Re: [FFmpeg-devel] [PATCH 3/4] ffmpeg: use av_buffersrc_close().

2014-08-02 Thread Michael Niedermayer 
On Wed, Jul 30, 2014 at 11:44:48PM +0200, Nicolas George wrote:
 Signed-off-by: Nicolas George geo...@nsup.org
 ---
  ffmpeg.c | 14 --
  1 file changed, 4 insertions(+), 10 deletions(-)

LGTM

[...]
-- 
Michael GnuPG fingerprint: 9FF2128B147EF6730BADF133611EC787040B0FAB

Asymptotically faster algorithms should always be preferred if you have
asymptotical amounts of data


signature.asc
Description: Digital signature
___
ffmpeg-devel mailing list
ffmpeg-devel@ffmpeg.org
http://ffmpeg.org/mailman/listinfo/ffmpeg-devel

Re: [FFmpeg-devel] [PATCH 2/4] lavfi/buffersrc: add add av_buffersrc_close().

2014-08-02 Thread Michael Niedermayer 
On Wed, Jul 30, 2014 at 11:44:47PM +0200, Nicolas George wrote:
 Also deprecate adding a NULL frame to mark EOF.
 
 TODO APIchanges entry, version bump.
 
 Signed-off-by: Nicolas George geo...@nsup.org
 ---
  libavfilter/buffersrc.c | 40 ++--
  libavfilter/buffersrc.h | 15 +++
  2 files changed, 45 insertions(+), 10 deletions(-)
 
 diff --git a/libavfilter/buffersrc.c b/libavfilter/buffersrc.c
 index 27d3db0..6d71587 100644
 --- a/libavfilter/buffersrc.c
 +++ b/libavfilter/buffersrc.c
 @@ -63,6 +63,7 @@ typedef struct BufferSourceContext {
  char*channel_layout_str;
  
  int eof;
 +int64_t eof_pts;
  } BufferSourceContext;
  
  #define CHECK_VIDEO_PARAM_CHANGE(s, c, width, height, format)\
 @@ -77,6 +78,27 @@ typedef struct BufferSourceContext {
  return AVERROR(EINVAL);\
  }
  
 +static int push_if_flag(AVFilterContext *ctx, int flags)
 +{
 +return (flags  AV_BUFFERSRC_FLAG_PUSH) ?
 +ctx-output_pads[0].request_frame(ctx-outputs[0]) : 0;
 +}
 +
 +int av_buffersrc_close(AVFilterContext *ctx, int64_t pts, int flags)
 +{
 +BufferSourceContext *s = ctx-priv;
 +
 +if (pts == AV_NOPTS_VALUE) {
 +av_log(ctx, AV_LOG_WARNING, No EOF timestamp\n);
 +/* FIXME use duration for audio */
 +pts = av_rescale_q(ctx-outputs[0]-current_pts,
 +   AV_TIME_BASE_Q, ctx-outputs[0]-time_base) + 1;
 +}
 +s-eof_pts = pts;
 +s-eof = 1;
 +return push_if_flag(ctx, flags);
 +}
 +
  int attribute_align_arg av_buffersrc_write_frame(AVFilterContext *ctx, const 
 AVFrame *frame)
  {
  return av_buffersrc_add_frame_flags(ctx, (AVFrame *)frame,
 @@ -125,8 +147,7 @@ static int 
 av_buffersrc_add_frame_internal(AVFilterContext *ctx,
  s-nb_failed_requests = 0;
  
  if (!frame) {
 -s-eof = 1;
 -return 0;
 +return av_buffersrc_close(ctx, AV_NOPTS_VALUE, flags);
  } else if (s-eof)
  return AVERROR(EINVAL);
  
 @@ -177,11 +198,7 @@ static int 
 av_buffersrc_add_frame_internal(AVFilterContext *ctx,
  return ret;
  }
  
 -if ((flags  AV_BUFFERSRC_FLAG_PUSH))
 -if ((ret = ctx-output_pads[0].request_frame(ctx-outputs[0]))  0)
 -return ret;
 -
 -return 0;
 +return push_if_flag(ctx, flags);
  }

some of this patch seems to be factorizing code,
other add fuctionality like ff_filter_link_close()

all LGTM i think but maybe should be split

[...]

-- 
Michael GnuPG fingerprint: 9FF2128B147EF6730BADF133611EC787040B0FAB

When you are offended at any man's fault, turn to yourself and study your
own failings. Then you will forget your anger. -- Epictetus


signature.asc
Description: Digital signature
___
ffmpeg-devel mailing list
ffmpeg-devel@ffmpeg.org
http://ffmpeg.org/mailman/listinfo/ffmpeg-devel

Re: [FFmpeg-devel] HEVC SIMD Optimization on PowerPC

2014-08-02 Thread compn 
On Sat, 2 Aug 2014 13:37:10 +0800
Zhenan Lin linzhe...@pku.edu.cn wrote:

 Hi, all,
 
  
 
 Our team plans to contribute SIMD optimization on PowerPC for HEVC

i thought powerpc was dead?
i dont know if anyone is working on it.

-compn
___
ffmpeg-devel mailing list
ffmpeg-devel@ffmpeg.org
http://ffmpeg.org/mailman/listinfo/ffmpeg-devel

Re: [FFmpeg-devel] [PATCH 0/4] Exploit compile-time constant

2014-08-02 Thread Michael Niedermayer 
On Mon, Jul 28, 2014 at 05:17:24PM +, Christophe Gisquet wrote:
 MAX_PB_SIZE is used or assumed for various buffer strides. In some cases,
 it is used as constant parameter(s) to functions.
 
 Make use of that knowledge to:
 - not pass the parameter
 - avoid extra GPR usage
 - precompute addresses / offsets
 

 Premature optimization and overall not that useful.
 Before:
 46092 decicycles in oma, 1028766 runs, 19810 skips
 10174 decicycles in chroma, 2065859 runs, 31293 skips
 
 After:
 45634 decicycles in luma, 1027414 runs, 21162 skips
 9932 decicycles in chroma, 2063780 runs, 33372 skips

is this for apply/push or just RFC/WIP ?
you say Premature optimization and overall not that useful.
i would tend to suggest to apply it as it improves speed ...


 
 Christophe Gisquet (4):
   hevc: move MAX_PB_SIZE declaration
   hevcdsp: remove compilation-time-fixed parameter
   hevcdsp: remove more instances of compile-time-fixed parameters
   x86: hevcdsp: use compilation-time-fixed constant

seems to fail with
libavcodec/x86/hevc_mc.asm:1258: error: (add:2) cannot reference symbol 
`MAX_PB_SIZE' in preprocessor


[...]
-- 
Michael GnuPG fingerprint: 9FF2128B147EF6730BADF133611EC787040B0FAB

Republics decline into democracies and democracies degenerate into
despotisms. -- Aristotle


signature.asc
Description: Digital signature
___
ffmpeg-devel mailing list
ffmpeg-devel@ffmpeg.org
http://ffmpeg.org/mailman/listinfo/ffmpeg-devel

Re: [FFmpeg-devel] HEVC SIMD Optimization on PowerPC

2014-08-02 Thread Michael Niedermayer 
On Sat, Aug 02, 2014 at 07:18:22AM -0400, compn wrote:
 On Sat, 2 Aug 2014 13:37:10 +0800
 Zhenan Lin linzhe...@pku.edu.cn wrote:
 
  Hi, all,
  
   
  
  Our team plans to contribute SIMD optimization on PowerPC for HEVC
 
 i thought powerpc was dead?

i thought so too, but powerpc optimizations certainly are welcome,
if someone wants to write some

[...]
-- 
Michael GnuPG fingerprint: 9FF2128B147EF6730BADF133611EC787040B0FAB

it is not once nor twice but times without number that the same ideas make
their appearance in the world. -- Aristotle


signature.asc
Description: Digital signature
___
ffmpeg-devel mailing list
ffmpeg-devel@ffmpeg.org
http://ffmpeg.org/mailman/listinfo/ffmpeg-devel

Re: [FFmpeg-devel] [PATCH] msvc: fix implicitly declared read/close.

2014-08-02 Thread Michael Niedermayer 
On Sat, Aug 02, 2014 at 12:01:37PM +1000, Matt Oliver wrote:
 MSVC/ICL does not have unistd.h as standard and instead declares read/close
 functions in io.h. This is currently handled in some files but not in a
 couple. Adding an include for io.h where needed adds the correct
 declarations.

  libavcodec/libxvid.c|4 
  libavcodec/libxvid_rc.c |4 
  libavutil/random_seed.c |3 +++
  3 files changed, 11 insertions(+)
 9df30ccc9f607fe515c1058f85e6f6bce0e07984  
 0001-msvc-fix-implicitly-declared-read-close.patch
 From a9df2f27cf88e5e2e0a518daaba41b1b9ce8bdcb Mon Sep 17 00:00:00 2001
 From: Matthew Oliver protogo...@gmail.com
 Date: Sat, 2 Aug 2014 11:08:29 +1000
 Subject: [PATCH] msvc: fix implicitly declared read/close.

applied

thanks

[...]

-- 
Michael GnuPG fingerprint: 9FF2128B147EF6730BADF133611EC787040B0FAB

I do not agree with what you have to say, but I'll defend to the death your
right to say it. -- Voltaire


signature.asc
Description: Digital signature
___
ffmpeg-devel mailing list
ffmpeg-devel@ffmpeg.org
http://ffmpeg.org/mailman/listinfo/ffmpeg-devel

[FFmpeg-devel] [PATCH] ffserver: deprecate Port and BindAddress options in favor of HTTPPort and HTTPBindAddress

2014-08-02 Thread Stefano Sabatini 
The new option names are more explicit.
---
 doc/ffserver.conf |  4 ++--
 doc/ffserver.texi | 19 ++-
 ffserver.c| 10 +++---
 3 files changed, 23 insertions(+), 10 deletions(-)

diff --git a/doc/ffserver.conf b/doc/ffserver.conf
index 094c093..b756961 100644
--- a/doc/ffserver.conf
+++ b/doc/ffserver.conf
@@ -1,11 +1,11 @@
 # Port on which the server is listening. You must select a different
 # port from your standard HTTP web server if it is running on the same
 # computer.
-Port 8090
+HTTPPort 8090
 
 # Address on which the server is bound. Only useful if you have
 # several network interfaces.
-BindAddress 0.0.0.0
+HTTPBindAddress 0.0.0.0
 
 # Number of simultaneous HTTP connections that can be handled. It has
 # to be defined *before* the MaxClients parameter, since it defines the
diff --git a/doc/ffserver.texi b/doc/ffserver.texi
index 2bf8152..77273d2 100644
--- a/doc/ffserver.texi
+++ b/doc/ffserver.texi
@@ -66,7 +66,7 @@ 
http://@var{ffserver_ip_address}:@var{http_port}/@var{feed_name}
 
 where @var{ffserver_ip_address} is the IP address of the machine where
 @command{ffserver} is installed, @var{http_port} is the port number of
-the HTTP server (configured through the @option{Port} option), and
+the HTTP server (configured through the @option{HTTPPort} option), and
 @var{feed_name} is the name of the corresponding feed defined in the
 configuration file.
 
@@ -101,7 +101,7 @@ 
http://@var{ffserver_ip_address}:@var{rtsp_port}/@var{stream_name}[@var{options}
 the configuration file. @var{options} is a list of options specified
 after the URL which affects how the stream is served by
 @command{ffserver}. @var{http_port} and @var{rtsp_port} are the HTTP
-and RTSP ports configured with the options @var{Port} and
+and RTSP ports configured with the options @var{HTTPPort} and
 @var{RTSPPort} respectively.
 
 In case the stream is associated to a feed, the encoding parameters
@@ -352,20 +352,29 @@ allow everybody else.
 
 @section Global options
 @table @option
+@item HTTPPort @var{port_number}
 @item Port @var{port_number}
 @item RTSPPort @var{port_number}
 
-Set TCP port number on which the HTTP/RTSP server is listening. You
-must select a different port from your standard HTTP web server if it
-is running on the same computer.
+@var{HTTPPort} sets the HTTP server listening TCP port number,
+@var{RTSPPort} sets the RTSP server listening TCP port number.
+
+@var{Port} is the equivalent of @var{HTTPPort} and is deprecated.
+
+You must select a different port from your standard HTTP web server if
+it is running on the same computer.
 
 If not specified, no corresponding server will be created.
 
+@item HTTPBindAddress @var{ip_address}
 @item BindAddress @var{ip_address}
 @item RTSPBindAddress @var{ip_address}
 Set address on which the HTTP/RTSP server is bound. Only useful if you
 have several network interfaces.
 
+@var{BindAddress} is the equivalent of @var{HTTPBindAddress} and is
+deprecated.
+
 @item MaxHTTPConnections @var{n}
 Set number of simultaneous HTTP connections that can be handled. It
 has to be defined @emph{before} the @option{MaxClients} parameter,
diff --git a/ffserver.c b/ffserver.c
index 00153b4..75049c1 100644
--- a/ffserver.c
+++ b/ffserver.c
@@ -4079,14 +4079,18 @@ static int parse_ffconfig(const char *filename)
 
 get_arg(cmd, sizeof(cmd), p);
 
-if (!av_strcasecmp(cmd, Port)) {
+if (!av_strcasecmp(cmd, Port) || !av_strcasecmp(cmd, HTTPPort)) {
+if (!av_strcasecmp(cmd, Port))
+WARNING(Port option is deprecated, use HTTPPort instead\n);
 get_arg(arg, sizeof(arg), p);
 val = atoi(arg);
 if (val  1 || val  65536) {
-ERROR(Invalid_port: %s\n, arg);
+ERROR(Invalid port: %s\n, arg);
 }
 my_http_addr.sin_port = htons(val);
-} else if (!av_strcasecmp(cmd, BindAddress)) {
+} else if (!av_strcasecmp(cmd, HTTPBindAddress) || 
!av_strcasecmp(cmd, BindAddress)) {
+if (!av_strcasecmp(cmd, BindAddress))
+WARNING(BindAddress option is deprecated, use HTTPBindAddress 
instead\n);
 get_arg(arg, sizeof(arg), p);
 if (resolve_host(my_http_addr.sin_addr, arg) != 0) {
 ERROR(%s:%d: Invalid host/IP address: %s\n, arg);
-- 
1.8.3.2

___
ffmpeg-devel mailing list
ffmpeg-devel@ffmpeg.org
http://ffmpeg.org/mailman/listinfo/ffmpeg-devel

[FFmpeg-devel] [PATCH] doc/ffserver: merge paragraph starting with What happens next? with previous one

2014-08-02 Thread Stefano Sabatini 
The name of the paragraph sounded a bit silly, and its text is small so
it's better to merge it with the previous paragraph.
---
 doc/ffserver.texi | 8 +++-
 1 file changed, 3 insertions(+), 5 deletions(-)

diff --git a/doc/ffserver.texi b/doc/ffserver.texi
index fb90dd6..2bf8152 100644
--- a/doc/ffserver.texi
+++ b/doc/ffserver.texi
@@ -203,11 +203,9 @@ WARNING: trying to stream test1.mpg doesn't work with WMP 
as it tries to
 transfer the entire file before starting to play.
 The same is true of AVI files.
 
-@section What happens next?
-
-You should edit the ffserver.conf file to suit your needs (in terms of
-frame rates etc). Then install ffserver and ffmpeg, write a script to start
-them up, and off you go.
+You should edit the @file{ffserver.conf} file to suit your needs (in
+terms of frame rates etc). Then install @command{ffserver} and
+@command{ffmpeg}, write a script to start them up, and off you go.
 
 @section What else can it do?
 
-- 
1.8.3.2

___
ffmpeg-devel mailing list
ffmpeg-devel@ffmpeg.org
http://ffmpeg.org/mailman/listinfo/ffmpeg-devel

Re: [FFmpeg-devel] [PATCH] doc/ffserver: merge paragraph starting with What happens next? with previous one

2014-08-02 Thread Timothy Gu 
On Aug 2, 2014 7:40 AM, Stefano Sabatini stefa...@gmail.com wrote:

 The name of the paragraph sounded a bit silly, and its text is small so
 it's better to merge it with the previous paragraph.
 ---
  doc/ffserver.texi | 8 +++-
  1 file changed, 3 insertions(+), 5 deletions(-)

OK.

[...]

Timothy
___
ffmpeg-devel mailing list
ffmpeg-devel@ffmpeg.org
http://ffmpeg.org/mailman/listinfo/ffmpeg-devel

Re: [FFmpeg-devel] [PATCH] ffserver: initialize pbuffer in prepare_sdp_description()

2014-08-02 Thread Stefano Sabatini 
On date Saturday 2014-08-02 14:25:21 +0200, Michael Niedermayer encoded:
 also check pbuffer before use
 
 Found-by: CSA
 Signed-off-by: Michael Niedermayer michae...@gmx.at
 ---
  ffserver.c |4 +++-
  1 file changed, 3 insertions(+), 1 deletion(-)
 
 diff --git a/ffserver.c b/ffserver.c
 index 00153b4..956d9f5 100644
 --- a/ffserver.c
 +++ b/ffserver.c
 @@ -2977,6 +2977,8 @@ static int prepare_sdp_description(FFStream *stream, 
 uint8_t **pbuffer,
  AVDictionaryEntry *entry = av_dict_get(stream-metadata, title, NULL, 
 0);
  int i;
  
 +*pbuffer = NULL;
 +
  avc =  avformat_alloc_context();
  if (avc == NULL || !rtp_format) {
  return -1;
 @@ -3013,7 +3015,7 @@ static int prepare_sdp_description(FFStream *stream, 
 uint8_t **pbuffer,
  av_free(avc);
  av_free(avs);
  
 -return strlen(*pbuffer);
 +return *pbuffer ? strlen(*pbuffer) : AVERROR(ENOMEM);
  }
  
  static void rtsp_cmd_options(HTTPContext *c, const char *url)

LGTM.
-- 
FFmpeg = Faithless Fostering Magnificient Portable Egregious Geek
___
ffmpeg-devel mailing list
ffmpeg-devel@ffmpeg.org
http://ffmpeg.org/mailman/listinfo/ffmpeg-devel

Re: [FFmpeg-devel] [PATCH 1/4] lavfi: add filter metaframes infrastructure.

2014-08-02 Thread Stefano Sabatini 
On date Wednesday 2014-07-30 23:44:46 +0200, Nicolas George encoded:
 Metaframes are frames without data, identified by a negative
 format code, used to carry special conditions.
 They are sent only to filter that declare supporting them.
 The only metaframe for now is EOF; this mechanism augments
 the current mechanism based on request_frame() returning
 AVERROR_EOF, with the advantage that the EOF metaframe carries
 a timestamp.
 The metaframes are a purely internal API and do not leak to
 the application.
 
 Signed-off-by: Nicolas George geo...@nsup.org
 ---
  libavfilter/avfilter.c | 73 
 +-
  libavfilter/internal.h | 34 +++
  2 files changed, 100 insertions(+), 7 deletions(-)
 
 
 Changed the name to metaframes.
 Allocate the frame with the classic method.
 
 
 diff --git a/libavfilter/avfilter.c b/libavfilter/avfilter.c
 index 7b11467..7894173 100644
 --- a/libavfilter/avfilter.c
 +++ b/libavfilter/avfilter.c
 @@ -346,15 +346,16 @@ int ff_request_frame(AVFilterLink *link)
  ret = link-srcpad-request_frame(link);
  else if (link-src-inputs[0])
  ret = ff_request_frame(link-src-inputs[0]);
 -if (ret == AVERROR_EOF  link-partial_buf) {
 -AVFrame *pbuf = link-partial_buf;
 -link-partial_buf = NULL;
 -ret = ff_filter_frame_framed(link, pbuf);
 -}
  if (ret  0) {
 +if (!link-frame_requested) {
 +av_assert0(ret == AVERROR_EOF);
 +ret = 0;
 +}
  link-frame_requested = 0;
 -if (ret == AVERROR_EOF)
 -link-closed = 1;
 +if (ret == AVERROR_EOF) {
 +ret = ff_filter_link_close(link, AV_NOPTS_VALUE);
 +return ret  0 ? ret : AVERROR_EOF;
 +}
  } else {
  av_assert0(!link-frame_requested ||
 link-flags  FF_LINK_FLAG_REQUEST_LOOP);
 @@ -1132,10 +1133,52 @@ static int ff_filter_frame_needs_framing(AVFilterLink 
 *link, AVFrame *frame)
  return ret;
  }
  
 +static int ff_filter_metaframe(AVFilterLink *link, AVFrame *frame)
 +{
 +AVFrame *pbuf = link-partial_buf;
 +int ret;
 +

 +if (pbuf) {
 +link-partial_buf = NULL;
 +if ((ret = ff_filter_frame_framed(link, pbuf))  0)
 +return ret;
 +}
 +
 +if ((link-dst-filter-flags  FF_FILTER_FLAG_SUPPORT_METAFRAMES)) {
 +ret = link-dstpad-filter_frame ?
 +  link-dstpad-filter_frame(link, frame) :
 +  default_filter_frame(link, frame);
 +if (ret  0)
 +return ret;

It shall probably execute the remaining part of the function even in
case of failure.

 +}
 +
 +switch (frame-format) {
 +
 +case FF_METAFRAME_EOF:
 +link-closed = 1;
 +break;
 +
 +case 0:

 +case FF_METAFRAME_NOP:
 +/* Not implemented yet because not used either for now.
 +   Caveat: if the same metaframe is forwarded to the next filter
 +   and the next filter changes the type, the type change must not be
 +   taken into account for the first link. */

I find this a bit confusing. Can you explain why the NOP metaframe is
needed?

 +
 +default:
 +av_assert0(!reached);
 +}
 +
 +return ret;
 +}
 +
  int ff_filter_frame(AVFilterLink *link, AVFrame *frame)
  {
  FF_TPRINTF_START(NULL, filter_frame); ff_tlog_link(NULL, link, 1); 
 ff_tlog(NULL,  ); ff_tlog_ref(NULL, frame, 1);
  
 +if (frame-format  -1)
 +return ff_filter_metaframe(link, frame);
 +
  /* Consistency checks */
  if (link-type == AVMEDIA_TYPE_VIDEO) {
  if (strcmp(link-dst-filter-name, scale)) {
 @@ -1162,6 +1205,22 @@ int ff_filter_frame(AVFilterLink *link, AVFrame *frame)
  }
  }
  
 +int ff_filter_link_close(AVFilterLink *link, int64_t pts)
 +{
 +AVFrame *frame;
 +int ret;
 +
 +if (link-closed)
 +return 0;
 +if (!(frame = av_frame_alloc()))
 +return AVERROR(ENOMEM);
 +frame-format = FF_METAFRAME_EOF;
 +frame-pts = pts;
 +ret = ff_filter_frame(link, frame);
 +av_frame_free(frame);
 +return ret;
 +}
 +
  const AVClass *avfilter_get_class(void)
  {
  return avfilter_class;
 diff --git a/libavfilter/internal.h b/libavfilter/internal.h
 index 308b115..fbe603a 100644
 --- a/libavfilter/internal.h
 +++ b/libavfilter/internal.h
 @@ -374,4 +374,38 @@ AVFilterContext *ff_filter_alloc(const AVFilter *filter, 
 const char *inst_name);
   */
  void ff_filter_graph_remove_filter(AVFilterGraph *graph, AVFilterContext 
 *filter);
  
 +/**
 + * The filter can accept metaframes.
 + * Metaframes are AVFrame structures with a negative format field.
 + * The framework will take default actions based on the metaframe type.
 + * The destination filter is allowed to reset the type to inhibit the
 + * default actions.
 + */
 +#define

Re: [FFmpeg-devel] HEVC SIMD Optimization on PowerPC

2014-08-02 Thread compn 
On Sat, 2 Aug 2014 14:50:52 +0200
Michael Niedermayer michae...@gmx.at wrote:

 On Sat, Aug 02, 2014 at 07:18:22AM -0400, compn wrote:
  On Sat, 2 Aug 2014 13:37:10 +0800
  Zhenan Lin linzhe...@pku.edu.cn wrote:
  
   Hi, all,
   

   
   Our team plans to contribute SIMD optimization on PowerPC for HEVC
  
  i thought powerpc was dead?
 
 i thought so too, but powerpc optimizations certainly are welcome,
 if someone wants to write some

ubitux linked me to the latest powerpc
https://en.wikipedia.org/wiki/POWER8 

so i guess ppc isnt dead yet.

-compn
___
ffmpeg-devel mailing list
ffmpeg-devel@ffmpeg.org
http://ffmpeg.org/mailman/listinfo/ffmpeg-devel

[FFmpeg-devel] [PATCH 1/5] avutil: add pixelutils API

2014-08-02 Thread Clément Bœsch 
---
 configure   |   2 +
 doc/APIchanges  |   3 ++
 libavutil/Makefile  |   2 +
 libavutil/pixelutils.c  |  85 ++
 libavutil/pixelutils.h  |  76 +++
 libavutil/version.h |   2 +-
 libavutil/x86/Makefile  |   4 ++
 libavutil/x86/pixelutils.asm| 111 
 libavutil/x86/pixelutils.h  |  26 ++
 libavutil/x86/pixelutils_init.c |  54 +++
 10 files changed, 364 insertions(+), 1 deletion(-)
 create mode 100644 libavutil/pixelutils.c
 create mode 100644 libavutil/pixelutils.h
 create mode 100644 libavutil/x86/pixelutils.asm
 create mode 100644 libavutil/x86/pixelutils.h
 create mode 100644 libavutil/x86/pixelutils_init.c

diff --git a/configure b/configure
index 9c3af50..57edd1d 100755
--- a/configure
+++ b/configure
@@ -144,6 +144,7 @@ Component options:
   --disable-mdct   disable MDCT code
   --disable-rdft   disable RDFT code
   --disable-fftdisable FFT code
+  --disable-pixelutils disable pixel utils in libavutil
 
 Hardware accelerators:
   --disable-dxva2  disable DXVA2 code [autodetect]
@@ -1451,6 +1452,7 @@ SUBSYSTEM_LIST=
 lsp
 lzo
 mdct
+pixelutils
 network
 rdft
 
diff --git a/doc/APIchanges b/doc/APIchanges
index abca377..69ca682 100644
--- a/doc/APIchanges
+++ b/doc/APIchanges
@@ -15,6 +15,9 @@ libavutil: 2012-10-22
 
 API changes, most recent first:
 
+2014-08-02 - xxx - lavu 52.95.100 - pixelutils.h
+  Add pixelutils API with SAD functions
+
 2014-07-30 - ba3e331 - lavu 52.94.100 - frame.h
   Add av_frame_side_data_name()
 
diff --git a/libavutil/Makefile b/libavutil/Makefile
index 91751dc..6114cc9 100644
--- a/libavutil/Makefile
+++ b/libavutil/Makefile
@@ -44,6 +44,7 @@ HEADERS = adler32.h   
  \
   opt.h \
   parseutils.h  \
   pixdesc.h \
+  pixelutils.h  \
   pixfmt.h  \
   random_seed.h \
   replaygain.h  \
@@ -113,6 +114,7 @@ OBJS = adler32.o
\
opt.o\
parseutils.o \
pixdesc.o\
+   pixelutils.o \
random_seed.o\
rational.o   \
rc4.o\
diff --git a/libavutil/pixelutils.c b/libavutil/pixelutils.c
new file mode 100644
index 000..146a882
--- /dev/null
+++ b/libavutil/pixelutils.c
@@ -0,0 +1,85 @@
+/*
+ * 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 config.h
+#include common.h
+#include pixelutils.h
+
+#if CONFIG_PIXELUTILS
+
+#include x86/pixelutils.h
+
+static av_always_inline int sad_wxh(const uint8_t *src1, ptrdiff_t stride1,
+const uint8_t *src2, ptrdiff_t stride2,
+int w, int h)
+{
+int x, y, sum = 0;
+
+for (y = 0; y  h; y++) {
+for (x = 0; x  w; x++)
+sum += abs(src1[x] - src2[x]);
+src1 += stride1;
+src2 += stride2;
+}
+return sum;
+}
+
+#define DECLARE_BLOCK_FUNCTIONS(size)  
 \
+static int block_sad_##size##x##size##_c(const uint8_t *src1, ptrdiff_t 
stride1,\
+ const uint8_t *src2, ptrdiff_t 
stride2)\
+{  
 \
+return sad_wxh(src1, stride1, src2, stride2, size,

[FFmpeg-devel] [PATCH 4/5] avfilter/mp_decimate: remove avcodec dependency

2014-08-02 Thread Clément Bœsch 
---
 configure   |  4 ++--
 libavfilter/vf_mpdecimate.c | 44 
 2 files changed, 22 insertions(+), 26 deletions(-)

diff --git a/configure b/configure
index bc187a5..210cafb 100755
--- a/configure
+++ b/configure
@@ -2549,8 +2549,8 @@ ladspa_filter_extralibs='$ldl'
 mcdeint_filter_deps=avcodec gpl
 movie_filter_deps=avcodec avformat
 mp_filter_deps=gpl avcodec swscale inline_asm
-mpdecimate_filter_deps=gpl avcodec
-mpdecimate_filter_select=me_cmp pixblockdsp
+mpdecimate_filter_deps=gpl
+mpdecimate_filter_select=pixelutils
 mptestsrc_filter_deps=gpl
 negate_filter_deps=lut_filter
 perspective_filter_deps=gpl
diff --git a/libavfilter/vf_mpdecimate.c b/libavfilter/vf_mpdecimate.c
index c667a9f..76af175 100644
--- a/libavfilter/vf_mpdecimate.c
+++ b/libavfilter/vf_mpdecimate.c
@@ -26,9 +26,8 @@
 
 #include libavutil/opt.h
 #include libavutil/pixdesc.h
+#include libavutil/pixelutils.h
 #include libavutil/timestamp.h
-#include libavcodec/dsputil.h
-#include libavcodec/pixblockdsp.h
 #include avfilter.h
 #include internal.h
 #include formats.h
@@ -49,9 +48,7 @@ typedef struct {
 
 int hsub, vsub;/// chroma subsampling values
 AVFrame *ref;  /// reference picture
-DSPContext dspctx; /// context providing optimized diff 
routines
-PixblockDSPContext pdsp;
-AVCodecContext *avctx; /// codec context required for the 
DSPContext
+av_pixelutils_sad_fn sad;  /// sum of absolute difference function
 } DecimateContext;
 
 #define OFFSET(x) offsetof(DecimateContext, x)
@@ -72,25 +69,21 @@ AVFILTER_DEFINE_CLASS(mpdecimate);
  * Return 1 if the two planes are different, 0 otherwise.
  */
 static int diff_planes(AVFilterContext *ctx,
-   uint8_t *cur, uint8_t *ref, int linesize,
+   uint8_t *cur, int cur_linesize,
+   uint8_t *ref, int ref_linesize,
int w, int h)
 {
 DecimateContext *decimate = ctx-priv;
-DSPContext *dspctx = decimate-dspctx;
-PixblockDSPContext *pdsp = decimate-pdsp;
 
 int x, y;
 int d, c = 0;
 int t = (w/16)*(h/16)*decimate-frac;
-int16_t block[8*8];
 
 /* compute difference for blocks of 8x8 bytes */
 for (y = 0; y  h-7; y += 4) {
 for (x = 8; x  w-7; x += 4) {
-pdsp-diff_pixels(block,
-cur+x+y*linesize,
-ref+x+y*linesize, linesize);
-d = dspctx-sum_abs_dctelem(block);
+d = decimate-sad(cur + y*cur_linesize + x, cur_linesize,
+  ref + y*ref_linesize + x, ref_linesize);
 if (d  decimate-hi)
 return 1;
 if (d  decimate-lo) {
@@ -124,7 +117,8 @@ static int decimate_frame(AVFilterContext *ctx,
 int vsub = plane == 1 || plane == 2 ? decimate-vsub : 0;
 int hsub = plane == 1 || plane == 2 ? decimate-hsub : 0;
 if (diff_planes(ctx,
-cur-data[plane], ref-data[plane], 
ref-linesize[plane],
+cur-data[plane], cur-linesize[plane],
+ref-data[plane], ref-linesize[plane],
 FF_CEIL_RSHIFT(ref-width,  hsub),
 FF_CEIL_RSHIFT(ref-height, vsub)))
 return 0;
@@ -135,17 +129,23 @@ static int decimate_frame(AVFilterContext *ctx,
 
 static av_cold int init(AVFilterContext *ctx)
 {
+int ret;
 DecimateContext *decimate = ctx-priv;
 
+AVPixelUtils *pixels = av_pixelutils_alloc();
+if (!pixels)
+return AVERROR(ENOMEM);
+ret = av_pixelutils_init(pixels, decimate);
+if (ret  0) {
+av_free(pixels);
+return ret;
+}
+decimate-sad = pixels-sad_u[AV_PIXEL_BLOCKSZ_8x8];
+av_free(pixels);
+
 av_log(ctx, AV_LOG_VERBOSE, max_drop_count:%d hi:%d lo:%d frac:%f\n,
decimate-max_drop_count, decimate-hi, decimate-lo, 
decimate-frac);
 
-decimate-avctx = avcodec_alloc_context3(NULL);
-if (!decimate-avctx)
-return AVERROR(ENOMEM);
-avpriv_dsputil_init(decimate-dspctx, decimate-avctx);
-ff_pixblockdsp_init(decimate-pdsp, decimate-avctx);
-
 return 0;
 }
 
@@ -153,10 +153,6 @@ static av_cold void uninit(AVFilterContext *ctx)
 {
 DecimateContext *decimate = ctx-priv;
 av_frame_free(decimate-ref);
-if (decimate-avctx) {
-avcodec_close(decimate-avctx);
-av_freep(decimate-avctx);
-}
 }
 
 static int query_formats(AVFilterContext *ctx)
-- 
2.0.2

___
ffmpeg-devel mailing list
ffmpeg-devel@ffmpeg.org
http://ffmpeg.org/mailman/listinfo/ffmpeg-devel

[FFmpeg-devel] [PATCH 3/5] avfilter/deshake: remove avcodec dependency

2014-08-02 Thread Clément Bœsch 
---
 configure|  4 +---
 libavfilter/deshake.h|  5 ++---
 libavfilter/vf_deshake.c | 27 +++
 3 files changed, 18 insertions(+), 18 deletions(-)

diff --git a/configure b/configure
index 5159f10..bc187a5 100755
--- a/configure
+++ b/configure
@@ -2531,8 +2531,7 @@ cropdetect_filter_deps=gpl
 dctdnoiz_filter_deps=avcodec
 dctdnoiz_filter_select=dct
 delogo_filter_deps=gpl
-deshake_filter_deps=avcodec
-deshake_filter_select=me_cmp
+deshake_filter_select=pixelutils
 drawtext_filter_deps=libfreetype
 ebur128_filter_deps=gpl
 flite_filter_deps=libflite
@@ -5241,7 +5240,6 @@ enabled amovie_filterprepend avfilter_deps 
avformat avcodec
 enabled aresample_filter prepend avfilter_deps swresample
 enabled asyncts_filter   prepend avfilter_deps avresample
 enabled atempo_filterprepend avfilter_deps avcodec
-enabled deshake_filter   prepend avfilter_deps avcodec
 enabled ebur128_filter  enabled swresample  prepend avfilter_deps 
swresample
 enabled elbg_filter  prepend avfilter_deps avcodec
 enabled mcdeint_filter   prepend avfilter_deps avcodec
diff --git a/libavfilter/deshake.h b/libavfilter/deshake.h
index 615953c..f61ed80 100644
--- a/libavfilter/deshake.h
+++ b/libavfilter/deshake.h
@@ -24,8 +24,8 @@
 
 #include config.h
 #include avfilter.h
-#include libavcodec/dsputil.h
 #include transform.h
+#include libavutil/pixelutils.h
 #if CONFIG_OPENCL
 #include libavutil/opencl.h
 #endif
@@ -80,8 +80,7 @@ typedef struct {
 int blocksize; /// Size of blocks to compare
 int contrast;  /// Contrast threshold
 int search;/// Motion search method
-AVCodecContext *avctx;
-DSPContext c;  /// Context providing optimized SAD methods
+av_pixelutils_sad_fn sad;  /// Sum of the absolute difference function
 Transform last;/// Transform from last frame
 int refcount;  /// Number of reference frames (defines 
averaging window)
 FILE *fp;
diff --git a/libavfilter/vf_deshake.c b/libavfilter/vf_deshake.c
index 50aa451..cd32ecf 100644
--- a/libavfilter/vf_deshake.c
+++ b/libavfilter/vf_deshake.c
@@ -57,7 +57,6 @@
 #include libavutil/mem.h
 #include libavutil/opt.h
 #include libavutil/pixdesc.h
-#include libavcodec/dsputil.h
 
 #include deshake.h
 #include deshake_opencl.h
@@ -132,9 +131,8 @@ static void find_block_motion(DeshakeContext *deshake, 
uint8_t *src1,
 int smallest = INT_MAX;
 int tmp, tmp2;
 
-#define CMP(i, j) deshake-c.sad[0](NULL, src1 + cy * stride + cx, \
-src2 + (j) * stride + (i), stride, \
-deshake-blocksize)
+#define CMP(i, j) deshake-sad(src1 + cy  * stride + cx,  stride,\
+   src2 + (j) * stride + (i), stride)
 
 if (deshake-search == EXHAUSTIVE) {
 // Compare every possible position - this is slw!
@@ -201,7 +199,7 @@ static int block_contrast(uint8_t *src, int x, int y, int 
stride, int blocksize)
 int i, j, pos;
 
 for (i = 0; i = blocksize * 2; i++) {
-// We use a width of 16 here to match the libavcodec sad functions
+// We use a width of 16 here to match the sad function
 for (j = 0; j = 15; j++) {
 pos = (y - i) * stride + (x - j);
 if (src[pos]  lowest)
@@ -263,7 +261,7 @@ static void find_motion(DeshakeContext *deshake, uint8_t 
*src1, uint8_t *src2,
 pos = 0;
 // Find motion for every block and store the motion vector in the counts
 for (y = deshake-ry; y  height - deshake-ry - (deshake-blocksize * 2); 
y += deshake-blocksize * 2) {
-// We use a width of 16 here to match the libavcodec sad functions
+// We use a width of 16 here to match the sad function
 for (x = deshake-rx; x  width - deshake-rx - 16; x += 16) {
 // If the contrast is too low, just skip this block as it probably
 // won't be very useful to us.
@@ -351,6 +349,17 @@ static av_cold int init(AVFilterContext *ctx)
 int ret;
 DeshakeContext *deshake = ctx-priv;
 
+AVPixelUtils *pixels = av_pixelutils_alloc();
+if (!pixels)
+return AVERROR(ENOMEM);
+ret = av_pixelutils_init(pixels, deshake);
+if (ret  0) {
+av_free(pixels);
+return ret;
+}
+deshake-sad = pixels-sad_u[AV_PIXEL_BLOCKSZ_16x16]; // XXX: configurable?
+av_free(pixels);
+
 deshake-refcount = 20; // XXX: add to options?
 deshake-blocksize /= 2;
 deshake-blocksize = av_clip(deshake-blocksize, 4, 128);
@@ -413,9 +422,6 @@ static int config_props(AVFilterLink *link)
 deshake-last.angle = 0;
 deshake-last.zoom = 0;
 
-deshake-avctx = avcodec_alloc_context3(NULL);
-avpriv_dsputil_init(deshake-c, deshake-avctx);
-
 return 0;
 }
 
@@ -428,9 +434,6 @@ static av_cold void uninit(AVFilterContext *ctx)
 av_frame_free(deshake-ref);
 if (deshake-fp)

Re: [FFmpeg-devel] [PATCH 1/5] avutil: add pixelutils API

2014-08-02 Thread Lukasz Marek 

+static av_always_inline int sad_wxh(const uint8_t *src1, ptrdiff_t stride1,
+const uint8_t *src2, ptrdiff_t stride2,
+int w, int h)
+{
+int x, y, sum = 0;
+
+for (y = 0; y  h; y++) {
+for (x = 0; x  w; x++)
+sum += abs(src1[x] - src2[x]);


minor: I came out of nowhere to this patch, but maybe FFABS? :)


+src1 += stride1;
+src2 += stride2;
+}
+return sum;
+}


___
ffmpeg-devel mailing list
ffmpeg-devel@ffmpeg.org
http://ffmpeg.org/mailman/listinfo/ffmpeg-devel

Re: [FFmpeg-devel] [PATCH 2/5] build: remove wrong avcodec dep in decimate

2014-08-02 Thread Paul B Mahol 
On Sat, Aug 2, 2014 at 8:20 PM, Clément Bœsch u...@pkh.me wrote:

 This was probably a mistake with mpdecimate. The decimate filter doesn't
 depend on avcodec.
 ---
  configure | 1 -
  1 file changed, 1 deletion(-)

lgtm
___
ffmpeg-devel mailing list
ffmpeg-devel@ffmpeg.org
http://ffmpeg.org/mailman/listinfo/ffmpeg-devel

Re: [FFmpeg-devel] [PATCH 1/5] avutil: add pixelutils API

2014-08-02 Thread Michael Niedermayer 
On Sat, Aug 02, 2014 at 08:20:17PM +0200, Clément Bœsch wrote:
 ---
  configure   |   2 +
  doc/APIchanges  |   3 ++
  libavutil/Makefile  |   2 +
  libavutil/pixelutils.c  |  85 ++
  libavutil/pixelutils.h  |  76 +++
  libavutil/version.h |   2 +-
  libavutil/x86/Makefile  |   4 ++
  libavutil/x86/pixelutils.asm| 111 
 
  libavutil/x86/pixelutils.h  |  26 ++
  libavutil/x86/pixelutils_init.c |  54 +++
  10 files changed, 364 insertions(+), 1 deletion(-)
  create mode 100644 libavutil/pixelutils.c
  create mode 100644 libavutil/pixelutils.h
  create mode 100644 libavutil/x86/pixelutils.asm
  create mode 100644 libavutil/x86/pixelutils.h
  create mode 100644 libavutil/x86/pixelutils_init.c
[...]
 diff --git a/libavutil/pixelutils.h b/libavutil/pixelutils.h
 new file mode 100644
 index 000..862d607
 --- /dev/null
 +++ b/libavutil/pixelutils.h
 @@ -0,0 +1,76 @@
 +/*
 + * 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_PIXELUTILS_H
 +#define AVUTIL_PIXELUTILS_H
 +
 +#include stddef.h
 +#include stdint.h
 +#include common.h
 +
 +enum AVPixelUtilsBlockSize {
 +// Warning: only add entries at the end for ABI compatibility.
 +// Ideally, keep only squared dimensions in that enum.
 +AV_PIXEL_BLOCKSZ_2x2,
 +AV_PIXEL_BLOCKSZ_4x4,
 +AV_PIXEL_BLOCKSZ_8x8,
 +AV_PIXEL_BLOCKSZ_16x16,
 +AV_PIXEL_BLOCKSZ_NB
 +};
 +
 +/**
 + * Sum of abs(src1[x] - src2[x])
 + */
 +typedef int (*av_pixelutils_sad_fn)(const uint8_t *src1, ptrdiff_t stride1,
 +const uint8_t *src2, ptrdiff_t stride2);
 +
 +typedef struct AVPixelUtils {
 +/**
 + * Sum of absolute differences block functions.
 + * src1 and src2 addresses need to be aligned to the block size.
 + */
 +av_pixelutils_sad_fn sad[AV_PIXEL_BLOCKSZ_NB];
 +
 +/**
 + * Sum of absolute differences block functions, (half) unaligned version.
 + * src1 address needs to be aligned to the block size.
 + * src2 has no alignment requirement; the 'u' refers to this second 
 source.
 + */
 +av_pixelutils_sad_fn sad_u[AV_PIXEL_BLOCKSZ_NB];

the ABI depends on AV_PIXEL_BLOCKSZ_NB with this design as the arrays
in the struct depend on its value so no enum values could be added

[...]
-- 
Michael GnuPG fingerprint: 9FF2128B147EF6730BADF133611EC787040B0FAB

Complexity theory is the science of finding the exact solution to an
approximation. Benchmarking OTOH is finding an approximation of the exact


signature.asc
Description: Digital signature
___
ffmpeg-devel mailing list
ffmpeg-devel@ffmpeg.org
http://ffmpeg.org/mailman/listinfo/ffmpeg-devel

Re: [FFmpeg-devel] [PATCH 1/5] avutil: add pixelutils API

2014-08-02 Thread Clément Bœsch 
On Sat, Aug 02, 2014 at 08:30:39PM +0200, Lukasz Marek wrote:
 +static av_always_inline int sad_wxh(const uint8_t *src1, ptrdiff_t stride1,
 +const uint8_t *src2, ptrdiff_t stride2,
 +int w, int h)
 +{
 +int x, y, sum = 0;
 +
 +for (y = 0; y  h; y++) {
 +for (x = 0; x  w; x++)
 +sum += abs(src1[x] - src2[x]);
 
 minor: I came out of nowhere to this patch, but maybe FFABS? :)
 

No difference in generated ASM here

-- 
Clément B.


pgpFDI_gArcbX.pgp
Description: PGP signature
___
ffmpeg-devel mailing list
ffmpeg-devel@ffmpeg.org
http://ffmpeg.org/mailman/listinfo/ffmpeg-devel

[FFmpeg-devel] [PATCH] avcodec/avdct: Add get_pixels()

2014-08-02 Thread Michael Niedermayer 
TODO: version bump, update libavfilters to use it

Suggested-by: ubitux
Signed-off-by: Michael Niedermayer michae...@gmx.at
---
 libavcodec/avdct.c |9 +
 libavcodec/avdct.h |4 
 2 files changed, 13 insertions(+)

diff --git a/libavcodec/avdct.c b/libavcodec/avdct.c
index 58f4974..2521256 100644
--- a/libavcodec/avdct.c
+++ b/libavcodec/avdct.c
@@ -21,6 +21,7 @@
 #include avcodec.h
 #include idctdsp.h
 #include fdctdsp.h
+#include pixblockdsp.h
 #include avdct.h
 
 #define OFFSET(x) offsetof(AVDCT,x)
@@ -117,6 +118,14 @@ int avcodec_dct_init(AVDCT *dsp)
 }
 #endif
 
+#if CONFIG_PIXBLOCKDSP
+{
+PixblockDSPContext pdsp;
+ff_pixblockdsp_init(pdsp, avctx);
+COPY(pdsp, get_pixels);
+}
+#endif
+
 avcodec_close(avctx);
 av_free(avctx);
 
diff --git a/libavcodec/avdct.h b/libavcodec/avdct.h
index 4c9d00f..4190203 100644
--- a/libavcodec/avdct.h
+++ b/libavcodec/avdct.h
@@ -61,6 +61,10 @@ typedef struct AVDCT {
  * must use AVOptions to set this field.
  */
 int idct_algo;
+
+void (*get_pixels)(int16_t *block /* align 16 */,
+   const uint8_t *pixels /* align 8 */,
+   int line_size);
 } AVDCT;
 
 /**
-- 
1.7.9.5

___
ffmpeg-devel mailing list
ffmpeg-devel@ffmpeg.org
http://ffmpeg.org/mailman/listinfo/ffmpeg-devel

Re: [FFmpeg-devel] [PATCH 1/5] avutil: add pixelutils API

2014-08-02 Thread James Almer 
On 02/08/14 3:20 PM, Clément Bœsch wrote:
 +psrlq   m0, m6, 32
 +paddw   m6, m0
 +psrlq   m0, m6, 16
 +paddw   m6, m0
 +movdeax, m6
 +movzx   eax, ax

You could use the HADDW macro here.

 +;---
 +; int ff_pixelutils_sad_8x8_mmxext(const uint8_t *src1, ptrdiff_t stride1,
 +;  const uint8_t *src2, ptrdiff_t stride2);
 +;---
 +INIT_MMX mmxext
 +cglobal pixelutils_sad_8x8, 4,4,0, src1, stride1, src2, stride2
 +pxorm2, m2
 +%rep 4
 +movam0, [src1q]
 +movam1, [src1q + stride1q]
 +psadbw  m0, [src2q]
 +psadbw  m1, [src2q + stride2q]
 +paddw   m2, m0
 +paddw   m2, m1
 +lea src1q, [src1q + 2*stride1q]
 +lea src2q, [src2q + 2*stride2q]
 +%endrep
 +movdeax, m2
 +RET

Adding sad16x16 mmxext should be a matter of using add instead of lea, changing 
the %rep amount, and using 8 instead of stride[12]q for the mova and psadbw.

 --- /dev/null
 +++ b/libavutil/x86/pixelutils.h
 @@ -0,0 +1,26 @@
 +/*
 + * 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_X86_PIXELUTILS_H
 +#define AVUTIL_X86_PIXELUTILS_H
 +
 +#include libavutil/pixelutils.h
 +
 +void ff_pixelutils_init_x86(AVPixelUtils *s);

This prototype should be in libavutil/pixelutils.h
No need to make a whole new header just for it.

Maybe you could add a quick test for these functions? Look at 
lavc/motion-test.c and 
lavu/float-dsp.c
___
ffmpeg-devel mailing list
ffmpeg-devel@ffmpeg.org
http://ffmpeg.org/mailman/listinfo/ffmpeg-devel

[FFmpeg-devel] [PATCH 2/2] avcodec/takdec: remove unused variable

2014-08-02 Thread Michael Niedermayer 
Found-by: CSA
Signed-off-by: Michael Niedermayer michae...@gmx.at
---
 libavcodec/takdec.c |1 -
 1 file changed, 1 deletion(-)

diff --git a/libavcodec/takdec.c b/libavcodec/takdec.c
index 83abab6..5810a01 100644
--- a/libavcodec/takdec.c
+++ b/libavcodec/takdec.c
@@ -448,7 +448,6 @@ static int decode_subframe(TAKDecContext *s, int32_t 
*decoded,
 x = 1  (32 - (15 - filter_quant));
 y = 1  ((15 - filter_quant) - 1);
 for (i = 0, j = filter_order - 1; i  filter_order / 2; i++, j--) {
-int tmp = y + tfilter[j];
 s-filter[j] = x - ((tfilter[i] + y)  (15 - filter_quant));
 s-filter[i] = x - ((tfilter[j] + y)  (15 - filter_quant));
 }
-- 
1.7.9.5

___
ffmpeg-devel mailing list
ffmpeg-devel@ffmpeg.org
http://ffmpeg.org/mailman/listinfo/ffmpeg-devel

[FFmpeg-devel] [PATCH] avformat/tee: flip assigment direction

2014-08-02 Thread Michael Niedermayer 
Found-by: CSA
Signed-off-by: Michael Niedermayer michae...@gmx.at
---
 libavformat/tee.c |2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/libavformat/tee.c b/libavformat/tee.c
index f26e157..681f943 100644
--- a/libavformat/tee.c
+++ b/libavformat/tee.c
@@ -467,7 +467,7 @@ static int tee_write_packet(AVFormatContext *avf, AVPacket 
*pkt)
 if ((ret = av_copy_packet(pkt2, pkt))  0 ||
 (ret = av_dup_packet(pkt2)) 0)
 if (!ret_all) {
-ret = ret_all;
+ret_all = ret;
 continue;
 }
 tb  = avf -streams[s ]-time_base;
-- 
1.7.9.5

___
ffmpeg-devel mailing list
ffmpeg-devel@ffmpeg.org
http://ffmpeg.org/mailman/listinfo/ffmpeg-devel

Re: [FFmpeg-devel] [PATCH 1/5] avutil: add pixelutils API

2014-08-02 Thread Clément Bœsch 
On Sat, Aug 02, 2014 at 04:29:39PM -0300, James Almer wrote:
 On 02/08/14 3:20 PM, Clément Bœsch wrote:
  +psrlq   m0, m6, 32
  +paddw   m6, m0
  +psrlq   m0, m6, 16
  +paddw   m6, m0
  +movdeax, m6
  +movzx   eax, ax
 
 You could use the HADDW macro here.
 

error: undefined symbol `pw_1' (first use)

sounds somehow constraining. I'll keep my version until you benchmark to
prove me HADDW is faster on an old MMX cpu ;)

  +;---
  +; int ff_pixelutils_sad_8x8_mmxext(const uint8_t *src1, ptrdiff_t stride1,
  +;  const uint8_t *src2, ptrdiff_t stride2);
  +;---
  +INIT_MMX mmxext
  +cglobal pixelutils_sad_8x8, 4,4,0, src1, stride1, src2, stride2
  +pxorm2, m2
  +%rep 4
  +movam0, [src1q]
  +movam1, [src1q + stride1q]
  +psadbw  m0, [src2q]
  +psadbw  m1, [src2q + stride2q]
  +paddw   m2, m0
  +paddw   m2, m1
  +lea src1q, [src1q + 2*stride1q]
  +lea src2q, [src2q + 2*stride2q]
  +%endrep
  +movdeax, m2
  +RET
 
 Adding sad16x16 mmxext should be a matter of using add instead of lea, 
 changing 
 the %rep amount, and using 8 instead of stride[12]q for the mova and psadbw.
 

Yeah right, added. Thanks.

  --- /dev/null
  +++ b/libavutil/x86/pixelutils.h
  @@ -0,0 +1,26 @@
  +/*
  + * 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_X86_PIXELUTILS_H
  +#define AVUTIL_X86_PIXELUTILS_H
  +
  +#include libavutil/pixelutils.h
  +
  +void ff_pixelutils_init_x86(AVPixelUtils *s);
 
 This prototype should be in libavutil/pixelutils.h
 No need to make a whole new header just for it.
 

No, libavutil/pixelutils.h is public, I don't want to have private
prototypes in it.

 Maybe you could add a quick test for these functions? Look at 
 lavc/motion-test.c and 
 lavu/float-dsp.c

Added.

I'll resubmit a patchset in a moment.

-- 
Clément B.


pgp450g7nvNHd.pgp
Description: PGP signature
___
ffmpeg-devel mailing list
ffmpeg-devel@ffmpeg.org
http://ffmpeg.org/mailman/listinfo/ffmpeg-devel

Re: [FFmpeg-devel] [PATCH 1/5] avutil: add pixelutils API

2014-08-02 Thread James Almer 
On 02/08/14 6:13 PM, Clément Bœsch wrote:
 On Sat, Aug 02, 2014 at 04:29:39PM -0300, James Almer wrote:
 On 02/08/14 3:20 PM, Clément Bœsch wrote:
 +psrlq   m0, m6, 32
 +paddw   m6, m0
 +psrlq   m0, m6, 16
 +paddw   m6, m0
 +movdeax, m6
 +movzx   eax, ax

 You could use the HADDW macro here.

 
 error: undefined symbol `pw_1' (first use)
 
 sounds somehow constraining. I'll keep my version until you benchmark to
 prove me HADDW is faster on an old MMX cpu ;)

I have no idea if it's faster, nor a way to test that for that matter.
It's four instructions instead of six, but pmaddwd + memory operand is probably 
not fast enough on old cpus.

 
 +;---
 +; int ff_pixelutils_sad_8x8_mmxext(const uint8_t *src1, ptrdiff_t stride1,
 +;  const uint8_t *src2, ptrdiff_t stride2);
 +;---
 +INIT_MMX mmxext
 +cglobal pixelutils_sad_8x8, 4,4,0, src1, stride1, src2, stride2
 +pxorm2, m2
 +%rep 4
 +movam0, [src1q]
 +movam1, [src1q + stride1q]
 +psadbw  m0, [src2q]
 +psadbw  m1, [src2q + stride2q]
 +paddw   m2, m0
 +paddw   m2, m1
 +lea src1q, [src1q + 2*stride1q]
 +lea src2q, [src2q + 2*stride2q]
 +%endrep
 +movdeax, m2
 +RET

 Adding sad16x16 mmxext should be a matter of using add instead of lea, 
 changing 
 the %rep amount, and using 8 instead of stride[12]q for the mova and psadbw.

 
 Yeah right, added. Thanks.
 
 --- /dev/null
 +++ b/libavutil/x86/pixelutils.h
 @@ -0,0 +1,26 @@
 +/*
 + * 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_X86_PIXELUTILS_H
 +#define AVUTIL_X86_PIXELUTILS_H
 +
 +#include libavutil/pixelutils.h
 +
 +void ff_pixelutils_init_x86(AVPixelUtils *s);

 This prototype should be in libavutil/pixelutils.h
 No need to make a whole new header just for it.

 
 No, libavutil/pixelutils.h is public, I don't want to have private
 prototypes in it.

Right, forgot it was public. I had lavc dsp stuff in mind when i said that.

 
 Maybe you could add a quick test for these functions? Look at 
 lavc/motion-test.c and 
 lavu/float-dsp.c
 
 Added.
 
 I'll resubmit a patchset in a moment.
 
 
 
 ___
 ffmpeg-devel mailing list
 ffmpeg-devel@ffmpeg.org
 http://ffmpeg.org/mailman/listinfo/ffmpeg-devel
 

___
ffmpeg-devel mailing list
ffmpeg-devel@ffmpeg.org
http://ffmpeg.org/mailman/listinfo/ffmpeg-devel

Re: [FFmpeg-devel] [PATCH] avformat/tee: flip assigment direction

2014-08-02 Thread Nicolas George 
Le quintidi 15 thermidor, an CCXXII, Michael Niedermayer a écrit :
 Found-by: CSA
 Signed-off-by: Michael Niedermayer michae...@gmx.at
 ---
  libavformat/tee.c |2 +-
  1 file changed, 1 insertion(+), 1 deletion(-)

That looks clearly right. Thanks.

Regards,

-- 
  Nicolas George


signature.asc
Description: Digital signature
___
ffmpeg-devel mailing list
ffmpeg-devel@ffmpeg.org
http://ffmpeg.org/mailman/listinfo/ffmpeg-devel

[FFmpeg-devel] [PATCH 2/5] build: remove wrong avcodec dep in decimate

2014-08-02 Thread Clément Bœsch 
This was probably a mistake with mpdecimate. The decimate filter doesn't
depend on avcodec.
---
 configure | 1 -
 1 file changed, 1 deletion(-)

diff --git a/configure b/configure
index 57edd1d..5159f10 100755
--- a/configure
+++ b/configure
@@ -5241,7 +5241,6 @@ enabled amovie_filterprepend avfilter_deps 
avformat avcodec
 enabled aresample_filter prepend avfilter_deps swresample
 enabled asyncts_filter   prepend avfilter_deps avresample
 enabled atempo_filterprepend avfilter_deps avcodec
-enabled decimate_filter  prepend avfilter_deps avcodec
 enabled deshake_filter   prepend avfilter_deps avcodec
 enabled ebur128_filter  enabled swresample  prepend avfilter_deps 
swresample
 enabled elbg_filter  prepend avfilter_deps avcodec
-- 
2.0.2

___
ffmpeg-devel mailing list
ffmpeg-devel@ffmpeg.org
http://ffmpeg.org/mailman/listinfo/ffmpeg-devel

[FFmpeg-devel] [PATCH 3/5] avfilter/deshake: remove avcodec dependency

2014-08-02 Thread Clément Bœsch 
---
 configure|  4 +---
 libavfilter/deshake.h|  5 ++---
 libavfilter/vf_deshake.c | 20 
 3 files changed, 11 insertions(+), 18 deletions(-)

diff --git a/configure b/configure
index 5159f10..bc187a5 100755
--- a/configure
+++ b/configure
@@ -2531,8 +2531,7 @@ cropdetect_filter_deps=gpl
 dctdnoiz_filter_deps=avcodec
 dctdnoiz_filter_select=dct
 delogo_filter_deps=gpl
-deshake_filter_deps=avcodec
-deshake_filter_select=me_cmp
+deshake_filter_select=pixelutils
 drawtext_filter_deps=libfreetype
 ebur128_filter_deps=gpl
 flite_filter_deps=libflite
@@ -5241,7 +5240,6 @@ enabled amovie_filterprepend avfilter_deps 
avformat avcodec
 enabled aresample_filter prepend avfilter_deps swresample
 enabled asyncts_filter   prepend avfilter_deps avresample
 enabled atempo_filterprepend avfilter_deps avcodec
-enabled deshake_filter   prepend avfilter_deps avcodec
 enabled ebur128_filter  enabled swresample  prepend avfilter_deps 
swresample
 enabled elbg_filter  prepend avfilter_deps avcodec
 enabled mcdeint_filter   prepend avfilter_deps avcodec
diff --git a/libavfilter/deshake.h b/libavfilter/deshake.h
index 615953c..f61ed80 100644
--- a/libavfilter/deshake.h
+++ b/libavfilter/deshake.h
@@ -24,8 +24,8 @@
 
 #include config.h
 #include avfilter.h
-#include libavcodec/dsputil.h
 #include transform.h
+#include libavutil/pixelutils.h
 #if CONFIG_OPENCL
 #include libavutil/opencl.h
 #endif
@@ -80,8 +80,7 @@ typedef struct {
 int blocksize; /// Size of blocks to compare
 int contrast;  /// Contrast threshold
 int search;/// Motion search method
-AVCodecContext *avctx;
-DSPContext c;  /// Context providing optimized SAD methods
+av_pixelutils_sad_fn sad;  /// Sum of the absolute difference function
 Transform last;/// Transform from last frame
 int refcount;  /// Number of reference frames (defines 
averaging window)
 FILE *fp;
diff --git a/libavfilter/vf_deshake.c b/libavfilter/vf_deshake.c
index 50aa451..95a6c51 100644
--- a/libavfilter/vf_deshake.c
+++ b/libavfilter/vf_deshake.c
@@ -57,7 +57,6 @@
 #include libavutil/mem.h
 #include libavutil/opt.h
 #include libavutil/pixdesc.h
-#include libavcodec/dsputil.h
 
 #include deshake.h
 #include deshake_opencl.h
@@ -132,9 +131,8 @@ static void find_block_motion(DeshakeContext *deshake, 
uint8_t *src1,
 int smallest = INT_MAX;
 int tmp, tmp2;
 
-#define CMP(i, j) deshake-c.sad[0](NULL, src1 + cy * stride + cx, \
-src2 + (j) * stride + (i), stride, \
-deshake-blocksize)
+#define CMP(i, j) deshake-sad(src1 + cy  * stride + cx,  stride,\
+   src2 + (j) * stride + (i), stride)
 
 if (deshake-search == EXHAUSTIVE) {
 // Compare every possible position - this is slw!
@@ -201,7 +199,7 @@ static int block_contrast(uint8_t *src, int x, int y, int 
stride, int blocksize)
 int i, j, pos;
 
 for (i = 0; i = blocksize * 2; i++) {
-// We use a width of 16 here to match the libavcodec sad functions
+// We use a width of 16 here to match the sad function
 for (j = 0; j = 15; j++) {
 pos = (y - i) * stride + (x - j);
 if (src[pos]  lowest)
@@ -263,7 +261,7 @@ static void find_motion(DeshakeContext *deshake, uint8_t 
*src1, uint8_t *src2,
 pos = 0;
 // Find motion for every block and store the motion vector in the counts
 for (y = deshake-ry; y  height - deshake-ry - (deshake-blocksize * 2); 
y += deshake-blocksize * 2) {
-// We use a width of 16 here to match the libavcodec sad functions
+// We use a width of 16 here to match the sad function
 for (x = deshake-rx; x  width - deshake-rx - 16; x += 16) {
 // If the contrast is too low, just skip this block as it probably
 // won't be very useful to us.
@@ -351,6 +349,10 @@ static av_cold int init(AVFilterContext *ctx)
 int ret;
 DeshakeContext *deshake = ctx-priv;
 
+deshake-sad = av_pixelutils_get_sad_fn(4, 4, 1, deshake); // 16x16, 2nd 
source unaligned
+if (!deshake-sad)
+return AVERROR(EINVAL);
+
 deshake-refcount = 20; // XXX: add to options?
 deshake-blocksize /= 2;
 deshake-blocksize = av_clip(deshake-blocksize, 4, 128);
@@ -413,9 +415,6 @@ static int config_props(AVFilterLink *link)
 deshake-last.angle = 0;
 deshake-last.zoom = 0;
 
-deshake-avctx = avcodec_alloc_context3(NULL);
-avpriv_dsputil_init(deshake-c, deshake-avctx);
-
 return 0;
 }
 
@@ -428,9 +427,6 @@ static av_cold void uninit(AVFilterContext *ctx)
 av_frame_free(deshake-ref);
 if (deshake-fp)
 fclose(deshake-fp);
-if (deshake-avctx)
-avcodec_close(deshake-avctx);
-av_freep(deshake-avctx);
 }
 
 static int filter_frame(AVFilterLink *link, AVFrame

[FFmpeg-devel] pixelutils API [v2]

2014-08-02 Thread Clément Bœsch 
Second attempt.

Since previously:
 - much simpler usage (you basically just call av_pixelutils_get_sad_fn() as a
   user); filters patches updated accordingly.
 - 16x16 mmxext added (thanks jamrial)
 - FATE tests

Note for the context since I didn't write anything about it in the previous
patchset: this is in continuation to avcodec/AVDCT in order to drop dsputils
which has been removed from Libav (but kept in FFmpeg for compatibility). After
this patchset, libavfilter should be free of dsputils leftovers.

BTW, the scope of the pixelutils is to have a bunch of optimized pixel
utilities for filters. Currently, we only have SAD utils, more functions are
welcome as long as it sticks to this very generic pixels scope: for example,
DCT stuff belongs to avcodec/AVDCT (at least for now), and an optimized pixel
function that won't be useful for multiple filters belongs in the local context
of that filter. A good candidate could be SAD for full-frames, or maybe pixel
lines functions for interlaced stuff.

pixelutils are optional. It will be enabled in libavutil only on demand
(typically if a filter uses it). The default full build will contain it, but if
you --disable-filters, libavutil will no longer integrate it.

Ah, and before I forget; we are kind of in a hurry since this is meant to avoid
some ABI issues in the next release. So if you want to review it, please do it
quickly.

Thanks,

___
ffmpeg-devel mailing list
ffmpeg-devel@ffmpeg.org
http://ffmpeg.org/mailman/listinfo/ffmpeg-devel

[FFmpeg-devel] [PATCH 5/5] avfilter/select: remove optional avcodec dependency

2014-08-02 Thread Clément Bœsch 
---
 configure  |  1 +
 libavfilter/f_select.c | 50 --
 2 files changed, 13 insertions(+), 38 deletions(-)

diff --git a/configure b/configure
index 210cafb..b9b1e3b 100755
--- a/configure
+++ b/configure
@@ -2564,6 +2564,7 @@ removelogo_filter_deps=avcodec avformat swscale
 resample_filter_deps=avresample
 sab_filter_deps=gpl swscale
 scale_filter_deps=swscale
+select_filter_select=pixelutils
 smartblur_filter_deps=gpl swscale
 showspectrum_filter_deps=avcodec
 showspectrum_filter_select=rdft
diff --git a/libavfilter/f_select.c b/libavfilter/f_select.c
index 8d2c860..d951b0f 100644
--- a/libavfilter/f_select.c
+++ b/libavfilter/f_select.c
@@ -28,16 +28,13 @@
 #include libavutil/fifo.h
 #include libavutil/internal.h
 #include libavutil/opt.h
+#include libavutil/pixelutils.h
 #include avfilter.h
 #include audio.h
 #include formats.h
 #include internal.h
 #include video.h
 
-#if CONFIG_AVCODEC
-#include libavcodec/dsputil.h
-#endif
-
 static const char *const var_names[] = {
 TB,/// timebase
 
@@ -144,11 +141,8 @@ typedef struct SelectContext {
 AVExpr *expr;
 double var_values[VAR_VARS_NB];
 int do_scene_detect;/// 1 if the expression requires scene 
detection variables, 0 otherwise
-#if CONFIG_AVCODEC
-AVCodecContext *avctx;  /// codec context required for the 
DSPContext (scene detect only)
-DSPContext c;   /// context providing optimized SAD 
methods   (scene detect only)
+av_pixelutils_sad_fn sad;   /// Sum of the absolute difference 
function (scene detect only)
 double prev_mafd;   /// previous MAFD 
(scene detect only)
-#endif
 AVFrame *prev_picref; /// previous frame
(scene detect only)
 double select;
 int select_out; /// mark the selected output pad index
@@ -240,18 +234,14 @@ static int config_input(AVFilterLink *inlink)
 select-var_values[VAR_SAMPLE_RATE] =
 inlink-type == AVMEDIA_TYPE_AUDIO ? inlink-sample_rate : NAN;
 
-#if CONFIG_AVCODEC
 if (select-do_scene_detect) {
-select-avctx = avcodec_alloc_context3(NULL);
-if (!select-avctx)
-return AVERROR(ENOMEM);
-avpriv_dsputil_init(select-c, select-avctx);
+select-sad = av_pixelutils_get_sad_fn(3, 3, 2, select); // 8x8 both 
sources aligned
+if (!select-sad)
+return AVERROR(EINVAL);
 }
-#endif
 return 0;
 }
 
-#if CONFIG_AVCODEC
 static double get_scene_score(AVFilterContext *ctx, AVFrame *frame)
 {
 double ret = 0;
@@ -259,24 +249,23 @@ static double get_scene_score(AVFilterContext *ctx, 
AVFrame *frame)
 AVFrame *prev_picref = select-prev_picref;
 
 if (prev_picref 
-frame-height== prev_picref-height 
-frame-width== prev_picref-width 
-frame-linesize[0] == prev_picref-linesize[0]) {
+frame-height == prev_picref-height 
+frame-width  == prev_picref-width) {
 int x, y, nb_sad = 0;
 int64_t sad = 0;
 double mafd, diff;
 uint8_t *p1 =  frame-data[0];
 uint8_t *p2 = prev_picref-data[0];
-const int linesize = frame-linesize[0];
+const int p1_linesize =   frame-linesize[0];
+const int p2_linesize = prev_picref-linesize[0];
 
 for (y = 0; y  frame-height - 8; y += 8) {
 for (x = 0; x  frame-width*3 - 8; x += 8) {
-sad += select-c.sad[1](NULL, p1 + x, p2 + x,
-linesize, 8);
+sad += select-sad(p1 + x, p1_linesize, p2 + x, p2_linesize);
 nb_sad += 8 * 8;
 }
-p1 += 8 * linesize;
-p2 += 8 * linesize;
+p1 += 8 * p1_linesize;
+p2 += 8 * p2_linesize;
 }
 emms_c();
 mafd = nb_sad ? (double)sad / nb_sad : 0;
@@ -288,7 +277,6 @@ static double get_scene_score(AVFilterContext *ctx, AVFrame 
*frame)
 select-prev_picref = av_frame_clone(frame);
 return ret;
 }
-#endif
 
 #define D2TS(d)  (isnan(d) ? AV_NOPTS_VALUE : (int64_t)(d))
 #define TS2D(ts) ((ts) == AV_NOPTS_VALUE ? NAN : (double)(ts))
@@ -320,7 +308,6 @@ static void select_frame(AVFilterContext *ctx, AVFrame 
*frame)
 !frame-interlaced_frame ? INTERLACE_TYPE_P :
 frame-top_field_first ? INTERLACE_TYPE_T : INTERLACE_TYPE_B;
 select-var_values[VAR_PICT_TYPE] = frame-pict_type;
-#if CONFIG_AVCODEC
 if (select-do_scene_detect) {
 char buf[32];
 select-var_values[VAR_SCENE] = get_scene_score(ctx, frame);
@@ -328,7 +315,6 @@ static void select_frame(AVFilterContext *ctx, AVFrame 
*frame)
 snprintf(buf, sizeof(buf), %f, select-var_values[VAR_SCENE]);
 av_dict_set(avpriv_frame_get_metadatap(frame), 
lavfi.scene_score, buf, 0);
 }
-#endif
 break;
 }
 
@@ -418,15

[FFmpeg-devel] [PATCH 1/5] avutil: add pixelutils API

2014-08-02 Thread Clément Bœsch 
---
 configure   |   2 +
 doc/APIchanges  |   3 +
 libavutil/Makefile  |   3 +
 libavutil/pixelutils.c  | 142 
 libavutil/pixelutils.h  |  52 ++
 libavutil/version.h |   2 +-
 libavutil/x86/Makefile  |   4 ++
 libavutil/x86/pixelutils.asm| 155 
 libavutil/x86/pixelutils.h  |  26 +++
 libavutil/x86/pixelutils_init.c |  58 +++
 tests/fate/libavutil.mak|   5 ++
 tests/ref/fate/pixelutils   |  15 
 12 files changed, 466 insertions(+), 1 deletion(-)
 create mode 100644 libavutil/pixelutils.c
 create mode 100644 libavutil/pixelutils.h
 create mode 100644 libavutil/x86/pixelutils.asm
 create mode 100644 libavutil/x86/pixelutils.h
 create mode 100644 libavutil/x86/pixelutils_init.c
 create mode 100644 tests/ref/fate/pixelutils

diff --git a/configure b/configure
index 9c3af50..57edd1d 100755
--- a/configure
+++ b/configure
@@ -144,6 +144,7 @@ Component options:
   --disable-mdct   disable MDCT code
   --disable-rdft   disable RDFT code
   --disable-fftdisable FFT code
+  --disable-pixelutils disable pixel utils in libavutil
 
 Hardware accelerators:
   --disable-dxva2  disable DXVA2 code [autodetect]
@@ -1451,6 +1452,7 @@ SUBSYSTEM_LIST=
 lsp
 lzo
 mdct
+pixelutils
 network
 rdft
 
diff --git a/doc/APIchanges b/doc/APIchanges
index abca377..69ca682 100644
--- a/doc/APIchanges
+++ b/doc/APIchanges
@@ -15,6 +15,9 @@ libavutil: 2012-10-22
 
 API changes, most recent first:
 
+2014-08-02 - xxx - lavu 52.95.100 - pixelutils.h
+  Add pixelutils API with SAD functions
+
 2014-07-30 - ba3e331 - lavu 52.94.100 - frame.h
   Add av_frame_side_data_name()
 
diff --git a/libavutil/Makefile b/libavutil/Makefile
index 91751dc..d57a741 100644
--- a/libavutil/Makefile
+++ b/libavutil/Makefile
@@ -44,6 +44,7 @@ HEADERS = adler32.h   
  \
   opt.h \
   parseutils.h  \
   pixdesc.h \
+  pixelutils.h  \
   pixfmt.h  \
   random_seed.h \
   replaygain.h  \
@@ -113,6 +114,7 @@ OBJS = adler32.o
\
opt.o\
parseutils.o \
pixdesc.o\
+   pixelutils.o \
random_seed.o\
rational.o   \
rc4.o\
@@ -170,6 +172,7 @@ TESTPROGS = adler32 
\
 pca \
 parseutils  \
 pixdesc \
+pixelutils  \
 random_seed \
 rational\
 ripemd  \
diff --git a/libavutil/pixelutils.c b/libavutil/pixelutils.c
new file mode 100644
index 000..278aa80
--- /dev/null
+++ b/libavutil/pixelutils.c
@@ -0,0 +1,142 @@
+/*
+ * 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 config.h
+#include common.h
+#include pixelutils.h
+
+#if CONFIG_PIXELUTILS
+
+#include x86/pixelutils.h
+
+static av_always_inline int sad_wxh(const uint8_t *src1, ptrdiff_t stride1,
+

[FFmpeg-devel] [PATCH 4/5] avfilter/mp_decimate: remove avcodec dependency

2014-08-02 Thread Clément Bœsch 
---
 configure   |  4 ++--
 libavfilter/vf_mpdecimate.c | 36 
 2 files changed, 14 insertions(+), 26 deletions(-)

diff --git a/configure b/configure
index bc187a5..210cafb 100755
--- a/configure
+++ b/configure
@@ -2549,8 +2549,8 @@ ladspa_filter_extralibs='$ldl'
 mcdeint_filter_deps=avcodec gpl
 movie_filter_deps=avcodec avformat
 mp_filter_deps=gpl avcodec swscale inline_asm
-mpdecimate_filter_deps=gpl avcodec
-mpdecimate_filter_select=me_cmp pixblockdsp
+mpdecimate_filter_deps=gpl
+mpdecimate_filter_select=pixelutils
 mptestsrc_filter_deps=gpl
 negate_filter_deps=lut_filter
 perspective_filter_deps=gpl
diff --git a/libavfilter/vf_mpdecimate.c b/libavfilter/vf_mpdecimate.c
index c667a9f..3ed9602 100644
--- a/libavfilter/vf_mpdecimate.c
+++ b/libavfilter/vf_mpdecimate.c
@@ -26,9 +26,8 @@
 
 #include libavutil/opt.h
 #include libavutil/pixdesc.h
+#include libavutil/pixelutils.h
 #include libavutil/timestamp.h
-#include libavcodec/dsputil.h
-#include libavcodec/pixblockdsp.h
 #include avfilter.h
 #include internal.h
 #include formats.h
@@ -49,9 +48,7 @@ typedef struct {
 
 int hsub, vsub;/// chroma subsampling values
 AVFrame *ref;  /// reference picture
-DSPContext dspctx; /// context providing optimized diff 
routines
-PixblockDSPContext pdsp;
-AVCodecContext *avctx; /// codec context required for the 
DSPContext
+av_pixelutils_sad_fn sad;  /// sum of absolute difference function
 } DecimateContext;
 
 #define OFFSET(x) offsetof(DecimateContext, x)
@@ -72,25 +69,21 @@ AVFILTER_DEFINE_CLASS(mpdecimate);
  * Return 1 if the two planes are different, 0 otherwise.
  */
 static int diff_planes(AVFilterContext *ctx,
-   uint8_t *cur, uint8_t *ref, int linesize,
+   uint8_t *cur, int cur_linesize,
+   uint8_t *ref, int ref_linesize,
int w, int h)
 {
 DecimateContext *decimate = ctx-priv;
-DSPContext *dspctx = decimate-dspctx;
-PixblockDSPContext *pdsp = decimate-pdsp;
 
 int x, y;
 int d, c = 0;
 int t = (w/16)*(h/16)*decimate-frac;
-int16_t block[8*8];
 
 /* compute difference for blocks of 8x8 bytes */
 for (y = 0; y  h-7; y += 4) {
 for (x = 8; x  w-7; x += 4) {
-pdsp-diff_pixels(block,
-cur+x+y*linesize,
-ref+x+y*linesize, linesize);
-d = dspctx-sum_abs_dctelem(block);
+d = decimate-sad(cur + y*cur_linesize + x, cur_linesize,
+  ref + y*ref_linesize + x, ref_linesize);
 if (d  decimate-hi)
 return 1;
 if (d  decimate-lo) {
@@ -124,7 +117,8 @@ static int decimate_frame(AVFilterContext *ctx,
 int vsub = plane == 1 || plane == 2 ? decimate-vsub : 0;
 int hsub = plane == 1 || plane == 2 ? decimate-hsub : 0;
 if (diff_planes(ctx,
-cur-data[plane], ref-data[plane], 
ref-linesize[plane],
+cur-data[plane], cur-linesize[plane],
+ref-data[plane], ref-linesize[plane],
 FF_CEIL_RSHIFT(ref-width,  hsub),
 FF_CEIL_RSHIFT(ref-height, vsub)))
 return 0;
@@ -137,15 +131,13 @@ static av_cold int init(AVFilterContext *ctx)
 {
 DecimateContext *decimate = ctx-priv;
 
+decimate-sad = av_pixelutils_get_sad_fn(3, 3, 0, decimate); // 8x8, not 
aligned on blocksize
+if (!decimate-sad)
+return AVERROR(EINVAL);
+
 av_log(ctx, AV_LOG_VERBOSE, max_drop_count:%d hi:%d lo:%d frac:%f\n,
decimate-max_drop_count, decimate-hi, decimate-lo, 
decimate-frac);
 
-decimate-avctx = avcodec_alloc_context3(NULL);
-if (!decimate-avctx)
-return AVERROR(ENOMEM);
-avpriv_dsputil_init(decimate-dspctx, decimate-avctx);
-ff_pixblockdsp_init(decimate-pdsp, decimate-avctx);
-
 return 0;
 }
 
@@ -153,10 +145,6 @@ static av_cold void uninit(AVFilterContext *ctx)
 {
 DecimateContext *decimate = ctx-priv;
 av_frame_free(decimate-ref);
-if (decimate-avctx) {
-avcodec_close(decimate-avctx);
-av_freep(decimate-avctx);
-}
 }
 
 static int query_formats(AVFilterContext *ctx)
-- 
2.0.2

___
ffmpeg-devel mailing list
ffmpeg-devel@ffmpeg.org
http://ffmpeg.org/mailman/listinfo/ffmpeg-devel

Re: [FFmpeg-devel] rectification filter

2014-08-02 Thread Clément Bœsch 
On Fri, Aug 01, 2014 at 06:47:38PM +0200, Daniel Oberhoff wrote:
[...]
  Did you already explain if / how this filter is 
  different from / better than the vignette filter?
 
 Its completely differen. Vignette modulates brightness, this one geometry, 
 but I will mention it.
 

So this is a lens correction filter? Maybe rename rectification into
lens or something (the name is kind of really generic).

-- 
Clément B.


pgpV_Vn3ol9A8.pgp
Description: PGP signature
___
ffmpeg-devel mailing list
ffmpeg-devel@ffmpeg.org
http://ffmpeg.org/mailman/listinfo/ffmpeg-devel

Re: [FFmpeg-devel] rectification filter

2014-08-02 Thread Clément Bœsch 
On Fri, Aug 01, 2014 at 01:08:58PM +0200, Daniel Oberhoff wrote:
 
 Am 01.08.2014 um 13:07 schrieb Carl Eugen Hoyos ceho...@ag.or.at:
 
  Daniel Oberhoff danieloberhoff at gmail.com writes:
  
  all work now, except vuy444p
  
  It's yuv444p
 
 oops :).
 
 And how about that test/data/fate/filter-pixfmts-rectification.rep file? 
 Should I commit that?
 

This is your run results; the references should be in tests/ref/fate.

-- 
Clément B.


pgp6w7gRPN_7z.pgp
Description: PGP signature
___
ffmpeg-devel mailing list
ffmpeg-devel@ffmpeg.org
http://ffmpeg.org/mailman/listinfo/ffmpeg-devel

Re: [FFmpeg-devel] [PATCH 3/5] avfilter/deshake: remove avcodec dependency

2014-08-02 Thread Michael Niedermayer 
On Sat, Aug 02, 2014 at 11:34:09PM +0200, Clément Bœsch wrote:
 ---
  configure|  4 +---
  libavfilter/deshake.h|  5 ++---
  libavfilter/vf_deshake.c | 20 
  3 files changed, 11 insertions(+), 18 deletions(-)
 
 diff --git a/configure b/configure
 index 5159f10..bc187a5 100755

LGTM

[...]
-- 
Michael GnuPG fingerprint: 9FF2128B147EF6730BADF133611EC787040B0FAB

Many things microsoft did are stupid, but not doing something just because
microsoft did it is even more stupid. If everything ms did were stupid they
would be bankrupt already.


signature.asc
Description: Digital signature
___
ffmpeg-devel mailing list
ffmpeg-devel@ffmpeg.org
http://ffmpeg.org/mailman/listinfo/ffmpeg-devel

Re: [FFmpeg-devel] rectification filter

2014-08-02 Thread Daniel Oberhoff 
Am 03.08.2014 um 00:12 schrieb Clément Bœsch u...@pkh.me:

 On Fri, Aug 01, 2014 at 12:56:42PM +0200, Daniel Oberhoff wrote:
 
 Am 01.08.2014 um 12:22 schrieb Clément Bœsch u...@pkh.me:
 
 On Fri, Aug 01, 2014 at 12:13:22PM +0200, Daniel Oberhoff wrote:
 
 Am 29.07.2014 um 09:54 schrieb Carl Eugen Hoyos ceho...@ag.or.at:
 
 Daniel Oberhoff danieloberhoff at gmail.com writes:
 
 OBJS-$(CONFIG_ZOOMPAN_FILTER)+= vf_zoompan.o
 +OBJS-$(CONFIG_RECTIFICATION_FILTER)  += vf_rectification.o
 
   REGISTER_FILTER(ZOOMPAN,zoompan,vf);
 +REGISTER_FILTER(RECTIFICATION,  rectification,  vf);
 
 Keep the alphabetic ordering please.
 
 
 Ok
 
 +AV_PIX_FMT_YUV410P,
 +AV_PIX_FMT_YUV444P,  AV_PIX_FMT_YUVJ444P,
 +AV_PIX_FMT_YUV420P,  AV_PIX_FMT_YUVJ420P,
 +AV_PIX_FMT_YUVA444P, AV_PIX_FMT_YUVA420P,
 +AV_PIX_FMT_NONE
 
 I have no idea what this filter does and since the 
 most important format (yuv420p) is supported, this 
 doesn't really matter but yuv410p and yuva444p are 
 quite exotic formats, I would at least have 
 expected yuv422p, yuva420p and yuv444p in the list.
 
 Did you test all above formats (at least the left row)?
 
 Added the one from your list that was missing. Can you suggest an easy way 
 of testing all these formats?
 
 
 
 add a format filter before yours in the filtergraph (-vf
 format=yuv410p,rectification …)
 
 Ok, borders can be a bitt different, but otherwise all work now, except 
 vuy444p, which ffmpeg reports as invalid. btw, how can I get the components 
 of a given color in the colorspace?
 
 If you want to add a FATE test (which would be really awesome), you can
 grep for video_filter in tests/fate/filter-video.mak (to test all pixel
 formats the filter supports), or just make a standard test with framecrc.
 
 [...]
 
 I did that by adding this:
 
 FATE_FILTER_PIXFMTS-$(CONFIG_RECTIFICATION_FILTER) += 
 fate-filter-pixfmts-rectification
 fate-filter-pixfmts-rectification: CMD = pixfmts 0.6:0.4:0.65:0.4
 
 then I ran 
 
 make fate-filter-pixfmts-rectification GEN=1
 
 which made a file
 
 test/data/fate/filter-pixfmts-rectification.rep
 
 
 This is the result of a normal run to be compared by with the reference
 which should have also been generated with the help of GEN=1 for your
 first run. The file(s) generated by GEN=1 need to be tracked.
 
 do I check this in or not?
 
 
 Of course you are supposed to check if the results make sense.
 
 Use make fate-filter-pixfmts-rectifications V=1 to see what's going on.
 
 running
 
 make fate-filter-pixfmts-rectification GEN=1
 
 runs fine…
 
 
 Of course, it doesn't do the comparison, it generates the reference.
 
 Best
 
 Daniel
 
 -- 
 Clément B.
 ___
 ffmpeg-devel mailing list
 ffmpeg-devel@ffmpeg.org
 http://ffmpeg.org/mailman/listinfo/ffmpeg-devel

Alright, will get that cleaned up. Since I seem to have all consent I need to 
also change the license I will see that I get a new patch ready soon.

Best and thanks for feedback!

Daniel
___
ffmpeg-devel mailing list
ffmpeg-devel@ffmpeg.org
http://ffmpeg.org/mailman/listinfo/ffmpeg-devel

Re: [FFmpeg-devel] [PATCH] x86/hevc_deblock: improve 8bit transpose store macros

2014-08-02 Thread Michael Niedermayer 
On Sat, Aug 02, 2014 at 11:21:31PM -0300, James Almer wrote:
 Up to four instructions less depending on function and instruction set.
 
 Signed-off-by: James Almer jamr...@gmail.com
 ---
  libavcodec/x86/hevc_deblock.asm | 70 
 +
  libavutil/x86/x86util.asm   |  9 ++
  2 files changed, 31 insertions(+), 48 deletions(-)

applied

thanks

[...]
-- 
Michael GnuPG fingerprint: 9FF2128B147EF6730BADF133611EC787040B0FAB

Why not whip the teacher when the pupil misbehaves? -- Diogenes of Sinope


signature.asc
Description: Digital signature
___
ffmpeg-devel mailing list
ffmpeg-devel@ffmpeg.org
http://ffmpeg.org/mailman/listinfo/ffmpeg-devel