 |
Results & Tool Screenshot
|
25 January 2000 (7:15pm) - Tuesday Evening
|
Since the release of that compatibility demo a few days ago, I received some great feedback
from you guys. Thank you. Even if I didn't get a chance to reply to you personally, your
feedback is greatly appreciated. It looks like most of the mainstream cards are working very well
with the multi-texturing, and it apparently screams on a geForce :). I'm still a bit concerned because
if the demo detects that it can't pull off the proper combination of blending in a single pass for a given
card, its supposed to default to two passes with alpha blending. However I had two reports from people
mentioning that they were seeing only textures, one with an Intel i740 and the other with an ATI Rage
Pro. Another person with a Rage Pro mentioned that the demo did work for him, so I don't
know what's going on there either. I'll have to investigate this further, and I'll try to get these
problems resolved before the next release.
In other news, I wrote a nicer version of my lightmap building tool. Here's a shot, followed by a description:
Its a multithreaded application which allows me to view the lightmaps as they're
being built (including zooming in), so I can check for anomalies and such. It also
allows me to rebuild a single lightmap, or cancel the process at any
time (and keep what was calculated so far). The application mode, as shown in this
screenshot with the progress bars and such, lets me play around with the options while its working, but it slows down
the calculation speed. For this reason I've also added a mode just for die-hard
calcuations, like if I wanted to leave a huge data set to calculate high-quality
maps over night or unattended. This utility is actually turning out to be very handy for what I'm
trying to do.
Anyway, more later on. I have a LOT of engine optimizing to do. I'm nowhere
near happy with the engine's performance yet, but I've got plenty of speed-up techniques still to implement.
|
|
Grain Engine Compatibility Demo
|
23 January 2000 (4:50pm) - Sunday Evening
|
I've posted a compatibility pre-release of the grain engine.
There's not much of an actual
test level. The purpose of this demo is mostly
to get feedback from you on whether it worked
without error.
I should be making incremental updates to the
engine from time to time as more features make
it in. Expect a much larger and more impressive
map in the next release.
Reporting bugs - I know there are a few issues of preference
and a couple anomalies including very sensitive control, a 120 degree fov,
and an occasional 'pop' in the collision detection. These will be fixed
or changed and the rest will be optimized (so ignore the frame rate counter,
which isn't exact anyway). What I'm mainly interested in is whether the demo LOOKS RIGHT,
not so much performance at this time. The demo tries to use single-pass multitexturing,
but will resort to multipass if it can't. If it says "multipass" in the top left corner,
please send me an e-mail letting me know what kind of
card you have, whether the blending looked correct, and a rough average of the frame
rate. If you notice any other bugs or the blending doesn't look correct, please also
report this. When I say "looks correct", scroll down on this page and compare your results
with the screenshot listed. It should look very similar (besides the texture on the pillars).
Any comments welcome. Just don't forget to tell me what 3D card you have! Thank you very much.
Download the demo here: grain01a.zip (502k), or from the downloads page. Please unzip with directories and view readme.txt.
|
|
Octrees And Such
|
17 January 2000 (3:59am) - Early Monday
|
A few more things have been added to the engine. I wrote a pretty nice
texture sprite class and also a new octree implementation. Here's another screenshot, this
one of the octrees "in action" (showing the octree cubes).
You can view it, with a description right here. Its sort of hard to see
what's going on in the image without being able to move around for yourself, but it'll have to do.
Its really pretty meaningless with such a simple world anyway, but that's something I'm working on.
Octrees are pretty neat. For those who don't know, an octree is a data structure that can
be used for a variety of applications, ranging from color quantization to assisting in 3D
visibility or more efficient collision detection.
For more information, perhaps try Jaap Suter's tutorial on flipCode (which you can
find here), or send me an e-mail.
|
|
Early Screenshot
|
16 January 2000 (12:35am) - Early Sunday
|
Just to show you a bit of what the engine's looking like, I took an early screenshot.
Its not much of a level (as I mentioned in my previous update), and its kind of dark,
but you can see the lightmaps and curved thing in the back. There are 2 lights in the scene.
Much is still in the works or to be changed.
Click the image above for a larger version of the screenshot as well as a brief description.
|
|
General Progress
|
15 January 2000 (3:33pm) - Saturday Afternoon
|
So I've been making a lot of progress recently with the engine. Collision detection
is in, as are lit curved surfaces. I'm still using lightmaps for my engine. I've
seen people ramble to no end about many great "new" lighting algorithms for real-time
dynamic lighting, and that's fine. I do plan to add dynamic lighting to the engine,
but I will do so on top of the static lightmap architecture. I just can't pass up such
great quality in a pre-processing step, and still, for my purposes, I don't plan on
having too much dynamic world geometry. Walls aren't going to be up and
moving to other sides of the world, and pillars shouldn't be walking around too much.
For these reasons I've kept the dynamic lighting (and probably shadowing) system open
and seperate.
I purchased a TNT2 recently. I'm waiting a bit on the geForce cards, especially since I got
a great price on the Diamond V770. I must say that this hunk of hardware is amazingly
fun to work with. My old setup, I had a Matrox Millenium II as my primary, and an 8
meg Voodoo2 (Monster2) as my accelerator. That was fast, but also annoying for various
reasons, one being that it (the Voodoo2) couldn't render in a window. Every time I
ran a DirectX app, it did this horrible freeze-black-screen-flicker when it enumerated
devices, and when the program initialized. That's all gone now and lends to much
faster testing of changes. With my new card I can just run it in a window with no
delay. Very convenient, smooth, and the rendering is beautiful. I'd recommend this card to anyone.
Other things. I started work on a tech file update about collision detection, but I don't
know if I'll bother post it. I'm also still working on a demo of the engine to upload,
but I need to finish a few things, such as making a good test level when I have some time. I've
been working with very simple worlds so far, but the engine can handle some pretty complex
things. I'll of course update this page when I have a preliminary demo.
|
|
Back At School
|
05 January 2000 (6:35pm) - Wednesday Evening
|
|
Well, I'm back at school (from winter break) which means I'll have far fewer distractions
to keep me from my code. Hopefully I'll get some real work done soon. More updates to come.
|
Previous Pages Of Update Entries:
December 1999
November 1999
|
Screenshots For This Month:
|
copyright (c) 1999 kurt miller. all rights reserved.
|
