Oh no! It's finally happening!
Morrowind will become more sophisticated than my computer!!!1 :banghead:

And I thought as long as I could run Crysis well I would be set for years...

I use XP, but voted for DirectX 10.0, because I can put the windows 7. Also, I've seen DirectX 10 GraphicX, adds support for DirectX 10 for XP. (Not set)
And will be animated trees in the next version? as grass is now. Or is it impossible without the use of tessellation of the above?

Voted

SM 4.0
DX 10
Windows 7

Additional data:
32 bit system
2gb ram

My oldie results

http://www.gpureview.com/GeForce-9400-GT-card-578.html

(oh no! my card is two years old!) XD

DirectX 11
SM 5.0
Windows 7

Also as a side note, it's interesting to see that 7 has inched out XP.

It is good to see you Liztail. Welcome back. :celebration: I started to write simple shaders since you're gone.

If you need ideas, I have this huge list which I paint in green more and more. Thanks to MGE team.

http://mgeshaderlib.50.forumer.com/viewtopic.php?p=385#p385


I can't see a mesh overhaul for Morrowind, it would be hard for modders I think. Because of that tessellation would be awesome. But the problem with tessellation is how to tell hardware what not to tessellate.

For terrain shadows, you mentioned the problem: "distant terrain is a different shape than the local terrain". If we fix it, we can have terrain shadows then. (plain logic :toughninja: ) I know it would be impossible, but how bad could it be? How does Crysis do it?

For statics shadows, creating a light perspective camera for distant land would be the start as you mentioned. I know they are going to be static but they are already statics. For creatures and NPCs maybe we can capture Morrowind shadows and apply some blur on them. Blurring shadows is very common. Probably we will do something like VSM shadowing.(actually they are all the same thing )

From top of my personal list,
true HDR, even better than any game, (Right now, the light info is between 0-1, doing tone-mapping and HDR is futile. There is not enough light info for good results. If we can give original lights and reflections more brightness(and sky less) so they exceed 1, after that we can do proper HDR lighting. No more over-exposured NPC faces. )

Shadows,

and whatever other games are doing to run 2000-5000 lines of shader code in real time. I have six post-process shaders and they hurt my performance a lot. And I want to add more!


It is more like "working for the dark side." I will finish the cloud shadows shader.

BTW, I'm in the lucky minority, I have Nvidia GeForce 240M GT, DX 10.1 and SM 4.1.

Im just curious, but what will happen to MGE development when/if OpenMW is released? Are the two projects in cahoots atall?

HD4770, so DX10.1 and SM4.1. But err... I've seen some people having DX10-capable hardware voting DX9, because they use Windows XP. The poll however specifically states the maximum DirectX level you card supports. Isn't this kinda... distorting the poll results?

About Multi-threaded rendering, if it would someday be implemented, this would probably only affect MGE-related stuff, and not Morrowind itself right?

And yay, LizTail's back

Not sure what you voted, but I'm one of the three that has a card that only supports SM 2.0 or older

(Actually, only 1.4 :bolt:)

Voted:

DirectX 9.0c
SM 3.0
XP SP2

Soon i'll be switching to a better GPU, with DX 10 and SM 4.0 (9800GT), though I won't use DX 10 because of XP.

Yup, exactly! I hadn't heard of anyone implementing global illumination. Do you have a link to it?



It already supports it, but Vality never finished releasing the version of his mod that supports it, AFAIK. You can find videos of it on YouTube, though.



Yeah, and Vista + Win7 is almost 2/3 so far. That's interesting because DX10+ uses a new and better API which also has backward compatibility for DX9 hardware, but it's only availiable in Vista and Win7.



Thanks! ^_^ To to make it possible to cast terrain shadows, we'd probably need to have a full resolution copy of the height map available to MGE at runtime and then create and destroy chunks of it as you walk around and use those to cast shadows on Morrowind terrain.


Last I checked it still had a long way to go. Both projects are open source so they can steal code from MGE if they want, but I don't plan to work on OpenMW right now.


Yeah, please vote based on what your card supports, not what your OS supports. OS is a separate question and I'm fully aware that XP doesn't support DX10+. Thanks!

Sorry if my comment was misleading, I suppose how I meant to word it was that Peachykeen wrote code pass lighting information to shaders -- his intent was to get normal mapping working, and from what I understand, he got fairly far along. If he (or someone else) doesn't post a link to his source by the time I get in later, I'll see if I can dig it up for you.

Same here.
:sadvaultboy:

WinXP SP3
3 GB ram
nVidia GeForce 9800 GT
SM 4.0
DX 9.0c card is capable of DX 10

Ati Radeon HD 4850 and Windows XP 64 bit, so that's DirectX 10.1 and SM 4.1. I hope to upgrade to Windows 7 soon to take advantage of DX 10 features.

DX 10
SM 4.0
Windows Vista 64....will fully upgrade to 7 next year when I build my new rig...waiting on AMD.....

just recently upgraded from the geforce 7600 to a geforce 9600 GT so im one of the direct X 10 and VS PS 4.0 voters though im not about to switch OS anytime soon so Vista was my vote thanks for all the hard work on the MGE project i wouldnt play morrowind without it .

Win XP
GTX 260
Phenom II X4 955
4GB RAM

With detail textures I mean that MGE onlyl seem to render vertex colors and the default base map. It doesn't seem to support any other .NIF properties, such as dark maps or detail maps or for all that matters multiple uvs. This would help immensly to get some meshes to look good in the distance and close up at the same time.

I shouldn't have even asked if it was an inferior version of the card - after all, you recommended it to me! I cannot doubt your wisdom! :bowdown:

EDIT: I have two old cards sitting around, an XFX 7900 GTX and an XFX 8800 GS. I'd be happy to donate them if anyone is interested. Just make sure your other hardware can integrate with it properly.

Ah, gotcha. Unfortunately, limiting the shader complexity of the distant objects and sorting them by texture while rendering to minimize switches are some of the most important ways that MGE can manage to still run at playable speeds while issuing so many thousands of draw calls, so I doubt this will be able to change =/ Personally, I think of it as a form of LOD where it makes sense for distant objects that are small on the screen to be less detailed than close up ones.

While partially true, the color of a mesh with both applied looks far different from only the base texture so you have a really ugly seam between that LOD. This whole thing began with dark-maps tbh. the idea was to be able to see fake shadows from afar (and close up) in order to break up the very noticeable tiling effect you can get on excessively large meshes. (in this example mountains).

While I understand it is ace for performance would it really be that much work if you can tag them in the exceptions list as some kind of special objects? so that only the meshes with a specific tag will render that way? this way it would be optional and a win/win for anyone who wants some things to just be.. super quality so to speak.

Again, just speculating on feature ideas here, but I think it would actually be quite a bit of work. You would need to adjust the static generation process to export several more textures and sets of UV coordinates for every static, and adjust the static format to flag them with the type of maps that they contained. This would dramatically increase the size of the static file and thus load time and GPU memory usage at runtime. You would also lose some or all of the effectiveness of the optimization that MGEgui currently does where it merges meshes with the same base map to cut down on draw calls. At runtime you'd either need to generate various shader permutations, or a big uber-shader with if statements to handle all the possibilities. If you wanted to support everything that Morrowind does, then you'd have almost no chance of sorting the render list in a way that would reduce texture swaps, so the best you could do would be to render identical objects together. To get to the point where the objects looked the same, you'd need to set several textures and several material colors before each group of identical meshes, which I'm pretty sure would completely tank performance. Even if someone did volunteer to do this, it would also mean a lot more optional code paths which is always a recipe for difficult bugs that only affect a small handful of people. Of course, I wouldn't ever say that it is impossible that it could ever happen. It just seems unlikely to be worth the effort on current hardware to me.

Voted! Can't wait to see what happens with MGE.

Gotcha, i'll have to do with what I have now then. Thanks for your very in-depth answer! =)

I talked briefly with Peachykeen about this, but I might as well also pitch the discussion here. Can it be done shaderwise instead? like we are doing with say, grass for example? So you will tag it as something and it will appy a shader that in it's turn applies the darkmap/detailmap? Sounds more feasible?

Whee, getting close to 100 votes now, thanks so much every one!


I'm not sure I fully understand what you're suggesting, but it seems to me that flagging a mesh as using a special shader and then trying to figure out what that shader needs so the runtime C++ code can pass in the right shader constants and set the right textures and so the static generation and loading process would know what to do would be even more work than reading it from the NIF file and coming up with a standardized encoding sceme for the static file.

If you're asking whether it would be easier to implement if the goal was to support dark and detail maps, and not all types of maps and material colors (glow and emissive color, for example), then I agree that it would be slightly easier, but but I don't think it would be by much since you would have to write a lot of the same code either way.