*row[5] + W3*row[7]; - b3 = W7*row[1] - W5*row[3] + W3*row[5] - W1*row[7]; - } - - row[0] = (a0 + b0) >> ROW_SHIFT; - row[7] = (a0 - b0) >> ROW_SHIFT; - row[1] = (a1 + b1) >> ROW_SHIFT; - row[6] = (a1 - b1) >> ROW_SHIFT; - row[2] = (a2 + b2) >> ROW_SHIFT; - row[5] = (a2 - b2) >> ROW_SHIFT; - row[3] = (a3 + b3) >> ROW_SHIFT; - row[4] = (a3 - b3) >> ROW_SHIFT; - - return 1; + int a0, a1, a2, a3, b0, b1, b2, b3; + uint8_t *cm = ff_cropTbl + MAX_NEG_CROP; + + /* XXX: I did that only to give same values as previous code */ + a0 = W4 * (col[8*0] + ((1<<(COL_SHIFT-1))/W4)); + a1 = a0; + a2 = a0; + a3 = a0; + + a0 += + W2*col[8*2]; + a1 += + W6*col[8*2]; + a2 += - W6*col[8*2]; + a3 += - W2*col[8*2]; + + MUL16(b0, W1, col[8*1]); + MUL16(b1, W3, col[8*1]); + MUL16(b2, W5, col[8*1]); + MUL16(b3, W7, col[8*1]); + + MAC16(b0, + W3, col[8*3]); + MAC16(b1, - W7, col[8*3]); + MAC16(b2, - W1, col[8*3]); + MAC16(b3, - W5, col[8*3]); + + if(col[8*4]){ + a0 += + W4*col[8*4]; + a1 += - W4*col[8*4]; + a2 += - W4*col[8*4]; + a3 += + W4*col[8*4]; + } + + if (col[8*5]) { + MAC16(b0, + W5, col[8*5]); + MAC16(b1, - W1, col[8*5]); + MAC16(b2, + W7, col[8*5]); + MAC16(b3, + W3, col[8*5]); + } + + if(col[8*6]){ + a0 += + W6*col[8*6]; + a1 += - W2*col[8*6]; + a2 += + W2*col[8*6]; + a3 += - W6*col[8*6]; + } + + if (col[8*7]) { + MAC16(b0, + W7, col[8*7]); + MAC16(b1, - W5, col[8*7]); + MAC16(b2, + W3, col[8*7]); + MAC16(b3, - W1, col[8*7]); + } + + dest[0] = cm[dest[0] + ((a0 + b0) >> COL_SHIFT)]; + dest += line_size; + dest[0] = cm[dest[0] + ((a1 + b1) >> COL_SHIFT)]; + dest += line_size; + dest[0] = cm[dest[0] + ((a2 + b2) >> COL_SHIFT)]; + dest += line_size; + dest[0] = cm[dest[0] + ((a3 + b3) >> COL_SHIFT)]; + dest += line_size; + dest[0] = cm[dest[0] + ((a3 - b3) >> COL_SHIFT)]; + dest += line_size; + dest[0] = cm[dest[0] + ((a2 - b2) >> COL_SHIFT)]; + dest += line_size; + dest[0] = cm[dest[0] + ((a1 - b1) >> COL_SHIFT)]; + dest += line_size; + dest[0] = cm[dest[0] + ((a0 - b0) >> COL_SHIFT)]; } -#endif /* not ARCH_ALPHA */ -static void inline idctCol (int16_t * col) +static inline void idctSparseCol (DCTELEM * col) { + int a0, a1, a2, a3, b0, b1, b2, b3; + + /* XXX: I did that only to give same values as previous code */ + a0 = W4 * (col[8*0] + ((1<<(COL_SHIFT-1))/W4)); + a1 = a0; + a2 = a0; + a3 = a0; + + a0 += + W2*col[8*2]; + a1 += + W6*col[8*2]; + a2 += - W6*col[8*2]; + a3 += - W2*col[8*2]; + + MUL16(b0, W1, col[8*1]); + MUL16(b1, W3, col[8*1]); + MUL16(b2, W5, col[8*1]); + MUL16(b3, W7, col[8*1]); + + MAC16(b0, + W3, col[8*3]); + MAC16(b1, - W7, col[8*3]); + MAC16(b2, - W1, col[8*3]); + MAC16(b3, - W5, col[8*3]); + + if(col[8*4]){ + a0 += + W4*col[8*4]; + a1 += - W4*col[8*4]; + a2 += - W4*col[8*4]; + a3 += + W4*col[8*4]; + } + + if (col[8*5]) { + MAC16(b0, + W5, col[8*5]); + MAC16(b1, - W1, col[8*5]); + MAC16(b2, + W7, col[8*5]); + MAC16(b3, + W3, col[8*5]); + } + + if(col[8*6]){ + a0 += + W6*col[8*6]; + a1 += - W2*col[8*6]; + a2 += + W2*col[8*6]; + a3 += - W6*col[8*6]; + } + + if (col[8*7]) { + MAC16(b0, + W7, col[8*7]); + MAC16(b1, - W5, col[8*7]); + MAC16(b2, + W3, col[8*7]); + MAC16(b3, - W1, col[8*7]); + } + + col[0 ] = ((a0 + b0) >> COL_SHIFT); + col[8 ] = ((a1 + b1) >> COL_SHIFT); + col[16] = ((a2 + b2) >> COL_SHIFT); + col[24] = ((a3 + b3) >> COL_SHIFT); + col[32] = ((a3 - b3) >> COL_SHIFT); + col[40] = ((a2 - b2) >> COL_SHIFT); + col[48] = ((a1 - b1) >> COL_SHIFT); + col[56] = ((a0 - b0) >> COL_SHIFT); +} -/* - if( !(col[8*1] | col[8*2] |col[8*3] |col[8*4] |col[8*5] |col[8*6] | col[8*7])) { - col[8*0] = col[8*1] = col[8*2] = col[8*3] = col[8*4] = - col[8*5] = col[8*6] = col[8*7] = col[8*0]<<3; - return; - }*/ - - int a0, a1, a2, a3, b0, b1, b2, b3; - col[0] += (1<<(COL_SHIFT-1))/W4; - a0 = W4*col[8*0] + W2*col[8*2] + W4*col[8*4] + W6*col[8*6]; - a1 = W4*col[8*0] + W6*col[8*2] - W4*col[8*4] - W2*col[8*6]; - a2 = W4*col[8*0] - W6*col[8*2] - W4*col[8*4] + W2*col[8*6]; - a3 = W4*col[8*0] - W2*col[8*2] + W4*col[8*4] - W6*col[8*6]; - - b0 = W1*col[8*1] + W3*col[8*3] + W5*col[8*5] + W7*col[8*7]; - b1 = W3*col[8*1] - W7*col[8*3] - W1*col[8*5] - W5*col[8*7]; - b2 = W5*col[8*1] - W1*col[8*3] + W7*col[8*5] + W3*col[8*7]; - b3 = W7*col[8*1] - W5*col[8*3] + W3*col[8*5] - W1*col[8*7]; - - col[8*0] = (a0 + b0) >> COL_SHIFT; - col[8*7] = (a0 - b0) >> COL_SHIFT; - col[8*1] = (a1 + b1) >> COL_SHIFT; - col[8*6] = (a1 - b1) >> COL_SHIFT; - col[8*2] = (a2 + b2) >> COL_SHIFT; - col[8*5] = (a2 - b2) >> COL_SHIFT; - col[8*3] = (a3 + b3) >> COL_SHIFT; - col[8*4] = (a3 - b3) >> COL_SHIFT; +void ff_simple_idct_put(uint8_t *dest, int line_size, DCTELEM *block) +{ + int i; + for(i=0; i<8; i++) + idctRowCondDC(block + i*8); + + for(i=0; i<8; i++) + idctSparseColPut(dest + i, line_size, block + i); } -static void inline idctSparseCol (int16_t * col) +void ff_simple_idct_add(uint8_t *dest, int line_size, DCTELEM *block) { - int a0, a1, a2, a3, b0, b1, b2, b3; - col[0] += (1<<(COL_SHIFT-1))/W4; - a0 = W4*col[8*0]; - a1 = W4*col[8*0]; - a2 = W4*col[8*0]; - a3 = W4*col[8*0]; - - if(col[8*2]){ - a0 += + W2*col[8*2]; - a1 += + W6*col[8*2]; - a2 += - W6*col[8*2]; - a3 += - W2*col[8*2]; - } - - if(col[8*4]){ - a0 += + W4*col[8*4]; - a1 += - W4*col[8*4]; - a2 += - W4*col[8*4]; - a3 += + W4*col[8*4]; - } - - if(col[8*6]){ - a0 += + W6*col[8*6]; - a1 += - W2*col[8*6]; - a2 += + W2*col[8*6]; - a3 += - W6*col[8*6]; - } - - if(col[8*1]){ - b0 = W1*col[8*1]; - b1 = W3*col[8*1]; - b2 = W5*col[8*1]; - b3 = W7*col[8*1]; - }else{ - b0 = - b1 = - b2 = - b3 = 0; - } - - if(col[8*3]){ - b0 += + W3*col[8*3]; - b1 += - W7*col[8*3]; - b2 += - W1*col[8*3]; - b3 += - W5*col[8*3]; - } - - if(col[8*5]){ - b0 += + W5*col[8*5]; - b1 += - W1*col[8*5]; - b2 += + W7*col[8*5]; - b3 += + W3*col[8*5]; - } - - if(col[8*7]){ - b0 += + W7*col[8*7]; - b1 += - W5*col[8*7]; - b2 += + W3*col[8*7]; - b3 += - W1*col[8*7]; - } - -#ifndef ARCH_ALPHA - if(!(b0|b1|b2|b3)){ - col[8*0] = (a0) >> COL_SHIFT; - col[8*7] = (a0) >> COL_SHIFT; - col[8*1] = (a1) >> COL_SHIFT; - col[8*6] = (a1) >> COL_SHIFT; - col[8*2] = (a2) >> COL_SHIFT; - col[8*5] = (a2) >> COL_SHIFT; - col[8*3] = (a3) >> COL_SHIFT; - col[8*4] = (a3) >> COL_SHIFT; - }else{ -#endif - col[8*0] = (a0 + b0) >> COL_SHIFT; - col[8*7] = (a0 - b0) >> COL_SHIFT; - col[8*1] = (a1 + b1) >> COL_SHIFT; - col[8*6] = (a1 - b1) >> COL_SHIFT; - col[8*2] = (a2 + b2) >> COL_SHIFT; - col[8*5] = (a2 - b2) >> COL_SHIFT; - col[8*3] = (a3 + b3) >> COL_SHIFT; - col[8*4] = (a3 - b3) >> COL_SHIFT; -#ifndef ARCH_ALPHA - } -#endif + int i; + for(i=0; i<8; i++) + idctRowCondDC(block + i*8); + + for(i=0; i<8; i++) + idctSparseColAdd(dest + i, line_size, block + i); } -static void inline idctSparse2Col (int16_t * col) +void ff_simple_idct(DCTELEM *block) { - int a0, a1, a2, a3, b0, b1, b2, b3; - col[0] += (1<<(COL_SHIFT-1))/W4; - a0 = W4*col[8*0]; - a1 = W4*col[8*0]; - a2 = W4*col[8*0]; - a3 = W4*col[8*0]; - - if(col[8*2]){ - a0 += + W2*col[8*2]; - a1 += + W6*col[8*2]; - a2 += - W6*col[8*2]; - a3 += - W2*col[8*2]; - } - - if(col[8*4]){ - a0 += + W4*col[8*4]; - a1 += - W4*col[8*4]; - a2 += - W4*col[8*4]; - a3 += + W4*col[8*4]; - } - - if(col[8*6]){ - a0 += + W6*col[8*6]; - a1 += - W2*col[8*6]; - a2 += + W2*col[8*6]; - a3 += - W6*col[8*6]; - } - - if(col[8*1] || 1){ - b0 = W1*col[8*1]; - b1 = W3*col[8*1]; - b2 = W5*col[8*1]; - b3 = W7*col[8*1]; - }else{ - b0 = - b1 = - b2 = - b3 = 0; - } - - if(col[8*3]){ - b0 += + W3*col[8*3]; - b1 += - W7*col[8*3]; - b2 += - W1*col[8*3]; - b3 += - W5*col[8*3]; - } - - if(col[8*5]){ - b0 += + W5*col[8*5]; - b1 += - W1*col[8*5]; - b2 += + W7*col[8*5]; - b3 += + W3*col[8*5]; - } - - if(col[8*7]){ - b0 += + W7*col[8*7]; - b1 += - W5*col[8*7]; - b2 += + W3*col[8*7]; - b3 += - W1*col[8*7]; - } - - col[8*0] = (a0 + b0) >> COL_SHIFT; - col[8*7] = (a0 - b0) >> COL_SHIFT; - col[8*1] = (a1 + b1) >> COL_SHIFT; - col[8*6] = (a1 - b1) >> COL_SHIFT; - col[8*2] = (a2 + b2) >> COL_SHIFT; - col[8*5] = (a2 - b2) >> COL_SHIFT; - col[8*3] = (a3 + b3) >> COL_SHIFT; - col[8*4] = (a3 - b3) >> COL_SHIFT; + int i; + for(i=0; i<8; i++) + idctRowCondDC(block + i*8); + + for(i=0; i<8; i++) + idctSparseCol(block + i); } -#ifdef ARCH_ALPHA -/* If all rows but the first one are zero after row transformation, - all rows will be identical after column transformation. */ -static inline void idctCol2(int16_t *col) +/* 2x4x8 idct */ + +#define CN_SHIFT 12 +#define C_FIX(x) ((int)((x) * (1 << CN_SHIFT) + 0.5)) +#define C1 C_FIX(0.6532814824) +#define C2 C_FIX(0.2705980501) + +/* row idct is multiple by 16 * sqrt(2.0), col idct4 is normalized, + and the butterfly must be multiplied by 0.5 * sqrt(2.0) */ +#define C_SHIFT (4+1+12) + +static inline void idct4col_put(uint8_t *dest, int line_size, const DCTELEM *col) { - int i; - uint64_t l, r; - uint64_t *lcol = (uint64_t *) col; - - for (i = 0; i < 8; ++i) { - int a0 = col[0] + (1 << (COL_SHIFT - 1)) / W4; - - a0 *= W4; - col[0] = a0 >> COL_SHIFT; - ++col; - } - - l = lcol[0]; - r = lcol[1]; - lcol[ 2] = l; lcol[ 3] = r; - lcol[ 4] = l; lcol[ 5] = r; - lcol[ 6] = l; lcol[ 7] = r; - lcol[ 8] = l; lcol[ 9] = r; - lcol[10] = l; lcol[11] = r; - lcol[12] = l; lcol[13] = r; - lcol[14] = l; lcol[15] = r; + int c0, c1, c2, c3, a0, a1, a2, a3; + const uint8_t *cm = ff_cropTbl + MAX_NEG_CROP; + + a0 = col[8*0]; + a1 = col[8*2]; + a2 = col[8*4]; + a3 = col[8*6]; + c0 = ((a0 + a2) << (CN_SHIFT - 1)) + (1 << (C_SHIFT - 1)); + c2 = ((a0 - a2) << (CN_SHIFT - 1)) + (1 << (C_SHIFT - 1)); + c1 = a1 * C1 + a3 * C2; + c3 = a1 * C2 - a3 * C1; + dest[0] = cm[(c0 + c1) >> C_SHIFT]; + dest += line_size; + dest[0] = cm[(c2 + c3) >> C_SHIFT]; + dest += line_size; + dest[0] = cm[(c2 - c3) >> C_SHIFT]; + dest += line_size; + dest[0] = cm[(c0 - c1) >> C_SHIFT]; +} + +#define BF(k) \ +{\ + int a0, a1;\ + a0 = ptr[k];\ + a1 = ptr[8 + k];\ + ptr[k] = a0 + a1;\ + ptr[8 + k] = a0 - a1;\ } -#endif -void simple_idct (short *block) +/* only used by DV codec. The input must be interlaced. 128 is added + to the pixels before clamping to avoid systematic error + (1024*sqrt(2)) offset would be needed otherwise. */ +/* XXX: I think a 1.0/sqrt(2) normalization should be needed to + compensate the extra butterfly stage - I don't have the full DV + specification */ +void ff_simple_idct248_put(uint8_t *dest, int line_size, DCTELEM *block) { + int i; + DCTELEM *ptr; + + /* butterfly */ + ptr = block; + for(i=0;i<4;i++) { + BF(0); + BF(1); + BF(2); + BF(3); + BF(4); + BF(5); + BF(6); + BF(7); + ptr += 2 * 8; + } + + /* IDCT8 on each line */ + for(i=0; i<8; i++) { + idctRowCondDC(block + i*8); + } + + /* IDCT4 and store */ + for(i=0;i<8;i++) { + idct4col_put(dest + i, 2 * line_size, block + i); + idct4col_put(dest + line_size + i, 2 * line_size, block + 8 + i); + } +} - int i; - -#if 0 - int nonZero[8]; - int buffer[64]; - int nNonZero=0; - - idctRowCondDC(block); - - for(i=1; i<8; i++) - { - nonZero[nNonZero]=i; - nNonZero+= idctRowCondZ(block + i*8); - } - - if(nNonZero==0) - { - for(i=0; i<8; i++) - { - block[i ]= - block[i+8 ]= - block[i+16]= - block[i+24]= - block[i+32]= - block[i+40]= - block[i+48]= - block[i+56]= (W4*block[i] + (1<<(COL_SHIFT-1))) >> COL_SHIFT; - } - } - else if(nNonZero==1) - { - int index= nonZero[0]*8; - for(i=0; i<8; i++) - { - int bias= W4*block[i] + (1<<(COL_SHIFT-1)); - int c= block[i + index]; - block[i ]= (c*coeff[index ] + bias) >> COL_SHIFT; - block[i+8 ]= (c*coeff[index+1] + bias) >> COL_SHIFT; - block[i+16]= (c*coeff[index+2] + bias) >> COL_SHIFT; - block[i+24]= (c*coeff[index+3] + bias) >> COL_SHIFT; - block[i+32]= (c*coeff[index+4] + bias) >> COL_SHIFT; - block[i+40]= (c*coeff[index+5] + bias) >> COL_SHIFT; - block[i+48]= (c*coeff[index+6] + bias) >> COL_SHIFT; - block[i+56]= (c*coeff[index+7] + bias) >> COL_SHIFT; - } - } -/* else if(nNonZero==2) - { - int index1= nonZero[0]*8; - int index2= nonZero[1]*8; - for(i=0; i<8; i++) - { - int bias= W4*block[i] + (1<<(COL_SHIFT-1)); - int c1= block[i + index1]; - int c2= block[i + index2]; - block[i ]= (c1*coeff[index1 ] + c2*coeff[index2 ] + bias) >> COL_SHIFT; - block[i+8 ]= (c1*coeff[index1+1] + c2*coeff[index2+1] + bias) >> COL_SHIFT; - block[i+16]= (c1*coeff[index1+2] + c2*coeff[index2+2] + bias) >> COL_SHIFT; - block[i+24]= (c1*coeff[index1+3] + c2*coeff[index2+3] + bias) >> COL_SHIFT; - block[i+32]= (c1*coeff[index1+4] + c2*coeff[index2+4] + bias) >> COL_SHIFT; - block[i+40]= (c1*coeff[index1+5] + c2*coeff[index2+5] + bias) >> COL_SHIFT; - block[i+48]= (c1*coeff[index1+6] + c2*coeff[index2+6] + bias) >> COL_SHIFT; - block[i+56]= (c1*coeff[index1+7] + c2*coeff[index2+7] + bias) >> COL_SHIFT; - } - }*/ - else - { - for(i=0; i<8; i++) - idctSparse2Col(block + i); - } -#elif defined(ARCH_ALPHA) - int rowsZero = 1; /* all rows except row 0 zero */ - int rowsConstant = 1; /* all rows consist of a constant value */ - - for (i = 0; i < 8; i++) { - int sparseness = idctRowCondDC(block + 8 * i); - - if (i > 0 && sparseness > 0) - rowsZero = 0; - if (sparseness == 2) - rowsConstant = 0; - } - - if (rowsZero) { - idctCol2(block); - } else if (rowsConstant) { - uint64_t *lblock = (uint64_t *) block; - - idctSparseCol(block); - for (i = 0; i < 8; i++) { - uint64_t v = (uint16_t) block[i * 8]; - - v += v << 16; - v += v << 32; - lblock[0] = v; - lblock[1] = v; - lblock += 2; - } - } else { - for (i = 0; i < 8; i++) - idctSparseCol(block + i); - } -#else - for(i=0; i<8; i++) - idctRowCondDC(block + i*8); - - for(i=0; i<8; i++) - idctSparseCol(block + i); -#endif +/* 8x4 & 4x8 WMV2 IDCT */ +#undef CN_SHIFT +#undef C_SHIFT +#undef C_FIX +#undef C1 +#undef C2 +#define CN_SHIFT 12 +#define C_FIX(x) ((int)((x) * 1.414213562 * (1 << CN_SHIFT) + 0.5)) +#define C1 C_FIX(0.6532814824) +#define C2 C_FIX(0.2705980501) +#define C3 C_FIX(0.5) +#define C_SHIFT (4+1+12) +static inline void idct4col_add(uint8_t *dest, int line_size, const DCTELEM *col) +{ + int c0, c1, c2, c3, a0, a1, a2, a3; + const uint8_t *cm = ff_cropTbl + MAX_NEG_CROP; + + a0 = col[8*0]; + a1 = col[8*1]; + a2 = col[8*2]; + a3 = col[8*3]; + c0 = (a0 + a2)*C3 + (1 << (C_SHIFT - 1)); + c2 = (a0 - a2)*C3 + (1 << (C_SHIFT - 1)); + c1 = a1 * C1 + a3 * C2; + c3 = a1 * C2 - a3 * C1; + dest[0] = cm[dest[0] + ((c0 + c1) >> C_SHIFT)]; + dest += line_size; + dest[0] = cm[dest[0] + ((c2 + c3) >> C_SHIFT)]; + dest += line_size; + dest[0] = cm[dest[0] + ((c2 - c3) >> C_SHIFT)]; + dest += line_size; + dest[0] = cm[dest[0] + ((c0 - c1) >> C_SHIFT)]; +} + +#define RN_SHIFT 15 +#define R_FIX(x) ((int)((x) * 1.414213562 * (1 << RN_SHIFT) + 0.5)) +#define R1 R_FIX(0.6532814824) +#define R2 R_FIX(0.2705980501) +#define R3 R_FIX(0.5) +#define R_SHIFT 11 +static inline void idct4row(DCTELEM *row) +{ + int c0, c1, c2, c3, a0, a1, a2, a3; + //const uint8_t *cm = ff_cropTbl + MAX_NEG_CROP; + + a0 = row[0]; + a1 = row[1]; + a2 = row[2]; + a3 = row[3]; + c0 = (a0 + a2)*R3 + (1 << (R_SHIFT - 1)); + c2 = (a0 - a2)*R3 + (1 << (R_SHIFT - 1)); + c1 = a1 * R1 + a3 * R2; + c3 = a1 * R2 - a3 * R1; + row[0]= (c0 + c1) >> R_SHIFT; + row[1]= (c2 + c3) >> R_SHIFT; + row[2]= (c2 - c3) >> R_SHIFT; + row[3]= (c0 - c1) >> R_SHIFT; +} + +void ff_simple_idct84_add(uint8_t *dest, int line_size, DCTELEM *block) +{ + int i; + + /* IDCT8 on each line */ + for(i=0; i<4; i++) { + idctRowCondDC(block + i*8); + } + + /* IDCT4 and store */ + for(i=0;i<8;i++) { + idct4col_add(dest + i, line_size, block + i); + } +} + +void ff_simple_idct48_add(uint8_t *dest, int line_size, DCTELEM *block) +{ + int i; + + /* IDCT4 on each line */ + for(i=0; i<8; i++) { + idct4row(block + i*8); + } + + /* IDCT8 and store */ + for(i=0; i<4; i++){ + idctSparseColAdd(dest + i, line_size, block + i); + } +} + +void ff_simple_idct44_add(uint8_t *dest, int line_size, DCTELEM *block) +{ + int i; + + /* IDCT4 on each line */ + for(i=0; i<4; i++) { + idct4row(block + i*8); + } + + /* IDCT4 and store */ + for(i=0; i<4; i++){ + idct4col_add(dest + i, line_size, block + i); + } }