src/third_party/libvpx/source/libvpx/vp9/common/mips/dspr2/vp9_common_dspr2.h

   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 #ifndef VP9_COMMON_MIPS_DSPR2_VP9_COMMON_DSPR2_H_
  12 #define VP9_COMMON_MIPS_DSPR2_VP9_COMMON_DSPR2_H_
  13
  14 #include <assert.h>
  15
  16 #include "./vpx_config.h"
  17 #include "vpx/vpx_integer.h"
  18 #include "vp9/common/vp9_common.h"
  19
  20 #ifdef __cplusplus
  21 extern "C" {
  22 #endif
  23
  24 #if HAVE_DSPR2
  25 #define CROP_WIDTH 512
  26 extern uint8_t *vp9_ff_cropTbl;
  27
  28 #define DCT_CONST_ROUND_SHIFT_TWICE_COSPI_16_64(input)                    ({   \
  29                                                                                \
  30   int32_t tmp, out;                                                            \
  31   int     dct_cost_rounding = DCT_CONST_ROUNDING;                              \
  32   int     in = input;                                                          \
  33                                                                                \
  34   __asm__ __volatile__ (                                                       \
  35       /* out = dct_const_round_shift(input_dc * cospi_16_64); */               \
  36       "mtlo     %[dct_cost_rounding],   $ac1                              \n\t"\
  37       "mthi     $zero,                  $ac1                              \n\t"\
  38       "madd     $ac1,                   %[in],            %[cospi_16_64]  \n\t"\
  39       "extp     %[tmp],                 $ac1,             31              \n\t"\
  40                                                                                \
  41       /* out = dct_const_round_shift(out * cospi_16_64); */                    \
  42       "mtlo     %[dct_cost_rounding],   $ac2                              \n\t"\
  43       "mthi     $zero,                  $ac2                              \n\t"\
  44       "madd     $ac2,                   %[tmp],           %[cospi_16_64]  \n\t"\
  45       "extp     %[out],                 $ac2,             31              \n\t"\
  46                                                                                \
  47       : [tmp] "=&r" (tmp), [out] "=r" (out)                                    \
  48       : [in] "r" (in),                                                         \
  49         [dct_cost_rounding] "r" (dct_cost_rounding),                           \
  50         [cospi_16_64] "r" (cospi_16_64)                                        \
  51    );                                                                          \
  52   out;                                                                    })
  53
  54 static INLINE void vp9_prefetch_load(const unsigned char *src) {
  55   __asm__ __volatile__ (
  56       "pref   0,  0(%[src])   \n\t"
  57       :
  58       : [src] "r" (src)
  59   );
  60 }
  61
  62 /* prefetch data for store */
  63 static INLINE void vp9_prefetch_store(unsigned char *dst) {
  64   __asm__ __volatile__ (
  65       "pref   1,  0(%[dst])   \n\t"
  66       :
  67       : [dst] "r" (dst)
  68   );
  69 }
  70
  71 static INLINE void vp9_prefetch_load_streamed(const unsigned char *src) {
  72   __asm__ __volatile__ (
  73       "pref   4,  0(%[src])   \n\t"
  74       :
  75       : [src] "r" (src)
  76   );
  77 }
  78
  79 /* prefetch data for store */
  80 static INLINE void vp9_prefetch_store_streamed(unsigned char *dst) {
  81   __asm__ __volatile__ (
  82       "pref   5,  0(%[dst])   \n\t"
  83       :
  84       : [dst] "r" (dst)
  85   );
  86 }
  87
  88 void vp9_idct32_cols_add_blk_dspr2(int16_t *input, uint8_t *dest,
  89                                    int dest_stride);
  90
  91 void vp9_convolve2_horiz_dspr2(const uint8_t *src, ptrdiff_t src_stride,
  92                                uint8_t *dst, ptrdiff_t dst_stride,
  93                                const int16_t *filter_x, int x_step_q4,
  94                                const int16_t *filter_y, int y_step_q4,
  95                                int w, int h);
  96
  97 void vp9_convolve2_avg_horiz_dspr2(const uint8_t *src, ptrdiff_t src_stride,
  98                                    uint8_t *dst, ptrdiff_t dst_stride,
  99                                    const int16_t *filter_x, int x_step_q4,
 100                                    const int16_t *filter_y, int y_step_q4,
 101                                    int w, int h);
 102
 103 void vp9_convolve2_avg_vert_dspr2(const uint8_t *src, ptrdiff_t src_stride,
 104                                   uint8_t *dst, ptrdiff_t dst_stride,
 105                                   const int16_t *filter_x, int x_step_q4,
 106                                   const int16_t *filter_y, int y_step_q4,
 107                                   int w, int h);
 108
 109 void vp9_convolve2_dspr2(const uint8_t *src, ptrdiff_t src_stride,
 110                          uint8_t *dst, ptrdiff_t dst_stride,
 111                          const int16_t *filter,
 112                          int w, int h);
 113
 114 void vp9_convolve2_vert_dspr2(const uint8_t *src, ptrdiff_t src_stride,
 115                               uint8_t *dst, ptrdiff_t dst_stride,
 116                               const int16_t *filter_x, int x_step_q4,
 117                               const int16_t *filter_y, int y_step_q4,
 118                               int w, int h);
 119
 120 #endif  // #if HAVE_DSPR2
 121 #ifdef __cplusplus
 122 }  // extern "C"
 123 #endif
 124
 125 #endif  // VP9_COMMON_MIPS_DSPR2_VP9_COMMON_DSPR2_H_