libavcodec/alpha/mpegvideo_alpha.c

   1 /*
   2  * Alpha optimized DSP utils
   3  * Copyright (c) 2002 Falk Hueffner <falk@debian.org>
   4  *
   5  * This file is part of FFmpeg.
   6  *
   7  * FFmpeg is free software; you can redistribute it and/or
   8  * modify it under the terms of the GNU Lesser General Public
   9  * License as published by the Free Software Foundation; either
  10  * version 2.1 of the License, or (at your option) any later version.
  11  *
  12  * FFmpeg is distributed in the hope that it will be useful,
  13  * but WITHOUT ANY WARRANTY; without even the implied warranty of
  14  * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
  15  * Lesser General Public License for more details.
  16  *
  17  * You should have received a copy of the GNU Lesser General Public
  18  * License along with FFmpeg; if not, write to the Free Software
  19  * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
  20  */
  21
  22 #include "libavcodec/dsputil.h"
  23 #include "libavcodec/mpegvideo.h"
  24 #include "asm.h"
  25
  26 static void dct_unquantize_h263_intra_axp(MpegEncContext *s, DCTELEM *block,
  27                                     int n, int qscale)
  28 {
  29     int i, n_coeffs;
  30     uint64_t qmul, qadd;
  31     uint64_t correction;
  32     DCTELEM *orig_block = block;
  33     DCTELEM block0;             /* might not be used uninitialized */
  34
  35     qadd = WORD_VEC((qscale - 1) | 1);
  36     qmul = qscale << 1;
  37     /* This mask kills spill from negative subwords to the next subword.  */
  38     correction = WORD_VEC((qmul - 1) + 1); /* multiplication / addition */
  39
  40     if (!s->h263_aic) {
  41         if (n < 4)
  42             block0 = block[0] * s->y_dc_scale;
  43         else
  44             block0 = block[0] * s->c_dc_scale;
  45     } else {
  46         qadd = 0;
  47     }
  48     n_coeffs = 63; // does not always use zigzag table
  49
  50     for(i = 0; i <= n_coeffs; block += 4, i += 4) {
  51         uint64_t levels, negmask, zeros, add;
  52
  53         levels = ldq(block);
  54         if (levels == 0)
  55             continue;
  56
  57 #ifdef __alpha_max__
  58         /* I don't think the speed difference justifies runtime
  59            detection.  */
  60         negmask = maxsw4(levels, -1); /* negative -> ffff (-1) */
  61         negmask = minsw4(negmask, 0); /* positive -> 0000 (0) */
  62 #else
  63         negmask = cmpbge(WORD_VEC(0x7fff), levels);
  64         negmask &= (negmask >> 1) | (1 << 7);
  65         negmask = zap(-1, negmask);
  66 #endif
  67
  68         zeros = cmpbge(0, levels);
  69         zeros &= zeros >> 1;
  70         /* zeros |= zeros << 1 is not needed since qadd <= 255, so
  71            zapping the lower byte suffices.  */
  72
  73         levels *= qmul;
  74         levels -= correction & (negmask << 16);
  75
  76         /* Negate qadd for negative levels.  */
  77         add = qadd ^ negmask;
  78         add += WORD_VEC(0x0001) & negmask;
  79         /* Set qadd to 0 for levels == 0.  */
  80         add = zap(add, zeros);
  81
  82         levels += add;
  83
  84         stq(levels, block);
  85     }
  86
  87     if (s->mb_intra && !s->h263_aic)
  88         orig_block[0] = block0;
  89 }
  90
  91 static void dct_unquantize_h263_inter_axp(MpegEncContext *s, DCTELEM *block,
  92                                     int n, int qscale)
  93 {
  94     int i, n_coeffs;
  95     uint64_t qmul, qadd;
  96     uint64_t correction;
  97
  98     qadd = WORD_VEC((qscale - 1) | 1);
  99     qmul = qscale << 1;
 100     /* This mask kills spill from negative subwords to the next subword.  */
 101     correction = WORD_VEC((qmul - 1) + 1); /* multiplication / addition */
 102
 103     n_coeffs = s->intra_scantable.raster_end[s->block_last_index[n]];
 104
 105     for(i = 0; i <= n_coeffs; block += 4, i += 4) {
 106         uint64_t levels, negmask, zeros, add;
 107
 108         levels = ldq(block);
 109         if (levels == 0)
 110             continue;
 111
 112 #ifdef __alpha_max__
 113         /* I don't think the speed difference justifies runtime
 114            detection.  */
 115         negmask = maxsw4(levels, -1); /* negative -> ffff (-1) */
 116         negmask = minsw4(negmask, 0); /* positive -> 0000 (0) */
 117 #else
 118         negmask = cmpbge(WORD_VEC(0x7fff), levels);
 119         negmask &= (negmask >> 1) | (1 << 7);
 120         negmask = zap(-1, negmask);
 121 #endif
 122
 123         zeros = cmpbge(0, levels);
 124         zeros &= zeros >> 1;
 125         /* zeros |= zeros << 1 is not needed since qadd <= 255, so
 126            zapping the lower byte suffices.  */
 127
 128         levels *= qmul;
 129         levels -= correction & (negmask << 16);
 130
 131         /* Negate qadd for negative levels.  */
 132         add = qadd ^ negmask;
 133         add += WORD_VEC(0x0001) & negmask;
 134         /* Set qadd to 0 for levels == 0.  */
 135         add = zap(add, zeros);
 136
 137         levels += add;
 138
 139         stq(levels, block);
 140     }
 141 }
 142
 143 void MPV_common_init_axp(MpegEncContext *s)
 144 {
 145     s->dct_unquantize_h263_intra = dct_unquantize_h263_intra_axp;
 146     s->dct_unquantize_h263_inter = dct_unquantize_h263_inter_axp;
 147 }