Forum Discussion
TimeWarp not working when running with emulated device
Hi,
I have some issues with the TimeWarp shader when running with the emulated device, i.e. ovrHmd_CreateDebug(ovrHmd_DK1)
If I run with the time warp shader I get the following look:
If I run with ovrHmd_DK2 as a emulated device I get the same incorrect behavior (but in higher resolution).
If I connect my device (a DK1 in this case) I get the correct look:
Running emulated without TimeWarp also works fine.
I don't know if this is related but if I try running in debug mode with emulated device, I am running into the following ASSERT in the FrameTimeManager::BeginFrame(unsigned frameIndex) function:
I have some issues with the TimeWarp shader when running with the emulated device, i.e. ovrHmd_CreateDebug(ovrHmd_DK1)
If I run with the time warp shader I get the following look:
If I run with ovrHmd_DK2 as a emulated device I get the same incorrect behavior (but in higher resolution).
If I connect my device (a DK1 in this case) I get the correct look:
Running emulated without TimeWarp also works fine.
I don't know if this is related but if I try running in debug mode with emulated device, I am running into the following ASSERT in the FrameTimeManager::BeginFrame(unsigned frameIndex) function:
OVR_ASSERT(FrameTiming.Inputs.ScreenDelay != 0);
3 Replies
"pixelminer" wrote:
I don't know if this is related but if I try running in debug mode with emulated device, I am running into the following ASSERT in the FrameTimeManager::BeginFrame(unsigned frameIndex) function:OVR_ASSERT(FrameTiming.Inputs.ScreenDelay != 0);
You have to call the ovrHmd_ResetFrameTiming method at least once before you call ovrHmd_BeginFrameTiming method.- This does not solve the problems with the TimeWarp shader when using an emulated device (It still looks like the above pictures).
But it do solve the ASSERT problem. A problem I would say comes from a inconsistency in the SDK. Calling ovrHmd_Create() will automatically call ovrHmd_ResetFrameTiming() from inside that function (if we have a valid device connected). But ovrHmd_ResetFrameTiming() is oddly enough not called from inside ovrHmd_CreateDebug().
A bit inconsistent but easy to fix:
// Get first available HMD
hmd = ovrHmd_Create(0);
// If no HMD is found try an emulated device
if (!hmd) {
std::cout << "Warning: No device could be found. Creating emulated device " << std::endl;
hmd = ovrHmd_CreateDebug(ovrHmd_DK1);
ovrHmd_ResetFrameTiming(hmd, 0);
} - I finally figured out why I was getting bogus matrixes as response from when running an emulated device:
ovrHmd_GetEyeTimewarpMatrices(ovrHmd hmddesc, ovrEyeType eye, ovrPosef renderPose, ovrMatrix4f twmOut[2])
The reason was that I was not initializing the renderPose struct correctly, so the ovrPosef struct was containing uninitialized variables.
When running with a real device connected this struct was populated with data from the ovrHmd_GetEyePose() function. But I had that function inside conditional that never happened when running an emulated device:ovrPoseStatef headpose = ts.HeadPose;
if (ts.StatusFlags & (ovrStatus_OrientationTracked | ovrStatus_PositionTracked)) {
...
for (int eyeIndex = 0; eyeIndex < ovrEye_Count; ++eyeIndex) {
ovrEyeType eye = m_hmdDevice->EyeRenderOrder[eyeIndex];
m_headPose[eye] = ovrHmd_GetEyePose(m_hmdDevice, eye);
}
}
Removing this conditional statement solved my issue.
