[gl4java-usergroup] gl*Pointer (Vertex,Normal,Color,...) / Array implementation ...

[Sven G. <sgo...@ja...>]

Dear Users, Dear Rob,

I analysed the problem of using VertexPointer (-> Arrays) !

Now, we do have a problem :-)

The semantics of the gl*Pointer methods is,
that the data referenced by the pointer (which is passed to GL as an
argument), is used LATER !

One GL Context hold's the data pointer for each:
	- vertex array
	- normal array
	- color array
	- ...

The data's pointer can be hold, after calling
	glEnableClientState(GLEnum.GL_VERTEX_ARRAY);    

So - at all time, the data, to which the reference points,
must be AVAIABLE till the next call of the appropiate:
	glDisableClientState(..)

In GL4Java, we use the following code:

        JNIEXPORT void JNICALL     
	Java_gl4java_GLFuncJauJNInf_glVertexPointer__III_3B (                 
		JNIEnv *env, jobject obj,                 
		jint size,                 
		jint type,                 
		jint stride,                 
		jbyteArray ptr)         
	{                 
		static void (CALLBACK *__func_ptr__)(GLint, GLenum, GLsizei, const
GLvoid *) = NULL;                 jbyte *ptr3 = NULL;  
                if(__func_ptr__==NULL) {                         
			__func_ptr__ = 
			     (void (CALLBACK *)(GLint, GLenum, GLsizei, const GLvoid
*))                                
getGLProcAddressHelper("glVertexPointer", NULL, 1, 0);        
                if(__func_ptr__==NULL)                                 
			return;                 
		}                 
		if(ptr!=NULL)                 
		{                         
			ptr3 = (*env)->GetByteArrayElements(env, ptr, 0);                 
		}                 
		__func_ptr__ ( (GLint) size, (GLenum) type, (GLsizei)
stride,                         
			       (const GLvoid *) ptr3                 );  
                if(ptr!=NULL)                 
		{                         
			(*env)->ReleaseByteArrayElements(env,  ptr, ptr3, 0);           
		}         
	}
   
The problem is, that the JVM can implement two ways for accessing
arrays:
	- make a copy
	- "pinning" - just pass the real pointer
Both methods use a garbage collection lock on the used array !

A JVM, which uses pinning, just passes us the pointer of such an object
!
The GL machine can use the data later - well, if the garbage collector
does not removed the non-locked data (by us) !

Well, if the copy method is used, a horrible scenario exists !
The GL machine accesses data, referenced by the given pointer before,
but the data is de-allocated / freed by the Release call !

The solution is:
	IF used the copy method, 
	for each type (Vertex, Color, ...) AND for each GL-Context:
		- not call Release, till:
			- a new gl*Pointer call is made
			- a glDisableClientState(*) is called
		So the data must being hold that long !

	IF used the pinning method:
		- nothing must be changed ...
		But we can use the upper implementation, to be sure,
		that the garbage collector has not release the ressources !

Generally, I am currently optimizing the array implementation !
The usage of identifying the method (copy,pinning) and using
this for the Release call (0, JNI_ABORT) !
The redundant call of Set*ArrayRegion is removed now !

This optimisation and bugfix, will be released with 
GL4Java 2.4.1 !

Yours, Sven Goethel
Rob Grzywinski wrote:
> 
> I am using MSJVM v5.00.3167 under Win95 and I just
> upgraded to GL4Java2.4.0.0.  My application, which
> uses display lists extensively, now displays all
> objects as if their 1st coordinate was at infinity (so
> all objects have an infinite line shooting out from
> them).  Also selection against any object causes an
> invalid page fault in module MSJAVA.DLL.  I'm working
> to isolate the problem and post the code, but I wanted
> to get the word out to see if anyone else has a
> similar problem.
> 
> BTW:  when I use Sun's JRE1.1.8 on the same machine
> with 2.4.0.0 everything is fine (all objects render as
> they did under 2.3.X.X).
> 
> BTW2:  yes, I recompiled and restarted the machine :)
> 
> Rob Grzywinski
> rgr...@ya...
> 
> __________________________________________________
> Do You Yahoo!?
> Get Yahoo! Mail - Free email you can access from anywhere!
> http://mail.yahoo.com/
> 
> _______________________________________________
> gl4java-usergroup mailing list
> gl4...@li...
> http://lists.sourceforge.net/mailman/listinfo/gl4java-usergroup

-- 
mailto:sgo...@ja...
www   : http://www.jausoft.com
voice : +49-521-2399440, +49-170-2115963; fax: +49-521-2399442