[FFmpeg-devel,3/3] tools: add an AV_CODEC_CAP_ENCODER_RECON_FRAME test tool

---
 Makefile                     |   2 +
 tools/Makefile               |   3 +-
 tools/enc_recon_frame_test.c | 388 +++++++++++++++++++++++++++++++++++
 3 files changed, 392 insertions(+), 1 deletion(-)
 create mode 100644 tools/enc_recon_frame_test.c

On 10/18/2022 8:51 AM, Anton Khirnov wrote:
> ---
>   Makefile                     |   2 +
>   tools/Makefile               |   3 +-
>   tools/enc_recon_frame_test.c | 388 +++++++++++++++++++++++++++++++++++
>   3 files changed, 392 insertions(+), 1 deletion(-)
>   create mode 100644 tools/enc_recon_frame_test.c
> 
> diff --git a/Makefile b/Makefile
> index 1fb742f390..bf1b69f96b 100644
> --- a/Makefile
> +++ b/Makefile
> @@ -67,6 +67,8 @@ tools/target_io_dem_fuzzer$(EXESUF): tools/target_io_dem_fuzzer.o $(FF_DEP_LIBS)
>   
>   tools/enum_options$(EXESUF): ELIBS = $(FF_EXTRALIBS)
>   tools/enum_options$(EXESUF): $(FF_DEP_LIBS)
> +tools/enc_recon_frame_test$(EXESUF): $(FF_DEP_LIBS)
> +tools/enc_recon_frame_test$(EXESUF): ELIBS = $(FF_EXTRALIBS)
>   tools/scale_slice_test$(EXESUF): $(FF_DEP_LIBS)
>   tools/scale_slice_test$(EXESUF): ELIBS = $(FF_EXTRALIBS)
>   tools/sofa2wavs$(EXESUF): ELIBS = $(FF_EXTRALIBS)
> diff --git a/tools/Makefile b/tools/Makefile
> index 4afa23342d..dee6a41668 100644
> --- a/tools/Makefile
> +++ b/tools/Makefile
> @@ -1,4 +1,4 @@
> -TOOLS = enum_options qt-faststart scale_slice_test trasher uncoded_frame
> +TOOLS = enc_recon_frame_test enum_options qt-faststart scale_slice_test trasher uncoded_frame
>   TOOLS-$(CONFIG_LIBMYSOFA) += sofa2wavs
>   TOOLS-$(CONFIG_ZLIB) += cws2fws
>   
> @@ -17,6 +17,7 @@ tools/target_dem_fuzzer.o: tools/target_dem_fuzzer.c
>   tools/target_io_dem_fuzzer.o: tools/target_dem_fuzzer.c
>   	$(COMPILE_C) -DIO_FLAT=0
>   
> +tools/enc_recon_frame_test$(EXESUF): tools/decode_simple.o
>   tools/venc_data_dump$(EXESUF): tools/decode_simple.o
>   tools/scale_slice_test$(EXESUF): tools/decode_simple.o
>   
> diff --git a/tools/enc_recon_frame_test.c b/tools/enc_recon_frame_test.c
> new file mode 100644
> index 0000000000..0d3b8e88e3
> --- /dev/null
> +++ b/tools/enc_recon_frame_test.c
> @@ -0,0 +1,388 @@
> +/*
> + * copyright (c) 2022 Anton Khirnov <anton@khirnov.net>
> + *
> + * This file is part of FFmpeg.
> + *
> + * FFmpeg is free software; you can redistribute it and/or
> + * modify it under the terms of the GNU Lesser General Public
> + * License as published by the Free Software Foundation; either
> + * version 2.1 of the License, or (at your option) any later version.
> + *
> + * FFmpeg is distributed in the hope that it will be useful,
> + * but WITHOUT ANY WARRANTY; without even the implied warranty of
> + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
> + * Lesser General Public License for more details.
> + *
> + * You should have received a copy of the GNU Lesser General Public
> + * License along with FFmpeg; if not, write to the Free Software
> + * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
> + */
> +
> +/* A test for AV_CODEC_FLAG_RECON_FRAME
> + * TODO: dump reconstructed frames to disk */
> +
> +#include <stdio.h>
> +#include <stdint.h>
> +#include <stdlib.h>
> +
> +#include "decode_simple.h"
> +
> +#include "libavutil/adler32.h"
> +#include "libavutil/common.h"
> +#include "libavutil/error.h"
> +#include "libavutil/frame.h"
> +#include "libavutil/imgutils.h"
> +#include "libavutil/opt.h"
> +
> +#include "libavformat/avformat.h"
> +
> +#include "libavcodec/avcodec.h"
> +#include "libavcodec/codec.h"
> +
> +#include "libswscale/swscale.h"
> +
> +typedef struct FrameChecksum {
> +    int64_t  ts;
> +    uint32_t checksum[4];
> +} FrameChecksum;
> +
> +typedef struct PrivData {
> +    AVCodecContext *enc;
> +    AVCodecContext *dec;
> +
> +    AVPacket *pkt;
> +    AVFrame  *frame, *frame_recon;
> +
> +    struct SwsContext *scaler;
> +
> +    FrameChecksum *checksums_decoded;
> +    size_t      nb_checksums_decoded;
> +    FrameChecksum *checksums_recon;
> +    size_t      nb_checksums_recon;
> +
> +    int dump;
> +} PrivData;
> +
> +static int frame_hash(FrameChecksum **pc, size_t *nb_c, int64_t ts,
> +                      const AVFrame *frame)
> +{
> +    FrameChecksum *c;
> +    int shift_h, shift_v;
> +
> +    c = av_realloc_array(*pc, *nb_c + 1, sizeof(*c));
> +    if (!c)
> +        return AVERROR(ENOMEM);
> +    *pc = c;
> +    (*nb_c)++;
> +
> +    c += *nb_c - 1;
> +    memset(c, 0, sizeof(*c));
> +
> +    av_pix_fmt_get_chroma_sub_sample(frame->format, &shift_h, &shift_v);
> +
> +    c->ts = ts;
> +    for (int p = 0; frame->data[p]; p++) {
> +        const uint8_t *data = frame->data[p];
> +        int linesize = av_image_get_linesize(frame->format, frame->width, p);
> +        uint32_t checksum = 0;
> +
> +        for (int j = 0; j < frame->height >> shift_v; j++) {
> +            checksum = av_adler32_update(checksum, data, linesize);
> +            data += frame->linesize[p];
> +        }
> +
> +        c->checksum[p] = checksum;
> +    }
> +
> +    return 0;
> +}
> +
> +static int recon_frame_process(PrivData *pd, const AVPacket *pkt)
> +{
> +    AVFrame *f = pd->frame_recon;
> +    int ret;
> +
> +    ret = avcodec_receive_frame(pd->enc, f);
> +    if (ret < 0) {
> +        fprintf(stderr, "Error retrieving a reconstructed frame\n");
> +        return ret;
> +    }
> +
> +    if (pd->dump) {
> +    }
> +
> +    // the encoder's internal format (in which the reconsturcted frames are
> +    // exported) may be different from the user-facing pixel format
> +    if (f->format != pd->enc->pix_fmt) {
> +        if (!pd->scaler) {
> +            pd->scaler = sws_getContext(f->width, f->height, f->format,
> +                                        f->width, f->height, pd->enc->pix_fmt,
> +                                        SWS_BITEXACT, NULL, NULL, NULL);
> +            if (!pd->scaler)
> +                return AVERROR(ENOMEM);
> +        }
> +
> +        ret = sws_scale_frame(pd->scaler, pd->frame, f);
> +        if (ret < 0) {
> +            fprintf(stderr, "Error converting pixel formats\n");
> +            return ret;
> +        }
> +
> +        av_frame_unref(f);
> +        f = pd->frame;
> +    }
> +
> +    ret = frame_hash(&pd->checksums_recon, &pd->nb_checksums_recon,
> +                     pkt->pts, f);
> +    av_frame_unref(f);
> +
> +    return 0;
> +}
> +
> +static int process_frame(DecodeContext *dc, AVFrame *frame)
> +{
> +    PrivData *pd = dc->opaque;
> +    int ret;
> +
> +    if (!avcodec_is_open(pd->enc)) {
> +        if (!frame) {
> +            fprintf(stderr, "No frames were decoded\n");
> +            return AVERROR_INVALIDDATA;
> +        }
> +
> +        pd->enc->width        = frame->width;
> +        pd->enc->height       = frame->height;
> +        pd->enc->pix_fmt      = frame->format;
> +        pd->enc->time_base    = dc->stream->time_base;
> +        pd->enc->thread_count = dc->decoder->thread_count;
> +        pd->enc->thread_type  = dc->decoder->thread_type;
> +
> +        ret = avcodec_open2(pd->enc, NULL, NULL);
> +        if (ret < 0) {
> +            fprintf(stderr, "Error opening the encoder\n");
> +            return ret;
> +        }
> +    }
> +
> +    if (frame)
> +        frame->pict_type = AV_PICTURE_TYPE_NONE;
> +
> +    ret = avcodec_send_frame(pd->enc, frame);
> +    if (ret < 0) {
> +        fprintf(stderr, "Error submitting a frame for encoding\n");
> +        return ret;
> +    }
> +
> +    while (1) {
> +        AVPacket *pkt = pd->pkt;
> +
> +        ret = avcodec_receive_packet(pd->enc, pkt);
> +        if (ret == AVERROR(EAGAIN))
> +            break;
> +        else if (ret == AVERROR_EOF)
> +            pkt = NULL;
> +        else if (ret < 0 && ret != AVERROR_EOF) {
> +            fprintf(stderr, "Error receiving a frame from the encoder\n");
> +            return ret;
> +        }
> +
> +        if (pkt) {
> +            ret = recon_frame_process(pd, pkt);
> +            if (ret < 0)
> +                return ret;
> +        }
> +
> +        if (!avcodec_is_open(pd->dec)) {
> +            if (!pkt) {
> +                fprintf(stderr, "No packets were received from the encoder\n");
> +                return AVERROR(EINVAL);
> +            }
> +
> +            pd->dec->width        = pd->enc->width;
> +            pd->dec->height       = pd->enc->height;
> +            pd->dec->pix_fmt      = pd->enc->pix_fmt;
> +            pd->dec->thread_count = dc->decoder->thread_count;
> +            pd->dec->thread_type  = dc->decoder->thread_type;
> +            if (pd->enc->extradata_size) {
> +                pd->dec->extradata = av_memdup(pd->enc->extradata,
> +                                               pd->enc->extradata_size + AV_INPUT_BUFFER_PADDING_SIZE);
> +                if (!pd->dec->extradata)
> +                    return AVERROR(ENOMEM);
> +            }
> +
> +            ret = avcodec_open2(pd->dec, NULL, NULL);
> +            if (ret < 0) {
> +                fprintf(stderr, "Error opening the decoder\n");
> +                return ret;
> +            }
> +        }
> +
> +        ret = avcodec_send_packet(pd->dec, pkt);
> +        if (ret < 0) {
> +            fprintf(stderr, "Error sending a frame to decoder\n");
> +            return ret;
> +        }
> +
> +        while (1) {
> +            ret = avcodec_receive_frame(pd->dec, pd->frame);
> +            if (ret == AVERROR(EAGAIN))
> +                break;
> +            else if (ret == AVERROR_EOF)
> +                return 0;
> +            else if (ret < 0) {
> +                fprintf(stderr, "Error receving a frame from decoder\n");
> +                return ret;
> +            }
> +
> +            ret = frame_hash(&pd->checksums_decoded, &pd->nb_checksums_decoded,
> +                             pd->frame->pts, pd->frame);
> +            av_frame_unref(pd->frame);
> +            if (ret < 0)
> +                return ret;
> +        }
> +
> +    }
> +
> +    return 0;
> +}
> +
> +static int frame_checksum_compare(const void *a, const void *b)
> +{
> +    const FrameChecksum *ca = a;
> +    const FrameChecksum *cb = b;
> +    if (ca->ts == cb->ts)
> +        return 0;
> +    return FFSIGN(ca->ts - cb->ts);
> +}
> +
> +int main(int argc, char **argv)
> +{
> +    PrivData      pd;
> +    DecodeContext dc;
> +
> +    const char *filename, *enc_name, *enc_opts, *thread_type = NULL, *nb_threads = NULL;
> +    const AVCodec *enc, *dec;
> +    int ret = 0, max_frames = 0, dump = 0;
> +
> +    if (argc <= 4) {
> +        fprintf(stderr,
> +                "Usage: %s <input file> <encoder> <encoder options> <max frame count> "
> +                "[<thread count> <thread type> <dump>]\n", argv[0]);
> +        return 0;
> +    }
> +
> +    filename     = argv[1];
> +    enc_name     = argv[2];
> +    enc_opts     = argv[3];
> +    max_frames   = strtol(argv[4], NULL, 0);

Maybe this should be optional too. If i want to test encoding an entire 
sample, and i don't know its exact length, then this option is not 
really helpful.

> +    if (argc >= 6)
> +        nb_threads = argv[5];
> +    if (argc >= 7)
> +        thread_type = argv[6];
> +    if (argc >= 8)
> +        dump = strtol(argv[7], NULL, 0);

Probably best to not add the option until it's implemented. Ignoring it 
silently is not nice.

> +
> +    memset(&dc, 0, sizeof(dc));
> +    memset(&pd, 0, sizeof(pd));
> +
> +    pd.dump = dump;
> +
> +    enc = avcodec_find_encoder_by_name(enc_name);
> +    if (!enc) {
> +        fprintf(stderr, "No such encoder: %s\n", enc_name);
> +        return 1;
> +    }
> +    if (!(enc->capabilities & AV_CODEC_CAP_ENCODER_RECON_FRAME)) {
> +        fprintf(stderr, "Encoder '%s' cannot ouput reconstructed frames\n",
> +                enc->name);
> +        return 1;
> +    }
> +
> +    dec = avcodec_find_decoder(enc->id);
> +    if (!dec) {
> +        fprintf(stderr, "No decoder for: %s\n", avcodec_get_name(enc->id));
> +        return 1;
> +    }
> +
> +    pd.enc = avcodec_alloc_context3(enc);
> +    if (!pd.enc) {
> +        fprintf(stderr, "Error allocating encoder\n");
> +        return 1;
> +    }
> +
> +    ret = av_set_options_string(pd.enc, enc_opts, "=", ",");
> +    if (ret < 0) {
> +        fprintf(stderr, "Error setting encoder options\n");
> +        goto fail;
> +    }
> +    pd.enc->flags |= AV_CODEC_FLAG_RECON_FRAME | AV_CODEC_FLAG_BITEXACT;
> +
> +    pd.dec = avcodec_alloc_context3(dec);
> +    if (!pd.dec) {
> +        fprintf(stderr, "Error allocating decoder\n");
> +        goto fail;
> +    }
> +
> +    pd.dec->flags |= AV_CODEC_FLAG_BITEXACT;

nit: Maybe enable AV_EF_CRCCHECK in err_recognition for extra checks in 
the decoder (It's not too important since no decoder for a codec with a 
currently supported recon encoder uses it right now).

> +
> +    pd.frame       = av_frame_alloc();
> +    pd.frame_recon = av_frame_alloc();
> +    pd.pkt         = av_packet_alloc();
> +    if (!pd.frame ||!pd.frame_recon || !pd.pkt) {
> +        ret = 1;
> +        goto fail;
> +    }
> +
> +    ret = ds_open(&dc, filename, 0);
> +    if (ret < 0) {
> +        fprintf(stderr, "Error opening the file\n");
> +        goto fail;
> +    }
> +
> +    dc.process_frame = process_frame;
> +    dc.opaque        = &pd;
> +    dc.max_frames    = max_frames;
> +
> +    ret  = av_dict_set(&dc.decoder_opts, "threads",          nb_threads,    0);
> +    ret |= av_dict_set(&dc.decoder_opts, "thread_type",      thread_type,   0);
> +
> +    ret = ds_run(&dc);
> +    if (ret < 0)
> +        goto fail;
> +
> +    if (pd.nb_checksums_decoded != pd.nb_checksums_recon) {
> +        fprintf(stderr, "Mismatching frame counts: recon=%zu decoded=%zu\n",
> +                pd.nb_checksums_recon, pd.nb_checksums_decoded);
> +        ret = 1;
> +        goto fail;
> +    }
> +
> +    // reconstructed frames are in coded order, sort them by pts into presentation order
> +    qsort(pd.checksums_recon, pd.nb_checksums_recon, sizeof(*pd.checksums_recon),
> +          frame_checksum_compare);
> +
> +    for (size_t i = 0; i < pd.nb_checksums_decoded; i++) {
> +        const FrameChecksum *d = &pd.checksums_decoded[i];
> +        const FrameChecksum *r = &pd.checksums_recon[i];
> +
> +        for (int p = 0; p < FF_ARRAY_ELEMS(d->checksum); p++)
> +            if (d->checksum[p] != r->checksum[p]) {
> +                fprintf(stderr, "Checksum mismatch in frame ts=%"PRId64", plane %d\n",
> +                        d->ts, p);
> +                ret = 1;
> +                goto fail;
> +            }
> +    }
> +    fprintf(stderr, "All %zu encoded frames match\n", pd.nb_checksums_decoded);
> +
> +fail:
> +    avcodec_free_context(&pd.enc);
> +    avcodec_free_context(&pd.dec);
> +    av_freep(&pd.checksums_decoded);
> +    av_freep(&pd.checksums_recon);
> +    av_frame_free(&pd.frame);
> +    av_frame_free(&pd.frame_recon);
> +    av_packet_free(&pd.pkt);
> +    ds_free(&dc);
> +    return !!ret;
> +}

Tested and it seems to work, so LGTM.