Inspiring

JOVR 0.4.2.0: Java/Maven bindings for 0.4.2 (Updated 09-05)

52 replies
10880 views

I've updated my Java bindings for SDK 0.4.2. The source code can be found on Github here. Example code can be found here.

Developers using Maven (or a Maven compatible tool like Gradle) can declare a dependency on the following artifact:


   <dependency>
      <groupId>org.saintandreas</groupId>
      <artifactId>jovr</artifactId>
      <version>0.4.2.0</version>
    </dependency>

Update 2014-09-05: Release of 0.4.2.0
Changes since 0.4.2.0

Updated binaries to SDK 0.4.2

Added new HQ distortion flag (currently non-functional in OpenGL)

Update 2014-09-03: Release of 0.4.1.2
Changes since 0.4.1.1

Added function to enable/disable HSW rendering by the SDK

Update 2014-08-15: Release of 0.4.1.1
Changes since 0.4.1.0

The windows binaries were being built in release mode, but had some debug preprocessor macros enabled, so any asserts you hit would crash the application.

Known issues are unchanged.

Update 2014-08-13: Release of 0.4.1.0

Changes since 0.4.0.1

Added ovr_InitializeRenderingShim method to the OvrLibrary type.

Updated windows binaries to 0.4.1 SDK

Added Mac binaries

Known issues

Health and safety warning doesn't display properly for me on OSX

Continued issues with OpenGL and Direct HMD mode. I recommend using extended mode for the time being if you're having issues.

Judder issues are prominent in OpenGL in all my tests. I recommend running your headset @60Hz until Oculus or the video card manufacturers do whatever is required to resolve the problem.

Update 2014-08-11: Release of 0.4.0.1

Changes since 0.4.0.0

Renamed the HmdDesc type to Hmd to more closely match the SDK naming

Added some missing functions to Hmd required for client side distortion

Added some missing function to Hmd for health and safety warning management

Added some missing override annotations

Known issues

Unfortunately, like the current SDK release, they are windows only. I've built both the 64 bit and 32 bit DLL versions and they're embedded in the jar file.

I've been unable to get OpenGL to work properly with Direct HMD mode. I recommend using extended mode for the time being if you're having issues.

Show previous replies

E

elect

Up and Coming

"jherico" wrote:
"elect" wrote:
Ps: what about ovrHmd_GetRenderScaleAndOffset?

Looks like some of the client side distortion items were missed in the Hmd class. I'll fix that in the next revision. In the meantime you can still access it directly through the OvrLibrary interface.

How?

J

jhericoAuthor

Inspiring

If you look at the source for the Hmd class (called HmdDesc in JOVR 0.4.0.0, Hmd in 0.4.0.1) you'll see that it's full of methods that call methods on OvrLibrary.INSTANCE, passing itself as the first member.

Like so:


  public FrameTiming beginFrame(int frameIndex) {
    return OvrLibrary.INSTANCE.ovrHmd_BeginFrame(this, frameIndex);
  }

All of the C API functions are exposed as method on OvrLibrary, whether or not they're in the Hmd/HmdDesc helper class. So you could directly call


OvrLibrary.INSTANCE.ovrHmd_GetRenderScaleAndOffset(hmd, ...)

However, I've updated the module over the weekend to add the missing functions, so if you update to 0.4.0.1 you should be able to call getRenderScaleAndOffset(...) directly on the Hmd class.

Brad Davis - Developer for High Fidelity Co-author of Oculus Rift in Action

E

elect

Up and Coming

Nice, I'll give it a try tomorrow

thanks for your work, jherico

E

elect

Up and Coming

I cant see it in the jvr 0.4.0.1

Anyway I tried

            OvrLibrary.INSTANCE.ovrHmd_GetRenderScaleAndOffset((FovPort.ByValue) eyeFov[eyeNum], (OvrSizei.ByValue) frameBuffer.getSize(),
                    (OvrRecti.ByValue) eyeRenderViewport[eyeNum], uvScaleOffset[eyeNum]);

Exception in thread "main-AWTAnimator" java.lang.ClassCastException: com.oculusvr.capi.FovPort cannot be cast to com.oculusvr.capi.FovPort$ByValue

If I dont try the casting I get error directly in the ide

J

jhericoAuthor

Inspiring

"elect" wrote:
I cant see it in the jvr 0.4.0.1

Odd, I can definately see it in my copy. I just deleted it from my repository to force it to redownload and it's still there:


public static OvrVector2f[] com.oculusvr.capi.Hmd.getRenderScaleAndOffset(FovPort, OvrSizei, OvrRecti);

"elect" wrote:

Anyway I tried
            OvrLibrary.INSTANCE.ovrHmd_GetRenderScaleAndOffset((FovPort.ByValue) eyeFov[eyeNum], (OvrSizei.ByValue) frameBuffer.getSize(),
                    (OvrRecti.ByValue) eyeRenderViewport[eyeNum], uvScaleOffset[eyeNum]);    
Exception in thread "main-AWTAnimator" java.lang.ClassCastException: com.oculusvr.capi.FovPort cannot be cast to com.oculusvr.capi.FovPort$ByValue

If I dont try the casting I get error directly in the ide

You can't cast directly, but you should be able to do


new FovPort.ByValue(eyeFov[eyeNum]);

and


new OvrRecti.ByValue(eyeRenderViewport[eyeNum]);

Brad Davis - Developer for High Fidelity Co-author of Oculus Rift in Action

E

elect

Up and Coming

"jherico" wrote:

You can't cast directly, but you should be able to do
new FovPort.ByValue(eyeFov[eyeNum]);
and
new OvrRecti.ByValue(eyeRenderViewport[eyeNum]);

OvrLibrary.INSTANCE.ovrHmd_GetRenderScaleAndOffset(new FovPort.ByValue(eyeFov[eyeNum]), new OvrSizei.ByValue(frameBuffer.getSize()), new OvrRecti.ByValue(eyeRenderViewport[eyeNum]), uvScaleOffset[eyeNum]);

Exception in thread "main-AWTAnimator" java.lang.IllegalArgumentException: Structure array elements must use contiguous memory (bad backing address at Structure array index 1)

J

jhericoAuthor

Inspiring

"elect" wrote:
OvrLibrary.INSTANCE.ovrHmd_GetRenderScaleAndOffset(new FovPort.ByValue(eyeFov[eyeNum]), new OvrSizei.ByValue(frameBuffer.getSize()), new OvrRecti.ByValue(eyeRenderViewport[eyeNum]), uvScaleOffset[eyeNum]);
Exception in thread "main-AWTAnimator" java.lang.IllegalArgumentException: Structure array elements must use contiguous memory (bad backing address at Structure array index 1)

It's a quirk of JNA that arrays passed through the C API have to be contiguous in memory. You can create a contiguous array using the toArray() member in the structure base class like so:


    OvrVector2f results[] = (OvrVector2f[]) new OvrVector2f().toArray(2);

The casting here is because the base class will only return the type Structure[].

Assuming uvScaleOffset is declared as


OvrVector2f uvScaleOffset[][];

you should be able to initialize it as


OvrVector2f uvScaleOffset[][] = {
  (OvrVector2f[]) new OvrVector2f().toArray(2),
  (OvrVector2f[]) new OvrVector2f().toArray(2)
};

Of course the point of the Hmd helper method is that you don't have to jump through all these 'byvalue' and 'contiguous memory' hoops. The implementation in the Hmd class look like so:


  public static OvrVector2f[] getRenderScaleAndOffset(FovPort fov, OvrSizei textureSize,
      OvrRecti renderViewport) {
    OvrVector2f results[] = (OvrVector2f[]) new OvrVector2f().toArray(2);
    OvrLibrary.INSTANCE.ovrHmd_GetRenderScaleAndOffset(
        byValue(fov), 
        byValue(textureSize), 
        byValue(renderViewport), 
        results);
    return results;
  }

The byValue calls there are static members of the Hmd class (which are private, but should probably be public) that inspect the type and up-convert to a ByValue if the type isn't already one.

Brad Davis - Developer for High Fidelity Co-author of Oculus Rift in Action

E

elect

Up and Coming

Ok, I tried to detach the library containing the jovr and attach it again, I see it now.. maybe some caching, dont know

Anyway, I gave it a try declaring it like this

uvScaleOffset = new OvrVector2f[2][2];

and getting the scale and offset like

uvScaleOffset[eyeNum] = Hmd.getRenderScaleAndOffset(eyeFov[eyeNum], renderTargetSize, eyeRenderViewport[eyeNum]);

It is working.. and more important I think that small pincushion effect I still had is definitely gone now (first I hardcoded those value based on the c example and I gave it a check, they are different) and now the image looks perfect, no more line distortion at the borders

40.png

viewtopic.php?f=20&t=12015#wrap

Thanks! ^^