Hi devels
I fixed the problem with the direct_memcpy call. You were right. It had
to do with faulty pointer arithmetics. I now call:
bool sciensis_blit(
void *drv_data,
void *dev_data,
DFBRectangle *rect,
i32 dx,
i32 dy
)
{
u8 *dst, *src;
u32 i, j;
sciensis_drv_data *scdrv = (sciensis_drv_data *) drv_data;
sciensis_dev_data *scdev = (sciensis_dev_data *) dev_data;
/* Software blit/blend implementation */
dst = scdev->dst_addr + dy * scdev->dst_pitch;
dst += DFB_BYTES_PER_LINE(scdev->dst_fmt, dx);
src = scdev->src_addr + rect->y * scdev->src_pitch;
src += DFB_BYTES_PER_LINE(scdev->src_fmt, rect->x);
while (rect->h--) {
memcpy(dst, src, rect->w * scdev->dst_bpp);
dst += scdev->dst_pitch;
src += scdev->src_pitch;
}
return true;
}
That works fine. This is called when the blit is done with alphachannel
blend and i do see blended surfaces. What i don't understand is this:
memcpy raw copies the src buffer to the dst buffer. So where is the
blending done? I use surfaces where the dst and src colors are not
premultiplicated. What i want to do is to have the raw surface data in
order to implement the SRC_OVER method all on my own (in order to
hardware accelerate it using an MMX-like mechanism)
Best regards
Tasos Parisinos
_______________________________________________
directfb-dev mailing list
directfb-dev@directfb.org
http://mail.directfb.org/cgi-bin/mailman/listinfo/directfb-dev
