On Fri, Apr 11, 2008 at 6:17 AM, Claude Heiland-Allen <<a href="mailto:claudiusmaximus@goto10.org">claudiusmaximus@goto10.org</a>> wrote:<br><div class="gmail_quote"><blockquote class="gmail_quote" style="border-left: 1px solid rgb(204, 204, 204); margin: 0pt 0pt 0pt 0.8ex; padding-left: 1ex;">
<div class="Ih2E3d">And in my experience with oldish AGP graphics cards, GPU->CPU image<br></div>
copying is rather slow. PCIe may be better, I don't know.<br>
</blockquote><div><br>PCIe is not really better. With GEM we can't use the asynchronous fast paths since the typical method is to read a fully rendered frame. Someone could look into using FBOs and async read back.<br>
</div><blockquote class="gmail_quote" style="border-left: 1px solid rgb(204, 204, 204); margin: 0pt 0pt 0pt 0.8ex; padding-left: 1ex;">I don't know if there is any way to share GPU memory between independent<br>
processes on the same machine, would be cool.<br>
</blockquote><div> </div></div>You could try to grab screen coordinates other than the active window. On OSX 10.5 there is supposedly a flag that makes your context 'read only' which allows other apps to scrape the contents into a texture. <br>
<br>cgc<br>