Cagdas Ogut wrote:
> Hi,
> 
> We are writing an application which receives some image data by dma, 
> processes it and later displays it by locking and rendering into a 
> triple buffered primary surface (with DSCAPS_VIDEOONLY flag set). 
> Everything works very well upto this point.
> 
> Though our next task is to write this data to harddisk or send it to 
> further processing. We wanted to use dump of primary for this purpose 
> but we are not sure which of the tripple buffers it dumps. if it dumps 
> the front buffer this means too much traffic from/to video card (MGA4x0 
> at AGP).

Dump() locks the surface read-only and such locks always use the
front buffer.

> Is there a way to assure that a triple buffered primary keeps its front 
> and mid buffers in video memory but back buffer in AGP/system memory?

The only case where buffers of one surface have different policies is
the DLBM_BACKSYSTEM buffer mode for a layer, but that's only double
buffered. The purpose is to have fast transparent window composition
in system memory, when no hardware acceleration is available.

In your case it might be better to have an intermediate single buffered
system only surface, being your back buffer. You can use Blit() to show
the data on screen, that's also what DLBM_BACKSYSTEM does internally.
You can decide if you want a single, double or even triple buffered
video only surface for the layer to show your data.

Performance wise you won't loose anything with the intermediate surface
as opposed to an unimplemented DLBM_TRIPLEBACKSYSTEM :)

 > As well, is it possible to assure dumping the back buffer instead of 
front?

No, but using a single buffered intermediate surface would also solve this.

-- 
Best regards,
   Denis Oliver Kropp

.------------------------------------------.
| DirectFB - Hardware accelerated graphics |
| http://www.directfb.org/                 |
"------------------------------------------"

_______________________________________________
directfb-users mailing list
[email protected]
http://mail.directfb.org/cgi-bin/mailman/listinfo/directfb-users

Reply via email to