path: root/ffmpeg_backend.cpp

extern "C" {
#include <libavcodec/avcodec.h>
#include <libavformat/avformat.h>
#include <libavformat/avio.h>
#include <libavutil/avutil.h>
#include <libswscale/swscale.h>
}

// workaround to make libav error printing work
#ifdef av_err2str
#undef av_err2str
#include <string>
av_always_inline std::string av_err2string(int errnum) {
    char str[AV_ERROR_MAX_STRING_SIZE];
    return av_make_error_string(str, AV_ERROR_MAX_STRING_SIZE, errnum);
}
#define av_err2str(err) av_err2string(err).c_str()
#endif  // av_err2str

#include "ffmpeg_backend.h"

bool32 AV_TryFrame(av_info *AV, int32 *err)
{
    *err = av_read_frame(AV->FileFormatContext, AV->VideoPacket);
    if (*err >= 0 && AV->VideoPacket->stream_index != AV->VideoStream->index) {
        av_packet_unref(AV->VideoPacket);
        return 0;
    }

    if (*err < 0)
        *err = avcodec_send_packet(AV->VideoCodecContext, AV->VideoPacket);
    else {
        *err = avcodec_send_packet(AV->VideoCodecContext, AV->VideoPacket);
    }
    av_packet_unref(AV->VideoPacket);

    if (*err < 0)
    {
        fprintf(stderr, "Libav *error: (%s)\n", av_err2str(*err));
        Assert(0);
    }

    while (*err >= 0) {
        *err = avcodec_receive_frame(AV->VideoCodecContext, AV->VideoFrame);
        if (*err == AVERROR_EOF) {
        } else if (*err == AVERROR(EAGAIN)) {
            *err = 0;
            break;
        } else if (*err < 0) {
            Assert(0);
        }
        return 1;
    }
    return 0;
}

// TODO(fox): Could be combined into AV_Init once we have dealloc functions for
// the AVInfo allocation.
bool32 AV_IsFileSupported(char *filename)
{
    int32 err = 0;

    // enum AVHWDeviceType type;
    // while((type = av_hwdevice_iterate_types(type)) != AV_HWDEVICE_TYPE_NONE)
    //     printf("%s\n", av_hwdevice_get_type_name(type));

    AVFormatContext *temp = avformat_alloc_context();
    err = avformat_open_input(&temp, filename, NULL, NULL);;

    if (err < 0) {
            fprintf(stderr, "Libav error: (%s)\n", av_err2str(err));
            avformat_free_context(temp);
            return 0;
    }

    err = avformat_find_stream_info(temp,  NULL);

    if (err < 0) {
            fprintf(stderr, "Libav error: (%s)\n", av_err2str(err));
            avformat_free_context(temp);
            return 0;
    }

    avformat_free_context(temp);

    return 1;
}

void AV_Init(source *Source, layer_bitmap_info *BitmapInfo, memory *Memory)
{
    BitmapInfo->AVInfo = AllocateMemory(Memory, sizeof(av_info), P_AVInfo);
    av_info *AV = (av_info *)BitmapInfo->AVInfo;
    *AV = {};

    int32 err = 0;

    // enum AVHWDeviceType type;
    // while((type = av_hwdevice_iterate_types(type)) != AV_HWDEVICE_TYPE_NONE)
    //     printf("%s\n", av_hwdevice_get_type_name(type));

    // The two calls below theoretically shouldn't fail since we already tested them in IsFileSupported.

    AV->FileFormatContext = avformat_alloc_context();
    err = avformat_open_input(&AV->FileFormatContext, Source->Path, NULL, NULL);;
    if (err < 0) {
            fprintf(stderr, "Libav error: (%s)\n", av_err2str(err));
            Assert(0);
    }

    err = avformat_find_stream_info(AV->FileFormatContext,  NULL);

    if (err < 0) {
            fprintf(stderr, "Libav error: (%s)\n", av_err2str(err));
            Assert(0);
    }

    for (uint32 i = 0; i < AV->FileFormatContext->nb_streams; i++)
    {
        AVCodecParameters *LocalCodecParameters = NULL;
        LocalCodecParameters = AV->FileFormatContext->streams[i]->codecpar;
        if (LocalCodecParameters->codec_type == AVMEDIA_TYPE_VIDEO) {
            AV->VideoCodecParameters = LocalCodecParameters;
            AV->VideoStream =  AV->FileFormatContext->streams[i];
            break;
        }
    }

    if (!AV->VideoCodecParameters) {
        printf("Libav error: No video track found.");
    }

    AV->VideoCodec = avcodec_find_decoder(AV->VideoCodecParameters->codec_id);

    if (!AV->VideoCodec) {
        printf("Libav error: Video codec could not be identified.");
    }
/*
    int16 codecs = 0;
    for (;;) {
        AV->VideoHWConfig = avcodec_get_hw_config(AV->VideoCodec, codecs);
        if (!AV->VideoHWConfig) {
            printf("Libav error: Hardware acceleration not found for decoder %s.",
                    AV->VideoCodec->name);
            break;
        }
        AV->HWPixFormat = AV->VideoHWConfig->pix_fmt;
        break;
        // if (AV->VideoHWConfig->methods & AV_CODEC_HW_CONFIG_METHOD_HW_DEVICE_CTX &&
        //     AV->VideoHWConfig->device_type == type) {
        // }
        codecs++;
    }
*/
    AV->VideoCodecContext = avcodec_alloc_context3(AV->VideoCodec);
    if (!AV->VideoCodecContext) {
        printf("Libav error: Decoder context allocation failed.");
    }
    err = avcodec_parameters_to_context(AV->VideoCodecContext, AV->VideoCodecParameters);
    if (err < 0) {
            fprintf(stderr, "Libav error: (%s)\n", av_err2str(err));
    }
    avcodec_open2(AV->VideoCodecContext, AV->VideoCodec, NULL);
    if (err < 0) {
            fprintf(stderr, "Libav error: (%s)\n", av_err2str(err));
    }

    AV->VideoPacket = av_packet_alloc();
    if (err < 0) {
            fprintf(stderr, "Libav error: (%s)\n", av_err2str(err));
    }
    AV->VideoFrame = av_frame_alloc();
    if (err < 0) {
            fprintf(stderr, "Libav error: (%s)\n", av_err2str(err));
    }

    Source->Info.FPS = (real32)AV->VideoStream->r_frame_rate.num / AV->VideoStream->r_frame_rate.den;
    Source->Info.Width = AV->VideoCodecContext->width;
    Source->Info.Height = AV->VideoCodecContext->height;
};



void AV_GetPTSAverage(source *Source, av_info *AV, int32 *err)
{
    // TODO(fox): This PTS average isn't exact and causes an occasional
    // frame skip. See libav remarks in forum for more details.

    // TODO(fox): Handle footage under five seconds.
    int16 TestAmount = 5;

    real32 FPS = (real32)AV->VideoStream->r_frame_rate.num / AV->VideoStream->r_frame_rate.den;

    int16 i = 0;
    real32 AvgPTSPerSecond = 0;
    for (;;) {
        if (AV_TryFrame(AV, err)) {
            if (i >= FPS * TestAmount) {
                AvgPTSPerSecond = (real32)AV->VideoFrame->pts / TestAmount;
                printf("frame: %i, pts: %li\n", i, AV->VideoFrame->pts);
                break;
            }
            i++;
            av_frame_unref(AV->VideoFrame);
        }
    }

    Source->Info.AvgPTSPerFrame = (real32)AvgPTSPerSecond / (int32)(FPS + 0.5f);
    printf("Avg PTS per sec: %.06f, Avg PTS per frame: %.06f\n", AvgPTSPerSecond, Source->Info.AvgPTSPerFrame);

    av_seek_frame(AV->FileFormatContext, -1, 0, AVSEEK_FLAG_BACKWARD);
}


cached_bitmap * AV_LoadVideoFrame(source *Source, layer_bitmap_info *BitmapInfo, memory *Memory, int32 TimelineFrame)
{
    av_info *AV = (av_info *)BitmapInfo->AVInfo;
    int32 *CurrentlyRenderedFrame = &BitmapInfo->CurrentFrame;

    int32 err = 0;

    if (!Source->Info.AvgPTSPerFrame) {
        AV_GetPTSAverage(Source, AV, &err);
    }
    Assert(Source->Info.AvgPTSPerFrame);

    int32 FrameToSeek = TimelineFrame - BitmapInfo->FrameOffset;
    if (*CurrentlyRenderedFrame == FrameToSeek || FrameToSeek < 0)
        return 0;

    // NOTE(fox): The decoder automatically advances to the next frame, so we
    // don't need to call av_seek_frame under normal playback.
    // This function only seeks to the nearest "keyframe."

    if (*CurrentlyRenderedFrame != FrameToSeek - 1) {
        int64 SeekSeconds = (int64)(FrameToSeek / (int32)(Source->Info.FPS + 0.5f) * AV_TIME_BASE);
        av_seek_frame(AV->FileFormatContext, -1, SeekSeconds, AVSEEK_FLAG_BACKWARD);
        printf("Seek activated\n");
    } else if (*CurrentlyRenderedFrame < 0) {
        av_seek_frame(AV->FileFormatContext, -1, 0, AVSEEK_FLAG_BACKWARD);
    }

    *CurrentlyRenderedFrame = FrameToSeek;

    int64 SeekPTS = (int64)(Source->Info.AvgPTSPerFrame*FrameToSeek + 0.5f);

    while (err >= 0) {
        if (AV_TryFrame(AV, &err)) {

            // The first frame that gets loaded isn't always the actual
            // first frame, so we need to check until it's correct.
            if (FrameToSeek == 0 && AV->VideoFrame->pts != AV->VideoStream->start_time) {
                av_frame_unref(AV->VideoFrame);
                printf("NON-START: avg: %li, real pts: %li", SeekPTS, AV->VideoFrame->pts);
                continue;
            }

            int64 Difference = AV->VideoFrame->pts - SeekPTS;
            if (abs(Difference) < Source->Info.AvgPTSPerFrame)
            {
                if (AV->PreviousPTS == -1) {
                    AV->PreviousPTS = AV->VideoFrame->pts;
                    printf("avg: %li, real pts: %li, difference: %li\n", SeekPTS, AV->VideoFrame->pts, Difference);
                } else {
                    printf("avg: %li, real pts: %li, difference: %li difference from last pts: %li\n", SeekPTS, AV->VideoFrame->pts, AV->VideoFrame->pts - SeekPTS, AV->VideoFrame->pts - AV->PreviousPTS);
                    AV->PreviousPTS = AV->VideoFrame->pts;
                }

                uint16 Width = Source->Info.Width;
                uint16 Height = Source->Info.Height;
                uint16 BytesPerPixel = Source->Info.BytesPerPixel;
                int32 Pitch = Width*BytesPerPixel;

                cached_bitmap *Bitmap = Memory_RollingBitmap(Memory, Source, FrameToSeek);
                void *Buffer = Bitmap->Data;

                int out_linesize[4] = { Pitch, Pitch, Pitch, Pitch };
                uint8 *dst_data[4] = { (uint8 *)Buffer, (uint8 *)Buffer + Width*Height*BytesPerPixel,
                                       (uint8 *)Buffer + Width*Height*BytesPerPixel*2, (uint8 *)Buffer + Width*Height*BytesPerPixel*3 };

                // NOTE(fox): This function will be replaced in the future.
                AV->RGBContext = sws_getContext(AV->VideoFrame->width, AV->VideoFrame->height, (AVPixelFormat)AV->VideoFrame->format,
                                                AV->VideoFrame->width, AV->VideoFrame->height, AV_PIX_FMT_RGBA, SWS_BILINEAR,
                                                NULL, NULL, NULL);

                if(!AV->RGBContext) {
                    printf("Libav error: SwsContext creation failed.");
                }

                sws_scale(AV->RGBContext, AV->VideoFrame->data, AV->VideoFrame->linesize, 0, AV->VideoFrame->height,
                          dst_data, out_linesize);

                av_frame_unref(AV->VideoFrame);

                Assert(BitmapInfo->BitmapBuffer);
                return Bitmap;
            }
            else
            {
                // If this gets printed when not seeking, a frame has been skipped!
                printf("FRAME SKIP: avg: %li, real pts: %li, difference: %li\n", SeekPTS, AV->VideoFrame->pts, Difference);
            }
            av_frame_unref(AV->VideoFrame);
        }
    }
    return 0;
}