Do you ever write your own stuff, or do you just translate other people's work?
Posted on 2004-02-10 03:48:04 by Henk-Jan
Imitate until you can create. :)
Posted on 2004-02-10 09:31:09 by bitRAKE
Imitate until you can create

precisely :alright:

"Immitation does not kill creativity. It only kills what needs to die, then creativity will take the lead"
Posted on 2004-02-10 09:59:34 by Ultrano
I don't really see the point in translating existing examples to asm. The code will be harder to read, it will not be faster, and I don't think anyone will learn from it. Not to mention the fact that the examples aren't exactly good to learn from anyway. They often demonstrate one thing, in an overly simplistic manner (look at the shadowvolume example.. you do NOT want to do it that way, it's far too slow. It's just a simple way to demonstrate the idea, which makes the translation to asm even less useful, it's not good code). They don't make use of a decent 3d engine either, they can only do the trick they're supposed to do, and that's it. Combining multiple 'tricks' is a much bigger problem.
Posted on 2004-02-10 11:50:22 by Henk-Jan
Henk-Jan, there is nothing wrong with taking small steps - I tend to enjoy the view more. It might not be your way, but others certainly find it more to their liking. If you only can see a "trick" in the example maybe you have run by it too quickly? Learning is a process not a destination.
Posted on 2004-02-10 20:13:21 by bitRAKE
Immitation is the most sincere form of flattery.
This is my third attempt to get a skinned mesh animation engine working, the other two were also translations of C source which were in turn half-assed implementations of the sdk example.
I write plenty of my own code, which if you'd taken the time to actually look at the source itself you would see. I've left the variable names and silly loops in place for easy comparison to the original. Speaking of writing one's own code, I've been posting in here for quite some time now and I don't recall seeing anything at all from you but oblique commentary, but I'm sure that doesn't mean you don't write anything. The process of translation has been much more educational for me than reading reams of tutorials, especially since it seems all the tutorials are likewise geared toward the sdk example. Yes, this code is terrible and could be improved a hell of a lot, and have a lot more flexibility than it does now, and my intention is certainly to clean it up and extend it, for example weight-blended multiple animations, and partial-hierarchy animations, that will all come in time. Constructive criticism is always appreciated. Please try to restrain yourself to such, and you will find your criticism received warmly and with gratitude.
Posted on 2004-02-10 22:27:44 by Homer
If you only can see a "trick" in the example maybe you have run by it too quickly?


The example only demonstrates one "trick", in this case a skinned mesh. One trick does not an engine make, that is my point.

I've been posting in here for quite some time now and I don't recall seeing anything at all from you but oblique commentary, but I'm sure that doesn't mean you don't write anything.


From my comments alone it should be obvious that I too have worked with similar things, I suppose.

The process of translation has been much more educational for me than reading reams of tutorials, especially since it seems all the tutorials are likewise geared toward the sdk example.


I'm just saying that you shouldn't take the SDK examples as good examples of how an engine should be, because they aren't. They just demonstrate one thing, in a simple manner. You should design your own engine, and fit the technology in there, and optimize it yourself. And if you want to optimize things, you can get a lot of information from the NVIDIA website, they have a lot of papers and presentations, that are very useful.
Yes this is indeed constructive criticism, like it or not.
Posted on 2004-02-11 05:31:13 by Henk-Jan
Personally I see more in this example than just "one trick - skinned mesh"... There's four different blend methods being applied in this example, and a decent engine would scale to the available hardware by selecting the most appropriate method.
Hardware context-sensitive casecode is only one step away from the example posted - which makes this subengine a lot more useful than you give it credit for.
Render code was translated, still working out the issues created by my preferring to use existing camera code rather than implement arcball.
Have a nice day :)
Posted on 2004-02-15 21:56:16 by Homer