PDA

View Full Version : 2006 PGD Annual: stage 3 scores...



tux
27-03-2006, 04:33 PM
Those 2 Guys
Torture Tank (01 Mar, 2006)

Eric Grange:
Add basic enemies to game engine. 5 / 5
Add player pick-up items and/or weapons to engine. 5 / 5
Design a better level to test out new features. 5 / 5
Provide an 'exit' for the level. 0 / 5

Judge Total = 15 / 20

Judge Comments:
Couldn't find an exit, or it didn't trigger after killing all enemy tanks I could find.


the exit is right next to the power up. its a cube with the words "EXIT" written on it.

McCLaw
27-03-2006, 06:12 PM
Oh well, its only 5 points, better luck next time I say :wink:

Huehnerschaender
27-03-2006, 07:00 PM
Hmmm.... I think if the exit IS there (the judges can test it by playing again), then the 5 points should be granted. I don't see a reason why there is a substraction of points if the goal was achieved, but the judges just didn't see it.

Greetings,
Dirk

McCLaw
28-03-2006, 06:10 AM
Hmmm.... I think if the exit IS there (the judges can test it by playing again), then the 5 points should be granted. I don't see a reason why there is a substraction of points if the goal was achieved, but the judges just didn't see it.

Greetings,
Dirk

That is true, but its the judges decision, me thinks. :)
We'll wait and see what happens :)

Eric
28-03-2006, 07:00 AM
You mean that dark box behind the bright pickup?

http://upload4.postimage.org/84904/Clipboard01.jpg (http://upload4.postimage.org/84904/photo_hosting.html)

Couldn't read anything on it, and since it wasn't mentionned in the readme, I didn't really noticed or looked for it... After you pointed me to it I tried and it exits indeed, but there is only one position where you can read what's written on it: that's when coming downhill (ie. on the other side than the one on the screenshot), and then, only when you're very very close to the box (nVidia 6800GS, latest drivers). Since it is in the middle of the enemies, and dark like the ground, I guess the forever looping explosions from shooting the enemies also contributed to keep it hidden even more... So I just kept destroying enemies and touring the map looking for something more prominent...

I'll update the score, but try to mention "hidden" goal features in a readme next stage ;)

McCLaw
28-03-2006, 08:27 AM
You mean that dark box behind the bright pickup?

http://upload4.postimage.org/84904/Clipboard01.jpg (http://upload4.postimage.org/84904/photo_hosting.html)

Couldn't read anything on it, and since it wasn't mentionned in the readme, I didn't really noticed or looked for it... After you pointed me to it I tried and it exits indeed, but there is only one position where you can read what's written on it: that's when coming downhill (ie. on the other side than the one on the screenshot), and then, only when you're very very close to the box (nVidia 6800GS, latest drivers). Since it is in the middle of the enemies, and dark like the ground, I guess the forever looping explosions from shooting the enemies also contributed to keep it hidden even more... So I just kept destroying enemies and touring the map looking for something more prominent...

I'll update the score, but try to mention "hidden" goal features in a readme next stage ;)

HEHE, thanx Eric.

It was abit daft of us to not put it on the readme, LOL

McCLaw
28-03-2006, 08:29 AM
OUCH!!! :shock:

It does come out VERY dark on your screen. I will have to make a plan with that :shock:

Eric
28-03-2006, 08:55 AM
I'm getting the same luminosity - or lack of :wink: - on the 6800GS and the 5900XT, could be something with nVidia hardware/drivers, I can give it a try on the x600-based laptop later today.

Also on the AXP2200+ machine (512MB ram, with a 5900XT with 128 MB), I've had the game startup take 40-50 seconds on first execution (subsequent executions were faster) as it had to swap out other proggies from the memory: during load time, the game requests more than 300 MB (not sure of the exact figure, but there is a definite spike in the taskmanager). When running, it uses about 120 MB, which is reasonnable.
The 'texturem.jpg' is probably the one to watch, it should be about 50 MB when decompressed, chances are you have quite a few copies of it in memory at some point. I resampled it to 1024x1024 and the game loaded and started instantaneously (and framerate didn't change that much), so it's definetely seems to be something that happens at load time, as the hardware can handle the high-res texture fine at render-time.

tux
28-03-2006, 09:03 AM
on my dads computer (p4 2.4ghz radeon 9200) the level is really bright. on my computer (amd64 3000 radeon x800) its dark. but i can clearly see the exit on the box.

i think its just the texture causing the dark level. but it loads within 5 seconds on my machine :) but the image quality on that screenshot is really bad :/

McCLaw
28-03-2006, 09:05 AM
I'm getting the same luminosity - or lack of :wink: - on the 6800GS and the 5900XT, could be something with nVidia hardware/drivers, I can give it a try on the x600-based laptop later today.

Also on the AXP2200+ machine (512MB ram, with a 5900XT with 128 MB), I've had the game startup take 40-50 seconds on first execution (subsequent executions were faster) as it had to swap out other proggies from the memory: during load time, the game requests more than 300 MB (not sure of the exact figure, but there is a definite spike in the taskmanager). When running, it uses about 120 MB, which is reasonnable.
The 'texturem.jpg' is probably the one to watch, it should be about 50 MB when decompressed, chances are you have quite a few copies of it in memory at some point. I resampled it to 1024x1024 and the game loaded and started instantaneously (and framerate didn't change that much), so it's definetely seems to be something that happens at load time, as the hardware can handle the high-res texture fine at render-time.

I use a 5600Gt to test with, and it does look darker than my 6800GT (It died a few weeks ago)

I'll have to implement a gama correction thing in the settings for the end of the current stage.

Looking at the code 'texturem.jpg' is loaded into a data pointer, and then loaded into video memory and the the data pointer is nulled, so there could be a jump of 100Mg in the memory while it is being loaded. Rescaling the texture is probably the only way we'll get passed that problem.

We had a few setbacks for the last stage though :(

We lost our designer(he had to start working again)

And I lost my one PC 5 days before the deadline, so had to set up my old PC to carry on working.

Needless to say, We dont expect to do well on our last submission. :oops:

You will be pleased to know that we sorted out at least some of the explosion problems though :D

Eric
28-03-2006, 10:11 AM
Looking at the code 'texturem.jpg' is loaded into a data pointer, and then loaded into video memory and the the data pointer is nulled, so there could be a jump of 100Mg

I'm seeing 130 MB in-game, and a 300 MB spike at load time, so that's a 150MB+ spike, at least room for 3 uncompressed versions... and wouldn't freeing the datapointer only release about 50 MB?

:idea: if the texture isn't loaded directly but passed to a mipmap generator (like GLU's, or driver's mipmap autogeneration), and that generator works on 32bpp images only and you pass it a 24bpp bitmap, that might explain things, as we would then see the following temporaries:
- 50 MB (4096x4096 @24bpp) bitmap loaded by the code
- 64 MB (4096x4096 @32bpp) mipmapper's work copy of the bitmap
- 30 MB mipmap temporaries
Total: approx 150 MB...

McCLaw
28-03-2006, 10:15 AM
hmm, thats true

I do gen mipmaps using glu. Perhaps using DDS textures would be better, but those we'll only have after the PGD contest is complete

Eric
28-03-2006, 12:34 PM
Pretty much all hardware now supports SGIS_generate_mipmap extension, which asks the hardware to generate the mipmaps autmagically, it will probably be more efficient at it than GLU. Code snippet:


if GL_SGIS_generate_mipmap then begin
// hardware-accelerated mipmap generation
glTexParameteri(GL_TEXTURE_2D, GL_GENERATE_MIPMAP_SGIS, GL_TRUE);
glTexImage2d(GL_TEXTURE_2D, 0, textureFormat, width, height, 0, GL_RGBA, GL_UNSIGNED_BYTE, buffer);
end else begin
// software GLU-based mipmap generation
gluBuild2DMipmaps(GL_TEXTURE_2D, textureFormat, width, height, GL_RGBA, GL_UNSIGNED_BYTE, buffer);
end;

Replace GL_RGBA with whatever format the buffer is in (GL_BGR if a straight 24bpp windows bitmap f.i.), textureFormat is one of the hardware texture formats GL_RGB8, GL_RGBA8, etc.

McCLaw
28-03-2006, 12:45 PM
Thanx eric, I'll add it, and see if it makes a difference. :)

McCLaw
29-03-2006, 08:04 PM
Hey Eric

I sorted out the dark map problem using "some form of gamma correction".

It is now configurable from the setup sceen, and also changable by hitting "G".

Tried your fix for generating the mipmaps, and it is ALOT faster, there is one problem though :(

For some reason, when I use the GL_SGIS_generate_mipmap extenstion, some of the textures fail to render :shock:



if Format = GL_RGBA then
begin
if GL_SGIS_generate_mipmap then begin
// hardware-accelerated mipmap generation
glTexParameteri(pTarget, GL_GENERATE_MIPMAP_SGIS, GL_TRUE);
glTexImage2d(pTarget, 0, GL_RGBA8, Width, Height, 0, GL_RGBA, GL_UNSIGNED_BYTE, pData);
end
else
begin
// software GLU-based mipmap generation
gluBuild2DMipmaps(pTarget, GL_RGBA8, Width, Height, GL_RGBA, GL_UNSIGNED_BYTE, pData);
end;
// Original code
//gluBuild2DMipmaps(pTarget, GL_RGBA8, Width, Height, GL_RGBA, GL_UNSIGNED_BYTE, pData)
end
else
begin
if GL_SGIS_generate_mipmap then begin
// hardware-accelerated mipmap generation
glTexParameteri(pTarget, GL_GENERATE_MIPMAP_SGIS, GL_TRUE);
glTexImage2d(pTarget, 0, GL_RGB8, Width, Height, 0, GL_RGB, GL_UNSIGNED_BYTE, pData);
end
else
begin
// software GLU-based mipmap generation
gluBuild2DMipmaps(pTarget, GL_RGB8, Width, Height, GL_RGB, GL_UNSIGNED_BYTE, pData);
end;
//Original code
//gluBuild2DMipmaps(pTarget, GL_RGB8, Width, Height, GL_RGB, GL_UNSIGNED_BYTE, pData);
end;


Any ideas on what I'm doing wrong?

Eric
30-03-2006, 05:52 AM
some of the textures fail to render

As in "they're all black/white/transparent" or "I get some crash/exception"?
Does it affect some textures randomly or always the same? (if so which?) If it's "tankbarr.jpg" it may come from the image dimensions (which aren't powers of two).
Also if you have other textures that don't need mipmap (or don't even work with mipmaps), don't forget to


glTexParameteri(pTarget, GL_GENERATE_MIPMAP_SGIS, GL_FALSE);

at some point, or they'll get mipmapped too.

McCLaw
30-03-2006, 05:58 AM
You are right, all the textures that render white are non power of two.

I was hoping I could get away with not having to stick to that antiquated rule. :(