Merge remote-tracking branch 'qatar/master'

* qatar/master: (23 commits)
  x86inc: use sse versions of common macros instead of sse2 when applicable
  doc/APIchanges: add missing dates and hashes
  lavf: don't return from void av_update_cur_dts()
  Changelog: add more entries.
  Changelog: update ffmpeg/avconv incompatibility list.
  avconv: remove some redundant temporary variables.
  avconv: fix broken indentation
  avconv: move copy_initial_nonkeyframes to the options context.
  avconv: use file:stream instead of file.stream in log messages.
  doc/avconv: elaborate on basic functionality.
  doc/avconv: -sample_fmts, not -help sample_fmts prints the sample formats
  openssl: Only use CRYPTO_set_id_callback on OpenSSL < 1.0.0
  Call avformat_network_init/deinit in the programs
  Remove leftover includes of strings.h
  avutil: Don't allow using strcasecmp/strncasecmp
  Replace all usage of strcasecmp/strncasecmp
  avstring: Add locale independent implementations of strcasecmp/strncasecmp
  avstring: Add locale independent implementations of toupper/tolower
  cosmetics: insert some spaces in explicit enum value assignments
  move 8SVX audio codecs to the audio codec list part on the next bump
  ...

Conflicts:
	avprobe.c
	doc/APIchanges
	ffplay.c
	ffserver.c
	libavcodec/avcodec.h
	libavdevice/bktr.c
	libavdevice/v4l.c
	libavdevice/v4l2.c
	libavformat/matroskaenc.c
	libavformat/wtv.c
	libavutil/avstring.c
	libavutil/avstring.h
	libavutil/avutil.h
	libswscale/x86/swscale_template.c
Merged-by: Michael Niedermayer <michaelni@gmx.at>

Changelog

@@ -11,37 +11,72 @@ version next:
 - added avconv, which is almost the same for now, except
 for a few incompatible changes in the options, which will hopefully make them
 easier to use. The changes are:
-    * -newvideo/-newaudio/-newsubtitle are gone, because they were redundant and
-      worked in a nonstandard way. -map is sufficient to add streams to output
-      files.
-    * -map now has slightly different and more powerful syntax.
-        + it's possible to specify stream type. E.g. -map 0:a:2 means 'third
-          audio stream'.
-        + omitting the stream index now maps all the streams of the given
-          type, not just the first. E.g. -map 0:s maps all the subtitle streams.
-        + colons (':') are used to separate file index/stream type/stream
-          index. Comma (',') is used to separate the sync stream. This is done
-          for consistency with other options.
-        + since -map can now match multiple streams, negative mappings were
+    * The options placement is now strictly enforced! While in theory the
+      options for ffmpeg should be given in [input options] -i INPUT [output
+      options] OUTPUT order, in practice it was possible to give output options
+      before the -i and it mostly worked. Except when it didn't - the behavior was
+      a bit inconsistent. In avconv, it is not possible to mix input and output
+      options. All non-global options are reset after an input or output filename.
+    * All per-file options are now truly per-file - they apply only to the next
+      input or output file and specifying different values for different files
+      will now work properly (notably -ss and -t options).
+    * All per-stream options are now truly per-stream - it is possible to
+      specify which stream(s) should a given option apply to. See the Stream
+      specifiers section in the avconv manual for details.
+    * In ffmpeg some options (like -newvideo/-newaudio/...) are irregular in the
+      sense that they're specified after the output filename instead of before,
+      like all other options. In avconv this irregularity is removed, all options
+      apply to the next input or output file.
+    * -newvideo/-newaudio/-newsubtitle options were removed. Not only were they
+      irregular and highly confusing, they were also redundant. In avconv the -map
+      option will create new streams in the output file and map input streams to
+      them. E.g. avconv -i INPUT -map 0 OUTPUT will create an output stream for
+      each stream in the first input file.
+    * The -map option now has slightly different and more powerful syntax:
+        + Colons (':') are used to separate file index/stream type/stream index
+          instead of dots. Comma (',') is used to separate the sync stream instead
+          of colon.. This is done for consistency with other options.
+        + It's possible to specify stream type. E.g. -map 0:a:2 creates an
+          output stream from the third input audio stream.
+        + Omitting the stream index now maps all the streams of the given type,
+          not just the first. E.g. -map 0:s creates output streams for all the
+          subtitle streams in the first input file.
+        + Since -map can now match multiple streams, negative mappings were
           introduced. Negative mappings disable some streams from an already
-          defined map. E.g. '-map 0 -map -0:a:1' means 'map everything except
-          for the second audio stream'.
-    * -vcodec/-acodec/-scodec are replaced by -c (or -codec), which
-      allows to precisely specify target stream(s) consistently with other
-      options. E.g. '-c:v libx264' sets the codec for all video streams,
-      '-c:a:0 libvorbis' sets the codec for the first audio stream and '-c
-      copy' copies all the streams.
+          defined map. E.g. '-map 0 -map -0:a:1' means 'create output streams for
+          all the stream in the first input file, except for the second audio
+          stream'.
+    * There is a new option -c (or -codec) for choosing the decoder/encoder to
+      use, which allows to precisely specify target stream(s) consistently with
+      other options. E.g. -c:v lib264 sets the codec for all video streams, -c:a:0
+      libvorbis sets the codec for the first audio stream and -c copy copies all
+      the streams without reencoding. Old -vcodec/-acodec/-scodec options are now
+      aliases to -c:v/a/s
     * It is now possible to precisely specify which stream should an AVOption
-      apply to. See the manual for detailed explanation.
+      apply to. E.g. -b:v:0 2M sets the bitrate for the first video stream, while
+      -b:a 128k sets the bitrate for all audio streams. Note that the old -ab 128k
+      syntax is deprecated and will stop working soon.
     * -map_chapters now takes only an input file index and applies to the next
       output file. This is consistent with how all the other options work.
     * -map_metadata now takes only an input metadata specifier and applies to
       the next output file. Output metadata specifier is now part of the option
       name, similarly to the AVOptions/map/codec feature above.
-    * Presets in avconv are disabled, because only libx264 used them and
-      presets for libx264 can now be specified using a private option
-      '-preset <presetname>'.
-    * -intra option was removed, it's equivalent to -g 0.
+    * -metadata can now be used to set metadata on streams and chapters, e.g.
+      -metadata:s:1 language=eng sets the language of the first stream to 'eng'.
+      This made -vlang/-alang/-slang options redundant, so they were removed.
+    * -qscale option now uses stream specifiers and applies to all streams, not
+      just video. I.e. plain -qscale number would now apply to all streams. To get
+      the old behavior, use -qscale:v. Also there is now a shortcut -q for -qscale
+      and -aq is now an alias for -q:a.
+    * -vbsf/-absf/-sbsf options were removed and replaced by a -bsf option which
+      uses stream specifiers. Use -bsf:v/a/s instead of the old options.
+    * -itsscale option now uses stream specifiers, so its argument is only the
+      scale parameter.
+    * -intra option was removed, use -g 0 for the same effect.
+    * -psnr option was removed, use -flags +psnr for the same effect.
+    * -vf option is now an alias to the new -filter option, which uses stream specifiers.
+    * -vframes/-aframes/-dframes options are now aliases to the new -frames option.
+    * -vtag/-atag/-stag options are now aliases to the new -tag option.
 - XMV demuxer
 - LOAS demuxer
 - ashowinfo filter added
@@ -68,6 +103,7 @@ easier to use. The changes are:
 - Ut Video decoder
 - Speex encoding via libspeex
 - 4:2:2 H.264 decoding support
+- 4:2:2 and 4:4:4 H.264 encoding with libx264
 - Pulseaudio input device
 - Prores encoder
 - Video Decoder Acceleration (VDA) HWAccel module.
@@ -77,6 +113,11 @@ easier to use. The changes are:
 - earwax audio filter added
 - libv4l2 support (--enable-libv4l2)
 - TLS/SSL and HTTPS protocol support
+- AVOptions API rewritten and documented
+- most of CODEC_FLAG2_*, some CODEC_FLAG_* and many codec-specific fields in
+  AVCodecContext deprecated. Codec private options should be used instead.
+- Properly working defaults in libx264 wrapper, support for native presets.
+- Encrypted OMA files support
 
 
 version 0.8:

doc/APIchanges

@@ -19,7 +19,13 @@ API changes, most recent first:
 2011-10-20 - b35e9e1 - lavu 51.22.0
   Add av_strtok() to avstring.h.
 
-2011-11-xx - xxxxxxx - lavf 53.13.0
+2011-11-06 - ba04ecf - lavu 51.14.0
+  Add av_strcasecmp() and av_strncasecmp() to avstring.h.
+
+2011-11-06 - 07b172f - lavu 51.13.0
+  Add av_toupper()/av_tolower()
+
+2011-11-05 - b6d08f4 - lavf 53.13.0
   Add avformat_network_init()/avformat_network_uninit()
 
 2011-10-27 - 512557b - lavc 53.15.0

doc/avconv.texi

@@ -26,6 +26,23 @@ avconv is a very fast video and audio converter that can also grab from
 a live audio/video source. It can also convert between arbitrary sample
 rates and resize video on the fly with a high quality polyphase filter.
 
+avconv reads from an arbitrary number of input "files" (which can be regular
+files, pipes, network streams, grabbing devices, etc.), specified by the
+@code{-i} option, and writes to an arbitrary number of output "files", which are
+specified by a plain output filename. Anything found on the commandline which
+cannot be interpreted as an option is considered to be an output filename.
+
+Each input or output file can in principle contain any number of streams of
+different types (video/audio/subtitle/attachment/data). Allowed number and/or
+types of streams can be limited by the container format. Selecting, which
+streams from which inputs go into output, is done either automatically or with
+the @code{-map} option (see the Stream selection chapter).
+
+To refer to input files in options, you must use their indices (0-based). E.g.
+the first input file is @code{0}, the second is @code{1} etc. Similarly, streams
+within a file are referred to by their indices. E.g. @code{2:3} refers to the
+fourth stream in the third input file. See also the Stream specifiers chapter.
+
 As a general rule, options are applied to the next specified
 file. Therefore, order is important, and you can have the same
 option on the command line multiple times. Each occurrence is
@@ -33,6 +50,10 @@ then applied to the next input or output file.
 Exceptions from this rule are the global options (e.g. verbosity level),
 which should be specified first.
 
+Do not mix input and output files -- first specify all input files, then all
+output files. Also do not mix options which belong to different files. All
+options apply ONLY to the next input or output file and are reset between files.
+
 @itemize
 @item
 To set the video bitrate of the output file to 64kbit/s:
@@ -525,6 +546,10 @@ frames after each specified time.
 This option can be useful to ensure that a seek point is present at a
 chapter mark or any other designated place in the output file.
 The timestamps must be specified in ascending order.
+
+@item -copyinkf[:@var{stream_specifier}] (@emph{output,per-stream})
+When doing stream copy, copy also non-key frames found at the
+beginning.
 @end table
 
 @section Audio Options
@@ -549,7 +574,7 @@ Disable audio recording.
 @item -acodec @var{codec} (@emph{input/output})
 Set the audio codec. This is an alias for @code{-codec:a}.
 @item -sample_fmt[:@var{stream_specifier}] @var{sample_fmt} (@emph{output,per-stream})
-Set the audio sample format. Use @code{-help sample_fmts} to get a list
+Set the audio sample format. Use @code{-sample_fmts} to get a list
 of supported sample formats.
 @end table
 

ffplay.c

@@ -914,6 +914,7 @@ static void do_exit(VideoState *is)
 #if CONFIG_AVFILTER
     avfilter_uninit();
 #endif
+    avformat_network_deinit();
     if (show_status)
         printf("\n");
     SDL_Quit();
@@ -3143,6 +3144,7 @@ int main(int argc, char **argv)
     avfilter_register_all();
 #endif
     av_register_all();
+    avformat_network_init();
 
     init_opts();
 

ffprobe.c

@@ -936,6 +936,7 @@ int main(int argc, char **argv)
 
     parse_loglevel(argc, argv, options);
     av_register_all();
+    avformat_network_init();
     init_opts();
 #if CONFIG_AVDEVICE
     avdevice_register_all();
@@ -953,5 +954,7 @@ int main(int argc, char **argv)
 
     ret = probe_file(input_filename);
 
+    avformat_network_deinit();
+
     return ret;
 }

libavcodec/avcodec.h

@@ -153,7 +153,9 @@ enum CodecID {
     CODEC_ID_TIERTEXSEQVIDEO,
     CODEC_ID_TIFF,
     CODEC_ID_GIF,
+#if LIBAVCODEC_VERSION_MAJOR == 53
     CODEC_ID_FFH264,
+#endif
     CODEC_ID_DXA,
     CODEC_ID_DNXHD,
     CODEC_ID_THP,
@@ -171,8 +173,10 @@ enum CodecID {
     CODEC_ID_INDEO5,
     CODEC_ID_MIMIC,
     CODEC_ID_RL2,
+#if LIBAVCODEC_VERSION_MAJOR == 53
     CODEC_ID_8SVX_EXP,
     CODEC_ID_8SVX_FIB,
+#endif
     CODEC_ID_ESCAPE124,
     CODEC_ID_DIRAC,
     CODEC_ID_BFI,
@@ -222,7 +226,7 @@ enum CodecID {
 
     /* various PCM "codecs" */
     CODEC_ID_FIRST_AUDIO = 0x10000,     ///< A dummy id pointing at the start of audio codecs
-    CODEC_ID_PCM_S16LE= 0x10000,
+    CODEC_ID_PCM_S16LE = 0x10000,
     CODEC_ID_PCM_S16BE,
     CODEC_ID_PCM_U16LE,
     CODEC_ID_PCM_U16BE,
@@ -251,7 +255,7 @@ enum CodecID {
     CODEC_ID_S302M,
 
     /* various ADPCM codecs */
-    CODEC_ID_ADPCM_IMA_QT= 0x11000,
+    CODEC_ID_ADPCM_IMA_QT = 0x11000,
     CODEC_ID_ADPCM_IMA_WAV,
     CODEC_ID_ADPCM_IMA_DK3,
     CODEC_ID_ADPCM_IMA_DK4,
@@ -282,21 +286,21 @@ enum CodecID {
     CODEC_ID_ADPCM_G722,
 
     /* AMR */
-    CODEC_ID_AMR_NB= 0x12000,
+    CODEC_ID_AMR_NB = 0x12000,
     CODEC_ID_AMR_WB,
 
     /* RealAudio codecs*/
-    CODEC_ID_RA_144= 0x13000,
+    CODEC_ID_RA_144 = 0x13000,
     CODEC_ID_RA_288,
 
     /* various DPCM codecs */
-    CODEC_ID_ROQ_DPCM= 0x14000,
+    CODEC_ID_ROQ_DPCM = 0x14000,
     CODEC_ID_INTERPLAY_DPCM,
     CODEC_ID_XAN_DPCM,
     CODEC_ID_SOL_DPCM,
 
     /* audio codecs */
-    CODEC_ID_MP2= 0x15000,
+    CODEC_ID_MP2 = 0x15000,
     CODEC_ID_MP3, ///< preferred ID for decoding MPEG audio layer 1, 2 or 3
     CODEC_ID_AAC,
     CODEC_ID_AC3,
@@ -308,8 +312,10 @@ enum CodecID {
     CODEC_ID_MACE3,
     CODEC_ID_MACE6,
     CODEC_ID_VMDAUDIO,
+#if LIBAVCODEC_VERSION_MAJOR == 53
     CODEC_ID_SONIC,
     CODEC_ID_SONIC_LS,
+#endif
     CODEC_ID_FLAC,
     CODEC_ID_MP3ADU,
     CODEC_ID_MP3ON4,
@@ -354,6 +360,8 @@ enum CodecID {
 #if LIBAVCODEC_VERSION_MAJOR > 53
     CODEC_ID_G723_1_DEPRECATED,
     CODEC_ID_G729_DEPRECATED,
+    CODEC_ID_8SVX_EXP,
+    CODEC_ID_8SVX_FIB,
 #endif
     CODEC_ID_G729 = 0x15800,
     CODEC_ID_G723_1= 0x15801,
@@ -361,7 +369,7 @@ enum CodecID {
 
     /* subtitle codecs */
     CODEC_ID_FIRST_SUBTITLE = 0x17000,          ///< A dummy ID pointing at the start of subtitle codecs.
-    CODEC_ID_DVD_SUBTITLE= 0x17000,
+    CODEC_ID_DVD_SUBTITLE = 0x17000,
     CODEC_ID_DVB_SUBTITLE,
     CODEC_ID_TEXT,  ///< raw UTF-8 text
     CODEC_ID_XSUB,
@@ -374,18 +382,18 @@ enum CodecID {
 
     /* other specific kind of codecs (generally used for attachments) */
     CODEC_ID_FIRST_UNKNOWN = 0x18000,           ///< A dummy ID pointing at the start of various fake codecs.
-    CODEC_ID_TTF= 0x18000,
+    CODEC_ID_TTF = 0x18000,
     CODEC_ID_BINTEXT    = MKBETAG('B','T','X','T'),
     CODEC_ID_XBIN       = MKBETAG('X','B','I','N'),
     CODEC_ID_IDF        = MKBETAG( 0 ,'I','D','F'),
 
-    CODEC_ID_PROBE= 0x19000, ///< codec_id is not known (like CODEC_ID_NONE) but lavf should attempt to identify it
+    CODEC_ID_PROBE = 0x19000, ///< codec_id is not known (like CODEC_ID_NONE) but lavf should attempt to identify it
 
-    CODEC_ID_MPEG2TS= 0x20000, /**< _FAKE_ codec to indicate a raw MPEG-2 TS
+    CODEC_ID_MPEG2TS = 0x20000, /**< _FAKE_ codec to indicate a raw MPEG-2 TS
                                 * stream (only used by libavformat) */
     CODEC_ID_MPEG4SYSTEMS = 0x20001, /**< _FAKE_ codec to indicate a MPEG-4 Systems
                                 * stream (only used by libavformat) */
-    CODEC_ID_FFMETADATA=0x21000,   ///< Dummy codec for streams containing only metadata information.
+    CODEC_ID_FFMETADATA = 0x21000,   ///< Dummy codec for streams containing only metadata information.
 };
 
 #if FF_API_OLD_SAMPLE_FMT

libavdevice/bktr.c

@@ -46,7 +46,6 @@
 #include <sys/time.h>
 #include <signal.h>
 #include <stdint.h>
-#include <strings.h>
 #include "avdevice.h"
 
 typedef struct {

libavdevice/dv1394.c

@@ -28,7 +28,6 @@
 #include <sys/mman.h>
 #include <sys/time.h>
 #include <time.h>
-#include <strings.h>
 
 #include "libavutil/log.h"
 #include "libavutil/opt.h"

libavdevice/v4l.c

@@ -38,7 +38,6 @@
 #define _LINUX_TIME_H 1
 #include <linux/videodev.h>
 #include <time.h>
-#include <strings.h>
 #include "avdevice.h"
 
 typedef struct {

libavformat/network.c

@@ -46,7 +46,7 @@ static void openssl_lock(int mode, int type, const char *file, int line)
     else
         pthread_mutex_unlock(&openssl_mutexes[type]);
 }
-#ifndef WIN32
+#if !defined(WIN32) && OPENSSL_VERSION_NUMBER < 0x10000000
 static unsigned long openssl_thread_id(void)
 {
     return (intptr_t) pthread_self();
@@ -79,7 +79,7 @@ void ff_tls_init(void)
             for (i = 0; i < CRYPTO_num_locks(); i++)
                 pthread_mutex_init(&openssl_mutexes[i], NULL);
             CRYPTO_set_locking_callback(openssl_lock);
-#ifndef WIN32
+#if !defined(WIN32) && OPENSSL_VERSION_NUMBER < 0x10000000
             CRYPTO_set_id_callback(openssl_thread_id);
 #endif
         }

libavformat/os_support.c

@@ -71,7 +71,6 @@ int ff_win32_open(const char *filename_utf8, int oflag, int pmode)
 
 #if !HAVE_INET_ATON
 #include <stdlib.h>
-#include <strings.h>
 
 int ff_inet_aton (const char * str, struct in_addr * add)
 {

libavformat/rtpdec_latm.c

@@ -23,7 +23,6 @@
 #include "internal.h"
 #include "libavutil/avstring.h"
 #include "libavcodec/get_bits.h"
-#include <strings.h>
 
 struct PayloadContext {
     AVIOContext *dyn_buf;

libavformat/utils.c

@@ -1434,7 +1434,7 @@ void ff_read_frame_flush(AVFormatContext *s)
 #if FF_API_SEEK_PUBLIC
 void av_update_cur_dts(AVFormatContext *s, AVStream *ref_st, int64_t timestamp)
 {
-    return ff_update_cur_dts(s, ref_st, timestamp);
+    ff_update_cur_dts(s, ref_st, timestamp);
 }
 #endif
 

libavutil/avstring.c

@@ -189,16 +189,12 @@ char *av_strtok(char *s, const char *delim, char **saveptr)
     return tok;
 }
 
-#define TOUPPER(c) do { if (c >= 'a' && c <= 'z') c -= 'a' - 'A'; } while (0)
-
 int av_strcasecmp(const char *a, const char *b)
 {
     uint8_t c1, c2;
     do {
-        c1 = *a++;
-        c2 = *b++;
-        TOUPPER(c1);
-        TOUPPER(c2);
+        c1 = av_tolower(*a++);
+        c2 = av_tolower(*b++);
     } while (c1 && c1 == c2);
     return c1 - c2;
 }
@@ -208,10 +204,8 @@ int av_strncasecmp(const char *a, const char *b, size_t n)
     const char *end = a + n;
     uint8_t c1, c2;
     do {
-        c1 = *a++;
-        c2 = *b++;
-        TOUPPER(c1);
-        TOUPPER(c2);
+        c1 = av_tolower(*a++);
+        c2 = av_tolower(*b++);
     } while (a < end && c1 && c1 == c2);
     return c1 - c2;
 }

libavutil/avstring.h

@@ -165,6 +165,26 @@ char *av_get_token(const char **buf, const char *term);
  */
 char *av_strtok(char *s, const char *delim, char **saveptr);
 
+/**
+ * Locale independent conversion of ASCII characters to upper case.
+ */
+static inline int av_toupper(int c)
+{
+    if (c >= 'a' && c <= 'z')
+        c ^= 0x20;
+    return c;
+}
+
+/**
+ * Locale independent conversion of ASCII characters to lower case.
+ */
+static inline int av_tolower(int c)
+{
+    if (c >= 'A' && c <= 'Z')
+        c ^= 0x20;
+    return c;
+}
+
 /**
  * Locale independent case-insensitive compare.
  * Note: This means only ASCII-range characters are case-insensitive

libavutil/avutil.h

@@ -40,7 +40,7 @@
 #define AV_VERSION(a, b, c) AV_VERSION_DOT(a, b, c)
 
 #define LIBAVUTIL_VERSION_MAJOR 51
-#define LIBAVUTIL_VERSION_MINOR 23
+#define LIBAVUTIL_VERSION_MINOR 24
 #define LIBAVUTIL_VERSION_MICRO  0
 
 #define LIBAVUTIL_VERSION_INT   AV_VERSION_INT(LIBAVUTIL_VERSION_MAJOR, \

libavutil/dict.c

@@ -18,7 +18,6 @@
  * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
  */
 
-#include <strings.h>
 #include "avstring.h"
 #include "dict.h"
 #include "internal.h"

libswscale/x86/scale.asm

@@ -34,6 +34,12 @@ yuv2yuvX_10_start:  times 4 dd 0x10000
 yuv2yuvX_9_start:   times 4 dd 0x20000
 yuv2yuvX_10_upper:  times 8 dw 0x3ff
 yuv2yuvX_9_upper:   times 8 dw 0x1ff
+pd_4:          times 4 dd 4
+pd_4min0x40000:times 4 dd 4 - (0x40000)
+pw_16:         times 8 dw 16
+pw_32:         times 8 dw 32
+pw_512:        times 8 dw 512
+pw_1024:       times 8 dw 1024
 
 SECTION .text
 
@@ -665,3 +671,139 @@ INIT_AVX
 yuv2planeX_fn avx,   8, 10, 7
 yuv2planeX_fn avx,   9,  7, 5
 yuv2planeX_fn avx,  10,  7, 5
+
+; %1=outout-bpc, %2=alignment (u/a)
+%macro yuv2plane1_mainloop 2
+.loop_%2:
+%if %1 == 8
+    paddsw          m0, m2, [r0+r2*2+mmsize*0]
+    paddsw          m1, m3, [r0+r2*2+mmsize*1]
+    psraw           m0, 7
+    psraw           m1, 7
+    packuswb        m0, m1
+    mov%2      [r1+r2], m0
+%elif %1 == 16
+    paddd           m0, m4, [r0+r2*4+mmsize*0]
+    paddd           m1, m4, [r0+r2*4+mmsize*1]
+    paddd           m2, m4, [r0+r2*4+mmsize*2]
+    paddd           m3, m4, [r0+r2*4+mmsize*3]
+    psrad           m0, 3
+    psrad           m1, 3
+    psrad           m2, 3
+    psrad           m3, 3
+%if cpuflag(sse4) ; avx/sse4
+    packusdw        m0, m1
+    packusdw        m2, m3
+%else ; mmx/sse2
+    packssdw        m0, m1
+    packssdw        m2, m3
+    paddw           m0, m5
+    paddw           m2, m5
+%endif ; mmx/sse2/sse4/avx
+    mov%2    [r1+r2*2], m0
+    mov%2    [r1+r2*2+mmsize], m2
+%else
+    paddsw          m0, m2, [r0+r2*2+mmsize*0]
+    paddsw          m1, m2, [r0+r2*2+mmsize*1]
+    psraw           m0, 15 - %1
+    psraw           m1, 15 - %1
+    pmaxsw          m0, m4
+    pmaxsw          m1, m4
+    pminsw          m0, m3
+    pminsw          m1, m3
+    mov%2    [r1+r2*2], m0
+    mov%2    [r1+r2*2+mmsize], m1
+%endif
+    add             r2, mmsize
+    jl .loop_%2
+%endmacro
+
+%macro yuv2plane1_fn 3
+cglobal yuv2plane1_%1, %3, %3, %2
+%if %1 == 8
+    add             r1, r2
+%else ; %1 != 8
+    lea             r1, [r1+r2*2]
+%endif ; %1 == 8
+%if %1 == 16
+    lea             r0, [r0+r2*4]
+%else ; %1 != 16
+    lea             r0, [r0+r2*2]
+%endif ; %1 == 16
+    neg             r2
+
+%if %1 == 8
+    pxor            m4, m4               ; zero
+
+    ; create registers holding dither
+    movq            m3, [r3]             ; dither
+    test           r4d, r4d
+    jz              .no_rot
+%if mmsize == 16
+    punpcklqdq      m3, m3
+%endif ; mmsize == 16
+    PALIGNR_MMX     m3, m3, 3, m2
+.no_rot:
+%if mmsize == 8
+    mova            m2, m3
+    punpckhbw       m3, m4               ; byte->word
+    punpcklbw       m2, m4               ; byte->word
+%else
+    punpcklbw       m3, m4
+    mova            m2, m3
+%endif
+%elif %1 == 9
+    pxor            m4, m4
+    mova            m3, [pw_512]
+    mova            m2, [pw_32]
+%elif %1 == 10
+    pxor            m4, m4
+    mova            m3, [pw_1024]
+    mova            m2, [pw_16]
+%else ; %1 == 16
+%if cpuflag(sse4) ; sse4/avx
+    mova            m4, [pd_4]
+%else ; mmx/sse2
+    mova            m4, [pd_4min0x40000]
+    mova            m5, [minshort]
+%endif ; mmx/sse2/sse4/avx
+%endif ; %1 == ..
+
+    ; actual pixel scaling
+%if mmsize == 8
+    yuv2plane1_mainloop %1, a
+%else ; mmsize == 16
+    test            r1, 15
+    jnz .unaligned
+    yuv2plane1_mainloop %1, a
+    REP_RET
+.unaligned:
+    yuv2plane1_mainloop %1, u
+%endif ; mmsize == 8/16
+    REP_RET
+%endmacro
+
+%ifdef ARCH_X86_32
+INIT_MMX mmx
+yuv2plane1_fn  8, 0, 5
+yuv2plane1_fn 16, 0, 3
+
+INIT_MMX mmx2
+yuv2plane1_fn  9, 0, 3
+yuv2plane1_fn 10, 0, 3
+%endif
+
+INIT_XMM sse2
+yuv2plane1_fn  8, 5, 5
+yuv2plane1_fn  9, 5, 3
+yuv2plane1_fn 10, 5, 3
+yuv2plane1_fn 16, 6, 3
+
+INIT_XMM sse4
+yuv2plane1_fn 16, 5, 3
+
+INIT_XMM avx
+yuv2plane1_fn  8, 5, 5
+yuv2plane1_fn  9, 5, 3
+yuv2plane1_fn 10, 5, 3
+yuv2plane1_fn 16, 5, 3

libswscale/x86/swscale_mmx.c

@@ -289,6 +289,22 @@ VSCALEX_FUNCS(sse4, sse4);
 VSCALEX_FUNC(16, sse4);
 VSCALEX_FUNCS(avx,  avx);
 
+#define VSCALE_FUNC(size, opt) \
+extern void ff_yuv2plane1_ ## size ## _ ## opt(const int16_t *src, uint8_t *dst, int dstW, \
+                                               const uint8_t *dither, int offset)
+#define VSCALE_FUNCS(opt1, opt2) \
+    VSCALE_FUNC(8,  opt1); \
+    VSCALE_FUNC(9,  opt2); \
+    VSCALE_FUNC(10, opt2); \
+    VSCALE_FUNC(16, opt1)
+
+#if ARCH_X86_32
+VSCALE_FUNCS(mmx, mmx2);
+#endif
+VSCALE_FUNCS(sse2, sse2);
+VSCALE_FUNC(16, sse4);
+VSCALE_FUNCS(avx, avx);
+
 void ff_sws_init_swScale_mmx(SwsContext *c)
 {
     int cpu_flags = av_get_cpu_flags();
@@ -336,11 +352,19 @@ switch(c->dstBpc){ \
     case 9:  if (!isBE(c->dstFormat) && opt2chk) /*vscalefn = ff_yuv2planeX_9_  ## opt2;*/ break; \
     default:                                     /*vscalefn = ff_yuv2planeX_8_  ## opt1;*/ break; \
     }
+#define ASSIGN_VSCALE_FUNC(vscalefn, opt1, opt2, opt2chk) \
+    switch(c->dstBpc){ \
+    case 16: if (!isBE(c->dstFormat))            vscalefn = ff_yuv2plane1_16_ ## opt1; break; \
+    case 10: if (!isBE(c->dstFormat) && opt2chk) vscalefn = ff_yuv2plane1_10_ ## opt2; break; \
+    case 9:  if (!isBE(c->dstFormat) && opt2chk) vscalefn = ff_yuv2plane1_9_  ## opt2;  break; \
+    default:                                     vscalefn = ff_yuv2plane1_8_  ## opt1;  break; \
+    }
 #if ARCH_X86_32
     if (cpu_flags & AV_CPU_FLAG_MMX) {
         ASSIGN_MMX_SCALE_FUNC(c->hyScale, c->hLumFilterSize, mmx, mmx);
         ASSIGN_MMX_SCALE_FUNC(c->hcScale, c->hChrFilterSize, mmx, mmx);
         ASSIGN_VSCALEX_FUNC(c->yuv2planeX, mmx, mmx2, cpu_flags & AV_CPU_FLAG_MMX2,);
+        ASSIGN_VSCALE_FUNC(c->yuv2plane1, mmx, mmx2, cpu_flags & AV_CPU_FLAG_MMX2);
     }
 #endif
 #define ASSIGN_SSE_SCALE_FUNC(hscalefn, filtersize, opt1, opt2) \
@@ -355,6 +379,7 @@ switch(c->dstBpc){ \
         ASSIGN_SSE_SCALE_FUNC(c->hyScale, c->hLumFilterSize, sse2, sse2);
         ASSIGN_SSE_SCALE_FUNC(c->hcScale, c->hChrFilterSize, sse2, sse2);
         ASSIGN_VSCALEX_FUNC(c->yuv2planeX, sse2, sse2, 1,);
+        ASSIGN_VSCALE_FUNC(c->yuv2plane1, sse2, sse2, 1);
     }
     if (cpu_flags & AV_CPU_FLAG_SSSE3) {
         ASSIGN_SSE_SCALE_FUNC(c->hyScale, c->hLumFilterSize, ssse3, ssse3);
@@ -366,10 +391,13 @@ switch(c->dstBpc){ \
         ASSIGN_SSE_SCALE_FUNC(c->hcScale, c->hChrFilterSize, sse4, ssse3);
         ASSIGN_VSCALEX_FUNC(c->yuv2planeX, sse4, sse4, 1,
                             if (!isBE(c->dstFormat)) c->yuv2planeX = ff_yuv2planeX_16_sse4);
+        if (c->dstBpc == 16 && !isBE(c->dstFormat))
+            c->yuv2plane1 = ff_yuv2plane1_16_sse4;
     }
 
     if (cpu_flags & AV_CPU_FLAG_AVX) {
         ASSIGN_VSCALEX_FUNC(c->yuv2planeX, avx, avx, 1,);
+        ASSIGN_VSCALE_FUNC(c->yuv2plane1, avx, avx, 1);
     }
 #endif
 }

libswscale/x86/swscale_template.c

@@ -109,29 +109,6 @@ static void RENAME(yuv2yuvX)(const int16_t *filter, int filterSize,
     );
 }
 
-static void RENAME(yuv2yuv1_ar)(const int16_t *src, uint8_t *dst, int dstW, const uint8_t *dither, int offset)
-{
-    dither_8to16(dither, offset);
-    __asm__ volatile(
-        "mov %2, %%"REG_a"                    \n\t"
-        ".p2align                4            \n\t" /* FIXME Unroll? */
-        "1:                                   \n\t"
-        "movq  (%0, %%"REG_a", 2), %%mm0      \n\t"
-        "movq 8(%0, %%"REG_a", 2), %%mm1      \n\t"
-        "paddsw             %%mm3, %%mm0      \n\t"
-        "paddsw             %%mm4, %%mm1      \n\t"
-        "psraw                 $7, %%mm0      \n\t"
-        "psraw                 $7, %%mm1      \n\t"
-        "packuswb           %%mm1, %%mm0      \n\t"
-        MOVNTQ(%%mm0, (%1, %%REGa))
-        "add                   $8, %%"REG_a"  \n\t"
-        "jnc                   1b             \n\t"
-        :: "r" (src + dstW), "r" (dst + dstW),
-           "g" ((x86_reg)-dstW)
-        : "%"REG_a
-    );
-}
-
 #define YSCALEYUV2PACKEDX_UV \
     __asm__ volatile(\
         "xor                   %%"REG_a", %%"REG_a"     \n\t"\
@@ -1881,9 +1858,7 @@ static av_cold void RENAME(sws_init_swScale)(SwsContext *c)
     c->use_mmx_vfilter= 0;
     if (!is16BPS(dstFormat) && !is9_OR_10BPS(dstFormat) && dstFormat != PIX_FMT_NV12
         && dstFormat != PIX_FMT_NV21 && !(c->flags & SWS_BITEXACT)) {
-            c->yuv2plane1 = RENAME(yuv2yuv1_ar    );
             if (c->flags & SWS_ACCURATE_RND) {
-                //c->yuv2yuv1 = RENAME(yuv2yuv1_ar    );
                 if (!(c->flags & SWS_FULL_CHR_H_INT)) {
                     switch (c->dstFormat) {
                     case PIX_FMT_RGB32:   c->yuv2packedX = RENAME(yuv2rgb32_X_ar);   break;
@@ -1896,7 +1871,6 @@ static av_cold void RENAME(sws_init_swScale)(SwsContext *c)
                 }
             } else {
                 int should_dither= isNBPS(c->srcFormat) || is16BPS(c->srcFormat);
-                //c->yuv2plane1 = should_dither ? RENAME(yuv2yuv1_ar    ) : RENAME(yuv2yuv1    );
                 c->use_mmx_vfilter= 1;
                 c->yuv2planeX = RENAME(yuv2yuvX    );
                 if (!(c->flags & SWS_FULL_CHR_H_INT)) {