• Advertisement
Sign in to follow this  

New idea that may make possible RTRT

This topic is 3488 days old which is more than the 365 day threshold we allow for new replies. Please post a new topic.

If you intended to correct an error in the post then please contact us.

Recommended Posts

ive spent years trying to figure out ways to optimize ray tracing code to allow it to run in real time. the closest ive gotten was a ray tracer relying on pre-computed kd-trees. while this provides a huge performance boost, its not suitable for video games with lots of movement. yesterday, however, ive come up with a completely novel scheme which should allow superior performance to kd-trees, and doesnt require any kind of spacial partitioning. i need help with the mathmatics and theory behind it however, so if 2 or 3 serious experts could contact me privately im willing to share my thoughts. the kind of help i need involves cone- plane intersection tests, raster line and fill-in methods, view frustrum partitioning and pyramidal- sphere intersection tests. alot of math which im not too keen on. the reason i wont post the idea here is because i wish to keep it secret, in case something special developes from this, but i do need help implementing it.

Share this post


Link to post
Share on other sites
Advertisement
Quote:
Original post by DescentPro
ive come up with a completely novel scheme which should allow superior performance to kd-trees, and doesnt require any kind of spacial partitioning.


Quote:
Original post by DescentPro
alot of math which im not too keen on.


How do those two match up?

Share this post


Link to post
Share on other sites
You will probably get a lot of pms, but keep this in mind.
Quote:

Why was GameDev.net created?

GameDev.net was created with the sole purpose of providing a public site where game developers could freely exchange information. This site isn't powered by a large corporation. It is powered by game developers, real people with real experience and education in their fields. And lastly, it is powered by you the viewer, who may some day find a topic you'd be interested in sharing with the rest of us.



Of what use is a method that no one knows about?

I really wish you good luck with your idea.

Share this post


Link to post
Share on other sites
Quote:
Original post by agi_shi
Quote:
Original post by DescentPro
ive come up with a completely novel scheme which should allow superior performance to kd-trees, and doesnt require any kind of spacial partitioning.


Quote:
Original post by DescentPro
alot of math which im not too keen on.


How do those two match up?


it takes imagination and creativity to come up with new concepts; a good knowledge in mathematics to implement it. i have a lot in the former, but im lacking somewhat in the latter.

as for my idea, to get into a little detail it differs from standard ray tracing optimizations in that it involves a two-pass procedure which will cut down the intersection tests of primary, shadows, and reflection/refraction rays by thousands if not millions of times in complex scenes. and that is without complex premade octrees or kd-trees

Share this post


Link to post
Share on other sites
Quote:
Original post by DescentPro
as for my idea, to get into a little detail it differs from standard ray tracing optimizations in that it involves a two-pass procedure which will cut down the intersection tests of primary, shadows, and reflection/refraction rays by thousands if not millions of times in complex scenes. and that is without complex premade octrees or kd-trees


What a bold statement. I sincerly hope that your idea works out but I have my doubts. After all there are a lot of people working on this problem, e.g. there's a conference specific to real-time ray tracing taking place next week in LA (I'll be there too :-) ). Your announcement simply sounds too revolutionary to be true imho.

But I'm open minded and look forward to your results. If you want some additional input form ray tracing experts I suggest you head over to the ompf-forums (where the rtrt-elite meets, hehe). :-)

Share this post


Link to post
Share on other sites



Intel's Larabee chip ( 32(?) P5 cores together ) is allegedly to be used for real-time raytracing. Of course a new method which might speed up ray-tracing computation will always make it better.

Share this post


Link to post
Share on other sites
Don't forget to write a paper about your method, and then a call it with your name, like DescentPro-Raytracing, instead of Whitted raytracing !

More seriously, good luck for your idea, but whenever we have good ideas, we often feel so happy about that, and then when we try to make it real, we discover a lot of problems that make the method not practical, but I really hope that your method will "revolutionize" (although I have some doubts), I think this is so a fascinating thing to discover new methods ;).

Share this post


Link to post
Share on other sites
Sign in to follow this  

  • Advertisement