vpx_dsp/mips/itrans8_dspr2.c

   1 /*
   2  *  Copyright (c) 2015 The WebM project authors. All Rights Reserved.
   3  *
   4  *  Use of this source code is governed by a BSD-style license
   5  *  that can be found in the LICENSE file in the root of the source
   6  *  tree. An additional intellectual property rights grant can be found
   7  *  in the file PATENTS.  All contributing project authors may
   8  *  be found in the AUTHORS file in the root of the source tree.
   9  */
  10
  11 #include "./vpx_config.h"
  12 #include "./vpx_dsp_rtcd.h"
  13 #include "vpx_dsp/mips/inv_txfm_dspr2.h"
  14 #include "vpx_dsp/txfm_common.h"
  15
  16 #if HAVE_DSPR2
  17 void idct8_rows_dspr2(const int16_t *input, int16_t *output, uint32_t no_rows) {
  18   int step1_0, step1_1, step1_2, step1_3, step1_4, step1_5, step1_6, step1_7;
  19   const int const_2_power_13 = 8192;
  20   int Temp0, Temp1, Temp2, Temp3, Temp4;
  21   int i;
  22
  23   for (i = no_rows; i--; ) {
  24     __asm__ __volatile__ (
  25         /*
  26           temp_1 = (input[0] + input[4]) * cospi_16_64;
  27           step2_0 = dct_const_round_shift(temp_1);
  28
  29           temp_2 = (input[0] - input[4]) * cospi_16_64;
  30           step2_1 = dct_const_round_shift(temp_2);
  31         */
  32         "lh       %[Temp0],             0(%[input])                     \n\t"
  33         "lh       %[Temp1],             8(%[input])                     \n\t"
  34         "mtlo     %[const_2_power_13],  $ac0                            \n\t"
  35         "mthi     $zero,                $ac0                            \n\t"
  36         "mtlo     %[const_2_power_13],  $ac1                            \n\t"
  37         "mthi     $zero,                $ac1                            \n\t"
  38         "add      %[Temp2],             %[Temp0],       %[Temp1]        \n\t"
  39         "madd     $ac0,                 %[Temp2],       %[cospi_16_64]  \n\t"
  40         "extp     %[Temp4],             $ac0,           31              \n\t"
  41
  42         "sub      %[Temp3],             %[Temp0],       %[Temp1]        \n\t"
  43         "madd     $ac1,                 %[Temp3],       %[cospi_16_64]  \n\t"
  44         "mtlo     %[const_2_power_13],  $ac0                            \n\t"
  45         "mthi     $zero,                $ac0                            \n\t"
  46         "extp     %[Temp2],             $ac1,           31              \n\t"
  47
  48         /*
  49           temp_1 = input[2] * cospi_24_64 - input[6] * cospi_8_64;
  50           step2_2 = dct_const_round_shift(temp_1);
  51         */
  52         "lh       %[Temp0],             4(%[input])                     \n\t"
  53         "lh       %[Temp1],             12(%[input])                    \n\t"
  54         "madd     $ac0,                 %[Temp0],       %[cospi_24_64]  \n\t"
  55         "msub     $ac0,                 %[Temp1],       %[cospi_8_64]   \n\t"
  56         "mtlo     %[const_2_power_13],  $ac1                            \n\t"
  57         "mthi     $zero,                $ac1                            \n\t"
  58         "extp     %[Temp3],             $ac0,           31              \n\t"
  59
  60         /*
  61           step1_1 = step2_1 + step2_2;
  62           step1_2 = step2_1 - step2_2;
  63         */
  64         "add      %[step1_1],           %[Temp2],       %[Temp3]        \n\t"
  65         "sub      %[step1_2],           %[Temp2],       %[Temp3]        \n\t"
  66
  67         /*
  68           temp_2 = input[2] * cospi_8_64 + input[6] * cospi_24_64;
  69           step2_3 = dct_const_round_shift(temp_2);
  70         */
  71         "madd     $ac1,                 %[Temp0],       %[cospi_8_64]   \n\t"
  72         "madd     $ac1,                 %[Temp1],       %[cospi_24_64]  \n\t"
  73         "extp     %[Temp1],             $ac1,           31              \n\t"
  74
  75         "mtlo     %[const_2_power_13],  $ac0                            \n\t"
  76         "mthi     $zero,                $ac0                            \n\t"
  77
  78         /*
  79           step1_0 = step2_0 + step2_3;
  80           step1_3 = step2_0 - step2_3;
  81         */
  82         "add      %[step1_0],           %[Temp4],       %[Temp1]        \n\t"
  83         "sub      %[step1_3],           %[Temp4],       %[Temp1]        \n\t"
  84
  85         /*
  86           temp_1 = input[1] * cospi_28_64 - input[7] * cospi_4_64;
  87           step1_4 = dct_const_round_shift(temp_1);
  88         */
  89         "lh       %[Temp0],             2(%[input])                     \n\t"
  90         "madd     $ac0,                 %[Temp0],       %[cospi_28_64]  \n\t"
  91         "mtlo     %[const_2_power_13],  $ac1                            \n\t"
  92         "mthi     $zero,                $ac1                            \n\t"
  93         "lh       %[Temp1],             14(%[input])                    \n\t"
  94         "lh       %[Temp0],             2(%[input])                     \n\t"
  95         "msub     $ac0,                 %[Temp1],       %[cospi_4_64]   \n\t"
  96         "extp     %[step1_4],           $ac0,           31              \n\t"
  97
  98         /*
  99           temp_2 = input[1] * cospi_4_64 + input[7] * cospi_28_64;
 100           step1_7 = dct_const_round_shift(temp_2);
 101         */
 102         "madd     $ac1,                 %[Temp0],       %[cospi_4_64]   \n\t"
 103         "madd     $ac1,                 %[Temp1],       %[cospi_28_64]  \n\t"
 104         "extp     %[step1_7],           $ac1,           31              \n\t"
 105
 106         /*
 107           temp_1 = input[5] * cospi_12_64 - input[3] * cospi_20_64;
 108           step1_5 = dct_const_round_shift(temp_1);
 109         */
 110         "mtlo     %[const_2_power_13],  $ac0                            \n\t"
 111         "mthi     $zero,                $ac0                            \n\t"
 112         "lh       %[Temp0],             10(%[input])                    \n\t"
 113         "madd     $ac0,                 %[Temp0],       %[cospi_12_64]  \n\t"
 114         "lh       %[Temp1],             6(%[input])                     \n\t"
 115         "msub     $ac0,                 %[Temp1],       %[cospi_20_64]  \n\t"
 116         "extp     %[step1_5],           $ac0,           31              \n\t"
 117
 118         /*
 119           temp_2 = input[5] * cospi_20_64 + input[3] * cospi_12_64;
 120           step1_6 = dct_const_round_shift(temp_2);
 121         */
 122         "mtlo     %[const_2_power_13],  $ac1                            \n\t"
 123         "mthi     $zero,                $ac1                            \n\t"
 124         "lh       %[Temp0],             10(%[input])                    \n\t"
 125         "madd     $ac1,                 %[Temp0],       %[cospi_20_64]  \n\t"
 126         "lh       %[Temp1],             6(%[input])                     \n\t"
 127         "madd     $ac1,                 %[Temp1],       %[cospi_12_64]  \n\t"
 128         "extp     %[step1_6],           $ac1,           31              \n\t"
 129
 130         /*
 131           temp_1 = (step1_7 - step1_6 - step1_4 + step1_5) * cospi_16_64;
 132           temp_2 = (step1_4 - step1_5 - step1_6 + step1_7) * cospi_16_64;
 133         */
 134         "sub      %[Temp0],             %[step1_7],     %[step1_6]      \n\t"
 135         "sub      %[Temp0],             %[Temp0],       %[step1_4]      \n\t"
 136         "add      %[Temp0],             %[Temp0],       %[step1_5]      \n\t"
 137         "sub      %[Temp1],             %[step1_4],     %[step1_5]      \n\t"
 138         "sub      %[Temp1],             %[Temp1],       %[step1_6]      \n\t"
 139         "add      %[Temp1],             %[Temp1],       %[step1_7]      \n\t"
 140
 141         "mtlo     %[const_2_power_13],  $ac0                            \n\t"
 142         "mthi     $zero,                $ac0                            \n\t"
 143         "mtlo     %[const_2_power_13],  $ac1                            \n\t"
 144         "mthi     $zero,                $ac1                            \n\t"
 145
 146         "madd     $ac0,                 %[Temp0],       %[cospi_16_64]  \n\t"
 147         "madd     $ac1,                 %[Temp1],       %[cospi_16_64]  \n\t"
 148
 149         /*
 150           step1_4 = step1_4 + step1_5;
 151           step1_7 = step1_6 + step1_7;
 152         */
 153         "add      %[step1_4],           %[step1_4],     %[step1_5]      \n\t"
 154         "add      %[step1_7],           %[step1_7],     %[step1_6]      \n\t"
 155
 156         "extp     %[step1_5],           $ac0,           31              \n\t"
 157         "extp     %[step1_6],           $ac1,           31              \n\t"
 158
 159         "add      %[Temp0],             %[step1_0],     %[step1_7]      \n\t"
 160         "sh       %[Temp0],             0(%[output])                    \n\t"
 161         "add      %[Temp1],             %[step1_1],     %[step1_6]      \n\t"
 162         "sh       %[Temp1],             16(%[output])                   \n\t"
 163         "add      %[Temp0],             %[step1_2],     %[step1_5]      \n\t"
 164         "sh       %[Temp0],             32(%[output])                   \n\t"
 165         "add      %[Temp1],             %[step1_3],     %[step1_4]      \n\t"
 166         "sh       %[Temp1],             48(%[output])                   \n\t"
 167
 168         "sub      %[Temp0],             %[step1_3],     %[step1_4]      \n\t"
 169         "sh       %[Temp0],             64(%[output])                   \n\t"
 170         "sub      %[Temp1],             %[step1_2],     %[step1_5]      \n\t"
 171         "sh       %[Temp1],             80(%[output])                   \n\t"
 172         "sub      %[Temp0],             %[step1_1],     %[step1_6]      \n\t"
 173         "sh       %[Temp0],             96(%[output])                   \n\t"
 174         "sub      %[Temp1],             %[step1_0],     %[step1_7]      \n\t"
 175         "sh       %[Temp1],             112(%[output])                  \n\t"
 176
 177         : [step1_0] "=&r" (step1_0), [step1_1] "=&r" (step1_1),
 178           [step1_2] "=&r" (step1_2), [step1_3] "=&r" (step1_3),
 179           [step1_4] "=&r" (step1_4), [step1_5] "=&r" (step1_5),
 180           [step1_6] "=&r" (step1_6), [step1_7] "=&r" (step1_7),
 181           [Temp0] "=&r" (Temp0), [Temp1] "=&r" (Temp1),
 182           [Temp2] "=&r" (Temp2), [Temp3] "=&r" (Temp3),
 183           [Temp4] "=&r" (Temp4)
 184         : [const_2_power_13] "r" (const_2_power_13),
 185           [cospi_16_64] "r" (cospi_16_64), [cospi_28_64] "r" (cospi_28_64),
 186           [cospi_4_64] "r" (cospi_4_64), [cospi_12_64] "r" (cospi_12_64),
 187           [cospi_20_64] "r" (cospi_20_64), [cospi_8_64] "r" (cospi_8_64),
 188           [cospi_24_64] "r" (cospi_24_64),
 189           [output] "r" (output), [input] "r" (input)
 190     );
 191
 192     input += 8;
 193     output += 1;
 194   }
 195 }
 196
 197 void idct8_columns_add_blk_dspr2(int16_t *input, uint8_t *dest,
 198                                  int dest_stride) {
 199   int step1_0, step1_1, step1_2, step1_3, step1_4, step1_5, step1_6, step1_7;
 200   int Temp0, Temp1, Temp2, Temp3;
 201   int i;
 202   const int const_2_power_13 = 8192;
 203   uint8_t *dest_pix;
 204   uint8_t *cm = vpx_ff_cropTbl;
 205
 206   /* prefetch vpx_ff_cropTbl */
 207   prefetch_load(vpx_ff_cropTbl);
 208   prefetch_load(vpx_ff_cropTbl +  32);
 209   prefetch_load(vpx_ff_cropTbl +  64);
 210   prefetch_load(vpx_ff_cropTbl +  96);
 211   prefetch_load(vpx_ff_cropTbl + 128);
 212   prefetch_load(vpx_ff_cropTbl + 160);
 213   prefetch_load(vpx_ff_cropTbl + 192);
 214   prefetch_load(vpx_ff_cropTbl + 224);
 215
 216   for (i = 0; i < 8; ++i) {
 217       dest_pix = (dest + i);
 218
 219     __asm__ __volatile__ (
 220         /*
 221           temp_1 = (input[0] + input[4]) * cospi_16_64;
 222           step2_0 = dct_const_round_shift(temp_1);
 223
 224           temp_2 = (input[0] - input[4]) * cospi_16_64;
 225           step2_1 = dct_const_round_shift(temp_2);
 226         */
 227         "lh       %[Temp0],             0(%[input])                     \n\t"
 228         "lh       %[Temp1],             8(%[input])                     \n\t"
 229         "mtlo     %[const_2_power_13],  $ac0                            \n\t"
 230         "mthi     $zero,                $ac0                            \n\t"
 231         "mtlo     %[const_2_power_13],  $ac1                            \n\t"
 232         "mthi     $zero,                $ac1                            \n\t"
 233         "add      %[Temp2],             %[Temp0],       %[Temp1]        \n\t"
 234         "madd     $ac0,                 %[Temp2],       %[cospi_16_64]  \n\t"
 235         "extp     %[step1_6],           $ac0,           31              \n\t"
 236
 237         "sub      %[Temp3],             %[Temp0],       %[Temp1]        \n\t"
 238         "madd     $ac1,                 %[Temp3],       %[cospi_16_64]  \n\t"
 239         "mtlo     %[const_2_power_13],  $ac0                            \n\t"
 240         "mthi     $zero,                $ac0                            \n\t"
 241         "extp     %[Temp2],             $ac1,           31              \n\t"
 242
 243         /*
 244           temp_1 = input[2] * cospi_24_64 - input[6] * cospi_8_64;
 245           step2_2 = dct_const_round_shift(temp_1);
 246         */
 247         "lh       %[Temp0],             4(%[input])                     \n\t"
 248         "lh       %[Temp1],             12(%[input])                    \n\t"
 249         "madd     $ac0,                 %[Temp0],       %[cospi_24_64]  \n\t"
 250         "msub     $ac0,                 %[Temp1],       %[cospi_8_64]   \n\t"
 251         "mtlo     %[const_2_power_13],  $ac1                            \n\t"
 252         "mthi     $zero,                $ac1                            \n\t"
 253         "extp     %[Temp3],             $ac0,           31              \n\t"
 254
 255         /*
 256           step1_1 = step2_1 + step2_2;
 257           step1_2 = step2_1 - step2_2;
 258         */
 259         "add      %[step1_1],           %[Temp2],       %[Temp3]        \n\t"
 260         "sub      %[step1_2],           %[Temp2],       %[Temp3]        \n\t"
 261
 262         /*
 263           temp_2 = input[2] * cospi_8_64 + input[6] * cospi_24_64;
 264           step2_3 = dct_const_round_shift(temp_2);
 265         */
 266         "madd     $ac1,                 %[Temp0],       %[cospi_8_64]   \n\t"
 267         "madd     $ac1,                 %[Temp1],       %[cospi_24_64]  \n\t"
 268         "extp     %[Temp1],             $ac1,           31              \n\t"
 269
 270         "mtlo     %[const_2_power_13],  $ac0                            \n\t"
 271         "mthi     $zero,                $ac0                            \n\t"
 272
 273         /*
 274           step1_0 = step2_0 + step2_3;
 275           step1_3 = step2_0 - step2_3;
 276         */
 277         "add      %[step1_0],           %[step1_6],     %[Temp1]        \n\t"
 278         "sub      %[step1_3],           %[step1_6],     %[Temp1]        \n\t"
 279
 280         /*
 281           temp_1 = input[1] * cospi_28_64 - input[7] * cospi_4_64;
 282           step1_4 = dct_const_round_shift(temp_1);
 283         */
 284         "lh       %[Temp0],             2(%[input])                     \n\t"
 285         "madd     $ac0,                 %[Temp0],       %[cospi_28_64]  \n\t"
 286         "mtlo     %[const_2_power_13],  $ac1                            \n\t"
 287         "mthi     $zero,                $ac1                            \n\t"
 288         "lh       %[Temp1],             14(%[input])                    \n\t"
 289         "lh       %[Temp0],             2(%[input])                     \n\t"
 290         "msub     $ac0,                 %[Temp1],       %[cospi_4_64]   \n\t"
 291         "extp     %[step1_4],           $ac0,           31              \n\t"
 292
 293         /*
 294           temp_2 = input[1] * cospi_4_64 + input[7] * cospi_28_64;
 295           step1_7 = dct_const_round_shift(temp_2);
 296         */
 297         "madd     $ac1,                 %[Temp0],       %[cospi_4_64]   \n\t"
 298         "madd     $ac1,                 %[Temp1],       %[cospi_28_64]  \n\t"
 299         "extp     %[step1_7],           $ac1,           31              \n\t"
 300
 301         /*
 302           temp_1 = input[5] * cospi_12_64 - input[3] * cospi_20_64;
 303           step1_5 = dct_const_round_shift(temp_1);
 304         */
 305         "mtlo     %[const_2_power_13],  $ac0                            \n\t"
 306         "mthi     $zero,                $ac0                            \n\t"
 307         "lh       %[Temp0],             10(%[input])                    \n\t"
 308         "madd     $ac0,                 %[Temp0],       %[cospi_12_64]  \n\t"
 309         "lh       %[Temp1],             6(%[input])                     \n\t"
 310         "msub     $ac0,                 %[Temp1],       %[cospi_20_64]  \n\t"
 311         "extp     %[step1_5],           $ac0,           31              \n\t"
 312
 313         /*
 314           temp_2 = input[5] * cospi_20_64 + input[3] * cospi_12_64;
 315           step1_6 = dct_const_round_shift(temp_2);
 316         */
 317         "mtlo     %[const_2_power_13],  $ac1                            \n\t"
 318         "mthi     $zero,                $ac1                            \n\t"
 319         "lh       %[Temp0],             10(%[input])                    \n\t"
 320         "madd     $ac1,                 %[Temp0],       %[cospi_20_64]  \n\t"
 321         "lh       %[Temp1],             6(%[input])                     \n\t"
 322         "madd     $ac1,                 %[Temp1],       %[cospi_12_64]  \n\t"
 323         "extp     %[step1_6],           $ac1,           31              \n\t"
 324
 325         /*
 326           temp_1 = (step1_7 - step1_6 - step1_4 + step1_5) * cospi_16_64;
 327           temp_2 = (step1_4 - step1_5 - step1_6 + step1_7) * cospi_16_64;
 328         */
 329         "sub      %[Temp0],             %[step1_7],     %[step1_6]      \n\t"
 330         "sub      %[Temp0],             %[Temp0],       %[step1_4]      \n\t"
 331         "add      %[Temp0],             %[Temp0],       %[step1_5]      \n\t"
 332         "sub      %[Temp1],             %[step1_4],     %[step1_5]      \n\t"
 333         "sub      %[Temp1],             %[Temp1],       %[step1_6]      \n\t"
 334         "add      %[Temp1],             %[Temp1],       %[step1_7]      \n\t"
 335
 336         "mtlo     %[const_2_power_13],  $ac0                            \n\t"
 337         "mthi     $zero,                $ac0                            \n\t"
 338         "mtlo     %[const_2_power_13],  $ac1                            \n\t"
 339         "mthi     $zero,                $ac1                            \n\t"
 340
 341         "madd     $ac0,                 %[Temp0],       %[cospi_16_64]  \n\t"
 342         "madd     $ac1,                 %[Temp1],       %[cospi_16_64]  \n\t"
 343
 344         /*
 345           step1_4 = step1_4 + step1_5;
 346           step1_7 = step1_6 + step1_7;
 347         */
 348         "add      %[step1_4],           %[step1_4],     %[step1_5]      \n\t"
 349         "add      %[step1_7],           %[step1_7],     %[step1_6]      \n\t"
 350
 351         "extp     %[step1_5],           $ac0,           31              \n\t"
 352         "extp     %[step1_6],           $ac1,           31              \n\t"
 353
 354         /* add block */
 355         "lbu      %[Temp1],             0(%[dest_pix])                  \n\t"
 356         "add      %[Temp0],             %[step1_0],     %[step1_7]      \n\t"
 357         "addi     %[Temp0],             %[Temp0],       16              \n\t"
 358         "sra      %[Temp0],             %[Temp0],       5               \n\t"
 359         "add      %[Temp1],             %[Temp1],       %[Temp0]        \n\t"
 360         "add      %[Temp0],             %[step1_1],     %[step1_6]      \n\t"
 361         "lbux     %[Temp2],             %[Temp1](%[cm])                 \n\t"
 362         "sb       %[Temp2],             0(%[dest_pix])                  \n\t"
 363         "addu     %[dest_pix],          %[dest_pix],    %[dest_stride]  \n\t"
 364
 365         "lbu      %[Temp1],             0(%[dest_pix])                  \n\t"
 366         "addi     %[Temp0],             %[Temp0],       16              \n\t"
 367         "sra      %[Temp0],             %[Temp0],       5               \n\t"
 368         "add      %[Temp1],             %[Temp1],       %[Temp0]        \n\t"
 369         "add      %[Temp0],             %[step1_2],     %[step1_5]      \n\t"
 370         "lbux     %[Temp2],             %[Temp1](%[cm])                 \n\t"
 371         "sb       %[Temp2],             0(%[dest_pix])                  \n\t"
 372         "addu     %[dest_pix],          %[dest_pix],    %[dest_stride]  \n\t"
 373
 374         "lbu      %[Temp1],             0(%[dest_pix])                  \n\t"
 375         "addi     %[Temp0],             %[Temp0],       16              \n\t"
 376         "sra      %[Temp0],             %[Temp0],       5               \n\t"
 377         "add      %[Temp1],             %[Temp1],       %[Temp0]        \n\t"
 378         "add      %[Temp0],             %[step1_3],     %[step1_4]      \n\t"
 379         "lbux     %[Temp2],             %[Temp1](%[cm])                 \n\t"
 380         "sb       %[Temp2],             0(%[dest_pix])                  \n\t"
 381         "addu     %[dest_pix],          %[dest_pix],    %[dest_stride]  \n\t"
 382
 383         "lbu      %[Temp1],             0(%[dest_pix])                  \n\t"
 384         "addi     %[Temp0],             %[Temp0],       16              \n\t"
 385         "sra      %[Temp0],             %[Temp0],       5               \n\t"
 386         "add      %[Temp1],             %[Temp1],       %[Temp0]        \n\t"
 387         "sub      %[Temp0],             %[step1_3],     %[step1_4]      \n\t"
 388         "lbux     %[Temp2],             %[Temp1](%[cm])                 \n\t"
 389         "sb       %[Temp2],             0(%[dest_pix])                  \n\t"
 390         "addu     %[dest_pix],          %[dest_pix],    %[dest_stride]  \n\t"
 391
 392         "lbu      %[Temp1],             0(%[dest_pix])                  \n\t"
 393         "addi     %[Temp0],             %[Temp0],       16              \n\t"
 394         "sra      %[Temp0],             %[Temp0],       5               \n\t"
 395         "add      %[Temp1],             %[Temp1],       %[Temp0]        \n\t"
 396         "sub      %[Temp0],             %[step1_2],     %[step1_5]      \n\t"
 397         "lbux     %[Temp2],             %[Temp1](%[cm])                 \n\t"
 398         "sb       %[Temp2],             0(%[dest_pix])                  \n\t"
 399         "addu     %[dest_pix],          %[dest_pix],    %[dest_stride]  \n\t"
 400
 401         "lbu      %[Temp1],             0(%[dest_pix])                  \n\t"
 402         "addi     %[Temp0],             %[Temp0],       16              \n\t"
 403         "sra      %[Temp0],             %[Temp0],       5               \n\t"
 404         "add      %[Temp1],             %[Temp1],       %[Temp0]        \n\t"
 405         "sub      %[Temp0],             %[step1_1],     %[step1_6]      \n\t"
 406         "lbux     %[Temp2],             %[Temp1](%[cm])                 \n\t"
 407         "sb       %[Temp2],             0(%[dest_pix])                  \n\t"
 408         "addu     %[dest_pix],          %[dest_pix],    %[dest_stride]  \n\t"
 409
 410         "lbu      %[Temp1],             0(%[dest_pix])                  \n\t"
 411         "addi     %[Temp0],             %[Temp0],       16              \n\t"
 412         "sra      %[Temp0],             %[Temp0],       5               \n\t"
 413         "add      %[Temp1],             %[Temp1],       %[Temp0]        \n\t"
 414         "sub      %[Temp0],             %[step1_0],     %[step1_7]      \n\t"
 415         "lbux     %[Temp2],             %[Temp1](%[cm])                 \n\t"
 416         "sb       %[Temp2],             0(%[dest_pix])                  \n\t"
 417         "addu     %[dest_pix],          %[dest_pix],    %[dest_stride]  \n\t"
 418
 419         "lbu      %[Temp1],             0(%[dest_pix])                  \n\t"
 420         "addi     %[Temp0],             %[Temp0],       16              \n\t"
 421         "sra      %[Temp0],             %[Temp0],       5               \n\t"
 422         "add      %[Temp1],             %[Temp1],       %[Temp0]        \n\t"
 423         "lbux     %[Temp2],             %[Temp1](%[cm])                 \n\t"
 424         "sb       %[Temp2],             0(%[dest_pix])                  \n\t"
 425
 426         : [step1_0] "=&r" (step1_0), [step1_1] "=&r" (step1_1),
 427           [step1_2] "=&r" (step1_2), [step1_3] "=&r" (step1_3),
 428           [step1_4] "=&r" (step1_4), [step1_5] "=&r" (step1_5),
 429           [step1_6] "=&r" (step1_6), [step1_7] "=&r" (step1_7),
 430           [Temp0] "=&r" (Temp0), [Temp1] "=&r" (Temp1),
 431           [Temp2] "=&r" (Temp2), [Temp3] "=&r" (Temp3),
 432           [dest_pix] "+r" (dest_pix)
 433         : [const_2_power_13] "r" (const_2_power_13),
 434           [cospi_16_64] "r" (cospi_16_64), [cospi_28_64] "r" (cospi_28_64),
 435           [cospi_4_64] "r" (cospi_4_64), [cospi_12_64] "r" (cospi_12_64),
 436           [cospi_20_64] "r" (cospi_20_64), [cospi_8_64] "r" (cospi_8_64),
 437           [cospi_24_64] "r" (cospi_24_64),
 438           [input] "r" (input), [cm] "r" (cm), [dest_stride] "r" (dest_stride)
 439     );
 440
 441     input += 8;
 442   }
 443 }
 444
 445 void vp9_idct8x8_64_add_dspr2(const int16_t *input, uint8_t *dest,
 446                               int dest_stride) {
 447   DECLARE_ALIGNED(32, int16_t, out[8 * 8]);
 448   int16_t *outptr = out;
 449   uint32_t pos = 45;
 450
 451   /* bit positon for extract from acc */
 452   __asm__ __volatile__ (
 453     "wrdsp    %[pos],    1    \n\t"
 454     :
 455     : [pos] "r" (pos)
 456   );
 457
 458   // First transform rows
 459   idct8_rows_dspr2(input, outptr, 8);
 460
 461   // Then transform columns and add to dest
 462   idct8_columns_add_blk_dspr2(&out[0], dest, dest_stride);
 463 }
 464
 465 void vp9_idct8x8_12_add_dspr2(const int16_t *input, uint8_t *dest,
 466                               int dest_stride) {
 467   DECLARE_ALIGNED(32, int16_t, out[8 * 8]);
 468   int16_t *outptr = out;
 469   uint32_t pos = 45;
 470
 471   /* bit positon for extract from acc */
 472   __asm__ __volatile__ (
 473     "wrdsp    %[pos],    1    \n\t"
 474     :
 475     : [pos] "r" (pos)
 476   );
 477
 478   // First transform rows
 479   idct8_rows_dspr2(input, outptr, 4);
 480
 481   outptr += 4;
 482
 483   __asm__ __volatile__ (
 484       "sw  $zero,   0(%[outptr])  \n\t"
 485       "sw  $zero,   4(%[outptr])  \n\t"
 486       "sw  $zero,  16(%[outptr])  \n\t"
 487       "sw  $zero,  20(%[outptr])  \n\t"
 488       "sw  $zero,  32(%[outptr])  \n\t"
 489       "sw  $zero,  36(%[outptr])  \n\t"
 490       "sw  $zero,  48(%[outptr])  \n\t"
 491       "sw  $zero,  52(%[outptr])  \n\t"
 492       "sw  $zero,  64(%[outptr])  \n\t"
 493       "sw  $zero,  68(%[outptr])  \n\t"
 494       "sw  $zero,  80(%[outptr])  \n\t"
 495       "sw  $zero,  84(%[outptr])  \n\t"
 496       "sw  $zero,  96(%[outptr])  \n\t"
 497       "sw  $zero, 100(%[outptr])  \n\t"
 498       "sw  $zero, 112(%[outptr])  \n\t"
 499       "sw  $zero, 116(%[outptr])  \n\t"
 500
 501       :
 502       : [outptr] "r" (outptr)
 503   );
 504
 505
 506   // Then transform columns and add to dest
 507   idct8_columns_add_blk_dspr2(&out[0], dest, dest_stride);
 508 }
 509
 510 void vp9_idct8x8_1_add_dspr2(const int16_t *input, uint8_t *dest,
 511                              int dest_stride) {
 512   uint32_t pos = 45;
 513   int32_t out;
 514   int32_t r;
 515   int32_t a1, absa1;
 516   int32_t t1, t2, vector_a1, vector_1, vector_2;
 517
 518   /* bit positon for extract from acc */
 519   __asm__ __volatile__ (
 520     "wrdsp      %[pos],     1           \n\t"
 521
 522     :
 523     : [pos] "r" (pos)
 524   );
 525
 526   out = DCT_CONST_ROUND_SHIFT_TWICE_COSPI_16_64(input[0]);
 527   __asm__ __volatile__ (
 528       "addi     %[out],     %[out],     16      \n\t"
 529       "sra      %[a1],      %[out],     5       \n\t"
 530
 531       : [out] "+r" (out), [a1] "=r" (a1)
 532       :
 533   );
 534
 535   if (a1 < 0) {
 536     /* use quad-byte
 537      * input and output memory are four byte aligned */
 538     __asm__ __volatile__ (
 539         "abs        %[absa1],       %[a1]       \n\t"
 540         "replv.qb   %[vector_a1],   %[absa1]    \n\t"
 541
 542         : [absa1] "=r" (absa1), [vector_a1] "=r" (vector_a1)
 543         : [a1] "r" (a1)
 544     );
 545
 546     for (r = 8; r--;) {
 547       __asm__ __volatile__ (
 548           "lw           %[t1],          0(%[dest])                      \n\t"
 549           "lw           %[t2],          4(%[dest])                      \n\t"
 550           "subu_s.qb    %[vector_1],    %[t1],          %[vector_a1]    \n\t"
 551           "subu_s.qb    %[vector_2],    %[t2],          %[vector_a1]    \n\t"
 552           "sw           %[vector_1],    0(%[dest])                      \n\t"
 553           "sw           %[vector_2],    4(%[dest])                      \n\t"
 554           "add          %[dest],        %[dest],        %[dest_stride]  \n\t"
 555
 556           : [t1] "=&r" (t1), [t2] "=&r" (t2),
 557             [vector_1] "=&r" (vector_1), [vector_2] "=&r" (vector_2),
 558             [dest] "+&r" (dest)
 559           : [dest_stride] "r" (dest_stride), [vector_a1] "r" (vector_a1)
 560       );
 561     }
 562   } else {
 563     /* use quad-byte
 564      * input and output memory are four byte aligned */
 565     __asm__ __volatile__ (
 566         "replv.qb   %[vector_a1],   %[a1]   \n\t"
 567
 568         : [vector_a1] "=r" (vector_a1)
 569         : [a1] "r" (a1)
 570     );
 571
 572     for (r = 8; r--;) {
 573       __asm__ __volatile__ (
 574           "lw           %[t1],          0(%[dest])                      \n\t"
 575           "lw           %[t2],          4(%[dest])                      \n\t"
 576           "addu_s.qb    %[vector_1],    %[t1],          %[vector_a1]    \n\t"
 577           "addu_s.qb    %[vector_2],    %[t2],          %[vector_a1]    \n\t"
 578           "sw           %[vector_1],    0(%[dest])                      \n\t"
 579           "sw           %[vector_2],    4(%[dest])                      \n\t"
 580           "add          %[dest],        %[dest],        %[dest_stride]  \n\t"
 581
 582           : [t1] "=&r" (t1), [t2] "=&r" (t2),
 583             [vector_1] "=&r" (vector_1), [vector_2] "=&r" (vector_2),
 584             [dest] "+r" (dest)
 585           : [dest_stride] "r" (dest_stride), [vector_a1] "r" (vector_a1)
 586       );
 587     }
 588   }
 589 }
 590
 591 void iadst8_dspr2(const int16_t *input, int16_t *output) {
 592   int s0, s1, s2, s3, s4, s5, s6, s7;
 593   int x0, x1, x2, x3, x4, x5, x6, x7;
 594
 595   x0 = input[7];
 596   x1 = input[0];
 597   x2 = input[5];
 598   x3 = input[2];
 599   x4 = input[3];
 600   x5 = input[4];
 601   x6 = input[1];
 602   x7 = input[6];
 603
 604   if (!(x0 | x1 | x2 | x3 | x4 | x5 | x6 | x7)) {
 605     output[0] = output[1] = output[2] = output[3] = output[4]
 606               = output[5] = output[6] = output[7] = 0;
 607     return;
 608   }
 609
 610   // stage 1
 611   s0 = cospi_2_64  * x0 + cospi_30_64 * x1;
 612   s1 = cospi_30_64 * x0 - cospi_2_64  * x1;
 613   s2 = cospi_10_64 * x2 + cospi_22_64 * x3;
 614   s3 = cospi_22_64 * x2 - cospi_10_64 * x3;
 615   s4 = cospi_18_64 * x4 + cospi_14_64 * x5;
 616   s5 = cospi_14_64 * x4 - cospi_18_64 * x5;
 617   s6 = cospi_26_64 * x6 + cospi_6_64  * x7;
 618   s7 = cospi_6_64  * x6 - cospi_26_64 * x7;
 619
 620   x0 = ROUND_POWER_OF_TWO((s0 + s4), DCT_CONST_BITS);
 621   x1 = ROUND_POWER_OF_TWO((s1 + s5), DCT_CONST_BITS);
 622   x2 = ROUND_POWER_OF_TWO((s2 + s6), DCT_CONST_BITS);
 623   x3 = ROUND_POWER_OF_TWO((s3 + s7), DCT_CONST_BITS);
 624   x4 = ROUND_POWER_OF_TWO((s0 - s4), DCT_CONST_BITS);
 625   x5 = ROUND_POWER_OF_TWO((s1 - s5), DCT_CONST_BITS);
 626   x6 = ROUND_POWER_OF_TWO((s2 - s6), DCT_CONST_BITS);
 627   x7 = ROUND_POWER_OF_TWO((s3 - s7), DCT_CONST_BITS);
 628
 629   // stage 2
 630   s0 = x0;
 631   s1 = x1;
 632   s2 = x2;
 633   s3 = x3;
 634   s4 =  cospi_8_64  * x4 + cospi_24_64 * x5;
 635   s5 =  cospi_24_64 * x4 - cospi_8_64  * x5;
 636   s6 = -cospi_24_64 * x6 + cospi_8_64  * x7;
 637   s7 =  cospi_8_64  * x6 + cospi_24_64 * x7;
 638
 639   x0 = s0 + s2;
 640   x1 = s1 + s3;
 641   x2 = s0 - s2;
 642   x3 = s1 - s3;
 643   x4 = ROUND_POWER_OF_TWO((s4 + s6), DCT_CONST_BITS);
 644   x5 = ROUND_POWER_OF_TWO((s5 + s7), DCT_CONST_BITS);
 645   x6 = ROUND_POWER_OF_TWO((s4 - s6), DCT_CONST_BITS);
 646   x7 = ROUND_POWER_OF_TWO((s5 - s7), DCT_CONST_BITS);
 647
 648   // stage 3
 649   s2 = cospi_16_64 * (x2 + x3);
 650   s3 = cospi_16_64 * (x2 - x3);
 651   s6 = cospi_16_64 * (x6 + x7);
 652   s7 = cospi_16_64 * (x6 - x7);
 653
 654   x2 = ROUND_POWER_OF_TWO((s2), DCT_CONST_BITS);
 655   x3 = ROUND_POWER_OF_TWO((s3), DCT_CONST_BITS);
 656   x6 = ROUND_POWER_OF_TWO((s6), DCT_CONST_BITS);
 657   x7 = ROUND_POWER_OF_TWO((s7), DCT_CONST_BITS);
 658
 659   output[0] =  x0;
 660   output[1] = -x4;
 661   output[2] =  x6;
 662   output[3] = -x2;
 663   output[4] =  x3;
 664   output[5] = -x7;
 665   output[6] =  x5;
 666   output[7] = -x1;
 667 }
 668 #endif  // HAVE_DSPR2