p_dst_pic->p->i_pitch *
( i_y_offset + p_filter->fmt_out.video.i_y_offset );
- i_src1_pitch = p_dst_orig->p[Y_PLANE].i_pitch;
+ i_src1_pitch = p_dst_orig->p->i_pitch;
p_src1 = p_dst_orig->p->p_pixels + i_x_offset * i_pix_pitch +
p_filter->fmt_out.video.i_x_offset * i_pix_pitch +
p_dst_orig->p->i_pitch *
#define MAX_TRANS 255
#define TRANS_BITS 8
- /* Draw until we reach the bottom of the subtitle */
- for( i_y = 0; i_y < i_height; i_y++, p_trans += i_src2_pitch,
- p_dst += i_dst_pitch, p_src1 += i_src1_pitch,
- p_src2_y += i_src2_pitch, p_src2_u += i_src2_pitch,
- p_src2_v += i_src2_pitch )
+ if( (i_pix_pitch == 4)
+ && (((((int)p_dst)|((int)p_src1)|i_dst_pitch|i_src1_pitch) & 3) == 0) )
{
- /* Draw until we reach the end of the line */
- for( i_x = 0; i_x < i_width; i_x++ )
+ /*
+ ** if picture pixels are 32 bits long and lines addresses are 32 bit aligned,
+ ** optimize rendering
+ */
+ uint32_t *p32_dst = (uint32_t *)p_dst;
+ uint32_t i32_dst_pitch = (uint32_t)(i_dst_pitch>>2);
+ uint32_t *p32_src1 = (uint32_t *)p_src1;
+ uint32_t i32_src1_pitch = (uint32_t)(i_src1_pitch>>2);
+
+ int i_rshift, i_gshift, i_bshift;
+ uint32_t i_rmask, i_gmask, i_bmask;
+
+ if( p_dst_pic->p_heap )
{
- if( p_trans )
- i_trans = ( p_trans[i_x] * i_alpha ) / 255;
- if( !i_trans )
+ i_rmask = p_dst_pic->p_heap->i_rmask;
+ i_gmask = p_dst_pic->p_heap->i_gmask;
+ i_bmask = p_dst_pic->p_heap->i_bmask;
+ i_rshift = p_dst_pic->p_heap->i_lrshift;
+ i_gshift = p_dst_pic->p_heap->i_lgshift;
+ i_bshift = p_dst_pic->p_heap->i_lbshift;
+ }
+ else
+ {
+ i_rmask = p_dst_pic->format.i_rmask;
+ i_gmask = p_dst_pic->format.i_gmask;
+ i_bmask = p_dst_pic->format.i_bmask;
+
+ if( (i_rmask == 0x00FF0000)
+ && (i_gmask == 0x0000FF00)
+ && (i_bmask == 0x000000FF) )
{
- /* Completely transparent. Don't change pixel */
- continue;
+ /* X8R8G8B8 pixel layout */
+ i_rshift = 16;
+ i_bshift = 8;
+ i_gshift = 0;
}
- else if( i_trans == MAX_TRANS )
+ else if( (i_rmask == 0xFF000000)
+ && (i_gmask == 0x00FF0000)
+ && (i_bmask == 0x0000FF00) )
{
- /* Completely opaque. Completely overwrite underlying pixel */
- yuv_to_rgb( &r, &g, &b,
- p_src2_y[i_x], p_src2_u[i_x], p_src2_v[i_x] );
+ /* R8G8B8X8 pixel layout */
+ i_rshift = 24;
+ i_bshift = 16;
+ i_gshift = 8;
+ }
+ else
+ {
+ goto slower;
+ }
+ }
+ /* Draw until we reach the bottom of the subtitle */
+ for( i_y = 0; i_y < i_height; i_y++, p_trans += i_src2_pitch,
+ p32_dst += i32_dst_pitch, p32_src1 += i32_src1_pitch,
+ p_src2_y += i_src2_pitch, p_src2_u += i_src2_pitch,
+ p_src2_v += i_src2_pitch )
+ {
+ /* Draw until we reach the end of the line */
+ for( i_x = 0; i_x < i_width; i_x++ )
+ {
+ if( p_trans )
+ i_trans = ( p_trans[i_x] * i_alpha ) / 255;
+ if( !i_trans )
+ {
+ /* Completely transparent. Don't change pixel */
+ continue;
+ }
+ else if( i_trans == MAX_TRANS )
+ {
+ /* Completely opaque. Completely overwrite underlying pixel */
+ yuv_to_rgb( &r, &g, &b,
+ p_src2_y[i_x], p_src2_u[i_x], p_src2_v[i_x] );
- p_dst[i_x * i_pix_pitch] = r;
- p_dst[i_x * i_pix_pitch + 1] = g;
- p_dst[i_x * i_pix_pitch + 2] = b;
- continue;
+ p32_dst[i_x] = (r<<i_rshift)
+ + (g<<i_gshift)
+ + (b<<i_bshift);
+ }
+ else
+ {
+ /* Blending */
+ uint32_t i_pix_src1 = p32_src1[i_x];
+ yuv_to_rgb( &r, &g, &b,
+ p_src2_y[i_x], p_src2_u[i_x], p_src2_v[i_x] );
+
+ p32_dst[i_x] = ( ( r * i_trans +
+ (uint16_t)((i_pix_src1 & i_rmask)>>i_rshift) *
+ (MAX_TRANS - i_trans) ) >> TRANS_BITS) << i_rshift
+ | ( ( g * i_trans +
+ (uint16_t)((i_pix_src1 & i_gmask)>>i_gshift) *
+ (MAX_TRANS - i_trans) ) >> TRANS_BITS) << i_gshift
+ | ( ( b * i_trans +
+ (uint16_t)((i_pix_src1 & i_bmask)>>i_bshift) *
+ (MAX_TRANS - i_trans) ) >> TRANS_BITS) << i_bshift;
+ }
}
+ }
+ }
+ else
+ {
+ int i_rindex, i_bindex, i_gindex;
+ uint32_t i_rmask, i_gmask, i_bmask;
+
+ slower:
+
+ i_rmask = p_dst_pic->format.i_rmask;
+ i_gmask = p_dst_pic->format.i_gmask;
+ i_bmask = p_dst_pic->format.i_bmask;
+
+ /*
+ ** quick and dirty way to get byte index from mask
+ ** will only work correctly if mask are 8 bit aligned
+ ** and are 8 bit long
+ */
+#ifdef WORDS_BIGENDIAN
+ i_rindex = ((i_rmask>>16) & 1)
+ | ((i_rmask>>8) & 2)
+ | ((i_rmask) & 3);
+ i_gindex = ((i_gmask>>16) & 1)
+ | ((i_gmask>>8) & 2)
+ | ((i_gmask) & 3);
+ i_bindex = ((i_bmask>>16) & 1)
+ | ((i_bmask>>8) & 2)
+ | ((i_bmask) & 3);
+#else
+ i_rindex = ((i_rmask>>24) & 3)
+ | ((i_rmask>>16) & 2)
+ | ((i_rmask>>8) & 1);
+ i_gindex = ((i_gmask>>24) & 3)
+ | ((i_gmask>>16) & 2)
+ | ((i_gmask>>8) & 1);
+ i_bindex = ((i_bmask>>24) & 3)
+ | ((i_bmask>>16) & 2)
+ | ((i_bmask>>8) & 1);
+#endif
- /* Blending */
- yuv_to_rgb( &r, &g, &b,
- p_src2_y[i_x], p_src2_u[i_x], p_src2_v[i_x] );
+ /* Draw until we reach the bottom of the subtitle */
+ for( i_y = 0; i_y < i_height; i_y++, p_trans += i_src2_pitch,
+ p_dst += i_dst_pitch, p_src1 += i_src1_pitch,
+ p_src2_y += i_src2_pitch, p_src2_u += i_src2_pitch,
+ p_src2_v += i_src2_pitch )
+ {
+ /* Draw until we reach the end of the line */
+ for( i_x = 0; i_x < i_width; i_x++ )
+ {
+ if( p_trans )
+ i_trans = ( p_trans[i_x] * i_alpha ) / 255;
+ if( !i_trans )
+ {
+ /* Completely transparent. Don't change pixel */
+ continue;
+ }
+ else
+ {
+ int i_pos = i_x * i_pix_pitch;
+ if( i_trans == MAX_TRANS )
+ {
- p_dst[i_x * i_pix_pitch] = ( r * i_trans +
- (uint16_t)p_src1[i_x * i_pix_pitch] *
- (MAX_TRANS - i_trans) ) >> TRANS_BITS;
- p_dst[i_x * i_pix_pitch + 1] = ( g * i_trans +
- (uint16_t)p_src1[i_x * i_pix_pitch + 1] *
- (MAX_TRANS - i_trans) ) >> TRANS_BITS;
- p_dst[i_x * i_pix_pitch + 2] = ( b * i_trans +
- (uint16_t)p_src1[i_x * i_pix_pitch + 2] *
- (MAX_TRANS - i_trans) ) >> TRANS_BITS;
+ /* Completely opaque. Completely overwrite underlying pixel */
+ yuv_to_rgb( &r, &g, &b,
+ p_src2_y[i_x], p_src2_u[i_x], p_src2_v[i_x] );
+
+ p_dst[i_pos + i_rindex ] = r;
+ p_dst[i_pos + i_gindex ] = g;
+ p_dst[i_pos + i_bindex ] = b;
+ }
+ else
+ {
+ int i_rpos = i_pos + i_rindex;
+ int i_gpos = i_pos + i_gindex;
+ int i_bpos = i_pos + i_bindex;
+
+ /* Blending */
+ yuv_to_rgb( &r, &g, &b,
+ p_src2_y[i_x], p_src2_u[i_x], p_src2_v[i_x] );
+
+ p_dst[i_rpos] = ( r * i_trans +
+ (uint16_t)p_src1[i_rpos] *
+ (MAX_TRANS - i_trans) ) >> TRANS_BITS;
+ p_dst[i_gpos] = ( r * i_trans +
+ (uint16_t)p_src1[i_gpos] *
+ (MAX_TRANS - i_trans) ) >> TRANS_BITS;
+ p_dst[i_bpos] = ( r * i_trans +
+ (uint16_t)p_src1[i_gpos] *
+ (MAX_TRANS - i_trans) ) >> TRANS_BITS;
+ }
+ }
+ }
}
}
/* Completely opaque. Completely overwrite underlying pixel */
p_dst_y[i_x] = p_pal[p_src2[i_x]][0];
- if( b_even_scanline && i_x % 2 == 0 )
+ if( b_even_scanline && ((i_x % 2) == 0) )
{
p_dst_u[i_x/2] = p_pal[p_src2[i_x]][1];
p_dst_v[i_x/2] = p_pal[p_src2[i_x]][2];
(uint16_t)p_src1_y[i_x] * (MAX_TRANS - i_trans) )
>> TRANS_BITS;
- if( b_even_scanline && i_x % 2 == 0 )
+ if( b_even_scanline && ((i_x % 2) == 0) )
{
p_dst_u[i_x/2] = ( (uint16_t)p_pal[p_src2[i_x]][1] * i_trans +
(uint16_t)p_src1_u[i_x/2] * (MAX_TRANS - i_trans) )