Alex Deucher wrote:
On Sun, Feb 8, 2009 at 9:22 AM, Christian König
wrote: Hi Alex,
the attached patch implements a simple wait_vblank function, and utilize those in R600DisplayTexturedVideo just before draw_auto as Matthias suggested.
I am assuming that the Xv output goes to CRTC1 and redefining the DxCRTC_STATUS registers in r6xx_accel is very ugly, but it's at least a start and seems to work quite fine for me, reducing tearing in the output under a (for me) noticeable level.
This looks good, but you might want to look at stalling on a particular vline range rather than vblank in general like we did in the radeon driver. I suppose it's worth comparing which method works better.
Can i start cleaning up this mess a little bit? For example removing "#define uint32_t CARD32" or the double register definition in rhd_reg.h and r600_reg*.h.
I'm fine with cleaning up. I wanted to move to using std int types originally, but Egbert and others wanted to use CARD* types instead for compatibility with really old compilers that don't support standard int types.
On a side note, what is the policy with C++ comments in C code? I made a start in manually converting the // C++ comments in the AtomBIOS headers to /* C */ style, largely to get familiar with the code and reduce compile noise. But then I noticed that some contributors are adding new C++ comments to the driver source. Even though GCC tolerates C++ comments, this generates needless warnings, is technically wrong, and could also break on older compilers. I'd happily convert all such comments on a one-off basis, but there's no point if others insist on putting them in afresh. Tony.
Alex
-- To unsubscribe, e-mail: radeonhd+unsubscribe@opensuse.org For additional commands, e-mail: radeonhd+help@opensuse.org