Mercurial > mplayer.hg
view libmpcodecs/vf_expand.c @ 8781:8e3e71a8c414
Remove an obsolete link in the solaris section of the documentation
(mplayer should work out of the box with the included mpdvdkit2 on
solaris).
Mention the patches that should be installed to fix one of the two bugs
that prevent the use of DVDs >4GB on solaris 8.
author | jkeil |
---|---|
date | Sat, 04 Jan 2003 19:33:22 +0000 |
parents | f5223b78ebf6 |
children | 9df3c03b2ffe |
line wrap: on
line source
#define OSD_SUPPORT #include <stdio.h> #include <stdlib.h> #include <string.h> #include "../config.h" #include "../mp_msg.h" #include "img_format.h" #include "mp_image.h" #include "vf.h" #include "../libvo/fastmemcpy.h" #ifdef OSD_SUPPORT #include "../libvo/sub.h" #include "../libvo/osd.h" #endif #define MAX(a,b) ((a) > (b) ? (a) : (b)) struct vf_priv_s { int exp_w,exp_h; int exp_x,exp_y; mp_image_t *dmpi; int osd; unsigned char* fb_ptr; }; extern int opt_screen_size_x; extern int opt_screen_size_y; //===========================================================================// #ifdef OSD_SUPPORT static struct vf_instance_s* vf=NULL; // fixme (needs sub.c changes) static int orig_w,orig_h; static void remove_func_2(int x0,int y0, int w,int h){ // TODO: let's cleanup the place //printf("OSD clear: %d;%d %dx%d \n",x0,y0,w,h); vf_mpi_clear(vf->priv->dmpi,x0,y0,w,h); } static void remove_func(int x0,int y0, int w,int h){ if(!vo_osd_changed_flag) return; // split it to 4 parts: if(y0<vf->priv->exp_y){ // it has parts above the image: int y=y0+h; if(y>vf->priv->exp_y) y=vf->priv->exp_y; remove_func_2(x0,y0,w,y-y0); if(y0+h<=vf->priv->exp_y) return; h-=y-y0;y0=y; } if(y0+h>vf->priv->exp_y+orig_h){ // it has parts under the image: int y=y0; if(y<vf->priv->exp_y+orig_h) y=vf->priv->exp_y+orig_h; remove_func_2(x0,y,w,y0+h-y); if(y0>=vf->priv->exp_y+orig_h) return; h=y-y0; } if(x0>=vf->priv->exp_x || x0+w<=vf->priv->exp_x+orig_w) return; // TODO clear left and right side of the image if needed } static void draw_func(int x0,int y0, int w,int h,unsigned char* src, unsigned char *srca, int stride){ unsigned char* dst; if(!vo_osd_changed_flag && vf->priv->dmpi->planes[0]==vf->priv->fb_ptr){ // ok, enough to update the area inside the video, leave the black bands // untouched! if(x0<vf->priv->exp_x){ int tmp=vf->priv->exp_x-x0; w-=tmp; src+=tmp; srca+=tmp; x0+=tmp; } if(y0<vf->priv->exp_y){ int tmp=vf->priv->exp_y-y0; h-=tmp; src+=tmp*stride; srca+=tmp*stride; y0+=tmp; } if(x0+w>vf->priv->exp_x+orig_w){ w=vf->priv->exp_x+orig_w-x0; } if(y0+h>vf->priv->exp_y+orig_h){ h=vf->priv->exp_y+orig_h-y0; } } if(w<=0 || h<=0) return; // nothing to do... // printf("OSD redraw: %d;%d %dx%d \n",x0,y0,w,h); dst=vf->priv->dmpi->planes[0]+ vf->priv->dmpi->stride[0]*y0+ (vf->priv->dmpi->bpp>>3)*x0; switch(vf->priv->dmpi->imgfmt){ case IMGFMT_BGR15: case IMGFMT_RGB15: vo_draw_alpha_rgb15(w,h,src,srca,stride,dst,vf->priv->dmpi->stride[0]); break; case IMGFMT_BGR16: case IMGFMT_RGB16: vo_draw_alpha_rgb16(w,h,src,srca,stride,dst,vf->priv->dmpi->stride[0]); break; case IMGFMT_BGR24: case IMGFMT_RGB24: vo_draw_alpha_rgb24(w,h,src,srca,stride,dst,vf->priv->dmpi->stride[0]); break; case IMGFMT_BGR32: case IMGFMT_RGB32: vo_draw_alpha_rgb32(w,h,src,srca,stride,dst,vf->priv->dmpi->stride[0]); break; case IMGFMT_YV12: case IMGFMT_I420: case IMGFMT_IYUV: case IMGFMT_YVU9: case IMGFMT_IF09: case IMGFMT_Y800: case IMGFMT_Y8: vo_draw_alpha_yv12(w,h,src,srca,stride,dst,vf->priv->dmpi->stride[0]); break; case IMGFMT_YUY2: vo_draw_alpha_yuy2(w,h,src,srca,stride,dst,vf->priv->dmpi->stride[0]); break; case IMGFMT_UYVY: vo_draw_alpha_yuy2(w,h,src,srca,stride,dst+1,vf->priv->dmpi->stride[0]); break; } } static void draw_osd(struct vf_instance_s* vf_,int w,int h){ vf=vf_;orig_w=w;orig_h=h; // printf("======================================\n"); if(vf->priv->exp_w!=w || vf->priv->exp_h!=h || vf->priv->exp_x || vf->priv->exp_y){ // yep, we're expanding image, not just copy. if(vf->priv->dmpi->planes[0]!=vf->priv->fb_ptr){ // double buffering, so we need full clear :( remove_func(0,0,vf->priv->exp_w,vf->priv->exp_h); } else { // partial clear: vo_remove_text(vf->priv->exp_w,vf->priv->exp_h,remove_func); } } vo_draw_text(vf->priv->exp_w,vf->priv->exp_h,draw_func); // save buffer pointer for double buffering detection - yes, i know it's // ugly method, but note that codecs with DR support does the same... vf->priv->fb_ptr=vf->priv->dmpi->planes[0]; } #endif //===========================================================================// static int config(struct vf_instance_s* vf, int width, int height, int d_width, int d_height, unsigned int flags, unsigned int outfmt){ if (outfmt == IMGFMT_IF09) return 0; // calculate the missing parameters: #if 0 if(vf->priv->exp_w<width) vf->priv->exp_w=width; if(vf->priv->exp_h<height) vf->priv->exp_h=height; #else if ( vf->priv->exp_w == -1 ) vf->priv->exp_w=width; else if (vf->priv->exp_w < -1 ) vf->priv->exp_w=width - vf->priv->exp_w; else if ( vf->priv->exp_w<width ) vf->priv->exp_w=width; if ( vf->priv->exp_h == -1 ) vf->priv->exp_h=height; else if ( vf->priv->exp_h < -1 ) vf->priv->exp_h=height - vf->priv->exp_h; else if( vf->priv->exp_h<height ) vf->priv->exp_h=height; #endif if(vf->priv->exp_x<0 || vf->priv->exp_x+width>vf->priv->exp_w) vf->priv->exp_x=(vf->priv->exp_w-width)/2; if(vf->priv->exp_y<0 || vf->priv->exp_y+height>vf->priv->exp_h) vf->priv->exp_y=(vf->priv->exp_h-height)/2; vf->priv->fb_ptr=NULL; if(!opt_screen_size_x && !opt_screen_size_y){ d_width=d_width*vf->priv->exp_w/width; d_height=d_height*vf->priv->exp_h/height; } return vf_next_config(vf,vf->priv->exp_w,vf->priv->exp_h,d_width,d_height,flags,outfmt); } // there are 4 cases: // codec --DR--> expand --DR--> vo // codec --DR--> expand -copy-> vo // codec -copy-> expand --DR--> vo // codec -copy-> expand -copy-> vo (worst case) static void get_image(struct vf_instance_s* vf, mp_image_t *mpi){ // if(mpi->type==MP_IMGTYPE_IPB) return; // not yet working #ifdef OSD_SUPPORT if(vf->priv->osd && (mpi->flags&MP_IMGFLAG_PRESERVE)){ // check if we have to render osd! vo_update_osd(vf->priv->exp_w, vf->priv->exp_h); if(vo_osd_check_range_update(vf->priv->exp_x,vf->priv->exp_y, vf->priv->exp_x+mpi->w,vf->priv->exp_y+mpi->h)) return; } #endif if(vf->priv->exp_w==mpi->width || (mpi->flags&(MP_IMGFLAG_ACCEPT_STRIDE|MP_IMGFLAG_ACCEPT_WIDTH)) ){ // try full DR ! vf->priv->dmpi=vf_get_image(vf->next,mpi->imgfmt, mpi->type, mpi->flags, MAX(vf->priv->exp_w, mpi->width +vf->priv->exp_x), MAX(vf->priv->exp_h, mpi->height+vf->priv->exp_y)); // set up mpi as a cropped-down image of dmpi: if(mpi->flags&MP_IMGFLAG_PLANAR){ mpi->planes[0]=vf->priv->dmpi->planes[0]+ vf->priv->exp_y*vf->priv->dmpi->stride[0]+vf->priv->exp_x; mpi->planes[1]=vf->priv->dmpi->planes[1]+ (vf->priv->exp_y>>mpi->chroma_y_shift)*vf->priv->dmpi->stride[1]+(vf->priv->exp_x>>mpi->chroma_x_shift); mpi->planes[2]=vf->priv->dmpi->planes[2]+ (vf->priv->exp_y>>mpi->chroma_y_shift)*vf->priv->dmpi->stride[2]+(vf->priv->exp_x>>mpi->chroma_x_shift); mpi->stride[1]=vf->priv->dmpi->stride[1]; mpi->stride[2]=vf->priv->dmpi->stride[2]; } else { mpi->planes[0]=vf->priv->dmpi->planes[0]+ vf->priv->exp_y*vf->priv->dmpi->stride[0]+ vf->priv->exp_x*(vf->priv->dmpi->bpp/8); } mpi->stride[0]=vf->priv->dmpi->stride[0]; mpi->width=vf->priv->dmpi->width; mpi->flags|=MP_IMGFLAG_DIRECT; mpi->priv=(void*)vf->priv->dmpi; } } static int put_image(struct vf_instance_s* vf, mp_image_t *mpi){ if(mpi->flags&MP_IMGFLAG_DIRECT){ vf->priv->dmpi=mpi->priv; #ifdef OSD_SUPPORT if(vf->priv->osd) draw_osd(vf,mpi->w,mpi->h); #endif // we've used DR, so we're ready... return vf_next_put_image(vf,vf->priv->dmpi); } // hope we'll get DR buffer: vf->priv->dmpi=vf_get_image(vf->next,mpi->imgfmt, MP_IMGTYPE_TEMP, MP_IMGFLAG_ACCEPT_STRIDE, vf->priv->exp_w, vf->priv->exp_h); // copy mpi->dmpi... if(mpi->flags&MP_IMGFLAG_PLANAR){ memcpy_pic(vf->priv->dmpi->planes[0]+ vf->priv->exp_y*vf->priv->dmpi->stride[0]+vf->priv->exp_x, mpi->planes[0], mpi->w, mpi->h, vf->priv->dmpi->stride[0],mpi->stride[0]); memcpy_pic(vf->priv->dmpi->planes[1]+ (vf->priv->exp_y>>mpi->chroma_y_shift)*vf->priv->dmpi->stride[1]+(vf->priv->exp_x>>mpi->chroma_x_shift), mpi->planes[1], mpi->chroma_width, mpi->chroma_height, vf->priv->dmpi->stride[1],mpi->stride[1]); memcpy_pic(vf->priv->dmpi->planes[2]+ (vf->priv->exp_y>>mpi->chroma_y_shift)*vf->priv->dmpi->stride[2]+(vf->priv->exp_x>>mpi->chroma_x_shift), mpi->planes[2], mpi->chroma_width, mpi->chroma_height, vf->priv->dmpi->stride[2],mpi->stride[2]); } else { memcpy_pic(vf->priv->dmpi->planes[0]+ vf->priv->exp_y*vf->priv->dmpi->stride[0]+vf->priv->exp_x*(vf->priv->dmpi->bpp/8), mpi->planes[0], mpi->w*(vf->priv->dmpi->bpp/8), mpi->h, vf->priv->dmpi->stride[0],mpi->stride[0]); } #ifdef OSD_SUPPORT if(vf->priv->osd) draw_osd(vf,mpi->w,mpi->h); #endif return vf_next_put_image(vf,vf->priv->dmpi); } //===========================================================================// static int control(struct vf_instance_s* vf, int request, void* data){ #ifdef OSD_SUPPORT switch(request){ case VFCTRL_DRAW_OSD: if(vf->priv->osd) return CONTROL_TRUE; } #endif return vf_next_control(vf,request,data); } static int open(vf_instance_t *vf, char* args){ vf->config=config; vf->control=control; vf->get_image=get_image; vf->put_image=put_image; vf->priv=malloc(sizeof(struct vf_priv_s)); vf->priv->exp_x= vf->priv->exp_y= vf->priv->exp_w= vf->priv->exp_h=-1; vf->priv->osd=0; // parse args -> if(args) sscanf(args, "%d:%d:%d:%d:%d", &vf->priv->exp_w, &vf->priv->exp_h, &vf->priv->exp_x, &vf->priv->exp_y, &vf->priv->osd); printf("Expand: %d x %d, %d ; %d (-1=autodetect) osd: %d\n", vf->priv->exp_w, vf->priv->exp_h, vf->priv->exp_x, vf->priv->exp_y, vf->priv->osd); return 1; } vf_info_t vf_info_expand = { #ifdef OSD_SUPPORT "expanding & osd", #else "expanding", #endif "expand", "A'rpi", "", open }; //===========================================================================//