]> git.sesse.net Git - vlc/blobdiff - modules/video_filter/deinterlace.c
Removed now unused picture_t fields and associated enums.
[vlc] / modules / video_filter / deinterlace.c
index 91aefdd2b65f4de59a2c46f8400bdd250cca8e42..70d9b2ce4500ea35f2fd99728cc76fe08d3884ee 100644 (file)
@@ -1,7 +1,7 @@
 /*****************************************************************************
  * deinterlace.c : deinterlacer plugin for vlc
  *****************************************************************************
- * Copyright (C) 2000, 2001, 2002, 2003 VideoLAN
+ * Copyright (C) 2000-2009 the VideoLAN team
  * $Id$
  *
  * Author: Sam Hocevar <sam@zoy.org>
  *
  * You should have received a copy of the GNU General Public License
  * along with this program; if not, write to the Free Software
- * Foundation, Inc., 59 Temple Place - Suite 330, Boston, MA  02111, USA.
+ * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston MA 02110-1301, USA.
  *****************************************************************************/
 
 /*****************************************************************************
  * Preamble
  *****************************************************************************/
-#include <errno.h>
-#include <stdlib.h>                                      /* malloc(), free() */
-#include <string.h>
 
-#include <vlc/vlc.h>
-#include <vlc/vout.h>
+#ifdef HAVE_CONFIG_H
+# include "config.h"
+#endif
+
+#include <assert.h>
+
+#ifdef HAVE_ALTIVEC_H
+#   include <altivec.h>
+#endif
 
-#include "../filter_common.h"
+#include <vlc_common.h>
+#include <vlc_plugin.h>
+#include <vlc_filter.h>
+#include <vlc_cpu.h>
+
+#ifdef CAN_COMPILE_MMXEXT
+#   include "mmx.h"
+#endif
 
 #define DEINTERLACE_DISCARD 1
 #define DEINTERLACE_MEAN    2
 #define DEINTERLACE_BLEND   3
 #define DEINTERLACE_BOB     4
 #define DEINTERLACE_LINEAR  5
+#define DEINTERLACE_X       6
+#define DEINTERLACE_YADIF   7
+#define DEINTERLACE_YADIF2X 8
 
 /*****************************************************************************
- * Local protypes
+ * Module descriptor
  *****************************************************************************/
-static int  Create    ( vlc_object_t * );
-static void Destroy   ( vlc_object_t * );
+static int  Open ( vlc_object_t * );
+static void Close( vlc_object_t * );
 
-static int  Init      ( vout_thread_t * );
-static void End       ( vout_thread_t * );
-static void Render    ( vout_thread_t *, picture_t * );
+#define MODE_TEXT N_("Deinterlace mode")
+#define MODE_LONGTEXT N_("Deinterlace method to use for local playback.")
+
+#define SOUT_MODE_TEXT N_("Streaming deinterlace mode")
+#define SOUT_MODE_LONGTEXT N_("Deinterlace method to use for streaming.")
+
+#define FILTER_CFG_PREFIX "sout-deinterlace-"
+
+static const char *const mode_list[] = {
+    "discard", "blend", "mean", "bob", "linear", "x", "yadif", "yadif2x" };
+static const char *const mode_list_text[] = {
+    N_("Discard"), N_("Blend"), N_("Mean"), N_("Bob"), N_("Linear"), "X", "Yadif", "Yadif (2x)" };
 
-static void RenderDiscard( vout_thread_t *, picture_t *, picture_t *, int );
-static void RenderBob    ( vout_thread_t *, picture_t *, picture_t *, int );
-static void RenderMean   ( vout_thread_t *, picture_t *, picture_t * );
-static void RenderBlend  ( vout_thread_t *, picture_t *, picture_t * );
-static void RenderLinear ( vout_thread_t *, picture_t *, picture_t *, int );
+vlc_module_begin ()
+    set_description( N_("Deinterlacing video filter") )
+    set_shortname( N_("Deinterlace" ))
+    set_capability( "video filter", 0 )
+    set_category( CAT_VIDEO )
+    set_subcategory( SUBCAT_VIDEO_VFILTER )
+
+    set_capability( "video filter2", 0 )
+    add_string( FILTER_CFG_PREFIX "mode", "blend", NULL, SOUT_MODE_TEXT,
+                SOUT_MODE_LONGTEXT, false )
+        change_string_list( mode_list, mode_list_text, 0 )
+        change_safe ()
+    add_shortcut( "deinterlace" )
+    set_callbacks( Open, Close )
+vlc_module_end ()
+
+
+/*****************************************************************************
+ * Local protypes
+ *****************************************************************************/
+static void RenderDiscard( filter_t *, picture_t *, picture_t *, int );
+static void RenderBob    ( filter_t *, picture_t *, picture_t *, int );
+static void RenderMean   ( filter_t *, picture_t *, picture_t * );
+static void RenderBlend  ( filter_t *, picture_t *, picture_t * );
+static void RenderLinear ( filter_t *, picture_t *, picture_t *, int );
+static void RenderX      ( picture_t *, picture_t * );
+static void RenderYadif  ( filter_t *, picture_t *, picture_t *, int, int );
 
 static void MergeGeneric ( void *, const void *, const void *, size_t );
 #if defined(CAN_COMPILE_C_ALTIVEC)
 static void MergeAltivec ( void *, const void *, const void *, size_t );
 #endif
-#if defined(CAN_COMPILE_MMX)
-static void MergeMMX     ( void *, const void *, const void *, size_t );
+#if defined(CAN_COMPILE_MMXEXT)
+static void MergeMMXEXT  ( void *, const void *, const void *, size_t );
+#endif
+#if defined(CAN_COMPILE_3DNOW)
+static void Merge3DNow   ( void *, const void *, const void *, size_t );
 #endif
 #if defined(CAN_COMPILE_SSE)
 static void MergeSSE2    ( void *, const void *, const void *, size_t );
 #endif
-#if defined(CAN_COMPILE_MMX) || defined(CAN_COMPILE_SSE)
+#if defined(CAN_COMPILE_MMXEXT) || defined(CAN_COMPILE_SSE)
 static void EndMMX       ( void );
 #endif
+#if defined(CAN_COMPILE_3DNOW)
+static void End3DNow     ( void );
+#endif
+#if defined __ARM_NEON__
+static void MergeNEON (void *, const void *, const void *, size_t);
+#endif
 
-static int  SendEvents   ( vlc_object_t *, char const *,
-                           vlc_value_t, vlc_value_t, void * );
-
-static void SetFilterMethod( vout_thread_t *p_vout, char *psz_method );
-static vout_thread_t *SpawnRealVout( vout_thread_t *p_vout );
-
-/*****************************************************************************
- * Callback prototypes
- *****************************************************************************/
-static int FilterCallback ( vlc_object_t *, char const *,
-                            vlc_value_t, vlc_value_t, void * );
-
-/*****************************************************************************
- * Module descriptor
- *****************************************************************************/
-#define MODE_TEXT N_("Deinterlace mode")
-#define MODE_LONGTEXT N_("You can choose the default deinterlace mode")
-
-static char *mode_list[] = { "discard", "blend", "mean", "bob", "linear" };
-static char *mode_list_text[] = { N_("Discard"), N_("Blend"), N_("Mean"),
-                                  N_("Bob"), N_("Linear") };
-
-vlc_module_begin();
-    set_description( _("Deinterlacing video filter") );
-    set_capability( "video filter", 0 );
-
-    add_string( "deinterlace-mode", "discard", NULL, MODE_TEXT,
-                MODE_LONGTEXT, VLC_FALSE );
-        change_string_list( mode_list, mode_list_text, 0 );
-
-    add_shortcut( "deinterlace" );
-    set_callbacks( Create, Destroy );
-vlc_module_end();
+static const char *const ppsz_filter_options[] = {
+    "mode", NULL
+};
 
-/*****************************************************************************
- * vout_sys_t: Deinterlace video output method descriptor
- *****************************************************************************
- * This structure is part of the video output thread descriptor.
- * It describes the Deinterlace specific properties of an output thread.
- *****************************************************************************/
-struct vout_sys_t
+#define HISTORY_SIZE (3)
+struct filter_sys_t
 {
-    int        i_mode;        /* Deinterlace mode */
-    vlc_bool_t b_double_rate; /* Shall we double the framerate? */
-
-    mtime_t    last_date;
-    mtime_t    next_date;
-
-    vout_thread_t *p_vout;
-
-    vlc_mutex_t filter_lock;
+    int  i_mode;        /* Deinterlace mode */
+    bool b_double_rate; /* Shall we double the framerate? */
+    bool b_half_height; /* Shall be divide the height by 2 */
 
     void (*pf_merge) ( void *, const void *, const void *, size_t );
     void (*pf_end_merge) ( void );
+
+    /* Yadif */
+    picture_t *pp_history[HISTORY_SIZE];
 };
 
 /*****************************************************************************
- * Create: allocates Deinterlace video thread output method
- *****************************************************************************
- * This function allocates and initializes a Deinterlace vout method.
+ * SetFilterMethod: setup the deinterlace method to use.
  *****************************************************************************/
-static int Create( vlc_object_t *p_this )
+static void SetFilterMethod( filter_t *p_filter, const char *psz_method, vlc_fourcc_t i_chroma )
 {
-    vout_thread_t *p_vout = (vout_thread_t *)p_this;
-    vlc_value_t val;
-
-    /* Allocate structure */
-    p_vout->p_sys = malloc( sizeof( vout_sys_t ) );
-    if( p_vout->p_sys == NULL )
-    {
-        msg_Err( p_vout, "out of memory" );
-        return VLC_ENOMEM;
-    }
+    filter_sys_t *p_sys = p_filter->p_sys;
 
-    p_vout->pf_init = Init;
-    p_vout->pf_end = End;
-    p_vout->pf_manage = NULL;
-    p_vout->pf_render = Render;
-    p_vout->pf_display = NULL;
+    if( !psz_method )
+        psz_method = "";
 
-    p_vout->p_sys->i_mode = DEINTERLACE_DISCARD;
-    p_vout->p_sys->b_double_rate = 0;
-    p_vout->p_sys->last_date = 0;
-    vlc_mutex_init( p_vout, &p_vout->p_sys->filter_lock );
-
-#if defined(CAN_COMPILE_C_ALTIVEC)
-    if( p_vout->p_libvlc->i_cpu & CPU_CAPABILITY_ALTIVEC )
-    {
-        p_vout->p_sys->pf_merge = MergeAltivec;
-        p_vout->p_sys->pf_end_merge = NULL;
-    }
-    else
-#endif
-#if defined(CAN_COMPILE_SSE)
-    if( p_vout->p_libvlc->i_cpu & CPU_CAPABILITY_SSE2 )
-    {
-        p_vout->p_sys->pf_merge = MergeSSE2;
-        p_vout->p_sys->pf_end_merge = EndMMX;
-    }
-    else
-#endif
-#if defined(CAN_COMPILE_MMX)
-    if( p_vout->p_libvlc->i_cpu & CPU_CAPABILITY_MMX )
-    {
-        p_vout->p_sys->pf_merge = MergeMMX;
-        p_vout->p_sys->pf_end_merge = EndMMX;
-    }
-    else
-#endif
+    if( !strcmp( psz_method, "mean" ) )
     {
-        p_vout->p_sys->pf_merge = MergeGeneric;
-        p_vout->p_sys->pf_end_merge = NULL;
+        p_sys->i_mode = DEINTERLACE_MEAN;
+        p_sys->b_double_rate = false;
+        p_sys->b_half_height = true;
     }
-
-    /* Look what method was requested */
-    var_Create( p_vout, "deinterlace-mode", VLC_VAR_STRING );
-    var_Change( p_vout, "deinterlace-mode", VLC_VAR_INHERITVALUE, &val, NULL );
-
-    if( val.psz_string == NULL )
+    else if( !strcmp( psz_method, "bob" )
+             || !strcmp( psz_method, "progressive-scan" ) )
     {
-        msg_Err( p_vout, "configuration variable deinterlace-mode empty" );
-        msg_Err( p_vout, "no deinterlace mode provided, using \"discard\"" );
-
-        val.psz_string = strdup( "discard" );
+        p_sys->i_mode = DEINTERLACE_BOB;
+        p_sys->b_double_rate = true;
+        p_sys->b_half_height = false;
     }
-
-    msg_Dbg( p_vout, "using %s deinterlace mode", val.psz_string );
-
-    SetFilterMethod( p_vout, val.psz_string );
-
-    free( val.psz_string );
-
-    var_AddCallback( p_vout, "deinterlace-mode", FilterCallback, NULL );
-
-    return VLC_SUCCESS;
-}
-
-/*****************************************************************************
- * SetFilterMethod: setup the deinterlace method to use.
- *****************************************************************************/
-static void SetFilterMethod( vout_thread_t *p_vout, char *psz_method )
-{
-    if( !strcmp( psz_method, "discard" ) )
+    else if( !strcmp( psz_method, "linear" ) )
     {
-        p_vout->p_sys->i_mode = DEINTERLACE_DISCARD;
-        p_vout->p_sys->b_double_rate = 0;
+        p_sys->i_mode = DEINTERLACE_LINEAR;
+        p_sys->b_double_rate = true;
+        p_sys->b_half_height = false;
     }
-    else if( !strcmp( psz_method, "mean" ) )
+    else if( !strcmp( psz_method, "x" ) )
     {
-        p_vout->p_sys->i_mode = DEINTERLACE_MEAN;
-        p_vout->p_sys->b_double_rate = 0;
+        p_sys->i_mode = DEINTERLACE_X;
+        p_sys->b_double_rate = false;
+        p_sys->b_half_height = false;
     }
-    else if( !strcmp( psz_method, "blend" )
-             || !strcmp( psz_method, "average" )
-             || !strcmp( psz_method, "combine-fields" ) )
+    else if( !strcmp( psz_method, "yadif" ) )
     {
-        p_vout->p_sys->i_mode = DEINTERLACE_BLEND;
-        p_vout->p_sys->b_double_rate = 0;
+        p_sys->i_mode = DEINTERLACE_YADIF;
+        p_sys->b_double_rate = false;
+        p_sys->b_half_height = false;
     }
-    else if( !strcmp( psz_method, "bob" )
-             || !strcmp( psz_method, "progressive-scan" ) )
+    else if( !strcmp( psz_method, "yadif2x" ) )
     {
-        p_vout->p_sys->i_mode = DEINTERLACE_BOB;
-        p_vout->p_sys->b_double_rate = 1;
+        p_sys->i_mode = DEINTERLACE_YADIF2X;
+        p_sys->b_double_rate = true;
+        p_sys->b_half_height = false;
     }
-    else if( !strcmp( psz_method, "linear" ) )
+    else if( !strcmp( psz_method, "discard" ) )
     {
-        p_vout->p_sys->i_mode = DEINTERLACE_LINEAR;
-        p_vout->p_sys->b_double_rate = 1;
+        const bool b_i422 = i_chroma == VLC_CODEC_I422 ||
+                            i_chroma == VLC_CODEC_J422;
+
+        p_sys->i_mode = DEINTERLACE_DISCARD;
+        p_sys->b_double_rate = false;
+        p_sys->b_half_height = !b_i422;
     }
     else
     {
-        msg_Err( p_vout, "no valid deinterlace mode provided, "
-                 "using \"discard\"" );
+        if( strcmp( psz_method, "blend" ) )
+            msg_Err( p_filter,
+                     "no valid deinterlace mode provided, using \"blend\"" );
+
+        p_sys->i_mode = DEINTERLACE_BLEND;
+        p_sys->b_double_rate = false;
+        p_sys->b_half_height = false;
     }
 
-    msg_Dbg( p_vout, "using %s deinterlace method", psz_method );
+    msg_Dbg( p_filter, "using %s deinterlace method", psz_method );
 }
 
-/*****************************************************************************
- * Init: initialize Deinterlace video thread output method
- *****************************************************************************/
-static int Init( vout_thread_t *p_vout )
+static void GetOutputFormat( filter_t *p_filter,
+                             video_format_t *p_dst, const video_format_t *p_src )
 {
-    int i_index;
-    picture_t *p_pic;
+    filter_sys_t *p_sys = p_filter->p_sys;
+    *p_dst = *p_src;
 
-    I_OUTPUTPICTURES = 0;
-
-    /* Initialize the output structure, full of directbuffers since we want
-     * the decoder to output directly to our structures. */
-    switch( p_vout->render.i_chroma )
+    if( p_sys->b_half_height )
     {
-        case VLC_FOURCC('I','4','2','0'):
-        case VLC_FOURCC('I','Y','U','V'):
-        case VLC_FOURCC('Y','V','1','2'):
-        case VLC_FOURCC('I','4','2','2'):
-            p_vout->output.i_chroma = p_vout->render.i_chroma;
-            p_vout->output.i_width  = p_vout->render.i_width;
-            p_vout->output.i_height = p_vout->render.i_height;
-            p_vout->output.i_aspect = p_vout->render.i_aspect;
-            break;
-
-        default:
-            return VLC_EGENERIC; /* unknown chroma */
-            break;
+        p_dst->i_height /= 2;
+        p_dst->i_visible_height /= 2;
+        p_dst->i_y_offset /= 2;
+        p_dst->i_sar_den *= 2;
     }
 
-    /* Try to open the real video output */
-    p_vout->p_sys->p_vout = SpawnRealVout( p_vout );
-
-    if( p_vout->p_sys->p_vout == NULL )
+    if( p_src->i_chroma == VLC_CODEC_I422 ||
+        p_src->i_chroma == VLC_CODEC_J422 )
     {
-        /* Everything failed */
-        msg_Err( p_vout, "cannot open vout, aborting" );
-
-        return VLC_EGENERIC;
-    }
-
-    ALLOCATE_DIRECTBUFFERS( VOUT_MAX_PICTURES );
-
-    ADD_CALLBACKS( p_vout->p_sys->p_vout, SendEvents );
-
-    ADD_PARENT_CALLBACKS( SendEventsToChild );
-
-    return VLC_SUCCESS;
-}
-
-/*****************************************************************************
- * SpawnRealVout: spawn the real video output.
- *****************************************************************************/
-static vout_thread_t *SpawnRealVout( vout_thread_t *p_vout )
-{
-    vout_thread_t *p_real_vout = NULL;
-
-    msg_Dbg( p_vout, "spawning the real video output" );
-
-    switch( p_vout->render.i_chroma )
-    {
-    case VLC_FOURCC('I','4','2','0'):
-    case VLC_FOURCC('I','Y','U','V'):
-    case VLC_FOURCC('Y','V','1','2'):
-        switch( p_vout->p_sys->i_mode )
+        switch( p_sys->i_mode )
         {
         case DEINTERLACE_MEAN:
-        case DEINTERLACE_DISCARD:
-            p_real_vout =
-                vout_Create( p_vout,
-                       p_vout->output.i_width, p_vout->output.i_height / 2,
-                       p_vout->output.i_chroma, p_vout->output.i_aspect );
-            break;
-
-        case DEINTERLACE_BOB:
-        case DEINTERLACE_BLEND:
         case DEINTERLACE_LINEAR:
-            p_real_vout =
-                vout_Create( p_vout,
-                       p_vout->output.i_width, p_vout->output.i_height,
-                       p_vout->output.i_chroma, p_vout->output.i_aspect );
+        case DEINTERLACE_X:
+        case DEINTERLACE_YADIF:
+        case DEINTERLACE_YADIF2X:
+            p_dst->i_chroma = p_src->i_chroma;
+            break;
+        default:
+            p_dst->i_chroma = p_src->i_chroma == VLC_CODEC_I422 ? VLC_CODEC_I420 :
+                                                                  VLC_CODEC_J420;
             break;
         }
-        break;
-
-    case VLC_FOURCC('I','4','2','2'):
-        p_real_vout =
-            vout_Create( p_vout,
-                       p_vout->output.i_width, p_vout->output.i_height,
-                       VLC_FOURCC('I','4','2','0'), p_vout->output.i_aspect );
-        break;
-
-    default:
-        break;
-    }
-
-    return p_real_vout;
-}
-
-/*****************************************************************************
- * End: terminate Deinterlace video thread output method
- *****************************************************************************/
-static void End( vout_thread_t *p_vout )
-{
-    int i_index;
-
-    /* Free the fake output buffers we allocated */
-    for( i_index = I_OUTPUTPICTURES ; i_index ; )
-    {
-        i_index--;
-        free( PP_OUTPUTPICTURE[ i_index ]->p_data_orig );
     }
 }
 
-/*****************************************************************************
- * Destroy: destroy Deinterlace video thread output method
- *****************************************************************************
- * Terminate an output method created by DeinterlaceCreateOutputMethod
- *****************************************************************************/
-static void Destroy( vlc_object_t *p_this )
-{
-    vout_thread_t *p_vout = (vout_thread_t *)p_this;
-
-    DEL_CALLBACKS( p_vout->p_sys->p_vout, SendEvents );
-
-    vlc_object_detach( p_vout->p_sys->p_vout );
-    vout_Destroy( p_vout->p_sys->p_vout );
-
-    DEL_PARENT_CALLBACKS( SendEventsToChild );
-
-    free( p_vout->p_sys );
-}
-
-/*****************************************************************************
- * Render: displays previously rendered output
- *****************************************************************************
- * This function send the currently rendered image to Deinterlace image,
- * waits until it is displayed and switch the two rendering buffers, preparing
- * next frame.
- *****************************************************************************/
-static void Render ( vout_thread_t *p_vout, picture_t *p_pic )
+static bool IsChromaSupported( vlc_fourcc_t i_chroma )
 {
-    picture_t *pp_outpic[2];
-
-    vlc_mutex_lock( &p_vout->p_sys->filter_lock );
-
-    /* Get a new picture */
-    while( ( pp_outpic[0] = vout_CreatePicture( p_vout->p_sys->p_vout,
-                                             0, 0, 0 ) )
-              == NULL )
-    {
-        if( p_vout->b_die || p_vout->b_error )
-        {
-            vlc_mutex_unlock( &p_vout->p_sys->filter_lock );
-            return;
-        }
-        msleep( VOUT_OUTMEM_SLEEP );
-     }
-
-    vout_DatePicture( p_vout->p_sys->p_vout, pp_outpic[0], p_pic->date );
-
-    /* If we are using double rate, get an additional new picture */
-    if( p_vout->p_sys->b_double_rate )
-    {
-        while( ( pp_outpic[1] = vout_CreatePicture( p_vout->p_sys->p_vout,
-                                                 0, 0, 0 ) )
-                  == NULL )
-        {
-            if( p_vout->b_die || p_vout->b_error )
-            {
-                vout_DestroyPicture( p_vout->p_sys->p_vout, pp_outpic[0] );
-                vlc_mutex_unlock( &p_vout->p_sys->filter_lock );
-                return;
-            }
-            msleep( VOUT_OUTMEM_SLEEP );
-        }
-
-        /* 20ms is a bit arbitrary, but it's only for the first image we get */
-        if( !p_vout->p_sys->last_date )
-        {
-            vout_DatePicture( p_vout->p_sys->p_vout, pp_outpic[1],
-                              p_pic->date + 20000 );
-        }
-        else
-        {
-            vout_DatePicture( p_vout->p_sys->p_vout, pp_outpic[1],
-                      (3 * p_pic->date - p_vout->p_sys->last_date) / 2 );
-        }
-        p_vout->p_sys->last_date = p_pic->date;
-    }
-
-    switch( p_vout->p_sys->i_mode )
-    {
-        case DEINTERLACE_DISCARD:
-            RenderDiscard( p_vout, pp_outpic[0], p_pic, 0 );
-            vout_DisplayPicture( p_vout->p_sys->p_vout, pp_outpic[0] );
-            break;
-
-        case DEINTERLACE_BOB:
-            RenderBob( p_vout, pp_outpic[0], p_pic, 0 );
-            vout_DisplayPicture( p_vout->p_sys->p_vout, pp_outpic[0] );
-            RenderBob( p_vout, pp_outpic[1], p_pic, 1 );
-            vout_DisplayPicture( p_vout->p_sys->p_vout, pp_outpic[1] );
-            break;
-
-        case DEINTERLACE_LINEAR:
-            RenderLinear( p_vout, pp_outpic[0], p_pic, 0 );
-            vout_DisplayPicture( p_vout->p_sys->p_vout, pp_outpic[0] );
-            RenderLinear( p_vout, pp_outpic[1], p_pic, 1 );
-            vout_DisplayPicture( p_vout->p_sys->p_vout, pp_outpic[1] );
-            break;
-
-        case DEINTERLACE_MEAN:
-            RenderMean( p_vout, pp_outpic[0], p_pic );
-            vout_DisplayPicture( p_vout->p_sys->p_vout, pp_outpic[0] );
-            break;
-
-        case DEINTERLACE_BLEND:
-            RenderBlend( p_vout, pp_outpic[0], p_pic );
-            vout_DisplayPicture( p_vout->p_sys->p_vout, pp_outpic[0] );
-            break;
-    }
-
-    vlc_mutex_unlock( &p_vout->p_sys->filter_lock );
+    return i_chroma == VLC_CODEC_I420 ||
+           i_chroma == VLC_CODEC_J420 ||
+           i_chroma == VLC_CODEC_YV12 ||
+           i_chroma == VLC_CODEC_I422 ||
+           i_chroma == VLC_CODEC_J422;
 }
 
 /*****************************************************************************
  * RenderDiscard: only keep TOP or BOTTOM field, discard the other.
  *****************************************************************************/
-static void RenderDiscard( vout_thread_t *p_vout,
+static void RenderDiscard( filter_t *p_filter,
                            picture_t *p_outpic, picture_t *p_pic, int i_field )
 {
     int i_plane;
@@ -492,25 +273,25 @@ static void RenderDiscard( vout_thread_t *p_vout,
 
         p_out = p_outpic->p[i_plane].p_pixels;
         p_out_end = p_out + p_outpic->p[i_plane].i_pitch
-                             * p_outpic->p[i_plane].i_lines;
+                             * p_outpic->p[i_plane].i_visible_lines;
 
-        switch( p_vout->render.i_chroma )
+        switch( p_filter->fmt_in.video.i_chroma )
         {
-        case VLC_FOURCC('I','4','2','0'):
-        case VLC_FOURCC('I','Y','U','V'):
-        case VLC_FOURCC('Y','V','1','2'):
+        case VLC_CODEC_I420:
+        case VLC_CODEC_J420:
+        case VLC_CODEC_YV12:
 
             for( ; p_out < p_out_end ; )
             {
-                p_vout->p_vlc->pf_memcpy( p_out, p_in,
-                                          p_pic->p[i_plane].i_pitch );
+                vlc_memcpy( p_out, p_in, p_pic->p[i_plane].i_pitch );
 
-                p_out += p_pic->p[i_plane].i_pitch;
+                p_out += p_outpic->p[i_plane].i_pitch;
                 p_in += 2 * p_pic->p[i_plane].i_pitch;
             }
             break;
 
-        case VLC_FOURCC('I','4','2','2'):
+        case VLC_CODEC_I422:
+        case VLC_CODEC_J422:
 
             i_increment = 2 * p_pic->p[i_plane].i_pitch;
 
@@ -518,12 +299,10 @@ static void RenderDiscard( vout_thread_t *p_vout,
             {
                 for( ; p_out < p_out_end ; )
                 {
-                    p_vout->p_vlc->pf_memcpy( p_out, p_in,
-                                              p_pic->p[i_plane].i_pitch );
-                    p_out += p_pic->p[i_plane].i_pitch;
-                    p_vout->p_vlc->pf_memcpy( p_out, p_in,
-                                              p_pic->p[i_plane].i_pitch );
-                    p_out += p_pic->p[i_plane].i_pitch;
+                    vlc_memcpy( p_out, p_in, p_pic->p[i_plane].i_pitch );
+                    p_out += p_outpic->p[i_plane].i_pitch;
+                    vlc_memcpy( p_out, p_in, p_pic->p[i_plane].i_pitch );
+                    p_out += p_outpic->p[i_plane].i_pitch;
                     p_in += i_increment;
                 }
             }
@@ -531,9 +310,8 @@ static void RenderDiscard( vout_thread_t *p_vout,
             {
                 for( ; p_out < p_out_end ; )
                 {
-                    p_vout->p_vlc->pf_memcpy( p_out, p_in,
-                                              p_pic->p[i_plane].i_pitch );
-                    p_out += p_pic->p[i_plane].i_pitch;
+                    vlc_memcpy( p_out, p_in, p_pic->p[i_plane].i_pitch );
+                    p_out += p_outpic->p[i_plane].i_pitch;
                     p_in += i_increment;
                 }
             }
@@ -548,7 +326,7 @@ static void RenderDiscard( vout_thread_t *p_vout,
 /*****************************************************************************
  * RenderBob: renders a BOB picture - simple copy
  *****************************************************************************/
-static void RenderBob( vout_thread_t *p_vout,
+static void RenderBob( filter_t *p_filter,
                        picture_t *p_outpic, picture_t *p_pic, int i_field )
 {
     int i_plane;
@@ -561,59 +339,54 @@ static void RenderBob( vout_thread_t *p_vout,
         p_in = p_pic->p[i_plane].p_pixels;
         p_out = p_outpic->p[i_plane].p_pixels;
         p_out_end = p_out + p_outpic->p[i_plane].i_pitch
-                             * p_outpic->p[i_plane].i_lines;
+                             * p_outpic->p[i_plane].i_visible_lines;
 
-        switch( p_vout->render.i_chroma )
+        switch( p_filter->fmt_in.video.i_chroma )
         {
-            case VLC_FOURCC('I','4','2','0'):
-            case VLC_FOURCC('I','Y','U','V'):
-            case VLC_FOURCC('Y','V','1','2'):
+            case VLC_CODEC_I420:
+            case VLC_CODEC_J420:
+            case VLC_CODEC_YV12:
                 /* For BOTTOM field we need to add the first line */
                 if( i_field == 1 )
                 {
-                    p_vout->p_vlc->pf_memcpy( p_out, p_in,
-                                              p_pic->p[i_plane].i_pitch );
+                    vlc_memcpy( p_out, p_in, p_pic->p[i_plane].i_pitch );
                     p_in += p_pic->p[i_plane].i_pitch;
-                    p_out += p_pic->p[i_plane].i_pitch;
+                    p_out += p_outpic->p[i_plane].i_pitch;
                 }
 
                 p_out_end -= 2 * p_outpic->p[i_plane].i_pitch;
 
                 for( ; p_out < p_out_end ; )
                 {
-                    p_vout->p_vlc->pf_memcpy( p_out, p_in,
-                                              p_pic->p[i_plane].i_pitch );
+                    vlc_memcpy( p_out, p_in, p_pic->p[i_plane].i_pitch );
 
-                    p_out += p_pic->p[i_plane].i_pitch;
+                    p_out += p_outpic->p[i_plane].i_pitch;
 
-                    p_vout->p_vlc->pf_memcpy( p_out, p_in,
-                                              p_pic->p[i_plane].i_pitch );
+                    vlc_memcpy( p_out, p_in, p_pic->p[i_plane].i_pitch );
 
                     p_in += 2 * p_pic->p[i_plane].i_pitch;
-                    p_out += p_pic->p[i_plane].i_pitch;
+                    p_out += p_outpic->p[i_plane].i_pitch;
                 }
 
-                p_vout->p_vlc->pf_memcpy( p_out, p_in,
-                                          p_pic->p[i_plane].i_pitch );
+                vlc_memcpy( p_out, p_in, p_pic->p[i_plane].i_pitch );
 
                 /* For TOP field we need to add the last line */
                 if( i_field == 0 )
                 {
                     p_in += p_pic->p[i_plane].i_pitch;
-                    p_out += p_pic->p[i_plane].i_pitch;
-                    p_vout->p_vlc->pf_memcpy( p_out, p_in,
-                                              p_pic->p[i_plane].i_pitch );
+                    p_out += p_outpic->p[i_plane].i_pitch;
+                    vlc_memcpy( p_out, p_in, p_pic->p[i_plane].i_pitch );
                 }
                 break;
 
-            case VLC_FOURCC('I','4','2','2'):
+            case VLC_CODEC_I422:
+            case VLC_CODEC_J422:
                 /* For BOTTOM field we need to add the first line */
                 if( i_field == 1 )
                 {
-                    p_vout->p_vlc->pf_memcpy( p_out, p_in,
-                                              p_pic->p[i_plane].i_pitch );
+                    vlc_memcpy( p_out, p_in, p_pic->p[i_plane].i_pitch );
                     p_in += p_pic->p[i_plane].i_pitch;
-                    p_out += p_pic->p[i_plane].i_pitch;
+                    p_out += p_outpic->p[i_plane].i_pitch;
                 }
 
                 p_out_end -= 2 * p_outpic->p[i_plane].i_pitch;
@@ -622,53 +395,48 @@ static void RenderBob( vout_thread_t *p_vout,
                 {
                     for( ; p_out < p_out_end ; )
                     {
-                        p_vout->p_vlc->pf_memcpy( p_out, p_in,
-                                                  p_pic->p[i_plane].i_pitch );
+                        vlc_memcpy( p_out, p_in, p_pic->p[i_plane].i_pitch );
 
-                        p_out += p_pic->p[i_plane].i_pitch;
+                        p_out += p_outpic->p[i_plane].i_pitch;
 
-                        p_vout->p_vlc->pf_memcpy( p_out, p_in,
-                                                  p_pic->p[i_plane].i_pitch );
+                        vlc_memcpy( p_out, p_in, p_pic->p[i_plane].i_pitch );
 
                         p_in += 2 * p_pic->p[i_plane].i_pitch;
-                        p_out += p_pic->p[i_plane].i_pitch;
+                        p_out += p_outpic->p[i_plane].i_pitch;
                     }
                 }
                 else
                 {
                     for( ; p_out < p_out_end ; )
                     {
-                        p_vout->p_vlc->pf_memcpy( p_out, p_in,
-                                                  p_pic->p[i_plane].i_pitch );
+                        vlc_memcpy( p_out, p_in, p_pic->p[i_plane].i_pitch );
 
-                        p_out += p_pic->p[i_plane].i_pitch;
+                        p_out += p_outpic->p[i_plane].i_pitch;
                         p_in += 2 * p_pic->p[i_plane].i_pitch;
                     }
                 }
 
-                p_vout->p_vlc->pf_memcpy( p_out, p_in,
-                                          p_pic->p[i_plane].i_pitch );
+                vlc_memcpy( p_out, p_in, p_pic->p[i_plane].i_pitch );
 
                 /* For TOP field we need to add the last line */
                 if( i_field == 0 )
                 {
                     p_in += p_pic->p[i_plane].i_pitch;
-                    p_out += p_pic->p[i_plane].i_pitch;
-                    p_vout->p_vlc->pf_memcpy( p_out, p_in,
-                                              p_pic->p[i_plane].i_pitch );
+                    p_out += p_outpic->p[i_plane].i_pitch;
+                    vlc_memcpy( p_out, p_in, p_pic->p[i_plane].i_pitch );
                 }
                 break;
         }
     }
 }
 
-#define Merge p_vout->p_sys->pf_merge
-#define EndMerge if(p_vout->p_sys->pf_end_merge) p_vout->p_sys->pf_end_merge
+#define Merge p_filter->p_sys->pf_merge
+#define EndMerge if(p_filter->p_sys->pf_end_merge) p_filter->p_sys->pf_end_merge
 
 /*****************************************************************************
  * RenderLinear: BOB with linear interpolation
  *****************************************************************************/
-static void RenderLinear( vout_thread_t *p_vout,
+static void RenderLinear( filter_t *p_filter,
                           picture_t *p_outpic, picture_t *p_pic, int i_field )
 {
     int i_plane;
@@ -681,49 +449,45 @@ static void RenderLinear( vout_thread_t *p_vout,
         p_in = p_pic->p[i_plane].p_pixels;
         p_out = p_outpic->p[i_plane].p_pixels;
         p_out_end = p_out + p_outpic->p[i_plane].i_pitch
-                             * p_outpic->p[i_plane].i_lines;
+                             * p_outpic->p[i_plane].i_visible_lines;
 
         /* For BOTTOM field we need to add the first line */
         if( i_field == 1 )
         {
-            p_vout->p_vlc->pf_memcpy( p_out, p_in,
-                                      p_pic->p[i_plane].i_pitch );
+            vlc_memcpy( p_out, p_in, p_pic->p[i_plane].i_pitch );
             p_in += p_pic->p[i_plane].i_pitch;
-            p_out += p_pic->p[i_plane].i_pitch;
+            p_out += p_outpic->p[i_plane].i_pitch;
         }
 
         p_out_end -= 2 * p_outpic->p[i_plane].i_pitch;
 
         for( ; p_out < p_out_end ; )
         {
-            p_vout->p_vlc->pf_memcpy( p_out, p_in,
-                                      p_pic->p[i_plane].i_pitch );
+            vlc_memcpy( p_out, p_in, p_pic->p[i_plane].i_pitch );
 
-            p_out += p_pic->p[i_plane].i_pitch;
+            p_out += p_outpic->p[i_plane].i_pitch;
 
             Merge( p_out, p_in, p_in + 2 * p_pic->p[i_plane].i_pitch,
                    p_pic->p[i_plane].i_pitch );
 
             p_in += 2 * p_pic->p[i_plane].i_pitch;
-            p_out += p_pic->p[i_plane].i_pitch;
+            p_out += p_outpic->p[i_plane].i_pitch;
         }
 
-        p_vout->p_vlc->pf_memcpy( p_out, p_in,
-                                  p_pic->p[i_plane].i_pitch );
+        vlc_memcpy( p_out, p_in, p_pic->p[i_plane].i_pitch );
 
         /* For TOP field we need to add the last line */
         if( i_field == 0 )
         {
             p_in += p_pic->p[i_plane].i_pitch;
-            p_out += p_pic->p[i_plane].i_pitch;
-            p_vout->p_vlc->pf_memcpy( p_out, p_in,
-                                      p_pic->p[i_plane].i_pitch );
+            p_out += p_outpic->p[i_plane].i_pitch;
+            vlc_memcpy( p_out, p_in, p_pic->p[i_plane].i_pitch );
         }
     }
     EndMerge();
 }
 
-static void RenderMean( vout_thread_t *p_vout,
+static void RenderMean( filter_t *p_filter,
                         picture_t *p_outpic, picture_t *p_pic )
 {
     int i_plane;
@@ -737,7 +501,7 @@ static void RenderMean( vout_thread_t *p_vout,
 
         p_out = p_outpic->p[i_plane].p_pixels;
         p_out_end = p_out + p_outpic->p[i_plane].i_pitch
-                             * p_outpic->p[i_plane].i_lines;
+                             * p_outpic->p[i_plane].i_visible_lines;
 
         /* All lines: mean value */
         for( ; p_out < p_out_end ; )
@@ -745,14 +509,14 @@ static void RenderMean( vout_thread_t *p_vout,
             Merge( p_out, p_in, p_in + p_pic->p[i_plane].i_pitch,
                    p_pic->p[i_plane].i_pitch );
 
-            p_out += p_pic->p[i_plane].i_pitch;
+            p_out += p_outpic->p[i_plane].i_pitch;
             p_in += 2 * p_pic->p[i_plane].i_pitch;
         }
     }
     EndMerge();
 }
 
-static void RenderBlend( vout_thread_t *p_vout,
+static void RenderBlend( filter_t *p_filter,
                          picture_t *p_outpic, picture_t *p_pic )
 {
     int i_plane;
@@ -766,34 +530,33 @@ static void RenderBlend( vout_thread_t *p_vout,
 
         p_out = p_outpic->p[i_plane].p_pixels;
         p_out_end = p_out + p_outpic->p[i_plane].i_pitch
-                             * p_outpic->p[i_plane].i_lines;
+                             * p_outpic->p[i_plane].i_visible_lines;
 
-        switch( p_vout->render.i_chroma )
+        switch( p_filter->fmt_in.video.i_chroma )
         {
-            case VLC_FOURCC('I','4','2','0'):
-            case VLC_FOURCC('I','Y','U','V'):
-            case VLC_FOURCC('Y','V','1','2'):
+            case VLC_CODEC_I420:
+            case VLC_CODEC_J420:
+            case VLC_CODEC_YV12:
                 /* First line: simple copy */
-                p_vout->p_vlc->pf_memcpy( p_out, p_in,
-                                          p_pic->p[i_plane].i_pitch );
-                p_out += p_pic->p[i_plane].i_pitch;
+                vlc_memcpy( p_out, p_in, p_pic->p[i_plane].i_pitch );
+                p_out += p_outpic->p[i_plane].i_pitch;
 
                 /* Remaining lines: mean value */
                 for( ; p_out < p_out_end ; )
                 {
-                   Merge( p_out, p_in, p_in + p_pic->p[i_plane].i_pitch,
-                          p_pic->p[i_plane].i_pitch );
+                    Merge( p_out, p_in, p_in + p_pic->p[i_plane].i_pitch,
+                           p_pic->p[i_plane].i_pitch );
 
-                    p_out += p_pic->p[i_plane].i_pitch;
+                    p_out += p_outpic->p[i_plane].i_pitch;
                     p_in += p_pic->p[i_plane].i_pitch;
                 }
                 break;
 
-            case VLC_FOURCC('I','4','2','2'):
+            case VLC_CODEC_I422:
+            case VLC_CODEC_J422:
                 /* First line: simple copy */
-                p_vout->p_vlc->pf_memcpy( p_out, p_in,
-                                          p_pic->p[i_plane].i_pitch );
-                p_out += p_pic->p[i_plane].i_pitch;
+                vlc_memcpy( p_out, p_in, p_pic->p[i_plane].i_pitch );
+                p_out += p_outpic->p[i_plane].i_pitch;
 
                 /* Remaining lines: mean value */
                 if( i_plane == Y_PLANE )
@@ -803,7 +566,7 @@ static void RenderBlend( vout_thread_t *p_vout,
                         Merge( p_out, p_in, p_in + p_pic->p[i_plane].i_pitch,
                                p_pic->p[i_plane].i_pitch );
 
-                        p_out += p_pic->p[i_plane].i_pitch;
+                        p_out += p_outpic->p[i_plane].i_pitch;
                         p_in += p_pic->p[i_plane].i_pitch;
                     }
                 }
@@ -815,7 +578,7 @@ static void RenderBlend( vout_thread_t *p_vout,
                         Merge( p_out, p_in, p_in + p_pic->p[i_plane].i_pitch,
                                p_pic->p[i_plane].i_pitch );
 
-                        p_out += p_pic->p[i_plane].i_pitch;
+                        p_out += p_outpic->p[i_plane].i_pitch;
                         p_in += 2*p_pic->p[i_plane].i_pitch;
                     }
                 }
@@ -855,9 +618,9 @@ static void MergeGeneric( void *_p_dest, const void *_p_s1,
     }
 }
 
-#if defined(CAN_COMPILE_MMX)
-static void MergeMMX( void *_p_dest, const void *_p_s1, const void *_p_s2,
-                      size_t i_bytes )
+#if defined(CAN_COMPILE_MMXEXT)
+static void MergeMMXEXT( void *_p_dest, const void *_p_s1, const void *_p_s2,
+                         size_t i_bytes )
 {
     uint8_t* p_dest = (uint8_t*)_p_dest;
     const uint8_t *p_s1 = (const uint8_t *)_p_s1;
@@ -884,6 +647,35 @@ static void MergeMMX( void *_p_dest, const void *_p_s1, const void *_p_s2,
 }
 #endif
 
+#if defined(CAN_COMPILE_3DNOW)
+static void Merge3DNow( void *_p_dest, const void *_p_s1, const void *_p_s2,
+                        size_t i_bytes )
+{
+    uint8_t* p_dest = (uint8_t*)_p_dest;
+    const uint8_t *p_s1 = (const uint8_t *)_p_s1;
+    const uint8_t *p_s2 = (const uint8_t *)_p_s2;
+    uint8_t* p_end = p_dest + i_bytes - 8;
+    while( p_dest < p_end )
+    {
+        __asm__  __volatile__( "movq %2,%%mm1;"
+                               "pavgusb %1, %%mm1;"
+                               "movq %%mm1, %0" :"=m" (*p_dest):
+                                                 "m" (*p_s1),
+                                                 "m" (*p_s2) );
+        p_dest += 8;
+        p_s1 += 8;
+        p_s2 += 8;
+    }
+
+    p_end += 8;
+
+    while( p_dest < p_end )
+    {
+        *p_dest++ = ( (uint16_t)(*p_s1++) + (uint16_t)(*p_s2++) ) >> 1;
+    }
+}
+#endif
+
 #if defined(CAN_COMPILE_SSE)
 static void MergeSSE2( void *_p_dest, const void *_p_s1, const void *_p_s2,
                        size_t i_bytes )
@@ -891,11 +683,12 @@ static void MergeSSE2( void *_p_dest, const void *_p_s1, const void *_p_s2,
     uint8_t* p_dest = (uint8_t*)_p_dest;
     const uint8_t *p_s1 = (const uint8_t *)_p_s1;
     const uint8_t *p_s2 = (const uint8_t *)_p_s2;
-    while( (int)p_s1 % 16 )
+    uint8_t* p_end;
+    while( (uintptr_t)p_s1 % 16 )
     {
         *p_dest++ = ( (uint16_t)(*p_s1++) + (uint16_t)(*p_s2++) ) >> 1;
-    }        
-    uint8_t* p_end = p_dest + i_bytes - 16;
+    }
+    p_end = p_dest + i_bytes - 16;
     while( p_dest < p_end )
     {
         __asm__  __volatile__( "movdqu %2,%%xmm1;"
@@ -917,40 +710,82 @@ static void MergeSSE2( void *_p_dest, const void *_p_s1, const void *_p_s2,
 }
 #endif
 
-#if defined(CAN_COMPILE_MMX) || defined(CAN_COMPILE_SSE)
+#if defined(CAN_COMPILE_MMXEXT) || defined(CAN_COMPILE_SSE)
 static void EndMMX( void )
 {
     __asm__ __volatile__( "emms" :: );
 }
 #endif
 
+#if defined(CAN_COMPILE_3DNOW)
+static void End3DNow( void )
+{
+    __asm__ __volatile__( "femms" :: );
+}
+#endif
+
 #ifdef CAN_COMPILE_C_ALTIVEC
 static void MergeAltivec( void *_p_dest, const void *_p_s1,
                           const void *_p_s2, size_t i_bytes )
 {
-    uint8_t *p_dest = (uint8_t*)_p_dest;
-    const uint8_t *p_s1 = (const uint8_t *)_p_s1;
-    const uint8_t *p_s2 = (const uint8_t *)_p_s2;
-    uint8_t *p_end = p_dest + i_bytes - 16;
+    uint8_t *p_dest = (uint8_t *)_p_dest;
+    uint8_t *p_s1   = (uint8_t *)_p_s1;
+    uint8_t *p_s2   = (uint8_t *)_p_s2;
+    uint8_t *p_end  = p_dest + i_bytes - 15;
 
-    if( ( (int)p_s1 & 0xF ) | ( (int)p_s2 & 0xF ) |
-        ( (int)p_dest & 0xF ) )
+    /* Use C until the first 16-bytes aligned destination pixel */
+    while( (uintptr_t)p_dest & 0xF )
     {
-        /* TODO Handle non 16-bytes aligned planes */
-        MergeGeneric( _p_dest, _p_s1, _p_s2, i_bytes );
-        return;
+        *p_dest++ = ( (uint16_t)(*p_s1++) + (uint16_t)(*p_s2++) ) >> 1;
     }
 
-    while( p_dest < p_end )
+    if( ( (int)p_s1 & 0xF ) | ( (int)p_s2 & 0xF ) )
     {
-        vec_st( vec_avg( vec_ld( 0, p_s1 ), vec_ld( 0, p_s2 ) ),
-                0, p_dest );
-        p_s1   += 16;
-        p_s2   += 16;
-        p_dest += 16;
+        /* Unaligned source */
+        vector unsigned char s1v, s2v, destv;
+        vector unsigned char s1oldv, s2oldv, s1newv, s2newv;
+        vector unsigned char perm1v, perm2v;
+
+        perm1v = vec_lvsl( 0, p_s1 );
+        perm2v = vec_lvsl( 0, p_s2 );
+        s1oldv = vec_ld( 0, p_s1 );
+        s2oldv = vec_ld( 0, p_s2 );
+
+        while( p_dest < p_end )
+        {
+            s1newv = vec_ld( 16, p_s1 );
+            s2newv = vec_ld( 16, p_s2 );
+            s1v    = vec_perm( s1oldv, s1newv, perm1v );
+            s2v    = vec_perm( s2oldv, s2newv, perm2v );
+            s1oldv = s1newv;
+            s2oldv = s2newv;
+            destv  = vec_avg( s1v, s2v );
+            vec_st( destv, 0, p_dest );
+
+            p_s1   += 16;
+            p_s2   += 16;
+            p_dest += 16;
+        }
     }
+    else
+    {
+        /* Aligned source */
+        vector unsigned char s1v, s2v, destv;
 
-    p_end += 16;
+        while( p_dest < p_end )
+        {
+            s1v   = vec_ld( 0, p_s1 );
+            s2v   = vec_ld( 0, p_s2 );
+            destv = vec_avg( s1v, s2v );
+            vec_st( destv, 0, p_dest );
+
+            p_s1   += 16;
+            p_s2   += 16;
+            p_dest += 16;
+        }
+    }
+
+    p_end += 15;
 
     while( p_dest < p_end )
     {
@@ -959,119 +794,920 @@ static void MergeAltivec( void *_p_dest, const void *_p_s1,
 }
 #endif
 
+#ifdef __ARM_NEON__
+static void MergeNEON (void *restrict out, const void *in1,
+                       const void *in2, size_t n)
+{
+    uint8_t *outp = out;
+    const uint8_t *in1p = in1;
+    const uint8_t *in2p = in2;
+    size_t mis = ((uintptr_t)outp) & 15;
+
+    if (mis)
+    {
+        MergeGeneric (outp, in1p, in2p, mis);
+        outp += mis;
+        in1p += mis;
+        in2p += mis;
+        n -= mis;
+    }
+
+    uint8_t *end = outp + (n & ~15);
+
+    if ((((uintptr_t)in1p)|((uintptr_t)in2p)) & 15)
+        while (outp < end)
+            asm volatile (
+                "vld1.u8  {q0-q1}, [%[in1]]!\n"
+                "vld1.u8  {q2-q3}, [%[in2]]!\n"
+                "vhadd.u8 q4, q0, q2\n"
+                "vld1.u8  {q6-q7}, [%[in1]]!\n"
+                "vhadd.u8 q5, q1, q3\n"
+                "vld1.u8  {q8-q9}, [%[in2]]!\n"
+                "vhadd.u8 q10, q6, q8\n"
+                "vhadd.u8 q11, q7, q9\n"
+                "vst1.u8  {q4-q5}, [%[out],:128]!\n"
+                "vst1.u8  {q10-q11}, [%[out],:128]!\n"
+                : [out] "+r" (outp), [in1] "+r" (in1p), [in2] "+r" (in2p)
+                :
+                : "q0", "q1", "q2", "memory");
+    else
+         while (outp < end)
+            asm volatile (
+                "vld1.u8  {q0-q1}, [%[in1],:128]!\n"
+                "vld1.u8  {q2-q3}, [%[in2],:128]!\n"
+                "vhadd.u8 q4, q0, q2\n"
+                "vld1.u8  {q6-q7}, [%[in1],:128]!\n"
+                "vhadd.u8 q5, q1, q3\n"
+                "vld1.u8  {q8-q9}, [%[in2],:128]!\n"
+                "vhadd.u8 q10, q6, q8\n"
+                "vhadd.u8 q11, q7, q9\n"
+                "vst1.u8  {q4-q5}, [%[out],:128]!\n"
+                "vst1.u8  {q10-q11}, [%[out],:128]!\n"
+                : [out] "+r" (outp), [in1] "+r" (in1p), [in2] "+r" (in2p)
+                :
+                : "q0", "q1", "q2", "memory");
+    n &= 15;
+    if (n)
+        MergeGeneric (outp, in1p, in2p, n);
+}
+#endif
+
 /*****************************************************************************
- * SendEvents: forward mouse and keyboard events to the parent p_vout
- *****************************************************************************/
-static int SendEvents( vlc_object_t *p_this, char const *psz_var,
-                       vlc_value_t oldval, vlc_value_t newval, void *_p_vout )
+ * RenderX: This algo works on a 8x8 block basic, it copies the top field
+ * and apply a process to recreate the bottom field :
+ *  If a 8x8 block is classified as :
+ *   - progressive: it applies a small blend (1,6,1)
+ *   - interlaced:
+ *    * in the MMX version: we do a ME between the 2 fields, if there is a
+ *    good match we use MC to recreate the bottom field (with a small
+ *    blend (1,6,1) )
+ *    * otherwise: it recreates the bottom field by an edge oriented
+ *    interpolation.
+  *****************************************************************************/
+
+/* XDeint8x8Detect: detect if a 8x8 block is interlaced.
+ * XXX: It need to access to 8x10
+ * We use more than 8 lines to help with scrolling (text)
+ * (and because XDeint8x8Frame use line 9)
+ * XXX: smooth/uniform area with noise detection doesn't works well
+ * but it's not really a problem because they don't have much details anyway
+ */
+static inline int ssd( int a ) { return a*a; }
+static inline int XDeint8x8DetectC( uint8_t *src, int i_src )
 {
-    vout_thread_t *p_vout = (vout_thread_t *)_p_vout;
-    vlc_value_t sentval = newval;
+    int y, x;
+    int ff, fr;
+    int fc;
 
-    if( !strcmp( psz_var, "mouse-y" ) )
+    /* Detect interlacing */
+    fc = 0;
+    for( y = 0; y < 7; y += 2 )
     {
-        switch( p_vout->p_sys->i_mode )
+        ff = fr = 0;
+        for( x = 0; x < 8; x++ )
         {
-            case DEINTERLACE_MEAN:
-            case DEINTERLACE_DISCARD:
-                sentval.i_int *= 2;
-                break;
+            fr += ssd(src[      x] - src[1*i_src+x]) +
+                  ssd(src[i_src+x] - src[2*i_src+x]);
+            ff += ssd(src[      x] - src[2*i_src+x]) +
+                  ssd(src[i_src+x] - src[3*i_src+x]);
         }
+        if( ff < 6*fr/8 && fr > 32 )
+            fc++;
+
+        src += 2*i_src;
     }
 
-    var_Set( p_vout, psz_var, sentval );
+    return fc < 1 ? false : true;
+}
+#ifdef CAN_COMPILE_MMXEXT
+static inline int XDeint8x8DetectMMXEXT( uint8_t *src, int i_src )
+{
 
-    return VLC_SUCCESS;
+    int y, x;
+    int32_t ff, fr;
+    int fc;
+
+    /* Detect interlacing */
+    fc = 0;
+    pxor_r2r( mm7, mm7 );
+    for( y = 0; y < 9; y += 2 )
+    {
+        ff = fr = 0;
+        pxor_r2r( mm5, mm5 );
+        pxor_r2r( mm6, mm6 );
+        for( x = 0; x < 8; x+=4 )
+        {
+            movd_m2r( src[        x], mm0 );
+            movd_m2r( src[1*i_src+x], mm1 );
+            movd_m2r( src[2*i_src+x], mm2 );
+            movd_m2r( src[3*i_src+x], mm3 );
+
+            punpcklbw_r2r( mm7, mm0 );
+            punpcklbw_r2r( mm7, mm1 );
+            punpcklbw_r2r( mm7, mm2 );
+            punpcklbw_r2r( mm7, mm3 );
+
+            movq_r2r( mm0, mm4 );
+
+            psubw_r2r( mm1, mm0 );
+            psubw_r2r( mm2, mm4 );
+
+            psubw_r2r( mm1, mm2 );
+            psubw_r2r( mm1, mm3 );
+
+            pmaddwd_r2r( mm0, mm0 );
+            pmaddwd_r2r( mm4, mm4 );
+            pmaddwd_r2r( mm2, mm2 );
+            pmaddwd_r2r( mm3, mm3 );
+            paddd_r2r( mm0, mm2 );
+            paddd_r2r( mm4, mm3 );
+            paddd_r2r( mm2, mm5 );
+            paddd_r2r( mm3, mm6 );
+        }
+
+        movq_r2r( mm5, mm0 );
+        psrlq_i2r( 32, mm0 );
+        paddd_r2r( mm0, mm5 );
+        movd_r2m( mm5, fr );
+
+        movq_r2r( mm6, mm0 );
+        psrlq_i2r( 32, mm0 );
+        paddd_r2r( mm0, mm6 );
+        movd_r2m( mm6, ff );
+
+        if( ff < 6*fr/8 && fr > 32 )
+            fc++;
+
+        src += 2*i_src;
+    }
+    return fc;
+}
+#endif
+
+static inline void XDeint8x8MergeC( uint8_t *dst, int i_dst,
+                                    uint8_t *src1, int i_src1,
+                                    uint8_t *src2, int i_src2 )
+{
+    int y, x;
+
+    /* Progressive */
+    for( y = 0; y < 8; y += 2 )
+    {
+        memcpy( dst, src1, 8 );
+        dst  += i_dst;
+
+        for( x = 0; x < 8; x++ )
+            dst[x] = (src1[x] + 6*src2[x] + src1[i_src1+x] + 4 ) >> 3;
+        dst += i_dst;
+
+        src1 += i_src1;
+        src2 += i_src2;
+    }
+}
+
+#ifdef CAN_COMPILE_MMXEXT
+static inline void XDeint8x8MergeMMXEXT( uint8_t *dst, int i_dst,
+                                         uint8_t *src1, int i_src1,
+                                         uint8_t *src2, int i_src2 )
+{
+    static const uint64_t m_4 = INT64_C(0x0004000400040004);
+    int y, x;
+
+    /* Progressive */
+    pxor_r2r( mm7, mm7 );
+    for( y = 0; y < 8; y += 2 )
+    {
+        for( x = 0; x < 8; x +=4 )
+        {
+            movd_m2r( src1[x], mm0 );
+            movd_r2m( mm0, dst[x] );
+
+            movd_m2r( src2[x], mm1 );
+            movd_m2r( src1[i_src1+x], mm2 );
+
+            punpcklbw_r2r( mm7, mm0 );
+            punpcklbw_r2r( mm7, mm1 );
+            punpcklbw_r2r( mm7, mm2 );
+            paddw_r2r( mm1, mm1 );
+            movq_r2r( mm1, mm3 );
+            paddw_r2r( mm3, mm3 );
+            paddw_r2r( mm2, mm0 );
+            paddw_r2r( mm3, mm1 );
+            paddw_m2r( m_4, mm1 );
+            paddw_r2r( mm1, mm0 );
+            psraw_i2r( 3, mm0 );
+            packuswb_r2r( mm7, mm0 );
+            movd_r2m( mm0, dst[i_dst+x] );
+        }
+        dst += 2*i_dst;
+        src1 += i_src1;
+        src2 += i_src2;
+    }
+}
+
+#endif
+
+/* For debug */
+static inline void XDeint8x8Set( uint8_t *dst, int i_dst, uint8_t v )
+{
+    int y;
+    for( y = 0; y < 8; y++ )
+        memset( &dst[y*i_dst], v, 8 );
+}
+
+/* XDeint8x8FieldE: Stupid deinterlacing (1,0,1) for block that miss a
+ * neighbour
+ * (Use 8x9 pixels)
+ * TODO: a better one for the inner part.
+ */
+static inline void XDeint8x8FieldEC( uint8_t *dst, int i_dst,
+                                     uint8_t *src, int i_src )
+{
+    int y, x;
+
+    /* Interlaced */
+    for( y = 0; y < 8; y += 2 )
+    {
+        memcpy( dst, src, 8 );
+        dst += i_dst;
+
+        for( x = 0; x < 8; x++ )
+            dst[x] = (src[x] + src[2*i_src+x] ) >> 1;
+        dst += 1*i_dst;
+        src += 2*i_src;
+    }
+}
+#ifdef CAN_COMPILE_MMXEXT
+static inline void XDeint8x8FieldEMMXEXT( uint8_t *dst, int i_dst,
+                                          uint8_t *src, int i_src )
+{
+    int y;
+
+    /* Interlaced */
+    for( y = 0; y < 8; y += 2 )
+    {
+        movq_m2r( src[0], mm0 );
+        movq_r2m( mm0, dst[0] );
+        dst += i_dst;
+
+        movq_m2r( src[2*i_src], mm1 );
+        pavgb_r2r( mm1, mm0 );
+
+        movq_r2m( mm0, dst[0] );
+
+        dst += 1*i_dst;
+        src += 2*i_src;
+    }
+}
+#endif
+
+/* XDeint8x8Field: Edge oriented interpolation
+ * (Need -4 and +5 pixels H, +1 line)
+ */
+static inline void XDeint8x8FieldC( uint8_t *dst, int i_dst,
+                                    uint8_t *src, int i_src )
+{
+    int y, x;
+
+    /* Interlaced */
+    for( y = 0; y < 8; y += 2 )
+    {
+        memcpy( dst, src, 8 );
+        dst += i_dst;
+
+        for( x = 0; x < 8; x++ )
+        {
+            uint8_t *src2 = &src[2*i_src];
+            /* I use 8 pixels just to match the MMX version, but it's overkill
+             * 5 would be enough (less isn't good) */
+            const int c0 = abs(src[x-4]-src2[x-2]) + abs(src[x-3]-src2[x-1]) +
+                           abs(src[x-2]-src2[x+0]) + abs(src[x-1]-src2[x+1]) +
+                           abs(src[x+0]-src2[x+2]) + abs(src[x+1]-src2[x+3]) +
+                           abs(src[x+2]-src2[x+4]) + abs(src[x+3]-src2[x+5]);
+
+            const int c1 = abs(src[x-3]-src2[x-3]) + abs(src[x-2]-src2[x-2]) +
+                           abs(src[x-1]-src2[x-1]) + abs(src[x+0]-src2[x+0]) +
+                           abs(src[x+1]-src2[x+1]) + abs(src[x+2]-src2[x+2]) +
+                           abs(src[x+3]-src2[x+3]) + abs(src[x+4]-src2[x+4]);
+
+            const int c2 = abs(src[x-2]-src2[x-4]) + abs(src[x-1]-src2[x-3]) +
+                           abs(src[x+0]-src2[x-2]) + abs(src[x+1]-src2[x-1]) +
+                           abs(src[x+2]-src2[x+0]) + abs(src[x+3]-src2[x+1]) +
+                           abs(src[x+4]-src2[x+2]) + abs(src[x+5]-src2[x+3]);
+
+            if( c0 < c1 && c1 <= c2 )
+                dst[x] = (src[x-1] + src2[x+1]) >> 1;
+            else if( c2 < c1 && c1 <= c0 )
+                dst[x] = (src[x+1] + src2[x-1]) >> 1;
+            else
+                dst[x] = (src[x+0] + src2[x+0]) >> 1;
+        }
+
+        dst += 1*i_dst;
+        src += 2*i_src;
+    }
+}
+#ifdef CAN_COMPILE_MMXEXT
+static inline void XDeint8x8FieldMMXEXT( uint8_t *dst, int i_dst,
+                                         uint8_t *src, int i_src )
+{
+    int y, x;
+
+    /* Interlaced */
+    for( y = 0; y < 8; y += 2 )
+    {
+        memcpy( dst, src, 8 );
+        dst += i_dst;
+
+        for( x = 0; x < 8; x++ )
+        {
+            uint8_t *src2 = &src[2*i_src];
+            int32_t c0, c1, c2;
+
+            movq_m2r( src[x-2], mm0 );
+            movq_m2r( src[x-3], mm1 );
+            movq_m2r( src[x-4], mm2 );
+
+            psadbw_m2r( src2[x-4], mm0 );
+            psadbw_m2r( src2[x-3], mm1 );
+            psadbw_m2r( src2[x-2], mm2 );
+
+            movd_r2m( mm0, c2 );
+            movd_r2m( mm1, c1 );
+            movd_r2m( mm2, c0 );
+
+            if( c0 < c1 && c1 <= c2 )
+                dst[x] = (src[x-1] + src2[x+1]) >> 1;
+            else if( c2 < c1 && c1 <= c0 )
+                dst[x] = (src[x+1] + src2[x-1]) >> 1;
+            else
+                dst[x] = (src[x+0] + src2[x+0]) >> 1;
+        }
+
+        dst += 1*i_dst;
+        src += 2*i_src;
+    }
+}
+#endif
+
+/* NxN arbitray size (and then only use pixel in the NxN block)
+ */
+static inline int XDeintNxNDetect( uint8_t *src, int i_src,
+                                   int i_height, int i_width )
+{
+    int y, x;
+    int ff, fr;
+    int fc;
+
+
+    /* Detect interlacing */
+    /* FIXME way too simple, need to be more like XDeint8x8Detect */
+    ff = fr = 0;
+    fc = 0;
+    for( y = 0; y < i_height - 2; y += 2 )
+    {
+        const uint8_t *s = &src[y*i_src];
+        for( x = 0; x < i_width; x++ )
+        {
+            fr += ssd(s[      x] - s[1*i_src+x]);
+            ff += ssd(s[      x] - s[2*i_src+x]);
+        }
+        if( ff < fr && fr > i_width / 2 )
+            fc++;
+    }
+
+    return fc < 2 ? false : true;
+}
+
+static inline void XDeintNxNFrame( uint8_t *dst, int i_dst,
+                                   uint8_t *src, int i_src,
+                                   int i_width, int i_height )
+{
+    int y, x;
+
+    /* Progressive */
+    for( y = 0; y < i_height; y += 2 )
+    {
+        memcpy( dst, src, i_width );
+        dst += i_dst;
+
+        if( y < i_height - 2 )
+        {
+            for( x = 0; x < i_width; x++ )
+                dst[x] = (src[x] + 2*src[1*i_src+x] + src[2*i_src+x] + 2 ) >> 2;
+        }
+        else
+        {
+            /* Blend last line */
+            for( x = 0; x < i_width; x++ )
+                dst[x] = (src[x] + src[1*i_src+x] ) >> 1;
+        }
+        dst += 1*i_dst;
+        src += 2*i_src;
+    }
+}
+
+static inline void XDeintNxNField( uint8_t *dst, int i_dst,
+                                   uint8_t *src, int i_src,
+                                   int i_width, int i_height )
+{
+    int y, x;
+
+    /* Interlaced */
+    for( y = 0; y < i_height; y += 2 )
+    {
+        memcpy( dst, src, i_width );
+        dst += i_dst;
+
+        if( y < i_height - 2 )
+        {
+            for( x = 0; x < i_width; x++ )
+                dst[x] = (src[x] + src[2*i_src+x] ) >> 1;
+        }
+        else
+        {
+            /* Blend last line */
+            for( x = 0; x < i_width; x++ )
+                dst[x] = (src[x] + src[i_src+x]) >> 1;
+        }
+        dst += 1*i_dst;
+        src += 2*i_src;
+    }
+}
+
+static inline void XDeintNxN( uint8_t *dst, int i_dst, uint8_t *src, int i_src,
+                              int i_width, int i_height )
+{
+    if( XDeintNxNDetect( src, i_src, i_width, i_height ) )
+        XDeintNxNField( dst, i_dst, src, i_src, i_width, i_height );
+    else
+        XDeintNxNFrame( dst, i_dst, src, i_src, i_width, i_height );
+}
+
+
+static inline int median( int a, int b, int c )
+{
+    int min = a, max =a;
+    if( b < min )
+        min = b;
+    else
+        max = b;
+
+    if( c < min )
+        min = c;
+    else if( c > max )
+        max = c;
+
+    return a + b + c - min - max;
+}
+
+
+/* XDeintBand8x8:
+ */
+static inline void XDeintBand8x8C( uint8_t *dst, int i_dst,
+                                   uint8_t *src, int i_src,
+                                   const int i_mbx, int i_modx )
+{
+    int x;
+
+    for( x = 0; x < i_mbx; x++ )
+    {
+        int s;
+        if( ( s = XDeint8x8DetectC( src, i_src ) ) )
+        {
+            if( x == 0 || x == i_mbx - 1 )
+                XDeint8x8FieldEC( dst, i_dst, src, i_src );
+            else
+                XDeint8x8FieldC( dst, i_dst, src, i_src );
+        }
+        else
+        {
+            XDeint8x8MergeC( dst, i_dst,
+                             &src[0*i_src], 2*i_src,
+                             &src[1*i_src], 2*i_src );
+        }
+
+        dst += 8;
+        src += 8;
+    }
+
+    if( i_modx )
+        XDeintNxN( dst, i_dst, src, i_src, i_modx, 8 );
+}
+#ifdef CAN_COMPILE_MMXEXT
+static inline void XDeintBand8x8MMXEXT( uint8_t *dst, int i_dst,
+                                        uint8_t *src, int i_src,
+                                        const int i_mbx, int i_modx )
+{
+    int x;
+
+    /* Reset current line */
+    for( x = 0; x < i_mbx; x++ )
+    {
+        int s;
+        if( ( s = XDeint8x8DetectMMXEXT( src, i_src ) ) )
+        {
+            if( x == 0 || x == i_mbx - 1 )
+                XDeint8x8FieldEMMXEXT( dst, i_dst, src, i_src );
+            else
+                XDeint8x8FieldMMXEXT( dst, i_dst, src, i_src );
+        }
+        else
+        {
+            XDeint8x8MergeMMXEXT( dst, i_dst,
+                                  &src[0*i_src], 2*i_src,
+                                  &src[1*i_src], 2*i_src );
+        }
+
+        dst += 8;
+        src += 8;
+    }
+
+    if( i_modx )
+        XDeintNxN( dst, i_dst, src, i_src, i_modx, 8 );
+}
+#endif
+
+static void RenderX( picture_t *p_outpic, picture_t *p_pic )
+{
+    int i_plane;
+
+    /* Copy image and skip lines */
+    for( i_plane = 0 ; i_plane < p_pic->i_planes ; i_plane++ )
+    {
+        const int i_mby = ( p_outpic->p[i_plane].i_visible_lines + 7 )/8 - 1;
+        const int i_mbx = p_outpic->p[i_plane].i_visible_pitch/8;
+
+        const int i_mody = p_outpic->p[i_plane].i_visible_lines - 8*i_mby;
+        const int i_modx = p_outpic->p[i_plane].i_visible_pitch - 8*i_mbx;
+
+        const int i_dst = p_outpic->p[i_plane].i_pitch;
+        const int i_src = p_pic->p[i_plane].i_pitch;
+
+        int y, x;
+
+        for( y = 0; y < i_mby; y++ )
+        {
+            uint8_t *dst = &p_outpic->p[i_plane].p_pixels[8*y*i_dst];
+            uint8_t *src = &p_pic->p[i_plane].p_pixels[8*y*i_src];
+
+#ifdef CAN_COMPILE_MMXEXT
+            if( vlc_CPU() & CPU_CAPABILITY_MMXEXT )
+                XDeintBand8x8MMXEXT( dst, i_dst, src, i_src, i_mbx, i_modx );
+            else
+#endif
+                XDeintBand8x8C( dst, i_dst, src, i_src, i_mbx, i_modx );
+        }
+
+        /* Last line (C only)*/
+        if( i_mody )
+        {
+            uint8_t *dst = &p_outpic->p[i_plane].p_pixels[8*y*i_dst];
+            uint8_t *src = &p_pic->p[i_plane].p_pixels[8*y*i_src];
+
+            for( x = 0; x < i_mbx; x++ )
+            {
+                XDeintNxN( dst, i_dst, src, i_src, 8, i_mody );
+
+                dst += 8;
+                src += 8;
+            }
+
+            if( i_modx )
+                XDeintNxN( dst, i_dst, src, i_src, i_modx, i_mody );
+        }
+    }
+
+#ifdef CAN_COMPILE_MMXEXT
+    if( vlc_CPU() & CPU_CAPABILITY_MMXEXT )
+        emms();
+#endif
 }
 
 /*****************************************************************************
- * FilterCallback: called when changing the deinterlace method on the fly.
+ * Yadif (Yet Another DeInterlacing Filter).
  *****************************************************************************/
-static int FilterCallback( vlc_object_t *p_this, char const *psz_cmd,
-                           vlc_value_t oldval, vlc_value_t newval,
-                           void *p_data )
+/* */
+struct vf_priv_s {
+    /*
+     * 0: Output 1 frame for each frame.
+     * 1: Output 1 frame for each field.
+     * 2: Like 0 but skips spatial interlacing check.
+     * 3: Like 1 but skips spatial interlacing check.
+     *
+     * In vlc, only & 0x02 has meaning, as we do the & 0x01 ourself.
+     */
+    int mode;
+};
+
+/* I am unsure it is the right one */
+typedef intptr_t x86_reg;
+
+#define FFABS(a) ((a) >= 0 ? (a) : (-(a)))
+#define FFMAX(a,b)      __MAX(a,b)
+#define FFMAX3(a,b,c)   FFMAX(FFMAX(a,b),c)
+#define FFMIN(a,b)      __MIN(a,b)
+#define FFMIN3(a,b,c)   FFMIN(FFMIN(a,b),c)
+
+/* yadif.h comes from vf_yadif.c of mplayer project */
+#include "yadif.h"
+
+static void RenderYadif( filter_t *p_filter, picture_t *p_dst, picture_t *p_src, int i_order, int i_field )
 {
-    vout_thread_t * p_vout = (vout_thread_t *)p_this;
-    int i_old_mode = p_vout->p_sys->i_mode;
+    filter_sys_t *p_sys = p_filter->p_sys;
 
-    msg_Dbg( p_vout, "using %s deinterlace mode", newval.psz_string );
+    /* */
+    assert( i_order == 0 || i_order == 1 );
+    assert( i_field == 0 || i_field == 1 );
 
-    vlc_mutex_lock( &p_vout->p_sys->filter_lock );
+    if( i_order == 0 )
+    {
+        /* Duplicate the picture
+         * TODO when the vout rework is finished, picture_Hold() might be enough
+         * but becarefull, the pitches must match */
+        picture_t *p_dup = picture_NewFromFormat( &p_src->format );
+        if( p_dup )
+            picture_Copy( p_dup, p_src );
+
+        /* Slide the history */
+        if( p_sys->pp_history[0] )
+            picture_Release( p_sys->pp_history[0]  );
+        for( int i = 1; i < HISTORY_SIZE; i++ )
+            p_sys->pp_history[i-1] = p_sys->pp_history[i];
+        p_sys->pp_history[HISTORY_SIZE-1] = p_dup;
+    }
 
-    SetFilterMethod( p_vout, newval.psz_string );
+    /* As the pitches must match, use ONLY pictures coming from picture_New()! */
+    picture_t *p_prev = p_sys->pp_history[0];
+    picture_t *p_cur  = p_sys->pp_history[1];
+    picture_t *p_next = p_sys->pp_history[2];
 
-    switch( p_vout->render.i_chroma )
+    /* Filter if we have all the pictures we need */
+    if( p_prev && p_cur && p_next )
     {
-    case VLC_FOURCC('I','4','2','2'):
-        vlc_mutex_unlock( &p_vout->p_sys->filter_lock );
-        return VLC_SUCCESS;
-        break;
-
-    case VLC_FOURCC('I','4','2','0'):
-    case VLC_FOURCC('I','Y','U','V'):
-    case VLC_FOURCC('Y','V','1','2'):
-        switch( p_vout->p_sys->i_mode )
+        /* */
+        void (*filter)(struct vf_priv_s *p, uint8_t *dst, uint8_t *prev, uint8_t *cur, uint8_t *next, int w, int refs, int parity);
+#if defined(HAVE_YADIF_SSE2)
+        if( vlc_CPU() & CPU_CAPABILITY_SSE2 )
+            filter = yadif_filter_line_mmx2;
+        else
+#endif
+            filter = yadif_filter_line_c;
+
+        for( int n = 0; n < p_dst->i_planes; n++ )
         {
-        case DEINTERLACE_MEAN:
-        case DEINTERLACE_DISCARD:
-            if( ( i_old_mode == DEINTERLACE_MEAN )
-                || ( i_old_mode == DEINTERLACE_DISCARD ) )
+            const plane_t *prevp = &p_prev->p[n];
+            const plane_t *curp  = &p_cur->p[n];
+            const plane_t *nextp = &p_next->p[n];
+            plane_t *dstp        = &p_dst->p[n];
+
+            for( int y = 1; y < dstp->i_visible_lines - 1; y++ )
             {
-                vlc_mutex_unlock( &p_vout->p_sys->filter_lock );
-                return VLC_SUCCESS;
+                if( (y % 2) == i_field )
+                {
+                    vlc_memcpy( &dstp->p_pixels[y * dstp->i_pitch],
+                                &curp->p_pixels[y * curp->i_pitch], dstp->i_visible_pitch );
+                }
+                else
+                {
+                    struct vf_priv_s cfg;
+                    /* Spatial checks only when enough data */
+                    cfg.mode = (y >= 2 && y < dstp->i_visible_lines - 2) ? 0 : 2;
+
+                    assert( prevp->i_pitch == curp->i_pitch && curp->i_pitch == nextp->i_pitch );
+                    filter( &cfg,
+                            &dstp->p_pixels[y * dstp->i_pitch],
+                            &prevp->p_pixels[y * prevp->i_pitch],
+                            &curp->p_pixels[y * curp->i_pitch],
+                            &nextp->p_pixels[y * nextp->i_pitch],
+                            dstp->i_visible_pitch,
+                            curp->i_pitch,
+                            (i_field ^ (i_order == i_field)) & 1 );
+                }
+
+                /* We duplicate the first and last lines */
+                if( y == 1 )
+                    vlc_memcpy(&dstp->p_pixels[(y-1) * dstp->i_pitch], &dstp->p_pixels[y * dstp->i_pitch], dstp->i_pitch);
+                else if( y == dstp->i_visible_lines - 2 )
+                    vlc_memcpy(&dstp->p_pixels[(y+1) * dstp->i_pitch], &dstp->p_pixels[y * dstp->i_pitch], dstp->i_pitch);
             }
+        }
+
+        /* */
+        p_dst->date = (p_next->date - p_cur->date) * i_order / 2 + p_cur->date;
+    }
+    else
+    {
+        /* Fallback to something simple
+         * XXX it is wrong when we have 2 pictures, we should not output a picture */
+        RenderX( p_dst, p_src );
+    }
+}
+
+/*****************************************************************************
+ * video filter2 functions
+ *****************************************************************************/
+static picture_t *Deinterlace( filter_t *p_filter, picture_t *p_pic )
+{
+    filter_sys_t *p_sys = p_filter->p_sys;
+    picture_t *p_pic_dst;
+
+    /* Request output picture */
+    p_pic_dst = filter_NewPicture( p_filter );
+    if( p_pic_dst == NULL )
+    {
+        picture_Release( p_pic );
+        return NULL;
+    }
+
+    switch( p_sys->i_mode )
+    {
+        case DEINTERLACE_DISCARD:
+            RenderDiscard( p_filter, p_pic_dst, p_pic, 0 );
             break;
 
         case DEINTERLACE_BOB:
-        case DEINTERLACE_BLEND:
+#if 0
+            RenderBob( p_filter, pp_outpic[0], p_pic, !p_pic->b_top_field_first );
+            RenderBob( p_filter, pp_outpic[1], p_pic, p_pic->b_top_field_first );
+            break;
+#endif
+
         case DEINTERLACE_LINEAR:
-            if( ( i_old_mode == DEINTERLACE_BOB )
-                || ( i_old_mode == DEINTERLACE_BLEND )
-                || ( i_old_mode == DEINTERLACE_LINEAR ) )
-            {
-                vlc_mutex_unlock( &p_vout->p_sys->filter_lock );
-                return VLC_SUCCESS;
-            }
+#if 0
+            RenderLinear( p_filter, pp_outpic[0], p_pic, !p_pic->b_top_field_first );
+            RenderLinear( p_filter, pp_outpic[1], p_pic, p_pic->b_top_field_first );
+#endif
+            msg_Err( p_filter, "doubling the frame rate is not supported yet" );
+            picture_Release( p_pic_dst );
+            picture_Release( p_pic );
+            return NULL;
+
+        case DEINTERLACE_MEAN:
+            RenderMean( p_filter, p_pic_dst, p_pic );
             break;
-        }
-        break;
 
-    default:
-        break;
+        case DEINTERLACE_BLEND:
+            RenderBlend( p_filter, p_pic_dst, p_pic );
+            break;
+
+        case DEINTERLACE_X:
+            RenderX( p_pic_dst, p_pic );
+            break;
+
+        case DEINTERLACE_YADIF:
+            msg_Err( p_filter, "delaying frames is not supported yet" );
+            //RenderYadif( p_vout, pp_outpic[0], p_pic, 0, 0 );
+            picture_Release( p_pic_dst );
+            picture_Release( p_pic );
+            return NULL;
+
+        case DEINTERLACE_YADIF2X:
+            msg_Err( p_filter, "doubling the frame rate is not supported yet" );
+            //RenderYadif( p_vout, pp_outpic[0], p_pic, 0, !p_pic->b_top_field_first );
+            //RenderYadif( p_vout, pp_outpic[1], p_pic, 1, p_pic->b_top_field_first );
+            picture_Release( p_pic_dst );
+            picture_Release( p_pic );
+            return NULL;
     }
 
-    /* We need to kill the old vout */
+    picture_CopyProperties( p_pic_dst, p_pic );
+    p_pic_dst->b_progressive = true;
+
+    picture_Release( p_pic );
+    return p_pic_dst;
+}
+
+static int Mouse( filter_t *p_filter,
+                  vlc_mouse_t *p_mouse, const vlc_mouse_t *p_old, const vlc_mouse_t *p_new )
+{
+    *p_mouse = *p_new;
+    if( p_filter->p_sys->b_half_height )
+        p_mouse->i_y *= 2;
+    return VLC_SUCCESS;
+}
 
-    DEL_CALLBACKS( p_vout->p_sys->p_vout, SendEvents );
 
-    vlc_object_detach( p_vout->p_sys->p_vout );
-    vout_Destroy( p_vout->p_sys->p_vout );
+/*****************************************************************************
+ * Open
+ *****************************************************************************/
+static int Open( vlc_object_t *p_this )
+{
+    filter_t *p_filter = (filter_t*)p_this;
+    filter_sys_t *p_sys;
 
-    /* Try to open a new video output */
-    p_vout->p_sys->p_vout = SpawnRealVout( p_vout );
+    if( !IsChromaSupported( p_filter->fmt_in.video.i_chroma ) )
+        return VLC_EGENERIC;
 
-    if( p_vout->p_sys->p_vout == NULL )
+    /* */
+    p_sys = p_filter->p_sys = malloc( sizeof( *p_sys ) );
+    if( !p_sys )
+        return VLC_ENOMEM;
+
+    p_sys->i_mode = DEINTERLACE_BLEND;
+    p_sys->b_double_rate = false;
+    p_sys->b_half_height = true;
+
+#if defined(CAN_COMPILE_C_ALTIVEC)
+    if( vlc_CPU() & CPU_CAPABILITY_ALTIVEC )
+    {
+        p_sys->pf_merge = MergeAltivec;
+        p_sys->pf_end_merge = NULL;
+    }
+    else
+#endif
+#if defined(CAN_COMPILE_SSE)
+    if( vlc_CPU() & CPU_CAPABILITY_SSE2 )
+    {
+        p_sys->pf_merge = MergeSSE2;
+        p_sys->pf_end_merge = EndMMX;
+    }
+    else
+#endif
+#if defined(CAN_COMPILE_MMXEXT)
+    if( vlc_CPU() & CPU_CAPABILITY_MMXEXT )
+    {
+        p_sys->pf_merge = MergeMMXEXT;
+        p_sys->pf_end_merge = EndMMX;
+    }
+    else
+#endif
+#if defined(CAN_COMPILE_3DNOW)
+    if( vlc_CPU() & CPU_CAPABILITY_3DNOW )
     {
-        /* Everything failed */
-        msg_Err( p_vout, "cannot open vout, aborting" );
+        p_sys->pf_merge = Merge3DNow;
+        p_sys->pf_end_merge = End3DNow;
+    }
+    else
+#endif
+#if defined __ARM_NEON__
+    if( vlc_CPU() & CPU_CAPABILITY_NEON )
+    {
+        p_sys->pf_merge = MergeNEON;
+        p_sys->pf_end_merge = NULL;
+    }
+    else
+#endif
+    {
+        p_sys->pf_merge = MergeGeneric;
+        p_sys->pf_end_merge = NULL;
+    }
+
+    /* */
+    config_ChainParse( p_filter, FILTER_CFG_PREFIX, ppsz_filter_options,
+                       p_filter->p_cfg );
 
-        vlc_mutex_unlock( &p_vout->p_sys->filter_lock );
+    char *psz_mode = var_GetNonEmptyString( p_filter, FILTER_CFG_PREFIX "mode" );
+    SetFilterMethod( p_filter, psz_mode, p_filter->fmt_in.video.i_chroma );
+    free( psz_mode );
+
+    /* */
+    video_format_t fmt;
+    GetOutputFormat( p_filter, &fmt, &p_filter->fmt_in.video );
+    if( !p_filter->b_allow_fmt_out_change &&
+        ( fmt.i_chroma != p_filter->fmt_in.video.i_chroma ||
+          fmt.i_height != p_filter->fmt_in.video.i_height ) )
+    {
+        Close( VLC_OBJECT(p_filter) );
         return VLC_EGENERIC;
     }
+    p_filter->fmt_out.video = fmt;
+    p_filter->fmt_out.i_codec = fmt.i_chroma;
+    p_filter->pf_video_filter = Deinterlace;
+    p_filter->pf_video_mouse  = Mouse;
 
-    ADD_CALLBACKS( p_vout->p_sys->p_vout, SendEvents );
+    msg_Dbg( p_filter, "deinterlacing" );
 
-    vlc_mutex_unlock( &p_vout->p_sys->filter_lock );
     return VLC_SUCCESS;
 }
 
 /*****************************************************************************
- * SendEventsToChild: forward events to the child/children vout
+ * Close: clean up the filter
  *****************************************************************************/
-static int SendEventsToChild( vlc_object_t *p_this, char const *psz_var,
-                       vlc_value_t oldval, vlc_value_t newval, void *p_data )
+static void Close( vlc_object_t *p_this )
 {
-    vout_thread_t *p_vout = (vout_thread_t *)p_this;
-    var_Set( p_vout->p_sys->p_vout, psz_var, newval );
-    return VLC_SUCCESS;
+    filter_t *p_filter = (filter_t*)p_this;
+
+    free( p_filter->p_sys );
 }
+