annotate libvo/vo_tdfx_vid.c @ 14742:76d461a061df

Unified colorkey code for vo xv and vo xvmc. Made the code also more flexible. Colorkey drawing is now by default done as proposed by Marko Macek. Patch also approved by iive.

/* 
 *  vo_tdfx_vid.c
 *
 *	Copyright (C) Alban Bedel - 03/2003
 *
 *  This file is part of MPlayer, a free movie player.
 *	
 *  MPlayer is free software; you can redistribute it and/or modify
 *  it under the terms of the GNU General Public License as published by
 *  the Free Software Foundation; either version 2, or (at your option)
 *  any later version.
 *   
 *  MPlayer is distributed in the hope that it will be useful,
 *  but WITHOUT ANY WARRANTY; without even the implied warranty of
 *  MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
 *  GNU General Public License for more details.
 *   
 *  You should have received a copy of the GNU General Public License
 *  along with GNU Make; see the file COPYING.  If not, write to
 *  the Free Software Foundation, 675 Mass Ave, Cambridge, MA 02139, USA. 
 *
 */

#include <stdio.h>
#include <stdlib.h>
#include <string.h>
#include <errno.h>
#include <sys/types.h>
#include <sys/stat.h>
#include <fcntl.h>
#include <unistd.h>
#include <sys/ioctl.h>
#include <sys/mman.h>
#include "config.h"
#include "video_out.h"
#include "video_out_internal.h"
#include "aspect.h"
#include "mp_msg.h"

#include "fastmemcpy.h"
#include "drivers/tdfx_vid.h"


static vo_info_t info = 
{
	"tdfx vid",
	"tdfx_vid",
	"Albeu",
	""
};

//#define VERBOSE

LIBVO_EXTERN(tdfx_vid)

static tdfx_vid_config_t tdfx_cfg;

static unsigned char* agp_mem = NULL;
static int tdfx_fd = -1;

static uint32_t img_fmt; // The real input format
static uint32_t src_width, src_height, src_fmt, src_bpp, src_stride;
static uint32_t dst_width, dst_height, dst_fmt, dst_bpp, dst_stride;

static uint32_t tdfx_page;
static uint32_t front_buffer;
static uint32_t back_buffer;
static uint8_t num_buffer = 3;
static uint32_t buffer_size; // Max size
static uint8_t current_buffer = 0;
static uint8_t current_ip_buf = 0;
static uint32_t buffer_stride[3];

static int use_overlay = 1;
static tdfx_vid_overlay_t tdfx_ov;

// FIXME
static void clear_screen(void) {
  tdfx_vid_agp_move_t mov;

  memset(agp_mem,0,tdfx_cfg.screen_width*dst_bpp*tdfx_cfg.screen_height);
  
  mov.move2 = TDFX_VID_MOVE_2_PACKED;
  mov.width = tdfx_cfg.screen_width*dst_bpp;
  mov.height = tdfx_cfg.screen_height;
  mov.src = 0;
  mov.src_stride = tdfx_cfg.screen_width*dst_bpp;
  mov.dst = front_buffer;
  mov.dst_stride = tdfx_cfg.screen_stride;

  printf("Move %d(%d) x %d => %d \n", mov.width,mov.src_stride,mov.height,mov.dst_stride);

  if(ioctl(tdfx_fd,TDFX_VID_AGP_MOVE,&mov))
    printf("tdfx_vid: AGP move failed to clear the screen\n");
  
}

static uint32_t draw_slice(uint8_t *image[], int stride[], int w,int h,int x,int y)
{
  uint8_t* ptr[3];

#ifdef VERBOSE
  printf("Draw slices %d\n",current_buffer);
#endif
  switch(img_fmt) {
  case IMGFMT_YUY2:
  case IMGFMT_UYVY:
  case IMGFMT_BGR8:
  case IMGFMT_BGR16:
  case IMGFMT_BGR24:
  case IMGFMT_BGR32:
    // copy :( to agp_mem
    // still faster than tdfxfb wich directly copy to the video mem :)
    mem2agpcpy_pic(agp_mem + current_buffer * buffer_size + 
	       y*buffer_stride[0] + x * src_bpp,
	       image[0],
	       src_bpp*w,h,buffer_stride[0],stride[0]);
    return 0;

  case IMGFMT_YV12:
  case IMGFMT_I420:
    // Copy to agp mem
    ptr[0] = agp_mem + current_buffer * buffer_size;
    mem2agpcpy_pic(ptr[0] + y * buffer_stride[0] + x,image[0],w,h,
		   buffer_stride[0],stride[0]);
    ptr[1] = ptr[0] + (src_height*src_width);
    mem2agpcpy_pic(ptr[1] + y/2 * buffer_stride[1] + x/2,image[1],w/2,h/2,
		   buffer_stride[1],stride[1]);
    ptr[2] = ptr[1] + (src_height*src_width/4);
    mem2agpcpy_pic(ptr[2] + y/2 * buffer_stride[2] + x/2,image[2],w/2,h/2,
		   buffer_stride[2],stride[2]);
    return 0;
  }
    
  return 1;
}

static void draw_osd(void)
{
}

static void
flip_page(void)
{
  tdfx_vid_blit_t blit;
  //return;
  // Scale convert
#ifdef VERBOSE
  printf("Flip\n");
#endif
  if(use_overlay) {
    // TDFX_VID_OVERLAY_ON does nothing if the overlay is alredy on
    if(!ioctl(tdfx_fd,TDFX_VID_OVERLAY_ON)) { // X11 killed the overlay :(
      if(ioctl(tdfx_fd,TDFX_VID_SET_OVERLAY,&tdfx_ov))
	mp_msg(MSGT_VO, MSGL_ERR, "tdfx_vid: set_overlay failed\n");
    }
    // These formats need conversion
    switch(src_fmt) {
    case IMGFMT_BGR8:
    case IMGFMT_BGR24:
    case IMGFMT_BGR32:
      memset(&blit,0,sizeof(tdfx_vid_blit_t));
      blit.src = back_buffer;
      blit.src_stride = src_stride;
      blit.src_x = 0;
      blit.src_y = 0;
      blit.src_w = src_width;
      blit.src_h = src_height;
      blit.src_format = src_fmt;
      
      blit.dst = front_buffer;
      blit.dst_stride = dst_stride;
      blit.dst_x = 0;
      blit.dst_y = 0;
      blit.dst_w = src_width;
      blit.dst_h = src_height;
      blit.dst_format = IMGFMT_BGR16;
      if(ioctl(tdfx_fd,TDFX_VID_BLIT,&blit))
	printf("tdfx_vid: Blit failed\n");
    }
    return;
  }
  memset(&blit,0,sizeof(tdfx_vid_blit_t));
  blit.src = back_buffer;
  blit.src_stride = src_stride;
  blit.src_x = 0;
  blit.src_y = 0;
  blit.src_w = src_width;
  blit.src_h = src_height;
  blit.src_format = src_fmt;

  blit.dst = front_buffer;
  blit.dst_stride = dst_stride;
  blit.dst_x = 0;
  blit.dst_y = 0;
  blit.dst_w = dst_width;
  blit.dst_h = dst_height;
  blit.dst_format = dst_fmt;

  if(ioctl(tdfx_fd,TDFX_VID_BLIT,&blit))
    printf("tdfx_vid: Blit failed\n");
}

static uint32_t
draw_frame(uint8_t *src[])
{
  int stride[] = { src_stride, 0, 0};
  return draw_slice(src,stride,src_width, src_height,0,0);
}

static uint32_t
query_format(uint32_t format)
{
  switch(format) {
  case IMGFMT_BGR8:
    if(tdfx_cfg.screen_format == TDFX_VID_FORMAT_BGR8)
      return 3 | VFCAP_HWSCALE_UP | VFCAP_HWSCALE_DOWN | VFCAP_ACCEPT_STRIDE;
    return 0;
  case IMGFMT_YUY2:
  case IMGFMT_UYVY:
  case IMGFMT_BGR15:
  case IMGFMT_BGR16:
  case IMGFMT_BGR24:
  case IMGFMT_BGR32:
  case IMGFMT_YV12:
  case IMGFMT_I420:
    if(tdfx_cfg.screen_format == TDFX_VID_FORMAT_BGR8)
      return 0;
    return 3 | VFCAP_HWSCALE_UP | VFCAP_HWSCALE_DOWN | VFCAP_ACCEPT_STRIDE;
  }
  return 0;
}

static uint32_t
config(uint32_t width, uint32_t height, uint32_t d_width, uint32_t d_height, uint32_t fullscreen, char *title, uint32_t format)
{

  if(tdfx_fd < 0)
    return 1;
  // When we are run as sub vo we must follow the size gaven to us
  if(!(fullscreen & VOFLAG_XOVERLAY_SUB_VO)) {
    if(!vo_screenwidth)
      vo_screenwidth = tdfx_cfg.screen_width;
    if(!vo_screenheight)
      vo_screenheight = tdfx_cfg.screen_height;
    
    aspect_save_orig(width,height);
    aspect_save_prescale(d_width,d_height);
    aspect_save_screenres(vo_screenwidth,vo_screenheight);
    
    if(fullscreen&0x01) { /* -fs */
      aspect(&d_width,&d_height,A_ZOOM);
      vo_fs = VO_TRUE;
    } else {
      aspect(&d_width,&d_height,A_NOZOOM);
      vo_fs = VO_FALSE;
    }
  }
  src_width = width;
  src_height = height;
  buffer_size = 0;
  buffer_stride[0] = 0;
  src_fmt = 0;
  switch(format) {
  case IMGFMT_BGR8:
  case IMGFMT_BGR24:
  case IMGFMT_BGR32:
    if(use_overlay)
      printf("tdfx_vid: Non-native overlay format need conversion\n");
  case IMGFMT_BGR15:
  case IMGFMT_BGR16:
    src_bpp = ((format & 0x3F)+7)/8; 
    break;
  case IMGFMT_YV12:
  case IMGFMT_I420:
    buffer_size = src_width * src_height * 3 / 2;
    buffer_stride[0] = ((src_width+1)/2)*2;
    buffer_stride[1] = buffer_stride[2] = buffer_stride[0]/2;
    src_fmt = TDFX_VID_FORMAT_YUY2;
  case IMGFMT_YUY2:
  case IMGFMT_UYVY: 
    src_bpp = 2;
    break;
  default:
    printf("Unsupported input format 0x%x\n",format);
    return 1;
  }

  img_fmt = format;
  src_stride = src_width*src_bpp;
  // The overlay need a 4 bytes aligned stride
  if(use_overlay)
    src_stride = ((src_stride+3)/4)*4;
  if(!src_fmt)
    src_fmt = format;
  if(!buffer_size)
    buffer_size = src_stride*src_height;
  if(!buffer_stride[0])
    buffer_stride[0] = src_stride;

  dst_fmt = tdfx_cfg.screen_format;
  dst_bpp = ((dst_fmt & 0x3F)+7)/8;
  dst_width = d_width;
  dst_height = d_height;
  dst_stride = tdfx_cfg.screen_stride;

  tdfx_page =  tdfx_cfg.screen_stride*tdfx_cfg.screen_height;
  front_buffer = tdfx_cfg.screen_start;
  back_buffer = front_buffer + tdfx_page;

  while(use_overlay) {
    tdfx_vid_overlay_t ov;
    uint32_t ov_fmt = src_fmt, ov_stride = src_stride;
    // Align the buffer
    back_buffer = (((back_buffer+3)/4)*4);
    // With the overlay the front buffer is not on the screen
    // so we take the back buffer
    front_buffer = back_buffer;
    switch(src_fmt) {
    case IMGFMT_BGR8:
    case IMGFMT_BGR24:
    case IMGFMT_BGR32:
      back_buffer = front_buffer + 2*(src_stride*src_height);
      ov_stride = dst_stride = src_width<<1;
      ov_fmt = IMGFMT_BGR16;
      break;
    }
    ov.src[0] = front_buffer;
    ov.src[1] = front_buffer + (src_stride*src_height);
    ov.src_width = src_width;
    ov.src_height = src_height;
    ov.src_stride = ov_stride;
    ov.format = ov_fmt;
    ov.dst_width = dst_width;
    ov.dst_height = dst_height;
    ov.dst_x = vo_dx;
    ov.dst_y = vo_dy;
    ov.use_colorkey = 0;

    if(ioctl(tdfx_fd,TDFX_VID_SET_OVERLAY,&ov)) {
      printf("tdfxvid: Overlay setup failed\n");
      use_overlay = 0;
      break;
    }
    tdfx_ov = ov;
    if(use_overlay == 1) {
      if(ioctl(tdfx_fd,TDFX_VID_OVERLAY_ON)) {
	printf("tdfx_vid: Overlay on failed\n");
	use_overlay = 0;
	break;
      }
      use_overlay++;
    }
    
    printf("tdfx_vid: Overlay ready : %d(%d) x %d @ %d => %d(%d) x %d @ %d\n",
	   src_width,src_stride,src_height,src_bpp,
	   dst_width,dst_stride,dst_height,dst_bpp);
    break;
  }

  if(!use_overlay)
    printf("tdfx_vid : Texture blit ready %d(%d) x %d @ %d => %d(%d) x %d @ %d\n",
	   src_width,src_stride,src_height,src_bpp,
	   dst_width,dst_stride,dst_height,dst_bpp);
  
  return 0;
}

static void
uninit(void)
{
  if(use_overlay == 2) {
    if(ioctl(tdfx_fd,TDFX_VID_OVERLAY_OFF))
      printf("tdfx_vid: Overlay off failed\n");
    use_overlay--;
  }
  close(tdfx_fd);
  tdfx_fd = -1;
}


static void check_events(void)
{
}

static uint32_t preinit(const char *arg)
{
 
  tdfx_fd = open(arg ? arg : "/dev/tdfx_vid", O_RDWR);
  if(tdfx_fd < 0) {
    printf("tdfx_vid: Can't open %s: %s\n",arg ? arg : "/dev/tdfx_vid",
	   strerror(errno));
    return 1;
  }

  if(ioctl(tdfx_fd,TDFX_VID_GET_CONFIG,&tdfx_cfg)) {
    printf("tdfx_vid: Can't get current cfg: %s\n",strerror(errno));
    return 1;
  }

  printf("tdfx_vid version %d\n"
	 "  Ram: %d\n"
	 "  Screen: %d x %d\n"
	 "  Format: %c%c%c%d\n",
	 tdfx_cfg.version,
	 tdfx_cfg.ram_size,
	 tdfx_cfg.screen_width, tdfx_cfg.screen_height,
	 tdfx_cfg.screen_format>>24,(tdfx_cfg.screen_format>>16)&0xFF,
	 (tdfx_cfg.screen_format>>8)&0xFF,tdfx_cfg.screen_format&0xFF);

  // For now just allocate more than i ever need
  agp_mem = mmap( NULL, 1024*768*4, PROT_READ | PROT_WRITE, MAP_SHARED,
		  tdfx_fd, 0);

  if(agp_mem == MAP_FAILED) {
    printf("tdfx_vid: Memmap failed !!!!!\n");
    return 1;
  }
  
  memset(agp_mem,0,1024*768*4);

  return 0;
}

static uint32_t get_image(mp_image_t *mpi) {
  int buf = 0;

#ifdef VERBOSE
  printf("Get image %d\n",buf);
#endif

  // Currently read are too slow bcs we read from the
  // agp aperture and not the memory directly
  //if(mpi->flags&MP_IMGFLAG_READABLE) return VO_FALSE;

  if(mpi->flags&MP_IMGFLAG_READABLE &&
     (mpi->type==MP_IMGTYPE_IPB || mpi->type==MP_IMGTYPE_IP)){
    // reference (I/P) frame of IP or IPB:
    if(num_buffer<2) return VO_FALSE; // not enough
    current_ip_buf^=1;
    // for IPB with 2 buffers we can DR only one of the 2 P frames:
    if(mpi->type==MP_IMGTYPE_IPB && num_buffer<3 && current_ip_buf) return VO_FALSE;
    buf=current_ip_buf;
    if(mpi->type==MP_IMGTYPE_IPB) ++buf; // preserve space for B
  }

  switch(mpi->imgfmt) {
  case IMGFMT_YUY2:
  case IMGFMT_UYVY:
  case IMGFMT_BGR8:
  case IMGFMT_BGR15:
  case IMGFMT_BGR16:
  case IMGFMT_BGR24:
  case IMGFMT_BGR32:
    mpi->planes[0] = agp_mem + buf * buffer_size;
    mpi->stride[0] = src_stride;
    break;
  case IMGFMT_YV12:
  case IMGFMT_I420:
    mpi->planes[0] = agp_mem + buf * buffer_size;
    mpi->stride[0] = mpi->width;
    mpi->planes[1] = mpi->planes[0] + mpi->stride[0]*mpi->height;
    mpi->stride[1] = mpi->chroma_width;
    mpi->planes[2] = mpi->planes[1] + mpi->stride[1]*mpi->chroma_height;
    mpi->stride[2] = mpi->chroma_width;
    break;
  default:
    printf("Get image todo\n");
    return VO_FALSE;
  }
  mpi->flags |= MP_IMGFLAG_DIRECT;
  mpi->priv = (void*)buf;
  
  return VO_TRUE;
}

static uint32_t start_slice(mp_image_t *mpi){
  int buf = 0;

#ifdef VERBOSE
  printf("Start slices %d\n",buf);
#endif

  if(mpi->flags & MP_IMGFLAG_DIRECT)
    buf = (int)mpi->priv;
  current_buffer = buf;

  return VO_TRUE;
}

static uint32_t draw_image(mp_image_t *mpi){
  int buf = 0;
  tdfx_vid_agp_move_t mov;
  tdfx_vid_yuv_t yuv;
  int p;
  uint8_t* planes[3];
  int stride[3];

#ifdef VERBOSE
  printf("Draw image %d\n",buf);
#endif

  if(mpi->flags & MP_IMGFLAG_DIRECT)
    buf = (int)mpi->priv;

  switch(mpi->imgfmt) {
  case IMGFMT_YUY2:
  case IMGFMT_UYVY:
  case IMGFMT_BGR8:
  case IMGFMT_BGR15:
  case IMGFMT_BGR16:
  case IMGFMT_BGR24:
  case IMGFMT_BGR32:
    if(!(mpi->flags&(MP_IMGFLAG_DIRECT|MP_IMGFLAG_DRAW_CALLBACK))) {
      // copy to agp_mem
#ifdef VERBOSE
      printf("Memcpy\n");
#endif
      planes[0] = agp_mem + buf * buffer_size;
      mem2agpcpy_pic(planes[0],mpi->planes[0],src_bpp*mpi->width,mpi->height,
		     buffer_stride[0],mpi->stride[0]);
    } else
      planes[0] = agp_mem + buf * buffer_size;

    mov.move2 = TDFX_VID_MOVE_2_PACKED;
    mov.width = mpi->width*((mpi->bpp+7)/8);
    mov.height = mpi->height;
    mov.src = planes[0] - agp_mem;
    mov.src_stride = buffer_stride[0];
  
    mov.dst = back_buffer;
    mov.dst_stride = src_stride;
	 
    if(ioctl(tdfx_fd,TDFX_VID_AGP_MOVE,&mov))
      printf("tdfx_vid: AGP move failed\n");
    break;

  case IMGFMT_YV12:
  case IMGFMT_I420:
    if(!(mpi->flags&(MP_IMGFLAG_DIRECT|MP_IMGFLAG_DRAW_CALLBACK))) {
      // Copy to agp mem
#ifdef VERBOSE
      printf("Memcpy\n");
#endif
      planes[0] = agp_mem + buf * buffer_size;
      memcpy_pic(planes[0],mpi->planes[0],mpi->width,mpi->height,
		 buffer_stride[0],mpi->stride[0]);
      planes[1] = planes[0] + (mpi->height*mpi->stride[0]);
      memcpy_pic(planes[1],mpi->planes[1],mpi->chroma_width,mpi->chroma_height,
		 buffer_stride[1],mpi->stride[1]);
      planes[2] = planes[1] + (mpi->chroma_height*mpi->stride[1]);
      memcpy_pic(planes[2],mpi->planes[2],mpi->chroma_width,mpi->chroma_height,
		 buffer_stride[2],mpi->stride[2]);
    } else {
      planes[0] = agp_mem + buf * buffer_size;
      planes[1] = planes[0] + buffer_stride[0] * src_height;
      planes[2] = planes[1] + buffer_stride[1] * src_height/2;
    }

    // Setup the yuv thing
    yuv.base = back_buffer;
    yuv.stride = src_stride;
    if(ioctl(tdfx_fd,TDFX_VID_SET_YUV,&yuv)) {
      printf("tdfx_vid: Set yuv failed\n");
      break;
    }
    

    // Now agp move that
    // Y
    mov.move2 = TDFX_VID_MOVE_2_YUV;
    mov.width = mpi->width;
    mov.height = mpi->height;
    mov.src = planes[0] - agp_mem;
    mov.src_stride =  buffer_stride[0];
    mov.dst = 0x0;
    mov.dst_stride = TDFX_VID_YUV_STRIDE;

    if(ioctl(tdfx_fd,TDFX_VID_AGP_MOVE,&mov)) {
      printf("tdfx_vid: AGP move failed on Y plane\n");
      break;
    }
    //return 0;
    // U
    p = mpi->imgfmt == IMGFMT_YV12 ? 1 : 2;
    mov.width = mpi->chroma_width;
    mov.height = mpi->chroma_height;
    mov.src = planes[p] - agp_mem;
    mov.src_stride = buffer_stride[p];
    mov.dst += TDFX_VID_YUV_PLANE_SIZE;
    if(ioctl(tdfx_fd,TDFX_VID_AGP_MOVE,&mov)) {
      printf("tdfx_vid: AGP move failed on U plane\n");
      break;
    }
    // V
    p = mpi->imgfmt == IMGFMT_YV12 ? 2 : 1;
    mov.src = planes[p] - agp_mem;
    mov.src_stride = buffer_stride[p];
    mov.dst += TDFX_VID_YUV_PLANE_SIZE;
    if(ioctl(tdfx_fd,TDFX_VID_AGP_MOVE,&mov)) {
      printf("tdfx_vid: AGP move failed on U plane\n");
      break;
    }
    break;
  default:
    printf("What's that for a format 0x%x\n",mpi->imgfmt);
    return VO_TRUE;
  }

  return VO_TRUE;
}

static uint32_t fullscreen(void) {
  vo_fs ^= 1;
  aspect(&dst_width,&dst_height,vo_fs ? A_ZOOM : A_NOZOOM);
  // This does not work :((
  //clear_screen();
  return VO_TRUE;
}

static uint32_t set_window(mp_win_t* w) {
  if(!use_overlay) return VO_FALSE;

  tdfx_ov.dst_x = w->x;
  tdfx_ov.dst_y = w->y;
  tdfx_ov.dst_width = w->w;
  tdfx_ov.dst_height = w->h;

  if(ioctl(tdfx_fd,TDFX_VID_SET_OVERLAY,&tdfx_ov))
    mp_msg(MSGT_VO, MSGL_V, "tdfx_vid: set window failed\n");

  return VO_TRUE;
}

static uint32_t set_colorkey(mp_colorkey_t* colork) {
  if(!use_overlay) return VO_FALSE;

  tdfx_ov.colorkey[0] = tdfx_ov.colorkey[1] = colork->x11;
  tdfx_ov.use_colorkey = 1;
  tdfx_ov.invert_colorkey = 0;

  if(ioctl(tdfx_fd,TDFX_VID_SET_OVERLAY,&tdfx_ov))
    mp_msg(MSGT_VO, MSGL_V, "tdfx_vid: set colorkey failed\n");

  return VO_TRUE;
}

static uint32_t control(uint32_t request, void *data, ...)
{
  switch (request) {
  case VOCTRL_QUERY_FORMAT:
    return query_format(*((uint32_t*)data));
  case VOCTRL_GET_IMAGE:
    return get_image(data);
  case VOCTRL_DRAW_IMAGE:
    return draw_image(data);
  case VOCTRL_START_SLICE:
    return start_slice(data);
  case VOCTRL_FULLSCREEN:
    return fullscreen();
  case VOCTRL_XOVERLAY_SUPPORT:
    return VO_TRUE;
  case VOCTRL_XOVERLAY_SET_COLORKEY:
    return set_colorkey(data);
  case VOCTRL_XOVERLAY_SET_WIN:
    return set_window(data);
  }
  return VO_NOTIMPL;
}