[vbox-dev] How to access host vertical blanking intervals

Klaus Espenlaub klaus.espenlaub at oracle.com
Thu Sep 13 17:58:43 GMT 2018


Hello Steve,

this "do what I want" request is tons more effort than what you think.

1. What is "the host display"? By default the VM is shown in a window,
which can span multiple physical screens. Also, what should happen in
the headless VM case when the screen is accessed using an RDP client?

2. Getting the vsync state is extremely host OS dependent, so it'd need
a whole machinery for abstracting this enough. Likely this would need to
bypass Qt and then life gets very hard.

3. While page flipping is minimizing tearing, it also means that on
every flip the entire VM screen content will be updated (no easy way to
figure out what actually changed). This means the screen updating will
be less efficient than when doing the updates in place.

All of this is fixable. After all we're talking about software. It's a
cartload of work for very little gain, though.

Just my 2c,
Klaus

On 03.09.2018 10:26, Michael Thayer wrote:
> Hello Steve,
> 
> Nothing at the moment I'm afraid.  It might be something we could
> consider adding some time (though it would be far from trivial, since we
> have four supported and additional unofficial host ports to make it work
> on).  It is also something we might accept a patch for.  One of the
> relevant code-owners is not available for the next week, so I can't say
> for sure.
> 
> Regards
> Michael
> 
> 31.08.2018 22:54, Steve Longerbeam wrote:
>> Hello VBOX developers,
>>
>> I recently contributed a patch to in-kernel vboxvideo DRM driver which
>> adds support for page flip. However the implementation does not
>> make any attempt to sync page flip with host vertical blanking
>> interval, so it's possible a DRM page flip will result in tearing
>> artifacts.
>>
>> So is there a call that can be made into the hypervisor that would allow
>> receiving an event when host display has entered a vblank interval?
>> Something in the VBOX Main API perhaps?
>>
>> Thanks in advance,
>> Steve-- 



More information about the vbox-dev mailing list