git.sesse.net Git - ffmpeg/blob - tests/checkasm/v210enc.c

   1 /*
   2  * Copyright (c) 2015 Henrik Gramner
   3  *
   4  * This file is part of FFmpeg.
   5  *
   6  * FFmpeg is free software; you can redistribute it and/or modify
   7  * it under the terms of the GNU General Public License as published by
   8  * the Free Software Foundation; either version 2 of the License, or
   9  * (at your option) any later version.
  10  *
  11  * FFmpeg 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
  14  * GNU General Public License for more details.
  15  *
  16  * You should have received a copy of the GNU General Public License along
  17  * with FFmpeg; if not, write to the Free Software Foundation, Inc.,
  18  * 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA.
  19  */
  20
  21 #include <string.h>
  22 #include "checkasm.h"
  23 #include "libavcodec/v210enc.h"
  24 #include "libavutil/common.h"
  25 #include "libavutil/internal.h"
  26 #include "libavutil/intreadwrite.h"
  27 #include "libavutil/mem_internal.h"
  28
  29 #define BUF_SIZE 512
  30
  31 #define randomize_buffers(mask)                        \
  32     do {                                               \
  33         int i, size = sizeof(*y0);                     \
  34         for (i = 0; i < BUF_SIZE; i += 4 / size) {     \
  35             uint32_t r = rnd() & mask;                 \
  36             AV_WN32A(y0 + i, r);                       \
  37             AV_WN32A(y1 + i, r);                       \
  38         }                                              \
  39         for (i = 0; i < BUF_SIZE / 2; i += 4 / size) { \
  40             uint32_t r = rnd() & mask;                 \
  41             AV_WN32A(u0 + i, r);                       \
  42             AV_WN32A(u1 + i, r);                       \
  43             r = rnd() & mask;                          \
  44             AV_WN32A(v0 + i, r);                       \
  45             AV_WN32A(v1 + i, r);                       \
  46         }                                              \
  47         for (i = 0; i < width * 8 / 3; i += 4) {       \
  48             uint32_t r = rnd();                        \
  49             AV_WN32A(dst0 + i, r);                     \
  50             AV_WN32A(dst1 + i, r);                     \
  51         }                                              \
  52     } while (0)
  53
  54 #define check_pack_line(type, mask)                                                \
  55     do {                                                                           \
  56         LOCAL_ALIGNED_16(type, y0, [BUF_SIZE]);                                    \
  57         LOCAL_ALIGNED_16(type, y1, [BUF_SIZE]);                                    \
  58         LOCAL_ALIGNED_16(type, u0, [BUF_SIZE / 2]);                                \
  59         LOCAL_ALIGNED_16(type, u1, [BUF_SIZE / 2]);                                \
  60         LOCAL_ALIGNED_16(type, v0, [BUF_SIZE / 2]);                                \
  61         LOCAL_ALIGNED_16(type, v1, [BUF_SIZE / 2]);                                \
  62         LOCAL_ALIGNED_16(uint8_t, dst0, [BUF_SIZE * 8 / 3]);                       \
  63         LOCAL_ALIGNED_16(uint8_t, dst1, [BUF_SIZE * 8 / 3]);                       \
  64                                                                                    \
  65         declare_func(void, const type * y, const type * u, const type * v,         \
  66                      uint8_t * dst, ptrdiff_t width);                              \
  67         ptrdiff_t width, step = 12 / sizeof(type);                                 \
  68                                                                                    \
  69         for (width = step; width < BUF_SIZE - 15; width += step) {                 \
  70             int y_offset  = rnd() & 15;                                            \
  71             int uv_offset = y_offset / 2;                                          \
  72             randomize_buffers(mask);                                               \
  73             call_ref(y0 + y_offset, u0 + uv_offset, v0 + uv_offset, dst0, width);  \
  74             call_new(y1 + y_offset, u1 + uv_offset, v1 + uv_offset, dst1, width);  \
  75             if (memcmp(y0, y1, BUF_SIZE) || memcmp(u0, u1, BUF_SIZE / 2) ||        \
  76                 memcmp(v0, v1, BUF_SIZE / 2) || memcmp(dst0, dst1, width * 8 / 3)) \
  77                 fail();                                                            \
  78             bench_new(y1 + y_offset, u1 + uv_offset, v1 + uv_offset, dst1, width); \
  79         }                                                                          \
  80     } while (0)
  81
  82 void checkasm_check_v210enc(void)
  83 {
  84     V210EncContext h;
  85
  86     ff_v210enc_init(&h);
  87
  88     if (check_func(h.pack_line_8, "v210_planar_pack_8"))
  89         check_pack_line(uint8_t, 0xffffffff);
  90
  91     if (check_func(h.pack_line_10, "v210_planar_pack_10"))
  92         check_pack_line(uint16_t, 0x03ff03ff);
  93
  94     report("planar_pack");
  95 }