On 6/27/20, C Hanish Menon <hanish...@gmail.com> wrote:
> Hi,
>
> It is a new video filter which I created to do detailing of the Intel
> Tile-X and Tile-Y framebuffer layouts into linear layout using a logic
> which runs on the cpu. It can be used if one uses kmsgrab and hwdownload to
> capture screen on a Intel GPU based system, so that one can get proper
> screen capture.
>
> Without this kmsgrab will generate a unusable/scrambled capture, because
> the contents will be tiled. I had this issue few days back when trying to
> capture screen with wayland, so created this.
>
> In the patch submitted, I have added the doc/filters.texi, which mentions
> the same.
Filter is marginally useful, it is done in CPU, completely
invalidating any possible gain using hw path.
>
>
>
> On Sun, Jun 28, 2020 at 1:30 AM Paul B Mahol <one...@gmail.com> wrote:
>
>> What is this?
>>
>> Missing documentation.
>> NAK
>>
>> On 6/27/20, hanishkvc <hanish...@gmail.com> wrote:
>> > v02-20200627IST2331
>> >
>> > Unrolled Intel Legacy Tile-Y detiling logic.
>> >
>> > Also a consolidated patch file, instead of the previous development
>> > flow based multiple patch files.
>> >
>> > v01-20200627IST1308
>> >
>> > Implemented Intel Legacy Tile-X and Tile-Y detiling logic
>> >
>> > NOTES:
>> >
>> > This video filter allows framebuffers which are tiled to be detiled
>> > using logic running on the cpu, into a linear layout.
>> >
>> > Currently it supports Intel Legacy Tile-X and Tile-Y layout detiling.
>> > THis should help one to work with frames captured (say using kmsgrab)
>> > on laptops having Intel GPU.
>> >
>> > Tile-X conversion logic has been explicitly cross checked, with Tile-X
>> > based frames. However Tile-Y conv logic hasnt been tested with Tile-Y
>> > based frames, but it should potentially do the job, based on my current
>> > understanding of the Tile-Y layout format.
>> >
>> > TODO1: At a later time have to generate Tile-Y based frames, and then
>> > cross check the corresponding logic explicitly.
>> >
>> > TODO2: May be use OpenGL or Vulcan buffer helper routines to do the
>> > layout conversion. But some online discussions from sometime back seem
>> > to indicate that this path is not fully bug free currently.
>> > ---
>> > Changelog | 1 +
>> > doc/filters.texi | 62 ++++++++
>> > libavfilter/Makefile | 1 +
>> > libavfilter/allfilters.c | 1 +
>> > libavfilter/vf_fbdetile.c | 309 ++++++++++++++++++++++++++++++++++++++
>> > 5 files changed, 374 insertions(+)
>> > create mode 100644 libavfilter/vf_fbdetile.c
>> >
>> > diff --git a/Changelog b/Changelog
>> > index a60e7d2eb8..0e03491f6a 100644
>> > --- a/Changelog
>> > +++ b/Changelog
>> > @@ -2,6 +2,7 @@ Entries are sorted chronologically from oldest to
>> youngest
>> > within each release,
>> > releases are sorted from youngest to oldest.
>> >
>> > version <next>:
>> > +- fbdetile cpu based framebuffer layout detiling video filter
>> > - AudioToolbox output device
>> > - MacCaption demuxer
>> >
>> > diff --git a/doc/filters.texi b/doc/filters.texi
>> > index 3c2dd2eb90..73ba21af89 100644
>> > --- a/doc/filters.texi
>> > +++ b/doc/filters.texi
>> > @@ -12210,6 +12210,68 @@ It accepts the following optional parameters:
>> > The number of the CUDA device to use
>> > @end table
>> >
>> > +@anchor{fbdetile}
>> > +@section fbdetile
>> > +
>> > +Detiles the Framebuffer tile layout into a linear layout using CPU.
>> > +
>> > +It currently supports conversion from Intel legacy tile-x and tile-y
>> > layouts
>> > +into a linear layout. This is useful if one is using kmsgrab and
>> hwdownload
>> > +to capture a screen which is using one of these non-linear layouts.
>> > +
>> > +Currently it expects the data to be a 32bit RGB based pixel format.
>> However
>> > +the logic doesnt do any pixel format conversion or so. Later will be
>> > enabling
>> > +16bit RGB data also, as the logic is transparent to it at one level.
>> > +
>> > +One could either insert this into the filter chain while capturing
>> itself,
>> > +or else, if it is slowing things down or so, then one could instead
>> insert
>> > +it into the filter chain during playback or transcoding or so.
>> > +
>> > +It supports the following optional parameters
>> > +
>> > +@table @option
>> > +@item type
>> > +Specify which detiling conversion to apply. The supported values are
>> > +@table @var
>> > +@item 0
>> > +intel tile-x to linear conversion (the default)
>> > +@item 1
>> > +intel tile-y to linear conversion.
>> > +@end table
>> > +@end table
>> > +
>> > +If one wants to convert during capture itself, one could do
>> > +@example
>> > +ffmpeg -f kmsgrab -i - -vf "hwdownload, fbdetile" OUTPUT
>> > +@end example
>> > +
>> > +However if one wants to convert after the tiled data has been already
>> > captured
>> > +@example
>> > +ffmpeg -i INPUT -vf "fbdetile" OUTPUT
>> > +@end example
>> > +@example
>> > +ffplay -i INPUT -vf "fbdetile"
>> > +@end example
>> > +
>> > +NOTE: While transcoding a test 1080p h264 stream, with 276 frames,
>> > with
>> two
>> > +runs of each situation, the performance was has given below. However
>> this
>> > +was for the older | initial version of the logic, as well as it was
>> > run
>> on
>> > +the default linux chromebook->vm->container, so the perf values need
>> not be
>> > +proper. But in a relative sense the overhead would be similar.
>> > +@example
>> > +rm out.mp4; time ./ffmpeg -i input.mp4 out.mp4
>> > +rm out.mp4; time ./ffmpeg -i input.mp4 -vf fbdetile=0 out.mp4
>> > +rm out.mp4; time ./ffmpeg -i input.mp4 -vf fbdetile=1 out.mp4
>> > +@end example
>> > +@table @option
>> > +@item with no fbdetile filter
>> > +it took ~7.28 secs,
>> > +@item with fbdetile=0 filter
>> > +it took ~8.69 secs,
>> > +@item with fbdetile=1 filter
>> > +it took ~9.20 secs.
>> > +@end table
>> > +
>> > @section hqx
>> >
>> > Apply a high-quality magnification filter designed for pixel art. This
>> > filter
>> > diff --git a/libavfilter/Makefile b/libavfilter/Makefile
>> > index 5123540653..bdb0c379ae 100644
>> > --- a/libavfilter/Makefile
>> > +++ b/libavfilter/Makefile
>> > @@ -280,6 +280,7 @@ OBJS-$(CONFIG_HWDOWNLOAD_FILTER) +=
>> > vf_hwdownload.o
>> > OBJS-$(CONFIG_HWMAP_FILTER) += vf_hwmap.o
>> > OBJS-$(CONFIG_HWUPLOAD_CUDA_FILTER) += vf_hwupload_cuda.o
>> > OBJS-$(CONFIG_HWUPLOAD_FILTER) += vf_hwupload.o
>> > +OBJS-$(CONFIG_FBDETILE_FILTER) += vf_fbdetile.o
>> > OBJS-$(CONFIG_HYSTERESIS_FILTER) += vf_hysteresis.o
>> framesync.o
>> > OBJS-$(CONFIG_IDET_FILTER) += vf_idet.o
>> > OBJS-$(CONFIG_IL_FILTER) += vf_il.o
>> > diff --git a/libavfilter/allfilters.c b/libavfilter/allfilters.c
>> > index 1183e40267..f8dceb2a88 100644
>> > --- a/libavfilter/allfilters.c
>> > +++ b/libavfilter/allfilters.c
>> > @@ -265,6 +265,7 @@ extern AVFilter ff_vf_hwdownload;
>> > extern AVFilter ff_vf_hwmap;
>> > extern AVFilter ff_vf_hwupload;
>> > extern AVFilter ff_vf_hwupload_cuda;
>> > +extern AVFilter ff_vf_fbdetile;
>> > extern AVFilter ff_vf_hysteresis;
>> > extern AVFilter ff_vf_idet;
>> > extern AVFilter ff_vf_il;
>> > diff --git a/libavfilter/vf_fbdetile.c b/libavfilter/vf_fbdetile.c
>> > new file mode 100644
>> > index 0000000000..8b20c96d2c
>> > --- /dev/null
>> > +++ b/libavfilter/vf_fbdetile.c
>> > @@ -0,0 +1,309 @@
>> > +/*
>> > + * Copyright (c) 2020 HanishKVC
>> > + *
>> > + * 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
>> > + */
>> > +
>> > +/**
>> > + * @file
>> > + * Detile the Frame buffer's tile layout using the cpu
>> > + * Currently it supports the legacy Intel Tile X layout detiling.
>> > + *
>> > + */
>> > +
>> > +/*
>> > + * ToThink|Check: Optimisations
>> > + *
>> > + * Does gcc setting used by ffmpeg allows memcpy | stringops inlining,
>> > + * loop unrolling, better native matching instructions, additional
>> > + * optimisations, ...
>> > + *
>> > + * Does gcc map to optimal memcpy logic, based on the situation it is
>> > + * used in.
>> > + *
>> > + * If not, may be look at vector_size or intrinsics or appropriate
>> > arch
>> > + * and cpu specific inline asm or ...
>> > + *
>> > + */
>> > +
>> > +#include "libavutil/avassert.h"
>> > +#include "libavutil/imgutils.h"
>> > +#include "libavutil/opt.h"
>> > +#include "avfilter.h"
>> > +#include "formats.h"
>> > +#include "internal.h"
>> > +#include "video.h"
>> > +
>> > +enum FilterMode {
>> > + TYPE_INTELX,
>> > + TYPE_INTELY,
>> > + NB_TYPE
>> > +};
>> > +
>> > +typedef struct FBDetileContext {
>> > + const AVClass *class;
>> > + int width, height;
>> > + int type;
>> > +} FBDetileContext;
>> > +
>> > +#define OFFSET(x) offsetof(FBDetileContext, x)
>> > +#define FLAGS AV_OPT_FLAG_FILTERING_PARAM|AV_OPT_FLAG_VIDEO_PARAM
>> > +static const AVOption fbdetile_options[] = {
>> > + { "type", "set framebuffer format_modifier type", OFFSET(type),
>> > AV_OPT_TYPE_INT, {.i64=TYPE_INTELX}, 0, NB_TYPE-1, FLAGS, "type" },
>> > + { "intelx", "Intel Tile-X layout", 0, AV_OPT_TYPE_CONST,
>> > {.i64=TYPE_INTELX}, INT_MIN, INT_MAX, FLAGS, "type" },
>> > + { "intely", "Intel Tile-Y layout", 0, AV_OPT_TYPE_CONST,
>> > {.i64=TYPE_INTELY}, INT_MIN, INT_MAX, FLAGS, "type" },
>> > + { NULL }
>> > +};
>> > +
>> > +AVFILTER_DEFINE_CLASS(fbdetile);
>> > +
>> > +static av_cold int init(AVFilterContext *ctx)
>> > +{
>> > + FBDetileContext *fbdetile = ctx->priv;
>> > +
>> > + if (fbdetile->type == TYPE_INTELX) {
>> > + fprintf(stderr,"INFO:fbdetile:init: Intel tile-x to
>> > linear\n");
>> > + } else if (fbdetile->type == TYPE_INTELY) {
>> > + fprintf(stderr,"INFO:fbdetile:init: Intel tile-y to
>> > linear\n");
>> > + } else {
>> > + fprintf(stderr,"DBUG:fbdetile:init: Unknown Tile format
>> specified,
>> > shouldnt reach here\n");
>> > + }
>> > + fbdetile->width = 1920;
>> > + fbdetile->height = 1080;
>> > + return 0;
>> > +}
>> > +
>> > +static int query_formats(AVFilterContext *ctx)
>> > +{
>> > + // Currently only RGB based 32bit formats are specified
>> > + // TODO: Technically the logic is transparent to 16bit RGB formats
>> also
>> > + static const enum AVPixelFormat pix_fmts[] = {AV_PIX_FMT_RGB0,
>> > AV_PIX_FMT_0RGB, AV_PIX_FMT_BGR0, AV_PIX_FMT_0BGR,
>> > + AV_PIX_FMT_RGBA,
>> > AV_PIX_FMT_ARGB, AV_PIX_FMT_BGRA, AV_PIX_FMT_ABGR,
>> > + AV_PIX_FMT_NONE};
>> > + AVFilterFormats *fmts_list;
>> > +
>> > + fmts_list = ff_make_format_list(pix_fmts);
>> > + if (!fmts_list)
>> > + return AVERROR(ENOMEM);
>> > + return ff_set_common_formats(ctx, fmts_list);
>> > +}
>> > +
>> > +static int config_props(AVFilterLink *inlink)
>> > +{
>> > + AVFilterContext *ctx = inlink->dst;
>> > + FBDetileContext *fbdetile = ctx->priv;
>> > +
>> > + fbdetile->width = inlink->w;
>> > + fbdetile->height = inlink->h;
>> > + fprintf(stderr,"DBUG:fbdetile:config_props: %d x %d\n",
>> > fbdetile->width, fbdetile->height);
>> > +
>> > + return 0;
>> > +}
>> > +
>> > +static void detile_intelx(AVFilterContext *ctx, int w, int h,
>> > + uint8_t *dst, int dstLineSize,
>> > + const uint8_t *src, int srcLineSize)
>> > +{
>> > + // Offsets and LineSize are in bytes
>> > + int tileW = 128; // For a 32Bit / Pixel framebuffer, 512/4
>> > + int tileH = 8;
>> > +
>> > + if (w*4 != srcLineSize) {
>> > + fprintf(stderr,"DBUG:fbdetile:intelx: w%dxh%d, dL%d, sL%d\n",
>> w, h,
>> > dstLineSize, srcLineSize);
>> > + fprintf(stderr,"ERRR:fbdetile:intelx: dont support LineSize |
>> Pitch
>> > going beyond width\n");
>> > + }
>> > + int sO = 0;
>> > + int dX = 0;
>> > + int dY = 0;
>> > + int nTRows = (w*h)/tileW;
>> > + int cTR = 0;
>> > + while (cTR < nTRows) {
>> > + int dO = dY*dstLineSize + dX*4;
>> > +#ifdef DEBUG_FBTILE
>> > + fprintf(stderr,"DBUG:fbdetile:intelx: dX%d dY%d, sO%d,
>> > dO%d\n",
>> dX,
>> > dY, sO, dO);
>> > +#endif
>> > + memcpy(dst+dO+0*dstLineSize, src+sO+0*512, 512);
>> > + memcpy(dst+dO+1*dstLineSize, src+sO+1*512, 512);
>> > + memcpy(dst+dO+2*dstLineSize, src+sO+2*512, 512);
>> > + memcpy(dst+dO+3*dstLineSize, src+sO+3*512, 512);
>> > + memcpy(dst+dO+4*dstLineSize, src+sO+4*512, 512);
>> > + memcpy(dst+dO+5*dstLineSize, src+sO+5*512, 512);
>> > + memcpy(dst+dO+6*dstLineSize, src+sO+6*512, 512);
>> > + memcpy(dst+dO+7*dstLineSize, src+sO+7*512, 512);
>> > + dX += tileW;
>> > + if (dX >= w) {
>> > + dX = 0;
>> > + dY += 8;
>> > + }
>> > + sO = sO + 8*512;
>> > + cTR += 8;
>> > + }
>> > +}
>> > +
>> > +/*
>> > + * Intel Legacy Tile-Y layout conversion support
>> > + *
>> > + * currently done in a simple dumb way. Two low hanging optimisations
>> > + * that could be readily applied are
>> > + *
>> > + * a) unrolling the inner for loop
>> > + * --- Given small size memcpy, should help, DONE
>> > + *
>> > + * b) using simd based 128bit loading and storing along with prefetch
>> > + * hinting.
>> > + *
>> > + * TOTHINK|CHECK: Does memcpy already does this and more if
>> > situation
>> > + * is right?!
>> > + *
>> > + * As code (or even intrinsics) would be specific to each
>> architecture,
>> > + * avoiding for now. Later have to check if vector_size attribute
>> > and
>> > + * corresponding implementation by gcc can handle different
>> > architectures
>> > + * properly, such that it wont become worse than memcpy provided
>> > for
>> > that
>> > + * architecture.
>> > + *
>> > + * Or maybe I could even merge the two intel detiling logics into one,
>> as
>> > + * the semantic and flow is almost same for both logics.
>> > + *
>> > + */
>> > +static void detile_intely(AVFilterContext *ctx, int w, int h,
>> > + uint8_t *dst, int dstLineSize,
>> > + const uint8_t *src, int srcLineSize)
>> > +{
>> > + // Offsets and LineSize are in bytes
>> > + int tileW = 4; // For a 32Bit / Pixel framebuffer, 16/4
>> > + int tileH = 32;
>> > +
>> > + if (w*4 != srcLineSize) {
>> > + fprintf(stderr,"DBUG:fbdetile:intely: w%dxh%d, dL%d, sL%d\n",
>> w, h,
>> > dstLineSize, srcLineSize);
>> > + fprintf(stderr,"ERRR:fbdetile:intely: dont support LineSize |
>> Pitch
>> > going beyond width\n");
>> > + }
>> > + int sO = 0;
>> > + int dX = 0;
>> > + int dY = 0;
>> > + int nTRows = (w*h)/tileW;
>> > + int cTR = 0;
>> > + while (cTR < nTRows) {
>> > + int dO = dY*dstLineSize + dX*4;
>> > +#ifdef DEBUG_FBTILE
>> > + fprintf(stderr,"DBUG:fbdetile:intely: dX%d dY%d, sO%d,
>> > dO%d\n",
>> dX,
>> > dY, sO, dO);
>> > +#endif
>> > +
>> > + memcpy(dst+dO+0*dstLineSize, src+sO+0*16, 16);
>> > + memcpy(dst+dO+1*dstLineSize, src+sO+1*16, 16);
>> > + memcpy(dst+dO+2*dstLineSize, src+sO+2*16, 16);
>> > + memcpy(dst+dO+3*dstLineSize, src+sO+3*16, 16);
>> > + memcpy(dst+dO+4*dstLineSize, src+sO+4*16, 16);
>> > + memcpy(dst+dO+5*dstLineSize, src+sO+5*16, 16);
>> > + memcpy(dst+dO+6*dstLineSize, src+sO+6*16, 16);
>> > + memcpy(dst+dO+7*dstLineSize, src+sO+7*16, 16);
>> > + memcpy(dst+dO+8*dstLineSize, src+sO+8*16, 16);
>> > + memcpy(dst+dO+9*dstLineSize, src+sO+9*16, 16);
>> > + memcpy(dst+dO+10*dstLineSize, src+sO+10*16, 16);
>> > + memcpy(dst+dO+11*dstLineSize, src+sO+11*16, 16);
>> > + memcpy(dst+dO+12*dstLineSize, src+sO+12*16, 16);
>> > + memcpy(dst+dO+13*dstLineSize, src+sO+13*16, 16);
>> > + memcpy(dst+dO+14*dstLineSize, src+sO+14*16, 16);
>> > + memcpy(dst+dO+15*dstLineSize, src+sO+15*16, 16);
>> > + memcpy(dst+dO+16*dstLineSize, src+sO+16*16, 16);
>> > + memcpy(dst+dO+17*dstLineSize, src+sO+17*16, 16);
>> > + memcpy(dst+dO+18*dstLineSize, src+sO+18*16, 16);
>> > + memcpy(dst+dO+19*dstLineSize, src+sO+19*16, 16);
>> > + memcpy(dst+dO+20*dstLineSize, src+sO+20*16, 16);
>> > + memcpy(dst+dO+21*dstLineSize, src+sO+21*16, 16);
>> > + memcpy(dst+dO+22*dstLineSize, src+sO+22*16, 16);
>> > + memcpy(dst+dO+23*dstLineSize, src+sO+23*16, 16);
>> > + memcpy(dst+dO+24*dstLineSize, src+sO+24*16, 16);
>> > + memcpy(dst+dO+25*dstLineSize, src+sO+25*16, 16);
>> > + memcpy(dst+dO+26*dstLineSize, src+sO+26*16, 16);
>> > + memcpy(dst+dO+27*dstLineSize, src+sO+27*16, 16);
>> > + memcpy(dst+dO+28*dstLineSize, src+sO+28*16, 16);
>> > + memcpy(dst+dO+29*dstLineSize, src+sO+29*16, 16);
>> > + memcpy(dst+dO+30*dstLineSize, src+sO+30*16, 16);
>> > + memcpy(dst+dO+31*dstLineSize, src+sO+31*16, 16);
>> > +
>> > + dX += tileW;
>> > + if (dX >= w) {
>> > + dX = 0;
>> > + dY += 32;
>> > + }
>> > + sO = sO + 32*16;
>> > + cTR += 32;
>> > + }
>> > +}
>> > +
>> > +static int filter_frame(AVFilterLink *inlink, AVFrame *in)
>> > +{
>> > + AVFilterContext *ctx = inlink->dst;
>> > + FBDetileContext *fbdetile = ctx->priv;
>> > + AVFilterLink *outlink = ctx->outputs[0];
>> > + AVFrame *out;
>> > +
>> > + out = ff_get_video_buffer(outlink, outlink->w, outlink->h);
>> > + if (!out) {
>> > + av_frame_free(&in);
>> > + return AVERROR(ENOMEM);
>> > + }
>> > + av_frame_copy_props(out, in);
>> > +
>> > + if (fbdetile->type == TYPE_INTELX) {
>> > + detile_intelx(ctx, fbdetile->width, fbdetile->height,
>> > + out->data[0], out->linesize[0],
>> > + in->data[0], in->linesize[0]);
>> > + } else if (fbdetile->type == TYPE_INTELY) {
>> > + detile_intely(ctx, fbdetile->width, fbdetile->height,
>> > + out->data[0], out->linesize[0],
>> > + in->data[0], in->linesize[0]);
>> > + }
>> > +
>> > + av_frame_free(&in);
>> > + return ff_filter_frame(outlink, out);
>> > +}
>> > +
>> > +static av_cold void uninit(AVFilterContext *ctx)
>> > +{
>> > +
>> > +}
>> > +
>> > +static const AVFilterPad fbdetile_inputs[] = {
>> > + {
>> > + .name = "default",
>> > + .type = AVMEDIA_TYPE_VIDEO,
>> > + .config_props = config_props,
>> > + .filter_frame = filter_frame,
>> > + },
>> > + { NULL }
>> > +};
>> > +
>> > +static const AVFilterPad fbdetile_outputs[] = {
>> > + {
>> > + .name = "default",
>> > + .type = AVMEDIA_TYPE_VIDEO,
>> > + },
>> > + { NULL }
>> > +};
>> > +
>> > +AVFilter ff_vf_fbdetile = {
>> > + .name = "fbdetile",
>> > + .description = NULL_IF_CONFIG_SMALL("Detile Framebuffer using
>> CPU"),
>> > + .priv_size = sizeof(FBDetileContext),
>> > + .init = init,
>> > + .uninit = uninit,
>> > + .query_formats = query_formats,
>> > + .inputs = fbdetile_inputs,
>> > + .outputs = fbdetile_outputs,
>> > + .priv_class = &fbdetile_class,
>> > +};
>> > --
>> > 2.20.1
>> >
>> > _______________________________________________
>> > ffmpeg-devel mailing list
>> > ffmpeg-devel@ffmpeg.org
>> > https://ffmpeg.org/mailman/listinfo/ffmpeg-devel
>> >
>> > To unsubscribe, visit link above, or email
>> > ffmpeg-devel-requ...@ffmpeg.org with subject "unsubscribe".
>>
>
>
> --
> Keep ;-)
> HanishKVC
>
_______________________________________________
ffmpeg-devel mailing list
ffmpeg-devel@ffmpeg.org
https://ffmpeg.org/mailman/listinfo/ffmpeg-devel
To unsubscribe, visit link above, or email
ffmpeg-devel-requ...@ffmpeg.org with subject "unsubscribe".
