I did not expect the data to be used until R600DoneCopy was called. This makes it necessary to finish the copy. The end result is much simpler. Mark.
On Fri, Feb 13, 2009 at 5:28 PM, Mark van Doesburg
I did not expect the data to be used until R600DoneCopy was called. This makes it necessary to finish the copy. The end result is much simpler.
Thanks Mark. I've applied this patch and made a few optimizations as well. Alex -- To unsubscribe, e-mail: radeonhd+unsubscribe@opensuse.org For additional commands, e-mail: radeonhd+help@opensuse.org
Alex Deucher
On Tue, Feb 24, 2009 at 1:15 PM, Mark van Doesburg
Alex Deucher
wrote: Thanks Mark. I've applied this patch and made a few optimizations as well.
Thanks for patching. Your version handles the out of memory case gracefully :-) I am not completely sure if moving the exaOffscreenAlloc to R600PrepareCopy is a good idea however. In many cases the copies will not overlap and there is no reason to call exaOffscreenAlloc and exaOffscreenFree at all.
Initially I never freed the memory after allocating it, unless the pre-allocated memory was too small, and the result was noticably faster.
I thought about doing that as well, but decided to stick with alloc/free to keep more offscreen memory available for pixmaps. It's probably worth profiling over longer periods however. Alex -- To unsubscribe, e-mail: radeonhd+unsubscribe@opensuse.org For additional commands, e-mail: radeonhd+help@opensuse.org
participants (2)
-
Alex Deucher
-
Mark van Doesburg