From: Mark Thompson <sw@jkqxz.net> To: ffmpeg-devel@ffmpeg.org Subject: Re: [FFmpeg-devel] [PATCH 1/9] libavutil: add hwcontext_amf. Date: Wed, 14 Feb 2024 20:56:28 +0000 Message-ID: <3fd27a93-d2bd-41f8-8572-ef4ddfddc871@jkqxz.net> (raw) In-Reply-To: <20240214015515.1027-1-ovchinnikov.dmitrii@gmail.com> On 14/02/2024 01:55, Dmitrii Ovchinnikov wrote: > Adds hwcontext_amf, which allows to use shared AMF > context for the encoder, decoder and AMF-based filters, > without copy to the host memory. > It will also allow you to use some optimizations in > the interaction of components (for example, SAV) and make a more > manageable and optimal setup for using GPU devices with AMF > in the case of a fully AMF pipeline. > It will be a significant performance uplift when full AMF pipeline > with filters is used. > > We also plan to add Compression artefact removal filter in near feature. > --- > libavutil/Makefile | 3 + > libavutil/hwcontext.c | 4 + > libavutil/hwcontext.h | 1 + > libavutil/hwcontext_amf.c | 580 +++++++++++++++++++++++++++++++++ > libavutil/hwcontext_amf.h | 105 ++++++ > libavutil/hwcontext_internal.h | 1 + > libavutil/pixdesc.c | 4 + > libavutil/pixfmt.h | 4 + > 8 files changed, 702 insertions(+) > create mode 100644 libavutil/hwcontext_amf.c > create mode 100644 libavutil/hwcontext_amf.h > > ... > diff --git a/libavutil/hwcontext_amf.h b/libavutil/hwcontext_amf.h > new file mode 100644 > index 0000000000..0161b9a29c > --- /dev/null > +++ b/libavutil/hwcontext_amf.h > @@ -0,0 +1,105 @@ > +/* > + * 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 > + */ > + > + > +#ifndef AVUTIL_HWCONTEXT_AMF_H > +#define AVUTIL_HWCONTEXT_AMF_H > +#include <AMF/core/Factory.h> > +#include <AMF/core/Context.h> > +#include <AMF/core/Trace.h> > +#include "pixfmt.h" > + > +#include "libavformat/avformat.h" > +#include "libavutil/hwcontext.h" > + > +#define FFMPEG_AMF_WRITER_ID L"ffmpeg_amf" > + > +typedef struct AmfTraceWriter { > + AMFTraceWriterVtbl *vtbl; > + void *avctx; > + void *avcl; > +} AmfTraceWriter; > + > +typedef struct AVAMFDeviceContextInternal { > + amf_handle library; ///< handle to DLL library > + AMFFactory *factory; ///< pointer to AMF factory > + AMFDebug *debug; ///< pointer to AMF debug interface > + AMFTrace *trace; ///< pointer to AMF trace interface > + > + amf_uint64 version; ///< version of AMF runtime > + AMFContext *context; ///< AMF context > + AMF_MEMORY_TYPE mem_type; > +} AVAMFDeviceContextInternal; > + > +/** > + * This struct is allocated as AVHWDeviceContext.hwctx > + */ > + > +typedef struct AVAMFDeviceContext { > + AVBufferRef *internal; > +} AVAMFDeviceContext; > + > +typedef struct AMFFramesContext { > + AMFSurface * surfaces; > + int nb_surfaces; > +} AMFFramesContext; > + > +/** > +* Error handling helper > +*/ > +#define AMF_RETURN_IF_FALSE(avctx, exp, ret_value, /*message,*/ ...) \ > + if (!(exp)) { \ > + av_log(avctx, AV_LOG_ERROR, __VA_ARGS__); \ > + return ret_value; \ > + } > + > +#define AMF_GOTO_FAIL_IF_FALSE(avctx, exp, ret_value, /*message,*/ ...) \ > + if (!(exp)) { \ > + av_log(avctx, AV_LOG_ERROR, __VA_ARGS__); \ > + ret = ret_value; \ > + goto fail; \ > + } > + > +#define AMF_TIME_BASE_Q (AVRational){1, AMF_SECOND} > + > +typedef struct FormatMap { > + enum AVPixelFormat av_format; > + enum AMF_SURFACE_FORMAT amf_format; > +} FormatMap; > + > +extern const FormatMap format_map[]; > +enum AMF_SURFACE_FORMAT av_amf_av_to_amf_format(enum AVPixelFormat fmt); > +enum AVPixelFormat av_amf_to_av_format(enum AMF_SURFACE_FORMAT fmt); > +extern AmfTraceWriter av_amf_trace_writer; > + > +int av_amf_context_init(AVAMFDeviceContextInternal* internal, void* avcl); > +int av_amf_load_library(AVAMFDeviceContextInternal* internal, void* avcl); > +int av_amf_create_context( AVAMFDeviceContextInternal * internal, > + void* avcl, > + const char *device, > + AVDictionary *opts, int flags); > +int av_amf_context_internal_create(AVAMFDeviceContextInternal * internal, > + void* avcl, > + const char *device, > + AVDictionary *opts, int flags); > +void av_amf_context_internal_free(void *opaque, uint8_t *data); > +int av_amf_context_derive(AVAMFDeviceContextInternal * internal, > + AVHWDeviceContext *child_device_ctx, AVDictionary *opts, > + int flags); > + > +#endif /* AVUTIL_HWCONTEXT_AMF_H */ We need to sort out the content of the public header file before doing anything else. It needs to contain exactly what a user has to see to set up an AMF hwcontext instance: * A device hwctx structure, with notes on how to set the fields to put an existing device into the device context before calling av_hwdevice_ctx_init(). * Notes on what the buffer pool in a frames context contains, which might involve another structure to act as the buffer entry. * Optionally a frame hwctx structure, with notes on how to set the fields when putting a pool of existing frames into the frame context before calling av_hwframe_ctx_init() (omit if it would contain nothing). * Optionally a hwconfig structure, with notes on how to set the fields when requesting the capabilities of some particular operation (omit if all operations have identical capabilities). * If the pixfmt needs to define additional structures, they go here as well. It shouldn't contain anything else unless there are hard ABI requirements (e.g. to allocate some sort of non-fixed structure). > diff --git a/libavutil/hwcontext_internal.h b/libavutil/hwcontext_internal.h > index 4df516ee6a..48d2dc012c 100644 > --- a/libavutil/hwcontext_internal.h > +++ b/libavutil/hwcontext_internal.h > @@ -175,5 +175,6 @@ extern const HWContextType ff_hwcontext_type_vdpau; > extern const HWContextType ff_hwcontext_type_videotoolbox; > extern const HWContextType ff_hwcontext_type_mediacodec; > extern const HWContextType ff_hwcontext_type_vulkan; > +extern const HWContextType ff_hwcontext_type_amf; > > #endif /* AVUTIL_HWCONTEXT_INTERNAL_H */ > diff --git a/libavutil/pixdesc.c b/libavutil/pixdesc.c > index f6d4d01460..ebc79b0c74 100644 > --- a/libavutil/pixdesc.c > +++ b/libavutil/pixdesc.c > @@ -2125,6 +2125,10 @@ static const AVPixFmtDescriptor av_pix_fmt_descriptors[AV_PIX_FMT_NB] = { > .name = "cuda", > .flags = AV_PIX_FMT_FLAG_HWACCEL, > }, > + [AV_PIX_FMT_AMF] = { > + .name = "amf", > + .flags = AV_PIX_FMT_FLAG_HWACCEL, > + }, > [AV_PIX_FMT_AYUV64LE] = { > .name = "ayuv64le", > .nb_components = 4, > diff --git a/libavutil/pixfmt.h b/libavutil/pixfmt.h > index 9c87571f49..06459be62c 100644 > --- a/libavutil/pixfmt.h > +++ b/libavutil/pixfmt.h > @@ -251,6 +251,10 @@ enum AVPixelFormat { > * exactly as for system memory frames. > */ > AV_PIX_FMT_CUDA, > + /** > + * HW acceleration through AMF. data[3] contain AMFSurface pointer > + */ > + AV_PIX_FMT_AMF, IMO naming pixel formats to be identical to the API hosting them was a mistake. It's an AMFSurface, call the format AV_PIX_FMT_AMF_SURFACE. Also no reason to copy old formats by putting your pointer in data[3], put it in data[0]. Thanks, - Mark _______________________________________________ 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".
prev parent reply other threads:[~2024-02-14 20:56 UTC|newest] Thread overview: 18+ messages / expand[flat|nested] mbox.gz Atom feed top 2024-02-14 1:55 Dmitrii Ovchinnikov 2024-02-14 1:55 ` [FFmpeg-devel] [PATCH 2/9] libavcodec: add amfdec Dmitrii Ovchinnikov 2024-02-14 23:41 ` Mark Thompson 2024-02-14 1:55 ` [FFmpeg-devel] [PATCH 3/9] avcodec/amfenc: Fixes the color information in the output Dmitrii Ovchinnikov 2024-02-14 1:55 ` [FFmpeg-devel] [PATCH 4/9] avcodec/amfenc: HDR metadata Dmitrii Ovchinnikov 2024-02-14 1:55 ` [FFmpeg-devel] [PATCH 5/9] avcodec/amfenc: add 10 bit encoding in av1_amf Dmitrii Ovchinnikov 2024-02-14 1:55 ` [FFmpeg-devel] [PATCH 6/9] avcodec/amfenc: add smart access video option Dmitrii Ovchinnikov 2024-02-14 1:55 ` [FFmpeg-devel] [PATCH 7/9] libavcodec/amfenc: redesign to use hwcontext_amf Dmitrii Ovchinnikov 2024-02-14 1:55 ` [FFmpeg-devel] [PATCH 8/9] avfilter/scale_amf: Add AMF HW scaler & color converter Dmitrii Ovchinnikov 2024-02-14 15:08 ` Timo Rothenpieler 2024-02-14 15:27 ` Evgeny Pavlov 2024-02-14 16:26 ` Dennis Mungai 2024-02-19 11:18 ` Evgeny Pavlov 2024-02-19 14:43 ` Dennis Mungai 2024-02-14 1:55 ` [FFmpeg-devel] [PATCH 9/9] doc/filters: Add documentation for AMF filters Dmitrii Ovchinnikov 2024-02-14 2:56 ` [FFmpeg-devel] [PATCH 1/9] libavutil: add hwcontext_amf James Almer 2024-02-14 16:48 ` Dmitrii Ovchinnikov 2024-02-14 20:56 ` Mark Thompson [this message]
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=3fd27a93-d2bd-41f8-8572-ef4ddfddc871@jkqxz.net \ --to=sw@jkqxz.net \ --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