diff options
Diffstat (limited to 'src/main.cpp')
| -rw-r--r-- | src/main.cpp | 1485 |
1 files changed, 1088 insertions, 397 deletions
diff --git a/src/main.cpp b/src/main.cpp index 48a4891..98740c3 100644 --- a/src/main.cpp +++ b/src/main.cpp @@ -3,8 +3,10 @@ extern "C" { #include "../include/capture/xcomposite_cuda.h" #include "../include/capture/xcomposite_vaapi.h" #include "../include/capture/kms_vaapi.h" +#include "../include/capture/kms_cuda.h" #include "../include/egl.h" #include "../include/utils.h" +#include "../include/color_conversion.h" } #include <assert.h> @@ -18,6 +20,9 @@ extern "C" { #include <map> #include <signal.h> #include <sys/stat.h> +#include <unistd.h> +#include <sys/wait.h> +#include <libgen.h> #include "../include/sound.hpp" @@ -45,10 +50,19 @@ static const int VIDEO_STREAM_INDEX = 0; static thread_local char av_error_buffer[AV_ERROR_MAX_STRING_SIZE]; -static void monitor_output_callback_print(const XRROutputInfo *output_info, const XRRCrtcInfo *crt_info, const XRRModeInfo *mode_info, void *userdata) { - (void)mode_info; +static void monitor_output_callback_print(const gsr_monitor *monitor, void *userdata) { (void)userdata; - fprintf(stderr, " \"%.*s\" (%dx%d+%d+%d)\n", output_info->nameLen, output_info->name, (int)crt_info->width, (int)crt_info->height, crt_info->x, crt_info->y); + fprintf(stderr, " \"%.*s\" (%dx%d+%d+%d)\n", monitor->name_len, monitor->name, monitor->size.x, monitor->size.y, monitor->pos.x, monitor->pos.y); +} + +typedef struct { + const char *output_name; +} FirstOutputCallback; + +static void get_first_output(const gsr_monitor *monitor, void *userdata) { + FirstOutputCallback *first_output = (FirstOutputCallback*)userdata; + if(!first_output->output_name) + first_output->output_name = strndup(monitor->name, monitor->name_len + 1); } static char* av_error_to_string(int err) { @@ -66,7 +80,10 @@ enum class VideoQuality { enum class VideoCodec { H264, - H265 + HEVC, + HEVC_HDR, + AV1, + AV1_HDR }; enum class AudioCodec { @@ -93,16 +110,38 @@ static int x11_io_error_handler(Display*) { return 0; } +static bool video_codec_is_hdr(VideoCodec video_codec) { + switch(video_codec) { + case VideoCodec::HEVC_HDR: + case VideoCodec::AV1_HDR: + return true; + default: + return false; + } +} + +struct PacketData { + PacketData() {} + PacketData(const PacketData&) = delete; + PacketData& operator=(const PacketData&) = delete; + + ~PacketData() { + av_free(data.data); + } + + AVPacket data; +}; + // |stream| is only required for non-replay mode static void receive_frames(AVCodecContext *av_codec_context, int stream_index, AVStream *stream, int64_t pts, AVFormatContext *av_format_context, double replay_start_time, - std::deque<AVPacket> &frame_data_queue, + std::deque<std::shared_ptr<PacketData>> &frame_data_queue, int replay_buffer_size_secs, bool &frames_erased, - std::mutex &write_output_mutex) { + std::mutex &write_output_mutex, + double paused_time_offset) { for (;;) { - // TODO: Use av_packet_alloc instead because sizeof(av_packet) might not be future proof(?) AVPacket *av_packet = av_packet_alloc(); if(!av_packet) break; @@ -117,14 +156,20 @@ static void receive_frames(AVCodecContext *av_codec_context, int stream_index, A std::lock_guard<std::mutex> lock(write_output_mutex); if(replay_buffer_size_secs != -1) { - double time_now = clock_get_monotonic_seconds(); + // TODO: Preallocate all frames data and use those instead. + // Why are we doing this you ask? there is a new ffmpeg bug that causes cpu usage to increase over time when you have + // packets that are not being free'd until later. So we copy the packet data, free the packet and then reconstruct + // the packet later on when we need it, to keep packets alive only for a short period. + auto new_packet = std::make_shared<PacketData>(); + new_packet->data = *av_packet; + new_packet->data.data = (uint8_t*)av_malloc(av_packet->size); + memcpy(new_packet->data.data, av_packet->data, av_packet->size); + + double time_now = clock_get_monotonic_seconds() - paused_time_offset; double replay_time_elapsed = time_now - replay_start_time; - AVPacket new_pack; - av_packet_move_ref(&new_pack, av_packet); - frame_data_queue.push_back(std::move(new_pack)); + frame_data_queue.push_back(std::move(new_packet)); if(replay_time_elapsed >= replay_buffer_size_secs) { - av_packet_unref(&frame_data_queue.front()); frame_data_queue.pop_front(); frames_erased = true; } @@ -132,22 +177,25 @@ static void receive_frames(AVCodecContext *av_codec_context, int stream_index, A av_packet_rescale_ts(av_packet, av_codec_context->time_base, stream->time_base); av_packet->stream_index = stream->index; // TODO: Is av_interleaved_write_frame needed? - int ret = av_interleaved_write_frame(av_format_context, av_packet); + int ret = av_write_frame(av_format_context, av_packet); if(ret < 0) { fprintf(stderr, "Error: Failed to write frame index %d to muxer, reason: %s (%d)\n", av_packet->stream_index, av_error_to_string(ret), ret); } } + av_packet_free(&av_packet); } else if (res == AVERROR(EAGAIN)) { // we have no packet // fprintf(stderr, "No packet!\n"); + av_packet_free(&av_packet); break; } else if (res == AVERROR_EOF) { // this is the end of the stream + av_packet_free(&av_packet); fprintf(stderr, "End of stream!\n"); break; } else { + av_packet_free(&av_packet); fprintf(stderr, "Unexpected error: %d\n", res); break; } - av_packet_free(&av_packet); } } @@ -171,7 +219,7 @@ static AVCodecID audio_codec_get_id(AudioCodec audio_codec) { return AV_CODEC_ID_AAC; } -static AVSampleFormat audio_codec_get_sample_format(AudioCodec audio_codec, const AVCodec *codec) { +static AVSampleFormat audio_codec_get_sample_format(AudioCodec audio_codec, const AVCodec *codec, bool mix_audio) { switch(audio_codec) { case AudioCodec::AAC: { return AV_SAMPLE_FMT_FLTP; @@ -188,8 +236,14 @@ static AVSampleFormat audio_codec_get_sample_format(AudioCodec audio_codec, cons } } + // Amix only works with float audio + if(mix_audio) + supports_s16 = false; + if(!supports_s16 && !supports_flt) { - fprintf(stderr, "Warning: opus audio codec is chosen but your ffmpeg version does not support s16/flt sample format and performance might be slightly worse. You can either rebuild ffmpeg with libopus instead of the built-in opus, use the flatpak version of gpu screen recorder or record with flac audio codec instead (-ac flac). Falling back to fltp audio sample format instead.\n"); + fprintf(stderr, "Warning: opus audio codec is chosen but your ffmpeg version does not support s16/flt sample format and performance might be slightly worse.\n"); + fprintf(stderr, " You can either rebuild ffmpeg with libopus instead of the built-in opus, use the flatpak version of gpu screen recorder or record with aac audio codec instead (-ac aac).\n"); + fprintf(stderr, " Falling back to fltp audio sample format instead.\n"); } if(supports_s16) @@ -209,12 +263,12 @@ static AVSampleFormat audio_codec_get_sample_format(AudioCodec audio_codec, cons static int64_t audio_codec_get_get_bitrate(AudioCodec audio_codec) { switch(audio_codec) { - case AudioCodec::AAC: return 128000; - case AudioCodec::OPUS: return 96000; - case AudioCodec::FLAC: return 96000; + case AudioCodec::AAC: return 160000; + case AudioCodec::OPUS: return 128000; + case AudioCodec::FLAC: return 128000; } assert(false); - return 96000; + return 128000; } static AudioFormat audio_codec_context_get_audio_format(const AVCodecContext *audio_codec_context) { @@ -237,7 +291,8 @@ static AVSampleFormat audio_format_to_sample_format(const AudioFormat audio_form return AV_SAMPLE_FMT_S16; } -static AVCodecContext* create_audio_codec_context(int fps, AudioCodec audio_codec) { +static AVCodecContext* create_audio_codec_context(int fps, AudioCodec audio_codec, bool mix_audio, int audio_bitrate) { + (void)fps; const AVCodec *codec = avcodec_find_encoder(audio_codec_get_id(audio_codec)); if (!codec) { fprintf(stderr, "Error: Could not find %s audio encoder\n", audio_codec_get_name(audio_codec)); @@ -248,8 +303,8 @@ static AVCodecContext* create_audio_codec_context(int fps, AudioCodec audio_code assert(codec->type == AVMEDIA_TYPE_AUDIO); codec_context->codec_id = codec->id; - codec_context->sample_fmt = audio_codec_get_sample_format(audio_codec, codec); - codec_context->bit_rate = audio_codec_get_get_bitrate(audio_codec); + codec_context->sample_fmt = audio_codec_get_sample_format(audio_codec, codec, mix_audio); + codec_context->bit_rate = audio_bitrate == 0 ? audio_codec_get_get_bitrate(audio_codec) : audio_bitrate; codec_context->sample_rate = 48000; if(audio_codec == AudioCodec::AAC) codec_context->profile = FF_PROFILE_AAC_LOW; @@ -262,8 +317,7 @@ static AVCodecContext* create_audio_codec_context(int fps, AudioCodec audio_code codec_context->time_base.num = 1; codec_context->time_base.den = AV_TIME_BASE; - codec_context->framerate.num = fps; - codec_context->framerate.den = 1; + codec_context->thread_count = 1; codec_context->flags |= AV_CODEC_FLAG_GLOBAL_HEADER; return codec_context; @@ -271,7 +325,8 @@ static AVCodecContext* create_audio_codec_context(int fps, AudioCodec audio_code static AVCodecContext *create_video_codec_context(AVPixelFormat pix_fmt, VideoQuality video_quality, - int fps, const AVCodec *codec, bool is_livestream, gsr_gpu_vendor vendor, FramerateMode framerate_mode) { + int fps, const AVCodec *codec, bool is_livestream, gsr_gpu_vendor vendor, FramerateMode framerate_mode, + bool hdr, gsr_color_range color_range) { AVCodecContext *codec_context = avcodec_alloc_context3(codec); @@ -289,7 +344,7 @@ static AVCodecContext *create_video_codec_context(AVPixelFormat pix_fmt, codec_context->framerate.den = 1; codec_context->sample_aspect_ratio.num = 0; codec_context->sample_aspect_ratio.den = 0; - // High values reeduce file size but increases time it takes to seek + // High values reduce file size but increases time it takes to seek if(is_livestream) { codec_context->flags |= (AV_CODEC_FLAG_CLOSED_GOP | AV_CODEC_FLAG_LOW_DELAY); codec_context->flags2 |= AV_CODEC_FLAG2_FAST; @@ -301,10 +356,16 @@ static AVCodecContext *create_video_codec_context(AVPixelFormat pix_fmt, } codec_context->max_b_frames = 0; codec_context->pix_fmt = pix_fmt; - //codec_context->color_range = AVCOL_RANGE_JPEG; // TODO: Amd/nvidia? - //codec_context->color_primaries = AVCOL_PRI_BT709; - //codec_context->color_trc = AVCOL_TRC_BT709; - //codec_context->colorspace = AVCOL_SPC_BT709; + codec_context->color_range = color_range == GSR_COLOR_RANGE_LIMITED ? AVCOL_RANGE_MPEG : AVCOL_RANGE_JPEG; + if(hdr) { + codec_context->color_primaries = AVCOL_PRI_BT2020; + codec_context->color_trc = AVCOL_TRC_SMPTE2084; + codec_context->colorspace = AVCOL_SPC_BT2020_NCL; + } else { + codec_context->color_primaries = AVCOL_PRI_BT709; + codec_context->color_trc = AVCOL_TRC_BT709; + codec_context->colorspace = AVCOL_SPC_BT709; + } //codec_context->chroma_sample_location = AVCHROMA_LOC_CENTER; if(codec->id == AV_CODEC_ID_HEVC) codec_context->codec_tag = MKTAG('h', 'v', 'c', '1'); @@ -353,6 +414,23 @@ static AVCodecContext *create_video_codec_context(AVPixelFormat pix_fmt, codec_context->bit_rate = 0; #endif + if(vendor != GSR_GPU_VENDOR_NVIDIA) { + switch(video_quality) { + case VideoQuality::MEDIUM: + codec_context->global_quality = 180; + break; + case VideoQuality::HIGH: + codec_context->global_quality = 140; + break; + case VideoQuality::VERY_HIGH: + codec_context->global_quality = 120; + break; + case VideoQuality::ULTRA: + codec_context->global_quality = 100; + break; + } + } + av_opt_set_int(codec_context->priv_data, "b_ref_mode", 0, 0); //av_opt_set_int(codec_context->priv_data, "cbr", true, 0); @@ -364,9 +442,13 @@ static AVCodecContext *create_video_codec_context(AVPixelFormat pix_fmt, //codec_context->compression_level = 2; } + //av_opt_set(codec_context->priv_data, "bsf", "hevc_metadata=colour_primaries=9:transfer_characteristics=16:matrix_coefficients=9", 0); + //codec_context->rc_max_rate = codec_context->bit_rate; //codec_context->rc_min_rate = codec_context->bit_rate; //codec_context->rc_buffer_size = codec_context->bit_rate / 10; + // TODO: Do this when not using cqp + //codec_context->rc_initial_buffer_occupancy = codec_context->bit_rate * 1000; codec_context->flags |= AV_CODEC_FLAG_GLOBAL_HEADER; @@ -374,8 +456,14 @@ static AVCodecContext *create_video_codec_context(AVPixelFormat pix_fmt, } static bool vaapi_create_codec_context(AVCodecContext *video_codec_context, const char *card_path) { + char render_path[128]; + if(!gsr_card_path_get_render_path(card_path, render_path)) { + fprintf(stderr, "gsr error: failed to get /dev/dri/renderDXXX file from %s\n", card_path); + return false; + } + AVBufferRef *device_ctx; - if(av_hwdevice_ctx_create(&device_ctx, AV_HWDEVICE_TYPE_VAAPI, card_path, NULL, 0) < 0) { + if(av_hwdevice_ctx_create(&device_ctx, AV_HWDEVICE_TYPE_VAAPI, render_path, NULL, 0) < 0) { fprintf(stderr, "Error: Failed to create hardware device context\n"); return false; } @@ -396,7 +484,7 @@ static bool vaapi_create_codec_context(AVCodecContext *video_codec_context, cons hw_frame_context->device_ref = device_ctx; hw_frame_context->device_ctx = (AVHWDeviceContext*)device_ctx->data; - hw_frame_context->initial_pool_size = 1; + //hw_frame_context->initial_pool_size = 1; if (av_hwframe_ctx_init(frame_context) < 0) { fprintf(stderr, "Error: Failed to initialize hardware frame context " @@ -413,7 +501,7 @@ static bool vaapi_create_codec_context(AVCodecContext *video_codec_context, cons static bool check_if_codec_valid_for_hardware(const AVCodec *codec, gsr_gpu_vendor vendor, const char *card_path) { // Do not use AV_PIX_FMT_CUDA because we dont want to do full check with hardware context - AVCodecContext *codec_context = create_video_codec_context(vendor == GSR_GPU_VENDOR_NVIDIA ? AV_PIX_FMT_YUV420P : AV_PIX_FMT_VAAPI, VideoQuality::VERY_HIGH, 60, codec, false, vendor, FramerateMode::CONSTANT); + AVCodecContext *codec_context = create_video_codec_context(vendor == GSR_GPU_VENDOR_NVIDIA ? AV_PIX_FMT_YUV420P : AV_PIX_FMT_VAAPI, VideoQuality::VERY_HIGH, 60, codec, false, vendor, FramerateMode::CONSTANT, false, GSR_COLOR_RANGE_LIMITED); if(!codec_context) return false; @@ -473,7 +561,31 @@ static const AVCodec* find_h265_encoder(gsr_gpu_vendor vendor, const char *card_ return checked_success ? codec : nullptr; } -static AVFrame* open_audio(AVCodecContext *audio_codec_context) { +static const AVCodec* find_av1_encoder(gsr_gpu_vendor vendor, const char *card_path) { + // Workaround bug with av1 nvidia in older ffmpeg versions that causes the whole application to crash + // when avcodec_open2 is opened with av1_nvenc + if(vendor == GSR_GPU_VENDOR_NVIDIA && LIBAVCODEC_BUILD < AV_VERSION_INT(60, 30, 100)) { + return nullptr; + } + + const AVCodec *codec = avcodec_find_encoder_by_name(vendor == GSR_GPU_VENDOR_NVIDIA ? "av1_nvenc" : "av1_vaapi"); + if(!codec) + codec = avcodec_find_encoder_by_name(vendor == GSR_GPU_VENDOR_NVIDIA ? "nvenc_av1" : "vaapi_av1"); + + if(!codec) + return nullptr; + + static bool checked = false; + static bool checked_success = true; + if(!checked) { + checked = true; + if(!check_if_codec_valid_for_hardware(codec, vendor, card_path)) + checked_success = false; + } + return checked_success ? codec : nullptr; +} + +static void open_audio(AVCodecContext *audio_codec_context) { AVDictionary *options = nullptr; av_dict_set(&options, "strict", "experimental", 0); @@ -483,7 +595,9 @@ static AVFrame* open_audio(AVCodecContext *audio_codec_context) { fprintf(stderr, "failed to open codec, reason: %s\n", av_error_to_string(ret)); _exit(1); } +} +static AVFrame* create_audio_frame(AVCodecContext *audio_codec_context) { AVFrame *frame = av_frame_alloc(); if(!frame) { fprintf(stderr, "failed to allocate audio frame\n"); @@ -500,7 +614,7 @@ static AVFrame* open_audio(AVCodecContext *audio_codec_context) { av_channel_layout_copy(&frame->ch_layout, &audio_codec_context->ch_layout); #endif - ret = av_frame_get_buffer(frame, 0); + int ret = av_frame_get_buffer(frame, 0); if(ret < 0) { fprintf(stderr, "failed to allocate audio data buffers, reason: %s\n", av_error_to_string(ret)); _exit(1); @@ -509,23 +623,38 @@ static AVFrame* open_audio(AVCodecContext *audio_codec_context) { return frame; } -static void open_video(AVCodecContext *codec_context, VideoQuality video_quality, bool very_old_gpu, gsr_gpu_vendor vendor, PixelFormat pixel_format) { +static void open_video(AVCodecContext *codec_context, VideoQuality video_quality, bool very_old_gpu, gsr_gpu_vendor vendor, PixelFormat pixel_format, bool hdr) { AVDictionary *options = nullptr; if(vendor == GSR_GPU_VENDOR_NVIDIA) { bool supports_p4 = false; - bool supports_p6 = false; + bool supports_p5 = false; const AVOption *opt = nullptr; while((opt = av_opt_next(codec_context->priv_data, opt))) { if(opt->type == AV_OPT_TYPE_CONST) { if(strcmp(opt->name, "p4") == 0) supports_p4 = true; - else if(strcmp(opt->name, "p6") == 0) - supports_p6 = true; + else if(strcmp(opt->name, "p5") == 0) + supports_p5 = true; } } - if(very_old_gpu) { + if(codec_context->codec_id == AV_CODEC_ID_AV1) { + switch(video_quality) { + case VideoQuality::MEDIUM: + av_dict_set_int(&options, "qp", 37, 0); + break; + case VideoQuality::HIGH: + av_dict_set_int(&options, "qp", 32, 0); + break; + case VideoQuality::VERY_HIGH: + av_dict_set_int(&options, "qp", 28, 0); + break; + case VideoQuality::ULTRA: + av_dict_set_int(&options, "qp", 24, 0); + break; + } + } else if(very_old_gpu || codec_context->codec_id == AV_CODEC_ID_H264) { switch(video_quality) { case VideoQuality::MEDIUM: av_dict_set_int(&options, "qp", 37, 0); @@ -543,13 +672,13 @@ static void open_video(AVCodecContext *codec_context, VideoQuality video_quality } else { switch(video_quality) { case VideoQuality::MEDIUM: - av_dict_set_int(&options, "qp", 40, 0); + av_dict_set_int(&options, "qp", 37, 0); break; case VideoQuality::HIGH: - av_dict_set_int(&options, "qp", 35, 0); + av_dict_set_int(&options, "qp", 32, 0); break; case VideoQuality::VERY_HIGH: - av_dict_set_int(&options, "qp", 30, 0); + av_dict_set_int(&options, "qp", 28, 0); break; case VideoQuality::ULTRA: av_dict_set_int(&options, "qp", 24, 0); @@ -557,7 +686,7 @@ static void open_video(AVCodecContext *codec_context, VideoQuality video_quality } } - if(!supports_p4 && !supports_p6) + if(!supports_p4 && !supports_p5) fprintf(stderr, "Info: your ffmpeg version is outdated. It's recommended that you use the flatpak version of gpu-screen-recorder version instead, which you can find at https://flathub.org/apps/details/com.dec05eba.gpu_screen_recorder\n"); //if(is_livestream) { @@ -571,10 +700,14 @@ static void open_video(AVCodecContext *codec_context, VideoQuality video_quality // older gpus p5-p7 slow the gpu down to a crawl... // "hq" is now just an alias for p7 in ffmpeg :( // TODO: Temporary disable because of stuttering? + + // TODO: Preset is set to p5 for now but it should ideally be p6 or p7. + // This change is needed because for certain sizes of a window (or monitor?) such as 971x780 causes encoding to freeze + // when using h264 codec. This is a new(?) nvidia driver bug. if(very_old_gpu) av_dict_set(&options, "preset", supports_p4 ? "p4" : "medium", 0); else - av_dict_set(&options, "preset", supports_p6 ? "p6" : "slow", 0); + av_dict_set(&options, "preset", supports_p5 ? "p5" : "slow", 0); av_dict_set(&options, "tune", "hq", 0); av_dict_set(&options, "rc", "constqp", 0); @@ -588,24 +721,57 @@ static void open_video(AVCodecContext *codec_context, VideoQuality video_quality av_dict_set(&options, "profile", "high444p", 0); break; } + } else if(codec_context->codec_id == AV_CODEC_ID_AV1) { + switch(pixel_format) { + case PixelFormat::YUV420: + av_dict_set(&options, "rgb_mode", "yuv420", 0); + break; + case PixelFormat::YUV444: + av_dict_set(&options, "rgb_mode", "yuv444", 0); + break; + } } else { //av_dict_set(&options, "profile", "main10", 0); //av_dict_set(&options, "pix_fmt", "yuv420p16le", 0); + if(hdr) { + av_dict_set(&options, "profile", "main10", 0); + } else { + av_dict_set(&options, "profile", "main", 0); + } } } else { - switch(video_quality) { - case VideoQuality::MEDIUM: - av_dict_set_int(&options, "qp", 32, 0); - break; - case VideoQuality::HIGH: - av_dict_set_int(&options, "qp", 28, 0); - break; - case VideoQuality::VERY_HIGH: - av_dict_set_int(&options, "qp", 24, 0); - break; - case VideoQuality::ULTRA: - av_dict_set_int(&options, "qp", 18, 0); - break; + if(codec_context->codec_id == AV_CODEC_ID_AV1) { + // Using global_quality option + } else if(codec_context->codec_id == AV_CODEC_ID_H264) { + switch(video_quality) { + case VideoQuality::MEDIUM: + av_dict_set_int(&options, "qp", 34, 0); + break; + case VideoQuality::HIGH: + av_dict_set_int(&options, "qp", 30, 0); + break; + case VideoQuality::VERY_HIGH: + av_dict_set_int(&options, "qp", 26, 0); + break; + case VideoQuality::ULTRA: + av_dict_set_int(&options, "qp", 22, 0); + break; + } + } else { + switch(video_quality) { + case VideoQuality::MEDIUM: + av_dict_set_int(&options, "qp", 37, 0); + break; + case VideoQuality::HIGH: + av_dict_set_int(&options, "qp", 32, 0); + break; + case VideoQuality::VERY_HIGH: + av_dict_set_int(&options, "qp", 28, 0); + break; + case VideoQuality::ULTRA: + av_dict_set_int(&options, "qp", 24, 0); + break; + } } // TODO: More quality options @@ -614,9 +780,17 @@ static void open_video(AVCodecContext *codec_context, VideoQuality video_quality if(codec_context->codec_id == AV_CODEC_ID_H264) { av_dict_set(&options, "profile", "high", 0); - av_dict_set_int(&options, "quality", 7, 0); + //av_dict_set_int(&options, "quality", 5, 0); // quality preset + } else if(codec_context->codec_id == AV_CODEC_ID_AV1) { + av_dict_set(&options, "profile", "main", 0); // TODO: use professional instead? + av_dict_set(&options, "tier", "main", 0); } else { - av_dict_set(&options, "profile", "main", 0); + if(hdr) { + av_dict_set(&options, "profile", "main10", 0); + av_dict_set(&options, "sei", "hdr", 0); + } else { + av_dict_set(&options, "profile", "main", 0); + } } } @@ -634,23 +808,26 @@ static void open_video(AVCodecContext *codec_context, VideoQuality video_quality } static void usage_header() { - fprintf(stderr, "usage: gpu-screen-recorder -w <window_id|monitor|focused> [-c <container_format>] [-s WxH] -f <fps> [-a <audio_input>] [-q <quality>] [-r <replay_buffer_size_sec>] [-k h264|h265] [-ac aac|opus|flac] [-oc yes|no] [-fm cfr|vfr] [-v yes|no] [-h|--help] [-o <output_file>]\n"); + const bool inside_flatpak = getenv("FLATPAK_ID") != NULL; + const char *program_name = inside_flatpak ? "flatpak run --command=gpu-screen-recorder com.dec05eba.gpu_screen_recorder" : "gpu-screen-recorder"; + fprintf(stderr, "usage: %s -w <window_id|monitor|focused> [-c <container_format>] [-s WxH] -f <fps> [-a <audio_input>] [-q <quality>] [-r <replay_buffer_size_sec>] [-k h264|hevc|hevc_hdr|av1|av1_hdr] [-ac aac|opus|flac] [-ab <bitrate>] [-oc yes|no] [-fm cfr|vfr] [-cr limited|full] [-v yes|no] [-h|--help] [-o <output_file>] [-mf yes|no] [-sc <script_path>] [-cursor yes|no]\n", program_name); } static void usage_full() { + const bool inside_flatpak = getenv("FLATPAK_ID") != NULL; + const char *program_name = inside_flatpak ? "flatpak run --command=gpu-screen-recorder com.dec05eba.gpu_screen_recorder" : "gpu-screen-recorder"; usage_header(); fprintf(stderr, "\n"); fprintf(stderr, "OPTIONS:\n"); - fprintf(stderr, " -w Window to record, a display, \"screen\", \"screen-direct\", \"screen-direct-force\" or \"focused\".\n"); - fprintf(stderr, " The display is the display (monitor) name in xrandr and if \"screen\", \"screen-direct\" or \"screen-direct-force\" is selected then all displays are recorded.\n"); - fprintf(stderr, " If this is \"focused\" then the currently focused window is recorded. When recording the focused window then the -s option has to be used as well.\n"); - fprintf(stderr, " \"screen-direct\"/\"screen-direct-force\" skips one texture copy for fullscreen applications so it may lead to better performance and it works with VRR monitors\n"); - fprintf(stderr, " when recording fullscreen application but may break some applications, such as mpv in fullscreen mode. Direct mode doesn't capture cursor either.\n"); - fprintf(stderr, " \"screen-direct-force\" is not recommended unless you use a VRR monitor because there might be driver issues that cause the video to stutter or record a black screen.\n"); - fprintf(stderr, " On AMD/Intel, capturing a monitor might have better performance than recording a single window.\n"); + fprintf(stderr, " -w Window id to record, a display (monitor name), \"screen\", \"screen-direct-force\" or \"focused\".\n"); + fprintf(stderr, " If this is \"screen\" or \"screen-direct-force\" then all monitors are recorded.\n"); + fprintf(stderr, " \"screen-direct-force\" is not recommended unless you use a VRR monitor on Nvidia X11 and you are aware that using this option can cause games to freeze/crash or other issues because of Nvidia driver issues.\n"); + fprintf(stderr, " \"screen-direct-force\" option is only available on Nvidia X11. VRR works without this option on other systems.\n"); fprintf(stderr, "\n"); fprintf(stderr, " -c Container format for output file, for example mp4, or flv. Only required if no output file is specified or if recording in replay buffer mode.\n"); fprintf(stderr, " If an output file is specified and -c is not used then the container format is determined from the output filename extension.\n"); + fprintf(stderr, " Only containers that support h264, hevc or av1 are supported, which means that only mp4, mkv, flv (and some others) are supported.\n"); + fprintf(stderr, " WebM is not supported yet (most hardware doesn't support WebM video encoding).\n"); fprintf(stderr, "\n"); fprintf(stderr, " -s The size (area) to record at in the format WxH, for example 1920x1080. This option is only supported (and required) when -w is \"focused\".\n"); fprintf(stderr, "\n"); @@ -668,33 +845,58 @@ static void usage_full() { fprintf(stderr, " and the video will only be saved when the gpu-screen-recorder is closed. This feature is similar to Nvidia's instant replay feature.\n"); fprintf(stderr, " This option has be between 5 and 1200. Note that the replay buffer size will not always be precise, because of keyframes. Optional, disabled by default.\n"); fprintf(stderr, "\n"); - fprintf(stderr, " -k Video codec to use. Should be either 'auto', 'h264' or 'h265'. Defaults to 'auto' which defaults to 'h265' unless recording at fps higher than 60. Defaults to 'h264' on intel.\n"); - fprintf(stderr, " Forcefully set to 'h264' if -c is 'flv'.\n"); + fprintf(stderr, " -k Video codec to use. Should be either 'auto', 'h264', 'hevc', 'av1', 'hevc_hdr' or 'av1_hdr'. Defaults to 'auto' which defaults to 'hevc' on AMD/Nvidia and 'h264' on intel.\n"); + fprintf(stderr, " Forcefully set to 'h264' if the file container type is 'flv'.\n"); + fprintf(stderr, " Forcefully set to 'hevc' on AMD/intel if video codec is 'h264' and if the file container type is 'mkv'.\n"); + fprintf(stderr, " 'hevc_hdr' and 'av1_hdr' option is not available on X11.\n"); + fprintf(stderr, " Note: hdr metadata is not included in the video when recording with 'hevc_hdr'/'av1_hdr' because of bugs in AMD, Intel and NVIDIA drivers (amazin', they are bugged).\n"); fprintf(stderr, "\n"); fprintf(stderr, " -ac Audio codec to use. Should be either 'aac', 'opus' or 'flac'. Defaults to 'opus' for .mp4/.mkv files, otherwise defaults to 'aac'.\n"); fprintf(stderr, " 'opus' and 'flac' is only supported by .mp4/.mkv files. 'opus' is recommended for best performance and smallest audio size.\n"); fprintf(stderr, "\n"); - fprintf(stderr, " -oc Overclock memory transfer rate to the maximum performance level. This only applies to NVIDIA and exists to overcome a bug in NVIDIA driver where performance level\n"); - fprintf(stderr, " is dropped when you record a game. Only needed if you are recording a game that is bottlenecked by GPU.\n"); + fprintf(stderr, " -ab Audio bitrate to use. Optional, by default the bitrate is 128000 for opus and flac and 160000 for aac.\n"); + fprintf(stderr, " If this is set to 0 then it's the same as if it's absent, in which case the bitrate is determined automatically depending on the audio codec.\n"); + fprintf(stderr, "\n"); + fprintf(stderr, " -oc Overclock memory transfer rate to the maximum performance level. This only applies to NVIDIA on X11 and exists to overcome a bug in NVIDIA driver where performance level\n"); + fprintf(stderr, " is dropped when you record a game. Only needed if you are recording a game that is bottlenecked by GPU. The same issue exists on Wayland but overclocking is not possible on Wayland.\n"); fprintf(stderr, " Works only if your have \"Coolbits\" set to \"12\" in NVIDIA X settings, see README for more information. Note! use at your own risk! Optional, disabled by default.\n"); fprintf(stderr, "\n"); - fprintf(stderr, " -fm Framerate mode. Should be either 'cfr' or 'vfr'. Defaults to 'cfr' on NVIDIA and 'vfr' on AMD/Intel.\n"); + fprintf(stderr, " -fm Framerate mode. Should be either 'cfr' or 'vfr'. Defaults to 'vfr'.\n"); + fprintf(stderr, "\n"); + fprintf(stderr, " -cr Color range. Should be either 'limited' (aka mpeg) or 'full' (aka jpeg). Defaults to 'limited'.\n"); + fprintf(stderr, " Limited color range means that colors are in range 16-235 while full color range means that colors are in range 0-255 (when not recording with hdr).\n"); + fprintf(stderr, " Note that some buggy video players (such as vlc) are unable to correctly display videos in full color range.\n"); fprintf(stderr, "\n"); fprintf(stderr, " -v Prints per second, fps updates. Optional, set to 'yes' by default.\n"); fprintf(stderr, "\n"); - fprintf(stderr, " -h Show this help.\n"); + fprintf(stderr, " -h, --help\n"); + fprintf(stderr, " Show this help.\n"); + fprintf(stderr, "\n"); + fprintf(stderr, " -mf Organise replays in folders based on the current date.\n"); + fprintf(stderr, "\n"); + fprintf(stderr, " -sc Run a script on the saved video file (non-blocking). The first argument to the script is the filepath to the saved video file and the second argument is the recording type (either \"regular\" or \"replay\").\n"); + fprintf(stderr, " Not applicable for live streams.\n"); + fprintf(stderr, "\n"); + fprintf(stderr, " -cursor\n"); + fprintf(stderr, " Record cursor. Defaults to 'yes'.\n"); + fprintf(stderr, "\n"); + fprintf(stderr, " --list-supported-video-codecs\n"); + fprintf(stderr, " List supported video codecs and exits. Prints h264, hevc, hevc_hdr, av1 and av1_hdr (if supported).\n"); fprintf(stderr, "\n"); //fprintf(stderr, " -pixfmt The pixel format to use for the output video. yuv420 is the most common format and is best supported, but the color is compressed, so colors can look washed out and certain colors of text can look bad. Use yuv444 for no color compression, but the video may not work everywhere and it may not work with hardware video decoding. Optional, defaults to yuv420\n"); fprintf(stderr, " -o The output file path. If omitted then the encoded data is sent to stdout. Required in replay mode (when using -r).\n"); - fprintf(stderr, " In replay mode this has to be an existing directory instead of a file.\n"); + fprintf(stderr, " In replay mode this has to be a directory instead of a file.\n"); + fprintf(stderr, " The directory to the file is created (recursively) if it doesn't already exist.\n"); fprintf(stderr, "\n"); fprintf(stderr, "NOTES:\n"); - fprintf(stderr, " Send signal SIGINT to gpu-screen-recorder (Ctrl+C, or killall gpu-screen-recorder) to stop and save the recording (when not using replay mode).\n"); + fprintf(stderr, " Send signal SIGINT to gpu-screen-recorder (Ctrl+C, or killall -SIGINT gpu-screen-recorder) to stop and save the recording. When in replay mode this stops recording without saving.\n"); fprintf(stderr, " Send signal SIGUSR1 to gpu-screen-recorder (killall -SIGUSR1 gpu-screen-recorder) to save a replay (when in replay mode).\n"); + fprintf(stderr, " Send signal SIGUSR2 to gpu-screen-recorder (killall -SIGUSR2 gpu-screen-recorder) to pause/unpause recording. Only applicable and useful when recording (not streaming nor replay).\n"); fprintf(stderr, "\n"); - fprintf(stderr, "EXAMPLES\n"); - fprintf(stderr, " gpu-screen-recorder -w screen -f 60 -a \"$(pactl get-default-sink).monitor\" -o video.mp4\n"); - fprintf(stderr, " gpu-screen-recorder -w screen -f 60 -a \"$(pactl get-default-sink).monitor|$(pactl get-default-source)\" -o video.mp4\n"); + fprintf(stderr, "EXAMPLES:\n"); + fprintf(stderr, " %s -w screen -f 60 -a \"$(pactl get-default-sink).monitor\" -o \"$HOME/Videos/video.mp4\"\n", program_name); + fprintf(stderr, " %s -w screen -f 60 -a \"$(pactl get-default-sink).monitor|$(pactl get-default-source)\" -o \"$HOME/Videos/video.mp4\"\n", program_name); + fprintf(stderr, " %s -w screen -f 60 -a \"$(pactl get-default-sink).monitor\" -c mkv -r 60 -o \"$HOME/Videos\"\n", program_name); //fprintf(stderr, " gpu-screen-recorder -w screen -f 60 -q ultra -pixfmt yuv444 -o video.mp4\n"); _exit(1); } @@ -706,8 +908,9 @@ static void usage() { static sig_atomic_t running = 1; static sig_atomic_t save_replay = 0; +static sig_atomic_t toggle_pause = 0; -static void int_handler(int) { +static void stop_handler(int) { running = 0; } @@ -715,37 +918,35 @@ static void save_replay_handler(int) { save_replay = 1; } -struct Arg { - std::vector<const char*> values; - bool optional = false; - bool list = false; - - const char* value() const { - if(values.empty()) - return nullptr; - return values.front(); - } -}; +static void toggle_pause_handler(int) { + toggle_pause = 1; +} static bool is_hex_num(char c) { return (c >= 'A' && c <= 'F') || (c >= 'a' && c <= 'f') || (c >= '0' && c <= '9'); } static bool contains_non_hex_number(const char *str) { + bool hex_start = false; size_t len = strlen(str); if(len >= 2 && memcmp(str, "0x", 2) == 0) { str += 2; len -= 2; + hex_start = true; } + bool is_hex = false; for(size_t i = 0; i < len; ++i) { char c = str[i]; if(c == '\0') return false; if(!is_hex_num(c)) return true; + if((c >= 'A' && c <= 'F') || (c >= 'a' && c <= 'f')) + is_hex = true; } - return false; + + return is_hex && !hex_start; } static std::string get_date_str() { @@ -756,6 +957,22 @@ static std::string get_date_str() { return str; } +static std::string get_date_only_str() { + char str[128]; + time_t now = time(NULL); + struct tm *t = localtime(&now); + strftime(str, sizeof(str)-1, "%Y-%m-%d", t); + return str; +} + +static std::string get_time_only_str() { + char str[128]; + time_t now = time(NULL); + struct tm *t = localtime(&now); + strftime(str, sizeof(str)-1, "%H-%M-%S", t); + return str; +} + static AVStream* create_stream(AVFormatContext *av_format_context, AVCodecContext *codec_context) { AVStream *stream = avformat_new_stream(av_format_context, nullptr); if (!stream) { @@ -768,30 +985,109 @@ static AVStream* create_stream(AVFormatContext *av_format_context, AVCodecContex return stream; } +static void run_recording_saved_script_async(const char *script_file, const char *video_file, const char *type) { + char script_file_full[PATH_MAX]; + script_file_full[0] = '\0'; + if(!realpath(script_file, script_file_full)) { + fprintf(stderr, "Error: script file not found: %s\n", script_file); + return; + } + + const char *args[6]; + const bool inside_flatpak = getenv("FLATPAK_ID") != NULL; + + if(inside_flatpak) { + args[0] = "flatpak-spawn"; + args[1] = "--host"; + args[2] = script_file_full; + args[3] = video_file; + args[4] = type; + args[5] = NULL; + } else { + args[0] = script_file_full; + args[1] = video_file; + args[2] = type; + args[3] = NULL; + } + + pid_t pid = fork(); + if(pid == -1) { + perror(script_file_full); + return; + } else if(pid == 0) { // child + setsid(); + signal(SIGHUP, SIG_IGN); + + pid_t second_child = fork(); + if(second_child == 0) { // child + execvp(args[0], (char* const*)args); + perror(script_file_full); + _exit(127); + } else if(second_child != -1) { // parent + _exit(0); + } + } else { // parent + waitpid(pid, NULL, 0); + } +} + struct AudioDevice { SoundDevice sound_device; AudioInput audio_input; AVFilterContext *src_filter_ctx = nullptr; + AVFrame *frame = nullptr; std::thread thread; // TODO: Instead of having a thread for each track, have one thread for all threads and read the data with non-blocking read }; // TODO: Cleanup struct AudioTrack { AVCodecContext *codec_context = nullptr; - AVFrame *frame = nullptr; AVStream *stream = nullptr; std::vector<AudioDevice> audio_devices; AVFilterGraph *graph = nullptr; AVFilterContext *sink = nullptr; int stream_index = 0; + int64_t pts = 0; }; static std::future<void> save_replay_thread; -static std::vector<AVPacket> save_replay_packets; +static std::vector<std::shared_ptr<PacketData>> save_replay_packets; static std::string save_replay_output_filepath; -static void save_replay_async(AVCodecContext *video_codec_context, int video_stream_index, std::vector<AudioTrack> &audio_tracks, const std::deque<AVPacket> &frame_data_queue, bool frames_erased, std::string output_dir, const char *container_format, const std::string &file_extension, std::mutex &write_output_mutex) { +static int create_directory_recursive(char *path) { + int path_len = strlen(path); + char *p = path; + char *end = path + path_len; + for(;;) { + char *slash_p = strchr(p, '/'); + + // Skips first '/', we don't want to try and create the root directory + if(slash_p == path) { + ++p; + continue; + } + + if(!slash_p) + slash_p = end; + + char prev_char = *slash_p; + *slash_p = '\0'; + int err = mkdir(path, S_IRWXU); + *slash_p = prev_char; + + if(err == -1 && errno != EEXIST) + return err; + + if(slash_p == end) + break; + else + p = slash_p + 1; + } + return 0; +} + +static void save_replay_async(AVCodecContext *video_codec_context, int video_stream_index, std::vector<AudioTrack> &audio_tracks, std::deque<std::shared_ptr<PacketData>> &frame_data_queue, bool frames_erased, std::string output_dir, const char *container_format, const std::string &file_extension, std::mutex &write_output_mutex, bool make_folders) { if(save_replay_thread.valid()) return; @@ -803,7 +1099,7 @@ static void save_replay_async(AVCodecContext *video_codec_context, int video_str std::lock_guard<std::mutex> lock(write_output_mutex); start_index = (size_t)-1; for(size_t i = 0; i < frame_data_queue.size(); ++i) { - const AVPacket &av_packet = frame_data_queue[i]; + const AVPacket &av_packet = frame_data_queue[i]->data; if((av_packet.flags & AV_PKT_FLAG_KEY) && av_packet.stream_index == video_stream_index) { start_index = i; break; @@ -814,11 +1110,11 @@ static void save_replay_async(AVCodecContext *video_codec_context, int video_str return; if(frames_erased) { - video_pts_offset = frame_data_queue[start_index].pts; + video_pts_offset = frame_data_queue[start_index]->data.pts; // Find the next audio packet to use as audio pts offset for(size_t i = start_index; i < frame_data_queue.size(); ++i) { - const AVPacket &av_packet = frame_data_queue[i]; + const AVPacket &av_packet = frame_data_queue[i]->data; if(av_packet.stream_index != video_stream_index) { audio_pts_offset = av_packet.pts; break; @@ -830,11 +1126,19 @@ static void save_replay_async(AVCodecContext *video_codec_context, int video_str save_replay_packets.resize(frame_data_queue.size()); for(size_t i = 0; i < frame_data_queue.size(); ++i) { - av_packet_ref(&save_replay_packets[i], &frame_data_queue[i]); + save_replay_packets[i] = frame_data_queue[i]; } } - save_replay_output_filepath = output_dir + "/Replay_" + get_date_str() + "." + file_extension; + if (make_folders) { + std::string output_folder = output_dir + '/' + get_date_only_str(); + create_directory_recursive(&output_folder[0]); + save_replay_output_filepath = output_folder + "/Replay_" + get_time_only_str() + "." + file_extension; + } else { + create_directory_recursive(&output_dir[0]); + save_replay_output_filepath = output_dir + "/Replay_" + get_date_str() + "." + file_extension; + } + save_replay_thread = std::async(std::launch::async, [video_stream_index, container_format, start_index, video_pts_offset, audio_pts_offset, video_codec_context, &audio_tracks]() mutable { AVFormatContext *av_format_context; avformat_alloc_output_context2(&av_format_context, nullptr, container_format, nullptr); @@ -866,7 +1170,16 @@ static void save_replay_async(AVCodecContext *video_codec_context, int video_str } for(size_t i = start_index; i < save_replay_packets.size(); ++i) { - AVPacket &av_packet = save_replay_packets[i]; + // TODO: Check if successful + AVPacket av_packet; + memset(&av_packet, 0, sizeof(av_packet)); + //av_packet_from_data(av_packet, save_replay_packets[i]->data.data, save_replay_packets[i]->data.size); + av_packet.data = save_replay_packets[i]->data.data; + av_packet.size = save_replay_packets[i]->data.size; + av_packet.stream_index = save_replay_packets[i]->data.stream_index; + av_packet.pts = save_replay_packets[i]->data.pts; + av_packet.dts = save_replay_packets[i]->data.pts; + av_packet.flags = save_replay_packets[i]->data.flags; AVStream *stream = video_stream; AVCodecContext *codec_context = video_codec_context; @@ -886,9 +1199,11 @@ static void save_replay_async(AVCodecContext *video_codec_context, int video_str av_packet.stream_index = stream->index; av_packet_rescale_ts(&av_packet, codec_context->time_base, stream->time_base); - int ret = av_interleaved_write_frame(av_format_context, &av_packet); + ret = av_write_frame(av_format_context, &av_packet); if(ret < 0) fprintf(stderr, "Error: Failed to write frame index %d to muxer, reason: %s (%d)\n", stream->index, av_error_to_string(ret), ret); + + //av_packet_free(&av_packet); } if (av_write_trailer(av_format_context) != 0) @@ -949,6 +1264,7 @@ static bool is_livestream_path(const char *str) { static int init_filter_graph(AVCodecContext *audio_codec_context, AVFilterGraph **graph, AVFilterContext **sink, std::vector<AVFilterContext*> &src_filter_ctx, size_t num_sources) { char ch_layout[64]; int err = 0; + ch_layout[0] = '\0'; AVFilterGraph *filter_graph = avfilter_graph_alloc(); if (!filter_graph) { @@ -1048,9 +1364,11 @@ static int init_filter_graph(AVCodecContext *audio_codec_context, AVFilterGraph return 0; } -static void xwayland_check_callback(const XRROutputInfo *output_info, const XRRCrtcInfo*, const XRRModeInfo*, void *userdata) { +static void xwayland_check_callback(const gsr_monitor *monitor, void *userdata) { bool *xwayland_found = (bool*)userdata; - if(output_info->nameLen >= 8 && strncmp(output_info->name, "XWAYLAND", 8) == 0) + if(monitor->name_len >= 8 && strncmp(monitor->name, "XWAYLAND", 8) == 0) + *xwayland_found = true; + else if(memmem(monitor->name, monitor->name_len, "X11", 3)) *xwayland_found = true; } @@ -1060,20 +1378,252 @@ static bool is_xwayland(Display *display) { return true; bool xwayland_found = false; - for_each_active_monitor_output(display, xwayland_check_callback, &xwayland_found); + for_each_active_monitor_output_x11(display, xwayland_check_callback, &xwayland_found); return xwayland_found; } -struct ReceivePacketData { - AVCodecContext *codec_context; - int stream_index; - AVStream *stream; - int64_t pts; +static void list_supported_video_codecs() { + bool wayland = false; + Display *dpy = XOpenDisplay(nullptr); + if (!dpy) { + wayland = true; + fprintf(stderr, "Warning: failed to connect to the X server. Assuming wayland is running without Xwayland\n"); + } + + XSetErrorHandler(x11_error_handler); + XSetIOErrorHandler(x11_io_error_handler); + + if(!wayland) + wayland = is_xwayland(dpy); + + gsr_egl egl; + if(!gsr_egl_load(&egl, dpy, wayland, false)) { + fprintf(stderr, "gsr error: failed to load opengl\n"); + _exit(1); + } + + char card_path[128]; + card_path[0] = '\0'; + if(wayland || egl.gpu_info.vendor != GSR_GPU_VENDOR_NVIDIA) { + // TODO: Allow specifying another card, and in other places + if(!gsr_get_valid_card_path(&egl, card_path)) { + fprintf(stderr, "Error: no /dev/dri/cardX device found. If you are running GPU Screen Recorder with prime-run then try running without it\n"); + _exit(2); + } + } + + av_log_set_level(AV_LOG_FATAL); + + // TODO: Output hdr + if(find_h264_encoder(egl.gpu_info.vendor, card_path)) + puts("h264"); + if(find_h265_encoder(egl.gpu_info.vendor, card_path)) + puts("hevc"); + if(find_av1_encoder(egl.gpu_info.vendor, card_path)) + puts("av1"); + + fflush(stdout); + + gsr_egl_unload(&egl); + if(dpy) + XCloseDisplay(dpy); +} + +static gsr_capture* create_capture_impl(const char *window_str, const char *screen_region, bool wayland, gsr_gpu_info gpu_inf, gsr_egl &egl, int fps, bool overclock, VideoCodec video_codec, gsr_color_range color_range, bool record_cursor) { + vec2i region_size = { 0, 0 }; + Window src_window_id = None; + bool follow_focused = false; + + gsr_capture *capture = nullptr; + if(strcmp(window_str, "focused") == 0) { + if(wayland) { + fprintf(stderr, "Error: GPU Screen Recorder window capture only works in a pure X11 session. Xwayland is not supported. You can record a monitor instead on wayland\n"); + _exit(2); + } + + if(!screen_region) { + fprintf(stderr, "Error: option -s is required when using -w focused\n"); + usage(); + } + + if(sscanf(screen_region, "%dx%d", ®ion_size.x, ®ion_size.y) != 2) { + fprintf(stderr, "Error: invalid value for option -s '%s', expected a value in format WxH\n", screen_region); + usage(); + } + + if(region_size.x <= 0 || region_size.y <= 0) { + fprintf(stderr, "Error: invalud value for option -s '%s', expected width and height to be greater than 0\n", screen_region); + usage(); + } + + follow_focused = true; + } else if(contains_non_hex_number(window_str)) { + if(wayland || egl.gpu_info.vendor != GSR_GPU_VENDOR_NVIDIA) { + if(strcmp(window_str, "screen") == 0) { + FirstOutputCallback first_output; + first_output.output_name = NULL; + for_each_active_monitor_output(&egl, GSR_CONNECTION_DRM, get_first_output, &first_output); + + if(first_output.output_name) { + window_str = first_output.output_name; + } else { + fprintf(stderr, "Error: no available output found\n"); + } + } + + gsr_monitor gmon; + if(!get_monitor_by_name(&egl, GSR_CONNECTION_DRM, window_str, &gmon)) { + fprintf(stderr, "gsr error: display \"%s\" not found, expected one of:\n", window_str); + fprintf(stderr, " \"screen\"\n"); + for_each_active_monitor_output(&egl, GSR_CONNECTION_DRM, monitor_output_callback_print, NULL); + _exit(1); + } + } else { + if(strcmp(window_str, "screen") != 0 && strcmp(window_str, "screen-direct") != 0 && strcmp(window_str, "screen-direct-force") != 0) { + gsr_monitor gmon; + if(!get_monitor_by_name(&egl, GSR_CONNECTION_X11, window_str, &gmon)) { + const int screens_width = XWidthOfScreen(DefaultScreenOfDisplay(egl.x11.dpy)); + const int screens_height = XWidthOfScreen(DefaultScreenOfDisplay(egl.x11.dpy)); + fprintf(stderr, "gsr error: display \"%s\" not found, expected one of:\n", window_str); + fprintf(stderr, " \"screen\" (%dx%d+%d+%d)\n", screens_width, screens_height, 0, 0); + fprintf(stderr, " \"screen-direct\" (%dx%d+%d+%d)\n", screens_width, screens_height, 0, 0); + fprintf(stderr, " \"screen-direct-force\" (%dx%d+%d+%d)\n", screens_width, screens_height, 0, 0); + for_each_active_monitor_output(&egl, GSR_CONNECTION_X11, monitor_output_callback_print, NULL); + _exit(1); + } + } + } + + if(egl.gpu_info.vendor == GSR_GPU_VENDOR_NVIDIA) { + if(wayland) { + gsr_capture_kms_cuda_params kms_params; + kms_params.egl = &egl; + kms_params.display_to_capture = window_str; + kms_params.gpu_inf = gpu_inf; + kms_params.hdr = video_codec_is_hdr(video_codec); + kms_params.color_range = color_range; + kms_params.record_cursor = record_cursor; + capture = gsr_capture_kms_cuda_create(&kms_params); + if(!capture) + _exit(1); + } else { + const char *capture_target = window_str; + bool direct_capture = strcmp(window_str, "screen-direct") == 0; + if(direct_capture) { + capture_target = "screen"; + // TODO: Temporary disable direct capture because push model causes stuttering when it's direct capturing. This might be a nvfbc bug. This does not happen when using a compositor. + direct_capture = false; + fprintf(stderr, "Warning: screen-direct has temporary been disabled as it causes stuttering. This is likely a NvFBC bug. Falling back to \"screen\".\n"); + } + + if(strcmp(window_str, "screen-direct-force") == 0) { + direct_capture = true; + capture_target = "screen"; + } + + gsr_capture_nvfbc_params nvfbc_params; + nvfbc_params.egl = &egl; + nvfbc_params.display_to_capture = capture_target; + nvfbc_params.fps = fps; + nvfbc_params.pos = { 0, 0 }; + nvfbc_params.size = { 0, 0 }; + nvfbc_params.direct_capture = direct_capture; + nvfbc_params.overclock = overclock; + nvfbc_params.hdr = video_codec_is_hdr(video_codec); + nvfbc_params.color_range = color_range; + nvfbc_params.record_cursor = record_cursor; + capture = gsr_capture_nvfbc_create(&nvfbc_params); + if(!capture) + _exit(1); + } + } else { + gsr_capture_kms_vaapi_params kms_params; + kms_params.egl = &egl; + kms_params.display_to_capture = window_str; + kms_params.gpu_inf = gpu_inf; + kms_params.hdr = video_codec_is_hdr(video_codec); + kms_params.color_range = color_range; + kms_params.record_cursor = record_cursor; + capture = gsr_capture_kms_vaapi_create(&kms_params); + if(!capture) + _exit(1); + } + } else { + if(wayland) { + fprintf(stderr, "Error: GPU Screen Recorder window capture only works in a pure X11 session. Xwayland is not supported. You can record a monitor instead on wayland\n"); + _exit(2); + } + + errno = 0; + src_window_id = strtol(window_str, nullptr, 0); + if(src_window_id == None || errno == EINVAL) { + fprintf(stderr, "Invalid window number %s\n", window_str); + usage(); + } + } + + if(!capture) { + switch(egl.gpu_info.vendor) { + case GSR_GPU_VENDOR_AMD: + case GSR_GPU_VENDOR_INTEL: { + gsr_capture_xcomposite_vaapi_params xcomposite_params; + xcomposite_params.base.egl = &egl; + xcomposite_params.base.window = src_window_id; + xcomposite_params.base.follow_focused = follow_focused; + xcomposite_params.base.region_size = region_size; + xcomposite_params.base.color_range = color_range; + xcomposite_params.base.record_cursor = record_cursor; + capture = gsr_capture_xcomposite_vaapi_create(&xcomposite_params); + if(!capture) + _exit(1); + break; + } + case GSR_GPU_VENDOR_NVIDIA: { + gsr_capture_xcomposite_cuda_params xcomposite_params; + xcomposite_params.base.egl = &egl; + xcomposite_params.base.window = src_window_id; + xcomposite_params.base.follow_focused = follow_focused; + xcomposite_params.base.region_size = region_size; + xcomposite_params.base.color_range = color_range; + xcomposite_params.base.record_cursor = record_cursor; + xcomposite_params.overclock = overclock; + capture = gsr_capture_xcomposite_cuda_create(&xcomposite_params); + if(!capture) + _exit(1); + break; + } + } + } + + return capture; +} + +struct Arg { + std::vector<const char*> values; + bool optional = false; + bool list = false; + + const char* value() const { + if(values.empty()) + return nullptr; + return values.front(); + } }; int main(int argc, char **argv) { - signal(SIGINT, int_handler); + signal(SIGINT, stop_handler); signal(SIGUSR1, save_replay_handler); + signal(SIGUSR2, toggle_pause_handler); + + // Stop nvidia driver from buffering frames + setenv("__GL_MaxFramesAllowed", "1", true); + // If this is set to 1 then cuGraphicsGLRegisterImage will fail for egl context with error: invalid OpenGL or DirectX context, + // so we overwrite it + setenv("__GL_THREADED_OPTIMIZATIONS", "0", true); + // Some people set this to nvidia (for nvdec) or vdpau (for nvidia vdpau), which breaks gpu screen recorder since + // nvidia doesn't support vaapi and nvidia-vaapi-driver doesn't support encoding yet. + // Let vaapi find the match vaapi driver instead of forcing a specific one. + unsetenv("LIBVA_DRIVER_NAME"); if(argc <= 1) usage_full(); @@ -1081,6 +1631,11 @@ int main(int argc, char **argv) { if(argc == 2 && (strcmp(argv[1], "-h") == 0 || strcmp(argv[1], "--help") == 0)) usage_full(); + if(argc == 2 && strcmp(argv[1], "--list-supported-video-codecs") == 0) { + list_supported_video_codecs(); + _exit(0); + } + //av_log_set_level(AV_LOG_TRACE); std::map<std::string, Arg> args = { @@ -1094,10 +1649,15 @@ int main(int argc, char **argv) { { "-r", Arg { {}, true, false } }, { "-k", Arg { {}, true, false } }, { "-ac", Arg { {}, true, false } }, + { "-ab", Arg { {}, true, false } }, { "-oc", Arg { {}, true, false } }, { "-fm", Arg { {}, true, false } }, { "-pixfmt", Arg { {}, true, false } }, { "-v", Arg { {}, true, false } }, + { "-mf", Arg { {}, true, false } }, + { "-sc", Arg { {}, true, false } }, + { "-cr", Arg { {}, true, false } }, + { "-cursor", Arg { {}, true, false } }, }; for(int i = 1; i < argc; i += 2) { @@ -1127,21 +1687,27 @@ int main(int argc, char **argv) { } } - VideoCodec video_codec = VideoCodec::H265; + VideoCodec video_codec = VideoCodec::HEVC; const char *video_codec_to_use = args["-k"].value(); if(!video_codec_to_use) video_codec_to_use = "auto"; if(strcmp(video_codec_to_use, "h264") == 0) { video_codec = VideoCodec::H264; - } else if(strcmp(video_codec_to_use, "h265") == 0) { - video_codec = VideoCodec::H265; + } else if(strcmp(video_codec_to_use, "h265") == 0 || strcmp(video_codec_to_use, "hevc") == 0) { + video_codec = VideoCodec::HEVC; + } else if(strcmp(video_codec_to_use, "hevc_hdr") == 0) { + video_codec = VideoCodec::HEVC_HDR; + } else if(strcmp(video_codec_to_use, "av1") == 0) { + video_codec = VideoCodec::AV1; + } else if(strcmp(video_codec_to_use, "av1_hdr") == 0) { + video_codec = VideoCodec::AV1_HDR; } else if(strcmp(video_codec_to_use, "auto") != 0) { - fprintf(stderr, "Error: -k should either be either 'auto', 'h264' or 'h265', got: '%s'\n", video_codec_to_use); + fprintf(stderr, "Error: -k should either be either 'auto', 'h264', 'hevc', 'hevc_hdr', 'av1' or 'av1_hdr', got: '%s'\n", video_codec_to_use); usage(); } - AudioCodec audio_codec = AudioCodec::OPUS; + AudioCodec audio_codec = AudioCodec::AAC; const char *audio_codec_to_use = args["-ac"].value(); if(!audio_codec_to_use) audio_codec_to_use = "aac"; @@ -1157,10 +1723,19 @@ int main(int argc, char **argv) { usage(); } - if(audio_codec != AudioCodec::AAC) { + if(audio_codec == AudioCodec::OPUS || audio_codec == AudioCodec::FLAC) { + fprintf(stderr, "Warning: opus and flac audio codecs are temporary disabled, using aac audio codec instead\n"); audio_codec_to_use = "aac"; audio_codec = AudioCodec::AAC; - fprintf(stderr, "Info: audio codec is forcefully set to aac at the moment because of issues with opus/flac. This is a temporary issue\n"); + } + + int audio_bitrate = 0; + const char *audio_bitrate_str = args["-ab"].value(); + if(audio_bitrate_str) { + if(sscanf(audio_bitrate_str, "%d", &audio_bitrate) != 1) { + fprintf(stderr, "Error: -ab argument \"%s\" is not an integer\n", audio_bitrate_str); + usage(); + } } bool overclock = false; @@ -1191,6 +1766,48 @@ int main(int argc, char **argv) { usage(); } + bool record_cursor = true; + const char *record_cursor_str = args["-cursor"].value(); + if(!record_cursor_str) + record_cursor_str = "yes"; + + if(strcmp(record_cursor_str, "yes") == 0) { + record_cursor = true; + } else if(strcmp(record_cursor_str, "no") == 0) { + record_cursor = false; + } else { + fprintf(stderr, "Error: -cursor should either be either 'yes' or 'no', got: '%s'\n", record_cursor_str); + usage(); + } + + bool make_folders = false; + const char *make_folders_str = args["-mf"].value(); + if(!make_folders_str) + make_folders_str = "no"; + + if(strcmp(make_folders_str, "yes") == 0) { + make_folders = true; + } else if(strcmp(make_folders_str, "no") == 0) { + make_folders = false; + } else { + fprintf(stderr, "Error: -mf should either be either 'yes' or 'no', got: '%s'\n", make_folders_str); + usage(); + } + + const char *recording_saved_script = args["-sc"].value(); + if(recording_saved_script) { + struct stat buf; + if(stat(recording_saved_script, &buf) == -1 || !S_ISREG(buf.st_mode)) { + fprintf(stderr, "Error: Script \"%s\" either doesn't exist or it's not a file\n", recording_saved_script); + usage(); + } + + if(!(buf.st_mode & S_IXUSR)) { + fprintf(stderr, "Error: Script \"%s\" is not executable\n", recording_saved_script); + usage(); + } + } + PixelFormat pixel_format = PixelFormat::YUV420; const char *pixfmt = args["-pixfmt"].value(); if(!pixfmt) @@ -1206,14 +1823,23 @@ int main(int argc, char **argv) { } const Arg &audio_input_arg = args["-a"]; - const std::vector<AudioInput> audio_inputs = get_pulseaudio_inputs(); + std::vector<AudioInput> audio_inputs; + if(!audio_input_arg.values.empty()) + audio_inputs = get_pulseaudio_inputs(); std::vector<MergedAudioInputs> requested_audio_inputs; + bool uses_amix = false; // Manually check if the audio inputs we give exist. This is only needed for pipewire, not pulseaudio. // Pipewire instead DEFAULTS TO THE DEFAULT AUDIO INPUT. THAT'S RETARDED. // OH, YOU MISSPELLED THE AUDIO INPUT? FUCK YOU for(const char *audio_input : audio_input_arg.values) { + if(!audio_input || audio_input[0] == '\0') + continue; + requested_audio_inputs.push_back({parse_audio_input_arg(audio_input)}); + if(requested_audio_inputs.back().audio_inputs.size() > 1) + uses_amix = true; + for(AudioInput &request_audio_input : requested_audio_inputs.back().audio_inputs) { bool match = false; for(const auto &existing_audio_input : audio_inputs) { @@ -1229,7 +1855,7 @@ int main(int argc, char **argv) { if(!match) { fprintf(stderr, "Error: Audio input device '%s' is not a valid audio device, expected one of:\n", request_audio_input.name.c_str()); for(const auto &existing_audio_input : audio_inputs) { - fprintf(stderr, " %s\n", existing_audio_input.name.c_str()); + fprintf(stderr, " %s (%s)\n", existing_audio_input.name.c_str(), existing_audio_input.description.c_str()); } _exit(2); } @@ -1237,6 +1863,9 @@ int main(int argc, char **argv) { } const char *container_format = args["-c"].value(); + if(container_format && strcmp(container_format, "mkv") == 0) + container_format = "matroska"; + int fps = atoi(args["-f"].value()); if(fps == 0) { fprintf(stderr, "Invalid fps argument: %s\n", args["-f"].value()); @@ -1271,53 +1900,67 @@ int main(int argc, char **argv) { fprintf(stderr, "Error: option -r has to be between 5 and 1200, was: %s\n", replay_buffer_size_secs_str); _exit(1); } - replay_buffer_size_secs += 5; // Add a few seconds to account of lost packets because of non-keyframe packets skipped + replay_buffer_size_secs += 3; // Add a few seconds to account of lost packets because of non-keyframe packets skipped } + const char *window_str = strdup(args["-w"].value()); + + bool wayland = false; Display *dpy = XOpenDisplay(nullptr); if (!dpy) { - fprintf(stderr, "Error: Failed to open display. Make sure you are running x11\n"); - _exit(2); + wayland = true; + fprintf(stderr, "Warning: failed to connect to the X server. Assuming wayland is running without Xwayland\n"); } XSetErrorHandler(x11_error_handler); XSetIOErrorHandler(x11_io_error_handler); - if(is_xwayland(dpy)) { - fprintf(stderr, "Error: GPU Screen Recorder only works in a pure X11 session. Xwayland is not supported\n"); - _exit(2); + if(!wayland) + wayland = is_xwayland(dpy); + + if(video_codec_is_hdr(video_codec) && !wayland) { + fprintf(stderr, "Error: hdr video codec option %s is not available on X11\n", video_codec_to_use); + _exit(1); + } + + const bool is_monitor_capture = strcmp(window_str, "focused") != 0 && contains_non_hex_number(window_str); + gsr_egl egl; + if(!gsr_egl_load(&egl, dpy, wayland, is_monitor_capture)) { + fprintf(stderr, "gsr error: failed to load opengl\n"); + _exit(1); } - gsr_gpu_info gpu_inf; bool very_old_gpu = false; - if(!gl_get_gpu_info(dpy, &gpu_inf)) - _exit(2); - if(gpu_inf.vendor == GSR_GPU_VENDOR_NVIDIA && gpu_inf.gpu_version != 0 && gpu_inf.gpu_version < 900) { + if(egl.gpu_info.vendor == GSR_GPU_VENDOR_NVIDIA && egl.gpu_info.gpu_version != 0 && egl.gpu_info.gpu_version < 900) { fprintf(stderr, "Info: your gpu appears to be very old (older than maxwell architecture). Switching to lower preset\n"); very_old_gpu = true; } - if(gpu_inf.vendor != GSR_GPU_VENDOR_NVIDIA && overclock) { - fprintf(stderr, "Info: overclock option has no effect on amd/intel, ignoring option...\n"); + if(egl.gpu_info.vendor != GSR_GPU_VENDOR_NVIDIA && overclock) { + fprintf(stderr, "Info: overclock option has no effect on amd/intel, ignoring option\n"); } - char card_path[128]; - card_path[0] = '\0'; - if(gpu_inf.vendor != GSR_GPU_VENDOR_NVIDIA) { + if(egl.gpu_info.vendor == GSR_GPU_VENDOR_NVIDIA && overclock && wayland) { + fprintf(stderr, "Info: overclocking is not possible on nvidia on wayland, ignoring option\n"); + } + + egl.card_path[0] = '\0'; + if(wayland || egl.gpu_info.vendor != GSR_GPU_VENDOR_NVIDIA) { // TODO: Allow specifying another card, and in other places - if(!gsr_get_valid_card_path(card_path)) { - fprintf(stderr, "Error: no /dev/dri/cardX device found\n"); + if(!gsr_get_valid_card_path(&egl, egl.card_path)) { + fprintf(stderr, "Error: no /dev/dri/cardX device found. If you are running GPU Screen Recorder with prime-run then try running without it\n"); _exit(2); } } // TODO: Fix constant framerate not working properly on amd/intel because capture framerate gets locked to the same framerate as // game framerate, which doesn't work well when you need to encode multiple duplicate frames (AMD/Intel is slow at encoding!). + // It also appears to skip audio frames on nvidia wayland? why? that should be fine, but it causes video stuttering because of audio/video sync. FramerateMode framerate_mode; const char *framerate_mode_str = args["-fm"].value(); if(!framerate_mode_str) - framerate_mode_str = gpu_inf.vendor == GSR_GPU_VENDOR_NVIDIA ? "cfr" : "vfr"; + framerate_mode_str = "vfr"; if(strcmp(framerate_mode_str, "cfr") == 0) { framerate_mode = FramerateMode::CONSTANT; @@ -1328,148 +1971,58 @@ int main(int argc, char **argv) { usage(); } - const char *screen_region = args["-s"].value(); - const char *window_str = args["-w"].value(); + gsr_color_range color_range; + const char *color_range_str = args["-cr"].value(); + if(!color_range_str) + color_range_str = "limited"; - if(screen_region && strcmp(window_str, "focused") != 0) { - fprintf(stderr, "Error: option -s is only available when using -w focused\n"); + if(strcmp(color_range_str, "limited") == 0) { + color_range = GSR_COLOR_RANGE_LIMITED; + } else if(strcmp(color_range_str, "full") == 0) { + color_range = GSR_COLOR_RANGE_FULL; + } else { + fprintf(stderr, "Error: -cr should either be either 'limited' or 'full', got: '%s'\n", color_range_str); usage(); } - vec2i region_size = { 0, 0 }; - Window src_window_id = None; - bool follow_focused = false; - - gsr_capture *capture = nullptr; - if(strcmp(window_str, "focused") == 0) { - if(!screen_region) { - fprintf(stderr, "Error: option -s is required when using -w focused\n"); - usage(); - } - - if(sscanf(screen_region, "%dx%d", ®ion_size.x, ®ion_size.y) != 2) { - fprintf(stderr, "Error: invalid value for option -s '%s', expected a value in format WxH\n", screen_region); - usage(); - } - - if(region_size.x <= 0 || region_size.y <= 0) { - fprintf(stderr, "Error: invalud value for option -s '%s', expected width and height to be greater than 0\n", screen_region); - usage(); - } - - follow_focused = true; - } else if(contains_non_hex_number(window_str)) { - if(strcmp(window_str, "screen") != 0 && strcmp(window_str, "screen-direct") != 0 && strcmp(window_str, "screen-direct-force") != 0) { - gsr_monitor gmon; - if(!get_monitor_by_name(dpy, window_str, &gmon)) { - fprintf(stderr, "gsr error: display \"%s\" not found, expected one of:\n", window_str); - fprintf(stderr, " \"screen\" (%dx%d+%d+%d)\n", XWidthOfScreen(DefaultScreenOfDisplay(dpy)), XHeightOfScreen(DefaultScreenOfDisplay(dpy)), 0, 0); - fprintf(stderr, " \"screen-direct\" (%dx%d+%d+%d)\n", XWidthOfScreen(DefaultScreenOfDisplay(dpy)), XHeightOfScreen(DefaultScreenOfDisplay(dpy)), 0, 0); - fprintf(stderr, " \"screen-direct-force\" (%dx%d+%d+%d)\n", XWidthOfScreen(DefaultScreenOfDisplay(dpy)), XHeightOfScreen(DefaultScreenOfDisplay(dpy)), 0, 0); - for_each_active_monitor_output(dpy, monitor_output_callback_print, NULL); - _exit(1); - } - } - - if(gpu_inf.vendor == GSR_GPU_VENDOR_NVIDIA) { - const char *capture_target = window_str; - bool direct_capture = strcmp(window_str, "screen-direct") == 0; - if(direct_capture) { - capture_target = "screen"; - // TODO: Temporary disable direct capture because push model causes stuttering when it's direct capturing. This might be a nvfbc bug. This does not happen when using a compositor. - direct_capture = false; - fprintf(stderr, "Warning: screen-direct has temporary been disabled as it causes stuttering. This is likely a NvFBC bug. Falling back to \"screen\".\n"); - } - - if(strcmp(window_str, "screen-direct-force") == 0) { - direct_capture = true; - capture_target = "screen"; - } - - gsr_capture_nvfbc_params nvfbc_params; - nvfbc_params.dpy = dpy; - nvfbc_params.display_to_capture = capture_target; - nvfbc_params.fps = fps; - nvfbc_params.pos = { 0, 0 }; - nvfbc_params.size = { 0, 0 }; - nvfbc_params.direct_capture = direct_capture; - nvfbc_params.overclock = overclock; - capture = gsr_capture_nvfbc_create(&nvfbc_params); - if(!capture) - _exit(1); - } else { - const char *capture_target = window_str; - if(strcmp(window_str, "screen-direct") == 0 || strcmp(window_str, "screen-direct-force") == 0) { - capture_target = "screen"; - } - - gsr_capture_kms_vaapi_params kms_params; - kms_params.display_to_capture = capture_target; - kms_params.gpu_inf = gpu_inf; - kms_params.card_path = card_path; - capture = gsr_capture_kms_vaapi_create(&kms_params); - if(!capture) - _exit(1); - } - } else { - errno = 0; - src_window_id = strtol(window_str, nullptr, 0); - if(src_window_id == None || errno == EINVAL) { - fprintf(stderr, "Invalid window number %s\n", window_str); - usage(); - } - } + const char *screen_region = args["-s"].value(); - if(!capture) { - switch(gpu_inf.vendor) { - case GSR_GPU_VENDOR_AMD: { - gsr_capture_xcomposite_vaapi_params xcomposite_params; - xcomposite_params.window = src_window_id; - xcomposite_params.follow_focused = follow_focused; - xcomposite_params.region_size = region_size; - xcomposite_params.card_path = card_path; - capture = gsr_capture_xcomposite_vaapi_create(&xcomposite_params); - if(!capture) - _exit(1); - break; - } - case GSR_GPU_VENDOR_INTEL: { - gsr_capture_xcomposite_vaapi_params xcomposite_params; - xcomposite_params.window = src_window_id; - xcomposite_params.follow_focused = follow_focused; - xcomposite_params.region_size = region_size; - xcomposite_params.card_path = card_path; - capture = gsr_capture_xcomposite_vaapi_create(&xcomposite_params); - if(!capture) - _exit(1); - break; - } - case GSR_GPU_VENDOR_NVIDIA: { - gsr_capture_xcomposite_cuda_params xcomposite_params; - xcomposite_params.window = src_window_id; - xcomposite_params.follow_focused = follow_focused; - xcomposite_params.region_size = region_size; - xcomposite_params.overclock = overclock; - capture = gsr_capture_xcomposite_cuda_create(&xcomposite_params); - if(!capture) - _exit(1); - break; - } - } + if(screen_region && strcmp(window_str, "focused") != 0) { + fprintf(stderr, "Error: option -s is only available when using -w focused\n"); + usage(); } + bool is_livestream = false; const char *filename = args["-o"].value(); if(filename) { - if(replay_buffer_size_secs != -1) { - if(!container_format) { - fprintf(stderr, "Error: option -c is required when using option -r\n"); - usage(); + is_livestream = is_livestream_path(filename); + if(is_livestream) { + if(replay_buffer_size_secs != -1) { + fprintf(stderr, "Error: replay mode is not applicable to live streaming\n"); + _exit(1); } + } else { + if(replay_buffer_size_secs == -1) { + char directory_buf[PATH_MAX]; + strcpy(directory_buf, filename); + char *directory = dirname(directory_buf); + if(strcmp(directory, ".") != 0 && strcmp(directory, "/") != 0) { + if(create_directory_recursive(directory) != 0) { + fprintf(stderr, "Error: failed to create directory for output file: %s\n", filename); + _exit(1); + } + } + } else { + if(!container_format) { + fprintf(stderr, "Error: option -c is required when using option -r\n"); + usage(); + } - struct stat buf; - if(stat(filename, &buf) == -1 || !S_ISDIR(buf.st_mode)) { - fprintf(stderr, "Error: directory \"%s\" does not exist or is not a directory\n", filename); - usage(); + struct stat buf; + if(stat(filename, &buf) != -1 && !S_ISDIR(buf.st_mode)) { + fprintf(stderr, "Error: File \"%s\" exists but it's not a directory\n", filename); + usage(); + } } } } else { @@ -1490,7 +2043,10 @@ int main(int argc, char **argv) { // The output format is automatically guessed by the file extension avformat_alloc_output_context2(&av_format_context, nullptr, container_format, filename); if (!av_format_context) { - fprintf(stderr, "Error: Failed to deduce container format from file extension\n"); + if(container_format) + fprintf(stderr, "Error: Container format '%s' (argument -c) is not valid\n", container_format); + else + fprintf(stderr, "Error: Failed to deduce container format from file extension\n"); _exit(1); } @@ -1503,11 +2059,18 @@ int main(int argc, char **argv) { file_extension = file_extension.substr(0, comma_index); } + if(egl.gpu_info.vendor != GSR_GPU_VENDOR_NVIDIA && file_extension == "mkv" && strcmp(video_codec_to_use, "h264") == 0) { + video_codec_to_use = "hevc"; + video_codec = VideoCodec::HEVC; + fprintf(stderr, "Warning: video codec was forcefully set to hevc because mkv container is used and mesa (AMD and Intel driver) does not support h264 in mkv files\n"); + } + switch(audio_codec) { case AudioCodec::AAC: { break; } case AudioCodec::OPUS: { + // TODO: Also check mpegts? if(file_extension != "mp4" && file_extension != "mkv") { audio_codec_to_use = "aac"; audio_codec = AudioCodec::AAC; @@ -1516,10 +2079,15 @@ int main(int argc, char **argv) { break; } case AudioCodec::FLAC: { + // TODO: Also check mpegts? if(file_extension != "mp4" && file_extension != "mkv") { audio_codec_to_use = "aac"; audio_codec = AudioCodec::AAC; fprintf(stderr, "Warning: flac audio codec is only supported by .mp4 and .mkv files, falling back to aac instead\n"); + } else if(uses_amix) { + audio_codec_to_use = "opus"; + audio_codec = AudioCodec::OPUS; + fprintf(stderr, "Warning: flac audio codec is not supported when mixing audio sources, falling back to opus instead\n"); } break; } @@ -1527,68 +2095,125 @@ int main(int argc, char **argv) { const double target_fps = 1.0 / (double)fps; - if(strcmp(video_codec_to_use, "auto") == 0) { - if(gpu_inf.vendor == GSR_GPU_VENDOR_INTEL) { - const AVCodec *h264_codec = find_h264_encoder(gpu_inf.vendor, card_path); + const bool video_codec_auto = strcmp(video_codec_to_use, "auto") == 0; + if(video_codec_auto) { + if(egl.gpu_info.vendor == GSR_GPU_VENDOR_INTEL) { + const AVCodec *h264_codec = find_h264_encoder(egl.gpu_info.vendor, egl.card_path); if(!h264_codec) { - fprintf(stderr, "Info: using h265 encoder because a codec was not specified and your gpu does not support h264\n"); - video_codec_to_use = "h265"; - video_codec = VideoCodec::H265; + fprintf(stderr, "Info: using hevc encoder because a codec was not specified and your gpu does not support h264\n"); + video_codec_to_use = "hevc"; + video_codec = VideoCodec::HEVC; } else { fprintf(stderr, "Info: using h264 encoder because a codec was not specified\n"); video_codec_to_use = "h264"; video_codec = VideoCodec::H264; } } else { - const AVCodec *h265_codec = find_h265_encoder(gpu_inf.vendor, card_path); + const AVCodec *h265_codec = find_h265_encoder(egl.gpu_info.vendor, egl.card_path); - // h265 generally allows recording at a higher resolution than h264 on nvidia cards. On a gtx 1080 4k is the max resolution for h264 but for h265 it's 8k. - // Another important info is that when recording at a higher fps than.. 60? h265 has very bad performance. For example when recording at 144 fps the fps drops to 1 + if(h265_codec && fps > 60) { + fprintf(stderr, "Warning: recording at higher fps than 60 with hevc might result in recording at a very low fps. If this happens, switch to h264 or av1\n"); + } + + // hevc generally allows recording at a higher resolution than h264 on nvidia cards. On a gtx 1080 4k is the max resolution for h264 but for hevc it's 8k. + // Another important info is that when recording at a higher fps than.. 60? hevc has very bad performance. For example when recording at 144 fps the fps drops to 1 // while with h264 the fps doesn't drop. if(!h265_codec) { - fprintf(stderr, "Info: using h264 encoder because a codec was not specified and your gpu does not support h265\n"); - video_codec_to_use = "h264"; - video_codec = VideoCodec::H264; - } else if(fps > 60) { - fprintf(stderr, "Info: using h264 encoder because a codec was not specified and fps is more than 60\n"); + fprintf(stderr, "Info: using h264 encoder because a codec was not specified and your gpu does not support hevc\n"); video_codec_to_use = "h264"; video_codec = VideoCodec::H264; } else { - fprintf(stderr, "Info: using h265 encoder because a codec was not specified\n"); - video_codec_to_use = "h265"; - video_codec = VideoCodec::H265; + fprintf(stderr, "Info: using hevc encoder because a codec was not specified\n"); + video_codec_to_use = "hevc"; + video_codec = VideoCodec::HEVC; } } } - //bool use_hevc = strcmp(window_str, "screen") == 0 || strcmp(window_str, "screen-direct") == 0; - if(video_codec != VideoCodec::H264 && strcmp(file_extension.c_str(), "flv") == 0) { + // TODO: Allow hevc, vp9 and av1 in (enhanced) flv (supported since ffmpeg 6.1) + const bool is_flv = strcmp(file_extension.c_str(), "flv") == 0; + if(video_codec != VideoCodec::H264 && is_flv) { video_codec_to_use = "h264"; video_codec = VideoCodec::H264; - fprintf(stderr, "Warning: h265 is not compatible with flv, falling back to h264 instead.\n"); + fprintf(stderr, "Warning: hevc/av1 is not compatible with flv, falling back to h264 instead.\n"); } const AVCodec *video_codec_f = nullptr; switch(video_codec) { case VideoCodec::H264: - video_codec_f = find_h264_encoder(gpu_inf.vendor, card_path); + video_codec_f = find_h264_encoder(egl.gpu_info.vendor, egl.card_path); break; - case VideoCodec::H265: - video_codec_f = find_h265_encoder(gpu_inf.vendor, card_path); + case VideoCodec::HEVC: + case VideoCodec::HEVC_HDR: + video_codec_f = find_h265_encoder(egl.gpu_info.vendor, egl.card_path); + break; + case VideoCodec::AV1: + case VideoCodec::AV1_HDR: + video_codec_f = find_av1_encoder(egl.gpu_info.vendor, egl.card_path); break; } + if(!video_codec_auto && !video_codec_f && !is_flv) { + switch(video_codec) { + case VideoCodec::H264: { + fprintf(stderr, "Warning: selected video codec h264 is not supported, trying hevc instead\n"); + video_codec_to_use = "hevc"; + video_codec = VideoCodec::HEVC; + video_codec_f = find_h265_encoder(egl.gpu_info.vendor, egl.card_path); + break; + } + case VideoCodec::HEVC: + case VideoCodec::HEVC_HDR: { + fprintf(stderr, "Warning: selected video codec hevc is not supported, trying h264 instead\n"); + video_codec_to_use = "h264"; + video_codec = VideoCodec::H264; + video_codec_f = find_h264_encoder(egl.gpu_info.vendor, egl.card_path); + break; + } + case VideoCodec::AV1: + case VideoCodec::AV1_HDR: { + fprintf(stderr, "Warning: selected video codec av1 is not supported, trying h264 instead\n"); + video_codec_to_use = "h264"; + video_codec = VideoCodec::H264; + video_codec_f = find_h264_encoder(egl.gpu_info.vendor, egl.card_path); + break; + } + } + } + if(!video_codec_f) { - const char *video_codec_name = video_codec == VideoCodec::H264 ? "h264" : "h265"; + const char *video_codec_name = ""; + switch(video_codec) { + case VideoCodec::H264: { + video_codec_name = "h264"; + break; + } + case VideoCodec::HEVC: + case VideoCodec::HEVC_HDR: { + video_codec_name = "hevc"; + break; + } + case VideoCodec::AV1: + case VideoCodec::AV1_HDR: { + video_codec_name = "av1"; + break; + } + } + fprintf(stderr, "Error: your gpu does not support '%s' video codec. If you are sure that your gpu does support '%s' video encoding and you are using an AMD/Intel GPU,\n" - " then it's possible that your distro has disabled hardware accelerated video encoding for '%s' video codec.\n" - " This may be the case on corporate distros such as Manjaro.\n" - " You can test this by running 'vainfo | grep VAEntrypointEncSlice' to see if it matches any H264/HEVC profile. vainfo is part of libva-utils.\n" - " On such distros, you need to manually install mesa from source to enable H264/HEVC hardware acceleration, or use a more user friendly distro.\n", video_codec_name, video_codec_name, video_codec_name); + " then make sure you have installed the GPU specific vaapi packages (intel-media-driver, libva-intel-driver or libva-mesa-driver).\n" + " It's also possible that your distro has disabled hardware accelerated video encoding for '%s' video codec.\n" + " This may be the case on corporate distros such as Manjaro, Fedora or OpenSUSE.\n" + " You can test this by running 'vainfo | grep VAEntrypointEncSlice' to see if it matches any H264/HEVC profile.\n" + " On such distros, you need to manually install mesa from source to enable H264/HEVC hardware acceleration, or use a more user friendly distro. Alternatively record with AV1 if supported by your GPU.\n" + " You can alternatively use the flatpak version of GPU Screen Recorder (https://flathub.org/apps/com.dec05eba.gpu_screen_recorder) which bypasses system issues with patented H264/HEVC codecs.\n" + " Make sure you have mesa-extra freedesktop runtime installed when using the flatpak (this should be the default), which can be installed with this command:\n" + " flatpak install --system org.freedesktop.Platform.GL.default//23.08-extra", video_codec_name, video_codec_name, video_codec_name); _exit(2); } - const bool is_livestream = is_livestream_path(filename); + gsr_capture *capture = create_capture_impl(window_str, screen_region, wayland, egl.gpu_info, egl, fps, overclock, video_codec, color_range, record_cursor); + // (Some?) livestreaming services require at least one audio track to work. // If not audio is provided then create one silent audio track. if(is_livestream && requested_audio_inputs.empty()) { @@ -1598,31 +2223,59 @@ int main(int argc, char **argv) { requested_audio_inputs.push_back(std::move(mai)); } + if(is_livestream && framerate_mode != FramerateMode::CONSTANT) { + fprintf(stderr, "Info: framerate mode was forcefully set to \"cfr\" because live streaming was detected\n"); + framerate_mode = FramerateMode::CONSTANT; + framerate_mode_str = "cfr"; + } + + if(is_livestream && recording_saved_script) { + fprintf(stderr, "Warning: live stream detected, -sc script is ignored\n"); + recording_saved_script = nullptr; + } + AVStream *video_stream = nullptr; std::vector<AudioTrack> audio_tracks; + const bool hdr = video_codec_is_hdr(video_codec); - AVCodecContext *video_codec_context = create_video_codec_context(gpu_inf.vendor == GSR_GPU_VENDOR_NVIDIA ? AV_PIX_FMT_CUDA : AV_PIX_FMT_VAAPI, quality, fps, video_codec_f, is_livestream, gpu_inf.vendor, framerate_mode); + AVCodecContext *video_codec_context = create_video_codec_context(egl.gpu_info.vendor == GSR_GPU_VENDOR_NVIDIA ? AV_PIX_FMT_CUDA : AV_PIX_FMT_VAAPI, quality, fps, video_codec_f, is_livestream, egl.gpu_info.vendor, framerate_mode, hdr, color_range); if(replay_buffer_size_secs == -1) video_stream = create_stream(av_format_context, video_codec_context); - if(gsr_capture_start(capture, video_codec_context) != 0) { - fprintf(stderr, "gsr error: gsr_capture_start failed\n"); + AVFrame *video_frame = av_frame_alloc(); + if(!video_frame) { + fprintf(stderr, "Error: Failed to allocate video frame\n"); _exit(1); } + video_frame->format = video_codec_context->pix_fmt; + video_frame->width = video_codec_context->width; + video_frame->height = video_codec_context->height; + video_frame->color_range = video_codec_context->color_range; + video_frame->color_primaries = video_codec_context->color_primaries; + video_frame->color_trc = video_codec_context->color_trc; + video_frame->colorspace = video_codec_context->colorspace; + video_frame->chroma_location = video_codec_context->chroma_sample_location; + + int capture_result = gsr_capture_start(capture, video_codec_context, video_frame); + if(capture_result != 0) { + fprintf(stderr, "gsr error: gsr_capture_start failed\n"); + _exit(capture_result); + } - open_video(video_codec_context, quality, very_old_gpu, gpu_inf.vendor, pixel_format); + open_video(video_codec_context, quality, very_old_gpu, egl.gpu_info.vendor, pixel_format, hdr); if(video_stream) avcodec_parameters_from_context(video_stream->codecpar, video_codec_context); int audio_stream_index = VIDEO_STREAM_INDEX + 1; for(const MergedAudioInputs &merged_audio_inputs : requested_audio_inputs) { - AVCodecContext *audio_codec_context = create_audio_codec_context(fps, audio_codec); + const bool use_amix = merged_audio_inputs.audio_inputs.size() > 1; + AVCodecContext *audio_codec_context = create_audio_codec_context(fps, audio_codec, use_amix, audio_bitrate); AVStream *audio_stream = nullptr; if(replay_buffer_size_secs == -1) audio_stream = create_stream(av_format_context, audio_codec_context); - AVFrame *audio_frame = open_audio(audio_codec_context); + open_audio(audio_codec_context); if(audio_stream) avcodec_parameters_from_context(audio_stream->codecpar, audio_codec_context); @@ -1637,7 +2290,6 @@ int main(int argc, char **argv) { std::vector<AVFilterContext*> src_filter_ctx; AVFilterGraph *graph = nullptr; AVFilterContext *sink = nullptr; - bool use_amix = merged_audio_inputs.audio_inputs.size() > 1; if(use_amix) { int err = init_filter_graph(audio_codec_context, &graph, &sink, src_filter_ctx, merged_audio_inputs.audio_inputs.size()); if(err < 0) { @@ -1669,12 +2321,14 @@ int main(int argc, char **argv) { } } + audio_device.frame = create_audio_frame(audio_codec_context); + audio_device.frame->pts = 0; + audio_devices.push_back(std::move(audio_device)); } AudioTrack audio_track; audio_track.codec_context = audio_codec_context; - audio_track.frame = audio_frame; audio_track.stream = audio_stream; audio_track.audio_devices = std::move(audio_devices); audio_track.graph = graph; @@ -1710,29 +2364,19 @@ int main(int argc, char **argv) { const double start_time_pts = clock_get_monotonic_seconds(); - double start_time = clock_get_monotonic_seconds(); // todo - target_fps to make first frame start immediately? - double frame_timer_start = start_time; + double start_time = clock_get_monotonic_seconds(); + double frame_timer_start = start_time - target_fps; // We want to capture the first frame immediately int fps_counter = 0; - AVFrame *frame = av_frame_alloc(); - if (!frame) { - fprintf(stderr, "Error: Failed to allocate frame\n"); - _exit(1); - } - frame->format = video_codec_context->pix_fmt; - frame->width = video_codec_context->width; - frame->height = video_codec_context->height; - frame->color_range = video_codec_context->color_range; - frame->color_primaries = video_codec_context->color_primaries; - frame->color_trc = video_codec_context->color_trc; - frame->colorspace = video_codec_context->colorspace; - frame->chroma_location = video_codec_context->chroma_sample_location; + bool paused = false; + double paused_time_offset = 0.0; + double paused_time_start = 0.0; std::mutex write_output_mutex; std::mutex audio_filter_mutex; const double record_start_time = clock_get_monotonic_seconds(); - std::deque<AVPacket> frame_data_queue; + std::deque<std::shared_ptr<PacketData>> frame_data_queue; bool frames_erased = false; const size_t audio_buffer_size = 1024 * 4 * 2; // max 4 bytes/sample, 2 channels @@ -1743,11 +2387,14 @@ int main(int argc, char **argv) { } memset(empty_audio, 0, audio_buffer_size); + const double audio_startup_time_seconds = std::max(0.0, 0.089166 - target_fps); + for(AudioTrack &audio_track : audio_tracks) { for(AudioDevice &audio_device : audio_track.audio_devices) { audio_device.thread = std::thread([&]() mutable { const AVSampleFormat sound_device_sample_format = audio_format_to_sample_format(audio_codec_context_get_audio_format(audio_track.codec_context)); - const bool needs_audio_conversion = audio_track.codec_context->sample_fmt != sound_device_sample_format; + // TODO: Always do conversion for now. This fixes issue with stuttering audio on pulseaudio with opus + multiple audio sources merged + const bool needs_audio_conversion = true;//audio_track.codec_context->sample_fmt != sound_device_sample_format; SwrContext *swr = nullptr; if(needs_audio_conversion) { swr = swr_alloc(); @@ -1755,8 +2402,13 @@ int main(int argc, char **argv) { fprintf(stderr, "Failed to create SwrContext\n"); _exit(1); } - av_opt_set_int(swr, "in_channel_layout", AV_CH_LAYOUT_STEREO, 0); - av_opt_set_int(swr, "out_channel_layout", AV_CH_LAYOUT_STEREO, 0); + #if LIBAVUTIL_VERSION_MAJOR <= 56 + av_opt_set_channel_layout(swr, "in_channel_layout", AV_CH_LAYOUT_STEREO, 0); + av_opt_set_channel_layout(swr, "out_channel_layout", AV_CH_LAYOUT_STEREO, 0); + #else + av_opt_set_chlayout(swr, "in_channel_layout", &audio_track.codec_context->ch_layout, 0); + av_opt_set_chlayout(swr, "out_channel_layout", &audio_track.codec_context->ch_layout, 0); + #endif av_opt_set_int(swr, "in_sample_rate", audio_track.codec_context->sample_rate, 0); av_opt_set_int(swr, "out_sample_rate", audio_track.codec_context->sample_rate, 0); av_opt_set_sample_fmt(swr, "in_sample_fmt", sound_device_sample_format, 0); @@ -1764,37 +2416,56 @@ int main(int argc, char **argv) { swr_init(swr); } - const double target_audio_hz = 1.0 / (double)audio_track.codec_context->sample_rate; double received_audio_time = clock_get_monotonic_seconds(); - const int64_t timeout_ms = std::round((1000.0 / (double)audio_track.codec_context->sample_rate) * 1000.0); - int64_t prev_pts = 0; + const double timeout_sec = 1000.0 / (double)audio_track.codec_context->sample_rate; + const int64_t timeout_ms = std::round(timeout_sec * 1000.0); while(running) { void *sound_buffer; int sound_buffer_size = -1; - if(audio_device.sound_device.handle) - sound_buffer_size = sound_device_read_next_chunk(&audio_device.sound_device, &sound_buffer); + //const double time_before_read_seconds = clock_get_monotonic_seconds(); + if(audio_device.sound_device.handle) { + // TODO: use this instead of calculating time to read. But this can fluctuate and we dont want to go back in time, + // also it's 0.0 for some users??? + double latency_seconds = 0.0; + sound_buffer_size = sound_device_read_next_chunk(&audio_device.sound_device, &sound_buffer, timeout_sec, &latency_seconds); + } + const bool got_audio_data = sound_buffer_size >= 0; + //const double time_after_read_seconds = clock_get_monotonic_seconds(); + //const double time_to_read_seconds = time_after_read_seconds - time_before_read_seconds; + const double this_audio_frame_time = (clock_get_monotonic_seconds() - audio_startup_time_seconds) - paused_time_offset; - const double this_audio_frame_time = clock_get_monotonic_seconds(); - if(got_audio_data) - received_audio_time = this_audio_frame_time; + if(paused) { + if(got_audio_data) + received_audio_time = this_audio_frame_time; - int ret = av_frame_make_writable(audio_track.frame); + if(!audio_device.sound_device.handle) + usleep(timeout_ms * 1000); + + continue; + } + + int ret = av_frame_make_writable(audio_device.frame); if (ret < 0) { fprintf(stderr, "Failed to make audio frame writable\n"); break; } // TODO: Is this |received_audio_time| really correct? - int64_t num_missing_frames = std::round((this_audio_frame_time - received_audio_time) / target_audio_hz / (int64_t)audio_track.frame->nb_samples); + const double prev_audio_time = received_audio_time; + const double audio_receive_time_diff = this_audio_frame_time - received_audio_time; + int64_t num_missing_frames = std::round(audio_receive_time_diff / timeout_sec); if(got_audio_data) num_missing_frames = std::max((int64_t)0, num_missing_frames - 1); if(!audio_device.sound_device.handle) num_missing_frames = std::max((int64_t)1, num_missing_frames); - // Jesus is there a better way to do this? I JUST WANT TO KEEP VIDEO AND AUDIO SYNCED HOLY FUCK I WANT TO KILL MYSELF NOW. + if(got_audio_data) + received_audio_time = this_audio_frame_time; + + // Fucking hell is there a better way to do this? I JUST WANT TO KEEP VIDEO AND AUDIO SYNCED HOLY FUCK I WANT TO KILL MYSELF NOW. // THIS PIECE OF SHIT WANTS EMPTY FRAMES OTHERWISE VIDEO PLAYS TOO FAST TO KEEP UP WITH AUDIO OR THE AUDIO PLAYS TOO EARLY. // BUT WE CANT USE DELAYS TO GIVE DUMMY DATA BECAUSE PULSEAUDIO MIGHT GIVE AUDIO A BIG DELAYED!!! // This garbage is needed because we want to produce constant frame rate videos instead of variable frame rate @@ -1806,29 +2477,28 @@ int main(int argc, char **argv) { //audio_track.frame->data[0] = empty_audio; received_audio_time = this_audio_frame_time; if(needs_audio_conversion) - swr_convert(swr, &audio_track.frame->data[0], audio_track.frame->nb_samples, (const uint8_t**)&empty_audio, audio_track.codec_context->frame_size); + swr_convert(swr, &audio_device.frame->data[0], audio_track.codec_context->frame_size, (const uint8_t**)&empty_audio, audio_track.codec_context->frame_size); else - audio_track.frame->data[0] = empty_audio; + audio_device.frame->data[0] = empty_audio; // TODO: Check if duplicate frame can be saved just by writing it with a different pts instead of sending it again std::lock_guard<std::mutex> lock(audio_filter_mutex); for(int i = 0; i < num_missing_frames; ++i) { + const int64_t new_pts = ((prev_audio_time - record_start_time) + timeout_sec * i) * AV_TIME_BASE; + if(new_pts == audio_device.frame->pts) + continue; + + audio_device.frame->pts = new_pts; if(audio_track.graph) { // TODO: av_buffersrc_add_frame - if(av_buffersrc_write_frame(audio_device.src_filter_ctx, audio_track.frame) < 0) { + if(av_buffersrc_write_frame(audio_device.src_filter_ctx, audio_device.frame) < 0) { fprintf(stderr, "Error: failed to add audio frame to filter\n"); } } else { - audio_track.frame->pts = (this_audio_frame_time - record_start_time) * (double)AV_TIME_BASE; - const bool same_pts = audio_track.frame->pts == prev_pts; - prev_pts = audio_track.frame->pts; - if(same_pts) - continue; - - ret = avcodec_send_frame(audio_track.codec_context, audio_track.frame); + ret = avcodec_send_frame(audio_track.codec_context, audio_device.frame); if(ret >= 0) { // TODO: Move to separate thread because this could write to network (for example when livestreaming) - receive_frames(audio_track.codec_context, audio_track.stream_index, audio_track.stream, audio_track.frame->pts, av_format_context, record_start_time, frame_data_queue, replay_buffer_size_secs, frames_erased, write_output_mutex); + receive_frames(audio_track.codec_context, audio_track.stream_index, audio_track.stream, audio_device.frame->pts, av_format_context, record_start_time, frame_data_queue, replay_buffer_size_secs, frames_erased, write_output_mutex, paused_time_offset); } else { fprintf(stderr, "Failed to encode audio!\n"); } @@ -1842,29 +2512,28 @@ int main(int argc, char **argv) { if(got_audio_data) { // TODO: Instead of converting audio, get float audio from alsa. Or does alsa do conversion internally to get this format? if(needs_audio_conversion) - swr_convert(swr, &audio_track.frame->data[0], audio_track.frame->nb_samples, (const uint8_t**)&sound_buffer, audio_track.codec_context->frame_size); + swr_convert(swr, &audio_device.frame->data[0], audio_track.codec_context->frame_size, (const uint8_t**)&sound_buffer, audio_track.codec_context->frame_size); else - audio_track.frame->data[0] = (uint8_t*)sound_buffer; + audio_device.frame->data[0] = (uint8_t*)sound_buffer; - audio_track.frame->pts = (this_audio_frame_time - record_start_time) * (double)AV_TIME_BASE; - const bool same_pts = audio_track.frame->pts == prev_pts; - prev_pts = audio_track.frame->pts; - if(same_pts) - continue; + const int64_t new_pts = (this_audio_frame_time - record_start_time) * AV_TIME_BASE; + if(new_pts != audio_device.frame->pts) { + audio_device.frame->pts = new_pts; - if(audio_track.graph) { - std::lock_guard<std::mutex> lock(audio_filter_mutex); - // TODO: av_buffersrc_add_frame - if(av_buffersrc_write_frame(audio_device.src_filter_ctx, audio_track.frame) < 0) { - fprintf(stderr, "Error: failed to add audio frame to filter\n"); - } - } else { - ret = avcodec_send_frame(audio_track.codec_context, audio_track.frame); - if(ret >= 0) { - // TODO: Move to separate thread because this could write to network (for example when livestreaming) - receive_frames(audio_track.codec_context, audio_track.stream_index, audio_track.stream, audio_track.frame->pts, av_format_context, record_start_time, frame_data_queue, replay_buffer_size_secs, frames_erased, write_output_mutex); + if(audio_track.graph) { + std::lock_guard<std::mutex> lock(audio_filter_mutex); + // TODO: av_buffersrc_add_frame + if(av_buffersrc_write_frame(audio_device.src_filter_ctx, audio_device.frame) < 0) { + fprintf(stderr, "Error: failed to add audio frame to filter\n"); + } } else { - fprintf(stderr, "Failed to encode audio!\n"); + ret = avcodec_send_frame(audio_track.codec_context, audio_device.frame); + if(ret >= 0) { + // TODO: Move to separate thread because this could write to network (for example when livestreaming) + receive_frames(audio_track.codec_context, audio_track.stream_index, audio_track.stream, audio_device.frame->pts, av_format_context, record_start_time, frame_data_queue, replay_buffer_size_secs, frames_erased, write_output_mutex, paused_time_offset); + } else { + fprintf(stderr, "Failed to encode audio!\n"); + } } } } @@ -1884,12 +2553,11 @@ int main(int argc, char **argv) { int64_t video_pts_counter = 0; int64_t video_prev_pts = 0; - int64_t audio_prev_pts = 0; while(running) { double frame_start = clock_get_monotonic_seconds(); - gsr_capture_tick(capture, video_codec_context, &frame); + gsr_capture_tick(capture, video_codec_context); should_stop_error = false; if(gsr_capture_should_stop(capture, &should_stop_error)) { running = 0; @@ -1905,19 +2573,15 @@ int main(int argc, char **argv) { int err = 0; while ((err = av_buffersink_get_frame(audio_track.sink, aframe)) >= 0) { - const double this_audio_frame_time = clock_get_monotonic_seconds(); - aframe->pts = (this_audio_frame_time - record_start_time) * (double)AV_TIME_BASE; - const bool same_pts = aframe->pts == audio_prev_pts; - audio_prev_pts = aframe->pts; - if(same_pts) { - av_frame_unref(aframe); + const double this_audio_frame_time = (clock_get_monotonic_seconds() - audio_startup_time_seconds) - paused_time_offset; + const int64_t new_pts = (this_audio_frame_time - record_start_time) * AV_TIME_BASE; + if(new_pts == aframe->pts) continue; - } - + aframe->pts = new_pts; err = avcodec_send_frame(audio_track.codec_context, aframe); if(err >= 0){ // TODO: Move to separate thread because this could write to network (for example when livestreaming) - receive_frames(audio_track.codec_context, audio_track.stream_index, audio_track.stream, aframe->pts, av_format_context, record_start_time, frame_data_queue, replay_buffer_size_secs, frames_erased, write_output_mutex); + receive_frames(audio_track.codec_context, audio_track.stream_index, audio_track.stream, aframe->pts, av_format_context, record_start_time, frame_data_queue, replay_buffer_size_secs, frames_erased, write_output_mutex, paused_time_offset); } else { fprintf(stderr, "Failed to encode audio!\n"); } @@ -1942,51 +2606,67 @@ int main(int argc, char **argv) { frame_time_overflow = std::min(frame_time_overflow, target_fps); frame_timer_start = time_now - frame_time_overflow; - const double this_video_frame_time = clock_get_monotonic_seconds(); + const double this_video_frame_time = clock_get_monotonic_seconds() - paused_time_offset; const int64_t expected_frames = std::round((this_video_frame_time - start_time_pts) / target_fps); - const int num_frames = framerate_mode == FramerateMode::CONSTANT ? std::max(0L, expected_frames - video_pts_counter) : 1; + const int num_frames = framerate_mode == FramerateMode::CONSTANT ? std::max((int64_t)0LL, expected_frames - video_pts_counter) : 1; - if(num_frames > 0) { - gsr_capture_capture(capture, frame); + if(num_frames > 0 && !paused) { + gsr_capture_capture(capture, video_frame); // TODO: Check if duplicate frame can be saved just by writing it with a different pts instead of sending it again for(int i = 0; i < num_frames; ++i) { if(framerate_mode == FramerateMode::CONSTANT) { - frame->pts = video_pts_counter + i; + video_frame->pts = video_pts_counter + i; } else { - frame->pts = (this_video_frame_time - record_start_time) * (double)AV_TIME_BASE; - const bool same_pts = frame->pts == video_prev_pts; - video_prev_pts = frame->pts; + video_frame->pts = (this_video_frame_time - record_start_time) * (double)AV_TIME_BASE; + const bool same_pts = video_frame->pts == video_prev_pts; + video_prev_pts = video_frame->pts; if(same_pts) continue; } - int ret = avcodec_send_frame(video_codec_context, frame); + int ret = avcodec_send_frame(video_codec_context, video_frame); if(ret == 0) { // TODO: Move to separate thread because this could write to network (for example when livestreaming) - receive_frames(video_codec_context, VIDEO_STREAM_INDEX, video_stream, frame->pts, av_format_context, - record_start_time, frame_data_queue, replay_buffer_size_secs, frames_erased, write_output_mutex); + receive_frames(video_codec_context, VIDEO_STREAM_INDEX, video_stream, video_frame->pts, av_format_context, + record_start_time, frame_data_queue, replay_buffer_size_secs, frames_erased, write_output_mutex, paused_time_offset); } else { fprintf(stderr, "Error: avcodec_send_frame failed, error: %s\n", av_error_to_string(ret)); } } + + gsr_capture_end(capture, video_frame); video_pts_counter += num_frames; } } + if(toggle_pause == 1) { + const bool new_paused_state = !paused; + if(new_paused_state) { + paused_time_start = clock_get_monotonic_seconds(); + fprintf(stderr, "Paused\n"); + } else { + paused_time_offset += (clock_get_monotonic_seconds() - paused_time_start); + fprintf(stderr, "Unpaused\n"); + } + + toggle_pause = 0; + paused = !paused; + } + if(save_replay_thread.valid() && save_replay_thread.wait_for(std::chrono::seconds(0)) == std::future_status::ready) { save_replay_thread.get(); puts(save_replay_output_filepath.c_str()); + fflush(stdout); + if(recording_saved_script) + run_recording_saved_script_async(recording_saved_script, save_replay_output_filepath.c_str(), "replay"); std::lock_guard<std::mutex> lock(write_output_mutex); - for(AVPacket &packet : save_replay_packets) { - av_packet_unref(&packet); - } save_replay_packets.clear(); } if(save_replay == 1 && !save_replay_thread.valid() && replay_buffer_size_secs != -1) { save_replay = 0; - save_replay_async(video_codec_context, VIDEO_STREAM_INDEX, audio_tracks, frame_data_queue, frames_erased, filename, container_format, file_extension, write_output_mutex); + save_replay_async(video_codec_context, VIDEO_STREAM_INDEX, audio_tracks, frame_data_queue, frames_erased, filename, container_format, file_extension, write_output_mutex, make_folders); } double frame_end = clock_get_monotonic_seconds(); @@ -2001,10 +2681,10 @@ int main(int argc, char **argv) { if(save_replay_thread.valid()) { save_replay_thread.get(); puts(save_replay_output_filepath.c_str()); + fflush(stdout); + if(recording_saved_script) + run_recording_saved_script_async(recording_saved_script, save_replay_output_filepath.c_str(), "replay"); std::lock_guard<std::mutex> lock(write_output_mutex); - for(AVPacket &packet : save_replay_packets) { - av_packet_unref(&packet); - } save_replay_packets.clear(); } @@ -2026,11 +2706,22 @@ int main(int argc, char **argv) { gsr_capture_destroy(capture, video_codec_context); + if(replay_buffer_size_secs == -1 && recording_saved_script) + run_recording_saved_script_async(recording_saved_script, filename, "regular"); + if(dpy) { // TODO: This causes a crash, why? maybe some other library dlclose xlib and that also happened to unload this??? //XCloseDisplay(dpy); } + //av_frame_free(&video_frame); + free((void*)window_str); free(empty_audio); + // We do an _exit here because cuda uses at_exit to do _something_ that causes the program to freeze, + // but only on some nvidia driver versions on some gpus (RTX?), and _exit exits the program without calling + // the at_exit registered functions. + // Cuda (cuvid library in this case) seems to be waiting for a thread that never finishes execution. + // Maybe this happens because we dont clean up all ffmpeg resources? + // TODO: Investigate this. _exit(should_stop_error ? 3 : 0); } |