I hit upon some good idea and listen. My opinion is that talking about 3D in English will be work fine as a good filter which removes tyu--bo-- away such as ▼ﾟД｡▼ｶﾌ and so on. If you understand me or agree with this , thanks for my proposal. I'll be bery glad if you begin in English.
Generally speaking in todays we must calculate and get z-value for every vetexes when converting the view coordinate to the screen one but I heard there is a good method which doesn't need to seek concrete z-value (or w). If there is,we can omit division-calc for every vertexes. Also I heard that using this method we will be able to clip or cull polygons by Z-near/far plane but I couldn't understand at all!
200-question exam on MATH! Urgh... measuring angles... vetexes...line segments...order of operation... transfering metrics to custormary and vice versa... easy alone, but when they throw every single thing we learned the past 2 quarters together, your head goes swirling around and around and around...
Perhaps you mean triangle scan conversion using 2D homogeneous coordinates,don't you? I can't tell exactly but I've found a quick explanation when came across the university of north Calolina's pages a few monthes ago. The important thing is that every homogeneous triangles can be defined or showed as a ratio linear combination of three vertexes.
Does anyone know any efficient method of ordering polygons those are once coordinated by VU1's micro-programs and out to the VU1's local memory.Ofcource in this case these polygons must be sorted without using Z-Buffer.
I'm having some questions about mipmap maybe which doesn't matter where the plat-form is. I would like to dynamically load mipmap levels only if they must be needed and also would like to delete them just when they aren't needed anymore. The question is that I must compromise to delete some textures when many maximam size of texture are requested at the same time,you know, if the sum size of tex becomes more than VRAM can cover.
How can I get to solve this problem even if allowed to compromise.
Can I use trianlge strips when using octrees algo? If a strip goes through several nodes and for example only one node is visible should I clip the strip?
Dynaminc texture loading suppose to be done by rather dynamic texture management than mipmap. I do NOT think a specialized texture management for mipmap is worthy.
>>415, I quite agree with you. Although I have no idea what platform >>413 is targetting, I guess that loading and offloading individual mipmap levels would generally be very inefficient and easily cause CPU or GPU to stall.
I have a question regarding GeForce4 Ti. Does someone have any idea what GL_DOT_PRODUCT_AFFINE_DEPTH_REPLACE_NV does? This new token seems to be defined in GL_NV_texture_shader3 OpenGL extension, which is available only on GeForce4 Ti series. Does it have something to do with "z-correct bump mapping" mentioned in the recent review at Tom's Hardware? www.tomshardware.com/graphic/02q1/020206/geforce4-03.html
It seems that whereas the actual geometry is not displaced, the depth value of the water surface is replaced on per-pixel basis according to the height of the surface. I think that the demo is using AFFINE_DEPTH_REPLACE operation or something.
Anyone knows what "z-correct bump mapping" means at all?
>>413 I had faced to the similar case as you. There is a good solution for running short of VRAM when using like as mipmap but still it is not a ideal method.(as 415 said it can be remarkely said dynamic texture management)
If you don't have to sort polygons such as no-semitrans polygons, sort these polygons not by Z-value but by texture,yes as you guess it is oftenly called "Texture sort"
To do like this , we can strongly decrease the traffic between MAIN-RAM and VRAM.
A certain Japanese console has a unusual need for its clipping matrix, it needs -|w| < z < +|w| rather then the usual 0 < z < w.
This means any text book clipping matrix, needs modifing for GS chip what we call FUCH-KUTARAGI-CHIP.
The best reference IMHO for clipping and perspective transforms is Jim Blin`s Trip down the Graphics Pipeline which describes the perspective system in details, and it also explains how to create your own clipping space.
I'm having a hard problem in transforming a direction vector from one coordinate space to the another. I need to transform the direction of the ray from world space to the object space so I can do an intersection on the triangles in a mesh.
to 327 I think you did a very great job but it seems your translation into the Japanese is calling some suck guys like as 436. I highly recommend to stop your translating,sorry.
>>434 OpenGL, the most popular graphics library in the world, requires that z value in homogeneous clipping space be within the range between -w and w, as is the case with your platform. Historically, I believe, this type of projection matrices rather than DirectX-style projection matrices (0 < z < w) has been widly used.
>>437 You can use the upper-left 3x3 matrix of your transform matrix (from world space to object space) to transform a direction vector into object space if and only if the 3x3 matrix is orthonormal, which means the 3x3 matrix represents pure rotation without any scaling or shearing. Otherwise (if the 3x3 matrix is not orthonormal), you need to use the inverse transpose of the upper-left 3x3 matrix instead.
>>438 I agree. No Japanese translation is needed for us. 327, we hope you meet our request. Thanks in advance.
