PDA

View Full Version : I have a demo, and it needs testing



Twixn
05-02-2005, 09:01 PM
Hey all.

I've been working on a demo for quite a while. and mostly i wanna see if it works on systems other than my own.

Subliminal (http://www.twixn.com/downloads/SBML.zip)

Its an FPS, Using Cg, OpenGL, OpenAL and Newton. Its about ~7.5mb.

So please, if you test it can you post your stats and if it worked.

and if you have any comments (i hope you do, IMO the demo is pretty cool) please share.

ATM, the demo has RagDoll and rigid body physics, Sound and 4 levels of Detail for the Map Model (Bump mapping, Bump and Parralax mapping, Bump and Relief Mapping and Bump Self-Shadowing Relief mapping), Hardware Accelerated Stencil Shadowing, and 3 different guns, switch using the 1,2,3 keys...1. Steyr Aug, 2. Grav Gun (has 2 fire modes, use RMB to 'grab' objects) and 3. Pulse Rifle.

And also some post-processing effects to make the Grav Gun cooler :p

Thanks for any input.

-Twixn-

RigidBody
05-02-2005, 09:19 PM
sorry that i can't say anything about framerates etc.- i have linux running and no windows installed...but i looked at the screenshots, looks good.
although my first impression was: the gun is too coarse. the walls are really cool to look at, but actually you move around and don't stare at the walls all the time.
what you see most of the time is your gun, so it should be modelled very fine.

Twixn
05-02-2005, 09:25 PM
Yeah, the guns are fairly low-poly...but they're not mine, they belong to a friend of mine who is letting me use them.

But i should point out that the screenshots are old and outdated...they dont include the grav gun nor the Pulse rifle, the grav gun doesnt look so good, im hopeless at texturing, so i tried to make up for it with visual effects :p .

-Twixn-

RigidBody
05-02-2005, 09:37 PM
tell your friend he's a rotten egg- can't be that hard to make a telescopic sight with more than 8 edges on the circumference :D

well, sorry i can't say more. maybe i'll put windows on my HD again some day... :p

Twixn
05-02-2005, 10:10 PM
Well, its too early in development for me to care so much about content...its actually a game engine instead of a game itself.

Mostly, im trying to see if it works on ATI cards, but im interested to see its performance on other NV cards (works on 6600GT PCIe).

-Twixn-

Ffelagund
05-03-2005, 01:17 AM
Works perfectly on my WildCat Realizm 100. Only has one annoying behaviour: try to set the cursor position at the center of the game window on every frame, because when the cursor arrow goes out of the window, the user cant fire without bringing up a background window, loosing the focus of the game window.

uruk
05-03-2005, 01:30 AM
~25 fps on FX 5200
Realy cool
May I suggest adding a crosshair?

Twixn
05-03-2005, 01:56 AM
Thanks for your suggestions (im writing them all down 'todo.txt' :p ).

Just make sure you try the Gravity Gun (pressing the '2' key brings it up, right mouse button to 'grab', left to 'push' away), im especially proud of that.

And im glad to hear it works on other cards. Still need ATI to complete the major vender compatability list :p

-Twixn-

sqrt[-1]
05-03-2005, 04:47 AM
Seems to run OK on my 6800GT.
But I noticed a few OpenGL errors keep spewing when running:

GL ERROR - Function glFrontFace(GL_LESS) generated error GL_INVALID_ENUM
GL ERROR - Function glMatrixMode(GL_MODELVIEW_MATRIX) generated error GL_INVALID_ENUM
GL ERROR - Function glDisable(GL_ZERO) generated error GL_INVALID_ENUM
GL ERROR - Function glPopMatrix() generated error GL_STACK_UNDERFLOW

The above was generated using GLIntercept (http://glintercept.nutty.org ) which can be configured to break into your debugger when these GL errors occur.

Edit:
Found one more:
GL ERROR - Function glBindTexture(GL_TEXTURE_2D,6) generated error GL_INVALID_OPERATION

This occurs because Texture ID 6 is a Rectangular texture. You app only works as it look like the texture is still bound from when it was sub-loaded from the screen.

You also make a lot of render calls using the same textures/shaders straight after each other (mainly drawing the gravity gun etc) Since these are raw triangles (ie not in a strip), you can easily batch them together in one render call. (not necessary, but better for performance)

Twixn
05-03-2005, 05:30 AM
Ok, thanks for that.

I didnt realize it was THAT messy inside.

-Twixn-

BoSoft
05-03-2005, 06:04 AM
Doesn't work on my Radeon 9800Pro.
Almost everything is black. I can just see the lights and the boxes, but it seems like your modelview matrix isn't correct ( maybe caused by the errors posted by sqrt[-1]? ). If you look up and down the boxes get distorted.
The weapon also doesn't move correctly with the view position and direction (maybe also caused by the matrix error?)
This happens on detail level 1 and 2.

-FA

Sunray
05-03-2005, 06:55 AM
http://sunray.cplusplus.se/dump/sbml.png

If this is how it should look, then I'm not impressed. :-)

ATI Radeon 9700 Pro, Cat 5.3.

Zengar
05-03-2005, 07:06 AM
Well, it's pretty nice, but too dark

and wooden boxes behave like rubber :-)

ZbuffeR
05-03-2005, 10:08 AM
GF6800le Pretty nice, but if you can do self shadowing relief mapping, you can surely do trilinear and anisotropic texturing ... :)

And if you have dynamic lighting, adding a light for a gun shot would be easy to add. And what about a shadow for the player ? I found it sad that Doom3 player shadow was not 'on' by default.

Twixn
05-03-2005, 12:14 PM
Bugger...i was hopeing that wouldnt happen.

Thats the reason i wanted ATI cards to test it so much, im pretty sure sqrt[-1] got it right.

I'll add the modifications to the lighing soon enough :p

-Twixn-

Twixn
05-03-2005, 01:06 PM
Ok, if you have an ATI card give this a try...its fixed (with reguard to the problems sqrt[-1] spotted)
SubliminalDev.exe (http://www.twixn.com/downloads/SubliminalDev.exe)

The whole demo has been updated, but if you already have it, theres no point downloading ~7mb again.

And also, im pretty sure Detail Levels 3+4 dont work on non-GF6 cards, Subliminal should tell you, and revert back to the hightest one you can handle.

-Twixn-

SirKnight
05-03-2005, 03:23 PM
Cool gravity gun! If I only had some enemies to throw sh*t at. :D

My fps were around 40 to 50 on my 6800 GT with 71.84 drivers at 1024x768 resolution. Seems a bit low for what was in the scene but I'm sure there is a good reason for it. ;) Still pretty cool though.

-SirKnight

Twixn
05-03-2005, 03:37 PM
Well, hunt down the ragdoll, and grab his foot :p (altho, no cross hair makes it abit hard, takes about 2 or 3 goes to grab it properly)

I originlly set 'Level of detail' to level 4 (which is self shadowed relief mapping, on high res that would eat FPS), so the low FPS is not surprising...goto the config file, and change it there, you should get about 200fps (i get about 250 on my 6600GT, 800*600)

-Twixn-

SirKnight
05-03-2005, 04:04 PM
Ah that was the ticket! Ok so yeah since relief mapping was on it makes more sense now. :) I did mess with the config file before I ran the engine but for some reason I forgot that 3 and 4 turn on relief mapping, heh. On setting 2 I then got 60 to 120 (still at 1024 res).

Also I noticed that you can't launch the ragdoll with the LMB like you can with the boxes. Is that a "feature?" ;)

Oh and one last thing. When using level 4, I do not see the bumps casting any shadows (selfshadow).

-SirKnight

Twixn
05-03-2005, 04:10 PM
The ragdoll is stil abit buggy, when huge forces (like throwing it) are put onto the ragdoll, it freaks out and becomes unstable...turning that off seemed the right thing to do for now :p

The shadows are there, press the 'L' key to make the light move around and have a look at some of the walls.

-Twixn-

fpo
05-03-2005, 05:38 PM
Good work... looks great on my 6800GT!! The relief mapping option rocks! And can not see artifacts... how many samples in lenear/binary search do you use?

Just can not move around as I'm used to inverted mouse in Y dir... can you add a mouse Y invert flag?

sqrt[-1]
05-03-2005, 06:46 PM
I quickly tried the new demo on a 9800, same as the screen shot above. It looks like the vertex buffers are junk?

(Note that Nvidia usually does a lot of "nice" things for programmers like initializing data to some state. Perhaps you are not setting some shader varible correctly?)

SirKnight
05-03-2005, 07:41 PM
Ah there's those pesky shadows. Making the light animate did the trick. :)

-SirKnight

michagl
05-03-2005, 08:12 PM
The above was generated using GLIntercept (http://glintercept.nutty.org ) which can be configured to break into your debugger when these GL errors occur.
examining the source it looks like this system could be used to intercept for instance geometry dispatches and redirect the output to a disk or just about anything else as it is already capable of doing with texture data.

are commercial opengl apps able to side step the interceptor? or is this a serious legal dispute? if it is an issue, how might it effect the adoption of opengl as a graphics platform?

Twixn
05-03-2005, 10:44 PM
Bugger...i've been working with GLintercept and i cant find any more bugs with that (out of 100mb of GLintercept log, there are only warnings about not releasing texture memory).

I'll check the shader inputs, hopefully thats it.

Fpo: I'll ad an inverter flag for you :p
And it uses 10 linear steps, and 5 binary steps. It does get some artifacts, but with Multi-sampling (and any other crazy filtering techniques, it works great).

-Twixn-

Well, i posted another exe:
SubliminalDev.exe (http://www.twixn.com/downloads/SubliminalDev.exe)

This one is more debug...Cg will not try go beyond ARB_f/v_p (like NV extensions), and it will print all debug output from compiling shaders.

It does of course produce no errors for me (a few warnings about stuff Cg shoves in, and then cant recognize, but that shouldnt do anything).

I don know if it will help, but getting any ATI debug might be useful.

V-man
05-04-2005, 06:48 AM
Originally posted by michagl:


The above was generated using GLIntercept (http://glintercept.nutty.org ) which can be configured to break into your debugger when these GL errors occur.
examining the source it looks like this system could be used to intercept for instance geometry dispatches and redirect the output to a disk or just about anything else as it is already capable of doing with texture data.

are commercial opengl apps able to side step the interceptor? or is this a serious legal dispute? if it is an issue, how might it effect the adoption of opengl as a graphics platform?Anybody can side step a DLL. For example, use FindFile for opengl32.dll. If it's there, then close the app.

BoSoft
05-04-2005, 07:08 AM
Well, i posted another exe:
SubliminalDev.exe (http://www.twixn.com/downloads/SubliminalDev.exe)
...
Compilation Report for "Media\ModelBumpParrallax.cg->Vertex":
(0): warning C6504: Unknown profile option 'NumTemps' ignored
(0): warning C6504: Unknown profile option 'MaxInstructions' ignored
(0): warning C6504: Unknown profile option 'NumAddressRegs' ignored
(0): warning C6504: Unknown profile option 'MaxLocalParams' ignored
215 lines, 4 warnings, 0 errors

same warnings on "Media\ModelBump.cg->Vertex"
and "Media\ModelShad.cg->Vertex"

fpo
05-04-2005, 10:31 AM
I have another feature request for you demo... could you allow changing the levelDetail parameter while running the demo in order to compare results from differen techinques at the same view point?

Another suggestions would be the invert mouse y (can't move around like that) and some brighter lighting (could use light color >1 for overbright maybe if color textures are too dark).

And last version gets me many warning message boxes until the app runs... annoying.

Again good work there!! It is fast and looks good. I have a new version for the relief mapping using objects space depth (specify depth in scene units and not the aspect ratio factor) and more correct shadows (only with object space depth you can get the exact self shadows). I will post it asap... testing a new hierachical ray intersect routine now...

Twixn
05-04-2005, 12:05 PM
To FullAmmo..thats expected, thats what i was talking about. Thanks anyway.

To Fpo...Those warning messages were the reason i posted it, i enabled the debug messages to hope it will help with fixing the ATI problem. I'll get to inverting the y, and perhaps the lighting, but atm im concentrating on making it work for ATI. And i look forward to your new demo :p

-Twixn-

michagl
05-04-2005, 12:14 PM
Originally posted by V-man:
Anybody can side step a DLL. For example, use FindFile for opengl32.dll. If it's there, then close the app.isn't a dll named opengl32.dll necesarry though? couldn't you stick the intercepter in your system path where it would expect to find opengl32.dll. how could it tell the difference?

are opengl apps not required to go through opengl32.dll?

knackered
05-04-2005, 12:48 PM
All opengl32.dll does is call LoadLibrary with the filename of the installable-client-driver which it gets from the registry. It then uses GetProcAddress on that ICD to fill in the function pointers it exports to applications.
This means all an application has to do is link to the opengl32.lib export library to make an explicit dependency on opengl32.dll (explicit means the application will automatically attempt to load opengl32.dll).
DLL's are searched for in the following order (I think):-
- current modules directory
- every path in the 'path' environment variable (which will be the user path concatenated onto the system path).

But yes, michagl, I suppose you're right - to get around the application searching its current directory for a copy of opengl32.dll you could indeed put the glinterceptor frigged opengl32.dll somewhere else in the system path, so long as it's before the system32 directory.

dorbie
05-04-2005, 01:57 PM
You can explicitly load a dll and prevent this that's what's being suggested, heck glquake (quake2) did this on the fly from day one with a local minigl lib, however you could equally well replace the opengl32.dll in the system directory and/or rename it and still call thorugh it.

Yup your stuff can be nicked and there's not much you can do to prevent it if you want code to work on a bunch of PCs & drivers, get over it, your low level 3D code ain't that valuable or useful and anyone who thinks it is needs a reality check. There are hundreds of games out there, you don't see their content & low level 3D code being ripped off willy nilly and where it is used to a limited extend it is practically encouraged through mods etc.

michagl
05-04-2005, 05:38 PM
Originally posted by dorbie:
You can explicitly load a dll and prevent this that's what's being suggested, heck glquake (quake2) did this on the fly from day one with a local minigl lib, however you could equally well replace the opengl32.dll in the system directory and/or rename it and still call thorugh it.

Yup your stuff can be nicked and there's not much you can do to prevent it if you want code to work on a bunch of PCs & drivers, get over it, your low level 3D code ain't that valuable or useful and anyone who thinks it is needs a reality check. There are hundreds of games out there, you don't see their content & low level 3D code being ripped off willy nilly and where it is used to a limited extend it is practically encouraged through mods etc.i'm not at all concerned about any of my data being 'nicked'... if any thing i'm an advocate of open data and the enevitability of all data being open. i wasn't thinking so much about low level code, but i figured the possibility of content being ripped and traded about, or even stuck in repositories would cause an uproar amongst commercial entities.

as i understand you opengl geometry can be intercepted unless the drivers are compiled into the code and hince limited in portability.

i'm curious what percentage of games run on opengl drivers? and is there an equivalent intercept method with directx?

do the games that allow modding provide export utilities to common file formats?

EDIT: provided it is indeed possible to easilly rip data from games, especially games predating tools like the interceptor... personally i would think it would be more interesting to rip generic geometry rather than necesarrily intellectual property. would there be a big political uproar over people ripping generic geometry like furniture for instance and sticking it all in a publicly accessible database? anything like this going on?

dorbie
05-04-2005, 05:45 PM
content = data and the comment applies in general terms, sigh.

P.S. it's been possible for years, where's the uproar?

P.P.S. there's no such thing as generic geometry, 3D content is copyright the progenitor unless otherwise licensed. I've personally written code to rationalize vertex array dispatch into header files and array calls to send it. You absolutely can intercept everything, and write it of a file format or just write compilable OpenGL code, as I did. Even apparent problems like unpredictable array dispatch and data bloat can be corrected with the right approach.

Twixn
05-04-2005, 05:48 PM
LOL...something tells me that we're shifting off-topic here :p

I'll just put out another call to ATI Graphics Card owners, to help me by testing my Engine.
Download the 'SBML.zip' in my first post, and the 'Subliminal.exe' update in one of my early posts (or later :p )...i'd really like to get it working on ATI.

Thanks

-Twixn-

SirKnight
05-04-2005, 06:01 PM
Originally posted by Twixn:
LOL...something tells me that we're shifting off-topic here :p Gee I wonder why, lol. :)

I wonder if ATI could help you debug it if you sent them your code? Where's Humus when you need him. ;)

That's what's bad about not being able to have both ATI and NVIDIA cards to develop on. It's almost always the case where you dev on NV and it won't work on ATI, dev on ATI and it won't work on NV. Frustrating...

-SirKnight

Twixn
05-04-2005, 06:05 PM
Lol...what we need is NV and ATI creating Software Emulation for their drivers :cool: so i can test ATI compatability on my 6600GT :D

-Twixn-

michagl
05-04-2005, 06:10 PM
P.P.S. there's no such thing as generic geometry, 3D content is copyright the progenitor unless otherwise licensed.
i didn't say copyrighted geometry, just generic geometry, an iron short sword versus some super fantasy sword. you can't say that there is any original content in the generic data to copyright. anyone has a right to use that sort of data under any legal system. whether or not it can necesarrily be proven that the data was ripped is another issue. but all i was saying is would anyone care? care so little they would just let it go by without even an attempt to stifle sharing.



I've personally written code to rationalize vertex array dispatch into header files and array calls to send it.sorry, but i can't really understand what you are saying here? would you mind rephrasing if it is worth it?

PS: sorry about swaying a little bit off context, but glIntercept was brought up and used, and people would have a lot easier time producing eye catching demos such as yours if there was a fair quality public repository of free open format data on the internet. i mean isn't it time to start moving generic data into the public sphere in an equal opertunity way?

dorbie
05-04-2005, 08:43 PM
i didn't say copyrighted geometry, just generic geometry, an iron short sword versus some super fantasy store. you can't say that there is any original content in the generic data to copyright. anyone has a right to use that sort of data under any legal system.I know what you said, you're wrong.

There's no such thing as generic geometry, someone made it and unless they explicitly release it under some liberal copyright you don't have permission to use it.

sqrt[-1]
05-05-2005, 12:47 AM
OK, I tried it out agin on an ATI.
By setting the shader level to 1 (just bump mapping) and by making the following changes, it all seems to work. If you want further help from me, disable the mouse "system wide locking" as I cannot run my tools (ie runtime shader editor) when in this mode.

ModelBump.cg
Change the lines:

float4 tempPosition = float4( IN.position.x,IN.position.y,IN.position.z,1.0f );
OUT.position = mul( modelViewProj, tempPosition );
to


OUT.position = mul( modelViewProj, IN.position);(I have no idea why you would want to do the moving of 1 to w as there should already be a 1 there)

Performance tip: Always write the output position as soon as possible in the vertex program.

However, this seems like an ATI bug. So perhaps Humus can take a look? Below is the problem cg compile output:


PARAM c[6] = { program.local[0..3],
{ 1 },
program.local[5] };
TEMP R0;
TEMP R1;
MOV result.color, vertex.color;
MOV result.texcoord[0], vertex.texcoord[0];
ADD R1.xyz, -vertex.position, c[5];
MOV R0.xyz, vertex.position;
MOV R0.w, c[4].x;
DP4 result.position.w, R0, c[3];
DP4 result.position.z, R0, c[2];
DP4 result.position.y, R0, c[1];
DP4 result.position.x, R0, c[0];
DP3 result.texcoord[1].z, vertex.normal, R1;
DP3 result.texcoord[1].y, vertex.texcoord[3], R1;
DP3 result.texcoord[1].x, vertex.texcoord[2], R1;
END
@ The side thread in the thread:
My tool is not the first or only one. (and there are some nice D3D ones also) No professional programmer would risk using stolen assets anyway. (Virtually all game file formats are known anyway)
Want free models/textures? goto http://www.turbosquid.com/ (Just do an advanced search for models with starting and ending price of 0)

Twixn
05-05-2005, 01:02 AM
Ok...sorry, i'll put the mouse locking as a flag.

I'll put up a new zip of the whole thing in a min, with the mouse locking as an initially disabled flag in the config, no Cg debug crap, y-axis (well, more x-axis, but up<=>down) flag which is initially set as off in the config and the fixed shaders.

-Twixn-

Ok the new demo is up:
Subliminal (http://www.twixn.com/downloads/SBML.zip)

the changes are listed above.

dorbie
05-05-2005, 02:01 AM
How about some trilinear MIP map filtering and some and some anisotropic filtering. It needs it.

Ffelagund
05-05-2005, 02:06 AM
Las time I forgot to test the application with all levels of detail. Here are the result (I am using a Realizm 100 card)

level 1: all looks perfectly.
level 2 (parallax): the parallax effect only looks good when the camera vector is close to be perpendicular with the wall that has the parallax applied, here is a screenshot (http://www.typhoonlabs.com/~ffelagund/screen3.JPG)

level 3: The effect simple doesn't look as it should: screenshot (http://www.typhoonlabs.com/~ffelagund/screen1.jpg)

level 4: worst than level 3: screenshot (http://www.typhoonlabs.com/~ffelagund/screen2.jpg)

Twixn
05-05-2005, 02:21 AM
Dorbie: I'll get to that...just tryin to get ATI workin.

Ffelagund: :eek: bugger, the parallax looks alright, its just it has an unusually high bump scale.
Go to 'Media\ModelBumpParallax.Cg' and in the function 'Fragment' change

Offset = Offset * 0.03f - 0.015f;with something lower like:

Offset = Offset * 0.02f- 0.01f;As for Levels 3 and 4, i have no idea whats going on there...have you tried Fpo's Relief Mapping demo? or the Relief mapping shader that comes with 'Shader Designer' and 'Render Monkey 1.5'? if not, then give them a try...see if you get the same results.

Thanks for your help.

-Twixn-

dpoon
05-05-2005, 02:46 AM
Nice job. Runs at around 20-30 fps on my 6800GT using forceware 71.89 @ 1280 x 1024 using bump+relief+selfshadow. A few suggestions though: hide the mouse cursor, and add a crosshair; and smooth out the mouse movement - a bit too jerky for my taste.

Twixn
05-05-2005, 03:02 AM
Originally posted by dpoon:
hide the mouse cursor, and add a crosshair; and smooth out the mouse movement - a bit too jerky for my taste.I'll hide the cursor when the mouse is locked to the screen flag is on.

the jerkyness is mostly due to my temporary timing system.
I havent implimented high precision timers yet, so atm the just the amount of time passed per sec is just a flipped FPS counter (1/FPS ~= time-in-sec) it produces smooth (but varying) movement (as opposed to 'GetTickCount()' which is choppy)...that wll be taken care of soon enough.

Thanks for your help.

-Twixn-

Ffelagund
05-05-2005, 03:11 AM
Originally posted by Twixn:
Dorbie: I'll get to that...just tryin to get ATI workin.

Ffelagund: :eek: bugger, the parallax looks alright, its just it has an unusually high bump scale.Sorry, these last days I'm seeing bugs lurking me around the corners. :rolleyes:


As for Levels 3 and 4, i have no idea whats going on there...have you tried Fpo's Relief Mapping demo? or the Relief mapping shader that comes with 'Shader Designer' and 'Render Monkey 1.5'? if not, then give them a try...see if you get the same results.
Thanks for your help.
-Twixn-I'm the writter of Shader Designer(SD) :) ,and I ported the relief shader to the SD(it works perfectly here)

Twixn
05-05-2005, 03:52 AM
Originally posted by Ffelagund:
I'm the writter of Shader Designer(SD) :) ,and I ported the relief shader to the SD(it works perfectly here)LOL...ok, well...nice job :p

-Twixn-

dpoon
05-05-2005, 05:39 AM
There appears to be some clipping issues with the weapons. This is especially noticeable when you move up close to a wall or a crate. Anyway keep up the good work.

michagl
05-05-2005, 11:03 AM
dorbie i get your copyright to the point that i don't feel it is worth disputing here. if you want to copyright bits and bytes that is up to you, but you can't copyright a wooden chair like you can mickey mouse. simply moving vertices around and smapping out and modifying textures would probably be enough to totally remove any chance of provability save for some people trying to put unremovable undeniable watermarks in textures and significantly high resolution geometry.

as for turbosquid i ended up there once. the free data was all extremely low quality and too numerous to browse reasonably. if nothing else it needs to be able to sort the data against quality (at least i couldn't find such a feature as i recall and if was there it didn't work). the price tags are rediculous as well. i was looking for a fair humanoid model. its depressing that by the 21st century there is no freely available fair dataset of our very species publicly available and easilly accessible.

EDIT: glInterceptor and ShaderDesigner engineers... i should remember these names. i downloaded SD once but the file turned out corrupted and a bit too big for another try. how does SD stack up to nvidiaFXcomposer? as for intercepter just curious if there are already plugins available for capturing the geometry per frame and writing it out to a file?

dorbie
05-05-2005, 11:15 AM
I wouldn't dispute it either if you weren't repeatedly posting false information that needs to be corrected, it's not what I want, it's the law and copyright automatically kicks in these days(n/m the license). You cannot get a wooden chair from a software intercept. You can only get a polygonal chair model that someone created with their time and effort and owns the copyright to, same for the textures. The rest of your post is a description of how to cover up the deliberate theft of someone's work. That too is flawed information, there are hidden geometry & texture watermarking schemes (in addition to obvious flaws).

Ffelagund
05-05-2005, 12:38 PM
Originally posted by michagl:
glInterceptor and ShaderDesigner engineers... i should remember these names. i downloaded SD once but the file turned out corrupted and a bit too big for another try. how does SD stack up to nvidiaFXcomposer? as for intercepter just curious if there are already plugins available for capturing the geometry per frame and writing it out to a file?Shader Designer was made for me on my free time for free. It has already accomplished its objectives. I'll maintain and upgrade it, but do not expect big changes in the future.

We are trying to get enought funds to start another application, but this time will be a LARGE application. Think in the whole FX composer as a "feature" (and not the biggest one) of the "new shader designer" (it has no name yet). By this way you can imagine the magnitude of the application. But before, we need some kind of finantiation.

V-man
05-05-2005, 07:07 PM
I just tried it and it works fine except I saw a wall that was flickering. This is on ATI 9500 Cat 5.4. Either it's a multipass related thing or you have 2 walls in the same place.
How good is the newton?
I didn't see any enemies or dolls.

Also I noticed that you have the billboard clip problem that I wanted to solve :)

Good job indeed.

Twixn
05-05-2005, 08:17 PM
The walls are only drawn once (no multi pass)...and there are no 2 poly's in the same place, except for the shadow volumes...so if you can take a screen shot (of that and the billboard problem), i would appreciate it.

There are no enemies, but there is one ragdoll in the main room, close to one of the 'entrances'.

And also..did you try the old demo on your card an get crazy problems?

-Twixn-

PS, glad to hear its at least running on ATI, Thanks for your input, newton is very good IMO, its fairly fast but i think Novodex and maybe tokamak are faster...but newton is stable, AND easyer to use, than the both of them :p

Twixn
05-06-2005, 04:11 PM
Just informing you that there is a new demo (~7.5mb) (http://www.twixn.com/downloads/SBML.zip)
and a new bug fix (~50k) (http://www.twixn.com/downloads/SBMLfix.zip) . The Demo (SBML.zip) is new, but the bugfix is newer, so download both if your going to help me test it.

It should work on ATI cards...the main issue being that the shadows from the boxes may freak out.

Anyway, id appreciate any more input/comments.

Thanks again

-Twixn-

fpo
05-07-2005, 05:54 AM
Nice new demo... thanks for thy mouse flip Y option (now I can move around normaly).

Looks faster now and the alternative fire mode for the grav gun is great (with some sort of distortion shader around the blue aura). Cool effect... and it is nice to be able to grab things and then throw them around... even works with the rag doll. Good work!

Twixn
05-07-2005, 12:33 PM
Thanks, im particularly proud of the Grav Gun :D , although some ATI cards cant use the special effects because i forgot to support EXT_tex_rect as well as ARB_tex_rect, OpenGL2.0 and ARB_tex_npot.

-Twixn-

ok, the new fix is up:
SBMLfix.zip (http://www.twixn.com/downloads/SBMLfix.zip)
It just adds support for EXT_tex_rect so if you have a 9500+, and dont have OpenGL 2.0,you can see the post-processing effects, which is pretty much what fpo was talkin about.

Twixn
05-07-2005, 02:56 PM
For some reason, the shadows are having problems on ATI cards (just the boxes).

Here is a pic of it: http://img213.echo.cx/img213/8821/sbml7oz.jpg

I have no idea whats happening with it...has anyone else has a problem like this (Hardware Extruded Carmack's Reverse Stencil Shadows on ATI cards).

Any suggestions?

-Twixn-

dorbie
05-07-2005, 03:15 PM
Check the number of stencil bits you're getting on that card. I suspect that's your problem.

Twixn
05-07-2005, 03:37 PM
It only happends on the boxes tho...the Map and the ragdoll are unaffected...which makes me think its some kind of matrix problem...i'll post my shader here:


struct App2Vert
{
float4 Pos : POSITION;
float4 Nor : NORMAL;
};

struct Vert2Frag
{
float4 Pos : POSITION;
};

Vert2Frag Vertex( App2Vert IN,
uniform float4x4 ModelViewProj,
uniform float3 LightPos,
uniform float extrude)
{
Vert2Frag OUT;

float3 tlight;

tlight = normalize(LightPos-IN.Pos.xyz);
float compnl=dot(tlight.xyz,IN.Nor.xyz);

float4 mult;
mult = float(compnl < 0)*extrude;

float4 endVec = IN.Pos;
endVec.xyz -= tlight*mult;

OUT.Pos=mul( ModelViewProj,endVec );
return OUT;
}

float4 Fragment( Vert2Frag IN) : COLOR0
{
return float4(1.0f,1.0f,1.0f,1.0f);
} All the uniforms are being passed correctly...And it works fine on NV cards, and some ATI cards (not the card itself, some cards work for one person, and dont for another) i suspect its a driver problem.

-Twixn-

Twixn
05-08-2005, 11:30 AM
Well, is seems it was a driver problem...getting Catalyst 5.4 (6.14.10.6525) fixes it.

But, can anyone having problems with the shadows give them a try and tell me if it works.

Better to be safe than sorry.

-Twixn-

Elixer
06-16-2005, 10:43 AM
Originally posted by Twixn:
LOL...something tells me that we're shifting off-topic here :p

I'll just put out another call to ATI Graphics Card owners, to help me by testing my Engine.
Download the 'SBML.zip' in my first post, and the 'Subliminal.exe' update in one of my early posts (or later :p )...i'd really like to get it working on ATI.

Thanks

-Twixn-I just tried it with Catalyst 5.6, and still didn't work. :(
I get a black window, with a white streak at the top.

Twixn
06-16-2005, 10:05 PM
Well, i've been testing it on an ATI 9800, and it works on 5.4...install that and try it (also, it has changed quite a bit since then).

But first, the 'subliminal.exe' is outdated...try just the straight zip file (SBML.zip), no updates. Try this before you play around with your drivers.

-Twixn-

Elixer
06-17-2005, 08:44 AM
Originally posted by Twixn:
Well, i've been testing it on an ATI 9800, and it works on 5.4...install that and try it (also, it has changed quite a bit since then).

But first, the 'subliminal.exe' is outdated...try just the straight zip file (SBML.zip), no updates. Try this before you play around with your drivers.

-Twixn-Ok, tried the straight zip, and that works OK. :) Guess the issue was with the subliminal.exe that I overwrote from the zip.
Looks good. :)

gator
06-17-2005, 04:10 PM
Originally posted by sqrt[-1]:

The above was generated using GLIntercept (http://glintercept.nutty.org ) which can be configured to break into your debugger when these GL errors occur.
Excuse me, but what is a '7z' compressed file and how do I decompress it?
Why don't these people just use 'zip'?

Twixn
06-17-2005, 04:49 PM
Yeah, the subliminal.exe was a patch to an older version of the engine (and the patch didnt work anyway), so im not surprised it caused trouble :p , glad to hear the new one works with the newer ATI drivers too, thanks.

Although, i have been working on some work arounds in case it doesnt work, primarily if the user has a crappy card then i made this: SBMLPGL.zip (http://www.twixn.com/downloads/SBMLPGL.zip)
It uses PseudoGL rather than OpenGL, (PseudoGL being a software renderer i made).

-Twixn-

HellKnight
06-18-2005, 03:53 AM
~10 frames on a GF6600GT, Athlon64 2800+ 640x480 with relief mapping.

I got a little bit more (maybe 20) without RM (LOD = 1) with a resolution of 800x600.

Indeed that seems VERY low to me, since the map geometry is surely not hight poly. You do frustum culling ? :)

I'd like to know whether you're using lightmaps for the soft shadows, cause it seems so, in the other case you are throwing in a lot of lights in order to approximate soft shadows, which would explain th low fps :)

Trahern
06-18-2005, 04:28 AM
You have tested the last .zip he posted here?
As he said its software renderer so its not a surprise its slow :) .
When I tested the normal version ( on 6600 GT ) everything was running quite smoothly

HellKnight
06-18-2005, 05:13 AM
No, I tested the first one posted here... What you say is really interesting, but maybe the demo's bottleneck is the CPU, and come on, an Athlon64 2800+ isn't relly a fast one (maybe so much time is spent computing the physics, f.e.). In such a case optimzing them wouldn't be a bad idea...

Another possibility is a different interpretation of "smooth". As a Quaker, "smooth" means to me at least 60 - 70 fps :) However, I know "smooth" is understood to mean at least 20 - 30fps or so among mortals :) , which was definately NOT the case.

Trahern
06-18-2005, 05:30 AM
I downloaded the new version of this demo and now I see that its quite different from the previous one ( the previous one contained only some crates and a ragdoll and the speed diference between these versions is quite big. So you are right it is very slow now ( not smooth :) ).

Twixn
06-18-2005, 02:10 PM
Well, yeah is it slow atm, i have made some optimizations since then (Frustrum Culling, VBO's and 2-sided stencil as extrusion is done in shaders), however they are not in that zip, i havent uploaded a new version yet, i plan to do so when i get a few more weapons done and a vehicle...i have yet to do any technique like PVS or Portal (if you know any good articles on PVS please tell me :p ).

i have a GF 6600 GT, AMD64 3500, and it runs fine now (ranges from 50-120fps, 800*600 LOD=1), when i release the next beta it will be faster, not only because of the graphics optimizations, but the next version of Newton (the Physics engine i use) will have some good speed improvements.

I can get about 30-40 fps on the software rendered one (rendering AND physics on the CPU).

Anyway, Thanks for your input.

-Twixn-

Also, the soft shadows are just cubemaps, nothing special about them, cept they are a cheap and good effect

Elixer
06-19-2005, 09:56 AM
Originally posted by gator:

Originally posted by sqrt[-1]:

The above was generated using GLIntercept (http://glintercept.nutty.org ) which can be configured to break into your debugger when these GL errors occur.
Excuse me, but what is a '7z' compressed file and how do I decompress it?
Why don't these people just use 'zip'?7z is 7zip. Why use it? It offers MUCH better compression, is open sourced, and is pretty nice.

The same argument was made when zip took over for .arc files, people were complaining why don't they use arc instead of zip. :)