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: /* yading@10: based upon some outcommented c code from mpeg2dec (idct_mmx.c yading@10: written by Aaron Holtzman ) yading@10: */ yading@10: yading@10: #include "bit_depth_template.c" yading@10: yading@10: #undef W1 yading@10: #undef W2 yading@10: #undef W3 yading@10: #undef W4 yading@10: #undef W5 yading@10: #undef W6 yading@10: #undef W7 yading@10: #undef ROW_SHIFT yading@10: #undef COL_SHIFT yading@10: #undef DC_SHIFT yading@10: #undef MUL yading@10: #undef MAC yading@10: yading@10: #if BIT_DEPTH == 8 yading@10: yading@10: #define W1 22725 //cos(i*M_PI/16)*sqrt(2)*(1<<14) + 0.5 yading@10: #define W2 21407 //cos(i*M_PI/16)*sqrt(2)*(1<<14) + 0.5 yading@10: #define W3 19266 //cos(i*M_PI/16)*sqrt(2)*(1<<14) + 0.5 yading@10: #define W4 16383 //cos(i*M_PI/16)*sqrt(2)*(1<<14) + 0.5 yading@10: #define W5 12873 //cos(i*M_PI/16)*sqrt(2)*(1<<14) + 0.5 yading@10: #define W6 8867 //cos(i*M_PI/16)*sqrt(2)*(1<<14) + 0.5 yading@10: #define W7 4520 //cos(i*M_PI/16)*sqrt(2)*(1<<14) + 0.5 yading@10: yading@10: #define ROW_SHIFT 11 yading@10: #define COL_SHIFT 20 yading@10: #define DC_SHIFT 3 yading@10: yading@10: #define MUL(a, b) MUL16(a, b) yading@10: #define MAC(a, b, c) MAC16(a, b, c) yading@10: yading@10: #elif BIT_DEPTH == 10 yading@10: yading@10: #define W1 90901 yading@10: #define W2 85627 yading@10: #define W3 77062 yading@10: #define W4 65535 yading@10: #define W5 51491 yading@10: #define W6 35468 yading@10: #define W7 18081 yading@10: yading@10: #define ROW_SHIFT 15 yading@10: #define COL_SHIFT 20 yading@10: #define DC_SHIFT 1 yading@10: yading@10: #define MUL(a, b) ((a) * (b)) yading@10: #define MAC(a, b, c) ((a) += (b) * (c)) yading@10: yading@10: #else yading@10: yading@10: #error "Unsupported bitdepth" yading@10: yading@10: #endif yading@10: yading@10: static inline void FUNC(idctRowCondDC)(int16_t *row, int extra_shift) yading@10: { yading@10: int a0, a1, a2, a3, b0, b1, b2, b3; yading@10: yading@10: #if HAVE_FAST_64BIT yading@10: #define ROW0_MASK (0xffffLL << 48 * HAVE_BIGENDIAN) yading@10: if (((((uint64_t *)row)[0] & ~ROW0_MASK) | ((uint64_t *)row)[1]) == 0) { yading@10: uint64_t temp; yading@10: if (DC_SHIFT - extra_shift > 0) { yading@10: temp = (row[0] << (DC_SHIFT - extra_shift)) & 0xffff; yading@10: } else { yading@10: temp = (row[0] >> (extra_shift - DC_SHIFT)) & 0xffff; yading@10: } yading@10: temp += temp << 16; yading@10: temp += temp << 32; yading@10: ((uint64_t *)row)[0] = temp; yading@10: ((uint64_t *)row)[1] = temp; yading@10: return; yading@10: } yading@10: #else yading@10: if (!(((uint32_t*)row)[1] | yading@10: ((uint32_t*)row)[2] | yading@10: ((uint32_t*)row)[3] | yading@10: row[1])) { yading@10: uint32_t temp; yading@10: if (DC_SHIFT - extra_shift > 0) { yading@10: temp = (row[0] << (DC_SHIFT - extra_shift)) & 0xffff; yading@10: } else { yading@10: temp = (row[0] >> (extra_shift - DC_SHIFT)) & 0xffff; yading@10: } yading@10: temp += temp << 16; yading@10: ((uint32_t*)row)[0]=((uint32_t*)row)[1] = yading@10: ((uint32_t*)row)[2]=((uint32_t*)row)[3] = temp; yading@10: return; yading@10: } yading@10: #endif yading@10: yading@10: a0 = (W4 * row[0]) + (1 << (ROW_SHIFT - 1)); yading@10: a1 = a0; yading@10: a2 = a0; yading@10: a3 = a0; yading@10: yading@10: a0 += W2 * row[2]; yading@10: a1 += W6 * row[2]; yading@10: a2 -= W6 * row[2]; yading@10: a3 -= W2 * row[2]; yading@10: yading@10: b0 = MUL(W1, row[1]); yading@10: MAC(b0, W3, row[3]); yading@10: b1 = MUL(W3, row[1]); yading@10: MAC(b1, -W7, row[3]); yading@10: b2 = MUL(W5, row[1]); yading@10: MAC(b2, -W1, row[3]); yading@10: b3 = MUL(W7, row[1]); yading@10: MAC(b3, -W5, row[3]); yading@10: yading@10: if (AV_RN64A(row + 4)) { yading@10: a0 += W4*row[4] + W6*row[6]; yading@10: a1 += - W4*row[4] - W2*row[6]; yading@10: a2 += - W4*row[4] + W2*row[6]; yading@10: a3 += W4*row[4] - W6*row[6]; yading@10: yading@10: MAC(b0, W5, row[5]); yading@10: MAC(b0, W7, row[7]); yading@10: yading@10: MAC(b1, -W1, row[5]); yading@10: MAC(b1, -W5, row[7]); yading@10: yading@10: MAC(b2, W7, row[5]); yading@10: MAC(b2, W3, row[7]); yading@10: yading@10: MAC(b3, W3, row[5]); yading@10: MAC(b3, -W1, row[7]); yading@10: } yading@10: yading@10: row[0] = (a0 + b0) >> (ROW_SHIFT + extra_shift); yading@10: row[7] = (a0 - b0) >> (ROW_SHIFT + extra_shift); yading@10: row[1] = (a1 + b1) >> (ROW_SHIFT + extra_shift); yading@10: row[6] = (a1 - b1) >> (ROW_SHIFT + extra_shift); yading@10: row[2] = (a2 + b2) >> (ROW_SHIFT + extra_shift); yading@10: row[5] = (a2 - b2) >> (ROW_SHIFT + extra_shift); yading@10: row[3] = (a3 + b3) >> (ROW_SHIFT + extra_shift); yading@10: row[4] = (a3 - b3) >> (ROW_SHIFT + extra_shift); yading@10: } yading@10: yading@10: #define IDCT_COLS do { \ yading@10: a0 = W4 * (col[8*0] + ((1<<(COL_SHIFT-1))/W4)); \ yading@10: a1 = a0; \ yading@10: a2 = a0; \ yading@10: a3 = a0; \ yading@10: \ yading@10: a0 += W2*col[8*2]; \ yading@10: a1 += W6*col[8*2]; \ yading@10: a2 += -W6*col[8*2]; \ yading@10: a3 += -W2*col[8*2]; \ yading@10: \ yading@10: b0 = MUL(W1, col[8*1]); \ yading@10: b1 = MUL(W3, col[8*1]); \ yading@10: b2 = MUL(W5, col[8*1]); \ yading@10: b3 = MUL(W7, col[8*1]); \ yading@10: \ yading@10: MAC(b0, W3, col[8*3]); \ yading@10: MAC(b1, -W7, col[8*3]); \ yading@10: MAC(b2, -W1, col[8*3]); \ yading@10: MAC(b3, -W5, col[8*3]); \ yading@10: \ yading@10: if (col[8*4]) { \ yading@10: a0 += W4*col[8*4]; \ yading@10: a1 += -W4*col[8*4]; \ yading@10: a2 += -W4*col[8*4]; \ yading@10: a3 += W4*col[8*4]; \ yading@10: } \ yading@10: \ yading@10: if (col[8*5]) { \ yading@10: MAC(b0, W5, col[8*5]); \ yading@10: MAC(b1, -W1, col[8*5]); \ yading@10: MAC(b2, W7, col[8*5]); \ yading@10: MAC(b3, W3, col[8*5]); \ yading@10: } \ yading@10: \ yading@10: if (col[8*6]) { \ yading@10: a0 += W6*col[8*6]; \ yading@10: a1 += -W2*col[8*6]; \ yading@10: a2 += W2*col[8*6]; \ yading@10: a3 += -W6*col[8*6]; \ yading@10: } \ yading@10: \ yading@10: if (col[8*7]) { \ yading@10: MAC(b0, W7, col[8*7]); \ yading@10: MAC(b1, -W5, col[8*7]); \ yading@10: MAC(b2, W3, col[8*7]); \ yading@10: MAC(b3, -W1, col[8*7]); \ yading@10: } \ yading@10: } while (0) yading@10: yading@10: static inline void FUNC(idctSparseColPut)(pixel *dest, int line_size, yading@10: int16_t *col) yading@10: { yading@10: int a0, a1, a2, a3, b0, b1, b2, b3; yading@10: yading@10: IDCT_COLS; yading@10: yading@10: dest[0] = av_clip_pixel((a0 + b0) >> COL_SHIFT); yading@10: dest += line_size; yading@10: dest[0] = av_clip_pixel((a1 + b1) >> COL_SHIFT); yading@10: dest += line_size; yading@10: dest[0] = av_clip_pixel((a2 + b2) >> COL_SHIFT); yading@10: dest += line_size; yading@10: dest[0] = av_clip_pixel((a3 + b3) >> COL_SHIFT); yading@10: dest += line_size; yading@10: dest[0] = av_clip_pixel((a3 - b3) >> COL_SHIFT); yading@10: dest += line_size; yading@10: dest[0] = av_clip_pixel((a2 - b2) >> COL_SHIFT); yading@10: dest += line_size; yading@10: dest[0] = av_clip_pixel((a1 - b1) >> COL_SHIFT); yading@10: dest += line_size; yading@10: dest[0] = av_clip_pixel((a0 - b0) >> COL_SHIFT); yading@10: } yading@10: yading@10: static inline void FUNC(idctSparseColAdd)(pixel *dest, int line_size, yading@10: int16_t *col) yading@10: { yading@10: int a0, a1, a2, a3, b0, b1, b2, b3; yading@10: yading@10: IDCT_COLS; yading@10: yading@10: dest[0] = av_clip_pixel(dest[0] + ((a0 + b0) >> COL_SHIFT)); yading@10: dest += line_size; yading@10: dest[0] = av_clip_pixel(dest[0] + ((a1 + b1) >> COL_SHIFT)); yading@10: dest += line_size; yading@10: dest[0] = av_clip_pixel(dest[0] + ((a2 + b2) >> COL_SHIFT)); yading@10: dest += line_size; yading@10: dest[0] = av_clip_pixel(dest[0] + ((a3 + b3) >> COL_SHIFT)); yading@10: dest += line_size; yading@10: dest[0] = av_clip_pixel(dest[0] + ((a3 - b3) >> COL_SHIFT)); yading@10: dest += line_size; yading@10: dest[0] = av_clip_pixel(dest[0] + ((a2 - b2) >> COL_SHIFT)); yading@10: dest += line_size; yading@10: dest[0] = av_clip_pixel(dest[0] + ((a1 - b1) >> COL_SHIFT)); yading@10: dest += line_size; yading@10: dest[0] = av_clip_pixel(dest[0] + ((a0 - b0) >> COL_SHIFT)); yading@10: } yading@10: yading@10: static inline void FUNC(idctSparseCol)(int16_t *col) yading@10: { yading@10: int a0, a1, a2, a3, b0, b1, b2, b3; yading@10: yading@10: IDCT_COLS; yading@10: yading@10: col[0 ] = ((a0 + b0) >> COL_SHIFT); yading@10: col[8 ] = ((a1 + b1) >> COL_SHIFT); yading@10: col[16] = ((a2 + b2) >> COL_SHIFT); yading@10: col[24] = ((a3 + b3) >> COL_SHIFT); yading@10: col[32] = ((a3 - b3) >> COL_SHIFT); yading@10: col[40] = ((a2 - b2) >> COL_SHIFT); yading@10: col[48] = ((a1 - b1) >> COL_SHIFT); yading@10: col[56] = ((a0 - b0) >> COL_SHIFT); yading@10: } yading@10: yading@10: void FUNC(ff_simple_idct_put)(uint8_t *dest_, int line_size, int16_t *block) yading@10: { yading@10: pixel *dest = (pixel *)dest_; yading@10: int i; yading@10: yading@10: line_size /= sizeof(pixel); yading@10: yading@10: for (i = 0; i < 8; i++) yading@10: FUNC(idctRowCondDC)(block + i*8, 0); yading@10: yading@10: for (i = 0; i < 8; i++) yading@10: FUNC(idctSparseColPut)(dest + i, line_size, block + i); yading@10: } yading@10: yading@10: void FUNC(ff_simple_idct_add)(uint8_t *dest_, int line_size, int16_t *block) yading@10: { yading@10: pixel *dest = (pixel *)dest_; yading@10: int i; yading@10: yading@10: line_size /= sizeof(pixel); yading@10: yading@10: for (i = 0; i < 8; i++) yading@10: FUNC(idctRowCondDC)(block + i*8, 0); yading@10: yading@10: for (i = 0; i < 8; i++) yading@10: FUNC(idctSparseColAdd)(dest + i, line_size, block + i); yading@10: } yading@10: yading@10: void FUNC(ff_simple_idct)(int16_t *block) yading@10: { yading@10: int i; yading@10: yading@10: for (i = 0; i < 8; i++) yading@10: FUNC(idctRowCondDC)(block + i*8, 0); yading@10: yading@10: for (i = 0; i < 8; i++) yading@10: FUNC(idctSparseCol)(block + i); yading@10: }