/*
- Copyright (C) 2001 Michael Niedermayer (michaelni@gmx.at)
+ * Simple IDCT
+ *
+ * Copyright (c) 2001 Michael Niedermayer <michaelni@gmx.at>
+ *
+ * This file is part of Libav.
+ *
+ * Libav 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 2.1 of the License, or (at your option) any later version.
+ *
+ * Libav 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 Libav; if not, write to the Free Software
+ * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
+ */
+
+/**
+ * @file
+ * simpleidct in C.
+ */
+
+#include "libavutil/intreadwrite.h"
+#include "avcodec.h"
+#include "mathops.h"
+#include "simple_idct.h"
- This program is free software; you can redistribute it and/or modify
- it under the terms of the GNU General Public License as published by
- the Free Software Foundation; either version 2 of the License, or
- (at your option) any later version.
+#define BIT_DEPTH 8
+#include "simple_idct_template.c"
+#undef BIT_DEPTH
- This program 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 General Public License for more details.
+#define BIT_DEPTH 10
+#include "simple_idct_template.c"
+#undef BIT_DEPTH
- You should have received a copy of the GNU General Public License
- along with this program; if not, write to the Free Software
- Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA
-*/
+/* 2x4x8 idct */
-/*
- based upon some outcommented c code from mpeg2dec (idct_mmx.c written by Aaron Holtzman <aholtzma@ess.engr.uvic.ca>)
-*/
+#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)
-#include <inttypes.h>
+/* 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)
-#include "simple_idct.h"
+static inline void idct4col_put(uint8_t *dest, int line_size, const int16_t *col)
+{
+ int c0, c1, c2, c3, a0, a1, a2, a3;
+
+ 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] = av_clip_uint8((c0 + c1) >> C_SHIFT);
+ dest += line_size;
+ dest[0] = av_clip_uint8((c2 + c3) >> C_SHIFT);
+ dest += line_size;
+ dest[0] = av_clip_uint8((c2 - c3) >> C_SHIFT);
+ dest += line_size;
+ dest[0] = av_clip_uint8((c0 - c1) >> C_SHIFT);
+}
-#if 0
-#define W1 2841 /* 2048*sqrt (2)*cos (1*pi/16) */
-#define W2 2676 /* 2048*sqrt (2)*cos (2*pi/16) */
-#define W3 2408 /* 2048*sqrt (2)*cos (3*pi/16) */
-#define W4 2048 /* 2048*sqrt (2)*cos (4*pi/16) */
-#define W5 1609 /* 2048*sqrt (2)*cos (5*pi/16) */
-#define W6 1108 /* 2048*sqrt (2)*cos (6*pi/16) */
-#define W7 565 /* 2048*sqrt (2)*cos (7*pi/16) */
-#define ROW_SHIFT 8
-#define COL_SHIFT 17
-#else
-#define W1 22725 //cos(i*M_PI/16)*sqrt(2)*(1<<14) + 0.5
-#define W2 21407 //cos(i*M_PI/16)*sqrt(2)*(1<<14) + 0.5
-#define W3 19266 //cos(i*M_PI/16)*sqrt(2)*(1<<14) + 0.5
-#define W4 16384 //cos(i*M_PI/16)*sqrt(2)*(1<<14) + 0.5
-#define W5 12873 //cos(i*M_PI/16)*sqrt(2)*(1<<14) + 0.5
-#define W6 8867 //cos(i*M_PI/16)*sqrt(2)*(1<<14) + 0.5
-#define W7 4520 //cos(i*M_PI/16)*sqrt(2)*(1<<14) + 0.5
-#define ROW_SHIFT 11
-#define COL_SHIFT 20 // 6
-#endif
-#if 1
-static void inline idctRow (int16_t * row)
+#define BF(k) \
+{\
+ int a0, a1;\
+ a0 = ptr[k];\
+ a1 = ptr[8 + k];\
+ ptr[k] = a0 + a1;\
+ ptr[8 + k] = a0 - a1;\
+}
+
+/* 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, int16_t *block)
{
- int a0, a1, a2, a3, b0, b1, b2, b3;
- const int C1 =W1;
- const int C2 =W2;
- const int C3 =W3;
- const int C4 =W4;
- const int C5 =W5;
- const int C6 =W6;
- const int C7 =W7;
-
- 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;
- }
-
- a0 = C4*row[0] + C2*row[2] + C4*row[4] + C6*row[6] + (1<<(ROW_SHIFT-1));
- a1 = C4*row[0] + C6*row[2] - C4*row[4] - C2*row[6] + (1<<(ROW_SHIFT-1));
- a2 = C4*row[0] - C6*row[2] - C4*row[4] + C2*row[6] + (1<<(ROW_SHIFT-1));
- a3 = C4*row[0] - C2*row[2] + C4*row[4] - C6*row[6] + (1<<(ROW_SHIFT-1));
-
- b0 = C1*row[1] + C3*row[3] + C5*row[5] + C7*row[7];
- b1 = C3*row[1] - C7*row[3] - C1*row[5] - C5*row[7];
- b2 = C5*row[1] - C1*row[3] + C7*row[5] + C3*row[7];
- b3 = C7*row[1] - C5*row[3] + C3*row[5] - C1*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;
+ int i;
+ int16_t *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_8(block + i*8, 0);
+ }
+
+ /* 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);
+ }
}
-static void inline idctCol (int16_t * col)
+/* 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 int16_t *col)
{
- int a0, a1, a2, a3, b0, b1, b2, b3;
- const int C1 =W1;
- const int C2 =W2;
- const int C3 =W3;
- const int C4 =W4;
- const int C5 =W5;
- const int C6 =W6;
- const int C7 =W7;
-/*
- 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;
- }*/
- col[0] += (1<<(COL_SHIFT-1))/W4;
- a0 = C4*col[8*0] + C2*col[8*2] + C4*col[8*4] + C6*col[8*6];
- a1 = C4*col[8*0] + C6*col[8*2] - C4*col[8*4] - C2*col[8*6];
- a2 = C4*col[8*0] - C6*col[8*2] - C4*col[8*4] + C2*col[8*6];
- a3 = C4*col[8*0] - C2*col[8*2] + C4*col[8*4] - C6*col[8*6];
-
- b0 = C1*col[8*1] + C3*col[8*3] + C5*col[8*5] + C7*col[8*7];
- b1 = C3*col[8*1] - C7*col[8*3] - C1*col[8*5] - C5*col[8*7];
- b2 = C5*col[8*1] - C1*col[8*3] + C7*col[8*5] + C3*col[8*7];
- b3 = C7*col[8*1] - C5*col[8*3] + C3*col[8*5] - C1*col[8*7];
-
- col[8*0] = (a0 + b0) >> COL_SHIFT;
- col[8*1] = (a1 + b1) >> COL_SHIFT;
- col[8*2] = (a2 + b2) >> COL_SHIFT;
- col[8*3] = (a3 + b3) >> COL_SHIFT;
- col[8*4] = (a3 - b3) >> COL_SHIFT;
- col[8*5] = (a2 - b2) >> COL_SHIFT;
- col[8*6] = (a1 - b1) >> COL_SHIFT;
- col[8*7] = (a0 - b0) >> COL_SHIFT;
+ int c0, c1, c2, c3, a0, a1, a2, a3;
+
+ 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] = av_clip_uint8(dest[0] + ((c0 + c1) >> C_SHIFT));
+ dest += line_size;
+ dest[0] = av_clip_uint8(dest[0] + ((c2 + c3) >> C_SHIFT));
+ dest += line_size;
+ dest[0] = av_clip_uint8(dest[0] + ((c2 - c3) >> C_SHIFT));
+ dest += line_size;
+ dest[0] = av_clip_uint8(dest[0] + ((c0 - c1) >> C_SHIFT));
}
-void simple_idct (short *block)
+#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(int16_t *row)
{
- int i;
- for(i=0; i<8; i++)
- idctRow(block + 8*i);
+ int c0, c1, c2, c3, a0, a1, a2, a3;
+
+ 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;
+}
- for(i=0; i<8; i++)
- idctCol(block + i);
+void ff_simple_idct84_add(uint8_t *dest, int line_size, int16_t *block)
+{
+ int i;
+ /* IDCT8 on each line */
+ for(i=0; i<4; i++) {
+ idctRowCondDC_8(block + i*8, 0);
+ }
+
+ /* IDCT4 and store */
+ for(i=0;i<8;i++) {
+ idct4col_add(dest + i, line_size, block + i);
+ }
}
-#else
+void ff_simple_idct48_add(uint8_t *dest, int line_size, int16_t *block)
+{
+ int i;
-#define W1 22725 //cos(i*M_PI/16)*sqrt(2)*(1<<14) + 0.5
-#define W2 21407 //cos(i*M_PI/16)*sqrt(2)*(1<<14) + 0.5
-#define W3 19266 //cos(i*M_PI/16)*sqrt(2)*(1<<14) + 0.5
-#define W4 16384 //cos(i*M_PI/16)*sqrt(2)*(1<<14) + 0.5
-#define W5 12873 //cos(i*M_PI/16)*sqrt(2)*(1<<14) + 0.5
-#define W6 8867 //cos(i*M_PI/16)*sqrt(2)*(1<<14) + 0.5
-#define W7 4520 //cos(i*M_PI/16)*sqrt(2)*(1<<14) + 0.5
-#define COL_SHIFT 31 // 6
+ /* IDCT4 on each line */
+ for(i=0; i<8; i++) {
+ idct4row(block + i*8);
+ }
-static void inline idctRow (int32_t *out, int16_t * row)
-{
- int a0, a1, a2, a3, b0, b1, b2, b3;
- const int C1 =W1;
- const int C2 =W2;
- const int C3 =W3;
- const int C4 =W4;
- const int C5 =W5;
- const int C6 =W6;
- const int C7 =W7;
-/*
- 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]<<14;
- return;
- }
-*/
- a0 = C4*row[0] + C2*row[2] + C4*row[4] + C6*row[6];
- a1 = C4*row[0] + C6*row[2] - C4*row[4] - C2*row[6];
- a2 = C4*row[0] - C6*row[2] - C4*row[4] + C2*row[6];
- a3 = C4*row[0] - C2*row[2] + C4*row[4] - C6*row[6];
-
- b0 = C1*row[1] + C3*row[3] + C5*row[5] + C7*row[7];
- b1 = C3*row[1] - C7*row[3] - C1*row[5] - C5*row[7];
- b2 = C5*row[1] - C1*row[3] + C7*row[5] + C3*row[7];
- b3 = C7*row[1] - C5*row[3] + C3*row[5] - C1*row[7];
-
- out[0] = (a0 + b0);
- out[1] = (a1 + b1);
- out[2] = (a2 + b2);
- out[3] = (a3 + b3);
- out[4] = (a3 - b3);
- out[5] = (a2 - b2);
- out[6] = (a1 - b1);
- out[7] = (a0 - b0);
+ /* IDCT8 and store */
+ for(i=0; i<4; i++){
+ idctSparseColAdd_8(dest + i, line_size, block + i);
+ }
}
-static void inline idctCol (int32_t *in, int16_t * col)
+void ff_simple_idct44_add(uint8_t *dest, int line_size, int16_t *block)
{
- int64_t a0, a1, a2, a3, b0, b1, b2, b3;
- const int64_t C1 =W1;
- const int64_t C2 =W2;
- const int64_t C3 =W3;
- const int64_t C4 =W4;
- const int64_t C5 =W5;
- const int64_t C6 =W6;
- const int64_t C7 =W7;
-/*
- 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;
- }*/
- in[0] += (1<<(COL_SHIFT-1))/W4;
- a0 = C4*in[8*0] + C2*in[8*2] + C4*in[8*4] + C6*in[8*6];
- a1 = C4*in[8*0] + C6*in[8*2] - C4*in[8*4] - C2*in[8*6];
- a2 = C4*in[8*0] - C6*in[8*2] - C4*in[8*4] + C2*in[8*6];
- a3 = C4*in[8*0] - C2*in[8*2] + C4*in[8*4] - C6*in[8*6];
-
- b0 = C1*in[8*1] + C3*in[8*3] + C5*in[8*5] + C7*in[8*7];
- b1 = C3*in[8*1] - C7*in[8*3] - C1*in[8*5] - C5*in[8*7];
- b2 = C5*in[8*1] - C1*in[8*3] + C7*in[8*5] + C3*in[8*7];
- b3 = C7*in[8*1] - C5*in[8*3] + C3*in[8*5] - C1*in[8*7];
-
- col[8*0] = (a0 + b0) >> COL_SHIFT;
- col[8*1] = (a1 + b1) >> COL_SHIFT;
- col[8*2] = (a2 + b2) >> COL_SHIFT;
- col[8*3] = (a3 + b3) >> COL_SHIFT;
- col[8*4] = (a3 - b3) >> COL_SHIFT;
- col[8*5] = (a2 - b2) >> COL_SHIFT;
- col[8*6] = (a1 - b1) >> COL_SHIFT;
- col[8*7] = (a0 - b0) >> COL_SHIFT;
+ 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);
+ }
}
-void simple_idct (short *block)
+void ff_prores_idct(int16_t *block, const int16_t *qmat)
{
- int i;
- int32_t temp[64];
- for(i=0; i<8; i++)
- idctRow(temp+8*i, block + 8*i);
+ int i;
- for(i=0; i<8; i++)
- idctCol(temp+i, block + i);
+ for (i = 0; i < 64; i++)
+ block[i] *= qmat[i];
-}
+ for (i = 0; i < 8; i++)
+ idctRowCondDC_10(block + i*8, 2);
-#endif
+ for (i = 0; i < 8; i++)
+ idctSparseCol_10(block + i);
+}