Imported nx-X11-3.1.0-1.tar.gznx-X11/3.1.0-1

Summary: Imported nx-X11-3.1.0-1.tar.gz
Keywords:

Imported nx-X11-3.1.0-1.tar.gz
into Git repository

15 files changed, 11248 insertions, 0 deletions

diff --git a/nx-X11/lib/XvMC/hw/i810/I810XvMC.c b/nx-X11/lib/XvMC/hw/i810/I810XvMC.c
new file mode 100644
index 000000000..89aa3ea4b
--- /dev/null
+++ b/nx-X11/lib/XvMC/hw/i810/I810XvMC.c
@@ -0,0 +1,4507 @@
+/***************************************************************************
+
+Copyright 2001 Intel Corporation.  All Rights Reserved.
+
+Permission is hereby granted, free of charge, to any person obtaining a
+copy of this software and associated documentation files (the
+"Software"), to deal in the Software without restriction, including
+without limitation the rights to use, copy, modify, merge, publish,
+distribute, sub license, and/or sell copies of the Software, and to
+permit persons to whom the Software is furnished to do so, subject to
+the following conditions:
+
+The above copyright notice and this permission notice (including the
+next paragraph) shall be included in all copies or substantial portions
+of the Software.
+
+THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+IN NO EVENT SHALL INTEL, AND/OR ITS SUPPLIERS BE LIABLE FOR ANY CLAIM,
+DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR
+OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR
+THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+
+**************************************************************************/
+
+/*************************************************************************
+** File libI810XvMC.c
+**
+** Authors:
+**      Matt Sottek <matthew.j.sottek@intel.com>
+**      Bob Paauwe  <bob.j.paauwe@intel.com>
+**
+**
+***************************************************************************/
+/* $XFree86: xc/lib/XvMC/hw/i810/I810XvMC.c,v 1.10 2002/10/30 12:52:01 alanh Exp $ */
+#include <stdio.h>
+#include <stdlib.h>
+#include <unistd.h>
+#include <errno.h>
+#include <signal.h>
+#include <fcntl.h>
+#include <dirent.h>
+#include <string.h>
+
+#include <sys/ioctl.h>
+#include <X11/Xlibint.h>
+#include <fourcc.h>
+#include <X11/extensions/Xv.h>
+#include <X11/extensions/Xvlib.h>
+#include <X11/extensions/XvMC.h>
+#include <X11/extensions/XvMClib.h>
+#include "I810XvMC.h"
+
+static int error_base;
+static int event_base;
+
+/***************************************************************************
+// Function: i810_get_free_buffer
+// Description: Allocates a free dma page using kernel ioctls, then
+//   programs the data into the already allocated dma buffer list.
+// Arguments: pI810XvMC private data structure from the current context.
+// Notes: We faked the drmMapBufs for the i810's security so now we have
+//   to insert an allocated page into the correct spot in the faked
+//   list to keep up appearences.
+//   Concept for this function was taken from Mesa sources.
+// Returns: drmBufPtr containing the information about the allocated page.
+***************************************************************************/
+drmBufPtr i810_get_free_buffer(i810XvMCContext *pI810XvMC) {
+   drmI810DMA dma;
+   drmBufPtr buf;
+
+   dma.granted = 0;
+   dma.request_size = 4096;
+   while(!dma.granted) {
+     if(GET_BUFFER(pI810XvMC, dma) || !dma.granted)
+       FLUSH(pI810XvMC);
+   } /* No DMA granted */
+
+   buf = &(pI810XvMC->dmabufs->list[dma.request_idx]);
+   buf->idx = dma.request_idx;
+   buf->used = 0;
+   buf->total = dma.request_size;
+   buf->address = (drmAddress)dma.virtual;
+   return buf;
+}
+
+/***************************************************************************
+// Function: free_privContext
+// Description: Free's the private context structure if the reference
+//  count is 0.
+***************************************************************************/
+void i810_free_privContext(i810XvMCContext *pI810XvMC) {
+
+  I810_LOCK(pI810XvMC,DRM_LOCK_QUIESCENT);
+
+
+  pI810XvMC->ref--;
+  if(!pI810XvMC->ref) {
+    drmUnmapBufs(pI810XvMC->dmabufs);
+    drmUnmap(pI810XvMC->overlay.address,pI810XvMC->overlay.size);
+    drmUnmap(pI810XvMC->surfaces.address,pI810XvMC->surfaces.size);
+    drmClose(pI810XvMC->fd);
+
+    free(pI810XvMC->dmabufs->list);
+    free(pI810XvMC);
+  }
+
+  I810_UNLOCK(pI810XvMC);
+}
+
+
+/***************************************************************************
+// Function: XvMCCreateContext
+// Description: Create a XvMC context for the given surface parameters.
+// Arguments:
+//   display - Connection to the X server.
+//   port - XvPortID to use as avertised by the X connection.
+//   surface_type_id - Unique identifier for the Surface type.
+//   width - Width of the surfaces.
+//   height - Height of the surfaces.
+//   flags - one or more of the following
+//      XVMC_DIRECT - A direct rendered context is requested.
+//
+// Notes: surface_type_id and width/height parameters must match those
+//        returned by XvMCListSurfaceTypes.
+// Returns: Status
+***************************************************************************/
+Status XvMCCreateContext(Display *display, XvPortID port,
+			 int surface_type_id, int width, int height, int flags,
+			 XvMCContext *context) {  
+  i810XvMCContext *pI810XvMC;
+  char busIdString[10];
+  int priv_count;
+  uint *priv_data;
+  uint magic;
+  Status ret;
+  int major, minor;
+
+  /* Verify Obvious things first */
+  if(context == NULL) {
+    return XvMCBadContext;
+  }
+
+  if(!(flags & XVMC_DIRECT)) {
+    /* Indirect */
+    printf("Indirect Rendering not supported!\nUsing Direct.");
+  }
+
+  /* Limit use to root for now */
+  if(geteuid()) {
+    printf("Use of XvMC on i810 is currently limited to root\n");
+    return BadAccess;
+  }
+
+  /* FIXME: Check $DISPLAY for legal values here */
+
+  context->surface_type_id = surface_type_id;
+  context->width = (unsigned short)width;
+  context->height = (unsigned short)height;
+  context->flags = flags;
+  context->port = port;
+  /* 
+     Width, Height, and flags are checked against surface_type_id
+     and port for validity inside the X server, no need to check
+     here.
+  */
+
+  /* Allocate private Context data */
+  context->privData = (void *)malloc(sizeof(i810XvMCContext));
+  if(!context->privData) {
+    printf("Unable to allocate resources for XvMC context.\n");
+    return BadAlloc;
+  }
+  pI810XvMC = (i810XvMCContext *)context->privData;
+
+
+  /* Verify the XvMC extension exists */
+  if(! XvMCQueryExtension(display, &event_base,
+			  &error_base)) {
+    printf("XvMC Extension is not available!\n");
+    return BadAlloc;
+  }
+  /* Verify XvMC version */
+  ret = XvMCQueryVersion(display, &major, &minor);
+  if(ret) {
+    printf("XvMCQuery Version Failed, unable to determine protocol version\n");
+  }
+  /* FIXME: Check Major and Minor here */
+
+  /* Check for drm */
+  if(! drmAvailable()) {
+    printf("Direct Rendering is not avilable on this system!\n");
+    return BadAlloc;
+  }
+
+  /* 
+     Build the Attribute Atoms, and Initialize the ones that exist
+     in Xv.
+  */
+  pI810XvMC->xv_colorkey = XInternAtom(display,"XV_COLORKEY",0);
+  if(!pI810XvMC->xv_colorkey) {
+    return XvBadPort;
+  }
+  ret = XvGetPortAttribute(display,port,pI810XvMC->xv_colorkey,
+			   &pI810XvMC->colorkey);
+  if(ret) {
+    return ret;
+  }
+  pI810XvMC->xv_brightness = XInternAtom(display,"XV_BRIGHTNESS",0);
+  pI810XvMC->xv_saturation = XInternAtom(display,"XV_SATURATION",0);
+  pI810XvMC->xv_contrast = XInternAtom(display,"XV_CONTRAST",0);
+  pI810XvMC->brightness = 0;
+  pI810XvMC->saturation = 0x80;  /* 1.0 in 3.7 format */
+  pI810XvMC->contrast = 0x40; /* 1.0 in 3.6 format */
+
+  /* Open DRI Device */
+  if((pI810XvMC->fd = drmOpen("i810",NULL)) < 0) {
+    printf("DRM Device for i810 could not be opened.\n");
+    free(busIdString);
+    free(pI810XvMC);
+    return BadAccess;
+  } /* !pI810XvMC->fd */
+
+  /* Get magic number and put it in privData for passing */
+  drmGetMagic(pI810XvMC->fd,&magic);
+  context->flags = (unsigned long)magic;
+
+  /*
+    Pass control to the X server to create a drm_context_t for us and
+    validate the with/height and flags.
+  */
+  if((ret = _xvmc_create_context(display, context, &priv_count, &priv_data))) {
+    printf("Unable to create XvMC Context.\n");
+    return ret;
+  }
+
+  /* 
+     X server returns a structure like this:
+     drm_context_t
+     fbBase
+     OverlayOffset
+     OverlaySize
+     SurfacesOffset
+     SurfacesSize
+     busIdString = 9 char + 1
+  */
+  if(priv_count != 9) {
+    printf("_xvmc_create_context() returned incorrect data size!\n");
+    printf("\tExpected 9, got %d\n",priv_count);
+    _xvmc_destroy_context(display, context);
+    free(pI810XvMC);
+    return BadAlloc;
+  }
+  pI810XvMC->drmcontext = priv_data[0];
+  pI810XvMC->fb_base = priv_data[1];
+  pI810XvMC->overlay.offset = priv_data[2] + priv_data[1];
+  pI810XvMC->overlay.size = priv_data[3];
+  pI810XvMC->surfaces.offset = priv_data[4] + priv_data[1];
+  pI810XvMC->surfaces.size = priv_data[5];
+  strncpy(pI810XvMC->busIdString,(char *)&priv_data[6],9);
+  pI810XvMC->busIdString[9] = '\0';
+
+  /* Must free the private data we were passed from X */
+  free(priv_data);
+  
+  /* Initialize private context values */
+  pI810XvMC->current = 0;
+  pI810XvMC->lock = 0;
+  pI810XvMC->last_flip = 0;
+  pI810XvMC->dual_prime = 0;
+
+  /* 
+     Map dma Buffers: Not really, this would be a drmMapBufs
+     but due to the i810 security model we have to just create an
+     empty data structure to fake it.
+  */
+  pI810XvMC->dmabufs = (drmBufMapPtr)malloc(sizeof(drmBufMap));
+  if(pI810XvMC->dmabufs == NULL) {
+    printf("Dma Bufs could not be mapped.\n");
+    _xvmc_destroy_context(display, context);
+    free(pI810XvMC);
+    return BadAlloc;
+  } /* pI810XvMC->dmabufs == NULL */
+  memset(pI810XvMC->dmabufs, 0, sizeof(drmBufMap));
+  pI810XvMC->dmabufs->list = (drmBufPtr)malloc(sizeof(drmBuf) *
+					       I810_DMA_BUF_NR);
+  if(pI810XvMC->dmabufs->list == NULL) {
+    printf("Dma Bufs could not be mapped.\n");
+    _xvmc_destroy_context(display, context);
+    free(pI810XvMC);
+    return BadAlloc;
+  } /* pI810XvMC->dmabufs->list == NULL */
+  memset(pI810XvMC->dmabufs->list, 0, sizeof(drmBuf) * I810_DMA_BUF_NR);
+  
+  /* Map the Overlay memory */
+  if(drmMap(pI810XvMC->fd,pI810XvMC->overlay.offset,
+	    pI810XvMC->overlay.size,&(pI810XvMC->overlay.address)) < 0) {
+    printf("Unable to map Overlay at offset 0x%x and size 0x%x\n",
+	   (unsigned int)pI810XvMC->overlay.offset,pI810XvMC->overlay.size);
+    _xvmc_destroy_context(display, context);
+    free(pI810XvMC->dmabufs->list);
+    free(pI810XvMC);
+    return BadAlloc;
+  } /* drmMap() < 0 */
+  
+  /* Overlay Regs are offset 1024 into Overlay Map */
+  pI810XvMC->oregs = (i810OverlayRec *)
+    ((unsigned char *)pI810XvMC->overlay.address + 1024);
+
+  /* Map Surfaces */
+  if(drmMap(pI810XvMC->fd,pI810XvMC->surfaces.offset,
+	    pI810XvMC->surfaces.size,&(pI810XvMC->surfaces.address)) < 0) {
+    printf("Unable to map XvMC Surfaces.\n");
+    _xvmc_destroy_context(display, context);
+    free(pI810XvMC->dmabufs->list);
+    free(pI810XvMC);
+    return BadAlloc;
+  } /* drmMap() < 0 */
+
+  /*
+    There is a tiny chance that someone was using the overlay and
+    issued a flip that hasn't finished. To be 100% sure I'll just
+    take the lock and sleep for the worst case time for a flip.
+  */
+  I810_LOCK(pI810XvMC,DRM_LOCK_QUIESCENT);
+  usleep(20000);  /* 1/50th Sec for 50hz refresh */
+
+  /* Set up Overlay regs with Initial Values */
+  pI810XvMC->oregs->YRGB_VPH = 0;
+  pI810XvMC->oregs->UV_VPH = 0;
+  pI810XvMC->oregs->HORZ_PH = 0;
+  pI810XvMC->oregs->INIT_PH = 0;
+  pI810XvMC->oregs->DWINPOS = 0;
+  pI810XvMC->oregs->DWINSZ = (I810_XVMC_MAXHEIGHT << 16) |
+    I810_XVMC_MAXWIDTH;
+  pI810XvMC->oregs->SWID =  I810_XVMC_MAXWIDTH | (I810_XVMC_MAXWIDTH << 15);
+  pI810XvMC->oregs->SWIDQW = (I810_XVMC_MAXWIDTH >> 3) |
+    (I810_XVMC_MAXWIDTH << 12);
+  pI810XvMC->oregs->SHEIGHT = I810_XVMC_MAXHEIGHT |
+    (I810_XVMC_MAXHEIGHT << 15);
+  pI810XvMC->oregs->YRGBSCALE = 0x80004000; /* scale factor 1 */
+  pI810XvMC->oregs->UVSCALE = 0x80004000;   /* scale factor 1 */
+  pI810XvMC->oregs->OV0CLRC0 = 0x4000;      /* brightness: 0 contrast: 1.0 */
+  pI810XvMC->oregs->OV0CLRC1 = 0x80;        /* saturation: bypass */
+  
+  /* Destination Colorkey Setup */
+  pI810XvMC->oregs->DCLRKV = RGB16ToColorKey(pI810XvMC->colorkey);
+  pI810XvMC->oregs->DCLRKM = 0x80070307;
+
+  
+  pI810XvMC->oregs->SCLRKVH = 0;
+  pI810XvMC->oregs->SCLRKVL = 0;
+  pI810XvMC->oregs->SCLRKM = 0; /* source color key disable */
+  pI810XvMC->oregs->OV0CONF = 0; /* two 720 pixel line buffers */
+  
+  pI810XvMC->oregs->OV0CMD = VC_UP_INTERPOLATION | HC_UP_INTERPOLATION | 
+    Y_ADJUST | YUV_420;
+
+  pI810XvMC->ref = 1;
+
+  I810_UNLOCK(pI810XvMC);
+
+  return Success;
+
+}
+
+/***************************************************************************
+// Function: XvMCDestroyContext
+// Description: Destorys the specified context.
+//
+// Arguments:
+//   display - Specifies the connection to the server.
+//   context - The context to be destroyed.
+//
+// Returns: Status
+***************************************************************************/
+Status XvMCDestroyContext(Display *display, XvMCContext *context) {
+  i810XvMCContext *pI810XvMC;
+
+  if(context == NULL) {
+    return (error_base + XvMCBadContext);
+  }
+  if(context->privData == NULL) {
+    return (error_base + XvMCBadContext);
+  }
+  pI810XvMC = (i810XvMCContext *)context->privData;
+
+  /* Turn off the overlay */
+  if(pI810XvMC->last_flip) {
+    I810_LOCK(pI810XvMC,DRM_LOCK_QUIESCENT);
+
+    /* Make sure last flip is done */
+    BLOCK_OVERLAY(pI810XvMC,pI810XvMC->current);
+
+    pI810XvMC->oregs->OV0CMD = VC_UP_INTERPOLATION | HC_UP_INTERPOLATION |
+      Y_ADJUST;
+    pI810XvMC->current = !pI810XvMC->current;
+    if(pI810XvMC->current == 1) {
+      pI810XvMC->oregs->OV0CMD |= BUFFER1_FIELD0;
+    }
+    else {
+      pI810XvMC->oregs->OV0CMD |= BUFFER0_FIELD0;
+    }
+    OVERLAY_FLIP(pI810XvMC);
+    pI810XvMC->last_flip++;
+  
+    /* Wait for the flip */
+    BLOCK_OVERLAY(pI810XvMC,pI810XvMC->current);
+
+    I810_UNLOCK(pI810XvMC);
+  }
+
+  /* Pass Control to the X server to destroy the drm_context_t */
+  _xvmc_destroy_context(display, context);
+
+  i810_free_privContext(pI810XvMC);
+  context->privData = NULL;
+
+  return Success;
+}
+
+
+/***************************************************************************
+// Function: XvMCCreateSurface
+***************************************************************************/
+Status XvMCCreateSurface( Display *display, XvMCContext *context,
+			  XvMCSurface *surface) {
+  i810XvMCContext *pI810XvMC;
+  i810XvMCSurface *pI810Surface;
+  int priv_count;
+  uint *priv_data;
+  Status ret;
+
+  if((surface == NULL) || (context == NULL) || (display == NULL)){
+    return BadValue;
+  }
+  
+  pI810XvMC = (i810XvMCContext *)context->privData;
+  if(pI810XvMC == NULL) {
+    return (error_base + XvMCBadContext);
+  }
+
+
+  surface->privData = (i810XvMCSurface *)malloc(sizeof(i810XvMCSurface));
+  if(!surface->privData) {
+    return BadAlloc;
+  }
+  pI810Surface = (i810XvMCSurface *)surface->privData;
+
+  /* Initialize private values */
+  pI810Surface->privContext = pI810XvMC;
+  pI810Surface->last_render = 0;
+  pI810Surface->last_flip = 0;
+  pI810Surface->second_field = 0;
+
+  if((ret = _xvmc_create_surface(display, context, surface,
+				&priv_count, &priv_data))) {
+    free(pI810Surface);
+    printf("Unable to create XvMCSurface.\n");
+    return ret;
+  }
+
+  /*
+    _xvmc_create_subpicture returns 2 uints with the offset into
+    the DRM map for the Y surface and UV surface.
+  */
+  if(priv_count != 2) {
+    printf("_xvmc_create_surface() return incorrect data size.\n");
+    printf("Expected 2 got %d\n",priv_count);
+    free(priv_data);
+    free(pI810Surface);
+    return BadAlloc;
+  }
+  /* Data == Client Address, offset == Physical address offset */
+  pI810Surface->data = pI810XvMC->surfaces.address;
+  pI810Surface->offset = pI810XvMC->surfaces.offset;
+
+
+  /* 
+     i810's MC Engine needs surfaces of 2^x (x= 9,10,11,12) pitch
+     and the Tiler need 512k aligned surfaces, basically we are
+     stuck with fixed memory with pitch 1024 for Y data. UV = 512.
+  */
+  pI810Surface->pitch = 10;
+  if((surface->surface_type_id == FOURCC_UYVY) ||
+     (surface->surface_type_id == FOURCC_YUY2)) {
+    /* This is not implemented server side. */
+    pI810Surface->pitch++;
+  }
+
+  /*
+    offsets[0,1,2] == Offsets from either data or offset for the Y
+    U and V surfaces.
+  */
+  pI810Surface->offsets[0] = priv_data[0];
+  if(((unsigned long)pI810Surface->data + pI810Surface->offsets[0]) & 4095) {
+    printf("XvMCCreateSurface: Surface offset 0 is not 4096 aligned\n");
+  }
+
+  if((surface->surface_type_id == FOURCC_UYVY) ||
+     (surface->surface_type_id == FOURCC_YUY2)) {
+    /* Packed surface, not fully implemented */
+    pI810Surface->offsets[1] = 0;
+    pI810Surface->offsets[2] = 0;
+  }
+  else {
+    /* Planar surface */
+    pI810Surface->offsets[1] = priv_data[1];
+    if(((unsigned long)pI810Surface->data + pI810Surface->offsets[1]) & 2047) {
+      printf("XvMCCreateSurface: Surface offset 1 is not 2048 aligned\n");
+    }
+
+    pI810Surface->offsets[2] = ((unsigned long)pI810Surface->offsets[1] +
+				(1<<(pI810Surface->pitch - 1)) * 288);
+    if(((unsigned long)pI810Surface->data + pI810Surface->offsets[2]) & 2047) {
+      printf("XvMCCreateSurface: Surface offset 2 is not 2048 aligned\n");
+    }
+
+  }
+
+  /* Free data returned from xvmc_create_surface */
+  free(priv_data);
+
+  /* Clear the surface to 0 */
+  memset((void *)((unsigned long)pI810Surface->data + (unsigned long)pI810Surface->offsets[0]),
+	 0, ((1<<pI810Surface->pitch) * surface->height));
+
+  switch(surface->surface_type_id) {
+  case FOURCC_YV12:
+  case FOURCC_I420:
+    /* Destination buffer info command */
+    pI810Surface->dbi1y = ((((unsigned int)pI810Surface->offset +
+			     pI810Surface->offsets[0]) & ~0xfc000fff) |
+			   (pI810Surface->pitch - 9));
+    pI810Surface->dbi1u = ((((unsigned int)pI810Surface->offset +
+			     pI810Surface->offsets[1]) & ~0xfc000fff) |
+			   (pI810Surface->pitch - 10));
+    pI810Surface->dbi1v = ((((unsigned int)pI810Surface->offset +
+			     pI810Surface->offsets[2]) & ~0xfc000fff) |
+			   (pI810Surface->pitch - 10));
+
+    /* Destination buffer variables command */
+    pI810Surface->dbv1 = (0x8<<20) | (0x8<<16);
+    /* Map info command */
+    pI810Surface->mi1y = (0x1<<24) | (1<<9) | (pI810Surface->pitch - 3);
+    pI810Surface->mi1u = (0x1<<24) | (1<<9) | (pI810Surface->pitch - 4);
+    pI810Surface->mi1v = (0x1<<24) | (1<<9) | (pI810Surface->pitch - 4);
+
+    pI810Surface->mi2y = (((unsigned int)surface->height - 1)<<16) |
+      ((unsigned int)surface->width - 1);
+    pI810Surface->mi2u = (((unsigned int)surface->height - 1)<<15) |
+      (((unsigned int)surface->width - 1)>>1);
+    pI810Surface->mi2v = pI810Surface->mi2u;
+
+    pI810Surface->mi3y = ((unsigned int)pI810Surface->offset +
+			  pI810Surface->offsets[0]) & ~0x0000000f;
+    pI810Surface->mi3u = ((unsigned int)pI810Surface->offset +
+			  pI810Surface->offsets[1]) & ~0x0000000f;
+    pI810Surface->mi3v = ((unsigned int)pI810Surface->offset +
+			  pI810Surface->offsets[2]) & ~0x0000000f;
+    break;
+  case FOURCC_UYVY:
+  case FOURCC_YUY2:
+  default:
+    /* Destination buffer info command */
+    pI810Surface->dbi1y = ((((unsigned int)pI810Surface->offset +
+			     pI810Surface->offsets[0]) & ~0xfc000fff) |
+			   (pI810Surface->pitch - 9));
+    /* Destination buffer variables command */
+    if(surface->surface_type_id == FOURCC_YUY2) {
+      pI810Surface->dbv1 = 0x5<<8;
+      pI810Surface->mi1y = 0x5<<24 | pI810Surface->pitch | 0x1<<21;
+    }
+    else {
+      pI810Surface->dbv1 = 0x4<<8;
+      pI810Surface->mi1y = 0x5<<24 | (pI810Surface->pitch - 3);
+    }
+    pI810Surface->mi2y = (((unsigned int)surface->width - 1)<<16) |
+      ((unsigned int)surface->height - 1);
+    pI810Surface->mi3y = ((unsigned int)pI810Surface->offset +
+			  pI810Surface->offsets[0]) & ~0xfc000fff;
+    break;
+  }
+  pI810XvMC->ref++;
+
+  return Success;
+}
+
+
+/***************************************************************************
+// Function: XvMCDestroySurface
+***************************************************************************/
+Status XvMCDestroySurface(Display *display, XvMCSurface *surface) {
+  i810XvMCSurface *pI810Surface;
+  i810XvMCContext *pI810XvMC;
+
+  if((display == NULL) || (surface == NULL)) {
+    return BadValue;
+  }
+  if(surface->privData == NULL) {
+    return (error_base + XvMCBadSurface);
+  }
+
+  pI810Surface = (i810XvMCSurface *)surface->privData;
+  if(pI810Surface->last_flip) {
+    XvMCSyncSurface(display,surface);
+  }
+  pI810XvMC = (i810XvMCContext *)pI810Surface->privContext;
+
+  _xvmc_destroy_surface(display,surface);
+
+  i810_free_privContext(pI810XvMC);
+
+  free(pI810Surface);
+  surface->privData = NULL;
+  return Success;
+}
+
+/***************************************************************************
+// Function: XvMCCreateBlocks
+***************************************************************************/
+Status XvMCCreateBlocks(Display *display, XvMCContext *context,
+			unsigned int num_blocks,
+			XvMCBlockArray *block) {
+
+  if((display == NULL) || (context == NULL) || (num_blocks == 0)) {
+    return BadValue;
+  }
+
+  block->blocks = (short *)malloc(num_blocks<<6 * sizeof(short));
+  if(block->blocks == NULL) {
+    return BadAlloc;
+  }
+
+  block->num_blocks = num_blocks;
+  block->context_id = context->context_id;
+
+  block->privData = NULL;
+
+  return Success;
+}
+
+/***************************************************************************
+// Function: XvMCDestroyBlocks
+***************************************************************************/
+Status XvMCDestroyBlocks(Display *display, XvMCBlockArray *block) {
+  if(display == NULL) {
+    return BadValue;
+  }
+
+  free(block->blocks);
+  block->num_blocks = 0;
+  block->context_id = 0;
+  block->privData = NULL;
+  return Success;
+}
+
+/***************************************************************************
+// Function: XvMCCreateMacroBlocks
+***************************************************************************/
+Status XvMCCreateMacroBlocks(Display *display, XvMCContext *context,
+			     unsigned int num_blocks,
+			     XvMCMacroBlockArray *blocks) {
+
+  if((display == NULL) || (context == NULL) || (blocks == NULL) ||
+      (num_blocks == 0)) {
+    return BadValue;
+  }
+  memset(blocks,0,sizeof(XvMCMacroBlockArray));
+  blocks->context_id = context->context_id;
+  blocks->privData = NULL;
+  
+  blocks->macro_blocks = (XvMCMacroBlock *)
+    malloc(num_blocks * sizeof(XvMCMacroBlock));
+  if(blocks->macro_blocks == NULL) {
+    return BadAlloc;
+  }
+  blocks->num_blocks = num_blocks;
+
+  return Success;
+}
+
+/***************************************************************************
+// Function: XvMCDestroyMacroBlocks
+***************************************************************************/
+Status XvMCDestroyMacroBlocks(Display *display, XvMCMacroBlockArray *block) {
+  if((display == NULL) || (block == NULL)) {
+    return BadValue;
+  }
+  if(block->macro_blocks) {
+    free(block->macro_blocks);
+  }
+  block->context_id = 0;
+  block->num_blocks = 0;
+  block->privData = NULL;
+
+  return Success;
+}
+
+
+/***************************************************************************
+// Function: dp (Debug Print)
+// Description: This function prints out in hex i * uint32_t at the address
+//  supplied. This enables you to print out the dma buffers from
+//  within the debugger even though they are not in your address space.
+***************************************************************************/
+void dp(unsigned int *address, unsigned int i) {
+  int j;
+
+  printf("DebugPrint:\n");
+  for(j=0; j<i; j++) {
+    printf("0x%8.8x ",address[j]);
+    if(j && !(j & 7)) { printf("\n");}
+  }
+}
+
+/***************************************************************************
+// Macro: PACK_*
+// Description: Packs 16bit signed data from blocks into either 8bit unsigned
+//  intra data or 16bit signed correction data, both packed into
+//  32 bit integers.
+***************************************************************************/
+#define PACK_INTRA_DATA(d,b,n)                  \
+  do {                                          \
+   char *dp = (char *)d;                        \
+   char *bp = (char *)b;                        \
+   int counter;                                 \
+   for(counter = 0; counter < n; counter++) {   \
+     *dp++ = *bp;                               \
+     bp += 2;                                   \
+   }                                            \
+  }while(0);
+
+#define PACK_CORR_DATA(d,b,n)          \
+            memcpy(d,b,n);             \
+            d = (uint *)((unsigned long)d + n);
+
+#define MARK_CORR_DATA(d,n)                          \
+            do {                                     \
+              uint* q = (uint*)((unsigned long)d - n);        \
+              while((unsigned long)q < (unsigned long)d) {             \
+               *q++ += 0x00330033;                   \
+              }                                      \
+	    }while(0);
+
+#define MARK_INTRA_BLOCK(d)         \
+           do {                     \
+             int q;                 \
+             for(q=0; q<16; q++) {  \
+               d[q] += 0x33333333;  \
+             }                      \
+	    }while(0);
+
+/*
+  Used for DCT 1 when we need DCT 0. Instead
+  of reading from one block we read from two and
+  interlace.
+*/
+#define PACK_CORR_DATA_1to0(d,top,bottom)            \
+            do {                                     \
+              short *t = top,*b = bottom;            \
+              PACK_CORR_DATA(d,t,16);                \
+              t = (short *)((unsigned long)t + 16);           \
+              PACK_CORR_DATA(d,b,16);                \
+              b = (short *)((unsigned long)b + 16);           \
+              PACK_CORR_DATA(d,t,16);                \
+              t = (short *)((unsigned long)t + 16);           \
+              PACK_CORR_DATA(d,b,16);                \
+              b = (short *)((unsigned long)b + 16);           \
+              PACK_CORR_DATA(d,t,16);                \
+              t = (short *)((unsigned long)t + 16);           \
+              PACK_CORR_DATA(d,b,16);                \
+              b = (short *)((unsigned long)b + 16);           \
+              PACK_CORR_DATA(d,t,16);                \
+              t = (short *)((unsigned long)t + 16);           \
+              PACK_CORR_DATA(d,b,16);                \
+              b = (short *)((unsigned long)b + 16);           \
+            }while(0);
+
+/* Used for DCT 0 when we need DCT 1. */
+#define PACK_CORR_DATA_0to1(d,top,bottom)            \
+            do{                                      \
+              short *t = top,*b = bottom;            \
+              PACK_CORR_DATA(d,t,16);                \
+              t = (short *)((unsigned long)t + 32);           \
+              PACK_CORR_DATA(d,t,16);                \
+              t = (short *)((unsigned long)t + 32);           \
+              PACK_CORR_DATA(d,t,16);                \
+              t = (short *)((unsigned long)t + 32);           \
+              PACK_CORR_DATA(d,t,16);                \
+              t = (short *)((unsigned long)t + 32);           \
+              PACK_CORR_DATA(d,b,16);                \
+              b = (short *)((unsigned long)b + 32);           \
+              PACK_CORR_DATA(d,b,16);                \
+              b = (short *)((unsigned long)b + 32);           \
+              PACK_CORR_DATA(d,b,16);                \
+              b = (short *)((unsigned long)b + 32);           \
+              PACK_CORR_DATA(d,b,16);                \
+              b = (short *)((unsigned long)b + 32);           \
+            }while(0);
+
+#define PACK_CORR_DATA_SHORT(d,block)                \
+            do {                                     \
+              short *b = block;                      \
+              PACK_CORR_DATA(d,b,16);                \
+              b = (short *)((unsigned long)b + 32);           \
+              PACK_CORR_DATA(d,b,16);                \
+              b = (short *)((unsigned long)b + 32);           \
+              PACK_CORR_DATA(d,b,16);                \
+              b = (short *)((unsigned long)b + 32);           \
+              PACK_CORR_DATA(d,b,16);                \
+              b = (short *)((unsigned long)b + 32);           \
+            }while(0);
+
+/* Lookup tables to speed common calculations */
+static unsigned int drps_table[] = {2<<6,3<<6};
+
+static unsigned int mvfs_table[] = {
+  0x12,
+  0x1a,
+  0x13,
+  0x1b
+};
+
+static unsigned int type_table[] = {
+  0x1<<12,   /* This is an error so make it Forward motion */
+  0x1<<12,
+  0x1<<12,
+  0x1<<12,
+  0x2<<12,
+  0x2<<12,
+  0x3<<12,
+  0x3<<12,
+  0x1<<12,  /* Pattern but no Motion, Make motion Forward */
+  0x1<<12,
+  0x1<<12,
+  0x1<<12,
+  0x2<<12,
+  0x2<<12,
+  0x3<<12,
+  0x3<<12
+};
+
+static unsigned int y_frame_bytes[] = {
+  0,0,0,0,128,128,128,128,
+  128,128,128,128,256,256,256,256,
+  128,128,128,128,256,256,256,256,
+  256,256,256,256,384,384,384,384,
+  128,128,128,128,256,256,256,256,
+  256,256,256,256,384,384,384,384,
+  256,256,256,256,384,384,384,384,
+  384,384,384,384,512,512,512,512
+};
+
+static unsigned int u_frame_bytes[] = {
+  0,0,128,128,0,0,128,128,
+  0,0,128,128,0,0,128,128,
+  0,0,128,128,0,0,128,128,
+  0,0,128,128,0,0,128,128,
+  0,0,128,128,0,0,128,128,
+  0,0,128,128,0,0,128,128,
+  0,0,128,128,0,0,128,128,
+  0,0,128,128,0,0,128,128
+};
+
+static unsigned int v_frame_bytes[] = {
+  0,128,0,128,0,128,0,128,
+  0,128,0,128,0,128,0,128,
+  0,128,0,128,0,128,0,128,
+  0,128,0,128,0,128,0,128,
+  0,128,0,128,0,128,0,128,
+  0,128,0,128,0,128,0,128,
+  0,128,0,128,0,128,0,128,
+  0,128,0,128,0,128,0,128
+};
+
+static unsigned int y_first_field_bytes[] = {
+  0,0,0,0,0,0,0,0,
+  0,0,0,0,0,0,0,0,
+  128,128,128,128,128,128,128,128,
+  128,128,128,128,128,128,128,128,
+  128,128,128,128,128,128,128,128,
+  128,128,128,128,128,128,128,128,
+  256,256,256,256,256,256,256,256,
+  256,256,256,256,256,256,256,256
+};
+
+static unsigned int y_second_field_bytes[] = {
+  0,0,0,0,128,128,128,128,
+  128,128,128,128,256,256,256,256,
+  0,0,0,0,128,128,128,128,
+  128,128,128,128,256,256,256,256,
+  0,0,0,0,128,128,128,128,
+  128,128,128,128,256,256,256,256,
+  0,0,0,0,128,128,128,128,
+  128,128,128,128,256,256,256,256
+};
+
+static unsigned int y_dct0_field_bytes[] = {
+  0,0,0,0,128,128,128,128,
+  128,128,128,128,256,256,256,256,
+  128,128,128,128,128,128,128,128,
+  256,256,256,256,256,256,256,256,
+  128,128,128,128,256,256,256,256,
+  128,128,128,128,256,256,256,256,
+  256,256,256,256,256,256,256,256,
+  256,256,256,256,256,256,256,256
+};
+
+static unsigned int y_dct1_frame_bytes[] = {
+  0,0,0,0,256,256,256,256,
+  256,256,256,256,512,512,512,512,
+  256,256,256,256,256,256,256,256,
+  512,512,512,512,512,512,512,512,
+  256,256,256,256,512,512,512,512,
+  256,256,256,256,512,512,512,512,
+  512,512,512,512,512,512,512,512,
+  512,512,512,512,512,512,512,512
+};
+
+static unsigned int u_field_bytes[] = {
+  0,0,64,64,0,0,64,64,
+  0,0,64,64,0,0,64,64,
+  0,0,64,64,0,0,64,64,
+  0,0,64,64,0,0,64,64,
+  0,0,64,64,0,0,64,64,
+  0,0,64,64,0,0,64,64,
+  0,0,64,64,0,0,64,64,
+  0,0,64,64,0,0,64,64
+};
+
+static unsigned int v_field_bytes[] = {
+  0,64,0,64,0,64,0,64,
+  0,64,0,64,0,64,0,64,
+  0,64,0,64,0,64,0,64,
+  0,64,0,64,0,64,0,64,
+  0,64,0,64,0,64,0,64,
+  0,64,0,64,0,64,0,64,
+  0,64,0,64,0,64,0,64,
+  0,64,0,64,0,64,0,64
+};
+
+static short empty_block[] = {
+  0,0,0,0,0,0,0,0,
+  0,0,0,0,0,0,0,0,
+  0,0,0,0,0,0,0,0,
+  0,0,0,0,0,0,0,0,
+  0,0,0,0,0,0,0,0,
+  0,0,0,0,0,0,0,0,
+  0,0,0,0,0,0,0,0,
+  0,0,0,0,0,0,0,0
+};
+
+
+/***************************************************************************
+// Function: dispatchYContext
+// Description: Allocate a DMA buffer write the Y MC Context info in it,
+//  and dispatch it to hardware.
+***************************************************************************/
+
+static __inline__ void dispatchYContext(i810XvMCSurface *privTarget,
+					i810XvMCSurface *privPast,
+					i810XvMCSurface *privFuture,
+					i810XvMCContext *pI810XvMC) {
+  uint *data;
+  drmBufPtr pDMA;
+  drm_i810_mc_t mc;
+
+  pDMA = i810_get_free_buffer(pI810XvMC);
+  data = pDMA->address;
+  *data++ = CMD_FLUSH;
+  *data++ = BOOLEAN_ENA_2;
+  *data++ = CMD_FLUSH;
+  *data++ = DEST_BUFFER_INFO;
+  *data++ = privTarget->dbi1y;
+  *data++ = DEST_BUFFER_VAR;
+  *data++ = privTarget->dbv1;
+  /* Past Surface */
+  *data++ = CMD_MAP_INFO;
+  *data++ = privPast->mi1y;
+  *data++ = privPast->mi2y;
+  *data++ = privPast->mi3y;
+  /* Future Surface */
+  *data++ = CMD_MAP_INFO;
+  *data++ = privFuture->mi1y | 0x1<<28;
+  *data++ = privFuture->mi2y;
+  *data++ = privFuture->mi3y;
+
+  mc.idx = pDMA->idx;
+  mc.used = (unsigned long)data - (unsigned long)pDMA->address;
+  mc.last_render = ++pI810XvMC->last_render;
+  privTarget->last_render = pI810XvMC->last_render;
+  I810_MC(pI810XvMC,mc);
+}
+
+static __inline__ void renderError(void) {
+  printf("Invalid Macroblock Parameters found.\n");
+  return;
+}
+
+/***************************************************************************
+// Function: renderIntrainFrame
+// Description: inline function that sets hardware parameters for an Intra
+//  encoded macroblock in a Frame picture.
+***************************************************************************/
+static __inline__ void renderIntrainFrame(uint **datay,uint **datau,
+					  uint **datav,
+					  XvMCMacroBlock *mb,
+					  short *block_ptr) {
+
+  register uint *dy = *datay;
+  register uint *du = *datau;
+  register uint *dv = *datav;
+
+  /* Y Blocks */
+  *dy++ = GFXBLOCK + 68;
+  *dy++ = (1<<30) | (3<<28) | (0xf<<24);
+  *dy++ = ((uint)mb->x<<20) | ((uint)mb->y<<4);
+  *dy++ = (16<<16) | 16;
+  *dy++ = 0;
+  *dy++ = 0;
+  PACK_INTRA_DATA(dy,block_ptr,256);
+  dy += 64;
+  block_ptr += 256;
+  /* End Y Blocks */
+
+  /* U Block */
+  *du++ = GFXBLOCK + 20;
+  *du++ = (2<<30) | (1<<28) | (1<<23);
+  *du++ = (((uint)mb->x)<<19) | (((uint)mb->y)<<3);
+  *du++ = (8<<16) | 8;
+  *du++ = 0;
+  *du++ = 0;
+  PACK_INTRA_DATA(du,block_ptr,64);
+  du += 16;
+  block_ptr += 64;
+  
+  /* V Block */
+  *dv++ = GFXBLOCK + 20;
+  *dv++ = (3<<30) | (1<<28) | (1<<22);
+  *dv++ = (((uint)mb->x)<<19) | (((uint)mb->y)<<3);
+  *dv++ = (8<<16) | 8;
+  *dv++ = 0;
+  *dv++ = 0;
+  PACK_INTRA_DATA(dv,block_ptr,64);
+  dv += 16;
+  block_ptr += 64;
+
+  *datay = dy;
+  *datau = du;
+  *datav = dv;
+}
+
+/***************************************************************************
+// Function: renderIntrainFrameDCT1
+// Description: inline function that sets hardware parameters for an Intra
+//  encoded macroblock in a Frame picture with DCT type 1.
+***************************************************************************/
+static __inline__ void renderIntrainFrameDCT1(uint **datay,uint **datau,
+					      uint **datav,XvMCMacroBlock *mb,
+					      short *block_ptr,uint flags) {
+
+  register uint *dy = *datay;
+  register uint *du = *datau;
+  register uint *dv = *datav;
+
+
+  /* Y Blocks */
+  *dy++ = GFXBLOCK + 36;
+  *dy++ = (1<<30) | (2<<28) | (0x3<<26) | (0x2<<6);
+  *dy++ = ((uint)mb->x<<20) | ((uint)mb->y<<3);
+  *dy++ = (8<<16) | 16;
+  *dy++ = 0;
+  *dy++ = 0;
+  PACK_INTRA_DATA(dy,block_ptr,128);
+  dy += 32;
+  block_ptr += 128;
+
+  /* Second Y block */
+  *dy++ = GFXBLOCK + 36;
+  *dy++ = (1<<30) | (2<<28) | (0x3<<26) | (0x3<<6);
+  *dy++ = ((uint)mb->x<<20) | ((uint)mb->y<<3);
+  *dy++ = (8<<16) | 16;
+  *dy++ = 0;
+  *dy++ = 0;
+  PACK_INTRA_DATA(dy,block_ptr,128);
+  dy += 32;
+  block_ptr += 128;
+  /* End Y Blocks */
+
+    
+  /* U Block */
+  *du++ = GFXBLOCK + 20;
+  *du++ = (2<<30) | (1<<28) | (1<<23);
+  *du++ = (((uint)mb->x)<<19) | (((uint)mb->y)<<3);
+  *du++ = (8<<16) | 8;
+  *du++ = 0;
+  *du++ = 0;
+  PACK_INTRA_DATA(du,block_ptr,64);
+  du += 16;
+  block_ptr += 64;
+  
+  /* V Block */
+  *dv++ = GFXBLOCK + 20;
+  *dv++ = (3<<30) | (1<<28) | (1<<22);
+  *dv++ = (((uint)mb->x)<<19) | (((uint)mb->y)<<3);
+  *dv++ = (8<<16) | 8;
+  *dv++ = 0;
+  *dv++ = 0;
+  PACK_INTRA_DATA(dv,block_ptr,64);
+  dv += 16;
+  block_ptr += 64;
+
+  *datay = dy;
+  *datau = du;
+  *datav = dv;
+}
+
+
+/***************************************************************************
+// Function: renderIntrainField
+// Description: inline function that sets hardware parameters for an Intra
+//  encoded macroblock in Field pictures.
+***************************************************************************/
+static __inline__ void renderIntrainField(uint **datay,uint **datau,
+					  uint **datav,
+					  XvMCMacroBlock *mb,short *block_ptr,
+					  uint ps) {
+
+  register uint *dy = *datay;
+  register uint *du = *datau;
+  register uint *dv = *datav;
+
+  uint xy = ((uint)mb->x<<20) | ((uint)mb->y<<4);
+  uint dw1 = drps_table[~ps & 0x1];
+
+  /* Y Blocks */
+  *dy++ = GFXBLOCK + 68;
+  *dy++ = (1<<30) | (3<<28) | (0xf<<24) | dw1;
+  *dy++ = xy;
+  *dy++ = (16<<16) | 16;
+  *dy++ = 0;
+  *dy++ = 0;
+  PACK_INTRA_DATA(dy,block_ptr,256);
+  dy += 64;
+  block_ptr += 256;
+  /* End Y Blocks */
+
+  xy >>= 1;
+
+  /* U Block */
+  *du++ = GFXBLOCK + 20;
+  *du++ = (2<<30) | (1<<28) | (1<<23) | dw1;
+  *du++ = xy;
+  *du++ = (8<<16) | 8;
+  *du++ = 0;
+  *du++ = 0;
+  PACK_INTRA_DATA(du,block_ptr,64);
+  du += 16;
+  block_ptr += 64;
+  
+  /* V Block */
+  *dv++ = GFXBLOCK + 20;
+  *dv++ = (3<<30) | (1<<28) | (1<<22) | dw1;
+  *dv++ = xy;
+  *dv++ = (8<<16) | 8;
+  *dv++ = 0;
+  *dv++ = 0;
+  PACK_INTRA_DATA(dv,block_ptr,64);
+  dv += 16;
+  block_ptr += 64;
+
+  *datay = dy;
+  *datau = du;
+  *datav = dv;
+}
+
+
+/***************************************************************************
+// Function: renderFieldinField
+// Description: inline function that sets hardware parameters for a Field
+//  encoded macroblock in a Field Picture.
+***************************************************************************/
+static __inline__ void renderFieldinField(uint **datay,uint **datau,
+					  uint **datav,
+					  XvMCMacroBlock *mb,short *block_ptr,
+					  uint ps, uint flags) {
+
+  register uint *dy = *datay;
+  register uint *du = *datau;
+  register uint *dv = *datav;
+
+  /* Motion Vectors */
+  short fmv[2];
+  short bmv[2];
+  /* gfxblock dword 1 */
+  uint dw1;
+
+  uint parity = ~ps & XVMC_TOP_FIELD;
+
+  uint ysize = y_frame_bytes[mb->coded_block_pattern];
+  uint usize = u_frame_bytes[mb->coded_block_pattern];
+  uint vsize = v_frame_bytes[mb->coded_block_pattern];
+
+  uint xy = ((uint)mb->x<<20) | ((uint)mb->y<<4);
+
+  /* i810 Specific flag used to identify the second field in a P frame */
+  if(flags & 0x80000000) {
+    /* P Frame */
+    if((mb->motion_vertical_field_select & XVMC_SELECT_FIRST_FORWARD) ==
+      parity) {
+      /* Same parity, use reference field (map0) */
+      dw1 = 1<<12 | ((0x2 + parity)<<6) | ((0x2 + parity)<<3) |
+	(((uint)mb->coded_block_pattern)<<22);
+      fmv[0] = mb->PMV[0][0][1];
+      fmv[1] = mb->PMV[0][0][0];
+      bmv[0] = 0;
+      bmv[1] = 0;
+    }
+    else {
+      /*
+	Opposite parity, set up as if it were backward
+	motion and use map1.
+      */
+      dw1 = 2<<12 | ((0x2 + parity)<<6) | (0x3 - parity) |
+	(((uint)mb->coded_block_pattern)<<22);
+      bmv[0] = mb->PMV[0][0][1];
+      bmv[1] = mb->PMV[0][0][0];
+      fmv[0] = 0;
+      fmv[1] = 0;
+    }
+  }
+  else {
+    dw1 = type_table[mb->macroblock_type & 0xf] |
+      drps_table[~ps & 0x1] |
+      mvfs_table[mb->motion_vertical_field_select & 3] |
+      (((uint)mb->coded_block_pattern)<<22);
+    
+    fmv[0] = mb->PMV[0][0][1];
+    fmv[1] = mb->PMV[0][0][0];
+    
+    bmv[0] = mb->PMV[0][1][1];
+    bmv[1] = mb->PMV[0][1][0];
+  }
+
+  /* Y Block */
+  *dy++ = GFXBLOCK + 4 + (ysize>>2);
+  *dy++ = (1<<30) | (3<<28) | dw1;
+  *dy++ = xy;
+  *dy++ = (16<<16) | 16;
+  *dy++ = *(uint *)fmv;
+  *dy++ = *(uint *)bmv;
+  PACK_CORR_DATA(dy,block_ptr,ysize);
+  block_ptr = (short *)((unsigned long)block_ptr + ysize);
+  /* End Y Blocks */
+
+  fmv[0] /= 2;
+  fmv[1] /= 2;
+  bmv[0] /= 2;
+  bmv[1] /= 2;
+  xy >>= 1;
+
+  /* U Block */
+  *du++ = GFXBLOCK + 4 + (usize>>2);
+  *du++ = (2<<30) | (1<<28) | dw1;
+  *du++ = xy;
+  *du++ = (8<<16) | 8;
+  *du++ = *(uint *)fmv;
+  *du++ = *(uint *)bmv;
+  PACK_CORR_DATA(du,block_ptr,usize);
+  block_ptr = (short *)((unsigned long)block_ptr + usize);
+
+  /* V Block */
+  *dv++ = GFXBLOCK + 4 + (vsize>>2);
+  *dv++ = (3<<30) | (1<<28) | dw1;
+  *dv++ = xy;
+  *dv++ = (8<<16) | 8;
+  *dv++ = *(uint *)fmv;
+  *dv++ = *(uint *)bmv;
+  PACK_CORR_DATA(dv,block_ptr,vsize);
+  block_ptr = (short *)((unsigned long)block_ptr + vsize);
+
+  *datay = dy;
+  *datau = du;
+  *datav = dv;
+}
+
+/***************************************************************************
+// Function: render16x8inField
+// Description: inline function that sets hardware parameters for a 16x8
+//  encoded macroblock in a field picture.
+***************************************************************************/
+static __inline__ void render16x8inField(uint **datay,uint **datau,
+					 uint **datav,
+					 XvMCMacroBlock *mb,short *block_ptr,
+					 uint ps, uint flags) {
+
+  register uint *dy = *datay;
+  register uint *du = *datau;
+  register uint *dv = *datav;
+
+  /* Motion Vectors */
+  short fmv[4];
+  short bmv[4];
+  /* gfxblock dword 1 */
+  uint dw1[2];
+
+  uint y1size = y_first_field_bytes[mb->coded_block_pattern];
+  uint y2size = y_second_field_bytes[mb->coded_block_pattern];
+  uint usize = u_field_bytes[mb->coded_block_pattern];
+  uint vsize = v_field_bytes[mb->coded_block_pattern];
+
+  uint parity = ~ps & XVMC_TOP_FIELD;
+
+  uint xy = ((uint)mb->x<<20) | ((uint)mb->y<<4);
+
+  /* i810 Specific flag used to identify the second field in a P frame */
+  if(flags & 0x80000000) {
+    /* P Frame */
+    if((mb->motion_vertical_field_select & XVMC_SELECT_FIRST_FORWARD) ==
+      parity) {
+      /* Same parity, use reference field (map0) */
+      dw1[0] = 1<<12 | ((0x2 + parity)<<6) | ((0x2 + parity)<<3) |
+	(((uint)mb->coded_block_pattern)<<22);
+
+      fmv[0] = mb->PMV[0][0][1];
+      fmv[1] = mb->PMV[0][0][0];
+      bmv[0] = 0;
+      bmv[1] = 0;
+    }
+    else {
+      /*
+	Opposite parity, set up as if it were backward
+	motion and use map1.
+      */
+      dw1[0] = 2<<12 | ((0x2 + parity)<<6) | (0x3 - parity) |
+	(((uint)mb->coded_block_pattern)<<22);
+
+      bmv[0] = mb->PMV[0][0][1];
+      bmv[1] = mb->PMV[0][0][0];
+      fmv[0] = 0;
+      fmv[1] = 0;
+    }
+    if((mb->motion_vertical_field_select & XVMC_SELECT_SECOND_FORWARD) ==
+       (parity<<2)) {
+      /* Same parity, use reference field (map0) */
+      dw1[1] = 1<<12 | ((0x2 + parity)<<6) | ((0x2 + parity)<<3) |
+	((((uint)mb->coded_block_pattern<<22) & (0x3<<22)) |
+	 (((uint)mb->coded_block_pattern<<24) & (0x3<<26)));
+
+      fmv[2] = mb->PMV[1][0][1];
+      fmv[3] = mb->PMV[1][0][0];
+      bmv[2] = 0;
+      bmv[3] = 0;
+    }
+    else {
+      /* 
+	 Opposite parity, set up as if it were backward
+	 motion and use map1.
+      */
+      dw1[1] = 2<<12 | ((0x2 + parity)<<6) | (0x3 - parity) |
+	((((uint)mb->coded_block_pattern<<22) & (0x3<<22)) |
+	 (((uint)mb->coded_block_pattern<<24) & (0x3<<26)));
+
+      bmv[2] = mb->PMV[1][0][1];
+      bmv[3] = mb->PMV[1][0][0];
+      fmv[2] = 0;
+      fmv[3] = 0;
+    }
+  }
+  else {
+    dw1[0] = type_table[mb->macroblock_type & 0xf] |
+      drps_table[~ps & 0x1] |
+      mvfs_table[mb->motion_vertical_field_select & 3] |
+      (((uint)mb->coded_block_pattern)<<22);
+
+    dw1[1] = type_table[mb->macroblock_type & 0xf] |
+      drps_table[~ps & 0x1] |
+      mvfs_table[(mb->motion_vertical_field_select>>2) & 0x3] |
+      ((((uint)mb->coded_block_pattern<<22) & (0x3<<22)) |
+       (((uint)mb->coded_block_pattern<<24) & (0x3<<26)));
+
+    fmv[0] = mb->PMV[0][0][1];
+    fmv[1] = mb->PMV[0][0][0];
+    fmv[2] = mb->PMV[1][0][1];
+    fmv[3] = mb->PMV[1][0][0];
+    
+    bmv[0] = mb->PMV[0][1][1];
+    bmv[1] = mb->PMV[0][1][0];
+    bmv[2] = mb->PMV[1][1][1];
+    bmv[3] = mb->PMV[1][1][0];
+  }
+
+  /* First Y Block */
+  *dy++ = GFXBLOCK + 4 + (y1size>>2);
+  *dy++ = (1<<30) | (2<<28) | dw1[0];
+  *dy++ = xy;
+  *dy++ = (8<<16) | 16;
+  *dy++ = *(uint *)fmv;
+  *dy++ = *(uint *)bmv;
+  PACK_CORR_DATA(dy,block_ptr,y1size);
+  block_ptr = (short *)((unsigned long)block_ptr + y1size);
+
+  /* Second Y Block */
+  *dy++ = GFXBLOCK + 4 + (y2size>>2);
+  *dy++ = (1<<30) | (2<<28) | dw1[1];
+  *dy++ = (xy + 8);
+  *dy++ = (8<<16) | 16;
+  *dy++ = *(uint *)&fmv[2];
+  *dy++ = *(uint *)&bmv[2];
+  PACK_CORR_DATA(dy,block_ptr,y2size);
+  block_ptr = (short *)((unsigned long)block_ptr + y2size);
+  /* End Y Blocks */
+
+  fmv[0] /= 2;
+  fmv[1] /= 2;
+  fmv[2] /= 2;
+  fmv[3] /= 2;
+  
+  bmv[0] /= 2;
+  bmv[1] /= 2;
+  bmv[2] /= 2;
+  bmv[3] /= 2;
+
+  xy >>= 1;
+
+  /* U Blocks */
+  *du++ = GFXBLOCK + 4 + (usize>>2);
+  *du++ = (2<<30) | (1<<28) | dw1[0];
+  *du++ = xy;
+  *du++ = (4<<16) | 8;
+  *du++ = *(uint *)fmv;
+  *du++ = *(uint *)bmv;
+  PACK_CORR_DATA(du,block_ptr,usize);
+  block_ptr = (short *)((unsigned long)block_ptr + usize);
+
+  /* Second U block */
+  *du++ = GFXBLOCK + 4 + (usize>>2);
+  *du++ = (2<<30) | (1<<28) | dw1[1];
+  *du++ = (xy + 4);
+  *du++ = (4<<16) | 8;
+  *du++ = *(uint *)&fmv[2];
+  *du++ = *(uint *)&bmv[2];
+  PACK_CORR_DATA(du,block_ptr,usize);
+  block_ptr = (short *)((unsigned long)block_ptr + usize);
+  /* End U Blocks */
+
+  /* V Blocks */
+  *dv++ = GFXBLOCK + 4 + (vsize>>2);
+  *dv++ = (3<<30) | (1<<28) | dw1[0];
+  *dv++ = xy;
+  *dv++ = (4<<16) | 8;
+  *dv++ = *(uint *)fmv;
+  *dv++ = *(uint *)bmv;
+  PACK_CORR_DATA(dv,block_ptr,vsize);
+  block_ptr = (short *)((unsigned long)block_ptr + vsize);
+
+  /* Second V Block */
+  *dv++ = GFXBLOCK + 4 + (vsize>>2);
+  *dv++ = (3<<30) | (1<<28) | dw1[1];
+  *dv++ = (xy + 4);
+  *dv++ = (4<<16) | 8;
+  *dv++ = *(uint *)&fmv[2];
+  *dv++ = *(uint *)&bmv[2];
+  PACK_CORR_DATA(dv,block_ptr,vsize);
+  block_ptr = (short *)((unsigned long)block_ptr + vsize);
+  /* End V Blocks */
+
+  *datay = dy;
+  *datau = du;
+  *datav = dv;
+}
+
+/***************************************************************************
+// Function: renderDualPrimeinField
+// Description: inline function that sets hardware parameters for a Dual
+//  prime encoded macroblock in a field picture.
+***************************************************************************/
+static __inline__ void renderDualPrimeinField(uint **datay,uint **datau,
+					      uint **datav,XvMCMacroBlock *mb,
+					      short *block_ptr,uint ps,
+					      uint flags) {
+
+  register uint *dy = *datay;
+  register uint *du = *datau;
+  register uint *dv = *datav;
+
+  /* Motion Vectors */
+  short fmv[2];
+  short bmv[2];
+  /* gfxblock dword 1 */
+  uint dw1;
+
+
+  uint ysize = y_frame_bytes[mb->coded_block_pattern];
+  uint usize = u_frame_bytes[mb->coded_block_pattern];
+  uint vsize = v_frame_bytes[mb->coded_block_pattern];
+
+  uint xy = ((uint)mb->x<<20) | ((uint)mb->y<<4);
+
+
+  if(ps & XVMC_TOP_FIELD) {
+    dw1 = (mb->coded_block_pattern<<22) | 3<<12 | 2<<6 | 2<<3 | 3;
+  }
+  else {
+    dw1 = (mb->coded_block_pattern<<22) | 3<<12 | 3<<6 | 3<<3 | 2;
+  }
+  fmv[0] = mb->PMV[0][0][1];
+  fmv[1] = mb->PMV[0][0][0];
+  bmv[0] = mb->PMV[0][1][1];
+  bmv[1] = mb->PMV[0][1][0];
+
+  /* Y Block */
+  *dy++ = GFXBLOCK + 4 + (ysize>>2);
+  *dy++ = (1<<30) | (3<<28) | dw1;
+  *dy++ = xy;
+  *dy++ = (16<<16) | 16;
+  *dy++ = *(uint *)fmv;
+  *dy++ = *(uint *)bmv;
+  PACK_CORR_DATA(dy,block_ptr,ysize);
+  block_ptr = (short *)((unsigned long)block_ptr + ysize);
+  /* End Y Blocks */
+ 
+  fmv[0] /= 2;
+  fmv[1] /= 2;
+  bmv[0] /= 2;
+  bmv[1] /= 2;
+  xy >>= 1;
+
+  /* U Block */
+  *du++ = GFXBLOCK + 4 + (usize>>2);
+  *du++ = (2<<30) | (1<<28) | dw1;
+  *du++ = xy;
+  *du++ = (8<<16) | 8;
+  *du++ = *(uint *)fmv;
+  *du++ = *(uint *)bmv;
+  PACK_CORR_DATA(du,block_ptr,usize);
+  block_ptr = (short *)((unsigned long)block_ptr + usize);
+
+  /* V Block */
+  *dv++ = GFXBLOCK + 4 + (vsize>>2);
+  *dv++ = (3<<30) | (1<<28) | dw1;
+  *dv++ = xy;
+  *dv++ = (8<<16) | 8;
+  *dv++ = *(uint *)fmv;
+  *dv++ = *(uint *)bmv;
+  PACK_CORR_DATA(dv,block_ptr,vsize);
+  block_ptr = (short *)((unsigned long)block_ptr + vsize);
+
+  *datay = dy;
+  *datau = du;
+  *datav = dv;
+}
+
+/***************************************************************************
+// Function: renderFieldinFrame
+// Description: inline function that sets hardware parameters for a Field
+//  encoded macroblock in a frame picture.
+***************************************************************************/
+static __inline__ void renderFieldinFrame(uint **datay,uint **datau,
+					  uint **datav,
+					  XvMCMacroBlock *mb,short *block_ptr,
+					  uint flags) {
+
+  register uint *dy = *datay;
+  register uint *du = *datau;
+  register uint *dv = *datav;
+
+  /* Motion Vectors */
+  short fmv[4];
+  short bmv[4];
+  /* gfxblock dword 1 */
+  uint dw1[2];
+
+  uint y1size = y_first_field_bytes[mb->coded_block_pattern];
+  uint y2size = y_second_field_bytes[mb->coded_block_pattern];
+  uint usize = u_field_bytes[mb->coded_block_pattern];
+  uint vsize = v_field_bytes[mb->coded_block_pattern];
+
+  uint xy = ((uint)mb->x<<20) | ((uint)mb->y<<3);
+
+  dw1[0] = type_table[mb->macroblock_type & 0xf] | (0x2<<6) |
+    mvfs_table[mb->motion_vertical_field_select & 3] |
+    (((uint)mb->coded_block_pattern)<<22);
+  
+  dw1[1] = type_table[mb->macroblock_type & 0xf] | (0x3<<6) |
+    mvfs_table[mb->motion_vertical_field_select>>2] |
+    (((mb->coded_block_pattern & 0x3) |
+      ((mb->coded_block_pattern & 0xc)<<2))<<22);
+
+  fmv[0] = mb->PMV[0][0][1]/2;
+  fmv[1] = mb->PMV[0][0][0];
+  fmv[2] = mb->PMV[1][0][1]/2;
+  fmv[3] = mb->PMV[1][0][0];
+  
+  bmv[0] = mb->PMV[0][1][1]/2;
+  bmv[1] = mb->PMV[0][1][0];
+  bmv[2] = mb->PMV[1][1][1]/2;
+  bmv[3] = mb->PMV[1][1][0];
+
+  /* First Y Block */
+  *dy++ = GFXBLOCK + 4 + (y1size>>2);
+  *dy++ = (1<<30) | (2<<28) | dw1[0];
+  *dy++ = xy;
+  *dy++ = (8<<16) | 16;
+  *dy++ = *(uint *)&fmv[0];
+  *dy++ = *(uint *)&bmv[0];
+  PACK_CORR_DATA(dy,block_ptr,y1size);
+  block_ptr = (short *)((unsigned long)block_ptr + y1size);
+
+  /* Second Y Block */
+  *dy++ = GFXBLOCK + 4 + (y2size>>2);
+  *dy++ = (1<<30) | (2<<28) | dw1[1];
+  *dy++ = xy;
+  *dy++ = (8<<16) | 16;
+  *dy++ = *(uint *)&fmv[2];
+  *dy++ = *(uint *)&bmv[2];
+  PACK_CORR_DATA(dy,block_ptr,y2size);
+  block_ptr = (short *)((unsigned long)block_ptr + y2size);
+  /* End Y Blocks */
+
+  fmv[0] /= 2;
+  fmv[1] /= 2;
+  fmv[2] /= 2;
+  fmv[3] /= 2;
+  
+  bmv[0] /= 2;
+  bmv[1] /= 2;
+  bmv[2] /= 2;
+  bmv[3] /= 2;
+
+  xy >>= 1;
+
+  /* U Blocks */
+  *du++ = GFXBLOCK + 4 + (usize>>2);
+  *du++ = (2<<30) | (1<<28) | dw1[0];
+  *du++ = xy;
+  *du++ = (4<<16) | 8;
+  *du++ = *(uint *)&fmv[0];
+  *du++ = *(uint *)&bmv[0];
+  if(usize) {
+    PACK_CORR_DATA_SHORT(du,block_ptr);
+  }
+
+  /* Second U Block */
+  *du++ = GFXBLOCK + 4 + (usize>>2);
+  *du++ = (2<<30) | (1<<28) | dw1[1];
+  *du++ = xy;
+  *du++ = (4<<16) | 8;
+  *du++ = *(uint *)&fmv[2];
+  *du++ = *(uint *)&bmv[2];
+  if(usize) {
+    block_ptr = (short *)((unsigned long)block_ptr + 16);
+    PACK_CORR_DATA_SHORT(du,block_ptr);
+    block_ptr = (short *)((unsigned long)block_ptr + 112);
+  }
+  /* End U Blocks */
+
+  /* V Blocks */
+  *dv++ = GFXBLOCK + 4 + (vsize>>2);
+  *dv++ = (3<<30) | (1<<28) | dw1[0];
+  *dv++ = xy;
+  *dv++ = (4<<16) | 8;
+  *dv++ = *(uint *)&fmv[0];
+  *dv++ = *(uint *)&bmv[0];
+  if(vsize) {
+    PACK_CORR_DATA_SHORT(dv,block_ptr);
+  }
+
+  /* Second V Block */
+  *dv++ = GFXBLOCK + 4 + (vsize>>2);
+  *dv++ = (3<<30) | (1<<28) | dw1[1];
+  *dv++ = xy;
+  *dv++ = (4<<16) | 8;
+  *dv++ = *(uint *)&fmv[2];
+  *dv++ = *(uint *)&bmv[2];
+  if(vsize) {
+    block_ptr = (short *)((unsigned long)block_ptr + 16);
+    PACK_CORR_DATA_SHORT(dv,block_ptr);
+    block_ptr = (short *)((unsigned long)block_ptr + 112);
+  }
+  /* End V Blocks */
+
+  *datay = dy;
+  *datau = du;
+  *datav = dv;
+}
+
+/***************************************************************************
+// Function: renderFieldinFrameDCT0
+// Description: inline function that sets hardware parameters for a Field
+//  encoded macroblock in a frame picture with DCT0.
+***************************************************************************/
+static __inline__ void renderFieldinFrameDCT0(uint **datay,uint **datau,
+					      uint **datav,XvMCMacroBlock *mb,
+					      short *block_ptr,uint flags) {
+
+  register uint *dy = *datay;
+  register uint *du = *datau;
+  register uint *dv = *datav;
+
+  /* Motion Vectors */
+  short fmv[4];
+  short bmv[4];
+  /* CBP */
+  uint cbp = (uint)mb->coded_block_pattern;
+  /* gfxblock dword 1 */
+  uint dw1[2];
+
+  short * top_left_b = NULL;
+  short * top_right_b = NULL;
+  short * bottom_left_b = NULL;
+  short * bottom_right_b = NULL;
+
+  unsigned int ysize = y_dct0_field_bytes[cbp];
+  unsigned int usize = u_field_bytes[cbp];
+  unsigned int vsize = v_field_bytes[cbp];
+
+  uint xy = ((uint)mb->x<<20) | ((uint)mb->y<<3);
+
+  dw1[0] = type_table[mb->macroblock_type & 0xf] | (0x2<<6) |
+    mvfs_table[mb->motion_vertical_field_select & 3] |
+    ((cbp | ((cbp<<2) & 0x30))<<22);
+  
+  dw1[1] = type_table[mb->macroblock_type & 0xf] | (0x3<<6) |
+    mvfs_table[mb->motion_vertical_field_select>>2] |
+    ((cbp | ((cbp<<2) & 0x30))<<22);
+
+
+  fmv[0] = mb->PMV[0][0][1]/2;
+  fmv[1] = mb->PMV[0][0][0];
+  fmv[2] = mb->PMV[1][0][1]/2;
+  fmv[3] = mb->PMV[1][0][0];
+  
+  bmv[0] = mb->PMV[0][1][1]/2;
+  bmv[1] = mb->PMV[0][1][0];
+  bmv[2] = mb->PMV[1][1][1]/2;
+  bmv[3] = mb->PMV[1][1][0];
+
+  /*
+    The i810 cannot use DCT0 directly with field motion, we have to
+    interlace the data for it. We use a zero block when the CBP has
+    one half of the to-be-interlaced data but not the other half.
+  */
+  top_left_b = &empty_block[0];
+  if(cbp & 0x20) {
+    top_left_b = block_ptr;
+    block_ptr += 64;
+  }
+
+  top_right_b = &empty_block[0];
+  if(cbp & 0x10) {
+    top_right_b = block_ptr;
+    block_ptr += 64;
+  }
+
+  bottom_left_b = &empty_block[0];
+  if(cbp & 0x8) {
+    bottom_left_b = block_ptr;
+    block_ptr += 64;
+  }
+
+  bottom_right_b = &empty_block[0];
+  if(cbp & 0x4) {
+    bottom_right_b = block_ptr;
+    block_ptr += 64;
+  }
+
+  /* First Y Block */
+  *dy++ = GFXBLOCK + 4 + (ysize>>2);
+  *dy++ = (1<<30) | (2<<28) | dw1[0];
+  *dy++ = xy;
+  *dy++ = (8<<16) | 16;
+  *dy++ = *(uint *)&fmv[0];
+  *dy++ = *(uint *)&bmv[0];
+  if(dw1[0] & (1<<27)) {
+    PACK_CORR_DATA_0to1(dy,top_left_b,bottom_left_b);
+  }
+  if(dw1[0] & (1<<26)) {
+    PACK_CORR_DATA_0to1(dy,top_right_b,bottom_right_b);
+  }
+
+  /* Second Y Block */
+  *dy++ = GFXBLOCK + 4 + (ysize>>2);
+  *dy++ = (1<<30) | (2<<28) | dw1[1];
+  *dy++ = xy;
+  *dy++ = (8<<16) | 16;
+  *dy++ = *(uint *)&fmv[2];
+  *dy++ = *(uint *)&bmv[2];
+  if(dw1[1] & (1<<27)) {
+    top_left_b = (short *)((unsigned long)top_left_b + 16);
+    bottom_left_b = (short *)((unsigned long)bottom_left_b + 16);
+    PACK_CORR_DATA_0to1(dy,top_left_b,bottom_left_b);
+  }
+  if(dw1[1] & (1<<26)) {
+    top_right_b = (short *)((unsigned long)top_right_b + 16);
+    bottom_right_b = (short *)((unsigned long)bottom_right_b + 16);
+    PACK_CORR_DATA_0to1(dy,top_right_b,bottom_right_b);
+  }
+  /* End Y Blocks */
+
+  fmv[0] /= 2;
+  fmv[1] /= 2;
+  fmv[2] /= 2;
+  fmv[3] /= 2;
+  
+  bmv[0] /= 2;
+  bmv[1] /= 2;
+  bmv[2] /= 2;
+  bmv[3] /= 2;
+
+  xy >>= 1;
+
+  /* U Blocks */
+  *du++ = GFXBLOCK + 4 + (usize>>2);
+  *du++ = (2<<30) | (1<<28) | dw1[0];
+  *du++ = xy;
+  *du++ = (4<<16) | 8;
+  *du++ = *(uint *)&fmv[0];
+  *du++ = *(uint *)&bmv[0];
+  if(usize) {
+    PACK_CORR_DATA_SHORT(du,block_ptr);
+  }
+
+  /* Second U Block */
+  *du++ = GFXBLOCK + 4 + (usize>>2);
+  *du++ = (2<<30) | (1<<28) | dw1[1];
+  *du++ = xy;
+  *du++ = (4<<16) | 8;
+  *du++ = *(uint *)&fmv[2];
+  *du++ = *(uint *)&bmv[2];
+  if(usize) {
+    block_ptr = (short *)((unsigned long)block_ptr + 16);
+    PACK_CORR_DATA_SHORT(du,block_ptr);
+    block_ptr = (short *)((unsigned long)block_ptr + 112);
+  }
+  /* End U Blocks */
+
+  /* V Blocks */
+  *dv++ = GFXBLOCK + 4 + (vsize>>2);
+  *dv++ = (3<<30) | (1<<28) | dw1[0];
+  *dv++ = xy;
+  *dv++ = (4<<16) | 8;
+  *dv++ = *(uint *)&fmv[0];
+  *dv++ = *(uint *)&bmv[0];
+  if(vsize) {
+    PACK_CORR_DATA_SHORT(dv,block_ptr);
+  }
+
+  /* Second V Block */
+  *dv++ = GFXBLOCK + 4 + (vsize>>2);
+  *dv++ = (3<<30) | (1<<28) | dw1[1];
+  *dv++ = xy;
+  *dv++ = (4<<16) | 8;
+  *dv++ = *(uint *)&fmv[2];
+  *dv++ = *(uint *)&bmv[2];
+  if(vsize) {
+    block_ptr = (short *)((unsigned long)block_ptr + 16);
+    PACK_CORR_DATA_SHORT(dv,block_ptr);
+    block_ptr = (short *)((unsigned long)block_ptr + 112);
+  }
+  /* End V Blocks */
+
+  *datay = dy;
+  *datau = du;
+  *datav = dv;
+}
+
+/***************************************************************************
+// Function: renderFrameinFrame
+// Description: inline function that sets hardware parameters for a Frame
+//  encoded macroblock in a frame picture.
+***************************************************************************/
+static __inline__ void renderFrameinFrame(uint **datay,uint **datau,
+					  uint **datav,
+					  XvMCMacroBlock *mb,short *block_ptr,
+					  uint flags) {
+
+  register uint *dy = *datay;
+  register uint *du = *datau;
+  register uint *dv = *datav;
+
+  /* Motion Vectors */
+  short fmv[2];
+  short bmv[2];
+  /* gfxblock dword 1 */
+  uint dw1;
+
+  unsigned int ysize = y_frame_bytes[mb->coded_block_pattern];
+  unsigned int usize = u_frame_bytes[mb->coded_block_pattern];
+  unsigned int vsize = v_frame_bytes[mb->coded_block_pattern];
+
+  uint xy = ((uint)mb->x<<20) | ((uint)mb->y<<4);
+
+  dw1 = type_table[mb->macroblock_type & 0xf] |
+    (((uint)mb->coded_block_pattern)<<22);
+
+
+  fmv[0] = mb->PMV[0][0][1];
+  fmv[1] = mb->PMV[0][0][0];
+  
+  bmv[0] = mb->PMV[0][1][1];
+  bmv[1] = mb->PMV[0][1][0];
+
+  /* Y Block */
+  *dy++ = GFXBLOCK + 4 + (ysize>>2);
+  *dy++ = (1<<30) | (3<<28) | dw1;
+  *dy++ = xy;
+  *dy++ = (16<<16) | 16;
+  *dy++ = *(uint *)fmv;
+  *dy++ = *(uint *)bmv;
+  PACK_CORR_DATA(dy,block_ptr,ysize);
+  block_ptr = (short *)((unsigned long)block_ptr + ysize);
+  /* End Y Blocks */
+
+  fmv[0] /= 2;
+  fmv[1] /= 2;
+  
+  bmv[0] /= 2;
+  bmv[1] /= 2;
+
+  xy >>= 1;
+
+  /* U Block */
+  *du++ = GFXBLOCK + 4 + (usize>>2);
+  *du++ = (2<<30) | (1<<28) | dw1;
+  *du++ = xy;
+  *du++ = (8<<16) | 8;
+  *du++ = *(uint *)fmv;
+  *du++ = *(uint *)bmv;
+  PACK_CORR_DATA(du,block_ptr,usize);
+  block_ptr = (short *)((unsigned long)block_ptr + usize);
+  /* End U Block */
+
+  /* V Block */
+  *dv++ = GFXBLOCK + 4 + (vsize>>2);
+  *dv++ = (3<<30) | (1<<28) | dw1;
+  *dv++ = xy;
+  *dv++ = (8<<16) | 8;
+  *dv++ = *(uint *)fmv;
+  *dv++ = *(uint *)bmv;
+  PACK_CORR_DATA(dv,block_ptr,vsize);
+  block_ptr = (short *)((unsigned long)block_ptr + vsize);
+  /* End V Block */
+
+  *datay = dy;
+  *datau = du;
+  *datav = dv;
+}
+
+/***************************************************************************
+// Function: renderFrameinFrameDCT1
+// Description: inline function that sets hardware parameters for a Frame
+//  encoded macroblock in a frame picture with DCT type 1.
+***************************************************************************/
+static __inline__ void renderFrameinFrameDCT1(uint **datay,uint **datau,
+					      uint **datav,XvMCMacroBlock *mb,
+					      short *block_ptr,uint flags) {
+
+  register uint *dy = *datay;
+  register uint *du = *datau;
+  register uint *dv = *datav;
+
+  /* Motion Vectors */
+  short fmv[4];
+  short bmv[4];
+
+  short * top_left_b = NULL;
+  short * top_right_b = NULL;
+  short * bottom_left_b = NULL;
+  short * bottom_right_b = NULL;
+
+  uint temp_bp = 0;
+
+  uint ysize = y_dct1_frame_bytes[mb->coded_block_pattern];
+  uint usize = u_frame_bytes[mb->coded_block_pattern];
+  uint vsize = v_frame_bytes[mb->coded_block_pattern];
+
+  uint xy = ((uint)mb->x<<20) | ((uint)mb->y<<4);
+
+  uint dw1 = type_table[mb->macroblock_type & 0xf] |
+    (((uint)mb->coded_block_pattern)<<22);
+
+  fmv[0] = mb->PMV[0][0][1];
+  fmv[1] = mb->PMV[0][0][0];
+  
+  bmv[0] = mb->PMV[0][1][1];
+  bmv[1] = mb->PMV[0][1][0];
+
+  /*
+    It is easiest to find out what blocks are in need of reading first
+    rather than as we go.
+  */
+  top_left_b = &empty_block[0];
+  if(dw1 & (1<<27)) {
+    temp_bp |= (1<<25);
+    top_left_b = block_ptr;
+    block_ptr += 64;
+  }
+
+  top_right_b = &empty_block[0];
+  if(dw1 & (1<<26)) {
+    temp_bp |= (1<<24);
+    top_right_b = block_ptr;
+    block_ptr += 64;
+  }
+
+  bottom_left_b = &empty_block[0];
+  if(dw1 & (1<<25)) {
+    temp_bp |= (1<<27);
+    bottom_left_b = block_ptr;
+    block_ptr += 64;
+  }
+
+  bottom_right_b = &empty_block[0];
+  if(dw1 & (1<<24)) {
+    temp_bp |= (1<<26);
+    bottom_right_b = block_ptr;
+    block_ptr += 64;
+  }
+  dw1 |= temp_bp;
+
+  /* Y Block */
+  *dy++ = GFXBLOCK + 4 + (ysize>>2);
+  *dy++ = (1<<30) | (3<<28) | dw1;
+  *dy++ = xy;
+  *dy++ = (16<<16) | 16;
+  *dy++ = *(uint *)fmv;
+  *dy++ = *(uint *)bmv;
+  if(dw1 & (1<<27)) {
+    PACK_CORR_DATA_1to0(dy,top_left_b,bottom_left_b);
+    top_left_b = (short *)((unsigned long)top_left_b + 64);
+    bottom_left_b = (short *)((unsigned long)bottom_left_b + 64);
+  }
+  if(dw1 & (1<<26)) {
+    PACK_CORR_DATA_1to0(dy,top_right_b,bottom_right_b);
+    top_right_b = (short *)((unsigned long)top_right_b + 64);
+    bottom_right_b = (short *)((unsigned long)bottom_right_b + 64);
+  }
+  if(dw1 & (1<<27)) {
+    PACK_CORR_DATA_1to0(dy,top_left_b,bottom_left_b);
+  }
+  if(dw1 & (1<<26)) {
+    PACK_CORR_DATA_1to0(dy,top_right_b,bottom_right_b);
+  }
+  /* End Y Block */
+
+  fmv[0] /= 2;
+  fmv[1] /= 2;
+  
+  bmv[0] /= 2;
+  bmv[1] /= 2;
+
+  xy >>= 1;
+
+  /* U Block */
+  *du++ = GFXBLOCK + 4 + (usize>>2);
+  *du++ = (2<<30) | (1<<28) | dw1;
+  *du++ = xy;
+  *du++ = (8<<16) | 8;
+  *du++ = *(uint *)fmv;
+  *du++ = *(uint *)bmv;
+  PACK_CORR_DATA(du,block_ptr,usize);
+  block_ptr = (short *)((unsigned long)block_ptr + usize);
+
+  /* V Block */
+  *dv++ = GFXBLOCK + 4 + (vsize>>2);
+  *dv++ = (3<<30) | (1<<28) | dw1;
+  *dv++ = xy;
+  *dv++ = (8<<16) | 8;
+  *dv++ = *(uint *)fmv;
+  *dv++ = *(uint *)bmv;
+  PACK_CORR_DATA(dv,block_ptr,vsize);
+  block_ptr = (short *)((unsigned long)block_ptr + vsize);
+
+  *datay = dy;
+  *datau = du;
+  *datav = dv;
+}
+
+/***************************************************************************
+// Function: renderDualPrimeinFrame
+// Description: inline function that sets hardware parameters for a Dual
+//  Prime encoded macroblock in a frame picture with dct 1.
+***************************************************************************/
+static __inline__ void renderDualPrimeinFrame(uint **datay,uint **datau,
+					      uint **datav,XvMCMacroBlock *mb,
+					      short *block_ptr,uint flags) {
+
+  register uint *dy = *datay;
+  register uint *du = *datau;
+  register uint *dv = *datav;
+
+  /* Motion Vectors */
+  short fmv[4];
+  short bmv[4];
+  /* gfxblock dword 1 */
+  uint dw1[2];
+
+  uint y1size = y_first_field_bytes[mb->coded_block_pattern];
+  uint y2size = y_second_field_bytes[mb->coded_block_pattern];
+  uint usize = u_field_bytes[mb->coded_block_pattern];
+  uint vsize = v_field_bytes[mb->coded_block_pattern];
+
+  uint xy = ((uint)mb->x<<20) | ((uint)mb->y<<3);
+
+  /*
+    Past Surface (map 0) is used for same parity prediction,
+    Future surface (map 1) is used for opposite.
+  */
+  dw1[0] = (((uint)mb->coded_block_pattern)<<22) |
+    3<<12 | 2<<6 | 2<<3 | 3;
+  dw1[1] = (((mb->coded_block_pattern & 0x3) |
+	     ((mb->coded_block_pattern & 0xc)<<2))<<22) |
+    3<<12 | 3<<6 | 3<<3 | 2;
+  
+  fmv[0] = mb->PMV[0][0][1];
+  fmv[1] = mb->PMV[0][0][0];
+  bmv[0] = mb->PMV[1][0][1];
+  bmv[1] = mb->PMV[1][0][0];
+  
+  fmv[2] = mb->PMV[0][0][1];
+  fmv[3] = mb->PMV[0][0][0];
+  bmv[2] = mb->PMV[1][1][1];
+  bmv[3] = mb->PMV[1][1][0];
+
+  /* First Y Block */
+  *dy++ = GFXBLOCK + 4 + (y1size>>2);
+  *dy++ = (1<<30) | (2<<28) | dw1[0];
+  *dy++ = xy;
+  *dy++ = (8<<16) | 16;
+  *dy++ = *(uint *)fmv;
+  *dy++ = *(uint *)bmv;;
+  PACK_CORR_DATA(dy,block_ptr,y1size);
+  block_ptr = (short *)((unsigned long)block_ptr + y1size);
+
+  /* Second Y Block */
+  *dy++ = GFXBLOCK + 4 + (y2size>>2);
+  *dy++ = (1<<30) | (2<<28) | dw1[1];
+  *dy++ = xy;
+  *dy++ = (8<<16) | 16;
+  *dy++ = *(uint *)&fmv[2];
+  *dy++ = *(uint *)&bmv[2];
+  PACK_CORR_DATA(dy,block_ptr,y2size);
+  block_ptr = (short *)((unsigned long)block_ptr + y2size);
+
+  fmv[0] /= 2;
+  fmv[1] /= 2;
+  bmv[0] /= 2;
+  bmv[1] /= 2;
+  
+  fmv[2] /= 2;
+  fmv[3] /= 2;
+  bmv[2] /= 2;
+  bmv[3] /= 2;
+
+  xy >>= 1;
+
+  /* U Blocks */
+  *du++ = GFXBLOCK + 4 + (usize>>2);
+  *du++ = (2<<30) | (1<<28) | dw1[0];
+  *du++ = xy;
+  *du++ = (4<<16) | 8;
+  *du++ = *(uint *)fmv;
+  *du++ = *(uint *)bmv;
+  if(dw1[0] & (1<<23)) {
+    PACK_CORR_DATA_SHORT(du,block_ptr);
+  }
+
+  /* Second U Block */
+  *du++ = GFXBLOCK + 4 + (usize>>2);
+  *du++ = (2<<30) | (1<<28) | dw1[1];
+  *du++ = xy;
+  *du++ = (4<<16) | 8;
+  *du++ = *(uint *)&fmv[2];
+  *du++ = *(uint *)&bmv[2];
+  if(dw1[1] & (1<<23)) {
+    block_ptr = (short *)((unsigned long)block_ptr + 16);
+    PACK_CORR_DATA_SHORT(du,block_ptr);
+    block_ptr = (short *)((unsigned long)block_ptr + 112);
+  }
+  /* End U Blocks */
+
+  /* V Blocks */
+  *dv++ = GFXBLOCK + 4 + (vsize>>2);
+  *dv++ = (3<<30) | (1<<28) | dw1[0];
+  *dv++ = xy;
+  *dv++ = (4<<16) | 8;
+  *dv++ = *(uint *)fmv;
+  *dv++ = *(uint *)bmv;
+  if(dw1[0] & (1<<22)) {
+    PACK_CORR_DATA_SHORT(dv,block_ptr);
+  }
+
+  /* Second V Block */
+  *dv++ = GFXBLOCK + 4 + (vsize>>2);
+  *dv++ = (3<<30) | (1<<28) | dw1[1];
+  *dv++ = xy;
+  *dv++ = (4<<16) | 8;
+  *dv++ = *(uint *)&fmv[2];
+  *dv++ = *(uint *)&bmv[2];
+  if(dw1[1] & (1<<22)) {
+    block_ptr = (short *)((unsigned long)block_ptr + 16);
+    PACK_CORR_DATA_SHORT(dv,block_ptr);
+    block_ptr = (short *)((unsigned long)block_ptr + 112);
+  }
+  /* End V Blocks */
+
+  *datay = dy;
+  *datau = du;
+  *datav = dv;
+}
+
+/***************************************************************************
+// Function: renderDualPrimeinFrameDCT0
+// Description: inline function that sets hardware parameters for a Dual
+//  Prime encoded macroblock in a frame picture with dct 0.
+***************************************************************************/
+static __inline__ void renderDualPrimeinFrameDCT0(uint **datay,uint **datau,
+						  uint **datav,
+						  XvMCMacroBlock *mb,
+						  short *block_ptr,
+						  uint flags) {
+
+  register uint *dy = *datay;
+  register uint *du = *datau;
+  register uint *dv = *datav;
+
+  /* Motion Vectors */
+  short fmv[4];
+  short bmv[4];
+  /* gfxblock dword 1 */
+  uint dw1[2];
+
+  short * top_left_b = NULL;
+  short * top_right_b = NULL;
+  short * bottom_left_b = NULL;
+  short * bottom_right_b = NULL;
+
+  uint cbp = (uint)mb->coded_block_pattern;
+
+  uint ysize = y_dct0_field_bytes[cbp];
+  uint usize = u_field_bytes[cbp];
+  uint vsize = v_field_bytes[cbp];
+
+  uint xy = ((uint)mb->x<<20) | ((uint)mb->y<<3);
+
+  /*
+    Past Surface (map 0) is used for same parity prediction,
+    Future surface (map 1) is used for opposite.
+  */
+  dw1[0] = ((cbp | ((cbp<<2) & 0x30))<<22) |
+    3<<12 | 2<<6 | 2<<3 | 3;
+  dw1[1] = ((cbp | ((cbp<<2) & 0x30))<<22) |
+    3<<12 | 3<<6 | 3<<3 | 2;
+  
+  fmv[0] = mb->PMV[0][0][1];
+  fmv[1] = mb->PMV[0][0][0];
+  bmv[0] = mb->PMV[1][0][1];
+  bmv[1] = mb->PMV[1][0][0];
+  
+  fmv[2] = mb->PMV[0][0][1];
+  fmv[3] = mb->PMV[0][0][0];
+  bmv[2] = mb->PMV[1][1][1];
+  bmv[3] = mb->PMV[1][1][0];
+  
+  /*
+    The i810 cannot use DCT0 directly with field motion, we have to
+    interlace the data for it. We use a zero block when the CBP has
+    one half of the to-be-interlaced data but not the other half.
+  */
+  top_left_b = &empty_block[0];
+  if(cbp & 0x20) {
+    top_left_b = block_ptr;
+    block_ptr += 64;
+  }
+
+  top_right_b = &empty_block[0];
+  if(cbp & 0x10) {
+    top_right_b = block_ptr;
+    block_ptr += 64;
+  }
+
+  bottom_left_b = &empty_block[0];
+  if(cbp & 0x8) {
+    bottom_left_b = block_ptr;
+    block_ptr += 64;
+  }
+
+  bottom_right_b = &empty_block[0];
+  if(cbp & 0x4) {
+    bottom_right_b = block_ptr;
+    block_ptr += 64;
+  }
+
+  /* First Y Block */
+  *dy++ = GFXBLOCK + 4 + (ysize>>2);
+  *dy++ = (1<<30) | (2<<28) | dw1[0];
+  *dy++ = xy;
+  *dy++ = (8<<16) | 16;
+  *dy++ = *(uint *)fmv;
+  *dy++ = *(uint *)bmv;
+  if(cbp & 0x20) {
+    PACK_CORR_DATA_0to1(dy,top_left_b,bottom_left_b);
+  }
+  if(cbp & 0x10) {
+    PACK_CORR_DATA_0to1(dy,top_right_b,bottom_right_b);
+  }
+
+  /* Second Y Block */
+  *dy++ = GFXBLOCK + 4 + (ysize>>2);
+  *dy++ = (1<<30) | (2<<28) | dw1[1];
+  *dy++ = xy;
+  *dy++ = (8<<16) | 16;
+  *dy++ = *(uint *)&fmv[2];
+  *dy++ = *(uint *)&bmv[2];
+  if(cbp & 0x20) {
+    top_left_b = (short *)((unsigned long)top_left_b + 16);
+    bottom_left_b = (short *)((unsigned long)bottom_left_b + 16);
+    PACK_CORR_DATA_0to1(dy,top_left_b,bottom_left_b);
+  }
+  if(cbp & 0x10) {
+    top_right_b = (short *)((unsigned long)top_right_b + 16);
+    bottom_right_b = (short *)((unsigned long)bottom_right_b + 16);
+    PACK_CORR_DATA_0to1(dy,top_right_b,bottom_right_b);
+  }
+  /* End Y Blocks */
+
+
+  fmv[0] /= 2;
+  fmv[1] /= 2;
+  bmv[0] /= 2;
+  bmv[1] /= 2;
+  
+  fmv[2] /= 2;
+  fmv[3] /= 2;
+  bmv[2] /= 2;
+  bmv[3] /= 2;
+
+  xy >>= 1;
+
+  /* U Blocks */
+  *du++ = GFXBLOCK + 4 + (usize>>2);
+  *du++ = (2<<30) | (1<<28) | dw1[0];
+  *du++ = xy;
+  *du++ = (4<<16) | 8;
+  *du++ = *(uint *)fmv;
+  *du++ = *(uint *)bmv;
+  if(cbp & (1<<23)) {
+    PACK_CORR_DATA_SHORT(du,block_ptr);
+  }
+
+  /* Second U Block */
+  *du++ = GFXBLOCK + 4 + (usize>>2);
+  *du++ = (2<<30) | (1<<28) | dw1[1];
+  *du++ = xy;
+  *du++ = (4<<16) | 8;
+  *du++ = *(uint *)&fmv[2];
+  *du++ = *(uint *)&bmv[2];
+  if(cbp & (1<<23)) {
+    block_ptr = (short *)((unsigned long)block_ptr + 16);
+    PACK_CORR_DATA_SHORT(du,block_ptr);
+    block_ptr = (short *)((unsigned long)block_ptr + 112);
+  }
+  /* End U Blocks */
+
+  /* V Blocks */
+  *dv++ = GFXBLOCK + 4 + (vsize>>2);
+  *dv++ = (3<<30) | (1<<28) | dw1[0];
+  *dv++ = xy;
+  *dv++ = (4<<16) | 8;
+  *dv++ = *(uint *)fmv;
+  *dv++ = *(uint *)bmv;
+  if(cbp & (1<<22)) {
+    PACK_CORR_DATA_SHORT(dv,block_ptr);
+  }
+
+  /* Second V Block */
+  *dv++ = GFXBLOCK + 4 + (vsize>>2);
+  *dv++ = (3<<30) | (1<<28) | dw1[1];
+  *dv++ = xy;
+  *dv++ = (4<<16) | 8;
+  *dv++ = *(uint *)&fmv[2];
+  *dv++ = *(uint *)&bmv[2];
+  if(cbp & (1<<22)) {
+    block_ptr = (short *)((unsigned long)block_ptr + 16);
+    PACK_CORR_DATA_SHORT(dv,block_ptr);
+    block_ptr = (short *)((unsigned long)block_ptr + 112);
+  }
+  /* End V Blocks */
+
+  *datay = dy;
+  *datau = du;
+  *datav = dv;
+}
+
+
+/***************************************************************************
+// Function: XvMCRenderSurface
+// Description: This function does the actual HWMC. Given a list of
+//  macroblock structures it dispatched the hardware commands to execute
+//  them. DMA buffer containing Y data are dispatched as they fill up
+//  U and V DMA buffers are queued until all Y's are done. This minimizes
+//  the context flipping and flushing required when switching between Y
+//  U and V surfaces.
+***************************************************************************/
+#define UV_QUEUE 14
+Status XvMCRenderSurface(Display *display, XvMCContext *context,
+			 unsigned int picture_structure,
+			 XvMCSurface *target_surface,
+			 XvMCSurface *past_surface,
+			 XvMCSurface *future_surface,
+			 unsigned int flags,
+			 unsigned int num_macroblocks,
+			 unsigned int first_macroblock,
+			 XvMCMacroBlockArray *macroblock_array,
+			 XvMCBlockArray *blocks) {
+  /* Dma Data Structures */
+  drmBufPtr pDMAy = NULL,pDMAu[UV_QUEUE],pDMAv[UV_QUEUE];
+  int u_index = 0,v_index = 0;
+  int dirty_context = 1;
+
+  /* Block Pointer */
+  short *block_ptr;
+  /* Current Macroblock Pointer */
+  XvMCMacroBlock *mb;
+
+  drm_i810_mc_t mc;
+  int i,j;
+  i810XvMCSurface *privTarget;
+  i810XvMCSurface *privFuture = NULL;
+  i810XvMCSurface *privPast = NULL;
+  i810XvMCContext *pI810XvMC;
+
+  /* DMA Pointers set to NULL */
+  uint *datay = NULL;
+  uint *datau = NULL;
+  uint *datav = NULL;
+
+
+  /* Check Parameters for validity */
+  if((target_surface == NULL) || (context == NULL) || (display == NULL)) {
+    printf("Error, Invalid Target,Context, or DIsplay!\n");   
+    return BadValue;
+  }
+
+  if(num_macroblocks == 0) {return Success;}
+  if((macroblock_array == NULL) || (blocks == NULL)) {return BadValue;}
+  if(context->privData == NULL) {return BadValue;}
+  pI810XvMC = (i810XvMCContext *)context->privData;
+
+
+  if(target_surface->privData == NULL) {
+    printf("Error, Invalid Target Surface!\n");
+    return BadValue;
+  }
+  privTarget = (i810XvMCSurface *)target_surface->privData;
+
+  if(macroblock_array->num_blocks < (num_macroblocks + first_macroblock)) {
+    printf("Error, Too many macroblocks requested for MB array size.\n");
+    return BadValue;
+  }
+
+  /* Test For YV12 Surface */
+  if(context->surface_type_id != FOURCC_YV12) {
+    printf("Error, HWMC only possible on YV12 Surfaces\n");
+    return BadValue;
+  }
+
+  /* P Frame Test */
+  if(past_surface == NULL) {
+    /* Just to avoid some ifs later. */
+    privPast = privTarget;
+  }
+  else {
+    if(past_surface->privData == NULL) {
+      printf("Error, Invalid Past Surface!\n");
+      return BadValue;
+    }
+    privPast = (i810XvMCSurface *)past_surface->privData;
+  }
+
+
+  /* B Frame Test */
+  if(future_surface == NULL) {
+    privFuture = privTarget;
+    if(pI810XvMC->dual_prime) {
+      privFuture = privPast;
+      /* I810 Specific flag for marking when dual prime is in use. */
+      flags |= 0x40000000;
+    }
+
+    /*
+      References are different for the Second Field Picture. The
+      i810 needs to know if it is the second field picture in a
+      P picture. We use a Private flag to mark this.
+    */
+    if(flags & XVMC_SECOND_FIELD) {
+      /* I810 Specific flag for marking second fields. */
+      flags |= 0x80000000;
+    }
+  }
+  else {
+    if((future_surface->privData == NULL) || (past_surface == NULL)) {
+      printf("Error, Invalid Future Surface or No Past Surface!\n");
+      return BadValue;
+    }
+    privFuture = (i810XvMCSurface *)future_surface->privData;
+
+    /*
+      Undo Second Field flag since the second field in B frames is just like
+      the first.
+    */
+    flags &= ~0x80000000;
+  }
+
+  /* Lock For DMA */
+  I810_LOCK(pI810XvMC,0);
+
+  for(i=first_macroblock; i<(num_macroblocks + first_macroblock); i++) {
+    /* Set up values needed for each macroblock */
+    mb = &macroblock_array->macro_blocks[i];
+    block_ptr = &(blocks->blocks[mb->index<<6]);
+
+    /* Lockup can happen if the coordinates are too far out of range */
+    if(mb->x > target_surface->width>>4) {
+      mb->x = 0;
+    }
+    if(mb->y > target_surface->height>>4) {
+      mb->y = 0;
+    }
+
+    /* If buffers are almost full dispatch them */
+    if(datay) {
+      pDMAy->used = (unsigned long)datay - (unsigned long)pDMAy->address;
+      if(pDMAy->used  > 3520) {
+	if(dirty_context) {
+	  dispatchYContext(privTarget,privPast,privFuture,pI810XvMC);
+	}
+	dirty_context = 0;
+	mc.idx = pDMAy->idx;
+	mc.used = pDMAy->used;
+	datay = NULL;
+	mc.last_render = ++pI810XvMC->last_render;
+	privTarget->last_render = pI810XvMC->last_render;
+	I810_MC(pI810XvMC,mc);
+      } /* datay near full */
+    } /* if(datay) */
+    if(datau) {
+      pDMAu[u_index]->used = (unsigned long)datau - (unsigned long)pDMAu[u_index]->address;
+      if(pDMAu[u_index]->used > 3904) {
+	u_index++;
+	datau = NULL;
+	if(u_index == UV_QUEUE) {
+	  for(j=0; j<UV_QUEUE; j++) {
+	    mc.idx = pDMAu[j]->idx;
+	    mc.used = pDMAu[j]->used;
+	    mc.last_render = ++pI810XvMC->last_render;
+	    privTarget->last_render = pI810XvMC->last_render;
+	    I810_MC(pI810XvMC,mc);
+	  }
+	  u_index = 0;
+	  dirty_context = 1;
+	} /* if(u_index == UV_QUEUE) */
+      } /* datau near full */
+    } /* if(datau) */
+    if(datav) {
+      pDMAv[v_index]->used = (unsigned long)datav - (unsigned long)pDMAv[v_index]->address;
+      if(pDMAv[v_index]->used > 3904) {
+	v_index++;
+	datav = NULL;
+	if(v_index == UV_QUEUE) {
+	  for(j=0; j<UV_QUEUE; j++) {
+	    mc.idx = pDMAv[j]->idx;
+	    mc.used = pDMAv[j]->used;
+	    mc.last_render = ++pI810XvMC->last_render;
+	    privTarget->last_render = pI810XvMC->last_render;
+	    I810_MC(pI810XvMC,mc);
+	  }
+	  v_index = 0;
+	  dirty_context = 1;
+	} /* if(v_index == UV_QUEUE) */
+      } /* datav near full */
+    } /* if(datav) */
+
+    /* Allocate buffers if this is the first loop,or if we just dispatched */
+    if(datay == NULL) {
+      pDMAy = i810_get_free_buffer(pI810XvMC);
+      datay = pDMAy->address;
+    }/* if(datay == NULL) */
+    if(datau == NULL) {
+      pDMAu[u_index] = i810_get_free_buffer(pI810XvMC);
+      datau = pDMAu[u_index]->address;
+      if(u_index == 0) {
+	*datau++ = CMD_FLUSH;
+	*datau++ = BOOLEAN_ENA_2;
+	*datau++ = CMD_FLUSH;
+	*datau++ = DEST_BUFFER_INFO;
+	*datau++ = privTarget->dbi1u;
+	*datau++ = DEST_BUFFER_VAR;
+	*datau++ = privTarget->dbv1;
+	/* Past Surface */
+	*datau++ = CMD_MAP_INFO;
+	*datau++ = privPast->mi1u;
+	*datau++ = privPast->mi2u;
+	*datau++ = privPast->mi3u;
+	/* Future Surface */
+	*datau++ = CMD_MAP_INFO;
+	*datau++ = privFuture->mi1u | 0x1<<28;
+	*datau++ = privFuture->mi2u;
+	*datau++ = privFuture->mi3u;
+      }
+    } /* if(datau == NULL) */
+    if(datav == NULL) {
+      pDMAv[v_index] = i810_get_free_buffer(pI810XvMC);
+      datav = pDMAv[v_index]->address;
+      if(v_index == 0) {
+	*datav++ = CMD_FLUSH;
+	*datav++ = BOOLEAN_ENA_2;
+	*datav++ = CMD_FLUSH;
+	*datav++ = DEST_BUFFER_INFO;
+	*datav++ = privTarget->dbi1v;
+	*datav++ = DEST_BUFFER_VAR;
+	*datav++ = privTarget->dbv1;
+	/* Past Surface */
+	*datav++ = CMD_MAP_INFO;
+	*datav++ = privPast->mi1v;
+	*datav++ = privPast->mi2v;
+	*datav++ = privPast->mi3v;
+	/* Future Surface */
+	*datav++ = CMD_MAP_INFO;
+	*datav++ = privFuture->mi1v | 0x1<<28;
+	*datav++ = privFuture->mi2v;
+	*datav++ = privFuture->mi3v;
+      }
+    }/* if(datav == NULL) */
+
+    /* Catch no pattern case */
+    if(!(mb->macroblock_type & 0x8)) {
+      mb->coded_block_pattern = 0;
+    }
+
+
+    if(mb->motion_type == XVMC_PREDICTION_DUAL_PRIME) {
+      /*
+	By default the maps will not be set up for dual
+	prime. We have to change them.
+      */
+      if(!pI810XvMC->dual_prime) {
+	pI810XvMC->dual_prime = 1;
+	privFuture = privPast;
+	/* Y */
+	*datay++ = CMD_MAP_INFO;
+	*datay++ = privFuture->mi1y | 0x1<<28;
+	*datay++ = privFuture->mi2y;
+	*datay++ = privFuture->mi3y;
+	/* U */
+	*datau++ = CMD_MAP_INFO;
+	*datau++ = privFuture->mi1u | 0x1<<28;
+	*datau++ = privFuture->mi2u;
+	*datau++ = privFuture->mi3u;
+	/* V */
+	*datav++ = CMD_MAP_INFO;
+	*datav++ = privFuture->mi1v | 0x1<<28;
+	*datav++ = privFuture->mi2v;
+	*datav++ = privFuture->mi3v;
+      }
+    }
+    if((pI810XvMC->dual_prime) &&
+       (mb->motion_type != XVMC_PREDICTION_DUAL_PRIME)) {
+      	pI810XvMC->dual_prime = 0;
+	privFuture = privTarget;
+	/* Y */
+	*datay++ = CMD_MAP_INFO;
+	*datay++ = privFuture->mi1y | 0x1<<28;
+	*datay++ = privFuture->mi2y;
+	*datay++ = privFuture->mi3y;
+	/* U */
+	*datau++ = CMD_MAP_INFO;
+	*datau++ = privFuture->mi1u | 0x1<<28;
+	*datau++ = privFuture->mi2u;
+	*datau++ = privFuture->mi3u;
+	/* V */
+	*datav++ = CMD_MAP_INFO;
+	*datav++ = privFuture->mi1v | 0x1<<28;
+	*datav++ = privFuture->mi2v;
+	*datav++ = privFuture->mi3v;
+    }
+
+
+    /* Frame Picture */
+    if((picture_structure & XVMC_FRAME_PICTURE) == XVMC_FRAME_PICTURE) {
+      /* Intra Blocks */
+      if(mb->macroblock_type & XVMC_MB_TYPE_INTRA) {
+	if(mb->dct_type) {
+	  renderIntrainFrameDCT1(&datay,&datau,&datav,mb,block_ptr,flags);
+	  continue;
+	}
+	renderIntrainFrame(&datay,&datau,&datav,mb,block_ptr);
+	continue;
+      }
+      switch((mb->motion_type & 0x3) | (mb->dct_type<<2)) {
+      case 0x2:  /* Frame DCT0 */
+	renderFrameinFrame(&datay,&datau,&datav,mb,block_ptr,flags);
+	continue;
+      case 0x5:  /* Field DCT1 */
+	renderFieldinFrame(&datay,&datau,&datav,mb,block_ptr,flags);
+	continue;
+      case 0x6: /* Frame DCT1 */
+	renderFrameinFrameDCT1(&datay,&datau,&datav,mb,block_ptr,flags);
+	continue;
+      case 0x1: /* Field DCT0 */
+	renderFieldinFrameDCT0(&datay,&datau,&datav,mb,block_ptr,flags);
+	continue;
+      case 0x3:  /* Dual Prime DCT0 */
+	renderDualPrimeinFrame(&datay,&datau,&datav,mb,block_ptr,flags);
+	continue;
+      case 0x7:  /* Dual Prime DCT1 */
+	renderDualPrimeinFrameDCT0(&datay,&datau,&datav,mb,block_ptr,flags);
+	continue;
+      default:  /* No Motion Type */
+	renderError();
+	continue;
+      } /* Switch */
+    } /* Frame Picture */
+
+    /* Field Pictures */
+    if(mb->macroblock_type & XVMC_MB_TYPE_INTRA) {
+      renderIntrainField(&datay,&datau,&datav,mb,block_ptr,picture_structure);
+      continue;
+    }
+    switch(mb->motion_type & 0x3) {
+    case 0x1: /* Field Motion */
+      renderFieldinField(&datay,&datau,&datav,mb,block_ptr,picture_structure,
+			 flags);
+      continue;
+    case 0x2: /* 16x8 Motion */
+      render16x8inField(&datay,&datau,&datav,mb,block_ptr,picture_structure,
+			flags);
+      continue;
+    case 0x3: /* Dual Prime */
+      renderDualPrimeinField(&datay,&datau,&datav,mb,block_ptr,
+			     picture_structure,flags);
+      continue;
+    default:  /* No Motion Type */
+      renderError();
+      continue;
+    }
+    continue;
+
+  } /* for each Macroblock */
+
+  /* Dispatch remaining DMA buffers */
+  if(dirty_context) {
+    dispatchYContext(privTarget,privPast,privFuture,pI810XvMC);
+  }
+  mc.idx = pDMAy->idx;
+  mc.used = (unsigned long)datay - (unsigned long)pDMAy->address;
+  mc.last_render = ++pI810XvMC->last_render;
+  privTarget->last_render = pI810XvMC->last_render;
+  I810_MC(pI810XvMC,mc);
+
+  pDMAu[u_index]->used = (unsigned long)datau - (unsigned long)pDMAu[u_index]->address;
+  for(j=0; j<=u_index; j++) {
+    mc.idx = pDMAu[j]->idx;
+    mc.used = pDMAu[j]->used;
+    mc.last_render = ++pI810XvMC->last_render;
+    privTarget->last_render = pI810XvMC->last_render;
+    I810_MC(pI810XvMC,mc);
+  }
+  pDMAv[v_index]->used = (unsigned long)datav - (unsigned long)pDMAv[v_index]->address;
+  for(j=0; j<=v_index; j++) {
+    mc.idx = pDMAv[j]->idx;
+    mc.used = pDMAv[j]->used;
+    mc.last_render = ++pI810XvMC->last_render;
+    privTarget->last_render = pI810XvMC->last_render;
+    I810_MC(pI810XvMC,mc);
+  }
+
+  I810_UNLOCK(pI810XvMC);
+
+  return Success;
+}
+
+/***************************************************************************
+// Function: XvMCPutSurface
+// Description:
+// Arguments:
+//  display: Connection to X server
+//  surface: Surface to be displayed
+//  draw: X Drawable on which to display the surface
+//  srcx: X coordinate of the top left corner of the region to be
+//          displayed within the surface.
+//  srcy: Y coordinate of the top left corner of the region to be
+//          displayed within the surface.
+//  srcw: Width of the region to be displayed.
+//  srch: Height of the region to be displayed.
+//  destx: X cordinate of the top left corner of the destination region
+//         in the drawable coordinates.
+//  desty: Y cordinate of the top left corner of the destination region
+//         in the drawable coordinates.
+//  destw: Width of the destination region.
+//  desth: Height of the destination region.
+//  flags: One or more of the following.
+//     XVMC_TOP_FIELD - Display only the Top field of the surface.
+//     XVMC_BOTTOM_FIELD - Display only the Bottom Field of the surface.
+//     XVMC_FRAME_PICTURE - Display both fields or frame.
+//
+// Info: Portions of this function derived from i810_video.c (XFree86)
+//
+//   This function is organized so that we wait as long as possible before
+//   touching the overlay registers. Since we don't know that the last
+//   flip has happened yet we want to give the overlay as long as
+//   possible to catch up before we have to check on its progress. This
+//   makes it unlikely that we have to wait on the last flip.
+***************************************************************************/
+Status XvMCPutSurface(Display *display,XvMCSurface *surface,
+		      Drawable draw, short srcx, short srcy,
+		      unsigned short srcw, unsigned short srch,
+		      short destx, short desty,
+		      unsigned short destw, unsigned short desth,
+		      int flags) {
+  i810XvMCContext *pI810XvMC;
+  i810XvMCSurface *pI810Surface;
+  i810OverlayRecPtr pORegs;
+  unsigned int ysrc_offset,uvsrc_offset;
+  Box extents;
+  uint window_width,window_height;
+  unsigned int xscaleInt = 0,xscaleFract = 0,yscaleInt = 0,yscaleFract = 0;
+  unsigned int xscaleFractUV = 0,xscaleIntUV = 0,yscaleFractUV = 0;
+  unsigned int yscaleIntUV = 0,yPitch = 0,uvPitch = 0;
+  unsigned int ovcmd = 0;
+  uint d;
+  double xscale,yscale;
+  int diff;
+  int clipped_srcx, clipped_srcy, clipped_destx, clipped_desty;
+  int clipped_srcw, clipped_srch, clipped_destw, clipped_desth;
+  uint x1,y1,root_width,root_height;
+  int x2 = 0, y2 = 0,unused;
+  uint nChilds;
+  int stat;
+  Window win,root,parent,*pChilds;
+
+
+  if((display == NULL) || (surface == NULL)) {
+    return BadValue;
+  }
+
+  if(surface->privData == NULL) {
+    return (error_base + XvMCBadSurface);
+  }
+  pI810Surface = (i810XvMCSurface *)surface->privData;
+  pI810XvMC = (i810XvMCContext *)pI810Surface->privContext;
+  pORegs = (i810OverlayRecPtr)pI810XvMC->oregs;
+
+
+  switch(surface->surface_type_id) {
+  case FOURCC_YV12:
+  case FOURCC_I420:
+    yPitch = (srcw + 7) & ~7;
+    uvPitch = ((srcw>>1) + 7) & ~7;
+    if((flags & XVMC_FRAME_PICTURE) != XVMC_FRAME_PICTURE) {
+      srch = srch>>1;
+    }
+    break;
+  case FOURCC_UYVY:
+  case FOURCC_YUY2:
+  default:
+    /* FIXME: Non Planar not fully implemented. */
+    return BadValue;
+    yPitch = ((srcw + 7) & ~7) << 1;
+    break;
+  }/* switch(surface->surface_type_id) */
+
+  /*
+    FIXME: This should be using the DRI's clip rect but that isn't
+    all hooked up yet. This has some latency but we get by.
+  */
+  win = draw;
+  XQueryTree(display,win,&root,&parent,&pChilds,&nChilds);
+  if(nChilds) XFree(pChilds);
+  XGetGeometry(display,win, &root, &x2, &y2, &window_width,
+	       &window_height, &d, &d);
+  x1 = x2;
+  y1 = y2;
+  win = parent;
+  do {
+    XQueryTree(display,win,&root,&parent,&pChilds,&nChilds);
+    if(nChilds) XFree(pChilds);
+    XGetGeometry(display,win, &root, &x2, &y2, &d, &d, &d, &d);
+    x1 += x2;
+    y1 += y2;
+    win = parent;
+  }while(win != root);
+  XGetGeometry(display,root, &root, &unused, &unused,
+               &root_width, &root_height, &d, &d);
+
+  /* Left edge of Video window clipped to screen */
+  extents.x1 = 0;
+  if(x1 > extents.x1) {
+    extents.x1 = x1;
+  }
+  /* Right edge of Video window clipped to screen */
+  extents.x2 = root_width;
+  if(extents.x2 > (x1 + window_width)) {
+    extents.x2 = x1 + window_width;
+  }
+  /* Top edge of Video window clipped to screen */
+  extents.y1 = 0;
+  if(y1 > extents.y1) {
+    extents.y1 = y1;
+  }
+  /* Bottom edge of Video window clipped to screen */
+  extents.y2 = root_height;
+  if(extents.y2 > (y1 + window_height)) {
+    extents.y2 = y1 + window_height;
+  }
+
+  /*
+    Clipping is more difficult than is seems. We need to keep the
+    scaling factors even if the destination window needs to be clipped.
+    We clip the destination window first then apply a scaled version
+    to the source window.
+  */
+
+  /* Put destination coords in screen coords */
+  destx += x1;
+  desty += y1;
+
+  /* Scale factors requested */
+  xscale = (double)srcw / (double)destw;
+  yscale = (double)srch / (double)desth;
+
+  /*
+    If destination window needs to be clipped we actually adjust both
+    the src and dest window so as to keep the scaling that was requested
+  */
+  clipped_srcx = srcx;
+  clipped_srcy = srcy;
+  clipped_destx = destx;
+  clipped_desty = desty;
+  clipped_srcw = srcw;
+  clipped_srch = srch;
+  clipped_destw = destw;
+  clipped_desth = desth;
+
+  /* Clip to the source surface boundaries */
+  if(clipped_srcx < 0) {
+    clipped_destx += (0 - clipped_srcx) / xscale;
+    clipped_srcw -= clipped_srcx;
+    clipped_destw -= clipped_srcx / xscale;
+    clipped_srcx = 0;
+  }
+  if((clipped_srcw + clipped_srcx) > surface->width) {
+    clipped_srcw = surface->width - clipped_srcx;
+    clipped_destw -= (clipped_srcw - srcw) / xscale;
+  }
+  if(clipped_srcy < 0) {
+    clipped_desty += (0 - clipped_srcy) / yscale;
+    clipped_srch -= clipped_srcy;
+    clipped_desth -= clipped_srcy / yscale;
+    clipped_srcy = 0;
+  }
+  if((clipped_srch + clipped_srcy) > surface->height) {
+    clipped_srch = surface->height - clipped_srcy;
+    clipped_desth -= (clipped_srch - srch) / yscale;
+  }
+
+  /* Clip to the extents */
+  if(clipped_destx < extents.x1) {
+    diff = extents.x1 - clipped_destx;
+    clipped_srcx += diff * xscale;
+    clipped_srcw -= diff * xscale;
+    clipped_destw -= diff;
+    clipped_destx = extents.x1;
+  }
+
+  diff = (clipped_destx + clipped_destw) - extents.x2;
+  if(diff > 0) {
+    clipped_destw -= diff;
+    clipped_srcw -= diff * xscale;
+  }
+
+  if(clipped_desty < extents.y1) {
+    diff = extents.y1 - clipped_desty;
+    clipped_srcy += diff * yscale;
+    clipped_srch -= diff * yscale;
+    clipped_desth -= diff;
+    clipped_desty = 0;
+  }
+
+  diff = (clipped_desty + clipped_desth) - extents.y2;
+  if(diff > 0) {
+    clipped_desth -= diff;
+    clipped_srch -= diff * yscale;
+  }
+
+  /* If the whole window is clipped turn off the overlay */
+  if((clipped_destx + clipped_destw < extents.x1) ||
+     (clipped_desty + clipped_desth < extents.y1) ||
+     (clipped_destx > extents.x2) ||
+     (clipped_desty > extents.y2)) {
+    return XvMCHideSurface(display, surface);
+  }
+
+  /*
+    Adjust the source offset width and height according to the clipped
+    destination window.
+  */
+  ysrc_offset = ((clipped_srcx + 1) & ~1) +
+    ((clipped_srcy + 1) & ~1) * (1<<pI810Surface->pitch);
+  uvsrc_offset = (clipped_srcx>>1) +
+    (clipped_srcy>>1) * (1<<(pI810Surface->pitch - 1));
+
+  /*
+    Initially, YCbCr and Overlay Enable and
+    vertical chrominance up interpolation and horozontal chrominance
+    up interpolation
+  */
+  ovcmd = VC_UP_INTERPOLATION | HC_UP_INTERPOLATION |
+    Y_ADJUST | OVERLAY_ENABLE;
+
+  if ((clipped_destw != clipped_srcw) ||
+      (clipped_desth != clipped_srch)) {
+    xscaleInt = (clipped_srcw / clipped_destw) & 0x3;
+    xscaleFract = (clipped_srcw << 12) / clipped_destw;
+    yscaleInt = (clipped_srch / clipped_desth) & 0x3;
+    yscaleFract = (clipped_srch << 12) / clipped_desth;
+   
+    if (clipped_destw > clipped_srcw) {
+      /* horizontal up-scaling */
+      ovcmd &= ~HORIZONTAL_CHROMINANCE_FILTER;
+      ovcmd &= ~HORIZONTAL_LUMINANCE_FILTER;
+      ovcmd |= (HC_UP_INTERPOLATION | HL_UP_INTERPOLATION);
+    }
+
+    if (clipped_desth > clipped_srch) {
+      /* vertical up-scaling */
+      ovcmd &= ~VERTICAL_CHROMINANCE_FILTER;
+      ovcmd &= ~VERTICAL_LUMINANCE_FILTER;
+      ovcmd |= (VC_UP_INTERPOLATION | VL_UP_INTERPOLATION);
+    }
+
+    if (clipped_destw < clipped_srcw) {
+      /* horizontal down-scaling */
+      ovcmd &= ~HORIZONTAL_CHROMINANCE_FILTER;
+      ovcmd &= ~HORIZONTAL_LUMINANCE_FILTER;
+      ovcmd |= (HC_DOWN_INTERPOLATION | HL_DOWN_INTERPOLATION);
+    }
+
+    if (clipped_desth < clipped_srch) {
+      /* vertical down-scaling */
+      ovcmd &= ~VERTICAL_CHROMINANCE_FILTER;
+      ovcmd &= ~VERTICAL_LUMINANCE_FILTER;
+      ovcmd |= (VC_DOWN_INTERPOLATION | VL_DOWN_INTERPOLATION);
+    }
+
+    /* now calculate the UV scaling factor */
+    if (xscaleFract) {
+      xscaleFractUV = xscaleFract >> MINUV_SCALE;
+      ovcmd &= ~HC_DOWN_INTERPOLATION;
+      ovcmd |= HC_UP_INTERPOLATION;
+    }
+
+    if (xscaleInt) {
+      xscaleIntUV = xscaleInt >> MINUV_SCALE;
+      if (xscaleIntUV) {
+	ovcmd &= ~HC_UP_INTERPOLATION;
+      }
+    }
+
+    if (yscaleFract) {
+      yscaleFractUV = yscaleFract >> MINUV_SCALE;
+      ovcmd &= ~VC_DOWN_INTERPOLATION;
+      ovcmd |= VC_UP_INTERPOLATION;
+    }
+
+    if (yscaleInt) {
+      yscaleIntUV = yscaleInt >> MINUV_SCALE;
+      if (yscaleIntUV) {
+	ovcmd &= ~VC_UP_INTERPOLATION;
+	ovcmd |= VC_DOWN_INTERPOLATION;
+      }
+    }
+
+  }/* if((destw != srcw) || (desth != srch)) */
+
+  /* Lock the DRM */
+  I810_LOCK(pI810XvMC,0);
+
+  /* Block until rendering on this surface is finished */
+  stat = XVMC_RENDERING;
+  while(stat & XVMC_RENDERING) {
+    XvMCGetSurfaceStatus(display,surface,&stat);
+  }
+  /* Block until the last flip is finished */
+  if(pI810XvMC->last_flip) {
+    BLOCK_OVERLAY(pI810XvMC,pI810XvMC->current);
+  }
+
+  pI810XvMC->current = !pI810XvMC->current;
+  pORegs->OV0CMD = ovcmd;
+
+  if ((clipped_destw != clipped_srcw) ||
+      (clipped_desth != clipped_srch)) {
+    pORegs->YRGBSCALE = (xscaleInt << 15) |
+      ((xscaleFract & 0xFFF) << 3) | (yscaleInt) |
+      ((yscaleFract & 0xFFF) << 20);
+
+    pORegs->UVSCALE = yscaleIntUV | ((xscaleFractUV & 0xFFF) << 3) |
+      ((yscaleFractUV & 0xFFF) << 20);
+  }
+  else {
+    /* Normal 1:1 scaling */
+    pORegs->YRGBSCALE = 0x80004000;
+    pORegs->UVSCALE = 0x80004000;
+  }
+
+  pORegs->SHEIGHT = clipped_srch | (clipped_srch << 15);
+  pORegs->DWINPOS = (clipped_desty << 16) | clipped_destx;
+  pORegs->DWINSZ = ((clipped_desth<< 16) | (clipped_destw));
+
+  /* Attributes */
+  pORegs->OV0CLRC0 = ((pI810XvMC->contrast & 0x1ff)<<8) |
+    (pI810XvMC->brightness & 0xff);
+  pORegs->OV0CLRC1 = (pI810XvMC->saturation & 0x3ff);
+  
+  /* Destination Colorkey Setup */
+  pI810XvMC->oregs->DCLRKV = RGB16ToColorKey(pI810XvMC->colorkey);
+
+  /* buffer locations, add the offset from the clipping */
+  if(pI810XvMC->current) {
+    pORegs->OBUF_1Y = (unsigned long)pI810Surface->offset +
+      (unsigned long)pI810Surface->offsets[0] + ysrc_offset;
+    pORegs->OBUF_1V = (unsigned long)pI810Surface->offset +
+      (unsigned long)pI810Surface->offsets[2] + uvsrc_offset;
+    pORegs->OBUF_1U = (unsigned long)pI810Surface->offset +
+      (unsigned long)pI810Surface->offsets[1] + uvsrc_offset;
+  }
+  else {
+    pORegs->OBUF_0Y = (unsigned long)pI810Surface->offset +
+      (unsigned long)pI810Surface->offsets[0] + ysrc_offset;
+    pORegs->OBUF_0V = (unsigned long)pI810Surface->offset +
+      (unsigned long)pI810Surface->offsets[2] + uvsrc_offset;
+    pORegs->OBUF_0U = (unsigned long)pI810Surface->offset +
+      (unsigned long)pI810Surface->offsets[1] + uvsrc_offset;
+  }
+
+  switch(surface->surface_type_id) {
+  case FOURCC_YV12:
+  case FOURCC_I420:
+    pORegs->SWID = (uvPitch << 16) | yPitch;
+    pORegs->SWIDQW = (uvPitch << 13) | (yPitch >> 3);
+    pORegs->OV0STRIDE = (1<<pI810Surface->pitch) | 
+      ((1<<pI810Surface->pitch) << 15);
+    pORegs->OV0CMD &= ~SOURCE_FORMAT;
+    pORegs->OV0CMD |= YUV_420;
+    if((flags & XVMC_FRAME_PICTURE) != XVMC_FRAME_PICTURE) {
+      /* Top Field Only */
+      if(flags & XVMC_TOP_FIELD) {
+	if(pI810XvMC->current == 1) {
+	  pORegs->OV0CMD |= (VERTICAL_PHASE_BOTH | FLIP_TYPE_FIELD |
+			     BUFFER1_FIELD0);
+	}
+	else {
+	  pORegs->OV0CMD |= (VERTICAL_PHASE_BOTH | FLIP_TYPE_FIELD |
+			     BUFFER0_FIELD0);
+	}
+	pORegs->YRGB_VPH = 1<<15 | 1<<31;
+	pORegs->UV_VPH = 3<<14 | 3<<30;
+	pORegs->INIT_PH = 0x06 | 0x18;
+      }
+      /* Bottom Field Only */
+      else {
+	if(pI810XvMC->current == 1) {
+	  pORegs->OV0CMD |= (VERTICAL_PHASE_BOTH | FLIP_TYPE_FIELD |
+			     BUFFER1_FIELD1);
+	}
+	else {
+	  pORegs->OV0CMD |= (VERTICAL_PHASE_BOTH | FLIP_TYPE_FIELD |
+			     BUFFER0_FIELD1);
+	}
+	pORegs->YRGB_VPH = 0;
+	pORegs->UV_VPH = 7<<29 | 7<<13;
+	pORegs->INIT_PH = 0x06;
+      }
+    }
+    /* Frame Picture */
+    else {
+      if(pI810XvMC->current == 1) {
+	pORegs->OV0CMD |= BUFFER1_FIELD0;
+      }
+      else {
+	pORegs->OV0CMD |= BUFFER0_FIELD0;
+      }
+      pORegs->YRGB_VPH = 0;
+      pORegs->UV_VPH = 0;
+      pORegs->INIT_PH = 0;
+    }
+    break;
+  case FOURCC_UYVY:
+  case FOURCC_YUY2:
+  default:
+    pORegs->SWID = srcw;
+    pORegs->SWIDQW = srcw >> 3;
+    pORegs->OV0STRIDE = pI810Surface->pitch;
+    pORegs->OV0CMD &= ~SOURCE_FORMAT;
+    pORegs->OV0CMD |= YUV_422;
+    pORegs->OV0CMD &= ~OV_BYTE_ORDER;
+    if (surface->surface_type_id == FOURCC_UYVY) {
+      pORegs->OV0CMD |= Y_SWAP;
+    }
+
+    pORegs->OV0CMD &= ~BUFFER_AND_FIELD;
+    if(pI810XvMC->current == 1) {
+      pORegs->OV0CMD |= BUFFER1_FIELD0;
+    }
+    else {
+      pORegs->OV0CMD |= BUFFER0_FIELD0;
+    }
+
+    break;
+  } /* switch(surface->surface_type_id) */
+  
+
+
+  OVERLAY_FLIP(pI810XvMC);
+
+  /*
+    The Overlay only flips when it knows you changed
+    something. So the first time change stuff while it
+    is watching to be sure.
+  */
+  if(!pI810XvMC->last_flip) {
+    pORegs->OV0CMD &= ~0x4;
+    if(pI810XvMC->current == 1) {
+      pORegs->OV0CMD |= BUFFER1_FIELD0;
+    }
+    else {
+      pORegs->OV0CMD |= BUFFER0_FIELD0;
+    }
+  }
+  pI810Surface->last_flip = ++pI810XvMC->last_flip;
+  I810_UNLOCK(pI810XvMC);
+
+  return Success;
+}
+
+/***************************************************************************
+// Function: XvMCSyncSurface
+// Arguments:
+//   display - Connection to the X server
+//   surface - The surface to synchronize
+// Info:
+// Returns: Status
+***************************************************************************/
+Status XvMCSyncSurface(Display *display,XvMCSurface *surface) {
+  Status ret;
+  int stat=0;
+  /*
+    FIXME: Perhaps a timer here to prevent lockup?
+    FIXME: Perhaps a usleep to not be busy waiting?
+  */
+  do {
+    ret = XvMCGetSurfaceStatus(display,surface,&stat);
+  }while(!ret && (stat & XVMC_RENDERING));
+  return ret;
+}
+
+/***************************************************************************
+// Function: XvMCFlushSurface
+// Description:
+//   This function commits pending rendering requests to ensure that they
+//   wll be completed in a finite amount of time.
+// Arguments:
+//   display - Connection to X server
+//   surface - Surface to flush
+// Info:
+//   This command is a noop for i810 becuase we always dispatch buffers in
+//   render. There is little gain to be had with 4k buffers.
+// Returns: Status
+***************************************************************************/
+Status XvMCFlushSurface(Display * display, XvMCSurface *surface) {
+  return Success;
+}
+
+/***************************************************************************
+// Function: XvMCGetSurfaceStatus
+// Description:
+// Arguments:
+//  display: connection to X server
+//  surface: The surface to query
+//  stat: One of the Following
+//    XVMC_RENDERING - The last XvMCRenderSurface command has not
+//                     completed.
+//    XVMC_DISPLAYING - The surface is currently being displayed or a
+//                     display is pending.
+***************************************************************************/
+Status XvMCGetSurfaceStatus(Display *display, XvMCSurface *surface,
+			    int *stat) {
+  i810XvMCSurface *privSurface;
+  i810XvMCContext *pI810XvMC;
+  int temp;
+
+  if((display == NULL) || (surface == NULL) || (stat == NULL)) {
+    return BadValue;
+  }
+  if(surface->privData == NULL) {
+    return BadValue;
+  }
+  *stat = 0;
+  privSurface = surface->privData;
+
+  pI810XvMC = privSurface->privContext;
+  if(pI810XvMC == NULL) {
+    return (error_base + XvMCBadSurface);
+  }
+
+  I810_LOCK(pI810XvMC,0);
+  if(privSurface->last_flip) {
+    /* This can not happen */
+    if(pI810XvMC->last_flip < privSurface->last_flip) {
+      printf("Error: Context last flip is less than surface last flip.\n");
+      return BadValue;
+    }
+    /*
+      If the context has 2 or more flips after this surface it
+      cannot be displaying. Don't bother to check.
+    */
+    if(!(pI810XvMC->last_flip > (privSurface->last_flip + 1))) {
+      /*
+	If this surface was the last flipped it is either displaying
+	or about to be so don't bother checking.
+      */
+      if(pI810XvMC->last_flip == privSurface->last_flip) {
+	*stat |= XVMC_DISPLAYING;
+      }
+      else {
+	/*
+	  In this case there has been one more flip since our surface's
+	  but we need to check if it is finished or not.
+	*/
+	temp = GET_FSTATUS(pI810XvMC);
+	if(((temp & (1<<20))>>20) != pI810XvMC->current) { 
+	  *stat |= XVMC_DISPLAYING;
+	}
+      }
+    }
+  }
+
+  if(privSurface->last_render &&
+      (privSurface->last_render > GET_RSTATUS(pI810XvMC))) {
+    *stat |= XVMC_RENDERING;
+  }
+  I810_UNLOCK(pI810XvMC);
+
+  return Success;
+}
+
+/***************************************************************************
+// 
+//  Surface manipulation functions
+//
+***************************************************************************/
+
+/***************************************************************************
+// Function: XvMCHideSurface
+// Description: Stops the display of a surface.
+// Arguments:
+//   display - Connection to the X server.
+//   surface - surface to be hidden.
+//
+// Returns: Status
+***************************************************************************/
+Status XvMCHideSurface(Display *display, XvMCSurface *surface) {
+  i810XvMCSurface *pI810Surface;
+  i810XvMCContext *pI810XvMC;
+  int ss, xx;
+
+  /* Did we get a good display and surface passed into us? */
+  if(display == NULL) {
+    return BadValue;
+  }
+
+  if(surface == NULL) {
+    return (error_base + XvMCBadSurface);
+  }
+
+  XvMCSyncSurface(display, surface);
+
+  /* Get surface private data pointer */
+  if(surface->privData == NULL) {
+    return (error_base + XvMCBadSurface);
+  }
+  pI810Surface = (i810XvMCSurface *)surface->privData;
+
+  /*
+    Get the status of the surface, if it is not currently displayed
+    we don't need to worry about it.
+  */
+  if((xx = XvMCGetSurfaceStatus(display, surface, &ss)) != Success) {
+    return xx;
+  }
+  if(! (ss & XVMC_DISPLAYING)) {
+    return Success;
+  }
+ 
+  /* Get the associated context pointer */
+  pI810XvMC = (i810XvMCContext *)pI810Surface->privContext;
+  if(pI810XvMC == NULL) {
+    return (error_base + XvMCBadSurface);
+  }
+
+  if(pI810XvMC->last_flip) {
+    I810_LOCK(pI810XvMC,DRM_LOCK_QUIESCENT);
+    
+    /* Make sure last flip is done */
+    BLOCK_OVERLAY(pI810XvMC,pI810XvMC->current);
+
+    /* Set the registers to turn the overlay off */
+    pI810XvMC->oregs->OV0CMD = VC_UP_INTERPOLATION | HC_UP_INTERPOLATION |
+      Y_ADJUST;
+    pI810XvMC->current = !pI810XvMC->current;
+    if(pI810XvMC->current == 1) {
+      pI810XvMC->oregs->OV0CMD |= BUFFER1_FIELD0;
+    }
+    else {
+      pI810XvMC->oregs->OV0CMD |= BUFFER0_FIELD0;
+    }
+    OVERLAY_FLIP(pI810XvMC);
+    /*
+      Increment the context flip but not the surface. This way no
+      surface has the last flip #.
+    */
+    pI810XvMC->last_flip++;
+
+
+    /* Now wait until the hardware reads the registers and makes the change. */
+    BLOCK_OVERLAY(pI810XvMC,pI810XvMC->current)
+
+      I810_UNLOCK(pI810XvMC);
+  }
+
+  return Success;
+}
+
+
+
+
+/***************************************************************************
+//
+// Functions that deal with subpictures
+//
+***************************************************************************/
+
+
+
+/***************************************************************************
+// Function: XvMCCreateSubpicture
+// Description: This creates a subpicture by filling out the XvMCSubpicture
+//              structure passed to it and returning Success.
+// Arguments:
+//   display - Connection to the X server.
+//   context - The context to create the subpicture for.
+//   subpicture - Pre-allocated XvMCSubpicture structure to be filled in.
+//   width - of subpicture
+//   height - of subpicture
+//   xvimage_id - The id describing the XvImage format.
+//
+// Returns: Status
+***************************************************************************/
+Status XvMCCreateSubpicture(Display *display, XvMCContext *context,
+                          XvMCSubpicture *subpicture,
+                          unsigned short width, unsigned short height,
+                          int xvimage_id) {
+  i810XvMCContext *pI810XvMC;
+  i810XvMCSubpicture *pI810Subpicture;
+  int priv_count;
+  uint *priv_data;
+  Status ret;
+
+  if((subpicture == NULL) || (context == NULL) || (display == NULL)){
+    return BadValue;
+  }
+  
+  pI810XvMC = (i810XvMCContext *)context->privData;
+  if(pI810XvMC == NULL) {
+    return (error_base + XvMCBadContext);
+  }
+
+
+  subpicture->context_id = context->context_id;
+  subpicture->xvimage_id = xvimage_id;
+  
+  /* These need to be checked to make sure they are not too big! */
+  subpicture->width = width;
+  subpicture->height = height;
+
+  subpicture->privData =
+    (i810XvMCSubpicture *)malloc(sizeof(i810XvMCSubpicture));
+
+  if(!subpicture->privData) {
+    return BadAlloc;
+  }
+  pI810Subpicture = (i810XvMCSubpicture *)subpicture->privData;
+
+
+  if((ret = _xvmc_create_subpicture(display, context, subpicture,
+				    &priv_count, &priv_data))) {
+    printf("Unable to create XvMCSubpicture.\n");
+    return ret;
+  }
+
+  if(priv_count != 1) {
+    printf("_xvmc_create_subpicture() returned incorrect data size.\n");
+    printf("Expected 1 got %d\n",priv_count);
+    free(priv_data);
+    return BadAlloc;
+  }
+  /* Data == Client Address, offset == Physical address offset */
+  pI810Subpicture->data = pI810XvMC->surfaces.address;
+  pI810Subpicture->offset = pI810XvMC->surfaces.offset;
+
+  /* Initialize private values */
+  pI810Subpicture->privContext = pI810XvMC;
+
+  pI810Subpicture->last_render = 0;
+  pI810Subpicture->last_flip = 0;
+
+  /* Based on the xvimage_id we will need to set the other values */
+  subpicture->num_palette_entries = 16;
+  subpicture->entry_bytes = 3;
+  strcpy(subpicture->component_order,"YUV");
+
+  /*
+    i810's MC Engine needs surfaces of 2^x (x= 9,10,11,12) pitch
+    and the Tiler need 512k aligned surfaces, basically we are
+    stuck with fixed memory with pitch 1024.
+  */
+  pI810Subpicture->pitch = 10;
+
+  /* 
+     offsets[0] == offset into the map described by either
+     address (Client memeory address) or offset (physical offset from fb base)
+  */
+  pI810Subpicture->offsets[0] = priv_data[0];
+  if(((unsigned long)pI810Subpicture->data + pI810Subpicture->offsets[0]) & 4095) {
+    printf("XvMCCreateSubpicture: Subpicture offset 0 is not 4096 aligned\n");
+  }
+
+  /* Free data returned from xvmc_create_surface */
+  free(priv_data);
+
+  /* Clear the surface to 0 */
+  memset((void *)((unsigned long)pI810Subpicture->data + (unsigned long)pI810Subpicture->offsets[0]),
+	 0, ((1<<pI810Subpicture->pitch) * subpicture->height));
+
+  switch(subpicture->xvimage_id) {
+  case FOURCC_IA44:
+  case FOURCC_AI44:
+    /* Destination buffer info command */
+    pI810Subpicture->dbi1 = ((((unsigned int)pI810Subpicture->offset +
+			       pI810Subpicture->offsets[0]) & ~0xfc000fff) |
+			     (pI810Subpicture->pitch - 9));
+    
+    /* Destination buffer variables command */
+    pI810Subpicture->dbv1 = (0x8<<20) | (0x8<<16);
+
+    /* Map info command */
+    pI810Subpicture->mi1 = (0x0<<24) | (3<<21) | (1<<9) |
+      (pI810Subpicture->pitch - 3);
+
+    pI810Subpicture->mi2 = (((unsigned int)subpicture->height - 1)<<16) |
+      ((unsigned int)subpicture->width - 1);
+
+    pI810Subpicture->mi3 = ((unsigned int)pI810Subpicture->offset +
+			 pI810Subpicture->offsets[0]) & ~0xfc00000f;
+    break;
+  default:
+    free(subpicture->privData);
+    return BadMatch;
+  }
+
+  pI810XvMC->ref++;
+  return Success;
+}
+
+
+
+/***************************************************************************
+// Function: XvMCClearSubpicture
+// Description: Clear the area of the given subpicture to "color".
+//              structure passed to it and returning Success.
+// Arguments:
+//   display - Connection to the X server.
+//   subpicture - Subpicture to clear.
+//   x, y, width, height - rectangle in the subpicture to clear.
+//   color - The data to file the rectangle with.
+//
+// Returns: Status
+***************************************************************************/
+Status XvMCClearSubpicture(Display *display, XvMCSubpicture *subpicture,
+                          short x, short y,
+                          unsigned short width, unsigned short height,
+                          unsigned int color) {
+
+  i810XvMCContext *pI810XvMC;
+  i810XvMCSubpicture *pI810Subpicture;
+  int i;
+
+  if((subpicture == NULL) || (display == NULL)){
+    return BadValue;
+  }
+
+  if(!subpicture->privData) {
+    return (error_base + XvMCBadSubpicture);
+  }
+  pI810Subpicture = (i810XvMCSubpicture *)subpicture->privData;
+
+  pI810XvMC = (i810XvMCContext *)pI810Subpicture->privContext;
+  if(pI810XvMC == NULL) {
+    return (error_base + XvMCBadSubpicture);
+  }
+
+  if((x < 0) || (x + width > subpicture->width)) {
+    return BadValue;
+  }
+
+  if((y < 0) || (y + height > subpicture->height)) {
+    return BadValue;
+  }
+
+  for(i=y; i<y + height; i++) {
+    memset((void *)((unsigned long)pI810Subpicture->data +
+		    (unsigned long)pI810Subpicture->offsets[0] + x +
+		    (1<<pI810Subpicture->pitch) * i),(char)color,width);
+  }
+
+  return Success;
+}
+
+/***************************************************************************
+// Function: XvMCCompositeSubpicture
+// Description: Composite the XvImae on the subpicture. This composit uses
+//              non-premultiplied alpha. Destination alpha is utilized
+//              except for with indexed subpictures. Indexed subpictures
+//              use a simple "replace".
+// Arguments:
+//   display - Connection to the X server.
+//   subpicture - Subpicture to clear.
+//   image - the XvImage to be used as the source of the composite.
+//   srcx, srcy, width, height - The rectangle from the image to be used.
+//   dstx, dsty - location in the subpicture to composite the source.
+//
+// Returns: Status
+***************************************************************************/
+Status XvMCCompositeSubpicture(Display *display, XvMCSubpicture *subpicture,
+                               XvImage *image,
+                               short srcx, short srcy,
+                               unsigned short width, unsigned short height,
+                               short dstx, short dsty) {
+
+  i810XvMCContext *pI810XvMC;
+  i810XvMCSubpicture *pI810Subpicture;
+  int i;
+
+  if((subpicture == NULL) || (display == NULL)){
+    return BadValue;
+  }
+
+  if(!subpicture->privData) {
+    return (error_base + XvMCBadSubpicture);
+  }
+  pI810Subpicture = (i810XvMCSubpicture *)subpicture->privData;
+
+  pI810XvMC = (i810XvMCContext *)pI810Subpicture->privContext;
+  if(pI810XvMC == NULL) {
+    return (error_base + XvMCBadSubpicture);
+  }
+
+  if((srcx < 0) || (srcx + width > image->width)) {
+    return BadValue;
+  }
+
+  if((dstx < 0) || (dstx + width > subpicture->width)) {
+    return BadValue;
+  }
+
+  if((srcy < 0) || (srcy + height > image->height)) {
+    return BadValue;
+  }
+
+  if((dsty < 0) || (dsty + height > subpicture->height)) {
+    return BadValue;
+  }
+
+  for(i=0; i<height; i++) {
+    memcpy((void *)((unsigned long)pI810Subpicture->data +
+		    (unsigned long)pI810Subpicture->offsets[0] + dstx +
+		    (1<<pI810Subpicture->pitch) * (i + dsty)),
+	   (void *)((unsigned long)image->data +
+		    (unsigned long)image->offsets[0] + srcx +
+		    image->pitches[0] * (i + srcy))
+	   ,width);
+  }
+
+  return Success;
+
+}
+
+
+/***************************************************************************
+// Function: XvMCDestroySubpicture
+// Description: Destroys the specified subpicture.
+// Arguments:
+//   display - Connection to the X server.
+//   subpicture - Subpicture to be destroyed.
+//
+// Returns: Status
+***************************************************************************/
+Status XvMCDestroySubpicture(Display *display, XvMCSubpicture *subpicture) {
+
+  i810XvMCSubpicture *pI810Subpicture;
+  i810XvMCContext *pI810XvMC;
+
+  if((display == NULL) || (subpicture == NULL)) {
+    return BadValue;
+  }
+  if(!subpicture->privData) {
+    return (error_base + XvMCBadSubpicture);
+  }
+  pI810Subpicture = (i810XvMCSubpicture *)subpicture->privData;
+
+  pI810XvMC = (i810XvMCContext *)pI810Subpicture->privContext;
+  if(!pI810XvMC) {
+    return (error_base + XvMCBadSubpicture);
+  }
+
+
+  if(pI810Subpicture->last_render) {
+    XvMCSyncSubpicture(display,subpicture);
+  }
+
+  _xvmc_destroy_subpicture(display,subpicture);
+
+  i810_free_privContext(pI810XvMC);
+
+  free(pI810Subpicture);
+  subpicture->privData = NULL;
+  return Success;
+}
+
+
+/***************************************************************************
+// Function: XvMCSetSubpicturePalette
+// Description: Set the subpictures palette
+// Arguments:
+//   display - Connection to the X server.
+//   subpicture - Subpiture to set palette for.
+//   palette - A pointer to an array holding the palette data. The array
+//     is num_palette_entries * entry_bytes in size.
+// Returns: Status
+***************************************************************************/
+
+Status XvMCSetSubpicturePalette(Display *display, XvMCSubpicture *subpicture,
+				unsigned char *palette) {
+ i810XvMCSubpicture *privSubpicture;
+ int i,j;
+
+ if((display == NULL) || (subpicture == NULL)) {
+   return BadValue;
+ }
+ if(subpicture->privData == NULL) {
+   return (error_base + XvMCBadSubpicture);
+ }
+ privSubpicture = (i810XvMCSubpicture *)subpicture->privData;
+
+ j=0;
+ for(i=0; i<16; i++) {
+   privSubpicture->palette[0][i] = palette[j++];
+   privSubpicture->palette[1][i] = palette[j++];
+   privSubpicture->palette[2][i] = palette[j++];
+ }
+ return Success;
+}
+
+/***************************************************************************
+// Function: XvMCBlendSubpicture
+// Description: 
+//    The behavior of this function is different depending on whether
+//    or not the XVMC_BACKEND_SUBPICTURE flag is set in the XvMCSurfaceInfo.
+//    i810 only support frontend behavior.
+//  
+//    XVMC_BACKEND_SUBPICTURE not set ("frontend" behavior):
+//   
+//    XvMCBlendSubpicture is a no-op in this case.
+//   
+// Arguments:
+//   display - Connection to the X server.
+//   subpicture - The subpicture to be blended into the video.
+//   target_surface - The surface to be displayed with the blended subpic.
+//   source_surface - Source surface prior to blending.
+//   subx, suby, subw, subh - The rectangle from the subpicture to use.
+//   surfx, surfy, surfw, surfh - The rectangle in the surface to blend
+//      blend the subpicture rectangle into. Scaling can ocure if 
+//      XVMC_SUBPICTURE_INDEPENDENT_SCALING is set.
+//
+// Returns: Status
+***************************************************************************/
+Status XvMCBlendSubpicture(Display *display, XvMCSurface *target_surface,
+                         XvMCSubpicture *subpicture,
+                         short subx, short suby,
+                         unsigned short subw, unsigned short subh,
+                         short surfx, short surfy,
+                         unsigned short surfw, unsigned short surfh) {
+
+  return BadMatch;
+}
+
+
+
+/***************************************************************************
+// Function: XvMCBlendSubpicture2
+// Description: 
+//    The behavior of this function is different depending on whether
+//    or not the XVMC_BACKEND_SUBPICTURE flag is set in the XvMCSurfaceInfo.
+//    i810 only supports frontend blending.
+//  
+//    XVMC_BACKEND_SUBPICTURE not set ("frontend" behavior):
+//   
+//    XvMCBlendSubpicture2 blends the source_surface and subpicture and
+//    puts it in the target_surface.  This does not effect the status of
+//    the source surface but will cause the target_surface to query
+//    XVMC_RENDERING until the blend is completed.
+//   
+// Arguments:
+//   display - Connection to the X server.
+//   subpicture - The subpicture to be blended into the video.
+//   target_surface - The surface to be displayed with the blended subpic.
+//   source_surface - Source surface prior to blending.
+//   subx, suby, subw, subh - The rectangle from the subpicture to use.
+//   surfx, surfy, surfw, surfh - The rectangle in the surface to blend
+//      blend the subpicture rectangle into. Scaling can ocure if 
+//      XVMC_SUBPICTURE_INDEPENDENT_SCALING is set.
+//
+// Returns: Status
+***************************************************************************/
+Status XvMCBlendSubpicture2(Display *display, 
+                          XvMCSurface *source_surface,
+                          XvMCSurface *target_surface,
+                          XvMCSubpicture *subpicture,
+                          short subx, short suby,
+                          unsigned short subw, unsigned short subh,
+                          short surfx, short surfy,
+                          unsigned short surfw, unsigned short surfh) {
+  drmBufPtr pDMA;
+  unsigned int *data;
+  i810XvMCContext *pI810XvMC;
+  i810XvMCSubpicture *privSubpicture;
+  i810XvMCSurface *privTarget;
+  i810XvMCSurface *privSource;
+  drm_i810_mc_t mc;
+  int i,j;
+
+  if(display == NULL) {
+    return BadValue;
+  }
+
+  if(subpicture == NULL) {
+    return (error_base + XvMCBadSubpicture);
+  }
+
+  if((target_surface == NULL) || (source_surface == NULL)) {
+    return (error_base + XvMCBadSurface);
+  }
+
+  if((subpicture->xvimage_id != FOURCC_AI44) &&
+     (subpicture->xvimage_id != FOURCC_IA44)) {
+      return (error_base + XvMCBadSubpicture);
+  }
+
+  if(!subpicture->privData) {
+    return (error_base + XvMCBadSubpicture);
+  }
+  privSubpicture = (i810XvMCSubpicture *)subpicture->privData;
+
+  pI810XvMC = (i810XvMCContext *)privSubpicture->privContext;
+  if(pI810XvMC == NULL) {
+    return (error_base + XvMCBadSubpicture);
+  }
+
+  if(!target_surface->privData) {
+    return (error_base + XvMCBadSurface);
+  }
+  privTarget = (i810XvMCSurface *)target_surface->privData;
+  
+  if(!source_surface->privData) {
+    return (error_base + XvMCBadSurface);
+  }
+  privSource = (i810XvMCSurface *)source_surface->privData;
+ 
+
+  /* Check that size isn't bigger than subpicture */
+  if((subx + subw) > subpicture->width) {
+    return BadValue;
+  }
+  if((suby + subh) > subpicture->height) {
+    return BadValue;
+  }
+  /* Check that dest isn't bigger than surface */
+  if((surfx + surfw) > target_surface->width) {
+    return BadValue;
+  }
+  if((surfy + surfh) > target_surface->height) {
+    return BadValue;
+  }
+  /* Make sure surfaces match */
+  if(target_surface->width != source_surface->width) {
+    return BadValue;
+  }
+  if(target_surface->height != source_surface->height) {
+    return BadValue;
+  }
+
+  /* Lock For DMA */
+  I810_LOCK(pI810XvMC,0);
+
+  /* Allocate DMA buffer */
+  pDMA = i810_get_free_buffer(pI810XvMC);
+  data = pDMA->address;
+
+  /* Copy Y data first */
+  /* SOURCE_COPY_BLT */
+  *data++ = (2<<29) | (0x43<<22) | 0x4;
+  *data++ = (0xcc<<16) | (1<<26) | (1<<privTarget->pitch);
+  *data++ = (target_surface->height<<16) | target_surface->width;
+  *data++ = privTarget->offset + privTarget->offsets[0];
+  *data++ = (1<<privSource->pitch);
+  *data++ = privSource->offset + privSource->offsets[0];
+
+  /* Select Context 1 for loading */
+  *data++ = CMD_FLUSH;
+  *data++ = (5<<23) | (1<<17) | (1<<8);
+  *data++ = CMD_FLUSH;
+
+  /* Load Palette */
+  *data++ = MAP_PALETTE_LOAD;
+  /* 16 levels of alpha for each Y */
+  switch(subpicture->xvimage_id) {
+  case FOURCC_IA44:
+    for(i=0; i<16; i++) {
+      for(j=0; j<16; j++) {
+        *data++ = (j<<12) | (j<<8) | privSubpicture->palette[0][i];
+      }
+    }
+    break;
+  case FOURCC_AI44:
+    for(i=0; i<16; i++) {
+      for(j=0; j<16; j++) {
+        *data++ = (i<<12) | (i<<8) | privSubpicture->palette[0][j];
+      }
+    }
+    break;
+  }
+  /* TARGET */
+  /* *data++ = CMD_FLUSH; */
+  /* *data++ = BOOLEAN_ENA_2; */
+  *data++ = CMD_FLUSH;
+  *data++ = DEST_BUFFER_INFO;
+  *data++ = privTarget->dbi1y;
+  *data++ = DEST_BUFFER_VAR;
+  *data++ = privTarget->dbv1;
+
+  /* ALPHA */
+  *data++ = CMD_MAP_INFO;
+  *data++ = privSubpicture->mi1;
+  *data++ = privSubpicture->mi2;
+  *data++ = privSubpicture->mi3;
+
+  *data++ = VERTEX_FORMAT | (1<<8) | (3<<1);
+  *data++ = BOOLEAN_ENA_1;
+  *data++ = SRC_DEST_BLEND_MONO | (0x940);
+  /* Map Filter */
+  *data++ = (3<<29) | (0x1c<<24) | (2<<19) | (0x224);
+
+  /* Use context 1 */
+  *data++ = CMD_FLUSH;
+  *data++ = (5<<23) | (1<<16) | 1;
+  *data++ = CMD_FLUSH;
+
+  /* Drawing Rect Info */
+  *data++ = DRAWING_RECT_INFO;
+  *data++ = 0x0;
+  *data++ = 0x0;
+  *data++ = 0x0;
+  *data++ = 0x0;
+  *data++ = 0x0;
+
+  /* GFXPRIMITIVE RECTANGLE */
+  *data++ = (3<<29) | (0x1f<<24) | (0x7<<18) | 11;
+  /* Bottom Right Vertex */
+  *(float *)data++ = (float) (surfx + surfw);
+  *(float *)data++ = (float) (surfy + surfh);
+  *(float *)data++ = (float) (subx + subw);
+  *(float *)data++ = (float) (suby + subh);
+  /* Bottom Left Vertex */
+  *(float *)data++ = (float) surfx;
+  *(float *)data++ = (float) (surfy + surfh);
+  *(float *)data++ = (float) subx;
+  *(float *)data++ = (float) (suby + subh);
+  /* Top Left Vertex */
+  *(float *)data++ = (float) surfx;
+  *(float *)data++ = (float) surfy;
+  *(float *)data++ = (float) subx;
+  *(float *)data++ = (float) suby;
+
+  /* Load and Use Context 0 */
+  *data++ = CMD_FLUSH;
+  *data++ = (5<<23) | (1<<17) | (1<<16);
+  *data++ = CMD_FLUSH;
+
+  /* U data */
+  /* SOURCE_COPY_BLT */
+  *data++ = (2<<29) | (0x43<<22) | 0x4;
+  *data++ = (0xcc<<16) | (1<<26) | (1<<(privTarget->pitch - 1));
+  *data++ = (target_surface->height<<15) | (target_surface->width>>1);
+  *data++ = (unsigned long)privTarget->offset + (unsigned long)privTarget->offsets[1];
+  *data++ = (1<<(privSource->pitch - 1));
+  *data++ = (unsigned long)privSource->offset + (unsigned long)privSource->offsets[1];
+
+  /* Context 1 select */
+  *data++ = CMD_FLUSH;
+  *data++ = (5<<23) | (1<<17) | (1<<8);
+  *data++ = CMD_FLUSH;
+  /* ALPHA PALETTE */
+  *data++ = MAP_PALETTE_LOAD;
+  /* 16 levels of alpha for each Y */
+  switch(subpicture->xvimage_id) {
+  case FOURCC_IA44:
+    for(i=0; i<16; i++) {
+      for(j=0; j<16; j++) {
+        *data++ = (j<<12) | (j<<8) | privSubpicture->palette[2][i];
+      }
+    }
+    break;
+  case FOURCC_AI44:
+    for(i=0; i<16; i++) {
+      for(j=0; j<16; j++) {
+        *data++ = (i<<12) | (i<<8) | privSubpicture->palette[2][j];
+      }
+    }
+    break;
+  }
+  /* TARGET */
+  *data++ = CMD_FLUSH;
+  *data++ = BOOLEAN_ENA_2;
+  *data++ = CMD_FLUSH;
+  *data++ = DEST_BUFFER_INFO;
+  *data++ = privTarget->dbi1u;
+  *data++ = DEST_BUFFER_VAR;
+  *data++ = privTarget->dbv1;
+
+  /* ALPHA */
+  *data++ = CMD_MAP_INFO;
+  *data++ = privSubpicture->mi1;
+  *data++ = privSubpicture->mi2;
+  *data++ = privSubpicture->mi3;
+
+  *data++ = VERTEX_FORMAT | (1<<8) | (3<<1);
+  *data++ = BOOLEAN_ENA_1;
+  *data++ = SRC_DEST_BLEND_MONO | (0x940);
+  /* Map Filter */
+  *data++ = (3<<29) | (0x1c<<24) | (2<<19) | (1<<16) | (0x224);
+
+  /* Use context 1 */
+  *data++ = CMD_FLUSH;
+  *data++ = (5<<23) | (1<<16) | 1;
+  *data++ = CMD_FLUSH;
+
+  /* Drawing Rect Info */
+  *data++ = (3<<29) | (0x1d<<24) | (0x80<<16) | 3;
+  *data++ = 0;
+  *data++ = 0;
+  *data++ = 0;
+  *data++ = 0;
+  *data++ = 0;
+
+  /* Rectangle */
+  *data++ = (3<<29) | (0x1f<<24) | (0x7<<18) | 11;
+  /* Bottom Right */
+  *(float *)data++ = (float) ((surfx + surfw)>>1);
+  *(float *)data++ = (float) ((surfy + surfh)>>1);
+  *(float *)data++ = (float) subx + subw;
+  *(float *)data++ = (float) suby + subh;
+  /* Bottom Left */
+  *(float *)data++ = (float) (surfx>>1);
+  *(float *)data++ = (float) ((surfy + surfh)>>1);
+  *(float *)data++ = (float) subx;
+  *(float *)data++ = (float) suby + subh;
+  /* Top Left */
+  *(float *)data++ = (float) (surfx>>1);
+  *(float *)data++ = (float) (surfy>>1);
+  *(float *)data++ = (float) subx;
+  *(float *)data++ = (float) suby;
+
+  /* Load and Use Context 0 */
+  *data++ = CMD_FLUSH;
+  *data++ = (5<<23) | (1<<17) | (1<<16);
+  *data++ = CMD_FLUSH;
+
+  /* V data */
+  /* SOURCE_COPY_BLT */
+  *data++ = (2<<29) | (0x43<<22) | 0x4;
+  *data++ = (0xcc<<16) | (1<<26) | (1<<(privTarget->pitch - 1));
+  *data++ = (target_surface->height<<15) | (target_surface->width>>1);
+  *data++ = (unsigned long)privTarget->offset + (unsigned long)privTarget->offsets[2];
+  *data++ = (1<<(privSource->pitch - 1));
+  *data++ = (unsigned long)privSource->offset + (unsigned long)privSource->offsets[2];
+
+  /* Context 1 select */
+  *data++ = CMD_FLUSH;
+  *data++ = (5<<23) | (1<<17) | (1<<8);
+  *data++ = CMD_FLUSH;
+
+  /* ALPHA PALETTE */
+  *data++ = MAP_PALETTE_LOAD;
+  /* 16 levels of alpha for each Y */
+  switch(subpicture->xvimage_id) {
+  case FOURCC_IA44:
+    for(i=0; i<16; i++) {
+      for(j=0; j<16; j++) {
+        *data++ = (j<<12) | (j<<8) | privSubpicture->palette[1][i];
+      }
+    }
+    break;
+  case FOURCC_AI44:
+    for(i=0; i<16; i++) {
+      for(j=0; j<16; j++) {
+        *data++ = (i<<12) | (i<<8) | privSubpicture->palette[1][j];
+      }
+    }
+    break;
+  }
+  /* TARGET */
+  *data++ = CMD_FLUSH;
+  *data++ = BOOLEAN_ENA_2;
+  *data++ = CMD_FLUSH;
+  *data++ = DEST_BUFFER_INFO;
+  *data++ = privTarget->dbi1v;
+  *data++ = DEST_BUFFER_VAR;
+  *data++ = privTarget->dbv1;
+
+  /* ALPHA */
+  *data++ = CMD_MAP_INFO;
+  *data++ = privSubpicture->mi1;
+  *data++ = privSubpicture->mi2;
+  *data++ = privSubpicture->mi3;
+
+  *data++ = VERTEX_FORMAT | (1<<8) | (3<<1);
+  *data++ = BOOLEAN_ENA_1;
+  *data++ = SRC_DEST_BLEND_MONO | (0x940);
+  /* Map Filter */
+  *data++ = (3<<29) | (0x1c<<24) | (2<<19) | (1<<16) | (0x224);
+
+  /* Use context 1 */
+  *data++ = CMD_FLUSH;
+  *data++ = (5<<23) | (1<<16) | 1;
+  *data++ = CMD_FLUSH;
+
+  /* Drawing Rect Info */
+  *data++ = (3<<29) | (0x1d<<24) | (0x80<<16) | 3;
+  *data++ = 0;
+  *data++ = 0;
+  *data++ = 0;
+  *data++ = 0;
+  *data++ = 0;
+
+  /* Rectangle */
+  *data++ = (3<<29) | (0x1f<<24) | (0x7<<18) | 11;
+  /* Bottom Right */
+  *(float *)data++ = (float) ((surfx + surfw)>>1);
+  *(float *)data++ = (float) ((surfy + surfh)>>1);
+  *(float *)data++ = (float) subx + subw;
+  *(float *)data++ = (float) suby + subh;
+  /* Bottom Left */
+  *(float *)data++ = (float) (surfx>>1);
+  *(float *)data++ = (float) ((surfy + surfh)>>1);
+  *(float *)data++ = (float) subx;
+  *(float *)data++ = (float) suby + subh;
+  /* Top Left */
+  *(float *)data++ = (float) (surfx>>1);
+  *(float *)data++ = (float) (surfy>>1);
+  *(float *)data++ = (float) subx;
+  *(float *)data++ = (float) suby;
+
+  /* Load and Use Context 0 */
+  *data++ = CMD_FLUSH;
+  *data++ = (5<<23) | (1<<17) | (1<<16);
+  *data++ = CMD_FLUSH;
+
+
+  /* Dispatch */
+  pDMA->used = (unsigned long)data - (unsigned long)pDMA->address;
+  mc.idx = pDMA->idx;
+  mc.used = pDMA->used;
+  mc.last_render = ++pI810XvMC->last_render;
+  privTarget->last_render = pI810XvMC->last_render;
+  I810_MC(pI810XvMC,mc);
+
+  I810_UNLOCK(pI810XvMC);
+  return Success;
+}
+
+
+
+/***************************************************************************
+// Function: XvMCSyncSubpicture
+// Description: This function blocks until all composite/clear requests on
+//              the subpicture have been complete.
+// Arguments:
+//   display - Connection to the X server.
+//   subpicture - The subpicture to synchronize
+//
+// Returns: Status
+***************************************************************************/
+Status XvMCSyncSubpicture(Display *display, XvMCSubpicture *subpicture) {
+  Status ret;
+  int stat=0;
+  do {
+    ret = XvMCGetSubpictureStatus(display,subpicture,&stat);
+  }while(!ret && (stat & XVMC_RENDERING));
+  return ret;
+}
+
+
+
+/***************************************************************************
+// Function: XvMCFlushSubpicture
+// Description: This function commits pending composite/clear requests to
+//              ensure that they will be completed in a finite amount of
+//              time.
+// Arguments:
+//   display - Connection to the X server.
+//   subpicture - The subpicture whos compsiting should be flushed
+//
+// Returns: Status
+// NOTES: i810 always dispatches commands so flush is a no-op
+***************************************************************************/
+Status XvMCFlushSubpicture(Display *display, XvMCSubpicture *subpicture) {
+  if(display == NULL) {
+    return BadValue;
+  }
+  if(subpicture == NULL) {
+    return (error_base + XvMCBadSubpicture);
+  }
+
+  return Success;
+}
+
+
+
+/***************************************************************************
+// Function: XvMCGetSubpictureStatus
+// Description: This function gets the current status of a subpicture
+//
+// Arguments:
+//   display - Connection to the X server.
+//   subpicture - The subpicture whos status is being queried
+//   stat - The status of the subpicture. It can be any of the following
+//          OR'd together:
+//          XVMC_RENDERING  - Last composite or clear request not completed
+//          XVMC_DISPLAYING - Suppicture currently being displayed.
+//
+// Returns: Status
+// Notes: i810 always blends into a third surface so the subpicture is
+//  never actually displaying, only a copy of it is displaying. We only
+//  have to worry about the rendering case.
+***************************************************************************/
+Status XvMCGetSubpictureStatus(Display *display, XvMCSubpicture *subpicture,
+                             int *stat) {
+
+  i810XvMCSubpicture *privSubpicture;
+  i810XvMCContext *pI810XvMC;
+
+  if((display == NULL) || (stat == NULL)) {
+    return BadValue;
+  }
+  if((subpicture == NULL) || (subpicture->privData == NULL)) {
+    return (error_base + XvMCBadSubpicture);
+  }
+  *stat = 0;
+  privSubpicture = (i810XvMCSubpicture *)subpicture->privData;
+
+  pI810XvMC = (i810XvMCContext *)privSubpicture->privContext;
+  if(pI810XvMC == NULL) {
+    return (error_base + XvMCBadSubpicture);
+  }
+
+  I810_LOCK(pI810XvMC,0);
+
+  if(privSubpicture->last_render &&
+      (privSubpicture->last_render > GET_RSTATUS(pI810XvMC))) {
+    *stat |= XVMC_RENDERING;
+  }
+  I810_UNLOCK(pI810XvMC);
+
+  return Success;
+}
+
+#define NUM_XVMC_ATTRIBUTES 4
+static XvAttribute I810_XVMC_ATTRIBUTES[] = {
+  {XvGettable | XvSettable, 0, 0xffffff, "XV_COLORKEY"},
+  {XvGettable | XvSettable, -127, +127, "XV_BRIGHTNESS"},
+  {XvGettable | XvSettable, 0, 0x1ff, "XV_CONTRAST"},
+  {XvGettable | XvSettable, 0, 0x3ff, "XV_SATURATION"}
+};
+
+
+/***************************************************************************
+// Function: XvMCQueryAttributes
+// Description: An array of XvAttributes of size "number" is returned by
+//   this function. If there are no attributes, NULL is returned and number
+//   is set to 0. The array may be freed with xfree().
+//
+// Arguments:
+//   display - Connection to the X server.
+//   context - The context whos attributes we are querying.
+//   number - The number of returned atoms.
+//
+// Returns:
+//  An array of XvAttributes.
+// Notes:
+//   For i810 we support these Attributes:
+//    XV_COLORKEY: The colorkey value, initialized from the Xv value at
+//                 context creation time.
+//    XV_BRIGHTNESS
+//    XV_CONTRAST
+//    XV_SATURATION
+***************************************************************************/
+XvAttribute *XvMCQueryAttributes(Display *display, XvMCContext *context,
+				 int *number) {
+  i810XvMCContext *pI810XvMC;
+  XvAttribute *attributes;
+
+  if(number == NULL) {
+    return NULL;
+  }
+  if(display == NULL) {
+    *number = 0;
+    return NULL;
+  }
+  if(context == NULL) {
+    *number = 0;
+    return NULL;
+  }
+  pI810XvMC = context->privData;
+  if(pI810XvMC == NULL) {
+    *number = 0;
+    return NULL;
+  }
+
+  attributes = (XvAttribute *)malloc(NUM_XVMC_ATTRIBUTES *
+				     sizeof(XvAttribute));
+  if(attributes == NULL) {
+    *number = 0;
+    return NULL;
+  }
+
+  memcpy(attributes,I810_XVMC_ATTRIBUTES,(NUM_XVMC_ATTRIBUTES *
+					  sizeof(XvAttribute)));
+
+  *number = NUM_XVMC_ATTRIBUTES;
+  return attributes;
+}
+
+/***************************************************************************
+// Function: XvMCSetAttribute
+// Description: This function sets a context-specific attribute.
+//
+// Arguments:
+//   display - Connection to the X server.
+//   context - The context whos attributes we are querying.
+//   attribute - The X atom of the attribute to be changed.
+//   value - The new value for the attribute.
+//
+// Returns:
+//  Status
+// Notes:
+//   For i810 we support these Attributes:
+//    XV_COLORKEY: The colorkey value, initialized from the Xv value at
+//                 context creation time.
+//    XV_BRIGHTNESS
+//    XV_CONTRAST
+//    XV_SATURATION
+***************************************************************************/
+Status XvMCSetAttribute(Display *display, XvMCContext *context,
+			Atom attribute, int value) {
+  i810XvMCContext *pI810XvMC;
+
+  if(display == NULL) {
+    return BadValue;
+  }
+  if(context == NULL) {
+    return (error_base + XvMCBadContext);
+  }
+  pI810XvMC = context->privData;
+  if(pI810XvMC == NULL) {
+    return (error_base + XvMCBadContext);
+  }
+
+  if(attribute == pI810XvMC->xv_colorkey) {
+    if((value < I810_XVMC_ATTRIBUTES[0].min_value) ||
+       (value > I810_XVMC_ATTRIBUTES[0].max_value)) {
+      return BadValue;
+    }
+    pI810XvMC->colorkey = value;
+    return Success;
+  }
+  if(attribute == pI810XvMC->xv_brightness) {
+    if((value < I810_XVMC_ATTRIBUTES[1].min_value) ||
+       (value > I810_XVMC_ATTRIBUTES[1].max_value)) {
+      return BadValue;
+    }
+    pI810XvMC->brightness = value;
+    return Success;
+  }
+  if(attribute == pI810XvMC->xv_saturation) {
+    if((value < I810_XVMC_ATTRIBUTES[2].min_value) ||
+       (value > I810_XVMC_ATTRIBUTES[2].max_value)) {
+      return BadValue;
+    }
+    pI810XvMC->saturation = value;
+    return Success;
+  }
+  if(attribute == pI810XvMC->xv_contrast) {
+    if((value < I810_XVMC_ATTRIBUTES[3].min_value) ||
+       (value > I810_XVMC_ATTRIBUTES[3].max_value)) {
+      return BadValue;
+    }
+    pI810XvMC->contrast = value;
+    return Success;
+  }
+  return BadValue;
+}
+
+/***************************************************************************
+// Function: XvMCGetAttribute
+// Description: This function queries a context-specific attribute and
+//   returns the value.
+//
+// Arguments:
+//   display - Connection to the X server.
+//   context - The context whos attributes we are querying.
+//   attribute - The X atom of the attribute to be queried
+//   value - The returned attribute value
+//
+// Returns:
+//  Status
+// Notes:
+//   For i810 we support these Attributes:
+//    XV_COLORKEY: The colorkey value, initialized from the Xv value at
+//                 context creation time.
+//    XV_BRIGHTNESS
+//    XV_CONTRAST
+//    XV_SATURATION
+***************************************************************************/
+Status XvMCGetAttribute(Display *display, XvMCContext *context,
+			Atom attribute, int *value) {
+  i810XvMCContext *pI810XvMC;
+
+  if(display == NULL) {
+    return BadValue;
+  }
+  if(context == NULL) {
+    return (error_base + XvMCBadContext);
+  }
+  pI810XvMC = context->privData;
+  if(pI810XvMC == NULL) {
+    return (error_base + XvMCBadContext);
+  }
+  if(value == NULL) {
+    return BadValue;
+  }
+
+  if(attribute == pI810XvMC->xv_colorkey) {
+    *value = pI810XvMC->colorkey;
+    return Success;
+  }
+  if(attribute == pI810XvMC->xv_brightness) {
+    *value = pI810XvMC->brightness;
+    return Success;
+  }
+  if(attribute == pI810XvMC->xv_saturation) {
+    *value = pI810XvMC->saturation;
+    return Success;
+  }
+  if(attribute == pI810XvMC->xv_contrast) {
+    *value = pI810XvMC->contrast;
+    return Success;
+  }
+  return BadValue;
+}
+
+
+
+
diff --git a/nx-X11/lib/XvMC/hw/i810/I810XvMC.h b/nx-X11/lib/XvMC/hw/i810/I810XvMC.h
new file mode 100644
index 000000000..ba8c792d0
--- /dev/null
+++ b/nx-X11/lib/XvMC/hw/i810/I810XvMC.h
@@ -0,0 +1,469 @@
+/***************************************************************************
+
+Copyright 2001 Intel Corporation.  All Rights Reserved.
+
+Permission is hereby granted, free of charge, to any person obtaining a
+copy of this software and associated documentation files (the
+"Software"), to deal in the Software without restriction, including
+without limitation the rights to use, copy, modify, merge, publish,
+distribute, sub license, and/or sell copies of the Software, and to
+permit persons to whom the Software is furnished to do so, subject to
+the following conditions:
+
+The above copyright notice and this permission notice (including the
+next paragraph) shall be included in all copies or substantial portions
+of the Software.
+
+THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+IN NO EVENT SHALL INTEL, AND/OR ITS SUPPLIERS BE LIABLE FOR ANY CLAIM,
+DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR
+OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR
+THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+
+**************************************************************************/
+
+/***************************************************************************
+ * libI810XvMC.h: MC Driver i810 includes
+ *
+ * Authors:
+ *      Matt Sottek <matthew.j.sottek@intel.com>
+ *
+ *
+ ***************************************************************************/
+/* $XFree86: xc/lib/XvMC/hw/i810/I810XvMC.h,v 1.3 2002/10/30 12:52:02 alanh Exp $ */
+
+
+#ifndef I810XVMC_H
+#define I810XVMC_H
+
+/* #define XVMC_DEBUG(x) do {x; }while(0); */
+#define XVMC_DEBUG(x)
+
+#include "xf86drm.h"
+#include "i810_common.h"
+#include <X11/Xlibint.h>
+
+
+/***************************************************************************
+// i810OverlayRec: Structure that is used to reference the overlay
+//  register memory. A i810OverlayRecPtr is set to the address of the
+//  allocated overlay registers.
+***************************************************************************/
+typedef struct _i810OverlayRec {
+  unsigned int OBUF_0Y;
+  unsigned int OBUF_1Y;
+  unsigned int OBUF_0U;
+  unsigned int OBUF_0V;
+  unsigned int OBUF_1U;
+  unsigned int OBUF_1V;
+  unsigned int OV0STRIDE;
+  unsigned int YRGB_VPH;
+  unsigned int UV_VPH;
+  unsigned int HORZ_PH;
+  unsigned int INIT_PH;
+  unsigned int DWINPOS;
+  unsigned int DWINSZ;
+  unsigned int SWID;
+  unsigned int SWIDQW;
+  unsigned int SHEIGHT;
+  unsigned int YRGBSCALE;
+  unsigned int UVSCALE;
+  unsigned int OV0CLRC0;
+  unsigned int OV0CLRC1;
+  unsigned int DCLRKV;
+  unsigned int DCLRKM;
+  unsigned int SCLRKVH;
+  unsigned int SCLRKVL;
+  unsigned int SCLRKM;
+  unsigned int OV0CONF;
+  unsigned int OV0CMD;
+} i810OverlayRec, *i810OverlayRecPtr;
+
+
+/***************************************************************************
+// i810XvMCDrmMap: Holds the data about the DRM maps
+***************************************************************************/
+typedef struct _i810XvMCDrmMap {
+  drm_handle_t offset;
+  drmAddress address;
+  unsigned int size;
+} i810XvMCDrmMap, *i810XvMCDrmMapPtr;
+
+/***************************************************************************
+// i810XvMCContext: Private Context data referenced via the privData
+//  pointer in the XvMCContext structure.
+***************************************************************************/
+typedef struct _i810XvMCContext {
+  int fd;   /* File descriptor for /dev/dri */
+  i810XvMCDrmMap overlay;
+  i810XvMCDrmMap surfaces;
+  drmBufMapPtr dmabufs;  /* Data structure to hold available dma buffers */
+  drm_context_t drmcontext;
+  unsigned int last_render;
+  unsigned int last_flip;
+  unsigned short ref;
+  unsigned short current;
+  int lock;   /* Lightweight lock to avoid locking twice */
+  char busIdString[10]; /* PCI:0:1:0 or PCI:0:2:0 */
+  i810OverlayRecPtr oregs;
+  unsigned int dual_prime; /* Flag to identify when dual prime is in use. */
+  unsigned int fb_base;
+  Atom xv_colorkey;
+  Atom xv_brightness;
+  Atom xv_contrast;
+  Atom xv_saturation;
+  int brightness;
+  int saturation;
+  int contrast;
+  int colorkey;
+} i810XvMCContext;
+
+/***************************************************************************
+// i810XvMCSurface: Private data structure for each XvMCSurface. This
+//  structure is referenced by the privData pointer in the XvMCSurface
+//  structure.
+***************************************************************************/
+typedef struct _i810XvMCSurface {
+  unsigned int pitch;
+  unsigned int dbi1y;  /* Destination buffer info command word 1 for Y */
+  unsigned int dbi1u; /* Destination buffer info command word 1 for U */
+  unsigned int dbi1v; /* Destination buffer info command word 1 for V */
+  unsigned int dbv1;  /* Destination buffer variables command word 1 */
+  unsigned int mi1y;   /* Map Info command word 1 (Minus bit 28) for Y */
+  unsigned int mi1u;   /* Map Info command word 1 (Minus bit 28) for U */
+  unsigned int mi1v;   /* Map Info command word 1 (Minus bit 28) for V */
+  unsigned int mi2y;   /* Map info command word 2 for y */
+  unsigned int mi2u;   /* Map info command word 2 for y */
+  unsigned int mi2v;   /* Map info command word 2 for y */
+  unsigned int mi3y;   /* Map info command word 3 */
+  unsigned int mi3u;   /* Map info command word 3 */
+  unsigned int mi3v;   /* Map info command word 3 */
+  unsigned int last_render;
+  unsigned int last_flip;
+  unsigned int second_field; /* Flags a surface that is only half done */
+  drmAddress data;
+  drm_handle_t offset;
+  unsigned int offsets[3];
+  i810XvMCContext *privContext;
+} i810XvMCSurface;
+
+/***************************************************************************
+// i810XvMCSubpicture: Private data structure for each XvMCSubpicture. This
+//  structure is referenced by the privData pointer in the XvMCSubpicture
+//  structure.
+***************************************************************************/
+typedef struct _i810XvMCSubpicture {
+  unsigned int pitch;
+  unsigned int dbi1;  /* Destination buffer info command word 1 */
+  unsigned int dbv1;  /* Destination buffer variables command word */
+  unsigned int mi1;   /* Map Info command word 1 (Minus bit 28) */
+  unsigned int mi2;   /* Map info command word 2 */
+  unsigned int mi3;   /* Map info command word 3 */
+  unsigned int last_render;
+  unsigned int last_flip;
+  drmAddress data;
+  drm_handle_t offset;
+  unsigned int offsets[1];
+  unsigned char palette[3][16];
+  i810XvMCContext *privContext;
+} i810XvMCSubpicture;
+
+typedef struct _Box {
+  int x1,y1,x2,y2;
+} Box, *BoxPtr;
+
+/***************************************************************************
+// drm_i810_overlay_t: Structure returned by overlay info ioctl.
+// NOTE: If you change this structure you will have to change the equiv.
+//  structure in the kernel.
+***************************************************************************/
+typedef struct _drm_i810_overlay_t {
+  unsigned int offset;
+  unsigned int physical;
+} drm_i810_overlay_t;
+
+/***************************************************************************
+// drm_i810_dma_t: Structure used by dma allocation ioctl.
+// NOTE: If you change this structure you will have to change the equiv.
+//  structure in the kernel.
+***************************************************************************/
+typedef struct _drm_i810_dma {
+  void *virtual;
+  int request_idx;
+  int request_size;
+  int granted;
+} drm_i810_dma_t;
+
+/***************************************************************************
+// drm_i810_mc_t: Structure used by mc dispatch ioctl.
+// NOTE: If you change this structure you will have to change the equiv.
+//  structure in the kernel.
+***************************************************************************/
+typedef struct _drm_i810_mc {
+  int idx;		/* buffer index */
+  int used;		/* nr bytes in use */
+  int num_blocks;         /* number of GFXBlocks */
+  int *length;	        /* List of lengths for GFXBlocks */
+  unsigned int last_render; /* Last render request */
+} drm_i810_mc_t;
+
+/* Subpicture fourcc */
+#define FOURCC_IA44 0x34344149
+
+/* Static Parameters */
+#define I810_XVMC_MAXWIDTH  720
+#define I810_XVMC_MAXHEIGHT  576
+#define I810_DEFAULT16_COLORKEY 31
+#define I810_DMA_BUF_NR 256
+
+/* COMMANDS */
+#define CMD_FLUSH        ((4<<23) | 0x1)
+#define BOOLEAN_ENA_1    ((3<<29) | (3<<24) | (3<<2))
+#define BOOLEAN_ENA_2    ((3<<29) | (4<<24) | (3<<16) | (1<<3) | (1<<2))
+#define DEST_BUFFER_INFO (0x15<<23)
+#define DEST_BUFFER_VAR  ((0x3<<29) | (0x1d<<24) | (0x85<<16))
+#define DRAWING_RECT_INFO ((3<<29) | (0x1d<<24) | (0x80<<16) | 3)
+#define GFXBLOCK         ((0x3<<29) | (0x1e<<24))
+#define CMD_MAP_INFO     ((0x3<<29) | (0x1d<<24) | 0x2)
+#define MAP_PALETTE_LOAD ((3<<29) | (0x1d<<24) | (0x82<<16) | 0xff)
+#define VERTEX_FORMAT    ((3<<29) | (0x5<<24))
+#define SRC_DEST_BLEND_MONO ((3<<29) | (8<<24))
+
+/* Bit Patterns */
+
+/*
+ * OV0CMD - Overlay Command Register
+ */
+#define VERTICAL_CHROMINANCE_FILTER     0x70000000
+#define VC_SCALING_OFF          0x00000000
+#define VC_LINE_REPLICATION     0x10000000
+#define VC_UP_INTERPOLATION     0x20000000
+#define VC_PIXEL_DROPPING       0x50000000
+#define VC_DOWN_INTERPOLATION   0x60000000
+#define VERTICAL_LUMINANCE_FILTER       0x0E000000
+#define VL_SCALING_OFF          0x00000000
+#define VL_LINE_REPLICATION     0x02000000
+#define VL_UP_INTERPOLATION     0x04000000
+#define VL_PIXEL_DROPPING       0x0A000000
+#define VL_DOWN_INTERPOLATION   0x0C000000
+#define HORIZONTAL_CHROMINANCE_FILTER   0x01C00000
+#define HC_SCALING_OFF          0x00000000
+#define HC_LINE_REPLICATION     0x00400000
+#define HC_UP_INTERPOLATION     0x00800000
+#define HC_PIXEL_DROPPING       0x01400000
+#define HC_DOWN_INTERPOLATION   0x01800000
+#define HORIZONTAL_LUMINANCE_FILTER     0x00380000
+#define HL_SCALING_OFF          0x00000000
+#define HL_LINE_REPLICATION     0x00080000
+#define HL_UP_INTERPOLATION     0x00100000
+#define HL_PIXEL_DROPPING       0x00280000
+#define HL_DOWN_INTERPOLATION   0x00300000
+
+#define Y_ADJUST                0x00010000
+#define OV_BYTE_ORDER           0x0000C000
+#define UV_SWAP                 0x00004000
+#define Y_SWAP                  0x00008000
+#define Y_AND_UV_SWAP           0x0000C000
+#define SOURCE_FORMAT           0x00003C00
+#define RGB_555                 0x00000800
+#define RGB_565                 0x00000C00
+#define YUV_422                 0x00002000
+#define YUV_411                 0x00002400
+#define YUV_420                 0x00003000
+#define YUV_410                 0x00003800
+#define VERTICAL_PHASE_BOTH     0x00000020
+#define FLIP_TYPE_FIELD         0x00000020
+#define FLIP_TYPE_FRAME         0x00000000
+#define BUFFER_AND_FIELD        0x00000006
+#define BUFFER0_FIELD0          0x00000000
+#define BUFFER0_FIELD1          0x00000002
+#define BUFFER1_FIELD0          0x00000004
+#define BUFFER1_FIELD1          0x00000006
+#define OVERLAY_ENABLE          0x00000001
+
+/*
+ * DOV0STA - Display/Overlay 0 Status Register
+ */
+#define DOV0STA         0x30008
+#define OV0ADD         0x30000
+#define MINUV_SCALE     0x1
+
+#define RGB16ToColorKey(c) \
+        (((c & 0xF800) << 8) | ((c & 0x07E0) << 5) | ((c & 0x001F) << 3))
+
+/* Locking Macros lightweight lock used to prevent relocking */
+#define I810_LOCK(c,f)                     \
+  if(!c->lock) {                           \
+    drmGetLock(c->fd, c->drmcontext, f);   \
+  }                                        \
+  c->lock++;
+
+#define I810_UNLOCK(c)                     \
+  c->lock--;                               \
+  if(!c->lock) {                           \
+    drmUnlock(c->fd, c->drmcontext);       \
+  }
+
+/*
+  Block until the passed in value (n) is the active
+  buffer on the overlay.
+*/
+#define BLOCK_OVERLAY(c,n)                        \
+  do {                                            \
+   int temp,i=0;                                  \
+   while(i < 100000) {                            \
+     temp = GET_FSTATUS(c);                       \
+     if(((temp & (1<<20))>>20) == n) {            \
+       break;                                     \
+     }                                            \
+     usleep(10);                                  \
+   }                                              \
+   if(i == 100000) {                              \
+     printf("Overlay Lockup.\n");                 \
+     return BadAlloc;                             \
+   }                                              \
+ }while(0);
+
+#define OVERLAY_INFO(c,i) drmCommandRead(c->fd, DRM_I810_OV0INFO, &i, sizeof(i))
+#define OVERLAY_FLIP(c) drmCommandNone(c->fd, DRM_I810_OV0FLIP)
+#define GET_FSTATUS(c) drmCommandNone(c->fd, DRM_I810_FSTATUS)
+#define I810_MC(c,mc) drmCommandWrite(c->fd, DRM_I810_MC, &mc, sizeof(mc))
+#define GET_RSTATUS(c) drmCommandNone(c->fd, DRM_I810_RSTATUS)
+#define GET_BUFFER(c,dma) drmCommandWriteRead(c->fd, DRM_I810_GETBUF, &dma, sizeof(drmI810DMA))
+#define FLUSH(c) drmCommandNone(c->fd, DRM_I810_FLUSH)
+
+/*
+  Definitions for temporary wire protocol hooks to be replaced
+  when a HW independent libXvMC is created.
+*/
+extern Status _xvmc_create_context(Display *dpy, XvMCContext *context,
+				   int *priv_count, uint **priv_data);
+
+extern Status _xvmc_destroy_context(Display *dpy, XvMCContext *context);
+
+extern Status _xvmc_create_surface(Display *dpy, XvMCContext *context,
+				   XvMCSurface *surface, int *priv_count,
+				   uint **priv_data);
+
+extern Status _xvmc_destroy_surface(Display *dpy, XvMCSurface *surface);
+
+extern Status  _xvmc_create_subpicture(Display *dpy, XvMCContext *context,
+				       XvMCSubpicture *subpicture,
+				       int *priv_count, uint **priv_data);
+
+extern Status   _xvmc_destroy_subpicture(Display *dpy,
+					 XvMCSubpicture *subpicture);
+
+/*
+  Prototypes
+*/
+drmBufPtr i810_get_free_buffer(i810XvMCContext *pI810XvMC);
+void i810_free_privContext(i810XvMCContext *pI810XvMC);
+void dp(unsigned int *address, unsigned int i);
+
+#endif
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
diff --git a/nx-X11/lib/XvMC/hw/i810/Imakefile b/nx-X11/lib/XvMC/hw/i810/Imakefile
new file mode 100644
index 000000000..b7956d01e
--- /dev/null
+++ b/nx-X11/lib/XvMC/hw/i810/Imakefile
@@ -0,0 +1,44 @@
+XCOMM $XFree86: xc/lib/XvMC/hw/i810/Imakefile,v 1.2 2002/10/30 12:52:02 alanh Exp $
+#define DoNormalLib NormalLibXvMC
+#define DoSharedLib SharedLibXvMC
+#define DoDebugLib DebugLibXvMC
+#define DoProfileLib ProfileLibXvMC
+#define LibName I810XvMC
+#define SoRev SOXVMCREV
+#define LibHeaders NO
+
+#include <Threads.tmpl>
+
+#ifdef SharedXvMCReqs
+REQUIREDLIBS = SharedXvMCReqs
+#endif
+
+#if Malloc0ReturnsNull
+ALLOC_DEFINES = -DMALLOC_0_RETURNS_NULL
+#endif
+
+      DEFINES = $(ALLOC_DEFINES) -DTRUE=1 -DFALSE=0
+     INCLUDES = -I$(XINCLUDESRC) -I$(INCLUDESRC) -I$(XLIBSRC) -I$(EXTINCSRC) \
+		-I$(XF86COMSRC) -I$(XF86OSSRC) -I$(DRMSRCDIR)/shared-core \
+		-I$(XF86DRIVERSRC)/i810
+         SRCS = I810XvMC.c
+         OBJS = I810XvMC.o xf86drm.o xf86drmHash.o xf86drmRandom.o xf86drmSL.o
+     LINTLIBS = $(LINTXLIB)
+
+
+#include <Library.tmpl>
+
+
+#if defined(LinuxArchitecture)
+
+LIBDRM_DIR = ../../../../extras/drm/libdrm
+
+LinkSourceFile(xf86drm.c,$(LIBDRM_DIR))
+LinkSourceFile(xf86drmHash.c,$(LIBDRM_DIR))
+LinkSourceFile(xf86drmRandom.c,$(LIBDRM_DIR))
+LinkSourceFile(xf86drmSL.c,$(LIBDRM_DIR))
+
+#endif
+
+DependTarget()
+
diff --git a/nx-X11/lib/XvMC/hw/via/driDrawable.c b/nx-X11/lib/XvMC/hw/via/driDrawable.c
new file mode 100644
index 000000000..35934d89d
--- /dev/null
+++ b/nx-X11/lib/XvMC/hw/via/driDrawable.c
@@ -0,0 +1,176 @@
+/*****************************************************************************
+ * driDrawable.c: Lean Version of DRI utilities.
+ *
+ * Copyright (c) 2005 Thomas Hellstrom. All rights reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice shall be included in
+ * all copies or substantial portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
+ * AUTHOR(S) OR COPYRIGHT HOLDER(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+ * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING
+ * FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER
+ * DEALINGS IN THE SOFTWARE.
+ */
+
+#ifdef HAVE_CONFIG_H
+#include "config.h"
+#endif
+
+#include <X11/Xlibint.h>
+#include <X11/Xutil.h>
+#include "xf86drm.h"
+#include "drm.h"
+#include "xf86dri.h"
+#include "drm_sarea.h"
+#include "driDrawable.h"
+
+
+static unsigned
+drawStamp(volatile drm_sarea_t *pSarea, int index)
+{
+    return pSarea->drawableTable[index].stamp;
+}
+
+int
+getDRIDrawableInfoLocked(void *drawHash, Display *display, int screen, Drawable draw,
+			 unsigned lockFlags, int drmFD, drm_context_t drmContext, 
+			 drmAddress sarea, Bool updateInfo, drawableInfo **info, 
+			 unsigned long infoSize)
+{
+  drawableInfo *drawInfo;
+  void *res;
+  drm_drawable_t drmDraw=0;
+  volatile drm_sarea_t *pSarea = (drm_sarea_t *) sarea;
+  drm_clip_rect_t *clipFront, *clipBack;
+
+  int ret;
+
+
+  if (drmHashLookup(drawHash, (unsigned long) draw, &res)) {
+
+      /*
+       * The drawable is unknown to us. Create it and put it in the
+       * hash table.
+       */
+
+      DRM_UNLOCK(drmFD, &pSarea->lock, drmContext); 
+      if (!uniDRICreateDrawable(display, screen, draw,
+				&drmDraw)) {
+	DRM_LOCK(drmFD, &pSarea->lock, drmContext, lockFlags); 
+	  return 1;
+      }
+      DRM_LOCK(drmFD, &pSarea->lock, drmContext, lockFlags); 
+
+
+      drawInfo = (drawableInfo *) malloc(infoSize);
+      if (!drawInfo) return 1;
+
+      drawInfo->drmDraw = drmDraw;
+      drawInfo->stamp = 0;
+      drawInfo->clipFront = 0;
+      drawInfo->clipBack = 0;
+
+      drmHashInsert( drawHash, (unsigned long) draw, drawInfo);
+
+  } else {
+      drawInfo = res;
+  }
+
+
+  drawInfo->touched = FALSE;
+  while (!drawInfo->clipFront || drawInfo->stamp != drawStamp(pSarea, drawInfo->index)) {
+
+      /*
+       * The drawable has been touched since we last got info about it.
+       * obtain new info from the X server.
+       */
+
+      drawInfo->touched = TRUE;
+
+      if (updateInfo || !drawInfo->clipFront) {
+	  DRM_UNLOCK(drmFD, &pSarea->lock, drmContext); 
+
+	  ret = uniDRIGetDrawableInfo(display, screen, draw,
+				      &drawInfo->index, &drawInfo->stamp, &drawInfo->x, 
+				      &drawInfo->y, &drawInfo->w, &drawInfo->h,
+				      &drawInfo->numClipFront, &clipFront, 
+				      &drawInfo->backX, &drawInfo->backY,
+				      &drawInfo->numClipBack, &clipBack);
+	  
+	  DRM_LIGHT_LOCK(drmFD, &pSarea->lock, drmContext); 
+
+	  /*
+	   * Error. Probably the drawable is destroyed. Return error and old values.
+	   */
+
+	  if (!ret) {
+	      free(drawInfo);
+	      drawInfo = NULL;
+	      drmHashDelete(drawHash, (unsigned long) draw);
+	      
+	      DRM_UNLOCK(drmFD, &pSarea->lock, drmContext);
+	      uniDRIDestroyDrawable( display, screen, draw);
+	      DRM_LOCK(drmFD, &pSarea->lock, drmContext, lockFlags);
+
+	      return 1;
+	  }
+	  
+	  if (drawInfo->stamp != drawStamp(pSarea, drawInfo->index)) {
+
+	      /*
+	       * The info is already outdated. Sigh. Have another go.
+	       */
+
+	      XFree(clipFront);
+	      XFree(clipBack);
+	      continue;
+	  }
+
+	  if (drawInfo->clipFront) XFree(drawInfo->clipFront);
+	  drawInfo->clipFront = clipFront;
+	  if (drawInfo->clipBack) XFree(drawInfo->clipBack);
+	  drawInfo->clipBack = clipBack;
+      } else {
+	  if (!drawInfo->clipFront) drawInfo->clipFront = (drm_clip_rect_t *) ~0UL;
+	  drawInfo->stamp = drawStamp(pSarea, drawInfo->index);
+      }
+  }
+  *info = drawInfo;
+  return 0;
+}
+
+void 
+driDestroyHashContents(void *drawHash)
+{
+    unsigned long key;
+    void *content;
+    drawableInfo *drawInfo;
+
+    
+    if (drmHashFirst(drawHash, &key, &content) < 1) 
+	return;
+    drawInfo = (drawableInfo *) content;
+    if (drawInfo->clipBack) XFree(drawInfo->clipBack);
+    if (drawInfo->clipFront) XFree(drawInfo->clipFront);
+    free(drawInfo);
+    while(drmHashNext(drawHash, &key, &content) == 1) {
+	drawInfo = (drawableInfo *) content;
+	if (drawInfo->clipBack) XFree(drawInfo->clipBack);
+	if (drawInfo->clipFront) XFree(drawInfo->clipFront);
+	free(drawInfo);
+    }
+	
+    return ;
+}
+    
+    
diff --git a/nx-X11/lib/XvMC/hw/via/driDrawable.h b/nx-X11/lib/XvMC/hw/via/driDrawable.h
new file mode 100644
index 000000000..e5651058b
--- /dev/null
+++ b/nx-X11/lib/XvMC/hw/via/driDrawable.h
@@ -0,0 +1,64 @@
+/*****************************************************************************
+ * driDrawable.h: Lean Version of DRI utilities.
+ *
+ * Copyright (c) 2005 Thomas Hellstrom. All rights reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice shall be included in
+ * all copies or substantial portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
+ * AUTHOR(S) OR COPYRIGHT HOLDER(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+ * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING
+ * FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER
+ * DEALINGS IN THE SOFTWARE.
+ */
+
+#ifndef _DRIDRAWABLE_H
+#define _DRIDRAWABLE_H
+
+typedef struct _drawableInfo {
+    drm_drawable_t drmDraw;
+    unsigned stamp;
+    unsigned index;
+    drm_clip_rect_t *clipFront;
+    drm_clip_rect_t *clipBack;
+    int x;
+    int y;
+    int w;
+    int h;
+    int backX;
+    int backY;
+    int numClipFront;
+    int numClipBack;
+    Bool touched;
+} drawableInfo;
+
+/*
+ * Get updated info about the drawable "draw". The drawableInfo record returned is malloced
+ * and administrated internally. Never free it unless you know exactly what you are doing.
+ * The drm hash table "drawHash" needs to be initialized externally.
+ */
+
+extern int
+getDRIDrawableInfoLocked (void *drawHash, Display *display, int screen, Drawable draw,
+			  unsigned lockFlags, int drmFD, drm_context_t drmContext,
+			  drmAddress sarea, Bool updateInfo, drawableInfo **info, 
+			  unsigned long infoSize);
+
+/*
+ * Free all resources created by the above function. Typically done on exit.
+ */
+
+extern void 
+driDestroyHashContents(void *drawHash);
+
+#endif
diff --git a/nx-X11/lib/XvMC/hw/via/unichrome/Imakefile b/nx-X11/lib/XvMC/hw/via/unichrome/Imakefile
new file mode 100644
index 000000000..bc9120972
--- /dev/null
+++ b/nx-X11/lib/XvMC/hw/via/unichrome/Imakefile
@@ -0,0 +1,51 @@
+#define DoNormalLib NormalLibXvMC
+#define DoSharedLib SharedLibXvMC
+#define DoDebugLib DebugLibXvMC
+#define DoProfileLib ProfileLibXvMC
+#define LibName viaXvMC
+#define SoRev SOXVMCREV
+#define LibHeaders NO
+
+#include <Threads.tmpl>
+
+#ifdef SharedXvMCReqs
+REQUIREDLIBS = SharedXvMCReqs -lXv
+#endif
+
+#if Malloc0ReturnsNull
+ALLOC_DEFINES = -DMALLOC_0_RETURNS_NULL
+#endif
+
+      DEFINES = $(ALLOC_DEFINES) -DTRUE=1 -DFALSE=0
+     INCLUDES = -I$(XINCLUDESRC) -I$(INCLUDESRC) -I$(XLIBSRC) -I$(EXTINCSRC) \
+		-I$(XF86COMSRC) -I$(XF86OSSRC) -I$(DRMSRCDIR)/shared-core \
+		-I$(XF86OSSRC)/linux/drm/kernel -I$(XF86DRIVERSRC)/via
+         SRCS = viaLowLevel.c 
+         OBJS = viaXvMC.o viaLowLevel.o xf86drm.o xf86drmHash.o \
+	        xf86drmRandom.o xf86drmSL.o xf86dri.o driDrawable.o
+     LINTLIBS = $(LINTXLIB)
+
+
+#include <Library.tmpl>
+
+
+#if defined(LinuxArchitecture)
+
+LIBDRM_DIR = ../../../../../extras/drm/libdrm
+
+LinkSourceFile(xf86drm.c,$(LIBDRM_DIR))
+LinkSourceFile(xf86drmHash.c,$(LIBDRM_DIR))
+LinkSourceFile(xf86drmRandom.c,$(LIBDRM_DIR))
+LinkSourceFile(xf86drmSL.c,$(LIBDRM_DIR))
+LinkSourceFile(viaLowLevel.h,..)
+LinkSourceFile(driDrawable.h,..)
+LinkSourceFile(viaXvMCPriv.h,..)
+LinkSourceFile(xf86dri.h,..)
+LinkSourceFile(xf86dristr.h,..)
+LinkSourceFile(xf86dri.c,..)
+LinkSourceFile(driDrawable.c,..)
+LinkSourceFile(viaXvMC.c,..)
+#endif
+
+DependTarget()
+
diff --git a/nx-X11/lib/XvMC/hw/via/unichrome/viaLowLevel.c b/nx-X11/lib/XvMC/hw/via/unichrome/viaLowLevel.c
new file mode 100644
index 000000000..1add1ca5d
--- /dev/null
+++ b/nx-X11/lib/XvMC/hw/via/unichrome/viaLowLevel.c
@@ -0,0 +1,1056 @@
+/*****************************************************************************
+ * VIA Unichrome XvMC extension client lib.
+ *
+ * Copyright (c) 2004 Thomas Hellström. All rights reserved.
+ * Copyright (c) 2003 Andreas Robinson. All rights reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice shall be included in
+ * all copies or substantial portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
+ * AUTHOR(S) OR COPYRIGHT HOLDER(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+ * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING
+ * FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER
+ * DEALINGS IN THE SOFTWARE.
+ */
+
+
+/*
+ * Low-level functions that deal directly with the hardware. In the future,
+ * these functions might be implemented in a kernel module. Also, some of them
+ * would benefit from DMA.
+ *
+ * Authors: Andreas Robinson 2003. Thomas Hellström 2004.
+ */
+
+
+
+#include "viaXvMCPriv.h"
+#include "viaLowLevel.h"
+#include <time.h>
+#include <sys/time.h>
+#include <stdio.h>
+
+typedef struct{
+    CARD32 agp_buffer[LL_AGP_CMDBUF_SIZE];
+    CARD32 pci_buffer[LL_PCI_CMDBUF_SIZE];
+    unsigned agp_pos;
+    unsigned pci_pos;
+    unsigned flip_pos;
+    int use_agp;   
+    int agp_mode;
+    int agp_header_start;
+    int agp_index;
+    int fd;
+    drm_context_t *drmcontext;
+    drmLockPtr hwLock;
+    drmAddress mmioAddress;
+    drmAddress fbAddress;
+    unsigned fbStride;
+    unsigned fbDepth;
+    unsigned width;
+    unsigned height;
+    unsigned curWaitFlags;
+    int performLocking;
+    unsigned errors;
+    drm_via_mem_t tsMem;
+    CARD32 tsOffset;
+    volatile CARD32 *tsP;
+    CARD32 curTimeStamp;
+    CARD32 lastReadTimeStamp;
+    int agpSync;
+    CARD32 agpSyncTimeStamp;
+    unsigned chipId;
+}XvMCLowLevel;
+
+
+
+
+
+/*
+ * For Other architectures than i386 these might have to be modified for
+ * bigendian etc.
+ */
+
+#define MPEGIN(xl,reg)							\
+    *((volatile CARD32 *)(((CARD8 *)(xl)->mmioAddress) + 0xc00 + (reg)))
+
+#define VIDIN(ctx,reg)							\
+    *((volatile CARD32 *)(((CARD8 *)(ctx)->mmioAddress) + 0x200 + (reg)))
+
+#define REGIN(ctx,reg)							\
+    *((volatile CARD32 *)(((CARD8 *)(ctx)->mmioAddress) + 0x0000 + (reg)))
+
+#define HQV_CONTROL             0x1D0
+#define HQV_SRC_STARTADDR_Y     0x1D4
+#define HQV_SRC_STARTADDR_U     0x1D8
+#define HQV_SRC_STARTADDR_V     0x1DC
+#define HQV_MINIFY_DEBLOCK      0x1E8    
+
+#define HQV_SW_FLIP         0x00000010
+#define HQV_FLIP_STATUS     0x00000001
+#define HQV_SUBPIC_FLIP     0x00008000
+#define HQV_FLIP_ODD        0x00000020
+#define HQV_DEINTERLACE     0x00010000
+#define HQV_FIELD_2_FRAME   0x00020000
+#define HQV_FRAME_2_FIELD   0x00040000
+#define HQV_FIELD_UV        0x00100000
+#define HQV_DEBLOCK_HOR     0x00008000
+#define HQV_DEBLOCK_VER     0x80000000
+
+#define V_COMPOSE_MODE          0x98
+#define V1_COMMAND_FIRE         0x80000000  
+#define V3_COMMAND_FIRE         0x40000000  
+
+/* SUBPICTURE Registers */
+#define SUBP_CONTROL_STRIDE     0x1C0
+#define SUBP_STARTADDR          0x1C4
+#define RAM_TABLE_CONTROL       0x1C8
+#define RAM_TABLE_READ          0x1CC
+
+/* SUBP_CONTROL_STRIDE              0x3c0 */
+#define SUBP_HQV_ENABLE             0x00010000
+#define SUBP_IA44                   0x00020000
+#define SUBP_AI44                   0x00000000
+#define SUBP_STRIDE_MASK            0x00001fff
+#define SUBP_CONTROL_MASK           0x00070000
+
+/* RAM_TABLE_CONTROL                0x3c8 */
+#define RAM_TABLE_RGB_ENABLE        0x00000007
+
+
+#define VIA_REG_STATUS          0x400
+#define VIA_REG_GEMODE          0x004
+#define VIA_REG_SRCBASE         0x030
+#define VIA_REG_DSTBASE         0x034
+#define VIA_REG_PITCH           0x038      
+#define VIA_REG_SRCCOLORKEY     0x01C      
+#define VIA_REG_KEYCONTROL      0x02C       
+#define VIA_REG_SRCPOS          0x008
+#define VIA_REG_DSTPOS          0x00C
+#define VIA_REG_GECMD           0x000
+#define VIA_REG_DIMENSION       0x010       /* width and height */
+#define VIA_REG_FGCOLOR         0x018
+
+
+#define VIA_VR_QUEUE_BUSY       0x00020000 /* Virtual Queue is busy */
+#define VIA_CMD_RGTR_BUSY       0x00000080  /* Command Regulator is busy */
+#define VIA_2D_ENG_BUSY         0x00000001  /* 2D Engine is busy */
+#define VIA_3D_ENG_BUSY         0x00000002  /* 3D Engine is busy */
+#define VIA_GEM_8bpp            0x00000000
+#define VIA_GEM_16bpp           0x00000100
+#define VIA_GEM_32bpp           0x00000300
+#define VIA_GEC_BLT             0x00000001
+#define VIA_PITCH_ENABLE        0x80000000
+#define VIA_GEC_INCX            0x00000000
+#define VIA_GEC_DECY            0x00004000
+#define VIA_GEC_INCY            0x00000000
+#define VIA_GEC_DECX            0x00008000
+#define VIA_GEC_FIXCOLOR_PAT    0x00002000
+
+
+#define VIA_BLIT_CLEAR 0x00
+#define VIA_BLIT_COPY 0xCC
+#define VIA_BLIT_FILL 0xF0
+#define VIA_BLIT_SET 0xFF
+
+#define VIA_SYNCWAITTIMEOUT 50000 /* Might be a bit conservative */
+#define VIA_DMAWAITTIMEOUT 150000
+#define VIA_VIDWAITTIMEOUT 50000
+#define VIA_XVMC_DECODERTIMEOUT 50000 /*(microseconds)*/
+
+#define H1_ADDR(val) (((val) >> 2) | 0xF0000000)
+#define WAITFLAGS(xl, flags)			\
+    (xl)->curWaitFlags |= (flags)
+#define BEGIN_RING_AGP(xl,size)						\
+    do {								\
+	if ((xl)->agp_pos > (LL_AGP_CMDBUF_SIZE-(size))) {		\
+	    agpFlush(xl);						\
+	}								\
+    } while(0)
+#define OUT_RING_AGP(xl, val)			\
+    (xl)->agp_buffer[(xl)->agp_pos++] = (val)
+#define OUT_RING_QW_AGP(xl, val1, val2)			\
+    do {						\
+	(xl)->agp_buffer[(xl)->agp_pos++] = (val1);	\
+	(xl)->agp_buffer[(xl)->agp_pos++] = (val2);	\
+    } while (0)
+  
+
+#define LL_HW_LOCK(xl)							\
+    do {								\
+	DRM_LOCK((xl)->fd,(xl)->hwLock,*(xl)->drmcontext,0);		\
+    } while(0);
+#define LL_HW_UNLOCK(xl)					\
+    do {							\
+	DRM_UNLOCK((xl)->fd,(xl)->hwLock,*(xl)->drmcontext);	\
+    } while(0);
+
+
+/*
+ * We want to have two concurrent types of thread taking the hardware
+ * lock simulataneously. One is the video out thread that needs immediate 
+ * access to flip an image. The other is everything else which may have
+ * the lock for quite some time. This is only so the video out thread can
+ * sneak in and display an image while other resources are busy.
+ */
+
+
+void 
+hwlLock(void *xlp, int videoLock) 
+{
+    XvMCLowLevel *xl = (XvMCLowLevel *) xlp;
+
+    LL_HW_LOCK(xl);
+}
+
+void 
+hwlUnlock(void *xlp, int videoLock) 
+{
+    XvMCLowLevel *xl = (XvMCLowLevel *) xlp;
+
+    LL_HW_UNLOCK(xl);
+}
+    
+static unsigned 
+timeDiff(struct timeval *now,struct timeval *then) {
+    return (now->tv_usec >= then->tv_usec) ?
+	now->tv_usec - then->tv_usec : 
+	1000000 - (then->tv_usec - now->tv_usec);
+}
+
+void 
+setAGPSyncLowLevel(void *xlp, int val, CARD32 timeStamp)
+{
+    XvMCLowLevel *xl = (XvMCLowLevel *) xlp;
+
+    xl->agpSync = val;
+    xl->agpSyncTimeStamp = timeStamp;
+}
+
+CARD32 
+viaDMATimeStampLowLevel(void *xlp)
+{
+    XvMCLowLevel *xl = (XvMCLowLevel *) xlp;
+
+    if (xl->use_agp) {
+	viaBlit(xl, 32, xl->tsOffset, 1, xl->tsOffset, 1, 1, 1, 0, 0, 
+		VIABLIT_FILL, xl->curTimeStamp);
+	return xl->curTimeStamp++;
+    }
+    return 0;
+}
+
+static void 
+viaDMAWaitTimeStamp(XvMCLowLevel *xl, CARD32 timeStamp, int doSleep) 
+{
+    struct timeval now, then;
+    struct timezone here;
+    struct timespec sleep, rem;
+
+    if (xl->use_agp && (timeStamp > xl->lastReadTimeStamp)) {
+	sleep.tv_nsec = 1;
+	sleep.tv_sec = 0;
+	here.tz_minuteswest = 0;
+	here.tz_dsttime = 0;
+	gettimeofday(&then,&here);
+ 
+	while(timeStamp > (xl->lastReadTimeStamp = *xl->tsP)) {
+	    gettimeofday(&now,&here); 
+	    if (timeDiff(&now,&then) > VIA_DMAWAITTIMEOUT) {
+		if((timeStamp > (xl->lastReadTimeStamp = *xl->tsP))) {
+		    xl->errors |= LL_DMA_TIMEDOUT;
+		    break;
+		}
+	    }
+	    if (doSleep) nanosleep(&sleep, &rem);
+	}
+    }
+}
+
+static int 
+viaDMAInitTimeStamp(XvMCLowLevel *xl) 
+{
+    int ret = 0;
+
+    if (xl->use_agp) {
+	xl->tsMem.context = *(xl->drmcontext);
+	xl->tsMem.size = 64;
+	xl->tsMem.type = VIA_MEM_VIDEO;
+	if (drmCommandWriteRead(xl->fd, DRM_VIA_ALLOCMEM, &xl->tsMem, sizeof(xl->tsMem)) < 0) 
+	    return ret;
+	if (xl->tsMem.size != 64)
+	    return -1;
+	xl->tsOffset = (xl->tsMem.offset + 31) & ~31; 
+	xl->tsP = (CARD32 *)xl->fbAddress + (xl->tsOffset >> 2); 
+	xl->curTimeStamp = 1;
+	*xl->tsP = 0;
+    }
+    return 0;
+}
+
+static int 
+viaDMACleanupTimeStamp(XvMCLowLevel *xl) 
+{
+
+    if (!(xl->tsMem.size) || !xl->use_agp) return 0;
+    return drmCommandWrite(xl->fd, DRM_VIA_FREEMEM, &xl->tsMem, sizeof(xl->tsMem)); 
+}
+
+
+static CARD32 
+viaMpegGetStatus(XvMCLowLevel *xl)
+{
+    return MPEGIN(xl,0x54);
+}
+
+static int 
+viaMpegIsBusy(XvMCLowLevel *xl, CARD32 mask, CARD32 idle)
+{
+    CARD32 tmp = viaMpegGetStatus(xl);
+
+    /*
+     * Error detected. 
+     * FIXME: Are errors really shown when error concealment is on?
+     */
+
+    if (tmp & 0x70) return 0; 
+
+    return (tmp & mask) != idle;
+}
+
+
+static void 
+syncDMA(XvMCLowLevel *xl, unsigned int doSleep)
+{
+
+    /*
+     * Ideally, we'd like to have an interrupt wait here, but, according to second hand
+     * information, the hardware does not support this, although earlier S3 chips do that.
+     * It is therefore not implemented into the DRM, and we'll do a user space wait here.
+     */
+
+    struct timeval now, then;
+    struct timezone here;
+    struct timespec sleep, rem;
+
+    sleep.tv_nsec = 1;
+    sleep.tv_sec = 0;
+    here.tz_minuteswest = 0;
+    here.tz_dsttime = 0;
+    gettimeofday(&then,&here);
+    while( !(REGIN(xl, VIA_REG_STATUS) & VIA_VR_QUEUE_BUSY)) {
+	gettimeofday(&now,&here); 
+	if (timeDiff(&now,&then) > VIA_DMAWAITTIMEOUT) {
+	    if( !(REGIN(xl, VIA_REG_STATUS) & VIA_VR_QUEUE_BUSY)) {
+		xl->errors |= LL_DMA_TIMEDOUT;
+		break;
+	    }
+	}
+	if (doSleep) nanosleep(&sleep, &rem);
+    }
+    while( REGIN(xl, VIA_REG_STATUS) & VIA_CMD_RGTR_BUSY ) {
+	gettimeofday(&now,&here); 
+	if (timeDiff(&now,&then) > VIA_DMAWAITTIMEOUT) {
+	    if( REGIN(xl, VIA_REG_STATUS) & VIA_CMD_RGTR_BUSY ) {
+		xl->errors |= LL_DMA_TIMEDOUT;
+		break;
+	    }
+	}
+	if (doSleep) nanosleep(&sleep, &rem);
+    }
+}
+
+static void 
+syncVideo(XvMCLowLevel *xl, unsigned int doSleep)
+{
+    /*
+     * Wait for HQV completion. Nothing strange here. We assume that the HQV
+     * Handles syncing to the V1 / V3 engines by itself. It should be safe to
+     * always wait for SUBPIC_FLIP completion although subpictures are not always
+     * used. 
+     */
+    
+    struct timeval now, then;
+    struct timezone here;
+    struct timespec sleep, rem;
+
+    sleep.tv_nsec = 1;
+    sleep.tv_sec = 0;
+    here.tz_minuteswest = 0;
+    here.tz_dsttime = 0;
+    gettimeofday(&then,&here); 
+    while(VIDIN(xl, HQV_CONTROL) & (HQV_SW_FLIP | HQV_SUBPIC_FLIP)) {
+	gettimeofday(&now,&here); 
+	if (timeDiff(&now,&then) > VIA_SYNCWAITTIMEOUT) {
+	    if(VIDIN(xl, HQV_CONTROL) & (HQV_SW_FLIP | HQV_SUBPIC_FLIP)) {
+		xl->errors |= LL_VIDEO_TIMEDOUT;
+		break;
+	    }
+	}
+	if (doSleep) nanosleep(&sleep, &rem);
+    }
+}
+
+static void 
+syncAccel(XvMCLowLevel *xl, unsigned int mode, unsigned int doSleep)
+{
+    struct timeval now, then;
+    struct timezone here;
+    struct timespec sleep, rem;
+    CARD32 mask = ((mode & LL_MODE_2D) ? VIA_2D_ENG_BUSY : 0) |
+	((mode & LL_MODE_3D) ? VIA_3D_ENG_BUSY : 0);
+
+    sleep.tv_nsec = 1;
+    sleep.tv_sec = 0;
+    here.tz_minuteswest = 0;
+    here.tz_dsttime = 0;
+    gettimeofday(&then,&here); 
+    while( REGIN(xl, VIA_REG_STATUS) & mask) {
+	gettimeofday(&now,&here); 
+	if (timeDiff(&now,&then) > VIA_SYNCWAITTIMEOUT) {
+	    if( REGIN(xl, VIA_REG_STATUS) & mask) {
+		xl->errors |= LL_ACCEL_TIMEDOUT;
+		break;
+	    }
+	}
+	if (doSleep) nanosleep(&sleep, &rem);
+    }
+}
+
+
+static void 
+syncMpeg(XvMCLowLevel *xl, unsigned int mode, unsigned int doSleep)
+{
+    /*
+     * Ideally, we'd like to have an interrupt wait here, but from information from VIA
+     * at least the MPEG completion interrupt is broken on the CLE266, which was
+     * discovered during validation of the chip.
+     */
+
+    struct timeval now, then;
+    struct timezone here;
+    struct timespec sleep, rem;
+    CARD32 busyMask = 0;
+    CARD32 idleVal = 0;
+    CARD32 ret;
+
+    sleep.tv_nsec = 1;
+    sleep.tv_sec = 0;
+    here.tz_minuteswest = 0;
+    here.tz_dsttime = 0;
+    gettimeofday(&then,&here); 
+    if (mode & LL_MODE_DECODER_SLICE) {
+	busyMask = VIA_SLICEBUSYMASK;
+	idleVal = VIA_SLICEIDLEVAL;
+    }
+    if (mode & LL_MODE_DECODER_IDLE) {
+	busyMask |= VIA_BUSYMASK;
+	idleVal = VIA_IDLEVAL;
+    }
+    while(viaMpegIsBusy(xl, busyMask, idleVal)) {
+	gettimeofday(&now,&here); 
+	if (timeDiff(&now,&then) > VIA_XVMC_DECODERTIMEOUT) {
+	    if (viaMpegIsBusy(xl, busyMask, idleVal)) {
+		xl->errors |= LL_DECODER_TIMEDOUT;
+	    }
+	    break;
+	}
+	if (doSleep) nanosleep(&sleep, &rem);
+    }
+  
+    ret = viaMpegGetStatus(xl);
+    if (ret & 0x70) {
+	xl->errors |= ((ret & 0x70) >> 3);
+    }
+    return;
+}
+
+static void 
+pciFlush(XvMCLowLevel *xl)
+{
+    int ret;
+    drm_via_cmdbuffer_t b;
+    unsigned mode=xl->curWaitFlags;
+  
+    b.buf = (char *)xl->pci_buffer;
+    b.size = xl->pci_pos * sizeof(CARD32);
+    if (xl->performLocking) hwlLock(xl,0);
+    if ((mode != LL_MODE_VIDEO) && (mode != 0)) 
+        syncDMA(xl, 0);
+    if ((mode & LL_MODE_2D) || (mode & LL_MODE_3D)) 
+	syncAccel(xl, mode, 0);
+    if (mode & LL_MODE_VIDEO) 
+	syncVideo(xl, 0);
+    if (mode & (LL_MODE_DECODER_SLICE | LL_MODE_DECODER_IDLE))
+	syncMpeg(xl, mode, 0);
+    ret = drmCommandWrite(xl->fd, DRM_VIA_PCICMD, &b, sizeof(b));
+    if (xl->performLocking) hwlUnlock(xl,0);
+    if (ret) {
+	xl->errors |= LL_PCI_COMMAND_ERR;
+    }
+    xl->pci_pos = 0;
+    xl->curWaitFlags = 0;
+}
+  
+static void 
+agpFlush(XvMCLowLevel *xl)
+{
+    drm_via_cmdbuffer_t b;
+    int ret;
+
+    if (xl->use_agp) {
+	b.buf = (char *)xl->agp_buffer;
+	b.size = xl->agp_pos * sizeof(CARD32);
+	if (xl->agpSync) {
+	    syncXvMCLowLevel(xl, LL_MODE_DECODER_IDLE, 1, xl->agpSyncTimeStamp);
+	    xl->agpSync = 0;
+	}
+	if (xl->performLocking) hwlLock(xl,0);
+	do {
+	    ret = drmCommandWrite(xl->fd, DRM_VIA_CMDBUFFER, &b, sizeof(b));
+	} while (-EAGAIN == ret);
+	if (xl->performLocking) hwlUnlock(xl,0);
+      
+	if (ret) {
+	    xl->errors |= LL_AGP_COMMAND_ERR;
+	} else {
+	    xl->agp_pos = 0;
+	}
+	xl->curWaitFlags &= LL_MODE_VIDEO;
+    } else {
+	unsigned mode=xl->curWaitFlags;
+  
+	b.buf = (char *)xl->agp_buffer;
+	b.size = xl->agp_pos * sizeof(CARD32);
+	if (xl->performLocking) hwlLock(xl,0);
+	if ((mode != LL_MODE_VIDEO) && (mode != 0)) 
+	    syncDMA(xl, 0);
+	if ((mode & LL_MODE_2D) || (mode & LL_MODE_3D)) 
+	    syncAccel(xl, mode, 0);
+	if (mode & LL_MODE_VIDEO) 
+	    syncVideo(xl, 0);
+	if (mode & (LL_MODE_DECODER_SLICE | LL_MODE_DECODER_IDLE))
+	    syncMpeg(xl, mode, 0);
+	ret = drmCommandWrite(xl->fd, DRM_VIA_PCICMD, &b, sizeof(b));
+	if (xl->performLocking) hwlUnlock(xl,0);
+	if (ret) {
+	    xl->errors |= LL_PCI_COMMAND_ERR;
+	}
+	xl->agp_pos = 0;
+	xl->curWaitFlags = 0;
+    }      
+}
+
+unsigned 
+flushXvMCLowLevel(void *xlp) 
+{
+    unsigned 
+	errors;
+    XvMCLowLevel *xl = (XvMCLowLevel *) xlp;
+    
+    if(xl->pci_pos) pciFlush(xl);
+    if(xl->agp_pos) agpFlush(xl);
+    errors = xl->errors;
+    xl->errors = 0;
+    return errors;
+}
+
+void 
+flushPCIXvMCLowLevel(void  *xlp) 
+{
+    XvMCLowLevel *xl = (XvMCLowLevel *) xlp;
+
+    if(xl->pci_pos) pciFlush(xl); 
+    if (!xl->use_agp && xl->agp_pos) agpFlush(xl);
+}
+
+
+__inline static void pciCommand(XvMCLowLevel *xl, unsigned offset, unsigned value, unsigned flags)
+{
+    if (xl->pci_pos > (LL_PCI_CMDBUF_SIZE-2)) pciFlush(xl); 
+    if (flags) xl->curWaitFlags |= flags;
+    xl->pci_buffer[xl->pci_pos++] = (offset >> 2) | 0xF0000000;
+    xl->pci_buffer[xl->pci_pos++] = value;
+}
+
+void 
+viaMpegSetSurfaceStride(void *xlp, ViaXvMCContext *ctx)
+{
+    CARD32 y_stride = ctx->yStride;
+    CARD32 uv_stride = y_stride >> 1;
+    XvMCLowLevel *xl = (XvMCLowLevel *) xlp;
+
+    
+    BEGIN_RING_AGP(xl, 2);
+    OUT_RING_QW_AGP(xl, H1_ADDR(0xc50), (y_stride >> 3) | ((uv_stride >> 3) << 16)); 
+    WAITFLAGS(xl, LL_MODE_DECODER_IDLE);
+}
+
+
+void 
+viaVideoSetSWFLipLocked(void *xlp, unsigned yOffs, unsigned uOffs,
+			unsigned vOffs, unsigned yStride, unsigned uvStride)
+{
+    XvMCLowLevel *xl = (XvMCLowLevel *) xlp;
+
+    pciCommand(xl,HQV_SRC_STARTADDR_Y | 0x200,yOffs,LL_MODE_VIDEO);
+    pciCommand(xl,HQV_SRC_STARTADDR_U | 0x200,uOffs,0);
+    pciCommand(xl,HQV_SRC_STARTADDR_V | 0x200,vOffs,0);
+}
+    
+void 
+viaVideoSWFlipLocked(void *xlp, unsigned flags,
+		     int progressiveSequence)
+{
+    CARD32 andWd,orWd;
+    XvMCLowLevel *xl = (XvMCLowLevel *) xlp;
+
+    andWd = 0;
+    orWd = 0;
+
+    if ((flags & XVMC_FRAME_PICTURE) == XVMC_BOTTOM_FIELD) {
+	andWd = 0xFFFFFFFFU;
+	orWd = HQV_FIELD_UV   |
+	    HQV_DEINTERLACE   |
+	    HQV_FIELD_2_FRAME |
+	    HQV_FRAME_2_FIELD |
+	    HQV_SW_FLIP       |
+	    HQV_FLIP_ODD      |
+	    HQV_FLIP_STATUS   |
+	    HQV_SUBPIC_FLIP;
+    } else if ((flags & XVMC_FRAME_PICTURE) == XVMC_TOP_FIELD) {
+	andWd = ~HQV_FLIP_ODD;
+	orWd = HQV_FIELD_UV   |
+	    HQV_DEINTERLACE   |
+	    HQV_FIELD_2_FRAME |
+	    HQV_FRAME_2_FIELD |
+	    HQV_SW_FLIP       |
+	    HQV_FLIP_STATUS   |
+	    HQV_SUBPIC_FLIP;
+    } else if ((flags & XVMC_FRAME_PICTURE) == XVMC_FRAME_PICTURE) {
+	andWd = ~(HQV_DEINTERLACE   | 
+		  HQV_FRAME_2_FIELD |
+		  HQV_FIELD_2_FRAME |
+		  HQV_FIELD_UV);
+	orWd = HQV_SW_FLIP   |
+	    HQV_FLIP_STATUS  |
+	    HQV_SUBPIC_FLIP;
+    }	    
+    if (progressiveSequence) {
+        andWd &= ~HQV_FIELD_UV;
+        orWd &= ~HQV_FIELD_UV;
+    }
+    
+    pciCommand(xl,HQV_CONTROL | 0x200,(VIDIN(xl,HQV_CONTROL) & andWd) | orWd, 0);
+}
+	
+void 
+viaMpegSetFB(void *xlp,unsigned i,
+	     unsigned yOffs,
+	     unsigned uOffs,
+	     unsigned vOffs) 
+{
+    XvMCLowLevel *xl = (XvMCLowLevel *) xlp;
+
+    i *= 12;
+    BEGIN_RING_AGP(xl, 6);
+    OUT_RING_QW_AGP(xl, H1_ADDR(0xc20 + i), yOffs >> 3);
+    OUT_RING_QW_AGP(xl, H1_ADDR(0xc24 + i), uOffs >> 3);
+    OUT_RING_QW_AGP(xl, H1_ADDR(0xc28 + i), vOffs >> 3);
+    WAITFLAGS(xl, LL_MODE_DECODER_IDLE);
+}
+
+void 
+viaMpegBeginPicture(void *xlp,ViaXvMCContext *ctx,
+		    unsigned width,
+		    unsigned height,
+		    const XvMCMpegControl *control) {
+				  
+    unsigned j, mb_width, mb_height;
+    XvMCLowLevel *xl = (XvMCLowLevel *) xlp;
+
+    mb_width = (width + 15) >> 4;
+
+    mb_height =
+	((control->mpeg_coding == XVMC_MPEG_2) && 
+	 (control->flags & XVMC_PROGRESSIVE_SEQUENCE)) ?
+	2*((height+31) >> 5) : (((height+15) >> 4));
+
+    BEGIN_RING_AGP(xl, 144);
+    WAITFLAGS(xl, LL_MODE_DECODER_IDLE);
+
+    OUT_RING_QW_AGP(xl, H1_ADDR(0xc00),
+		    ((control->picture_structure & XVMC_FRAME_PICTURE) << 2) |
+		    ((control->picture_coding_type & 3) << 4) |
+		    ((control->flags & XVMC_ALTERNATE_SCAN) ? (1 << 6) : 0));
+    
+    if (!(ctx->intraLoaded)) {
+	OUT_RING_QW_AGP(xl, H1_ADDR(0xc5c), 0);
+	for (j = 0; j < 64; j += 4) {
+            OUT_RING_QW_AGP(xl, H1_ADDR(0xc60),  
+			    ctx->intra_quantiser_matrix[j] |
+			    (ctx->intra_quantiser_matrix[j+1] << 8) |
+			    (ctx->intra_quantiser_matrix[j+2] << 16) |
+			    (ctx->intra_quantiser_matrix[j+3] << 24));
+        }
+	ctx->intraLoaded = 1;
+    }
+
+    if (!(ctx->nonIntraLoaded)) {
+	OUT_RING_QW_AGP(xl, H1_ADDR(0xc5c), 1);
+	for (j = 0; j < 64; j += 4) {
+            OUT_RING_QW_AGP(xl, H1_ADDR(0xc60),  
+			    ctx->non_intra_quantiser_matrix[j] |
+			    (ctx->non_intra_quantiser_matrix[j+1] << 8) |
+			    (ctx->non_intra_quantiser_matrix[j+2] << 16) |
+			    (ctx->non_intra_quantiser_matrix[j+3] << 24));
+        }
+	ctx->nonIntraLoaded = 1;
+    }
+
+    if (!(ctx->chromaIntraLoaded)) {
+	OUT_RING_QW_AGP(xl, H1_ADDR(0xc5c), 2);
+	for (j = 0; j < 64; j += 4) {
+            OUT_RING_QW_AGP(xl, H1_ADDR(0xc60),  
+			    ctx->chroma_intra_quantiser_matrix[j] |
+			    (ctx->chroma_intra_quantiser_matrix[j+1] << 8) |
+			    (ctx->chroma_intra_quantiser_matrix[j+2] << 16) |
+			    (ctx->chroma_intra_quantiser_matrix[j+3] << 24));
+        }
+	ctx->chromaIntraLoaded = 1;
+    }
+
+    if (!(ctx->chromaNonIntraLoaded)) {
+	OUT_RING_QW_AGP(xl, H1_ADDR(0xc5c), 3);
+	for (j = 0; j < 64; j += 4) {
+            OUT_RING_QW_AGP(xl, H1_ADDR(0xc60),  
+			    ctx->chroma_non_intra_quantiser_matrix[j] |
+			    (ctx->chroma_non_intra_quantiser_matrix[j+1] << 8) |
+			    (ctx->chroma_non_intra_quantiser_matrix[j+2] << 16) |
+			    (ctx->chroma_non_intra_quantiser_matrix[j+3] << 24));
+        }
+	ctx->chromaNonIntraLoaded = 1;
+    }
+    
+    OUT_RING_QW_AGP(xl, H1_ADDR(0xc90),  
+		    ((mb_width * mb_height) & 0x3fff) |
+		    ((control->flags & XVMC_PRED_DCT_FRAME) ? ( 1 << 14)  : 0) |
+		    ((control->flags & XVMC_TOP_FIELD_FIRST) ? (1 << 15) : 0 )  |
+		    ((control->mpeg_coding == XVMC_MPEG_2) ? (1 << 16) : 0) |
+		    ((mb_width & 0xff) << 18));
+    	
+    OUT_RING_QW_AGP(xl, H1_ADDR(0xc94),
+		    ((control->flags & XVMC_CONCEALMENT_MOTION_VECTORS) ? 1 : 0) |
+		    ((control->flags & XVMC_Q_SCALE_TYPE) ? 2 : 0) |
+		    ((control->intra_dc_precision & 3) << 2) |
+		    (((1 + 0x100000 / mb_width) & 0xfffff) << 4) |
+		    ((control->flags & XVMC_INTRA_VLC_FORMAT) ? (1 << 24) : 0));
+
+    OUT_RING_QW_AGP(xl, H1_ADDR(0xc98), 
+		    (((control->FHMV_range) & 0xf) << 0) |
+		    (((control->FVMV_range) & 0xf) << 4) |
+		    (((control->BHMV_range) & 0xf) << 8) |
+		    (((control->BVMV_range) & 0xf) << 12) |
+		    ((control->flags & XVMC_SECOND_FIELD) ?  (1 << 20) : 0) | 
+		    (0x0a6 << 16));
+    
+}
+
+
+
+void 
+viaMpegReset(void *xlp)
+{
+    int i,j;
+    XvMCLowLevel *xl = (XvMCLowLevel *) xlp;
+
+    BEGIN_RING_AGP(xl, 100);
+    WAITFLAGS(xl, LL_MODE_DECODER_IDLE);
+
+    for (i = 0; i < 14; i++)
+        OUT_RING_QW_AGP(xl, H1_ADDR(0xc08) ,0);
+
+    OUT_RING_QW_AGP(xl, H1_ADDR(0xc98), 0x400000);
+
+    for (i = 0; i < 6; i++) {
+        OUT_RING_QW_AGP(xl, H1_ADDR(0xc0c), 0x43 | 0x20);
+        for (j = 0xc10; j < 0xc20; j += 4)
+	    OUT_RING_QW_AGP(xl, H1_ADDR(j), 0);
+    }
+
+    
+    OUT_RING_QW_AGP(xl,H1_ADDR(0xc0c), 0xc3 | 0x20);
+    for (j = 0xc10; j < 0xc20; j += 4)
+	OUT_RING_QW_AGP(xl,H1_ADDR(j),0);
+    
+}
+
+void 
+viaMpegWriteSlice(void *xlp, CARD8* slice, int nBytes, CARD32 sCode)
+{
+    int i, n, r;
+    CARD32* buf;
+    int count;
+    XvMCLowLevel *xl = (XvMCLowLevel *) xlp;
+
+
+    if (xl->errors & (LL_DECODER_TIMEDOUT |
+		      LL_IDCT_FIFO_ERROR  |
+		      LL_SLICE_FIFO_ERROR |
+		      LL_SLICE_FAULT)) return;
+
+    n = nBytes >> 2;
+    if (sCode) nBytes += 4;
+    r = nBytes & 3;
+    buf = (CARD32*) slice;
+
+    if (r) nBytes += 4 - r;
+
+    nBytes += 8;
+
+    BEGIN_RING_AGP(xl, 4);
+    WAITFLAGS(xl, LL_MODE_DECODER_IDLE);
+
+    OUT_RING_QW_AGP(xl, H1_ADDR(0xc9c), nBytes);
+    
+    if (sCode) OUT_RING_QW_AGP(xl, H1_ADDR(0xca0), sCode);
+
+    i = 0;
+    count = 0;
+
+    do {
+	count += (LL_AGP_CMDBUF_SIZE -20) >> 1;
+	count = (count > n) ? n : count;
+	BEGIN_RING_AGP(xl, (count - i) << 1);
+
+	for (; i < count; i++) {
+	    OUT_RING_QW_AGP(xl, H1_ADDR(0xca0), *buf++);
+	}
+    } while (i < n);
+
+    BEGIN_RING_AGP(xl, 6);
+
+    if (r) {
+	OUT_RING_QW_AGP(xl, H1_ADDR(0xca0), *buf & ((1 << (r << 3)) - 1));
+    }
+    OUT_RING_QW_AGP(xl, H1_ADDR(0xca0), 0);
+    OUT_RING_QW_AGP(xl, H1_ADDR(0xca0), 0);
+
+}
+
+void 
+viaVideoSubPictureOffLocked(void *xlp) {
+
+    CARD32 stride;
+    XvMCLowLevel *xl = (XvMCLowLevel *) xlp;
+
+
+    stride = VIDIN(xl,SUBP_CONTROL_STRIDE);
+
+    pciCommand(xl, SUBP_CONTROL_STRIDE | 0x200, stride & ~SUBP_HQV_ENABLE,LL_MODE_VIDEO);
+}
+
+void 
+viaVideoSubPictureLocked(void *xlp,ViaXvMCSubPicture *pViaSubPic) {
+
+    unsigned i;
+    CARD32 cWord;
+    XvMCLowLevel *xl = (XvMCLowLevel *) xlp;
+
+
+    for (i=0; i<VIA_SUBPIC_PALETTE_SIZE; ++i) {
+	pciCommand(xl, RAM_TABLE_CONTROL | 0x200, pViaSubPic->palette[i],LL_MODE_VIDEO);
+    }
+
+    pciCommand(xl, SUBP_STARTADDR | 0x200, pViaSubPic->offset, 0);
+    cWord = (pViaSubPic->stride & SUBP_STRIDE_MASK) | SUBP_HQV_ENABLE;
+    cWord |= (pViaSubPic->ia44) ? SUBP_IA44 : SUBP_AI44;
+    pciCommand(xl, SUBP_CONTROL_STRIDE | 0x200, cWord, 0);
+}
+
+void 
+viaBlit(void *xlp,unsigned bpp,unsigned srcBase,
+	unsigned srcPitch,unsigned dstBase,unsigned dstPitch,
+	unsigned w,unsigned h,int xdir,int ydir, unsigned blitMode, 
+	unsigned color) 
+{
+
+    CARD32 dwGEMode = 0, srcY=0, srcX, dstY=0, dstX;
+    CARD32 cmd;
+    XvMCLowLevel *xl = (XvMCLowLevel *) xlp;
+
+
+    if (!w || !h)
+        return;
+
+    switch (bpp) {
+    case 16:
+        dwGEMode |= VIA_GEM_16bpp;
+        break;
+    case 32:
+        dwGEMode |= VIA_GEM_32bpp;
+	break;
+    default:
+        dwGEMode |= VIA_GEM_8bpp;
+        break;
+    }
+
+    srcX = srcBase & 31;
+    dstX = dstBase & 31;
+    switch (bpp) {
+    case 16:
+        dwGEMode |= VIA_GEM_16bpp;
+	srcX >>= 2;
+	dstX >>= 2;
+        break;
+    case 32:
+        dwGEMode |= VIA_GEM_32bpp;
+	srcX >>= 4;
+	dstX >>= 4;
+	break;
+    default:
+        dwGEMode |= VIA_GEM_8bpp;
+        break;
+    }
+
+    BEGIN_RING_AGP(xl, 20);
+    WAITFLAGS(xl, LL_MODE_2D);
+
+
+    OUT_RING_QW_AGP(xl, H1_ADDR(VIA_REG_GEMODE), dwGEMode);
+    cmd = 0; 
+
+    if (xdir < 0) {
+        cmd |= VIA_GEC_DECX;
+        srcX += (w - 1);
+        dstX += (w - 1);
+    }
+    if (ydir < 0) {
+        cmd |= VIA_GEC_DECY;
+        srcY += (h - 1);
+        dstY += (h - 1);
+    }
+
+    switch(blitMode) {
+    case VIABLIT_TRANSCOPY:
+	OUT_RING_QW_AGP(xl, H1_ADDR(VIA_REG_SRCCOLORKEY), color);
+	OUT_RING_QW_AGP(xl, H1_ADDR(VIA_REG_KEYCONTROL), 0x4000);
+	cmd |= VIA_GEC_BLT | (VIA_BLIT_COPY << 24);
+	break;
+    case VIABLIT_FILL:
+	OUT_RING_QW_AGP(xl, H1_ADDR(VIA_REG_FGCOLOR), color);
+	cmd |= VIA_GEC_BLT | VIA_GEC_FIXCOLOR_PAT | (VIA_BLIT_FILL << 24);
+	break;
+    default:
+	OUT_RING_QW_AGP(xl, H1_ADDR(VIA_REG_KEYCONTROL), 0x0);
+	cmd |= VIA_GEC_BLT | (VIA_BLIT_COPY << 24);
+    }	
+
+    OUT_RING_QW_AGP(xl, H1_ADDR(VIA_REG_SRCBASE), (srcBase & ~31) >> 3);
+    OUT_RING_QW_AGP(xl, H1_ADDR(VIA_REG_DSTBASE), (dstBase & ~31) >> 3);
+    OUT_RING_QW_AGP(xl, H1_ADDR(VIA_REG_PITCH), VIA_PITCH_ENABLE |
+		    (srcPitch >> 3) | (((dstPitch) >> 3) << 16));
+    OUT_RING_QW_AGP(xl, H1_ADDR(VIA_REG_SRCPOS), ((srcY << 16) | srcX));
+    OUT_RING_QW_AGP(xl, H1_ADDR(VIA_REG_DSTPOS), ((dstY << 16) | dstX));
+    OUT_RING_QW_AGP(xl, H1_ADDR(VIA_REG_DIMENSION), (((h - 1) << 16) | (w - 1)));
+    OUT_RING_QW_AGP(xl, H1_ADDR(VIA_REG_GECMD), cmd);
+}
+
+unsigned 
+syncXvMCLowLevel(void *xlp, unsigned int mode, unsigned int doSleep,
+		 CARD32 timeStamp)
+{
+    unsigned
+	errors;
+    XvMCLowLevel *xl = (XvMCLowLevel *) xlp;
+
+    if (mode == 0) {
+	errors = xl->errors;
+	xl->errors = 0;
+	return errors;
+    }
+
+    if ((mode & (LL_MODE_VIDEO | LL_MODE_3D)) || !xl->use_agp) {
+	if (xl->performLocking) 
+	    hwlLock(xl,0);
+	if ((mode != LL_MODE_VIDEO))
+	    syncDMA(xl, doSleep);
+	if (mode & LL_MODE_3D) 
+	    syncAccel(xl, mode, doSleep);
+	if (mode & LL_MODE_VIDEO)
+	    syncVideo(xl, doSleep);
+	if (xl->performLocking) 
+	    hwlUnlock(xl,0);
+    } else {
+	viaDMAWaitTimeStamp(xl, timeStamp, doSleep);
+    }
+
+    if (mode & (LL_MODE_DECODER_SLICE | LL_MODE_DECODER_IDLE)) 
+	syncMpeg(xl, mode, doSleep);
+
+    errors = xl->errors;
+    xl->errors = 0;
+    
+    return errors;
+}
+
+
+extern void
+*initXvMCLowLevel(int fd, drm_context_t *ctx,
+		 drmLockPtr hwLock, drmAddress mmioAddress, 
+		 drmAddress fbAddress, unsigned fbStride, unsigned fbDepth,
+		 unsigned width, unsigned height, int useAgp, unsigned chipId )
+
+{
+    int ret;
+  
+    XvMCLowLevel *xl = (XvMCLowLevel *)malloc(sizeof(XvMCLowLevel));
+    if (!xl) return NULL;
+
+    xl->agp_pos = 0;
+    xl->pci_pos = 0;
+    xl->use_agp = useAgp;
+    xl->fd = fd;
+    xl->drmcontext = ctx;
+    xl->hwLock = hwLock;
+    xl->mmioAddress = mmioAddress;
+    xl->fbAddress = fbAddress;
+    xl->curWaitFlags = 0;
+    xl->performLocking = 1;
+    xl->errors = 0;
+    xl->agpSync = 0;
+    ret = viaDMAInitTimeStamp(xl); 
+    if (ret) {
+	free(xl);
+	return NULL;
+    }
+    return xl;
+}
+
+void 
+setLowLevelLocking(void *xlp, int performLocking)
+{
+    XvMCLowLevel *xl = (XvMCLowLevel *) xlp;
+
+    xl->performLocking = performLocking;
+}
+
+void 
+closeXvMCLowLevel(void *xlp) 
+{
+    XvMCLowLevel *xl = (XvMCLowLevel *) xlp;
+
+    viaDMACleanupTimeStamp(xl); 
+    free(xl);
+}
+
diff --git a/nx-X11/lib/XvMC/hw/via/unichromeProA/Imakefile b/nx-X11/lib/XvMC/hw/via/unichromeProA/Imakefile
new file mode 100644
index 000000000..f87ae9b05
--- /dev/null
+++ b/nx-X11/lib/XvMC/hw/via/unichromeProA/Imakefile
@@ -0,0 +1,50 @@
+#define DoNormalLib NormalLibXvMC
+#define DoSharedLib SharedLibXvMC
+#define DoDebugLib DebugLibXvMC
+#define DoProfileLib ProfileLibXvMC
+#define LibName viaXvMCPro
+#define SoRev SOXVMCREV
+#define LibHeaders NO
+
+#include <Threads.tmpl>
+
+#ifdef SharedXvMCReqs
+REQUIREDLIBS = SharedXvMCReqs -lXv
+#endif
+
+#if Malloc0ReturnsNull
+ALLOC_DEFINES = -DMALLOC_0_RETURNS_NULL
+#endif
+
+      DEFINES = $(ALLOC_DEFINES) -DTRUE=1 -DFALSE=0
+     INCLUDES = -I$(XINCLUDESRC) -I$(INCLUDESRC) -I$(XLIBSRC) -I$(EXTINCSRC) \
+		-I$(XF86COMSRC) -I$(XF86OSSRC) -I$(DRMSRCDIR)/shared-core \
+		-I$(XF86OSSRC)/linux/drm/kernel -I$(XF86DRIVERSRC)/via
+         SRCS = viaLowLevelPro.c 
+         OBJS = viaXvMC.o viaLowLevelPro.o xf86drm.o xf86drmHash.o \
+	        xf86drmRandom.o xf86drmSL.o xf86dri.o driDrawable.o
+     LINTLIBS = $(LINTXLIB)
+
+
+#include <Library.tmpl>
+
+
+#if defined(LinuxArchitecture)
+LIBDRM_DIR = ../../../../../extras/drm/libdrm
+
+LinkSourceFile(xf86drm.c,$(LIBDRM_DIR))
+LinkSourceFile(xf86drmHash.c,$(LIBDRM_DIR))
+LinkSourceFile(xf86drmRandom.c,$(LIBDRM_DIR))
+LinkSourceFile(xf86drmSL.c,$(LIBDRM_DIR))
+LinkSourceFile(viaLowLevel.h,..)
+LinkSourceFile(driDrawable.h,..)
+LinkSourceFile(viaXvMCPriv.h,..)
+LinkSourceFile(xf86dri.h,..)
+LinkSourceFile(xf86dristr.h,..)
+LinkSourceFile(xf86dri.c,..)
+LinkSourceFile(driDrawable.c,..)
+LinkSourceFile(viaXvMC.c,..)
+#endif
+
+DependTarget()
+
diff --git a/nx-X11/lib/XvMC/hw/via/unichromeProA/viaLowLevelPro.c b/nx-X11/lib/XvMC/hw/via/unichromeProA/viaLowLevelPro.c
new file mode 100644
index 000000000..666166c96
--- /dev/null
+++ b/nx-X11/lib/XvMC/hw/via/unichromeProA/viaLowLevelPro.c
@@ -0,0 +1,1465 @@
+/*****************************************************************************
+ * VIA Unichrome XvMC extension client lib.
+ *
+ * Copyright (c) 2004 Thomas Hellström. All rights reserved.
+ * Copyright (c) 2003 Andreas Robinson. All rights reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice shall be included in
+ * all copies or substantial portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
+ * AUTHOR(S) OR COPYRIGHT HOLDER(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+ * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING
+ * FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER
+ * DEALINGS IN THE SOFTWARE.
+ */
+
+
+/*
+ * Low-level functions that deal directly with the hardware. In the future,
+ * these functions might be implemented in a kernel module. Also, some of them
+ * would benefit from DMA.
+ *
+ * Authors: 
+ *  Andreas Robinson 2003. (Initial decoder interface functions).
+ *  Thomas Hellstrom 2004, 2005 (Blitting functions, AGP and locking, Unichrome Pro Video AGP).
+ *  Ivor Hewitt 2005 (Unichrome Pro modifications and merging).
+ */
+
+/* IH
+ * I've left the proReg or-ing in case we need/want to implement the V1/V3
+ * register toggle too, which also moves the register locations.
+ * The CN400 has dual mpeg decoders, not sure at the moment whether these
+ * are also operated through independent registers also.
+ */
+
+#undef VIDEO_DMA
+#define HQV_USE_IRQ
+#define UNICHROME_PRO
+
+#include "viaXvMCPriv.h"
+#include "viaLowLevel.h"
+#include "driDrawable.h"
+#include <time.h>
+#include <sys/time.h>
+#include <stdio.h>
+
+typedef enum {ll_init, ll_agpBuf, ll_pciBuf, ll_timeStamp, ll_llBuf} 
+    LLState;
+
+
+typedef struct {
+    drm_via_mem_t mem;
+    unsigned offset;
+    unsigned stride;
+    unsigned height;
+} LowLevelBuffer;
+
+struct _XvMCLowLevel;
+
+typedef struct _ViaCommandBuffer {
+    CARD32 *buf;
+    CARD32 waitFlags;
+    unsigned pos;
+    unsigned bufSize;
+    int mode;
+    int header_start;
+    int rindex;
+    void (*flushFunc)(struct _ViaCommandBuffer *cb, struct _XvMCLowLevel *xl);
+} ViaCommandBuffer;
+    
+typedef struct _XvMCLowLevel{
+    ViaCommandBuffer agpBuf, pciBuf, *videoBuf;
+    int use_agp;   
+    int fd;
+    drm_context_t *drmcontext;
+    drmLockPtr hwLock;
+    drmAddress mmioAddress;
+    drmAddress fbAddress;
+    unsigned fbStride;
+    unsigned fbDepth;
+    unsigned width;
+    unsigned height;
+    int performLocking;
+    unsigned errors;
+    drm_via_mem_t tsMem;
+    CARD32 tsOffset;
+    volatile CARD32 *tsP;
+    CARD32 curTimeStamp;
+    CARD32 lastReadTimeStamp;
+    int agpSync;
+    CARD32 agpSyncTimeStamp;
+    unsigned chipId;
+
+    /*
+     * Data for video-engine less display
+     */
+
+    XvMCRegion sRegion;
+    XvMCRegion dRegion;
+    LowLevelBuffer scale;
+    LowLevelBuffer back;
+    Bool downScaling;
+    CARD32 downScaleW;
+    CARD32 downScaleH;
+    CARD32 upScaleW;
+    CARD32 upScaleH;
+    unsigned fetch;
+    unsigned line;
+    LLState state;
+}XvMCLowLevel;
+
+
+/*
+ * For Other architectures than i386 these might have to be modified for
+ * bigendian etc.
+ */
+
+
+#define MPEGIN(xl,reg)							\
+    *((volatile CARD32 *)(((CARD8 *)(xl)->mmioAddress) + 0xc00 + (reg)))
+
+#define VIDIN(ctx,reg)							\
+    *((volatile CARD32 *)(((CARD8 *)(ctx)->mmioAddress) + 0x200 + (reg)))
+
+#define REGIN(ctx,reg)							\
+    *((volatile CARD32 *)(((CARD8 *)(ctx)->mmioAddress) + 0x0000 + (reg)))
+
+#define HQV_CONTROL             0x1D0
+#define HQV_SRC_OFFSET          0x1CC
+#define HQV_SRC_STARTADDR_Y     0x1D4
+#define HQV_SRC_STARTADDR_U     0x1D8
+#define HQV_SRC_STARTADDR_V     0x1DC
+#define HQV_MINIFY_DEBLOCK      0x1E8    
+
+#define REG_HQV1_INDEX      0x00001000
+    
+#define HQV_SW_FLIP         0x00000010
+#define HQV_FLIP_STATUS     0x00000001
+#define HQV_SUBPIC_FLIP     0x00008000
+#define HQV_FLIP_ODD        0x00000020
+#define HQV_DEINTERLACE     0x00010000
+#define HQV_FIELD_2_FRAME   0x00020000
+#define HQV_FRAME_2_FIELD   0x00040000
+#define HQV_FIELD_UV        0x00100000
+#define HQV_DEBLOCK_HOR     0x00008000
+#define HQV_DEBLOCK_VER     0x80000000
+#define HQV_YUV420          0xC0000000
+#define HQV_YUV422          0x80000000
+#define HQV_ENABLE          0x08000000
+#define HQV_GEN_IRQ         0x00000080
+
+#define HQV_SCALE_ENABLE    0x00000800
+#define HQV_SCALE_DOWN      0x00001000
+
+#define V_COMPOSE_MODE          0x98
+#define V1_COMMAND_FIRE         0x80000000  
+#define V3_COMMAND_FIRE         0x40000000  
+
+/* SUBPICTURE Registers */
+#define SUBP_CONTROL_STRIDE     0x1C0
+#define SUBP_STARTADDR          0x1C4
+#define RAM_TABLE_CONTROL       0x1C8
+#define RAM_TABLE_READ          0x1CC
+
+/* SUBP_CONTROL_STRIDE              0x3c0 */
+#define SUBP_HQV_ENABLE             0x00010000
+#define SUBP_IA44                   0x00020000
+#define SUBP_AI44                   0x00000000
+#define SUBP_STRIDE_MASK            0x00001fff
+#define SUBP_CONTROL_MASK           0x00070000
+
+/* RAM_TABLE_CONTROL                0x3c8 */
+#define RAM_TABLE_RGB_ENABLE        0x00000007
+
+
+#define VIA_REG_STATUS          0x400
+#define VIA_REG_GEMODE          0x004
+#define VIA_REG_SRCBASE         0x030
+#define VIA_REG_DSTBASE         0x034
+#define VIA_REG_PITCH           0x038      
+#define VIA_REG_SRCCOLORKEY     0x01C      
+#define VIA_REG_KEYCONTROL      0x02C       
+#define VIA_REG_SRCPOS          0x008
+#define VIA_REG_DSTPOS          0x00C
+#define VIA_REG_GECMD           0x000
+#define VIA_REG_DIMENSION       0x010       /* width and height */
+#define VIA_REG_FGCOLOR         0x018
+
+
+#define VIA_VR_QUEUE_BUSY       0x00020000 /* Virtual Queue is busy */
+#define VIA_CMD_RGTR_BUSY       0x00000080  /* Command Regulator is busy */
+#define VIA_2D_ENG_BUSY         0x00000001  /* 2D Engine is busy */
+#define VIA_3D_ENG_BUSY         0x00000002  /* 3D Engine is busy */
+#define VIA_GEM_8bpp            0x00000000
+#define VIA_GEM_16bpp           0x00000100
+#define VIA_GEM_32bpp           0x00000300
+#define VIA_GEC_BLT             0x00000001
+#define VIA_PITCH_ENABLE        0x80000000
+#define VIA_GEC_INCX            0x00000000
+#define VIA_GEC_DECY            0x00004000
+#define VIA_GEC_INCY            0x00000000
+#define VIA_GEC_DECX            0x00008000
+#define VIA_GEC_FIXCOLOR_PAT    0x00002000
+
+
+#define VIA_BLIT_CLEAR 0x00
+#define VIA_BLIT_COPY 0xCC
+#define VIA_BLIT_FILL 0xF0
+#define VIA_BLIT_SET 0xFF
+
+#define VIA_SYNCWAITTIMEOUT 50000 /* Might be a bit conservative */
+#define VIA_DMAWAITTIMEOUT 150000
+#define VIA_VIDWAITTIMEOUT 50000
+#define VIA_XVMC_DECODERTIMEOUT 50000 /*(microseconds)*/
+
+#define VIA_AGP_HEADER5 0xFE040000
+#define VIA_AGP_HEADER6 0xFE050000
+
+typedef struct{
+    CARD32 data;
+    Bool set;
+} HQVRegister;
+
+
+#define H1_ADDR(val) (((val) >> 2) | 0xF0000000)
+#define WAITFLAGS(cb, flags)			\
+    (cb)->waitFlags |= (flags)
+#define BEGIN_RING_AGP(cb, xl, size)					\
+    do {								\
+	if ((cb)->pos > ((cb)->bufSize-(size))) {			\
+	    cb->flushFunc(cb, xl);					\
+	}								\
+    } while(0)
+#define OUT_RING_AGP(cb, val) do{			\
+	(cb)->buf[(cb)->pos++] = (val);	\
+  } while(0);		
+
+#define OUT_RING_QW_AGP(cb, val1, val2)			\
+    do {						\
+	(cb)->buf[(cb)->pos++] = (val1);	\
+	(cb)->buf[(cb)->pos++] = (val2);	\
+    } while (0)
+  
+
+#define BEGIN_HEADER5_AGP(cb, xl, index)	\
+    do {					\
+	BEGIN_RING_AGP(cb, xl, 8);		\
+	(cb)->mode = VIA_AGP_HEADER5;		\
+        (cb)->rindex = (index);			\
+	(cb)->header_start = (cb)->pos;		\
+	(cb)->pos += 4;				\
+    } while (0)
+
+#define BEGIN_HEADER6_AGP(cb, xl)		\
+    do {					\
+	BEGIN_RING_AGP(cb, xl, 8);		\
+	(cb)->mode = VIA_AGP_HEADER6;	\
+	(cb)->header_start = (cb)->pos; \
+	(cb)->pos += 4;			\
+    } while (0)
+
+#define BEGIN_HEADER5_DATA(cb, xl, size, index)				\
+    do {								\
+	if ((cb)->pos > ((cb)->bufSize - ((size) + 16))) {		\
+	    cb->flushFunc(cb, xl);					\
+	    BEGIN_HEADER5_AGP(cb, xl, index);				\
+	} else if ((cb)->mode && (((cb)->mode != VIA_AGP_HEADER5) ||	\
+				  ((cb)->rindex != index))) {		\
+	    finish_header_agp(cb);					\
+	    BEGIN_HEADER5_AGP((cb), xl, (index));			\
+	} else if (cb->mode != VIA_AGP_HEADER5) {			\
+	    BEGIN_HEADER5_AGP((cb), xl, (index)); 			\
+	}								\
+    }while(0)
+
+#define BEGIN_HEADER6_DATA(cb, xl, size)				\
+    do{									\
+	if ((cb)->pos > (cb->bufSize-(((size) << 1) + 16))) {		\
+	    cb->flushFunc(cb, xl);					\
+	    BEGIN_HEADER6_AGP(cb, xl);					\
+	} else	if ((cb)->mode && ((cb)->mode != VIA_AGP_HEADER6)) {	\
+	    finish_header_agp(cb);					\
+	    BEGIN_HEADER6_AGP(cb, xl);					\
+	}								\
+	else if ((cb->mode != VIA_AGP_HEADER6)) {			\
+	    BEGIN_HEADER6_AGP(cb, (xl));				\
+	}								\
+    }while(0)
+
+#define HQV_SHADOW_BASE 0x3CC
+#define HQV_SHADOW_SIZE 13
+
+#define SETHQVSHADOW(shadow, offset, value)				\
+    do {								\
+	HQVRegister *r = (shadow) + (((offset) - HQV_SHADOW_BASE) >> 2); \
+	r->data = (value);						\
+	r->set = TRUE;							\
+    } while(0)
+
+#define GETHQVSHADOW(shadow, offset)  ((shadow)[(offset - HQV_SHADOW_BASE) >> 2].data)
+
+#define LL_HW_LOCK(xl)							\
+    do {								\
+	DRM_LOCK((xl)->fd,(xl)->hwLock,*(xl)->drmcontext,0);		\
+    } while(0);
+#define LL_HW_UNLOCK(xl)					\
+    do {							\
+	DRM_UNLOCK((xl)->fd,(xl)->hwLock,*(xl)->drmcontext);	\
+    } while(0);
+
+static HQVRegister hqvShadow[HQV_SHADOW_SIZE];
+
+
+static void 
+initHQVShadow(HQVRegister *r)
+{
+    int i;
+    
+    for(i=0; i<HQV_SHADOW_SIZE; ++i) {
+	r->data = 0;
+	r++->set = FALSE;
+    }
+}
+
+static void
+setHQVDeblocking(HQVRegister *shadow, Bool on, Bool lowPass) 
+{
+    CARD32 tmp = GETHQVSHADOW(shadow, 0x3DC);
+    
+    if (!on) {
+	tmp &= ~(1 << 27);
+	SETHQVSHADOW(shadow, 0x3DC , tmp);
+	return;
+    }
+
+    tmp |= (8 << 16) | (1 << 27);
+    if (lowPass) 
+	tmp |= (1 << 26);
+    SETHQVSHADOW(shadow, 0x3DC , tmp);
+
+    tmp = GETHQVSHADOW(shadow, 0x3D4);
+    tmp |= (6 << 27);
+    SETHQVSHADOW(shadow, 0x3D4, tmp);
+
+    tmp = GETHQVSHADOW(shadow, 0x3D8);
+    tmp |= (19 << 27);
+    SETHQVSHADOW(shadow, 0x3D8, tmp);
+} 
+ 
+static void 
+setHQVStartAddress(HQVRegister *shadow, unsigned yOffs, unsigned uOffs,
+		   unsigned stride, unsigned format)
+{
+    CARD32 tmp = GETHQVSHADOW(shadow, 0x3D4);
+    
+    tmp |= yOffs & 0x03FFFFF0;
+    SETHQVSHADOW(shadow, 0x3D4, tmp);
+    tmp = GETHQVSHADOW(shadow, 0x3D8);
+    tmp |= uOffs & 0x03FFFFF0;
+    SETHQVSHADOW(shadow, 0x3D8, tmp);
+    tmp = GETHQVSHADOW(shadow, 0x3F8);
+    tmp |= (stride & 0x1FF8);
+    SETHQVSHADOW(shadow, 0x3F8, tmp);
+    tmp = GETHQVSHADOW(shadow, 0x3D0);
+
+    if (format == 0) {
+	/*
+	 * NV12
+	 */ 
+	tmp |= (0x0C << 28);
+    } else if (format == 1) {
+	/*
+	 * RGB16
+	 */
+	tmp |= (0x02 << 28);
+    } else if (format == 2) {
+	/*
+	 * RGB32
+	 */
+	;
+    }
+    SETHQVSHADOW(shadow, 0x3D0, tmp);
+}
+
+static void
+setHQVDeinterlacing(HQVRegister *shadow, CARD32 frameType) 
+{
+    CARD32 tmp = GETHQVSHADOW(shadow, 0x3D0);
+ 
+
+    if ((frameType & XVMC_FRAME_PICTURE) == XVMC_TOP_FIELD) {
+	tmp |= HQV_FIELD_UV   |
+	    HQV_DEINTERLACE   |
+	    HQV_FIELD_2_FRAME |
+	    HQV_FRAME_2_FIELD;
+    } else if ((frameType & XVMC_FRAME_PICTURE) == XVMC_BOTTOM_FIELD) {
+	tmp |= HQV_FIELD_UV   |
+	    HQV_DEINTERLACE   |
+	    HQV_FIELD_2_FRAME |
+	    HQV_FRAME_2_FIELD |
+	    HQV_FLIP_ODD;
+    } 
+    SETHQVSHADOW(shadow, 0x3D0, tmp);
+}
+
+static void
+setHQVTripleBuffer(HQVRegister *shadow, Bool on) 
+{
+    CARD32 tmp = GETHQVSHADOW(shadow, 0x3D0);
+
+    if (on)
+	tmp |= ( 1 << 26 );
+    else 
+	tmp &= ~( 1 << 26 );
+    SETHQVSHADOW(shadow, 0x3D0, tmp);
+}
+
+static void
+finish_header_agp(ViaCommandBuffer *cb) 
+{
+    int
+      numDWords,i;
+    CARD32
+	*hb;
+
+    if (!cb->mode) return;
+    numDWords = cb->pos - cb->header_start - 4; 
+    hb = cb->buf + cb->header_start;
+    switch (cb->mode) {
+    case VIA_AGP_HEADER5:
+	hb[0] = VIA_AGP_HEADER5 | cb->rindex;
+	hb[1] = numDWords ;
+	hb[2] = 0x00F50000; /* SW debug flag. (?) */
+	break;
+    default:
+	hb[0] = VIA_AGP_HEADER6;
+	hb[1] = numDWords >> 1 ;
+	hb[2] = 0x00F60000; /* SW debug flag. (?) */
+	break;
+    }
+    hb[3] = 0;
+    if (numDWords & 3) {
+	for (i=0; i<(4 - (numDWords & 3)); ++i) 
+	    OUT_RING_AGP(cb, 0x00000000);
+    }
+    cb->mode = 0;
+}
+
+void 
+hwlLock(void *xlp, int videoLock) 
+{
+    XvMCLowLevel *xl = (XvMCLowLevel *) xlp;
+
+    LL_HW_LOCK(xl);
+}
+
+void 
+hwlUnlock(void *xlp, int videoLock) 
+{
+    XvMCLowLevel *xl = (XvMCLowLevel *) xlp;
+
+    LL_HW_UNLOCK(xl);
+}
+    
+static unsigned 
+timeDiff(struct timeval *now,struct timeval *then) {
+    return (now->tv_usec >= then->tv_usec) ?
+	now->tv_usec - then->tv_usec : 
+	1000000 - (then->tv_usec - now->tv_usec);
+}
+
+void 
+setAGPSyncLowLevel(void *xlp, int val, CARD32 timeStamp)
+{
+    XvMCLowLevel *xl = (XvMCLowLevel *) xlp;
+
+    xl->agpSync = val;
+    xl->agpSyncTimeStamp = timeStamp;
+}
+
+CARD32 
+viaDMATimeStampLowLevel(void *xlp)
+{
+    XvMCLowLevel *xl = (XvMCLowLevel *) xlp;
+
+    if (xl->use_agp) {
+	viaBlit(xl, 32, xl->tsOffset, 1, xl->tsOffset, 1, 1, 1, 0, 0, 
+		VIABLIT_FILL, xl->curTimeStamp);
+	return xl->curTimeStamp++;
+    }
+    return 0;
+}
+
+static void 
+viaDMAWaitTimeStamp(XvMCLowLevel *xl, CARD32 timeStamp, int doSleep) 
+{
+    struct timeval now, then;
+    struct timezone here;
+    struct timespec sleep, rem;
+
+    if (xl->use_agp && (xl->lastReadTimeStamp - timeStamp > (1 << 23))) {
+	sleep.tv_nsec = 1;
+	sleep.tv_sec = 0;
+	here.tz_minuteswest = 0;
+	here.tz_dsttime = 0;
+	gettimeofday(&then,&here);
+ 
+	while(((xl->lastReadTimeStamp = *xl->tsP) - timeStamp) > (1 << 23)) {
+	    gettimeofday(&now,&here); 
+	    if (timeDiff(&now,&then) > VIA_DMAWAITTIMEOUT) {
+		if(((xl->lastReadTimeStamp = *xl->tsP) - timeStamp) > (1 << 23)) {
+		    xl->errors |= LL_DMA_TIMEDOUT;
+		    break;
+		}
+	    }
+	    if (doSleep) nanosleep(&sleep, &rem);
+	}
+    }
+}
+
+static int 
+viaDMAInitTimeStamp(XvMCLowLevel *xl) 
+{
+    int ret = 0;
+
+    if (xl->use_agp) {
+	xl->tsMem.context = *(xl->drmcontext);
+	xl->tsMem.size = 64;
+	xl->tsMem.type = VIA_MEM_VIDEO;
+	if ((ret = drmCommandWriteRead(xl->fd, DRM_VIA_ALLOCMEM, 
+				       &xl->tsMem, sizeof(xl->tsMem))) < 0) 
+	    return ret;
+	if (xl->tsMem.size != 64)
+	    return -1;
+	xl->tsOffset = (xl->tsMem.offset + 31) & ~31; 
+	xl->tsP = (CARD32 *)xl->fbAddress + (xl->tsOffset >> 2); 
+	xl->curTimeStamp = 1;
+	*xl->tsP = 0;
+    }
+    return 0;
+}
+
+static int 
+viaDMACleanupTimeStamp(XvMCLowLevel *xl) 
+{
+
+    if (!(xl->tsMem.size) || !xl->use_agp) return 0;
+    return drmCommandWrite(xl->fd, DRM_VIA_FREEMEM, &xl->tsMem, sizeof(xl->tsMem)); 
+}
+
+
+static CARD32 
+viaMpegGetStatus(XvMCLowLevel *xl)
+{
+    return MPEGIN(xl,0x54);
+}
+
+static int 
+viaMpegIsBusy(XvMCLowLevel *xl, CARD32 mask, CARD32 idle)
+{
+    CARD32 tmp = viaMpegGetStatus(xl);
+
+    /*
+     * Error detected. 
+     * FIXME: Are errors really shown when error concealment is on?
+     */
+
+    if (tmp & 0x70) return 0; 
+
+    return (tmp & mask) != idle;
+}
+
+
+static void 
+syncDMA(XvMCLowLevel *xl, unsigned int doSleep)
+{
+  
+    /*
+     * Ideally, we'd like to have an interrupt wait here, but, according to second hand
+     * information, the hardware does not support this, although earlier S3 chips do that.
+     * It is therefore not implemented into the DRM, and we'll do a user space wait here.
+     */
+
+    struct timeval now, then;
+    struct timezone here;
+    struct timespec sleep, rem;
+
+    sleep.tv_nsec = 1;
+    sleep.tv_sec = 0;
+    here.tz_minuteswest = 0;
+    here.tz_dsttime = 0;
+    gettimeofday(&then,&here);
+    while( !(REGIN(xl, VIA_REG_STATUS) & VIA_VR_QUEUE_BUSY)) {
+	gettimeofday(&now,&here); 
+	if (timeDiff(&now,&then) > VIA_DMAWAITTIMEOUT) {
+	    if( !(REGIN(xl, VIA_REG_STATUS) & VIA_VR_QUEUE_BUSY)) {
+		xl->errors |= LL_DMA_TIMEDOUT;
+		break;
+	    }
+	}
+	if (doSleep) nanosleep(&sleep, &rem);
+    }
+    while( REGIN(xl, VIA_REG_STATUS) & VIA_CMD_RGTR_BUSY ) {
+	gettimeofday(&now,&here); 
+	if (timeDiff(&now,&then) > VIA_DMAWAITTIMEOUT) {
+	    if( REGIN(xl, VIA_REG_STATUS) & VIA_CMD_RGTR_BUSY ) {
+		xl->errors |= LL_DMA_TIMEDOUT;
+		break;
+	    }
+	}
+	if (doSleep) nanosleep(&sleep, &rem);
+    }
+}
+
+#ifdef HQV_USE_IRQ
+static void 
+syncVideo(XvMCLowLevel *xl, unsigned int doSleep)
+{
+    int proReg = REG_HQV1_INDEX;
+
+    /*
+     * Wait for HQV completion using completion interrupt. Nothing strange here. 
+     * Note that the interrupt handler clears the HQV_FLIP_STATUS bit, so we 
+     * can't wait on that one.
+     */
+    
+    if ((VIDIN(xl, HQV_CONTROL|proReg) & (HQV_SW_FLIP | HQV_SUBPIC_FLIP))) {
+	drm_via_irqwait_t irqw;
+	irqw.request.irq = 1;
+	irqw.request.type = VIA_IRQ_ABSOLUTE;
+	if (drmCommandWriteRead(xl->fd, DRM_VIA_WAIT_IRQ, &irqw, sizeof(irqw)) < 0) 
+	    xl->errors |= LL_VIDEO_TIMEDOUT;
+    }
+}
+#else
+static void 
+syncVideo(XvMCLowLevel *xl, unsigned int doSleep)
+{
+    /*
+     * Wait for HQV completion. Nothing strange here. We assume that the HQV
+     * Handles syncing to the V1 / V3 engines by itself. It should be safe to
+     * always wait for SUBPIC_FLIP completion although subpictures are not always
+     * used. 
+     */
+    
+    struct timeval now, then;
+    struct timezone here;
+    struct timespec sleep, rem;
+
+    int proReg = REG_HQV1_INDEX;
+
+    sleep.tv_nsec = 1;
+    sleep.tv_sec = 0;
+    here.tz_minuteswest = 0;
+    here.tz_dsttime = 0;
+    gettimeofday(&then,&here); 
+    while((VIDIN(xl, HQV_CONTROL|proReg) & (HQV_SW_FLIP | HQV_SUBPIC_FLIP  )) ) {
+	gettimeofday(&now,&here); 
+	if (timeDiff(&now,&then) > VIA_SYNCWAITTIMEOUT) {
+	  if((VIDIN(xl, HQV_CONTROL|proReg) & (HQV_SW_FLIP | HQV_SUBPIC_FLIP )) ) {
+		xl->errors |= LL_VIDEO_TIMEDOUT;
+		break;
+	    }
+	}
+	if (doSleep) nanosleep(&sleep, &rem);
+    }
+}
+#endif
+
+static void 
+syncAccel(XvMCLowLevel *xl, unsigned int mode, unsigned int doSleep)
+{
+    struct timeval now, then;
+    struct timezone here;
+    struct timespec sleep, rem;
+    CARD32 mask = ((mode & LL_MODE_2D) ? VIA_2D_ENG_BUSY : 0) |
+	((mode & LL_MODE_3D) ? VIA_3D_ENG_BUSY : 0);
+
+    sleep.tv_nsec = 1;
+    sleep.tv_sec = 0;
+    here.tz_minuteswest = 0;
+    here.tz_dsttime = 0;
+    gettimeofday(&then,&here); 
+    while( REGIN(xl, VIA_REG_STATUS) & mask) {
+	gettimeofday(&now,&here); 
+	if (timeDiff(&now,&then) > VIA_SYNCWAITTIMEOUT) {
+	    if( REGIN(xl, VIA_REG_STATUS) & mask) {
+		xl->errors |= LL_ACCEL_TIMEDOUT;
+		break;
+	    }
+	}
+	if (doSleep) nanosleep(&sleep, &rem);
+    }
+}
+
+
+static void 
+syncMpeg(XvMCLowLevel *xl, unsigned int mode, unsigned int doSleep)
+{
+    /*
+     * Ideally, we'd like to have an interrupt wait here, but from information from VIA
+     * at least the MPEG completion interrupt is broken on the CLE266, which was
+     * discovered during validation of the chip.
+     */
+
+    struct timeval now, then;
+    struct timezone here;
+    struct timespec sleep, rem;
+    CARD32 busyMask = 0;
+    CARD32 idleVal = 0;
+    CARD32 ret;
+
+    sleep.tv_nsec = 1;
+    sleep.tv_sec = 0;
+    here.tz_minuteswest = 0;
+    here.tz_dsttime = 0;
+    gettimeofday(&then,&here); 
+    if (mode & LL_MODE_DECODER_SLICE) {
+	busyMask = VIA_SLICEBUSYMASK;
+	idleVal = VIA_SLICEIDLEVAL;
+    }
+    if (mode & LL_MODE_DECODER_IDLE) {
+	busyMask |= VIA_BUSYMASK;
+	idleVal = VIA_IDLEVAL;
+    }
+    while(viaMpegIsBusy(xl, busyMask, idleVal)) {
+	gettimeofday(&now,&here); 
+	if (timeDiff(&now,&then) > VIA_XVMC_DECODERTIMEOUT) {
+	    if (viaMpegIsBusy(xl, busyMask, idleVal)) {
+		xl->errors |= LL_DECODER_TIMEDOUT;
+	    }
+	    break;
+	}
+	if (doSleep) nanosleep(&sleep, &rem);
+    }
+  
+    ret = viaMpegGetStatus(xl);
+    if (ret & 0x70) {
+	xl->errors |= ((ret & 0x70) >> 3);
+    }
+    return;
+}
+
+static void 
+pciFlush(ViaCommandBuffer *cb, XvMCLowLevel *xl)
+{
+    int ret;
+    drm_via_cmdbuffer_t b;
+    unsigned mode = cb->waitFlags;
+  
+    finish_header_agp(cb);
+    b.buf = (char *)cb->buf;
+    b.size = cb->pos * sizeof(CARD32);
+    if (xl->performLocking) hwlLock(xl,0);
+    if (((mode == LL_MODE_VIDEO) && (xl->videoBuf == &xl->agpBuf)) ||
+	((mode != LL_MODE_VIDEO) && (mode != 0)))
+	syncDMA(xl, 0);
+    if ((mode & LL_MODE_2D) || (mode & LL_MODE_3D)) {
+      syncAccel(xl, mode, 0); 
+    }
+    if (mode & LL_MODE_VIDEO) {
+      syncVideo(xl, 1);
+    }
+    if (mode & (LL_MODE_DECODER_SLICE | LL_MODE_DECODER_IDLE)) {
+      syncMpeg(xl, mode, 0); 
+    }
+    ret = drmCommandWrite(xl->fd, DRM_VIA_PCICMD, &b, sizeof(b));
+    if (xl->performLocking) hwlUnlock(xl,0);
+    if (ret) {
+	xl->errors |= LL_PCI_COMMAND_ERR;
+    }
+    cb->pos = 0;
+    cb->waitFlags = 0;
+}
+  
+static void 
+agpFlush(ViaCommandBuffer *cb, XvMCLowLevel *xl)
+{
+    drm_via_cmdbuffer_t b;
+    int ret;
+    int i;
+
+    finish_header_agp(cb);
+    if (xl->use_agp) {
+	b.buf = (char *)cb->buf;
+	b.size = cb->pos * sizeof(CARD32);
+	if (xl->agpSync) {
+	    syncXvMCLowLevel(xl, LL_MODE_DECODER_IDLE, 1, xl->agpSyncTimeStamp);
+	    xl->agpSync = 0;
+	}
+	if (xl->performLocking) hwlLock(xl,0);
+	do {
+	    ret = drmCommandWrite(xl->fd, DRM_VIA_CMDBUFFER, &b, sizeof(b));
+	} while (-EAGAIN == ret);
+	if (xl->performLocking) hwlUnlock(xl,0);
+      
+	if (ret) {
+	    xl->errors |= LL_AGP_COMMAND_ERR;
+	    for(i=0; i<cb->pos; i+=2) {
+	      printf("0x%x, 0x%x\n", (unsigned) cb->buf[i], (unsigned) cb->buf[i+1]);
+	    }
+	    exit(-1);
+	} else {
+	    cb->pos = 0;
+	}
+	cb->waitFlags &= LL_MODE_VIDEO; /* FIXME: Check this! */
+    } else {
+	unsigned mode=cb->waitFlags;
+  
+	b.buf = (char *)cb->buf;
+	b.size = cb->pos * sizeof(CARD32);
+	if (xl->performLocking) hwlLock(xl,0);
+	if (((mode == LL_MODE_VIDEO) && (cb == &xl->agpBuf)) ||
+	    ((mode != LL_MODE_VIDEO) && (mode != 0)))
+	    syncDMA(xl, 0);
+	if ((mode & LL_MODE_2D) || (mode & LL_MODE_3D)) 
+	    syncAccel(xl, mode, 0);
+	if (mode & LL_MODE_VIDEO) 
+	    syncVideo(xl, 1);
+	if (mode & (LL_MODE_DECODER_SLICE | LL_MODE_DECODER_IDLE))
+	    syncMpeg(xl, mode, 0);
+	ret = drmCommandWrite(xl->fd, DRM_VIA_PCICMD, &b, sizeof(b));
+	if (xl->performLocking) hwlUnlock(xl,0);
+	if (ret) {
+	    xl->errors |= LL_PCI_COMMAND_ERR;
+	}
+	cb->pos = 0;
+	cb->waitFlags = 0;
+    }      
+}
+
+static void
+uploadHQVShadow(XvMCLowLevel *xl, unsigned offset, HQVRegister *shadow,
+		Bool flip)
+{
+    int i;
+    CARD32 tmp;
+    ViaCommandBuffer *cb = xl->videoBuf;
+
+    BEGIN_HEADER6_DATA(cb, xl, HQV_SHADOW_SIZE);
+    WAITFLAGS(cb, LL_MODE_VIDEO);
+
+    if (shadow[0].set)
+	OUT_RING_QW_AGP(cb, 0x3CC + offset, 0);
+    
+    for (i=2; i < HQV_SHADOW_SIZE; ++i) {
+      if (shadow[i].set) {
+	  OUT_RING_QW_AGP(cb, offset + HQV_SHADOW_BASE + ( i << 2) , shadow[i].data);  
+	  shadow[i].set = FALSE;
+      }
+    }
+
+    /*
+     * Finally the control register for flip.
+     */
+    
+    if (flip) {
+	tmp = GETHQVSHADOW( shadow, 0x3D0);
+        OUT_RING_QW_AGP(cb, offset + HQV_CONTROL + 0x200 , 
+			HQV_ENABLE | HQV_GEN_IRQ | HQV_SUBPIC_FLIP | HQV_SW_FLIP | tmp); 
+    }
+    shadow[0].set = FALSE;
+    shadow[1].set = FALSE;
+}
+    
+
+
+
+unsigned 
+flushXvMCLowLevel(void *xlp) 
+{
+    unsigned 
+	errors;
+    XvMCLowLevel *xl = (XvMCLowLevel *) xlp;
+
+
+    if(xl->pciBuf.pos) pciFlush(&xl->pciBuf, xl);
+    if(xl->agpBuf.pos) agpFlush(&xl->agpBuf, xl);
+    errors = xl->errors;
+    if (errors) printf("Error 0x%x\n", errors);
+    xl->errors = 0;
+    return errors;
+}
+
+void 
+flushPCIXvMCLowLevel(void *xlp) 
+{
+    XvMCLowLevel *xl = (XvMCLowLevel *) xlp;
+
+
+    if(xl->pciBuf.pos) pciFlush(&xl->pciBuf, xl); 
+    if ((!xl->use_agp &&  xl->agpBuf.pos)) agpFlush(&xl->agpBuf, xl);
+}
+
+
+void 
+viaMpegSetSurfaceStride(void *xlp, ViaXvMCContext *ctx)
+{
+    CARD32 y_stride = ctx->yStride;
+    CARD32 uv_stride = y_stride >> 1;
+    XvMCLowLevel *xl = (XvMCLowLevel *) xlp;    
+    ViaCommandBuffer *cb = &xl->agpBuf;
+
+    BEGIN_HEADER6_DATA(cb, xl, 1);
+    OUT_RING_QW_AGP(cb, 0xc50, (y_stride >> 3) | ((uv_stride >> 3) << 16)); 
+    WAITFLAGS(cb, LL_MODE_DECODER_IDLE);
+}
+
+
+void 
+viaVideoSetSWFLipLocked(void *xlp, unsigned yOffs, unsigned uOffs,
+			unsigned vOffs, unsigned yStride, unsigned uvStride) 
+{
+    XvMCLowLevel *xl = (XvMCLowLevel *) xlp;
+
+    initHQVShadow(hqvShadow);
+    setHQVStartAddress(hqvShadow, yOffs, vOffs, yStride, 0);
+    if (xl->videoBuf == &xl->agpBuf)
+	syncDMA(xl, 1);
+    syncVideo(xl, 1);
+    uploadHQVShadow(xl, REG_HQV1_INDEX, hqvShadow, FALSE);
+    xl->videoBuf->flushFunc(xl->videoBuf, xl);
+}
+    
+void 
+viaVideoSWFlipLocked(void *xlp, unsigned flags,
+		     Bool progressiveSequence)
+{
+    XvMCLowLevel *xl = (XvMCLowLevel *) xlp;
+
+    setHQVDeinterlacing(hqvShadow, flags);
+    setHQVDeblocking(hqvShadow,( (flags & XVMC_FRAME_PICTURE) == XVMC_FRAME_PICTURE), TRUE);
+    setHQVTripleBuffer(hqvShadow, TRUE);
+    if (xl->videoBuf == &xl->agpBuf)
+	syncDMA(xl, 1);
+    syncVideo(xl, 1);
+    uploadHQVShadow(xl, REG_HQV1_INDEX, hqvShadow, TRUE);    
+    xl->videoBuf->flushFunc(xl->videoBuf, xl);
+}
+
+	
+void 
+viaMpegSetFB(void *xlp,unsigned i,
+	     unsigned yOffs,
+	     unsigned uOffs,
+	     unsigned vOffs) 
+{
+    XvMCLowLevel *xl = (XvMCLowLevel *) xlp;
+    ViaCommandBuffer *cb = &xl->agpBuf;
+
+    i *= (4*2);
+    BEGIN_HEADER6_DATA(cb,xl, 2);
+    OUT_RING_QW_AGP(cb, 0xc28 + i, yOffs >> 3);
+    OUT_RING_QW_AGP(cb, 0xc2c + i, vOffs >> 3);
+
+    WAITFLAGS(cb, LL_MODE_DECODER_IDLE);
+}
+
+void 
+viaMpegBeginPicture(void *xlp,ViaXvMCContext *ctx,
+		    unsigned width,
+		    unsigned height,
+		    const XvMCMpegControl *control) {
+				  
+    unsigned j, mb_width, mb_height;
+    XvMCLowLevel *xl = (XvMCLowLevel *) xlp;
+    ViaCommandBuffer *cb = &xl->agpBuf;
+
+    mb_width = (width + 15) >> 4;
+
+    mb_height =
+	((control->mpeg_coding == XVMC_MPEG_2) && 
+	 (control->flags & XVMC_PROGRESSIVE_SEQUENCE)) ?
+	2*((height+31) >> 5) : (((height+15) >> 4));
+
+    BEGIN_HEADER6_DATA(cb,xl, 72);
+    WAITFLAGS(cb, LL_MODE_DECODER_IDLE);
+
+    OUT_RING_QW_AGP(cb, 0xc00,
+		    ((control->picture_structure & XVMC_FRAME_PICTURE) << 2) |
+		    ((control->picture_coding_type & 3) << 4) |
+		    ((control->flags & XVMC_ALTERNATE_SCAN) ? (1 << 6) : 0));
+   
+    if (!(ctx->intraLoaded)) {
+	OUT_RING_QW_AGP(cb, 0xc5c, 0);
+	for (j = 0; j < 64; j += 4) {
+            OUT_RING_QW_AGP(cb, 0xc60,  
+			    ctx->intra_quantiser_matrix[j] |
+			    (ctx->intra_quantiser_matrix[j+1] << 8) |
+			    (ctx->intra_quantiser_matrix[j+2] << 16) |
+			    (ctx->intra_quantiser_matrix[j+3] << 24));
+        }
+	ctx->intraLoaded = 1;
+    }
+
+    if (!(ctx->nonIntraLoaded)) {
+	OUT_RING_QW_AGP(cb, 0xc5c, 1);
+	for (j = 0; j < 64; j += 4) {
+            OUT_RING_QW_AGP(cb, 0xc60,  
+			    ctx->non_intra_quantiser_matrix[j] |
+			    (ctx->non_intra_quantiser_matrix[j+1] << 8) |
+			    (ctx->non_intra_quantiser_matrix[j+2] << 16) |
+			    (ctx->non_intra_quantiser_matrix[j+3] << 24));
+        }
+	ctx->nonIntraLoaded = 1;
+    }
+
+    if (!(ctx->chromaIntraLoaded)) {
+	OUT_RING_QW_AGP(cb, 0xc5c, 2);
+	for (j = 0; j < 64; j += 4) {
+            OUT_RING_QW_AGP(cb, 0xc60,  
+			    ctx->chroma_intra_quantiser_matrix[j] |
+			    (ctx->chroma_intra_quantiser_matrix[j+1] << 8) |
+			    (ctx->chroma_intra_quantiser_matrix[j+2] << 16) |
+			    (ctx->chroma_intra_quantiser_matrix[j+3] << 24));
+        }
+	ctx->chromaIntraLoaded = 1;
+    }
+
+    if (!(ctx->chromaNonIntraLoaded)) {
+	OUT_RING_QW_AGP(cb, 0xc5c, 3);
+	for (j = 0; j < 64; j += 4) {
+            OUT_RING_QW_AGP(cb, 0xc60,  
+			    ctx->chroma_non_intra_quantiser_matrix[j] |
+			    (ctx->chroma_non_intra_quantiser_matrix[j+1] << 8) |
+			    (ctx->chroma_non_intra_quantiser_matrix[j+2] << 16) |
+			    (ctx->chroma_non_intra_quantiser_matrix[j+3] << 24));
+        }
+	ctx->chromaNonIntraLoaded = 1;
+    }
+    
+    OUT_RING_QW_AGP(cb, 0xc90,  
+		    ((mb_width * mb_height) & 0x3fff) |
+		    ((control->flags & XVMC_PRED_DCT_FRAME) ? ( 1 << 14)  : 0) |
+		    ((control->flags & XVMC_TOP_FIELD_FIRST) ? (1 << 15) : 0 )  |
+		    ((control->mpeg_coding == XVMC_MPEG_2) ? (1 << 16) : 0) |
+		    ((mb_width & 0xff) << 18));
+    	
+    OUT_RING_QW_AGP(cb, 0xc94,
+		    ((control->flags & XVMC_CONCEALMENT_MOTION_VECTORS) ? 1 : 0) |
+		    ((control->flags & XVMC_Q_SCALE_TYPE) ? 2 : 0) |
+		    ((control->intra_dc_precision & 3) << 2) |
+		    (((1 + 0x100000 / mb_width) & 0xfffff) << 4) |
+		    ((control->flags & XVMC_INTRA_VLC_FORMAT) ? (1 << 24) : 0));
+
+    OUT_RING_QW_AGP(cb, 0xc98, 
+		    (((control->FHMV_range) & 0xf) << 0) |
+		    (((control->FVMV_range) & 0xf) << 4) |
+		    (((control->BHMV_range) & 0xf) << 8) |
+		    (((control->BVMV_range) & 0xf) << 12) |
+		    ((control->flags & XVMC_SECOND_FIELD) ?  (1 << 20) : 0) | 
+		    (0x0a6 << 16));
+
+}
+
+
+
+void 
+viaMpegReset(void *xlp)
+{
+    int i,j;
+    XvMCLowLevel *xl = (XvMCLowLevel *) xlp;
+    ViaCommandBuffer *cb = &xl->agpBuf;
+
+    BEGIN_HEADER6_DATA(cb,xl, 99);
+    WAITFLAGS(cb, LL_MODE_DECODER_IDLE);
+
+    OUT_RING_QW_AGP(cb, 0xcf0 ,0);
+
+    for (i = 0; i < 6; i++) {
+        OUT_RING_QW_AGP(cb, 0xcc0 ,0);
+        OUT_RING_QW_AGP(cb, 0xc0c, 0x43|0x20 );
+        for (j = 0xc10; j < 0xc20; j += 4)
+	        OUT_RING_QW_AGP(cb, j, 0);
+    }
+
+    OUT_RING_QW_AGP(cb, 0xc0c, 0x1c3);
+    for (j = 0xc10; j < 0xc20; j += 4)
+	    OUT_RING_QW_AGP(cb,j,0);
+
+    for (i = 0; i < 19; i++)
+        OUT_RING_QW_AGP(cb, 0xc08 ,0);
+    
+    OUT_RING_QW_AGP(cb, 0xc98, 0x400000);
+
+    for (i = 0; i < 6; i++) {
+    OUT_RING_QW_AGP(cb, 0xcc0 ,0);
+    OUT_RING_QW_AGP(cb, 0xc0c, 0x1c3|0x20);
+    for (j = 0xc10; j < 0xc20; j += 4)
+	    OUT_RING_QW_AGP(cb,j,0);
+    }
+    OUT_RING_QW_AGP(cb, 0xcf0 ,0);
+
+}
+
+void 
+viaMpegWriteSlice(void *xlp, CARD8* slice, int nBytes, CARD32 sCode)
+{
+    int i, n, r;
+    CARD32* buf;
+    int count;
+    XvMCLowLevel *xl = (XvMCLowLevel *) xlp;
+    ViaCommandBuffer *cb = &xl->agpBuf;
+
+    if (xl->errors & (LL_DECODER_TIMEDOUT |
+		      LL_IDCT_FIFO_ERROR  |
+		      LL_SLICE_FIFO_ERROR |
+		      LL_SLICE_FAULT)) return;
+
+    n = nBytes >> 2;
+    if (sCode) nBytes += 4;
+    r = nBytes & 3;
+    buf = (CARD32*) slice;
+
+    if (r) nBytes += 4 - r;
+
+    nBytes += 8;
+
+    BEGIN_HEADER6_DATA(cb,xl, 2);
+    WAITFLAGS(cb, LL_MODE_DECODER_IDLE);
+    OUT_RING_QW_AGP(cb, 0xc9c, nBytes);
+    
+    if (sCode) OUT_RING_QW_AGP(cb, 0xca0, sCode);
+
+    i = 0;
+    count = 0;
+
+    do {
+	count += (LL_AGP_CMDBUF_SIZE -20);
+	count = (count > n) ? n : count;
+	BEGIN_HEADER5_DATA(cb, xl, (count - i), 0xca0);
+
+	for (; i < count; i++) {
+	    OUT_RING_AGP(cb, *buf++);
+	}
+	finish_header_agp(cb);
+    } while (i < n);
+
+    BEGIN_HEADER5_DATA(cb, xl, 3, 0xca0);
+
+    if (r) {
+	OUT_RING_AGP(cb, *buf & ((1 << (r << 3)) - 1));
+    }
+    OUT_RING_AGP(cb,0);
+    OUT_RING_AGP(cb,0);
+    finish_header_agp(cb);
+}
+
+void 
+viaVideoSubPictureOffLocked(void *xlp) {
+
+    CARD32 stride;
+    int proReg = REG_HQV1_INDEX;
+    XvMCLowLevel *xl = (XvMCLowLevel *) xlp;
+    ViaCommandBuffer *cb = xl->videoBuf;
+
+    if (xl->videoBuf == &xl->agpBuf)
+	syncDMA(xl, 1);
+    stride = VIDIN(xl,proReg|SUBP_CONTROL_STRIDE);
+    WAITFLAGS(cb, LL_MODE_VIDEO);
+    BEGIN_HEADER6_DATA(cb, xl, 1);
+    OUT_RING_QW_AGP(cb, proReg|SUBP_CONTROL_STRIDE | 0x200, stride & ~SUBP_HQV_ENABLE);
+}
+
+void 
+viaVideoSubPictureLocked(void *xlp, ViaXvMCSubPicture *pViaSubPic) {
+
+    unsigned i;
+    CARD32 cWord;    
+    XvMCLowLevel *xl = (XvMCLowLevel *) xlp;
+    int proReg = REG_HQV1_INDEX;    
+    ViaCommandBuffer *cb = xl->videoBuf;
+
+    if (xl->videoBuf == &xl->agpBuf)
+	syncDMA(xl, 1);
+    WAITFLAGS(cb, LL_MODE_VIDEO);
+    BEGIN_HEADER6_DATA(cb, xl, VIA_SUBPIC_PALETTE_SIZE + 2);
+    for (i=0; i<VIA_SUBPIC_PALETTE_SIZE; ++i) {
+	OUT_RING_QW_AGP(cb, proReg|RAM_TABLE_CONTROL | 0x200, pViaSubPic->palette[i]);
+    }
+
+    cWord = (pViaSubPic->stride & SUBP_STRIDE_MASK) | SUBP_HQV_ENABLE;
+    cWord |= (pViaSubPic->ia44) ? SUBP_IA44 : SUBP_AI44;
+    OUT_RING_QW_AGP(cb, proReg|SUBP_STARTADDR | 0x200, pViaSubPic->offset);
+    OUT_RING_QW_AGP(cb, proReg|SUBP_CONTROL_STRIDE | 0x200, cWord);
+}
+
+void 
+viaBlit(void *xlp,unsigned bpp,unsigned srcBase,
+	unsigned srcPitch,unsigned dstBase,unsigned dstPitch,
+	unsigned w,unsigned h,int xdir,int ydir, unsigned blitMode, 
+	unsigned color) 
+{
+
+    CARD32 dwGEMode = 0, srcY=0, srcX, dstY=0, dstX;
+    CARD32 cmd;
+    XvMCLowLevel *xl = (XvMCLowLevel *) xlp;
+    ViaCommandBuffer *cb = &xl->agpBuf;
+
+    if (!w || !h)
+        return;
+
+    finish_header_agp(cb);
+
+    switch (bpp) {
+    case 16:
+        dwGEMode |= VIA_GEM_16bpp;
+        break;
+    case 32:
+        dwGEMode |= VIA_GEM_32bpp;
+	break;
+    default:
+        dwGEMode |= VIA_GEM_8bpp;
+        break;
+    }
+
+    srcX = srcBase & 31;
+    dstX = dstBase & 31;
+    switch (bpp) {
+    case 16:
+        dwGEMode |= VIA_GEM_16bpp;
+	srcX >>= 2;
+	dstX >>= 2;
+        break;
+    case 32:
+        dwGEMode |= VIA_GEM_32bpp;
+	srcX >>= 4;
+	dstX >>= 4;
+	break;
+    default:
+        dwGEMode |= VIA_GEM_8bpp;
+        break;
+    }
+
+    BEGIN_RING_AGP(cb, xl, 20);
+    WAITFLAGS(cb, LL_MODE_2D);
+
+
+    OUT_RING_QW_AGP(cb, H1_ADDR(VIA_REG_GEMODE), dwGEMode);
+    cmd = 0; 
+
+    if (xdir < 0) {
+        cmd |= VIA_GEC_DECX;
+        srcX += (w - 1);
+        dstX += (w - 1);
+    }
+    if (ydir < 0) {
+        cmd |= VIA_GEC_DECY;
+        srcY += (h - 1);
+        dstY += (h - 1);
+    }
+
+    switch(blitMode) {
+    case VIABLIT_TRANSCOPY:
+	OUT_RING_QW_AGP(cb, H1_ADDR(VIA_REG_SRCCOLORKEY), color);
+	OUT_RING_QW_AGP(cb, H1_ADDR(VIA_REG_KEYCONTROL), 0x4000);
+	cmd |= VIA_GEC_BLT | (VIA_BLIT_COPY << 24);
+	break;
+    case VIABLIT_FILL:
+	OUT_RING_QW_AGP(cb, H1_ADDR(VIA_REG_FGCOLOR), color);
+	cmd |= VIA_GEC_BLT | VIA_GEC_FIXCOLOR_PAT | (VIA_BLIT_FILL << 24);
+	break;
+    default:
+	OUT_RING_QW_AGP(cb, H1_ADDR(VIA_REG_KEYCONTROL), 0x0);
+	cmd |= VIA_GEC_BLT | (VIA_BLIT_COPY << 24);
+    }	
+
+    OUT_RING_QW_AGP(cb, H1_ADDR(VIA_REG_SRCBASE), (srcBase & ~31) >> 3);
+    OUT_RING_QW_AGP(cb, H1_ADDR(VIA_REG_DSTBASE), (dstBase & ~31) >> 3);
+    OUT_RING_QW_AGP(cb, H1_ADDR(VIA_REG_PITCH), VIA_PITCH_ENABLE |
+		    (srcPitch >> 3) | (((dstPitch) >> 3) << 16));
+    OUT_RING_QW_AGP(cb, H1_ADDR(VIA_REG_SRCPOS), ((srcY << 16) | srcX));
+    OUT_RING_QW_AGP(cb, H1_ADDR(VIA_REG_DSTPOS), ((dstY << 16) | dstX));
+    OUT_RING_QW_AGP(cb, H1_ADDR(VIA_REG_DIMENSION), (((h - 1) << 16) | (w - 1)));
+    OUT_RING_QW_AGP(cb, H1_ADDR(VIA_REG_GECMD), cmd);
+}
+
+unsigned 
+syncXvMCLowLevel(void *xlp, unsigned int mode, unsigned int doSleep,
+		 CARD32 timeStamp)
+{
+    unsigned
+	errors;
+    XvMCLowLevel *xl = (XvMCLowLevel *) xlp;
+
+    if (mode == 0) {
+	errors = xl->errors;
+	xl->errors = 0;
+	return errors;
+    }
+
+    if ((mode & (LL_MODE_VIDEO | LL_MODE_3D)) || !xl->use_agp) {
+	if (xl->performLocking) 
+	    hwlLock(xl,0);	
+	if ((xl->videoBuf == &xl->agpBuf) ||  (mode != LL_MODE_VIDEO))
+	    syncDMA(xl, doSleep);
+	if (mode & LL_MODE_3D) 
+	    syncAccel(xl, mode, doSleep);
+	if (mode & LL_MODE_VIDEO)
+	    syncVideo(xl, doSleep);
+	if (xl->performLocking) 
+	    hwlUnlock(xl,0);
+    } else {
+	viaDMAWaitTimeStamp(xl, timeStamp, doSleep);
+    }
+
+    if (mode & (LL_MODE_DECODER_SLICE | LL_MODE_DECODER_IDLE)) 
+	syncMpeg(xl, mode, doSleep);
+
+    errors = xl->errors;
+    xl->errors = 0;
+
+    return errors;
+}
+
+static int
+updateLowLevelBuf(XvMCLowLevel *xl, LowLevelBuffer *buf, 
+		  unsigned width, unsigned height)
+{
+    unsigned
+	stride, size;
+    drm_via_mem_t *mem = &buf->mem;
+    int ret;
+
+    stride = (width + 31) & ~31;
+    size = stride * height + (xl->fbDepth >> 3);
+    
+    if (size != mem->size) {
+	if (mem->size) 
+	    drmCommandWrite(xl->fd, DRM_VIA_FREEMEM, mem, sizeof(*mem)); 
+	mem->context = *(xl->drmcontext);
+	mem->size = size;
+	mem->type = VIA_MEM_VIDEO;
+	
+	if (((ret = drmCommandWriteRead(xl->fd, DRM_VIA_ALLOCMEM, mem, sizeof(mem))) < 0)  ||
+	    mem->size != size) {
+	    mem->size = 0;
+	    return -1;	
+	}    
+    }
+    
+    buf->offset = (mem->offset + 31) & ~31;
+    buf->stride = stride;
+    buf->height = height;
+    return 0;
+}
+
+static void
+cleanupLowLevelBuf(XvMCLowLevel *xl, LowLevelBuffer *buf)
+{
+    drm_via_mem_t *mem = &buf->mem;
+
+    if (mem->size) 
+	drmCommandWrite(xl->fd, DRM_VIA_FREEMEM, mem, sizeof(mem)); 
+    mem->size = 0;
+}
+    
+
+static void
+*releaseXvMCLowLevel(XvMCLowLevel *xl)
+{
+    switch(xl->state) {
+    case ll_llBuf:
+	cleanupLowLevelBuf(xl, &xl->scale);
+    case ll_timeStamp:
+	viaDMACleanupTimeStamp(xl);
+    case ll_pciBuf:
+	free(xl->pciBuf.buf);
+    case ll_agpBuf:
+	free(xl->agpBuf.buf);
+    case ll_init:
+	free(xl);
+    default:
+	;
+    }
+    return NULL;
+}
+	
+	
+void 
+*initXvMCLowLevel(int fd, drm_context_t *ctx,
+		  drmLockPtr hwLock, drmAddress mmioAddress, 
+		  drmAddress fbAddress, unsigned fbStride, unsigned fbDepth,
+		  unsigned width, unsigned height, int useAgp, unsigned chipId ) 
+{
+    XvMCLowLevel *xl = (XvMCLowLevel *)malloc(sizeof(XvMCLowLevel));
+
+    if (!xl) return NULL;
+    xl->state = ll_init;
+
+    xl->agpBuf.buf = (CARD32 *)malloc(LL_AGP_CMDBUF_SIZE * sizeof(CARD32));
+    if (!xl->agpBuf.buf) return releaseXvMCLowLevel(xl);
+    xl->state = ll_agpBuf;
+    xl->agpBuf.bufSize = LL_AGP_CMDBUF_SIZE;
+    xl->agpBuf.flushFunc = &agpFlush;
+    xl->agpBuf.pos = 0;
+    xl->agpBuf.mode = 0;
+    xl->agpBuf.waitFlags = 0;
+
+    xl->pciBuf.buf = (CARD32 *)malloc(LL_PCI_CMDBUF_SIZE * sizeof(CARD32));
+    if (!xl->pciBuf.buf) return releaseXvMCLowLevel(xl);
+    xl->state = ll_pciBuf;
+    xl->pciBuf.bufSize = LL_PCI_CMDBUF_SIZE;
+    xl->pciBuf.flushFunc = &pciFlush;
+    xl->pciBuf.pos = 0;
+    xl->pciBuf.mode = 0;
+    xl->pciBuf.waitFlags = 0;
+
+    xl->use_agp = useAgp;
+    xl->fd = fd;
+    xl->drmcontext = ctx;
+    xl->hwLock = hwLock;
+    xl->mmioAddress = mmioAddress;
+    xl->fbAddress = fbAddress;
+    xl->fbDepth = fbDepth;
+    xl->fbStride = fbStride;
+    xl->width = width;
+    xl->height = height;
+    xl->performLocking = 1;
+    xl->errors = 0;
+    xl->agpSync = 0;
+    xl->chipId = chipId;
+
+    if (viaDMAInitTimeStamp(xl)) 
+	return releaseXvMCLowLevel(xl);
+    xl->state = ll_timeStamp;
+    
+    xl->scale.mem.size = 0;
+    xl->back.mem.size = 0;
+
+    if (updateLowLevelBuf(xl, &xl->scale, width, height)) 
+	return releaseXvMCLowLevel(xl);
+    xl->state = ll_llBuf;
+
+#ifdef VIDEO_DMA
+    xl->videoBuf = &xl->agpBuf;
+#else
+    xl->videoBuf = &xl->pciBuf;
+#endif
+
+    return xl;
+}
+
+void 
+setLowLevelLocking(void *xlp, int performLocking)
+{
+    XvMCLowLevel *xl = (XvMCLowLevel *) xlp;
+    xl->performLocking = performLocking;
+}
+
+void 
+closeXvMCLowLevel(void *xlp) 
+{
+    XvMCLowLevel *xl = (XvMCLowLevel *) xlp;
+    releaseXvMCLowLevel(xl);
+}
diff --git a/nx-X11/lib/XvMC/hw/via/viaLowLevel.h b/nx-X11/lib/XvMC/hw/via/viaLowLevel.h
new file mode 100644
index 000000000..12e5fe373
--- /dev/null
+++ b/nx-X11/lib/XvMC/hw/via/viaLowLevel.h
@@ -0,0 +1,148 @@
+/*****************************************************************************
+ * VIA Unichrome XvMC extension client lib.
+ *
+ * Copyright (c) 2004 The Unichrome Project. All rights reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice shall be included in
+ * all copies or substantial portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
+ * AUTHOR(S) OR COPYRIGHT HOLDER(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+ * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING
+ * FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER
+ * DEALINGS IN THE SOFTWARE.
+ */
+
+
+/*
+ * Authors: Thomas Hellström 2004 - 2005.
+ */
+
+
+#ifndef VIA_LOWLEVEL_H                  
+#define VIA_LOWLEVEL_H
+
+
+/*
+ * The below define is cache size sensitive. Increasing the AGP buffer size
+ * will enable the library to do deeper pipelining, but will degrade the
+ * performance in the drm dma command verifier. 
+ */
+
+#define LL_AGP_CMDBUF_SIZE (4096*2)
+#define LL_PCI_CMDBUF_SIZE (4096)
+
+#define LL_MODE_DECODER_SLICE 0x01
+#define LL_MODE_DECODER_IDLE 0x02
+#define LL_MODE_VIDEO   0x04
+#define LL_MODE_2D      0x08
+#define LL_MODE_3D      0x10
+
+/*
+ * Errors
+ */
+
+#define LL_DECODER_TIMEDOUT 0x00000001
+#define LL_IDCT_FIFO_ERROR  0x00000002
+#define LL_SLICE_FIFO_ERROR 0x00000004
+#define LL_SLICE_FAULT      0x00000008
+#define LL_DMA_TIMEDOUT     0x00000010
+#define LL_VIDEO_TIMEDOUT   0x00000020
+#define LL_ACCEL_TIMEDOUT   0x00000040
+#define LL_PCI_COMMAND_ERR  0x00000080
+#define LL_AGP_COMMAND_ERR  0x00000100
+
+#define VIA_SLICEBUSYMASK        0x00000200
+#define VIA_BUSYMASK             0x00000207
+#define VIA_SLICEIDLEVAL         0x00000200
+#define VIA_IDLEVAL              0x00000204   
+
+#include "via_drm.h"
+#include "viaXvMCPriv.h"
+
+#define setRegion(xx,yy,ww,hh,region) \
+    do {			  \
+	(region).x = (xx);	  \
+	(region).y = (yy);	  \
+	(region).w = (ww);	  \
+	(region).h = (hh);	  \
+    } while(0)
+
+#define regionEqual(r1, r2)				\
+    ((r1).x == (r2).x &&				\
+     (r1).y == (r2).y &&				\
+     (r1).w == (r2).w &&				\
+     (r1).h == (r2).h)
+
+
+
+extern void
+*initXvMCLowLevel(int fd, drm_context_t *ctx,
+		 drmLockPtr hwLock, drmAddress mmioAddress, 
+		 drmAddress fbAddress, unsigned fbStride, unsigned fbDepth,
+		 unsigned width, unsigned height, int useAgp, unsigned chipId );
+
+extern void setLowLevelLocking(void *xlp, int perFormLocking);
+extern void closeXvMCLowLevel(void *xlp);
+extern void flushPCIXvMCLowLevel(void *xlp);
+extern CARD32 viaDMATimeStampLowLevel(void *xlp);
+extern void setAGPSyncLowLevel(void *xlp, int val, CARD32 timeStamp);
+
+
+/*
+ * These two functions also return and clear the current error status.
+ */
+
+extern unsigned flushXvMCLowLevel(void *xlp);
+extern unsigned syncXvMCLowLevel(void *xlp, unsigned int mode,
+				 unsigned int doSleep, CARD32 timeStamp);
+
+extern void hwlUnlock(void *xlp, int videoLock); 
+extern void hwlLock(void *xlp, int videoLock); 
+
+extern void viaVideoSetSWFLipLocked(void *xlp, unsigned yOffs, unsigned uOffs,
+				    unsigned vOffs, unsigned yStride, unsigned uvStride);
+
+extern void viaMpegReset(void *xlp);
+extern void viaMpegWriteSlice(void *xlp, CARD8* slice, 
+				    int nBytes, CARD32 sCode);
+extern void viaMpegSetSurfaceStride(void *xlp, ViaXvMCContext *ctx);
+extern void viaMpegSetFB(void *xlp,unsigned i, unsigned yOffs,
+			       unsigned uOffs, unsigned vOffs);
+extern void viaMpegBeginPicture(void *xlp, ViaXvMCContext *ctx,unsigned width,
+				unsigned height,const XvMCMpegControl *control);
+
+/*
+ * Low-level Video functions in viaLowLevel.c
+ */ 
+
+
+extern void viaBlit(void *xlp,unsigned bpp,unsigned srcBase,
+		    unsigned srcPitch,unsigned dstBase,unsigned dstPitch,
+		    unsigned w,unsigned h,int xdir,int ydir, 
+		    unsigned blitMode, unsigned color); 
+
+extern void viaVideoSWFlipLocked(void *xlp, unsigned flags,
+				 int progressiveSequence);
+
+extern void viaVideoSubPictureLocked(void *xlp,ViaXvMCSubPicture *pViaSubPic);
+extern void viaVideoSubPictureOffLocked(void *xlp);
+
+
+
+
+#define PCI_CHIP_VT3204         0x3108 /* K8M800 */
+#define PCI_CHIP_VT3259         0x3118 /* PM800/PM880/CN400 */
+#define PCI_CHIP_CLE3122        0x3122 /* CLE266 */
+#define PCI_CHIP_VT3205         0x7205 /* KM400 */
+
+#endif
diff --git a/nx-X11/lib/XvMC/hw/via/viaXvMC.c b/nx-X11/lib/XvMC/hw/via/viaXvMC.c
new file mode 100644
index 000000000..eebc87ea8
--- /dev/null
+++ b/nx-X11/lib/XvMC/hw/via/viaXvMC.c
@@ -0,0 +1,1964 @@
+/*****************************************************************************
+ * VIA Unichrome XvMC extension client lib.
+ *
+ * Copyright (c) 2004-2005 Thomas Hellström. All rights reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice shall be included in
+ * all copies or substantial portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
+ * AUTHOR(S) OR COPYRIGHT HOLDER(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+ * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING
+ * FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER
+ * DEALINGS IN THE SOFTWARE.
+ */
+
+
+/*
+ *Author: Thomas Hellström, 2004.
+ *Bugfixes by among others Pascal Brisset and Terry Barnaby.
+ *DRI protocol support by Thomas Hellström, 2005.
+ */
+
+#undef WAITPAUSE
+
+#include "viaXvMCPriv.h"
+#include "viaLowLevel.h"
+#include <stdio.h>
+#include <sys/ioctl.h>
+#include <sys/time.h>
+#include <time.h>
+#include <fourcc.h>
+#include <X11/extensions/Xv.h>
+#include <xf86drm.h>
+#include <pthread.h>
+#include <X11/extensions/vldXvMC.h>
+#include "xf86dri.h"
+#include "driDrawable.h"
+    
+#define SAREAPTR(ctx) ((ViaXvMCSAreaPriv *)			\
+		       (((CARD8 *)(ctx)->sAreaAddress) +	\
+			(ctx)->sAreaPrivOffset))
+
+
+
+static int error_base;
+static int event_base;
+static unsigned numContexts = 0;
+static int globalFD;
+static drmAddress sAreaAddress;
+static drmAddress fbAddress;
+static drmAddress mmioAddress;
+
+
+#define FOURCC_XVMC (('C' << 24) + ('M' << 16) + ('V' << 8) + 'X')
+
+#define ppthread_mutex_lock(arg)		\
+  {						\
+    pthread_mutex_lock(arg);			\
+  }						\
+
+#define ppthread_mutex_unlock(arg)		\
+  {						\
+    pthread_mutex_unlock(arg);			\
+  }						\
+
+static unsigned yOffs (ViaXvMCSurface *srf) 
+{
+    return srf->offsets[0];
+}
+
+static unsigned vOffs (ViaXvMCSurface *srf) 
+{
+    return srf->offsets[0] + srf->yStride * srf->height;
+}
+
+static unsigned uOffs (ViaXvMCSurface *srf) 
+{
+    return srf->offsets[0] + ( srf->yStride * srf->height) + 
+	(srf->yStride >> 1) * (srf->height >> 1);
+}
+
+
+static void defaultQMatrices(ViaXvMCContext *ctx) 
+{
+    int i;
+
+    static const char intra[64] = {
+	8, 16, 19, 22, 26, 27, 29, 34, 16, 16, 22, 24, 27, 29, 34, 37,
+        19, 22, 26, 27, 29, 34, 34, 38, 22, 22, 26, 27, 29, 34, 37, 40,
+        22, 26, 27, 29, 32, 35, 40, 48, 26, 27, 29, 32, 35, 40, 48, 58,
+        26, 27, 29, 34, 38, 46, 56, 69, 27, 29, 35, 38, 46, 56, 69, 83
+    };
+    
+    for( i=0; i<64; ++i) {
+	ctx->intra_quantiser_matrix[i] = intra[i];
+	ctx->non_intra_quantiser_matrix[i] = 16;
+    }
+    ctx->intraLoaded = 0;
+    ctx->nonIntraLoaded = 0;
+}
+
+
+static void releaseDecoder(ViaXvMCContext *ctx,int clearCtx) 
+{
+    volatile ViaXvMCSAreaPriv *sAPriv;
+
+    sAPriv = SAREAPTR(ctx);
+    UNICHROME_UNLOCK(ctx->fd, UNICHROME_LOCK_DECODER1, sAPriv, ctx->drmcontext);
+}
+
+
+static int grabDecoder(	ViaXvMCContext *ctx, int *hadLastLock) 
+{
+    volatile ViaXvMCSAreaPriv *sAPriv =  SAREAPTR(ctx);
+    int retFtx, lc;
+
+    /*
+     * Try to grab the decoder. If it is not available we will sleep until
+     * it becomes available or for a maximum of 20 ms. 
+     * Then try to grab it again, unless a timeout occured. If the decoder is
+     * available, the lock should be reasonably fast.
+     */
+
+    if (ctx->haveDecoder) {
+        flushXvMCLowLevel(ctx->xl);  /* Ignore errors here. */
+	
+	/*fprintf(stderr,"ViaXvMC: ERROR: Trying to re-lock decoder.\n"); */
+	*hadLastLock = 1;
+        return 0;
+    }
+    UNICHROME_LOCK(ctx->fd, UNICHROME_LOCK_DECODER1, sAPriv, ctx->drmcontext, lc,
+		   retFtx);
+    *hadLastLock = (ctx->drmcontext == lc);
+
+    return retFtx;
+}
+	    
+static void setupAttribDesc(Display *display, XvPortID port,
+			    const ViaXvMCAttrHolder *attrib,
+			    XvAttribute attribDesc[]) 
+{
+    XvAttribute *XvAttribs,*curAD;
+    int num;
+    unsigned i,j;
+
+    XLockDisplay(display);
+    XvAttribs = XvQueryPortAttributes(display, port, &num);
+    for(i=0; i<attrib->numAttr; ++i) {
+	curAD = attribDesc + i;
+	curAD->flags = 0;
+	curAD->min_value = 0;
+	curAD->max_value = 0;
+	curAD->name = NULL;
+	for(j=0; j<num; ++j) {
+	    if (attrib->attributes[i].attribute == 
+		XInternAtom(display,XvAttribs[j].name,TRUE)) {
+		*curAD = XvAttribs[j];
+		curAD->name = strdup(XvAttribs[j].name);
+		break;
+	    }
+	}
+    }
+    if (XvAttribs) XFree(XvAttribs);
+    XUnlockDisplay(display);
+
+}
+
+static void releaseAttribDesc(int numAttr, XvAttribute attribDesc[]) 
+{
+    int i;
+
+    for (i=0; i<numAttr; ++i) {
+	if (attribDesc[i].name)
+	    free(attribDesc[i].name);
+    }
+}
+    
+static Status releaseContextResources(Display *display, XvMCContext *context,
+				      int freePrivate, Status errType)
+{
+    ViaXvMCContext *pViaXvMC = (ViaXvMCContext *) context->privData;
+
+    switch(pViaXvMC->resources) {
+    case context_drawHash:
+        driDestroyHashContents( pViaXvMC->drawHash );
+	drmHashDestroy( pViaXvMC->drawHash );
+    case context_lowLevel:
+	closeXvMCLowLevel(pViaXvMC->xl);
+    case context_mutex:
+	pthread_mutex_destroy(&pViaXvMC->ctxMutex);
+    case context_drmContext:
+	XLockDisplay(display);
+	uniDRIDestroyContext(display, pViaXvMC->screen, pViaXvMC->id);
+	XUnlockDisplay(display);
+    case context_sAreaMap:
+	numContexts--;
+	if (numContexts == 0)
+	    drmUnmap(pViaXvMC->sAreaAddress,pViaXvMC->sAreaSize);
+    case context_fbMap:
+	if (numContexts == 0)
+	    drmUnmap(pViaXvMC->fbAddress,pViaXvMC->fbSize);
+    case context_mmioMap:
+	if (numContexts == 0)
+	    drmUnmap(pViaXvMC->mmioAddress,pViaXvMC->mmioSize);
+    case context_fd:
+	if (numContexts == 0) {
+	    if (pViaXvMC->fd >= 0)
+		drmClose(pViaXvMC->fd);
+	}
+	pViaXvMC->fd = -1;
+    case context_driConnection:
+	if (numContexts == 0) {
+	    XLockDisplay(display);
+	    uniDRICloseConnection(display, pViaXvMC->screen);
+	    XUnlockDisplay(display);
+	}
+    case context_context:
+	XLockDisplay(display);
+	_xvmc_destroy_context(display, context);
+	XUnlockDisplay(display);
+	if (!freePrivate) break;
+    default:
+	free(pViaXvMC);
+    }
+    return errType;
+}
+
+Status XvMCCreateContext(Display *display, XvPortID port,
+			 int surface_type_id, int width, int height, int flags,
+			 XvMCContext *context) 
+{  
+    ViaXvMCContext *pViaXvMC;
+    int priv_count;
+    uint *priv_data;
+    uint magic;
+    unsigned i;
+    Status ret;
+    int major, minor;
+    ViaXvMCCreateContextRec *tmpComm;
+    drmVersionPtr drmVer;
+    char *curBusID;
+    int isCapable;
+
+    /* 
+     * Verify Obvious things first 
+     */
+
+    if(context == NULL) {
+	return XvMCBadContext;
+    }
+
+    if(!(flags & XVMC_DIRECT)) {
+	fprintf(stderr,"Indirect Rendering not supported! Using Direct.\n");
+    }
+
+    /* 
+     *FIXME: Check $DISPLAY for legal values here 
+     */
+
+    context->surface_type_id = surface_type_id;
+    context->width = (unsigned short)((width + 15) & ~15);
+    context->height = (unsigned short)((height + 15) & ~15);
+    context->flags = flags;
+    context->port = port;
+
+    /* 
+     *  Width, Height, and flags are checked against surface_type_id
+     *  and port for validity inside the X server, no need to check
+     *  here.
+     */
+
+    /* Allocate private Context data */
+    context->privData = (void *)malloc(sizeof(ViaXvMCContext));
+    if(!context->privData) {
+	fprintf(stderr,"Unable to allocate resources for XvMC context.\n");
+	return BadAlloc;
+    }
+  
+    pViaXvMC = (ViaXvMCContext *)context->privData;
+    pViaXvMC->resources = context_none;
+
+    /* Verify the XvMC extension exists */
+
+    XLockDisplay(display);
+    if(! XvMCQueryExtension(display, &event_base,
+			    &error_base)) {
+	fprintf(stderr,"XvMC Extension is not available!\n");
+	free(pViaXvMC);
+	XUnlockDisplay(display);
+	return BadAlloc;
+    }
+
+    /* Verify XvMC version */
+    ret = XvMCQueryVersion(display, &major, &minor);
+    if(ret) {
+	fprintf(stderr,"XvMCQuery Version Failed, unable to determine "
+		"protocol version!\n");
+    }
+    XUnlockDisplay(display);
+
+    /* FIXME: Check Major and Minor here */
+
+    XLockDisplay(display);
+    if((ret = _xvmc_create_context(display, context, &priv_count, 
+				       &priv_data))) {
+      XUnlockDisplay(display);
+      fprintf(stderr,"Unable to create XvMC Context.\n");
+      return releaseContextResources(display, context, 1, BadAlloc);
+    }
+    XUnlockDisplay(display);
+
+    /*
+     * Check size and version of returned data.
+     */
+
+    tmpComm = (  ViaXvMCCreateContextRec *) priv_data;
+    if(priv_count != (sizeof(ViaXvMCCreateContextRec) >> 2)) {
+	fprintf(stderr,"_xvmc_create_context() returned incorrect "
+		"data size!\n");
+	fprintf(stderr,"\tExpected %d, got %d\n",
+		(int) (sizeof(ViaXvMCCreateContextRec) >> 2),
+		(int) priv_count);
+	XFree(priv_data);
+	return releaseContextResources(display, context, 1, BadAlloc);
+    }
+    pViaXvMC->resources = context_context;
+
+    if ((tmpComm->major != VIAXVMC_MAJOR) ||
+	(tmpComm->minor != VIAXVMC_MINOR)) {
+	fprintf(stderr,"Version mismatch between the X via driver\n"
+		"and the XvMC library. Cannot continue!\n");
+	XFree(priv_data);
+	return releaseContextResources(display, context, 1, BadAlloc);
+    }
+
+    pViaXvMC->ctxNo = tmpComm->ctxNo;
+    pViaXvMC->fbOffset = tmpComm->fbOffset;
+    pViaXvMC->fbSize = tmpComm->fbSize;
+    pViaXvMC->mmioOffset = tmpComm->mmioOffset;
+    pViaXvMC->mmioSize = tmpComm->mmioSize;
+    pViaXvMC->sAreaSize = tmpComm->sAreaSize;
+    pViaXvMC->sAreaPrivOffset = tmpComm->sAreaPrivOffset;
+    pViaXvMC->decoderOn = 0;
+    pViaXvMC->xvMCPort = tmpComm->xvmc_port;
+    pViaXvMC->useAGP = tmpComm->useAGP;
+    pViaXvMC->attrib = tmpComm->initAttrs;
+    pViaXvMC->screen = tmpComm->screen;
+    pViaXvMC->depth = tmpComm->depth;
+    pViaXvMC->stride = tmpComm->stride;
+    pViaXvMC->chipId = tmpComm->chipId;
+
+    /* 
+     * Must free the private data we were passed from X 
+     */
+
+    XFree(priv_data);
+
+    /*
+     * Check for direct rendering capable, establish DRI and DRM connections,
+     * map framebuffer, DRI shared area and read-only register areas. 
+     * Initial checking for drm has already been done by the server. 
+     * Only do this for the first context we create.
+     */ 
+
+    if (numContexts == 0) {
+	XLockDisplay(display);
+	ret = uniDRIQueryDirectRenderingCapable(display, pViaXvMC->screen, &isCapable);
+	if (!ret || !isCapable) {
+	    XUnlockDisplay(display);
+	    fprintf(stderr,"Direct Rendering is not available on this system!\n");
+	    return releaseContextResources(display, context, 1, BadAlloc);
+	}
+
+	if (!uniDRIOpenConnection(display, pViaXvMC->screen, &pViaXvMC->sAreaOffset,
+				   &curBusID)) {
+	    XUnlockDisplay(display);
+	    fprintf(stderr,"Could not open DRI connection to X server!\n");
+	    return releaseContextResources(display, context, 1, BadAlloc);
+	}	
+	XUnlockDisplay(display);
+
+	strncpy(pViaXvMC->busIdString,curBusID,20);
+	pViaXvMC->busIdString[20] = '\0';    
+	XFree(curBusID);
+
+	pViaXvMC->resources = context_driConnection;    
+
+	if((pViaXvMC->fd = drmOpen("via",pViaXvMC->busIdString)) < 0) {
+	    fprintf(stderr,"DRM Device for via could not be opened.\n");
+	    return releaseContextResources(display, context, 1, BadAlloc);
+	}
+	globalFD = pViaXvMC->fd;
+	pViaXvMC->resources = context_fd;
+
+	if (NULL == (drmVer = drmGetVersion(pViaXvMC->fd))) {
+	    fprintf(stderr, 
+		    "viaXvMC: Could not get drm version.");
+	    return releaseContextResources(display, context, 1, BadAlloc);
+	}
+	if (((drmVer->version_major != 2 ) || (drmVer->version_minor < 0))) {
+	    fprintf(stderr, 
+		    "viaXvMC: Kernel drm is not compatible with XvMC.\n"); 
+	    fprintf(stderr, 
+		    "viaXvMC: Kernel drm version: %d.%d.%d "
+		    "and I need at least version 2.0.0.\n"
+		    "Please update.\n",
+		    drmVer->version_major,drmVer->version_minor,
+		    drmVer->version_patchlevel); 
+	    drmFreeVersion(drmVer);
+	    return releaseContextResources(display, context, 1, BadAlloc);
+	} 
+	drmFreeVersion(drmVer);
+	drmGetMagic(pViaXvMC->fd,&magic);
+	
+	XLockDisplay(display);
+	if (!uniDRIAuthConnection(display, pViaXvMC->screen, magic)) {
+	    XUnlockDisplay(display);
+	    fprintf(stderr, "viaXvMC: X server did not allow DRI. Check permissions.\n");
+	    XFree(priv_data);
+	    return releaseContextResources(display, context, 1, BadAlloc);
+	}
+	XUnlockDisplay(display);
+
+	/* 
+	 * Map the register memory 
+	 */
+
+	if(drmMap(pViaXvMC->fd,pViaXvMC->mmioOffset,
+		  pViaXvMC->mmioSize,&mmioAddress) < 0) {
+	    fprintf(stderr,"Unable to map the display chip mmio registers.\n");
+	    return releaseContextResources(display, context, 1, BadAlloc);
+	}   
+	pViaXvMC->mmioAddress = mmioAddress;
+	pViaXvMC->resources = context_mmioMap;
+
+	/* 
+	 * Map Framebuffer memory 
+	 */
+
+	if(drmMap(pViaXvMC->fd,pViaXvMC->fbOffset,
+		  pViaXvMC->fbSize,&fbAddress) < 0) {
+	    fprintf(stderr,"Unable to map XvMC Framebuffer.\n");
+	    return releaseContextResources(display, context, 1, BadAlloc);
+	} 
+	pViaXvMC->fbAddress = fbAddress;
+	pViaXvMC->resources = context_fbMap;
+
+
+	/*
+	 * Map DRI Sarea.
+	 */
+
+	if(drmMap(pViaXvMC->fd,pViaXvMC->sAreaOffset,
+		  pViaXvMC->sAreaSize,&sAreaAddress) < 0) {
+	    fprintf(stderr,"Unable to map DRI SAREA.\n");
+	    return releaseContextResources(display, context, 1, BadAlloc);
+	} 
+    } else {
+	pViaXvMC->fd = globalFD;
+	pViaXvMC->mmioAddress = mmioAddress;
+	pViaXvMC->fbAddress = fbAddress;
+    }
+
+    pViaXvMC->sAreaAddress = sAreaAddress;
+    pViaXvMC->resources = context_sAreaMap;
+    numContexts++;
+
+    /*
+     * Find a matching visual. Important only for direct drawing to the visible
+     * frame-buffer.
+     */
+
+    XLockDisplay(display);
+    ret = XMatchVisualInfo(display, pViaXvMC->screen, 
+			   (pViaXvMC->depth == 32) ? 24 : pViaXvMC->depth, TrueColor, 
+			   &pViaXvMC->visualInfo);
+    XUnlockDisplay(display);
+    if (!ret) {
+	fprintf(stderr, "viaXvMC: Could not find a matching TrueColor visual.\n");
+	return releaseContextResources(display, context, 1, BadAlloc);
+    }	
+
+    if (!uniDRICreateContext(display, pViaXvMC->screen, pViaXvMC->visualInfo.visual, 
+			      &pViaXvMC->id, &pViaXvMC->drmcontext)) {
+	
+	fprintf(stderr, "viaXvMC: Could not create DRI context.\n");
+	return releaseContextResources(display, context, 1, BadAlloc);
+    }
+
+    pViaXvMC->resources = context_drmContext;
+
+    for (i=0; i<VIA_MAX_RENDSURF; ++i) {
+	pViaXvMC->rendSurf[i] = 0;
+    }
+    pViaXvMC->lastSrfDisplaying = ~0;
+    setupAttribDesc(display, port, &pViaXvMC->attrib, pViaXvMC->attribDesc);
+
+    pViaXvMC->hwLock = (drmLockPtr) pViaXvMC->sAreaAddress;
+    defaultQMatrices(pViaXvMC);
+    pViaXvMC->chromaIntraLoaded = 1;
+    pViaXvMC->chromaNonIntraLoaded = 1;
+    pViaXvMC->yStride = (width + 31) & ~31;
+    pViaXvMC->haveDecoder = 0;
+    pViaXvMC->attribChanged = 1;
+    pViaXvMC->haveXv = 0;
+    pViaXvMC->port = context->port;
+    pthread_mutex_init(&pViaXvMC->ctxMutex,NULL);
+    pViaXvMC->resources = context_mutex;
+    pViaXvMC->timeStamp = 0;
+    setRegion(0,0,-1,-1,pViaXvMC->sRegion);
+    setRegion(0,0,-1,-1,pViaXvMC->dRegion);
+
+    if (NULL == (pViaXvMC->xl = 
+		 initXvMCLowLevel(pViaXvMC->fd, &pViaXvMC->drmcontext,
+				  pViaXvMC->hwLock, pViaXvMC->mmioAddress, 
+				  pViaXvMC->fbAddress, pViaXvMC->stride, pViaXvMC->depth,
+				  context->width, context->height,
+				  pViaXvMC->useAGP, pViaXvMC->chipId))) {
+
+	fprintf(stderr,"ViaXvMC: Could not allocate timestamp blit area.\n");
+	return releaseContextResources(display, context, 1, BadAlloc);
+    }       
+    pViaXvMC->resources = context_lowLevel;
+    setAGPSyncLowLevel(pViaXvMC->xl, 1, 0);
+    
+    if (NULL == (pViaXvMC->drawHash = drmHashCreate())) {
+	fprintf(stderr,"ViaXvMC: Could not allocate drawable hash table.\n");
+	return releaseContextResources(display, context, 1, BadAlloc);
+    }   
+    pViaXvMC->resources = context_drawHash;
+
+
+    if (numContexts == 1) {
+	hwlLock(pViaXvMC->xl,1); 
+	setLowLevelLocking(pViaXvMC->xl,0);
+	viaVideoSubPictureOffLocked(pViaXvMC->xl);
+	flushXvMCLowLevel(pViaXvMC->xl);  
+	setLowLevelLocking(pViaXvMC->xl,1);
+	hwlUnlock(pViaXvMC->xl,1);
+    }
+
+    return Success;
+}
+
+
+Status XvMCDestroyContext(Display *display, XvMCContext *context) 
+{
+    ViaXvMCContext *pViaXvMC;
+
+
+    if(context == NULL) {
+	return (error_base + XvMCBadContext);
+    }
+    if(NULL == (pViaXvMC = context->privData)) {
+	return (error_base + XvMCBadContext);
+    }
+
+    /*
+     * Release decoder if we have it. In case of crash or termination
+     * before XvMCDestroyContext, the X server will take care of this.
+     */
+
+    releaseAttribDesc(pViaXvMC->attrib.numAttr,pViaXvMC->attribDesc);
+    releaseDecoder(pViaXvMC,1);
+    return releaseContextResources(display, context, 1, Success);
+}
+
+Status XvMCCreateSurface( Display *display, XvMCContext *context,
+			  XvMCSurface *surface) 
+{
+    ViaXvMCContext *pViaXvMC;
+    ViaXvMCSurface *pViaSurface;
+    int priv_count;
+    unsigned *priv_data;
+    unsigned i;
+    Status ret;
+
+    if((surface == NULL) || (context == NULL) || (display == NULL)){
+	return BadValue;
+    }
+  
+    pViaXvMC = (ViaXvMCContext *)context->privData;
+    ppthread_mutex_lock( &pViaXvMC->ctxMutex );
+
+    if(pViaXvMC == NULL) {
+        ppthread_mutex_unlock( &pViaXvMC->ctxMutex );
+	return (error_base + XvMCBadContext);
+    }
+
+    pViaSurface = surface->privData = (ViaXvMCSurface *)malloc(sizeof(ViaXvMCSurface));
+    if(!surface->privData) {
+        ppthread_mutex_unlock( &pViaXvMC->ctxMutex );
+	return BadAlloc;
+    }
+    XLockDisplay(display);
+    if((ret = _xvmc_create_surface(display, context, surface,
+				   &priv_count, &priv_data))) {
+	XUnlockDisplay(display);
+	free(pViaSurface);
+	fprintf(stderr,"Unable to create XvMC Surface.\n");
+	ppthread_mutex_unlock( &pViaXvMC->ctxMutex );
+	return ret;
+    }
+    XUnlockDisplay(display);
+
+    pViaSurface->srfNo = priv_data[0];
+
+    /*
+     * Store framebuffer offsets to the buffers allocated for this surface.
+     * For some chipset revisions, surfaces may be double-buffered.
+     */
+
+    pViaSurface->numBuffers = priv_data[1];
+    for (i=0; i < pViaSurface->numBuffers; ++i) {
+	pViaSurface->offsets[i] = priv_data[i+2];
+    }
+    pViaSurface->curBuf = 0;
+
+
+    /* Free data returned from xvmc_create_surface */
+
+    XFree(priv_data);
+
+    pViaSurface->width = context->width;
+    pViaSurface->height = context->height;
+    pViaSurface->yStride = pViaXvMC->yStride;
+    pViaSurface->privContext = pViaXvMC;
+    pViaSurface->privSubPic = NULL;
+    pViaSurface->needsSync = 0;
+    ppthread_mutex_unlock( &pViaXvMC->ctxMutex );
+    return Success;
+}
+
+Status XvMCDestroySurface(Display *display, XvMCSurface *surface) 
+{
+    ViaXvMCSurface *pViaSurface;
+
+    if((display == NULL) || (surface == NULL)) {
+	return BadValue;
+    }
+    if(surface->privData == NULL) {
+	return (error_base + XvMCBadSurface);
+    }
+
+    pViaSurface = (ViaXvMCSurface *)surface->privData;
+
+    XLockDisplay(display);
+    _xvmc_destroy_surface(display,surface);
+    XUnlockDisplay(display);
+
+    free(pViaSurface);
+    surface->privData = NULL;
+    return Success;
+}
+
+Status XvMCPutSlice2(Display *display,XvMCContext *context, char *slice,
+		     int nBytes, int sliceCode) 
+{
+    ViaXvMCContext *pViaXvMC;
+    CARD32 sCode = 0x00010000 | (sliceCode & 0xFF) << 24;
+
+    if((display == NULL) || (context == NULL)) {
+	return BadValue;
+    }
+    if(NULL == (pViaXvMC = context->privData)) {
+	return (error_base + XvMCBadContext);
+    }
+    ppthread_mutex_lock( &pViaXvMC->ctxMutex );
+    if (!pViaXvMC->haveDecoder) {
+	fprintf(stderr,"XvMCPutSlice: This context does not own decoder!\n");
+	ppthread_mutex_unlock( &pViaXvMC->ctxMutex );
+	return BadAlloc;
+    }
+
+    viaMpegWriteSlice(pViaXvMC->xl, (CARD8 *)slice, nBytes, sCode);
+
+    flushPCIXvMCLowLevel(pViaXvMC->xl); 
+    ppthread_mutex_unlock( &pViaXvMC->ctxMutex );
+    return Success;
+}
+
+Status XvMCPutSlice(Display *display,XvMCContext *context, char *slice,
+		    int nBytes) 
+{
+    ViaXvMCContext *pViaXvMC;
+
+    if((display == NULL) || (context == NULL)) {
+	return BadValue;
+    }
+    if(NULL == (pViaXvMC = context->privData)) {
+	return (error_base + XvMCBadContext);
+    }
+    ppthread_mutex_lock( &pViaXvMC->ctxMutex );
+
+    if (!pViaXvMC->haveDecoder) {
+	fprintf(stderr,"XvMCPutSlice: This context does not own decoder!\n");
+	ppthread_mutex_unlock( &pViaXvMC->ctxMutex );
+	return BadAlloc;
+    }
+
+    viaMpegWriteSlice(pViaXvMC->xl, (CARD8 *)slice, nBytes, 0);
+    flushPCIXvMCLowLevel(pViaXvMC->xl);
+    ppthread_mutex_unlock( &pViaXvMC->ctxMutex );
+    return Success;
+}
+
+
+static Status updateXVOverlay(Display *display,ViaXvMCContext *pViaXvMC, 
+			      ViaXvMCSurface *pViaSurface, Drawable draw,
+			      short srcx, short srcy, unsigned short srcw, 
+			      unsigned short srch,short destx,short desty,
+			      unsigned short destw,unsigned short desth) 
+{
+    ViaXvMCCommandBuffer buf;
+    ViaXvMCSubPicture *pViaSubPic;
+    Status ret;
+
+    if (!pViaXvMC->haveXv) {
+	pViaXvMC->xvImage = 
+	    XvCreateImage(display,pViaXvMC->port,FOURCC_XVMC,
+			  (char *)&buf,pViaSurface->width,
+			  pViaSurface->height);
+	pViaXvMC->gc = XCreateGC(display,draw,0,0);
+	pViaXvMC->haveXv = 1;
+    }    
+    pViaXvMC->draw = draw;
+    pViaXvMC->xvImage->data = (char *)&buf;
+    
+    buf.command = (pViaXvMC->attribChanged) ? 
+	VIA_XVMC_COMMAND_FDISPLAY : VIA_XVMC_COMMAND_DISPLAY;
+    buf.ctxNo = pViaXvMC->ctxNo | VIA_XVMC_VALID;
+    buf.srfNo = pViaSurface->srfNo | VIA_XVMC_VALID;
+    pViaSubPic = pViaSurface->privSubPic;
+    buf.subPicNo = ((NULL == pViaSubPic) ? 0 : pViaSubPic->srfNo  ) 
+	| VIA_XVMC_VALID;
+    buf.attrib = pViaXvMC->attrib;
+    
+    XLockDisplay(display);
+
+    if ((ret = XvPutImage(display,pViaXvMC->port,draw,pViaXvMC->gc,
+			  pViaXvMC->xvImage,srcx,srcy,srcw,srch,
+			  destx,desty,destw,desth))) {
+	XUnlockDisplay(display);
+	return ret;
+    }
+    XSync(display, 0);
+    XUnlockDisplay(display);
+    pViaXvMC->attribChanged = 0;
+    return Success;
+}
+
+
+Status XvMCPutSurface(Display *display,XvMCSurface *surface,Drawable draw,
+		      short srcx, short srcy, unsigned short srcw, 
+		      unsigned short srch,short destx,short desty,
+		      unsigned short destw,unsigned short desth, int flags) 
+{
+    /*
+     * This function contains some hairy locking logic. What we really want to
+     * do is to flip the picture ASAP, to get a low latency and smooth playback.
+     * However, if somebody else used the overlay since we used it last or if it is
+     * our first time, we'll have to call X to update the overlay first. Otherwise 
+     * we'll do the overlay update once we've flipped. Since we release the hardware
+     * lock when we call X, X needs to verify using the SAREA that nobody else flipped
+     * in a picture between the lock release and the X server control. Similarly
+     * when the overlay update returns, we have to make sure that we still own the
+     * overlay.
+     */
+
+    ViaXvMCSurface *pViaSurface;
+    ViaXvMCContext *pViaXvMC;
+    ViaXvMCSubPicture *pViaSubPic;
+    volatile ViaXvMCSAreaPriv *sAPriv;
+    Status ret;
+    unsigned dispSurface, lastSurface;
+    int overlayUpdated;
+    drawableInfo *drawInfo;
+    XvMCRegion sReg, dReg;
+    Bool forceUpdate = FALSE;
+
+    if((display == NULL) || (surface == NULL)) {
+	return BadValue;
+    }
+    if(NULL == (pViaSurface = surface->privData )) {
+	return (error_base + XvMCBadSurface);
+    }
+    if (NULL == (pViaXvMC = pViaSurface->privContext)) {
+	return (error_base + XvMCBadContext);
+    }
+
+    ppthread_mutex_lock( &pViaXvMC->ctxMutex );
+    pViaSubPic = pViaSurface->privSubPic;
+    sAPriv = SAREAPTR( pViaXvMC );
+
+    setRegion(srcx, srcy, srcw, srch, sReg);
+    setRegion(destx, desty, destw, desth, dReg);
+    
+
+    if ((!regionEqual(sReg, pViaXvMC->sRegion)) ||
+	(!regionEqual(dReg, pViaXvMC->dRegion))) {
+    
+	/*
+	 * Force update of the video overlay to match the new format.
+	 */
+
+	pViaXvMC->sRegion = sReg;
+	pViaXvMC->dRegion = dReg;
+	forceUpdate = TRUE;
+    }
+    
+
+    hwlLock(pViaXvMC->xl,1);
+    
+    if (getDRIDrawableInfoLocked(pViaXvMC->drawHash, display, pViaXvMC->screen, draw, 0,
+				 pViaXvMC->fd, pViaXvMC->drmcontext, pViaXvMC->sAreaAddress, 
+				 FALSE, &drawInfo, sizeof(*drawInfo))) {
+
+	hwlUnlock(pViaXvMC->xl,1);
+	ppthread_mutex_unlock( &pViaXvMC->ctxMutex );
+	return BadAccess;
+    }
+
+    setLowLevelLocking(pViaXvMC->xl,0);
+	
+
+    /*
+     * Put a surface ID in the SAREA to "authenticate" to the 
+     * X server.
+     */
+
+    dispSurface = sAPriv->XvMCDisplaying[pViaXvMC->xvMCPort];
+    lastSurface = pViaXvMC->lastSrfDisplaying;
+    sAPriv->XvMCDisplaying[pViaXvMC->xvMCPort] = 
+	pViaXvMC->lastSrfDisplaying = pViaSurface->srfNo | VIA_XVMC_VALID;
+    overlayUpdated = 0;
+
+    viaVideoSetSWFLipLocked(pViaXvMC->xl, yOffs(pViaSurface), uOffs(pViaSurface), 
+			    vOffs(pViaSurface), pViaSurface->yStride, pViaSurface->yStride >> 1);
+
+    while ((lastSurface != dispSurface) || forceUpdate) {
+
+	forceUpdate = FALSE;
+	flushPCIXvMCLowLevel(pViaXvMC->xl);
+	setLowLevelLocking(pViaXvMC->xl,1);
+	hwlUnlock(pViaXvMC->xl,1);
+
+	/*
+	 * We weren't the last to display. Update the overlay before flipping.
+	 */
+
+	ret = updateXVOverlay(display,pViaXvMC,pViaSurface,draw,srcx,srcy,srcw, 
+			      srch,destx,desty,destw,desth);
+	if (ret) {
+	    ppthread_mutex_unlock( &pViaXvMC->ctxMutex );
+	    return ret;
+	}
+
+	hwlLock(pViaXvMC->xl,1);
+
+	if (getDRIDrawableInfoLocked(pViaXvMC->drawHash, display, pViaXvMC->screen, draw, 0,
+				     pViaXvMC->fd, pViaXvMC->drmcontext, pViaXvMC->sAreaAddress, 
+				     FALSE, &drawInfo, sizeof(*drawInfo))) {
+	    
+	    hwlUnlock(pViaXvMC->xl,1);
+	    ppthread_mutex_unlock( &pViaXvMC->ctxMutex );
+	    return BadAccess;
+	}
+	
+	setLowLevelLocking(pViaXvMC->xl,0);
+	lastSurface = pViaSurface->srfNo | VIA_XVMC_VALID;
+	dispSurface = sAPriv->XvMCDisplaying[pViaXvMC->xvMCPort];
+	overlayUpdated = 1;
+    } 
+
+
+    /*
+     * Subpictures
+     */
+
+    if (NULL != pViaSubPic) {
+	if (sAPriv->XvMCSubPicOn[pViaXvMC->xvMCPort] 
+	    != (pViaSubPic->srfNo | VIA_XVMC_VALID)) {
+	    sAPriv->XvMCSubPicOn[pViaXvMC->xvMCPort] = 
+		pViaSubPic->srfNo | VIA_XVMC_VALID; 
+	    viaVideoSubPictureLocked(pViaXvMC->xl, pViaSubPic);
+	}  
+    } else {
+	if (sAPriv->XvMCSubPicOn[pViaXvMC->xvMCPort] & VIA_XVMC_VALID) {
+	    viaVideoSubPictureOffLocked(pViaXvMC->xl);
+	    sAPriv->XvMCSubPicOn[pViaXvMC->xvMCPort] &= ~VIA_XVMC_VALID;
+	}  
+    }
+
+    /*
+     * Flip
+     */
+
+    viaVideoSWFlipLocked(pViaXvMC->xl, flags, pViaSurface->progressiveSequence);
+    flushXvMCLowLevel(pViaXvMC->xl);  
+
+    setLowLevelLocking(pViaXvMC->xl,1);
+    hwlUnlock(pViaXvMC->xl,1);
+
+    if (overlayUpdated || !drawInfo->touched ) {
+        ppthread_mutex_unlock( &pViaXvMC->ctxMutex );
+	return Success;
+    }
+    
+    /*
+     * Update overlay
+     */
+
+    ret = updateXVOverlay(display,pViaXvMC,pViaSurface,draw,srcx,srcy,srcw, 
+			  srch,destx,desty,destw,desth);
+    ppthread_mutex_unlock( &pViaXvMC->ctxMutex );
+    return ret;
+
+}
+
+Status XvMCBeginSurface(Display *display,
+			XvMCContext *context,
+			XvMCSurface *target_surface,
+			XvMCSurface *past_surface,
+			XvMCSurface *future_surface,
+			const XvMCMpegControl *control) 
+{
+    ViaXvMCSurface *targS,*futS,*pastS;
+    ViaXvMCContext *pViaXvMC;
+    int hadDecoderLast;
+    CARD32 timeStamp;
+
+    if((display == NULL) || (context == NULL) || (target_surface == NULL)) {
+	return BadValue;
+    }
+     
+    pViaXvMC = context->privData;     
+
+    ppthread_mutex_lock( &pViaXvMC->ctxMutex );
+    if (grabDecoder(pViaXvMC, &hadDecoderLast)) {
+      ppthread_mutex_unlock( &pViaXvMC->ctxMutex );
+      return BadAlloc;
+    }
+    pViaXvMC->haveDecoder = 1;
+ 
+    /*
+     * We need to wait for decoder idle at next flush, since hardware doesn't queue 
+     * beginsurface requests until the decoder is idle. This is  
+     * done by waiting on the last previous timestamp, or if there was another context
+     * having the decoder before us, by emitting a new one.
+     */
+
+    if (pViaXvMC->useAGP) {
+        if (!hadDecoderLast || pViaXvMC->timeStamp == 0) {
+	    timeStamp = viaDMATimeStampLowLevel(pViaXvMC->xl);
+	    if (flushXvMCLowLevel(pViaXvMC->xl)) {
+	        releaseDecoder(pViaXvMC, 0);
+		ppthread_mutex_unlock( &pViaXvMC->ctxMutex );
+		return BadAlloc;
+	    } 	      
+	    pViaXvMC->timeStamp = timeStamp;
+	} else {
+	    timeStamp = pViaXvMC->timeStamp;
+        }
+	setAGPSyncLowLevel(pViaXvMC->xl, 1, timeStamp);
+    }
+
+    if (!hadDecoderLast || !pViaXvMC->decoderOn) {
+	pViaXvMC->intraLoaded = 0;
+	pViaXvMC->nonIntraLoaded = 0;
+    }
+
+    viaMpegReset(pViaXvMC->xl);
+
+    targS = (ViaXvMCSurface *)target_surface->privData;
+    futS = NULL;
+    pastS = NULL;
+
+ 
+    pViaXvMC->rendSurf[0] = targS->srfNo | VIA_XVMC_VALID;
+    if (future_surface) {
+	futS = (ViaXvMCSurface *)future_surface->privData;
+	futS->needsSync = 0;
+    }
+    if (past_surface) {	
+	pastS = (ViaXvMCSurface *)past_surface->privData;
+	pastS->needsSync = 0;
+    }
+
+
+    targS->progressiveSequence = (control->flags & XVMC_PROGRESSIVE_SEQUENCE);
+    targS->topFieldFirst = (control->flags & XVMC_TOP_FIELD_FIRST);
+    targS->privSubPic = NULL;
+    
+    viaMpegSetSurfaceStride(pViaXvMC->xl,pViaXvMC);
+
+    viaMpegSetFB(pViaXvMC->xl,0,yOffs(targS),uOffs(targS),vOffs(targS));
+    if (past_surface) {
+	viaMpegSetFB(pViaXvMC->xl,1,yOffs(pastS),uOffs(pastS),vOffs(pastS));
+    } else {
+	viaMpegSetFB(pViaXvMC->xl,1,0,0,0);
+    }
+    
+    if (future_surface) {
+	viaMpegSetFB(pViaXvMC->xl,2,yOffs(futS),uOffs(futS),vOffs(futS));
+    } else {
+	viaMpegSetFB(pViaXvMC->xl,2,0,0,0);
+    }
+
+    viaMpegBeginPicture(pViaXvMC->xl,pViaXvMC,context->width,context->height,control);
+    flushPCIXvMCLowLevel(pViaXvMC->xl);
+    targS->needsSync = 1;
+    targS->syncMode = LL_MODE_DECODER_IDLE;
+    pViaXvMC->decoderOn = 1;
+    ppthread_mutex_unlock( &pViaXvMC->ctxMutex );
+    return Success;
+}
+    
+  
+Status XvMCSyncSurface(Display *display,XvMCSurface *surface)
+{
+    ViaXvMCSurface *pViaSurface;
+    ViaXvMCContext *pViaXvMC;
+    unsigned i;
+
+    if((display == NULL) || (surface == NULL)) {
+	return BadValue;
+    }
+    if(surface->privData == NULL) {
+	return (error_base + XvMCBadSurface);
+    }
+    
+    pViaSurface = (ViaXvMCSurface *)surface->privData;
+    pViaXvMC = pViaSurface->privContext;
+
+    if(pViaXvMC == NULL) {
+	return (error_base + XvMCBadSurface);
+    }
+
+    ppthread_mutex_lock( &pViaXvMC->ctxMutex );
+
+    if (pViaSurface->needsSync) {
+        CARD32 timeStamp = pViaSurface->timeStamp;
+	int syncMode = pViaSurface->syncMode;
+
+	if (pViaXvMC->useAGP) {
+
+	    syncMode = (pViaSurface->syncMode == LL_MODE_2D || 
+			pViaSurface->timeStamp < pViaXvMC->timeStamp) ? 
+		LL_MODE_2D : LL_MODE_DECODER_IDLE;
+	    if (pViaSurface->syncMode != LL_MODE_2D) 
+	        timeStamp = pViaXvMC->timeStamp;
+
+	} else if (syncMode != LL_MODE_2D && 
+		   pViaXvMC->rendSurf[0] != (pViaSurface->srfNo | VIA_XVMC_VALID)) {
+
+	  pViaSurface->needsSync = 0;
+	  ppthread_mutex_unlock( &pViaXvMC->ctxMutex );
+	  return Success;
+	}
+
+	if (syncXvMCLowLevel(pViaXvMC->xl, syncMode, 1, 
+			     pViaSurface->timeStamp)) {
+	    ppthread_mutex_unlock( &pViaXvMC->ctxMutex );
+	    return BadValue;
+	}
+	pViaSurface->needsSync = 0;
+    }
+
+    if (pViaXvMC->rendSurf[0] == (pViaSurface->srfNo | VIA_XVMC_VALID)) {
+	pViaSurface->needsSync = 0;
+	for (i=0; i<VIA_MAX_RENDSURF; ++i) {
+	    pViaXvMC->rendSurf[i] = 0;
+	}
+    }
+
+    ppthread_mutex_unlock( &pViaXvMC->ctxMutex );
+    return Success;
+}
+
+Status XvMCLoadQMatrix(Display *display, XvMCContext *context,
+		       const XvMCQMatrix *qmx) 
+{	
+    ViaXvMCContext
+	*pViaXvMC;
+
+    if((display == NULL) || (context == NULL)) {
+	return BadValue;
+    }
+    if(NULL == (pViaXvMC = context->privData)) {
+	return (error_base + XvMCBadContext);
+    }
+
+    ppthread_mutex_lock( &pViaXvMC->ctxMutex );
+    if (qmx->load_intra_quantiser_matrix) {
+	memcpy(pViaXvMC->intra_quantiser_matrix,
+	       qmx->intra_quantiser_matrix,
+	       sizeof(qmx->intra_quantiser_matrix));
+	pViaXvMC->intraLoaded = 0;
+    }
+
+    if (qmx->load_non_intra_quantiser_matrix) {
+	memcpy(pViaXvMC->non_intra_quantiser_matrix,
+	       qmx->non_intra_quantiser_matrix,
+	       sizeof(qmx->non_intra_quantiser_matrix));
+	pViaXvMC->nonIntraLoaded = 0;
+    }
+
+    if (qmx->load_chroma_intra_quantiser_matrix) {
+	memcpy(pViaXvMC->chroma_intra_quantiser_matrix,
+	       qmx->chroma_intra_quantiser_matrix,
+	       sizeof(qmx->chroma_intra_quantiser_matrix));
+	pViaXvMC->chromaIntraLoaded = 0;
+    }
+
+    if (qmx->load_chroma_non_intra_quantiser_matrix) {
+	memcpy(pViaXvMC->chroma_non_intra_quantiser_matrix,
+	       qmx->chroma_non_intra_quantiser_matrix,
+	       sizeof(qmx->chroma_non_intra_quantiser_matrix));
+	pViaXvMC->chromaNonIntraLoaded = 0;
+    }
+    ppthread_mutex_unlock( &pViaXvMC->ctxMutex );
+
+    return Success;
+}    
+
+/*
+ * Below, we provide functions unusable for this implementation, but for
+ * standard completeness.
+ */
+
+  
+Status XvMCRenderSurface
+(
+    Display *display,
+    XvMCContext *context,
+    unsigned int picture_structure,
+    XvMCSurface *target_surface,
+    XvMCSurface *past_surface,
+    XvMCSurface *future_surface,
+    unsigned int flags,
+    unsigned int num_macroblocks,
+    unsigned int first_macroblock,
+    XvMCMacroBlockArray *macroblock_array,
+    XvMCBlockArray *blocks
+    ) 
+{
+    return (error_base + XvMCBadContext);
+}
+
+Status XvMCCreateBlocks 
+(
+    Display *display, 
+    XvMCContext *context,
+    unsigned int num_blocks,
+    XvMCBlockArray * block
+    )
+{
+    return (error_base + XvMCBadContext);
+}
+
+Status XvMCDestroyBlocks (Display *display, XvMCBlockArray * block) 
+{
+    return Success;
+}
+
+Status XvMCCreateMacroBlocks 
+(
+    Display *display, 
+    XvMCContext *context,
+    unsigned int num_blocks,
+    XvMCMacroBlockArray * blocks
+    )
+{
+    return (error_base + XvMCBadContext);
+}
+
+Status XvMCDestroyMacroBlocks(Display *display, XvMCMacroBlockArray * block)
+{
+    return (error_base + XvMCBadContext);
+}
+
+Status XvMCCreateSubpicture( Display *display, 
+			     XvMCContext *context,
+			     XvMCSubpicture *subpicture,
+			     unsigned short width,
+			     unsigned short height,
+			     int xvimage_id) 
+{
+    ViaXvMCContext *pViaXvMC;
+    ViaXvMCSubPicture *pViaSubPic;
+    int priv_count;
+    unsigned *priv_data;
+    Status ret;
+
+    if((subpicture == NULL) || (context == NULL) || (display == NULL)){
+	return BadValue;
+    }
+  
+    pViaXvMC = (ViaXvMCContext *)context->privData;
+    if(pViaXvMC == NULL) {
+	return (error_base + XvMCBadContext);
+    }
+
+    subpicture->privData = (ViaXvMCSubPicture *)
+	malloc(sizeof(ViaXvMCSubPicture));
+    if(!subpicture->privData) {
+	return BadAlloc;
+    }
+
+    ppthread_mutex_lock( &pViaXvMC->ctxMutex );
+    subpicture->width = context->width;
+    subpicture->height = context->height;
+    subpicture->xvimage_id = xvimage_id;
+    pViaSubPic = (ViaXvMCSubPicture *)subpicture->privData;
+
+    XLockDisplay(display);
+    if((ret = _xvmc_create_subpicture(display, context, subpicture,
+				      &priv_count, &priv_data))) {
+	XUnlockDisplay(display);
+	free(pViaSubPic);
+	fprintf(stderr,"Unable to create XvMC Subpicture.\n");
+	ppthread_mutex_unlock( &pViaXvMC->ctxMutex );
+	return ret;
+    }
+    XUnlockDisplay(display);
+
+
+    subpicture->num_palette_entries = VIA_SUBPIC_PALETTE_SIZE;
+    subpicture->entry_bytes = 3;
+    strncpy(subpicture->component_order,"YUV",4); 
+    pViaSubPic->srfNo = priv_data[0];
+    pViaSubPic->offset = priv_data[1];
+    pViaSubPic->stride = (subpicture->width + 31) & ~31;
+    pViaSubPic->privContext = pViaXvMC;
+    pViaSubPic->ia44 = (xvimage_id == FOURCC_IA44);
+    pViaSubPic->needsSync = 0;
+
+    /* Free data returned from _xvmc_create_subpicture */
+
+    XFree(priv_data);
+    ppthread_mutex_unlock( &pViaXvMC->ctxMutex );
+    return Success;
+}
+
+Status
+XvMCSetSubpicturePalette (Display *display, XvMCSubpicture *subpicture, 
+			  unsigned char *palette) 
+{    
+    ViaXvMCSubPicture *pViaSubPic;
+    ViaXvMCContext *pViaXvMC;
+    volatile ViaXvMCSAreaPriv *sAPriv;
+    unsigned i;
+    CARD32 tmp;
+
+    if((subpicture == NULL) || (display == NULL)){
+	return BadValue;
+    }
+    if(subpicture->privData == NULL) {
+	return (error_base + XvMCBadSubpicture);
+    }
+    pViaSubPic = (ViaXvMCSubPicture *) subpicture->privData;
+    for (i=0; i < VIA_SUBPIC_PALETTE_SIZE; ++i) {
+	tmp = *palette++ << 8;
+	tmp |= *palette++ << 16;
+	tmp |= *palette++ << 24;
+	tmp |= ((i & 0x0f) << 4) | 0x07;
+	pViaSubPic->palette[i] = tmp;
+    }
+
+    pViaXvMC = pViaSubPic->privContext;
+    ppthread_mutex_lock( &pViaXvMC->ctxMutex );
+    sAPriv = SAREAPTR( pViaXvMC );
+    hwlLock(pViaXvMC->xl,1);
+    setLowLevelLocking(pViaXvMC->xl,0);
+
+    /*
+     * If the subpicture is displaying, Immeadiately update it with the
+     * new palette.
+     */
+
+    if (sAPriv->XvMCSubPicOn[pViaXvMC->xvMCPort] == 
+	(pViaSubPic->srfNo | VIA_XVMC_VALID)) {
+        viaVideoSubPictureLocked(pViaXvMC->xl,pViaSubPic);
+    }
+    flushPCIXvMCLowLevel(pViaXvMC->xl);
+    setLowLevelLocking(pViaXvMC->xl,1);
+    hwlUnlock(pViaXvMC->xl,1);
+    ppthread_mutex_unlock( &pViaXvMC->ctxMutex );
+    return Success;
+}
+
+
+static int findOverlap(unsigned width,unsigned height,
+		       short *dstX, short *dstY,
+		       short *srcX, short *srcY,
+		       unsigned short *areaW, unsigned short *areaH) 
+{
+    int
+	w,h;
+    unsigned
+	mWidth,mHeight;
+    
+    w = *areaW;
+    h = *areaH;
+
+    if ((*dstX >= width) || (*dstY >= height)) 
+	return 1;
+    if (*dstX < 0) {
+	w += *dstX;
+	*srcX -= *dstX;
+	*dstX = 0;
+    }
+    if (*dstY < 0) {
+	h += *dstY;
+	*srcY -= *dstY;
+	*dstY = 0;
+    }
+    if ((w <= 0) || ((h <= 0))) 
+	return 1;
+    mWidth = width - *dstX;
+    mHeight = height - *dstY;
+    *areaW = (w <= mWidth) ? w : mWidth;
+    *areaH = (h <= mHeight) ? h : mHeight; 
+    return 0;
+}
+    
+
+
+Status XvMCClearSubpicture (
+    Display *display,
+    XvMCSubpicture *subpicture,
+    short x,
+    short y,
+    unsigned short width,
+    unsigned short height,
+    unsigned int color
+    ) 
+{
+
+    ViaXvMCContext *pViaXvMC;
+    ViaXvMCSubPicture *pViaSubPic;
+    short dummyX,dummyY;
+    unsigned long bOffs;
+
+    if((subpicture == NULL) || (display == NULL)) {
+	return BadValue;
+    }
+    if(subpicture->privData == NULL) {
+	return (error_base + XvMCBadSubpicture);
+    }
+    pViaSubPic = (ViaXvMCSubPicture *) subpicture->privData;
+    pViaXvMC = pViaSubPic->privContext;
+    ppthread_mutex_lock( &pViaXvMC->ctxMutex );
+
+    /* Clip clearing area so that it fits inside subpicture. */
+
+    if (findOverlap(subpicture->width, subpicture->height, &x, &y,
+		    &dummyX, &dummyY, &width, &height)) {
+	ppthread_mutex_unlock( &pViaXvMC->ctxMutex );
+	return Success;
+    }
+		    
+    bOffs = pViaSubPic->offset + y*pViaSubPic->stride + x;
+    viaBlit(pViaXvMC->xl, 8, 0, pViaSubPic->stride, bOffs, pViaSubPic->stride,
+	    width, height, 1, 1, VIABLIT_FILL, color);
+    pViaSubPic->needsSync = 1;
+    pViaSubPic->timeStamp = viaDMATimeStampLowLevel(pViaXvMC->xl);
+    if (flushXvMCLowLevel(pViaXvMC->xl)) {
+	ppthread_mutex_unlock( &pViaXvMC->ctxMutex );
+	return BadValue;
+    }
+    ppthread_mutex_unlock( &pViaXvMC->ctxMutex );
+    return Success;
+}
+
+Status
+XvMCCompositeSubpicture (
+    Display *display,
+    XvMCSubpicture *subpicture,
+    XvImage *image,
+    short srcx,
+    short srcy,
+    unsigned short width,
+    unsigned short height,
+    short dstx,
+    short dsty
+    ) 
+{
+    
+    unsigned i;
+    ViaXvMCContext *pViaXvMC;
+    ViaXvMCSubPicture *pViaSubPic;
+    CARD8 *dAddr, *sAddr;
+
+    if((subpicture == NULL) || (display == NULL) || (image == NULL)){
+	return BadValue;
+    }
+    if(NULL == (pViaSubPic = (ViaXvMCSubPicture *)subpicture->privData)) {
+	return (error_base + XvMCBadSubpicture);
+    }
+
+    pViaXvMC = pViaSubPic->privContext;
+
+
+    if (image->id != subpicture->xvimage_id) 
+	return BadMatch;
+
+    ppthread_mutex_lock( &pViaXvMC->ctxMutex );
+    
+
+    /*
+     * Clip copy area so that it fits inside subpicture and image.
+     */
+
+    if (findOverlap(subpicture->width, subpicture->height, 
+		    &dstx, &dsty, &srcx, &srcy, &width, &height)) {
+	ppthread_mutex_unlock( &pViaXvMC->ctxMutex );
+	return Success;
+    }
+    if (findOverlap(image->width, image->height, 
+		    &srcx, &srcy, &dstx, &dsty, &width, &height)) {
+	ppthread_mutex_unlock( &pViaXvMC->ctxMutex );
+	return Success;
+    }
+
+    if (pViaSubPic->needsSync) {
+	if (syncXvMCLowLevel(pViaXvMC->xl, LL_MODE_2D, 0, pViaSubPic->timeStamp)) {
+	    ppthread_mutex_unlock( &pViaXvMC->ctxMutex );
+	    return BadValue;
+	}
+	pViaSubPic->needsSync = 0;
+    }
+
+    for(i=0; i<height; ++i) {
+        dAddr = (((CARD8 *)pViaXvMC->fbAddress) + 
+	         (pViaSubPic->offset + (dsty+i)*pViaSubPic->stride + dstx));
+        sAddr = (((CARD8 *)image->data) + 
+		 (image->offsets[0] + (srcy+i)*image->pitches[0] + srcx));
+	memcpy(dAddr,sAddr,width);
+    }
+
+    ppthread_mutex_unlock( &pViaXvMC->ctxMutex );
+    return Success;
+}
+
+
+
+Status
+XvMCBlendSubpicture (
+    Display *display,
+    XvMCSurface *target_surface,
+    XvMCSubpicture *subpicture,
+    short subx,
+    short suby,
+    unsigned short subw,
+    unsigned short subh,
+    short surfx,
+    short surfy,
+    unsigned short surfw,
+    unsigned short surfh
+    ) 
+{    
+    ViaXvMCSurface *pViaSurface;
+    ViaXvMCSubPicture *pViaSubPic;
+
+    if((display == NULL) || target_surface == NULL){
+	return BadValue;
+    }
+    
+    if (subx || suby || surfx || surfy ||
+	(subw != surfw) || (subh != surfh))  {
+	fprintf(stderr,"ViaXvMC: Only completely overlapping subpicture "
+		"supported.\n");
+	return BadValue;
+    }
+
+    if(NULL == (pViaSurface = target_surface->privData)) {
+	return (error_base + XvMCBadSurface);
+    }
+
+    if (subpicture) {
+
+	if(NULL == (pViaSubPic = subpicture->privData)) {
+	    return (error_base + XvMCBadSubpicture);
+	}
+	
+	pViaSurface->privSubPic = pViaSubPic;
+    } else {
+	pViaSurface->privSubPic = NULL;
+    }
+    return Success;
+}
+
+Status
+XvMCBlendSubpicture2 (
+    Display *display,
+    XvMCSurface *source_surface,
+    XvMCSurface *target_surface,
+    XvMCSubpicture *subpicture,
+    short subx,
+    short suby,
+    unsigned short subw,
+    unsigned short subh,
+    short surfx,
+    short surfy,
+    unsigned short surfw,
+    unsigned short surfh
+    ) 
+{    
+    ViaXvMCSurface *pViaSurface,*pViaSSurface;
+    ViaXvMCSubPicture *pViaSubPic;
+    ViaXvMCContext *pViaXvMC;
+
+    unsigned width,height;
+
+    if((display == NULL) || target_surface == NULL || source_surface == NULL){
+	return BadValue;
+    }
+    
+    if (subx || suby || surfx || surfy ||
+	(subw != surfw) || (subh != surfh))  {
+	fprintf(stderr,"ViaXvMC: Only completely overlapping subpicture "
+		"supported.\n");
+	return BadMatch;
+    }
+
+    if(NULL == (pViaSurface = target_surface->privData)) {
+	return (error_base + XvMCBadSurface);
+    }
+
+    if(NULL == (pViaSSurface = source_surface->privData)) {
+	return (error_base + XvMCBadSurface);
+    }
+    pViaXvMC = pViaSurface->privContext;
+    width = pViaSSurface->width;
+    height = pViaSSurface->height;
+    if (width != pViaSurface->width || height != pViaSSurface->height) {
+	return BadMatch;
+    }
+
+    if (XvMCSyncSurface(display,source_surface)) {
+        return BadValue;
+    }
+
+    ppthread_mutex_lock( &pViaXvMC->ctxMutex );
+    viaBlit(pViaXvMC->xl, 8, yOffs(pViaSSurface), pViaSSurface->yStride,
+	    yOffs(pViaSurface), pViaSurface->yStride,
+	    width, height, 1, 1, VIABLIT_COPY, 0);
+    flushPCIXvMCLowLevel(pViaXvMC->xl);
+    if (pViaXvMC->chipId != PCI_CHIP_VT3259) {
+
+	/*
+	 * YV12 Chroma blit.
+	 */
+	
+	viaBlit(pViaXvMC->xl, 8, uOffs(pViaSSurface), pViaSSurface->yStride >> 1,
+		uOffs(pViaSurface), pViaSurface->yStride >> 1,
+		width >> 1, height >> 1, 1, 1, VIABLIT_COPY, 0);
+	flushPCIXvMCLowLevel(pViaXvMC->xl);
+	viaBlit(pViaXvMC->xl, 8, vOffs(pViaSSurface), pViaSSurface->yStride >> 1,
+		vOffs(pViaSurface), pViaSurface->yStride >> 1,
+		width >> 1, height >> 1, 1, 1, VIABLIT_COPY, 0);
+    } else {
+	
+	/*
+	 * NV12 Chroma blit.
+	 */
+
+	viaBlit(pViaXvMC->xl, 8, vOffs(pViaSSurface), pViaSSurface->yStride,
+		vOffs(pViaSurface), pViaSurface->yStride,
+		width, height >> 1, 1, 1, VIABLIT_COPY, 0);
+    }
+    pViaSurface->needsSync = 1;
+    pViaSurface->syncMode = LL_MODE_2D;
+    pViaSurface->timeStamp = viaDMATimeStampLowLevel(pViaXvMC->xl);
+    if (flushXvMCLowLevel(pViaXvMC->xl)) {
+	ppthread_mutex_unlock( &pViaXvMC->ctxMutex );
+	return BadValue;
+    }
+    if (subpicture) {
+
+	if(NULL == (pViaSubPic = subpicture->privData)) {
+	    ppthread_mutex_unlock( &pViaXvMC->ctxMutex );
+	    return (error_base + XvMCBadSubpicture);
+	}
+	
+	pViaSurface->privSubPic = pViaSubPic;
+    } else {
+	pViaSurface->privSubPic = NULL;
+    }
+    ppthread_mutex_unlock( &pViaXvMC->ctxMutex );
+    return Success;
+}
+
+
+Status
+XvMCSyncSubpicture (Display *display, XvMCSubpicture *subpicture) 
+{
+    ViaXvMCSubPicture *pViaSubPic;
+    ViaXvMCContext *pViaXvMC;
+    Status retVal=0;
+
+    if((display == NULL) || subpicture == NULL){
+	return BadValue;
+    }
+    if(NULL == (pViaSubPic = subpicture->privData)) {
+	return (error_base + XvMCBadSubpicture);
+    }
+
+    pViaXvMC = pViaSubPic->privContext;
+    ppthread_mutex_lock( &pViaXvMC->ctxMutex );
+    if (pViaSubPic->needsSync) {
+	if (syncXvMCLowLevel(pViaXvMC->xl, LL_MODE_2D, 
+			     0, pViaSubPic->timeStamp)) {
+	    retVal = BadValue;
+	}
+	pViaSubPic->needsSync = 0;
+    }
+    ppthread_mutex_unlock( &pViaXvMC->ctxMutex );
+    return retVal;
+}
+
+Status
+XvMCFlushSubpicture (Display *display, XvMCSubpicture *subpicture) 
+{
+    ViaXvMCSubPicture *pViaSubPic;
+
+    if((display == NULL) || subpicture == NULL){
+	return BadValue;
+    }
+    if(NULL == (pViaSubPic = subpicture->privData)) {
+	return (error_base + XvMCBadSubpicture);
+    }
+    
+    return Success;
+}
+
+Status
+XvMCDestroySubpicture (Display *display, XvMCSubpicture *subpicture) 
+{
+    ViaXvMCSubPicture *pViaSubPic;
+    ViaXvMCContext *pViaXvMC;
+    volatile ViaXvMCSAreaPriv *sAPriv;
+
+    if((display == NULL) || subpicture == NULL){
+	return BadValue;
+    }
+    if(NULL == (pViaSubPic = subpicture->privData)) {
+	return (error_base + XvMCBadSubpicture);
+    }
+    pViaXvMC = pViaSubPic->privContext;
+    ppthread_mutex_lock( &pViaXvMC->ctxMutex );
+
+    
+    sAPriv = SAREAPTR(pViaXvMC);
+    hwlLock(pViaXvMC->xl,1); 
+    setLowLevelLocking(pViaXvMC->xl,0);
+    if (sAPriv->XvMCSubPicOn[pViaXvMC->xvMCPort] == 
+	( pViaSubPic->srfNo | VIA_XVMC_VALID )) {
+        viaVideoSubPictureOffLocked(pViaXvMC->xl);
+	sAPriv->XvMCSubPicOn[pViaXvMC->xvMCPort] = 0;
+    }
+    flushPCIXvMCLowLevel(pViaXvMC->xl);
+    setLowLevelLocking(pViaXvMC->xl,1);
+    hwlUnlock(pViaXvMC->xl,1);
+
+    XLockDisplay(display);
+    _xvmc_destroy_subpicture(display,subpicture);
+    XUnlockDisplay(display);
+
+    free(pViaSubPic);
+    subpicture->privData = NULL;
+    ppthread_mutex_unlock( &pViaXvMC->ctxMutex );
+
+    return Success;
+}
+
+Status
+XvMCGetSubpictureStatus (Display *display, XvMCSubpicture *subpic, int *stat) 
+{
+    ViaXvMCSubPicture *pViaSubPic;
+    ViaXvMCContext *pViaXvMC;
+    volatile ViaXvMCSAreaPriv *sAPriv;
+   
+
+    if((display == NULL) || subpic == NULL){
+	return BadValue;
+    }
+    if(NULL == (pViaSubPic = subpic->privData)) {
+	return (error_base + XvMCBadSubpicture);
+    }
+    if (stat) {
+	*stat = 0;
+	pViaXvMC = pViaSubPic->privContext;
+	sAPriv = SAREAPTR( pViaXvMC );
+	if (sAPriv->XvMCSubPicOn[pViaXvMC->xvMCPort] == 
+	    (pViaSubPic->srfNo | VIA_XVMC_VALID)) 
+	    *stat |= XVMC_DISPLAYING;	    
+    }
+    return Success;
+}
+
+Status
+XvMCFlushSurface (Display *display, XvMCSurface *surface)
+{
+    ViaXvMCSurface *pViaSurface;
+    ViaXvMCContext *pViaXvMC;
+    Status ret;
+
+    if((display == NULL) || surface == NULL){
+	return BadValue;
+    }
+    if(NULL == (pViaSurface = surface->privData)) {
+	return (error_base + XvMCBadSurface);
+    }
+
+    pViaXvMC = pViaSurface->privContext;
+    ppthread_mutex_lock( &pViaXvMC->ctxMutex );
+    if (pViaSurface->needsSync) 
+	pViaSurface->timeStamp = pViaXvMC->timeStamp = 
+	  viaDMATimeStampLowLevel(pViaXvMC->xl);
+    ret = (flushXvMCLowLevel(pViaXvMC->xl)) ? BadValue : Success;
+    if (pViaXvMC->rendSurf[0] == (pViaSurface->srfNo | VIA_XVMC_VALID)) {
+	hwlLock(pViaXvMC->xl,0);
+	pViaXvMC->haveDecoder = 0;
+	releaseDecoder(pViaXvMC, 0);
+	hwlUnlock(pViaXvMC->xl,0);
+    }
+    ppthread_mutex_unlock( &pViaXvMC->ctxMutex );
+    return ret;
+}
+
+Status
+XvMCGetSurfaceStatus (Display *display, XvMCSurface *surface, int *stat) 
+{
+    ViaXvMCSurface *pViaSurface;
+    ViaXvMCContext *pViaXvMC;
+    volatile ViaXvMCSAreaPriv *sAPriv;
+    unsigned i;
+    int ret = 0;
+
+    if((display == NULL) || surface == NULL){
+	return BadValue;
+    }
+    if(NULL == (pViaSurface = surface->privData)) {
+	return (error_base + XvMCBadSurface);
+    }
+    if (stat) {
+	*stat = 0;
+	pViaXvMC = pViaSurface->privContext;
+	ppthread_mutex_lock( &pViaXvMC->ctxMutex ); 
+	sAPriv = SAREAPTR( pViaXvMC );
+	if (sAPriv->XvMCDisplaying[pViaXvMC->xvMCPort] 
+	    == (pViaSurface->srfNo | VIA_XVMC_VALID)) 
+	    *stat |= XVMC_DISPLAYING;
+	for (i=0; i<VIA_MAX_RENDSURF; ++i) {
+	    if(pViaXvMC->rendSurf[i] == 
+	       (pViaSurface->srfNo | VIA_XVMC_VALID)) {
+		*stat |= XVMC_RENDERING;
+		break;
+	    }
+	}
+	ppthread_mutex_unlock( &pViaXvMC->ctxMutex ); 
+    }
+    return ret;
+}
+
+XvAttribute *
+XvMCQueryAttributes (
+    Display *display,
+    XvMCContext *context,
+    int *number
+    )
+{
+    ViaXvMCContext *pViaXvMC;
+    XvAttribute *ret;
+    unsigned long siz;
+
+    *number = 0;
+    if ((display == NULL) || (context == NULL)) {
+	return NULL;
+    }
+    
+    if (NULL == (pViaXvMC = context->privData)) {
+	return NULL;
+    }
+
+    ppthread_mutex_lock( &pViaXvMC->ctxMutex );
+    if (NULL != (ret = (XvAttribute *)
+		 malloc(siz = VIA_NUM_XVMC_ATTRIBUTES*sizeof(XvAttribute)))) {
+	memcpy(ret,pViaXvMC->attribDesc,siz);
+	*number =  VIA_NUM_XVMC_ATTRIBUTES;
+    }
+    ppthread_mutex_unlock( &pViaXvMC->ctxMutex );
+
+    return ret;    
+}
+
+Status
+XvMCSetAttribute (
+    Display *display,
+    XvMCContext *context, 
+    Atom attribute, 
+    int value
+    )
+{
+    int found;
+    unsigned i;
+    ViaXvMCContext *pViaXvMC;
+    ViaXvMCCommandBuffer buf;
+    
+    if ((display == NULL) || (context == NULL)) {
+	return (error_base + XvMCBadContext);
+    }
+    
+    if (NULL == (pViaXvMC = context->privData)) {
+	return (error_base + XvMCBadContext);
+    }
+
+    ppthread_mutex_lock( &pViaXvMC->ctxMutex );
+
+    found = 0;
+    for (i=0; i < pViaXvMC->attrib.numAttr; ++i) {
+	if (attribute == pViaXvMC->attrib.attributes[i].attribute) {
+	    if ((!(pViaXvMC->attribDesc[i].flags & XvSettable)) ||
+		value < pViaXvMC->attribDesc[i].min_value ||
+		value > pViaXvMC->attribDesc[i].max_value) 
+		return BadValue;
+	    pViaXvMC->attrib.attributes[i].value = value;
+	    found = 1;
+	    pViaXvMC->attribChanged = 1;
+	    break;
+	}
+    }
+    if (!found) {
+	ppthread_mutex_unlock( &pViaXvMC->ctxMutex );
+	return BadMatch;
+    }
+    if (pViaXvMC->haveXv) {
+	buf.command = VIA_XVMC_COMMAND_ATTRIBUTES;
+	pViaXvMC->xvImage->data = (char *)&buf;
+	buf.ctxNo = pViaXvMC->ctxNo | VIA_XVMC_VALID;
+	buf.attrib = pViaXvMC->attrib;
+	XLockDisplay(display);
+	pViaXvMC->attribChanged = 
+	    XvPutImage(display,pViaXvMC->port,pViaXvMC->draw,
+		       pViaXvMC->gc,
+		       pViaXvMC->xvImage,0,0,1,1,0,0,1,1);
+	XUnlockDisplay(display);
+    }
+    ppthread_mutex_unlock( &pViaXvMC->ctxMutex ); 
+    return Success;
+}
+    
+
+Status
+XvMCGetAttribute (
+    Display *display,
+    XvMCContext *context, 
+    Atom attribute, 
+    int *value
+    )
+{
+    int found;
+    unsigned i;
+    ViaXvMCContext *pViaXvMC;
+    
+    if ((display == NULL) || (context == NULL)) {
+	return (error_base + XvMCBadContext);
+    }
+    
+    if (NULL == (pViaXvMC = context->privData)) {
+	return (error_base + XvMCBadContext);
+    }
+
+    ppthread_mutex_lock( &pViaXvMC->ctxMutex ); 
+    found = 0;
+    for (i=0; i < pViaXvMC->attrib.numAttr; ++i) {
+	if (attribute == pViaXvMC->attrib.attributes[i].attribute) {
+	    if (pViaXvMC->attribDesc[i].flags & XvGettable) {
+		*value = pViaXvMC->attrib.attributes[i].value;
+		found = 1;
+		break;
+	    }
+	}
+    }
+    ppthread_mutex_unlock( &pViaXvMC->ctxMutex ); 
+
+    if (!found) 
+	return BadMatch;
+    return Success;
+}
+    
+
+Status XvMCHideSurface(Display *display,XvMCSurface *surface)
+{
+
+    ViaXvMCSurface *pViaSurface;
+    ViaXvMCContext *pViaXvMC;
+    ViaXvMCSubPicture *pViaSubPic;
+    volatile ViaXvMCSAreaPriv *sAPriv;
+    ViaXvMCCommandBuffer buf;
+    Status ret;
+
+    if ((display == NULL) || (surface == NULL)) {
+	return BadValue;
+    }
+    if (NULL == (pViaSurface = surface->privData )) {
+	return (error_base + XvMCBadSurface);
+    }
+    if (NULL == (pViaXvMC = pViaSurface->privContext)) {
+	return (error_base + XvMCBadContext);
+    }
+
+    ppthread_mutex_lock( &pViaXvMC->ctxMutex ); 
+    if (!pViaXvMC->haveXv) {
+	ppthread_mutex_unlock( &pViaXvMC->ctxMutex ); 
+	return Success;
+    }
+    
+    sAPriv = SAREAPTR( pViaXvMC );
+    hwlLock(pViaXvMC->xl,1); 
+
+    if (sAPriv->XvMCDisplaying[pViaXvMC->xvMCPort] != 
+	(pViaSurface->srfNo | VIA_XVMC_VALID)) {
+	hwlUnlock(pViaXvMC->xl,1);
+	ppthread_mutex_unlock( &pViaXvMC->ctxMutex ); 
+	return Success;
+    }
+    setLowLevelLocking(pViaXvMC->xl,0);
+    if (NULL != (pViaSubPic = pViaSurface->privSubPic)) {
+	if (sAPriv->XvMCSubPicOn[pViaXvMC->xvMCPort] ==
+	    (pViaSubPic->srfNo | VIA_XVMC_VALID)) {
+	    sAPriv->XvMCSubPicOn[pViaXvMC->xvMCPort] &= ~VIA_XVMC_VALID;
+	    viaVideoSubPictureOffLocked(pViaXvMC->xl); 
+	}
+    }
+    flushPCIXvMCLowLevel(pViaXvMC->xl);
+    setLowLevelLocking(pViaXvMC->xl,1);
+    hwlUnlock(pViaXvMC->xl,1);
+	    
+    buf.command = VIA_XVMC_COMMAND_UNDISPLAY;
+    buf.ctxNo = pViaXvMC->ctxNo | VIA_XVMC_VALID;
+    buf.srfNo = pViaSurface->srfNo | VIA_XVMC_VALID;
+    pViaXvMC->xvImage->data = (char *)&buf;
+    if ((ret = XvPutImage(display,pViaXvMC->port,pViaXvMC->draw,
+			  pViaXvMC->gc,
+			  pViaXvMC->xvImage,0,0,1,1,0,0,1,1))) {
+	fprintf(stderr,"XvMCPutSurface: Hiding overlay failed.\n");
+	ppthread_mutex_unlock( &pViaXvMC->ctxMutex ); 
+	return ret;
+    }
+    ppthread_mutex_unlock( &pViaXvMC->ctxMutex ); 
+    return Success;
+}
diff --git a/nx-X11/lib/XvMC/hw/via/viaXvMCPriv.h b/nx-X11/lib/XvMC/hw/via/viaXvMCPriv.h
new file mode 100644
index 000000000..de4976a25
--- /dev/null
+++ b/nx-X11/lib/XvMC/hw/via/viaXvMCPriv.h
@@ -0,0 +1,198 @@
+/*****************************************************************************
+ * VIA Unichrome XvMC extension client lib.
+ *
+ * Copyright (c) 2004 Thomas Hellström. All rights reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice shall be included in
+ * all copies or substantial portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
+ * AUTHOR(S) OR COPYRIGHT HOLDER(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+ * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING
+ * FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER
+ * DEALINGS IN THE SOFTWARE.
+ */
+
+#ifndef _VIAXVMCPRIV_H
+#define _VIAXVMCPRIV_H 1
+
+#include <X11/Xlibint.h>
+#include <X11/extensions/XvMC.h>
+#include <X11/extensions/XvMClib.h>
+#include <stdlib.h>
+#include <X11/Xutil.h>
+#include <X11/extensions/vldXvMC.h>
+#include "via_xvmc.h"
+
+typedef struct{
+    int x;
+    int y;
+    int w;
+    int h;
+} XvMCRegion;
+
+
+extern Status _xvmc_create_context(Display *dpy, XvMCContext *context,
+				   int *priv_count, uint **priv_data);
+extern Status _xvmc_destroy_context(Display *dpy, XvMCContext *context);
+extern Status _xvmc_create_surface(Display *dpy, XvMCContext *context,
+				   XvMCSurface *surface, int *priv_count,
+				   uint **priv_data);
+extern Status _xvmc_destroy_surface(Display *dpy, XvMCSurface *surface);
+extern Status _xvmc_create_subpicture(Display *dpy, XvMCContext *context,
+				      XvMCSubpicture *subpicture,
+				      int *priv_count, uint **priv_data);
+extern Status _xvmc_destroy_subpicture(Display *dpy,
+				       XvMCSubpicture *subpicture);
+
+#define VIA_SUBPIC_PALETTE_SIZE 16 /*Number of colors in subpicture palette*/
+#define VIA_CBUFFERSIZE 4096       /*Hardware command buffer size*/
+#define VIA_MAX_BUFS 2             /*Number of YUV buffers per surface*/
+#define VIA_MAX_RENDSURF 3         /*Maximum numbers of surfaces per context 
+				     that can answer RENDERING to a rendering 
+				     query*/
+
+typedef enum{
+  context_drawHash,
+  context_lowLevel,
+  context_mutex,
+  context_sAreaMap,
+  context_fbMap,
+  context_mmioMap,
+  context_drmContext,
+  context_fd,
+  context_driConnection,
+  context_context,
+  context_none
+} ContextRes;
+
+
+typedef struct{
+    unsigned ctxNo;                 /* XvMC private context reference number */
+    pthread_mutex_t ctxMutex;       /* Mutex for multi-threading. Not used */
+    drm_context_t drmcontext;       /* The drm context */
+    drm_handle_t fbOffset;          /* Handle to drm frame-buffer area */
+    drm_handle_t mmioOffset;        /* Handle to drm mmio area */
+    drm_handle_t sAreaOffset;       /* Handle to drm shared memory area */
+    unsigned fbSize;                /* Size of drm frame-buffer area */
+    unsigned mmioSize;              /* Size of drm mmio area */
+    unsigned sAreaSize;             /* Size of drm shared memory area */
+    unsigned sAreaPrivOffset;       /* Offset in sarea to private part */
+    drmAddress fbAddress;           /* Virtual address of frame buffer area */
+    drmAddress mmioAddress;         /* Virtual address of mmio area */
+    drmAddress sAreaAddress;        /* Virtual address of shared memory area */
+    char busIdString[21];           /* Busid of video card */
+    unsigned yStride;               /* Y stride of surfaces in this context */
+    int fd;                         /* FD for connection to drm module */
+    unsigned char intra_quantiser_matrix[64]; 
+    unsigned char non_intra_quantiser_matrix[64];
+    unsigned char chroma_intra_quantiser_matrix[64];
+    unsigned char chroma_non_intra_quantiser_matrix[64];
+    unsigned rendSurf[VIA_MAX_RENDSURF]; /* Which surfaces answer rendering to
+					    a rendering query */
+    int decoderOn;                       /* Decoder switched on ? */
+    int intraLoaded;                     /* Intra quantiser matrix loaded in 
+					    decoder? */
+    int nonIntraLoaded;                  /* Non-Intra quantiser matrix loaded
+					    in decoder */
+    int chromaIntraLoaded;
+    int chromaNonIntraLoaded;
+    int haveDecoder;                     /* Does this context own decoder? */
+    int attribChanged;                   /* Attributes have changed and need to
+					    be uploaded to Xv at next frame
+					    display */
+    drmLockPtr hwLock;                   /* Virtual address Pointer to the 
+					    heavyweight drm hardware lock */
+    unsigned xvMCPort;                   /* XvMC private port. Corresponds to
+					    an Xv port, but not by number */
+    ViaXvMCAttrHolder attrib;            /* This contexts attributes and their
+					    values */
+    XvAttribute attribDesc[VIA_NUM_XVMC_ATTRIBUTES]; /* Attribute decriptions */
+    int useAGP;                          /* Use the AGP ringbuffer to upload data to the chip */
+    void *xl;                            /* Lowlevel context. Opaque to us. */
+    int haveXv;                         /* Have I initialized the Xv 
+					   connection for this surface? */
+    XvImage *xvImage;                   /* Fake Xv Image used for command 
+					   buffer transport to the X server */
+    GC gc;                              /* X GC needed for displaying */
+    Drawable draw;                      /* Drawable to undisplay from */
+    XvPortID port;                      /* Xv Port ID when displaying */
+    int lastSrfDisplaying;
+    ContextRes resources;
+    CARD32 timeStamp;
+    CARD32 videoTimeStamp;
+    XID id;
+    unsigned screen;
+    unsigned depth;
+    unsigned stride;
+    XVisualInfo visualInfo;
+    void *drawHash;
+    CARD32 chipId;
+    XvMCRegion sRegion;
+    XvMCRegion dRegion;
+}ViaXvMCContext;
+
+typedef struct{
+    pthread_mutex_t subMutex;            /* Currently not used. */
+    unsigned srfNo;                      /* XvMC private surface number */
+    unsigned offset;                     /* Offset into frame-buffer area */
+    unsigned stride;                     /* Storage stride */
+    unsigned width;                      /* Width */
+    unsigned height;                     /* Height */
+    CARD32 palette[VIA_SUBPIC_PALETTE_SIZE]; /* YUV Palette */
+    ViaXvMCContext *privContext;         /* Pointer to context private data */
+    int ia44;                            /* IA44 or AI44 format */
+    int needsSync;
+    CARD32 timeStamp;
+}ViaXvMCSubPicture;
+
+
+typedef struct{
+    pthread_mutex_t srfMutex;           /* For multithreading. Not used. */
+    pthread_cond_t bufferAvailable;     /* For multithreading. Not used. */
+    unsigned srfNo;                     /* XvMC private surface numbers */
+    unsigned numBuffers;                /* Number of picture buffers */
+    unsigned curBuf;                    /* Which is the current buffer? */
+    unsigned offsets[VIA_MAX_BUFS];     /* Offsets of picture buffers 
+					   into the frame-buffer area */
+    unsigned yStride;                   /* Stride of YUV420 Y component. */ 
+    unsigned width;                     /* Dimensions */
+    unsigned height;
+    int progressiveSequence;            /* Mpeg progressive picture? Hmm? */
+    ViaXvMCContext *privContext;        /* XvMC context private part. */
+    ViaXvMCSubPicture *privSubPic;      /* Subpicture to be blended when 
+					   displaying. NULL if none. */
+    int needsSync;
+    int syncMode;
+    CARD32 timeStamp;
+    int topFieldFirst;
+}ViaXvMCSurface;
+
+/*
+ * Take and release the global drm hardware lock.
+ */
+
+#define HW_LOCK(ctx)						\
+    DRM_LOCK((ctx)->fd,(ctx)->hwLock,(ctx)->drmcontext,0)     
+#define HW_UNLOCK(ctx)						\
+    DRM_UNLOCK((ctx)->fd,(ctx->hwLock),(ctx)->drmcontext) 
+
+/*
+ * Low-level Mpeg functions in viaLowLevel.c
+ */ 
+
+#define VIABLIT_TRANSCOPY 0
+#define VIABLIT_COPY 1
+#define VIABLIT_FILL 2
+
+
+#endif
diff --git a/nx-X11/lib/XvMC/hw/via/xf86dri.c b/nx-X11/lib/XvMC/hw/via/xf86dri.c
new file mode 100644
index 000000000..dd1d1e072
--- /dev/null
+++ b/nx-X11/lib/XvMC/hw/via/xf86dri.c
@@ -0,0 +1,592 @@
+/* $XFree86: xc/lib/GL/dri/XF86dri.c,v 1.13 2002/10/30 12:51:25 alanh Exp $ */
+/**************************************************************************
+
+Copyright 1998-1999 Precision Insight, Inc., Cedar Park, Texas.
+Copyright 2000 VA Linux Systems, Inc.
+All Rights Reserved.
+
+Permission is hereby granted, free of charge, to any person obtaining a
+copy of this software and associated documentation files (the
+"Software"), to deal in the Software without restriction, including
+without limitation the rights to use, copy, modify, merge, publish,
+distribute, sub license, and/or sell copies of the Software, and to
+permit persons to whom the Software is furnished to do so, subject to
+the following conditions:
+
+The above copyright notice and this permission notice (including the
+next paragraph) shall be included in all copies or substantial portions
+of the Software.
+
+THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+IN NO EVENT SHALL PRECISION INSIGHT AND/OR ITS SUPPLIERS BE LIABLE FOR
+ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+
+**************************************************************************/
+
+/*
+ * Authors:
+ *   Kevin E. Martin <martin@valinux.com>
+ *   Jens Owen <jens@tungstengraphics.com>
+ *   Rickard E. (Rik) Faith <faith@valinux.com>
+ *
+ */
+
+/* THIS IS NOT AN X CONSORTIUM STANDARD */
+
+#define NEED_REPLIES
+#include <X11/Xlibint.h>
+#include <X11/extensions/Xext.h>
+#include <X11/extensions/extutil.h>
+/* #include "glheader.h" */
+#include "xf86dristr.h"
+
+static XExtensionInfo _xf86dri_info_data;
+static XExtensionInfo *xf86dri_info = &_xf86dri_info_data;
+static char xf86dri_extension_name[] = XF86DRINAME;
+
+#define uniDRICheckExtension(dpy,i,val) \
+  XextCheckExtension (dpy, i, xf86dri_extension_name, val)
+
+/*****************************************************************************
+ *                                                                           *
+ *			   private utility routines                          *
+ *                                                                           *
+ *****************************************************************************/
+
+static int close_display(Display *dpy, XExtCodes *extCodes);
+static /* const */ XExtensionHooks xf86dri_extension_hooks = {
+    NULL,				/* create_gc */
+    NULL,				/* copy_gc */
+    NULL,				/* flush_gc */
+    NULL,				/* free_gc */
+    NULL,				/* create_font */
+    NULL,				/* free_font */
+    close_display,			/* close_display */
+    NULL,				/* wire_to_event */
+    NULL,				/* event_to_wire */
+    NULL,				/* error */
+    NULL,				/* error_string */
+};
+
+static XEXT_GENERATE_FIND_DISPLAY (find_display, xf86dri_info, 
+				   xf86dri_extension_name, 
+				   &xf86dri_extension_hooks, 
+				   0, NULL)
+
+static XEXT_GENERATE_CLOSE_DISPLAY (close_display, xf86dri_info)
+
+
+/*****************************************************************************
+ *                                                                           *
+ *		    public XFree86-DRI Extension routines                    *
+ *                                                                           *
+ *****************************************************************************/
+
+#if 0
+#include <stdio.h>
+#define TRACE(msg)  fprintf(stderr,"uniDRI%s\n", msg);
+#else
+#define TRACE(msg)
+#endif
+
+
+ Bool uniDRIQueryExtension (dpy, event_basep, error_basep)
+    Display *dpy;
+    int *event_basep, *error_basep;
+{
+    XExtDisplayInfo *info = find_display (dpy);
+
+    TRACE("QueryExtension...");
+    if (XextHasExtension(info)) {
+	*event_basep = info->codes->first_event;
+	*error_basep = info->codes->first_error;
+        TRACE("QueryExtension... return True");
+	return True;
+    } else {
+        TRACE("QueryExtension... return False");
+	return False;
+    }
+}
+
+ Bool uniDRIQueryVersion(dpy, majorVersion, minorVersion, patchVersion)
+    Display* dpy;
+    int* majorVersion; 
+    int* minorVersion;
+    int* patchVersion;
+{
+    XExtDisplayInfo *info = find_display (dpy);
+    xXF86DRIQueryVersionReply rep;
+    xXF86DRIQueryVersionReq *req;
+
+    TRACE("QueryVersion...");
+    uniDRICheckExtension (dpy, info, False);
+
+    LockDisplay(dpy);
+    GetReq(XF86DRIQueryVersion, req);
+    req->reqType = info->codes->major_opcode;
+    req->driReqType = X_XF86DRIQueryVersion;
+    if (!_XReply(dpy, (xReply *)&rep, 0, xFalse)) {
+	UnlockDisplay(dpy);
+	SyncHandle();
+        TRACE("QueryVersion... return False");
+	return False;
+    }
+    *majorVersion = rep.majorVersion;
+    *minorVersion = rep.minorVersion;
+    *patchVersion = rep.patchVersion;
+    UnlockDisplay(dpy);
+    SyncHandle();
+    TRACE("QueryVersion... return True");
+    return True;
+}
+
+ Bool uniDRIQueryDirectRenderingCapable(dpy, screen, isCapable)
+    Display* dpy;
+    int screen;
+    Bool* isCapable;
+{
+    XExtDisplayInfo *info = find_display (dpy);
+    xXF86DRIQueryDirectRenderingCapableReply rep;
+    xXF86DRIQueryDirectRenderingCapableReq *req;
+
+    TRACE("QueryDirectRenderingCapable...");
+    uniDRICheckExtension (dpy, info, False);
+
+    LockDisplay(dpy);
+    GetReq(XF86DRIQueryDirectRenderingCapable, req);
+    req->reqType = info->codes->major_opcode;
+    req->driReqType = X_XF86DRIQueryDirectRenderingCapable;
+    req->screen = screen;
+    if (!_XReply(dpy, (xReply *)&rep, 0, xFalse)) {
+	UnlockDisplay(dpy);
+	SyncHandle();
+        TRACE("QueryDirectRenderingCapable... return False");
+	return False;
+    }
+    *isCapable = rep.isCapable;
+    UnlockDisplay(dpy);
+    SyncHandle();
+    TRACE("QueryDirectRenderingCapable... return True");
+    return True;
+}
+
+ Bool uniDRIOpenConnection(dpy, screen, hSAREA, busIdString)
+    Display* dpy;
+    int screen;
+    drm_handle_t * hSAREA;
+    char **busIdString;
+{
+    XExtDisplayInfo *info = find_display (dpy);
+    xXF86DRIOpenConnectionReply rep;
+    xXF86DRIOpenConnectionReq *req;
+
+    TRACE("OpenConnection...");
+    uniDRICheckExtension (dpy, info, False);
+
+    LockDisplay(dpy);
+    GetReq(XF86DRIOpenConnection, req);
+    req->reqType = info->codes->major_opcode;
+    req->driReqType = X_XF86DRIOpenConnection;
+    req->screen = screen;
+    if (!_XReply(dpy, (xReply *)&rep, 0, xFalse)) {
+	UnlockDisplay(dpy);
+	SyncHandle();
+        TRACE("OpenConnection... return False");
+	return False;
+    }
+
+    *hSAREA = rep.hSAREALow;
+#ifdef LONG64
+    *hSAREA |= ((drm_handle_t)rep.hSAREAHigh) << 32;
+#endif
+
+    if (rep.length) {
+        if (!(*busIdString = (char *)Xcalloc(rep.busIdStringLength + 1, 1))) {
+            _XEatData(dpy, ((rep.busIdStringLength+3) & ~3));
+            UnlockDisplay(dpy);
+            SyncHandle();
+            TRACE("OpenConnection... return False");
+            return False;
+        }
+	_XReadPad(dpy, *busIdString, rep.busIdStringLength);
+    } else {
+        *busIdString = NULL;
+    }
+    UnlockDisplay(dpy);
+    SyncHandle();
+    TRACE("OpenConnection... return True");
+    return True;
+}
+
+ Bool uniDRIAuthConnection(dpy, screen, magic)
+    Display* dpy;
+    int screen;
+    drm_magic_t magic;
+{
+    XExtDisplayInfo *info = find_display (dpy);
+    xXF86DRIAuthConnectionReq *req;
+    xXF86DRIAuthConnectionReply rep;
+
+    TRACE("AuthConnection...");
+    uniDRICheckExtension (dpy, info, False);
+
+    LockDisplay(dpy);
+    GetReq(XF86DRIAuthConnection, req);
+    req->reqType = info->codes->major_opcode;
+    req->driReqType = X_XF86DRIAuthConnection;
+    req->screen = screen;
+    req->magic = magic;
+    rep.authenticated = 0;
+    if (!_XReply(dpy, (xReply *)&rep, 0, xFalse) || !rep.authenticated) {
+	UnlockDisplay(dpy);
+	SyncHandle();
+        TRACE("AuthConnection... return False");
+	return False;
+    }
+    UnlockDisplay(dpy);
+    SyncHandle();
+    TRACE("AuthConnection... return True");
+    return True;
+}
+
+ Bool uniDRICloseConnection(dpy, screen)
+    Display* dpy;
+    int screen;
+{
+    XExtDisplayInfo *info = find_display (dpy);
+    xXF86DRICloseConnectionReq *req;
+
+    TRACE("CloseConnection...");
+
+    uniDRICheckExtension (dpy, info, False);
+
+    LockDisplay(dpy);
+    GetReq(XF86DRICloseConnection, req);
+    req->reqType = info->codes->major_opcode;
+    req->driReqType = X_XF86DRICloseConnection;
+    req->screen = screen;
+    UnlockDisplay(dpy);
+    SyncHandle();
+    TRACE("CloseConnection... return True");
+    return True;
+}
+
+ Bool uniDRIGetClientDriverName(dpy, screen, ddxDriverMajorVersion, 
+	ddxDriverMinorVersion, ddxDriverPatchVersion, clientDriverName)
+    Display* dpy;
+    int screen;
+    int* ddxDriverMajorVersion;
+    int* ddxDriverMinorVersion;
+    int* ddxDriverPatchVersion;
+    char** clientDriverName;
+{
+    XExtDisplayInfo *info = find_display (dpy);
+    xXF86DRIGetClientDriverNameReply rep;
+    xXF86DRIGetClientDriverNameReq *req;
+
+    TRACE("GetClientDriverName...");
+    uniDRICheckExtension (dpy, info, False);
+
+    LockDisplay(dpy);
+    GetReq(XF86DRIGetClientDriverName, req);
+    req->reqType = info->codes->major_opcode;
+    req->driReqType = X_XF86DRIGetClientDriverName;
+    req->screen = screen;
+    if (!_XReply(dpy, (xReply *)&rep, 0, xFalse)) {
+	UnlockDisplay(dpy);
+	SyncHandle();
+        TRACE("GetClientDriverName... return False");
+	return False;
+    }
+
+    *ddxDriverMajorVersion = rep.ddxDriverMajorVersion;
+    *ddxDriverMinorVersion = rep.ddxDriverMinorVersion;
+    *ddxDriverPatchVersion = rep.ddxDriverPatchVersion;
+
+    if (rep.length) {
+        if (!(*clientDriverName = (char *)Xcalloc(rep.clientDriverNameLength + 1, 1))) {
+            _XEatData(dpy, ((rep.clientDriverNameLength+3) & ~3));
+            UnlockDisplay(dpy);
+            SyncHandle();
+            TRACE("GetClientDriverName... return False");
+            return False;
+        }
+	_XReadPad(dpy, *clientDriverName, rep.clientDriverNameLength);
+    } else {
+        *clientDriverName = NULL;
+    }
+    UnlockDisplay(dpy);
+    SyncHandle();
+    TRACE("GetClientDriverName... return True");
+    return True;
+}
+
+ Bool uniDRICreateContextWithConfig(dpy, screen, configID, context,
+	hHWContext)
+    Display* dpy;
+    int screen;
+    int configID;
+    XID* context;
+    drm_context_t * hHWContext;
+{
+    XExtDisplayInfo *info = find_display (dpy);
+    xXF86DRICreateContextReply rep;
+    xXF86DRICreateContextReq *req;
+
+    TRACE("CreateContext...");
+    uniDRICheckExtension (dpy, info, False);
+
+    LockDisplay(dpy);
+    GetReq(XF86DRICreateContext, req);
+    req->reqType = info->codes->major_opcode;
+    req->driReqType = X_XF86DRICreateContext;
+    req->visual = configID;
+    req->screen = screen;
+    *context = XAllocID(dpy);
+    req->context = *context;
+    if (!_XReply(dpy, (xReply *)&rep, 0, xFalse)) {
+	UnlockDisplay(dpy);
+	SyncHandle();
+        TRACE("CreateContext... return False");
+	return False;
+    }
+    *hHWContext = rep.hHWContext;
+    UnlockDisplay(dpy);
+    SyncHandle();
+    TRACE("CreateContext... return True");
+    return True;
+}
+
+ Bool uniDRICreateContext(dpy, screen, visual, context, hHWContext)
+    Display* dpy;
+    int screen;
+    Visual* visual;
+    XID* context;
+    drm_context_t * hHWContext;
+{
+    return uniDRICreateContextWithConfig( dpy, screen, visual->visualid,
+					   context, hHWContext );
+}
+
+ Bool uniDRIDestroyContext( Display * ndpy, int screen, 
+    XID context )
+{
+    Display * const dpy = (Display *) ndpy;
+    XExtDisplayInfo *info = find_display (dpy);
+    xXF86DRIDestroyContextReq *req;
+
+    TRACE("DestroyContext...");
+    uniDRICheckExtension (dpy, info, False);
+
+    LockDisplay(dpy);
+    GetReq(XF86DRIDestroyContext, req);
+    req->reqType = info->codes->major_opcode;
+    req->driReqType = X_XF86DRIDestroyContext;
+    req->screen = screen;
+    req->context = context;
+    UnlockDisplay(dpy);
+    SyncHandle();
+    TRACE("DestroyContext... return True");
+    return True;
+}
+
+Bool uniDRICreateDrawable( Display * ndpy, int screen, 
+			    Drawable drawable, drm_drawable_t * hHWDrawable )
+{
+    Display * const dpy = (Display *) ndpy;
+    XExtDisplayInfo *info = find_display (dpy);
+    xXF86DRICreateDrawableReply rep;
+    xXF86DRICreateDrawableReq *req;
+
+    TRACE("CreateDrawable...");
+    uniDRICheckExtension (dpy, info, False);
+
+    LockDisplay(dpy);
+    GetReq(XF86DRICreateDrawable, req);
+    req->reqType = info->codes->major_opcode;
+    req->driReqType = X_XF86DRICreateDrawable;
+    req->screen = screen;
+    req->drawable = drawable;
+    if (!_XReply(dpy, (xReply *)&rep, 0, xFalse)) {
+	UnlockDisplay(dpy);
+	SyncHandle();
+        TRACE("CreateDrawable... return False");
+	return False;
+    }
+    *hHWDrawable = rep.hHWDrawable;
+    UnlockDisplay(dpy);
+    SyncHandle();
+    TRACE("CreateDrawable... return True");
+    return True;
+}
+
+Bool uniDRIDestroyDrawable( Display * ndpy, int screen,
+			     Drawable drawable )
+{
+    Display * const dpy = (Display *) ndpy;
+    XExtDisplayInfo *info = find_display (dpy);
+    xXF86DRIDestroyDrawableReq *req;
+
+    TRACE("DestroyDrawable...");
+    uniDRICheckExtension (dpy, info, False);
+
+    LockDisplay(dpy);
+    GetReq(XF86DRIDestroyDrawable, req);
+    req->reqType = info->codes->major_opcode;
+    req->driReqType = X_XF86DRIDestroyDrawable;
+    req->screen = screen;
+    req->drawable = drawable;
+    UnlockDisplay(dpy);
+    SyncHandle();
+    TRACE("DestroyDrawable... return True");
+    return True;
+}
+
+ Bool uniDRIGetDrawableInfo(Display* dpy, int screen, Drawable drawable,
+    unsigned int* index, unsigned int* stamp,
+    int* X, int* Y, int* W, int* H,
+    int* numClipRects, drm_clip_rect_t ** pClipRects,
+    int* backX, int* backY,
+    int* numBackClipRects, drm_clip_rect_t ** pBackClipRects )
+{
+    XExtDisplayInfo *info = find_display (dpy);
+    xXF86DRIGetDrawableInfoReply rep;
+    xXF86DRIGetDrawableInfoReq *req;
+    int total_rects;
+
+    TRACE("GetDrawableInfo...");
+    uniDRICheckExtension (dpy, info, False);
+
+    LockDisplay(dpy);
+    GetReq(XF86DRIGetDrawableInfo, req);
+    req->reqType = info->codes->major_opcode;
+    req->driReqType = X_XF86DRIGetDrawableInfo;
+    req->screen = screen;
+    req->drawable = drawable;
+
+    if (!_XReply(dpy, (xReply *)&rep, 1, xFalse)) 
+    {
+	UnlockDisplay(dpy);
+	SyncHandle();
+        TRACE("GetDrawableInfo... return False");
+	return False;
+    }
+    *index = rep.drawableTableIndex;
+    *stamp = rep.drawableTableStamp;
+    *X = (int)rep.drawableX;
+    *Y = (int)rep.drawableY;
+    *W = (int)rep.drawableWidth;
+    *H = (int)rep.drawableHeight;
+    *numClipRects = rep.numClipRects;
+    total_rects = *numClipRects;
+
+    *backX = rep.backX;
+    *backY = rep.backY;
+    *numBackClipRects = rep.numBackClipRects;
+    total_rects += *numBackClipRects;
+
+#if 0
+    /* Because of the fix in Xserver/GL/dri/xf86dri.c, this check breaks
+     * backwards compatibility (Because of the >> 2 shift) but the fix
+     * enables multi-threaded apps to work.
+     */
+    if (rep.length !=  ((((SIZEOF(xXF86DRIGetDrawableInfoReply) - 
+		       SIZEOF(xGenericReply) + 
+		       total_rects * sizeof(drm_clip_rect_t)) + 3) & ~3) >> 2)) {
+        _XEatData(dpy, rep.length);
+	UnlockDisplay(dpy);
+	SyncHandle();
+        TRACE("GetDrawableInfo... return False");
+        return False;
+    }
+#endif
+
+    if (*numClipRects) {
+       int len = sizeof(drm_clip_rect_t) * (*numClipRects);
+
+       *pClipRects = (drm_clip_rect_t *)Xcalloc(len, 1);
+       if (*pClipRects) 
+	  _XRead(dpy, (char*)*pClipRects, len);
+    } else {
+        *pClipRects = NULL;
+    }
+
+    if (*numBackClipRects) {
+       int len = sizeof(drm_clip_rect_t) * (*numBackClipRects);
+
+       *pBackClipRects = (drm_clip_rect_t *)Xcalloc(len, 1);
+       if (*pBackClipRects) 
+	  _XRead(dpy, (char*)*pBackClipRects, len);
+    } else {
+        *pBackClipRects = NULL;
+    }
+
+    UnlockDisplay(dpy);
+    SyncHandle();
+    TRACE("GetDrawableInfo... return True");
+    return True;
+}
+
+ Bool uniDRIGetDeviceInfo(dpy, screen, hFrameBuffer, 
+	fbOrigin, fbSize, fbStride, devPrivateSize, pDevPrivate)
+    Display* dpy;
+    int screen;
+    drm_handle_t * hFrameBuffer;
+    int* fbOrigin;
+    int* fbSize;
+    int* fbStride;
+    int* devPrivateSize;
+    void** pDevPrivate;
+{
+    XExtDisplayInfo *info = find_display (dpy);
+    xXF86DRIGetDeviceInfoReply rep;
+    xXF86DRIGetDeviceInfoReq *req;
+
+    TRACE("GetDeviceInfo...");
+    uniDRICheckExtension (dpy, info, False);
+
+    LockDisplay(dpy);
+    GetReq(XF86DRIGetDeviceInfo, req);
+    req->reqType = info->codes->major_opcode;
+    req->driReqType = X_XF86DRIGetDeviceInfo;
+    req->screen = screen;
+    if (!_XReply(dpy, (xReply *)&rep, 0, xFalse)) {
+	UnlockDisplay(dpy);
+	SyncHandle();
+        TRACE("GetDeviceInfo... return False");
+	return False;
+    }
+
+    *hFrameBuffer = rep.hFrameBufferLow;
+#ifdef LONG64
+    *hFrameBuffer |= ((drm_handle_t)rep.hFrameBufferHigh) << 32;
+#endif
+
+    *fbOrigin = rep.framebufferOrigin;
+    *fbSize = rep.framebufferSize;
+    *fbStride = rep.framebufferStride;
+    *devPrivateSize = rep.devPrivateSize;
+
+    if (rep.length) {
+        if (!(*pDevPrivate = (void *)Xcalloc(rep.devPrivateSize, 1))) {
+            _XEatData(dpy, ((rep.devPrivateSize+3) & ~3));
+            UnlockDisplay(dpy);
+            SyncHandle();
+            TRACE("GetDeviceInfo... return False");
+            return False;
+        }
+	_XRead(dpy, (char*)*pDevPrivate, rep.devPrivateSize);
+    } else {
+        *pDevPrivate = NULL;
+    }
+
+    UnlockDisplay(dpy);
+    SyncHandle();
+    TRACE("GetDeviceInfo... return True");
+    return True;
+}
+
diff --git a/nx-X11/lib/XvMC/hw/via/xf86dri.h b/nx-X11/lib/XvMC/hw/via/xf86dri.h
new file mode 100644
index 000000000..f82b573fe
--- /dev/null
+++ b/nx-X11/lib/XvMC/hw/via/xf86dri.h
@@ -0,0 +1,121 @@
+/* $XFree86: xc/lib/GL/dri/xf86dri.h,v 1.8 2002/10/30 12:51:25 alanh Exp $ */
+/**************************************************************************
+
+Copyright 1998-1999 Precision Insight, Inc., Cedar Park, Texas.
+Copyright 2000 VA Linux Systems, Inc.
+All Rights Reserved.
+
+Permission is hereby granted, free of charge, to any person obtaining a
+copy of this software and associated documentation files (the
+"Software"), to deal in the Software without restriction, including
+without limitation the rights to use, copy, modify, merge, publish,
+distribute, sub license, and/or sell copies of the Software, and to
+permit persons to whom the Software is furnished to do so, subject to
+the following conditions:
+
+The above copyright notice and this permission notice (including the
+next paragraph) shall be included in all copies or substantial portions
+of the Software.
+
+THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+IN NO EVENT SHALL PRECISION INSIGHT AND/OR ITS SUPPLIERS BE LIABLE FOR
+ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+
+**************************************************************************/
+
+/**
+ * \file xf86dri.h
+ * Protocol numbers and function prototypes for DRI X protocol.
+ *
+ * \author Kevin E. Martin <martin@valinux.com>
+ * \author Jens Owen <jens@tungstengraphics.com>
+ * \author Rickard E. (Rik) Faith <faith@valinux.com>
+ */
+
+#ifndef _XF86DRI_H_
+#define _XF86DRI_H_
+
+#include <X11/Xfuncproto.h>
+#include <xf86drm.h>
+
+#define X_XF86DRIQueryVersion			0
+#define X_XF86DRIQueryDirectRenderingCapable	1
+#define X_XF86DRIOpenConnection			2
+#define X_XF86DRICloseConnection		3
+#define X_XF86DRIGetClientDriverName		4
+#define X_XF86DRICreateContext			5
+#define X_XF86DRIDestroyContext			6
+#define X_XF86DRICreateDrawable			7
+#define X_XF86DRIDestroyDrawable		8
+#define X_XF86DRIGetDrawableInfo		9
+#define X_XF86DRIGetDeviceInfo			10
+#define X_XF86DRIAuthConnection                 11
+#define X_XF86DRIOpenFullScreen                 12   /* Deprecated */
+#define X_XF86DRICloseFullScreen                13   /* Deprecated */
+
+#define XF86DRINumberEvents		0
+
+#define XF86DRIClientNotLocal		0
+#define XF86DRIOperationNotSupported	1
+#define XF86DRINumberErrors		(XF86DRIOperationNotSupported + 1)
+
+#ifndef _XF86DRI_SERVER_
+
+
+_XFUNCPROTOBEGIN
+
+Bool uniDRIQueryExtension( Display *dpy, int *event_base, int *error_base );
+
+Bool uniDRIQueryVersion( Display *dpy, int *majorVersion, int *minorVersion,
+			 int *patchVersion );
+
+Bool uniDRIQueryDirectRenderingCapable( Display *dpy, int screen,
+					Bool *isCapable );
+
+Bool uniDRIOpenConnection( Display *dpy, int screen, drm_handle_t *hSAREA,
+			   char **busIDString );
+
+Bool uniDRIAuthConnection( Display *dpy, int screen, drm_magic_t magic );
+
+Bool uniDRICloseConnection( Display *dpy, int screen );
+
+Bool uniDRIGetClientDriverName( Display *dpy, int screen,
+				int *ddxDriverMajorVersion, int *ddxDriverMinorVersion,
+				int *ddxDriverPatchVersion, char **clientDriverName );
+
+Bool uniDRICreateContext( Display *dpy, int screen, Visual *visual,
+			  XID *ptr_to_returned_context_id, drm_context_t *hHWContext );
+
+Bool uniDRICreateContextWithConfig( Display *dpy, int screen, int configID,
+				    XID *ptr_to_returned_context_id, drm_context_t *hHWContext );
+
+extern Bool uniDRIDestroyContext( Display *dpy, int screen,
+				  XID context_id );
+
+extern Bool uniDRICreateDrawable( Display *dpy, int screen,
+				  Drawable drawable, drm_drawable_t *hHWDrawable );
+
+extern Bool uniDRIDestroyDrawable( Display *dpy, int screen, 
+				   Drawable drawable);
+
+Bool uniDRIGetDrawableInfo( Display *dpy, int screen, Drawable drawable,
+			    unsigned int *index, unsigned int *stamp, 
+			    int *X, int *Y, int *W, int *H,
+			    int *numClipRects, drm_clip_rect_t ** pClipRects,
+			    int *backX, int *backY,
+			    int *numBackClipRects, drm_clip_rect_t **pBackClipRects );
+
+Bool uniDRIGetDeviceInfo( Display *dpy, int screen,
+			  drm_handle_t *hFrameBuffer, int *fbOrigin, int *fbSize,
+			  int *fbStride, int *devPrivateSize, void **pDevPrivate );
+
+_XFUNCPROTOEND
+
+#endif /* _XF86DRI_SERVER_ */
+
+#endif /* _XF86DRI_H_ */
+
diff --git a/nx-X11/lib/XvMC/hw/via/xf86dristr.h b/nx-X11/lib/XvMC/hw/via/xf86dristr.h
new file mode 100644
index 000000000..ac05b183b
--- /dev/null
+++ b/nx-X11/lib/XvMC/hw/via/xf86dristr.h
@@ -0,0 +1,343 @@
+/* $XFree86: xc/lib/GL/dri/xf86dristr.h,v 1.10 2002/10/30 12:51:25 alanh Exp $ */
+/**************************************************************************
+
+Copyright 1998-1999 Precision Insight, Inc., Cedar Park, Texas.
+Copyright 2000 VA Linux Systems, Inc.
+All Rights Reserved.
+
+Permission is hereby granted, free of charge, to any person obtaining a
+copy of this software and associated documentation files (the
+"Software"), to deal in the Software without restriction, including
+without limitation the rights to use, copy, modify, merge, publish,
+distribute, sub license, and/or sell copies of the Software, and to
+permit persons to whom the Software is furnished to do so, subject to
+the following conditions:
+
+The above copyright notice and this permission notice (including the
+next paragraph) shall be included in all copies or substantial portions
+of the Software.
+
+THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+IN NO EVENT SHALL PRECISION INSIGHT AND/OR ITS SUPPLIERS BE LIABLE FOR
+ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+
+**************************************************************************/
+
+/*
+ * Authors:
+ *   Kevin E. Martin <martin@valinux.com>
+ *   Jens Owen <jens@tungstengraphics.com>
+ *   Rickard E. (Rik) Fiath <faith@valinux.com>
+ *
+ */
+
+#ifndef _XF86DRISTR_H_
+#define _XF86DRISTR_H_
+
+#include "xf86dri.h"
+
+#define XF86DRINAME "XFree86-DRI"
+
+/* The DRI version number.  This was originally set to be the same of the
+ * XFree86 version number.  However, this version is really indepedent of
+ * the XFree86 version.
+ *
+ * Version History:
+ *    4.0.0: Original
+ *    4.0.1: Patch to bump clipstamp when windows are destroyed, 28 May 02
+ *    4.1.0: Add transition from single to multi in DRMInfo rec, 24 Jun 02
+ */
+#define XF86DRI_MAJOR_VERSION	4
+#define XF86DRI_MINOR_VERSION	1
+#define XF86DRI_PATCH_VERSION	0
+
+typedef struct _XF86DRIQueryVersion {
+    CARD8	reqType;		/* always DRIReqCode */
+    CARD8	driReqType;		/* always X_DRIQueryVersion */
+    CARD16	length B16;
+} xXF86DRIQueryVersionReq;
+#define sz_xXF86DRIQueryVersionReq	4
+
+typedef struct {
+    BYTE	type;			/* X_Reply */
+    BOOL	pad1;
+    CARD16	sequenceNumber B16;
+    CARD32	length B32;
+    CARD16	majorVersion B16;	/* major version of DRI protocol */
+    CARD16	minorVersion B16;	/* minor version of DRI protocol */
+    CARD32	patchVersion B32;       /* patch version of DRI protocol */
+    CARD32	pad3 B32;
+    CARD32	pad4 B32;
+    CARD32	pad5 B32;
+    CARD32	pad6 B32;
+} xXF86DRIQueryVersionReply;
+#define sz_xXF86DRIQueryVersionReply	32
+
+typedef struct _XF86DRIQueryDirectRenderingCapable {
+    CARD8	reqType;		/* always DRIReqCode */
+    CARD8	driReqType;		/* X_DRIQueryDirectRenderingCapable */
+    CARD16	length B16;
+    CARD32	screen B32;
+} xXF86DRIQueryDirectRenderingCapableReq;
+#define sz_xXF86DRIQueryDirectRenderingCapableReq	8
+
+typedef struct {
+    BYTE	type;			/* X_Reply */
+    BOOL	pad1;
+    CARD16	sequenceNumber B16;
+    CARD32	length B32;
+    BOOL	isCapable;
+    BOOL	pad2;
+    BOOL	pad3;
+    BOOL	pad4;
+    CARD32	pad5 B32;
+    CARD32	pad6 B32;
+    CARD32	pad7 B32;
+    CARD32	pad8 B32;
+    CARD32	pad9 B32;
+} xXF86DRIQueryDirectRenderingCapableReply;
+#define sz_xXF86DRIQueryDirectRenderingCapableReply	32
+
+typedef struct _XF86DRIOpenConnection {
+    CARD8	reqType;		/* always DRIReqCode */
+    CARD8	driReqType;		/* always X_DRIOpenConnection */
+    CARD16	length B16;
+    CARD32	screen B32;
+} xXF86DRIOpenConnectionReq;
+#define sz_xXF86DRIOpenConnectionReq	8
+
+typedef struct {
+    BYTE	type;			/* X_Reply */
+    BOOL	pad1;
+    CARD16	sequenceNumber B16;
+    CARD32	length B32;
+    CARD32	hSAREALow B32;
+    CARD32	hSAREAHigh B32;
+    CARD32	busIdStringLength B32;
+    CARD32	pad6 B32;
+    CARD32	pad7 B32;
+    CARD32	pad8 B32;
+} xXF86DRIOpenConnectionReply;
+#define sz_xXF86DRIOpenConnectionReply	32
+
+typedef struct _XF86DRIAuthConnection {
+    CARD8	reqType;		/* always DRIReqCode */
+    CARD8	driReqType;		/* always X_DRICloseConnection */
+    CARD16	length B16;
+    CARD32	screen B32;
+    CARD32      magic B32;
+} xXF86DRIAuthConnectionReq;
+#define sz_xXF86DRIAuthConnectionReq	12
+
+typedef struct {
+    BYTE        type;
+    BOOL        pad1;
+    CARD16      sequenceNumber B16;
+    CARD32      length B32;
+    CARD32      authenticated B32;
+    CARD32      pad2 B32;
+    CARD32      pad3 B32;
+    CARD32      pad4 B32;
+    CARD32      pad5 B32;
+    CARD32      pad6 B32;
+} xXF86DRIAuthConnectionReply;
+#define zx_xXF86DRIAuthConnectionReply  32
+
+typedef struct _XF86DRICloseConnection {
+    CARD8	reqType;		/* always DRIReqCode */
+    CARD8	driReqType;		/* always X_DRICloseConnection */
+    CARD16	length B16;
+    CARD32	screen B32;
+} xXF86DRICloseConnectionReq;
+#define sz_xXF86DRICloseConnectionReq	8
+
+typedef struct _XF86DRIGetClientDriverName {
+    CARD8	reqType;		/* always DRIReqCode */
+    CARD8	driReqType;		/* always X_DRIGetClientDriverName */
+    CARD16	length B16;
+    CARD32	screen B32;
+} xXF86DRIGetClientDriverNameReq;
+#define sz_xXF86DRIGetClientDriverNameReq	8
+
+typedef struct {
+    BYTE	type;			/* X_Reply */
+    BOOL	pad1;
+    CARD16	sequenceNumber B16;
+    CARD32	length B32;
+    CARD32	ddxDriverMajorVersion B32;
+    CARD32	ddxDriverMinorVersion B32;
+    CARD32	ddxDriverPatchVersion B32;
+    CARD32	clientDriverNameLength B32;
+    CARD32	pad5 B32;
+    CARD32	pad6 B32;
+} xXF86DRIGetClientDriverNameReply;
+#define sz_xXF86DRIGetClientDriverNameReply	32
+
+typedef struct _XF86DRICreateContext {
+    CARD8	reqType;		/* always DRIReqCode */
+    CARD8	driReqType;		/* always X_DRICreateContext */
+    CARD16	length B16;
+    CARD32	screen B32;
+    CARD32	visual B32;
+    CARD32	context B32;
+} xXF86DRICreateContextReq;
+#define sz_xXF86DRICreateContextReq	16
+
+typedef struct {
+    BYTE	type;			/* X_Reply */
+    BOOL	pad1;
+    CARD16	sequenceNumber B16;
+    CARD32	length B32;
+    CARD32	hHWContext B32;
+    CARD32	pad2 B32;
+    CARD32	pad3 B32;
+    CARD32	pad4 B32;
+    CARD32	pad5 B32;
+    CARD32	pad6 B32;
+} xXF86DRICreateContextReply;
+#define sz_xXF86DRICreateContextReply	32
+
+typedef struct _XF86DRIDestroyContext {
+    CARD8	reqType;		/* always DRIReqCode */
+    CARD8	driReqType;		/* always X_DRIDestroyContext */
+    CARD16	length B16;
+    CARD32	screen B32;
+    CARD32	context B32;
+} xXF86DRIDestroyContextReq;
+#define sz_xXF86DRIDestroyContextReq	12
+
+typedef struct _XF86DRICreateDrawable {
+    CARD8	reqType;		/* always DRIReqCode */
+    CARD8	driReqType;		/* always X_DRICreateDrawable */
+    CARD16	length B16;
+    CARD32	screen B32;
+    CARD32	drawable B32;
+} xXF86DRICreateDrawableReq;
+#define sz_xXF86DRICreateDrawableReq	12
+
+typedef struct {
+    BYTE	type;			/* X_Reply */
+    BOOL	pad1;
+    CARD16	sequenceNumber B16;
+    CARD32	length B32;
+    CARD32	hHWDrawable B32;
+    CARD32	pad2 B32;
+    CARD32	pad3 B32;
+    CARD32	pad4 B32;
+    CARD32	pad5 B32;
+    CARD32	pad6 B32;
+} xXF86DRICreateDrawableReply;
+#define sz_xXF86DRICreateDrawableReply	32
+
+typedef struct _XF86DRIDestroyDrawable {
+    CARD8	reqType;		/* always DRIReqCode */
+    CARD8	driReqType;		/* always X_DRIDestroyDrawable */
+    CARD16	length B16;
+    CARD32	screen B32;
+    CARD32	drawable B32;
+} xXF86DRIDestroyDrawableReq;
+#define sz_xXF86DRIDestroyDrawableReq	12
+
+typedef struct _XF86DRIGetDrawableInfo {
+    CARD8	reqType;		/* always DRIReqCode */
+    CARD8	driReqType;		/* always X_DRIGetDrawableInfo */
+    CARD16	length B16;
+    CARD32	screen B32;
+    CARD32	drawable B32;
+} xXF86DRIGetDrawableInfoReq;
+#define sz_xXF86DRIGetDrawableInfoReq	12
+
+typedef struct {
+    BYTE	type;			/* X_Reply */
+    BOOL	pad1;
+    CARD16	sequenceNumber B16;
+    CARD32	length B32;
+    CARD32	drawableTableIndex B32;
+    CARD32	drawableTableStamp B32;
+    INT16	drawableX B16;
+    INT16	drawableY B16;
+    INT16	drawableWidth B16;
+    INT16	drawableHeight B16;
+    CARD32	numClipRects B32;
+    INT16       backX B16;
+    INT16       backY B16;
+    CARD32      numBackClipRects B32;
+} xXF86DRIGetDrawableInfoReply;
+
+#define sz_xXF86DRIGetDrawableInfoReply	36
+
+
+typedef struct _XF86DRIGetDeviceInfo {
+    CARD8	reqType;		/* always DRIReqCode */
+    CARD8	driReqType;		/* always X_DRIGetDeviceInfo */
+    CARD16	length B16;
+    CARD32	screen B32;
+} xXF86DRIGetDeviceInfoReq;
+#define sz_xXF86DRIGetDeviceInfoReq	8
+
+typedef struct {
+    BYTE	type;			/* X_Reply */
+    BOOL	pad1;
+    CARD16	sequenceNumber B16;
+    CARD32	length B32;
+    CARD32	hFrameBufferLow B32;
+    CARD32	hFrameBufferHigh B32;
+    CARD32	framebufferOrigin B32;
+    CARD32	framebufferSize B32;
+    CARD32	framebufferStride B32;
+    CARD32	devPrivateSize B32;
+} xXF86DRIGetDeviceInfoReply;
+#define sz_xXF86DRIGetDeviceInfoReply	32
+
+typedef struct _XF86DRIOpenFullScreen {
+    CARD8       reqType;	/* always DRIReqCode */
+    CARD8       driReqType;	/* always X_DRIOpenFullScreen */
+    CARD16      length B16;
+    CARD32      screen B32;
+    CARD32      drawable B32;
+} xXF86DRIOpenFullScreenReq;
+#define sz_xXF86DRIOpenFullScreenReq    12
+
+typedef struct {
+    BYTE        type;
+    BOOL        pad1;
+    CARD16      sequenceNumber B16;
+    CARD32      length B32;
+    CARD32      isFullScreen B32;
+    CARD32      pad2 B32;
+    CARD32      pad3 B32;
+    CARD32      pad4 B32;
+    CARD32      pad5 B32;
+    CARD32      pad6 B32;
+} xXF86DRIOpenFullScreenReply;
+#define sz_xXF86DRIOpenFullScreenReply  32
+
+typedef struct _XF86DRICloseFullScreen {
+    CARD8       reqType;	/* always DRIReqCode */
+    CARD8       driReqType;	/* always X_DRICloseFullScreen */
+    CARD16      length B16;
+    CARD32      screen B32;
+    CARD32      drawable B32;
+} xXF86DRICloseFullScreenReq;
+#define sz_xXF86DRICloseFullScreenReq   12
+
+typedef struct {
+    BYTE        type;
+    BOOL        pad1;
+    CARD16      sequenceNumber B16;
+    CARD32      length B32;
+    CARD32      pad2 B32;
+    CARD32      pad3 B32;
+    CARD32      pad4 B32;
+    CARD32      pad5 B32;
+    CARD32      pad6 B32;
+    CARD32      pad7 B32;
+} xXF86DRICloseFullScreenReply;
+#define sz_xXF86DRICloseFullScreenReply  32
+
+
+#endif /* _XF86DRISTR_H_ */