libavcodec/arm/mpegvideo_iwmmxt.c

   1 /*
   2  * copyright (c) 2004 AGAWA Koji
   3  *
   4  * This file is part of FFmpeg.
   5  *
   6  * FFmpeg is free software; you can redistribute it and/or
   7  * modify it under the terms of the GNU Lesser General Public
   8  * License as published by the Free Software Foundation; either
   9  * version 2.1 of the License, or (at your option) any later version.
  10  *
  11  * FFmpeg is distributed in the hope that it will be useful,
  12  * but WITHOUT ANY WARRANTY; without even the implied warranty of
  13  * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
  14  * Lesser General Public License for more details.
  15  *
  16  * You should have received a copy of the GNU Lesser General Public
  17  * License along with FFmpeg; if not, write to the Free Software
  18  * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
  19  */
  20
  21 #include "libavcodec/avcodec.h"
  22 #include "libavcodec/dsputil.h"
  23 #include "libavcodec/mpegvideo.h"
  24
  25 static void dct_unquantize_h263_intra_iwmmxt(MpegEncContext *s,
  26                                              DCTELEM *block, int n, int qscale)
  27 {
  28     int level, qmul, qadd;
  29     int nCoeffs;
  30     DCTELEM *block_orig = block;
  31
  32     assert(s->block_last_index[n]>=0);
  33
  34     qmul = qscale << 1;
  35
  36     if (!s->h263_aic) {
  37         if (n < 4)
  38             level = block[0] * s->y_dc_scale;
  39         else
  40             level = block[0] * s->c_dc_scale;
  41         qadd = (qscale - 1) | 1;
  42     }else{
  43         qadd = 0;
  44         level = block[0];
  45     }
  46     if(s->ac_pred)
  47         nCoeffs=63;
  48     else
  49         nCoeffs= s->inter_scantable.raster_end[ s->block_last_index[n] ];
  50
  51     __asm__ volatile (
  52 /*      "movd %1, %%mm6                 \n\t" //qmul */
  53 /*      "packssdw %%mm6, %%mm6          \n\t" */
  54 /*      "packssdw %%mm6, %%mm6          \n\t" */
  55         "tbcsth wr6, %[qmul]            \n\t"
  56 /*      "movd %2, %%mm5                 \n\t" //qadd */
  57 /*      "packssdw %%mm5, %%mm5          \n\t" */
  58 /*      "packssdw %%mm5, %%mm5          \n\t" */
  59         "tbcsth wr5, %[qadd]            \n\t"
  60         "wzero wr7                      \n\t" /* "pxor %%mm7, %%mm7             \n\t" */
  61         "wzero wr4                      \n\t" /* "pxor %%mm4, %%mm4             \n\t" */
  62         "wsubh wr7, wr5, wr7            \n\t" /* "psubw %%mm5, %%mm7            \n\t" */
  63         "1:                             \n\t"
  64         "wldrd wr2, [%[block]]          \n\t" /* "movq (%0, %3), %%mm0          \n\t" */
  65         "wldrd wr3, [%[block], #8]      \n\t" /* "movq 8(%0, %3), %%mm1         \n\t" */
  66         "wmulsl wr0, wr6, wr2           \n\t" /* "pmullw %%mm6, %%mm0           \n\t" */
  67         "wmulsl wr1, wr6, wr3           \n\t" /* "pmullw %%mm6, %%mm1           \n\t" */
  68 /*      "movq (%0, %3), %%mm2           \n\t" */
  69 /*      "movq 8(%0, %3), %%mm3          \n\t" */
  70         "wcmpgtsh wr2, wr4, wr2         \n\t" /* "pcmpgtw %%mm4, %%mm2          \n\t" // block[i] < 0 ? -1 : 0 */
  71         "wcmpgtsh wr3, wr4, wr2         \n\t" /* "pcmpgtw %%mm4, %%mm3          \n\t" // block[i] < 0 ? -1 : 0 */
  72         "wxor wr0, wr2, wr0             \n\t" /* "pxor %%mm2, %%mm0             \n\t" */
  73         "wxor wr1, wr3, wr1             \n\t" /* "pxor %%mm3, %%mm1             \n\t" */
  74         "waddh wr0, wr7, wr0            \n\t" /* "paddw %%mm7, %%mm0            \n\t" */
  75         "waddh wr1, wr7, wr1            \n\t" /* "paddw %%mm7, %%mm1            \n\t" */
  76         "wxor wr2, wr0, wr2             \n\t" /* "pxor %%mm0, %%mm2             \n\t" */
  77         "wxor wr3, wr1, wr3             \n\t" /* "pxor %%mm1, %%mm3             \n\t" */
  78         "wcmpeqh wr0, wr7, wr0          \n\t" /* "pcmpeqw %%mm7, %%mm0          \n\t" // block[i] == 0 ? -1 : 0 */
  79         "wcmpeqh wr1, wr7, wr1          \n\t" /* "pcmpeqw %%mm7, %%mm1          \n\t" // block[i] == 0 ? -1 : 0 */
  80         "wandn wr0, wr2, wr0            \n\t" /* "pandn %%mm2, %%mm0            \n\t" */
  81         "wandn wr1, wr3, wr1            \n\t" /* "pandn %%mm3, %%mm1            \n\t" */
  82         "wstrd wr0, [%[block]]          \n\t" /* "movq %%mm0, (%0, %3)          \n\t" */
  83         "wstrd wr1, [%[block], #8]      \n\t" /* "movq %%mm1, 8(%0, %3)         \n\t" */
  84         "add %[block], %[block], #16    \n\t" /* "addl $16, %3                  \n\t" */
  85         "subs %[i], %[i], #1            \n\t"
  86         "bne 1b                         \n\t" /* "jng 1b                                \n\t" */
  87         :[block]"+r"(block)
  88         :[i]"r"((nCoeffs + 8) / 8), [qmul]"r"(qmul), [qadd]"r"(qadd)
  89         :"memory");
  90
  91     block_orig[0] = level;
  92 }
  93
  94 #if 0
  95 static void dct_unquantize_h263_inter_iwmmxt(MpegEncContext *s,
  96                                              DCTELEM *block, int n, int qscale)
  97 {
  98     int nCoeffs;
  99
 100     assert(s->block_last_index[n]>=0);
 101
 102     if(s->ac_pred)
 103         nCoeffs=63;
 104     else
 105         nCoeffs= s->inter_scantable.raster_end[ s->block_last_index[n] ];
 106
 107     ippiQuantInvInter_Compact_H263_16s_I(block, nCoeffs+1, qscale);
 108 }
 109 #endif
 110
 111 void MPV_common_init_iwmmxt(MpegEncContext *s)
 112 {
 113     if (!(mm_flags & FF_MM_IWMMXT)) return;
 114
 115     s->dct_unquantize_h263_intra = dct_unquantize_h263_intra_iwmmxt;
 116 #if 0
 117     s->dct_unquantize_h263_inter = dct_unquantize_h263_inter_iwmmxt;
 118 #endif
 119 }