ANNOUNCING JGLMark!!!

General Discussions
#1

Well this thought had to pop in to everyone’s head at one moment in time… “We have a alot of good Windows based Direct3D benchmarks, but no GOOD OpenGL cross-platform benchmarks”

Well I propose A Java OpenGL benchmark and I dub it JGLMark (Simple yet adequate). Yes the aim is to stay along the same lines of functionality of 3DMark. Below are some links to pictures of the preliminary work. The featured API for the benchmark LWJGL of course!


I would like this benchmark to become a community effort so I am calling on the community to provide initial feedback on the idea as well as the initial look and feel. Imagine a Platform independent 3DMark!!!

I have sent and initial project proposal to javagaming, however I would like to recruit those interested in working on the project and garner feedback.

#2

[quote=“crash0veride007,post:1,topic:24669”]
It might make sense to code it with both LWJGL and JOGL. Then people can compare results between the two. I imagine that the difference won’t be that much, but having the ability to compare the two using exactly the same data would give us some real numbers. It seems a frequent question is ‘which API should I use?’ and that might help people decide.

#3

I know this has been beaten to death, but surely jogl & LWJGL are for practical purposes identical in their OpenGL wrapping expense. The only performance differences I would imagine would solely be from application implementation differences. Then again, I only have experience with LWJGL so I may be talking out of my ass.

Of course one feature I believe both are missing is a GL state cache on the java side. It is well worth it to cache & test state changes on the java side and avoid the calls through JNI when possible. Heck for most drivers it is worthwhile to do this even in a pure C++ implementation. Any really high end use of either JOGL or LWJGL is going to require the application writer to implement his own state cache, so why not include it in the API?

As far as the benchmark goes, I am interested in it primarily for testing different drivers & platforms side by side. I expect that attempting to measure JOGL vs LWJGL will be an excersise in futility.

#4

That’s the point… we could put it to rest finally. It’s one thing to talk about how there shouldn’t be any difference… but let’s see some evidence and be done with it.

#5

Well, LWJGL/JOGL benchmarker could benchmark whether it can start the program at first place in given hardware ;-). I have seen a stability over several hardware setups be difficult area for Java OpenGL wrappers.

#6

I’d be fine with having a JOGL and a LWJGL version of the benchmark. Although up to this point I myself have only coded in LWJGL and JAVA.

#7

How do you downlaod jglmark program?
Frontpage has a ridiciosly(!) large bitmap image as a imagemap. But download button does not have any hotspots. I would like to try it out on my machine, but cannot download files.
https://jglmark.dev.java.net/

#8

You have been granted observer role, so you may check it out from cvs, and compile and play. The JGLMark team is a group of us working on it in our spare time so their is no release yet, but we do soon hope to make an early Alpha Preview release soon.