1/*
2 *  Copyright (c) 2017 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 "./vp9_rtcd.h"
12#include "./vpx_dsp_rtcd.h"
13#include "./vpx_scale_rtcd.h"
14#include "vp9/common/vp9_blockd.h"
15#include "vpx_dsp/vpx_filter.h"
16#include "vpx_scale/yv12config.h"
17
18void vp9_scale_and_extend_frame_c(const YV12_BUFFER_CONFIG *src,
19                                  YV12_BUFFER_CONFIG *dst,
20                                  INTERP_FILTER filter_type, int phase_scaler) {
21  const int src_w = src->y_crop_width;
22  const int src_h = src->y_crop_height;
23  const uint8_t *const srcs[3] = { src->y_buffer, src->u_buffer,
24                                   src->v_buffer };
25  const int src_strides[3] = { src->y_stride, src->uv_stride, src->uv_stride };
26  uint8_t *const dsts[3] = { dst->y_buffer, dst->u_buffer, dst->v_buffer };
27  const int dst_strides[3] = { dst->y_stride, dst->uv_stride, dst->uv_stride };
28  const InterpKernel *const kernel = vp9_filter_kernels[filter_type];
29  int x, y, i;
30
31#if HAVE_SSSE3 || HAVE_NEON
32  // TODO(linfengz): The 4:3 specialized C code is disabled by default since
33  // it's much slower than the general version which calls vpx_scaled_2d() even
34  // if vpx_scaled_2d() is not optimized. It will only be enabled as a reference
35  // for the platforms which have faster optimization.
36  if (4 * dst->y_crop_width == 3 * src_w &&
37      4 * dst->y_crop_height == 3 * src_h) {
38    // Specialize 4 to 3 scaling.
39    // Example pixel locations.
40    // (O: Original pixel. S: Scaled pixel. X: Overlapped pixel.)
41    //      phase_scaler = 0               |      phase_scaler = 8
42    //                                     |
43    //      X     O S   O   S O     X      |      O     O     O     O     O
44    //                                     |
45    //                                     |
46    //                                     |         S       S       S
47    //                                     |
48    //                                     |
49    //      O     O     O     O     O      |      O     O     O     O     O
50    //                                     |
51    //      S       S       S       S      |
52    //                                     |
53    //                                     |
54    //                                     |         S       S       S
55    //      O     O     O     O     O      |      O     O     O     O     O
56    //                                     |
57    //                                     |
58    //                                     |
59    //      S       S       S       S      |
60    //                                     |
61    //      O     O     O     O     O      |      O     O     O     O     O
62    //                                     |         S       S       S
63    //                                     |
64    //                                     |
65    //                                     |
66    //                                     |
67    //      X     O S   O   S O     X      |      O     O     O     O     O
68
69    const int dst_ws[3] = { dst->y_crop_width, dst->uv_crop_width,
70                            dst->uv_crop_width };
71    const int dst_hs[3] = { dst->y_crop_height, dst->uv_crop_height,
72                            dst->uv_crop_height };
73    for (i = 0; i < MAX_MB_PLANE; ++i) {
74      const int dst_w = dst_ws[i];
75      const int dst_h = dst_hs[i];
76      const int src_stride = src_strides[i];
77      const int dst_stride = dst_strides[i];
78      for (y = 0; y < dst_h; y += 3) {
79        for (x = 0; x < dst_w; x += 3) {
80          const uint8_t *src_ptr = srcs[i] + 4 * y / 3 * src_stride + 4 * x / 3;
81          uint8_t *dst_ptr = dsts[i] + y * dst_stride + x;
82
83          // Must call c function because its optimization doesn't support 3x3.
84          vpx_scaled_2d_c(src_ptr, src_stride, dst_ptr, dst_stride, kernel,
85                          phase_scaler, 64 / 3, phase_scaler, 64 / 3, 3, 3);
86        }
87      }
88    }
89  } else
90#endif
91  {
92    const int dst_w = dst->y_crop_width;
93    const int dst_h = dst->y_crop_height;
94    for (i = 0; i < MAX_MB_PLANE; ++i) {
95      const int factor = (i == 0 || i == 3 ? 1 : 2);
96      const int src_stride = src_strides[i];
97      const int dst_stride = dst_strides[i];
98      for (y = 0; y < dst_h; y += 16) {
99        const int y_q4 = y * (16 / factor) * src_h / dst_h + phase_scaler;
100        for (x = 0; x < dst_w; x += 16) {
101          const int x_q4 = x * (16 / factor) * src_w / dst_w + phase_scaler;
102          const uint8_t *src_ptr = srcs[i] +
103                                   (y / factor) * src_h / dst_h * src_stride +
104                                   (x / factor) * src_w / dst_w;
105          uint8_t *dst_ptr = dsts[i] + (y / factor) * dst_stride + (x / factor);
106
107          vpx_scaled_2d(src_ptr, src_stride, dst_ptr, dst_stride, kernel,
108                        x_q4 & 0xf, 16 * src_w / dst_w, y_q4 & 0xf,
109                        16 * src_h / dst_h, 16 / factor, 16 / factor);
110        }
111      }
112    }
113  }
114
115  vpx_extend_frame_borders(dst);
116}
117