src/third_party/libvpx/source/libvpx/vp9/common/mips/dspr2/vp9_itrans32_cols_dspr2.c

   1 /*
   2  *  Copyright (c) 2013 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 <assert.h>
  12
  13 #include "./vpx_config.h"
  14 #include "./vp9_rtcd.h"
  15 #include "vp9/common/vp9_common.h"
  16 #include "vp9/common/vp9_blockd.h"
  17 #include "vp9/common/vp9_idct.h"
  18 #include "vp9/common/mips/dspr2/vp9_common_dspr2.h"
  19
  20 #if HAVE_DSPR2
  21 void vp9_idct32_cols_add_blk_dspr2(int16_t *input, uint8_t *dest,
  22                                    int dest_stride) {
  23   int16_t step1_0, step1_1, step1_2, step1_3, step1_4, step1_5, step1_6;
  24   int16_t step1_7, step1_8, step1_9, step1_10, step1_11, step1_12, step1_13;
  25   int16_t step1_14, step1_15, step1_16, step1_17, step1_18, step1_19;
  26   int16_t step1_20, step1_21, step1_22, step1_23, step1_24, step1_25, step1_26;
  27   int16_t step1_27, step1_28, step1_29, step1_30, step1_31;
  28   int16_t step2_0, step2_1, step2_2, step2_3, step2_4, step2_5, step2_6;
  29   int16_t step2_7, step2_8, step2_9, step2_10, step2_11, step2_12, step2_13;
  30   int16_t step2_14, step2_15, step2_16, step2_17, step2_18, step2_19, step2_20;
  31   int16_t step2_21, step2_22, step2_23, step2_24, step2_25, step2_26, step2_27;
  32   int16_t step2_28, step2_29, step2_30, step2_31;
  33   int16_t step3_8, step3_9, step3_10, step3_11, step3_12, step3_13, step3_14;
  34   int16_t step3_15, step3_16, step3_17, step3_18, step3_19, step3_20, step3_21;
  35   int16_t step3_22, step3_23, step3_24, step3_25, step3_26, step3_27;
  36   int16_t step3_28, step3_29, step3_30, step3_31;
  37   int temp0, temp1, temp2, temp3;
  38   int load1, load2, load3, load4;
  39   int result1, result2;
  40   int i, temp21;
  41   uint8_t *dest_pix, *dest_pix1;
  42   const int const_2_power_13 = 8192;
  43   uint8_t *cm = vp9_ff_cropTbl;
  44
  45   /* prefetch vp9_ff_cropTbl */
  46   vp9_prefetch_load(vp9_ff_cropTbl);
  47   vp9_prefetch_load(vp9_ff_cropTbl +  32);
  48   vp9_prefetch_load(vp9_ff_cropTbl +  64);
  49   vp9_prefetch_load(vp9_ff_cropTbl +  96);
  50   vp9_prefetch_load(vp9_ff_cropTbl + 128);
  51   vp9_prefetch_load(vp9_ff_cropTbl + 160);
  52   vp9_prefetch_load(vp9_ff_cropTbl + 192);
  53   vp9_prefetch_load(vp9_ff_cropTbl + 224);
  54
  55   for (i = 0; i < 32; ++i) {
  56     dest_pix = dest + i;
  57     dest_pix1 = dest + i + 31 * dest_stride;
  58
  59     __asm__ __volatile__ (
  60         "lh       %[load1],             2(%[input])                     \n\t"
  61         "lh       %[load2],             62(%[input])                    \n\t"
  62         "lh       %[load3],             34(%[input])                    \n\t"
  63         "lh       %[load4],             30(%[input])                    \n\t"
  64
  65         "mtlo     %[const_2_power_13],  $ac1                            \n\t"
  66         "mthi     $zero,                $ac1                            \n\t"
  67         "mtlo     %[const_2_power_13],  $ac3                            \n\t"
  68         "mthi     $zero,                $ac3                            \n\t"
  69
  70         "madd     $ac1,                 %[load1],       %[cospi_31_64]  \n\t"
  71         "msub     $ac1,                 %[load2],       %[cospi_1_64]   \n\t"
  72         "extp     %[temp0],             $ac1,           31              \n\t"
  73
  74         "madd     $ac3,                 %[load1],       %[cospi_1_64]   \n\t"
  75         "madd     $ac3,                 %[load2],       %[cospi_31_64]  \n\t"
  76         "extp     %[temp3],             $ac3,           31              \n\t"
  77
  78         "mtlo     %[const_2_power_13],  $ac1                            \n\t"
  79         "mthi     $zero,                $ac1                            \n\t"
  80         "mtlo     %[const_2_power_13],  $ac2                            \n\t"
  81         "mthi     $zero,                $ac2                            \n\t"
  82
  83         "madd     $ac2,                 %[load3],       %[cospi_15_64]  \n\t"
  84         "msub     $ac2,                 %[load4],       %[cospi_17_64]  \n\t"
  85         "extp     %[temp1],             $ac2,           31              \n\t"
  86
  87         "madd     $ac1,                 %[load3],       %[cospi_17_64]  \n\t"
  88         "madd     $ac1,                 %[load4],       %[cospi_15_64]  \n\t"
  89         "extp     %[temp2],             $ac1,           31              \n\t"
  90
  91         "mtlo     %[const_2_power_13],  $ac1                            \n\t"
  92         "mthi     $zero,                $ac1                            \n\t"
  93         "mtlo     %[const_2_power_13],  $ac3                            \n\t"
  94         "mthi     $zero,                $ac3                            \n\t"
  95
  96         "sub      %[load1],             %[temp3],       %[temp2]        \n\t"
  97         "sub      %[load2],             %[temp0],       %[temp1]        \n\t"
  98
  99         "madd     $ac1,                 %[load1],       %[cospi_28_64]  \n\t"
 100         "msub     $ac1,                 %[load2],       %[cospi_4_64]   \n\t"
 101         "madd     $ac3,                 %[load1],       %[cospi_4_64]   \n\t"
 102         "madd     $ac3,                 %[load2],       %[cospi_28_64]  \n\t"
 103
 104         "extp     %[step1_17],          $ac1,           31              \n\t"
 105         "extp     %[step1_30],          $ac3,           31              \n\t"
 106         "add      %[step1_16],          %[temp0],       %[temp1]        \n\t"
 107         "add      %[step1_31],          %[temp2],       %[temp3]        \n\t"
 108
 109         : [load1] "=&r" (load1), [load2] "=&r" (load2), [load3] "=&r" (load3),
 110           [load4] "=&r" (load4), [temp0] "=&r" (temp0), [temp1] "=&r" (temp1),
 111           [temp2] "=&r" (temp2), [temp3] "=&r" (temp3),
 112           [step1_16] "=r" (step1_16), [step1_17] "=r" (step1_17),
 113           [step1_30] "=r" (step1_30), [step1_31] "=r" (step1_31)
 114         : [const_2_power_13] "r" (const_2_power_13), [input] "r" (input),
 115           [cospi_31_64] "r" (cospi_31_64), [cospi_1_64] "r" (cospi_1_64),
 116           [cospi_4_64] "r" (cospi_4_64), [cospi_17_64] "r" (cospi_17_64),
 117           [cospi_15_64] "r" (cospi_15_64), [cospi_28_64] "r" (cospi_28_64)
 118     );
 119
 120     __asm__ __volatile__ (
 121         "lh       %[load1],             18(%[input])                    \n\t"
 122         "lh       %[load2],             46(%[input])                    \n\t"
 123         "lh       %[load3],             50(%[input])                    \n\t"
 124         "lh       %[load4],             14(%[input])                    \n\t"
 125
 126         "mtlo     %[const_2_power_13],  $ac1                            \n\t"
 127         "mthi     $zero,                $ac1                            \n\t"
 128         "mtlo     %[const_2_power_13],  $ac3                            \n\t"
 129         "mthi     $zero,                $ac3                            \n\t"
 130
 131         "madd     $ac1,                 %[load1],       %[cospi_23_64]  \n\t"
 132         "msub     $ac1,                 %[load2],       %[cospi_9_64]   \n\t"
 133         "extp     %[temp0],             $ac1,           31              \n\t"
 134
 135         "madd     $ac3,                 %[load1],       %[cospi_9_64]   \n\t"
 136         "madd     $ac3,                 %[load2],       %[cospi_23_64]  \n\t"
 137         "extp     %[temp3],             $ac3,           31              \n\t"
 138
 139         "mtlo     %[const_2_power_13],  $ac1                            \n\t"
 140         "mthi     $zero,                $ac1                            \n\t"
 141         "mtlo     %[const_2_power_13],  $ac2                            \n\t"
 142         "mthi     $zero,                $ac2                            \n\t"
 143
 144         "madd     $ac2,                 %[load3],       %[cospi_7_64]   \n\t"
 145         "msub     $ac2,                 %[load4],       %[cospi_25_64]  \n\t"
 146         "extp     %[temp1],             $ac2,           31              \n\t"
 147
 148         "madd     $ac1,                 %[load3],       %[cospi_25_64]  \n\t"
 149         "madd     $ac1,                 %[load4],       %[cospi_7_64]   \n\t"
 150         "extp     %[temp2],             $ac1,           31              \n\t"
 151
 152         "mtlo     %[const_2_power_13],  $ac1                            \n\t"
 153         "mthi     $zero,                $ac1                            \n\t"
 154         "mtlo     %[const_2_power_13],  $ac3                            \n\t"
 155         "mthi     $zero,                $ac3                            \n\t"
 156
 157         "sub      %[load1],             %[temp1],       %[temp0]        \n\t"
 158         "sub      %[load2],             %[temp2],       %[temp3]        \n\t"
 159
 160         "msub     $ac1,                 %[load1],       %[cospi_28_64]  \n\t"
 161         "msub     $ac1,                 %[load2],       %[cospi_4_64]   \n\t"
 162         "msub     $ac3,                 %[load1],       %[cospi_4_64]   \n\t"
 163         "madd     $ac3,                 %[load2],       %[cospi_28_64]  \n\t"
 164
 165         "extp     %[step1_18],          $ac1,           31              \n\t"
 166         "extp     %[step1_29],          $ac3,           31              \n\t"
 167         "add      %[step1_19],          %[temp0],       %[temp1]        \n\t"
 168         "add      %[step1_28],          %[temp2],       %[temp3]        \n\t"
 169
 170         : [load1] "=&r" (load1), [load2] "=&r" (load2), [load3] "=&r" (load3),
 171           [load4] "=&r" (load4), [temp0] "=&r" (temp0), [temp1] "=&r" (temp1),
 172           [temp2] "=&r" (temp2), [temp3] "=&r" (temp3),
 173           [step1_18] "=r" (step1_18), [step1_19] "=r" (step1_19),
 174           [step1_28] "=r" (step1_28), [step1_29] "=r" (step1_29)
 175         : [const_2_power_13] "r" (const_2_power_13), [input] "r" (input),
 176           [cospi_23_64] "r" (cospi_23_64), [cospi_9_64] "r" (cospi_9_64),
 177           [cospi_4_64] "r" (cospi_4_64), [cospi_7_64] "r" (cospi_7_64),
 178           [cospi_25_64] "r" (cospi_25_64), [cospi_28_64] "r" (cospi_28_64)
 179     );
 180
 181     __asm__ __volatile__ (
 182         "lh       %[load1],             10(%[input])                    \n\t"
 183         "lh       %[load2],             54(%[input])                    \n\t"
 184         "lh       %[load3],             42(%[input])                    \n\t"
 185         "lh       %[load4],             22(%[input])                    \n\t"
 186
 187         "mtlo     %[const_2_power_13],  $ac1                            \n\t"
 188         "mthi     $zero,                $ac1                            \n\t"
 189         "mtlo     %[const_2_power_13],  $ac3                            \n\t"
 190         "mthi     $zero,                $ac3                            \n\t"
 191
 192         "madd     $ac1,                 %[load1],       %[cospi_27_64]  \n\t"
 193         "msub     $ac1,                 %[load2],       %[cospi_5_64]   \n\t"
 194         "extp     %[temp0],             $ac1,           31              \n\t"
 195
 196         "madd     $ac3,                 %[load1],       %[cospi_5_64]   \n\t"
 197         "madd     $ac3,                 %[load2],       %[cospi_27_64]  \n\t"
 198         "extp     %[temp3],             $ac3,           31              \n\t"
 199
 200         "mtlo     %[const_2_power_13],  $ac1                            \n\t"
 201         "mthi     $zero,                $ac1                            \n\t"
 202         "mtlo     %[const_2_power_13],  $ac2                            \n\t"
 203         "mthi     $zero,                $ac2                            \n\t"
 204
 205         "madd     $ac2,                 %[load3],       %[cospi_11_64]  \n\t"
 206         "msub     $ac2,                 %[load4],       %[cospi_21_64]  \n\t"
 207         "extp     %[temp1],             $ac2,           31              \n\t"
 208
 209         "madd     $ac1,                 %[load3],       %[cospi_21_64]  \n\t"
 210         "madd     $ac1,                 %[load4],       %[cospi_11_64]  \n\t"
 211         "extp     %[temp2],             $ac1,           31              \n\t"
 212
 213         "mtlo     %[const_2_power_13],  $ac1                            \n\t"
 214         "mthi     $zero,                $ac1                            \n\t"
 215         "mtlo     %[const_2_power_13],  $ac3                            \n\t"
 216         "mthi     $zero,                $ac3                            \n\t"
 217
 218         "sub      %[load1],             %[temp0],       %[temp1]        \n\t"
 219         "sub      %[load2],             %[temp3],       %[temp2]        \n\t"
 220
 221         "madd     $ac1,                 %[load2],       %[cospi_12_64]  \n\t"
 222         "msub     $ac1,                 %[load1],       %[cospi_20_64]  \n\t"
 223         "madd     $ac3,                 %[load1],       %[cospi_12_64]  \n\t"
 224         "madd     $ac3,                 %[load2],       %[cospi_20_64]  \n\t"
 225
 226         "extp     %[step1_21],          $ac1,           31              \n\t"
 227         "extp     %[step1_26],          $ac3,           31              \n\t"
 228         "add      %[step1_20],          %[temp0],       %[temp1]        \n\t"
 229         "add      %[step1_27],          %[temp2],       %[temp3]        \n\t"
 230
 231         : [load1] "=&r" (load1), [load2] "=&r" (load2), [load3] "=&r" (load3),
 232           [load4] "=&r" (load4), [temp0] "=&r" (temp0), [temp1] "=&r" (temp1),
 233           [temp2] "=&r" (temp2), [temp3] "=&r" (temp3),
 234           [step1_20] "=r" (step1_20), [step1_21] "=r" (step1_21),
 235           [step1_26] "=r" (step1_26), [step1_27] "=r" (step1_27)
 236         : [const_2_power_13] "r" (const_2_power_13), [input] "r" (input),
 237           [cospi_27_64] "r" (cospi_27_64), [cospi_5_64] "r" (cospi_5_64),
 238           [cospi_11_64] "r" (cospi_11_64), [cospi_21_64] "r" (cospi_21_64),
 239           [cospi_12_64] "r" (cospi_12_64), [cospi_20_64] "r" (cospi_20_64)
 240     );
 241
 242     __asm__ __volatile__ (
 243         "lh       %[load1],             26(%[input])                    \n\t"
 244         "lh       %[load2],             38(%[input])                    \n\t"
 245         "lh       %[load3],             58(%[input])                    \n\t"
 246         "lh       %[load4],              6(%[input])                    \n\t"
 247
 248         "mtlo     %[const_2_power_13],  $ac1                            \n\t"
 249         "mthi     $zero,                $ac1                            \n\t"
 250         "mtlo     %[const_2_power_13],  $ac3                            \n\t"
 251         "mthi     $zero,                $ac3                            \n\t"
 252
 253         "madd     $ac1,                 %[load1],       %[cospi_19_64]  \n\t"
 254         "msub     $ac1,                 %[load2],       %[cospi_13_64]  \n\t"
 255         "extp     %[temp0],             $ac1,           31              \n\t"
 256         "madd     $ac3,                 %[load1],       %[cospi_13_64]  \n\t"
 257         "madd     $ac3,                 %[load2],       %[cospi_19_64]  \n\t"
 258         "extp     %[temp3],             $ac3,           31              \n\t"
 259
 260         "mtlo     %[const_2_power_13],  $ac1                            \n\t"
 261         "mthi     $zero,                $ac1                            \n\t"
 262         "mtlo     %[const_2_power_13],  $ac2                            \n\t"
 263         "mthi     $zero,                $ac2                            \n\t"
 264
 265         "madd     $ac2,                 %[load3],       %[cospi_3_64]   \n\t"
 266         "msub     $ac2,                 %[load4],       %[cospi_29_64]  \n\t"
 267         "extp     %[temp1],             $ac2,           31              \n\t"
 268         "madd     $ac1,                 %[load3],       %[cospi_29_64]  \n\t"
 269         "madd     $ac1,                 %[load4],       %[cospi_3_64]   \n\t"
 270         "extp     %[temp2],             $ac1,           31              \n\t"
 271
 272         "mtlo     %[const_2_power_13],  $ac1                            \n\t"
 273         "mthi     $zero,                $ac1                            \n\t"
 274         "mtlo     %[const_2_power_13],  $ac3                            \n\t"
 275         "mthi     $zero,                $ac3                            \n\t"
 276
 277         "sub      %[load1],             %[temp1],       %[temp0]        \n\t"
 278         "sub      %[load2],             %[temp2],       %[temp3]        \n\t"
 279         "msub     $ac1,                 %[load1],       %[cospi_12_64]  \n\t"
 280         "msub     $ac1,                 %[load2],       %[cospi_20_64]  \n\t"
 281         "msub     $ac3,                 %[load1],       %[cospi_20_64]  \n\t"
 282         "madd     $ac3,                 %[load2],       %[cospi_12_64]  \n\t"
 283         "extp     %[step1_22],          $ac1,           31              \n\t"
 284         "extp     %[step1_25],          $ac3,           31              \n\t"
 285         "add      %[step1_23],          %[temp0],       %[temp1]        \n\t"
 286         "add      %[step1_24],          %[temp2],       %[temp3]        \n\t"
 287
 288         : [load1] "=&r" (load1), [load2] "=&r" (load2), [load3] "=&r" (load3),
 289           [load4] "=&r" (load4), [temp0] "=&r" (temp0), [temp1] "=&r" (temp1),
 290           [temp2] "=&r" (temp2), [temp3] "=&r" (temp3),
 291           [step1_22] "=r" (step1_22), [step1_23] "=r" (step1_23),
 292           [step1_24] "=r" (step1_24), [step1_25] "=r" (step1_25)
 293         : [const_2_power_13] "r" (const_2_power_13), [input] "r" (input),
 294           [cospi_19_64] "r" (cospi_19_64), [cospi_13_64] "r" (cospi_13_64),
 295           [cospi_3_64] "r" (cospi_3_64), [cospi_29_64] "r" (cospi_29_64),
 296           [cospi_12_64] "r" (cospi_12_64), [cospi_20_64] "r" (cospi_20_64)
 297     );
 298
 299     __asm__ __volatile__ (
 300         "lh       %[load1],              4(%[input])                    \n\t"
 301         "lh       %[load2],             60(%[input])                    \n\t"
 302         "lh       %[load3],             36(%[input])                    \n\t"
 303         "lh       %[load4],             28(%[input])                    \n\t"
 304
 305         "mtlo     %[const_2_power_13],  $ac1                            \n\t"
 306         "mthi     $zero,                $ac1                            \n\t"
 307         "mtlo     %[const_2_power_13],  $ac3                            \n\t"
 308         "mthi     $zero,                $ac3                            \n\t"
 309
 310         "madd     $ac1,                 %[load1],       %[cospi_30_64]  \n\t"
 311         "msub     $ac1,                 %[load2],       %[cospi_2_64]   \n\t"
 312         "extp     %[temp0],             $ac1,           31              \n\t"
 313         "madd     $ac3,                 %[load1],       %[cospi_2_64]   \n\t"
 314         "madd     $ac3,                 %[load2],       %[cospi_30_64]  \n\t"
 315         "extp     %[temp3],             $ac3,           31              \n\t"
 316
 317         "mtlo     %[const_2_power_13],  $ac1                            \n\t"
 318         "mthi     $zero,                $ac1                            \n\t"
 319         "mtlo     %[const_2_power_13],  $ac2                            \n\t"
 320         "mthi     $zero,                $ac2                            \n\t"
 321
 322         "madd     $ac2,                 %[load3],       %[cospi_14_64]  \n\t"
 323         "msub     $ac2,                 %[load4],       %[cospi_18_64]  \n\t"
 324         "extp     %[temp1],             $ac2,           31              \n\t"
 325         "madd     $ac1,                 %[load3],       %[cospi_18_64]  \n\t"
 326         "madd     $ac1,                 %[load4],       %[cospi_14_64]  \n\t"
 327         "extp     %[temp2],             $ac1,           31              \n\t"
 328
 329         "mtlo     %[const_2_power_13],  $ac1                            \n\t"
 330         "mthi     $zero,                $ac1                            \n\t"
 331         "mtlo     %[const_2_power_13],  $ac3                            \n\t"
 332         "mthi     $zero,                $ac3                            \n\t"
 333
 334         "sub      %[load1],             %[temp0],       %[temp1]        \n\t"
 335         "sub      %[load2],             %[temp3],       %[temp2]        \n\t"
 336         "msub     $ac1,                 %[load1],       %[cospi_8_64]   \n\t"
 337         "madd     $ac1,                 %[load2],       %[cospi_24_64]  \n\t"
 338         "madd     $ac3,                 %[load1],       %[cospi_24_64]  \n\t"
 339         "madd     $ac3,                 %[load2],       %[cospi_8_64]   \n\t"
 340         "extp     %[step2_9],           $ac1,           31              \n\t"
 341         "extp     %[step2_14],          $ac3,           31              \n\t"
 342         "add      %[step2_8],           %[temp0],       %[temp1]        \n\t"
 343         "add      %[step2_15],          %[temp2],       %[temp3]        \n\t"
 344
 345         : [load1] "=&r" (load1), [load2] "=&r" (load2), [load3] "=&r" (load3),
 346           [load4] "=&r" (load4), [temp0] "=&r" (temp0), [temp1] "=&r" (temp1),
 347           [temp2] "=&r" (temp2), [temp3] "=&r" (temp3),
 348           [step2_8] "=r" (step2_8), [step2_9] "=r" (step2_9),
 349           [step2_14] "=r" (step2_14), [step2_15] "=r" (step2_15)
 350         : [const_2_power_13] "r" (const_2_power_13), [input] "r" (input),
 351           [cospi_30_64] "r" (cospi_30_64), [cospi_2_64] "r" (cospi_2_64),
 352           [cospi_14_64] "r" (cospi_14_64), [cospi_18_64] "r" (cospi_18_64),
 353           [cospi_8_64] "r" (cospi_8_64), [cospi_24_64] "r" (cospi_24_64)
 354     );
 355
 356     __asm__ __volatile__ (
 357         "lh       %[load1],             20(%[input])                    \n\t"
 358         "lh       %[load2],             44(%[input])                    \n\t"
 359         "lh       %[load3],             52(%[input])                    \n\t"
 360         "lh       %[load4],             12(%[input])                    \n\t"
 361
 362         "mtlo     %[const_2_power_13],  $ac1                            \n\t"
 363         "mthi     $zero,                $ac1                            \n\t"
 364         "mtlo     %[const_2_power_13],  $ac3                            \n\t"
 365         "mthi     $zero,                $ac3                            \n\t"
 366
 367         "madd     $ac1,                 %[load1],       %[cospi_22_64]  \n\t"
 368         "msub     $ac1,                 %[load2],       %[cospi_10_64]  \n\t"
 369         "extp     %[temp0],             $ac1,           31              \n\t"
 370         "madd     $ac3,                 %[load1],       %[cospi_10_64]  \n\t"
 371         "madd     $ac3,                 %[load2],       %[cospi_22_64]  \n\t"
 372         "extp     %[temp3],             $ac3,           31              \n\t"
 373
 374         "mtlo     %[const_2_power_13],  $ac1                            \n\t"
 375         "mthi     $zero,                $ac1                            \n\t"
 376         "mtlo     %[const_2_power_13],  $ac2                            \n\t"
 377         "mthi     $zero,                $ac2                            \n\t"
 378
 379         "madd     $ac2,                 %[load3],       %[cospi_6_64]   \n\t"
 380         "msub     $ac2,                 %[load4],       %[cospi_26_64]  \n\t"
 381         "extp     %[temp1],             $ac2,           31              \n\t"
 382         "madd     $ac1,                 %[load3],       %[cospi_26_64]  \n\t"
 383         "madd     $ac1,                 %[load4],       %[cospi_6_64]   \n\t"
 384         "extp     %[temp2],             $ac1,           31              \n\t"
 385
 386         "mtlo     %[const_2_power_13],  $ac1                            \n\t"
 387         "mthi     $zero,                $ac1                            \n\t"
 388         "mtlo     %[const_2_power_13],  $ac3                            \n\t"
 389         "mthi     $zero,                $ac3                            \n\t"
 390
 391         "sub      %[load1],             %[temp1],       %[temp0]        \n\t"
 392         "sub      %[load2],             %[temp2],       %[temp3]        \n\t"
 393         "msub     $ac1,                 %[load1],       %[cospi_24_64]  \n\t"
 394         "msub     $ac1,                 %[load2],       %[cospi_8_64]   \n\t"
 395         "madd     $ac3,                 %[load2],       %[cospi_24_64]  \n\t"
 396         "msub     $ac3,                 %[load1],       %[cospi_8_64]   \n\t"
 397         "extp     %[step2_10],          $ac1,           31              \n\t"
 398         "extp     %[step2_13],          $ac3,           31              \n\t"
 399         "add      %[step2_11],          %[temp0],       %[temp1]        \n\t"
 400         "add      %[step2_12],          %[temp2],       %[temp3]        \n\t"
 401
 402         : [load1] "=&r" (load1), [load2] "=&r" (load2), [load3] "=&r" (load3),
 403           [load4] "=&r" (load4), [temp0] "=&r" (temp0), [temp1] "=&r" (temp1),
 404           [temp2] "=&r" (temp2), [temp3] "=&r" (temp3),
 405           [step2_10] "=r" (step2_10), [step2_11] "=r" (step2_11),
 406           [step2_12] "=r" (step2_12), [step2_13] "=r" (step2_13)
 407         : [const_2_power_13] "r" (const_2_power_13), [input] "r" (input),
 408           [cospi_22_64] "r" (cospi_22_64), [cospi_10_64] "r" (cospi_10_64),
 409           [cospi_6_64] "r" (cospi_6_64), [cospi_26_64] "r" (cospi_26_64),
 410           [cospi_8_64] "r" (cospi_8_64), [cospi_24_64] "r" (cospi_24_64)
 411     );
 412
 413     __asm__ __volatile__ (
 414         "mtlo     %[const_2_power_13],  $ac0                            \n\t"
 415         "mthi     $zero,                $ac0                            \n\t"
 416         "sub      %[temp0],             %[step2_14],    %[step2_13]     \n\t"
 417         "sub      %[temp0],             %[temp0],       %[step2_9]      \n\t"
 418         "add      %[temp0],             %[temp0],       %[step2_10]     \n\t"
 419         "madd     $ac0,                 %[temp0],       %[cospi_16_64]  \n\t"
 420         "mtlo     %[const_2_power_13],  $ac1                            \n\t"
 421         "mthi     $zero,                $ac1                            \n\t"
 422         "sub      %[temp1],             %[step2_14],    %[step2_13]     \n\t"
 423         "add      %[temp1],             %[temp1],       %[step2_9]      \n\t"
 424         "sub      %[temp1],             %[temp1],       %[step2_10]     \n\t"
 425         "madd     $ac1,                 %[temp1],       %[cospi_16_64]  \n\t"
 426         "mtlo     %[const_2_power_13],  $ac2                            \n\t"
 427         "mthi     $zero,                $ac2                            \n\t"
 428         "sub      %[temp0],             %[step2_15],    %[step2_12]     \n\t"
 429         "sub      %[temp0],             %[temp0],       %[step2_8]      \n\t"
 430         "add      %[temp0],             %[temp0],       %[step2_11]     \n\t"
 431         "madd     $ac2,                 %[temp0],       %[cospi_16_64]  \n\t"
 432         "mtlo     %[const_2_power_13],  $ac3                            \n\t"
 433         "mthi     $zero,                $ac3                            \n\t"
 434         "sub      %[temp1],             %[step2_15],    %[step2_12]     \n\t"
 435         "add      %[temp1],             %[temp1],       %[step2_8]      \n\t"
 436         "sub      %[temp1],             %[temp1],       %[step2_11]     \n\t"
 437         "madd     $ac3,                 %[temp1],       %[cospi_16_64]  \n\t"
 438
 439         "add      %[step3_8],           %[step2_8],     %[step2_11]     \n\t"
 440         "add      %[step3_9],           %[step2_9],     %[step2_10]     \n\t"
 441         "add      %[step3_14],          %[step2_13],    %[step2_14]     \n\t"
 442         "add      %[step3_15],          %[step2_12],    %[step2_15]     \n\t"
 443         "extp     %[step3_10],          $ac0,           31              \n\t"
 444         "extp     %[step3_13],          $ac1,           31              \n\t"
 445         "extp     %[step3_11],          $ac2,           31              \n\t"
 446         "extp     %[step3_12],          $ac3,           31              \n\t"
 447
 448         : [temp0] "=&r" (temp0), [temp1] "=&r" (temp1),
 449           [step3_8] "=r" (step3_8), [step3_9] "=r" (step3_9),
 450           [step3_10] "=r" (step3_10), [step3_11] "=r" (step3_11),
 451           [step3_12] "=r" (step3_12), [step3_13] "=r" (step3_13),
 452           [step3_14] "=r" (step3_14), [step3_15] "=r" (step3_15)
 453         : [const_2_power_13] "r" (const_2_power_13), [step2_8] "r" (step2_8),
 454           [step2_9] "r" (step2_9), [step2_10] "r" (step2_10),
 455           [step2_11] "r" (step2_11), [step2_12] "r" (step2_12),
 456           [step2_13] "r" (step2_13), [step2_14] "r" (step2_14),
 457           [step2_15] "r" (step2_15), [cospi_16_64] "r" (cospi_16_64)
 458     );
 459
 460     step2_18 = step1_17 - step1_18;
 461     step2_29 = step1_30 - step1_29;
 462
 463     __asm__ __volatile__ (
 464         "mtlo     %[const_2_power_13],  $ac0                            \n\t"
 465         "mthi     $zero,                $ac0                            \n\t"
 466         "msub     $ac0,                 %[step2_18],    %[cospi_8_64]   \n\t"
 467         "madd     $ac0,                 %[step2_29],    %[cospi_24_64]  \n\t"
 468         "extp     %[step3_18],          $ac0,           31              \n\t"
 469
 470         : [step3_18] "=r" (step3_18)
 471         : [const_2_power_13] "r" (const_2_power_13),
 472           [step2_18] "r" (step2_18), [step2_29] "r" (step2_29),
 473           [cospi_24_64] "r" (cospi_24_64), [cospi_8_64] "r" (cospi_8_64)
 474     );
 475
 476     temp21 = step2_18 * cospi_24_64 + step2_29 * cospi_8_64;
 477     step3_29 = (temp21 + DCT_CONST_ROUNDING) >> DCT_CONST_BITS;
 478
 479     step2_19 = step1_16 - step1_19;
 480     step2_28 = step1_31 - step1_28;
 481
 482     __asm__ __volatile__ (
 483         "mtlo     %[const_2_power_13],  $ac0                            \n\t"
 484         "mthi     $zero,                $ac0                            \n\t"
 485         "msub     $ac0,                 %[step2_19],    %[cospi_8_64]   \n\t"
 486         "madd     $ac0,                 %[step2_28],    %[cospi_24_64]  \n\t"
 487         "extp     %[step3_19],          $ac0,           31              \n\t"
 488
 489         : [step3_19] "=r" (step3_19)
 490         : [const_2_power_13] "r" (const_2_power_13),
 491           [step2_19] "r" (step2_19), [step2_28] "r" (step2_28),
 492           [cospi_24_64] "r" (cospi_24_64), [cospi_8_64] "r" (cospi_8_64)
 493     );
 494
 495     temp21 = step2_19 * cospi_24_64 + step2_28 * cospi_8_64;
 496     step3_28 = (temp21 + DCT_CONST_ROUNDING) >> DCT_CONST_BITS;
 497
 498     step3_16 = step1_16 + step1_19;
 499     step3_17 = step1_17 + step1_18;
 500     step3_30 = step1_29 + step1_30;
 501     step3_31 = step1_28 + step1_31;
 502
 503     step2_20 = step1_23 - step1_20;
 504     step2_27 = step1_24 - step1_27;
 505
 506     __asm__ __volatile__ (
 507         "mtlo     %[const_2_power_13],  $ac0                            \n\t"
 508         "mthi     $zero,                $ac0                            \n\t"
 509         "msub     $ac0,                 %[step2_20],    %[cospi_24_64]  \n\t"
 510         "msub     $ac0,                 %[step2_27],    %[cospi_8_64]   \n\t"
 511         "extp     %[step3_20],          $ac0,           31              \n\t"
 512
 513         : [step3_20] "=r" (step3_20)
 514         : [const_2_power_13] "r" (const_2_power_13),
 515           [step2_20] "r" (step2_20), [step2_27] "r" (step2_27),
 516           [cospi_24_64] "r" (cospi_24_64), [cospi_8_64] "r" (cospi_8_64)
 517     );
 518
 519     temp21 = -step2_20 * cospi_8_64 + step2_27 * cospi_24_64;
 520     step3_27 = (temp21 + DCT_CONST_ROUNDING) >> DCT_CONST_BITS;
 521
 522     step2_21 = step1_22 - step1_21;
 523     step2_26 = step1_25 - step1_26;
 524
 525     __asm__ __volatile__ (
 526         "mtlo     %[const_2_power_13],  $ac1                            \n\t"
 527         "mthi     $zero,                $ac1                            \n\t"
 528         "msub     $ac1,                 %[step2_21],    %[cospi_24_64]  \n\t"
 529         "msub     $ac1,                 %[step2_26],    %[cospi_8_64]   \n\t"
 530         "extp     %[step3_21],          $ac1,           31              \n\t"
 531
 532         : [step3_21] "=r" (step3_21)
 533         : [const_2_power_13] "r" (const_2_power_13),
 534           [step2_21] "r" (step2_21), [step2_26] "r" (step2_26),
 535           [cospi_24_64] "r" (cospi_24_64), [cospi_8_64] "r" (cospi_8_64)
 536     );
 537
 538     temp21 = -step2_21 * cospi_8_64 + step2_26 * cospi_24_64;
 539     step3_26 = (temp21 + DCT_CONST_ROUNDING) >> DCT_CONST_BITS;
 540
 541     step3_22 = step1_21 + step1_22;
 542     step3_23 = step1_20 + step1_23;
 543     step3_24 = step1_24 + step1_27;
 544     step3_25 = step1_25 + step1_26;
 545
 546     step2_16 = step3_16 + step3_23;
 547     step2_17 = step3_17 + step3_22;
 548     step2_18 = step3_18 + step3_21;
 549     step2_19 = step3_19 + step3_20;
 550     step2_20 = step3_19 - step3_20;
 551     step2_21 = step3_18 - step3_21;
 552     step2_22 = step3_17 - step3_22;
 553     step2_23 = step3_16 - step3_23;
 554
 555     step2_24 = step3_31 - step3_24;
 556     step2_25 = step3_30 - step3_25;
 557     step2_26 = step3_29 - step3_26;
 558     step2_27 = step3_28 - step3_27;
 559     step2_28 = step3_28 + step3_27;
 560     step2_29 = step3_29 + step3_26;
 561     step2_30 = step3_30 + step3_25;
 562     step2_31 = step3_31 + step3_24;
 563
 564     __asm__ __volatile__ (
 565         "lh       %[load1],             0(%[input])                     \n\t"
 566         "lh       %[load2],             32(%[input])                    \n\t"
 567         "lh       %[load3],             16(%[input])                    \n\t"
 568         "lh       %[load4],             48(%[input])                    \n\t"
 569
 570         "mtlo     %[const_2_power_13],  $ac1                            \n\t"
 571         "mthi     $zero,                $ac1                            \n\t"
 572         "mtlo     %[const_2_power_13],  $ac2                            \n\t"
 573         "mthi     $zero,                $ac2                            \n\t"
 574         "add      %[result1],           %[load1],       %[load2]        \n\t"
 575         "sub      %[result2],           %[load1],       %[load2]        \n\t"
 576         "madd     $ac1,                 %[result1],     %[cospi_16_64]  \n\t"
 577         "madd     $ac2,                 %[result2],     %[cospi_16_64]  \n\t"
 578         "extp     %[temp0],             $ac1,           31              \n\t"
 579         "extp     %[temp1],             $ac2,           31              \n\t"
 580
 581         "mtlo     %[const_2_power_13],  $ac3                            \n\t"
 582         "mthi     $zero,                $ac3                            \n\t"
 583         "madd     $ac3,                 %[load3],       %[cospi_24_64]  \n\t"
 584         "msub     $ac3,                 %[load4],       %[cospi_8_64]   \n\t"
 585         "extp     %[temp2],             $ac3,           31              \n\t"
 586         "mtlo     %[const_2_power_13],  $ac1                            \n\t"
 587         "mthi     $zero,                $ac1                            \n\t"
 588         "madd     $ac1,                 %[load3],       %[cospi_8_64]   \n\t"
 589         "madd     $ac1,                 %[load4],       %[cospi_24_64]  \n\t"
 590         "extp     %[temp3],             $ac1,           31              \n\t"
 591         "add      %[step1_0],           %[temp0],       %[temp3]        \n\t"
 592         "add      %[step1_1],           %[temp1],       %[temp2]        \n\t"
 593         "sub      %[step1_2],           %[temp1],       %[temp2]        \n\t"
 594         "sub      %[step1_3],           %[temp0],       %[temp3]        \n\t"
 595
 596         : [load1] "=&r" (load1), [load2] "=&r" (load2),
 597           [load3] "=&r" (load3), [load4] "=&r" (load4),
 598           [result1] "=&r" (result1), [result2] "=&r" (result2),
 599           [temp0] "=&r" (temp0), [temp1] "=&r" (temp1),
 600           [temp2] "=&r" (temp2), [temp3] "=&r" (temp3),
 601           [step1_0] "=r" (step1_0), [step1_1] "=r" (step1_1),
 602           [step1_2] "=r" (step1_2), [step1_3] "=r" (step1_3)
 603         : [const_2_power_13] "r" (const_2_power_13), [input] "r" (input),
 604           [cospi_24_64] "r" (cospi_24_64), [cospi_8_64] "r" (cospi_8_64),
 605           [cospi_16_64] "r" (cospi_16_64)
 606     );
 607
 608     __asm__ __volatile__ (
 609         "lh       %[load1],             8(%[input])                     \n\t"
 610         "lh       %[load2],             56(%[input])                    \n\t"
 611         "lh       %[load3],             40(%[input])                    \n\t"
 612         "lh       %[load4],             24(%[input])                    \n\t"
 613
 614         "mtlo     %[const_2_power_13],  $ac1                            \n\t"
 615         "mthi     $zero,                $ac1                            \n\t"
 616         "mtlo     %[const_2_power_13],  $ac3                            \n\t"
 617         "mthi     $zero,                $ac3                            \n\t"
 618
 619         "madd     $ac1,                 %[load1],       %[cospi_28_64]  \n\t"
 620         "msub     $ac1,                 %[load2],       %[cospi_4_64]   \n\t"
 621         "extp     %[temp0],             $ac1,           31              \n\t"
 622         "madd     $ac3,                 %[load1],       %[cospi_4_64]   \n\t"
 623         "madd     $ac3,                 %[load2],       %[cospi_28_64]  \n\t"
 624         "extp     %[temp3],             $ac3,           31              \n\t"
 625
 626         "mtlo     %[const_2_power_13],  $ac1                            \n\t"
 627         "mthi     $zero,                $ac1                            \n\t"
 628         "mtlo     %[const_2_power_13],  $ac2                            \n\t"
 629         "mthi     $zero,                $ac2                            \n\t"
 630
 631         "madd     $ac2,                 %[load3],       %[cospi_12_64]  \n\t"
 632         "msub     $ac2,                 %[load4],       %[cospi_20_64]  \n\t"
 633         "extp     %[temp1],             $ac2,           31              \n\t"
 634         "madd     $ac1,                 %[load3],       %[cospi_20_64]  \n\t"
 635         "madd     $ac1,                 %[load4],       %[cospi_12_64]  \n\t"
 636         "extp     %[temp2],             $ac1,           31              \n\t"
 637
 638         "mtlo     %[const_2_power_13],  $ac1                            \n\t"
 639         "mthi     $zero,                $ac1                            \n\t"
 640         "mtlo     %[const_2_power_13],  $ac3                            \n\t"
 641         "mthi     $zero,                $ac3                            \n\t"
 642
 643         "sub      %[load1],             %[temp3],       %[temp2]        \n\t"
 644         "sub      %[load1],             %[load1],       %[temp0]        \n\t"
 645         "add      %[load1],             %[load1],       %[temp1]        \n\t"
 646         "sub      %[load2],             %[temp0],       %[temp1]        \n\t"
 647         "sub      %[load2],             %[load2],       %[temp2]        \n\t"
 648         "add      %[load2],             %[load2],       %[temp3]        \n\t"
 649         "madd     $ac1,                 %[load1],       %[cospi_16_64]  \n\t"
 650         "madd     $ac3,                 %[load2],       %[cospi_16_64]  \n\t"
 651
 652         "extp     %[step1_5],           $ac1,           31              \n\t"
 653         "extp     %[step1_6],           $ac3,           31              \n\t"
 654         "add      %[step1_4],           %[temp0],       %[temp1]        \n\t"
 655         "add      %[step1_7],           %[temp3],       %[temp2]        \n\t"
 656
 657         : [load1] "=&r" (load1), [load2] "=&r" (load2),
 658           [load3] "=&r" (load3), [load4] "=&r" (load4),
 659           [temp0] "=&r" (temp0), [temp1] "=&r" (temp1),
 660           [temp2] "=&r" (temp2), [temp3] "=&r" (temp3),
 661           [step1_4] "=r" (step1_4), [step1_5] "=r" (step1_5),
 662           [step1_6] "=r" (step1_6), [step1_7] "=r" (step1_7)
 663         : [const_2_power_13] "r" (const_2_power_13), [input] "r" (input),
 664           [cospi_20_64] "r" (cospi_20_64), [cospi_12_64] "r" (cospi_12_64),
 665           [cospi_4_64] "r" (cospi_4_64), [cospi_28_64] "r" (cospi_28_64),
 666           [cospi_16_64] "r" (cospi_16_64)
 667     );
 668
 669     step2_0 = step1_0 + step1_7;
 670     step2_1 = step1_1 + step1_6;
 671     step2_2 = step1_2 + step1_5;
 672     step2_3 = step1_3 + step1_4;
 673     step2_4 = step1_3 - step1_4;
 674     step2_5 = step1_2 - step1_5;
 675     step2_6 = step1_1 - step1_6;
 676     step2_7 = step1_0 - step1_7;
 677
 678     // stage 7
 679     step1_0 = step2_0 + step3_15;
 680     step1_1 = step2_1 + step3_14;
 681     step1_2 = step2_2 + step3_13;
 682     step1_3 = step2_3 + step3_12;
 683     step1_4 = step2_4 + step3_11;
 684     step1_5 = step2_5 + step3_10;
 685     step1_6 = step2_6 + step3_9;
 686     step1_7 = step2_7 + step3_8;
 687     step1_8 = step2_7 - step3_8;
 688     step1_9 = step2_6 - step3_9;
 689     step1_10 = step2_5 - step3_10;
 690     step1_11 = step2_4 - step3_11;
 691     step1_12 = step2_3 - step3_12;
 692     step1_13 = step2_2 - step3_13;
 693     step1_14 = step2_1 - step3_14;
 694     step1_15 = step2_0 - step3_15;
 695
 696     __asm__ __volatile__ (
 697         "sub      %[temp0],             %[step2_27],    %[step2_20]     \n\t"
 698         "mtlo     %[const_2_power_13],  $ac0                            \n\t"
 699         "mthi     $zero,                $ac0                            \n\t"
 700         "madd     $ac0,                 %[temp0],       %[cospi_16_64]  \n\t"
 701         "extp     %[step1_20],          $ac0,           31              \n\t"
 702
 703         : [temp0] "=&r" (temp0), [step1_20] "=r" (step1_20)
 704         : [const_2_power_13] "r" (const_2_power_13), [step2_20] "r" (step2_20),
 705           [step2_27] "r" (step2_27), [cospi_16_64] "r" (cospi_16_64)
 706     );
 707
 708     temp21 = (step2_20 + step2_27) * cospi_16_64;
 709     step1_27 = (temp21 + DCT_CONST_ROUNDING) >> DCT_CONST_BITS;
 710
 711     __asm__ __volatile__ (
 712         "sub      %[temp0],             %[step2_26],    %[step2_21]     \n\t"
 713         "mtlo     %[const_2_power_13],  $ac0                            \n\t"
 714         "mthi     $zero,                $ac0                            \n\t"
 715         "madd     $ac0,                 %[temp0],       %[cospi_16_64]  \n\t"
 716         "extp     %[step1_21],          $ac0,           31              \n\t"
 717
 718         : [temp0] "=&r" (temp0), [step1_21] "=r" (step1_21)
 719         : [const_2_power_13] "r" (const_2_power_13), [step2_26] "r" (step2_26),
 720           [step2_21] "r" (step2_21), [cospi_16_64] "r" (cospi_16_64)
 721     );
 722
 723     temp21 = (step2_21 + step2_26) * cospi_16_64;
 724     step1_26 = (temp21 + DCT_CONST_ROUNDING) >> DCT_CONST_BITS;
 725
 726     __asm__ __volatile__ (
 727         "sub      %[temp0],             %[step2_25],    %[step2_22]     \n\t"
 728         "mtlo     %[const_2_power_13],  $ac0                            \n\t"
 729         "mthi     $zero,                $ac0                            \n\t"
 730         "madd     $ac0,                 %[temp0],       %[cospi_16_64]  \n\t"
 731         "extp     %[step1_22],          $ac0,           31              \n\t"
 732
 733         : [temp0] "=&r" (temp0), [step1_22] "=r" (step1_22)
 734         : [const_2_power_13] "r" (const_2_power_13), [step2_25] "r" (step2_25),
 735           [step2_22] "r" (step2_22), [cospi_16_64] "r" (cospi_16_64)
 736     );
 737
 738     temp21 = (step2_22 + step2_25) * cospi_16_64;
 739     step1_25 = (temp21 + DCT_CONST_ROUNDING) >> DCT_CONST_BITS;
 740
 741     __asm__ __volatile__ (
 742         "sub      %[temp0],             %[step2_24],    %[step2_23]     \n\t"
 743         "mtlo     %[const_2_power_13],  $ac0                            \n\t"
 744         "mthi     $zero,                $ac0                            \n\t"
 745         "madd     $ac0,                 %[temp0],       %[cospi_16_64]  \n\t"
 746         "extp     %[step1_23],          $ac0,           31              \n\t"
 747
 748         : [temp0] "=&r" (temp0), [step1_23] "=r" (step1_23)
 749         : [const_2_power_13] "r" (const_2_power_13), [step2_24] "r" (step2_24),
 750           [step2_23] "r" (step2_23), [cospi_16_64] "r" (cospi_16_64)
 751     );
 752
 753     temp21 = (step2_23 + step2_24) * cospi_16_64;
 754     step1_24 = (temp21 + DCT_CONST_ROUNDING) >> DCT_CONST_BITS;
 755
 756     __asm__ __volatile__ (
 757         "lbu      %[temp2],         0(%[dest_pix])                      \n\t"
 758         "add      %[temp0],         %[step1_0],         %[step2_31]     \n\t"
 759         "addi     %[temp0],         %[temp0],           32              \n\t"
 760         "sra      %[temp0],         %[temp0],           6               \n\t"
 761         "add      %[temp2],         %[temp2],           %[temp0]        \n\t"
 762         "lbux     %[temp0],         %[temp2](%[cm])                     \n\t"
 763         "add      %[temp1],         %[step1_1],         %[step2_30]     \n\t"
 764         "sb       %[temp0],         0(%[dest_pix])                      \n\t"
 765         "addu     %[dest_pix],      %[dest_pix],        %[dest_stride]  \n\t"
 766         "lbu      %[temp3],         0(%[dest_pix])                      \n\t"
 767         "addi     %[temp1],         %[temp1],           32              \n\t"
 768         "sra      %[temp1],         %[temp1],           6               \n\t"
 769         "add      %[temp3],         %[temp3],           %[temp1]        \n\t"
 770         "lbux     %[temp1],         %[temp3](%[cm])                     \n\t"
 771         "sb       %[temp1],         0(%[dest_pix])                      \n\t"
 772         "addu     %[dest_pix],      %[dest_pix],        %[dest_stride]  \n\t"
 773
 774         "lbu      %[temp2],         0(%[dest_pix])                      \n\t"
 775         "add      %[temp0],         %[step1_2],         %[step2_29]     \n\t"
 776         "addi     %[temp0],         %[temp0],           32              \n\t"
 777         "sra      %[temp0],         %[temp0],           6               \n\t"
 778         "add      %[temp2],         %[temp2],           %[temp0]        \n\t"
 779         "lbux     %[temp0],         %[temp2](%[cm])                     \n\t"
 780         "add      %[temp1],         %[step1_3],         %[step2_28]     \n\t"
 781         "sb       %[temp0],         0(%[dest_pix])                      \n\t"
 782         "addu     %[dest_pix],      %[dest_pix],        %[dest_stride]  \n\t"
 783         "lbu      %[temp3],         0(%[dest_pix])                      \n\t"
 784         "addi     %[temp1],         %[temp1],           32              \n\t"
 785         "sra      %[temp1],         %[temp1],           6               \n\t"
 786         "add      %[temp3],         %[temp3],           %[temp1]        \n\t"
 787         "lbux     %[temp1],         %[temp3](%[cm])                     \n\t"
 788         "sb       %[temp1],         0(%[dest_pix])                      \n\t"
 789         "addu     %[dest_pix],      %[dest_pix],        %[dest_stride]  \n\t"
 790
 791         : [temp0] "=&r" (temp0), [temp1] "=&r" (temp1), [temp2] "=&r" (temp2),
 792           [temp3] "=&r" (temp3), [dest_pix] "+r" (dest_pix)
 793         : [cm] "r" (cm), [dest_stride] "r" (dest_stride),
 794           [step1_0] "r" (step1_0), [step1_1] "r" (step1_1),
 795           [step1_2] "r" (step1_2), [step1_3] "r" (step1_3),
 796           [step2_28] "r" (step2_28), [step2_29] "r" (step2_29),
 797           [step2_30] "r" (step2_30), [step2_31] "r" (step2_31)
 798     );
 799
 800     step3_12 = ROUND_POWER_OF_TWO((step1_3 - step2_28), 6);
 801     step3_13 = ROUND_POWER_OF_TWO((step1_2 - step2_29), 6);
 802     step3_14 = ROUND_POWER_OF_TWO((step1_1 - step2_30), 6);
 803     step3_15 = ROUND_POWER_OF_TWO((step1_0 - step2_31), 6);
 804
 805     __asm__ __volatile__ (
 806         "lbu      %[temp2],         0(%[dest_pix1])                     \n\t"
 807         "add      %[temp2],         %[temp2],           %[step3_15]     \n\t"
 808         "lbux     %[temp0],         %[temp2](%[cm])                     \n\t"
 809         "sb       %[temp0],         0(%[dest_pix1])                     \n\t"
 810         "subu     %[dest_pix1],     %[dest_pix1],       %[dest_stride]  \n\t"
 811         "lbu      %[temp3],         0(%[dest_pix1])                     \n\t"
 812         "add      %[temp3],         %[temp3],           %[step3_14]     \n\t"
 813         "lbux     %[temp1],         %[temp3](%[cm])                     \n\t"
 814         "sb       %[temp1],         0(%[dest_pix1])                     \n\t"
 815         "subu     %[dest_pix1],     %[dest_pix1],       %[dest_stride]  \n\t"
 816
 817         "lbu      %[temp2],         0(%[dest_pix1])                     \n\t"
 818         "add      %[temp2],         %[temp2],           %[step3_13]     \n\t"
 819         "lbux     %[temp0],         %[temp2](%[cm])                     \n\t"
 820         "sb       %[temp0],         0(%[dest_pix1])                     \n\t"
 821         "subu     %[dest_pix1],     %[dest_pix1],       %[dest_stride]  \n\t"
 822         "lbu      %[temp3],         0(%[dest_pix1])                     \n\t"
 823         "add      %[temp3],         %[temp3],           %[step3_12]     \n\t"
 824         "lbux     %[temp1],         %[temp3](%[cm])                     \n\t"
 825         "sb       %[temp1],         0(%[dest_pix1])                     \n\t"
 826         "subu     %[dest_pix1],     %[dest_pix1],       %[dest_stride]  \n\t"
 827
 828         : [temp0] "=&r" (temp0), [temp1] "=&r" (temp1), [temp2] "=&r" (temp2),
 829           [temp3] "=&r" (temp3), [dest_pix1] "+r" (dest_pix1)
 830         : [cm] "r" (cm), [dest_stride] "r" (dest_stride),
 831           [step3_12] "r" (step3_12), [step3_13] "r" (step3_13),
 832           [step3_14] "r" (step3_14), [step3_15] "r" (step3_15)
 833     );
 834
 835     __asm__ __volatile__ (
 836         "lbu      %[temp2],         0(%[dest_pix])                      \n\t"
 837         "add      %[temp0],         %[step1_4],         %[step1_27]     \n\t"
 838         "addi     %[temp0],         %[temp0],           32              \n\t"
 839         "sra      %[temp0],         %[temp0],           6               \n\t"
 840         "add      %[temp2],         %[temp2],           %[temp0]        \n\t"
 841         "lbux     %[temp0],         %[temp2](%[cm])                     \n\t"
 842         "add      %[temp1],         %[step1_5],         %[step1_26]     \n\t"
 843         "sb       %[temp0],         0(%[dest_pix])                      \n\t"
 844         "addu     %[dest_pix],      %[dest_pix],        %[dest_stride]  \n\t"
 845         "lbu      %[temp3],         0(%[dest_pix])                      \n\t"
 846         "addi     %[temp1],         %[temp1],           32              \n\t"
 847         "sra      %[temp1],         %[temp1],           6               \n\t"
 848         "add      %[temp3],         %[temp3],           %[temp1]        \n\t"
 849         "lbux     %[temp1],         %[temp3](%[cm])                     \n\t"
 850         "sb       %[temp1],         0(%[dest_pix])                      \n\t"
 851         "addu     %[dest_pix],      %[dest_pix],        %[dest_stride]  \n\t"
 852
 853         "lbu      %[temp2],         0(%[dest_pix])                      \n\t"
 854         "add      %[temp0],         %[step1_6],         %[step1_25]     \n\t"
 855         "addi     %[temp0],         %[temp0],           32              \n\t"
 856         "sra      %[temp0],         %[temp0],           6               \n\t"
 857         "add      %[temp2],         %[temp2],           %[temp0]        \n\t"
 858         "lbux     %[temp0],         %[temp2](%[cm])                     \n\t"
 859         "add      %[temp1],         %[step1_7],         %[step1_24]     \n\t"
 860         "sb       %[temp0],         0(%[dest_pix])                      \n\t"
 861         "addu     %[dest_pix],      %[dest_pix],        %[dest_stride]  \n\t"
 862         "lbu      %[temp3],         0(%[dest_pix])                      \n\t"
 863         "addi     %[temp1],         %[temp1],           32              \n\t"
 864         "sra      %[temp1],         %[temp1],           6               \n\t"
 865         "add      %[temp3],         %[temp3],           %[temp1]        \n\t"
 866         "lbux     %[temp1],         %[temp3](%[cm])                     \n\t"
 867         "sb       %[temp1],         0(%[dest_pix])                      \n\t"
 868         "addu     %[dest_pix],      %[dest_pix],        %[dest_stride]  \n\t"
 869
 870         : [temp0] "=&r" (temp0), [temp1] "=&r" (temp1), [temp2] "=&r" (temp2),
 871           [temp3] "=&r" (temp3), [dest_pix] "+r" (dest_pix)
 872         : [cm] "r" (cm), [dest_stride] "r" (dest_stride),
 873           [step1_4] "r" (step1_4), [step1_5] "r" (step1_5),
 874           [step1_6] "r" (step1_6), [step1_7] "r" (step1_7),
 875           [step1_24] "r" (step1_24), [step1_25] "r" (step1_25),
 876           [step1_26] "r" (step1_26), [step1_27] "r" (step1_27)
 877     );
 878
 879     step3_12 = ROUND_POWER_OF_TWO((step1_7 - step1_24), 6);
 880     step3_13 = ROUND_POWER_OF_TWO((step1_6 - step1_25), 6);
 881     step3_14 = ROUND_POWER_OF_TWO((step1_5 - step1_26), 6);
 882     step3_15 = ROUND_POWER_OF_TWO((step1_4 - step1_27), 6);
 883
 884     __asm__ __volatile__ (
 885         "lbu      %[temp2],         0(%[dest_pix1])                     \n\t"
 886         "add      %[temp2],         %[temp2],           %[step3_15]     \n\t"
 887         "lbux     %[temp0],         %[temp2](%[cm])                     \n\t"
 888         "sb       %[temp0],         0(%[dest_pix1])                     \n\t"
 889         "subu     %[dest_pix1],     %[dest_pix1],       %[dest_stride]  \n\t"
 890         "lbu      %[temp3],         0(%[dest_pix1])                     \n\t"
 891         "add      %[temp3],         %[temp3],           %[step3_14]     \n\t"
 892         "lbux     %[temp1],         %[temp3](%[cm])                     \n\t"
 893         "sb       %[temp1],         0(%[dest_pix1])                     \n\t"
 894         "subu     %[dest_pix1],     %[dest_pix1],       %[dest_stride]  \n\t"
 895
 896         "lbu      %[temp2],         0(%[dest_pix1])                     \n\t"
 897         "add      %[temp2],         %[temp2],           %[step3_13]     \n\t"
 898         "lbux     %[temp0],         %[temp2](%[cm])                     \n\t"
 899         "sb       %[temp0],         0(%[dest_pix1])                     \n\t"
 900         "subu     %[dest_pix1],     %[dest_pix1],       %[dest_stride]  \n\t"
 901         "lbu      %[temp3],         0(%[dest_pix1])                     \n\t"
 902         "add      %[temp3],         %[temp3],           %[step3_12]     \n\t"
 903         "lbux     %[temp1],         %[temp3](%[cm])                     \n\t"
 904         "sb       %[temp1],         0(%[dest_pix1])                     \n\t"
 905         "subu     %[dest_pix1],     %[dest_pix1],       %[dest_stride]  \n\t"
 906
 907         : [temp0] "=&r" (temp0), [temp1] "=&r" (temp1), [temp2] "=&r" (temp2),
 908           [temp3] "=&r" (temp3), [dest_pix1] "+r" (dest_pix1)
 909         : [cm] "r" (cm), [dest_stride] "r" (dest_stride),
 910           [step3_12] "r" (step3_12), [step3_13] "r" (step3_13),
 911           [step3_14] "r" (step3_14), [step3_15] "r" (step3_15)
 912     );
 913
 914     __asm__ __volatile__ (
 915         "lbu      %[temp2],         0(%[dest_pix])                      \n\t"
 916         "add      %[temp0],         %[step1_8],         %[step1_23]     \n\t"
 917         "addi     %[temp0],         %[temp0],           32              \n\t"
 918         "sra      %[temp0],         %[temp0],           6               \n\t"
 919         "add      %[temp2],         %[temp2],           %[temp0]        \n\t"
 920         "lbux     %[temp0],         %[temp2](%[cm])                     \n\t"
 921         "add      %[temp1],         %[step1_9],         %[step1_22]     \n\t"
 922         "sb       %[temp0],         0(%[dest_pix])                      \n\t"
 923         "addu     %[dest_pix],      %[dest_pix],        %[dest_stride]  \n\t"
 924         "lbu      %[temp3],         0(%[dest_pix])                      \n\t"
 925         "addi     %[temp1],         %[temp1],           32              \n\t"
 926         "sra      %[temp1],         %[temp1],           6               \n\t"
 927         "add      %[temp3],         %[temp3],           %[temp1]        \n\t"
 928         "lbux     %[temp1],         %[temp3](%[cm])                     \n\t"
 929         "sb       %[temp1],         0(%[dest_pix])                      \n\t"
 930         "addu     %[dest_pix],      %[dest_pix],        %[dest_stride]  \n\t"
 931
 932         "lbu      %[temp2],         0(%[dest_pix])                      \n\t"
 933         "add      %[temp0],         %[step1_10],        %[step1_21]     \n\t"
 934         "addi     %[temp0],         %[temp0],           32              \n\t"
 935         "sra      %[temp0],         %[temp0],           6               \n\t"
 936         "add      %[temp2],         %[temp2],           %[temp0]        \n\t"
 937         "lbux     %[temp0],         %[temp2](%[cm])                     \n\t"
 938         "add      %[temp1],         %[step1_11],        %[step1_20]     \n\t"
 939         "sb       %[temp0],         0(%[dest_pix])                      \n\t"
 940         "addu     %[dest_pix],      %[dest_pix],        %[dest_stride]  \n\t"
 941         "lbu      %[temp3],         0(%[dest_pix])                      \n\t"
 942         "addi     %[temp1],         %[temp1],           32              \n\t"
 943         "sra      %[temp1],         %[temp1],           6               \n\t"
 944         "add      %[temp3],         %[temp3],           %[temp1]        \n\t"
 945         "lbux     %[temp1],         %[temp3](%[cm])                     \n\t"
 946         "sb       %[temp1],         0(%[dest_pix])                      \n\t"
 947         "addu     %[dest_pix],      %[dest_pix],        %[dest_stride]  \n\t"
 948
 949         : [temp0] "=&r" (temp0), [temp1] "=&r" (temp1), [temp2] "=&r" (temp2),
 950           [temp3] "=&r" (temp3), [dest_pix] "+r" (dest_pix)
 951         : [cm] "r" (cm), [dest_stride] "r" (dest_stride),
 952           [step1_8] "r" (step1_8), [step1_9] "r" (step1_9),
 953           [step1_10] "r" (step1_10), [step1_11] "r" (step1_11),
 954           [step1_20] "r" (step1_20), [step1_21] "r" (step1_21),
 955           [step1_22] "r" (step1_22), [step1_23] "r" (step1_23)
 956     );
 957
 958     step3_12 = ROUND_POWER_OF_TWO((step1_11 - step1_20), 6);
 959     step3_13 = ROUND_POWER_OF_TWO((step1_10 - step1_21), 6);
 960     step3_14 = ROUND_POWER_OF_TWO((step1_9 - step1_22), 6);
 961     step3_15 = ROUND_POWER_OF_TWO((step1_8 - step1_23), 6);
 962
 963     __asm__ __volatile__ (
 964         "lbu      %[temp2],         0(%[dest_pix1])                     \n\t"
 965         "add      %[temp2],         %[temp2],           %[step3_15]     \n\t"
 966         "lbux     %[temp0],         %[temp2](%[cm])                     \n\t"
 967         "sb       %[temp0],         0(%[dest_pix1])                     \n\t"
 968         "subu     %[dest_pix1],     %[dest_pix1],       %[dest_stride]  \n\t"
 969         "lbu      %[temp3],         0(%[dest_pix1])                     \n\t"
 970         "add      %[temp3],         %[temp3],           %[step3_14]     \n\t"
 971         "lbux     %[temp1],         %[temp3](%[cm])                     \n\t"
 972         "sb       %[temp1],         0(%[dest_pix1])                     \n\t"
 973         "subu     %[dest_pix1],     %[dest_pix1],       %[dest_stride]  \n\t"
 974
 975         "lbu      %[temp2],         0(%[dest_pix1])                     \n\t"
 976         "add      %[temp2],         %[temp2],           %[step3_13]     \n\t"
 977         "lbux     %[temp0],         %[temp2](%[cm])                     \n\t"
 978         "sb       %[temp0],         0(%[dest_pix1])                     \n\t"
 979         "subu     %[dest_pix1],     %[dest_pix1],       %[dest_stride]  \n\t"
 980         "lbu      %[temp3],         0(%[dest_pix1])                     \n\t"
 981         "add      %[temp3],         %[temp3],           %[step3_12]     \n\t"
 982         "lbux     %[temp1],         %[temp3](%[cm])                     \n\t"
 983         "sb       %[temp1],         0(%[dest_pix1])                     \n\t"
 984         "subu     %[dest_pix1],     %[dest_pix1],       %[dest_stride]  \n\t"
 985
 986         : [temp0] "=&r" (temp0), [temp1] "=&r" (temp1), [temp2] "=&r" (temp2),
 987           [temp3] "=&r" (temp3), [dest_pix1] "+r" (dest_pix1)
 988         : [cm] "r" (cm), [dest_stride] "r" (dest_stride),
 989           [step3_12] "r" (step3_12), [step3_13] "r" (step3_13),
 990           [step3_14] "r" (step3_14), [step3_15] "r" (step3_15)
 991     );
 992
 993     __asm__ __volatile__ (
 994         "lbu      %[temp2],         0(%[dest_pix])                      \n\t"
 995         "add      %[temp0],         %[step1_12],        %[step2_19]     \n\t"
 996         "addi     %[temp0],         %[temp0],           32              \n\t"
 997         "sra      %[temp0],         %[temp0],           6               \n\t"
 998         "add      %[temp2],         %[temp2],           %[temp0]        \n\t"
 999         "lbux     %[temp0],         %[temp2](%[cm])                     \n\t"
1000         "add      %[temp1],         %[step1_13],        %[step2_18]     \n\t"
1001         "sb       %[temp0],         0(%[dest_pix])                      \n\t"
1002         "addu     %[dest_pix],      %[dest_pix],        %[dest_stride]  \n\t"
1003         "lbu      %[temp3],         0(%[dest_pix])                      \n\t"
1004         "addi     %[temp1],         %[temp1],           32              \n\t"
1005         "sra      %[temp1],         %[temp1],           6               \n\t"
1006         "add      %[temp3],         %[temp3],           %[temp1]        \n\t"
1007         "lbux     %[temp1],         %[temp3](%[cm])                     \n\t"
1008         "sb       %[temp1],         0(%[dest_pix])                      \n\t"
1009         "addu     %[dest_pix],      %[dest_pix],        %[dest_stride]  \n\t"
1010
1011         "lbu      %[temp2],         0(%[dest_pix])                      \n\t"
1012         "add      %[temp0],         %[step1_14],        %[step2_17]     \n\t"
1013         "addi     %[temp0],         %[temp0],           32              \n\t"
1014         "sra      %[temp0],         %[temp0],           6               \n\t"
1015         "add      %[temp2],         %[temp2],           %[temp0]        \n\t"
1016         "lbux     %[temp0],         %[temp2](%[cm])                     \n\t"
1017         "add      %[temp1],         %[step1_15],        %[step2_16]     \n\t"
1018         "sb       %[temp0],         0(%[dest_pix])                      \n\t"
1019         "addu     %[dest_pix],      %[dest_pix],        %[dest_stride]  \n\t"
1020         "lbu      %[temp3],         0(%[dest_pix])                      \n\t"
1021         "addi     %[temp1],         %[temp1],           32              \n\t"
1022         "sra      %[temp1],         %[temp1],           6               \n\t"
1023         "add      %[temp3],         %[temp3],           %[temp1]        \n\t"
1024         "lbux     %[temp1],         %[temp3](%[cm])                     \n\t"
1025         "sb       %[temp1],         0(%[dest_pix])                      \n\t"
1026
1027         : [temp0] "=&r" (temp0), [temp1] "=&r" (temp1), [temp2] "=&r" (temp2),
1028           [temp3] "=&r" (temp3), [dest_pix] "+r" (dest_pix)
1029         : [cm] "r" (cm), [dest_stride] "r" (dest_stride),
1030           [step1_12] "r" (step1_12), [step1_13] "r" (step1_13),
1031           [step1_14] "r" (step1_14), [step1_15] "r" (step1_15),
1032           [step2_16] "r" (step2_16), [step2_17] "r" (step2_17),
1033           [step2_18] "r" (step2_18), [step2_19] "r" (step2_19)
1034     );
1035
1036     step3_12 = ROUND_POWER_OF_TWO((step1_15 - step2_16), 6);
1037     step3_13 = ROUND_POWER_OF_TWO((step1_14 - step2_17), 6);
1038     step3_14 = ROUND_POWER_OF_TWO((step1_13 - step2_18), 6);
1039     step3_15 = ROUND_POWER_OF_TWO((step1_12 - step2_19), 6);
1040
1041     __asm__ __volatile__ (
1042         "lbu      %[temp2],         0(%[dest_pix1])                     \n\t"
1043         "add      %[temp2],         %[temp2],           %[step3_15]     \n\t"
1044         "lbux     %[temp0],         %[temp2](%[cm])                     \n\t"
1045         "sb       %[temp0],         0(%[dest_pix1])                     \n\t"
1046         "subu     %[dest_pix1],     %[dest_pix1],       %[dest_stride]  \n\t"
1047         "lbu      %[temp3],         0(%[dest_pix1])                     \n\t"
1048         "add      %[temp3],         %[temp3],           %[step3_14]     \n\t"
1049         "lbux     %[temp1],         %[temp3](%[cm])                     \n\t"
1050         "sb       %[temp1],         0(%[dest_pix1])                     \n\t"
1051         "subu     %[dest_pix1],     %[dest_pix1],       %[dest_stride]  \n\t"
1052
1053         "lbu      %[temp2],         0(%[dest_pix1])                     \n\t"
1054         "add      %[temp2],         %[temp2],           %[step3_13]     \n\t"
1055         "lbux     %[temp0],         %[temp2](%[cm])                     \n\t"
1056         "sb       %[temp0],         0(%[dest_pix1])                     \n\t"
1057         "subu     %[dest_pix1],     %[dest_pix1],       %[dest_stride]  \n\t"
1058         "lbu      %[temp3],         0(%[dest_pix1])                     \n\t"
1059         "add      %[temp3],         %[temp3],           %[step3_12]     \n\t"
1060         "lbux     %[temp1],         %[temp3](%[cm])                     \n\t"
1061         "sb       %[temp1],         0(%[dest_pix1])                     \n\t"
1062
1063         : [temp0] "=&r" (temp0), [temp1] "=&r" (temp1), [temp2] "=&r" (temp2),
1064           [temp3] "=&r" (temp3), [dest_pix1] "+r" (dest_pix1)
1065         : [cm] "r" (cm), [dest_stride] "r" (dest_stride),
1066           [step3_12] "r" (step3_12), [step3_13] "r" (step3_13),
1067           [step3_14] "r" (step3_14), [step3_15] "r" (step3_15)
1068     );
1069
1070     input += 32;
1071   }
1072 }
1073 #endif  // #if HAVE_DSPR2