quicktime/ffmpeg/libavcodec/armv4l/dsputil_iwmmxt.c

   1 /*
   2  * iWMMXt optimized DSP utils
   3  * Copyright (c) 2004 AGAWA Koji
   4  *
   5  * This library is free software; you can redistribute it and/or
   6  * modify it under the terms of the GNU Lesser General Public
   7  * License as published by the Free Software Foundation; either
   8  * version 2 of the License, or (at your option) any later version.
   9  *
  10  * This library is distributed in the hope that it will be useful,
  11  * but WITHOUT ANY WARRANTY; without even the implied warranty of
  12  * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
  13  * Lesser General Public License for more details.
  14  *
  15  * You should have received a copy of the GNU Lesser General Public
  16  * License along with this library; if not, write to the Free Software
  17  * Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA  02111-1307  USA
  18  */
  19
  20 #include "../dsputil.h"
  21
  22 #define DEF(x, y) x ## _no_rnd_ ## y ##_iwmmxt
  23 #define SET_RND(regd)  __asm__ __volatile__ ("mov r12, #1 \n\t tbcsth " #regd ", r12":::"r12");
  24 #define WAVG2B "wavg2b"
  25 #include "dsputil_iwmmxt_rnd.h"
  26 #undef DEF
  27 #undef SET_RND
  28 #undef WAVG2B
  29
  30 #define DEF(x, y) x ## _ ## y ##_iwmmxt
  31 #define SET_RND(regd)  __asm__ __volatile__ ("mov r12, #2 \n\t tbcsth " #regd ", r12":::"r12");
  32 #define WAVG2B "wavg2br"
  33 #include "dsputil_iwmmxt_rnd.h"
  34 #undef DEF
  35 #undef SET_RND
  36 #undef WAVG2BR
  37
  38 // need scheduling
  39 #define OP(AVG)                                         \
  40     asm volatile (                                      \
  41         /* alignment */                                 \
  42         "and r12, %[pixels], #7 \n\t"                   \
  43         "bic %[pixels], %[pixels], #7 \n\t"             \
  44         "tmcr wcgr1, r12 \n\t"                          \
  45                                                         \
  46         "wldrd wr0, [%[pixels]] \n\t"                   \
  47         "wldrd wr1, [%[pixels], #8] \n\t"               \
  48         "add %[pixels], %[pixels], %[line_size] \n\t"   \
  49         "walignr1 wr4, wr0, wr1 \n\t"                   \
  50                                                         \
  51         "1: \n\t"                                       \
  52                                                         \
  53         "wldrd wr2, [%[pixels]] \n\t"                   \
  54         "wldrd wr3, [%[pixels], #8] \n\t"               \
  55         "add %[pixels], %[pixels], %[line_size] \n\t"   \
  56         "pld [%[pixels]] \n\t"                          \
  57         "walignr1 wr5, wr2, wr3 \n\t"                   \
  58         AVG " wr6, wr4, wr5 \n\t"                       \
  59         "wstrd wr6, [%[block]] \n\t"                    \
  60         "add %[block], %[block], %[line_size] \n\t"     \
  61                                                         \
  62         "wldrd wr0, [%[pixels]] \n\t"                   \
  63         "wldrd wr1, [%[pixels], #8] \n\t"               \
  64         "add %[pixels], %[pixels], %[line_size] \n\t"   \
  65         "walignr1 wr4, wr0, wr1 \n\t"                   \
  66         "pld [%[pixels]] \n\t"                          \
  67         AVG " wr6, wr4, wr5 \n\t"                       \
  68         "wstrd wr6, [%[block]] \n\t"                    \
  69         "add %[block], %[block], %[line_size] \n\t"     \
  70                                                         \
  71         "subs %[h], %[h], #2 \n\t"                      \
  72         "bne 1b \n\t"                                   \
  73         : [block]"+r"(block), [pixels]"+r"(pixels), [h]"+r"(h)  \
  74         : [line_size]"r"(line_size) \
  75         : "memory", "r12");
  76 void put_pixels8_y2_iwmmxt(uint8_t *block, const uint8_t *pixels, const int line_size, int h)
  77 {
  78     OP("wavg2br");
  79 }
  80 void put_no_rnd_pixels8_y2_iwmmxt(uint8_t *block, const uint8_t *pixels, const int line_size, int h)
  81 {
  82     OP("wavg2b");
  83 }
  84 #undef OP
  85
  86 void add_pixels_clamped_iwmmxt(const DCTELEM *block, uint8_t *pixels, int line_size)
  87 {
  88     uint8_t *pixels2 = pixels + line_size;
  89
  90     __asm__ __volatile__ (
  91         "mov            r12, #4                 \n\t"
  92         "1:                                     \n\t"
  93         "pld            [%[pixels], %[line_size2]]              \n\t"
  94         "pld            [%[pixels2], %[line_size2]]             \n\t"
  95         "wldrd          wr4, [%[pixels]]        \n\t"
  96         "wldrd          wr5, [%[pixels2]]       \n\t"
  97         "pld            [%[block], #32]         \n\t"
  98         "wunpckelub     wr6, wr4                \n\t"
  99         "wldrd          wr0, [%[block]]         \n\t"
 100         "wunpckehub     wr7, wr4                \n\t"
 101         "wldrd          wr1, [%[block], #8]     \n\t"
 102         "wunpckelub     wr8, wr5                \n\t"
 103         "wldrd          wr2, [%[block], #16]    \n\t"
 104         "wunpckehub     wr9, wr5                \n\t"
 105         "wldrd          wr3, [%[block], #24]    \n\t"
 106         "add            %[block], %[block], #32 \n\t"
 107         "waddhss        wr10, wr0, wr6          \n\t"
 108         "waddhss        wr11, wr1, wr7          \n\t"
 109         "waddhss        wr12, wr2, wr8          \n\t"
 110         "waddhss        wr13, wr3, wr9          \n\t"
 111         "wpackhus       wr14, wr10, wr11        \n\t"
 112         "wpackhus       wr15, wr12, wr13        \n\t"
 113         "wstrd          wr14, [%[pixels]]       \n\t"
 114         "add            %[pixels], %[pixels], %[line_size2]     \n\t"
 115         "subs           r12, r12, #1            \n\t"
 116         "wstrd          wr15, [%[pixels2]]      \n\t"
 117         "add            %[pixels2], %[pixels2], %[line_size2]   \n\t"
 118         "bne            1b                      \n\t"
 119         : [block]"+r"(block), [pixels]"+r"(pixels), [pixels2]"+r"(pixels2)
 120         : [line_size2]"r"(line_size << 1)
 121         : "cc", "memory", "r12");
 122 }
 123
 124 static void nop(uint8_t *block, const uint8_t *pixels, int line_size, int h)
 125 {
 126     return;
 127 }
 128
 129 int mm_flags; /* multimedia extension flags */
 130
 131 int mm_support(void)
 132 {
 133     return 0; /* TODO, implement proper detection */
 134 }
 135
 136 void dsputil_init_iwmmxt(DSPContext* c, AVCodecContext *avctx)
 137 {
 138     mm_flags = mm_support();
 139
 140     if (avctx->dsp_mask) {
 141         if (avctx->dsp_mask & FF_MM_FORCE)
 142             mm_flags |= (avctx->dsp_mask & 0xffff);
 143         else
 144             mm_flags &= ~(avctx->dsp_mask & 0xffff);
 145     }
 146
 147     if (!(mm_flags & MM_IWMMXT)) return;
 148
 149     c->add_pixels_clamped = add_pixels_clamped_iwmmxt;
 150
 151     c->put_pixels_tab[0][0] = put_pixels16_iwmmxt;
 152     c->put_pixels_tab[0][1] = put_pixels16_x2_iwmmxt;
 153     c->put_pixels_tab[0][2] = put_pixels16_y2_iwmmxt;
 154     c->put_pixels_tab[0][3] = put_pixels16_xy2_iwmmxt;
 155     c->put_no_rnd_pixels_tab[0][0] = put_pixels16_iwmmxt;
 156     c->put_no_rnd_pixels_tab[0][1] = put_no_rnd_pixels16_x2_iwmmxt;
 157     c->put_no_rnd_pixels_tab[0][2] = put_no_rnd_pixels16_y2_iwmmxt;
 158     c->put_no_rnd_pixels_tab[0][3] = put_no_rnd_pixels16_xy2_iwmmxt;
 159
 160     c->put_pixels_tab[1][0] = put_pixels8_iwmmxt;
 161     c->put_pixels_tab[1][1] = put_pixels8_x2_iwmmxt;
 162     c->put_pixels_tab[1][2] = put_pixels8_y2_iwmmxt;
 163     c->put_pixels_tab[1][3] = put_pixels8_xy2_iwmmxt;
 164     c->put_no_rnd_pixels_tab[1][0] = put_pixels8_iwmmxt;
 165     c->put_no_rnd_pixels_tab[1][1] = put_no_rnd_pixels8_x2_iwmmxt;
 166     c->put_no_rnd_pixels_tab[1][2] = put_no_rnd_pixels8_y2_iwmmxt;
 167     c->put_no_rnd_pixels_tab[1][3] = put_no_rnd_pixels8_xy2_iwmmxt;
 168
 169     c->avg_pixels_tab[0][0] = avg_pixels16_iwmmxt;
 170     c->avg_pixels_tab[0][1] = avg_pixels16_x2_iwmmxt;
 171     c->avg_pixels_tab[0][2] = avg_pixels16_y2_iwmmxt;
 172     c->avg_pixels_tab[0][3] = avg_pixels16_xy2_iwmmxt;
 173     c->avg_no_rnd_pixels_tab[0][0] = avg_pixels16_iwmmxt;
 174     c->avg_no_rnd_pixels_tab[0][1] = avg_no_rnd_pixels16_x2_iwmmxt;
 175     c->avg_no_rnd_pixels_tab[0][2] = avg_no_rnd_pixels16_y2_iwmmxt;
 176     c->avg_no_rnd_pixels_tab[0][3] = avg_no_rnd_pixels16_xy2_iwmmxt;
 177
 178     c->avg_pixels_tab[1][0] = avg_pixels8_iwmmxt;
 179     c->avg_pixels_tab[1][1] = avg_pixels8_x2_iwmmxt;
 180     c->avg_pixels_tab[1][2] = avg_pixels8_y2_iwmmxt;
 181     c->avg_pixels_tab[1][3] = avg_pixels8_xy2_iwmmxt;
 182     c->avg_no_rnd_pixels_tab[1][0] = avg_no_rnd_pixels8_iwmmxt;
 183     c->avg_no_rnd_pixels_tab[1][1] = avg_no_rnd_pixels8_x2_iwmmxt;
 184     c->avg_no_rnd_pixels_tab[1][2] = avg_no_rnd_pixels8_y2_iwmmxt;
 185     c->avg_no_rnd_pixels_tab[1][3] = avg_no_rnd_pixels8_xy2_iwmmxt;
 186 }