Linear Interpolation
I want to develop my own animated file format.
Can I expect a significant speed increase, by increasing the number of key frames instead of using linear interpolation ?
The max number of triangles per mesh should be about 5000.
Is it true that the speed might even decrease if the size of the file exceed a certain limit? if so , what is about the max acceptable size of the file for each animation set?
Pentium 3 1000,GeForce 2.0
Thanks in advance
To answer your first question.
-Yes
But, you will eat up system memory. What you want to do is share vertices. This way when you interpolate you only have to interpolate shared vertices. When I share vertices the number of vertices usually come out to about 60% the number of triangles. So if I have a model with 800 triangles I can optimize it down to 480 vertices. Of course this is a model with a continues mesh...this means that all the triangles are joined to make a solid object.
Until you figure out a way to utilize skeletal animations you should interpolate between frames.
5000 tris seems a little high. Q3A model hover around 800. Of course newer cards are coming out so 5000 might be ok, but nor for a GF2.
-Yes
But, you will eat up system memory. What you want to do is share vertices. This way when you interpolate you only have to interpolate shared vertices. When I share vertices the number of vertices usually come out to about 60% the number of triangles. So if I have a model with 800 triangles I can optimize it down to 480 vertices. Of course this is a model with a continues mesh...this means that all the triangles are joined to make a solid object.
Until you figure out a way to utilize skeletal animations you should interpolate between frames.
5000 tris seems a little high. Q3A model hover around 800. Of course newer cards are coming out so 5000 might be ok, but nor for a GF2.
more keyframes? (youre talking the piss or what )
a simple linear lerp is very quick esp on a 1ghz machine (even on my celeron433)
>>5000 tris seems a little high. Q3A model hover around 800. Of course newer cards are coming out so 5000 might be ok, but nor for a GF2.<<
5000 tris a model to high for a gf2?
even with the works + then some enabled a gf2mx can do 3million tris a second 3million == 50,000 x 60fps
http://uk.geocities.com/sloppyturds/gotterdammerung.html
a simple linear lerp is very quick esp on a 1ghz machine (even on my celeron433)
>>5000 tris seems a little high. Q3A model hover around 800. Of course newer cards are coming out so 5000 might be ok, but nor for a GF2.<<
5000 tris a model to high for a gf2?
even with the works + then some enabled a gf2mx can do 3million tris a second 3million == 50,000 x 60fps
http://uk.geocities.com/sloppyturds/gotterdammerung.html
>>even with the works + then some enabled a gf2mx can do 3million tris a second 3million == 50,000 x 60fps
Yeah, but don''t forget you have your enviroment(map) to load, and you''re going to be overdrawing some stuff, and you have your HUD, and items scattered on the ground, and all sorts of other stuff that slowly sucks away your triangle rate.
Yeah, but don''t forget you have your enviroment(map) to load, and you''re going to be overdrawing some stuff, and you have your HUD, and items scattered on the ground, and all sorts of other stuff that slowly sucks away your triangle rate.
GF2 MX does 3 million tris/sec? I''ve managed to get mine to do 11 million
Death of one is a tragedy, death of a million is just a statistic.
Death of one is a tragedy, death of a million is just a statistic.
Thanks to everybody in particular to WhatEver.
Actually I intend to use the glDrawElements()opengl function.
The last parameter being an array of indices.
I start from continuos meshes in .3ds format as key-frames,then I convert each mesh in a opengl format, using a file converter.
I get a vertices , an indices and a texcoordinate array for each key frame,last I pack all the data in a file.
Now the problem.
To use a linear interpolation I must be sure that the indices array is the same for all the key frames, in other words the index , for example "3" must refer to the same vertex in all the key frames but it does not seem it is like that.
Any suggestion? Thanks in advance .
Actually I intend to use the glDrawElements()opengl function.
The last parameter being an array of indices.
I start from continuos meshes in .3ds format as key-frames,then I convert each mesh in a opengl format, using a file converter.
I get a vertices , an indices and a texcoordinate array for each key frame,last I pack all the data in a file.
Now the problem.
To use a linear interpolation I must be sure that the indices array is the same for all the key frames, in other words the index , for example "3" must refer to the same vertex in all the key frames but it does not seem it is like that.
Any suggestion? Thanks in advance .
Talking about triangles per second is abit naive don't you think? My software rasterizer can do 44.6 million polygons per second on a 1.3 Ghz athlon. If every polygon is 10 pixels that is. Fill rate is abit more accurate when talking about performance in my opnion
Edited by - Blitz on February 13, 2002 2:24:12 PM
Edited by - Blitz on February 13, 2002 2:24:12 PM
>>Talking about triangles per second is abit naive don''t you think?<<
yes thats why i wrote (with the works) ie real game data drawn on the screen + not benchmark results.
>>for example "3" must refer to the same vertex in all the key frames but it does not seem it is like that.<<
it should do perhaps the file conversion is stuffing up somewhere?
http://uk.geocities.com/sloppyturds/gotterdammerung.html
yes thats why i wrote (with the works) ie real game data drawn on the screen + not benchmark results.
>>for example "3" must refer to the same vertex in all the key frames but it does not seem it is like that.<<
it should do perhaps the file conversion is stuffing up somewhere?
http://uk.geocities.com/sloppyturds/gotterdammerung.html
quote:Original post by Blitz
Talking about triangles per second is abit naive don''t you think? My software rasterizer can do 44.6 million polygons per second on a 1.3 Ghz athlon. If every polygon is 10 pixels that is. Fill rate is abit more accurate when talking about performance in my opnion
Edited by - Blitz on February 13, 2002 2:24:12 PM
Indeed, I agree with you, I just have to post mindlessly now and then, to keep me just off sane...
Death of one is a tragedy, death of a million is just a statistic.
This topic is closed to new replies.
Advertisement
Popular Topics
Advertisement