Re: Hoi's wip thread
Posted: 20 Feb 2009, 15:07
Flozi is right.
Open Source Realtime Strategy Game Engine
https://springrts.com/phpbb/
3 cubesFLOZi wrote:I count 34 tris with the bottom culled...smoth wrote:5*2 + 3 -> 13 * 2 -> 26
the side towers are 5 faces each and 3 on the middle section if you remove unseen geometry 2 polies for each face.
you are not culling unseen geometry are you.
if this is true he really did the walls wrong.Ps: the wall is 44 tri's, and 34 without the stuff you don't see.
The 'wall' is 6 tris. Each end post is 14. 2 for the top, 2 for the end side, 2 for each side perpendicular to the wall, and 6 for the side connected to the wall.smoth wrote:3 cubesFLOZi wrote:I count 34 tris with the bottom culled...smoth wrote:5*2 + 3 -> 13 * 2 -> 26
the side towers are 5 faces each and 3 on the middle section if you remove unseen geometry 2 polies for each face.
you are not culling unseen geometry are you.
6 faces for each
3*6 = 18
X2 to transform square faces to tris
36 polies
/me sighs,
My example was if he made the center piece a strip of 3 faces and the two towers 5 sided with no bottom.
if this is true he really did the walls wrong.Ps: the wall is 44 tri's, and 34 without the stuff you don't see.
one face is 2 polies each.... I have no idea how it would be 34
I'm just making stuff to improve, this isn't really meant for a mod or something.daan 79 wrote:do you make stuff just for you?
It's not done, so not optimised, so I can't give an ansfer to that yet.daan 79 wrote:How many poly is that gun?
It depends on the size of the zbuffer and would only really be apparant at very acute angles to the wall itself.smoth wrote:I have never really seen perpendicular z-fighting. Parallel I see a lot but perpendicular i never see.
Also these walls there might be 100s. isn't it wise to save on that sort of thing?
+1, it's a non-issue, 'cept the shadowmap that results is a little funky. I think more of that is stuff that's currently borked with the shadowmap anyhow.I have never really seen perpendicular z-fighting. Parallel I see a lot but perpendicular i never see.
Fillrate's a factor after culling of backfaces, IIRC, so most of the time it's "how big is that texture" vs. "how much texture are we showing". I've been finding that tricount's less important than fillrate... if you don't have shaders on. With shaders, it's the inverse, because of tessellation. IOW, things with lots of triangles that are then using shaders are pretty expensive, because you're going to have some tessellation on anything but the smallest bits where it's just one texel. An object with big flat planes is getting broken up into dozens, if not hundreds of smaller triangles, it's easy to see in screenshots, if you play around a bit with the camera on reflective things.Which slows down fillrate more? More surface (because more exists, albeit hidden), or more polygons?
Vertex/Fragment shaders do not and can not generate new triangles.An object with big flat planes is getting broken up into dozens, if not hundreds of smaller triangles
I dunno if I can, all the stuff textured on the image below is on one uvmap and the gun is not uvmapped yet, and spring only supports 1 uvmap, so you'll either need to use it without the stand, or I could try to fit everything on one uvmap.daan 79 wrote:You basically made the perfect battlestar gun!
Tell me when you wanna share that nice piece!