[Ported Game] Duke Nukem 3D - EDuke32

[D4damage]

Very nice. Thanks mcobit.

Here's a keyboard reference for mapster:

http://wiki.eduke32.com/wiki/Build/Mapster32_Keyboard_Commands

 

[TrashyMG]

Very nice. Thanks mcobit.

Here's a keyboard reference for mapster:

http://wiki.eduke32.com/wiki/Build/Mapster32_Keyboard_Commands

Thanks, that isn't really the problem. lack of a keypad is.. The laptop mode I'm not sure maps it to anything useful.. it may need to be edited in the code.. but I'll check for sure.


Looks like I can have them remapped with the remap variable, the remapping is done by specifying what hex key-code you want substituted with another key separated by - sign and it has to be all on one line, comma delimited.. I will play with a config and share it when I'm done.

 

[Foxgod]

This game was awesome, i completely finished all 3 chapters on the Pandora.

Tnx again for porting it

 

[PokeParadox]

just a note if you are on 1GHz:


With the default driver the gl port has crapified textures... I used the driver switcher to the very latest driver and it seems to work perfectly albeit with a little tearing. So advice for everyone is to experiment with the driver switcher.

 

[mcobit]

Nice. Did you also get the blocky textures on the floors?


Did you try to enable vsync in the menu to prevent tearing?


I set the LIBGL_VSYNC flag, but this doesn't seem to affect it somehow.

 

[PokeParadox]

Yes I still got blocky textures, even with Bilinear filtering turned on. I believe the SGX drivers have broken bilinear which causes these weird squares to appear... Maybe one of the other drivers fixes this *shrug*

I did enable VSync, I don't remember if it did anything... I want to play around with this after work tonight.

I seems to remember Notaz putting a powervr.ini file in the same folder of a program to enable Vsync... maybe this is actually what's needed?

 

[mcobit]

If you find something, let me know. The blocky textures may be caused by lunixbochs libgl as he gets them in other apps using this lib, too. He didn't find a fix though.

 

[fusion_power]

I've done a ton of map building back in the day, mostly Blood levels as that was my favorite out of the bunch..I was amazed I could remember how to do things.. Quickly made a small map fairly easy on the Pandora once I got the 3D mode working.

I just did a few maps, not even real "playable" maps but I spent alot of time in detailing my very own "home", in the meaning of an big bunker system with alot of traps and Security stuff. One map is based on a real summer cabin from my youth but in the Build version, it has a giant Bunker system underneeth. I still have these Maps, they even worked on the GP2X port of Duke 3D.How do you actualy run Custom Maps on the EDuke Pandora Port? Command line or is there an ingame option?

Very nice. Thanks mcobit.

Here's a keyboard reference for mapster:

http://wiki.eduke32.com/wiki/Build/Mapster32_Keyboard_Commands

Yupp, that's a good bunch of hotkeys. Not really beginner friendly, I'm not even sure If I remmber just the basic keys (you "draw" your rooms with space bar in 2D mode I remember)

 

[mcobit]

You can select "Play user map" on New game dialog. IIRC

 

[PokeParadox]

not really been able to do some proper testing since my brother came round, but it seems that vsync is respected... but there is a weird floating invisible "bump" in the top right that looks like tearing and it's there with all drivers I've tested so far...

 

[lunixbochs]

Make sure you follow "I enabled vsync. Why do I still see tearing with framebuffer output!?" from my first post in the libGL thread: http://boards.openpandora.org/index.php/topic/11506-opengl-implementation-tldr-more-games/

You might not need LIBGL_VSYNC=1 if you use LIBGL_FB=1 and the powervr.ini file.

 

[crow_riot]

Yes I still got blocky textures, even with Bilinear filtering turned on.

Blocky textures? are they distorted? does it look like this: http://crow.riot.org/pnd/gles-driverbug02.png

If so, then it's a problem with the texture coordinates getting too big. I've fixed that in audiorace by wrapping the texture coordinates after some amount. but dont know if it's possible in your case.

I've also experienced this on my iPhone 3G.

 

[mcobit]

That's it crowriot


I wonder if this wrapping can be implemented by default in linixbochs lib. Some more programs seem to have those issues.


Edit: The iphone 3gs uses the same graphics chip as the pandora so no wonder that it behaves the same.


Edit2: Someone is having the same issues here:


http://stackoverflow.com/questions/15117917/opengl-es-texture-coordinates-mapping-and-limiting

 

[mcobit]

I am not using the pvrini file. Might try that.

 

[PokeParadox]

I am not using the pvrini file. Might try that.

It sounds like the best idea - I'm guessing Notaz used it with Super Hexagon for a reason.

I want to try using hi-res textures and possibly models tonight

EDIT: I'm thinking the textures are probably overkill for the resolution - we could probably do with a mid-res pack just for the Pandora: It would be an improvement but not be as large a budren on RAM.

 

[crow_riot]

That's it crowriot

I wonder if this wrapping can be implemented by default in linixbochs lib. Some more programs seem to have those issues.

i think it's possible - but will very likely add some overhead to the lib. cause extra vertices need to be added to the geometry and you need to keep track of the wrapping. but only lunixbochs can tell how much work it would be

 

[lunixbochs]

I already rebuild the vertices in many cases (like quads to tris).

Any special considerations I need to make when wrapping texture coords?

 

[crow_riot]

I already rebuild the vertices in many cases (like quads to tris).

Any special considerations I need to make when wrapping texture coords?

I've thought about it yesterday (and now while typing): I think it could be very easy by just checking the triangle's UV coordinates and offset them to be in some minor range ... let's say the coordinates are something like (just numbers here) [(16000,16000),(16002,16002),(16000,16002)] it would be easiest to subtract the lowest number of .u and .v from the higher numbers thus having [(0,0),(2,2),(0,2)] ...
of course this only helps if the meshes are layouted nicely. if there's only one big quad having UV cordinates spanning from 0 to some big value, only subdividing the quad will help

 

[mcobit]

An environmental variable could be used to enable this wrapping for games, that need it. So others won't have a performance penalty.

 

[lunixbochs]

I wrapped tex coords and it had no effect, so I looked at the actual values. They're all about -200 < x < 200 so I don't see how they'd be overflowing precision.

Saurbraten is thus far the best example of this corruption and those coords are all 0 < x < 1.

It doesn't *quite* look like crow riot's example, more like this:

http://bochs.info/img/Clipboard_2013-3-25_at_1.38_PM-20130325-134350.jpg

http://bochs.info/img/Clipboard_2013-3-19_at_5.43_PM-20130319-174648.png

http://s1.directupload.net/images/130326/lsyee8st.png

Note the perfectly-screen-aligned squares in all examples. Looks like the tiled renderer in the SGX showing its ugly face when it hits a bug.

There's a 50x30 grid of 16x16 pixel tiles (800x480), which matches the SGX spec of 16x16 pixel tiles.