Minor request.
Moderator: Moderators
Minor request.
Can we get a command /mipmap to disable/enable mips? so we can disable them if we want?
Re: Minor request.
Nvidia cp allows you to force them off iirc.smoth wrote:Can we get a command /mipmap to disable/enable mips? so we can disable them if we want?
e: ah, you can force them on but not off >:(
Last edited by Peet on 18 Nov 2007, 08:00, edited 1 time in total.
I think it's the way that Spring is handling them, Smoth, as well. It seems to be making mipmap determinations per polygon, instead of per model within the rendering space. This is probably a cause of a fair amount of rendering inefficiency, having to test per triangle.
Spring devs, I can go fetch some screens showing objects with two levels of mip shown in the same frame, if this is unclear.
Spring devs, I can go fetch some screens showing objects with two levels of mip shown in the same frame, if this is unclear.
Mipmaps are handled by OpenGL, it's always determined per fragment (pixel) which mipmap(s) to apply. I think there isn't even a way to change this using OpenGL.
AFAIK, the only things that can be wrong are:
AFAIK, the only things that can be wrong are:
- mipmapping isn't enabled
- incorrectly downsampled mipmaps are uploaded to your card if it doesn't support hardware generated mipmaps (any modern one does AFAIK)
- a lower then maximum quality mipmapping setting is used, ie. NEAREST_MIPMAP_LINEAR instead of LINEAR_MIPMAP_LINEAR.
- mipmaps of atlased textures (projectile sprites, 3do units) are calculated wrong
Last edited by Tobi on 18 Nov 2007, 22:20, edited 2 times in total.
No, Smoth- mips help performance a lot. If OpenGL is doing this per-fragment, it explains some of the things I've seen- I assumed that this was the case.
Freelancer handled mips the same way LOD was handled- by a distance from the camera function, so that mips were handled per-object, instead of per fragment. It looked better. However, most of the time, tbh, it's really not a problem, at least in PURE- I'm just being picky.
Freelancer handled mips the same way LOD was handled- by a distance from the camera function, so that mips were handled per-object, instead of per fragment. It looked better. However, most of the time, tbh, it's really not a problem, at least in PURE- I'm just being picky.
we also lose quality in the texture detail as we go to a lower resolution image. For me that is vexing. I have disabled bilinear and trilinear before in options to get all my sharp jaggies. When I look at spring textures I see heavily blurred images. I do not see this in other games and I thought I did not see it when I worked in the ogre engine but sure enough it is there.
*sighs* I really want some sm3
. Oh well. I have no idea what to think, maybe there is no way to improve render quality. It just seems like spring has muddy over all graphics. I know that DDS is killing the detail on my textures for my maps. They come off blurry and they lose contrast in a big way. I can have a beautiful map texture and it looks like arse in game.
I am asking all of this random crap because I am largely ignorant on how this stuff works and wanted to see if there was something that maybe was over looked. I suppose the only way to really test any of this is to go in and redefine parts of the engine to look into it. Sorry for wasting time. Feel free to lock the thread.
*sighs* I really want some sm3

I am asking all of this random crap because I am largely ignorant on how this stuff works and wanted to see if there was something that maybe was over looked. I suppose the only way to really test any of this is to go in and redefine parts of the engine to look into it. Sorry for wasting time. Feel free to lock the thread.
Modern cards love mipping too early to preserve fillrate. I guess you could try enabling aniso filtering.
I think modern cards no longer blur only within the 4 texel blocks or whatever, I've had texels on my models disappear at higher MIPs because the UVs were too close to the edge, suggesting that the algorithm did blend even across the texture's edge. The texture was power-of-two, of course.
I think modern cards no longer blur only within the 4 texel blocks or whatever, I've had texels on my models disappear at higher MIPs because the UVs were too close to the edge, suggesting that the algorithm did blend even across the texture's edge. The texture was power-of-two, of course.