git.sesse.net Git - ffmpeg/blob - libavcodec/qpel_template.c

   1 /*
   2  * quarterpel DSP function templates
   3  *
   4  * This file is part of Libav.
   5  *
   6  * Libav is free software; you can redistribute it and/or
   7  * modify it under the terms of the GNU Lesser General Public
   8  * License as published by the Free Software Foundation; either
   9  * version 2.1 of the License, or (at your option) any later version.
  10  *
  11  * Libav is distributed in the hope that it will be useful,
  12  * but WITHOUT ANY WARRANTY; without even the implied warranty of
  13  * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
  14  * Lesser General Public License for more details.
  15  *
  16  * You should have received a copy of the GNU Lesser General Public
  17  * License along with Libav; if not, write to the Free Software
  18  * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
  19  */
  20
  21 /**
  22  * @file
  23  * quarterpel DSP function templates
  24  */
  25
  26 #define PIXOP2(OPNAME, OP)                                              \
  27 static inline void OPNAME ## _no_rnd_pixels8_l2_8(uint8_t *dst,         \
  28                                                   const uint8_t *src1,  \
  29                                                   const uint8_t *src2,  \
  30                                                   int dst_stride,       \
  31                                                   int src_stride1,      \
  32                                                   int src_stride2,      \
  33                                                   int h)                \
  34 {                                                                       \
  35     int i;                                                              \
  36                                                                         \
  37     for (i = 0; i < h; i++) {                                           \
  38         uint32_t a, b;                                                  \
  39         a = AV_RN32(&src1[i * src_stride1]);                            \
  40         b = AV_RN32(&src2[i * src_stride2]);                            \
  41         OP(*((uint32_t *) &dst[i * dst_stride]),                        \
  42            no_rnd_avg32(a, b));                                         \
  43         a = AV_RN32(&src1[i * src_stride1 + 4]);                        \
  44         b = AV_RN32(&src2[i * src_stride2 + 4]);                        \
  45         OP(*((uint32_t *) &dst[i * dst_stride + 4]),                    \
  46            no_rnd_avg32(a, b));                                         \
  47     }                                                                   \
  48 }                                                                       \
  49                                                                         \
  50 static inline void OPNAME ## _no_rnd_pixels16_l2_8(uint8_t *dst,        \
  51                                                    const uint8_t *src1, \
  52                                                    const uint8_t *src2, \
  53                                                    int dst_stride,      \
  54                                                    int src_stride1,     \
  55                                                    int src_stride2,     \
  56                                                    int h)               \
  57 {                                                                       \
  58     OPNAME ## _no_rnd_pixels8_l2_8(dst, src1, src2, dst_stride,         \
  59                                    src_stride1, src_stride2, h);        \
  60     OPNAME ## _no_rnd_pixels8_l2_8(dst  + 8,                            \
  61                                    src1 + 8,                            \
  62                                    src2 + 8,                            \
  63                                    dst_stride, src_stride1,             \
  64                                    src_stride2, h);                     \
  65 }                                                                       \
  66                                                                         \
  67 static inline void OPNAME ## _pixels8_l4_8(uint8_t *dst,                \
  68                                            const uint8_t *src1,         \
  69                                            const uint8_t *src2,         \
  70                                            const uint8_t *src3,         \
  71                                            const uint8_t *src4,         \
  72                                            int dst_stride,              \
  73                                            int src_stride1,             \
  74                                            int src_stride2,             \
  75                                            int src_stride3,             \
  76                                            int src_stride4,             \
  77                                            int h)                       \
  78 {                                                                       \
  79     /* FIXME HIGH BIT DEPTH */                                          \
  80     int i;                                                              \
  81                                                                         \
  82     for (i = 0; i < h; i++) {                                           \
  83         uint32_t a, b, c, d, l0, l1, h0, h1;                            \
  84         a  = AV_RN32(&src1[i * src_stride1]);                           \
  85         b  = AV_RN32(&src2[i * src_stride2]);                           \
  86         c  = AV_RN32(&src3[i * src_stride3]);                           \
  87         d  = AV_RN32(&src4[i * src_stride4]);                           \
  88         l0 = (a & 0x03030303UL) +                                       \
  89              (b & 0x03030303UL) +                                       \
  90                   0x02020202UL;                                         \
  91         h0 = ((a & 0xFCFCFCFCUL) >> 2) +                                \
  92              ((b & 0xFCFCFCFCUL) >> 2);                                 \
  93         l1 = (c & 0x03030303UL) +                                       \
  94              (d & 0x03030303UL);                                        \
  95         h1 = ((c & 0xFCFCFCFCUL) >> 2) +                                \
  96              ((d & 0xFCFCFCFCUL) >> 2);                                 \
  97         OP(*((uint32_t *) &dst[i * dst_stride]),                        \
  98            h0 + h1 + (((l0 + l1) >> 2) & 0x0F0F0F0FUL));                \
  99         a  = AV_RN32(&src1[i * src_stride1 + 4]);                       \
 100         b  = AV_RN32(&src2[i * src_stride2 + 4]);                       \
 101         c  = AV_RN32(&src3[i * src_stride3 + 4]);                       \
 102         d  = AV_RN32(&src4[i * src_stride4 + 4]);                       \
 103         l0 = (a & 0x03030303UL) +                                       \
 104              (b & 0x03030303UL) +                                       \
 105                   0x02020202UL;                                         \
 106         h0 = ((a & 0xFCFCFCFCUL) >> 2) +                                \
 107              ((b & 0xFCFCFCFCUL) >> 2);                                 \
 108         l1 = (c & 0x03030303UL) +                                       \
 109              (d & 0x03030303UL);                                        \
 110         h1 = ((c & 0xFCFCFCFCUL) >> 2) +                                \
 111              ((d & 0xFCFCFCFCUL) >> 2);                                 \
 112         OP(*((uint32_t *) &dst[i * dst_stride + 4]),                    \
 113            h0 + h1 + (((l0 + l1) >> 2) & 0x0F0F0F0FUL));                \
 114     }                                                                   \
 115 }                                                                       \
 116                                                                         \
 117 static inline void OPNAME ## _no_rnd_pixels8_l4_8(uint8_t *dst,         \
 118                                                   const uint8_t *src1,  \
 119                                                   const uint8_t *src2,  \
 120                                                   const uint8_t *src3,  \
 121                                                   const uint8_t *src4,  \
 122                                                   int dst_stride,       \
 123                                                   int src_stride1,      \
 124                                                   int src_stride2,      \
 125                                                   int src_stride3,      \
 126                                                   int src_stride4,      \
 127                                                   int h)                \
 128 {                                                                       \
 129     /* FIXME HIGH BIT DEPTH */                                          \
 130     int i;                                                              \
 131                                                                         \
 132     for (i = 0; i < h; i++) {                                           \
 133         uint32_t a, b, c, d, l0, l1, h0, h1;                            \
 134         a  = AV_RN32(&src1[i * src_stride1]);                           \
 135         b  = AV_RN32(&src2[i * src_stride2]);                           \
 136         c  = AV_RN32(&src3[i * src_stride3]);                           \
 137         d  = AV_RN32(&src4[i * src_stride4]);                           \
 138         l0 = (a & 0x03030303UL) +                                       \
 139              (b & 0x03030303UL) +                                       \
 140                   0x01010101UL;                                         \
 141         h0 = ((a & 0xFCFCFCFCUL) >> 2) +                                \
 142              ((b & 0xFCFCFCFCUL) >> 2);                                 \
 143         l1 = (c & 0x03030303UL) +                                       \
 144              (d & 0x03030303UL);                                        \
 145         h1 = ((c & 0xFCFCFCFCUL) >> 2) +                                \
 146              ((d & 0xFCFCFCFCUL) >> 2);                                 \
 147         OP(*((uint32_t *) &dst[i * dst_stride]),                        \
 148            h0 + h1 + (((l0 + l1) >> 2) & 0x0F0F0F0FUL));                \
 149         a  = AV_RN32(&src1[i * src_stride1 + 4]);                       \
 150         b  = AV_RN32(&src2[i * src_stride2 + 4]);                       \
 151         c  = AV_RN32(&src3[i * src_stride3 + 4]);                       \
 152         d  = AV_RN32(&src4[i * src_stride4 + 4]);                       \
 153         l0 = (a & 0x03030303UL) +                                       \
 154              (b & 0x03030303UL) +                                       \
 155                   0x01010101UL;                                         \
 156         h0 = ((a & 0xFCFCFCFCUL) >> 2) +                                \
 157              ((b & 0xFCFCFCFCUL) >> 2);                                 \
 158         l1 = (c & 0x03030303UL) +                                       \
 159              (d & 0x03030303UL);                                        \
 160         h1 = ((c & 0xFCFCFCFCUL) >> 2) +                                \
 161              ((d & 0xFCFCFCFCUL) >> 2);                                 \
 162         OP(*((uint32_t *) &dst[i * dst_stride + 4]),                    \
 163            h0 + h1 + (((l0 + l1) >> 2) & 0x0F0F0F0FUL));                \
 164     }                                                                   \
 165 }                                                                       \
 166                                                                         \
 167 static inline void OPNAME ## _pixels16_l4_8(uint8_t *dst,               \
 168                                             const uint8_t *src1,        \
 169                                             const uint8_t *src2,        \
 170                                             const uint8_t *src3,        \
 171                                             const uint8_t *src4,        \
 172                                             int dst_stride,             \
 173                                             int src_stride1,            \
 174                                             int src_stride2,            \
 175                                             int src_stride3,            \
 176                                             int src_stride4,            \
 177                                             int h)                      \
 178 {                                                                       \
 179     OPNAME ## _pixels8_l4_8(dst, src1, src2, src3, src4, dst_stride,    \
 180                             src_stride1, src_stride2, src_stride3,      \
 181                             src_stride4, h);                            \
 182     OPNAME ## _pixels8_l4_8(dst  + 8,                                   \
 183                             src1 + 8, src2 + 8,                         \
 184                             src3 + 8, src4 + 8,                         \
 185                             dst_stride, src_stride1, src_stride2,       \
 186                             src_stride3, src_stride4, h);               \
 187 }                                                                       \
 188                                                                         \
 189 static inline void OPNAME ## _no_rnd_pixels16_l4_8(uint8_t *dst,        \
 190                                                    const uint8_t *src1, \
 191                                                    const uint8_t *src2, \
 192                                                    const uint8_t *src3, \
 193                                                    const uint8_t *src4, \
 194                                                    int dst_stride,      \
 195                                                    int src_stride1,     \
 196                                                    int src_stride2,     \
 197                                                    int src_stride3,     \
 198                                                    int src_stride4,     \
 199                                                    int h)               \
 200 {                                                                       \
 201     OPNAME ## _no_rnd_pixels8_l4_8(dst, src1, src2, src3, src4,         \
 202                                    dst_stride, src_stride1,             \
 203                                    src_stride2, src_stride3,            \
 204                                    src_stride4, h);                     \
 205     OPNAME ## _no_rnd_pixels8_l4_8(dst  + 8,                            \
 206                                    src1 + 8, src2 + 8,                  \
 207                                    src3 + 8, src4 + 8,                  \
 208                                    dst_stride, src_stride1,             \
 209                                    src_stride2, src_stride3,            \
 210                                    src_stride4, h);                     \
 211 }                                                                       \
 212
 213 #define op_avg(a, b) a = rnd_avg32(a, b)
 214 #define op_put(a, b) a = b
 215 #define put_no_rnd_pixels8_8_c put_pixels8_8_c
 216 PIXOP2(avg, op_avg)
 217 PIXOP2(put, op_put)
 218 #undef op_avg
 219 #undef op_put