simple_idct_alpha.c 7.68 KB
Newer Older
1 2 3 4 5
/*
 * Simple IDCT (Alpha optimized)
 *
 * Copyright (c) 2001 Michael Niedermayer <michaelni@gmx.at>
 *
6 7 8
 * based upon some outcommented C code from mpeg2dec (idct_mmx.c
 * written by Aaron Holtzman <aholtzma@ess.engr.uvic.ca>)
 *
9
 * Alpha optimizations by Måns Rullgård <mans@mansr.com>
10 11
 *                     and Falk Hueffner <falk@debian.org>
 *
12 13 14
 * This file is part of FFmpeg.
 *
 * FFmpeg is free software; you can redistribute it and/or
15 16
 * modify it under the terms of the GNU Lesser General Public
 * License as published by the Free Software Foundation; either
17
 * version 2.1 of the License, or (at your option) any later version.
18
 *
19
 * FFmpeg is distributed in the hope that it will be useful,
20 21 22 23 24
 * 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
25
 * License along with FFmpeg; if not, write to the Free Software
26
 * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
27 28
 */

29
#include "libavcodec/dsputil.h"
30 31
#include "asm.h"

32 33
extern void (*put_pixels_clamped_axp_p)(const DCTELEM *block, uint8_t *pixels,
                                        int line_size);
34
extern void (*add_pixels_clamped_axp_p)(const DCTELEM *block, uint8_t *pixels,
35 36
                                        int line_size);

37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53
// cos(i * M_PI / 16) * sqrt(2) * (1 << 14)
// W4 is actually exactly 16384, but using 16383 works around
// accumulating rounding errors for some encoders
#define W1 ((int_fast32_t) 22725)
#define W2 ((int_fast32_t) 21407)
#define W3 ((int_fast32_t) 19266)
#define W4 ((int_fast32_t) 16383)
#define W5 ((int_fast32_t) 12873)
#define W6 ((int_fast32_t)  8867)
#define W7 ((int_fast32_t)  4520)
#define ROW_SHIFT 11
#define COL_SHIFT 20

/* 0: all entries 0, 1: only first entry nonzero, 2: otherwise  */
static inline int idct_row(DCTELEM *row)
{
    int_fast32_t a0, a1, a2, a3, b0, b1, b2, b3, t;
54
    uint64_t l, r, t2;
55 56 57 58 59
    l = ldq(row);
    r = ldq(row + 4);

    if (l == 0 && r == 0)
        return 0;
60

61 62 63 64
    a0 = W4 * sextw(l) + (1 << (ROW_SHIFT - 1));

    if (((l & ~0xffffUL) | r) == 0) {
        a0 >>= ROW_SHIFT;
65 66 67
        t2 = (uint16_t) a0;
        t2 |= t2 << 16;
        t2 |= t2 << 32;
68

69 70
        stq(t2, row);
        stq(t2, row + 4);
71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127
        return 1;
    }

    a1 = a0;
    a2 = a0;
    a3 = a0;

    t = extwl(l, 4);            /* row[2] */
    if (t != 0) {
        t = sextw(t);
        a0 += W2 * t;
        a1 += W6 * t;
        a2 -= W6 * t;
        a3 -= W2 * t;
    }

    t = extwl(r, 0);            /* row[4] */
    if (t != 0) {
        t = sextw(t);
        a0 += W4 * t;
        a1 -= W4 * t;
        a2 -= W4 * t;
        a3 += W4 * t;
    }

    t = extwl(r, 4);            /* row[6] */
    if (t != 0) {
        t = sextw(t);
        a0 += W6 * t;
        a1 -= W2 * t;
        a2 += W2 * t;
        a3 -= W6 * t;
    }

    t = extwl(l, 2);            /* row[1] */
    if (t != 0) {
        t = sextw(t);
        b0 = W1 * t;
        b1 = W3 * t;
        b2 = W5 * t;
        b3 = W7 * t;
    } else {
        b0 = 0;
        b1 = 0;
        b2 = 0;
        b3 = 0;
    }

    t = extwl(l, 6);            /* row[3] */
    if (t) {
        t = sextw(t);
        b0 += W3 * t;
        b1 -= W7 * t;
        b2 -= W1 * t;
        b3 -= W5 * t;
    }

128

129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241
    t = extwl(r, 2);            /* row[5] */
    if (t) {
        t = sextw(t);
        b0 += W5 * t;
        b1 -= W1 * t;
        b2 += W7 * t;
        b3 += W3 * t;
    }

    t = extwl(r, 6);            /* row[7] */
    if (t) {
        t = sextw(t);
        b0 += W7 * t;
        b1 -= W5 * t;
        b2 += W3 * t;
        b3 -= W1 * t;
    }

    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;
}

static inline void idct_col(DCTELEM *col)
{
    int_fast32_t 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 = 0;
        b1 = 0;
        b2 = 0;
        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;
}

/* If all rows but the first one are zero after row transformation,
   all rows will be identical after column transformation.  */
static inline void idct_col2(DCTELEM *col)
{
    int i;
    uint64_t l, r;

    for (i = 0; i < 8; ++i) {
242
        int_fast32_t a0 = col[i] + (1 << (COL_SHIFT - 1)) / W4;
243 244

        a0 *= W4;
245
        col[i] = a0 >> COL_SHIFT;
246 247
    }

248 249 250 251 252 253 254 255
    l = ldq(col + 0 * 4); r = ldq(col + 1 * 4);
    stq(l, col +  2 * 4); stq(r, col +  3 * 4);
    stq(l, col +  4 * 4); stq(r, col +  5 * 4);
    stq(l, col +  6 * 4); stq(r, col +  7 * 4);
    stq(l, col +  8 * 4); stq(r, col +  9 * 4);
    stq(l, col + 10 * 4); stq(r, col + 11 * 4);
    stq(l, col + 12 * 4); stq(r, col + 13 * 4);
    stq(l, col + 14 * 4); stq(r, col + 15 * 4);
256 257 258 259 260 261 262 263 264 265 266 267 268 269 270 271 272 273 274 275 276 277 278
}

void simple_idct_axp(DCTELEM *block)
{

    int i;
    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 = idct_row(block + 8 * i);

        if (i > 0 && sparseness > 0)
            rowsZero = 0;
        if (sparseness == 2)
            rowsConstant = 0;
    }

    if (rowsZero) {
        idct_col2(block);
    } else if (rowsConstant) {
        idct_col(block);
        for (i = 0; i < 8; i += 2) {
279 280
            uint64_t v = (uint16_t) block[0];
            uint64_t w = (uint16_t) block[8];
281 282 283 284 285

            v |= v << 16;
            w |= w << 16;
            v |= v << 32;
            w |= w << 32;
286 287 288 289
            stq(v, block + 0 * 4);
            stq(v, block + 1 * 4);
            stq(w, block + 2 * 4);
            stq(w, block + 3 * 4);
290
            block += 4 * 4;
291 292 293 294 295 296 297 298 299 300
        }
    } else {
        for (i = 0; i < 8; i++)
            idct_col(block + i);
    }
}

void simple_idct_put_axp(uint8_t *dest, int line_size, DCTELEM *block)
{
    simple_idct_axp(block);
301
    put_pixels_clamped_axp_p(block, dest, line_size);
302 303 304 305 306
}

void simple_idct_add_axp(uint8_t *dest, int line_size, DCTELEM *block)
{
    simple_idct_axp(block);
307
    add_pixels_clamped_axp_p(block, dest, line_size);
308
}