• Content count

  • Joined

  • Last visited

Community Reputation

227 Neutral

About tomek_zielinski

  • Rank
  1. General optimisation questions!

    and I'm using raw pointer: ptr = &vec[0]; for( num=0; num<vec.size(); num++, ptr++ ) { .... } In know that it's not STL standar compliant, but in VC it works since Dinkumware has got continuos vector storage
  2. Stencil-masked shadowmapping for older hardware

    I implemented it and have 'real' shadowmapping on TNT:)
  3. New version of free physic engine!!

    Good news! In upcoming patch there will be no minumal damping (only 0.0001 to maintain stability) _Vlad => this means no more moon-like things :)
  4. What do you think about such approach in a racing game: - take low poly model of a car - extrude it according to light direction - mask the area in stencil buffer - render shadow cookie of a car - preoject shadow cookie with stencil testing on this way backprojection will be easily rejected for arbitrary level geometry and orthogonal light projection
  5. New version of free physic engine!!

    I'm glad of that too, because it enabled me to do efficient shadowmapping:)
  6. New version of free physic engine!!

    There is internal damping in Newton, even if you set it to 0 - this is the cost of stability, but you can always take it into account. Several people noticed that but only one guy returned back to ODE, the rest could live with that:) EDIT: I'm using gravity of 15-25 to prevent moon-feeling:)
  7. New version of free physic engine!!

    I was using both ODE and Newton. The only advantages of ODE are that it's open-source and faster. But this speed comes at price of being unstable most of the time and require tweaking, hacking etc. Of course if you've learned how to cope with ODE you may be happy with it, but in my opinion Newton is much more powerful. First, solver is exact, stable and accurate with 'normal' input, most of the time no need to tweak anything. Second, plenty of collision primitives(null, boxes, spheres, ellipsoids, cylinders, capped cylinders, chamfer cylinders, cones, general convex hulls, triangle meshes, user-defined meshes). Collisions are exact, no problems like with ODE. Unique feature is that you can apply distortion matrix to any primitive to compensate its speed or to achieve some special effects(for example crouching player) Third - vehicle container - just plug and play, adjust vehicle parameters and you have your own arcade Collin. With more sophisticated setup you can achieve even moderate complexity simulation. Fourth - character support - unique UpVector joint for keeping you character straight up without hacking and builtin dedicated ragdoll support Fifth - builtin general raycasting for whole world and individual rigid body Besides: powerful in my opinion material interaction, user defined joints and continuos collision coming soon, and what is extremely imnporoitant for me - very active community, fast replies from developers, fast bug fixes if you find any problem and all for free. I know it sounds like a commercial, but I'm doing a commercial game and ODE made me once cry, then I found Newton and I'm pretty damn well happy with it. The only downside is that LCP solvers are faster than Newton one, but for my purpose I agree to sacrifice some speed for a stabiliti and predictibility
  8. New version of free physic engine!!

    I switched to the new SDK, implemented vehicle and I can say that speed improved somehow due to new solver and collision system. I cannot give excat numbers because my renderer is a bottleneck, but whole game became 'less jaggy'
  9. New version of free physic engine!!

    I've noticed that there is new Newton SDK I'm using old 1.24 version with success in a commercial game development and new release has plenty of new functionality, including dedicated vehicle support, many collision primitives, fully funcitonal raycasting, standalone collision detection routines and UNIQUE feature - runtime collision primitive deformation! [Edited by - tomek_zielinski on November 22, 2004 1:06:05 PM]
  10. Making textures from photos

    Thanks to you all
  11. Making textures from photos

    Hi. We're now trying to do what you said. I thought that for example guys that were making Max Payne and had thousends of photos had to do this automaticaly. We haven;t got so many photos so it's not such a huge problem, but I asked in case auto-processing is somewhere out there
  12. Making textures from photos

    Thanx once again! And once again I've got question:): how to determine if image needs correction and how much? For example asphalt is much darker than marble with the same lighting, so how to find out if relation between their brightess is correct? Do I need to use histogram or some other tools. Or I need to learn to "feel" such things? EDIT: I need this form my game. I would like to have all textures lit at the same level so that lighitng in game engine will produce real results.
  13. Making textures from photos

    Thanks very much! I'll try it
  14. Making textures from photos

    Thanx for detailed answer! Thank god it's not black magic:) With that lighting I meant not lighting differences within single photo, but across several photos(for example one taken in the mornig and one in the evening). How to filter several photos and make them looking like shot in similar lighting conditions?
  15. Making textures from photos

    I've got a question - how are you doing this? I have a buch of photos each one with different lighting. First I need to make the lighting uniform. Then I need to make them a bit 'unreal' (like in all games - colors like Hawaian see etc). Is there a path to follow, some filters or it should be made by hand?