git.sesse.net Git - pkanalytics/blob - video_widget.cpp

   1 #define GL_GLEXT_PROTOTYPES
   2
   3 #include "video_widget.h"
   4
   5 #include <assert.h>
   6 #include <pthread.h>
   7 #include <stdint.h>
   8 #include <stdio.h>
   9 #include <stdlib.h>
  10 #include <string.h>
  11 #include <sys/stat.h>
  12 #include <unistd.h>
  13
  14 extern "C" {
  15 #include <libavcodec/avcodec.h>
  16 #include <libavformat/avformat.h>
  17 #include <libavutil/avutil.h>
  18 #include <libavutil/error.h>
  19 #include <libavutil/frame.h>
  20 #include <libavutil/imgutils.h>
  21 #include <libavutil/mem.h>
  22 #include <libavutil/pixfmt.h>
  23 #include <libavutil/opt.h>
  24 #include <libswscale/swscale.h>
  25 }
  26
  27 #include <chrono>
  28 #include <cstdint>
  29 #include <utility>
  30 #include <vector>
  31 #include <unordered_set>
  32
  33 #include "post_to_main_thread.h"
  34
  35 #include <QOpenGLFunctions>
  36 #include <QWheelEvent>
  37 #include <QMouseEvent>
  38 #include <QMouseEvent>
  39 #include <QHBoxLayout>
  40
  41 #define BUFFER_OFFSET(i) ((char *)nullptr + (i))
  42
  43 using namespace std;
  44 using namespace std::chrono;
  45
  46 namespace {
  47
  48 bool is_full_range(const AVPixFmtDescriptor *desc)
  49 {
  50         // This is horrible, but there's no better way that I know of.
  51         return (strchr(desc->name, 'j') != nullptr);
  52 }
  53
  54 AVPixelFormat decide_dst_format(AVPixelFormat src_format)
  55 {
  56         // If this is a non-Y'CbCr format, just convert to 4:4:4 Y'CbCr
  57         // and be done with it. It's too strange to spend a lot of time on.
  58         // (Let's hope there's no alpha.)
  59         const AVPixFmtDescriptor *src_desc = av_pix_fmt_desc_get(src_format);
  60         if (src_desc == nullptr ||
  61             src_desc->nb_components != 3 ||
  62             (src_desc->flags & AV_PIX_FMT_FLAG_RGB)) {
  63                 return AV_PIX_FMT_YUV444P;
  64         }
  65
  66         // The best for us would be Cb and Cr together if possible,
  67         // but FFmpeg doesn't support that except in the special case of
  68         // NV12, so we need to go to planar even for the case of NV12.
  69         // Thus, look for the closest (but no worse) 8-bit planar Y'CbCr format
  70         // that matches in color range. (This will also include the case of
  71         // the source format already being acceptable.)
  72         bool src_full_range = is_full_range(src_desc);
  73         const char *best_format = "yuv444p";
  74         unsigned best_score = numeric_limits<unsigned>::max();
  75         for (const AVPixFmtDescriptor *desc = av_pix_fmt_desc_next(nullptr);
  76              desc;
  77              desc = av_pix_fmt_desc_next(desc)) {
  78                 // Find planar Y'CbCr formats only.
  79                 if (desc->nb_components != 3) continue;
  80                 if (desc->flags & AV_PIX_FMT_FLAG_RGB) continue;
  81                 if (!(desc->flags & AV_PIX_FMT_FLAG_PLANAR)) continue;
  82                 if (desc->comp[0].plane != 0 ||
  83                     desc->comp[1].plane != 1 ||
  84                     desc->comp[2].plane != 2) continue;
  85
  86                 // 8-bit formats only.
  87                 if (desc->flags & AV_PIX_FMT_FLAG_BE) continue;
  88                 if (desc->comp[0].depth != 8) continue;
  89
  90                 // Same or better chroma resolution only.
  91                 int chroma_w_diff = src_desc->log2_chroma_w - desc->log2_chroma_w;
  92                 int chroma_h_diff = src_desc->log2_chroma_h - desc->log2_chroma_h;
  93                 if (chroma_w_diff < 0 || chroma_h_diff < 0)
  94                         continue;
  95
  96                 // Matching full/limited range only.
  97                 if (is_full_range(desc) != src_full_range)
  98                         continue;
  99
 100                 // Pick something with as little excess chroma resolution as possible.
 101                 unsigned score = (1 << (chroma_w_diff)) << chroma_h_diff;
 102                 if (score < best_score) {
 103                         best_score = score;
 104                         best_format = desc->name;
 105                 }
 106         }
 107         return av_get_pix_fmt(best_format);
 108 }
 109
 110 }  // namespace
 111
 112 bool VideoWidget::process_queued_commands(AVFormatContext *format_ctx, AVCodecContext *video_codec_ctx, int video_stream_index, bool *seeked)
 113 {
 114         // Process any queued commands from other threads.
 115         vector<QueuedCommand> commands;
 116         {
 117                 lock_guard<mutex> lock(queue_mu);
 118                 swap(commands, command_queue);
 119         }
 120
 121         for (const QueuedCommand &cmd : commands) {
 122                 switch (cmd.command) {
 123                 case QueuedCommand::PAUSE:
 124                         paused = true;
 125                         break;
 126                 case QueuedCommand::RESUME:
 127                         paused = false;
 128                         pts_origin = last_pts;
 129                         start = next_frame_start = steady_clock::now();
 130                         break;
 131                 case QueuedCommand::SEEK:
 132                 case QueuedCommand::SEEK_ABSOLUTE:
 133                         // Dealt with below.
 134                         break;
 135                 }
 136         }
 137
 138         // Combine all seeks into one big one. (There are edge cases where this is probably
 139         // subtly wrong, but we'll live with it.)
 140         int64_t base_pts = last_pts;
 141         int64_t relative_seek_ms = 0;
 142         int64_t relative_seek_frames = 0;
 143         for (const QueuedCommand &cmd : commands) {
 144                 if (cmd.command == QueuedCommand::SEEK) {
 145                         relative_seek_ms += cmd.relative_seek_ms;
 146                         relative_seek_frames += cmd.relative_seek_frames;
 147                 } else if (cmd.command == QueuedCommand::SEEK_ABSOLUTE) {
 148                         base_pts = av_rescale_q(cmd.seek_ms, AVRational{ 1, 1000 }, video_timebase);
 149                         relative_seek_ms = 0;
 150                         relative_seek_frames = 0;
 151                 }
 152         }
 153         int64_t relative_seek_pts = av_rescale_q(relative_seek_ms, AVRational{ 1, 1000 }, video_timebase);
 154         if (relative_seek_ms != 0 && relative_seek_pts == 0) {
 155                 // Just to be sure rounding errors don't move us into nothingness.
 156                 relative_seek_pts = (relative_seek_ms > 0) ? 1 : -1;
 157         }
 158         int64_t goal_pts = base_pts + relative_seek_pts;
 159         if (goal_pts != last_pts || relative_seek_frames < 0) {
 160                 avcodec_flush_buffers(video_codec_ctx);
 161                 queued_frames.clear();
 162
 163                 // Seek to the last keyframe before this point.
 164                 int64_t seek_pts = goal_pts;
 165                 if (relative_seek_frames < 0) {
 166                         // If we're frame-skipping backwards, add 100 ms of slop for each frame
 167                         // so we're fairly certain we are able to see the ones we want.
 168                         seek_pts -= av_rescale_q(-relative_seek_frames, AVRational{ 1, 10 }, video_timebase);
 169                 }
 170                 av_seek_frame(format_ctx, video_stream_index, seek_pts, AVSEEK_FLAG_BACKWARD);
 171
 172                 // Decode frames until EOF, or until we see something past our seek point.
 173                 std::deque<AVFrameWithDeleter> queue;
 174                 for ( ;; ) {
 175                         bool error = false;
 176                         AVFrameWithDeleter frame = decode_frame(format_ctx, video_codec_ctx,
 177                                 pathname, video_stream_index, &error);
 178                         if (frame == nullptr || error) {
 179                                 break;
 180                         }
 181
 182                         int64_t frame_pts = frame->pts;
 183                         if (relative_seek_frames < 0) {
 184                                 // Buffer this frame; don't display it unless we know it's the Nth-latest.
 185                                 queue.push_back(std::move(frame));
 186                                 if (queue.size() > uint64_t(-relative_seek_frames) + 1) {
 187                                         queue.pop_front();
 188                                 }
 189                         }
 190                         if (frame_pts >= goal_pts) {
 191                                 if (relative_seek_frames > 0) {
 192                                         --relative_seek_frames;
 193                                 } else {
 194                                         if (relative_seek_frames < 0) {
 195                                                 // Hope we have the right amount.
 196                                                 // The rest will remain in the queue for when we play forward again.
 197                                                 frame = std::move(queue.front());
 198                                                 queue.pop_front();
 199                                                 queued_frames = std::move(queue);
 200                                         }
 201                                         video_window->set_current_frame(make_video_frame(frame.get()));
 202                                         update();
 203                                         store_pts(frame->pts);
 204                                         break;
 205                                 }
 206                         }
 207                 }
 208
 209                 // NOTE: We keep pause status as-is.
 210
 211                 pts_origin = last_pts;
 212                 start = next_frame_start = last_frame = steady_clock::now();
 213                 if (seeked) {
 214                         *seeked = true;
 215                 }
 216         } else if (relative_seek_frames > 0) {
 217                 // The base PTS is fine, we only need to skip a few frames forwards.
 218                 while (relative_seek_frames > 1) {
 219                         // Eat a frame (ignore errors).
 220                         bool error;
 221                         decode_frame(format_ctx, video_codec_ctx, pathname, video_stream_index, &error);
 222                         --relative_seek_frames;
 223                 }
 224
 225                 // Display the last one.
 226                 bool error;
 227                 AVFrameWithDeleter frame = decode_frame(format_ctx, video_codec_ctx,
 228                         pathname, video_stream_index, &error);
 229                 if (frame == nullptr || error) {
 230                         return true;
 231                 }
 232                 video_window->set_current_frame(make_video_frame(frame.get()));
 233                 update();
 234                 store_pts(frame->pts);
 235         }
 236         return false;
 237 }
 238
 239 VideoWidget::VideoWidget(QWidget *parent)
 240         : QWidget(parent),
 241           video_window(new VideoWindow(this)) {
 242         setLayout(new QHBoxLayout);
 243         layout()->addWidget(QWidget::createWindowContainer(video_window));
 244
 245         // ...
 246         connect(video_window, &VideoWindow::mouse_wheel, this, &VideoWidget::wheelEvent);
 247         connect(video_window, &VideoWindow::mouse_pressed, this, &VideoWidget::mousePressEvent);
 248         connect(video_window, &VideoWindow::mouse_released, this, &VideoWidget::mouseReleaseEvent);
 249         connect(video_window, &VideoWindow::mouse_moved, this, &VideoWidget::mouseMoveEvent);
 250 }
 251
 252 VideoWidget::~VideoWidget()
 253 {
 254         stop();
 255
 256         // Qt will delete video_window for us after we're gone,
 257         // so make sure its destructor does not try to mess with
 258         // our freelist. The actual freelist frames will leak.
 259         video_window->set_current_frame(nullptr);
 260 }
 261
 262 GLuint compile_shader(const string &shader_src, GLenum type)
 263 {
 264         GLuint obj = glCreateShader(type);
 265         const GLchar* source[] = { shader_src.data() };
 266         const GLint length[] = { (GLint)shader_src.size() };
 267         glShaderSource(obj, 1, source, length);
 268         glCompileShader(obj);
 269
 270         GLchar info_log[4096];
 271         GLsizei log_length = sizeof(info_log) - 1;
 272         glGetShaderInfoLog(obj, log_length, &log_length, info_log);
 273         info_log[log_length] = 0;
 274         if (strlen(info_log) > 0) {
 275                 fprintf(stderr, "Shader compile log: %s\n", info_log);
 276         }
 277
 278         GLint status;
 279         glGetShaderiv(obj, GL_COMPILE_STATUS, &status);
 280         if (status == GL_FALSE) {
 281                 // Add some line numbers to easier identify compile errors.
 282                 string src_with_lines = "/*   1 */ ";
 283                 size_t lineno = 1;
 284                 for (char ch : shader_src) {
 285                         src_with_lines.push_back(ch);
 286                         if (ch == '\n') {
 287                                 char buf[32];
 288                                 snprintf(buf, sizeof(buf), "/* %3zu */ ", ++lineno);
 289                                 src_with_lines += buf;
 290                         }
 291                 }
 292
 293                 fprintf(stderr, "Failed to compile shader:\n%s\n", src_with_lines.c_str());
 294                 exit(1);
 295         }
 296
 297         return obj;
 298 }
 299
 300 void VideoWindow::initializeGL()
 301 {
 302         glDisable(GL_BLEND);
 303         glDisable(GL_DEPTH_TEST);
 304         glDepthMask(GL_FALSE);
 305         glCreateTextures(GL_TEXTURE_2D, 3, tex);
 306
 307         ycbcr_vertex_shader = compile_shader(R"(
 308 #version 460 core
 309
 310 layout(location = 0) in vec2 position;
 311 layout(location = 1) in vec2 texcoord;
 312 out vec2 tc;
 313
 314 void main()
 315 {
 316         // The result of glOrtho(0.0, 1.0, 0.0, 1.0, 0.0, 1.0) is:
 317         //
 318         //   2.000  0.000  0.000 -1.000
 319         //   0.000  2.000  0.000 -1.000
 320         //   0.000  0.000 -2.000 -1.000
 321         //   0.000  0.000  0.000  1.000
 322         gl_Position = vec4(2.0 * position.x - 1.0, 2.0 * position.y - 1.0, -1.0, 1.0);
 323         tc = texcoord;
 324         tc.y = 1.0f - tc.y;
 325 }
 326 )", GL_VERTEX_SHADER);
 327         ycbcr_fragment_shader = compile_shader(R"(
 328 #version 460 core
 329
 330 layout(location = 0) uniform sampler2D tex_y;
 331 layout(location = 1) uniform sampler2D tex_cb;
 332 layout(location = 2) uniform sampler2D tex_cr;
 333 layout(location = 3) uniform vec2 cbcr_offset;
 334
 335 in vec2 tc;
 336 out vec4 FragColor;
 337
 338 // Computed statically by Movit, for limited-range BT.709.
 339 // (We don't check whether the input could be BT.601 or BT.2020 currently, or full-range)
 340 const mat3 inv_ycbcr_matrix = mat3(
 341         1.16438f, 1.16438f, 1.16438f,
 342         0.0f, -0.21325f, 2.11240f,
 343         1.79274f, -0.53291f, 0.0f
 344 );
 345
 346 void main()
 347 {
 348         if (tc.x < 0.0 || tc.x > 1.0 || tc.y < 0.0 || tc.y > 1.0) {
 349                 FragColor.rgba = vec4(0.0f, 0.0f, 0.0f, 1.0f);
 350                 return;
 351         }
 352
 353         vec3 ycbcr;
 354         ycbcr.r = texture(tex_y, tc).r;
 355         ycbcr.g = texture(tex_cb, tc + cbcr_offset).r;
 356         ycbcr.b = texture(tex_cr, tc + cbcr_offset).r;
 357         ycbcr -= vec3(16.0f / 255.0f, 128.0f / 255.0f, 128.0f / 255.0f);
 358         FragColor.rgb = inv_ycbcr_matrix * ycbcr;
 359         FragColor.a = 1.0f;
 360 }
 361 )", GL_FRAGMENT_SHADER);
 362         ycbcr_program = glCreateProgram();
 363         glAttachShader(ycbcr_program, ycbcr_vertex_shader);
 364         glAttachShader(ycbcr_program, ycbcr_fragment_shader);
 365         glLinkProgram(ycbcr_program);
 366
 367         GLint success;
 368         glGetProgramiv(ycbcr_program, GL_LINK_STATUS, &success);
 369         if (success == GL_FALSE) {
 370                 GLchar error_log[1024] = {0};
 371                 glGetProgramInfoLog(ycbcr_program, 1024, nullptr, error_log);
 372                 fprintf(stderr, "Error linking program: %s\n", error_log);
 373                 exit(1);
 374         }
 375
 376         glCreateSamplers(1, &bilinear_sampler);
 377         glSamplerParameteri(bilinear_sampler, GL_TEXTURE_MIN_FILTER, GL_LINEAR_MIPMAP_NEAREST);
 378         glSamplerParameteri(bilinear_sampler, GL_TEXTURE_MAG_FILTER, GL_LINEAR);
 379         glSamplerParameteri(bilinear_sampler, GL_TEXTURE_WRAP_S, GL_CLAMP_TO_EDGE);
 380         glSamplerParameteri(bilinear_sampler, GL_TEXTURE_WRAP_T, GL_CLAMP_TO_EDGE);
 381 }
 382
 383 void VideoWindow::resizeGL(int w, int h)
 384 {
 385         glViewport(0, 0, w, h);
 386         display_aspect = double(w) / h;
 387 }
 388
 389 int num_levels(GLuint width, GLuint height)
 390 {
 391         int levels = 1;
 392         while (width > 1 || height > 1) {
 393                 width = max(width / 2, 1u);
 394                 height = max(height / 2, 1u);
 395                 ++levels;
 396         }
 397         return levels;
 398 }
 399
 400 void VideoWindow::paintGL()
 401 {
 402         std::shared_ptr<VideoWidget::Frame> frame;
 403         {
 404                 lock_guard lock(current_frame_mu);
 405                 frame = current_frame;
 406         }
 407         if (frame == nullptr) {
 408                 glClear(GL_COLOR_BUFFER_BIT);
 409                 return;
 410         }
 411
 412         glUseProgram(ycbcr_program);
 413         if (frame->width != last_width || frame->height != last_height) {
 414                 glTextureStorage2D(tex[0], num_levels(frame->width, frame->height), GL_R8, frame->width, frame->height);
 415         }
 416         if (frame->chroma_width != last_chroma_width || frame->chroma_height != last_chroma_height) {
 417                 for (GLuint num : { tex[1], tex[2] }) {
 418                         glTextureStorage2D(num, num_levels(frame->chroma_width, frame->chroma_height), GL_R8, frame->chroma_width, frame->chroma_height);
 419                 }
 420         }
 421
 422         glBindBuffer(GL_PIXEL_UNPACK_BUFFER, frame->pbo);
 423
 424         if (frame->need_flush_len > 0) {
 425                 glFlushMappedNamedBufferRange(frame->pbo, 0, frame->need_flush_len);
 426                 frame->need_flush_len = 0;
 427         }
 428
 429         glPixelStorei(GL_UNPACK_ROW_LENGTH, 0);
 430         glTextureSubImage2D(tex[0], 0, 0, 0, frame->width, frame->height, GL_RED, GL_UNSIGNED_BYTE, BUFFER_OFFSET(0));
 431         glGenerateTextureMipmap(tex[0]);
 432
 433         glTextureSubImage2D(tex[1], 0, 0, 0, frame->chroma_width, frame->chroma_height, GL_RED, GL_UNSIGNED_BYTE, BUFFER_OFFSET(frame->width * frame->height));
 434         glGenerateTextureMipmap(tex[1]);
 435
 436         glTextureSubImage2D(tex[2], 0, 0, 0, frame->chroma_width, frame->chroma_height, GL_RED, GL_UNSIGNED_BYTE, BUFFER_OFFSET(frame->width * frame->height + frame->chroma_width * frame->chroma_height));
 437         glGenerateTextureMipmap(tex[2]);
 438
 439         glBindBuffer(GL_PIXEL_UNPACK_BUFFER, 0);
 440
 441         glBindTextureUnit(0, tex[0]);
 442         glBindTextureUnit(1, tex[1]);
 443         glBindTextureUnit(2, tex[2]);
 444         glBindSampler(0, bilinear_sampler);
 445         glBindSampler(1, bilinear_sampler);
 446         glBindSampler(2, bilinear_sampler);
 447         glProgramUniform1i(ycbcr_program, 0, 0);
 448         glProgramUniform1i(ycbcr_program, 1, 1);
 449         glProgramUniform1i(ycbcr_program, 2, 2);
 450         glProgramUniform2f(ycbcr_program, 3, cbcr_offset[0], -cbcr_offset[1]);
 451
 452         float tx1 = 0.0f;
 453         float tx2 = 1.0f;
 454         float ty1 = 0.0f;
 455         float ty2 = 1.0f;
 456
 457         double video_aspect = double(frame->width) / frame->height;
 458         if (display_aspect > video_aspect) {
 459                 double extra_width = frame->height * display_aspect - frame->width;
 460                 tx1 = -0.5 * extra_width / frame->width;
 461                 tx2 = 1.0 + 0.5 * extra_width / frame->width;
 462         } else if (display_aspect < video_aspect) {
 463                 double extra_height = frame->width / display_aspect - frame->height;
 464                 ty1 = -0.5 * extra_height / frame->height;
 465                 ty2 = 1.0 + 0.5 * extra_height / frame->height;
 466         }
 467
 468         glBegin(GL_QUADS);
 469
 470         // (0,0)
 471         glVertexAttrib2f(1, tx1, ty1);
 472         glVertex2f(zoom_matrix[2 * 3 + 0], zoom_matrix[2 * 3 + 1]);
 473
 474         // (0,1)
 475         glVertexAttrib2f(1, tx1, ty2);
 476         glVertex2f(zoom_matrix[1 * 3 + 0] + zoom_matrix[2 * 3 + 0], zoom_matrix[1 * 3 + 1] + zoom_matrix[2 * 3 + 1]);
 477
 478         // (1,1)
 479         glVertexAttrib2f(1, tx2, ty2);
 480         glVertex2f(zoom_matrix[0 * 3 + 0] + zoom_matrix[1 * 3 + 0] + zoom_matrix[2 * 3 + 0],
 481                    zoom_matrix[1 * 3 + 0] + zoom_matrix[1 * 3 + 1] + zoom_matrix[2 * 3 + 1]);
 482
 483         // (1,0)
 484         glVertexAttrib2f(1, tx2, ty1);
 485         glVertex2f(zoom_matrix[0 * 3 + 0] + zoom_matrix[2 * 3 + 0],
 486                    zoom_matrix[1 * 3 + 0] + zoom_matrix[2 * 3 + 1]);
 487
 488         glEnd();
 489 }
 490
 491 void VideoWindow::set_current_frame(shared_ptr<VideoWidget::Frame> new_frame)
 492 {
 493         {
 494                 lock_guard lock(current_frame_mu);
 495                 current_frame = std::move(new_frame);
 496         }
 497         update();
 498 }
 499
 500 void matmul3x3(const double a[9], const double b[9], double res[9])
 501 {
 502         for (int i = 0; i < 3; ++i) {
 503                 for (int j = 0; j < 3; ++j) {
 504                         double sum = 0.0;
 505                         for (int k = 0; k < 3; ++k) {
 506                                 sum += a[i * 3 + k] * b[k * 3 + j];
 507                         }
 508                         res[i * 3 + j] = sum;
 509                 }
 510         }
 511 }
 512
 513 void VideoWidget::wheelEvent(QWheelEvent *event)
 514 {
 515         int delta = event->angleDelta().y();
 516         if (delta == 0) {
 517                 return;
 518         }
 519         double x = event->position().x() / width();
 520         double y = 1.0 - event->position().y() / height();
 521         double zoom = delta > 0 ? pow(1.005, delta) : pow(1/1.005, -delta);
 522
 523         const double inv_translation_matrix[9] = {
 524                 1.0, 0.0, 0.0,
 525                 0.0, 1.0, 0.0,
 526                 -x, -y, 1.0
 527         };
 528         const double scale_matrix[9] = {
 529                 zoom, 0.0, 0.0,
 530                 0.0, zoom, 0.0,
 531                 0.0, 0.0, 1.0
 532         };
 533         const double translation_matrix[9] = {
 534                 1.0, 0.0, 0.0,
 535                 0.0, 1.0, 0.0,
 536                 x, y, 1.0
 537         };
 538         double tmp1[9], tmp2[9];
 539         matmul3x3(zoom_matrix, inv_translation_matrix, tmp1);
 540         matmul3x3(tmp1, scale_matrix, tmp2);
 541         matmul3x3(tmp2, translation_matrix, zoom_matrix);
 542
 543         fixup_zoom_matrix();
 544         video_window->set_zoom_matrix(zoom_matrix);
 545         update();
 546 }
 547
 548 void VideoWidget::mousePressEvent(QMouseEvent *e)
 549 {
 550         if (e->button() == Qt::BackButton) {
 551                 emit mouse_back_clicked();
 552         } else if (e->button() == Qt::ForwardButton) {
 553                 emit mouse_forward_clicked();
 554         } else if (e->button() == Qt::LeftButton) {
 555                 dragging = true;
 556                 last_drag_x = e->position().x();
 557                 last_drag_y = e->position().y();
 558         }
 559 }
 560
 561 void VideoWidget::mouseReleaseEvent(QMouseEvent *e)
 562 {
 563         if (e->button() == Qt::LeftButton) {
 564                 dragging = false;
 565         }
 566 }
 567
 568 void VideoWidget::mouseMoveEvent(QMouseEvent *e)
 569 {
 570         if (!dragging) {
 571                 return;
 572         }
 573         float dx = (e->position().x() - last_drag_x) / width();
 574         float dy = (e->position().y() - last_drag_y) / height();
 575
 576         //zoom_matrix[6] += dx * zoom_matrix[0];
 577         //zoom_matrix[7] += dy * zoom_matrix[4];
 578         zoom_matrix[6] += dx;
 579         zoom_matrix[7] -= dy;
 580         fixup_zoom_matrix();
 581         video_window->set_zoom_matrix(zoom_matrix);
 582
 583         last_drag_x = e->position().x();
 584         last_drag_y = e->position().y();
 585
 586         update();
 587 }
 588
 589 // Normalize the matrix so that we never get skew or similar,
 590 // and also never can zoom or pan too far out.
 591 void VideoWidget::fixup_zoom_matrix()
 592 {
 593         // Correct for any numerical errors (we know the matrix must be orthogonal
 594         // and have zero rotation).
 595         zoom_matrix[4] = zoom_matrix[0];
 596         zoom_matrix[1] = zoom_matrix[2] = zoom_matrix[3] = zoom_matrix[5] = 0.0;
 597         zoom_matrix[8] = 1.0;
 598
 599         // We can't zoom further out than 1:1. (Perhaps it would be nice to
 600         // reuse the last zoom-in point to do this, but the center will have to do
 601         // for now.)
 602         if (zoom_matrix[0] < 1.0) {
 603                 const double zoom = 1.0 / zoom_matrix[0];
 604                 const double inv_translation_matrix[9] = {
 605                         1.0, 0.0, 0.0,
 606                         0.0, 1.0, 0.0,
 607                         -0.5, -0.5, 1.0
 608                 };
 609                 const double scale_matrix[9] = {
 610                         zoom, 0.0, 0.0,
 611                         0.0, zoom, 0.0,
 612                         0.0, 0.0, 1.0
 613                 };
 614                 const double translation_matrix[9] = {
 615                         1.0, 0.0, 0.0,
 616                         0.0, 1.0, 0.0,
 617                         0.5, 0.5, 1.0
 618                 };
 619                 double tmp1[9], tmp2[9];
 620                 matmul3x3(zoom_matrix, inv_translation_matrix, tmp1);
 621                 matmul3x3(tmp1, scale_matrix, tmp2);
 622                 matmul3x3(tmp2, translation_matrix, zoom_matrix);
 623         }
 624
 625         // Looking at the points we'll draw with glVertex2f(), make sure none of them are
 626         // inside the square (which would generally mean we've panned ourselves out-of-bounds).
 627         // We simply adjust the translation, which is possible because we fixed scaling above.
 628         zoom_matrix[6] = min(zoom_matrix[6], 0.0);  // Left side (x=0).
 629         zoom_matrix[7] = min(zoom_matrix[7], 0.0);  // Bottom side (y=0).
 630         zoom_matrix[6] = std::max(zoom_matrix[6], 1.0 - zoom_matrix[0]);  // Right side (x=1).
 631         zoom_matrix[7] = std::max(zoom_matrix[7], 1.0 - zoom_matrix[4]);  // Top side (y=1).
 632 }
 633
 634 bool VideoWidget::open(const string &filename)
 635 {
 636         stop();
 637         internal_rewind();
 638         pathname = filename;
 639         play();
 640
 641         while (running == STARTING) {
 642                 // Poor man's condition variable...
 643                 usleep(10000);
 644                 sched_yield();
 645         }
 646         return (running != VIDEO_FILE_ERROR);
 647 }
 648
 649 void VideoWidget::play()
 650 {
 651         if (running != NOT_RUNNING && running != VIDEO_FILE_ERROR) {
 652                 std::lock_guard<std::mutex> lock(queue_mu);
 653                 command_queue.push_back(QueuedCommand { QueuedCommand::RESUME });
 654                 producer_thread_should_quit.wakeup();
 655                 return;
 656         }
 657         running = STARTING;
 658         producer_thread_should_quit.unquit();
 659         if (producer_thread.joinable()) {
 660                 producer_thread.join();
 661         }
 662         producer_thread = std::thread(&VideoWidget::producer_thread_func, this);
 663 }
 664
 665 void VideoWidget::pause()
 666 {
 667         if (running == NOT_RUNNING || running == VIDEO_FILE_ERROR) {
 668                 return;
 669         }
 670         std::lock_guard<std::mutex> lock(queue_mu);
 671         command_queue.push_back(QueuedCommand { QueuedCommand::PAUSE });
 672         producer_thread_should_quit.wakeup();
 673 }
 674
 675 void VideoWidget::seek(int64_t relative_seek_ms)
 676 {
 677         if (running == NOT_RUNNING || running == VIDEO_FILE_ERROR) {
 678                 return;
 679         }
 680         std::lock_guard<std::mutex> lock(queue_mu);
 681         command_queue.push_back(QueuedCommand { QueuedCommand::SEEK, relative_seek_ms, 0, 0 });
 682         producer_thread_should_quit.wakeup();
 683 }
 684
 685 void VideoWidget::seek_frames(int64_t relative_seek_frames)
 686 {
 687         if (running == NOT_RUNNING || running == VIDEO_FILE_ERROR) {
 688                 return;
 689         }
 690         std::lock_guard<std::mutex> lock(queue_mu);
 691         command_queue.push_back(QueuedCommand { QueuedCommand::SEEK, 0, relative_seek_frames, 0 });
 692         producer_thread_should_quit.wakeup();
 693 }
 694
 695 void VideoWidget::seek_absolute(int64_t position_ms)
 696 {
 697         if (running == NOT_RUNNING || running == VIDEO_FILE_ERROR) {
 698                 return;
 699         }
 700         std::lock_guard<std::mutex> lock(queue_mu);
 701         command_queue.push_back(QueuedCommand { QueuedCommand::SEEK_ABSOLUTE, 0, 0, position_ms });
 702         producer_thread_should_quit.wakeup();
 703 }
 704
 705 void VideoWidget::stop()
 706 {
 707         if (running == NOT_RUNNING || running == VIDEO_FILE_ERROR) {
 708                 return;
 709         }
 710         producer_thread_should_quit.quit();
 711         producer_thread.join();
 712 }
 713
 714 void VideoWidget::producer_thread_func()
 715 {
 716         if (!producer_thread_should_quit.should_quit()) {
 717                 if (!play_video(pathname)) {
 718                         running = VIDEO_FILE_ERROR;
 719                 } else {
 720                         running = NOT_RUNNING;
 721                 }
 722         }
 723 }
 724
 725 void VideoWidget::internal_rewind()
 726 {
 727         pts_origin = last_pts = 0;
 728         last_position = 0;
 729         start = next_frame_start = steady_clock::now();
 730 }
 731
 732 template<AVHWDeviceType type>
 733 AVPixelFormat get_hw_format(AVCodecContext *ctx, const AVPixelFormat *fmt)
 734 {
 735         bool found_config_of_right_type = false;
 736         for (int i = 0;; ++i) {  // Termination condition inside loop.
 737                 const AVCodecHWConfig *config = avcodec_get_hw_config(ctx->codec, i);
 738                 if (config == nullptr) {  // End of list.
 739                         break;
 740                 }
 741                 if (!(config->methods & AV_CODEC_HW_CONFIG_METHOD_HW_DEVICE_CTX) ||
 742                     config->device_type != type) {
 743                         // Not interesting for us.
 744                         continue;
 745                 }
 746
 747                 // We have a config of the right type, but does it actually support
 748                 // the pixel format we want? (Seemingly, FFmpeg's way of signaling errors
 749                 // is to just replace the pixel format with a software-decoded one,
 750                 // such as yuv420p.)
 751                 found_config_of_right_type = true;
 752                 for (const AVPixelFormat *fmt_ptr = fmt; *fmt_ptr != -1; ++fmt_ptr) {
 753                         if (config->pix_fmt == *fmt_ptr) {
 754                                 fprintf(stderr, "Initialized '%s' hardware decoding for codec '%s'.\n",
 755                                         av_hwdevice_get_type_name(type), ctx->codec->name);
 756                                 if (ctx->profile == FF_PROFILE_H264_BASELINE) {
 757                                         fprintf(stderr, "WARNING: Stream claims to be H.264 Baseline, which is generally poorly supported in hardware decoders.\n");
 758                                         fprintf(stderr, "         Consider encoding it as Constrained Baseline, Main or High instead.\n");
 759                                         fprintf(stderr, "         Decoding might fail and fall back to software.\n");
 760                                 }
 761                                 return config->pix_fmt;
 762                         }
 763                 }
 764                 fprintf(stderr, "Decoder '%s' supports only these pixel formats:", ctx->codec->name);
 765                 unordered_set<AVPixelFormat> seen;
 766                 for (const AVPixelFormat *fmt_ptr = fmt; *fmt_ptr != -1; ++fmt_ptr) {
 767                         if (!seen.count(*fmt_ptr)) {
 768                                 fprintf(stderr, " %s", av_get_pix_fmt_name(*fmt_ptr));
 769                                 seen.insert(*fmt_ptr);
 770                         }
 771                 }
 772                 fprintf(stderr, " (wanted %s for hardware acceleration)\n", av_get_pix_fmt_name(config->pix_fmt));
 773
 774         }
 775
 776         if (!found_config_of_right_type) {
 777                 fprintf(stderr, "Decoder '%s' does not support device type '%s'.\n", ctx->codec->name, av_hwdevice_get_type_name(type));
 778         }
 779
 780         // We found no VA-API formats, so take the first software format.
 781         for (const AVPixelFormat *fmt_ptr = fmt; *fmt_ptr != -1; ++fmt_ptr) {
 782                 if ((av_pix_fmt_desc_get(*fmt_ptr)->flags & AV_PIX_FMT_FLAG_HWACCEL) == 0) {
 783                         fprintf(stderr, "Falling back to software format %s.\n", av_get_pix_fmt_name(*fmt_ptr));
 784                         return *fmt_ptr;
 785                 }
 786         }
 787
 788         // Fallback: Just return anything. (Should never really happen.)
 789         return fmt[0];
 790 }
 791
 792 AVFrameWithDeleter VideoWidget::decode_frame(AVFormatContext *format_ctx, AVCodecContext *video_codec_ctx,
 793         const std::string &pathname, int video_stream_index,
 794         bool *error)
 795 {
 796         *error = false;
 797
 798         if (!queued_frames.empty()) {
 799                 AVFrameWithDeleter frame = std::move(queued_frames.front());
 800                 queued_frames.pop_front();
 801                 return frame;
 802         }
 803
 804         // Read packets until we have a frame or there are none left.
 805         bool frame_finished = false;
 806         AVFrameWithDeleter video_avframe = av_frame_alloc_unique();
 807         bool eof = false;
 808         do {
 809                 AVPacket *pkt = av_packet_alloc();
 810                 unique_ptr<AVPacket, decltype(av_packet_unref)*> pkt_cleanup(
 811                         pkt, av_packet_unref);
 812                 pkt->data = nullptr;
 813                 pkt->size = 0;
 814                 if (av_read_frame(format_ctx, pkt) == 0) {
 815                         if (pkt->stream_index == video_stream_index) {
 816                                 if (avcodec_send_packet(video_codec_ctx, pkt) < 0) {
 817                                         fprintf(stderr, "%s: Cannot send packet to video codec.\n", pathname.c_str());
 818                                         *error = true;
 819                                         return AVFrameWithDeleter(nullptr);
 820                                 }
 821                         }
 822                 } else {
 823                         eof = true;  // Or error, but ignore that for the time being.
 824                 }
 825
 826                 // Decode video, if we have a frame.
 827                 int err = avcodec_receive_frame(video_codec_ctx, video_avframe.get());
 828                 if (err == 0) {
 829                         frame_finished = true;
 830                         break;
 831                 } else if (err != AVERROR(EAGAIN)) {
 832                         fprintf(stderr, "%s: Cannot receive frame from video codec.\n", pathname.c_str());
 833                         *error = true;
 834                         return AVFrameWithDeleter(nullptr);
 835                 }
 836         } while (!eof);
 837
 838         if (frame_finished)
 839                 return video_avframe;
 840         else
 841                 return AVFrameWithDeleter(nullptr);
 842 }
 843
 844 int find_stream_index(AVFormatContext *ctx, AVMediaType media_type)
 845 {
 846         for (unsigned i = 0; i < ctx->nb_streams; ++i) {
 847                 if (ctx->streams[i]->codecpar->codec_type == media_type) {
 848                         return i;
 849                 }
 850         }
 851         return -1;
 852 }
 853
 854 steady_clock::time_point compute_frame_start(int64_t frame_pts, int64_t pts_origin, const AVRational &video_timebase, const steady_clock::time_point &origin, double rate)
 855 {
 856         const duration<double> pts((frame_pts - pts_origin) * double(video_timebase.num) / double(video_timebase.den));
 857         return origin + duration_cast<steady_clock::duration>(pts / rate);
 858 }
 859
 860 bool VideoWidget::play_video(const string &pathname)
 861 {
 862         queued_frames.clear();
 863         AVFormatContextWithCloser format_ctx = avformat_open_input_unique(pathname.c_str(), /*fmt=*/nullptr,
 864                 /*options=*/nullptr);
 865         if (format_ctx == nullptr) {
 866                 fprintf(stderr, "%s: Error opening file\n", pathname.c_str());
 867                 return false;
 868         }
 869
 870         if (avformat_find_stream_info(format_ctx.get(), nullptr) < 0) {
 871                 fprintf(stderr, "%s: Error finding stream info\n", pathname.c_str());
 872                 return false;
 873         }
 874
 875         int video_stream_index = find_stream_index(format_ctx.get(), AVMEDIA_TYPE_VIDEO);
 876         if (video_stream_index == -1) {
 877                 fprintf(stderr, "%s: No video stream found\n", pathname.c_str());
 878                 return false;
 879         }
 880
 881         // Open video decoder.
 882         const AVCodecParameters *video_codecpar = format_ctx->streams[video_stream_index]->codecpar;
 883         const AVCodec *video_codec = avcodec_find_decoder(video_codecpar->codec_id);
 884
 885         video_timebase = format_ctx->streams[video_stream_index]->time_base;
 886         AVCodecContextWithDeleter video_codec_ctx = avcodec_alloc_context3_unique(nullptr);
 887         if (avcodec_parameters_to_context(video_codec_ctx.get(), video_codecpar) < 0) {
 888                 fprintf(stderr, "%s: Cannot fill video codec parameters\n", pathname.c_str());
 889                 return false;
 890         }
 891         if (video_codec == nullptr) {
 892                 fprintf(stderr, "%s: Cannot find video decoder\n", pathname.c_str());
 893                 return false;
 894         }
 895
 896         // Seemingly, it's not too easy to make something that just initializes
 897         // “whatever goes”, so we don't get CUDA or VULKAN or whatever here
 898         // without enumerating through several different types.
 899         // VA-API and VDPAU will do for now. We prioritize VDPAU for the
 900         // simple reason that there's a VA-API-via-VDPAU emulation for NVidia
 901         // cards that seems to work, but just hangs when trying to transfer the frame.
 902         //
 903         // Note that we don't actually check codec support beforehand,
 904         // so if you have a low-end VDPAU device but a high-end VA-API device,
 905         // you lose out on the extra codec support from the latter.
 906         AVBufferRef *hw_device_ctx = nullptr;
 907         if (av_hwdevice_ctx_create(&hw_device_ctx, AV_HWDEVICE_TYPE_VDPAU, nullptr, nullptr, 0) >= 0) {
 908                 video_codec_ctx->hw_device_ctx = av_buffer_ref(hw_device_ctx);
 909                 video_codec_ctx->get_format = get_hw_format<AV_HWDEVICE_TYPE_VDPAU>;
 910         } else if (av_hwdevice_ctx_create(&hw_device_ctx, AV_HWDEVICE_TYPE_VAAPI, nullptr, nullptr, 0) >= 0) {
 911                 video_codec_ctx->hw_device_ctx = av_buffer_ref(hw_device_ctx);
 912                 video_codec_ctx->get_format = get_hw_format<AV_HWDEVICE_TYPE_VAAPI>;
 913         } else {
 914                 fprintf(stderr, "Failed to initialize VA-API or VDPAU for FFmpeg acceleration. Decoding video in software.\n");
 915         }
 916
 917         if (avcodec_open2(video_codec_ctx.get(), video_codec, nullptr) < 0) {
 918                 fprintf(stderr, "%s: Cannot open video decoder\n", pathname.c_str());
 919                 return false;
 920         }
 921         unique_ptr<AVCodecContext, decltype(avcodec_close)*> video_codec_ctx_cleanup(
 922                 video_codec_ctx.get(), avcodec_close);
 923
 924         internal_rewind();
 925
 926         running = RUNNING;
 927
 928         // Main loop.
 929         int consecutive_errors = 0;
 930         double rate = 1.0;
 931         while (!producer_thread_should_quit.should_quit()) {
 932                 if (process_queued_commands(format_ctx.get(), video_codec_ctx.get(), video_stream_index, /*seeked=*/nullptr)) {
 933                         return true;
 934                 }
 935                 if (paused) {
 936                         producer_thread_should_quit.sleep_for(hours(1));
 937                         continue;
 938                 }
 939
 940                 bool error;
 941                 AVFrameWithDeleter frame = decode_frame(format_ctx.get(), video_codec_ctx.get(),
 942                         pathname, video_stream_index, &error);
 943                 if (error) {
 944                         if (++consecutive_errors >= 100) {
 945                                 fprintf(stderr, "More than 100 consecutive video frames, aborting playback.\n");
 946                                 return false;
 947                         } else {
 948                                 continue;
 949                         }
 950                 } else {
 951                         consecutive_errors = 0;
 952                 }
 953                 if (frame == nullptr) {
 954                         // EOF.
 955                         return false;
 956                 }
 957
 958                 // Sleep until it's time to present this frame.
 959                 for ( ;; ) {
 960                         if (last_pts == 0 && pts_origin == 0) {
 961                                 pts_origin = frame->pts;
 962                         }
 963                         steady_clock::time_point now = steady_clock::now();
 964                         next_frame_start = compute_frame_start(frame->pts, pts_origin, video_timebase, start, rate);
 965
 966                         if (duration<double>(now - next_frame_start).count() >= 0.1) {
 967                                 // If we don't have enough CPU to keep up, or if we have a live stream
 968                                 // where the initial origin was somehow wrong, we could be behind indefinitely.
 969                                 fprintf(stderr, "%s: Playback %.0f ms behind, resetting time scale\n",
 970                                         pathname.c_str(),
 971                                         1e3 * duration<double>(now - next_frame_start).count());
 972                                 pts_origin = frame->pts;
 973                                 start = next_frame_start = now;
 974                         }
 975                         bool finished_wakeup;
 976                         finished_wakeup = producer_thread_should_quit.sleep_until(next_frame_start);
 977                         if (finished_wakeup) {
 978                                 video_window->set_current_frame(make_video_frame(frame.get()));
 979                                 last_frame = steady_clock::now();
 980                                 update();
 981                                 break;
 982                         } else {
 983                                 if (producer_thread_should_quit.should_quit()) break;
 984
 985                                 bool seeked = false;
 986                                 if (process_queued_commands(format_ctx.get(), video_codec_ctx.get(), video_stream_index, &seeked)) {
 987                                         return true;
 988                                 }
 989
 990                                 if (paused) {
 991                                         // Just paused, so present the frame immediately and then go into deep sleep.
 992                                         video_window->set_current_frame(make_video_frame(frame.get()));
 993                                         last_frame = steady_clock::now();
 994                                         update();
 995                                         break;
 996                                 }
 997
 998                                 // If we just seeked, drop this frame on the floor and be done.
 999                                 if (seeked) {
1000                                         break;
1001                                 }
1002                         }
1003                 }
1004                 store_pts(frame->pts);
1005         }
1006         return true;
1007 }
1008
1009 void VideoWidget::store_pts(int64_t pts)
1010 {
1011         last_pts = pts;
1012         last_position = lrint(pts * double(video_timebase.num) / double(video_timebase.den) * 1000);
1013         post_to_main_thread([this, last_position{last_position.load()}] {
1014                 emit position_changed(last_position);
1015         });
1016 }
1017
1018 // Taken from Movit (see the comment there for explanation)
1019 float compute_chroma_offset(float pos, unsigned subsampling_factor, unsigned resolution)
1020 {
1021         float local_chroma_pos = (0.5 + pos * (subsampling_factor - 1)) / subsampling_factor;
1022         if (fabs(local_chroma_pos - 0.5) < 1e-10) {
1023                 // x + (-0) can be optimized away freely, as opposed to x + 0.
1024                 return -0.0;
1025         } else {
1026                 return (0.5 - local_chroma_pos) / resolution;
1027         }
1028 }
1029
1030 shared_ptr<VideoWidget::Frame> VideoWidget::alloc_frame(unsigned width, unsigned height, unsigned chroma_width, unsigned chroma_height)
1031 {
1032         lock_guard lock(freelist_mu);
1033         for (auto it = frame_freelist.begin(); it != frame_freelist.end(); ++it) {
1034                 if ((*it)->width == width &&
1035                     (*it)->height == height &&
1036                     (*it)->chroma_width == chroma_width &&
1037                     (*it)->chroma_height == chroma_height) {
1038                         Frame *frame = *it;
1039                         frame_freelist.erase(it);
1040                         return shared_ptr<Frame>{frame, free_frame};
1041                 }
1042         }
1043
1044         Frame *frame = new Frame;
1045         frame->owner = this;
1046         frame->width = width;
1047         frame->height = height;
1048         frame->chroma_width = chroma_width;
1049         frame->chroma_height = chroma_height;
1050
1051         size_t len = frame->width * frame->height + 2 * frame->chroma_width * frame->chroma_height;
1052
1053         while (!video_window->isValid()) {
1054                 usleep(100000);
1055         }
1056
1057         // Augh :-)
1058         mutex mu;
1059         condition_variable done_cv;
1060         bool done = false;
1061
1062         post_to_main_thread([this, &frame, len, &done, &mu, &done_cv]{
1063                 video_window->makeCurrent();
1064                 glCreateBuffers(1, &frame->pbo);
1065                 glNamedBufferStorage(frame->pbo, len, nullptr, GL_MAP_WRITE_BIT | GL_MAP_PERSISTENT_BIT);
1066                 frame->data = (uint8_t *)glMapNamedBufferRange(frame->pbo, 0, len, GL_MAP_WRITE_BIT | GL_MAP_FLUSH_EXPLICIT_BIT | GL_MAP_PERSISTENT_BIT);
1067                 video_window->doneCurrent();
1068
1069                 lock_guard lock(mu);
1070                 done = true;
1071                 done_cv.notify_all();
1072         });
1073         {
1074                 unique_lock lock(mu);
1075                 done_cv.wait(lock, [&done]{ return done; });
1076         }
1077
1078         return shared_ptr<Frame>{frame, free_frame};
1079 }
1080
1081 void VideoWidget::free_frame(VideoWidget::Frame *frame)
1082 {
1083         VideoWidget *self = frame->owner;
1084         lock_guard lock(self->freelist_mu);
1085         if (self->frame_freelist.size() >= 16) {
1086                 GLuint pbo = frame->pbo;
1087                 post_to_main_thread([self, pbo]{
1088                         self->video_window->makeCurrent();
1089                         glUnmapNamedBuffer(pbo);
1090                         glDeleteBuffers(1, &pbo);
1091                         self->video_window->doneCurrent();
1092                 });
1093                 delete self->frame_freelist.front();
1094                 self->frame_freelist.pop_front();
1095         }
1096         self->frame_freelist.push_back(frame);
1097 }
1098
1099 shared_ptr<VideoWidget::Frame> VideoWidget::make_video_frame(const AVFrame *frame)
1100 {
1101         AVFrameWithDeleter sw_frame;
1102
1103         if (frame->format == AV_PIX_FMT_VAAPI ||
1104             frame->format == AV_PIX_FMT_VDPAU) {
1105                 // Get the frame down to the CPU. (TODO: See if we can keep it
1106                 // on the GPU all the way, since it will be going up again later.
1107                 // However, this only works if the OpenGL GPU is the same one.)
1108                 sw_frame = av_frame_alloc_unique();
1109                 int err = av_hwframe_transfer_data(sw_frame.get(), frame, 0);
1110                 if (err != 0) {
1111                         fprintf(stderr, "%s: Cannot transfer hardware video frame to software.\n", pathname.c_str());
1112                 } else {
1113                         sw_frame->pts = frame->pts;
1114                         sw_frame->pkt_duration = frame->pkt_duration;
1115                         frame = sw_frame.get();
1116                 }
1117         }
1118
1119         if (sws_ctx == nullptr ||
1120             sws_last_width != frame->width ||
1121             sws_last_height != frame->height ||
1122             sws_last_src_format != frame->format) {
1123                 sws_dst_format = decide_dst_format(AVPixelFormat(frame->format));
1124                 sws_ctx.reset(
1125                         sws_getContext(frame->width, frame->height, AVPixelFormat(frame->format),
1126                                 frame->width, frame->height, sws_dst_format,
1127                                 SWS_BICUBIC, nullptr, nullptr, nullptr));
1128                 sws_last_width = frame->width;
1129                 sws_last_height = frame->height;
1130                 sws_last_src_format = frame->format;
1131         }
1132         if (sws_ctx == nullptr) {
1133                 fprintf(stderr, "Could not create scaler context\n");
1134                 abort();
1135         }
1136
1137         uint8_t *pic_data[4] = { nullptr, nullptr, nullptr, nullptr };
1138         int linesizes[4] = { 0, 0, 0, 0 };
1139         const AVPixFmtDescriptor *desc = av_pix_fmt_desc_get(sws_dst_format);
1140
1141         shared_ptr<Frame> video_frame = alloc_frame(
1142                 frame->width,
1143                 frame->height,
1144                 AV_CEIL_RSHIFT(int(frame->width), desc->log2_chroma_w),
1145                 AV_CEIL_RSHIFT(int(frame->height), desc->log2_chroma_h));
1146
1147         // We always assume left chroma placement for now.
1148         video_window->set_cbcr_offset(
1149                 compute_chroma_offset(0.0f, 1 << desc->log2_chroma_w, video_frame->chroma_width),
1150                 compute_chroma_offset(0.5f, 1 << desc->log2_chroma_h, video_frame->chroma_height)
1151         );
1152
1153         pic_data[0] = video_frame->data;
1154         linesizes[0] = frame->width;
1155
1156         pic_data[1] = pic_data[0] + frame->width * frame->height;
1157         linesizes[1] = video_frame->chroma_width;
1158
1159         pic_data[2] = pic_data[1] + video_frame->chroma_width * video_frame->chroma_height;
1160         linesizes[2] = video_frame->chroma_width;
1161
1162         sws_scale(sws_ctx.get(), frame->data, frame->linesize, 0, frame->height, pic_data, linesizes);
1163
1164         video_frame->need_flush_len = video_frame->width * video_frame->height + 2 * video_frame->chroma_width * video_frame->chroma_height;
1165
1166         return video_frame;
1167 }
1168