Please support canted displays, used in wide angle VR headsets like Pimax

I have a Pimax 5K+ which I can't run in full field of view, because Elite makes an incorrect assumption that both displays are pointed straight forward. That's not the case for Pimax, StarVR, Xtal, who knows who else. At least the optics of the Vive Index are also canted. Point is, GetProjectionMatrix and GetEyeToHeadTransform may contain rotation as well as translation. Supporting it is generally easier than not; it fails because you're specifically rebuilding the same transformation from incomplete data.
 
I have a Pimax 5K+ which I can't run in full field of view, because Elite makes an incorrect assumption that both displays are pointed straight forward. That's not the case for Pimax, StarVR, Xtal, who knows who else. At least the optics of the Vive Index are also canted. Point is, GetProjectionMatrix and GetEyeToHeadTransform may contain rotation as well as translation. Supporting it is generally easier than not; it fails because you're specifically rebuilding the same transformation from incomplete data.
An idea would do 2 bug reports. One on Pimax 5k VR+ Manufacture and one on Frontier.
 
I argue that it is a bug because it's making an assumption that AFAICT isn't even supported by the OpenVR API. You produce this type of result if you build the view matrices from the field of view statistics and an IPD, but the call that extracts the IPD (GetEyeToHeadTransform or GetProjectionMatrix) extracts a full transformation matrix. GetProjectionRaw, which produces the frustum boundaries, doesn't produce the origin point. Similarly, OpenXR reports both Pose and Fov; Pose is 6 degrees of freedom, not 1. Perhaps the assumption was inherited from Oculus? I expect it isn't a regression, though, just something that was working enough for some tested cases. Thanks for the link; I've upvoted it now.
 
Top Bottom