Git Inbox Mirror of the ffmpeg-devel mailing list - see https://ffmpeg.org/mailman/listinfo/ffmpeg-devel
 help / color / mirror / Atom feed
From: Nuo Mi <nuomi2021@gmail.com>
To: FFmpeg development discussions and patches <ffmpeg-devel@ffmpeg.org>
Subject: Re: [FFmpeg-devel] [PATCH 01/14] vvcdec: add thread executor
Date: Sun, 21 May 2023 22:24:56 +0800
Message-ID: <CAFXK13cuNwO0Ey--5P_yau7bjTq4AP6UPu6Ei1DSKxw-Qk6Vzw@mail.gmail.com> (raw)
In-Reply-To: <12209065.O9o76ZdvQC@basile.remlab.net>

Hi Rémi,
Thank you for the cmments

On Sun, May 21, 2023 at 9:15 PM Rémi Denis-Courmont <remi@remlab.net> wrote:

> Le sunnuntaina 21. toukokuuta 2023, 16.03.06 EEST Nuo Mi a écrit :
> > The executor design pattern was inroduced by java
> > <
> https://docs.oracle.com/en/java/javase/20/docs/api/java.base/java/util/conc
> > urrent/Executor.html> it also adapted by python
> > <https://docs.python.org/3/library/concurrent.futures.html>
> > Compared to handcrafted thread pool management, it greatly simplifies the
> > thread code. ---
> >  libavcodec/Makefile           |   1 +
> >  libavcodec/vvc/Makefile       |   4 +
> >  libavcodec/vvc/vvc_executor.c | 193 ++++++++++++++++++++++++++++++++++
> >  libavcodec/vvc/vvc_executor.h |  73 +++++++++++++
> >  4 files changed, 271 insertions(+)
> >  create mode 100644 libavcodec/vvc/Makefile
> >  create mode 100644 libavcodec/vvc/vvc_executor.c
> >  create mode 100644 libavcodec/vvc/vvc_executor.h
> >
> > diff --git a/libavcodec/Makefile b/libavcodec/Makefile
> > index dab09f483a..b1fcbf71b2 100644
> > --- a/libavcodec/Makefile
> > +++ b/libavcodec/Makefile
> > @@ -62,6 +62,7 @@ OBJS = ac3_parser.o
>
> >              \ xiph.o
>
> >     \
> >
> >  # subsystems
> > +include $(SRC_PATH)/libavcodec/vvc/Makefile
> >  OBJS-$(CONFIG_AANDCTTABLES)            += aandcttab.o
> >  OBJS-$(CONFIG_AC3DSP)                  += ac3dsp.o ac3.o ac3tab.o
> >  OBJS-$(CONFIG_ADTS_HEADER)             += adts_header.o
> > mpeg4audio_sample_rates.o diff --git a/libavcodec/vvc/Makefile
> > b/libavcodec/vvc/Makefile
> > new file mode 100644
> > index 0000000000..c4b93e0389
> > --- /dev/null
> > +++ b/libavcodec/vvc/Makefile
> > @@ -0,0 +1,4 @@
> > +clean::
> > +     $(RM) $(CLEANSUFFIXES:%=libavcodec/vvc/%)
> > +
> > +OBJS-$(CONFIG_VVC_DECODER)          +=  vvc/vvc_executor.o
> > diff --git a/libavcodec/vvc/vvc_executor.c
> b/libavcodec/vvc/vvc_executor.c
> > new file mode 100644
> > index 0000000000..f2afdf79ae
> > --- /dev/null
> > +++ b/libavcodec/vvc/vvc_executor.c
> > @@ -0,0 +1,193 @@
> > +/*
> > + * VVC video Decoder
> > + *
> > + * Copyright (C) 2022 Nuo Mi
> > + *
> > + * 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 + */
> > +#include "libavutil/avutil.h"
> > +#include "libavutil/thread.h"
> > +
> > +#include "vvc_executor.h"
>
> This does not seem specific to VVC in any way, so the naming (and choice
> of
> folder placement) is rather weird.
>
> > +
> > +typedef struct ThreadInfo {
> > +    int idx;
> > +    VVCExecutor *e;
> > +    pthread_t thread;
> > +} ThreadInfo;
> > +
> > +struct VVCExecutor {
> > +    VVCTaskCallbacks cb;
> > +    ThreadInfo *threads;
> > +    uint8_t *local_contexts;
>
> It seems odd and needless complex to separate this from the thread info.
> It
> looks like you could simply append a pointer or a flexible array to
> ThreadInfo
> instead.
>
 Do you mean in VVCTasklet? No, we do not want to expose ThreadInfo details
to users.


> > +    int thread_count;
> > +
> > +    pthread_mutex_t lock;
> > +    pthread_cond_t cond;
> > +    int die;
> > +    VVCTasklet *tasks;
> > +};
> > +
> > +static void remove_task(VVCTasklet **prev, VVCTasklet *t)
> > +{
> > +    *prev  = t->next;
> > +    t->next = NULL;
> > +}
> > +
> > +static void add_task(VVCTasklet **prev, VVCTasklet *t)
> > +{
> > +    t->next = *prev;
> > +    *prev   = t;
> > +}
> > +
> > +static void *executor_worker_task(void *data)
> > +{
> > +    ThreadInfo *ti = (ThreadInfo*)data;
> > +    VVCExecutor *e = ti->e;
> > +    void *lc       = e->local_contexts + ti->idx *
> > e->cb.local_context_size;
> > +    VVCTasklet **prev;
> > +    VVCTaskCallbacks *cb = &e->cb;
> > +
> > +    pthread_mutex_lock(&e->lock);
> > +    while (1) {
> > +        VVCTasklet* t = NULL;
> > +        if (e->die) break;
> > +
> > +        for (prev = &e->tasks; *prev; prev = &(*prev)->next) {
> > +            if (cb->ready(*prev, cb->user_data)) {
> > +                t = *prev;
> > +                break;
> > +            }
> > +        }
> > +        if (t) {
> > +            //found one task
> > +            remove_task(prev, t);
> > +            pthread_mutex_unlock(&e->lock);
> > +            cb->run(t, lc, cb->user_data);
> > +            pthread_mutex_lock(&e->lock);
> > +        } else {
> > +            //no task in one loop
> > +            pthread_cond_wait(&e->cond, &e->lock);
> > +        }
> > +    }
> > +    pthread_mutex_unlock(&e->lock);
> > +    return NULL;
> > +}
> > +
> > +VVCExecutor* ff_vvc_executor_alloc(const VVCTaskCallbacks *cb, int
> > thread_count) +{
> > +    VVCExecutor *e;
> > +    int i, j, ret;
> > +    if (!cb || !cb->user_data || !cb->ready || !cb->run ||
> > !cb->priority_higher) +        return NULL;
> > +    e = av_calloc(1, sizeof(*e));
> > +    if (!e)
> > +        return NULL;
> > +    e->cb = *cb;
> > +
> > +    e->local_contexts = av_malloc(thread_count *
> e->cb.local_context_size);
> > +    if (!e->local_contexts)
> > +        goto free_executor;
> > +
> > +    e->threads = av_calloc(thread_count, sizeof(*e->threads));
> > +    if (!e->threads)
> > +        goto free_contexts;
> > +    for (i = 0; i < thread_count; i++) {
> > +        ThreadInfo *ti = e->threads + i;
> > +        ti->e = e;
> > +        ti->idx = i;
> > +    }
> > +
> > +    ret = pthread_mutex_init(&e->lock, NULL);
> > +    if (ret)
> > +        goto free_threads;
> > +
> > +    ret = pthread_cond_init(&e->cond, NULL);
> > +    if (ret)
> > +        goto destroy_lock;
> > +
> > +    for (i = 0; i < thread_count; i++) {
> > +        ThreadInfo *ti = e->threads + i;
> > +        ret = pthread_create(&ti->thread, NULL, executor_worker_task,
> ti);
> > +        if (ret)
> > +            goto join_threads;
> > +    }
> > +    e->thread_count = thread_count;
> > +    return e;
> > +
> > +join_threads:
> > +    pthread_mutex_lock(&e->lock);
> > +    e->die = 1;
> > +    pthread_cond_broadcast(&e->cond);
> > +    pthread_mutex_unlock(&e->lock);
> > +    for (j = 0; j < i; j++)
> > +        pthread_join(e->threads[j].thread, NULL);
> > +    pthread_cond_destroy(&e->cond);
> > +destroy_lock:
> > +    pthread_mutex_destroy(&e->lock);
> > +free_threads:
> > +    av_free(e->threads);
> > +free_contexts:
> > +    av_free(e->local_contexts);
> > +free_executor:
> > +    free(e);
> > +    return NULL;
> > +}
> > +
> > +void ff_vvc_executor_free(VVCExecutor **executor)
> > +{
> > +    VVCExecutor *e;
> > +    if (!executor || !*executor)
> > +        return;
> > +    e = *executor;
> > +
> > +    //singal die
> > +    pthread_mutex_lock(&e->lock);
> > +    e->die = 1;
> > +    pthread_cond_broadcast(&e->cond);
> > +    pthread_mutex_unlock(&e->lock);
> > +
> > +    for (int i = 0; i < e->thread_count; i++)
> > +        pthread_join(e->threads[i].thread, NULL);
> > +    pthread_cond_destroy(&e->cond);
> > +    pthread_mutex_destroy(&e->lock);
> > +
> > +    av_free(e->threads);
> > +    av_free(e->local_contexts);
> > +
> > +    av_freep(executor);
> > +}
> > +
> > +void ff_vvc_executor_execute(VVCExecutor *e, VVCTasklet *t)
> > +{
> > +    VVCTaskCallbacks *cb = &e->cb;
> > +    VVCTasklet **prev;
> > +
> > +    pthread_mutex_lock(&e->lock);
> > +    for (prev = &e->tasks; *prev && cb->priority_higher(*prev, t); prev
> =
> > &(*prev)->next) +        /* nothing */;
> > +    add_task(prev, t);
> > +    pthread_cond_signal(&e->cond);
> > +    pthread_mutex_unlock(&e->lock);
> > +}
> > +
> > +void ff_vvc_executor_wakeup(VVCExecutor *e)
> > +{
> > +    pthread_mutex_lock(&e->lock);
> > +    pthread_cond_broadcast(&e->cond);
> > +    pthread_mutex_unlock(&e->lock);
>
> Signaling a condition variable without changing any state makes no sense.
>
It's for error handling.

Imaging last all thread are waiting for ready jobs except the last one.
The last one thread has some error, it reports an error to the caller.
The caller needs to wake up threads to flush the task list.


> > +}
>
> --
> Rémi Denis-Courmont
> http://www.remlab.net/
>
>
>
> _______________________________________________
> ffmpeg-devel mailing list
> ffmpeg-devel@ffmpeg.org
> https://ffmpeg.org/mailman/listinfo/ffmpeg-devel
>
> To unsubscribe, visit link above, or email
> ffmpeg-devel-request@ffmpeg.org with subject "unsubscribe".
>
_______________________________________________
ffmpeg-devel mailing list
ffmpeg-devel@ffmpeg.org
https://ffmpeg.org/mailman/listinfo/ffmpeg-devel

To unsubscribe, visit link above, or email
ffmpeg-devel-request@ffmpeg.org with subject "unsubscribe".

  reply	other threads:[~2023-05-21 14:25 UTC|newest]

Thread overview: 28+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2023-05-21 13:03 [FFmpeg-devel] [PATCH 00/14] add vvc decoder c code Nuo Mi
2023-05-21 13:03 ` [FFmpeg-devel] [PATCH 01/14] vvcdec: add thread executor Nuo Mi
2023-05-21 13:15   ` Rémi Denis-Courmont
2023-05-21 14:24     ` Nuo Mi [this message]
2023-05-22 16:29       ` Rémi Denis-Courmont
2023-05-23  6:23         ` Nuo Mi
2023-05-21 14:44     ` Nuo Mi
2023-05-21 14:11   ` Lynne
2023-05-21 15:03     ` Nuo Mi
2023-05-21 15:06       ` Lynne
2023-05-21 15:16         ` Nuo Mi
2023-05-21 13:03 ` [FFmpeg-devel] [PATCH 02/14] vvcdec: add vvc decoder stub Nuo Mi
2023-05-21 17:30   ` Michael Niedermayer
2023-05-22  1:55     ` Nuo Mi
2023-05-21 13:03 ` [FFmpeg-devel] [PATCH 03/14] vvcdec: add sps, pps, sh parser Nuo Mi
2023-05-21 14:25   ` James Almer
2023-06-10 11:37     ` Nuo Mi
2023-05-21 13:03 ` [FFmpeg-devel] [PATCH 04/14] vvcdec: add cabac decoder Nuo Mi
2023-05-21 13:03 ` [FFmpeg-devel] [PATCH 05/14] vvcdec: add reference management Nuo Mi
2023-05-21 13:03 ` [FFmpeg-devel] [PATCH 06/14] vvcdec: add motion vector decoder Nuo Mi
2023-05-21 13:03 ` [FFmpeg-devel] [PATCH 07/14] vvcdec: add inter prediction Nuo Mi
2023-05-21 13:03 ` [FFmpeg-devel] [PATCH 08/14] vvcdec: add inv transform 1d Nuo Mi
2023-05-21 13:03 ` [FFmpeg-devel] [PATCH 09/14] vvcdec: add intra prediction Nuo Mi
2023-05-21 13:03 ` [FFmpeg-devel] [PATCH 10/14] vvcdec: add LMCS, Deblocking, SAO, and ALF filters Nuo Mi
2023-05-21 13:03 ` [FFmpeg-devel] [PATCH 11/14] vvcdec: add dsp init and inv transform Nuo Mi
2023-05-21 13:03 ` [FFmpeg-devel] [PATCH 12/14] vvcdec: add CTU(Coding Tree Unit) parser Nuo Mi
2023-05-21 13:03 ` [FFmpeg-devel] [PATCH 13/14] vvcdec: add CTU thread logical Nuo Mi
2023-05-21 13:03 ` [FFmpeg-devel] [PATCH 14/14] vvcdec: add full vvc decoder Nuo Mi

Reply instructions:

You may reply publicly to this message via plain-text email
using any one of the following methods:

* Save the following mbox file, import it into your mail client,
  and reply-to-all from there: mbox

  Avoid top-posting and favor interleaved quoting:
  https://en.wikipedia.org/wiki/Posting_style#Interleaved_style

* Reply using the --to, --cc, and --in-reply-to
  switches of git-send-email(1):

  git send-email \
    --in-reply-to=CAFXK13cuNwO0Ey--5P_yau7bjTq4AP6UPu6Ei1DSKxw-Qk6Vzw@mail.gmail.com \
    --to=nuomi2021@gmail.com \
    --cc=ffmpeg-devel@ffmpeg.org \
    /path/to/YOUR_REPLY

  https://kernel.org/pub/software/scm/git/docs/git-send-email.html

* If your mail client supports setting the In-Reply-To header
  via mailto: links, try the mailto: link

Git Inbox Mirror of the ffmpeg-devel mailing list - see https://ffmpeg.org/mailman/listinfo/ffmpeg-devel

This inbox may be cloned and mirrored by anyone:

	git clone --mirror https://master.gitmailbox.com/ffmpegdev/0 ffmpegdev/git/0.git

	# If you have public-inbox 1.1+ installed, you may
	# initialize and index your mirror using the following commands:
	public-inbox-init -V2 ffmpegdev ffmpegdev/ https://master.gitmailbox.com/ffmpegdev \
		ffmpegdev@gitmailbox.com
	public-inbox-index ffmpegdev

Example config snippet for mirrors.


AGPL code for this site: git clone https://public-inbox.org/public-inbox.git