]> git.sesse.net Git - vlc/blobdiff - modules/video_chroma/i420_rgb16.c
i420_rgb: clobber lists for MMX and SSE2
[vlc] / modules / video_chroma / i420_rgb16.c
index 8e04c3856811144b4fe9ce97bba4332398b2c811..7a2a3fed5150fd01466720d9d8c6db18d9dc6b8d 100644 (file)
@@ -11,7 +11,7 @@
  * it under the terms of the GNU General Public License as published by
  * the Free Software Foundation; either version 2 of the License, or
  * (at your option) any later version.
- * 
+ *
  * This program is distributed in the hope that it will be useful,
  * but WITHOUT ANY WARRANTY; without even the implied warranty of
  * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
 /*****************************************************************************
  * Preamble
  *****************************************************************************/
-#include <string.h>                                            /* strerror() */
 
-#include <vlc/vlc.h>
-#include <vlc_vout.h>
+#ifdef HAVE_CONFIG_H
+# include "config.h"
+#endif
+
+#include <vlc_common.h>
+#include <vlc_filter.h>
+#include <vlc_cpu.h>
 
 #include "i420_rgb.h"
 #if defined (MODULE_NAME_IS_i420_rgb)
 #   include "i420_rgb_c.h"
+#   define VLC_TARGET
 #elif defined (MODULE_NAME_IS_i420_rgb_mmx)
-#   include "i420_rgb_mmx.h"
+#   include "../mmx/i420_rgb_mmx.h"
+#   define VLC_TARGET VLC_MMX
 #elif defined (MODULE_NAME_IS_i420_rgb_sse2)
-#   include "i420_rgb_mmx.h"
+#   include "../mmx/i420_rgb_mmx.h"
+#   define VLC_TARGET VLC_SSE
 #endif
 
-static void SetOffset( int, int, int, int, vlc_bool_t *,
+static void SetOffset( int, int, int, int, bool *,
                        unsigned int *, int * );
 
 #if defined (MODULE_NAME_IS_i420_rgb)
@@ -53,8 +60,8 @@ static void SetOffset( int, int, int, int, vlc_bool_t *,
  *  - input: 2 lines (2 Y lines, 1 U/V line)
  *  - output: 1 line
  *****************************************************************************/
-void E_(I420_RGB16_dither)( vout_thread_t *p_vout, picture_t *p_src,
-                                                      picture_t *p_dest )
+void I420_RGB16_dither( filter_t *p_filter, picture_t *p_src,
+                                                picture_t *p_dest )
 {
     /* We got this one from the old arguments */
     uint16_t *p_pic = (uint16_t*)p_dest->p->p_pixels;
@@ -62,7 +69,7 @@ void E_(I420_RGB16_dither)( vout_thread_t *p_vout, picture_t *p_src,
     uint8_t  *p_u   = p_src->U_PIXELS;
     uint8_t  *p_v   = p_src->V_PIXELS;
 
-    vlc_bool_t   b_hscale;                        /* horizontal scaling type */
+    bool   b_hscale;                        /* horizontal scaling type */
     unsigned int i_vscale;                          /* vertical scaling type */
     unsigned int i_x, i_y;                /* horizontal and vertical indexes */
     unsigned int i_real_y;                                          /* y % 4 */
@@ -70,19 +77,19 @@ void E_(I420_RGB16_dither)( vout_thread_t *p_vout, picture_t *p_src,
     int         i_right_margin;
     int         i_rewind;
     int         i_scale_count;                       /* scale modulo counter */
-    int         i_chroma_width = p_vout->render.i_width / 2; /* chroma width */
+    int         i_chroma_width = p_filter->fmt_in.video.i_width / 2; /* chroma width */
     uint16_t *  p_pic_start;       /* beginning of the current line for copy */
     int         i_uval, i_vval;                           /* U and V samples */
     int         i_red, i_green, i_blue;          /* U and V modified samples */
-    uint16_t *  p_yuv = p_vout->chroma.p_sys->p_rgb16;
+    uint16_t *  p_yuv = p_filter->p_sys->p_rgb16;
     uint16_t *  p_ybase;                     /* Y dependant conversion table */
 
     /* Conversion buffer pointer */
-    uint16_t *  p_buffer_start = (uint16_t*)p_vout->chroma.p_sys->p_buffer;
+    uint16_t *  p_buffer_start = (uint16_t*)p_filter->p_sys->p_buffer;
     uint16_t *  p_buffer;
 
     /* Offset array pointer */
-    int *       p_offset_start = p_vout->chroma.p_sys->p_offset;
+    int *       p_offset_start = p_filter->p_sys->p_offset;
     int *       p_offset;
 
     const int i_source_margin = p_src->p[0].i_pitch
@@ -98,42 +105,37 @@ void E_(I420_RGB16_dither)( vout_thread_t *p_vout, picture_t *p_src,
 
     for(i_x = 0; i_x < 4; i_x++)
     {
-        dither10[i_x] = dither10[i_x] << (SHIFT - 4 + p_vout->output.i_rrshift);
-        dither11[i_x] = dither11[i_x] << (SHIFT - 4 + p_vout->output.i_rrshift);
-        dither12[i_x] = dither12[i_x] << (SHIFT - 4 + p_vout->output.i_rrshift);
-        dither13[i_x] = dither13[i_x] << (SHIFT - 4 + p_vout->output.i_rrshift);
+        dither10[i_x] = dither10[i_x] << (SHIFT - 4 + p_filter->fmt_out.video.i_rrshift);
+        dither11[i_x] = dither11[i_x] << (SHIFT - 4 + p_filter->fmt_out.video.i_rrshift);
+        dither12[i_x] = dither12[i_x] << (SHIFT - 4 + p_filter->fmt_out.video.i_rrshift);
+        dither13[i_x] = dither13[i_x] << (SHIFT - 4 + p_filter->fmt_out.video.i_rrshift);
     }
 
     i_right_margin = p_dest->p->i_pitch - p_dest->p->i_visible_pitch;
-
-    if( p_vout->render.i_width & 7 )
-    {
-        i_rewind = 8 - ( p_vout->render.i_width & 7 );
-    }
-    else
-    {
-        i_rewind = 0;
-    }
+    i_rewind = (-p_filter->fmt_in.video.i_width) & 7;
 
     /* Rule: when a picture of size (x1,y1) with aspect ratio r1 is rendered
      * on a picture of size (x2,y2) with aspect ratio r2, if x1 grows to x1'
      * then y1 grows to y1' = x1' * y2/x2 * r2/r1 */
-    SetOffset( p_vout->render.i_width, p_vout->render.i_height,
-               p_vout->output.i_width, p_vout->output.i_height,
+    SetOffset( p_filter->fmt_in.video.i_width,
+               p_filter->fmt_in.video.i_height,
+               p_filter->fmt_out.video.i_width,
+               p_filter->fmt_out.video.i_height,
                &b_hscale, &i_vscale, p_offset_start );
 
     /*
      * Perform conversion
      */
     i_scale_count = ( i_vscale == 1 ) ?
-                    p_vout->output.i_height : p_vout->render.i_height;
-    for( i_y = 0; i_y < p_vout->render.i_height; i_y++ )
+                    p_filter->fmt_out.video.i_height :
+                    p_filter->fmt_in.video.i_height;
+    for( i_y = 0; i_y < p_filter->fmt_in.video.i_height; i_y++ )
     {
         i_real_y = i_y & 0x3;
         p_pic_start = p_pic;
         p_buffer = b_hscale ? p_buffer_start : p_pic;
 
-        for ( i_x = p_vout->render.i_width / 8; i_x--; )
+        for ( i_x = p_filter->fmt_in.video.i_width / 8; i_x--; )
         {
             int *p_dither = dither10;
             CONVERT_YUV_PIXEL_DITHER(2);
@@ -204,8 +206,7 @@ void E_(I420_RGB16_dither)( vout_thread_t *p_vout, picture_t *p_src,
 
 #if defined (MODULE_NAME_IS_i420_rgb)
 
-void E_(I420_RGB16)( vout_thread_t *p_vout, picture_t *p_src,
-                                            picture_t *p_dest )
+void I420_RGB16( filter_t *p_filter, picture_t *p_src, picture_t *p_dest )
 {
     /* We got this one from the old arguments */
     uint16_t *p_pic = (uint16_t*)p_dest->p->p_pixels;
@@ -213,26 +214,26 @@ void E_(I420_RGB16)( vout_thread_t *p_vout, picture_t *p_src,
     uint8_t  *p_u   = p_src->U_PIXELS;
     uint8_t  *p_v   = p_src->V_PIXELS;
 
-    vlc_bool_t  b_hscale;                         /* horizontal scaling type */
+    bool  b_hscale;                         /* horizontal scaling type */
     unsigned int i_vscale;                          /* vertical scaling type */
     unsigned int i_x, i_y;                /* horizontal and vertical indexes */
 
     int         i_right_margin;
     int         i_rewind;
     int         i_scale_count;                       /* scale modulo counter */
-    int         i_chroma_width = p_vout->render.i_width / 2; /* chroma width */
+    int         i_chroma_width = p_filter->fmt_in.video.i_width / 2; /* chroma width */
     uint16_t *  p_pic_start;       /* beginning of the current line for copy */
     int         i_uval, i_vval;                           /* U and V samples */
     int         i_red, i_green, i_blue;          /* U and V modified samples */
-    uint16_t *  p_yuv = p_vout->chroma.p_sys->p_rgb16;
+    uint16_t *  p_yuv = p_filter->p_sys->p_rgb16;
     uint16_t *  p_ybase;                     /* Y dependant conversion table */
 
     /* Conversion buffer pointer */
-    uint16_t *  p_buffer_start = (uint16_t*)p_vout->chroma.p_sys->p_buffer;
+    uint16_t *  p_buffer_start = (uint16_t*)p_filter->p_sys->p_buffer;
     uint16_t *  p_buffer;
 
     /* Offset array pointer */
-    int *       p_offset_start = p_vout->chroma.p_sys->p_offset;
+    int *       p_offset_start = p_filter->p_sys->p_offset;
     int *       p_offset;
 
     const int i_source_margin = p_src->p[0].i_pitch
@@ -241,34 +242,29 @@ void E_(I420_RGB16)( vout_thread_t *p_vout, picture_t *p_src,
                                  - p_src->p[1].i_visible_pitch;
 
     i_right_margin = p_dest->p->i_pitch - p_dest->p->i_visible_pitch;
-
-    if( p_vout->render.i_width & 7 )
-    {
-        i_rewind = 8 - ( p_vout->render.i_width & 7 );
-    }
-    else
-    {
-        i_rewind = 0;
-    }
+    i_rewind = (-p_filter->fmt_in.video.i_width) & 7;
 
     /* Rule: when a picture of size (x1,y1) with aspect ratio r1 is rendered
      * on a picture of size (x2,y2) with aspect ratio r2, if x1 grows to x1'
      * then y1 grows to y1' = x1' * y2/x2 * r2/r1 */
-    SetOffset( p_vout->render.i_width, p_vout->render.i_height,
-               p_vout->output.i_width, p_vout->output.i_height,
+    SetOffset( p_filter->fmt_in.video.i_width,
+               p_filter->fmt_in.video.i_height,
+               p_filter->fmt_out.video.i_width,
+               p_filter->fmt_out.video.i_height,
                &b_hscale, &i_vscale, p_offset_start );
 
     /*
      * Perform conversion
      */
     i_scale_count = ( i_vscale == 1 ) ?
-                    p_vout->output.i_height : p_vout->render.i_height;
-    for( i_y = 0; i_y < p_vout->render.i_height; i_y++ )
+                    p_filter->fmt_out.video.i_height :
+                    p_filter->fmt_in.video.i_height;
+    for( i_y = 0; i_y < p_filter->fmt_in.video.i_height; i_y++ )
     {
         p_pic_start = p_pic;
         p_buffer = b_hscale ? p_buffer_start : p_pic;
 
-        for ( i_x = p_vout->render.i_width / 8; i_x--; )
+        for ( i_x = p_filter->fmt_in.video.i_width / 8; i_x--; )
         {
             CONVERT_YUV_PIXEL(2);  CONVERT_Y_PIXEL(2);
             CONVERT_YUV_PIXEL(2);  CONVERT_Y_PIXEL(2);
@@ -304,8 +300,8 @@ void E_(I420_RGB16)( vout_thread_t *p_vout, picture_t *p_src,
 
 #else // ! defined (MODULE_NAME_IS_i420_rgb)
 
-void E_(I420_R5G5B5)( vout_thread_t *p_vout, picture_t *p_src,
-                                            picture_t *p_dest )
+VLC_TARGET
+void I420_R5G5B5( filter_t *p_filter, picture_t *p_src, picture_t *p_dest )
 {
     /* We got this one from the old arguments */
     uint16_t *p_pic = (uint16_t*)p_dest->p->p_pixels;
@@ -313,22 +309,22 @@ void E_(I420_R5G5B5)( vout_thread_t *p_vout, picture_t *p_src,
     uint8_t  *p_u   = p_src->U_PIXELS;
     uint8_t  *p_v   = p_src->V_PIXELS;
 
-    vlc_bool_t  b_hscale;                         /* horizontal scaling type */
+    bool  b_hscale;                         /* horizontal scaling type */
     unsigned int i_vscale;                          /* vertical scaling type */
     unsigned int i_x, i_y;                /* horizontal and vertical indexes */
 
     int         i_right_margin;
     int         i_rewind;
     int         i_scale_count;                       /* scale modulo counter */
-    int         i_chroma_width = p_vout->render.i_width / 2; /* chroma width */
+    int         i_chroma_width = p_filter->fmt_in.video.i_width / 2; /* chroma width */
     uint16_t *  p_pic_start;       /* beginning of the current line for copy */
 
     /* Conversion buffer pointer */
-    uint16_t *  p_buffer_start = (uint16_t*)p_vout->chroma.p_sys->p_buffer;
+    uint16_t *  p_buffer_start = (uint16_t*)p_filter->p_sys->p_buffer;
     uint16_t *  p_buffer;
 
     /* Offset array pointer */
-    int *       p_offset_start = p_vout->chroma.p_sys->p_offset;
+    int *       p_offset_start = p_filter->p_sys->p_offset;
     int *       p_offset;
 
     const int i_source_margin = p_src->p[0].i_pitch
@@ -341,8 +337,10 @@ void E_(I420_R5G5B5)( vout_thread_t *p_vout, picture_t *p_src,
     /* Rule: when a picture of size (x1,y1) with aspect ratio r1 is rendered
      * on a picture of size (x2,y2) with aspect ratio r2, if x1 grows to x1'
      * then y1 grows to y1' = x1' * y2/x2 * r2/r1 */
-    SetOffset( p_vout->render.i_width, p_vout->render.i_height,
-               p_vout->output.i_width, p_vout->output.i_height,
+    SetOffset( p_filter->fmt_in.video.i_width,
+               p_filter->fmt_in.video.i_height,
+               p_filter->fmt_out.video.i_width,
+               p_filter->fmt_out.video.i_height,
                &b_hscale, &i_vscale, p_offset_start );
 
 
@@ -350,36 +348,30 @@ void E_(I420_R5G5B5)( vout_thread_t *p_vout, picture_t *p_src,
      * Perform conversion
      */
     i_scale_count = ( i_vscale == 1 ) ?
-                    p_vout->output.i_height : p_vout->render.i_height;
+                    p_filter->fmt_out.video.i_height :
+                    p_filter->fmt_in.video.i_height;
 
 #if defined (MODULE_NAME_IS_i420_rgb_sse2)
 
-    if( p_vout->render.i_width & 15 )
-    {
-        i_rewind = 16 - ( p_vout->render.i_width & 15 );
-    }
-    else
-    {
-        i_rewind = 0;
-    }
+    i_rewind = (-p_filter->fmt_in.video.i_width) & 15;
 
     /*
-    ** SSE2 128 bits fetch/store instructions are faster 
+    ** SSE2 128 bits fetch/store instructions are faster
     ** if memory access is 16 bytes aligned
     */
 
     p_buffer = b_hscale ? p_buffer_start : p_pic;
     if( 0 == (15 & (p_src->p[Y_PLANE].i_pitch|
                     p_dest->p->i_pitch|
-                    ((int)p_y)|
-                    ((int)p_buffer))) )
+                    ((intptr_t)p_y)|
+                    ((intptr_t)p_buffer))) )
     {
         /* use faster SSE2 aligned fetch and store */
-        for( i_y = 0; i_y < p_vout->render.i_height; i_y++ )
+        for( i_y = 0; i_y < p_filter->fmt_in.video.i_height; i_y++ )
         {
             p_pic_start = p_pic;
 
-            for ( i_x = p_vout->render.i_width/16; i_x--; )
+            for ( i_x = p_filter->fmt_in.video.i_width/16; i_x--; )
             {
                 SSE2_CALL (
                     SSE2_INIT_16_ALIGNED
@@ -426,12 +418,12 @@ void E_(I420_R5G5B5)( vout_thread_t *p_vout, picture_t *p_src,
     else
     {
         /* use slower SSE2 unaligned fetch and store */
-        for( i_y = 0; i_y < p_vout->render.i_height; i_y++ )
+        for( i_y = 0; i_y < p_filter->fmt_in.video.i_height; i_y++ )
         {
             p_pic_start = p_pic;
             p_buffer = b_hscale ? p_buffer_start : p_pic;
 
-            for ( i_x = p_vout->render.i_width/16; i_x--; )
+            for ( i_x = p_filter->fmt_in.video.i_width/16; i_x--; )
             {
                 SSE2_CALL (
                     SSE2_INIT_16_UNALIGNED
@@ -481,21 +473,14 @@ void E_(I420_R5G5B5)( vout_thread_t *p_vout, picture_t *p_src,
 
 #else // defined (MODULE_NAME_IS_i420_rgb_mmx)
 
-    if( p_vout->render.i_width & 7 )
-    {
-        i_rewind = 8 - ( p_vout->render.i_width & 7 );
-    }
-    else
-    {
-        i_rewind = 0;
-    }
+    i_rewind = (-p_filter->fmt_in.video.i_width) & 7;
 
-    for( i_y = 0; i_y < p_vout->render.i_height; i_y++ )
+    for( i_y = 0; i_y < p_filter->fmt_in.video.i_height; i_y++ )
     {
         p_pic_start = p_pic;
         p_buffer = b_hscale ? p_buffer_start : p_pic;
 
-        for ( i_x = p_vout->render.i_width / 8; i_x--; )
+        for ( i_x = p_filter->fmt_in.video.i_width / 8; i_x--; )
         {
             MMX_CALL (
                 MMX_INIT_16
@@ -545,8 +530,8 @@ void E_(I420_R5G5B5)( vout_thread_t *p_vout, picture_t *p_src,
 #endif
 }
 
-void E_(I420_R5G6B5)( vout_thread_t *p_vout, picture_t *p_src,
-                                            picture_t *p_dest )
+VLC_TARGET
+void I420_R5G6B5( filter_t *p_filter, picture_t *p_src, picture_t *p_dest )
 {
     /* We got this one from the old arguments */
     uint16_t *p_pic = (uint16_t*)p_dest->p->p_pixels;
@@ -554,22 +539,22 @@ void E_(I420_R5G6B5)( vout_thread_t *p_vout, picture_t *p_src,
     uint8_t  *p_u   = p_src->U_PIXELS;
     uint8_t  *p_v   = p_src->V_PIXELS;
 
-    vlc_bool_t  b_hscale;                         /* horizontal scaling type */
+    bool  b_hscale;                         /* horizontal scaling type */
     unsigned int i_vscale;                          /* vertical scaling type */
     unsigned int i_x, i_y;                /* horizontal and vertical indexes */
 
     int         i_right_margin;
     int         i_rewind;
     int         i_scale_count;                       /* scale modulo counter */
-    int         i_chroma_width = p_vout->render.i_width / 2; /* chroma width */
+    int         i_chroma_width = p_filter->fmt_in.video.i_width / 2; /* chroma width */
     uint16_t *  p_pic_start;       /* beginning of the current line for copy */
 
     /* Conversion buffer pointer */
-    uint16_t *  p_buffer_start = (uint16_t*)p_vout->chroma.p_sys->p_buffer;
+    uint16_t *  p_buffer_start = (uint16_t*)p_filter->p_sys->p_buffer;
     uint16_t *  p_buffer;
 
     /* Offset array pointer */
-    int *       p_offset_start = p_vout->chroma.p_sys->p_offset;
+    int *       p_offset_start = p_filter->p_sys->p_offset;
     int *       p_offset;
 
     const int i_source_margin = p_src->p[0].i_pitch
@@ -582,8 +567,10 @@ void E_(I420_R5G6B5)( vout_thread_t *p_vout, picture_t *p_src,
     /* Rule: when a picture of size (x1,y1) with aspect ratio r1 is rendered
      * on a picture of size (x2,y2) with aspect ratio r2, if x1 grows to x1'
      * then y1 grows to y1' = x1' * y2/x2 * r2/r1 */
-    SetOffset( p_vout->render.i_width, p_vout->render.i_height,
-               p_vout->output.i_width, p_vout->output.i_height,
+    SetOffset( p_filter->fmt_in.video.i_width,
+               p_filter->fmt_in.video.i_height,
+               p_filter->fmt_out.video.i_width,
+               p_filter->fmt_out.video.i_height,
                &b_hscale, &i_vscale, p_offset_start );
 
 
@@ -591,36 +578,30 @@ void E_(I420_R5G6B5)( vout_thread_t *p_vout, picture_t *p_src,
      * Perform conversion
      */
     i_scale_count = ( i_vscale == 1 ) ?
-                    p_vout->output.i_height : p_vout->render.i_height;
+                    p_filter->fmt_out.video.i_height :
+                    p_filter->fmt_in.video.i_height;
 
 #if defined (MODULE_NAME_IS_i420_rgb_sse2)
 
-    if( p_vout->render.i_width & 15 )
-    {
-        i_rewind = 16 - ( p_vout->render.i_width & 15 );
-    }
-    else
-    {
-        i_rewind = 0;
-    }
+    i_rewind = (-p_filter->fmt_in.video.i_width) & 15;
 
     /*
-    ** SSE2 128 bits fetch/store instructions are faster 
+    ** SSE2 128 bits fetch/store instructions are faster
     ** if memory access is 16 bytes aligned
     */
 
     p_buffer = b_hscale ? p_buffer_start : p_pic;
     if( 0 == (15 & (p_src->p[Y_PLANE].i_pitch|
                     p_dest->p->i_pitch|
-                    ((int)p_y)|
-                    ((int)p_buffer))) )
+                    ((intptr_t)p_y)|
+                    ((intptr_t)p_buffer))) )
     {
         /* use faster SSE2 aligned fetch and store */
-        for( i_y = 0; i_y < p_vout->render.i_height; i_y++ )
+        for( i_y = 0; i_y < p_filter->fmt_in.video.i_height; i_y++ )
         {
             p_pic_start = p_pic;
 
-            for ( i_x = p_vout->render.i_width/16; i_x--; )
+            for ( i_x = p_filter->fmt_in.video.i_width/16; i_x--; )
             {
                 SSE2_CALL (
                     SSE2_INIT_16_ALIGNED
@@ -667,12 +648,12 @@ void E_(I420_R5G6B5)( vout_thread_t *p_vout, picture_t *p_src,
     else
     {
         /* use slower SSE2 unaligned fetch and store */
-        for( i_y = 0; i_y < p_vout->render.i_height; i_y++ )
+        for( i_y = 0; i_y < p_filter->fmt_in.video.i_height; i_y++ )
         {
             p_pic_start = p_pic;
             p_buffer = b_hscale ? p_buffer_start : p_pic;
 
-            for ( i_x = p_vout->render.i_width/16; i_x--; )
+            for ( i_x = p_filter->fmt_in.video.i_width/16; i_x--; )
             {
                 SSE2_CALL(
                     SSE2_INIT_16_UNALIGNED
@@ -722,21 +703,14 @@ void E_(I420_R5G6B5)( vout_thread_t *p_vout, picture_t *p_src,
 
 #else // defined (MODULE_NAME_IS_i420_rgb_mmx)
 
-    if( p_vout->render.i_width & 7 )
-    {
-        i_rewind = 8 - ( p_vout->render.i_width & 7 );
-    }
-    else
-    {
-        i_rewind = 0;
-    }
+    i_rewind = (-p_filter->fmt_in.video.i_width) & 7;
 
-    for( i_y = 0; i_y < p_vout->render.i_height; i_y++ )
+    for( i_y = 0; i_y < p_filter->fmt_in.video.i_height; i_y++ )
     {
         p_pic_start = p_pic;
         p_buffer = b_hscale ? p_buffer_start : p_pic;
 
-        for ( i_x = p_vout->render.i_width / 8; i_x--; )
+        for ( i_x = p_filter->fmt_in.video.i_width / 8; i_x--; )
         {
             MMX_CALL (
                 MMX_INIT_16
@@ -801,8 +775,7 @@ void E_(I420_R5G6B5)( vout_thread_t *p_vout, picture_t *p_src,
 
 #if defined (MODULE_NAME_IS_i420_rgb)
 
-void E_(I420_RGB32)( vout_thread_t *p_vout, picture_t *p_src,
-                                            picture_t *p_dest )
+void I420_RGB32( filter_t *p_filter, picture_t *p_src, picture_t *p_dest )
 {
     /* We got this one from the old arguments */
     uint32_t *p_pic = (uint32_t*)p_dest->p->p_pixels;
@@ -810,26 +783,26 @@ void E_(I420_RGB32)( vout_thread_t *p_vout, picture_t *p_src,
     uint8_t  *p_u   = p_src->U_PIXELS;
     uint8_t  *p_v   = p_src->V_PIXELS;
 
-    vlc_bool_t  b_hscale;                         /* horizontal scaling type */
+    bool  b_hscale;                         /* horizontal scaling type */
     unsigned int i_vscale;                          /* vertical scaling type */
     unsigned int i_x, i_y;                /* horizontal and vertical indexes */
 
     int         i_right_margin;
     int         i_rewind;
     int         i_scale_count;                       /* scale modulo counter */
-    int         i_chroma_width = p_vout->render.i_width / 2; /* chroma width */
+    int         i_chroma_width = p_filter->fmt_in.video.i_width / 2; /* chroma width */
     uint32_t *  p_pic_start;       /* beginning of the current line for copy */
     int         i_uval, i_vval;                           /* U and V samples */
     int         i_red, i_green, i_blue;          /* U and V modified samples */
-    uint32_t *  p_yuv = p_vout->chroma.p_sys->p_rgb32;
+    uint32_t *  p_yuv = p_filter->p_sys->p_rgb32;
     uint32_t *  p_ybase;                     /* Y dependant conversion table */
 
     /* Conversion buffer pointer */
-    uint32_t *  p_buffer_start = (uint32_t*)p_vout->chroma.p_sys->p_buffer;
+    uint32_t *  p_buffer_start = (uint32_t*)p_filter->p_sys->p_buffer;
     uint32_t *  p_buffer;
 
     /* Offset array pointer */
-    int *       p_offset_start = p_vout->chroma.p_sys->p_offset;
+    int *       p_offset_start = p_filter->p_sys->p_offset;
     int *       p_offset;
 
     const int i_source_margin = p_src->p[0].i_pitch
@@ -838,34 +811,29 @@ void E_(I420_RGB32)( vout_thread_t *p_vout, picture_t *p_src,
                                  - p_src->p[1].i_visible_pitch;
 
     i_right_margin = p_dest->p->i_pitch - p_dest->p->i_visible_pitch;
-
-    if( p_vout->render.i_width & 7 )
-    {
-        i_rewind = 8 - ( p_vout->render.i_width & 7 );
-    }
-    else
-    {
-        i_rewind = 0;
-    }
+    i_rewind = (-p_filter->fmt_in.video.i_width) & 7;
 
     /* Rule: when a picture of size (x1,y1) with aspect ratio r1 is rendered
      * on a picture of size (x2,y2) with aspect ratio r2, if x1 grows to x1'
      * then y1 grows to y1' = x1' * y2/x2 * r2/r1 */
-    SetOffset( p_vout->render.i_width, p_vout->render.i_height,
-               p_vout->output.i_width, p_vout->output.i_height,
+    SetOffset( p_filter->fmt_in.video.i_width,
+               p_filter->fmt_in.video.i_height,
+               p_filter->fmt_out.video.i_width,
+               p_filter->fmt_out.video.i_height,
                &b_hscale, &i_vscale, p_offset_start );
 
     /*
      * Perform conversion
      */
     i_scale_count = ( i_vscale == 1 ) ?
-                    p_vout->output.i_height : p_vout->render.i_height;
-    for( i_y = 0; i_y < p_vout->render.i_height; i_y++ )
+                    p_filter->fmt_out.video.i_height :
+                    p_filter->fmt_in.video.i_height;
+    for( i_y = 0; i_y < p_filter->fmt_in.video.i_height; i_y++ )
     {
         p_pic_start = p_pic;
         p_buffer = b_hscale ? p_buffer_start : p_pic;
 
-        for ( i_x = p_vout->render.i_width / 8; i_x--; )
+        for ( i_x = p_filter->fmt_in.video.i_width / 8; i_x--; )
         {
             CONVERT_YUV_PIXEL(4);  CONVERT_Y_PIXEL(4);
             CONVERT_YUV_PIXEL(4);  CONVERT_Y_PIXEL(4);
@@ -900,7 +868,8 @@ void E_(I420_RGB32)( vout_thread_t *p_vout, picture_t *p_src,
 
 #else // defined (MODULE_NAME_IS_i420_rgb_mmx) || defined (MODULE_NAME_IS_i420_rgb_sse2)
 
-void E_(I420_A8R8G8B8)( vout_thread_t *p_vout, picture_t *p_src,
+VLC_TARGET
+void I420_A8R8G8B8( filter_t *p_filter, picture_t *p_src,
                                             picture_t *p_dest )
 {
     /* We got this one from the old arguments */
@@ -909,21 +878,21 @@ void E_(I420_A8R8G8B8)( vout_thread_t *p_vout, picture_t *p_src,
     uint8_t  *p_u   = p_src->U_PIXELS;
     uint8_t  *p_v   = p_src->V_PIXELS;
 
-    vlc_bool_t  b_hscale;                         /* horizontal scaling type */
+    bool  b_hscale;                         /* horizontal scaling type */
     unsigned int i_vscale;                          /* vertical scaling type */
     unsigned int i_x, i_y;                /* horizontal and vertical indexes */
 
     int         i_right_margin;
     int         i_rewind;
     int         i_scale_count;                       /* scale modulo counter */
-    int         i_chroma_width = p_vout->render.i_width / 2; /* chroma width */
+    int         i_chroma_width = p_filter->fmt_in.video.i_width / 2; /* chroma width */
     uint32_t *  p_pic_start;       /* beginning of the current line for copy */
     /* Conversion buffer pointer */
-    uint32_t *  p_buffer_start = (uint32_t*)p_vout->chroma.p_sys->p_buffer;
+    uint32_t *  p_buffer_start = (uint32_t*)p_filter->p_sys->p_buffer;
     uint32_t *  p_buffer;
 
     /* Offset array pointer */
-    int *       p_offset_start = p_vout->chroma.p_sys->p_offset;
+    int *       p_offset_start = p_filter->p_sys->p_offset;
     int *       p_offset;
 
     const int i_source_margin = p_src->p[0].i_pitch
@@ -936,44 +905,40 @@ void E_(I420_A8R8G8B8)( vout_thread_t *p_vout, picture_t *p_src,
     /* Rule: when a picture of size (x1,y1) with aspect ratio r1 is rendered
      * on a picture of size (x2,y2) with aspect ratio r2, if x1 grows to x1'
      * then y1 grows to y1' = x1' * y2/x2 * r2/r1 */
-    SetOffset( p_vout->render.i_width, p_vout->render.i_height,
-               p_vout->output.i_width, p_vout->output.i_height,
+    SetOffset( p_filter->fmt_in.video.i_width,
+               p_filter->fmt_in.video.i_height,
+               p_filter->fmt_out.video.i_width,
+               p_filter->fmt_out.video.i_height,
                &b_hscale, &i_vscale, p_offset_start );
 
     /*
      * Perform conversion
      */
     i_scale_count = ( i_vscale == 1 ) ?
-                    p_vout->output.i_height : p_vout->render.i_height;
+                    p_filter->fmt_out.video.i_height :
+                    p_filter->fmt_in.video.i_height;
 
 #if defined (MODULE_NAME_IS_i420_rgb_sse2)
 
-    if( p_vout->render.i_width & 15 )
-    {
-        i_rewind = 16 - ( p_vout->render.i_width & 15 );
-    }
-    else
-    {
-        i_rewind = 0;
-    }
+    i_rewind = (-p_filter->fmt_in.video.i_width) & 15;
 
     /*
-    ** SSE2 128 bits fetch/store instructions are faster 
+    ** SSE2 128 bits fetch/store instructions are faster
     ** if memory access is 16 bytes aligned
     */
 
     p_buffer = b_hscale ? p_buffer_start : p_pic;
     if( 0 == (15 & (p_src->p[Y_PLANE].i_pitch|
                     p_dest->p->i_pitch|
-                    ((int)p_y)|
-                    ((int)p_buffer))) )
+                    ((intptr_t)p_y)|
+                    ((intptr_t)p_buffer))) )
     {
         /* use faster SSE2 aligned fetch and store */
-        for( i_y = 0; i_y < p_vout->render.i_height; i_y++ )
+        for( i_y = 0; i_y < p_filter->fmt_in.video.i_height; i_y++ )
         {
             p_pic_start = p_pic;
 
-            for ( i_x = p_vout->render.i_width / 16; i_x--; )
+            for ( i_x = p_filter->fmt_in.video.i_width / 16; i_x--; )
             {
                 SSE2_CALL (
                     SSE2_INIT_32_ALIGNED
@@ -1020,12 +985,12 @@ void E_(I420_A8R8G8B8)( vout_thread_t *p_vout, picture_t *p_src,
     else
     {
         /* use slower SSE2 unaligned fetch and store */
-        for( i_y = 0; i_y < p_vout->render.i_height; i_y++ )
+        for( i_y = 0; i_y < p_filter->fmt_in.video.i_height; i_y++ )
         {
             p_pic_start = p_pic;
             p_buffer = b_hscale ? p_buffer_start : p_pic;
 
-            for ( i_x = p_vout->render.i_width / 16; i_x--; )
+            for ( i_x = p_filter->fmt_in.video.i_width / 16; i_x--; )
             {
                 SSE2_CALL (
                     SSE2_INIT_32_UNALIGNED
@@ -1075,21 +1040,14 @@ void E_(I420_A8R8G8B8)( vout_thread_t *p_vout, picture_t *p_src,
 
 #else // defined (MODULE_NAME_IS_i420_rgb_mmx)
 
-    if( p_vout->render.i_width & 7 )
-    {
-        i_rewind = 8 - ( p_vout->render.i_width & 7 );
-    }
-    else
-    {
-        i_rewind = 0;
-    }
+    i_rewind = (-p_filter->fmt_in.video.i_width) & 7;
 
-    for( i_y = 0; i_y < p_vout->render.i_height; i_y++ )
+    for( i_y = 0; i_y < p_filter->fmt_in.video.i_height; i_y++ )
     {
         p_pic_start = p_pic;
         p_buffer = b_hscale ? p_buffer_start : p_pic;
 
-        for ( i_x = p_vout->render.i_width / 8; i_x--; )
+        for ( i_x = p_filter->fmt_in.video.i_width / 8; i_x--; )
         {
             MMX_CALL (
                 MMX_INIT_32
@@ -1139,8 +1097,8 @@ void E_(I420_A8R8G8B8)( vout_thread_t *p_vout, picture_t *p_src,
 #endif
 }
 
-void E_(I420_R8G8B8A8)( vout_thread_t *p_vout, picture_t *p_src,
-                                            picture_t *p_dest )
+VLC_TARGET
+void I420_R8G8B8A8( filter_t *p_filter, picture_t *p_src, picture_t *p_dest )
 {
     /* We got this one from the old arguments */
     uint32_t *p_pic = (uint32_t*)p_dest->p->p_pixels;
@@ -1148,21 +1106,21 @@ void E_(I420_R8G8B8A8)( vout_thread_t *p_vout, picture_t *p_src,
     uint8_t  *p_u   = p_src->U_PIXELS;
     uint8_t  *p_v   = p_src->V_PIXELS;
 
-    vlc_bool_t  b_hscale;                         /* horizontal scaling type */
+    bool  b_hscale;                         /* horizontal scaling type */
     unsigned int i_vscale;                          /* vertical scaling type */
     unsigned int i_x, i_y;                /* horizontal and vertical indexes */
 
     int         i_right_margin;
     int         i_rewind;
     int         i_scale_count;                       /* scale modulo counter */
-    int         i_chroma_width = p_vout->render.i_width / 2; /* chroma width */
+    int         i_chroma_width = p_filter->fmt_in.video.i_width / 2; /* chroma width */
     uint32_t *  p_pic_start;       /* beginning of the current line for copy */
     /* Conversion buffer pointer */
-    uint32_t *  p_buffer_start = (uint32_t*)p_vout->chroma.p_sys->p_buffer;
+    uint32_t *  p_buffer_start = (uint32_t*)p_filter->p_sys->p_buffer;
     uint32_t *  p_buffer;
 
     /* Offset array pointer */
-    int *       p_offset_start = p_vout->chroma.p_sys->p_offset;
+    int *       p_offset_start = p_filter->p_sys->p_offset;
     int *       p_offset;
 
     const int i_source_margin = p_src->p[0].i_pitch
@@ -1175,44 +1133,40 @@ void E_(I420_R8G8B8A8)( vout_thread_t *p_vout, picture_t *p_src,
     /* Rule: when a picture of size (x1,y1) with aspect ratio r1 is rendered
      * on a picture of size (x2,y2) with aspect ratio r2, if x1 grows to x1'
      * then y1 grows to y1' = x1' * y2/x2 * r2/r1 */
-    SetOffset( p_vout->render.i_width, p_vout->render.i_height,
-               p_vout->output.i_width, p_vout->output.i_height,
+    SetOffset( p_filter->fmt_in.video.i_width,
+               p_filter->fmt_in.video.i_height,
+               p_filter->fmt_out.video.i_width,
+               p_filter->fmt_out.video.i_height,
                &b_hscale, &i_vscale, p_offset_start );
 
     /*
      * Perform conversion
      */
     i_scale_count = ( i_vscale == 1 ) ?
-                    p_vout->output.i_height : p_vout->render.i_height;
+                    p_filter->fmt_out.video.i_height :
+                    p_filter->fmt_in.video.i_height;
 
 #if defined (MODULE_NAME_IS_i420_rgb_sse2)
 
-    if( p_vout->render.i_width & 15 )
-    {
-        i_rewind = 16 - ( p_vout->render.i_width & 15 );
-    }
-    else
-    {
-        i_rewind = 0;
-    }
+    i_rewind = (-p_filter->fmt_in.video.i_width) & 15;
 
     /*
-    ** SSE2 128 bits fetch/store instructions are faster 
+    ** SSE2 128 bits fetch/store instructions are faster
     ** if memory access is 16 bytes aligned
     */
 
     p_buffer = b_hscale ? p_buffer_start : p_pic;
     if( 0 == (15 & (p_src->p[Y_PLANE].i_pitch|
                     p_dest->p->i_pitch|
-                    ((int)p_y)|
-                    ((int)p_buffer))) )
+                    ((intptr_t)p_y)|
+                    ((intptr_t)p_buffer))) )
     {
         /* use faster SSE2 aligned fetch and store */
-        for( i_y = 0; i_y < p_vout->render.i_height; i_y++ )
+        for( i_y = 0; i_y < p_filter->fmt_in.video.i_height; i_y++ )
         {
             p_pic_start = p_pic;
 
-            for ( i_x = p_vout->render.i_width / 16; i_x--; )
+            for ( i_x = p_filter->fmt_in.video.i_width / 16; i_x--; )
             {
                 SSE2_CALL (
                     SSE2_INIT_32_ALIGNED
@@ -1259,12 +1213,12 @@ void E_(I420_R8G8B8A8)( vout_thread_t *p_vout, picture_t *p_src,
     else
     {
         /* use slower SSE2 unaligned fetch and store */
-        for( i_y = 0; i_y < p_vout->render.i_height; i_y++ )
+        for( i_y = 0; i_y < p_filter->fmt_in.video.i_height; i_y++ )
         {
             p_pic_start = p_pic;
             p_buffer = b_hscale ? p_buffer_start : p_pic;
 
-            for ( i_x = p_vout->render.i_width / 16; i_x--; )
+            for ( i_x = p_filter->fmt_in.video.i_width / 16; i_x--; )
             {
                 SSE2_CALL (
                     SSE2_INIT_32_UNALIGNED
@@ -1314,21 +1268,14 @@ void E_(I420_R8G8B8A8)( vout_thread_t *p_vout, picture_t *p_src,
 
 #else // defined (MODULE_NAME_IS_i420_rgb_mmx)
 
-    if( p_vout->render.i_width & 7 )
-    {
-        i_rewind = 8 - ( p_vout->render.i_width & 7 );
-    }
-    else
-    {
-        i_rewind = 0;
-    }
+    i_rewind = (-p_filter->fmt_in.video.i_width) & 7;
 
-    for( i_y = 0; i_y < p_vout->render.i_height; i_y++ )
+    for( i_y = 0; i_y < p_filter->fmt_in.video.i_height; i_y++ )
     {
         p_pic_start = p_pic;
         p_buffer = b_hscale ? p_buffer_start : p_pic;
 
-        for ( i_x = p_vout->render.i_width / 8; i_x--; )
+        for ( i_x = p_filter->fmt_in.video.i_width / 8; i_x--; )
         {
             MMX_CALL (
                 MMX_INIT_32
@@ -1378,8 +1325,8 @@ void E_(I420_R8G8B8A8)( vout_thread_t *p_vout, picture_t *p_src,
 #endif
 }
 
-void E_(I420_B8G8R8A8)( vout_thread_t *p_vout, picture_t *p_src,
-                                            picture_t *p_dest )
+VLC_TARGET
+void I420_B8G8R8A8( filter_t *p_filter, picture_t *p_src, picture_t *p_dest )
 {
     /* We got this one from the old arguments */
     uint32_t *p_pic = (uint32_t*)p_dest->p->p_pixels;
@@ -1387,21 +1334,21 @@ void E_(I420_B8G8R8A8)( vout_thread_t *p_vout, picture_t *p_src,
     uint8_t  *p_u   = p_src->U_PIXELS;
     uint8_t  *p_v   = p_src->V_PIXELS;
 
-    vlc_bool_t  b_hscale;                         /* horizontal scaling type */
+    bool  b_hscale;                         /* horizontal scaling type */
     unsigned int i_vscale;                          /* vertical scaling type */
     unsigned int i_x, i_y;                /* horizontal and vertical indexes */
 
     int         i_right_margin;
     int         i_rewind;
     int         i_scale_count;                       /* scale modulo counter */
-    int         i_chroma_width = p_vout->render.i_width / 2; /* chroma width */
+    int         i_chroma_width = p_filter->fmt_in.video.i_width / 2; /* chroma width */
     uint32_t *  p_pic_start;       /* beginning of the current line for copy */
     /* Conversion buffer pointer */
-    uint32_t *  p_buffer_start = (uint32_t*)p_vout->chroma.p_sys->p_buffer;
+    uint32_t *  p_buffer_start = (uint32_t*)p_filter->p_sys->p_buffer;
     uint32_t *  p_buffer;
 
     /* Offset array pointer */
-    int *       p_offset_start = p_vout->chroma.p_sys->p_offset;
+    int *       p_offset_start = p_filter->p_sys->p_offset;
     int *       p_offset;
 
     const int i_source_margin = p_src->p[0].i_pitch
@@ -1414,44 +1361,40 @@ void E_(I420_B8G8R8A8)( vout_thread_t *p_vout, picture_t *p_src,
     /* Rule: when a picture of size (x1,y1) with aspect ratio r1 is rendered
      * on a picture of size (x2,y2) with aspect ratio r2, if x1 grows to x1'
      * then y1 grows to y1' = x1' * y2/x2 * r2/r1 */
-    SetOffset( p_vout->render.i_width, p_vout->render.i_height,
-               p_vout->output.i_width, p_vout->output.i_height,
+    SetOffset( p_filter->fmt_in.video.i_width,
+               p_filter->fmt_in.video.i_height,
+               p_filter->fmt_out.video.i_width,
+               p_filter->fmt_out.video.i_height,
                &b_hscale, &i_vscale, p_offset_start );
 
     /*
      * Perform conversion
      */
     i_scale_count = ( i_vscale == 1 ) ?
-                    p_vout->output.i_height : p_vout->render.i_height;
+                    p_filter->fmt_out.video.i_height :
+                    p_filter->fmt_in.video.i_height;
 
 #if defined (MODULE_NAME_IS_i420_rgb_sse2)
 
-    if( p_vout->render.i_width & 15 )
-    {
-        i_rewind = 16 - ( p_vout->render.i_width & 15 );
-    }
-    else
-    {
-        i_rewind = 0;
-    }
+    i_rewind = (-p_filter->fmt_in.video.i_width) & 15;
 
     /*
-    ** SSE2 128 bits fetch/store instructions are faster 
+    ** SSE2 128 bits fetch/store instructions are faster
     ** if memory access is 16 bytes aligned
     */
 
     p_buffer = b_hscale ? p_buffer_start : p_pic;
     if( 0 == (15 & (p_src->p[Y_PLANE].i_pitch|
                     p_dest->p->i_pitch|
-                    ((int)p_y)|
-                    ((int)p_buffer))) )
+                    ((intptr_t)p_y)|
+                    ((intptr_t)p_buffer))) )
     {
         /* use faster SSE2 aligned fetch and store */
-        for( i_y = 0; i_y < p_vout->render.i_height; i_y++ )
+        for( i_y = 0; i_y < p_filter->fmt_in.video.i_height; i_y++ )
         {
             p_pic_start = p_pic;
 
-            for ( i_x = p_vout->render.i_width / 16; i_x--; )
+            for ( i_x = p_filter->fmt_in.video.i_width / 16; i_x--; )
             {
                 SSE2_CALL (
                     SSE2_INIT_32_ALIGNED
@@ -1498,12 +1441,12 @@ void E_(I420_B8G8R8A8)( vout_thread_t *p_vout, picture_t *p_src,
     else
     {
         /* use slower SSE2 unaligned fetch and store */
-        for( i_y = 0; i_y < p_vout->render.i_height; i_y++ )
+        for( i_y = 0; i_y < p_filter->fmt_in.video.i_height; i_y++ )
         {
             p_pic_start = p_pic;
             p_buffer = b_hscale ? p_buffer_start : p_pic;
 
-            for ( i_x = p_vout->render.i_width / 16; i_x--; )
+            for ( i_x = p_filter->fmt_in.video.i_width / 16; i_x--; )
             {
                 SSE2_CALL (
                     SSE2_INIT_32_UNALIGNED
@@ -1550,21 +1493,14 @@ void E_(I420_B8G8R8A8)( vout_thread_t *p_vout, picture_t *p_src,
 
 #else
 
-    if( p_vout->render.i_width & 7 )
-    {
-        i_rewind = 8 - ( p_vout->render.i_width & 7 );
-    }
-    else
-    {
-        i_rewind = 0;
-    }
+    i_rewind = (-p_filter->fmt_in.video.i_width) & 7;
 
-    for( i_y = 0; i_y < p_vout->render.i_height; i_y++ )
+    for( i_y = 0; i_y < p_filter->fmt_in.video.i_height; i_y++ )
     {
         p_pic_start = p_pic;
         p_buffer = b_hscale ? p_buffer_start : p_pic;
 
-        for ( i_x = p_vout->render.i_width / 8; i_x--; )
+        for ( i_x = p_filter->fmt_in.video.i_width / 8; i_x--; )
         {
             MMX_CALL (
                 MMX_INIT_32
@@ -1614,8 +1550,8 @@ void E_(I420_B8G8R8A8)( vout_thread_t *p_vout, picture_t *p_src,
 #endif
 }
 
-void E_(I420_A8B8G8R8)( vout_thread_t *p_vout, picture_t *p_src,
-                                            picture_t *p_dest )
+VLC_TARGET
+void I420_A8B8G8R8( filter_t *p_filter, picture_t *p_src, picture_t *p_dest )
 {
     /* We got this one from the old arguments */
     uint32_t *p_pic = (uint32_t*)p_dest->p->p_pixels;
@@ -1623,21 +1559,21 @@ void E_(I420_A8B8G8R8)( vout_thread_t *p_vout, picture_t *p_src,
     uint8_t  *p_u   = p_src->U_PIXELS;
     uint8_t  *p_v   = p_src->V_PIXELS;
 
-    vlc_bool_t  b_hscale;                         /* horizontal scaling type */
+    bool  b_hscale;                         /* horizontal scaling type */
     unsigned int i_vscale;                          /* vertical scaling type */
     unsigned int i_x, i_y;                /* horizontal and vertical indexes */
 
     int         i_right_margin;
     int         i_rewind;
     int         i_scale_count;                       /* scale modulo counter */
-    int         i_chroma_width = p_vout->render.i_width / 2; /* chroma width */
+    int         i_chroma_width = p_filter->fmt_in.video.i_width / 2; /* chroma width */
     uint32_t *  p_pic_start;       /* beginning of the current line for copy */
     /* Conversion buffer pointer */
-    uint32_t *  p_buffer_start = (uint32_t*)p_vout->chroma.p_sys->p_buffer;
+    uint32_t *  p_buffer_start = (uint32_t*)p_filter->p_sys->p_buffer;
     uint32_t *  p_buffer;
 
     /* Offset array pointer */
-    int *       p_offset_start = p_vout->chroma.p_sys->p_offset;
+    int *       p_offset_start = p_filter->p_sys->p_offset;
     int *       p_offset;
 
     const int i_source_margin = p_src->p[0].i_pitch
@@ -1650,44 +1586,40 @@ void E_(I420_A8B8G8R8)( vout_thread_t *p_vout, picture_t *p_src,
     /* Rule: when a picture of size (x1,y1) with aspect ratio r1 is rendered
      * on a picture of size (x2,y2) with aspect ratio r2, if x1 grows to x1'
      * then y1 grows to y1' = x1' * y2/x2 * r2/r1 */
-    SetOffset( p_vout->render.i_width, p_vout->render.i_height,
-               p_vout->output.i_width, p_vout->output.i_height,
+    SetOffset( p_filter->fmt_in.video.i_width,
+               p_filter->fmt_in.video.i_height,
+               p_filter->fmt_out.video.i_width,
+               p_filter->fmt_out.video.i_height,
                &b_hscale, &i_vscale, p_offset_start );
 
     /*
      * Perform conversion
      */
     i_scale_count = ( i_vscale == 1 ) ?
-                    p_vout->output.i_height : p_vout->render.i_height;
+                    p_filter->fmt_out.video.i_height :
+                    p_filter->fmt_in.video.i_height;
 
 #if defined (MODULE_NAME_IS_i420_rgb_sse2)
 
-    if( p_vout->render.i_width & 15 )
-    {
-        i_rewind = 16 - ( p_vout->render.i_width & 15 );
-    }
-    else
-    {
-        i_rewind = 0;
-    }
+    i_rewind = (-p_filter->fmt_in.video.i_width) & 15;
 
     /*
-    ** SSE2 128 bits fetch/store instructions are faster 
+    ** SSE2 128 bits fetch/store instructions are faster
     ** if memory access is 16 bytes aligned
     */
 
     p_buffer = b_hscale ? p_buffer_start : p_pic;
     if( 0 == (15 & (p_src->p[Y_PLANE].i_pitch|
                     p_dest->p->i_pitch|
-                    ((int)p_y)|
-                    ((int)p_buffer))) )
+                    ((intptr_t)p_y)|
+                    ((intptr_t)p_buffer))) )
     {
         /* use faster SSE2 aligned fetch and store */
-        for( i_y = 0; i_y < p_vout->render.i_height; i_y++ )
+        for( i_y = 0; i_y < p_filter->fmt_in.video.i_height; i_y++ )
         {
             p_pic_start = p_pic;
 
-            for ( i_x = p_vout->render.i_width / 16; i_x--; )
+            for ( i_x = p_filter->fmt_in.video.i_width / 16; i_x--; )
             {
                 SSE2_CALL (
                     SSE2_INIT_32_ALIGNED
@@ -1734,12 +1666,12 @@ void E_(I420_A8B8G8R8)( vout_thread_t *p_vout, picture_t *p_src,
     else
     {
         /* use slower SSE2 unaligned fetch and store */
-        for( i_y = 0; i_y < p_vout->render.i_height; i_y++ )
+        for( i_y = 0; i_y < p_filter->fmt_in.video.i_height; i_y++ )
         {
             p_pic_start = p_pic;
             p_buffer = b_hscale ? p_buffer_start : p_pic;
 
-            for ( i_x = p_vout->render.i_width / 16; i_x--; )
+            for ( i_x = p_filter->fmt_in.video.i_width / 16; i_x--; )
             {
                 SSE2_CALL (
                     SSE2_INIT_32_UNALIGNED
@@ -1786,21 +1718,14 @@ void E_(I420_A8B8G8R8)( vout_thread_t *p_vout, picture_t *p_src,
 
 #else
 
-    if( p_vout->render.i_width & 7 )
-    {
-        i_rewind = 8 - ( p_vout->render.i_width & 7 );
-    }
-    else
-    {
-        i_rewind = 0;
-    }
+    i_rewind = (-p_filter->fmt_in.video.i_width) & 7;
 
-    for( i_y = 0; i_y < p_vout->render.i_height; i_y++ )
+    for( i_y = 0; i_y < p_filter->fmt_in.video.i_height; i_y++ )
     {
         p_pic_start = p_pic;
         p_buffer = b_hscale ? p_buffer_start : p_pic;
 
-        for ( i_x = p_vout->render.i_width / 8; i_x--; )
+        for ( i_x = p_filter->fmt_in.video.i_width / 8; i_x--; )
         {
             MMX_CALL (
                 MMX_INIT_32
@@ -1861,7 +1786,7 @@ void E_(I420_A8B8G8R8)( vout_thread_t *p_vout, picture_t *p_src,
  * It will also set horizontal and vertical scaling indicators.
  *****************************************************************************/
 static void SetOffset( int i_width, int i_height, int i_pic_width,
-                       int i_pic_height, vlc_bool_t *pb_hscale,
+                       int i_pic_height, bool *pb_hscale,
                        unsigned int *pi_vscale, int *p_offset )
 {
     int i_x;                                    /* x position in destination */